diff options
Diffstat (limited to 'drivers/staging/lustre/lustre/libcfs')
30 files changed, 13709 insertions, 0 deletions
diff --git a/drivers/staging/lustre/lustre/libcfs/Makefile b/drivers/staging/lustre/lustre/libcfs/Makefile new file mode 100644 index 000000000000..bf5c563dcacc --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/Makefile @@ -0,0 +1,21 @@ +obj-$(CONFIG_LUSTRE_FS) += libcfs.o + +libcfs-linux-objs := linux-tracefile.o linux-debug.o +libcfs-linux-objs += linux-prim.o linux-cpu.o +libcfs-linux-objs += linux-tcpip.o +libcfs-linux-objs += linux-proc.o linux-curproc.o +libcfs-linux-objs += linux-module.o +libcfs-linux-objs += linux-crypto.o +libcfs-linux-objs += linux-crypto-adler.o + +libcfs-linux-objs := $(addprefix linux/,$(libcfs-linux-objs)) + +libcfs-all-objs := debug.o fail.o nidstrings.o module.o tracefile.o \ + watchdog.o libcfs_string.o hash.o kernel_user_comm.o \ + prng.o workitem.o upcall_cache.o libcfs_cpu.o \ + libcfs_mem.o libcfs_lock.o + +libcfs-objs := $(libcfs-linux-objs) $(libcfs-all-objs) + +ccflags-y := -I$(src)/../include +ccflags-y += -I$(src)/ diff --git a/drivers/staging/lustre/lustre/libcfs/debug.c b/drivers/staging/lustre/lustre/libcfs/debug.c new file mode 100644 index 000000000000..5a87b0832074 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/debug.c @@ -0,0 +1,476 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2011, 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/debug.c + * + * Author: Phil Schwan <phil@clusterfs.com> + * + */ + +# define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> +#include "tracefile.h" + +static char debug_file_name[1024]; + +unsigned int libcfs_subsystem_debug = ~0; +CFS_MODULE_PARM(libcfs_subsystem_debug, "i", int, 0644, + "Lustre kernel debug subsystem mask"); +EXPORT_SYMBOL(libcfs_subsystem_debug); + +unsigned int libcfs_debug = (D_CANTMASK | + D_NETERROR | D_HA | D_CONFIG | D_IOCTL); +CFS_MODULE_PARM(libcfs_debug, "i", int, 0644, + "Lustre kernel debug mask"); +EXPORT_SYMBOL(libcfs_debug); + +unsigned int libcfs_debug_mb = 0; +CFS_MODULE_PARM(libcfs_debug_mb, "i", uint, 0644, + "Total debug buffer size."); +EXPORT_SYMBOL(libcfs_debug_mb); + +unsigned int libcfs_printk = D_CANTMASK; +CFS_MODULE_PARM(libcfs_printk, "i", uint, 0644, + "Lustre kernel debug console mask"); +EXPORT_SYMBOL(libcfs_printk); + +unsigned int libcfs_console_ratelimit = 1; +CFS_MODULE_PARM(libcfs_console_ratelimit, "i", uint, 0644, + "Lustre kernel debug console ratelimit (0 to disable)"); +EXPORT_SYMBOL(libcfs_console_ratelimit); + +unsigned int libcfs_console_max_delay; +CFS_MODULE_PARM(libcfs_console_max_delay, "l", uint, 0644, + "Lustre kernel debug console max delay (jiffies)"); +EXPORT_SYMBOL(libcfs_console_max_delay); + +unsigned int libcfs_console_min_delay; +CFS_MODULE_PARM(libcfs_console_min_delay, "l", uint, 0644, + "Lustre kernel debug console min delay (jiffies)"); +EXPORT_SYMBOL(libcfs_console_min_delay); + +unsigned int libcfs_console_backoff = CDEBUG_DEFAULT_BACKOFF; +CFS_MODULE_PARM(libcfs_console_backoff, "i", uint, 0644, + "Lustre kernel debug console backoff factor"); +EXPORT_SYMBOL(libcfs_console_backoff); + +unsigned int libcfs_debug_binary = 1; +EXPORT_SYMBOL(libcfs_debug_binary); + +unsigned int libcfs_stack = 3 * THREAD_SIZE / 4; +EXPORT_SYMBOL(libcfs_stack); + +unsigned int portal_enter_debugger; +EXPORT_SYMBOL(portal_enter_debugger); + +unsigned int libcfs_catastrophe; +EXPORT_SYMBOL(libcfs_catastrophe); + +unsigned int libcfs_watchdog_ratelimit = 300; +EXPORT_SYMBOL(libcfs_watchdog_ratelimit); + +unsigned int libcfs_panic_on_lbug = 1; +CFS_MODULE_PARM(libcfs_panic_on_lbug, "i", uint, 0644, + "Lustre kernel panic on LBUG"); +EXPORT_SYMBOL(libcfs_panic_on_lbug); + +atomic_t libcfs_kmemory = ATOMIC_INIT(0); +EXPORT_SYMBOL(libcfs_kmemory); + +static wait_queue_head_t debug_ctlwq; + +char libcfs_debug_file_path_arr[PATH_MAX] = LIBCFS_DEBUG_FILE_PATH_DEFAULT; + +/* We need to pass a pointer here, but elsewhere this must be a const */ +char *libcfs_debug_file_path; +CFS_MODULE_PARM(libcfs_debug_file_path, "s", charp, 0644, + "Path for dumping debug logs, " + "set 'NONE' to prevent log dumping"); + +int libcfs_panic_in_progress; + +/* libcfs_debug_token2mask() expects the returned + * string in lower-case */ +const char * +libcfs_debug_subsys2str(int subsys) +{ + switch (1 << subsys) { + default: + return NULL; + case S_UNDEFINED: + return "undefined"; + case S_MDC: + return "mdc"; + case S_MDS: + return "mds"; + case S_OSC: + return "osc"; + case S_OST: + return "ost"; + case S_CLASS: + return "class"; + case S_LOG: + return "log"; + case S_LLITE: + return "llite"; + case S_RPC: + return "rpc"; + case S_LNET: + return "lnet"; + case S_LND: + return "lnd"; + case S_PINGER: + return "pinger"; + case S_FILTER: + return "filter"; + case S_ECHO: + return "echo"; + case S_LDLM: + return "ldlm"; + case S_LOV: + return "lov"; + case S_LQUOTA: + return "lquota"; + case S_OSD: + return "osd"; + case S_LMV: + return "lmv"; + case S_SEC: + return "sec"; + case S_GSS: + return "gss"; + case S_MGC: + return "mgc"; + case S_MGS: + return "mgs"; + case S_FID: + return "fid"; + case S_FLD: + return "fld"; + } +} + +/* libcfs_debug_token2mask() expects the returned + * string in lower-case */ +const char * +libcfs_debug_dbg2str(int debug) +{ + switch (1 << debug) { + default: + return NULL; + case D_TRACE: + return "trace"; + case D_INODE: + return "inode"; + case D_SUPER: + return "super"; + case D_EXT2: + return "ext2"; + case D_MALLOC: + return "malloc"; + case D_CACHE: + return "cache"; + case D_INFO: + return "info"; + case D_IOCTL: + return "ioctl"; + case D_NETERROR: + return "neterror"; + case D_NET: + return "net"; + case D_WARNING: + return "warning"; + case D_BUFFS: + return "buffs"; + case D_OTHER: + return "other"; + case D_DENTRY: + return "dentry"; + case D_NETTRACE: + return "nettrace"; + case D_PAGE: + return "page"; + case D_DLMTRACE: + return "dlmtrace"; + case D_ERROR: + return "error"; + case D_EMERG: + return "emerg"; + case D_HA: + return "ha"; + case D_RPCTRACE: + return "rpctrace"; + case D_VFSTRACE: + return "vfstrace"; + case D_READA: + return "reada"; + case D_MMAP: + return "mmap"; + case D_CONFIG: + return "config"; + case D_CONSOLE: + return "console"; + case D_QUOTA: + return "quota"; + case D_SEC: + return "sec"; + case D_LFSCK: + return "lfsck"; + } +} + +int +libcfs_debug_mask2str(char *str, int size, int mask, int is_subsys) +{ + const char *(*fn)(int bit) = is_subsys ? libcfs_debug_subsys2str : + libcfs_debug_dbg2str; + int len = 0; + const char *token; + int i; + + if (mask == 0) { /* "0" */ + if (size > 0) + str[0] = '0'; + len = 1; + } else { /* space-separated tokens */ + for (i = 0; i < 32; i++) { + if ((mask & (1 << i)) == 0) + continue; + + token = fn(i); + if (token == NULL) /* unused bit */ + continue; + + if (len > 0) { /* separator? */ + if (len < size) + str[len] = ' '; + len++; + } + + while (*token != 0) { + if (len < size) + str[len] = *token; + token++; + len++; + } + } + } + + /* terminate 'str' */ + if (len < size) + str[len] = 0; + else + str[size - 1] = 0; + + return len; +} + +int +libcfs_debug_str2mask(int *mask, const char *str, int is_subsys) +{ + const char *(*fn)(int bit) = is_subsys ? libcfs_debug_subsys2str : + libcfs_debug_dbg2str; + int m = 0; + int matched; + int n; + int t; + + /* Allow a number for backwards compatibility */ + + for (n = strlen(str); n > 0; n--) + if (!isspace(str[n-1])) + break; + matched = n; + + if ((t = sscanf(str, "%i%n", &m, &matched)) >= 1 && + matched == n) { + /* don't print warning for lctl set_param debug=0 or -1 */ + if (m != 0 && m != -1) + CWARN("You are trying to use a numerical value for the " + "mask - this will be deprecated in a future " + "release.\n"); + *mask = m; + return 0; + } + + return cfs_str2mask(str, fn, mask, is_subsys ? 0 : D_CANTMASK, + 0xffffffff); +} + +/** + * Dump Lustre log to ::debug_file_path by calling tracefile_dump_all_pages() + */ +void libcfs_debug_dumplog_internal(void *arg) +{ + DECL_JOURNAL_DATA; + + PUSH_JOURNAL; + + if (strncmp(libcfs_debug_file_path_arr, "NONE", 4) != 0) { + snprintf(debug_file_name, sizeof(debug_file_name) - 1, + "%s.%ld." LPLD, libcfs_debug_file_path_arr, + cfs_time_current_sec(), (long_ptr_t)arg); + printk(KERN_ALERT "LustreError: dumping log to %s\n", + debug_file_name); + cfs_tracefile_dump_all_pages(debug_file_name); + libcfs_run_debug_log_upcall(debug_file_name); + } + POP_JOURNAL; +} + +int libcfs_debug_dumplog_thread(void *arg) +{ + libcfs_debug_dumplog_internal(arg); + wake_up(&debug_ctlwq); + return 0; +} + +void libcfs_debug_dumplog(void) +{ + wait_queue_t wait; + task_t *dumper; + ENTRY; + + /* we're being careful to ensure that the kernel thread is + * able to set our state to running as it exits before we + * get to schedule() */ + init_waitqueue_entry_current(&wait); + set_current_state(TASK_INTERRUPTIBLE); + add_wait_queue(&debug_ctlwq, &wait); + + dumper = kthread_run(libcfs_debug_dumplog_thread, + (void *)(long)current_pid(), + "libcfs_debug_dumper"); + if (IS_ERR(dumper)) + printk(KERN_ERR "LustreError: cannot start log dump thread:" + " %ld\n", PTR_ERR(dumper)); + else + waitq_wait(&wait, TASK_INTERRUPTIBLE); + + /* be sure to teardown if cfs_create_thread() failed */ + remove_wait_queue(&debug_ctlwq, &wait); + set_current_state(TASK_RUNNING); +} +EXPORT_SYMBOL(libcfs_debug_dumplog); + +int libcfs_debug_init(unsigned long bufsize) +{ + int rc = 0; + unsigned int max = libcfs_debug_mb; + + init_waitqueue_head(&debug_ctlwq); + + if (libcfs_console_max_delay <= 0 || /* not set by user or */ + libcfs_console_min_delay <= 0 || /* set to invalid values */ + libcfs_console_min_delay >= libcfs_console_max_delay) { + libcfs_console_max_delay = CDEBUG_DEFAULT_MAX_DELAY; + libcfs_console_min_delay = CDEBUG_DEFAULT_MIN_DELAY; + } + + if (libcfs_debug_file_path != NULL) { + memset(libcfs_debug_file_path_arr, 0, PATH_MAX); + strncpy(libcfs_debug_file_path_arr, + libcfs_debug_file_path, PATH_MAX-1); + } + + /* If libcfs_debug_mb is set to an invalid value or uninitialized + * then just make the total buffers smp_num_cpus * TCD_MAX_PAGES */ + if (max > cfs_trace_max_debug_mb() || max < num_possible_cpus()) { + max = TCD_MAX_PAGES; + } else { + max = (max / num_possible_cpus()); + max = (max << (20 - PAGE_CACHE_SHIFT)); + } + rc = cfs_tracefile_init(max); + + if (rc == 0) + libcfs_register_panic_notifier(); + + return rc; +} + +int libcfs_debug_cleanup(void) +{ + libcfs_unregister_panic_notifier(); + cfs_tracefile_exit(); + return 0; +} + +int libcfs_debug_clear_buffer(void) +{ + cfs_trace_flush_pages(); + return 0; +} + +/* Debug markers, although printed by S_LNET + * should not be be marked as such. */ +#undef DEBUG_SUBSYSTEM +#define DEBUG_SUBSYSTEM S_UNDEFINED +int libcfs_debug_mark_buffer(const char *text) +{ + CDEBUG(D_TRACE,"***************************************************\n"); + LCONSOLE(D_WARNING, "DEBUG MARKER: %s\n", text); + CDEBUG(D_TRACE,"***************************************************\n"); + + return 0; +} +#undef DEBUG_SUBSYSTEM +#define DEBUG_SUBSYSTEM S_LNET + +void libcfs_debug_set_level(unsigned int debug_level) +{ + printk(KERN_WARNING "Lustre: Setting portals debug level to %08x\n", + debug_level); + libcfs_debug = debug_level; +} + +EXPORT_SYMBOL(libcfs_debug_set_level); + +long libcfs_log_return(struct libcfs_debug_msg_data *msgdata, long rc) +{ + libcfs_debug_msg(msgdata, "Process leaving (rc=%lu : %ld : %lx)\n", + rc, rc, rc); + return rc; +} +EXPORT_SYMBOL(libcfs_log_return); + +void libcfs_log_goto(struct libcfs_debug_msg_data *msgdata, const char *label, + long_ptr_t rc) +{ + libcfs_debug_msg(msgdata, "Process leaving via %s (rc=" LPLU " : " LPLD + " : " LPLX ")\n", label, (ulong_ptr_t)rc, rc, rc); +} +EXPORT_SYMBOL(libcfs_log_goto); diff --git a/drivers/staging/lustre/lustre/libcfs/fail.c b/drivers/staging/lustre/lustre/libcfs/fail.c new file mode 100644 index 000000000000..c54448d69008 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/fail.c @@ -0,0 +1,137 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see http://www.gnu.org/licenses + * + * Please contact Oracle Corporation, Inc., 500 Oracle Parkway, Redwood Shores, + * CA 94065 USA or visit www.oracle.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2011, 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Oracle Corporation, Inc. + */ + +#include <linux/libcfs/libcfs.h> + +unsigned long cfs_fail_loc = 0; +unsigned int cfs_fail_val = 0; +wait_queue_head_t cfs_race_waitq; +int cfs_race_state; + +EXPORT_SYMBOL(cfs_fail_loc); +EXPORT_SYMBOL(cfs_fail_val); +EXPORT_SYMBOL(cfs_race_waitq); +EXPORT_SYMBOL(cfs_race_state); + +int __cfs_fail_check_set(__u32 id, __u32 value, int set) +{ + static atomic_t cfs_fail_count = ATOMIC_INIT(0); + + LASSERT(!(id & CFS_FAIL_ONCE)); + + if ((cfs_fail_loc & (CFS_FAILED | CFS_FAIL_ONCE)) == + (CFS_FAILED | CFS_FAIL_ONCE)) { + atomic_set(&cfs_fail_count, 0); /* paranoia */ + return 0; + } + + /* Fail 1/cfs_fail_val times */ + if (cfs_fail_loc & CFS_FAIL_RAND) { + if (cfs_fail_val < 2 || cfs_rand() % cfs_fail_val > 0) + return 0; + } + + /* Skip the first cfs_fail_val, then fail */ + if (cfs_fail_loc & CFS_FAIL_SKIP) { + if (atomic_inc_return(&cfs_fail_count) <= cfs_fail_val) + return 0; + } + + /* check cfs_fail_val... */ + if (set == CFS_FAIL_LOC_VALUE) { + if (cfs_fail_val != -1 && cfs_fail_val != value) + return 0; + } + + /* Fail cfs_fail_val times, overridden by FAIL_ONCE */ + if (cfs_fail_loc & CFS_FAIL_SOME && + (!(cfs_fail_loc & CFS_FAIL_ONCE) || cfs_fail_val <= 1)) { + int count = atomic_inc_return(&cfs_fail_count); + + if (count >= cfs_fail_val) { + set_bit(CFS_FAIL_ONCE_BIT, &cfs_fail_loc); + atomic_set(&cfs_fail_count, 0); + /* we are lost race to increase */ + if (count > cfs_fail_val) + return 0; + } + } + + if ((set == CFS_FAIL_LOC_ORSET || set == CFS_FAIL_LOC_RESET) && + (value & CFS_FAIL_ONCE)) + set_bit(CFS_FAIL_ONCE_BIT, &cfs_fail_loc); + /* Lost race to set CFS_FAILED_BIT. */ + if (test_and_set_bit(CFS_FAILED_BIT, &cfs_fail_loc)) { + /* If CFS_FAIL_ONCE is valid, only one process can fail, + * otherwise multi-process can fail at the same time. */ + if (cfs_fail_loc & CFS_FAIL_ONCE) + return 0; + } + + switch (set) { + case CFS_FAIL_LOC_NOSET: + case CFS_FAIL_LOC_VALUE: + break; + case CFS_FAIL_LOC_ORSET: + cfs_fail_loc |= value & ~(CFS_FAILED | CFS_FAIL_ONCE); + break; + case CFS_FAIL_LOC_RESET: + cfs_fail_loc = value; + break; + default: + LASSERTF(0, "called with bad set %u\n", set); + break; + } + + return 1; +} +EXPORT_SYMBOL(__cfs_fail_check_set); + +int __cfs_fail_timeout_set(__u32 id, __u32 value, int ms, int set) +{ + int ret = 0; + + ret = __cfs_fail_check_set(id, value, set); + if (ret) { + CERROR("cfs_fail_timeout id %x sleeping for %dms\n", + id, ms); + schedule_timeout_and_set_state(TASK_UNINTERRUPTIBLE, + cfs_time_seconds(ms) / 1000); + set_current_state(TASK_RUNNING); + CERROR("cfs_fail_timeout id %x awake\n", id); + } + return ret; +} +EXPORT_SYMBOL(__cfs_fail_timeout_set); diff --git a/drivers/staging/lustre/lustre/libcfs/hash.c b/drivers/staging/lustre/lustre/libcfs/hash.c new file mode 100644 index 000000000000..98c76dfac3dd --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/hash.c @@ -0,0 +1,2123 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2011, 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/hash.c + * + * Implement a hash class for hash process in lustre system. + * + * Author: YuZhangyong <yzy@clusterfs.com> + * + * 2008-08-15: Brian Behlendorf <behlendorf1@llnl.gov> + * - Simplified API and improved documentation + * - Added per-hash feature flags: + * * CFS_HASH_DEBUG additional validation + * * CFS_HASH_REHASH dynamic rehashing + * - Added per-hash statistics + * - General performance enhancements + * + * 2009-07-31: Liang Zhen <zhen.liang@sun.com> + * - move all stuff to libcfs + * - don't allow cur_bits != max_bits without setting of CFS_HASH_REHASH + * - ignore hs_rwlock if without CFS_HASH_REHASH setting + * - buckets are allocated one by one(intead of contiguous memory), + * to avoid unnecessary cacheline conflict + * + * 2010-03-01: Liang Zhen <zhen.liang@sun.com> + * - "bucket" is a group of hlist_head now, user can speicify bucket size + * by bkt_bits of cfs_hash_create(), all hlist_heads in a bucket share + * one lock for reducing memory overhead. + * + * - support lockless hash, caller will take care of locks: + * avoid lock overhead for hash tables that are already protected + * by locking in the caller for another reason + * + * - support both spin_lock/rwlock for bucket: + * overhead of spinlock contention is lower than read/write + * contention of rwlock, so using spinlock to serialize operations on + * bucket is more reasonable for those frequently changed hash tables + * + * - support one-single lock mode: + * one lock to protect all hash operations to avoid overhead of + * multiple locks if hash table is always small + * + * - removed a lot of unnecessary addref & decref on hash element: + * addref & decref are atomic operations in many use-cases which + * are expensive. + * + * - support non-blocking cfs_hash_add() and cfs_hash_findadd(): + * some lustre use-cases require these functions to be strictly + * non-blocking, we need to schedule required rehash on a different + * thread on those cases. + * + * - safer rehash on large hash table + * In old implementation, rehash function will exclusively lock the + * hash table and finish rehash in one batch, it's dangerous on SMP + * system because rehash millions of elements could take long time. + * New implemented rehash can release lock and relax CPU in middle + * of rehash, it's safe for another thread to search/change on the + * hash table even it's in rehasing. + * + * - support two different refcount modes + * . hash table has refcount on element + * . hash table doesn't change refcount on adding/removing element + * + * - support long name hash table (for param-tree) + * + * - fix a bug for cfs_hash_rehash_key: + * in old implementation, cfs_hash_rehash_key could screw up the + * hash-table because @key is overwritten without any protection. + * Now we need user to define hs_keycpy for those rehash enabled + * hash tables, cfs_hash_rehash_key will overwrite hash-key + * inside lock by calling hs_keycpy. + * + * - better hash iteration: + * Now we support both locked iteration & lockless iteration of hash + * table. Also, user can break the iteration by return 1 in callback. + */ + +#include <linux/libcfs/libcfs.h> +#include <linux/seq_file.h> + +#if CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1 +static unsigned int warn_on_depth = 8; +CFS_MODULE_PARM(warn_on_depth, "i", uint, 0644, + "warning when hash depth is high."); +#endif + +struct cfs_wi_sched *cfs_sched_rehash; + +static inline void +cfs_hash_nl_lock(cfs_hash_lock_t *lock, int exclusive) {} + +static inline void +cfs_hash_nl_unlock(cfs_hash_lock_t *lock, int exclusive) {} + +static inline void +cfs_hash_spin_lock(cfs_hash_lock_t *lock, int exclusive) +{ + spin_lock(&lock->spin); +} + +static inline void +cfs_hash_spin_unlock(cfs_hash_lock_t *lock, int exclusive) +{ + spin_unlock(&lock->spin); +} + +static inline void +cfs_hash_rw_lock(cfs_hash_lock_t *lock, int exclusive) +{ + if (!exclusive) + read_lock(&lock->rw); + else + write_lock(&lock->rw); +} + +static inline void +cfs_hash_rw_unlock(cfs_hash_lock_t *lock, int exclusive) +{ + if (!exclusive) + read_unlock(&lock->rw); + else + write_unlock(&lock->rw); +} + +/** No lock hash */ +static cfs_hash_lock_ops_t cfs_hash_nl_lops = +{ + .hs_lock = cfs_hash_nl_lock, + .hs_unlock = cfs_hash_nl_unlock, + .hs_bkt_lock = cfs_hash_nl_lock, + .hs_bkt_unlock = cfs_hash_nl_unlock, +}; + +/** no bucket lock, one spinlock to protect everything */ +static cfs_hash_lock_ops_t cfs_hash_nbl_lops = +{ + .hs_lock = cfs_hash_spin_lock, + .hs_unlock = cfs_hash_spin_unlock, + .hs_bkt_lock = cfs_hash_nl_lock, + .hs_bkt_unlock = cfs_hash_nl_unlock, +}; + +/** spin bucket lock, rehash is enabled */ +static cfs_hash_lock_ops_t cfs_hash_bkt_spin_lops = +{ + .hs_lock = cfs_hash_rw_lock, + .hs_unlock = cfs_hash_rw_unlock, + .hs_bkt_lock = cfs_hash_spin_lock, + .hs_bkt_unlock = cfs_hash_spin_unlock, +}; + +/** rw bucket lock, rehash is enabled */ +static cfs_hash_lock_ops_t cfs_hash_bkt_rw_lops = +{ + .hs_lock = cfs_hash_rw_lock, + .hs_unlock = cfs_hash_rw_unlock, + .hs_bkt_lock = cfs_hash_rw_lock, + .hs_bkt_unlock = cfs_hash_rw_unlock, +}; + +/** spin bucket lock, rehash is disabled */ +static cfs_hash_lock_ops_t cfs_hash_nr_bkt_spin_lops = +{ + .hs_lock = cfs_hash_nl_lock, + .hs_unlock = cfs_hash_nl_unlock, + .hs_bkt_lock = cfs_hash_spin_lock, + .hs_bkt_unlock = cfs_hash_spin_unlock, +}; + +/** rw bucket lock, rehash is disabled */ +static cfs_hash_lock_ops_t cfs_hash_nr_bkt_rw_lops = +{ + .hs_lock = cfs_hash_nl_lock, + .hs_unlock = cfs_hash_nl_unlock, + .hs_bkt_lock = cfs_hash_rw_lock, + .hs_bkt_unlock = cfs_hash_rw_unlock, +}; + +static void +cfs_hash_lock_setup(cfs_hash_t *hs) +{ + if (cfs_hash_with_no_lock(hs)) { + hs->hs_lops = &cfs_hash_nl_lops; + + } else if (cfs_hash_with_no_bktlock(hs)) { + hs->hs_lops = &cfs_hash_nbl_lops; + spin_lock_init(&hs->hs_lock.spin); + + } else if (cfs_hash_with_rehash(hs)) { + rwlock_init(&hs->hs_lock.rw); + + if (cfs_hash_with_rw_bktlock(hs)) + hs->hs_lops = &cfs_hash_bkt_rw_lops; + else if (cfs_hash_with_spin_bktlock(hs)) + hs->hs_lops = &cfs_hash_bkt_spin_lops; + else + LBUG(); + } else { + if (cfs_hash_with_rw_bktlock(hs)) + hs->hs_lops = &cfs_hash_nr_bkt_rw_lops; + else if (cfs_hash_with_spin_bktlock(hs)) + hs->hs_lops = &cfs_hash_nr_bkt_spin_lops; + else + LBUG(); + } +} + +/** + * Simple hash head without depth tracking + * new element is always added to head of hlist + */ +typedef struct { + struct hlist_head hh_head; /**< entries list */ +} cfs_hash_head_t; + +static int +cfs_hash_hh_hhead_size(cfs_hash_t *hs) +{ + return sizeof(cfs_hash_head_t); +} + +static struct hlist_head * +cfs_hash_hh_hhead(cfs_hash_t *hs, cfs_hash_bd_t *bd) +{ + cfs_hash_head_t *head = (cfs_hash_head_t *)&bd->bd_bucket->hsb_head[0]; + + return &head[bd->bd_offset].hh_head; +} + +static int +cfs_hash_hh_hnode_add(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnode) +{ + hlist_add_head(hnode, cfs_hash_hh_hhead(hs, bd)); + return -1; /* unknown depth */ +} + +static int +cfs_hash_hh_hnode_del(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnode) +{ + hlist_del_init(hnode); + return -1; /* unknown depth */ +} + +/** + * Simple hash head with depth tracking + * new element is always added to head of hlist + */ +typedef struct { + struct hlist_head hd_head; /**< entries list */ + unsigned int hd_depth; /**< list length */ +} cfs_hash_head_dep_t; + +static int +cfs_hash_hd_hhead_size(cfs_hash_t *hs) +{ + return sizeof(cfs_hash_head_dep_t); +} + +static struct hlist_head * +cfs_hash_hd_hhead(cfs_hash_t *hs, cfs_hash_bd_t *bd) +{ + cfs_hash_head_dep_t *head; + + head = (cfs_hash_head_dep_t *)&bd->bd_bucket->hsb_head[0]; + return &head[bd->bd_offset].hd_head; +} + +static int +cfs_hash_hd_hnode_add(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnode) +{ + cfs_hash_head_dep_t *hh = container_of(cfs_hash_hd_hhead(hs, bd), + cfs_hash_head_dep_t, hd_head); + hlist_add_head(hnode, &hh->hd_head); + return ++hh->hd_depth; +} + +static int +cfs_hash_hd_hnode_del(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnode) +{ + cfs_hash_head_dep_t *hh = container_of(cfs_hash_hd_hhead(hs, bd), + cfs_hash_head_dep_t, hd_head); + hlist_del_init(hnode); + return --hh->hd_depth; +} + +/** + * double links hash head without depth tracking + * new element is always added to tail of hlist + */ +typedef struct { + struct hlist_head dh_head; /**< entries list */ + struct hlist_node *dh_tail; /**< the last entry */ +} cfs_hash_dhead_t; + +static int +cfs_hash_dh_hhead_size(cfs_hash_t *hs) +{ + return sizeof(cfs_hash_dhead_t); +} + +static struct hlist_head * +cfs_hash_dh_hhead(cfs_hash_t *hs, cfs_hash_bd_t *bd) +{ + cfs_hash_dhead_t *head; + + head = (cfs_hash_dhead_t *)&bd->bd_bucket->hsb_head[0]; + return &head[bd->bd_offset].dh_head; +} + +static int +cfs_hash_dh_hnode_add(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnode) +{ + cfs_hash_dhead_t *dh = container_of(cfs_hash_dh_hhead(hs, bd), + cfs_hash_dhead_t, dh_head); + + if (dh->dh_tail != NULL) /* not empty */ + hlist_add_after(dh->dh_tail, hnode); + else /* empty list */ + hlist_add_head(hnode, &dh->dh_head); + dh->dh_tail = hnode; + return -1; /* unknown depth */ +} + +static int +cfs_hash_dh_hnode_del(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnd) +{ + cfs_hash_dhead_t *dh = container_of(cfs_hash_dh_hhead(hs, bd), + cfs_hash_dhead_t, dh_head); + + if (hnd->next == NULL) { /* it's the tail */ + dh->dh_tail = (hnd->pprev == &dh->dh_head.first) ? NULL : + container_of(hnd->pprev, struct hlist_node, next); + } + hlist_del_init(hnd); + return -1; /* unknown depth */ +} + +/** + * double links hash head with depth tracking + * new element is always added to tail of hlist + */ +typedef struct { + struct hlist_head dd_head; /**< entries list */ + struct hlist_node *dd_tail; /**< the last entry */ + unsigned int dd_depth; /**< list length */ +} cfs_hash_dhead_dep_t; + +static int +cfs_hash_dd_hhead_size(cfs_hash_t *hs) +{ + return sizeof(cfs_hash_dhead_dep_t); +} + +static struct hlist_head * +cfs_hash_dd_hhead(cfs_hash_t *hs, cfs_hash_bd_t *bd) +{ + cfs_hash_dhead_dep_t *head; + + head = (cfs_hash_dhead_dep_t *)&bd->bd_bucket->hsb_head[0]; + return &head[bd->bd_offset].dd_head; +} + +static int +cfs_hash_dd_hnode_add(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnode) +{ + cfs_hash_dhead_dep_t *dh = container_of(cfs_hash_dd_hhead(hs, bd), + cfs_hash_dhead_dep_t, dd_head); + + if (dh->dd_tail != NULL) /* not empty */ + hlist_add_after(dh->dd_tail, hnode); + else /* empty list */ + hlist_add_head(hnode, &dh->dd_head); + dh->dd_tail = hnode; + return ++dh->dd_depth; +} + +static int +cfs_hash_dd_hnode_del(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnd) +{ + cfs_hash_dhead_dep_t *dh = container_of(cfs_hash_dd_hhead(hs, bd), + cfs_hash_dhead_dep_t, dd_head); + + if (hnd->next == NULL) { /* it's the tail */ + dh->dd_tail = (hnd->pprev == &dh->dd_head.first) ? NULL : + container_of(hnd->pprev, struct hlist_node, next); + } + hlist_del_init(hnd); + return --dh->dd_depth; +} + +static cfs_hash_hlist_ops_t cfs_hash_hh_hops = { + .hop_hhead = cfs_hash_hh_hhead, + .hop_hhead_size = cfs_hash_hh_hhead_size, + .hop_hnode_add = cfs_hash_hh_hnode_add, + .hop_hnode_del = cfs_hash_hh_hnode_del, +}; + +static cfs_hash_hlist_ops_t cfs_hash_hd_hops = { + .hop_hhead = cfs_hash_hd_hhead, + .hop_hhead_size = cfs_hash_hd_hhead_size, + .hop_hnode_add = cfs_hash_hd_hnode_add, + .hop_hnode_del = cfs_hash_hd_hnode_del, +}; + +static cfs_hash_hlist_ops_t cfs_hash_dh_hops = { + .hop_hhead = cfs_hash_dh_hhead, + .hop_hhead_size = cfs_hash_dh_hhead_size, + .hop_hnode_add = cfs_hash_dh_hnode_add, + .hop_hnode_del = cfs_hash_dh_hnode_del, +}; + +static cfs_hash_hlist_ops_t cfs_hash_dd_hops = { + .hop_hhead = cfs_hash_dd_hhead, + .hop_hhead_size = cfs_hash_dd_hhead_size, + .hop_hnode_add = cfs_hash_dd_hnode_add, + .hop_hnode_del = cfs_hash_dd_hnode_del, +}; + +static void +cfs_hash_hlist_setup(cfs_hash_t *hs) +{ + if (cfs_hash_with_add_tail(hs)) { + hs->hs_hops = cfs_hash_with_depth(hs) ? + &cfs_hash_dd_hops : &cfs_hash_dh_hops; + } else { + hs->hs_hops = cfs_hash_with_depth(hs) ? + &cfs_hash_hd_hops : &cfs_hash_hh_hops; + } +} + +static void +cfs_hash_bd_from_key(cfs_hash_t *hs, cfs_hash_bucket_t **bkts, + unsigned int bits, const void *key, cfs_hash_bd_t *bd) +{ + unsigned int index = cfs_hash_id(hs, key, (1U << bits) - 1); + + LASSERT(bits == hs->hs_cur_bits || bits == hs->hs_rehash_bits); + + bd->bd_bucket = bkts[index & ((1U << (bits - hs->hs_bkt_bits)) - 1)]; + bd->bd_offset = index >> (bits - hs->hs_bkt_bits); +} + +void +cfs_hash_bd_get(cfs_hash_t *hs, const void *key, cfs_hash_bd_t *bd) +{ + /* NB: caller should hold hs->hs_rwlock if REHASH is set */ + if (likely(hs->hs_rehash_buckets == NULL)) { + cfs_hash_bd_from_key(hs, hs->hs_buckets, + hs->hs_cur_bits, key, bd); + } else { + LASSERT(hs->hs_rehash_bits != 0); + cfs_hash_bd_from_key(hs, hs->hs_rehash_buckets, + hs->hs_rehash_bits, key, bd); + } +} +EXPORT_SYMBOL(cfs_hash_bd_get); + +static inline void +cfs_hash_bd_dep_record(cfs_hash_t *hs, cfs_hash_bd_t *bd, int dep_cur) +{ + if (likely(dep_cur <= bd->bd_bucket->hsb_depmax)) + return; + + bd->bd_bucket->hsb_depmax = dep_cur; +# if CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1 + if (likely(warn_on_depth == 0 || + max(warn_on_depth, hs->hs_dep_max) >= dep_cur)) + return; + + spin_lock(&hs->hs_dep_lock); + hs->hs_dep_max = dep_cur; + hs->hs_dep_bkt = bd->bd_bucket->hsb_index; + hs->hs_dep_off = bd->bd_offset; + hs->hs_dep_bits = hs->hs_cur_bits; + spin_unlock(&hs->hs_dep_lock); + + cfs_wi_schedule(cfs_sched_rehash, &hs->hs_dep_wi); +# endif +} + +void +cfs_hash_bd_add_locked(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnode) +{ + int rc; + + rc = hs->hs_hops->hop_hnode_add(hs, bd, hnode); + cfs_hash_bd_dep_record(hs, bd, rc); + bd->bd_bucket->hsb_version++; + if (unlikely(bd->bd_bucket->hsb_version == 0)) + bd->bd_bucket->hsb_version++; + bd->bd_bucket->hsb_count++; + + if (cfs_hash_with_counter(hs)) + atomic_inc(&hs->hs_count); + if (!cfs_hash_with_no_itemref(hs)) + cfs_hash_get(hs, hnode); +} +EXPORT_SYMBOL(cfs_hash_bd_add_locked); + +void +cfs_hash_bd_del_locked(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnode) +{ + hs->hs_hops->hop_hnode_del(hs, bd, hnode); + + LASSERT(bd->bd_bucket->hsb_count > 0); + bd->bd_bucket->hsb_count--; + bd->bd_bucket->hsb_version++; + if (unlikely(bd->bd_bucket->hsb_version == 0)) + bd->bd_bucket->hsb_version++; + + if (cfs_hash_with_counter(hs)) { + LASSERT(atomic_read(&hs->hs_count) > 0); + atomic_dec(&hs->hs_count); + } + if (!cfs_hash_with_no_itemref(hs)) + cfs_hash_put_locked(hs, hnode); +} +EXPORT_SYMBOL(cfs_hash_bd_del_locked); + +void +cfs_hash_bd_move_locked(cfs_hash_t *hs, cfs_hash_bd_t *bd_old, + cfs_hash_bd_t *bd_new, struct hlist_node *hnode) +{ + cfs_hash_bucket_t *obkt = bd_old->bd_bucket; + cfs_hash_bucket_t *nbkt = bd_new->bd_bucket; + int rc; + + if (cfs_hash_bd_compare(bd_old, bd_new) == 0) + return; + + /* use cfs_hash_bd_hnode_add/del, to avoid atomic & refcount ops + * in cfs_hash_bd_del/add_locked */ + hs->hs_hops->hop_hnode_del(hs, bd_old, hnode); + rc = hs->hs_hops->hop_hnode_add(hs, bd_new, hnode); + cfs_hash_bd_dep_record(hs, bd_new, rc); + + LASSERT(obkt->hsb_count > 0); + obkt->hsb_count--; + obkt->hsb_version++; + if (unlikely(obkt->hsb_version == 0)) + obkt->hsb_version++; + nbkt->hsb_count++; + nbkt->hsb_version++; + if (unlikely(nbkt->hsb_version == 0)) + nbkt->hsb_version++; +} +EXPORT_SYMBOL(cfs_hash_bd_move_locked); + +enum { + /** always set, for sanity (avoid ZERO intent) */ + CFS_HS_LOOKUP_MASK_FIND = 1 << 0, + /** return entry with a ref */ + CFS_HS_LOOKUP_MASK_REF = 1 << 1, + /** add entry if not existing */ + CFS_HS_LOOKUP_MASK_ADD = 1 << 2, + /** delete entry, ignore other masks */ + CFS_HS_LOOKUP_MASK_DEL = 1 << 3, +}; + +typedef enum cfs_hash_lookup_intent { + /** return item w/o refcount */ + CFS_HS_LOOKUP_IT_PEEK = CFS_HS_LOOKUP_MASK_FIND, + /** return item with refcount */ + CFS_HS_LOOKUP_IT_FIND = (CFS_HS_LOOKUP_MASK_FIND | + CFS_HS_LOOKUP_MASK_REF), + /** return item w/o refcount if existed, otherwise add */ + CFS_HS_LOOKUP_IT_ADD = (CFS_HS_LOOKUP_MASK_FIND | + CFS_HS_LOOKUP_MASK_ADD), + /** return item with refcount if existed, otherwise add */ + CFS_HS_LOOKUP_IT_FINDADD = (CFS_HS_LOOKUP_IT_FIND | + CFS_HS_LOOKUP_MASK_ADD), + /** delete if existed */ + CFS_HS_LOOKUP_IT_FINDDEL = (CFS_HS_LOOKUP_MASK_FIND | + CFS_HS_LOOKUP_MASK_DEL) +} cfs_hash_lookup_intent_t; + +static struct hlist_node * +cfs_hash_bd_lookup_intent(cfs_hash_t *hs, cfs_hash_bd_t *bd, + const void *key, struct hlist_node *hnode, + cfs_hash_lookup_intent_t intent) + +{ + struct hlist_head *hhead = cfs_hash_bd_hhead(hs, bd); + struct hlist_node *ehnode; + struct hlist_node *match; + int intent_add = (intent & CFS_HS_LOOKUP_MASK_ADD) != 0; + + /* with this function, we can avoid a lot of useless refcount ops, + * which are expensive atomic operations most time. */ + match = intent_add ? NULL : hnode; + hlist_for_each(ehnode, hhead) { + if (!cfs_hash_keycmp(hs, key, ehnode)) + continue; + + if (match != NULL && match != ehnode) /* can't match */ + continue; + + /* match and ... */ + if ((intent & CFS_HS_LOOKUP_MASK_DEL) != 0) { + cfs_hash_bd_del_locked(hs, bd, ehnode); + return ehnode; + } + + /* caller wants refcount? */ + if ((intent & CFS_HS_LOOKUP_MASK_REF) != 0) + cfs_hash_get(hs, ehnode); + return ehnode; + } + /* no match item */ + if (!intent_add) + return NULL; + + LASSERT(hnode != NULL); + cfs_hash_bd_add_locked(hs, bd, hnode); + return hnode; +} + +struct hlist_node * +cfs_hash_bd_lookup_locked(cfs_hash_t *hs, cfs_hash_bd_t *bd, const void *key) +{ + return cfs_hash_bd_lookup_intent(hs, bd, key, NULL, + CFS_HS_LOOKUP_IT_FIND); +} +EXPORT_SYMBOL(cfs_hash_bd_lookup_locked); + +struct hlist_node * +cfs_hash_bd_peek_locked(cfs_hash_t *hs, cfs_hash_bd_t *bd, const void *key) +{ + return cfs_hash_bd_lookup_intent(hs, bd, key, NULL, + CFS_HS_LOOKUP_IT_PEEK); +} +EXPORT_SYMBOL(cfs_hash_bd_peek_locked); + +struct hlist_node * +cfs_hash_bd_findadd_locked(cfs_hash_t *hs, cfs_hash_bd_t *bd, + const void *key, struct hlist_node *hnode, + int noref) +{ + return cfs_hash_bd_lookup_intent(hs, bd, key, hnode, + CFS_HS_LOOKUP_IT_ADD | + (!noref * CFS_HS_LOOKUP_MASK_REF)); +} +EXPORT_SYMBOL(cfs_hash_bd_findadd_locked); + +struct hlist_node * +cfs_hash_bd_finddel_locked(cfs_hash_t *hs, cfs_hash_bd_t *bd, + const void *key, struct hlist_node *hnode) +{ + /* hnode can be NULL, we find the first item with @key */ + return cfs_hash_bd_lookup_intent(hs, bd, key, hnode, + CFS_HS_LOOKUP_IT_FINDDEL); +} +EXPORT_SYMBOL(cfs_hash_bd_finddel_locked); + +static void +cfs_hash_multi_bd_lock(cfs_hash_t *hs, cfs_hash_bd_t *bds, + unsigned n, int excl) +{ + cfs_hash_bucket_t *prev = NULL; + int i; + + /** + * bds must be ascendantly ordered by bd->bd_bucket->hsb_index. + * NB: it's possible that several bds point to the same bucket but + * have different bd::bd_offset, so need take care of deadlock. + */ + cfs_hash_for_each_bd(bds, n, i) { + if (prev == bds[i].bd_bucket) + continue; + + LASSERT(prev == NULL || + prev->hsb_index < bds[i].bd_bucket->hsb_index); + cfs_hash_bd_lock(hs, &bds[i], excl); + prev = bds[i].bd_bucket; + } +} + +static void +cfs_hash_multi_bd_unlock(cfs_hash_t *hs, cfs_hash_bd_t *bds, + unsigned n, int excl) +{ + cfs_hash_bucket_t *prev = NULL; + int i; + + cfs_hash_for_each_bd(bds, n, i) { + if (prev != bds[i].bd_bucket) { + cfs_hash_bd_unlock(hs, &bds[i], excl); + prev = bds[i].bd_bucket; + } + } +} + +static struct hlist_node * +cfs_hash_multi_bd_lookup_locked(cfs_hash_t *hs, cfs_hash_bd_t *bds, + unsigned n, const void *key) +{ + struct hlist_node *ehnode; + unsigned i; + + cfs_hash_for_each_bd(bds, n, i) { + ehnode = cfs_hash_bd_lookup_intent(hs, &bds[i], key, NULL, + CFS_HS_LOOKUP_IT_FIND); + if (ehnode != NULL) + return ehnode; + } + return NULL; +} + +static struct hlist_node * +cfs_hash_multi_bd_findadd_locked(cfs_hash_t *hs, + cfs_hash_bd_t *bds, unsigned n, const void *key, + struct hlist_node *hnode, int noref) +{ + struct hlist_node *ehnode; + int intent; + unsigned i; + + LASSERT(hnode != NULL); + intent = CFS_HS_LOOKUP_IT_PEEK | (!noref * CFS_HS_LOOKUP_MASK_REF); + + cfs_hash_for_each_bd(bds, n, i) { + ehnode = cfs_hash_bd_lookup_intent(hs, &bds[i], key, + NULL, intent); + if (ehnode != NULL) + return ehnode; + } + + if (i == 1) { /* only one bucket */ + cfs_hash_bd_add_locked(hs, &bds[0], hnode); + } else { + cfs_hash_bd_t mybd; + + cfs_hash_bd_get(hs, key, &mybd); + cfs_hash_bd_add_locked(hs, &mybd, hnode); + } + + return hnode; +} + +static struct hlist_node * +cfs_hash_multi_bd_finddel_locked(cfs_hash_t *hs, cfs_hash_bd_t *bds, + unsigned n, const void *key, + struct hlist_node *hnode) +{ + struct hlist_node *ehnode; + unsigned i; + + cfs_hash_for_each_bd(bds, n, i) { + ehnode = cfs_hash_bd_lookup_intent(hs, &bds[i], key, hnode, + CFS_HS_LOOKUP_IT_FINDDEL); + if (ehnode != NULL) + return ehnode; + } + return NULL; +} + +static void +cfs_hash_bd_order(cfs_hash_bd_t *bd1, cfs_hash_bd_t *bd2) +{ + int rc; + + if (bd2->bd_bucket == NULL) + return; + + if (bd1->bd_bucket == NULL) { + *bd1 = *bd2; + bd2->bd_bucket = NULL; + return; + } + + rc = cfs_hash_bd_compare(bd1, bd2); + if (rc == 0) { + bd2->bd_bucket = NULL; + + } else if (rc > 0) { /* swab bd1 and bd2 */ + cfs_hash_bd_t tmp; + + tmp = *bd2; + *bd2 = *bd1; + *bd1 = tmp; + } +} + +void +cfs_hash_dual_bd_get(cfs_hash_t *hs, const void *key, cfs_hash_bd_t *bds) +{ + /* NB: caller should hold hs_lock.rw if REHASH is set */ + cfs_hash_bd_from_key(hs, hs->hs_buckets, + hs->hs_cur_bits, key, &bds[0]); + if (likely(hs->hs_rehash_buckets == NULL)) { + /* no rehash or not rehashing */ + bds[1].bd_bucket = NULL; + return; + } + + LASSERT(hs->hs_rehash_bits != 0); + cfs_hash_bd_from_key(hs, hs->hs_rehash_buckets, + hs->hs_rehash_bits, key, &bds[1]); + + cfs_hash_bd_order(&bds[0], &bds[1]); +} +EXPORT_SYMBOL(cfs_hash_dual_bd_get); + +void +cfs_hash_dual_bd_lock(cfs_hash_t *hs, cfs_hash_bd_t *bds, int excl) +{ + cfs_hash_multi_bd_lock(hs, bds, 2, excl); +} +EXPORT_SYMBOL(cfs_hash_dual_bd_lock); + +void +cfs_hash_dual_bd_unlock(cfs_hash_t *hs, cfs_hash_bd_t *bds, int excl) +{ + cfs_hash_multi_bd_unlock(hs, bds, 2, excl); +} +EXPORT_SYMBOL(cfs_hash_dual_bd_unlock); + +struct hlist_node * +cfs_hash_dual_bd_lookup_locked(cfs_hash_t *hs, cfs_hash_bd_t *bds, + const void *key) +{ + return cfs_hash_multi_bd_lookup_locked(hs, bds, 2, key); +} +EXPORT_SYMBOL(cfs_hash_dual_bd_lookup_locked); + +struct hlist_node * +cfs_hash_dual_bd_findadd_locked(cfs_hash_t *hs, cfs_hash_bd_t *bds, + const void *key, struct hlist_node *hnode, + int noref) +{ + return cfs_hash_multi_bd_findadd_locked(hs, bds, 2, key, + hnode, noref); +} +EXPORT_SYMBOL(cfs_hash_dual_bd_findadd_locked); + +struct hlist_node * +cfs_hash_dual_bd_finddel_locked(cfs_hash_t *hs, cfs_hash_bd_t *bds, + const void *key, struct hlist_node *hnode) +{ + return cfs_hash_multi_bd_finddel_locked(hs, bds, 2, key, hnode); +} +EXPORT_SYMBOL(cfs_hash_dual_bd_finddel_locked); + +static void +cfs_hash_buckets_free(cfs_hash_bucket_t **buckets, + int bkt_size, int prev_size, int size) +{ + int i; + + for (i = prev_size; i < size; i++) { + if (buckets[i] != NULL) + LIBCFS_FREE(buckets[i], bkt_size); + } + + LIBCFS_FREE(buckets, sizeof(buckets[0]) * size); +} + +/* + * Create or grow bucket memory. Return old_buckets if no allocation was + * needed, the newly allocated buckets if allocation was needed and + * successful, and NULL on error. + */ +static cfs_hash_bucket_t ** +cfs_hash_buckets_realloc(cfs_hash_t *hs, cfs_hash_bucket_t **old_bkts, + unsigned int old_size, unsigned int new_size) +{ + cfs_hash_bucket_t **new_bkts; + int i; + + LASSERT(old_size == 0 || old_bkts != NULL); + + if (old_bkts != NULL && old_size == new_size) + return old_bkts; + + LIBCFS_ALLOC(new_bkts, sizeof(new_bkts[0]) * new_size); + if (new_bkts == NULL) + return NULL; + + if (old_bkts != NULL) { + memcpy(new_bkts, old_bkts, + min(old_size, new_size) * sizeof(*old_bkts)); + } + + for (i = old_size; i < new_size; i++) { + struct hlist_head *hhead; + cfs_hash_bd_t bd; + + LIBCFS_ALLOC(new_bkts[i], cfs_hash_bkt_size(hs)); + if (new_bkts[i] == NULL) { + cfs_hash_buckets_free(new_bkts, cfs_hash_bkt_size(hs), + old_size, new_size); + return NULL; + } + + new_bkts[i]->hsb_index = i; + new_bkts[i]->hsb_version = 1; /* shouldn't be zero */ + new_bkts[i]->hsb_depmax = -1; /* unknown */ + bd.bd_bucket = new_bkts[i]; + cfs_hash_bd_for_each_hlist(hs, &bd, hhead) + INIT_HLIST_HEAD(hhead); + + if (cfs_hash_with_no_lock(hs) || + cfs_hash_with_no_bktlock(hs)) + continue; + + if (cfs_hash_with_rw_bktlock(hs)) + rwlock_init(&new_bkts[i]->hsb_lock.rw); + else if (cfs_hash_with_spin_bktlock(hs)) + spin_lock_init(&new_bkts[i]->hsb_lock.spin); + else + LBUG(); /* invalid use-case */ + } + return new_bkts; +} + +/** + * Initialize new libcfs hash, where: + * @name - Descriptive hash name + * @cur_bits - Initial hash table size, in bits + * @max_bits - Maximum allowed hash table resize, in bits + * @ops - Registered hash table operations + * @flags - CFS_HASH_REHASH enable synamic hash resizing + * - CFS_HASH_SORT enable chained hash sort + */ +static int cfs_hash_rehash_worker(cfs_workitem_t *wi); + +#if CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1 +static int cfs_hash_dep_print(cfs_workitem_t *wi) +{ + cfs_hash_t *hs = container_of(wi, cfs_hash_t, hs_dep_wi); + int dep; + int bkt; + int off; + int bits; + + spin_lock(&hs->hs_dep_lock); + dep = hs->hs_dep_max; + bkt = hs->hs_dep_bkt; + off = hs->hs_dep_off; + bits = hs->hs_dep_bits; + spin_unlock(&hs->hs_dep_lock); + + LCONSOLE_WARN("#### HASH %s (bits: %d): max depth %d at bucket %d/%d\n", + hs->hs_name, bits, dep, bkt, off); + spin_lock(&hs->hs_dep_lock); + hs->hs_dep_bits = 0; /* mark as workitem done */ + spin_unlock(&hs->hs_dep_lock); + return 0; +} + +static void cfs_hash_depth_wi_init(cfs_hash_t *hs) +{ + spin_lock_init(&hs->hs_dep_lock); + cfs_wi_init(&hs->hs_dep_wi, hs, cfs_hash_dep_print); +} + +static void cfs_hash_depth_wi_cancel(cfs_hash_t *hs) +{ + if (cfs_wi_deschedule(cfs_sched_rehash, &hs->hs_dep_wi)) + return; + + spin_lock(&hs->hs_dep_lock); + while (hs->hs_dep_bits != 0) { + spin_unlock(&hs->hs_dep_lock); + cond_resched(); + spin_lock(&hs->hs_dep_lock); + } + spin_unlock(&hs->hs_dep_lock); +} + +#else /* CFS_HASH_DEBUG_LEVEL < CFS_HASH_DEBUG_1 */ + +static inline void cfs_hash_depth_wi_init(cfs_hash_t *hs) {} +static inline void cfs_hash_depth_wi_cancel(cfs_hash_t *hs) {} + +#endif /* CFS_HASH_DEBUG_LEVEL >= CFS_HASH_DEBUG_1 */ + +cfs_hash_t * +cfs_hash_create(char *name, unsigned cur_bits, unsigned max_bits, + unsigned bkt_bits, unsigned extra_bytes, + unsigned min_theta, unsigned max_theta, + cfs_hash_ops_t *ops, unsigned flags) +{ + cfs_hash_t *hs; + int len; + + ENTRY; + + CLASSERT(CFS_HASH_THETA_BITS < 15); + + LASSERT(name != NULL); + LASSERT(ops != NULL); + LASSERT(ops->hs_key); + LASSERT(ops->hs_hash); + LASSERT(ops->hs_object); + LASSERT(ops->hs_keycmp); + LASSERT(ops->hs_get != NULL); + LASSERT(ops->hs_put_locked != NULL); + + if ((flags & CFS_HASH_REHASH) != 0) + flags |= CFS_HASH_COUNTER; /* must have counter */ + + LASSERT(cur_bits > 0); + LASSERT(cur_bits >= bkt_bits); + LASSERT(max_bits >= cur_bits && max_bits < 31); + LASSERT(ergo((flags & CFS_HASH_REHASH) == 0, cur_bits == max_bits)); + LASSERT(ergo((flags & CFS_HASH_REHASH) != 0, + (flags & CFS_HASH_NO_LOCK) == 0)); + LASSERT(ergo((flags & CFS_HASH_REHASH_KEY) != 0, + ops->hs_keycpy != NULL)); + + len = (flags & CFS_HASH_BIGNAME) == 0 ? + CFS_HASH_NAME_LEN : CFS_HASH_BIGNAME_LEN; + LIBCFS_ALLOC(hs, offsetof(cfs_hash_t, hs_name[len])); + if (hs == NULL) + RETURN(NULL); + + strncpy(hs->hs_name, name, len); + hs->hs_name[len - 1] = '\0'; + hs->hs_flags = flags; + + atomic_set(&hs->hs_refcount, 1); + atomic_set(&hs->hs_count, 0); + + cfs_hash_lock_setup(hs); + cfs_hash_hlist_setup(hs); + + hs->hs_cur_bits = (__u8)cur_bits; + hs->hs_min_bits = (__u8)cur_bits; + hs->hs_max_bits = (__u8)max_bits; + hs->hs_bkt_bits = (__u8)bkt_bits; + + hs->hs_ops = ops; + hs->hs_extra_bytes = extra_bytes; + hs->hs_rehash_bits = 0; + cfs_wi_init(&hs->hs_rehash_wi, hs, cfs_hash_rehash_worker); + cfs_hash_depth_wi_init(hs); + + if (cfs_hash_with_rehash(hs)) + __cfs_hash_set_theta(hs, min_theta, max_theta); + + hs->hs_buckets = cfs_hash_buckets_realloc(hs, NULL, 0, + CFS_HASH_NBKT(hs)); + if (hs->hs_buckets != NULL) + return hs; + + LIBCFS_FREE(hs, offsetof(cfs_hash_t, hs_name[len])); + RETURN(NULL); +} +EXPORT_SYMBOL(cfs_hash_create); + +/** + * Cleanup libcfs hash @hs. + */ +static void +cfs_hash_destroy(cfs_hash_t *hs) +{ + struct hlist_node *hnode; + struct hlist_node *pos; + cfs_hash_bd_t bd; + int i; + ENTRY; + + LASSERT(hs != NULL); + LASSERT(!cfs_hash_is_exiting(hs) && + !cfs_hash_is_iterating(hs)); + + /** + * prohibit further rehashes, don't need any lock because + * I'm the only (last) one can change it. + */ + hs->hs_exiting = 1; + if (cfs_hash_with_rehash(hs)) + cfs_hash_rehash_cancel(hs); + + cfs_hash_depth_wi_cancel(hs); + /* rehash should be done/canceled */ + LASSERT(hs->hs_buckets != NULL && + hs->hs_rehash_buckets == NULL); + + cfs_hash_for_each_bucket(hs, &bd, i) { + struct hlist_head *hhead; + + LASSERT(bd.bd_bucket != NULL); + /* no need to take this lock, just for consistent code */ + cfs_hash_bd_lock(hs, &bd, 1); + + cfs_hash_bd_for_each_hlist(hs, &bd, hhead) { + hlist_for_each_safe(hnode, pos, hhead) { + LASSERTF(!cfs_hash_with_assert_empty(hs), + "hash %s bucket %u(%u) is not " + " empty: %u items left\n", + hs->hs_name, bd.bd_bucket->hsb_index, + bd.bd_offset, bd.bd_bucket->hsb_count); + /* can't assert key valicate, because we + * can interrupt rehash */ + cfs_hash_bd_del_locked(hs, &bd, hnode); + cfs_hash_exit(hs, hnode); + } + } + LASSERT(bd.bd_bucket->hsb_count == 0); + cfs_hash_bd_unlock(hs, &bd, 1); + cond_resched(); + } + + LASSERT(atomic_read(&hs->hs_count) == 0); + + cfs_hash_buckets_free(hs->hs_buckets, cfs_hash_bkt_size(hs), + 0, CFS_HASH_NBKT(hs)); + i = cfs_hash_with_bigname(hs) ? + CFS_HASH_BIGNAME_LEN : CFS_HASH_NAME_LEN; + LIBCFS_FREE(hs, offsetof(cfs_hash_t, hs_name[i])); + + EXIT; +} + +cfs_hash_t *cfs_hash_getref(cfs_hash_t *hs) +{ + if (atomic_inc_not_zero(&hs->hs_refcount)) + return hs; + return NULL; +} +EXPORT_SYMBOL(cfs_hash_getref); + +void cfs_hash_putref(cfs_hash_t *hs) +{ + if (atomic_dec_and_test(&hs->hs_refcount)) + cfs_hash_destroy(hs); +} +EXPORT_SYMBOL(cfs_hash_putref); + +static inline int +cfs_hash_rehash_bits(cfs_hash_t *hs) +{ + if (cfs_hash_with_no_lock(hs) || + !cfs_hash_with_rehash(hs)) + return -EOPNOTSUPP; + + if (unlikely(cfs_hash_is_exiting(hs))) + return -ESRCH; + + if (unlikely(cfs_hash_is_rehashing(hs))) + return -EALREADY; + + if (unlikely(cfs_hash_is_iterating(hs))) + return -EAGAIN; + + /* XXX: need to handle case with max_theta != 2.0 + * and the case with min_theta != 0.5 */ + if ((hs->hs_cur_bits < hs->hs_max_bits) && + (__cfs_hash_theta(hs) > hs->hs_max_theta)) + return hs->hs_cur_bits + 1; + + if (!cfs_hash_with_shrink(hs)) + return 0; + + if ((hs->hs_cur_bits > hs->hs_min_bits) && + (__cfs_hash_theta(hs) < hs->hs_min_theta)) + return hs->hs_cur_bits - 1; + + return 0; +} + +/** + * don't allow inline rehash if: + * - user wants non-blocking change (add/del) on hash table + * - too many elements + */ +static inline int +cfs_hash_rehash_inline(cfs_hash_t *hs) +{ + return !cfs_hash_with_nblk_change(hs) && + atomic_read(&hs->hs_count) < CFS_HASH_LOOP_HOG; +} + +/** + * Add item @hnode to libcfs hash @hs using @key. The registered + * ops->hs_get function will be called when the item is added. + */ +void +cfs_hash_add(cfs_hash_t *hs, const void *key, struct hlist_node *hnode) +{ + cfs_hash_bd_t bd; + int bits; + + LASSERT(hlist_unhashed(hnode)); + + cfs_hash_lock(hs, 0); + cfs_hash_bd_get_and_lock(hs, key, &bd, 1); + + cfs_hash_key_validate(hs, key, hnode); + cfs_hash_bd_add_locked(hs, &bd, hnode); + + cfs_hash_bd_unlock(hs, &bd, 1); + + bits = cfs_hash_rehash_bits(hs); + cfs_hash_unlock(hs, 0); + if (bits > 0) + cfs_hash_rehash(hs, cfs_hash_rehash_inline(hs)); +} +EXPORT_SYMBOL(cfs_hash_add); + +static struct hlist_node * +cfs_hash_find_or_add(cfs_hash_t *hs, const void *key, + struct hlist_node *hnode, int noref) +{ + struct hlist_node *ehnode; + cfs_hash_bd_t bds[2]; + int bits = 0; + + LASSERT(hlist_unhashed(hnode)); + + cfs_hash_lock(hs, 0); + cfs_hash_dual_bd_get_and_lock(hs, key, bds, 1); + + cfs_hash_key_validate(hs, key, hnode); + ehnode = cfs_hash_dual_bd_findadd_locked(hs, bds, key, + hnode, noref); + cfs_hash_dual_bd_unlock(hs, bds, 1); + + if (ehnode == hnode) /* new item added */ + bits = cfs_hash_rehash_bits(hs); + cfs_hash_unlock(hs, 0); + if (bits > 0) + cfs_hash_rehash(hs, cfs_hash_rehash_inline(hs)); + + return ehnode; +} + +/** + * Add item @hnode to libcfs hash @hs using @key. The registered + * ops->hs_get function will be called if the item was added. + * Returns 0 on success or -EALREADY on key collisions. + */ +int +cfs_hash_add_unique(cfs_hash_t *hs, const void *key, struct hlist_node *hnode) +{ + return cfs_hash_find_or_add(hs, key, hnode, 1) != hnode ? + -EALREADY : 0; +} +EXPORT_SYMBOL(cfs_hash_add_unique); + +/** + * Add item @hnode to libcfs hash @hs using @key. If this @key + * already exists in the hash then ops->hs_get will be called on the + * conflicting entry and that entry will be returned to the caller. + * Otherwise ops->hs_get is called on the item which was added. + */ +void * +cfs_hash_findadd_unique(cfs_hash_t *hs, const void *key, + struct hlist_node *hnode) +{ + hnode = cfs_hash_find_or_add(hs, key, hnode, 0); + + return cfs_hash_object(hs, hnode); +} +EXPORT_SYMBOL(cfs_hash_findadd_unique); + +/** + * Delete item @hnode from the libcfs hash @hs using @key. The @key + * is required to ensure the correct hash bucket is locked since there + * is no direct linkage from the item to the bucket. The object + * removed from the hash will be returned and obs->hs_put is called + * on the removed object. + */ +void * +cfs_hash_del(cfs_hash_t *hs, const void *key, struct hlist_node *hnode) +{ + void *obj = NULL; + int bits = 0; + cfs_hash_bd_t bds[2]; + + cfs_hash_lock(hs, 0); + cfs_hash_dual_bd_get_and_lock(hs, key, bds, 1); + + /* NB: do nothing if @hnode is not in hash table */ + if (hnode == NULL || !hlist_unhashed(hnode)) { + if (bds[1].bd_bucket == NULL && hnode != NULL) { + cfs_hash_bd_del_locked(hs, &bds[0], hnode); + } else { + hnode = cfs_hash_dual_bd_finddel_locked(hs, bds, + key, hnode); + } + } + + if (hnode != NULL) { + obj = cfs_hash_object(hs, hnode); + bits = cfs_hash_rehash_bits(hs); + } + + cfs_hash_dual_bd_unlock(hs, bds, 1); + cfs_hash_unlock(hs, 0); + if (bits > 0) + cfs_hash_rehash(hs, cfs_hash_rehash_inline(hs)); + + return obj; +} +EXPORT_SYMBOL(cfs_hash_del); + +/** + * Delete item given @key in libcfs hash @hs. The first @key found in + * the hash will be removed, if the key exists multiple times in the hash + * @hs this function must be called once per key. The removed object + * will be returned and ops->hs_put is called on the removed object. + */ +void * +cfs_hash_del_key(cfs_hash_t *hs, const void *key) +{ + return cfs_hash_del(hs, key, NULL); +} +EXPORT_SYMBOL(cfs_hash_del_key); + +/** + * Lookup an item using @key in the libcfs hash @hs and return it. + * If the @key is found in the hash hs->hs_get() is called and the + * matching objects is returned. It is the callers responsibility + * to call the counterpart ops->hs_put using the cfs_hash_put() macro + * when when finished with the object. If the @key was not found + * in the hash @hs NULL is returned. + */ +void * +cfs_hash_lookup(cfs_hash_t *hs, const void *key) +{ + void *obj = NULL; + struct hlist_node *hnode; + cfs_hash_bd_t bds[2]; + + cfs_hash_lock(hs, 0); + cfs_hash_dual_bd_get_and_lock(hs, key, bds, 0); + + hnode = cfs_hash_dual_bd_lookup_locked(hs, bds, key); + if (hnode != NULL) + obj = cfs_hash_object(hs, hnode); + + cfs_hash_dual_bd_unlock(hs, bds, 0); + cfs_hash_unlock(hs, 0); + + return obj; +} +EXPORT_SYMBOL(cfs_hash_lookup); + +static void +cfs_hash_for_each_enter(cfs_hash_t *hs) +{ + LASSERT(!cfs_hash_is_exiting(hs)); + + if (!cfs_hash_with_rehash(hs)) + return; + /* + * NB: it's race on cfs_has_t::hs_iterating, but doesn't matter + * because it's just an unreliable signal to rehash-thread, + * rehash-thread will try to finsih rehash ASAP when seeing this. + */ + hs->hs_iterating = 1; + + cfs_hash_lock(hs, 1); + hs->hs_iterators++; + + /* NB: iteration is mostly called by service thread, + * we tend to cancel pending rehash-requst, instead of + * blocking service thread, we will relaunch rehash request + * after iteration */ + if (cfs_hash_is_rehashing(hs)) + cfs_hash_rehash_cancel_locked(hs); + cfs_hash_unlock(hs, 1); +} + +static void +cfs_hash_for_each_exit(cfs_hash_t *hs) +{ + int remained; + int bits; + + if (!cfs_hash_with_rehash(hs)) + return; + cfs_hash_lock(hs, 1); + remained = --hs->hs_iterators; + bits = cfs_hash_rehash_bits(hs); + cfs_hash_unlock(hs, 1); + /* NB: it's race on cfs_has_t::hs_iterating, see above */ + if (remained == 0) + hs->hs_iterating = 0; + if (bits > 0) { + cfs_hash_rehash(hs, atomic_read(&hs->hs_count) < + CFS_HASH_LOOP_HOG); + } +} + +/** + * For each item in the libcfs hash @hs call the passed callback @func + * and pass to it as an argument each hash item and the private @data. + * + * a) the function may sleep! + * b) during the callback: + * . the bucket lock is held so the callback must never sleep. + * . if @removal_safe is true, use can remove current item by + * cfs_hash_bd_del_locked + */ +static __u64 +cfs_hash_for_each_tight(cfs_hash_t *hs, cfs_hash_for_each_cb_t func, + void *data, int remove_safe) +{ + struct hlist_node *hnode; + struct hlist_node *pos; + cfs_hash_bd_t bd; + __u64 count = 0; + int excl = !!remove_safe; + int loop = 0; + int i; + ENTRY; + + cfs_hash_for_each_enter(hs); + + cfs_hash_lock(hs, 0); + LASSERT(!cfs_hash_is_rehashing(hs)); + + cfs_hash_for_each_bucket(hs, &bd, i) { + struct hlist_head *hhead; + + cfs_hash_bd_lock(hs, &bd, excl); + if (func == NULL) { /* only glimpse size */ + count += bd.bd_bucket->hsb_count; + cfs_hash_bd_unlock(hs, &bd, excl); + continue; + } + + cfs_hash_bd_for_each_hlist(hs, &bd, hhead) { + hlist_for_each_safe(hnode, pos, hhead) { + cfs_hash_bucket_validate(hs, &bd, hnode); + count++; + loop++; + if (func(hs, &bd, hnode, data)) { + cfs_hash_bd_unlock(hs, &bd, excl); + goto out; + } + } + } + cfs_hash_bd_unlock(hs, &bd, excl); + if (loop < CFS_HASH_LOOP_HOG) + continue; + loop = 0; + cfs_hash_unlock(hs, 0); + cond_resched(); + cfs_hash_lock(hs, 0); + } + out: + cfs_hash_unlock(hs, 0); + + cfs_hash_for_each_exit(hs); + RETURN(count); +} + +typedef struct { + cfs_hash_cond_opt_cb_t func; + void *arg; +} cfs_hash_cond_arg_t; + +static int +cfs_hash_cond_del_locked(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnode, void *data) +{ + cfs_hash_cond_arg_t *cond = data; + + if (cond->func(cfs_hash_object(hs, hnode), cond->arg)) + cfs_hash_bd_del_locked(hs, bd, hnode); + return 0; +} + +/** + * Delete item from the libcfs hash @hs when @func return true. + * The write lock being hold during loop for each bucket to avoid + * any object be reference. + */ +void +cfs_hash_cond_del(cfs_hash_t *hs, cfs_hash_cond_opt_cb_t func, void *data) +{ + cfs_hash_cond_arg_t arg = { + .func = func, + .arg = data, + }; + + cfs_hash_for_each_tight(hs, cfs_hash_cond_del_locked, &arg, 1); +} +EXPORT_SYMBOL(cfs_hash_cond_del); + +void +cfs_hash_for_each(cfs_hash_t *hs, + cfs_hash_for_each_cb_t func, void *data) +{ + cfs_hash_for_each_tight(hs, func, data, 0); +} +EXPORT_SYMBOL(cfs_hash_for_each); + +void +cfs_hash_for_each_safe(cfs_hash_t *hs, + cfs_hash_for_each_cb_t func, void *data) +{ + cfs_hash_for_each_tight(hs, func, data, 1); +} +EXPORT_SYMBOL(cfs_hash_for_each_safe); + +static int +cfs_hash_peek(cfs_hash_t *hs, cfs_hash_bd_t *bd, + struct hlist_node *hnode, void *data) +{ + *(int *)data = 0; + return 1; /* return 1 to break the loop */ +} + +int +cfs_hash_is_empty(cfs_hash_t *hs) +{ + int empty = 1; + + cfs_hash_for_each_tight(hs, cfs_hash_peek, &empty, 0); + return empty; +} +EXPORT_SYMBOL(cfs_hash_is_empty); + +__u64 +cfs_hash_size_get(cfs_hash_t *hs) +{ + return cfs_hash_with_counter(hs) ? + atomic_read(&hs->hs_count) : + cfs_hash_for_each_tight(hs, NULL, NULL, 0); +} +EXPORT_SYMBOL(cfs_hash_size_get); + +/* + * cfs_hash_for_each_relax: + * Iterate the hash table and call @func on each item without + * any lock. This function can't guarantee to finish iteration + * if these features are enabled: + * + * a. if rehash_key is enabled, an item can be moved from + * one bucket to another bucket + * b. user can remove non-zero-ref item from hash-table, + * so the item can be removed from hash-table, even worse, + * it's possible that user changed key and insert to another + * hash bucket. + * there's no way for us to finish iteration correctly on previous + * two cases, so iteration has to be stopped on change. + */ +static int +cfs_hash_for_each_relax(cfs_hash_t *hs, cfs_hash_for_each_cb_t func, void *data) +{ + struct hlist_node *hnode; + struct hlist_node *tmp; + cfs_hash_bd_t bd; + __u32 version; + int count = 0; + int stop_on_change; + int rc; + int i; + ENTRY; + + stop_on_change = cfs_hash_with_rehash_key(hs) || + !cfs_hash_with_no_itemref(hs) || + CFS_HOP(hs, put_locked) == NULL; + cfs_hash_lock(hs, 0); + LASSERT(!cfs_hash_is_rehashing(hs)); + + cfs_hash_for_each_bucket(hs, &bd, i) { + struct hlist_head *hhead; + + cfs_hash_bd_lock(hs, &bd, 0); + version = cfs_hash_bd_version_get(&bd); + + cfs_hash_bd_for_each_hlist(hs, &bd, hhead) { + for (hnode = hhead->first; hnode != NULL;) { + cfs_hash_bucket_validate(hs, &bd, hnode); + cfs_hash_get(hs, hnode); + cfs_hash_bd_unlock(hs, &bd, 0); + cfs_hash_unlock(hs, 0); + + rc = func(hs, &bd, hnode, data); + if (stop_on_change) + cfs_hash_put(hs, hnode); + cond_resched(); + count++; + + cfs_hash_lock(hs, 0); + cfs_hash_bd_lock(hs, &bd, 0); + if (!stop_on_change) { + tmp = hnode->next; + cfs_hash_put_locked(hs, hnode); + hnode = tmp; + } else { /* bucket changed? */ + if (version != + cfs_hash_bd_version_get(&bd)) + break; + /* safe to continue because no change */ + hnode = hnode->next; + } + if (rc) /* callback wants to break iteration */ + break; + } + } + cfs_hash_bd_unlock(hs, &bd, 0); + } + cfs_hash_unlock(hs, 0); + + return count; +} + +int +cfs_hash_for_each_nolock(cfs_hash_t *hs, + cfs_hash_for_each_cb_t func, void *data) +{ + ENTRY; + + if (cfs_hash_with_no_lock(hs) || + cfs_hash_with_rehash_key(hs) || + !cfs_hash_with_no_itemref(hs)) + RETURN(-EOPNOTSUPP); + + if (CFS_HOP(hs, get) == NULL || + (CFS_HOP(hs, put) == NULL && + CFS_HOP(hs, put_locked) == NULL)) + RETURN(-EOPNOTSUPP); + + cfs_hash_for_each_enter(hs); + cfs_hash_for_each_relax(hs, func, data); + cfs_hash_for_each_exit(hs); + + RETURN(0); +} +EXPORT_SYMBOL(cfs_hash_for_each_nolock); + +/** + * For each hash bucket in the libcfs hash @hs call the passed callback + * @func until all the hash buckets are empty. The passed callback @func + * or the previously registered callback hs->hs_put must remove the item + * from the hash. You may either use the cfs_hash_del() or hlist_del() + * functions. No rwlocks will be held during the callback @func it is + * safe to sleep if needed. This function will not terminate until the + * hash is empty. Note it is still possible to concurrently add new + * items in to the hash. It is the callers responsibility to ensure + * the required locking is in place to prevent concurrent insertions. + */ +int +cfs_hash_for_each_empty(cfs_hash_t *hs, + cfs_hash_for_each_cb_t func, void *data) +{ + unsigned i = 0; + ENTRY; + + if (cfs_hash_with_no_lock(hs)) + return -EOPNOTSUPP; + + if (CFS_HOP(hs, get) == NULL || + (CFS_HOP(hs, put) == NULL && + CFS_HOP(hs, put_locked) == NULL)) + return -EOPNOTSUPP; + + cfs_hash_for_each_enter(hs); + while (cfs_hash_for_each_relax(hs, func, data)) { + CDEBUG(D_INFO, "Try to empty hash: %s, loop: %u\n", + hs->hs_name, i++); + } + cfs_hash_for_each_exit(hs); + RETURN(0); +} +EXPORT_SYMBOL(cfs_hash_for_each_empty); + +void +cfs_hash_hlist_for_each(cfs_hash_t *hs, unsigned hindex, + cfs_hash_for_each_cb_t func, void *data) +{ + struct hlist_head *hhead; + struct hlist_node *hnode; + cfs_hash_bd_t bd; + + cfs_hash_for_each_enter(hs); + cfs_hash_lock(hs, 0); + if (hindex >= CFS_HASH_NHLIST(hs)) + goto out; + + cfs_hash_bd_index_set(hs, hindex, &bd); + + cfs_hash_bd_lock(hs, &bd, 0); + hhead = cfs_hash_bd_hhead(hs, &bd); + hlist_for_each(hnode, hhead) { + if (func(hs, &bd, hnode, data)) + break; + } + cfs_hash_bd_unlock(hs, &bd, 0); + out: + cfs_hash_unlock(hs, 0); + cfs_hash_for_each_exit(hs); +} + +EXPORT_SYMBOL(cfs_hash_hlist_for_each); + +/* + * For each item in the libcfs hash @hs which matches the @key call + * the passed callback @func and pass to it as an argument each hash + * item and the private @data. During the callback the bucket lock + * is held so the callback must never sleep. + */ +void +cfs_hash_for_each_key(cfs_hash_t *hs, const void *key, + cfs_hash_for_each_cb_t func, void *data) +{ + struct hlist_node *hnode; + cfs_hash_bd_t bds[2]; + unsigned i; + + cfs_hash_lock(hs, 0); + + cfs_hash_dual_bd_get_and_lock(hs, key, bds, 0); + + cfs_hash_for_each_bd(bds, 2, i) { + struct hlist_head *hlist = cfs_hash_bd_hhead(hs, &bds[i]); + + hlist_for_each(hnode, hlist) { + cfs_hash_bucket_validate(hs, &bds[i], hnode); + + if (cfs_hash_keycmp(hs, key, hnode)) { + if (func(hs, &bds[i], hnode, data)) + break; + } + } + } + + cfs_hash_dual_bd_unlock(hs, bds, 0); + cfs_hash_unlock(hs, 0); +} +EXPORT_SYMBOL(cfs_hash_for_each_key); + +/** + * Rehash the libcfs hash @hs to the given @bits. This can be used + * to grow the hash size when excessive chaining is detected, or to + * shrink the hash when it is larger than needed. When the CFS_HASH_REHASH + * flag is set in @hs the libcfs hash may be dynamically rehashed + * during addition or removal if the hash's theta value exceeds + * either the hs->hs_min_theta or hs->max_theta values. By default + * these values are tuned to keep the chained hash depth small, and + * this approach assumes a reasonably uniform hashing function. The + * theta thresholds for @hs are tunable via cfs_hash_set_theta(). + */ +void +cfs_hash_rehash_cancel_locked(cfs_hash_t *hs) +{ + int i; + + /* need hold cfs_hash_lock(hs, 1) */ + LASSERT(cfs_hash_with_rehash(hs) && + !cfs_hash_with_no_lock(hs)); + + if (!cfs_hash_is_rehashing(hs)) + return; + + if (cfs_wi_deschedule(cfs_sched_rehash, &hs->hs_rehash_wi)) { + hs->hs_rehash_bits = 0; + return; + } + + for (i = 2; cfs_hash_is_rehashing(hs); i++) { + cfs_hash_unlock(hs, 1); + /* raise console warning while waiting too long */ + CDEBUG(IS_PO2(i >> 3) ? D_WARNING : D_INFO, + "hash %s is still rehashing, rescheded %d\n", + hs->hs_name, i - 1); + cond_resched(); + cfs_hash_lock(hs, 1); + } +} +EXPORT_SYMBOL(cfs_hash_rehash_cancel_locked); + +void +cfs_hash_rehash_cancel(cfs_hash_t *hs) +{ + cfs_hash_lock(hs, 1); + cfs_hash_rehash_cancel_locked(hs); + cfs_hash_unlock(hs, 1); +} +EXPORT_SYMBOL(cfs_hash_rehash_cancel); + +int +cfs_hash_rehash(cfs_hash_t *hs, int do_rehash) +{ + int rc; + + LASSERT(cfs_hash_with_rehash(hs) && !cfs_hash_with_no_lock(hs)); + + cfs_hash_lock(hs, 1); + + rc = cfs_hash_rehash_bits(hs); + if (rc <= 0) { + cfs_hash_unlock(hs, 1); + return rc; + } + + hs->hs_rehash_bits = rc; + if (!do_rehash) { + /* launch and return */ + cfs_wi_schedule(cfs_sched_rehash, &hs->hs_rehash_wi); + cfs_hash_unlock(hs, 1); + return 0; + } + + /* rehash right now */ + cfs_hash_unlock(hs, 1); + + return cfs_hash_rehash_worker(&hs->hs_rehash_wi); +} +EXPORT_SYMBOL(cfs_hash_rehash); + +static int +cfs_hash_rehash_bd(cfs_hash_t *hs, cfs_hash_bd_t *old) +{ + cfs_hash_bd_t new; + struct hlist_head *hhead; + struct hlist_node *hnode; + struct hlist_node *pos; + void *key; + int c = 0; + + /* hold cfs_hash_lock(hs, 1), so don't need any bucket lock */ + cfs_hash_bd_for_each_hlist(hs, old, hhead) { + hlist_for_each_safe(hnode, pos, hhead) { + key = cfs_hash_key(hs, hnode); + LASSERT(key != NULL); + /* Validate hnode is in the correct bucket. */ + cfs_hash_bucket_validate(hs, old, hnode); + /* + * Delete from old hash bucket; move to new bucket. + * ops->hs_key must be defined. + */ + cfs_hash_bd_from_key(hs, hs->hs_rehash_buckets, + hs->hs_rehash_bits, key, &new); + cfs_hash_bd_move_locked(hs, old, &new, hnode); + c++; + } + } + + return c; +} + +static int +cfs_hash_rehash_worker(cfs_workitem_t *wi) +{ + cfs_hash_t *hs = container_of(wi, cfs_hash_t, hs_rehash_wi); + cfs_hash_bucket_t **bkts; + cfs_hash_bd_t bd; + unsigned int old_size; + unsigned int new_size; + int bsize; + int count = 0; + int rc = 0; + int i; + + LASSERT (hs != NULL && cfs_hash_with_rehash(hs)); + + cfs_hash_lock(hs, 0); + LASSERT(cfs_hash_is_rehashing(hs)); + + old_size = CFS_HASH_NBKT(hs); + new_size = CFS_HASH_RH_NBKT(hs); + + cfs_hash_unlock(hs, 0); + + /* + * don't need hs::hs_rwlock for hs::hs_buckets, + * because nobody can change bkt-table except me. + */ + bkts = cfs_hash_buckets_realloc(hs, hs->hs_buckets, + old_size, new_size); + cfs_hash_lock(hs, 1); + if (bkts == NULL) { + rc = -ENOMEM; + goto out; + } + + if (bkts == hs->hs_buckets) { + bkts = NULL; /* do nothing */ + goto out; + } + + rc = __cfs_hash_theta(hs); + if ((rc >= hs->hs_min_theta) && (rc <= hs->hs_max_theta)) { + /* free the new allocated bkt-table */ + old_size = new_size; + new_size = CFS_HASH_NBKT(hs); + rc = -EALREADY; + goto out; + } + + LASSERT(hs->hs_rehash_buckets == NULL); + hs->hs_rehash_buckets = bkts; + + rc = 0; + cfs_hash_for_each_bucket(hs, &bd, i) { + if (cfs_hash_is_exiting(hs)) { + rc = -ESRCH; + /* someone wants to destroy the hash, abort now */ + if (old_size < new_size) /* OK to free old bkt-table */ + break; + /* it's shrinking, need free new bkt-table */ + hs->hs_rehash_buckets = NULL; + old_size = new_size; + new_size = CFS_HASH_NBKT(hs); + goto out; + } + + count += cfs_hash_rehash_bd(hs, &bd); + if (count < CFS_HASH_LOOP_HOG || + cfs_hash_is_iterating(hs)) { /* need to finish ASAP */ + continue; + } + + count = 0; + cfs_hash_unlock(hs, 1); + cond_resched(); + cfs_hash_lock(hs, 1); + } + + hs->hs_rehash_count++; + + bkts = hs->hs_buckets; + hs->hs_buckets = hs->hs_rehash_buckets; + hs->hs_rehash_buckets = NULL; + + hs->hs_cur_bits = hs->hs_rehash_bits; + out: + hs->hs_rehash_bits = 0; + if (rc == -ESRCH) /* never be scheduled again */ + cfs_wi_exit(cfs_sched_rehash, wi); + bsize = cfs_hash_bkt_size(hs); + cfs_hash_unlock(hs, 1); + /* can't refer to @hs anymore because it could be destroyed */ + if (bkts != NULL) + cfs_hash_buckets_free(bkts, bsize, new_size, old_size); + if (rc != 0) + CDEBUG(D_INFO, "early quit of of rehashing: %d\n", rc); + /* return 1 only if cfs_wi_exit is called */ + return rc == -ESRCH; +} + +/** + * Rehash the object referenced by @hnode in the libcfs hash @hs. The + * @old_key must be provided to locate the objects previous location + * in the hash, and the @new_key will be used to reinsert the object. + * Use this function instead of a cfs_hash_add() + cfs_hash_del() + * combo when it is critical that there is no window in time where the + * object is missing from the hash. When an object is being rehashed + * the registered cfs_hash_get() and cfs_hash_put() functions will + * not be called. + */ +void cfs_hash_rehash_key(cfs_hash_t *hs, const void *old_key, + void *new_key, struct hlist_node *hnode) +{ + cfs_hash_bd_t bds[3]; + cfs_hash_bd_t old_bds[2]; + cfs_hash_bd_t new_bd; + + LASSERT(!hlist_unhashed(hnode)); + + cfs_hash_lock(hs, 0); + + cfs_hash_dual_bd_get(hs, old_key, old_bds); + cfs_hash_bd_get(hs, new_key, &new_bd); + + bds[0] = old_bds[0]; + bds[1] = old_bds[1]; + bds[2] = new_bd; + + /* NB: bds[0] and bds[1] are ordered already */ + cfs_hash_bd_order(&bds[1], &bds[2]); + cfs_hash_bd_order(&bds[0], &bds[1]); + + cfs_hash_multi_bd_lock(hs, bds, 3, 1); + if (likely(old_bds[1].bd_bucket == NULL)) { + cfs_hash_bd_move_locked(hs, &old_bds[0], &new_bd, hnode); + } else { + cfs_hash_dual_bd_finddel_locked(hs, old_bds, old_key, hnode); + cfs_hash_bd_add_locked(hs, &new_bd, hnode); + } + /* overwrite key inside locks, otherwise may screw up with + * other operations, i.e: rehash */ + cfs_hash_keycpy(hs, new_key, hnode); + + cfs_hash_multi_bd_unlock(hs, bds, 3, 1); + cfs_hash_unlock(hs, 0); +} +EXPORT_SYMBOL(cfs_hash_rehash_key); + +int cfs_hash_debug_header(struct seq_file *m) +{ + return seq_printf(m, "%-*s%6s%6s%6s%6s%6s%6s%6s%7s%8s%8s%8s%s\n", + CFS_HASH_BIGNAME_LEN, + "name", "cur", "min", "max", "theta", "t-min", "t-max", + "flags", "rehash", "count", "maxdep", "maxdepb", + " distribution"); +} +EXPORT_SYMBOL(cfs_hash_debug_header); + +static cfs_hash_bucket_t ** +cfs_hash_full_bkts(cfs_hash_t *hs) +{ + /* NB: caller should hold hs->hs_rwlock if REHASH is set */ + if (hs->hs_rehash_buckets == NULL) + return hs->hs_buckets; + + LASSERT(hs->hs_rehash_bits != 0); + return hs->hs_rehash_bits > hs->hs_cur_bits ? + hs->hs_rehash_buckets : hs->hs_buckets; +} + +static unsigned int +cfs_hash_full_nbkt(cfs_hash_t *hs) +{ + /* NB: caller should hold hs->hs_rwlock if REHASH is set */ + if (hs->hs_rehash_buckets == NULL) + return CFS_HASH_NBKT(hs); + + LASSERT(hs->hs_rehash_bits != 0); + return hs->hs_rehash_bits > hs->hs_cur_bits ? + CFS_HASH_RH_NBKT(hs) : CFS_HASH_NBKT(hs); +} + +int cfs_hash_debug_str(cfs_hash_t *hs, struct seq_file *m) +{ + int dist[8] = { 0, }; + int maxdep = -1; + int maxdepb = -1; + int total = 0; + int theta; + int i; + + cfs_hash_lock(hs, 0); + theta = __cfs_hash_theta(hs); + + seq_printf(m, "%-*s %5d %5d %5d %d.%03d %d.%03d %d.%03d 0x%02x %6d ", + CFS_HASH_BIGNAME_LEN, hs->hs_name, + 1 << hs->hs_cur_bits, 1 << hs->hs_min_bits, + 1 << hs->hs_max_bits, + __cfs_hash_theta_int(theta), __cfs_hash_theta_frac(theta), + __cfs_hash_theta_int(hs->hs_min_theta), + __cfs_hash_theta_frac(hs->hs_min_theta), + __cfs_hash_theta_int(hs->hs_max_theta), + __cfs_hash_theta_frac(hs->hs_max_theta), + hs->hs_flags, hs->hs_rehash_count); + + /* + * The distribution is a summary of the chained hash depth in + * each of the libcfs hash buckets. Each buckets hsb_count is + * divided by the hash theta value and used to generate a + * histogram of the hash distribution. A uniform hash will + * result in all hash buckets being close to the average thus + * only the first few entries in the histogram will be non-zero. + * If you hash function results in a non-uniform hash the will + * be observable by outlier bucks in the distribution histogram. + * + * Uniform hash distribution: 128/128/0/0/0/0/0/0 + * Non-Uniform hash distribution: 128/125/0/0/0/0/2/1 + */ + for (i = 0; i < cfs_hash_full_nbkt(hs); i++) { + cfs_hash_bd_t bd; + + bd.bd_bucket = cfs_hash_full_bkts(hs)[i]; + cfs_hash_bd_lock(hs, &bd, 0); + if (maxdep < bd.bd_bucket->hsb_depmax) { + maxdep = bd.bd_bucket->hsb_depmax; + maxdepb = ffz(~maxdep); + } + total += bd.bd_bucket->hsb_count; + dist[min(__cfs_fls(bd.bd_bucket->hsb_count/max(theta,1)),7)]++; + cfs_hash_bd_unlock(hs, &bd, 0); + } + + seq_printf(m, "%7d %7d %7d ", total, maxdep, maxdepb); + for (i = 0; i < 8; i++) + seq_printf(m, "%d%c", dist[i], (i == 7) ? '\n' : '/'); + + cfs_hash_unlock(hs, 0); + + return 0; +} +EXPORT_SYMBOL(cfs_hash_debug_str); diff --git a/drivers/staging/lustre/lustre/libcfs/heap.c b/drivers/staging/lustre/lustre/libcfs/heap.c new file mode 100644 index 000000000000..147e4fe4762d --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/heap.c @@ -0,0 +1,475 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + + * This program is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License version 2 for more details. A copy is + * included in the COPYING file that accompanied this code. + + * You should have received a copy of the GNU General Public License + * along with this program; if not, write to the Free Software + * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + * + * GPL HEADER END + */ +/* + * Copyright (c) 2011 Intel Corporation + */ +/* + * libcfs/libcfs/heap.c + * + * Author: Eric Barton <eeb@whamcloud.com> + * Liang Zhen <liang@whamcloud.com> + */ +/** \addtogroup heap + * + * @{ + */ + +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> + +#define CBH_ALLOC(ptr, h) \ +do { \ + if ((h)->cbh_flags & CBH_FLAG_ATOMIC_GROW) \ + LIBCFS_CPT_ALLOC_GFP((ptr), h->cbh_cptab, h->cbh_cptid, \ + CBH_NOB, GFP_ATOMIC); \ + else \ + LIBCFS_CPT_ALLOC((ptr), h->cbh_cptab, h->cbh_cptid, \ + CBH_NOB); \ +} while (0) + +#define CBH_FREE(ptr) LIBCFS_FREE(ptr, CBH_NOB) + +/** + * Grows the capacity of a binary heap so that it can handle a larger number of + * \e cfs_binheap_node_t objects. + * + * \param[in] h The binary heap + * + * \retval 0 Successfully grew the heap + * \retval -ENOMEM OOM error + */ +static int +cfs_binheap_grow(cfs_binheap_t *h) +{ + cfs_binheap_node_t ***frag1 = NULL; + cfs_binheap_node_t **frag2; + int hwm = h->cbh_hwm; + + /* need a whole new chunk of pointers */ + LASSERT((h->cbh_hwm & CBH_MASK) == 0); + + if (hwm == 0) { + /* first use of single indirect */ + CBH_ALLOC(h->cbh_elements1, h); + if (h->cbh_elements1 == NULL) + return -ENOMEM; + + goto out; + } + + hwm -= CBH_SIZE; + if (hwm < CBH_SIZE * CBH_SIZE) { + /* not filled double indirect */ + CBH_ALLOC(frag2, h); + if (frag2 == NULL) + return -ENOMEM; + + if (hwm == 0) { + /* first use of double indirect */ + CBH_ALLOC(h->cbh_elements2, h); + if (h->cbh_elements2 == NULL) { + CBH_FREE(frag2); + return -ENOMEM; + } + } + + h->cbh_elements2[hwm >> CBH_SHIFT] = frag2; + goto out; + } + + hwm -= CBH_SIZE * CBH_SIZE; +#if (CBH_SHIFT * 3 < 32) + if (hwm >= CBH_SIZE * CBH_SIZE * CBH_SIZE) { + /* filled triple indirect */ + return -ENOMEM; + } +#endif + CBH_ALLOC(frag2, h); + if (frag2 == NULL) + return -ENOMEM; + + if (((hwm >> CBH_SHIFT) & CBH_MASK) == 0) { + /* first use of this 2nd level index */ + CBH_ALLOC(frag1, h); + if (frag1 == NULL) { + CBH_FREE(frag2); + return -ENOMEM; + } + } + + if (hwm == 0) { + /* first use of triple indirect */ + CBH_ALLOC(h->cbh_elements3, h); + if (h->cbh_elements3 == NULL) { + CBH_FREE(frag2); + CBH_FREE(frag1); + return -ENOMEM; + } + } + + if (frag1 != NULL) { + LASSERT(h->cbh_elements3[hwm >> (2 * CBH_SHIFT)] == NULL); + h->cbh_elements3[hwm >> (2 * CBH_SHIFT)] = frag1; + } else { + frag1 = h->cbh_elements3[hwm >> (2 * CBH_SHIFT)]; + LASSERT(frag1 != NULL); + } + + frag1[(hwm >> CBH_SHIFT) & CBH_MASK] = frag2; + + out: + h->cbh_hwm += CBH_SIZE; + return 0; +} + +/** + * Creates and initializes a binary heap instance. + * + * \param[in] ops The operations to be used + * \param[in] flags The heap flags + * \parm[in] count The initial heap capacity in # of elements + * \param[in] arg An optional private argument + * \param[in] cptab The CPT table this heap instance will operate over + * \param[in] cptid The CPT id of \a cptab this heap instance will operate over + * + * \retval valid-pointer A newly-created and initialized binary heap object + * \retval NULL error + */ +cfs_binheap_t * +cfs_binheap_create(cfs_binheap_ops_t *ops, unsigned int flags, + unsigned count, void *arg, struct cfs_cpt_table *cptab, + int cptid) +{ + cfs_binheap_t *h; + + LASSERT(ops != NULL); + LASSERT(ops->hop_compare != NULL); + LASSERT(cptab != NULL); + LASSERT(cptid == CFS_CPT_ANY || + (cptid >= 0 && cptid < cptab->ctb_nparts)); + + LIBCFS_CPT_ALLOC(h, cptab, cptid, sizeof(*h)); + if (h == NULL) + return NULL; + + h->cbh_ops = ops; + h->cbh_nelements = 0; + h->cbh_hwm = 0; + h->cbh_private = arg; + h->cbh_flags = flags & (~CBH_FLAG_ATOMIC_GROW); + h->cbh_cptab = cptab; + h->cbh_cptid = cptid; + + while (h->cbh_hwm < count) { /* preallocate */ + if (cfs_binheap_grow(h) != 0) { + cfs_binheap_destroy(h); + return NULL; + } + } + + h->cbh_flags |= flags & CBH_FLAG_ATOMIC_GROW; + + return h; +} +EXPORT_SYMBOL(cfs_binheap_create); + +/** + * Releases all resources associated with a binary heap instance. + * + * Deallocates memory for all indirection levels and the binary heap object + * itself. + * + * \param[in] h The binary heap object + */ +void +cfs_binheap_destroy(cfs_binheap_t *h) +{ + int idx0; + int idx1; + int n; + + LASSERT(h != NULL); + + n = h->cbh_hwm; + + if (n > 0) { + CBH_FREE(h->cbh_elements1); + n -= CBH_SIZE; + } + + if (n > 0) { + for (idx0 = 0; idx0 < CBH_SIZE && n > 0; idx0++) { + CBH_FREE(h->cbh_elements2[idx0]); + n -= CBH_SIZE; + } + + CBH_FREE(h->cbh_elements2); + } + + if (n > 0) { + for (idx0 = 0; idx0 < CBH_SIZE && n > 0; idx0++) { + + for (idx1 = 0; idx1 < CBH_SIZE && n > 0; idx1++) { + CBH_FREE(h->cbh_elements3[idx0][idx1]); + n -= CBH_SIZE; + } + + CBH_FREE(h->cbh_elements3[idx0]); + } + + CBH_FREE(h->cbh_elements3); + } + + LIBCFS_FREE(h, sizeof(*h)); +} +EXPORT_SYMBOL(cfs_binheap_destroy); + +/** + * Obtains a double pointer to a heap element, given its index into the binary + * tree. + * + * \param[in] h The binary heap instance + * \param[in] idx The requested node's index + * + * \retval valid-pointer A double pointer to a heap pointer entry + */ +static cfs_binheap_node_t ** +cfs_binheap_pointer(cfs_binheap_t *h, unsigned int idx) +{ + if (idx < CBH_SIZE) + return &(h->cbh_elements1[idx]); + + idx -= CBH_SIZE; + if (idx < CBH_SIZE * CBH_SIZE) + return &(h->cbh_elements2[idx >> CBH_SHIFT][idx & CBH_MASK]); + + idx -= CBH_SIZE * CBH_SIZE; + return &(h->cbh_elements3[idx >> (2 * CBH_SHIFT)]\ + [(idx >> CBH_SHIFT) & CBH_MASK]\ + [idx & CBH_MASK]); +} + +/** + * Obtains a pointer to a heap element, given its index into the binary tree. + * + * \param[in] h The binary heap + * \param[in] idx The requested node's index + * + * \retval valid-pointer The requested heap node + * \retval NULL Supplied index is out of bounds + */ +cfs_binheap_node_t * +cfs_binheap_find(cfs_binheap_t *h, unsigned int idx) +{ + if (idx >= h->cbh_nelements) + return NULL; + + return *cfs_binheap_pointer(h, idx); +} +EXPORT_SYMBOL(cfs_binheap_find); + +/** + * Moves a node upwards, towards the root of the binary tree. + * + * \param[in] h The heap + * \param[in] e The node + * + * \retval 1 The position of \a e in the tree was changed at least once + * \retval 0 The position of \a e in the tree was not changed + */ +static int +cfs_binheap_bubble(cfs_binheap_t *h, cfs_binheap_node_t *e) +{ + unsigned int cur_idx = e->chn_index; + cfs_binheap_node_t **cur_ptr; + unsigned int parent_idx; + cfs_binheap_node_t **parent_ptr; + int did_sth = 0; + + cur_ptr = cfs_binheap_pointer(h, cur_idx); + LASSERT(*cur_ptr == e); + + while (cur_idx > 0) { + parent_idx = (cur_idx - 1) >> 1; + + parent_ptr = cfs_binheap_pointer(h, parent_idx); + LASSERT((*parent_ptr)->chn_index == parent_idx); + + if (h->cbh_ops->hop_compare(*parent_ptr, e)) + break; + + (*parent_ptr)->chn_index = cur_idx; + *cur_ptr = *parent_ptr; + cur_ptr = parent_ptr; + cur_idx = parent_idx; + did_sth = 1; + } + + e->chn_index = cur_idx; + *cur_ptr = e; + + return did_sth; +} + +/** + * Moves a node downwards, towards the last level of the binary tree. + * + * \param[in] h The heap + * \param[in] e The node + * + * \retval 1 The position of \a e in the tree was changed at least once + * \retval 0 The position of \a e in the tree was not changed + */ +static int +cfs_binheap_sink(cfs_binheap_t *h, cfs_binheap_node_t *e) +{ + unsigned int n = h->cbh_nelements; + unsigned int child_idx; + cfs_binheap_node_t **child_ptr; + cfs_binheap_node_t *child; + unsigned int child2_idx; + cfs_binheap_node_t **child2_ptr; + cfs_binheap_node_t *child2; + unsigned int cur_idx; + cfs_binheap_node_t **cur_ptr; + int did_sth = 0; + + cur_idx = e->chn_index; + cur_ptr = cfs_binheap_pointer(h, cur_idx); + LASSERT(*cur_ptr == e); + + while (cur_idx < n) { + child_idx = (cur_idx << 1) + 1; + if (child_idx >= n) + break; + + child_ptr = cfs_binheap_pointer(h, child_idx); + child = *child_ptr; + + child2_idx = child_idx + 1; + if (child2_idx < n) { + child2_ptr = cfs_binheap_pointer(h, child2_idx); + child2 = *child2_ptr; + + if (h->cbh_ops->hop_compare(child2, child)) { + child_idx = child2_idx; + child_ptr = child2_ptr; + child = child2; + } + } + + LASSERT(child->chn_index == child_idx); + + if (h->cbh_ops->hop_compare(e, child)) + break; + + child->chn_index = cur_idx; + *cur_ptr = child; + cur_ptr = child_ptr; + cur_idx = child_idx; + did_sth = 1; + } + + e->chn_index = cur_idx; + *cur_ptr = e; + + return did_sth; +} + +/** + * Sort-inserts a node into the binary heap. + * + * \param[in] h The heap + * \param[in] e The node + * + * \retval 0 Element inserted successfully + * \retval != 0 error + */ +int +cfs_binheap_insert(cfs_binheap_t *h, cfs_binheap_node_t *e) +{ + cfs_binheap_node_t **new_ptr; + unsigned int new_idx = h->cbh_nelements; + int rc; + + if (new_idx == h->cbh_hwm) { + rc = cfs_binheap_grow(h); + if (rc != 0) + return rc; + } + + if (h->cbh_ops->hop_enter) { + rc = h->cbh_ops->hop_enter(h, e); + if (rc != 0) + return rc; + } + + e->chn_index = new_idx; + new_ptr = cfs_binheap_pointer(h, new_idx); + h->cbh_nelements++; + *new_ptr = e; + + cfs_binheap_bubble(h, e); + + return 0; +} +EXPORT_SYMBOL(cfs_binheap_insert); + +/** + * Removes a node from the binary heap. + * + * \param[in] h The heap + * \param[in] e The node + */ +void +cfs_binheap_remove(cfs_binheap_t *h, cfs_binheap_node_t *e) +{ + unsigned int n = h->cbh_nelements; + unsigned int cur_idx = e->chn_index; + cfs_binheap_node_t **cur_ptr; + cfs_binheap_node_t *last; + + LASSERT(cur_idx != CBH_POISON); + LASSERT(cur_idx < n); + + cur_ptr = cfs_binheap_pointer(h, cur_idx); + LASSERT(*cur_ptr == e); + + n--; + last = *cfs_binheap_pointer(h, n); + h->cbh_nelements = n; + if (last == e) + return; + + last->chn_index = cur_idx; + *cur_ptr = last; + if (!cfs_binheap_bubble(h, *cur_ptr)) + cfs_binheap_sink(h, *cur_ptr); + + e->chn_index = CBH_POISON; + if (h->cbh_ops->hop_exit) + h->cbh_ops->hop_exit(h, e); +} +EXPORT_SYMBOL(cfs_binheap_remove); + +/** @} heap */ diff --git a/drivers/staging/lustre/lustre/libcfs/kernel_user_comm.c b/drivers/staging/lustre/lustre/libcfs/kernel_user_comm.c new file mode 100644 index 000000000000..d6d3b2e0f307 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/kernel_user_comm.c @@ -0,0 +1,346 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * Author: Nathan Rutman <nathan.rutman@sun.com> + * + * Kernel <-> userspace communication routines. + * Using pipes for all arches. + */ + +#define DEBUG_SUBSYSTEM S_CLASS +#define D_KUC D_OTHER + +#include <linux/libcfs/libcfs.h> + +#ifdef LUSTRE_UTILS +/* This is the userspace side. */ + +/** Start the userspace side of a KUC pipe. + * @param link Private descriptor for pipe/socket. + * @param groups KUC broadcast group to listen to + * (can be null for unicast to this pid) + */ +int libcfs_ukuc_start(lustre_kernelcomm *link, int group) +{ + int pfd[2]; + + if (pipe(pfd) < 0) + return -errno; + + memset(link, 0, sizeof(*link)); + link->lk_rfd = pfd[0]; + link->lk_wfd = pfd[1]; + link->lk_group = group; + link->lk_uid = getpid(); + return 0; +} + +int libcfs_ukuc_stop(lustre_kernelcomm *link) +{ + if (link->lk_wfd > 0) + close(link->lk_wfd); + return close(link->lk_rfd); +} + +#define lhsz sizeof(*kuch) + +/** Read a message from the link. + * Allocates memory, returns handle + * + * @param link Private descriptor for pipe/socket. + * @param buf Buffer to read into, must include size for kuc_hdr + * @param maxsize Maximum message size allowed + * @param transport Only listen to messages on this transport + * (and the generic transport) + */ +int libcfs_ukuc_msg_get(lustre_kernelcomm *link, char *buf, int maxsize, + int transport) +{ + struct kuc_hdr *kuch; + int rc = 0; + + memset(buf, 0, maxsize); + + CDEBUG(D_KUC, "Waiting for message from kernel on fd %d\n", + link->lk_rfd); + + while (1) { + /* Read header first to get message size */ + rc = read(link->lk_rfd, buf, lhsz); + if (rc <= 0) { + rc = -errno; + break; + } + kuch = (struct kuc_hdr *)buf; + + CDEBUG(D_KUC, "Received message mg=%x t=%d m=%d l=%d\n", + kuch->kuc_magic, kuch->kuc_transport, kuch->kuc_msgtype, + kuch->kuc_msglen); + + if (kuch->kuc_magic != KUC_MAGIC) { + CERROR("bad message magic %x != %x\n", + kuch->kuc_magic, KUC_MAGIC); + rc = -EPROTO; + break; + } + + if (kuch->kuc_msglen > maxsize) { + rc = -EMSGSIZE; + break; + } + + /* Read payload */ + rc = read(link->lk_rfd, buf + lhsz, kuch->kuc_msglen - lhsz); + if (rc < 0) { + rc = -errno; + break; + } + if (rc < (kuch->kuc_msglen - lhsz)) { + CERROR("short read: got %d of %d bytes\n", + rc, kuch->kuc_msglen); + rc = -EPROTO; + break; + } + + if (kuch->kuc_transport == transport || + kuch->kuc_transport == KUC_TRANSPORT_GENERIC) { + return 0; + } + /* Drop messages for other transports */ + } + return rc; +} + +#else /* LUSTRE_UTILS */ +/* This is the kernel side (liblustre as well). */ + +/** + * libcfs_kkuc_msg_put - send an message from kernel to userspace + * @param fp to send the message to + * @param payload Payload data. First field of payload is always + * struct kuc_hdr + */ +int libcfs_kkuc_msg_put(struct file *filp, void *payload) +{ + struct kuc_hdr *kuch = (struct kuc_hdr *)payload; + ssize_t count = kuch->kuc_msglen; + loff_t offset = 0; + mm_segment_t fs; + int rc = -ENOSYS; + + if (filp == NULL || IS_ERR(filp)) + return -EBADF; + + if (kuch->kuc_magic != KUC_MAGIC) { + CERROR("KernelComm: bad magic %x\n", kuch->kuc_magic); + return -ENOSYS; + } + + fs = get_fs(); + set_fs(KERNEL_DS); + while (count > 0) { + rc = vfs_write(filp, (void __force __user *)payload, + count, &offset); + if (rc < 0) + break; + count -= rc; + payload += rc; + rc = 0; + } + set_fs(fs); + + if (rc < 0) + CWARN("message send failed (%d)\n", rc); + else + CDEBUG(D_KUC, "Sent message rc=%d, fp=%p\n", rc, filp); + + return rc; +} +EXPORT_SYMBOL(libcfs_kkuc_msg_put); + +/* Broadcast groups are global across all mounted filesystems; + * i.e. registering for a group on 1 fs will get messages for that + * group from any fs */ +/** A single group reigstration has a uid and a file pointer */ +struct kkuc_reg { + struct list_head kr_chain; + int kr_uid; + struct file *kr_fp; + __u32 kr_data; +}; +static struct list_head kkuc_groups[KUC_GRP_MAX+1] = {}; +/* Protect message sending against remove and adds */ +static DECLARE_RWSEM(kg_sem); + +/** Add a receiver to a broadcast group + * @param filp pipe to write into + * @param uid identidier for this receiver + * @param group group number + */ +int libcfs_kkuc_group_add(struct file *filp, int uid, int group, __u32 data) +{ + struct kkuc_reg *reg; + + if (group > KUC_GRP_MAX) { + CDEBUG(D_WARNING, "Kernelcomm: bad group %d\n", group); + return -EINVAL; + } + + /* fput in group_rem */ + if (filp == NULL) + return -EBADF; + + /* freed in group_rem */ + reg = kmalloc(sizeof(*reg), 0); + if (reg == NULL) + return -ENOMEM; + + reg->kr_fp = filp; + reg->kr_uid = uid; + reg->kr_data = data; + + down_write(&kg_sem); + if (kkuc_groups[group].next == NULL) + INIT_LIST_HEAD(&kkuc_groups[group]); + list_add(®->kr_chain, &kkuc_groups[group]); + up_write(&kg_sem); + + CDEBUG(D_KUC, "Added uid=%d fp=%p to group %d\n", uid, filp, group); + + return 0; +} +EXPORT_SYMBOL(libcfs_kkuc_group_add); + +int libcfs_kkuc_group_rem(int uid, int group) +{ + struct kkuc_reg *reg, *next; + ENTRY; + + if (kkuc_groups[group].next == NULL) + RETURN(0); + + if (uid == 0) { + /* Broadcast a shutdown message */ + struct kuc_hdr lh; + + lh.kuc_magic = KUC_MAGIC; + lh.kuc_transport = KUC_TRANSPORT_GENERIC; + lh.kuc_msgtype = KUC_MSG_SHUTDOWN; + lh.kuc_msglen = sizeof(lh); + libcfs_kkuc_group_put(group, &lh); + } + + down_write(&kg_sem); + list_for_each_entry_safe(reg, next, &kkuc_groups[group], kr_chain) { + if ((uid == 0) || (uid == reg->kr_uid)) { + list_del(®->kr_chain); + CDEBUG(D_KUC, "Removed uid=%d fp=%p from group %d\n", + reg->kr_uid, reg->kr_fp, group); + if (reg->kr_fp != NULL) + fput(reg->kr_fp); + kfree(reg); + } + } + up_write(&kg_sem); + + RETURN(0); +} +EXPORT_SYMBOL(libcfs_kkuc_group_rem); + +int libcfs_kkuc_group_put(int group, void *payload) +{ + struct kkuc_reg *reg; + int rc = 0; + int one_success = 0; + ENTRY; + + down_read(&kg_sem); + list_for_each_entry(reg, &kkuc_groups[group], kr_chain) { + if (reg->kr_fp != NULL) { + rc = libcfs_kkuc_msg_put(reg->kr_fp, payload); + if (rc == 0) + one_success = 1; + else if (rc == -EPIPE) { + fput(reg->kr_fp); + reg->kr_fp = NULL; + } + } + } + up_read(&kg_sem); + + /* don't return an error if the message has been delivered + * at least to one agent */ + if (one_success) + rc = 0; + + RETURN(rc); +} +EXPORT_SYMBOL(libcfs_kkuc_group_put); + +/** + * Calls a callback function for each link of the given kuc group. + * @param group the group to call the function on. + * @param cb_func the function to be called. + * @param cb_arg iextra argument to be passed to the callback function. + */ +int libcfs_kkuc_group_foreach(int group, libcfs_kkuc_cb_t cb_func, + void *cb_arg) +{ + struct kkuc_reg *reg; + int rc = 0; + ENTRY; + + if (group > KUC_GRP_MAX) { + CDEBUG(D_WARNING, "Kernelcomm: bad group %d\n", group); + RETURN(-EINVAL); + } + + /* no link for this group */ + if (kkuc_groups[group].next == NULL) + RETURN(0); + + down_read(&kg_sem); + list_for_each_entry(reg, &kkuc_groups[group], kr_chain) { + if (reg->kr_fp != NULL) { + rc = cb_func(reg->kr_data, cb_arg); + } + } + up_read(&kg_sem); + + RETURN(rc); +} +EXPORT_SYMBOL(libcfs_kkuc_group_foreach); + +#endif /* LUSTRE_UTILS */ diff --git a/drivers/staging/lustre/lustre/libcfs/libcfs_cpu.c b/drivers/staging/lustre/lustre/libcfs/libcfs_cpu.c new file mode 100644 index 000000000000..8e88eb59dd51 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/libcfs_cpu.c @@ -0,0 +1,204 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA + * + * GPL HEADER END + */ +/* + * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * Please see comments in libcfs/include/libcfs/libcfs_cpu.h for introduction + * + * Author: liang@whamcloud.com + */ + +#ifndef EXPORT_SYMTAB +# define EXPORT_SYMTAB +#endif +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> + +/** Global CPU partition table */ +struct cfs_cpt_table *cfs_cpt_table __read_mostly = NULL; +EXPORT_SYMBOL(cfs_cpt_table); + +#ifndef HAVE_LIBCFS_CPT + +#define CFS_CPU_VERSION_MAGIC 0xbabecafe + +struct cfs_cpt_table * +cfs_cpt_table_alloc(unsigned int ncpt) +{ + struct cfs_cpt_table *cptab; + + if (ncpt != 1) { + CERROR("Can't support cpu partition number %d\n", ncpt); + return NULL; + } + + LIBCFS_ALLOC(cptab, sizeof(*cptab)); + if (cptab != NULL) { + cptab->ctb_version = CFS_CPU_VERSION_MAGIC; + cptab->ctb_nparts = ncpt; + } + + return cptab; +} +EXPORT_SYMBOL(cfs_cpt_table_alloc); + +void +cfs_cpt_table_free(struct cfs_cpt_table *cptab) +{ + LASSERT(cptab->ctb_version == CFS_CPU_VERSION_MAGIC); + + LIBCFS_FREE(cptab, sizeof(*cptab)); +} +EXPORT_SYMBOL(cfs_cpt_table_free); + +int +cfs_cpt_number(struct cfs_cpt_table *cptab) +{ + return 1; +} +EXPORT_SYMBOL(cfs_cpt_number); + +int +cfs_cpt_weight(struct cfs_cpt_table *cptab, int cpt) +{ + return 1; +} +EXPORT_SYMBOL(cfs_cpt_weight); + +int +cfs_cpt_online(struct cfs_cpt_table *cptab, int cpt) +{ + return 1; +} +EXPORT_SYMBOL(cfs_cpt_online); + +int +cfs_cpt_set_cpu(struct cfs_cpt_table *cptab, int cpt, int cpu) +{ + return 1; +} +EXPORT_SYMBOL(cfs_cpt_set_cpu); + +void +cfs_cpt_unset_cpu(struct cfs_cpt_table *cptab, int cpt, int cpu) +{ +} +EXPORT_SYMBOL(cfs_cpt_unset_cpu); + +int +cfs_cpt_set_cpumask(struct cfs_cpt_table *cptab, int cpt, cpumask_t *mask) +{ + return 1; +} +EXPORT_SYMBOL(cfs_cpt_set_cpumask); + +void +cfs_cpt_unset_cpumask(struct cfs_cpt_table *cptab, int cpt, cpumask_t *mask) +{ +} +EXPORT_SYMBOL(cfs_cpt_unset_cpumask); + +int +cfs_cpt_set_node(struct cfs_cpt_table *cptab, int cpt, int node) +{ + return 1; +} +EXPORT_SYMBOL(cfs_cpt_set_node); + +void +cfs_cpt_unset_node(struct cfs_cpt_table *cptab, int cpt, int node) +{ +} +EXPORT_SYMBOL(cfs_cpt_unset_node); + +int +cfs_cpt_set_nodemask(struct cfs_cpt_table *cptab, int cpt, nodemask_t *mask) +{ + return 1; +} +EXPORT_SYMBOL(cfs_cpt_set_nodemask); + +void +cfs_cpt_unset_nodemask(struct cfs_cpt_table *cptab, int cpt, nodemask_t *mask) +{ +} +EXPORT_SYMBOL(cfs_cpt_unset_nodemask); + +void +cfs_cpt_clear(struct cfs_cpt_table *cptab, int cpt) +{ +} +EXPORT_SYMBOL(cfs_cpt_clear); + +int +cfs_cpt_spread_node(struct cfs_cpt_table *cptab, int cpt) +{ + return 0; +} +EXPORT_SYMBOL(cfs_cpt_spread_node); + +int +cfs_cpt_current(struct cfs_cpt_table *cptab, int remap) +{ + return 0; +} +EXPORT_SYMBOL(cfs_cpt_current); + +int +cfs_cpt_of_cpu(struct cfs_cpt_table *cptab, int cpu) +{ + return 0; +} +EXPORT_SYMBOL(cfs_cpt_of_cpu); + +int +cfs_cpt_bind(struct cfs_cpt_table *cptab, int cpt) +{ + return 0; +} +EXPORT_SYMBOL(cfs_cpt_bind); + +void +cfs_cpu_fini(void) +{ + if (cfs_cpt_table != NULL) { + cfs_cpt_table_free(cfs_cpt_table); + cfs_cpt_table = NULL; + } +} + +int +cfs_cpu_init(void) +{ + cfs_cpt_table = cfs_cpt_table_alloc(1); + + return cfs_cpt_table != NULL ? 0 : -1; +} + +#endif /* HAVE_LIBCFS_CPT */ diff --git a/drivers/staging/lustre/lustre/libcfs/libcfs_lock.c b/drivers/staging/lustre/lustre/libcfs/libcfs_lock.c new file mode 100644 index 000000000000..8d6c4adf2ee6 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/libcfs_lock.c @@ -0,0 +1,192 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA + * + * GPL HEADER END + */ +/* Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * Author: liang@whamcloud.com + */ + +#ifndef EXPORT_SYMTAB +# define EXPORT_SYMTAB +#endif +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> + + +/** destroy cpu-partition lock, see libcfs_private.h for more detail */ +void +cfs_percpt_lock_free(struct cfs_percpt_lock *pcl) +{ + LASSERT(pcl->pcl_locks != NULL); + LASSERT(!pcl->pcl_locked); + + cfs_percpt_free(pcl->pcl_locks); + LIBCFS_FREE(pcl, sizeof(*pcl)); +} +EXPORT_SYMBOL(cfs_percpt_lock_free); + +/** + * create cpu-partition lock, see libcfs_private.h for more detail. + * + * cpu-partition lock is designed for large-scale SMP system, so we need to + * reduce cacheline conflict as possible as we can, that's the + * reason we always allocate cacheline-aligned memory block. + */ +struct cfs_percpt_lock * +cfs_percpt_lock_alloc(struct cfs_cpt_table *cptab) +{ + struct cfs_percpt_lock *pcl; + spinlock_t *lock; + int i; + + /* NB: cptab can be NULL, pcl will be for HW CPUs on that case */ + LIBCFS_ALLOC(pcl, sizeof(*pcl)); + if (pcl == NULL) + return NULL; + + pcl->pcl_cptab = cptab; + pcl->pcl_locks = cfs_percpt_alloc(cptab, sizeof(*lock)); + if (pcl->pcl_locks == NULL) { + LIBCFS_FREE(pcl, sizeof(*pcl)); + return NULL; + } + + cfs_percpt_for_each(lock, i, pcl->pcl_locks) + spin_lock_init(lock); + + return pcl; +} +EXPORT_SYMBOL(cfs_percpt_lock_alloc); + +/** + * lock a CPU partition + * + * \a index != CFS_PERCPT_LOCK_EX + * hold private lock indexed by \a index + * + * \a index == CFS_PERCPT_LOCK_EX + * exclusively lock @pcl and nobody can take private lock + */ +void +cfs_percpt_lock(struct cfs_percpt_lock *pcl, int index) +{ + int ncpt = cfs_cpt_number(pcl->pcl_cptab); + int i; + + LASSERT(index >= CFS_PERCPT_LOCK_EX && index < ncpt); + + if (ncpt == 1) { + index = 0; + } else { /* serialize with exclusive lock */ + while (pcl->pcl_locked) + cpu_relax(); + } + + if (likely(index != CFS_PERCPT_LOCK_EX)) { + spin_lock(pcl->pcl_locks[index]); + return; + } + + /* exclusive lock request */ + for (i = 0; i < ncpt; i++) { + spin_lock(pcl->pcl_locks[i]); + if (i == 0) { + LASSERT(!pcl->pcl_locked); + /* nobody should take private lock after this + * so I wouldn't starve for too long time */ + pcl->pcl_locked = 1; + } + } +} +EXPORT_SYMBOL(cfs_percpt_lock); + +/** unlock a CPU partition */ +void +cfs_percpt_unlock(struct cfs_percpt_lock *pcl, int index) +{ + int ncpt = cfs_cpt_number(pcl->pcl_cptab); + int i; + + index = ncpt == 1 ? 0 : index; + + if (likely(index != CFS_PERCPT_LOCK_EX)) { + spin_unlock(pcl->pcl_locks[index]); + return; + } + + for (i = ncpt - 1; i >= 0; i--) { + if (i == 0) { + LASSERT(pcl->pcl_locked); + pcl->pcl_locked = 0; + } + spin_unlock(pcl->pcl_locks[i]); + } +} +EXPORT_SYMBOL(cfs_percpt_unlock); + + +/** free cpu-partition refcount */ +void +cfs_percpt_atomic_free(atomic_t **refs) +{ + cfs_percpt_free(refs); +} +EXPORT_SYMBOL(cfs_percpt_atomic_free); + +/** allocate cpu-partition refcount with initial value @init_val */ +atomic_t ** +cfs_percpt_atomic_alloc(struct cfs_cpt_table *cptab, int init_val) +{ + atomic_t **refs; + atomic_t *ref; + int i; + + refs = cfs_percpt_alloc(cptab, sizeof(*ref)); + if (refs == NULL) + return NULL; + + cfs_percpt_for_each(ref, i, refs) + atomic_set(ref, init_val); + return refs; +} +EXPORT_SYMBOL(cfs_percpt_atomic_alloc); + +/** return sum of cpu-partition refs */ +int +cfs_percpt_atomic_summary(atomic_t **refs) +{ + atomic_t *ref; + int i; + int val = 0; + + cfs_percpt_for_each(ref, i, refs) + val += atomic_read(ref); + + return val; +} +EXPORT_SYMBOL(cfs_percpt_atomic_summary); diff --git a/drivers/staging/lustre/lustre/libcfs/libcfs_mem.c b/drivers/staging/lustre/lustre/libcfs/libcfs_mem.c new file mode 100644 index 000000000000..879137303482 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/libcfs_mem.c @@ -0,0 +1,205 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA + * + * GPL HEADER END + */ +/* + * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * Author: liang@whamcloud.com + */ + +#ifndef EXPORT_SYMTAB +# define EXPORT_SYMTAB +#endif +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> + +struct cfs_var_array { + unsigned int va_count; /* # of buffers */ + unsigned int va_size; /* size of each var */ + struct cfs_cpt_table *va_cptab; /* cpu partition table */ + void *va_ptrs[0]; /* buffer addresses */ +}; + +/* + * free per-cpu data, see more detail in cfs_percpt_free + */ +void +cfs_percpt_free(void *vars) +{ + struct cfs_var_array *arr; + int i; + + arr = container_of(vars, struct cfs_var_array, va_ptrs[0]); + + for (i = 0; i < arr->va_count; i++) { + if (arr->va_ptrs[i] != NULL) + LIBCFS_FREE(arr->va_ptrs[i], arr->va_size); + } + + LIBCFS_FREE(arr, offsetof(struct cfs_var_array, + va_ptrs[arr->va_count])); +} +EXPORT_SYMBOL(cfs_percpt_free); + +/* + * allocate per cpu-partition variables, returned value is an array of pointers, + * variable can be indexed by CPU partition ID, i.e: + * + * arr = cfs_percpt_alloc(cfs_cpu_pt, size); + * then caller can access memory block for CPU 0 by arr[0], + * memory block for CPU 1 by arr[1]... + * memory block for CPU N by arr[N]... + * + * cacheline aligned. + */ +void * +cfs_percpt_alloc(struct cfs_cpt_table *cptab, unsigned int size) +{ + struct cfs_var_array *arr; + int count; + int i; + + count = cfs_cpt_number(cptab); + + LIBCFS_ALLOC(arr, offsetof(struct cfs_var_array, va_ptrs[count])); + if (arr == NULL) + return NULL; + + arr->va_size = size = L1_CACHE_ALIGN(size); + arr->va_count = count; + arr->va_cptab = cptab; + + for (i = 0; i < count; i++) { + LIBCFS_CPT_ALLOC(arr->va_ptrs[i], cptab, i, size); + if (arr->va_ptrs[i] == NULL) { + cfs_percpt_free((void *)&arr->va_ptrs[0]); + return NULL; + } + } + + return (void *)&arr->va_ptrs[0]; +} +EXPORT_SYMBOL(cfs_percpt_alloc); + +/* + * return number of CPUs (or number of elements in per-cpu data) + * according to cptab of @vars + */ +int +cfs_percpt_number(void *vars) +{ + struct cfs_var_array *arr; + + arr = container_of(vars, struct cfs_var_array, va_ptrs[0]); + + return arr->va_count; +} +EXPORT_SYMBOL(cfs_percpt_number); + +/* + * return memory block shadowed from current CPU + */ +void * +cfs_percpt_current(void *vars) +{ + struct cfs_var_array *arr; + int cpt; + + arr = container_of(vars, struct cfs_var_array, va_ptrs[0]); + cpt = cfs_cpt_current(arr->va_cptab, 0); + if (cpt < 0) + return NULL; + + return arr->va_ptrs[cpt]; +} +EXPORT_SYMBOL(cfs_percpt_current); + +void * +cfs_percpt_index(void *vars, int idx) +{ + struct cfs_var_array *arr; + + arr = container_of(vars, struct cfs_var_array, va_ptrs[0]); + + LASSERT(idx >= 0 && idx < arr->va_count); + return arr->va_ptrs[idx]; +} +EXPORT_SYMBOL(cfs_percpt_index); + +/* + * free variable array, see more detail in cfs_array_alloc + */ +void +cfs_array_free(void *vars) +{ + struct cfs_var_array *arr; + int i; + + arr = container_of(vars, struct cfs_var_array, va_ptrs[0]); + + for (i = 0; i < arr->va_count; i++) { + if (arr->va_ptrs[i] == NULL) + continue; + + LIBCFS_FREE(arr->va_ptrs[i], arr->va_size); + } + LIBCFS_FREE(arr, offsetof(struct cfs_var_array, + va_ptrs[arr->va_count])); +} +EXPORT_SYMBOL(cfs_array_free); + +/* + * allocate a variable array, returned value is an array of pointers. + * Caller can specify length of array by @count, @size is size of each + * memory block in array. + */ +void * +cfs_array_alloc(int count, unsigned int size) +{ + struct cfs_var_array *arr; + int i; + + LIBCFS_ALLOC(arr, offsetof(struct cfs_var_array, va_ptrs[count])); + if (arr == NULL) + return NULL; + + arr->va_count = count; + arr->va_size = size; + + for (i = 0; i < count; i++) { + LIBCFS_ALLOC(arr->va_ptrs[i], size); + + if (arr->va_ptrs[i] == NULL) { + cfs_array_free((void *)&arr->va_ptrs[0]); + return NULL; + } + } + + return (void *)&arr->va_ptrs[0]; +} +EXPORT_SYMBOL(cfs_array_alloc); diff --git a/drivers/staging/lustre/lustre/libcfs/libcfs_string.c b/drivers/staging/lustre/lustre/libcfs/libcfs_string.c new file mode 100644 index 000000000000..9edccc99683e --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/libcfs_string.c @@ -0,0 +1,647 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * String manipulation functions. + * + * libcfs/libcfs/libcfs_string.c + * + * Author: Nathan Rutman <nathan.rutman@sun.com> + */ + +#include <linux/libcfs/libcfs.h> + +/* non-0 = don't match */ +int cfs_strncasecmp(const char *s1, const char *s2, size_t n) +{ + if (s1 == NULL || s2 == NULL) + return 1; + + if (n == 0) + return 0; + + while (n-- != 0 && tolower(*s1) == tolower(*s2)) { + if (n == 0 || *s1 == '\0' || *s2 == '\0') + break; + s1++; + s2++; + } + + return tolower(*(unsigned char *)s1) - tolower(*(unsigned char *)s2); +} +EXPORT_SYMBOL(cfs_strncasecmp); + +/* Convert a text string to a bitmask */ +int cfs_str2mask(const char *str, const char *(*bit2str)(int bit), + int *oldmask, int minmask, int allmask) +{ + const char *debugstr; + char op = 0; + int newmask = minmask, i, len, found = 0; + ENTRY; + + /* <str> must be a list of tokens separated by whitespace + * and optionally an operator ('+' or '-'). If an operator + * appears first in <str>, '*oldmask' is used as the starting point + * (relative), otherwise minmask is used (absolute). An operator + * applies to all following tokens up to the next operator. */ + while (*str != 0) { + while (isspace(*str)) + str++; + if (*str == 0) + break; + if (*str == '+' || *str == '-') { + op = *str++; + if (!found) + /* only if first token is relative */ + newmask = *oldmask; + while (isspace(*str)) + str++; + if (*str == 0) /* trailing op */ + return -EINVAL; + } + + /* find token length */ + for (len = 0; str[len] != 0 && !isspace(str[len]) && + str[len] != '+' && str[len] != '-'; len++); + + /* match token */ + found = 0; + for (i = 0; i < 32; i++) { + debugstr = bit2str(i); + if (debugstr != NULL && + strlen(debugstr) == len && + cfs_strncasecmp(str, debugstr, len) == 0) { + if (op == '-') + newmask &= ~(1 << i); + else + newmask |= (1 << i); + found = 1; + break; + } + } + if (!found && len == 3 && + (cfs_strncasecmp(str, "ALL", len) == 0)) { + if (op == '-') + newmask = minmask; + else + newmask = allmask; + found = 1; + } + if (!found) { + CWARN("unknown mask '%.*s'.\n" + "mask usage: [+|-]<all|type> ...\n", len, str); + return -EINVAL; + } + str += len; + } + + *oldmask = newmask; + return 0; +} +EXPORT_SYMBOL(cfs_str2mask); + +/* Duplicate a string in a platform-independent way */ +char *cfs_strdup(const char *str, u_int32_t flags) +{ + size_t lenz; /* length of str + zero byte */ + char *dup_str; + + lenz = strlen(str) + 1; + + dup_str = kmalloc(lenz, flags); + if (dup_str == NULL) + return NULL; + + memcpy(dup_str, str, lenz); + + return dup_str; +} +EXPORT_SYMBOL(cfs_strdup); + +/** + * cfs_{v}snprintf() return the actual size that is printed rather than + * the size that would be printed in standard functions. + */ +/* safe vsnprintf */ +int cfs_vsnprintf(char *buf, size_t size, const char *fmt, va_list args) +{ + int i; + + LASSERT(size > 0); + i = vsnprintf(buf, size, fmt, args); + + return (i >= size ? size - 1 : i); +} +EXPORT_SYMBOL(cfs_vsnprintf); + +/* safe snprintf */ +int cfs_snprintf(char *buf, size_t size, const char *fmt, ...) +{ + va_list args; + int i; + + va_start(args, fmt); + i = cfs_vsnprintf(buf, size, fmt, args); + va_end(args); + + return i; +} +EXPORT_SYMBOL(cfs_snprintf); + +/* get the first string out of @str */ +char *cfs_firststr(char *str, size_t size) +{ + size_t i = 0; + char *end; + + /* trim leading spaces */ + while (i < size && *str && isspace(*str)) { + ++i; + ++str; + } + + /* string with all spaces */ + if (*str == '\0') + goto out; + + end = str; + while (i < size && *end != '\0' && !isspace(*end)) { + ++i; + ++end; + } + + *end= '\0'; +out: + return str; +} +EXPORT_SYMBOL(cfs_firststr); + +char * +cfs_trimwhite(char *str) +{ + char *end; + + while (cfs_iswhite(*str)) + str++; + + end = str + strlen(str); + while (end > str) { + if (!cfs_iswhite(end[-1])) + break; + end--; + } + + *end = 0; + return str; +} +EXPORT_SYMBOL(cfs_trimwhite); + +/** + * Extracts tokens from strings. + * + * Looks for \a delim in string \a next, sets \a res to point to + * substring before the delimiter, sets \a next right after the found + * delimiter. + * + * \retval 1 if \a res points to a string of non-whitespace characters + * \retval 0 otherwise + */ +int +cfs_gettok(struct cfs_lstr *next, char delim, struct cfs_lstr *res) +{ + char *end; + + if (next->ls_str == NULL) + return 0; + + /* skip leading white spaces */ + while (next->ls_len) { + if (!cfs_iswhite(*next->ls_str)) + break; + next->ls_str++; + next->ls_len--; + } + + if (next->ls_len == 0) /* whitespaces only */ + return 0; + + if (*next->ls_str == delim) { + /* first non-writespace is the delimiter */ + return 0; + } + + res->ls_str = next->ls_str; + end = memchr(next->ls_str, delim, next->ls_len); + if (end == NULL) { + /* there is no the delimeter in the string */ + end = next->ls_str + next->ls_len; + next->ls_str = NULL; + } else { + next->ls_str = end + 1; + next->ls_len -= (end - res->ls_str + 1); + } + + /* skip ending whitespaces */ + while (--end != res->ls_str) { + if (!cfs_iswhite(*end)) + break; + } + + res->ls_len = end - res->ls_str + 1; + return 1; +} +EXPORT_SYMBOL(cfs_gettok); + +/** + * Converts string to integer. + * + * Accepts decimal and hexadecimal number recordings. + * + * \retval 1 if first \a nob chars of \a str convert to decimal or + * hexadecimal integer in the range [\a min, \a max] + * \retval 0 otherwise + */ +int +cfs_str2num_check(char *str, int nob, unsigned *num, + unsigned min, unsigned max) +{ + char *endp; + + str = cfs_trimwhite(str); + *num = strtoul(str, &endp, 0); + if (endp == str) + return 0; + + for (; endp < str + nob; endp++) { + if (!cfs_iswhite(*endp)) + return 0; + } + + return (*num >= min && *num <= max); +} +EXPORT_SYMBOL(cfs_str2num_check); + +/** + * Parses \<range_expr\> token of the syntax. If \a bracketed is false, + * \a src should only have a single token which can be \<number\> or \* + * + * \retval pointer to allocated range_expr and initialized + * range_expr::re_lo, range_expr::re_hi and range_expr:re_stride if \a + `* src parses to + * \<number\> | + * \<number\> '-' \<number\> | + * \<number\> '-' \<number\> '/' \<number\> + * \retval 0 will be returned if it can be parsed, otherwise -EINVAL or + * -ENOMEM will be returned. + */ +int +cfs_range_expr_parse(struct cfs_lstr *src, unsigned min, unsigned max, + int bracketed, struct cfs_range_expr **expr) +{ + struct cfs_range_expr *re; + struct cfs_lstr tok; + + LIBCFS_ALLOC(re, sizeof(*re)); + if (re == NULL) + return -ENOMEM; + + if (src->ls_len == 1 && src->ls_str[0] == '*') { + re->re_lo = min; + re->re_hi = max; + re->re_stride = 1; + goto out; + } + + if (cfs_str2num_check(src->ls_str, src->ls_len, + &re->re_lo, min, max)) { + /* <number> is parsed */ + re->re_hi = re->re_lo; + re->re_stride = 1; + goto out; + } + + if (!bracketed || !cfs_gettok(src, '-', &tok)) + goto failed; + + if (!cfs_str2num_check(tok.ls_str, tok.ls_len, + &re->re_lo, min, max)) + goto failed; + + /* <number> - */ + if (cfs_str2num_check(src->ls_str, src->ls_len, + &re->re_hi, min, max)) { + /* <number> - <number> is parsed */ + re->re_stride = 1; + goto out; + } + + /* go to check <number> '-' <number> '/' <number> */ + if (cfs_gettok(src, '/', &tok)) { + if (!cfs_str2num_check(tok.ls_str, tok.ls_len, + &re->re_hi, min, max)) + goto failed; + + /* <number> - <number> / ... */ + if (cfs_str2num_check(src->ls_str, src->ls_len, + &re->re_stride, min, max)) { + /* <number> - <number> / <number> is parsed */ + goto out; + } + } + + out: + *expr = re; + return 0; + + failed: + LIBCFS_FREE(re, sizeof(*re)); + return -EINVAL; +} +EXPORT_SYMBOL(cfs_range_expr_parse); + +/** + * Matches value (\a value) against ranges expression list \a expr_list. + * + * \retval 1 if \a value matches + * \retval 0 otherwise + */ +int +cfs_expr_list_match(__u32 value, struct cfs_expr_list *expr_list) +{ + struct cfs_range_expr *expr; + + list_for_each_entry(expr, &expr_list->el_exprs, re_link) { + if (value >= expr->re_lo && value <= expr->re_hi && + ((value - expr->re_lo) % expr->re_stride) == 0) + return 1; + } + + return 0; +} +EXPORT_SYMBOL(cfs_expr_list_match); + +/** + * Convert express list (\a expr_list) to an array of all matched values + * + * \retval N N is total number of all matched values + * \retval 0 if expression list is empty + * \retval < 0 for failure + */ +int +cfs_expr_list_values(struct cfs_expr_list *expr_list, int max, __u32 **valpp) +{ + struct cfs_range_expr *expr; + __u32 *val; + int count = 0; + int i; + + list_for_each_entry(expr, &expr_list->el_exprs, re_link) { + for (i = expr->re_lo; i <= expr->re_hi; i++) { + if (((i - expr->re_lo) % expr->re_stride) == 0) + count++; + } + } + + if (count == 0) /* empty expression list */ + return 0; + + if (count > max) { + CERROR("Number of values %d exceeds max allowed %d\n", + max, count); + return -EINVAL; + } + + LIBCFS_ALLOC(val, sizeof(val[0]) * count); + if (val == NULL) + return -ENOMEM; + + count = 0; + list_for_each_entry(expr, &expr_list->el_exprs, re_link) { + for (i = expr->re_lo; i <= expr->re_hi; i++) { + if (((i - expr->re_lo) % expr->re_stride) == 0) + val[count++] = i; + } + } + + *valpp = val; + return count; +} +EXPORT_SYMBOL(cfs_expr_list_values); + +/** + * Frees cfs_range_expr structures of \a expr_list. + * + * \retval none + */ +void +cfs_expr_list_free(struct cfs_expr_list *expr_list) +{ + while (!list_empty(&expr_list->el_exprs)) { + struct cfs_range_expr *expr; + + expr = list_entry(expr_list->el_exprs.next, + struct cfs_range_expr, re_link), + list_del(&expr->re_link); + LIBCFS_FREE(expr, sizeof(*expr)); + } + + LIBCFS_FREE(expr_list, sizeof(*expr_list)); +} +EXPORT_SYMBOL(cfs_expr_list_free); + +void +cfs_expr_list_print(struct cfs_expr_list *expr_list) +{ + struct cfs_range_expr *expr; + + list_for_each_entry(expr, &expr_list->el_exprs, re_link) { + CDEBUG(D_WARNING, "%d-%d/%d\n", + expr->re_lo, expr->re_hi, expr->re_stride); + } +} +EXPORT_SYMBOL(cfs_expr_list_print); + +/** + * Parses \<cfs_expr_list\> token of the syntax. + * + * \retval 1 if \a str parses to \<number\> | \<expr_list\> + * \retval 0 otherwise + */ +int +cfs_expr_list_parse(char *str, int len, unsigned min, unsigned max, + struct cfs_expr_list **elpp) +{ + struct cfs_expr_list *expr_list; + struct cfs_range_expr *expr; + struct cfs_lstr src; + int rc; + + LIBCFS_ALLOC(expr_list, sizeof(*expr_list)); + if (expr_list == NULL) + return -ENOMEM; + + src.ls_str = str; + src.ls_len = len; + + INIT_LIST_HEAD(&expr_list->el_exprs); + + if (src.ls_str[0] == '[' && + src.ls_str[src.ls_len - 1] == ']') { + src.ls_str++; + src.ls_len -= 2; + + rc = -EINVAL; + while (src.ls_str != NULL) { + struct cfs_lstr tok; + + if (!cfs_gettok(&src, ',', &tok)) { + rc = -EINVAL; + break; + } + + rc = cfs_range_expr_parse(&tok, min, max, 1, &expr); + if (rc != 0) + break; + + list_add_tail(&expr->re_link, + &expr_list->el_exprs); + } + } else { + rc = cfs_range_expr_parse(&src, min, max, 0, &expr); + if (rc == 0) { + list_add_tail(&expr->re_link, + &expr_list->el_exprs); + } + } + + if (rc != 0) + cfs_expr_list_free(expr_list); + else + *elpp = expr_list; + + return rc; +} +EXPORT_SYMBOL(cfs_expr_list_parse); + +/** + * Frees cfs_expr_list structures of \a list. + * + * For each struct cfs_expr_list structure found on \a list it frees + * range_expr list attached to it and frees the cfs_expr_list itself. + * + * \retval none + */ +void +cfs_expr_list_free_list(struct list_head *list) +{ + struct cfs_expr_list *el; + + while (!list_empty(list)) { + el = list_entry(list->next, + struct cfs_expr_list, el_link); + list_del(&el->el_link); + cfs_expr_list_free(el); + } +} +EXPORT_SYMBOL(cfs_expr_list_free_list); + +int +cfs_ip_addr_parse(char *str, int len, struct list_head *list) +{ + struct cfs_expr_list *el; + struct cfs_lstr src; + int rc; + int i; + + src.ls_str = str; + src.ls_len = len; + i = 0; + + while (src.ls_str != NULL) { + struct cfs_lstr res; + + if (!cfs_gettok(&src, '.', &res)) { + rc = -EINVAL; + goto out; + } + + rc = cfs_expr_list_parse(res.ls_str, res.ls_len, 0, 255, &el); + if (rc != 0) + goto out; + + list_add_tail(&el->el_link, list); + i++; + } + + if (i == 4) + return 0; + + rc = -EINVAL; + out: + cfs_expr_list_free_list(list); + + return rc; +} +EXPORT_SYMBOL(cfs_ip_addr_parse); + +/** + * Matches address (\a addr) against address set encoded in \a list. + * + * \retval 1 if \a addr matches + * \retval 0 otherwise + */ +int +cfs_ip_addr_match(__u32 addr, struct list_head *list) +{ + struct cfs_expr_list *el; + int i = 0; + + list_for_each_entry_reverse(el, list, el_link) { + if (!cfs_expr_list_match(addr & 0xff, el)) + return 0; + addr >>= 8; + i++; + } + + return i == 4; +} +EXPORT_SYMBOL(cfs_ip_addr_match); + +void +cfs_ip_addr_free(struct list_head *list) +{ + cfs_expr_list_free_list(list); +} +EXPORT_SYMBOL(cfs_ip_addr_free); diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-cpu.c b/drivers/staging/lustre/lustre/libcfs/linux/linux-cpu.c new file mode 100644 index 000000000000..95142d190971 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-cpu.c @@ -0,0 +1,1085 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; if not, write to the + * Free Software Foundation, Inc., 59 Temple Place - Suite 330, + * Boston, MA 021110-1307, USA + * + * GPL HEADER END + */ +/* + * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved. + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * Author: liang@whamcloud.com + */ + +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/cpu.h> +#include <linux/sched.h> +#include <linux/libcfs/libcfs.h> + +#ifdef CONFIG_SMP + +/** + * modparam for setting number of partitions + * + * 0 : estimate best value based on cores or NUMA nodes + * 1 : disable multiple partitions + * >1 : specify number of partitions + */ +static int cpu_npartitions; +CFS_MODULE_PARM(cpu_npartitions, "i", int, 0444, "# of CPU partitions"); + +/** + * modparam for setting CPU partitions patterns: + * + * i.e: "0[0,1,2,3] 1[4,5,6,7]", number before bracket is CPU partition ID, + * number in bracket is processor ID (core or HT) + * + * i.e: "N 0[0,1] 1[2,3]" the first character 'N' means numbers in bracket + * are NUMA node ID, number before bracket is CPU partition ID. + * + * NB: If user specified cpu_pattern, cpu_npartitions will be ignored + */ +static char *cpu_pattern = ""; +CFS_MODULE_PARM(cpu_pattern, "s", charp, 0444, "CPU partitions pattern"); + +struct cfs_cpt_data { + /* serialize hotplug etc */ + spinlock_t cpt_lock; + /* reserved for hotplug */ + unsigned long cpt_version; + /* mutex to protect cpt_cpumask */ + struct semaphore cpt_mutex; + /* scratch buffer for set/unset_node */ + cpumask_t *cpt_cpumask; +}; + +static struct cfs_cpt_data cpt_data; + +void +cfs_cpu_core_siblings(int cpu, cpumask_t *mask) +{ + /* return cpumask of cores in the same socket */ + cpumask_copy(mask, topology_core_cpumask(cpu)); +} +EXPORT_SYMBOL(cfs_cpu_core_siblings); + +/* return number of cores in the same socket of \a cpu */ +int +cfs_cpu_core_nsiblings(int cpu) +{ + int num; + + down(&cpt_data.cpt_mutex); + + cfs_cpu_core_siblings(cpu, cpt_data.cpt_cpumask); + num = cpus_weight(*cpt_data.cpt_cpumask); + + up(&cpt_data.cpt_mutex); + + return num; +} +EXPORT_SYMBOL(cfs_cpu_core_nsiblings); + +/* return cpumask of HTs in the same core */ +void +cfs_cpu_ht_siblings(int cpu, cpumask_t *mask) +{ + cpumask_copy(mask, topology_thread_cpumask(cpu)); +} +EXPORT_SYMBOL(cfs_cpu_ht_siblings); + +/* return number of HTs in the same core of \a cpu */ +int +cfs_cpu_ht_nsiblings(int cpu) +{ + int num; + + down(&cpt_data.cpt_mutex); + + cfs_cpu_ht_siblings(cpu, cpt_data.cpt_cpumask); + num = cpus_weight(*cpt_data.cpt_cpumask); + + up(&cpt_data.cpt_mutex); + + return num; +} +EXPORT_SYMBOL(cfs_cpu_ht_nsiblings); + +void +cfs_node_to_cpumask(int node, cpumask_t *mask) +{ + cpumask_copy(mask, cpumask_of_node(node)); +} +EXPORT_SYMBOL(cfs_node_to_cpumask); + +void +cfs_cpt_table_free(struct cfs_cpt_table *cptab) +{ + int i; + + if (cptab->ctb_cpu2cpt != NULL) { + LIBCFS_FREE(cptab->ctb_cpu2cpt, + num_possible_cpus() * + sizeof(cptab->ctb_cpu2cpt[0])); + } + + for (i = 0; cptab->ctb_parts != NULL && i < cptab->ctb_nparts; i++) { + struct cfs_cpu_partition *part = &cptab->ctb_parts[i]; + + if (part->cpt_nodemask != NULL) { + LIBCFS_FREE(part->cpt_nodemask, + sizeof(*part->cpt_nodemask)); + } + + if (part->cpt_cpumask != NULL) + LIBCFS_FREE(part->cpt_cpumask, cpumask_size()); + } + + if (cptab->ctb_parts != NULL) { + LIBCFS_FREE(cptab->ctb_parts, + cptab->ctb_nparts * sizeof(cptab->ctb_parts[0])); + } + + if (cptab->ctb_nodemask != NULL) + LIBCFS_FREE(cptab->ctb_nodemask, sizeof(*cptab->ctb_nodemask)); + if (cptab->ctb_cpumask != NULL) + LIBCFS_FREE(cptab->ctb_cpumask, cpumask_size()); + + LIBCFS_FREE(cptab, sizeof(*cptab)); +} +EXPORT_SYMBOL(cfs_cpt_table_free); + +struct cfs_cpt_table * +cfs_cpt_table_alloc(unsigned int ncpt) +{ + struct cfs_cpt_table *cptab; + int i; + + LIBCFS_ALLOC(cptab, sizeof(*cptab)); + if (cptab == NULL) + return NULL; + + cptab->ctb_nparts = ncpt; + + LIBCFS_ALLOC(cptab->ctb_cpumask, cpumask_size()); + LIBCFS_ALLOC(cptab->ctb_nodemask, sizeof(*cptab->ctb_nodemask)); + + if (cptab->ctb_cpumask == NULL || cptab->ctb_nodemask == NULL) + goto failed; + + LIBCFS_ALLOC(cptab->ctb_cpu2cpt, + num_possible_cpus() * sizeof(cptab->ctb_cpu2cpt[0])); + if (cptab->ctb_cpu2cpt == NULL) + goto failed; + + memset(cptab->ctb_cpu2cpt, -1, + num_possible_cpus() * sizeof(cptab->ctb_cpu2cpt[0])); + + LIBCFS_ALLOC(cptab->ctb_parts, ncpt * sizeof(cptab->ctb_parts[0])); + if (cptab->ctb_parts == NULL) + goto failed; + + for (i = 0; i < ncpt; i++) { + struct cfs_cpu_partition *part = &cptab->ctb_parts[i]; + + LIBCFS_ALLOC(part->cpt_cpumask, cpumask_size()); + LIBCFS_ALLOC(part->cpt_nodemask, sizeof(*part->cpt_nodemask)); + if (part->cpt_cpumask == NULL || part->cpt_nodemask == NULL) + goto failed; + } + + spin_lock(&cpt_data.cpt_lock); + /* Reserved for hotplug */ + cptab->ctb_version = cpt_data.cpt_version; + spin_unlock(&cpt_data.cpt_lock); + + return cptab; + + failed: + cfs_cpt_table_free(cptab); + return NULL; +} +EXPORT_SYMBOL(cfs_cpt_table_alloc); + +int +cfs_cpt_table_print(struct cfs_cpt_table *cptab, char *buf, int len) +{ + char *tmp = buf; + int rc = 0; + int i; + int j; + + for (i = 0; i < cptab->ctb_nparts; i++) { + if (len > 0) { + rc = snprintf(tmp, len, "%d\t: ", i); + len -= rc; + } + + if (len <= 0) { + rc = -EFBIG; + goto out; + } + + tmp += rc; + for_each_cpu_mask(j, *cptab->ctb_parts[i].cpt_cpumask) { + rc = snprintf(tmp, len, "%d ", j); + len -= rc; + if (len <= 0) { + rc = -EFBIG; + goto out; + } + tmp += rc; + } + + *tmp = '\n'; + tmp++; + len--; + } + + out: + if (rc < 0) + return rc; + + return tmp - buf; +} +EXPORT_SYMBOL(cfs_cpt_table_print); + +int +cfs_cpt_number(struct cfs_cpt_table *cptab) +{ + return cptab->ctb_nparts; +} +EXPORT_SYMBOL(cfs_cpt_number); + +int +cfs_cpt_weight(struct cfs_cpt_table *cptab, int cpt) +{ + LASSERT(cpt == CFS_CPT_ANY || (cpt >= 0 && cpt < cptab->ctb_nparts)); + + return cpt == CFS_CPT_ANY ? + cpus_weight(*cptab->ctb_cpumask) : + cpus_weight(*cptab->ctb_parts[cpt].cpt_cpumask); +} +EXPORT_SYMBOL(cfs_cpt_weight); + +int +cfs_cpt_online(struct cfs_cpt_table *cptab, int cpt) +{ + LASSERT(cpt == CFS_CPT_ANY || (cpt >= 0 && cpt < cptab->ctb_nparts)); + + return cpt == CFS_CPT_ANY ? + any_online_cpu(*cptab->ctb_cpumask) != NR_CPUS : + any_online_cpu(*cptab->ctb_parts[cpt].cpt_cpumask) != NR_CPUS; +} +EXPORT_SYMBOL(cfs_cpt_online); + +cpumask_t * +cfs_cpt_cpumask(struct cfs_cpt_table *cptab, int cpt) +{ + LASSERT(cpt == CFS_CPT_ANY || (cpt >= 0 && cpt < cptab->ctb_nparts)); + + return cpt == CFS_CPT_ANY ? + cptab->ctb_cpumask : cptab->ctb_parts[cpt].cpt_cpumask; +} +EXPORT_SYMBOL(cfs_cpt_cpumask); + +nodemask_t * +cfs_cpt_nodemask(struct cfs_cpt_table *cptab, int cpt) +{ + LASSERT(cpt == CFS_CPT_ANY || (cpt >= 0 && cpt < cptab->ctb_nparts)); + + return cpt == CFS_CPT_ANY ? + cptab->ctb_nodemask : cptab->ctb_parts[cpt].cpt_nodemask; +} +EXPORT_SYMBOL(cfs_cpt_nodemask); + +int +cfs_cpt_set_cpu(struct cfs_cpt_table *cptab, int cpt, int cpu) +{ + int node; + + LASSERT(cpt >= 0 && cpt < cptab->ctb_nparts); + + if (cpu < 0 || cpu >= NR_CPUS || !cpu_online(cpu)) { + CDEBUG(D_INFO, "CPU %d is invalid or it's offline\n", cpu); + return 0; + } + + if (cptab->ctb_cpu2cpt[cpu] != -1) { + CDEBUG(D_INFO, "CPU %d is already in partition %d\n", + cpu, cptab->ctb_cpu2cpt[cpu]); + return 0; + } + + cptab->ctb_cpu2cpt[cpu] = cpt; + + LASSERT(!cpu_isset(cpu, *cptab->ctb_cpumask)); + LASSERT(!cpu_isset(cpu, *cptab->ctb_parts[cpt].cpt_cpumask)); + + cpu_set(cpu, *cptab->ctb_cpumask); + cpu_set(cpu, *cptab->ctb_parts[cpt].cpt_cpumask); + + node = cpu_to_node(cpu); + + /* first CPU of @node in this CPT table */ + if (!node_isset(node, *cptab->ctb_nodemask)) + node_set(node, *cptab->ctb_nodemask); + + /* first CPU of @node in this partition */ + if (!node_isset(node, *cptab->ctb_parts[cpt].cpt_nodemask)) + node_set(node, *cptab->ctb_parts[cpt].cpt_nodemask); + + return 1; +} +EXPORT_SYMBOL(cfs_cpt_set_cpu); + +void +cfs_cpt_unset_cpu(struct cfs_cpt_table *cptab, int cpt, int cpu) +{ + int node; + int i; + + LASSERT(cpt == CFS_CPT_ANY || (cpt >= 0 && cpt < cptab->ctb_nparts)); + + if (cpu < 0 || cpu >= NR_CPUS) { + CDEBUG(D_INFO, "Invalid CPU id %d\n", cpu); + return; + } + + if (cpt == CFS_CPT_ANY) { + /* caller doesn't know the partition ID */ + cpt = cptab->ctb_cpu2cpt[cpu]; + if (cpt < 0) { /* not set in this CPT-table */ + CDEBUG(D_INFO, "Try to unset cpu %d which is " + "not in CPT-table %p\n", cpt, cptab); + return; + } + + } else if (cpt != cptab->ctb_cpu2cpt[cpu]) { + CDEBUG(D_INFO, + "CPU %d is not in cpu-partition %d\n", cpu, cpt); + return; + } + + LASSERT(cpu_isset(cpu, *cptab->ctb_parts[cpt].cpt_cpumask)); + LASSERT(cpu_isset(cpu, *cptab->ctb_cpumask)); + + cpu_clear(cpu, *cptab->ctb_parts[cpt].cpt_cpumask); + cpu_clear(cpu, *cptab->ctb_cpumask); + cptab->ctb_cpu2cpt[cpu] = -1; + + node = cpu_to_node(cpu); + + LASSERT(node_isset(node, *cptab->ctb_parts[cpt].cpt_nodemask)); + LASSERT(node_isset(node, *cptab->ctb_nodemask)); + + for_each_cpu_mask(i, *cptab->ctb_parts[cpt].cpt_cpumask) { + /* this CPT has other CPU belonging to this node? */ + if (cpu_to_node(i) == node) + break; + } + + if (i == NR_CPUS) + node_clear(node, *cptab->ctb_parts[cpt].cpt_nodemask); + + for_each_cpu_mask(i, *cptab->ctb_cpumask) { + /* this CPT-table has other CPU belonging to this node? */ + if (cpu_to_node(i) == node) + break; + } + + if (i == NR_CPUS) + node_clear(node, *cptab->ctb_nodemask); + + return; +} +EXPORT_SYMBOL(cfs_cpt_unset_cpu); + +int +cfs_cpt_set_cpumask(struct cfs_cpt_table *cptab, int cpt, cpumask_t *mask) +{ + int i; + + if (cpus_weight(*mask) == 0 || any_online_cpu(*mask) == NR_CPUS) { + CDEBUG(D_INFO, "No online CPU is found in the CPU mask " + "for CPU partition %d\n", cpt); + return 0; + } + + for_each_cpu_mask(i, *mask) { + if (!cfs_cpt_set_cpu(cptab, cpt, i)) + return 0; + } + + return 1; +} +EXPORT_SYMBOL(cfs_cpt_set_cpumask); + +void +cfs_cpt_unset_cpumask(struct cfs_cpt_table *cptab, int cpt, cpumask_t *mask) +{ + int i; + + for_each_cpu_mask(i, *mask) + cfs_cpt_unset_cpu(cptab, cpt, i); +} +EXPORT_SYMBOL(cfs_cpt_unset_cpumask); + +int +cfs_cpt_set_node(struct cfs_cpt_table *cptab, int cpt, int node) +{ + cpumask_t *mask; + int rc; + + if (node < 0 || node >= MAX_NUMNODES) { + CDEBUG(D_INFO, + "Invalid NUMA id %d for CPU partition %d\n", node, cpt); + return 0; + } + + down(&cpt_data.cpt_mutex); + + mask = cpt_data.cpt_cpumask; + cfs_node_to_cpumask(node, mask); + + rc = cfs_cpt_set_cpumask(cptab, cpt, mask); + + up(&cpt_data.cpt_mutex); + + return rc; +} +EXPORT_SYMBOL(cfs_cpt_set_node); + +void +cfs_cpt_unset_node(struct cfs_cpt_table *cptab, int cpt, int node) +{ + cpumask_t *mask; + + if (node < 0 || node >= MAX_NUMNODES) { + CDEBUG(D_INFO, + "Invalid NUMA id %d for CPU partition %d\n", node, cpt); + return; + } + + down(&cpt_data.cpt_mutex); + + mask = cpt_data.cpt_cpumask; + cfs_node_to_cpumask(node, mask); + + cfs_cpt_unset_cpumask(cptab, cpt, mask); + + up(&cpt_data.cpt_mutex); +} +EXPORT_SYMBOL(cfs_cpt_unset_node); + +int +cfs_cpt_set_nodemask(struct cfs_cpt_table *cptab, int cpt, nodemask_t *mask) +{ + int i; + + for_each_node_mask(i, *mask) { + if (!cfs_cpt_set_node(cptab, cpt, i)) + return 0; + } + + return 1; +} +EXPORT_SYMBOL(cfs_cpt_set_nodemask); + +void +cfs_cpt_unset_nodemask(struct cfs_cpt_table *cptab, int cpt, nodemask_t *mask) +{ + int i; + + for_each_node_mask(i, *mask) + cfs_cpt_unset_node(cptab, cpt, i); +} +EXPORT_SYMBOL(cfs_cpt_unset_nodemask); + +void +cfs_cpt_clear(struct cfs_cpt_table *cptab, int cpt) +{ + int last; + int i; + + if (cpt == CFS_CPT_ANY) { + last = cptab->ctb_nparts - 1; + cpt = 0; + } else { + last = cpt; + } + + for (; cpt <= last; cpt++) { + for_each_cpu_mask(i, *cptab->ctb_parts[cpt].cpt_cpumask) + cfs_cpt_unset_cpu(cptab, cpt, i); + } +} +EXPORT_SYMBOL(cfs_cpt_clear); + +int +cfs_cpt_spread_node(struct cfs_cpt_table *cptab, int cpt) +{ + nodemask_t *mask; + int weight; + int rotor; + int node; + + /* convert CPU partition ID to HW node id */ + + if (cpt < 0 || cpt >= cptab->ctb_nparts) { + mask = cptab->ctb_nodemask; + rotor = cptab->ctb_spread_rotor++; + } else { + mask = cptab->ctb_parts[cpt].cpt_nodemask; + rotor = cptab->ctb_parts[cpt].cpt_spread_rotor++; + } + + weight = nodes_weight(*mask); + LASSERT(weight > 0); + + rotor %= weight; + + for_each_node_mask(node, *mask) { + if (rotor-- == 0) + return node; + } + + LBUG(); + return 0; +} +EXPORT_SYMBOL(cfs_cpt_spread_node); + +int +cfs_cpt_current(struct cfs_cpt_table *cptab, int remap) +{ + int cpu = smp_processor_id(); + int cpt = cptab->ctb_cpu2cpt[cpu]; + + if (cpt < 0) { + if (!remap) + return cpt; + + /* don't return negative value for safety of upper layer, + * instead we shadow the unknown cpu to a valid partition ID */ + cpt = cpu % cptab->ctb_nparts; + } + + return cpt; +} +EXPORT_SYMBOL(cfs_cpt_current); + +int +cfs_cpt_of_cpu(struct cfs_cpt_table *cptab, int cpu) +{ + LASSERT(cpu >= 0 && cpu < NR_CPUS); + + return cptab->ctb_cpu2cpt[cpu]; +} +EXPORT_SYMBOL(cfs_cpt_of_cpu); + +int +cfs_cpt_bind(struct cfs_cpt_table *cptab, int cpt) +{ + cpumask_t *cpumask; + nodemask_t *nodemask; + int rc; + int i; + + LASSERT(cpt == CFS_CPT_ANY || (cpt >= 0 && cpt < cptab->ctb_nparts)); + + if (cpt == CFS_CPT_ANY) { + cpumask = cptab->ctb_cpumask; + nodemask = cptab->ctb_nodemask; + } else { + cpumask = cptab->ctb_parts[cpt].cpt_cpumask; + nodemask = cptab->ctb_parts[cpt].cpt_nodemask; + } + + if (any_online_cpu(*cpumask) == NR_CPUS) { + CERROR("No online CPU found in CPU partition %d, did someone " + "do CPU hotplug on system? You might need to reload " + "Lustre modules to keep system working well.\n", cpt); + return -EINVAL; + } + + for_each_online_cpu(i) { + if (cpu_isset(i, *cpumask)) + continue; + + rc = set_cpus_allowed_ptr(current, cpumask); + set_mems_allowed(*nodemask); + if (rc == 0) + schedule(); /* switch to allowed CPU */ + + return rc; + } + + /* don't need to set affinity because all online CPUs are covered */ + return 0; +} +EXPORT_SYMBOL(cfs_cpt_bind); + +/** + * Choose max to \a number CPUs from \a node and set them in \a cpt. + * We always prefer to choose CPU in the same core/socket. + */ +static int +cfs_cpt_choose_ncpus(struct cfs_cpt_table *cptab, int cpt, + cpumask_t *node, int number) +{ + cpumask_t *socket = NULL; + cpumask_t *core = NULL; + int rc = 0; + int cpu; + + LASSERT(number > 0); + + if (number >= cpus_weight(*node)) { + while (!cpus_empty(*node)) { + cpu = first_cpu(*node); + + rc = cfs_cpt_set_cpu(cptab, cpt, cpu); + if (!rc) + return -EINVAL; + cpu_clear(cpu, *node); + } + return 0; + } + + /* allocate scratch buffer */ + LIBCFS_ALLOC(socket, cpumask_size()); + LIBCFS_ALLOC(core, cpumask_size()); + if (socket == NULL || core == NULL) { + rc = -ENOMEM; + goto out; + } + + while (!cpus_empty(*node)) { + cpu = first_cpu(*node); + + /* get cpumask for cores in the same socket */ + cfs_cpu_core_siblings(cpu, socket); + cpus_and(*socket, *socket, *node); + + LASSERT(!cpus_empty(*socket)); + + while (!cpus_empty(*socket)) { + int i; + + /* get cpumask for hts in the same core */ + cfs_cpu_ht_siblings(cpu, core); + cpus_and(*core, *core, *node); + + LASSERT(!cpus_empty(*core)); + + for_each_cpu_mask(i, *core) { + cpu_clear(i, *socket); + cpu_clear(i, *node); + + rc = cfs_cpt_set_cpu(cptab, cpt, i); + if (!rc) { + rc = -EINVAL; + goto out; + } + + if (--number == 0) + goto out; + } + cpu = first_cpu(*socket); + } + } + + out: + if (socket != NULL) + LIBCFS_FREE(socket, cpumask_size()); + if (core != NULL) + LIBCFS_FREE(core, cpumask_size()); + return rc; +} + +#define CPT_WEIGHT_MIN 4u + +static unsigned int +cfs_cpt_num_estimate(void) +{ + unsigned nnode = num_online_nodes(); + unsigned ncpu = num_online_cpus(); + unsigned ncpt; + + if (ncpu <= CPT_WEIGHT_MIN) { + ncpt = 1; + goto out; + } + + /* generate reasonable number of CPU partitions based on total number + * of CPUs, Preferred N should be power2 and match this condition: + * 2 * (N - 1)^2 < NCPUS <= 2 * N^2 */ + for (ncpt = 2; ncpu > 2 * ncpt * ncpt; ncpt <<= 1) {} + + if (ncpt <= nnode) { /* fat numa system */ + while (nnode > ncpt) + nnode >>= 1; + + } else { /* ncpt > nnode */ + while ((nnode << 1) <= ncpt) + nnode <<= 1; + } + + ncpt = nnode; + + out: +#if (BITS_PER_LONG == 32) + /* config many CPU partitions on 32-bit system could consume + * too much memory */ + ncpt = min(2U, ncpt); +#endif + while (ncpu % ncpt != 0) + ncpt--; /* worst case is 1 */ + + return ncpt; +} + +static struct cfs_cpt_table * +cfs_cpt_table_create(int ncpt) +{ + struct cfs_cpt_table *cptab = NULL; + cpumask_t *mask = NULL; + int cpt = 0; + int num; + int rc; + int i; + + rc = cfs_cpt_num_estimate(); + if (ncpt <= 0) + ncpt = rc; + + if (ncpt > num_online_cpus() || ncpt > 4 * rc) { + CWARN("CPU partition number %d is larger than suggested " + "value (%d), your system may have performance" + "issue or run out of memory while under pressure\n", + ncpt, rc); + } + + if (num_online_cpus() % ncpt != 0) { + CERROR("CPU number %d is not multiple of cpu_npartition %d, " + "please try different cpu_npartitions value or" + "set pattern string by cpu_pattern=STRING\n", + (int)num_online_cpus(), ncpt); + goto failed; + } + + cptab = cfs_cpt_table_alloc(ncpt); + if (cptab == NULL) { + CERROR("Failed to allocate CPU map(%d)\n", ncpt); + goto failed; + } + + num = num_online_cpus() / ncpt; + if (num == 0) { + CERROR("CPU changed while setting CPU partition\n"); + goto failed; + } + + LIBCFS_ALLOC(mask, cpumask_size()); + if (mask == NULL) { + CERROR("Failed to allocate scratch cpumask\n"); + goto failed; + } + + for_each_online_node(i) { + cfs_node_to_cpumask(i, mask); + + while (!cpus_empty(*mask)) { + struct cfs_cpu_partition *part; + int n; + + if (cpt >= ncpt) + goto failed; + + part = &cptab->ctb_parts[cpt]; + + n = num - cpus_weight(*part->cpt_cpumask); + LASSERT(n > 0); + + rc = cfs_cpt_choose_ncpus(cptab, cpt, mask, n); + if (rc < 0) + goto failed; + + LASSERT(num >= cpus_weight(*part->cpt_cpumask)); + if (num == cpus_weight(*part->cpt_cpumask)) + cpt++; + } + } + + if (cpt != ncpt || + num != cpus_weight(*cptab->ctb_parts[ncpt - 1].cpt_cpumask)) { + CERROR("Expect %d(%d) CPU partitions but got %d(%d), " + "CPU hotplug/unplug while setting?\n", + cptab->ctb_nparts, num, cpt, + cpus_weight(*cptab->ctb_parts[ncpt - 1].cpt_cpumask)); + goto failed; + } + + LIBCFS_FREE(mask, cpumask_size()); + + return cptab; + + failed: + CERROR("Failed to setup CPU-partition-table with %d " + "CPU-partitions, online HW nodes: %d, HW cpus: %d.\n", + ncpt, num_online_nodes(), num_online_cpus()); + + if (mask != NULL) + LIBCFS_FREE(mask, cpumask_size()); + + if (cptab != NULL) + cfs_cpt_table_free(cptab); + + return NULL; +} + +static struct cfs_cpt_table * +cfs_cpt_table_create_pattern(char *pattern) +{ + struct cfs_cpt_table *cptab; + char *str = pattern; + int node = 0; + int high; + int ncpt; + int c; + + for (ncpt = 0;; ncpt++) { /* quick scan bracket */ + str = strchr(str, '['); + if (str == NULL) + break; + str++; + } + + str = cfs_trimwhite(pattern); + if (*str == 'n' || *str == 'N') { + pattern = str + 1; + node = 1; + } + + if (ncpt == 0 || + (node && ncpt > num_online_nodes()) || + (!node && ncpt > num_online_cpus())) { + CERROR("Invalid pattern %s, or too many partitions %d\n", + pattern, ncpt); + return NULL; + } + + high = node ? MAX_NUMNODES - 1 : NR_CPUS - 1; + + cptab = cfs_cpt_table_alloc(ncpt); + if (cptab == NULL) { + CERROR("Failed to allocate cpu partition table\n"); + return NULL; + } + + for (str = cfs_trimwhite(pattern), c = 0;; c++) { + struct cfs_range_expr *range; + struct cfs_expr_list *el; + char *bracket = strchr(str, '['); + int cpt; + int rc; + int i; + int n; + + if (bracket == NULL) { + if (*str != 0) { + CERROR("Invalid pattern %s\n", str); + goto failed; + } else if (c != ncpt) { + CERROR("expect %d partitions but found %d\n", + ncpt, c); + goto failed; + } + break; + } + + if (sscanf(str, "%u%n", &cpt, &n) < 1) { + CERROR("Invalid cpu pattern %s\n", str); + goto failed; + } + + if (cpt < 0 || cpt >= ncpt) { + CERROR("Invalid partition id %d, total partitions %d\n", + cpt, ncpt); + goto failed; + } + + if (cfs_cpt_weight(cptab, cpt) != 0) { + CERROR("Partition %d has already been set.\n", cpt); + goto failed; + } + + str = cfs_trimwhite(str + n); + if (str != bracket) { + CERROR("Invalid pattern %s\n", str); + goto failed; + } + + bracket = strchr(str, ']'); + if (bracket == NULL) { + CERROR("missing right bracket for cpt %d, %s\n", + cpt, str); + goto failed; + } + + if (cfs_expr_list_parse(str, (bracket - str) + 1, + 0, high, &el) != 0) { + CERROR("Can't parse number range: %s\n", str); + goto failed; + } + + list_for_each_entry(range, &el->el_exprs, re_link) { + for (i = range->re_lo; i <= range->re_hi; i++) { + if ((i - range->re_lo) % range->re_stride != 0) + continue; + + rc = node ? cfs_cpt_set_node(cptab, cpt, i) : + cfs_cpt_set_cpu(cptab, cpt, i); + if (!rc) { + cfs_expr_list_free(el); + goto failed; + } + } + } + + cfs_expr_list_free(el); + + if (!cfs_cpt_online(cptab, cpt)) { + CERROR("No online CPU is found on partition %d\n", cpt); + goto failed; + } + + str = cfs_trimwhite(bracket + 1); + } + + return cptab; + + failed: + cfs_cpt_table_free(cptab); + return NULL; +} + +#ifdef CONFIG_HOTPLUG_CPU +static int +cfs_cpu_notify(struct notifier_block *self, unsigned long action, void *hcpu) +{ + unsigned int cpu = (unsigned long)hcpu; + + switch (action) { + case CPU_DEAD: + case CPU_DEAD_FROZEN: + case CPU_ONLINE: + case CPU_ONLINE_FROZEN: + spin_lock(&cpt_data.cpt_lock); + cpt_data.cpt_version++; + spin_unlock(&cpt_data.cpt_lock); + default: + CWARN("Lustre: can't support CPU hotplug well now, " + "performance and stability could be impacted" + "[CPU %u notify: %lx]\n", cpu, action); + } + + return NOTIFY_OK; +} + +static struct notifier_block cfs_cpu_notifier = { + .notifier_call = cfs_cpu_notify, + .priority = 0 +}; + +#endif + +void +cfs_cpu_fini(void) +{ + if (cfs_cpt_table != NULL) + cfs_cpt_table_free(cfs_cpt_table); + +#ifdef CONFIG_HOTPLUG_CPU + unregister_hotcpu_notifier(&cfs_cpu_notifier); +#endif + if (cpt_data.cpt_cpumask != NULL) + LIBCFS_FREE(cpt_data.cpt_cpumask, cpumask_size()); +} + +int +cfs_cpu_init(void) +{ + LASSERT(cfs_cpt_table == NULL); + + memset(&cpt_data, 0, sizeof(cpt_data)); + + LIBCFS_ALLOC(cpt_data.cpt_cpumask, cpumask_size()); + if (cpt_data.cpt_cpumask == NULL) { + CERROR("Failed to allocate scratch buffer\n"); + return -1; + } + + spin_lock_init(&cpt_data.cpt_lock); + sema_init(&cpt_data.cpt_mutex, 1); + +#ifdef CONFIG_HOTPLUG_CPU + register_hotcpu_notifier(&cfs_cpu_notifier); +#endif + + if (*cpu_pattern != 0) { + cfs_cpt_table = cfs_cpt_table_create_pattern(cpu_pattern); + if (cfs_cpt_table == NULL) { + CERROR("Failed to create cptab from pattern %s\n", + cpu_pattern); + goto failed; + } + + } else { + cfs_cpt_table = cfs_cpt_table_create(cpu_npartitions); + if (cfs_cpt_table == NULL) { + CERROR("Failed to create ptable with npartitions %d\n", + cpu_npartitions); + goto failed; + } + } + + spin_lock(&cpt_data.cpt_lock); + if (cfs_cpt_table->ctb_version != cpt_data.cpt_version) { + spin_unlock(&cpt_data.cpt_lock); + CERROR("CPU hotplug/unplug during setup\n"); + goto failed; + } + spin_unlock(&cpt_data.cpt_lock); + + LCONSOLE(0, "HW CPU cores: %d, npartitions: %d\n", + num_online_cpus(), cfs_cpt_number(cfs_cpt_table)); + return 0; + + failed: + cfs_cpu_fini(); + return -1; +} + +#endif diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-crypto-adler.c b/drivers/staging/lustre/lustre/libcfs/linux/linux-crypto-adler.c new file mode 100644 index 000000000000..20b2d61d9ff2 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-crypto-adler.c @@ -0,0 +1,144 @@ +/* GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see http://www.gnu.org/licenses + * + * Please visit http://www.xyratex.com/contact if you need additional + * information or have any questions. + * + * GPL HEADER END + */ + +/* + * Copyright 2012 Xyratex Technology Limited + */ + +/* + * This is crypto api shash wrappers to zlib_adler32. + */ + +#include <linux/module.h> +#include <linux/zutil.h> +#include <crypto/internal/hash.h> + + +#define CHKSUM_BLOCK_SIZE 1 +#define CHKSUM_DIGEST_SIZE 4 + + +static u32 __adler32(u32 cksum, unsigned char const *p, size_t len) +{ + return zlib_adler32(cksum, p, len); +} + +static int adler32_cra_init(struct crypto_tfm *tfm) +{ + u32 *key = crypto_tfm_ctx(tfm); + + *key = 1; + + return 0; +} + +static int adler32_setkey(struct crypto_shash *hash, const u8 *key, + unsigned int keylen) +{ + u32 *mctx = crypto_shash_ctx(hash); + + if (keylen != sizeof(u32)) { + crypto_shash_set_flags(hash, CRYPTO_TFM_RES_BAD_KEY_LEN); + return -EINVAL; + } + *mctx = *(u32 *)key; + return 0; +} + +static int adler32_init(struct shash_desc *desc) +{ + u32 *mctx = crypto_shash_ctx(desc->tfm); + u32 *cksump = shash_desc_ctx(desc); + + *cksump = *mctx; + + return 0; +} + +static int adler32_update(struct shash_desc *desc, const u8 *data, + unsigned int len) +{ + u32 *cksump = shash_desc_ctx(desc); + + *cksump = __adler32(*cksump, data, len); + return 0; +} +static int __adler32_finup(u32 *cksump, const u8 *data, unsigned int len, + u8 *out) +{ + *(u32 *)out = __adler32(*cksump, data, len); + return 0; +} + +static int adler32_finup(struct shash_desc *desc, const u8 *data, + unsigned int len, u8 *out) +{ + return __adler32_finup(shash_desc_ctx(desc), data, len, out); +} + +static int adler32_final(struct shash_desc *desc, u8 *out) +{ + u32 *cksump = shash_desc_ctx(desc); + + *(u32 *)out = *cksump; + return 0; +} + +static int adler32_digest(struct shash_desc *desc, const u8 *data, + unsigned int len, u8 *out) +{ + return __adler32_finup(crypto_shash_ctx(desc->tfm), data, len, + out); +} +static struct shash_alg alg = { + .setkey = adler32_setkey, + .init = adler32_init, + .update = adler32_update, + .final = adler32_final, + .finup = adler32_finup, + .digest = adler32_digest, + .descsize = sizeof(u32), + .digestsize = CHKSUM_DIGEST_SIZE, + .base = { + .cra_name = "adler32", + .cra_driver_name = "adler32-zlib", + .cra_priority = 100, + .cra_blocksize = CHKSUM_BLOCK_SIZE, + .cra_ctxsize = sizeof(u32), + .cra_module = THIS_MODULE, + .cra_init = adler32_cra_init, + } +}; + + +int cfs_crypto_adler32_register(void) +{ + return crypto_register_shash(&alg); +} +EXPORT_SYMBOL(cfs_crypto_adler32_register); + +void cfs_crypto_adler32_unregister(void) +{ + crypto_unregister_shash(&alg); +} +EXPORT_SYMBOL(cfs_crypto_adler32_unregister); diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-crypto.c b/drivers/staging/lustre/lustre/libcfs/linux/linux-crypto.c new file mode 100644 index 000000000000..8e35777b4da5 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-crypto.c @@ -0,0 +1,289 @@ +/* GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see http://www.gnu.org/licenses + * + * Please visit http://www.xyratex.com/contact if you need additional + * information or have any questions. + * + * GPL HEADER END + */ + +/* + * Copyright 2012 Xyratex Technology Limited + * + * Copyright (c) 2012, Intel Corporation. + */ + +#include <linux/crypto.h> +#include <linux/scatterlist.h> +#include <linux/libcfs/libcfs.h> +#include <linux/libcfs/linux/linux-crypto.h> +/** + * Array of hash algorithm speed in MByte per second + */ +static int cfs_crypto_hash_speeds[CFS_HASH_ALG_MAX]; + + + +static int cfs_crypto_hash_alloc(unsigned char alg_id, + const struct cfs_crypto_hash_type **type, + struct hash_desc *desc, unsigned char *key, + unsigned int key_len) +{ + int err = 0; + + *type = cfs_crypto_hash_type(alg_id); + + if (*type == NULL) { + CWARN("Unsupported hash algorithm id = %d, max id is %d\n", + alg_id, CFS_HASH_ALG_MAX); + return -EINVAL; + } + desc->tfm = crypto_alloc_hash((*type)->cht_name, 0, 0); + + if (desc->tfm == NULL) + return -EINVAL; + + if (IS_ERR(desc->tfm)) { + CDEBUG(D_INFO, "Failed to alloc crypto hash %s\n", + (*type)->cht_name); + return PTR_ERR(desc->tfm); + } + + desc->flags = 0; + + /** Shash have different logic for initialization then digest + * shash: crypto_hash_setkey, crypto_hash_init + * digest: crypto_digest_init, crypto_digest_setkey + * Skip this function for digest, because we use shash logic at + * cfs_crypto_hash_alloc. + */ + if (key != NULL) { + err = crypto_hash_setkey(desc->tfm, key, key_len); + } else if ((*type)->cht_key != 0) { + err = crypto_hash_setkey(desc->tfm, + (unsigned char *)&((*type)->cht_key), + (*type)->cht_size); + } + + if (err != 0) { + crypto_free_hash(desc->tfm); + return err; + } + + CDEBUG(D_INFO, "Using crypto hash: %s (%s) speed %d MB/s\n", + (crypto_hash_tfm(desc->tfm))->__crt_alg->cra_name, + (crypto_hash_tfm(desc->tfm))->__crt_alg->cra_driver_name, + cfs_crypto_hash_speeds[alg_id]); + + return crypto_hash_init(desc); +} + +int cfs_crypto_hash_digest(unsigned char alg_id, + const void *buf, unsigned int buf_len, + unsigned char *key, unsigned int key_len, + unsigned char *hash, unsigned int *hash_len) +{ + struct scatterlist sl; + struct hash_desc hdesc; + int err; + const struct cfs_crypto_hash_type *type; + + if (buf == NULL || buf_len == 0 || hash_len == NULL) + return -EINVAL; + + err = cfs_crypto_hash_alloc(alg_id, &type, &hdesc, key, key_len); + if (err != 0) + return err; + + if (hash == NULL || *hash_len < type->cht_size) { + *hash_len = type->cht_size; + crypto_free_hash(hdesc.tfm); + return -ENOSPC; + } + sg_init_one(&sl, (void *)buf, buf_len); + + hdesc.flags = 0; + err = crypto_hash_digest(&hdesc, &sl, sl.length, hash); + crypto_free_hash(hdesc.tfm); + + return err; +} +EXPORT_SYMBOL(cfs_crypto_hash_digest); + +struct cfs_crypto_hash_desc * + cfs_crypto_hash_init(unsigned char alg_id, + unsigned char *key, unsigned int key_len) +{ + + struct hash_desc *hdesc; + int err; + const struct cfs_crypto_hash_type *type; + + hdesc = kmalloc(sizeof(*hdesc), 0); + if (hdesc == NULL) + return ERR_PTR(-ENOMEM); + + err = cfs_crypto_hash_alloc(alg_id, &type, hdesc, key, key_len); + + if (err) { + kfree(hdesc); + return ERR_PTR(err); + } + return (struct cfs_crypto_hash_desc *)hdesc; +} +EXPORT_SYMBOL(cfs_crypto_hash_init); + +int cfs_crypto_hash_update_page(struct cfs_crypto_hash_desc *hdesc, + struct page *page, unsigned int offset, + unsigned int len) +{ + struct scatterlist sl; + + sg_init_table(&sl, 1); + sg_set_page(&sl, page, len, offset & ~CFS_PAGE_MASK); + + return crypto_hash_update((struct hash_desc *)hdesc, &sl, sl.length); +} +EXPORT_SYMBOL(cfs_crypto_hash_update_page); + +int cfs_crypto_hash_update(struct cfs_crypto_hash_desc *hdesc, + const void *buf, unsigned int buf_len) +{ + struct scatterlist sl; + + sg_init_one(&sl, (void *)buf, buf_len); + + return crypto_hash_update((struct hash_desc *)hdesc, &sl, sl.length); +} +EXPORT_SYMBOL(cfs_crypto_hash_update); + +/* If hash_len pointer is NULL - destroy descriptor. */ +int cfs_crypto_hash_final(struct cfs_crypto_hash_desc *hdesc, + unsigned char *hash, unsigned int *hash_len) +{ + int err; + int size = crypto_hash_digestsize(((struct hash_desc *)hdesc)->tfm); + + if (hash_len == NULL) { + crypto_free_hash(((struct hash_desc *)hdesc)->tfm); + kfree(hdesc); + return 0; + } + if (hash == NULL || *hash_len < size) { + *hash_len = size; + return -ENOSPC; + } + err = crypto_hash_final((struct hash_desc *) hdesc, hash); + + if (err < 0) { + /* May be caller can fix error */ + return err; + } + crypto_free_hash(((struct hash_desc *)hdesc)->tfm); + kfree(hdesc); + return err; +} +EXPORT_SYMBOL(cfs_crypto_hash_final); + +static void cfs_crypto_performance_test(unsigned char alg_id, + const unsigned char *buf, + unsigned int buf_len) +{ + unsigned long start, end; + int bcount, err = 0; + int sec = 1; /* do test only 1 sec */ + unsigned char hash[64]; + unsigned int hash_len = 64; + + for (start = jiffies, end = start + sec * HZ, bcount = 0; + time_before(jiffies, end); bcount++) { + err = cfs_crypto_hash_digest(alg_id, buf, buf_len, NULL, 0, + hash, &hash_len); + if (err) + break; + + } + end = jiffies; + + if (err) { + cfs_crypto_hash_speeds[alg_id] = -1; + CDEBUG(D_INFO, "Crypto hash algorithm %s, err = %d\n", + cfs_crypto_hash_name(alg_id), err); + } else { + unsigned long tmp; + tmp = ((bcount * buf_len / jiffies_to_msecs(end - start)) * + 1000) / (1024 * 1024); + cfs_crypto_hash_speeds[alg_id] = (int)tmp; + } + CDEBUG(D_INFO, "Crypto hash algorithm %s speed = %d MB/s\n", + cfs_crypto_hash_name(alg_id), cfs_crypto_hash_speeds[alg_id]); +} + +int cfs_crypto_hash_speed(unsigned char hash_alg) +{ + if (hash_alg < CFS_HASH_ALG_MAX) + return cfs_crypto_hash_speeds[hash_alg]; + else + return -1; +} +EXPORT_SYMBOL(cfs_crypto_hash_speed); + +/** + * Do performance test for all hash algorithms. + */ +static int cfs_crypto_test_hashes(void) +{ + unsigned char i; + unsigned char *data; + unsigned int j; + /* Data block size for testing hash. Maximum + * kmalloc size for 2.6.18 kernel is 128K */ + unsigned int data_len = 1 * 128 * 1024; + + data = kmalloc(data_len, 0); + if (data == NULL) { + CERROR("Failed to allocate mem\n"); + return -ENOMEM; + } + + for (j = 0; j < data_len; j++) + data[j] = j & 0xff; + + for (i = 0; i < CFS_HASH_ALG_MAX; i++) + cfs_crypto_performance_test(i, data, data_len); + + kfree(data); + return 0; +} + +static int adler32; + +int cfs_crypto_register(void) +{ + adler32 = cfs_crypto_adler32_register(); + + /* check all algorithms and do performance test */ + cfs_crypto_test_hashes(); + return 0; +} +void cfs_crypto_unregister(void) +{ + if (adler32 == 0) + cfs_crypto_adler32_unregister(); + + return; +} diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-curproc.c b/drivers/staging/lustre/lustre/libcfs/linux/linux-curproc.c new file mode 100644 index 000000000000..f236510a2f3f --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-curproc.c @@ -0,0 +1,339 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2011, 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/linux/linux-curproc.c + * + * Lustre curproc API implementation for Linux kernel + * + * Author: Nikita Danilov <nikita@clusterfs.com> + */ + +#include <linux/sched.h> +#include <linux/fs_struct.h> + +#include <linux/compat.h> +#include <linux/thread_info.h> + +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> + +/* + * Implementation of cfs_curproc API (see portals/include/libcfs/curproc.h) + * for Linux kernel. + */ + +int cfs_curproc_groups_nr(void) +{ + int nr; + + task_lock(current); + nr = current_cred()->group_info->ngroups; + task_unlock(current); + return nr; +} + +void cfs_curproc_groups_dump(gid_t *array, int size) +{ + task_lock(current); + size = min_t(int, size, current_cred()->group_info->ngroups); + memcpy(array, current_cred()->group_info->blocks[0], size * sizeof(__u32)); + task_unlock(current); +} + + +int current_is_in_group(gid_t gid) +{ + return in_group_p(gid); +} + +/* Currently all the CFS_CAP_* defines match CAP_* ones. */ +#define cfs_cap_pack(cap) (cap) +#define cfs_cap_unpack(cap) (cap) + +void cfs_cap_raise(cfs_cap_t cap) +{ + struct cred *cred; + if ((cred = prepare_creds())) { + cap_raise(cred->cap_effective, cfs_cap_unpack(cap)); + commit_creds(cred); + } +} + +void cfs_cap_lower(cfs_cap_t cap) +{ + struct cred *cred; + if ((cred = prepare_creds())) { + cap_lower(cred->cap_effective, cfs_cap_unpack(cap)); + commit_creds(cred); + } +} + +int cfs_cap_raised(cfs_cap_t cap) +{ + return cap_raised(current_cap(), cfs_cap_unpack(cap)); +} + +void cfs_kernel_cap_pack(kernel_cap_t kcap, cfs_cap_t *cap) +{ +#if defined (_LINUX_CAPABILITY_VERSION) && _LINUX_CAPABILITY_VERSION == 0x19980330 + *cap = cfs_cap_pack(kcap); +#elif defined (_LINUX_CAPABILITY_VERSION) && _LINUX_CAPABILITY_VERSION == 0x20071026 + *cap = cfs_cap_pack(kcap[0]); +#elif defined(_KERNEL_CAPABILITY_VERSION) && _KERNEL_CAPABILITY_VERSION == 0x20080522 + /* XXX lost high byte */ + *cap = cfs_cap_pack(kcap.cap[0]); +#else + #error "need correct _KERNEL_CAPABILITY_VERSION " +#endif +} + +void cfs_kernel_cap_unpack(kernel_cap_t *kcap, cfs_cap_t cap) +{ +#if defined (_LINUX_CAPABILITY_VERSION) && _LINUX_CAPABILITY_VERSION == 0x19980330 + *kcap = cfs_cap_unpack(cap); +#elif defined (_LINUX_CAPABILITY_VERSION) && _LINUX_CAPABILITY_VERSION == 0x20071026 + (*kcap)[0] = cfs_cap_unpack(cap); +#elif defined(_KERNEL_CAPABILITY_VERSION) && _KERNEL_CAPABILITY_VERSION == 0x20080522 + kcap->cap[0] = cfs_cap_unpack(cap); +#else + #error "need correct _KERNEL_CAPABILITY_VERSION " +#endif +} + +cfs_cap_t cfs_curproc_cap_pack(void) +{ + cfs_cap_t cap; + cfs_kernel_cap_pack(current_cap(), &cap); + return cap; +} + +void cfs_curproc_cap_unpack(cfs_cap_t cap) +{ + struct cred *cred; + if ((cred = prepare_creds())) { + cfs_kernel_cap_unpack(&cred->cap_effective, cap); + commit_creds(cred); + } +} + +int cfs_capable(cfs_cap_t cap) +{ + return capable(cfs_cap_unpack(cap)); +} + +/* Check if task is running in 32-bit API mode, for the purpose of + * userspace binary interfaces. On 32-bit Linux this is (unfortunately) + * always true, even if the application is using LARGEFILE64 and 64-bit + * APIs, because Linux provides no way for the filesystem to know if it + * is called via 32-bit or 64-bit APIs. Other clients may vary. On + * 64-bit systems, this will only be true if the binary is calling a + * 32-bit system call. */ +int current_is_32bit(void) +{ + return is_compat_task(); +} + +static int cfs_access_process_vm(struct task_struct *tsk, unsigned long addr, + void *buf, int len, int write) +{ + /* Just copied from kernel for the kernels which doesn't + * have access_process_vm() exported */ + struct mm_struct *mm; + struct vm_area_struct *vma; + struct page *page; + void *old_buf = buf; + + mm = get_task_mm(tsk); + if (!mm) + return 0; + + down_read(&mm->mmap_sem); + /* ignore errors, just check how much was sucessfully transfered */ + while (len) { + int bytes, rc, offset; + void *maddr; + + rc = get_user_pages(tsk, mm, addr, 1, + write, 1, &page, &vma); + if (rc <= 0) + break; + + bytes = len; + offset = addr & (PAGE_SIZE-1); + if (bytes > PAGE_SIZE-offset) + bytes = PAGE_SIZE-offset; + + maddr = kmap(page); + if (write) { + copy_to_user_page(vma, page, addr, + maddr + offset, buf, bytes); + set_page_dirty_lock(page); + } else { + copy_from_user_page(vma, page, addr, + buf, maddr + offset, bytes); + } + kunmap(page); + page_cache_release(page); + len -= bytes; + buf += bytes; + addr += bytes; + } + up_read(&mm->mmap_sem); + mmput(mm); + + return buf - old_buf; +} + +/* Read the environment variable of current process specified by @key. */ +int cfs_get_environ(const char *key, char *value, int *val_len) +{ + struct mm_struct *mm; + char *buffer, *tmp_buf = NULL; + int buf_len = PAGE_CACHE_SIZE; + int key_len = strlen(key); + unsigned long addr; + int rc; + ENTRY; + + buffer = kmalloc(buf_len, GFP_USER); + if (!buffer) + RETURN(-ENOMEM); + + mm = get_task_mm(current); + if (!mm) { + kfree(buffer); + RETURN(-EINVAL); + } + + /* Avoid deadlocks on mmap_sem if called from sys_mmap_pgoff(), + * which is already holding mmap_sem for writes. If some other + * thread gets the write lock in the meantime, this thread will + * block, but at least it won't deadlock on itself. LU-1735 */ + if (down_read_trylock(&mm->mmap_sem) == 0) + return -EDEADLK; + up_read(&mm->mmap_sem); + + addr = mm->env_start; + while (addr < mm->env_end) { + int this_len, retval, scan_len; + char *env_start, *env_end; + + memset(buffer, 0, buf_len); + + this_len = min_t(int, mm->env_end - addr, buf_len); + retval = cfs_access_process_vm(current, addr, buffer, + this_len, 0); + if (retval != this_len) + break; + + addr += retval; + + /* Parse the buffer to find out the specified key/value pair. + * The "key=value" entries are separated by '\0'. */ + env_start = buffer; + scan_len = this_len; + while (scan_len) { + char *entry; + int entry_len; + + env_end = memscan(env_start, '\0', scan_len); + LASSERT(env_end >= env_start && + env_end <= env_start + scan_len); + + /* The last entry of this buffer cross the buffer + * boundary, reread it in next cycle. */ + if (unlikely(env_end - env_start == scan_len)) { + /* This entry is too large to fit in buffer */ + if (unlikely(scan_len == this_len)) { + CERROR("Too long env variable.\n"); + GOTO(out, rc = -EINVAL); + } + addr -= scan_len; + break; + } + + entry = env_start; + entry_len = env_end - env_start; + + /* Key length + length of '=' */ + if (entry_len > key_len + 1 && + !memcmp(entry, key, key_len)) { + entry += key_len + 1; + entry_len -= key_len + 1; + /* The 'value' buffer passed in is too small.*/ + if (entry_len >= *val_len) + GOTO(out, rc = -EOVERFLOW); + + memcpy(value, entry, entry_len); + *val_len = entry_len; + GOTO(out, rc = 0); + } + + scan_len -= (env_end - env_start + 1); + env_start = env_end + 1; + } + } + GOTO(out, rc = -ENOENT); + +out: + mmput(mm); + kfree((void *)buffer); + if (tmp_buf) + kfree((void *)tmp_buf); + return rc; +} +EXPORT_SYMBOL(cfs_get_environ); + +EXPORT_SYMBOL(cfs_curproc_groups_nr); +EXPORT_SYMBOL(cfs_curproc_groups_dump); +EXPORT_SYMBOL(current_is_in_group); +EXPORT_SYMBOL(cfs_cap_raise); +EXPORT_SYMBOL(cfs_cap_lower); +EXPORT_SYMBOL(cfs_cap_raised); +EXPORT_SYMBOL(cfs_curproc_cap_pack); +EXPORT_SYMBOL(cfs_curproc_cap_unpack); +EXPORT_SYMBOL(cfs_capable); +EXPORT_SYMBOL(current_is_32bit); + +/* + * Local variables: + * c-indentation-style: "K&R" + * c-basic-offset: 8 + * tab-width: 8 + * fill-column: 80 + * scroll-step: 1 + * End: + */ diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-debug.c b/drivers/staging/lustre/lustre/libcfs/linux/linux-debug.c new file mode 100644 index 000000000000..e2c195b8dd53 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-debug.c @@ -0,0 +1,264 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/linux/linux-debug.c + * + * Author: Phil Schwan <phil@clusterfs.com> + */ + +#include <linux/module.h> +#include <linux/kmod.h> +#include <linux/notifier.h> +#include <linux/kernel.h> +#include <linux/mm.h> +#include <linux/string.h> +#include <linux/stat.h> +#include <linux/errno.h> +#include <linux/unistd.h> +#include <linux/interrupt.h> +#include <asm/uaccess.h> +#include <linux/completion.h> + +#include <linux/fs.h> +#include <linux/stat.h> +#include <asm/uaccess.h> +#include <linux/miscdevice.h> +#include <linux/version.h> + +# define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> +#include <linux/libcfs/linux/portals_compat25.h> + +#include "tracefile.h" + +#include <linux/kallsyms.h> + +char lnet_upcall[1024] = "/usr/lib/lustre/lnet_upcall"; +char lnet_debug_log_upcall[1024] = "/usr/lib/lustre/lnet_debug_log_upcall"; + +/** + * Upcall function once a Lustre log has been dumped. + * + * \param file path of the dumped log + */ +void libcfs_run_debug_log_upcall(char *file) +{ + char *argv[3]; + int rc; + char *envp[] = { + "HOME=/", + "PATH=/sbin:/bin:/usr/sbin:/usr/bin", + NULL}; + ENTRY; + + argv[0] = lnet_debug_log_upcall; + + LASSERTF(file != NULL, "called on a null filename\n"); + argv[1] = file; //only need to pass the path of the file + + argv[2] = NULL; + + rc = USERMODEHELPER(argv[0], argv, envp); + if (rc < 0 && rc != -ENOENT) { + CERROR("Error %d invoking LNET debug log upcall %s %s; " + "check /proc/sys/lnet/debug_log_upcall\n", + rc, argv[0], argv[1]); + } else { + CDEBUG(D_HA, "Invoked LNET debug log upcall %s %s\n", + argv[0], argv[1]); + } + + EXIT; +} + +void libcfs_run_upcall(char **argv) +{ + int rc; + int argc; + char *envp[] = { + "HOME=/", + "PATH=/sbin:/bin:/usr/sbin:/usr/bin", + NULL}; + ENTRY; + + argv[0] = lnet_upcall; + argc = 1; + while (argv[argc] != NULL) + argc++; + + LASSERT(argc >= 2); + + rc = USERMODEHELPER(argv[0], argv, envp); + if (rc < 0 && rc != -ENOENT) { + CERROR("Error %d invoking LNET upcall %s %s%s%s%s%s%s%s%s; " + "check /proc/sys/lnet/upcall\n", + rc, argv[0], argv[1], + argc < 3 ? "" : ",", argc < 3 ? "" : argv[2], + argc < 4 ? "" : ",", argc < 4 ? "" : argv[3], + argc < 5 ? "" : ",", argc < 5 ? "" : argv[4], + argc < 6 ? "" : ",..."); + } else { + CDEBUG(D_HA, "Invoked LNET upcall %s %s%s%s%s%s%s%s%s\n", + argv[0], argv[1], + argc < 3 ? "" : ",", argc < 3 ? "" : argv[2], + argc < 4 ? "" : ",", argc < 4 ? "" : argv[3], + argc < 5 ? "" : ",", argc < 5 ? "" : argv[4], + argc < 6 ? "" : ",..."); + } +} + +void libcfs_run_lbug_upcall(struct libcfs_debug_msg_data *msgdata) +{ + char *argv[6]; + char buf[32]; + + ENTRY; + snprintf (buf, sizeof buf, "%d", msgdata->msg_line); + + argv[1] = "LBUG"; + argv[2] = (char *)msgdata->msg_file; + argv[3] = (char *)msgdata->msg_fn; + argv[4] = buf; + argv[5] = NULL; + + libcfs_run_upcall (argv); +} + +/* coverity[+kill] */ +void lbug_with_loc(struct libcfs_debug_msg_data *msgdata) +{ + libcfs_catastrophe = 1; + libcfs_debug_msg(msgdata, "LBUG\n"); + + if (in_interrupt()) { + panic("LBUG in interrupt.\n"); + /* not reached */ + } + + libcfs_debug_dumpstack(NULL); + if (!libcfs_panic_on_lbug) + libcfs_debug_dumplog(); + libcfs_run_lbug_upcall(msgdata); + if (libcfs_panic_on_lbug) + panic("LBUG"); + set_task_state(current, TASK_UNINTERRUPTIBLE); + while (1) + schedule(); +} + + +#include <linux/nmi.h> +#include <asm/stacktrace.h> + + +static int print_trace_stack(void *data, char *name) +{ + printk(" <%s> ", name); + return 0; +} + +# define RELIABLE reliable +# define DUMP_TRACE_CONST const +static void print_trace_address(void *data, unsigned long addr, int reliable) +{ + char fmt[32]; + touch_nmi_watchdog(); + sprintf(fmt, " [<%016lx>] %s%%s\n", addr, RELIABLE ? "": "? "); + __print_symbol(fmt, addr); +} + +static DUMP_TRACE_CONST struct stacktrace_ops print_trace_ops = { + .stack = print_trace_stack, + .address = print_trace_address, + .walk_stack = print_context_stack, +}; + +void libcfs_debug_dumpstack(struct task_struct *tsk) +{ + /* dump_stack() */ + /* show_trace() */ + if (tsk == NULL) + tsk = current; + printk("Pid: %d, comm: %.20s\n", tsk->pid, tsk->comm); + /* show_trace_log_lvl() */ + printk("\nCall Trace:\n"); + dump_trace(tsk, NULL, NULL, + 0, + &print_trace_ops, NULL); + printk("\n"); +} + +task_t *libcfs_current(void) +{ + CWARN("current task struct is %p\n", current); + return current; +} + +static int panic_notifier(struct notifier_block *self, unsigned long unused1, + void *unused2) +{ + if (libcfs_panic_in_progress) + return 0; + + libcfs_panic_in_progress = 1; + mb(); + + return 0; +} + +static struct notifier_block libcfs_panic_notifier = { + notifier_call : panic_notifier, + next : NULL, + priority : 10000 +}; + +void libcfs_register_panic_notifier(void) +{ + atomic_notifier_chain_register(&panic_notifier_list, &libcfs_panic_notifier); +} + +void libcfs_unregister_panic_notifier(void) +{ + atomic_notifier_chain_unregister(&panic_notifier_list, &libcfs_panic_notifier); +} + +EXPORT_SYMBOL(libcfs_debug_dumpstack); +EXPORT_SYMBOL(libcfs_current); + + +EXPORT_SYMBOL(libcfs_run_upcall); +EXPORT_SYMBOL(libcfs_run_lbug_upcall); +EXPORT_SYMBOL(lbug_with_loc); diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-module.c b/drivers/staging/lustre/lustre/libcfs/linux/linux-module.c new file mode 100644 index 000000000000..2c7d4a3d660f --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-module.c @@ -0,0 +1,183 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + */ + +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> + +#define LNET_MINOR 240 + +int libcfs_ioctl_getdata(char *buf, char *end, void *arg) +{ + struct libcfs_ioctl_hdr *hdr; + struct libcfs_ioctl_data *data; + int err; + ENTRY; + + hdr = (struct libcfs_ioctl_hdr *)buf; + data = (struct libcfs_ioctl_data *)buf; + + err = copy_from_user(buf, (void *)arg, sizeof(*hdr)); + if (err) + RETURN(err); + + if (hdr->ioc_version != LIBCFS_IOCTL_VERSION) { + CERROR("PORTALS: version mismatch kernel vs application\n"); + RETURN(-EINVAL); + } + + if (hdr->ioc_len + buf >= end) { + CERROR("PORTALS: user buffer exceeds kernel buffer\n"); + RETURN(-EINVAL); + } + + + if (hdr->ioc_len < sizeof(struct libcfs_ioctl_data)) { + CERROR("PORTALS: user buffer too small for ioctl\n"); + RETURN(-EINVAL); + } + + err = copy_from_user(buf, (void *)arg, hdr->ioc_len); + if (err) + RETURN(err); + + if (libcfs_ioctl_is_invalid(data)) { + CERROR("PORTALS: ioctl not correctly formatted\n"); + RETURN(-EINVAL); + } + + if (data->ioc_inllen1) + data->ioc_inlbuf1 = &data->ioc_bulk[0]; + + if (data->ioc_inllen2) + data->ioc_inlbuf2 = &data->ioc_bulk[0] + + cfs_size_round(data->ioc_inllen1); + + RETURN(0); +} + +int libcfs_ioctl_popdata(void *arg, void *data, int size) +{ + if (copy_to_user((char *)arg, data, size)) + return -EFAULT; + return 0; +} + +extern struct cfs_psdev_ops libcfs_psdev_ops; + +static int +libcfs_psdev_open(struct inode * inode, struct file * file) +{ + struct libcfs_device_userstate **pdu = NULL; + int rc = 0; + + if (!inode) + return (-EINVAL); + pdu = (struct libcfs_device_userstate **)&file->private_data; + if (libcfs_psdev_ops.p_open != NULL) + rc = libcfs_psdev_ops.p_open(0, (void *)pdu); + else + return (-EPERM); + return rc; +} + +/* called when closing /dev/device */ +static int +libcfs_psdev_release(struct inode * inode, struct file * file) +{ + struct libcfs_device_userstate *pdu; + int rc = 0; + + if (!inode) + return (-EINVAL); + pdu = file->private_data; + if (libcfs_psdev_ops.p_close != NULL) + rc = libcfs_psdev_ops.p_close(0, (void *)pdu); + else + rc = -EPERM; + return rc; +} + +static long libcfs_ioctl(struct file *file, + unsigned int cmd, unsigned long arg) +{ + struct cfs_psdev_file pfile; + int rc = 0; + + if (current_fsuid() != 0) + return -EACCES; + + if ( _IOC_TYPE(cmd) != IOC_LIBCFS_TYPE || + _IOC_NR(cmd) < IOC_LIBCFS_MIN_NR || + _IOC_NR(cmd) > IOC_LIBCFS_MAX_NR ) { + CDEBUG(D_IOCTL, "invalid ioctl ( type %d, nr %d, size %d )\n", + _IOC_TYPE(cmd), _IOC_NR(cmd), _IOC_SIZE(cmd)); + return (-EINVAL); + } + + /* Handle platform-dependent IOC requests */ + switch (cmd) { + case IOC_LIBCFS_PANIC: + if (!cfs_capable(CFS_CAP_SYS_BOOT)) + return (-EPERM); + panic("debugctl-invoked panic"); + return (0); + case IOC_LIBCFS_MEMHOG: + if (!cfs_capable(CFS_CAP_SYS_ADMIN)) + return -EPERM; + /* go thought */ + } + + pfile.off = 0; + pfile.private_data = file->private_data; + if (libcfs_psdev_ops.p_ioctl != NULL) + rc = libcfs_psdev_ops.p_ioctl(&pfile, cmd, (void *)arg); + else + rc = -EPERM; + return (rc); +} + +static struct file_operations libcfs_fops = { + unlocked_ioctl: libcfs_ioctl, + open : libcfs_psdev_open, + release : libcfs_psdev_release +}; + +psdev_t libcfs_dev = { + LNET_MINOR, + "lnet", + &libcfs_fops +}; diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-prim.c b/drivers/staging/lustre/lustre/libcfs/linux/linux-prim.c new file mode 100644 index 000000000000..b652a79a4811 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-prim.c @@ -0,0 +1,259 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2011, 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + */ + +#define DEBUG_SUBSYSTEM S_LNET +#include <linux/module.h> +#include <linux/kernel.h> +#include <linux/fs_struct.h> +#include <linux/sched.h> + +#include <linux/libcfs/libcfs.h> + +#if defined(CONFIG_KGDB) +#include <asm/kgdb.h> +#endif + +#define LINUX_WAITQ(w) ((wait_queue_t *) w) +#define LINUX_WAITQ_HEAD(w) ((wait_queue_head_t *) w) + +void +init_waitqueue_entry_current(wait_queue_t *link) +{ + init_waitqueue_entry(LINUX_WAITQ(link), current); +} +EXPORT_SYMBOL(init_waitqueue_entry_current); + +/** + * wait_queue_t of Linux (version < 2.6.34) is a FIFO list for exclusively + * waiting threads, which is not always desirable because all threads will + * be waken up again and again, even user only needs a few of them to be + * active most time. This is not good for performance because cache can + * be polluted by different threads. + * + * LIFO list can resolve this problem because we always wakeup the most + * recent active thread by default. + * + * NB: please don't call non-exclusive & exclusive wait on the same + * waitq if add_wait_queue_exclusive_head is used. + */ +void +add_wait_queue_exclusive_head(wait_queue_head_t *waitq, wait_queue_t *link) +{ + unsigned long flags; + + spin_lock_irqsave(&LINUX_WAITQ_HEAD(waitq)->lock, flags); + __add_wait_queue_exclusive(LINUX_WAITQ_HEAD(waitq), LINUX_WAITQ(link)); + spin_unlock_irqrestore(&LINUX_WAITQ_HEAD(waitq)->lock, flags); +} +EXPORT_SYMBOL(add_wait_queue_exclusive_head); + +void +waitq_wait(wait_queue_t *link, cfs_task_state_t state) +{ + schedule(); +} +EXPORT_SYMBOL(waitq_wait); + +int64_t +waitq_timedwait(wait_queue_t *link, cfs_task_state_t state, + int64_t timeout) +{ + return schedule_timeout(timeout); +} +EXPORT_SYMBOL(waitq_timedwait); + +void +schedule_timeout_and_set_state(cfs_task_state_t state, int64_t timeout) +{ + set_current_state(state); + schedule_timeout(timeout); +} +EXPORT_SYMBOL(schedule_timeout_and_set_state); + +/* deschedule for a bit... */ +void +cfs_pause(cfs_duration_t ticks) +{ + set_current_state(TASK_UNINTERRUPTIBLE); + schedule_timeout(ticks); +} +EXPORT_SYMBOL(cfs_pause); + +void cfs_init_timer(timer_list_t *t) +{ + init_timer(t); +} +EXPORT_SYMBOL(cfs_init_timer); + +void cfs_timer_init(timer_list_t *t, cfs_timer_func_t *func, void *arg) +{ + init_timer(t); + t->function = func; + t->data = (unsigned long)arg; +} +EXPORT_SYMBOL(cfs_timer_init); + +void cfs_timer_done(timer_list_t *t) +{ + return; +} +EXPORT_SYMBOL(cfs_timer_done); + +void cfs_timer_arm(timer_list_t *t, cfs_time_t deadline) +{ + mod_timer(t, deadline); +} +EXPORT_SYMBOL(cfs_timer_arm); + +void cfs_timer_disarm(timer_list_t *t) +{ + del_timer(t); +} +EXPORT_SYMBOL(cfs_timer_disarm); + +int cfs_timer_is_armed(timer_list_t *t) +{ + return timer_pending(t); +} +EXPORT_SYMBOL(cfs_timer_is_armed); + +cfs_time_t cfs_timer_deadline(timer_list_t *t) +{ + return t->expires; +} +EXPORT_SYMBOL(cfs_timer_deadline); + +void cfs_enter_debugger(void) +{ +#if defined(CONFIG_KGDB) +// BREAKPOINT(); +#else + /* nothing */ +#endif +} + + +sigset_t +cfs_block_allsigs(void) +{ + unsigned long flags; + sigset_t old; + + SIGNAL_MASK_LOCK(current, flags); + old = current->blocked; + sigfillset(¤t->blocked); + recalc_sigpending(); + SIGNAL_MASK_UNLOCK(current, flags); + + return old; +} + +sigset_t cfs_block_sigs(unsigned long sigs) +{ + unsigned long flags; + sigset_t old; + + SIGNAL_MASK_LOCK(current, flags); + old = current->blocked; + sigaddsetmask(¤t->blocked, sigs); + recalc_sigpending(); + SIGNAL_MASK_UNLOCK(current, flags); + return old; +} + +/* Block all signals except for the @sigs */ +sigset_t cfs_block_sigsinv(unsigned long sigs) +{ + unsigned long flags; + sigset_t old; + + SIGNAL_MASK_LOCK(current, flags); + old = current->blocked; + sigaddsetmask(¤t->blocked, ~sigs); + recalc_sigpending(); + SIGNAL_MASK_UNLOCK(current, flags); + + return old; +} + +void +cfs_restore_sigs (sigset_t old) +{ + unsigned long flags; + + SIGNAL_MASK_LOCK(current, flags); + current->blocked = old; + recalc_sigpending(); + SIGNAL_MASK_UNLOCK(current, flags); +} + +int +cfs_signal_pending(void) +{ + return signal_pending(current); +} + +void +cfs_clear_sigpending(void) +{ + unsigned long flags; + + SIGNAL_MASK_LOCK(current, flags); + clear_tsk_thread_flag(current, TIF_SIGPENDING); + SIGNAL_MASK_UNLOCK(current, flags); +} + +int +libcfs_arch_init(void) +{ + return 0; +} + +void +libcfs_arch_cleanup(void) +{ + return; +} + +EXPORT_SYMBOL(libcfs_arch_init); +EXPORT_SYMBOL(libcfs_arch_cleanup); +EXPORT_SYMBOL(cfs_enter_debugger); +EXPORT_SYMBOL(cfs_block_allsigs); +EXPORT_SYMBOL(cfs_block_sigs); +EXPORT_SYMBOL(cfs_block_sigsinv); +EXPORT_SYMBOL(cfs_restore_sigs); +EXPORT_SYMBOL(cfs_signal_pending); +EXPORT_SYMBOL(cfs_clear_sigpending); diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-proc.c b/drivers/staging/lustre/lustre/libcfs/linux/linux-proc.c new file mode 100644 index 000000000000..522b28e99e41 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-proc.c @@ -0,0 +1,580 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2011, 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/linux/linux-proc.c + * + * Author: Zach Brown <zab@zabbo.net> + * Author: Peter J. Braam <braam@clusterfs.com> + * Author: Phil Schwan <phil@clusterfs.com> + */ + +#include <linux/module.h> +#include <linux/kernel.h> +#include <linux/mm.h> +#include <linux/string.h> +#include <linux/stat.h> +#include <linux/errno.h> +#include <linux/unistd.h> +#include <net/sock.h> +#include <linux/uio.h> + +#include <asm/uaccess.h> + +#include <linux/fs.h> +#include <linux/file.h> +#include <linux/stat.h> +#include <linux/list.h> +#include <asm/uaccess.h> + +#include <linux/proc_fs.h> +#include <linux/sysctl.h> + +# define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> +#include <asm/div64.h> +#include "tracefile.h" + +#ifdef CONFIG_SYSCTL +static ctl_table_header_t *lnet_table_header = NULL; +#endif +extern char lnet_upcall[1024]; +/** + * The path of debug log dump upcall script. + */ +extern char lnet_debug_log_upcall[1024]; + +#define CTL_LNET (0x100) +enum { + PSDEV_DEBUG = 1, /* control debugging */ + PSDEV_SUBSYSTEM_DEBUG, /* control debugging */ + PSDEV_PRINTK, /* force all messages to console */ + PSDEV_CONSOLE_RATELIMIT, /* ratelimit console messages */ + PSDEV_CONSOLE_MAX_DELAY_CS, /* maximum delay over which we skip messages */ + PSDEV_CONSOLE_MIN_DELAY_CS, /* initial delay over which we skip messages */ + PSDEV_CONSOLE_BACKOFF, /* delay increase factor */ + PSDEV_DEBUG_PATH, /* crashdump log location */ + PSDEV_DEBUG_DUMP_PATH, /* crashdump tracelog location */ + PSDEV_CPT_TABLE, /* information about cpu partitions */ + PSDEV_LNET_UPCALL, /* User mode upcall script */ + PSDEV_LNET_MEMUSED, /* bytes currently PORTAL_ALLOCated */ + PSDEV_LNET_CATASTROPHE, /* if we have LBUGged or panic'd */ + PSDEV_LNET_PANIC_ON_LBUG, /* flag to panic on LBUG */ + PSDEV_LNET_DUMP_KERNEL, /* snapshot kernel debug buffer to file */ + PSDEV_LNET_DAEMON_FILE, /* spool kernel debug buffer to file */ + PSDEV_LNET_DEBUG_MB, /* size of debug buffer */ + PSDEV_LNET_DEBUG_LOG_UPCALL, /* debug log upcall script */ + PSDEV_LNET_WATCHDOG_RATELIMIT, /* ratelimit watchdog messages */ + PSDEV_LNET_FORCE_LBUG, /* hook to force an LBUG */ + PSDEV_LNET_FAIL_LOC, /* control test failures instrumentation */ + PSDEV_LNET_FAIL_VAL, /* userdata for fail loc */ +}; + +int +proc_call_handler(void *data, int write, + loff_t *ppos, void *buffer, size_t *lenp, + int (*handler)(void *data, int write, + loff_t pos, void *buffer, int len)) +{ + int rc = handler(data, write, *ppos, buffer, *lenp); + + if (rc < 0) + return rc; + + if (write) { + *ppos += *lenp; + } else { + *lenp = rc; + *ppos += rc; + } + return 0; +} +EXPORT_SYMBOL(proc_call_handler); + +static int __proc_dobitmasks(void *data, int write, + loff_t pos, void *buffer, int nob) +{ + const int tmpstrlen = 512; + char *tmpstr; + int rc; + unsigned int *mask = data; + int is_subsys = (mask == &libcfs_subsystem_debug) ? 1 : 0; + int is_printk = (mask == &libcfs_printk) ? 1 : 0; + + rc = cfs_trace_allocate_string_buffer(&tmpstr, tmpstrlen); + if (rc < 0) + return rc; + + if (!write) { + libcfs_debug_mask2str(tmpstr, tmpstrlen, *mask, is_subsys); + rc = strlen(tmpstr); + + if (pos >= rc) { + rc = 0; + } else { + rc = cfs_trace_copyout_string(buffer, nob, + tmpstr + pos, "\n"); + } + } else { + rc = cfs_trace_copyin_string(tmpstr, tmpstrlen, buffer, nob); + if (rc < 0) { + cfs_trace_free_string_buffer(tmpstr, tmpstrlen); + return rc; + } + + rc = libcfs_debug_str2mask(mask, tmpstr, is_subsys); + /* Always print LBUG/LASSERT to console, so keep this mask */ + if (is_printk) + *mask |= D_EMERG; + } + + cfs_trace_free_string_buffer(tmpstr, tmpstrlen); + return rc; +} + +DECLARE_PROC_HANDLER(proc_dobitmasks) + +static int min_watchdog_ratelimit = 0; /* disable ratelimiting */ +static int max_watchdog_ratelimit = (24*60*60); /* limit to once per day */ + +static int __proc_dump_kernel(void *data, int write, + loff_t pos, void *buffer, int nob) +{ + if (!write) + return 0; + + return cfs_trace_dump_debug_buffer_usrstr(buffer, nob); +} + +DECLARE_PROC_HANDLER(proc_dump_kernel) + +static int __proc_daemon_file(void *data, int write, + loff_t pos, void *buffer, int nob) +{ + if (!write) { + int len = strlen(cfs_tracefile); + + if (pos >= len) + return 0; + + return cfs_trace_copyout_string(buffer, nob, + cfs_tracefile + pos, "\n"); + } + + return cfs_trace_daemon_command_usrstr(buffer, nob); +} + +DECLARE_PROC_HANDLER(proc_daemon_file) + +static int __proc_debug_mb(void *data, int write, + loff_t pos, void *buffer, int nob) +{ + if (!write) { + char tmpstr[32]; + int len = snprintf(tmpstr, sizeof(tmpstr), "%d", + cfs_trace_get_debug_mb()); + + if (pos >= len) + return 0; + + return cfs_trace_copyout_string(buffer, nob, tmpstr + pos, + "\n"); + } + + return cfs_trace_set_debug_mb_usrstr(buffer, nob); +} + +DECLARE_PROC_HANDLER(proc_debug_mb) + +int LL_PROC_PROTO(proc_console_max_delay_cs) +{ + int rc, max_delay_cs; + ctl_table_t dummy = *table; + cfs_duration_t d; + + dummy.data = &max_delay_cs; + dummy.proc_handler = &proc_dointvec; + + if (!write) { /* read */ + max_delay_cs = cfs_duration_sec(libcfs_console_max_delay * 100); + rc = ll_proc_dointvec(&dummy, write, filp, buffer, lenp, ppos); + return rc; + } + + /* write */ + max_delay_cs = 0; + rc = ll_proc_dointvec(&dummy, write, filp, buffer, lenp, ppos); + if (rc < 0) + return rc; + if (max_delay_cs <= 0) + return -EINVAL; + + d = cfs_time_seconds(max_delay_cs) / 100; + if (d == 0 || d < libcfs_console_min_delay) + return -EINVAL; + libcfs_console_max_delay = d; + + return rc; +} + +int LL_PROC_PROTO(proc_console_min_delay_cs) +{ + int rc, min_delay_cs; + ctl_table_t dummy = *table; + cfs_duration_t d; + + dummy.data = &min_delay_cs; + dummy.proc_handler = &proc_dointvec; + + if (!write) { /* read */ + min_delay_cs = cfs_duration_sec(libcfs_console_min_delay * 100); + rc = ll_proc_dointvec(&dummy, write, filp, buffer, lenp, ppos); + return rc; + } + + /* write */ + min_delay_cs = 0; + rc = ll_proc_dointvec(&dummy, write, filp, buffer, lenp, ppos); + if (rc < 0) + return rc; + if (min_delay_cs <= 0) + return -EINVAL; + + d = cfs_time_seconds(min_delay_cs) / 100; + if (d == 0 || d > libcfs_console_max_delay) + return -EINVAL; + libcfs_console_min_delay = d; + + return rc; +} + +int LL_PROC_PROTO(proc_console_backoff) +{ + int rc, backoff; + ctl_table_t dummy = *table; + + dummy.data = &backoff; + dummy.proc_handler = &proc_dointvec; + + if (!write) { /* read */ + backoff= libcfs_console_backoff; + rc = ll_proc_dointvec(&dummy, write, filp, buffer, lenp, ppos); + return rc; + } + + /* write */ + backoff = 0; + rc = ll_proc_dointvec(&dummy, write, filp, buffer, lenp, ppos); + if (rc < 0) + return rc; + if (backoff <= 0) + return -EINVAL; + + libcfs_console_backoff = backoff; + + return rc; +} + +int LL_PROC_PROTO(libcfs_force_lbug) +{ + if (write) + LBUG(); + return 0; +} + +int LL_PROC_PROTO(proc_fail_loc) +{ + int rc; + long old_fail_loc = cfs_fail_loc; + + rc = ll_proc_dolongvec(table, write, filp, buffer, lenp, ppos); + if (old_fail_loc != cfs_fail_loc) + wake_up(&cfs_race_waitq); + return rc; +} + +static int __proc_cpt_table(void *data, int write, + loff_t pos, void *buffer, int nob) +{ + char *buf = NULL; + int len = 4096; + int rc = 0; + + if (write) + return -EPERM; + + LASSERT(cfs_cpt_table != NULL); + + while (1) { + LIBCFS_ALLOC(buf, len); + if (buf == NULL) + return -ENOMEM; + + rc = cfs_cpt_table_print(cfs_cpt_table, buf, len); + if (rc >= 0) + break; + + LIBCFS_FREE(buf, len); + if (rc == -EFBIG) { + len <<= 1; + continue; + } + goto out; + } + + if (pos >= rc) { + rc = 0; + goto out; + } + + rc = cfs_trace_copyout_string(buffer, nob, buf + pos, NULL); + out: + if (buf != NULL) + LIBCFS_FREE(buf, len); + return rc; +} +DECLARE_PROC_HANDLER(proc_cpt_table) + +static ctl_table_t lnet_table[] = { + /* + * NB No .strategy entries have been provided since sysctl(8) prefers + * to go via /proc for portability. + */ + { + INIT_CTL_NAME(PSDEV_DEBUG) + .procname = "debug", + .data = &libcfs_debug, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dobitmasks, + }, + { + INIT_CTL_NAME(PSDEV_SUBSYSTEM_DEBUG) + .procname = "subsystem_debug", + .data = &libcfs_subsystem_debug, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dobitmasks, + }, + { + INIT_CTL_NAME(PSDEV_PRINTK) + .procname = "printk", + .data = &libcfs_printk, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dobitmasks, + }, + { + INIT_CTL_NAME(PSDEV_CONSOLE_RATELIMIT) + .procname = "console_ratelimit", + .data = &libcfs_console_ratelimit, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dointvec + }, + { + INIT_CTL_NAME(PSDEV_CONSOLE_MAX_DELAY_CS) + .procname = "console_max_delay_centisecs", + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_console_max_delay_cs + }, + { + INIT_CTL_NAME(PSDEV_CONSOLE_MIN_DELAY_CS) + .procname = "console_min_delay_centisecs", + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_console_min_delay_cs + }, + { + INIT_CTL_NAME(PSDEV_CONSOLE_BACKOFF) + .procname = "console_backoff", + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_console_backoff + }, + + { + INIT_CTL_NAME(PSDEV_DEBUG_PATH) + .procname = "debug_path", + .data = libcfs_debug_file_path_arr, + .maxlen = sizeof(libcfs_debug_file_path_arr), + .mode = 0644, + .proc_handler = &proc_dostring, + }, + + { + INIT_CTL_NAME(PSDEV_CPT_TABLE) + .procname = "cpu_partition_table", + .maxlen = 128, + .mode = 0444, + .proc_handler = &proc_cpt_table, + }, + + { + INIT_CTL_NAME(PSDEV_LNET_UPCALL) + .procname = "upcall", + .data = lnet_upcall, + .maxlen = sizeof(lnet_upcall), + .mode = 0644, + .proc_handler = &proc_dostring, + }, + { + INIT_CTL_NAME(PSDEV_LNET_DEBUG_LOG_UPCALL) + .procname = "debug_log_upcall", + .data = lnet_debug_log_upcall, + .maxlen = sizeof(lnet_debug_log_upcall), + .mode = 0644, + .proc_handler = &proc_dostring, + }, + { + INIT_CTL_NAME(PSDEV_LNET_MEMUSED) + .procname = "lnet_memused", + .data = (int *)&libcfs_kmemory.counter, + .maxlen = sizeof(int), + .mode = 0444, + .proc_handler = &proc_dointvec, + INIT_STRATEGY(&sysctl_intvec) + }, + { + INIT_CTL_NAME(PSDEV_LNET_CATASTROPHE) + .procname = "catastrophe", + .data = &libcfs_catastrophe, + .maxlen = sizeof(int), + .mode = 0444, + .proc_handler = &proc_dointvec, + INIT_STRATEGY(&sysctl_intvec) + }, + { + INIT_CTL_NAME(PSDEV_LNET_PANIC_ON_LBUG) + .procname = "panic_on_lbug", + .data = &libcfs_panic_on_lbug, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dointvec, + INIT_STRATEGY(&sysctl_intvec) + }, + { + INIT_CTL_NAME(PSDEV_LNET_DUMP_KERNEL) + .procname = "dump_kernel", + .maxlen = 256, + .mode = 0200, + .proc_handler = &proc_dump_kernel, + }, + { + INIT_CTL_NAME(PSDEV_LNET_DAEMON_FILE) + .procname = "daemon_file", + .mode = 0644, + .maxlen = 256, + .proc_handler = &proc_daemon_file, + }, + { + INIT_CTL_NAME(PSDEV_LNET_DEBUG_MB) + .procname = "debug_mb", + .mode = 0644, + .proc_handler = &proc_debug_mb, + }, + { + INIT_CTL_NAME(PSDEV_LNET_WATCHDOG_RATELIMIT) + .procname = "watchdog_ratelimit", + .data = &libcfs_watchdog_ratelimit, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dointvec_minmax, + .extra1 = &min_watchdog_ratelimit, + .extra2 = &max_watchdog_ratelimit, + }, + { INIT_CTL_NAME(PSDEV_LNET_FORCE_LBUG) + .procname = "force_lbug", + .data = NULL, + .maxlen = 0, + .mode = 0200, + .proc_handler = &libcfs_force_lbug + }, + { + INIT_CTL_NAME(PSDEV_LNET_FAIL_LOC) + .procname = "fail_loc", + .data = &cfs_fail_loc, + .maxlen = sizeof(cfs_fail_loc), + .mode = 0644, + .proc_handler = &proc_fail_loc + }, + { + INIT_CTL_NAME(PSDEV_LNET_FAIL_VAL) + .procname = "fail_val", + .data = &cfs_fail_val, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = &proc_dointvec + }, + { + INIT_CTL_NAME(0) + } +}; + +#ifdef CONFIG_SYSCTL +static ctl_table_t top_table[] = { + { + INIT_CTL_NAME(CTL_LNET) + .procname = "lnet", + .mode = 0555, + .data = NULL, + .maxlen = 0, + .child = lnet_table, + }, + { + INIT_CTL_NAME(0) + } +}; +#endif + +int insert_proc(void) +{ +#ifdef CONFIG_SYSCTL + if (lnet_table_header == NULL) + lnet_table_header = cfs_register_sysctl_table(top_table, 0); +#endif + return 0; +} + +void remove_proc(void) +{ +#ifdef CONFIG_SYSCTL + if (lnet_table_header != NULL) + unregister_sysctl_table(lnet_table_header); + + lnet_table_header = NULL; +#endif +} diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-tcpip.c b/drivers/staging/lustre/lustre/libcfs/linux/linux-tcpip.c new file mode 100644 index 000000000000..855c7e87d96f --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-tcpip.c @@ -0,0 +1,659 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + */ +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> +#include <linux/libcfs/libcfs.h> + +#include <linux/if.h> +#include <linux/in.h> +#include <linux/file.h> +/* For sys_open & sys_close */ +#include <linux/syscalls.h> + +int +libcfs_sock_ioctl(int cmd, unsigned long arg) +{ + mm_segment_t oldmm = get_fs(); + struct socket *sock; + int rc; + struct file *sock_filp; + + rc = sock_create (PF_INET, SOCK_STREAM, 0, &sock); + if (rc != 0) { + CERROR ("Can't create socket: %d\n", rc); + return rc; + } + + sock_filp = sock_alloc_file(sock, 0, NULL); + if (IS_ERR(sock_filp)) { + sock_release(sock); + rc = PTR_ERR(sock_filp); + goto out; + } + + set_fs(KERNEL_DS); + if (sock_filp->f_op->unlocked_ioctl) + rc = sock_filp->f_op->unlocked_ioctl(sock_filp, cmd, arg); + set_fs(oldmm); + + fput(sock_filp); +out: + return rc; +} + +int +libcfs_ipif_query (char *name, int *up, __u32 *ip, __u32 *mask) +{ + struct ifreq ifr; + int nob; + int rc; + __u32 val; + + nob = strnlen(name, IFNAMSIZ); + if (nob == IFNAMSIZ) { + CERROR("Interface name %s too long\n", name); + return -EINVAL; + } + + CLASSERT (sizeof(ifr.ifr_name) >= IFNAMSIZ); + + strcpy(ifr.ifr_name, name); + rc = libcfs_sock_ioctl(SIOCGIFFLAGS, (unsigned long)&ifr); + + if (rc != 0) { + CERROR("Can't get flags for interface %s\n", name); + return rc; + } + + if ((ifr.ifr_flags & IFF_UP) == 0) { + CDEBUG(D_NET, "Interface %s down\n", name); + *up = 0; + *ip = *mask = 0; + return 0; + } + + *up = 1; + + strcpy(ifr.ifr_name, name); + ifr.ifr_addr.sa_family = AF_INET; + rc = libcfs_sock_ioctl(SIOCGIFADDR, (unsigned long)&ifr); + + if (rc != 0) { + CERROR("Can't get IP address for interface %s\n", name); + return rc; + } + + val = ((struct sockaddr_in *)&ifr.ifr_addr)->sin_addr.s_addr; + *ip = ntohl(val); + + strcpy(ifr.ifr_name, name); + ifr.ifr_addr.sa_family = AF_INET; + rc = libcfs_sock_ioctl(SIOCGIFNETMASK, (unsigned long)&ifr); + + if (rc != 0) { + CERROR("Can't get netmask for interface %s\n", name); + return rc; + } + + val = ((struct sockaddr_in *)&ifr.ifr_netmask)->sin_addr.s_addr; + *mask = ntohl(val); + + return 0; +} + +EXPORT_SYMBOL(libcfs_ipif_query); + +int +libcfs_ipif_enumerate (char ***namesp) +{ + /* Allocate and fill in 'names', returning # interfaces/error */ + char **names; + int toobig; + int nalloc; + int nfound; + struct ifreq *ifr; + struct ifconf ifc; + int rc; + int nob; + int i; + + + nalloc = 16; /* first guess at max interfaces */ + toobig = 0; + for (;;) { + if (nalloc * sizeof(*ifr) > PAGE_CACHE_SIZE) { + toobig = 1; + nalloc = PAGE_CACHE_SIZE/sizeof(*ifr); + CWARN("Too many interfaces: only enumerating first %d\n", + nalloc); + } + + LIBCFS_ALLOC(ifr, nalloc * sizeof(*ifr)); + if (ifr == NULL) { + CERROR ("ENOMEM enumerating up to %d interfaces\n", nalloc); + rc = -ENOMEM; + goto out0; + } + + ifc.ifc_buf = (char *)ifr; + ifc.ifc_len = nalloc * sizeof(*ifr); + + rc = libcfs_sock_ioctl(SIOCGIFCONF, (unsigned long)&ifc); + + if (rc < 0) { + CERROR ("Error %d enumerating interfaces\n", rc); + goto out1; + } + + LASSERT (rc == 0); + + nfound = ifc.ifc_len/sizeof(*ifr); + LASSERT (nfound <= nalloc); + + if (nfound < nalloc || toobig) + break; + + LIBCFS_FREE(ifr, nalloc * sizeof(*ifr)); + nalloc *= 2; + } + + if (nfound == 0) + goto out1; + + LIBCFS_ALLOC(names, nfound * sizeof(*names)); + if (names == NULL) { + rc = -ENOMEM; + goto out1; + } + /* NULL out all names[i] */ + memset (names, 0, nfound * sizeof(*names)); + + for (i = 0; i < nfound; i++) { + + nob = strnlen (ifr[i].ifr_name, IFNAMSIZ); + if (nob == IFNAMSIZ) { + /* no space for terminating NULL */ + CERROR("interface name %.*s too long (%d max)\n", + nob, ifr[i].ifr_name, IFNAMSIZ); + rc = -ENAMETOOLONG; + goto out2; + } + + LIBCFS_ALLOC(names[i], IFNAMSIZ); + if (names[i] == NULL) { + rc = -ENOMEM; + goto out2; + } + + memcpy(names[i], ifr[i].ifr_name, nob); + names[i][nob] = 0; + } + + *namesp = names; + rc = nfound; + + out2: + if (rc < 0) + libcfs_ipif_free_enumeration(names, nfound); + out1: + LIBCFS_FREE(ifr, nalloc * sizeof(*ifr)); + out0: + return rc; +} + +EXPORT_SYMBOL(libcfs_ipif_enumerate); + +void +libcfs_ipif_free_enumeration (char **names, int n) +{ + int i; + + LASSERT (n > 0); + + for (i = 0; i < n && names[i] != NULL; i++) + LIBCFS_FREE(names[i], IFNAMSIZ); + + LIBCFS_FREE(names, n * sizeof(*names)); +} + +EXPORT_SYMBOL(libcfs_ipif_free_enumeration); + +int +libcfs_sock_write (struct socket *sock, void *buffer, int nob, int timeout) +{ + int rc; + mm_segment_t oldmm = get_fs(); + long ticks = timeout * HZ; + unsigned long then; + struct timeval tv; + + LASSERT (nob > 0); + /* Caller may pass a zero timeout if she thinks the socket buffer is + * empty enough to take the whole message immediately */ + + for (;;) { + struct iovec iov = { + .iov_base = buffer, + .iov_len = nob + }; + struct msghdr msg = { + .msg_name = NULL, + .msg_namelen = 0, + .msg_iov = &iov, + .msg_iovlen = 1, + .msg_control = NULL, + .msg_controllen = 0, + .msg_flags = (timeout == 0) ? MSG_DONTWAIT : 0 + }; + + if (timeout != 0) { + /* Set send timeout to remaining time */ + tv = (struct timeval) { + .tv_sec = ticks / HZ, + .tv_usec = ((ticks % HZ) * 1000000) / HZ + }; + set_fs(KERNEL_DS); + rc = sock_setsockopt(sock, SOL_SOCKET, SO_SNDTIMEO, + (char *)&tv, sizeof(tv)); + set_fs(oldmm); + if (rc != 0) { + CERROR("Can't set socket send timeout " + "%ld.%06d: %d\n", + (long)tv.tv_sec, (int)tv.tv_usec, rc); + return rc; + } + } + + set_fs (KERNEL_DS); + then = jiffies; + rc = sock_sendmsg (sock, &msg, iov.iov_len); + ticks -= jiffies - then; + set_fs (oldmm); + + if (rc == nob) + return 0; + + if (rc < 0) + return rc; + + if (rc == 0) { + CERROR ("Unexpected zero rc\n"); + return (-ECONNABORTED); + } + + if (ticks <= 0) + return -EAGAIN; + + buffer = ((char *)buffer) + rc; + nob -= rc; + } + + return (0); +} +EXPORT_SYMBOL(libcfs_sock_write); + +int +libcfs_sock_read (struct socket *sock, void *buffer, int nob, int timeout) +{ + int rc; + mm_segment_t oldmm = get_fs(); + long ticks = timeout * HZ; + unsigned long then; + struct timeval tv; + + LASSERT (nob > 0); + LASSERT (ticks > 0); + + for (;;) { + struct iovec iov = { + .iov_base = buffer, + .iov_len = nob + }; + struct msghdr msg = { + .msg_name = NULL, + .msg_namelen = 0, + .msg_iov = &iov, + .msg_iovlen = 1, + .msg_control = NULL, + .msg_controllen = 0, + .msg_flags = 0 + }; + + /* Set receive timeout to remaining time */ + tv = (struct timeval) { + .tv_sec = ticks / HZ, + .tv_usec = ((ticks % HZ) * 1000000) / HZ + }; + set_fs(KERNEL_DS); + rc = sock_setsockopt(sock, SOL_SOCKET, SO_RCVTIMEO, + (char *)&tv, sizeof(tv)); + set_fs(oldmm); + if (rc != 0) { + CERROR("Can't set socket recv timeout %ld.%06d: %d\n", + (long)tv.tv_sec, (int)tv.tv_usec, rc); + return rc; + } + + set_fs(KERNEL_DS); + then = jiffies; + rc = sock_recvmsg(sock, &msg, iov.iov_len, 0); + ticks -= jiffies - then; + set_fs(oldmm); + + if (rc < 0) + return rc; + + if (rc == 0) + return -ECONNRESET; + + buffer = ((char *)buffer) + rc; + nob -= rc; + + if (nob == 0) + return 0; + + if (ticks <= 0) + return -ETIMEDOUT; + } +} + +EXPORT_SYMBOL(libcfs_sock_read); + +static int +libcfs_sock_create (struct socket **sockp, int *fatal, + __u32 local_ip, int local_port) +{ + struct sockaddr_in locaddr; + struct socket *sock; + int rc; + int option; + mm_segment_t oldmm = get_fs(); + + /* All errors are fatal except bind failure if the port is in use */ + *fatal = 1; + + rc = sock_create (PF_INET, SOCK_STREAM, 0, &sock); + *sockp = sock; + if (rc != 0) { + CERROR ("Can't create socket: %d\n", rc); + return (rc); + } + + set_fs (KERNEL_DS); + option = 1; + rc = sock_setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, + (char *)&option, sizeof (option)); + set_fs (oldmm); + if (rc != 0) { + CERROR("Can't set SO_REUSEADDR for socket: %d\n", rc); + goto failed; + } + + if (local_ip != 0 || local_port != 0) { + memset(&locaddr, 0, sizeof(locaddr)); + locaddr.sin_family = AF_INET; + locaddr.sin_port = htons(local_port); + locaddr.sin_addr.s_addr = (local_ip == 0) ? + INADDR_ANY : htonl(local_ip); + + rc = sock->ops->bind(sock, (struct sockaddr *)&locaddr, + sizeof(locaddr)); + if (rc == -EADDRINUSE) { + CDEBUG(D_NET, "Port %d already in use\n", local_port); + *fatal = 0; + goto failed; + } + if (rc != 0) { + CERROR("Error trying to bind to port %d: %d\n", + local_port, rc); + goto failed; + } + } + + return 0; + + failed: + sock_release(sock); + return rc; +} + +int +libcfs_sock_setbuf (struct socket *sock, int txbufsize, int rxbufsize) +{ + mm_segment_t oldmm = get_fs(); + int option; + int rc; + + if (txbufsize != 0) { + option = txbufsize; + set_fs (KERNEL_DS); + rc = sock_setsockopt(sock, SOL_SOCKET, SO_SNDBUF, + (char *)&option, sizeof (option)); + set_fs (oldmm); + if (rc != 0) { + CERROR ("Can't set send buffer %d: %d\n", + option, rc); + return (rc); + } + } + + if (rxbufsize != 0) { + option = rxbufsize; + set_fs (KERNEL_DS); + rc = sock_setsockopt (sock, SOL_SOCKET, SO_RCVBUF, + (char *)&option, sizeof (option)); + set_fs (oldmm); + if (rc != 0) { + CERROR ("Can't set receive buffer %d: %d\n", + option, rc); + return (rc); + } + } + + return 0; +} + +EXPORT_SYMBOL(libcfs_sock_setbuf); + +int +libcfs_sock_getaddr (struct socket *sock, int remote, __u32 *ip, int *port) +{ + struct sockaddr_in sin; + int len = sizeof (sin); + int rc; + + rc = sock->ops->getname (sock, (struct sockaddr *)&sin, &len, + remote ? 2 : 0); + if (rc != 0) { + CERROR ("Error %d getting sock %s IP/port\n", + rc, remote ? "peer" : "local"); + return rc; + } + + if (ip != NULL) + *ip = ntohl (sin.sin_addr.s_addr); + + if (port != NULL) + *port = ntohs (sin.sin_port); + + return 0; +} + +EXPORT_SYMBOL(libcfs_sock_getaddr); + +int +libcfs_sock_getbuf (struct socket *sock, int *txbufsize, int *rxbufsize) +{ + + if (txbufsize != NULL) { + *txbufsize = sock->sk->sk_sndbuf; + } + + if (rxbufsize != NULL) { + *rxbufsize = sock->sk->sk_rcvbuf; + } + + return 0; +} + +EXPORT_SYMBOL(libcfs_sock_getbuf); + +int +libcfs_sock_listen (struct socket **sockp, + __u32 local_ip, int local_port, int backlog) +{ + int fatal; + int rc; + + rc = libcfs_sock_create(sockp, &fatal, local_ip, local_port); + if (rc != 0) { + if (!fatal) + CERROR("Can't create socket: port %d already in use\n", + local_port); + return rc; + } + + rc = (*sockp)->ops->listen(*sockp, backlog); + if (rc == 0) + return 0; + + CERROR("Can't set listen backlog %d: %d\n", backlog, rc); + sock_release(*sockp); + return rc; +} + +EXPORT_SYMBOL(libcfs_sock_listen); + +int +libcfs_sock_accept (struct socket **newsockp, struct socket *sock) +{ + wait_queue_t wait; + struct socket *newsock; + int rc; + + init_waitqueue_entry(&wait, current); + + /* XXX this should add a ref to sock->ops->owner, if + * TCP could be a module */ + rc = sock_create_lite(PF_PACKET, sock->type, IPPROTO_TCP, &newsock); + if (rc) { + CERROR("Can't allocate socket\n"); + return rc; + } + + newsock->ops = sock->ops; + + set_current_state(TASK_INTERRUPTIBLE); + add_wait_queue(cfs_sk_sleep(sock->sk), &wait); + + rc = sock->ops->accept(sock, newsock, O_NONBLOCK); + if (rc == -EAGAIN) { + /* Nothing ready, so wait for activity */ + schedule(); + rc = sock->ops->accept(sock, newsock, O_NONBLOCK); + } + + remove_wait_queue(cfs_sk_sleep(sock->sk), &wait); + set_current_state(TASK_RUNNING); + + if (rc != 0) + goto failed; + + *newsockp = newsock; + return 0; + + failed: + sock_release(newsock); + return rc; +} + +EXPORT_SYMBOL(libcfs_sock_accept); + +void +libcfs_sock_abort_accept (struct socket *sock) +{ + wake_up_all(cfs_sk_sleep(sock->sk)); +} + +EXPORT_SYMBOL(libcfs_sock_abort_accept); + +int +libcfs_sock_connect (struct socket **sockp, int *fatal, + __u32 local_ip, int local_port, + __u32 peer_ip, int peer_port) +{ + struct sockaddr_in srvaddr; + int rc; + + rc = libcfs_sock_create(sockp, fatal, local_ip, local_port); + if (rc != 0) + return rc; + + memset (&srvaddr, 0, sizeof (srvaddr)); + srvaddr.sin_family = AF_INET; + srvaddr.sin_port = htons(peer_port); + srvaddr.sin_addr.s_addr = htonl(peer_ip); + + rc = (*sockp)->ops->connect(*sockp, + (struct sockaddr *)&srvaddr, sizeof(srvaddr), + 0); + if (rc == 0) + return 0; + + /* EADDRNOTAVAIL probably means we're already connected to the same + * peer/port on the same local port on a differently typed + * connection. Let our caller retry with a different local + * port... */ + *fatal = !(rc == -EADDRNOTAVAIL); + + CDEBUG_LIMIT(*fatal ? D_NETERROR : D_NET, + "Error %d connecting %u.%u.%u.%u/%d -> %u.%u.%u.%u/%d\n", rc, + HIPQUAD(local_ip), local_port, HIPQUAD(peer_ip), peer_port); + + sock_release(*sockp); + return rc; +} + +EXPORT_SYMBOL(libcfs_sock_connect); + +void +libcfs_sock_release (struct socket *sock) +{ + sock_release(sock); +} + +EXPORT_SYMBOL(libcfs_sock_release); diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-tracefile.c b/drivers/staging/lustre/lustre/libcfs/linux/linux-tracefile.c new file mode 100644 index 000000000000..6f563436a255 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-tracefile.c @@ -0,0 +1,275 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + */ + +#define DEBUG_SUBSYSTEM S_LNET +#define LUSTRE_TRACEFILE_PRIVATE + +#include <linux/libcfs/libcfs.h> +#include "tracefile.h" + +/* percents to share the total debug memory for each type */ +static unsigned int pages_factor[CFS_TCD_TYPE_MAX] = { + 80, /* 80% pages for CFS_TCD_TYPE_PROC */ + 10, /* 10% pages for CFS_TCD_TYPE_SOFTIRQ */ + 10 /* 10% pages for CFS_TCD_TYPE_IRQ */ +}; + +char *cfs_trace_console_buffers[NR_CPUS][CFS_TCD_TYPE_MAX]; + +struct rw_semaphore cfs_tracefile_sem; + +int cfs_tracefile_init_arch() +{ + int i; + int j; + struct cfs_trace_cpu_data *tcd; + + init_rwsem(&cfs_tracefile_sem); + + /* initialize trace_data */ + memset(cfs_trace_data, 0, sizeof(cfs_trace_data)); + for (i = 0; i < CFS_TCD_TYPE_MAX; i++) { + cfs_trace_data[i] = + kmalloc(sizeof(union cfs_trace_data_union) * + num_possible_cpus(), GFP_KERNEL); + if (cfs_trace_data[i] == NULL) + goto out; + + } + + /* arch related info initialized */ + cfs_tcd_for_each(tcd, i, j) { + spin_lock_init(&tcd->tcd_lock); + tcd->tcd_pages_factor = pages_factor[i]; + tcd->tcd_type = i; + tcd->tcd_cpu = j; + } + + for (i = 0; i < num_possible_cpus(); i++) + for (j = 0; j < 3; j++) { + cfs_trace_console_buffers[i][j] = + kmalloc(CFS_TRACE_CONSOLE_BUFFER_SIZE, + GFP_KERNEL); + + if (cfs_trace_console_buffers[i][j] == NULL) + goto out; + } + + return 0; + +out: + cfs_tracefile_fini_arch(); + printk(KERN_ERR "lnet: Not enough memory\n"); + return -ENOMEM; +} + +void cfs_tracefile_fini_arch() +{ + int i; + int j; + + for (i = 0; i < num_possible_cpus(); i++) + for (j = 0; j < 3; j++) + if (cfs_trace_console_buffers[i][j] != NULL) { + kfree(cfs_trace_console_buffers[i][j]); + cfs_trace_console_buffers[i][j] = NULL; + } + + for (i = 0; cfs_trace_data[i] != NULL; i++) { + kfree(cfs_trace_data[i]); + cfs_trace_data[i] = NULL; + } + + fini_rwsem(&cfs_tracefile_sem); +} + +void cfs_tracefile_read_lock() +{ + down_read(&cfs_tracefile_sem); +} + +void cfs_tracefile_read_unlock() +{ + up_read(&cfs_tracefile_sem); +} + +void cfs_tracefile_write_lock() +{ + down_write(&cfs_tracefile_sem); +} + +void cfs_tracefile_write_unlock() +{ + up_write(&cfs_tracefile_sem); +} + +cfs_trace_buf_type_t cfs_trace_buf_idx_get() +{ + if (in_irq()) + return CFS_TCD_TYPE_IRQ; + else if (in_softirq()) + return CFS_TCD_TYPE_SOFTIRQ; + else + return CFS_TCD_TYPE_PROC; +} + +/* + * The walking argument indicates the locking comes from all tcd types + * iterator and we must lock it and dissable local irqs to avoid deadlocks + * with other interrupt locks that might be happening. See LU-1311 + * for details. + */ +int cfs_trace_lock_tcd(struct cfs_trace_cpu_data *tcd, int walking) +{ + __LASSERT(tcd->tcd_type < CFS_TCD_TYPE_MAX); + if (tcd->tcd_type == CFS_TCD_TYPE_IRQ) + spin_lock_irqsave(&tcd->tcd_lock, tcd->tcd_lock_flags); + else if (tcd->tcd_type == CFS_TCD_TYPE_SOFTIRQ) + spin_lock_bh(&tcd->tcd_lock); + else if (unlikely(walking)) + spin_lock_irq(&tcd->tcd_lock); + else + spin_lock(&tcd->tcd_lock); + return 1; +} + +void cfs_trace_unlock_tcd(struct cfs_trace_cpu_data *tcd, int walking) +{ + __LASSERT(tcd->tcd_type < CFS_TCD_TYPE_MAX); + if (tcd->tcd_type == CFS_TCD_TYPE_IRQ) + spin_unlock_irqrestore(&tcd->tcd_lock, tcd->tcd_lock_flags); + else if (tcd->tcd_type == CFS_TCD_TYPE_SOFTIRQ) + spin_unlock_bh(&tcd->tcd_lock); + else if (unlikely(walking)) + spin_unlock_irq(&tcd->tcd_lock); + else + spin_unlock(&tcd->tcd_lock); +} + +int cfs_tcd_owns_tage(struct cfs_trace_cpu_data *tcd, + struct cfs_trace_page *tage) +{ + /* + * XXX nikita: do NOT call portals_debug_msg() (CDEBUG/ENTRY/EXIT) + * from here: this will lead to infinite recursion. + */ + return tcd->tcd_cpu == tage->cpu; +} + +void +cfs_set_ptldebug_header(struct ptldebug_header *header, + struct libcfs_debug_msg_data *msgdata, + unsigned long stack) +{ + struct timeval tv; + + do_gettimeofday(&tv); + + header->ph_subsys = msgdata->msg_subsys; + header->ph_mask = msgdata->msg_mask; + header->ph_cpu_id = smp_processor_id(); + header->ph_type = cfs_trace_buf_idx_get(); + header->ph_sec = (__u32)tv.tv_sec; + header->ph_usec = tv.tv_usec; + header->ph_stack = stack; + header->ph_pid = current->pid; + header->ph_line_num = msgdata->msg_line; + header->ph_extern_pid = 0; + return; +} + +static char * +dbghdr_to_err_string(struct ptldebug_header *hdr) +{ + switch (hdr->ph_subsys) { + + case S_LND: + case S_LNET: + return "LNetError"; + default: + return "LustreError"; + } +} + +static char * +dbghdr_to_info_string(struct ptldebug_header *hdr) +{ + switch (hdr->ph_subsys) { + + case S_LND: + case S_LNET: + return "LNet"; + default: + return "Lustre"; + } +} + +void cfs_print_to_console(struct ptldebug_header *hdr, int mask, + const char *buf, int len, const char *file, + const char *fn) +{ + char *prefix = "Lustre", *ptype = NULL; + + if ((mask & D_EMERG) != 0) { + prefix = dbghdr_to_err_string(hdr); + ptype = KERN_EMERG; + } else if ((mask & D_ERROR) != 0) { + prefix = dbghdr_to_err_string(hdr); + ptype = KERN_ERR; + } else if ((mask & D_WARNING) != 0) { + prefix = dbghdr_to_info_string(hdr); + ptype = KERN_WARNING; + } else if ((mask & (D_CONSOLE | libcfs_printk)) != 0) { + prefix = dbghdr_to_info_string(hdr); + ptype = KERN_INFO; + } + + if ((mask & D_CONSOLE) != 0) { + printk("%s%s: %.*s", ptype, prefix, len, buf); + } else { + printk("%s%s: %d:%d:(%s:%d:%s()) %.*s", ptype, prefix, + hdr->ph_pid, hdr->ph_extern_pid, file, hdr->ph_line_num, + fn, len, buf); + } + return; +} + +int cfs_trace_max_debug_mb(void) +{ + int total_mb = (num_physpages >> (20 - PAGE_SHIFT)); + + return MAX(512, (total_mb * 80)/100); +} diff --git a/drivers/staging/lustre/lustre/libcfs/linux/linux-tracefile.h b/drivers/staging/lustre/lustre/libcfs/linux/linux-tracefile.h new file mode 100644 index 000000000000..ba84e4ffddd1 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/linux/linux-tracefile.h @@ -0,0 +1,48 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2009, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + */ + +#ifndef __LIBCFS_LINUX_TRACEFILE_H__ +#define __LIBCFS_LINUX_TRACEFILE_H__ + +/** + * three types of trace_data in linux + */ +typedef enum { + CFS_TCD_TYPE_PROC = 0, + CFS_TCD_TYPE_SOFTIRQ, + CFS_TCD_TYPE_IRQ, + CFS_TCD_TYPE_MAX +} cfs_trace_buf_type_t; + +#endif diff --git a/drivers/staging/lustre/lustre/libcfs/lwt.c b/drivers/staging/lustre/lustre/libcfs/lwt.c new file mode 100644 index 000000000000..b631f7dde8e7 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/lwt.c @@ -0,0 +1,266 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/lwt.c + * + * Author: Eric Barton <eeb@clusterfs.com> + */ + +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> + +#if LWT_SUPPORT + +#if !KLWT_SUPPORT +int lwt_enabled; +lwt_cpu_t lwt_cpus[NR_CPUS]; +#endif + +int lwt_pages_per_cpu; + +/* NB only root is allowed to retrieve LWT info; it's an open door into the + * kernel... */ + +int +lwt_lookup_string (int *size, char *knl_ptr, + char *user_ptr, int user_size) +{ + int maxsize = 128; + + /* knl_ptr was retrieved from an LWT snapshot and the caller wants to + * turn it into a string. NB we can crash with an access violation + * trying to determine the string length, so we're trusting our + * caller... */ + + if (!cfs_capable(CFS_CAP_SYS_ADMIN)) + return (-EPERM); + + if (user_size > 0 && + maxsize > user_size) + maxsize = user_size; + + *size = strnlen (knl_ptr, maxsize - 1) + 1; + + if (user_ptr != NULL) { + if (user_size < 4) + return (-EINVAL); + + if (copy_to_user (user_ptr, knl_ptr, *size)) + return (-EFAULT); + + /* Did I truncate the string? */ + if (knl_ptr[*size - 1] != 0) + copy_to_user (user_ptr + *size - 4, "...", 4); + } + + return (0); +} + +int +lwt_control (int enable, int clear) +{ + lwt_page_t *p; + int i; + int j; + + if (!cfs_capable(CFS_CAP_SYS_ADMIN)) + return (-EPERM); + + if (!enable) { + LWT_EVENT(0,0,0,0); + lwt_enabled = 0; + mb(); + /* give people some time to stop adding traces */ + schedule_timeout(10); + } + + for (i = 0; i < num_online_cpus(); i++) { + p = lwt_cpus[i].lwtc_current_page; + + if (p == NULL) + return (-ENODATA); + + if (!clear) + continue; + + for (j = 0; j < lwt_pages_per_cpu; j++) { + memset (p->lwtp_events, 0, PAGE_CACHE_SIZE); + + p = list_entry (p->lwtp_list.next, + lwt_page_t, lwtp_list); + } + } + + if (enable) { + lwt_enabled = 1; + mb(); + LWT_EVENT(0,0,0,0); + } + + return (0); +} + +int +lwt_snapshot (cfs_cycles_t *now, int *ncpu, int *total_size, + void *user_ptr, int user_size) +{ + const int events_per_page = PAGE_CACHE_SIZE / sizeof(lwt_event_t); + const int bytes_per_page = events_per_page * sizeof(lwt_event_t); + lwt_page_t *p; + int i; + int j; + + if (!cfs_capable(CFS_CAP_SYS_ADMIN)) + return (-EPERM); + + *ncpu = num_online_cpus(); + *total_size = num_online_cpus() * lwt_pages_per_cpu * + bytes_per_page; + *now = get_cycles(); + + if (user_ptr == NULL) + return (0); + + for (i = 0; i < num_online_cpus(); i++) { + p = lwt_cpus[i].lwtc_current_page; + + if (p == NULL) + return (-ENODATA); + + for (j = 0; j < lwt_pages_per_cpu; j++) { + if (copy_to_user(user_ptr, p->lwtp_events, + bytes_per_page)) + return (-EFAULT); + + user_ptr = ((char *)user_ptr) + bytes_per_page; + p = list_entry(p->lwtp_list.next, + lwt_page_t, lwtp_list); + } + } + + return (0); +} + +int +lwt_init () +{ + int i; + int j; + + for (i = 0; i < num_online_cpus(); i++) + if (lwt_cpus[i].lwtc_current_page != NULL) + return (-EALREADY); + + LASSERT (!lwt_enabled); + + /* NULL pointers, zero scalars */ + memset (lwt_cpus, 0, sizeof (lwt_cpus)); + lwt_pages_per_cpu = + LWT_MEMORY / (num_online_cpus() * PAGE_CACHE_SIZE); + + for (i = 0; i < num_online_cpus(); i++) + for (j = 0; j < lwt_pages_per_cpu; j++) { + struct page *page = alloc_page (GFP_KERNEL); + lwt_page_t *lwtp; + + if (page == NULL) { + CERROR ("Can't allocate page\n"); + lwt_fini (); + return (-ENOMEM); + } + + LIBCFS_ALLOC(lwtp, sizeof (*lwtp)); + if (lwtp == NULL) { + CERROR ("Can't allocate lwtp\n"); + __free_page(page); + lwt_fini (); + return (-ENOMEM); + } + + lwtp->lwtp_page = page; + lwtp->lwtp_events = page_address(page); + memset (lwtp->lwtp_events, 0, PAGE_CACHE_SIZE); + + if (j == 0) { + INIT_LIST_HEAD (&lwtp->lwtp_list); + lwt_cpus[i].lwtc_current_page = lwtp; + } else { + list_add (&lwtp->lwtp_list, + &lwt_cpus[i].lwtc_current_page->lwtp_list); + } + } + + lwt_enabled = 1; + mb(); + + LWT_EVENT(0,0,0,0); + + return (0); +} + +void +lwt_fini () +{ + int i; + + lwt_control(0, 0); + + for (i = 0; i < num_online_cpus(); i++) + while (lwt_cpus[i].lwtc_current_page != NULL) { + lwt_page_t *lwtp = lwt_cpus[i].lwtc_current_page; + + if (list_empty (&lwtp->lwtp_list)) { + lwt_cpus[i].lwtc_current_page = NULL; + } else { + lwt_cpus[i].lwtc_current_page = + list_entry (lwtp->lwtp_list.next, + lwt_page_t, lwtp_list); + + list_del (&lwtp->lwtp_list); + } + + __free_page (lwtp->lwtp_page); + LIBCFS_FREE (lwtp, sizeof (*lwtp)); + } +} + +EXPORT_SYMBOL(lwt_enabled); +EXPORT_SYMBOL(lwt_cpus); + +EXPORT_SYMBOL(lwt_init); +EXPORT_SYMBOL(lwt_fini); +EXPORT_SYMBOL(lwt_lookup_string); +EXPORT_SYMBOL(lwt_control); +EXPORT_SYMBOL(lwt_snapshot); +#endif diff --git a/drivers/staging/lustre/lustre/libcfs/module.c b/drivers/staging/lustre/lustre/libcfs/module.c new file mode 100644 index 000000000000..3372537c6f3b --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/module.c @@ -0,0 +1,498 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + */ + +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> +#include <linux/libcfs/libcfs_crypto.h> +#include <linux/lnet/lib-lnet.h> +#include <linux/lnet/lnet.h> +#include "tracefile.h" + +void +kportal_memhog_free (struct libcfs_device_userstate *ldu) +{ + struct page **level0p = &ldu->ldu_memhog_root_page; + struct page **level1p; + struct page **level2p; + int count1; + int count2; + + if (*level0p != NULL) { + + level1p = (struct page **)page_address(*level0p); + count1 = 0; + + while (count1 < PAGE_CACHE_SIZE/sizeof(struct page *) && + *level1p != NULL) { + + level2p = (struct page **)page_address(*level1p); + count2 = 0; + + while (count2 < PAGE_CACHE_SIZE/sizeof(struct page *) && + *level2p != NULL) { + + __free_page(*level2p); + ldu->ldu_memhog_pages--; + level2p++; + count2++; + } + + __free_page(*level1p); + ldu->ldu_memhog_pages--; + level1p++; + count1++; + } + + __free_page(*level0p); + ldu->ldu_memhog_pages--; + + *level0p = NULL; + } + + LASSERT (ldu->ldu_memhog_pages == 0); +} + +int +kportal_memhog_alloc (struct libcfs_device_userstate *ldu, int npages, int flags) +{ + struct page **level0p; + struct page **level1p; + struct page **level2p; + int count1; + int count2; + + LASSERT (ldu->ldu_memhog_pages == 0); + LASSERT (ldu->ldu_memhog_root_page == NULL); + + if (npages < 0) + return -EINVAL; + + if (npages == 0) + return 0; + + level0p = &ldu->ldu_memhog_root_page; + *level0p = alloc_page(flags); + if (*level0p == NULL) + return -ENOMEM; + ldu->ldu_memhog_pages++; + + level1p = (struct page **)page_address(*level0p); + count1 = 0; + memset(level1p, 0, PAGE_CACHE_SIZE); + + while (ldu->ldu_memhog_pages < npages && + count1 < PAGE_CACHE_SIZE/sizeof(struct page *)) { + + if (cfs_signal_pending()) + return (-EINTR); + + *level1p = alloc_page(flags); + if (*level1p == NULL) + return -ENOMEM; + ldu->ldu_memhog_pages++; + + level2p = (struct page **)page_address(*level1p); + count2 = 0; + memset(level2p, 0, PAGE_CACHE_SIZE); + + while (ldu->ldu_memhog_pages < npages && + count2 < PAGE_CACHE_SIZE/sizeof(struct page *)) { + + if (cfs_signal_pending()) + return (-EINTR); + + *level2p = alloc_page(flags); + if (*level2p == NULL) + return (-ENOMEM); + ldu->ldu_memhog_pages++; + + level2p++; + count2++; + } + + level1p++; + count1++; + } + + return 0; +} + +/* called when opening /dev/device */ +static int libcfs_psdev_open(unsigned long flags, void *args) +{ + struct libcfs_device_userstate *ldu; + ENTRY; + + try_module_get(THIS_MODULE); + + LIBCFS_ALLOC(ldu, sizeof(*ldu)); + if (ldu != NULL) { + ldu->ldu_memhog_pages = 0; + ldu->ldu_memhog_root_page = NULL; + } + *(struct libcfs_device_userstate **)args = ldu; + + RETURN(0); +} + +/* called when closing /dev/device */ +static int libcfs_psdev_release(unsigned long flags, void *args) +{ + struct libcfs_device_userstate *ldu; + ENTRY; + + ldu = (struct libcfs_device_userstate *)args; + if (ldu != NULL) { + kportal_memhog_free(ldu); + LIBCFS_FREE(ldu, sizeof(*ldu)); + } + + module_put(THIS_MODULE); + RETURN(0); +} + +static struct rw_semaphore ioctl_list_sem; +static struct list_head ioctl_list; + +int libcfs_register_ioctl(struct libcfs_ioctl_handler *hand) +{ + int rc = 0; + + down_write(&ioctl_list_sem); + if (!list_empty(&hand->item)) + rc = -EBUSY; + else + list_add_tail(&hand->item, &ioctl_list); + up_write(&ioctl_list_sem); + + return rc; +} +EXPORT_SYMBOL(libcfs_register_ioctl); + +int libcfs_deregister_ioctl(struct libcfs_ioctl_handler *hand) +{ + int rc = 0; + + down_write(&ioctl_list_sem); + if (list_empty(&hand->item)) + rc = -ENOENT; + else + list_del_init(&hand->item); + up_write(&ioctl_list_sem); + + return rc; +} +EXPORT_SYMBOL(libcfs_deregister_ioctl); + +static int libcfs_ioctl_int(struct cfs_psdev_file *pfile,unsigned long cmd, + void *arg, struct libcfs_ioctl_data *data) +{ + int err = -EINVAL; + ENTRY; + + switch (cmd) { + case IOC_LIBCFS_CLEAR_DEBUG: + libcfs_debug_clear_buffer(); + RETURN(0); + /* + * case IOC_LIBCFS_PANIC: + * Handled in arch/cfs_module.c + */ + case IOC_LIBCFS_MARK_DEBUG: + if (data->ioc_inlbuf1 == NULL || + data->ioc_inlbuf1[data->ioc_inllen1 - 1] != '\0') + RETURN(-EINVAL); + libcfs_debug_mark_buffer(data->ioc_inlbuf1); + RETURN(0); +#if LWT_SUPPORT + case IOC_LIBCFS_LWT_CONTROL: + err = lwt_control ((data->ioc_flags & 1) != 0, + (data->ioc_flags & 2) != 0); + break; + + case IOC_LIBCFS_LWT_SNAPSHOT: { + cfs_cycles_t now; + int ncpu; + int total_size; + + err = lwt_snapshot (&now, &ncpu, &total_size, + data->ioc_pbuf1, data->ioc_plen1); + data->ioc_u64[0] = now; + data->ioc_u32[0] = ncpu; + data->ioc_u32[1] = total_size; + + /* Hedge against broken user/kernel typedefs (e.g. cycles_t) */ + data->ioc_u32[2] = sizeof(lwt_event_t); + data->ioc_u32[3] = offsetof(lwt_event_t, lwte_where); + + if (err == 0 && + libcfs_ioctl_popdata(arg, data, sizeof (*data))) + err = -EFAULT; + break; + } + + case IOC_LIBCFS_LWT_LOOKUP_STRING: + err = lwt_lookup_string (&data->ioc_count, data->ioc_pbuf1, + data->ioc_pbuf2, data->ioc_plen2); + if (err == 0 && + libcfs_ioctl_popdata(arg, data, sizeof (*data))) + err = -EFAULT; + break; +#endif + case IOC_LIBCFS_MEMHOG: + if (pfile->private_data == NULL) { + err = -EINVAL; + } else { + kportal_memhog_free(pfile->private_data); + /* XXX The ioc_flags is not GFP flags now, need to be fixed */ + err = kportal_memhog_alloc(pfile->private_data, + data->ioc_count, + data->ioc_flags); + if (err != 0) + kportal_memhog_free(pfile->private_data); + } + break; + + case IOC_LIBCFS_PING_TEST: { + extern void (kping_client)(struct libcfs_ioctl_data *); + void (*ping)(struct libcfs_ioctl_data *); + + CDEBUG(D_IOCTL, "doing %d pings to nid %s (%s)\n", + data->ioc_count, libcfs_nid2str(data->ioc_nid), + libcfs_nid2str(data->ioc_nid)); + ping = symbol_get(kping_client); + if (!ping) + CERROR("symbol_get failed\n"); + else { + ping(data); + symbol_put(kping_client); + } + RETURN(0); + } + + default: { + struct libcfs_ioctl_handler *hand; + err = -EINVAL; + down_read(&ioctl_list_sem); + list_for_each_entry(hand, &ioctl_list, item) { + err = hand->handle_ioctl(cmd, data); + if (err != -EINVAL) { + if (err == 0) + err = libcfs_ioctl_popdata(arg, + data, sizeof (*data)); + break; + } + } + up_read(&ioctl_list_sem); + break; + } + } + + RETURN(err); +} + +static int libcfs_ioctl(struct cfs_psdev_file *pfile, unsigned long cmd, void *arg) +{ + char *buf; + struct libcfs_ioctl_data *data; + int err = 0; + ENTRY; + + LIBCFS_ALLOC_GFP(buf, 1024, GFP_IOFS); + if (buf == NULL) + RETURN(-ENOMEM); + + /* 'cmd' and permissions get checked in our arch-specific caller */ + if (libcfs_ioctl_getdata(buf, buf + 800, (void *)arg)) { + CERROR("PORTALS ioctl: data error\n"); + GOTO(out, err = -EINVAL); + } + data = (struct libcfs_ioctl_data *)buf; + + err = libcfs_ioctl_int(pfile, cmd, arg, data); + +out: + LIBCFS_FREE(buf, 1024); + RETURN(err); +} + + +struct cfs_psdev_ops libcfs_psdev_ops = { + libcfs_psdev_open, + libcfs_psdev_release, + NULL, + NULL, + libcfs_ioctl +}; + +extern int insert_proc(void); +extern void remove_proc(void); +MODULE_AUTHOR("Peter J. Braam <braam@clusterfs.com>"); +MODULE_DESCRIPTION("Portals v3.1"); +MODULE_LICENSE("GPL"); + +extern psdev_t libcfs_dev; +extern struct rw_semaphore cfs_tracefile_sem; +extern struct mutex cfs_trace_thread_mutex; +extern struct cfs_wi_sched *cfs_sched_rehash; + +extern void libcfs_init_nidstrings(void); +extern int libcfs_arch_init(void); +extern void libcfs_arch_cleanup(void); + +static int init_libcfs_module(void) +{ + int rc; + + libcfs_arch_init(); + libcfs_init_nidstrings(); + init_rwsem(&cfs_tracefile_sem); + mutex_init(&cfs_trace_thread_mutex); + init_rwsem(&ioctl_list_sem); + INIT_LIST_HEAD(&ioctl_list); + init_waitqueue_head(&cfs_race_waitq); + + rc = libcfs_debug_init(5 * 1024 * 1024); + if (rc < 0) { + printk(KERN_ERR "LustreError: libcfs_debug_init: %d\n", rc); + return (rc); + } + + rc = cfs_cpu_init(); + if (rc != 0) + goto cleanup_debug; + +#if LWT_SUPPORT + rc = lwt_init(); + if (rc != 0) { + CERROR("lwt_init: error %d\n", rc); + goto cleanup_debug; + } +#endif + rc = misc_register(&libcfs_dev); + if (rc) { + CERROR("misc_register: error %d\n", rc); + goto cleanup_lwt; + } + + rc = cfs_wi_startup(); + if (rc) { + CERROR("initialize workitem: error %d\n", rc); + goto cleanup_deregister; + } + + /* max to 4 threads, should be enough for rehash */ + rc = min(cfs_cpt_weight(cfs_cpt_table, CFS_CPT_ANY), 4); + rc = cfs_wi_sched_create("cfs_rh", cfs_cpt_table, CFS_CPT_ANY, + rc, &cfs_sched_rehash); + if (rc != 0) { + CERROR("Startup workitem scheduler: error: %d\n", rc); + goto cleanup_deregister; + } + + rc = cfs_crypto_register(); + if (rc) { + CERROR("cfs_crypto_regster: error %d\n", rc); + goto cleanup_wi; + } + + + rc = insert_proc(); + if (rc) { + CERROR("insert_proc: error %d\n", rc); + goto cleanup_crypto; + } + + CDEBUG (D_OTHER, "portals setup OK\n"); + return 0; + cleanup_crypto: + cfs_crypto_unregister(); + cleanup_wi: + cfs_wi_shutdown(); + cleanup_deregister: + misc_deregister(&libcfs_dev); + cleanup_lwt: +#if LWT_SUPPORT + lwt_fini(); +#endif + cleanup_debug: + libcfs_debug_cleanup(); + return rc; +} + +static void exit_libcfs_module(void) +{ + int rc; + + remove_proc(); + + CDEBUG(D_MALLOC, "before Portals cleanup: kmem %d\n", + atomic_read(&libcfs_kmemory)); + + if (cfs_sched_rehash != NULL) { + cfs_wi_sched_destroy(cfs_sched_rehash); + cfs_sched_rehash = NULL; + } + + cfs_crypto_unregister(); + cfs_wi_shutdown(); + + rc = misc_deregister(&libcfs_dev); + if (rc) + CERROR("misc_deregister error %d\n", rc); + +#if LWT_SUPPORT + lwt_fini(); +#endif + cfs_cpu_fini(); + + if (atomic_read(&libcfs_kmemory) != 0) + CERROR("Portals memory leaked: %d bytes\n", + atomic_read(&libcfs_kmemory)); + + rc = libcfs_debug_cleanup(); + if (rc) + printk(KERN_ERR "LustreError: libcfs_debug_cleanup: %d\n", + rc); + + fini_rwsem(&ioctl_list_sem); + fini_rwsem(&cfs_tracefile_sem); + + libcfs_arch_cleanup(); +} + +cfs_module(libcfs, "1.0.0", init_libcfs_module, exit_libcfs_module); diff --git a/drivers/staging/lustre/lustre/libcfs/nidstrings.c b/drivers/staging/lustre/lustre/libcfs/nidstrings.c new file mode 100644 index 000000000000..ccfd1078a906 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/nidstrings.c @@ -0,0 +1,867 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2011, 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/nidstrings.c + * + * Author: Phil Schwan <phil@clusterfs.com> + */ + +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> +#include <linux/lnet/lnet.h> + +/* CAVEAT VENDITOR! Keep the canonical string representation of nets/nids + * consistent in all conversion functions. Some code fragments are copied + * around for the sake of clarity... + */ + +/* CAVEAT EMPTOR! Racey temporary buffer allocation! + * Choose the number of nidstrings to support the MAXIMUM expected number of + * concurrent users. If there are more, the returned string will be volatile. + * NB this number must allow for a process to be descheduled for a timeslice + * between getting its string and using it. + */ + +static char libcfs_nidstrings[LNET_NIDSTR_COUNT][LNET_NIDSTR_SIZE]; +static int libcfs_nidstring_idx = 0; + +static spinlock_t libcfs_nidstring_lock; + +void libcfs_init_nidstrings (void) +{ + spin_lock_init(&libcfs_nidstring_lock); +} + +# define NIDSTR_LOCK(f) spin_lock_irqsave(&libcfs_nidstring_lock, f) +# define NIDSTR_UNLOCK(f) spin_unlock_irqrestore(&libcfs_nidstring_lock, f) + +static char * +libcfs_next_nidstring (void) +{ + char *str; + unsigned long flags; + + NIDSTR_LOCK(flags); + + str = libcfs_nidstrings[libcfs_nidstring_idx++]; + if (libcfs_nidstring_idx == + sizeof(libcfs_nidstrings)/sizeof(libcfs_nidstrings[0])) + libcfs_nidstring_idx = 0; + + NIDSTR_UNLOCK(flags); + return str; +} + +static int libcfs_lo_str2addr(const char *str, int nob, __u32 *addr); +static void libcfs_ip_addr2str(__u32 addr, char *str); +static int libcfs_ip_str2addr(const char *str, int nob, __u32 *addr); +static void libcfs_decnum_addr2str(__u32 addr, char *str); +static void libcfs_hexnum_addr2str(__u32 addr, char *str); +static int libcfs_num_str2addr(const char *str, int nob, __u32 *addr); +static int libcfs_num_parse(char *str, int len, struct list_head *list); +static int libcfs_num_match(__u32 addr, struct list_head *list); + +struct netstrfns { + int nf_type; + char *nf_name; + char *nf_modname; + void (*nf_addr2str)(__u32 addr, char *str); + int (*nf_str2addr)(const char *str, int nob, __u32 *addr); + int (*nf_parse_addrlist)(char *str, int len, + struct list_head *list); + int (*nf_match_addr)(__u32 addr, struct list_head *list); +}; + +static struct netstrfns libcfs_netstrfns[] = { + {/* .nf_type */ LOLND, + /* .nf_name */ "lo", + /* .nf_modname */ "klolnd", + /* .nf_addr2str */ libcfs_decnum_addr2str, + /* .nf_str2addr */ libcfs_lo_str2addr, + /* .nf_parse_addr*/ libcfs_num_parse, + /* .nf_match_addr*/ libcfs_num_match}, + {/* .nf_type */ SOCKLND, + /* .nf_name */ "tcp", + /* .nf_modname */ "ksocklnd", + /* .nf_addr2str */ libcfs_ip_addr2str, + /* .nf_str2addr */ libcfs_ip_str2addr, + /* .nf_parse_addrlist*/ cfs_ip_addr_parse, + /* .nf_match_addr*/ cfs_ip_addr_match}, + {/* .nf_type */ O2IBLND, + /* .nf_name */ "o2ib", + /* .nf_modname */ "ko2iblnd", + /* .nf_addr2str */ libcfs_ip_addr2str, + /* .nf_str2addr */ libcfs_ip_str2addr, + /* .nf_parse_addrlist*/ cfs_ip_addr_parse, + /* .nf_match_addr*/ cfs_ip_addr_match}, + {/* .nf_type */ CIBLND, + /* .nf_name */ "cib", + /* .nf_modname */ "kciblnd", + /* .nf_addr2str */ libcfs_ip_addr2str, + /* .nf_str2addr */ libcfs_ip_str2addr, + /* .nf_parse_addrlist*/ cfs_ip_addr_parse, + /* .nf_match_addr*/ cfs_ip_addr_match}, + {/* .nf_type */ OPENIBLND, + /* .nf_name */ "openib", + /* .nf_modname */ "kopeniblnd", + /* .nf_addr2str */ libcfs_ip_addr2str, + /* .nf_str2addr */ libcfs_ip_str2addr, + /* .nf_parse_addrlist*/ cfs_ip_addr_parse, + /* .nf_match_addr*/ cfs_ip_addr_match}, + {/* .nf_type */ IIBLND, + /* .nf_name */ "iib", + /* .nf_modname */ "kiiblnd", + /* .nf_addr2str */ libcfs_ip_addr2str, + /* .nf_str2addr */ libcfs_ip_str2addr, + /* .nf_parse_addrlist*/ cfs_ip_addr_parse, + /* .nf_match_addr*/ cfs_ip_addr_match}, + {/* .nf_type */ VIBLND, + /* .nf_name */ "vib", + /* .nf_modname */ "kviblnd", + /* .nf_addr2str */ libcfs_ip_addr2str, + /* .nf_str2addr */ libcfs_ip_str2addr, + /* .nf_parse_addrlist*/ cfs_ip_addr_parse, + /* .nf_match_addr*/ cfs_ip_addr_match}, + {/* .nf_type */ RALND, + /* .nf_name */ "ra", + /* .nf_modname */ "kralnd", + /* .nf_addr2str */ libcfs_ip_addr2str, + /* .nf_str2addr */ libcfs_ip_str2addr, + /* .nf_parse_addrlist*/ cfs_ip_addr_parse, + /* .nf_match_addr*/ cfs_ip_addr_match}, + {/* .nf_type */ QSWLND, + /* .nf_name */ "elan", + /* .nf_modname */ "kqswlnd", + /* .nf_addr2str */ libcfs_decnum_addr2str, + /* .nf_str2addr */ libcfs_num_str2addr, + /* .nf_parse_addrlist*/ libcfs_num_parse, + /* .nf_match_addr*/ libcfs_num_match}, + {/* .nf_type */ GMLND, + /* .nf_name */ "gm", + /* .nf_modname */ "kgmlnd", + /* .nf_addr2str */ libcfs_hexnum_addr2str, + /* .nf_str2addr */ libcfs_num_str2addr, + /* .nf_parse_addrlist*/ libcfs_num_parse, + /* .nf_match_addr*/ libcfs_num_match}, + {/* .nf_type */ MXLND, + /* .nf_name */ "mx", + /* .nf_modname */ "kmxlnd", + /* .nf_addr2str */ libcfs_ip_addr2str, + /* .nf_str2addr */ libcfs_ip_str2addr, + /* .nf_parse_addrlist*/ cfs_ip_addr_parse, + /* .nf_match_addr*/ cfs_ip_addr_match}, + {/* .nf_type */ PTLLND, + /* .nf_name */ "ptl", + /* .nf_modname */ "kptllnd", + /* .nf_addr2str */ libcfs_decnum_addr2str, + /* .nf_str2addr */ libcfs_num_str2addr, + /* .nf_parse_addrlist*/ libcfs_num_parse, + /* .nf_match_addr*/ libcfs_num_match}, + {/* .nf_type */ GNILND, + /* .nf_name */ "gni", + /* .nf_modname */ "kgnilnd", + /* .nf_addr2str */ libcfs_decnum_addr2str, + /* .nf_str2addr */ libcfs_num_str2addr, + /* .nf_parse_addrlist*/ libcfs_num_parse, + /* .nf_match_addr*/ libcfs_num_match}, + /* placeholder for net0 alias. It MUST BE THE LAST ENTRY */ + {/* .nf_type */ -1}, +}; + +const int libcfs_nnetstrfns = sizeof(libcfs_netstrfns)/sizeof(libcfs_netstrfns[0]); + +int +libcfs_lo_str2addr(const char *str, int nob, __u32 *addr) +{ + *addr = 0; + return 1; +} + +void +libcfs_ip_addr2str(__u32 addr, char *str) +{ +#if 0 /* never lookup */ +#endif + snprintf(str, LNET_NIDSTR_SIZE, "%u.%u.%u.%u", + (addr >> 24) & 0xff, (addr >> 16) & 0xff, + (addr >> 8) & 0xff, addr & 0xff); +} + +/* CAVEAT EMPTOR XscanfX + * I use "%n" at the end of a sscanf format to detect trailing junk. However + * sscanf may return immediately if it sees the terminating '0' in a string, so + * I initialise the %n variable to the expected length. If sscanf sets it; + * fine, if it doesn't, then the scan ended at the end of the string, which is + * fine too :) */ + +int +libcfs_ip_str2addr(const char *str, int nob, __u32 *addr) +{ + int a; + int b; + int c; + int d; + int n = nob; /* XscanfX */ + + /* numeric IP? */ + if (sscanf(str, "%u.%u.%u.%u%n", &a, &b, &c, &d, &n) >= 4 && + n == nob && + (a & ~0xff) == 0 && (b & ~0xff) == 0 && + (c & ~0xff) == 0 && (d & ~0xff) == 0) { + *addr = ((a<<24)|(b<<16)|(c<<8)|d); + return 1; + } + + return 0; +} + +void +libcfs_decnum_addr2str(__u32 addr, char *str) +{ + snprintf(str, LNET_NIDSTR_SIZE, "%u", addr); +} + +void +libcfs_hexnum_addr2str(__u32 addr, char *str) +{ + snprintf(str, LNET_NIDSTR_SIZE, "0x%x", addr); +} + +int +libcfs_num_str2addr(const char *str, int nob, __u32 *addr) +{ + int n; + + n = nob; + if (sscanf(str, "0x%x%n", addr, &n) >= 1 && n == nob) + return 1; + + n = nob; + if (sscanf(str, "0X%x%n", addr, &n) >= 1 && n == nob) + return 1; + + n = nob; + if (sscanf(str, "%u%n", addr, &n) >= 1 && n == nob) + return 1; + + return 0; +} + +struct netstrfns * +libcfs_lnd2netstrfns(int lnd) +{ + int i; + + if (lnd >= 0) + for (i = 0; i < libcfs_nnetstrfns; i++) + if (lnd == libcfs_netstrfns[i].nf_type) + return &libcfs_netstrfns[i]; + + return NULL; +} + +struct netstrfns * +libcfs_namenum2netstrfns(const char *name) +{ + struct netstrfns *nf; + int i; + + for (i = 0; i < libcfs_nnetstrfns; i++) { + nf = &libcfs_netstrfns[i]; + if (nf->nf_type >= 0 && + !strncmp(name, nf->nf_name, strlen(nf->nf_name))) + return nf; + } + return NULL; +} + +struct netstrfns * +libcfs_name2netstrfns(const char *name) +{ + int i; + + for (i = 0; i < libcfs_nnetstrfns; i++) + if (libcfs_netstrfns[i].nf_type >= 0 && + !strcmp(libcfs_netstrfns[i].nf_name, name)) + return &libcfs_netstrfns[i]; + + return NULL; +} + +int +libcfs_isknown_lnd(int type) +{ + return libcfs_lnd2netstrfns(type) != NULL; +} + +char * +libcfs_lnd2modname(int lnd) +{ + struct netstrfns *nf = libcfs_lnd2netstrfns(lnd); + + return (nf == NULL) ? NULL : nf->nf_modname; +} + +char * +libcfs_lnd2str(int lnd) +{ + char *str; + struct netstrfns *nf = libcfs_lnd2netstrfns(lnd); + + if (nf != NULL) + return nf->nf_name; + + str = libcfs_next_nidstring(); + snprintf(str, LNET_NIDSTR_SIZE, "?%u?", lnd); + return str; +} + +int +libcfs_str2lnd(const char *str) +{ + struct netstrfns *nf = libcfs_name2netstrfns(str); + + if (nf != NULL) + return nf->nf_type; + + return -1; +} + +char * +libcfs_net2str(__u32 net) +{ + int lnd = LNET_NETTYP(net); + int num = LNET_NETNUM(net); + struct netstrfns *nf = libcfs_lnd2netstrfns(lnd); + char *str = libcfs_next_nidstring(); + + if (nf == NULL) + snprintf(str, LNET_NIDSTR_SIZE, "<%u:%u>", lnd, num); + else if (num == 0) + snprintf(str, LNET_NIDSTR_SIZE, "%s", nf->nf_name); + else + snprintf(str, LNET_NIDSTR_SIZE, "%s%u", nf->nf_name, num); + + return str; +} + +char * +libcfs_nid2str(lnet_nid_t nid) +{ + __u32 addr = LNET_NIDADDR(nid); + __u32 net = LNET_NIDNET(nid); + int lnd = LNET_NETTYP(net); + int nnum = LNET_NETNUM(net); + struct netstrfns *nf; + char *str; + int nob; + + if (nid == LNET_NID_ANY) + return "<?>"; + + nf = libcfs_lnd2netstrfns(lnd); + str = libcfs_next_nidstring(); + + if (nf == NULL) + snprintf(str, LNET_NIDSTR_SIZE, "%x@<%u:%u>", addr, lnd, nnum); + else { + nf->nf_addr2str(addr, str); + nob = strlen(str); + if (nnum == 0) + snprintf(str + nob, LNET_NIDSTR_SIZE - nob, "@%s", + nf->nf_name); + else + snprintf(str + nob, LNET_NIDSTR_SIZE - nob, "@%s%u", + nf->nf_name, nnum); + } + + return str; +} + +static struct netstrfns * +libcfs_str2net_internal(const char *str, __u32 *net) +{ + struct netstrfns *uninitialized_var(nf); + int nob; + int netnum; + int i; + + for (i = 0; i < libcfs_nnetstrfns; i++) { + nf = &libcfs_netstrfns[i]; + if (nf->nf_type >= 0 && + !strncmp(str, nf->nf_name, strlen(nf->nf_name))) + break; + } + + if (i == libcfs_nnetstrfns) + return NULL; + + nob = strlen(nf->nf_name); + + if (strlen(str) == (unsigned int)nob) { + netnum = 0; + } else { + if (nf->nf_type == LOLND) /* net number not allowed */ + return NULL; + + str += nob; + i = strlen(str); + if (sscanf(str, "%u%n", &netnum, &i) < 1 || + i != (int)strlen(str)) + return NULL; + } + + *net = LNET_MKNET(nf->nf_type, netnum); + return nf; +} + +__u32 +libcfs_str2net(const char *str) +{ + __u32 net; + + if (libcfs_str2net_internal(str, &net) != NULL) + return net; + + return LNET_NIDNET(LNET_NID_ANY); +} + +lnet_nid_t +libcfs_str2nid(const char *str) +{ + const char *sep = strchr(str, '@'); + struct netstrfns *nf; + __u32 net; + __u32 addr; + + if (sep != NULL) { + nf = libcfs_str2net_internal(sep + 1, &net); + if (nf == NULL) + return LNET_NID_ANY; + } else { + sep = str + strlen(str); + net = LNET_MKNET(SOCKLND, 0); + nf = libcfs_lnd2netstrfns(SOCKLND); + LASSERT (nf != NULL); + } + + if (!nf->nf_str2addr(str, (int)(sep - str), &addr)) + return LNET_NID_ANY; + + return LNET_MKNID(net, addr); +} + +char * +libcfs_id2str(lnet_process_id_t id) +{ + char *str = libcfs_next_nidstring(); + + if (id.pid == LNET_PID_ANY) { + snprintf(str, LNET_NIDSTR_SIZE, + "LNET_PID_ANY-%s", libcfs_nid2str(id.nid)); + return str; + } + + snprintf(str, LNET_NIDSTR_SIZE, "%s%u-%s", + ((id.pid & LNET_PID_USERFLAG) != 0) ? "U" : "", + (id.pid & ~LNET_PID_USERFLAG), libcfs_nid2str(id.nid)); + return str; +} + +int +libcfs_str2anynid(lnet_nid_t *nidp, const char *str) +{ + if (!strcmp(str, "*")) { + *nidp = LNET_NID_ANY; + return 1; + } + + *nidp = libcfs_str2nid(str); + return *nidp != LNET_NID_ANY; +} + +/** + * Nid range list syntax. + * \verbatim + * + * <nidlist> :== <nidrange> [ ' ' <nidrange> ] + * <nidrange> :== <addrrange> '@' <net> + * <addrrange> :== '*' | + * <ipaddr_range> | + * <cfs_expr_list> + * <ipaddr_range> :== <cfs_expr_list>.<cfs_expr_list>.<cfs_expr_list>. + * <cfs_expr_list> + * <cfs_expr_list> :== <number> | + * <expr_list> + * <expr_list> :== '[' <range_expr> [ ',' <range_expr>] ']' + * <range_expr> :== <number> | + * <number> '-' <number> | + * <number> '-' <number> '/' <number> + * <net> :== <netname> | <netname><number> + * <netname> :== "lo" | "tcp" | "o2ib" | "cib" | "openib" | "iib" | + * "vib" | "ra" | "elan" | "mx" | "ptl" + * \endverbatim + */ + +/** + * Structure to represent \<nidrange\> token of the syntax. + * + * One of this is created for each \<net\> parsed. + */ +struct nidrange { + /** + * Link to list of this structures which is built on nid range + * list parsing. + */ + struct list_head nr_link; + /** + * List head for addrrange::ar_link. + */ + struct list_head nr_addrranges; + /** + * Flag indicating that *@<net> is found. + */ + int nr_all; + /** + * Pointer to corresponding element of libcfs_netstrfns. + */ + struct netstrfns *nr_netstrfns; + /** + * Number of network. E.g. 5 if \<net\> is "elan5". + */ + int nr_netnum; +}; + +/** + * Structure to represent \<addrrange\> token of the syntax. + */ +struct addrrange { + /** + * Link to nidrange::nr_addrranges. + */ + struct list_head ar_link; + /** + * List head for cfs_expr_list::el_list. + */ + struct list_head ar_numaddr_ranges; +}; + +/** + * Nf_parse_addrlist method for networks using numeric addresses. + * + * Examples of such networks are gm and elan. + * + * \retval 0 if \a str parsed to numeric address + * \retval errno otherwise + */ +static int +libcfs_num_parse(char *str, int len, struct list_head *list) +{ + struct cfs_expr_list *el; + int rc; + + rc = cfs_expr_list_parse(str, len, 0, MAX_NUMERIC_VALUE, &el); + if (rc == 0) + list_add_tail(&el->el_link, list); + + return rc; +} + +/** + * Parses \<addrrange\> token on the syntax. + * + * Allocates struct addrrange and links to \a nidrange via + * (nidrange::nr_addrranges) + * + * \retval 1 if \a src parses to '*' | \<ipaddr_range\> | \<cfs_expr_list\> + * \retval 0 otherwise + */ +static int +parse_addrange(const struct cfs_lstr *src, struct nidrange *nidrange) +{ + struct addrrange *addrrange; + + if (src->ls_len == 1 && src->ls_str[0] == '*') { + nidrange->nr_all = 1; + return 1; + } + + LIBCFS_ALLOC(addrrange, sizeof(struct addrrange)); + if (addrrange == NULL) + return 0; + list_add_tail(&addrrange->ar_link, &nidrange->nr_addrranges); + INIT_LIST_HEAD(&addrrange->ar_numaddr_ranges); + + return nidrange->nr_netstrfns->nf_parse_addrlist(src->ls_str, + src->ls_len, + &addrrange->ar_numaddr_ranges); +} + +/** + * Finds or creates struct nidrange. + * + * Checks if \a src is a valid network name, looks for corresponding + * nidrange on the ist of nidranges (\a nidlist), creates new struct + * nidrange if it is not found. + * + * \retval pointer to struct nidrange matching network specified via \a src + * \retval NULL if \a src does not match any network + */ +static struct nidrange * +add_nidrange(const struct cfs_lstr *src, + struct list_head *nidlist) +{ + struct netstrfns *nf; + struct nidrange *nr; + int endlen; + unsigned netnum; + + if (src->ls_len >= LNET_NIDSTR_SIZE) + return NULL; + + nf = libcfs_namenum2netstrfns(src->ls_str); + if (nf == NULL) + return NULL; + endlen = src->ls_len - strlen(nf->nf_name); + if (endlen == 0) + /* network name only, e.g. "elan" or "tcp" */ + netnum = 0; + else { + /* e.g. "elan25" or "tcp23", refuse to parse if + * network name is not appended with decimal or + * hexadecimal number */ + if (!cfs_str2num_check(src->ls_str + strlen(nf->nf_name), + endlen, &netnum, 0, MAX_NUMERIC_VALUE)) + return NULL; + } + + list_for_each_entry(nr, nidlist, nr_link) { + if (nr->nr_netstrfns != nf) + continue; + if (nr->nr_netnum != netnum) + continue; + return nr; + } + + LIBCFS_ALLOC(nr, sizeof(struct nidrange)); + if (nr == NULL) + return NULL; + list_add_tail(&nr->nr_link, nidlist); + INIT_LIST_HEAD(&nr->nr_addrranges); + nr->nr_netstrfns = nf; + nr->nr_all = 0; + nr->nr_netnum = netnum; + + return nr; +} + +/** + * Parses \<nidrange\> token of the syntax. + * + * \retval 1 if \a src parses to \<addrrange\> '@' \<net\> + * \retval 0 otherwise + */ +static int +parse_nidrange(struct cfs_lstr *src, struct list_head *nidlist) +{ + struct cfs_lstr addrrange; + struct cfs_lstr net; + struct cfs_lstr tmp; + struct nidrange *nr; + + tmp = *src; + if (cfs_gettok(src, '@', &addrrange) == 0) + goto failed; + + if (cfs_gettok(src, '@', &net) == 0 || src->ls_str != NULL) + goto failed; + + nr = add_nidrange(&net, nidlist); + if (nr == NULL) + goto failed; + + if (parse_addrange(&addrrange, nr) != 0) + goto failed; + + return 1; + failed: + CWARN("can't parse nidrange: \"%.*s\"\n", tmp.ls_len, tmp.ls_str); + return 0; +} + +/** + * Frees addrrange structures of \a list. + * + * For each struct addrrange structure found on \a list it frees + * cfs_expr_list list attached to it and frees the addrrange itself. + * + * \retval none + */ +static void +free_addrranges(struct list_head *list) +{ + while (!list_empty(list)) { + struct addrrange *ar; + + ar = list_entry(list->next, struct addrrange, ar_link); + + cfs_expr_list_free_list(&ar->ar_numaddr_ranges); + list_del(&ar->ar_link); + LIBCFS_FREE(ar, sizeof(struct addrrange)); + } +} + +/** + * Frees nidrange strutures of \a list. + * + * For each struct nidrange structure found on \a list it frees + * addrrange list attached to it and frees the nidrange itself. + * + * \retval none + */ +void +cfs_free_nidlist(struct list_head *list) +{ + struct list_head *pos, *next; + struct nidrange *nr; + + list_for_each_safe(pos, next, list) { + nr = list_entry(pos, struct nidrange, nr_link); + free_addrranges(&nr->nr_addrranges); + list_del(pos); + LIBCFS_FREE(nr, sizeof(struct nidrange)); + } +} + +/** + * Parses nid range list. + * + * Parses with rigorous syntax and overflow checking \a str into + * \<nidrange\> [ ' ' \<nidrange\> ], compiles \a str into set of + * structures and links that structure to \a nidlist. The resulting + * list can be used to match a NID againts set of NIDS defined by \a + * str. + * \see cfs_match_nid + * + * \retval 1 on success + * \retval 0 otherwise + */ +int +cfs_parse_nidlist(char *str, int len, struct list_head *nidlist) +{ + struct cfs_lstr src; + struct cfs_lstr res; + int rc; + ENTRY; + + src.ls_str = str; + src.ls_len = len; + INIT_LIST_HEAD(nidlist); + while (src.ls_str) { + rc = cfs_gettok(&src, ' ', &res); + if (rc == 0) { + cfs_free_nidlist(nidlist); + RETURN(0); + } + rc = parse_nidrange(&res, nidlist); + if (rc == 0) { + cfs_free_nidlist(nidlist); + RETURN(0); + } + } + RETURN(1); +} + +/* + * Nf_match_addr method for networks using numeric addresses + * + * \retval 1 on match + * \retval 0 otherwise + */ +static int +libcfs_num_match(__u32 addr, struct list_head *numaddr) +{ + struct cfs_expr_list *el; + + LASSERT(!list_empty(numaddr)); + el = list_entry(numaddr->next, struct cfs_expr_list, el_link); + + return cfs_expr_list_match(addr, el); +} + +/** + * Matches a nid (\a nid) against the compiled list of nidranges (\a nidlist). + * + * \see cfs_parse_nidlist() + * + * \retval 1 on match + * \retval 0 otherwises + */ +int cfs_match_nid(lnet_nid_t nid, struct list_head *nidlist) +{ + struct nidrange *nr; + struct addrrange *ar; + ENTRY; + + list_for_each_entry(nr, nidlist, nr_link) { + if (nr->nr_netstrfns->nf_type != LNET_NETTYP(LNET_NIDNET(nid))) + continue; + if (nr->nr_netnum != LNET_NETNUM(LNET_NIDNET(nid))) + continue; + if (nr->nr_all) + RETURN(1); + list_for_each_entry(ar, &nr->nr_addrranges, ar_link) + if (nr->nr_netstrfns->nf_match_addr(LNET_NIDADDR(nid), + &ar->ar_numaddr_ranges)) + RETURN(1); + } + RETURN(0); +} + + +EXPORT_SYMBOL(libcfs_isknown_lnd); +EXPORT_SYMBOL(libcfs_lnd2modname); +EXPORT_SYMBOL(libcfs_lnd2str); +EXPORT_SYMBOL(libcfs_str2lnd); +EXPORT_SYMBOL(libcfs_net2str); +EXPORT_SYMBOL(libcfs_nid2str); +EXPORT_SYMBOL(libcfs_str2net); +EXPORT_SYMBOL(libcfs_str2nid); +EXPORT_SYMBOL(libcfs_id2str); +EXPORT_SYMBOL(libcfs_str2anynid); +EXPORT_SYMBOL(cfs_free_nidlist); +EXPORT_SYMBOL(cfs_parse_nidlist); +EXPORT_SYMBOL(cfs_match_nid); diff --git a/drivers/staging/lustre/lustre/libcfs/prng.c b/drivers/staging/lustre/lustre/libcfs/prng.c new file mode 100644 index 000000000000..69224d84bc4b --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/prng.c @@ -0,0 +1,139 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/prng.c + * + * concatenation of following two 16-bit multiply with carry generators + * x(n)=a*x(n-1)+carry mod 2^16 and y(n)=b*y(n-1)+carry mod 2^16, + * number and carry packed within the same 32 bit integer. + * algorithm recommended by Marsaglia +*/ + +#include <linux/libcfs/libcfs.h> + +/* +From: George Marsaglia <geo@stat.fsu.edu> +Newsgroups: sci.math +Subject: Re: A RANDOM NUMBER GENERATOR FOR C +Date: Tue, 30 Sep 1997 05:29:35 -0700 + + * You may replace the two constants 36969 and 18000 by any + * pair of distinct constants from this list: + * 18000 18030 18273 18513 18879 19074 19098 19164 19215 19584 + * 19599 19950 20088 20508 20544 20664 20814 20970 21153 21243 + * 21423 21723 21954 22125 22188 22293 22860 22938 22965 22974 + * 23109 23124 23163 23208 23508 23520 23553 23658 23865 24114 + * 24219 24660 24699 24864 24948 25023 25308 25443 26004 26088 + * 26154 26550 26679 26838 27183 27258 27753 27795 27810 27834 + * 27960 28320 28380 28689 28710 28794 28854 28959 28980 29013 + * 29379 29889 30135 30345 30459 30714 30903 30963 31059 31083 + * (or any other 16-bit constants k for which both k*2^16-1 + * and k*2^15-1 are prime) */ + +#define RANDOM_CONST_A 18030 +#define RANDOM_CONST_B 29013 + +static unsigned int seed_x = 521288629; +static unsigned int seed_y = 362436069; + +/** + * cfs_rand - creates new seeds + * + * First it creates new seeds from the previous seeds. Then it generates a + * new psuedo random number for use. + * + * Returns a pseudo-random 32-bit integer + */ +unsigned int cfs_rand(void) +{ + seed_x = RANDOM_CONST_A * (seed_x & 65535) + (seed_x >> 16); + seed_y = RANDOM_CONST_B * (seed_y & 65535) + (seed_y >> 16); + + return ((seed_x << 16) + (seed_y & 65535)); +} +EXPORT_SYMBOL(cfs_rand); + +/** + * cfs_srand - sets the inital seed + * @seed1 : (seed_x) should have the most entropy in the low bits of the word + * @seed2 : (seed_y) should have the most entropy in the high bits of the word + * + * Replaces the original seeds with new values. Used to generate a new pseudo + * random numbers. + */ +void cfs_srand(unsigned int seed1, unsigned int seed2) +{ + if (seed1) + seed_x = seed1; /* use default seeds if parameter is 0 */ + if (seed2) + seed_y = seed2; +} +EXPORT_SYMBOL(cfs_srand); + +/** + * cfs_get_random_bytes - generate a bunch of random numbers + * @buf : buffer to fill with random numbers + * @size: size of passed in buffer + * + * Fills a buffer with random bytes + */ +void cfs_get_random_bytes(void *buf, int size) +{ + int *p = buf; + int rem, tmp; + + LASSERT(size >= 0); + + rem = min((int)((unsigned long)buf & (sizeof(int) - 1)), size); + if (rem) { + get_random_bytes(&tmp, sizeof(tmp)); + tmp ^= cfs_rand(); + memcpy(buf, &tmp, rem); + p = buf + rem; + size -= rem; + } + + while (size >= sizeof(int)) { + get_random_bytes(&tmp, sizeof(tmp)); + *p = cfs_rand() ^ tmp; + size -= sizeof(int); + p++; + } + buf = p; + if (size) { + get_random_bytes(&tmp, sizeof(tmp)); + tmp ^= cfs_rand(); + memcpy(buf, &tmp, size); + } +} +EXPORT_SYMBOL(cfs_get_random_bytes); diff --git a/drivers/staging/lustre/lustre/libcfs/tracefile.c b/drivers/staging/lustre/lustre/libcfs/tracefile.c new file mode 100644 index 000000000000..439e71dfae33 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/tracefile.c @@ -0,0 +1,1195 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/tracefile.c + * + * Author: Zach Brown <zab@clusterfs.com> + * Author: Phil Schwan <phil@clusterfs.com> + */ + + +#define DEBUG_SUBSYSTEM S_LNET +#define LUSTRE_TRACEFILE_PRIVATE +#include "tracefile.h" + +#include <linux/libcfs/libcfs.h> + +/* XXX move things up to the top, comment */ +union cfs_trace_data_union (*cfs_trace_data[TCD_MAX_TYPES])[NR_CPUS] __cacheline_aligned; + +char cfs_tracefile[TRACEFILE_NAME_SIZE]; +long long cfs_tracefile_size = CFS_TRACEFILE_SIZE; +static struct tracefiled_ctl trace_tctl; +struct mutex cfs_trace_thread_mutex; +static int thread_running = 0; + +atomic_t cfs_tage_allocated = ATOMIC_INIT(0); + +static void put_pages_on_tcd_daemon_list(struct page_collection *pc, + struct cfs_trace_cpu_data *tcd); + +static inline struct cfs_trace_page * +cfs_tage_from_list(struct list_head *list) +{ + return list_entry(list, struct cfs_trace_page, linkage); +} + +static struct cfs_trace_page *cfs_tage_alloc(int gfp) +{ + struct page *page; + struct cfs_trace_page *tage; + + /* My caller is trying to free memory */ + if (!in_interrupt() && memory_pressure_get()) + return NULL; + + /* + * Don't spam console with allocation failures: they will be reported + * by upper layer anyway. + */ + gfp |= __GFP_NOWARN; + page = alloc_page(gfp); + if (page == NULL) + return NULL; + + tage = kmalloc(sizeof(*tage), gfp); + if (tage == NULL) { + __free_page(page); + return NULL; + } + + tage->page = page; + atomic_inc(&cfs_tage_allocated); + return tage; +} + +static void cfs_tage_free(struct cfs_trace_page *tage) +{ + __LASSERT(tage != NULL); + __LASSERT(tage->page != NULL); + + __free_page(tage->page); + kfree(tage); + atomic_dec(&cfs_tage_allocated); +} + +static void cfs_tage_to_tail(struct cfs_trace_page *tage, + struct list_head *queue) +{ + __LASSERT(tage != NULL); + __LASSERT(queue != NULL); + + list_move_tail(&tage->linkage, queue); +} + +int cfs_trace_refill_stock(struct cfs_trace_cpu_data *tcd, int gfp, + struct list_head *stock) +{ + int i; + + /* + * XXX nikita: do NOT call portals_debug_msg() (CDEBUG/ENTRY/EXIT) + * from here: this will lead to infinite recursion. + */ + + for (i = 0; i + tcd->tcd_cur_stock_pages < TCD_STOCK_PAGES ; ++ i) { + struct cfs_trace_page *tage; + + tage = cfs_tage_alloc(gfp); + if (tage == NULL) + break; + list_add_tail(&tage->linkage, stock); + } + return i; +} + +/* return a page that has 'len' bytes left at the end */ +static struct cfs_trace_page * +cfs_trace_get_tage_try(struct cfs_trace_cpu_data *tcd, unsigned long len) +{ + struct cfs_trace_page *tage; + + if (tcd->tcd_cur_pages > 0) { + __LASSERT(!list_empty(&tcd->tcd_pages)); + tage = cfs_tage_from_list(tcd->tcd_pages.prev); + if (tage->used + len <= PAGE_CACHE_SIZE) + return tage; + } + + if (tcd->tcd_cur_pages < tcd->tcd_max_pages) { + if (tcd->tcd_cur_stock_pages > 0) { + tage = cfs_tage_from_list(tcd->tcd_stock_pages.prev); + --tcd->tcd_cur_stock_pages; + list_del_init(&tage->linkage); + } else { + tage = cfs_tage_alloc(GFP_ATOMIC); + if (unlikely(tage == NULL)) { + if ((!memory_pressure_get() || + in_interrupt()) && printk_ratelimit()) + printk(KERN_WARNING + "cannot allocate a tage (%ld)\n", + tcd->tcd_cur_pages); + return NULL; + } + } + + tage->used = 0; + tage->cpu = smp_processor_id(); + tage->type = tcd->tcd_type; + list_add_tail(&tage->linkage, &tcd->tcd_pages); + tcd->tcd_cur_pages++; + + if (tcd->tcd_cur_pages > 8 && thread_running) { + struct tracefiled_ctl *tctl = &trace_tctl; + /* + * wake up tracefiled to process some pages. + */ + wake_up(&tctl->tctl_waitq); + } + return tage; + } + return NULL; +} + +static void cfs_tcd_shrink(struct cfs_trace_cpu_data *tcd) +{ + int pgcount = tcd->tcd_cur_pages / 10; + struct page_collection pc; + struct cfs_trace_page *tage; + struct cfs_trace_page *tmp; + + /* + * XXX nikita: do NOT call portals_debug_msg() (CDEBUG/ENTRY/EXIT) + * from here: this will lead to infinite recursion. + */ + + if (printk_ratelimit()) + printk(KERN_WARNING "debug daemon buffer overflowed; " + "discarding 10%% of pages (%d of %ld)\n", + pgcount + 1, tcd->tcd_cur_pages); + + INIT_LIST_HEAD(&pc.pc_pages); + spin_lock_init(&pc.pc_lock); + + list_for_each_entry_safe(tage, tmp, &tcd->tcd_pages, linkage) { + if (pgcount-- == 0) + break; + + list_move_tail(&tage->linkage, &pc.pc_pages); + tcd->tcd_cur_pages--; + } + put_pages_on_tcd_daemon_list(&pc, tcd); +} + +/* return a page that has 'len' bytes left at the end */ +static struct cfs_trace_page *cfs_trace_get_tage(struct cfs_trace_cpu_data *tcd, + unsigned long len) +{ + struct cfs_trace_page *tage; + + /* + * XXX nikita: do NOT call portals_debug_msg() (CDEBUG/ENTRY/EXIT) + * from here: this will lead to infinite recursion. + */ + + if (len > PAGE_CACHE_SIZE) { + printk(KERN_ERR + "cowardly refusing to write %lu bytes in a page\n", len); + return NULL; + } + + tage = cfs_trace_get_tage_try(tcd, len); + if (tage != NULL) + return tage; + if (thread_running) + cfs_tcd_shrink(tcd); + if (tcd->tcd_cur_pages > 0) { + tage = cfs_tage_from_list(tcd->tcd_pages.next); + tage->used = 0; + cfs_tage_to_tail(tage, &tcd->tcd_pages); + } + return tage; +} + +int libcfs_debug_msg(struct libcfs_debug_msg_data *msgdata, + const char *format, ...) +{ + va_list args; + int rc; + + va_start(args, format); + rc = libcfs_debug_vmsg2(msgdata, format, args, NULL); + va_end(args); + + return rc; +} +EXPORT_SYMBOL(libcfs_debug_msg); + +int libcfs_debug_vmsg2(struct libcfs_debug_msg_data *msgdata, + const char *format1, va_list args, + const char *format2, ...) +{ + struct cfs_trace_cpu_data *tcd = NULL; + struct ptldebug_header header = {0}; + struct cfs_trace_page *tage; + /* string_buf is used only if tcd != NULL, and is always set then */ + char *string_buf = NULL; + char *debug_buf; + int known_size; + int needed = 85; /* average message length */ + int max_nob; + va_list ap; + int depth; + int i; + int remain; + int mask = msgdata->msg_mask; + char *file = (char *)msgdata->msg_file; + cfs_debug_limit_state_t *cdls = msgdata->msg_cdls; + + if (strchr(file, '/')) + file = strrchr(file, '/') + 1; + + tcd = cfs_trace_get_tcd(); + + /* cfs_trace_get_tcd() grabs a lock, which disables preemption and + * pins us to a particular CPU. This avoids an smp_processor_id() + * warning on Linux when debugging is enabled. */ + cfs_set_ptldebug_header(&header, msgdata, CDEBUG_STACK()); + + if (tcd == NULL) /* arch may not log in IRQ context */ + goto console; + + if (tcd->tcd_cur_pages == 0) + header.ph_flags |= PH_FLAG_FIRST_RECORD; + + if (tcd->tcd_shutting_down) { + cfs_trace_put_tcd(tcd); + tcd = NULL; + goto console; + } + + depth = __current_nesting_level(); + known_size = strlen(file) + 1 + depth; + if (msgdata->msg_fn) + known_size += strlen(msgdata->msg_fn) + 1; + + if (libcfs_debug_binary) + known_size += sizeof(header); + + /*/ + * '2' used because vsnprintf return real size required for output + * _without_ terminating NULL. + * if needed is to small for this format. + */ + for (i = 0; i < 2; i++) { + tage = cfs_trace_get_tage(tcd, needed + known_size + 1); + if (tage == NULL) { + if (needed + known_size > PAGE_CACHE_SIZE) + mask |= D_ERROR; + + cfs_trace_put_tcd(tcd); + tcd = NULL; + goto console; + } + + string_buf = (char *)page_address(tage->page) + + tage->used + known_size; + + max_nob = PAGE_CACHE_SIZE - tage->used - known_size; + if (max_nob <= 0) { + printk(KERN_EMERG "negative max_nob: %d\n", + max_nob); + mask |= D_ERROR; + cfs_trace_put_tcd(tcd); + tcd = NULL; + goto console; + } + + needed = 0; + if (format1) { + va_copy(ap, args); + needed = vsnprintf(string_buf, max_nob, format1, ap); + va_end(ap); + } + + if (format2) { + remain = max_nob - needed; + if (remain < 0) + remain = 0; + + va_start(ap, format2); + needed += vsnprintf(string_buf + needed, remain, + format2, ap); + va_end(ap); + } + + if (needed < max_nob) /* well. printing ok.. */ + break; + } + + if (*(string_buf+needed-1) != '\n') + printk(KERN_INFO "format at %s:%d:%s doesn't end in " + "newline\n", file, msgdata->msg_line, msgdata->msg_fn); + + header.ph_len = known_size + needed; + debug_buf = (char *)page_address(tage->page) + tage->used; + + if (libcfs_debug_binary) { + memcpy(debug_buf, &header, sizeof(header)); + tage->used += sizeof(header); + debug_buf += sizeof(header); + } + + /* indent message according to the nesting level */ + while (depth-- > 0) { + *(debug_buf++) = '.'; + ++ tage->used; + } + + strcpy(debug_buf, file); + tage->used += strlen(file) + 1; + debug_buf += strlen(file) + 1; + + if (msgdata->msg_fn) { + strcpy(debug_buf, msgdata->msg_fn); + tage->used += strlen(msgdata->msg_fn) + 1; + debug_buf += strlen(msgdata->msg_fn) + 1; + } + + __LASSERT(debug_buf == string_buf); + + tage->used += needed; + __LASSERT (tage->used <= PAGE_CACHE_SIZE); + +console: + if ((mask & libcfs_printk) == 0) { + /* no console output requested */ + if (tcd != NULL) + cfs_trace_put_tcd(tcd); + return 1; + } + + if (cdls != NULL) { + if (libcfs_console_ratelimit && + cdls->cdls_next != 0 && /* not first time ever */ + !cfs_time_after(cfs_time_current(), cdls->cdls_next)) { + /* skipping a console message */ + cdls->cdls_count++; + if (tcd != NULL) + cfs_trace_put_tcd(tcd); + return 1; + } + + if (cfs_time_after(cfs_time_current(), cdls->cdls_next + + libcfs_console_max_delay + + cfs_time_seconds(10))) { + /* last timeout was a long time ago */ + cdls->cdls_delay /= libcfs_console_backoff * 4; + } else { + cdls->cdls_delay *= libcfs_console_backoff; + + if (cdls->cdls_delay < libcfs_console_min_delay) + cdls->cdls_delay = libcfs_console_min_delay; + else if (cdls->cdls_delay > libcfs_console_max_delay) + cdls->cdls_delay = libcfs_console_max_delay; + } + + /* ensure cdls_next is never zero after it's been seen */ + cdls->cdls_next = (cfs_time_current() + cdls->cdls_delay) | 1; + } + + if (tcd != NULL) { + cfs_print_to_console(&header, mask, string_buf, needed, file, + msgdata->msg_fn); + cfs_trace_put_tcd(tcd); + } else { + string_buf = cfs_trace_get_console_buffer(); + + needed = 0; + if (format1 != NULL) { + va_copy(ap, args); + needed = vsnprintf(string_buf, + CFS_TRACE_CONSOLE_BUFFER_SIZE, + format1, ap); + va_end(ap); + } + if (format2 != NULL) { + remain = CFS_TRACE_CONSOLE_BUFFER_SIZE - needed; + if (remain > 0) { + va_start(ap, format2); + needed += vsnprintf(string_buf+needed, remain, + format2, ap); + va_end(ap); + } + } + cfs_print_to_console(&header, mask, + string_buf, needed, file, msgdata->msg_fn); + + cfs_trace_put_console_buffer(string_buf); + } + + if (cdls != NULL && cdls->cdls_count != 0) { + string_buf = cfs_trace_get_console_buffer(); + + needed = snprintf(string_buf, CFS_TRACE_CONSOLE_BUFFER_SIZE, + "Skipped %d previous similar message%s\n", + cdls->cdls_count, + (cdls->cdls_count > 1) ? "s" : ""); + + cfs_print_to_console(&header, mask, + string_buf, needed, file, msgdata->msg_fn); + + cfs_trace_put_console_buffer(string_buf); + cdls->cdls_count = 0; + } + + return 0; +} +EXPORT_SYMBOL(libcfs_debug_vmsg2); + +void +cfs_trace_assertion_failed(const char *str, + struct libcfs_debug_msg_data *msgdata) +{ + struct ptldebug_header hdr; + + libcfs_panic_in_progress = 1; + libcfs_catastrophe = 1; + mb(); + + cfs_set_ptldebug_header(&hdr, msgdata, CDEBUG_STACK()); + + cfs_print_to_console(&hdr, D_EMERG, str, strlen(str), + msgdata->msg_file, msgdata->msg_fn); + + panic("Lustre debug assertion failure\n"); + + /* not reached */ +} + +static void +panic_collect_pages(struct page_collection *pc) +{ + /* Do the collect_pages job on a single CPU: assumes that all other + * CPUs have been stopped during a panic. If this isn't true for some + * arch, this will have to be implemented separately in each arch. */ + int i; + int j; + struct cfs_trace_cpu_data *tcd; + + INIT_LIST_HEAD(&pc->pc_pages); + + cfs_tcd_for_each(tcd, i, j) { + list_splice_init(&tcd->tcd_pages, &pc->pc_pages); + tcd->tcd_cur_pages = 0; + + if (pc->pc_want_daemon_pages) { + list_splice_init(&tcd->tcd_daemon_pages, + &pc->pc_pages); + tcd->tcd_cur_daemon_pages = 0; + } + } +} + +static void collect_pages_on_all_cpus(struct page_collection *pc) +{ + struct cfs_trace_cpu_data *tcd; + int i, cpu; + + spin_lock(&pc->pc_lock); + cfs_for_each_possible_cpu(cpu) { + cfs_tcd_for_each_type_lock(tcd, i, cpu) { + list_splice_init(&tcd->tcd_pages, &pc->pc_pages); + tcd->tcd_cur_pages = 0; + if (pc->pc_want_daemon_pages) { + list_splice_init(&tcd->tcd_daemon_pages, + &pc->pc_pages); + tcd->tcd_cur_daemon_pages = 0; + } + } + } + spin_unlock(&pc->pc_lock); +} + +static void collect_pages(struct page_collection *pc) +{ + INIT_LIST_HEAD(&pc->pc_pages); + + if (libcfs_panic_in_progress) + panic_collect_pages(pc); + else + collect_pages_on_all_cpus(pc); +} + +static void put_pages_back_on_all_cpus(struct page_collection *pc) +{ + struct cfs_trace_cpu_data *tcd; + struct list_head *cur_head; + struct cfs_trace_page *tage; + struct cfs_trace_page *tmp; + int i, cpu; + + spin_lock(&pc->pc_lock); + cfs_for_each_possible_cpu(cpu) { + cfs_tcd_for_each_type_lock(tcd, i, cpu) { + cur_head = tcd->tcd_pages.next; + + list_for_each_entry_safe(tage, tmp, &pc->pc_pages, + linkage) { + + __LASSERT_TAGE_INVARIANT(tage); + + if (tage->cpu != cpu || tage->type != i) + continue; + + cfs_tage_to_tail(tage, cur_head); + tcd->tcd_cur_pages++; + } + } + } + spin_unlock(&pc->pc_lock); +} + +static void put_pages_back(struct page_collection *pc) +{ + if (!libcfs_panic_in_progress) + put_pages_back_on_all_cpus(pc); +} + +/* Add pages to a per-cpu debug daemon ringbuffer. This buffer makes sure that + * we have a good amount of data at all times for dumping during an LBUG, even + * if we have been steadily writing (and otherwise discarding) pages via the + * debug daemon. */ +static void put_pages_on_tcd_daemon_list(struct page_collection *pc, + struct cfs_trace_cpu_data *tcd) +{ + struct cfs_trace_page *tage; + struct cfs_trace_page *tmp; + + spin_lock(&pc->pc_lock); + list_for_each_entry_safe(tage, tmp, &pc->pc_pages, linkage) { + + __LASSERT_TAGE_INVARIANT(tage); + + if (tage->cpu != tcd->tcd_cpu || tage->type != tcd->tcd_type) + continue; + + cfs_tage_to_tail(tage, &tcd->tcd_daemon_pages); + tcd->tcd_cur_daemon_pages++; + + if (tcd->tcd_cur_daemon_pages > tcd->tcd_max_pages) { + struct cfs_trace_page *victim; + + __LASSERT(!list_empty(&tcd->tcd_daemon_pages)); + victim = cfs_tage_from_list(tcd->tcd_daemon_pages.next); + + __LASSERT_TAGE_INVARIANT(victim); + + list_del(&victim->linkage); + cfs_tage_free(victim); + tcd->tcd_cur_daemon_pages--; + } + } + spin_unlock(&pc->pc_lock); +} + +static void put_pages_on_daemon_list(struct page_collection *pc) +{ + struct cfs_trace_cpu_data *tcd; + int i, cpu; + + cfs_for_each_possible_cpu(cpu) { + cfs_tcd_for_each_type_lock(tcd, i, cpu) + put_pages_on_tcd_daemon_list(pc, tcd); + } +} + +void cfs_trace_debug_print(void) +{ + struct page_collection pc; + struct cfs_trace_page *tage; + struct cfs_trace_page *tmp; + + spin_lock_init(&pc.pc_lock); + + pc.pc_want_daemon_pages = 1; + collect_pages(&pc); + list_for_each_entry_safe(tage, tmp, &pc.pc_pages, linkage) { + char *p, *file, *fn; + struct page *page; + + __LASSERT_TAGE_INVARIANT(tage); + + page = tage->page; + p = page_address(page); + while (p < ((char *)page_address(page) + tage->used)) { + struct ptldebug_header *hdr; + int len; + hdr = (void *)p; + p += sizeof(*hdr); + file = p; + p += strlen(file) + 1; + fn = p; + p += strlen(fn) + 1; + len = hdr->ph_len - (int)(p - (char *)hdr); + + cfs_print_to_console(hdr, D_EMERG, p, len, file, fn); + + p += len; + } + + list_del(&tage->linkage); + cfs_tage_free(tage); + } +} + +int cfs_tracefile_dump_all_pages(char *filename) +{ + struct page_collection pc; + struct file *filp; + struct cfs_trace_page *tage; + struct cfs_trace_page *tmp; + int rc; + + DECL_MMSPACE; + + cfs_tracefile_write_lock(); + + filp = filp_open(filename, O_CREAT|O_EXCL|O_WRONLY|O_LARGEFILE, 0600); + if (IS_ERR(filp)) { + rc = PTR_ERR(filp); + filp = NULL; + printk(KERN_ERR "LustreError: can't open %s for dump: rc %d\n", + filename, rc); + goto out; + } + + spin_lock_init(&pc.pc_lock); + pc.pc_want_daemon_pages = 1; + collect_pages(&pc); + if (list_empty(&pc.pc_pages)) { + rc = 0; + goto close; + } + + /* ok, for now, just write the pages. in the future we'll be building + * iobufs with the pages and calling generic_direct_IO */ + MMSPACE_OPEN; + list_for_each_entry_safe(tage, tmp, &pc.pc_pages, linkage) { + + __LASSERT_TAGE_INVARIANT(tage); + + rc = filp_write(filp, page_address(tage->page), + tage->used, filp_poff(filp)); + if (rc != (int)tage->used) { + printk(KERN_WARNING "wanted to write %u but wrote " + "%d\n", tage->used, rc); + put_pages_back(&pc); + __LASSERT(list_empty(&pc.pc_pages)); + break; + } + list_del(&tage->linkage); + cfs_tage_free(tage); + } + MMSPACE_CLOSE; + rc = filp_fsync(filp); + if (rc) + printk(KERN_ERR "sync returns %d\n", rc); +close: + filp_close(filp, NULL); +out: + cfs_tracefile_write_unlock(); + return rc; +} + +void cfs_trace_flush_pages(void) +{ + struct page_collection pc; + struct cfs_trace_page *tage; + struct cfs_trace_page *tmp; + + spin_lock_init(&pc.pc_lock); + + pc.pc_want_daemon_pages = 1; + collect_pages(&pc); + list_for_each_entry_safe(tage, tmp, &pc.pc_pages, linkage) { + + __LASSERT_TAGE_INVARIANT(tage); + + list_del(&tage->linkage); + cfs_tage_free(tage); + } +} + +int cfs_trace_copyin_string(char *knl_buffer, int knl_buffer_nob, + const char *usr_buffer, int usr_buffer_nob) +{ + int nob; + + if (usr_buffer_nob > knl_buffer_nob) + return -EOVERFLOW; + + if (copy_from_user((void *)knl_buffer, + (void *)usr_buffer, usr_buffer_nob)) + return -EFAULT; + + nob = strnlen(knl_buffer, usr_buffer_nob); + while (nob-- >= 0) /* strip trailing whitespace */ + if (!isspace(knl_buffer[nob])) + break; + + if (nob < 0) /* empty string */ + return -EINVAL; + + if (nob == knl_buffer_nob) /* no space to terminate */ + return -EOVERFLOW; + + knl_buffer[nob + 1] = 0; /* terminate */ + return 0; +} +EXPORT_SYMBOL(cfs_trace_copyin_string); + +int cfs_trace_copyout_string(char *usr_buffer, int usr_buffer_nob, + const char *knl_buffer, char *append) +{ + /* NB if 'append' != NULL, it's a single character to append to the + * copied out string - usually "\n", for /proc entries and "" (i.e. a + * terminating zero byte) for sysctl entries */ + int nob = strlen(knl_buffer); + + if (nob > usr_buffer_nob) + nob = usr_buffer_nob; + + if (copy_to_user(usr_buffer, knl_buffer, nob)) + return -EFAULT; + + if (append != NULL && nob < usr_buffer_nob) { + if (copy_to_user(usr_buffer + nob, append, 1)) + return -EFAULT; + + nob++; + } + + return nob; +} +EXPORT_SYMBOL(cfs_trace_copyout_string); + +int cfs_trace_allocate_string_buffer(char **str, int nob) +{ + if (nob > 2 * PAGE_CACHE_SIZE) /* string must be "sensible" */ + return -EINVAL; + + *str = kmalloc(nob, GFP_IOFS | __GFP_ZERO); + if (*str == NULL) + return -ENOMEM; + + return 0; +} + +void cfs_trace_free_string_buffer(char *str, int nob) +{ + kfree(str); +} + +int cfs_trace_dump_debug_buffer_usrstr(void *usr_str, int usr_str_nob) +{ + char *str; + int rc; + + rc = cfs_trace_allocate_string_buffer(&str, usr_str_nob + 1); + if (rc != 0) + return rc; + + rc = cfs_trace_copyin_string(str, usr_str_nob + 1, + usr_str, usr_str_nob); + if (rc != 0) + goto out; + + if (str[0] != '/') { + rc = -EINVAL; + goto out; + } + rc = cfs_tracefile_dump_all_pages(str); +out: + cfs_trace_free_string_buffer(str, usr_str_nob + 1); + return rc; +} + +int cfs_trace_daemon_command(char *str) +{ + int rc = 0; + + cfs_tracefile_write_lock(); + + if (strcmp(str, "stop") == 0) { + cfs_tracefile_write_unlock(); + cfs_trace_stop_thread(); + cfs_tracefile_write_lock(); + memset(cfs_tracefile, 0, sizeof(cfs_tracefile)); + + } else if (strncmp(str, "size=", 5) == 0) { + cfs_tracefile_size = simple_strtoul(str + 5, NULL, 0); + if (cfs_tracefile_size < 10 || cfs_tracefile_size > 20480) + cfs_tracefile_size = CFS_TRACEFILE_SIZE; + else + cfs_tracefile_size <<= 20; + + } else if (strlen(str) >= sizeof(cfs_tracefile)) { + rc = -ENAMETOOLONG; + } else if (str[0] != '/') { + rc = -EINVAL; + } else { + strcpy(cfs_tracefile, str); + + printk(KERN_INFO + "Lustre: debug daemon will attempt to start writing " + "to %s (%lukB max)\n", cfs_tracefile, + (long)(cfs_tracefile_size >> 10)); + + cfs_trace_start_thread(); + } + + cfs_tracefile_write_unlock(); + return rc; +} + +int cfs_trace_daemon_command_usrstr(void *usr_str, int usr_str_nob) +{ + char *str; + int rc; + + rc = cfs_trace_allocate_string_buffer(&str, usr_str_nob + 1); + if (rc != 0) + return rc; + + rc = cfs_trace_copyin_string(str, usr_str_nob + 1, + usr_str, usr_str_nob); + if (rc == 0) + rc = cfs_trace_daemon_command(str); + + cfs_trace_free_string_buffer(str, usr_str_nob + 1); + return rc; +} + +int cfs_trace_set_debug_mb(int mb) +{ + int i; + int j; + int pages; + int limit = cfs_trace_max_debug_mb(); + struct cfs_trace_cpu_data *tcd; + + if (mb < num_possible_cpus()) { + printk(KERN_WARNING + "Lustre: %d MB is too small for debug buffer size, " + "setting it to %d MB.\n", mb, num_possible_cpus()); + mb = num_possible_cpus(); + } + + if (mb > limit) { + printk(KERN_WARNING + "Lustre: %d MB is too large for debug buffer size, " + "setting it to %d MB.\n", mb, limit); + mb = limit; + } + + mb /= num_possible_cpus(); + pages = mb << (20 - PAGE_CACHE_SHIFT); + + cfs_tracefile_write_lock(); + + cfs_tcd_for_each(tcd, i, j) + tcd->tcd_max_pages = (pages * tcd->tcd_pages_factor) / 100; + + cfs_tracefile_write_unlock(); + + return 0; +} + +int cfs_trace_set_debug_mb_usrstr(void *usr_str, int usr_str_nob) +{ + char str[32]; + int rc; + + rc = cfs_trace_copyin_string(str, sizeof(str), usr_str, usr_str_nob); + if (rc < 0) + return rc; + + return cfs_trace_set_debug_mb(simple_strtoul(str, NULL, 0)); +} + +int cfs_trace_get_debug_mb(void) +{ + int i; + int j; + struct cfs_trace_cpu_data *tcd; + int total_pages = 0; + + cfs_tracefile_read_lock(); + + cfs_tcd_for_each(tcd, i, j) + total_pages += tcd->tcd_max_pages; + + cfs_tracefile_read_unlock(); + + return (total_pages >> (20 - PAGE_CACHE_SHIFT)) + 1; +} + +static int tracefiled(void *arg) +{ + struct page_collection pc; + struct tracefiled_ctl *tctl = arg; + struct cfs_trace_page *tage; + struct cfs_trace_page *tmp; + struct file *filp; + int last_loop = 0; + int rc; + + DECL_MMSPACE; + + /* we're started late enough that we pick up init's fs context */ + /* this is so broken in uml? what on earth is going on? */ + + spin_lock_init(&pc.pc_lock); + complete(&tctl->tctl_start); + + while (1) { + wait_queue_t __wait; + + pc.pc_want_daemon_pages = 0; + collect_pages(&pc); + if (list_empty(&pc.pc_pages)) + goto end_loop; + + filp = NULL; + cfs_tracefile_read_lock(); + if (cfs_tracefile[0] != 0) { + filp = filp_open(cfs_tracefile, + O_CREAT | O_RDWR | O_LARGEFILE, + 0600); + if (IS_ERR(filp)) { + rc = PTR_ERR(filp); + filp = NULL; + printk(KERN_WARNING "couldn't open %s: " + "%d\n", cfs_tracefile, rc); + } + } + cfs_tracefile_read_unlock(); + if (filp == NULL) { + put_pages_on_daemon_list(&pc); + __LASSERT(list_empty(&pc.pc_pages)); + goto end_loop; + } + + MMSPACE_OPEN; + + list_for_each_entry_safe(tage, tmp, &pc.pc_pages, + linkage) { + static loff_t f_pos; + + __LASSERT_TAGE_INVARIANT(tage); + + if (f_pos >= (off_t)cfs_tracefile_size) + f_pos = 0; + else if (f_pos > (off_t)filp_size(filp)) + f_pos = filp_size(filp); + + rc = filp_write(filp, page_address(tage->page), + tage->used, &f_pos); + if (rc != (int)tage->used) { + printk(KERN_WARNING "wanted to write %u " + "but wrote %d\n", tage->used, rc); + put_pages_back(&pc); + __LASSERT(list_empty(&pc.pc_pages)); + } + } + MMSPACE_CLOSE; + + filp_close(filp, NULL); + put_pages_on_daemon_list(&pc); + if (!list_empty(&pc.pc_pages)) { + int i; + + printk(KERN_ALERT "Lustre: trace pages aren't " + " empty\n"); + printk(KERN_ERR "total cpus(%d): ", + num_possible_cpus()); + for (i = 0; i < num_possible_cpus(); i++) + if (cpu_online(i)) + printk(KERN_ERR "%d(on) ", i); + else + printk(KERN_ERR "%d(off) ", i); + printk(KERN_ERR "\n"); + + i = 0; + list_for_each_entry_safe(tage, tmp, &pc.pc_pages, + linkage) + printk(KERN_ERR "page %d belongs to cpu " + "%d\n", ++i, tage->cpu); + printk(KERN_ERR "There are %d pages unwritten\n", + i); + } + __LASSERT(list_empty(&pc.pc_pages)); +end_loop: + if (atomic_read(&tctl->tctl_shutdown)) { + if (last_loop == 0) { + last_loop = 1; + continue; + } else { + break; + } + } + init_waitqueue_entry_current(&__wait); + add_wait_queue(&tctl->tctl_waitq, &__wait); + set_current_state(TASK_INTERRUPTIBLE); + waitq_timedwait(&__wait, TASK_INTERRUPTIBLE, + cfs_time_seconds(1)); + remove_wait_queue(&tctl->tctl_waitq, &__wait); + } + complete(&tctl->tctl_stop); + return 0; +} + +int cfs_trace_start_thread(void) +{ + struct tracefiled_ctl *tctl = &trace_tctl; + int rc = 0; + + mutex_lock(&cfs_trace_thread_mutex); + if (thread_running) + goto out; + + init_completion(&tctl->tctl_start); + init_completion(&tctl->tctl_stop); + init_waitqueue_head(&tctl->tctl_waitq); + atomic_set(&tctl->tctl_shutdown, 0); + + if (IS_ERR(kthread_run(tracefiled, tctl, "ktracefiled"))) { + rc = -ECHILD; + goto out; + } + + wait_for_completion(&tctl->tctl_start); + thread_running = 1; +out: + mutex_unlock(&cfs_trace_thread_mutex); + return rc; +} + +void cfs_trace_stop_thread(void) +{ + struct tracefiled_ctl *tctl = &trace_tctl; + + mutex_lock(&cfs_trace_thread_mutex); + if (thread_running) { + printk(KERN_INFO + "Lustre: shutting down debug daemon thread...\n"); + atomic_set(&tctl->tctl_shutdown, 1); + wait_for_completion(&tctl->tctl_stop); + thread_running = 0; + } + mutex_unlock(&cfs_trace_thread_mutex); +} + +int cfs_tracefile_init(int max_pages) +{ + struct cfs_trace_cpu_data *tcd; + int i; + int j; + int rc; + int factor; + + rc = cfs_tracefile_init_arch(); + if (rc != 0) + return rc; + + cfs_tcd_for_each(tcd, i, j) { + /* tcd_pages_factor is initialized int tracefile_init_arch. */ + factor = tcd->tcd_pages_factor; + INIT_LIST_HEAD(&tcd->tcd_pages); + INIT_LIST_HEAD(&tcd->tcd_stock_pages); + INIT_LIST_HEAD(&tcd->tcd_daemon_pages); + tcd->tcd_cur_pages = 0; + tcd->tcd_cur_stock_pages = 0; + tcd->tcd_cur_daemon_pages = 0; + tcd->tcd_max_pages = (max_pages * factor) / 100; + LASSERT(tcd->tcd_max_pages > 0); + tcd->tcd_shutting_down = 0; + } + + return 0; +} + +static void trace_cleanup_on_all_cpus(void) +{ + struct cfs_trace_cpu_data *tcd; + struct cfs_trace_page *tage; + struct cfs_trace_page *tmp; + int i, cpu; + + cfs_for_each_possible_cpu(cpu) { + cfs_tcd_for_each_type_lock(tcd, i, cpu) { + tcd->tcd_shutting_down = 1; + + list_for_each_entry_safe(tage, tmp, &tcd->tcd_pages, + linkage) { + __LASSERT_TAGE_INVARIANT(tage); + + list_del(&tage->linkage); + cfs_tage_free(tage); + } + + tcd->tcd_cur_pages = 0; + } + } +} + +static void cfs_trace_cleanup(void) +{ + struct page_collection pc; + + INIT_LIST_HEAD(&pc.pc_pages); + spin_lock_init(&pc.pc_lock); + + trace_cleanup_on_all_cpus(); + + cfs_tracefile_fini_arch(); +} + +void cfs_tracefile_exit(void) +{ + cfs_trace_stop_thread(); + cfs_trace_cleanup(); +} diff --git a/drivers/staging/lustre/lustre/libcfs/tracefile.h b/drivers/staging/lustre/lustre/libcfs/tracefile.h new file mode 100644 index 000000000000..7e8d17c12b5b --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/tracefile.h @@ -0,0 +1,340 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + */ + +#ifndef __LIBCFS_TRACEFILE_H__ +#define __LIBCFS_TRACEFILE_H__ + +#include <linux/libcfs/libcfs.h> + +#include "linux/linux-tracefile.h" + +/* trace file lock routines */ + +#define TRACEFILE_NAME_SIZE 1024 +extern char cfs_tracefile[TRACEFILE_NAME_SIZE]; +extern long long cfs_tracefile_size; + +extern void libcfs_run_debug_log_upcall(char *file); + +int cfs_tracefile_init_arch(void); +void cfs_tracefile_fini_arch(void); + +void cfs_tracefile_read_lock(void); +void cfs_tracefile_read_unlock(void); +void cfs_tracefile_write_lock(void); +void cfs_tracefile_write_unlock(void); + +int cfs_tracefile_dump_all_pages(char *filename); +void cfs_trace_debug_print(void); +void cfs_trace_flush_pages(void); +int cfs_trace_start_thread(void); +void cfs_trace_stop_thread(void); +int cfs_tracefile_init(int max_pages); +void cfs_tracefile_exit(void); + + + +int cfs_trace_copyin_string(char *knl_buffer, int knl_buffer_nob, + const char *usr_buffer, int usr_buffer_nob); +int cfs_trace_copyout_string(char *usr_buffer, int usr_buffer_nob, + const char *knl_str, char *append); +int cfs_trace_allocate_string_buffer(char **str, int nob); +void cfs_trace_free_string_buffer(char *str, int nob); +int cfs_trace_dump_debug_buffer_usrstr(void *usr_str, int usr_str_nob); +int cfs_trace_daemon_command(char *str); +int cfs_trace_daemon_command_usrstr(void *usr_str, int usr_str_nob); +int cfs_trace_set_debug_mb(int mb); +int cfs_trace_set_debug_mb_usrstr(void *usr_str, int usr_str_nob); +int cfs_trace_get_debug_mb(void); + +extern void libcfs_debug_dumplog_internal(void *arg); +extern void libcfs_register_panic_notifier(void); +extern void libcfs_unregister_panic_notifier(void); +extern int libcfs_panic_in_progress; +extern int cfs_trace_max_debug_mb(void); + +#define TCD_MAX_PAGES (5 << (20 - PAGE_CACHE_SHIFT)) +#define TCD_STOCK_PAGES (TCD_MAX_PAGES) +#define CFS_TRACEFILE_SIZE (500 << 20) + +#ifdef LUSTRE_TRACEFILE_PRIVATE + +/* + * Private declare for tracefile + */ +#define TCD_MAX_PAGES (5 << (20 - PAGE_CACHE_SHIFT)) +#define TCD_STOCK_PAGES (TCD_MAX_PAGES) + +#define CFS_TRACEFILE_SIZE (500 << 20) + +/* Size of a buffer for sprinting console messages if we can't get a page + * from system */ +#define CFS_TRACE_CONSOLE_BUFFER_SIZE 1024 + +union cfs_trace_data_union { + struct cfs_trace_cpu_data { + /* + * Even though this structure is meant to be per-CPU, locking + * is needed because in some places the data may be accessed + * from other CPUs. This lock is directly used in trace_get_tcd + * and trace_put_tcd, which are called in libcfs_debug_vmsg2 and + * tcd_for_each_type_lock + */ + spinlock_t tcd_lock; + unsigned long tcd_lock_flags; + + /* + * pages with trace records not yet processed by tracefiled. + */ + struct list_head tcd_pages; + /* number of pages on ->tcd_pages */ + unsigned long tcd_cur_pages; + + /* + * pages with trace records already processed by + * tracefiled. These pages are kept in memory, so that some + * portion of log can be written in the event of LBUG. This + * list is maintained in LRU order. + * + * Pages are moved to ->tcd_daemon_pages by tracefiled() + * (put_pages_on_daemon_list()). LRU pages from this list are + * discarded when list grows too large. + */ + struct list_head tcd_daemon_pages; + /* number of pages on ->tcd_daemon_pages */ + unsigned long tcd_cur_daemon_pages; + + /* + * Maximal number of pages allowed on ->tcd_pages and + * ->tcd_daemon_pages each. + * Always TCD_MAX_PAGES * tcd_pages_factor / 100 in current + * implementation. + */ + unsigned long tcd_max_pages; + + /* + * preallocated pages to write trace records into. Pages from + * ->tcd_stock_pages are moved to ->tcd_pages by + * portals_debug_msg(). + * + * This list is necessary, because on some platforms it's + * impossible to perform efficient atomic page allocation in a + * non-blockable context. + * + * Such platforms fill ->tcd_stock_pages "on occasion", when + * tracing code is entered in blockable context. + * + * trace_get_tage_try() tries to get a page from + * ->tcd_stock_pages first and resorts to atomic page + * allocation only if this queue is empty. ->tcd_stock_pages + * is replenished when tracing code is entered in blocking + * context (darwin-tracefile.c:trace_get_tcd()). We try to + * maintain TCD_STOCK_PAGES (40 by default) pages in this + * queue. Atomic allocation is only required if more than + * TCD_STOCK_PAGES pagesful are consumed by trace records all + * emitted in non-blocking contexts. Which is quite unlikely. + */ + struct list_head tcd_stock_pages; + /* number of pages on ->tcd_stock_pages */ + unsigned long tcd_cur_stock_pages; + + unsigned short tcd_shutting_down; + unsigned short tcd_cpu; + unsigned short tcd_type; + /* The factors to share debug memory. */ + unsigned short tcd_pages_factor; + } tcd; + char __pad[L1_CACHE_ALIGN(sizeof(struct cfs_trace_cpu_data))]; +}; + +#define TCD_MAX_TYPES 8 +extern union cfs_trace_data_union (*cfs_trace_data[TCD_MAX_TYPES])[NR_CPUS]; + +#define cfs_tcd_for_each(tcd, i, j) \ + for (i = 0; cfs_trace_data[i] != NULL; i++) \ + for (j = 0, ((tcd) = &(*cfs_trace_data[i])[j].tcd); \ + j < num_possible_cpus(); \ + j++, (tcd) = &(*cfs_trace_data[i])[j].tcd) + +#define cfs_tcd_for_each_type_lock(tcd, i, cpu) \ + for (i = 0; cfs_trace_data[i] && \ + (tcd = &(*cfs_trace_data[i])[cpu].tcd) && \ + cfs_trace_lock_tcd(tcd, 1); cfs_trace_unlock_tcd(tcd, 1), i++) + +/* XXX nikita: this declaration is internal to tracefile.c and should probably + * be moved there */ +struct page_collection { + struct list_head pc_pages; + /* + * spin-lock protecting ->pc_pages. It is taken by smp_call_function() + * call-back functions. XXX nikita: Which is horrible: all processors + * receive NMI at the same time only to be serialized by this + * lock. Probably ->pc_pages should be replaced with an array of + * NR_CPUS elements accessed locklessly. + */ + spinlock_t pc_lock; + /* + * if this flag is set, collect_pages() will spill both + * ->tcd_daemon_pages and ->tcd_pages to the ->pc_pages. Otherwise, + * only ->tcd_pages are spilled. + */ + int pc_want_daemon_pages; +}; + +/* XXX nikita: this declaration is internal to tracefile.c and should probably + * be moved there */ +struct tracefiled_ctl { + struct completion tctl_start; + struct completion tctl_stop; + wait_queue_head_t tctl_waitq; + pid_t tctl_pid; + atomic_t tctl_shutdown; +}; + +/* + * small data-structure for each page owned by tracefiled. + */ +/* XXX nikita: this declaration is internal to tracefile.c and should probably + * be moved there */ +struct cfs_trace_page { + /* + * page itself + */ + struct page *page; + /* + * linkage into one of the lists in trace_data_union or + * page_collection + */ + struct list_head linkage; + /* + * number of bytes used within this page + */ + unsigned int used; + /* + * cpu that owns this page + */ + unsigned short cpu; + /* + * type(context) of this page + */ + unsigned short type; +}; + +extern void cfs_set_ptldebug_header(struct ptldebug_header *header, + struct libcfs_debug_msg_data *m, + unsigned long stack); +extern void cfs_print_to_console(struct ptldebug_header *hdr, int mask, + const char *buf, int len, const char *file, + const char *fn); + +extern int cfs_trace_lock_tcd(struct cfs_trace_cpu_data *tcd, int walking); +extern void cfs_trace_unlock_tcd(struct cfs_trace_cpu_data *tcd, int walking); + +/** + * trace_buf_type_t, trace_buf_idx_get() and trace_console_buffers[][] + * are not public libcfs API; they should be defined in + * platform-specific tracefile include files + * (see, for example, linux-tracefile.h). + */ + +extern char *cfs_trace_console_buffers[NR_CPUS][CFS_TCD_TYPE_MAX]; +extern cfs_trace_buf_type_t cfs_trace_buf_idx_get(void); + +static inline char * +cfs_trace_get_console_buffer(void) +{ + unsigned int i = get_cpu(); + unsigned int j = cfs_trace_buf_idx_get(); + + return cfs_trace_console_buffers[i][j]; +} + +static inline void +cfs_trace_put_console_buffer(char *buffer) +{ + put_cpu(); +} + +static inline struct cfs_trace_cpu_data * +cfs_trace_get_tcd(void) +{ + struct cfs_trace_cpu_data *tcd = + &(*cfs_trace_data[cfs_trace_buf_idx_get()])[get_cpu()].tcd; + + cfs_trace_lock_tcd(tcd, 0); + + return tcd; +} + +static inline void +cfs_trace_put_tcd (struct cfs_trace_cpu_data *tcd) +{ + cfs_trace_unlock_tcd(tcd, 0); + + put_cpu(); +} + +int cfs_trace_refill_stock(struct cfs_trace_cpu_data *tcd, int gfp, + struct list_head *stock); + + +int cfs_tcd_owns_tage(struct cfs_trace_cpu_data *tcd, + struct cfs_trace_page *tage); + +extern void cfs_trace_assertion_failed(const char *str, + struct libcfs_debug_msg_data *m); + +/* ASSERTION that is safe to use within the debug system */ +#define __LASSERT(cond) \ +do { \ + if (unlikely(!(cond))) { \ + LIBCFS_DEBUG_MSG_DATA_DECL(msgdata, D_EMERG, NULL); \ + cfs_trace_assertion_failed("ASSERTION("#cond") failed", \ + &msgdata); \ + } \ +} while (0) + +#define __LASSERT_TAGE_INVARIANT(tage) \ +do { \ + __LASSERT(tage != NULL); \ + __LASSERT(tage->page != NULL); \ + __LASSERT(tage->used <= PAGE_CACHE_SIZE); \ + __LASSERT(page_count(tage->page) > 0); \ +} while (0) + +#endif /* LUSTRE_TRACEFILE_PRIVATE */ + +#endif /* __LIBCFS_TRACEFILE_H__ */ diff --git a/drivers/staging/lustre/lustre/libcfs/upcall_cache.c b/drivers/staging/lustre/lustre/libcfs/upcall_cache.c new file mode 100644 index 000000000000..18c68c3493b8 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/upcall_cache.c @@ -0,0 +1,462 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/upcall_cache.c + * + * Supplementary groups cache. + */ +#define DEBUG_SUBSYSTEM S_SEC + +#include <linux/libcfs/lucache.h> + +static struct upcall_cache_entry *alloc_entry(struct upcall_cache *cache, + __u64 key, void *args) +{ + struct upcall_cache_entry *entry; + + LIBCFS_ALLOC(entry, sizeof(*entry)); + if (!entry) + return NULL; + + UC_CACHE_SET_NEW(entry); + INIT_LIST_HEAD(&entry->ue_hash); + entry->ue_key = key; + atomic_set(&entry->ue_refcount, 0); + init_waitqueue_head(&entry->ue_waitq); + if (cache->uc_ops->init_entry) + cache->uc_ops->init_entry(entry, args); + return entry; +} + +/* protected by cache lock */ +static void free_entry(struct upcall_cache *cache, + struct upcall_cache_entry *entry) +{ + if (cache->uc_ops->free_entry) + cache->uc_ops->free_entry(cache, entry); + + list_del(&entry->ue_hash); + CDEBUG(D_OTHER, "destroy cache entry %p for key "LPU64"\n", + entry, entry->ue_key); + LIBCFS_FREE(entry, sizeof(*entry)); +} + +static inline int upcall_compare(struct upcall_cache *cache, + struct upcall_cache_entry *entry, + __u64 key, void *args) +{ + if (entry->ue_key != key) + return -1; + + if (cache->uc_ops->upcall_compare) + return cache->uc_ops->upcall_compare(cache, entry, key, args); + + return 0; +} + +static inline int downcall_compare(struct upcall_cache *cache, + struct upcall_cache_entry *entry, + __u64 key, void *args) +{ + if (entry->ue_key != key) + return -1; + + if (cache->uc_ops->downcall_compare) + return cache->uc_ops->downcall_compare(cache, entry, key, args); + + return 0; +} + +static inline void get_entry(struct upcall_cache_entry *entry) +{ + atomic_inc(&entry->ue_refcount); +} + +static inline void put_entry(struct upcall_cache *cache, + struct upcall_cache_entry *entry) +{ + if (atomic_dec_and_test(&entry->ue_refcount) && + (UC_CACHE_IS_INVALID(entry) || UC_CACHE_IS_EXPIRED(entry))) { + free_entry(cache, entry); + } +} + +static int check_unlink_entry(struct upcall_cache *cache, + struct upcall_cache_entry *entry) +{ + if (UC_CACHE_IS_VALID(entry) && + cfs_time_before(cfs_time_current(), entry->ue_expire)) + return 0; + + if (UC_CACHE_IS_ACQUIRING(entry)) { + if (entry->ue_acquire_expire == 0 || + cfs_time_before(cfs_time_current(), + entry->ue_acquire_expire)) + return 0; + + UC_CACHE_SET_EXPIRED(entry); + wake_up_all(&entry->ue_waitq); + } else if (!UC_CACHE_IS_INVALID(entry)) { + UC_CACHE_SET_EXPIRED(entry); + } + + list_del_init(&entry->ue_hash); + if (!atomic_read(&entry->ue_refcount)) + free_entry(cache, entry); + return 1; +} + +static inline int refresh_entry(struct upcall_cache *cache, + struct upcall_cache_entry *entry) +{ + LASSERT(cache->uc_ops->do_upcall); + return cache->uc_ops->do_upcall(cache, entry); +} + +struct upcall_cache_entry *upcall_cache_get_entry(struct upcall_cache *cache, + __u64 key, void *args) +{ + struct upcall_cache_entry *entry = NULL, *new = NULL, *next; + struct list_head *head; + wait_queue_t wait; + int rc, found; + ENTRY; + + LASSERT(cache); + + head = &cache->uc_hashtable[UC_CACHE_HASH_INDEX(key)]; +find_again: + found = 0; + spin_lock(&cache->uc_lock); + list_for_each_entry_safe(entry, next, head, ue_hash) { + /* check invalid & expired items */ + if (check_unlink_entry(cache, entry)) + continue; + if (upcall_compare(cache, entry, key, args) == 0) { + found = 1; + break; + } + } + + if (!found) { + if (!new) { + spin_unlock(&cache->uc_lock); + new = alloc_entry(cache, key, args); + if (!new) { + CERROR("fail to alloc entry\n"); + RETURN(ERR_PTR(-ENOMEM)); + } + goto find_again; + } else { + list_add(&new->ue_hash, head); + entry = new; + } + } else { + if (new) { + free_entry(cache, new); + new = NULL; + } + list_move(&entry->ue_hash, head); + } + get_entry(entry); + + /* acquire for new one */ + if (UC_CACHE_IS_NEW(entry)) { + UC_CACHE_SET_ACQUIRING(entry); + UC_CACHE_CLEAR_NEW(entry); + spin_unlock(&cache->uc_lock); + rc = refresh_entry(cache, entry); + spin_lock(&cache->uc_lock); + entry->ue_acquire_expire = + cfs_time_shift(cache->uc_acquire_expire); + if (rc < 0) { + UC_CACHE_CLEAR_ACQUIRING(entry); + UC_CACHE_SET_INVALID(entry); + wake_up_all(&entry->ue_waitq); + if (unlikely(rc == -EREMCHG)) { + put_entry(cache, entry); + GOTO(out, entry = ERR_PTR(rc)); + } + } + } + /* someone (and only one) is doing upcall upon this item, + * wait it to complete */ + if (UC_CACHE_IS_ACQUIRING(entry)) { + long expiry = (entry == new) ? + cfs_time_seconds(cache->uc_acquire_expire) : + MAX_SCHEDULE_TIMEOUT; + long left; + + init_waitqueue_entry_current(&wait); + add_wait_queue(&entry->ue_waitq, &wait); + set_current_state(TASK_INTERRUPTIBLE); + spin_unlock(&cache->uc_lock); + + left = waitq_timedwait(&wait, TASK_INTERRUPTIBLE, + expiry); + + spin_lock(&cache->uc_lock); + remove_wait_queue(&entry->ue_waitq, &wait); + if (UC_CACHE_IS_ACQUIRING(entry)) { + /* we're interrupted or upcall failed in the middle */ + rc = left > 0 ? -EINTR : -ETIMEDOUT; + CERROR("acquire for key "LPU64": error %d\n", + entry->ue_key, rc); + put_entry(cache, entry); + GOTO(out, entry = ERR_PTR(rc)); + } + } + + /* invalid means error, don't need to try again */ + if (UC_CACHE_IS_INVALID(entry)) { + put_entry(cache, entry); + GOTO(out, entry = ERR_PTR(-EIDRM)); + } + + /* check expired + * We can't refresh the existing one because some + * memory might be shared by multiple processes. + */ + if (check_unlink_entry(cache, entry)) { + /* if expired, try again. but if this entry is + * created by me but too quickly turn to expired + * without any error, should at least give a + * chance to use it once. + */ + if (entry != new) { + put_entry(cache, entry); + spin_unlock(&cache->uc_lock); + new = NULL; + goto find_again; + } + } + + /* Now we know it's good */ +out: + spin_unlock(&cache->uc_lock); + RETURN(entry); +} +EXPORT_SYMBOL(upcall_cache_get_entry); + +void upcall_cache_put_entry(struct upcall_cache *cache, + struct upcall_cache_entry *entry) +{ + ENTRY; + + if (!entry) { + EXIT; + return; + } + + LASSERT(atomic_read(&entry->ue_refcount) > 0); + spin_lock(&cache->uc_lock); + put_entry(cache, entry); + spin_unlock(&cache->uc_lock); + EXIT; +} +EXPORT_SYMBOL(upcall_cache_put_entry); + +int upcall_cache_downcall(struct upcall_cache *cache, __u32 err, __u64 key, + void *args) +{ + struct upcall_cache_entry *entry = NULL; + struct list_head *head; + int found = 0, rc = 0; + ENTRY; + + LASSERT(cache); + + head = &cache->uc_hashtable[UC_CACHE_HASH_INDEX(key)]; + + spin_lock(&cache->uc_lock); + list_for_each_entry(entry, head, ue_hash) { + if (downcall_compare(cache, entry, key, args) == 0) { + found = 1; + get_entry(entry); + break; + } + } + + if (!found) { + CDEBUG(D_OTHER, "%s: upcall for key "LPU64" not expected\n", + cache->uc_name, key); + /* haven't found, it's possible */ + spin_unlock(&cache->uc_lock); + RETURN(-EINVAL); + } + + if (err) { + CDEBUG(D_OTHER, "%s: upcall for key "LPU64" returned %d\n", + cache->uc_name, entry->ue_key, err); + GOTO(out, rc = -EINVAL); + } + + if (!UC_CACHE_IS_ACQUIRING(entry)) { + CDEBUG(D_RPCTRACE,"%s: found uptodate entry %p (key "LPU64")\n", + cache->uc_name, entry, entry->ue_key); + GOTO(out, rc = 0); + } + + if (UC_CACHE_IS_INVALID(entry) || UC_CACHE_IS_EXPIRED(entry)) { + CERROR("%s: found a stale entry %p (key "LPU64") in ioctl\n", + cache->uc_name, entry, entry->ue_key); + GOTO(out, rc = -EINVAL); + } + + spin_unlock(&cache->uc_lock); + if (cache->uc_ops->parse_downcall) + rc = cache->uc_ops->parse_downcall(cache, entry, args); + spin_lock(&cache->uc_lock); + if (rc) + GOTO(out, rc); + + entry->ue_expire = cfs_time_shift(cache->uc_entry_expire); + UC_CACHE_SET_VALID(entry); + CDEBUG(D_OTHER, "%s: created upcall cache entry %p for key "LPU64"\n", + cache->uc_name, entry, entry->ue_key); +out: + if (rc) { + UC_CACHE_SET_INVALID(entry); + list_del_init(&entry->ue_hash); + } + UC_CACHE_CLEAR_ACQUIRING(entry); + spin_unlock(&cache->uc_lock); + wake_up_all(&entry->ue_waitq); + put_entry(cache, entry); + + RETURN(rc); +} +EXPORT_SYMBOL(upcall_cache_downcall); + +static void cache_flush(struct upcall_cache *cache, int force) +{ + struct upcall_cache_entry *entry, *next; + int i; + ENTRY; + + spin_lock(&cache->uc_lock); + for (i = 0; i < UC_CACHE_HASH_SIZE; i++) { + list_for_each_entry_safe(entry, next, + &cache->uc_hashtable[i], ue_hash) { + if (!force && atomic_read(&entry->ue_refcount)) { + UC_CACHE_SET_EXPIRED(entry); + continue; + } + LASSERT(!atomic_read(&entry->ue_refcount)); + free_entry(cache, entry); + } + } + spin_unlock(&cache->uc_lock); + EXIT; +} + +void upcall_cache_flush_idle(struct upcall_cache *cache) +{ + cache_flush(cache, 0); +} +EXPORT_SYMBOL(upcall_cache_flush_idle); + +void upcall_cache_flush_all(struct upcall_cache *cache) +{ + cache_flush(cache, 1); +} +EXPORT_SYMBOL(upcall_cache_flush_all); + +void upcall_cache_flush_one(struct upcall_cache *cache, __u64 key, void *args) +{ + struct list_head *head; + struct upcall_cache_entry *entry; + int found = 0; + ENTRY; + + head = &cache->uc_hashtable[UC_CACHE_HASH_INDEX(key)]; + + spin_lock(&cache->uc_lock); + list_for_each_entry(entry, head, ue_hash) { + if (upcall_compare(cache, entry, key, args) == 0) { + found = 1; + break; + } + } + + if (found) { + CWARN("%s: flush entry %p: key "LPU64", ref %d, fl %x, " + "cur %lu, ex %ld/%ld\n", + cache->uc_name, entry, entry->ue_key, + atomic_read(&entry->ue_refcount), entry->ue_flags, + cfs_time_current_sec(), entry->ue_acquire_expire, + entry->ue_expire); + UC_CACHE_SET_EXPIRED(entry); + if (!atomic_read(&entry->ue_refcount)) + free_entry(cache, entry); + } + spin_unlock(&cache->uc_lock); +} +EXPORT_SYMBOL(upcall_cache_flush_one); + +struct upcall_cache *upcall_cache_init(const char *name, const char *upcall, + struct upcall_cache_ops *ops) +{ + struct upcall_cache *cache; + int i; + ENTRY; + + LIBCFS_ALLOC(cache, sizeof(*cache)); + if (!cache) + RETURN(ERR_PTR(-ENOMEM)); + + spin_lock_init(&cache->uc_lock); + rwlock_init(&cache->uc_upcall_rwlock); + for (i = 0; i < UC_CACHE_HASH_SIZE; i++) + INIT_LIST_HEAD(&cache->uc_hashtable[i]); + strncpy(cache->uc_name, name, sizeof(cache->uc_name) - 1); + /* upcall pathname proc tunable */ + strncpy(cache->uc_upcall, upcall, sizeof(cache->uc_upcall) - 1); + cache->uc_entry_expire = 20 * 60; + cache->uc_acquire_expire = 30; + cache->uc_ops = ops; + + RETURN(cache); +} +EXPORT_SYMBOL(upcall_cache_init); + +void upcall_cache_cleanup(struct upcall_cache *cache) +{ + if (!cache) + return; + upcall_cache_flush_all(cache); + LIBCFS_FREE(cache, sizeof(*cache)); +} +EXPORT_SYMBOL(upcall_cache_cleanup); diff --git a/drivers/staging/lustre/lustre/libcfs/watchdog.c b/drivers/staging/lustre/lustre/libcfs/watchdog.c new file mode 100644 index 000000000000..7c385ada3e10 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/watchdog.c @@ -0,0 +1,516 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/watchdog.c + * + * Author: Jacob Berkman <jacob@clusterfs.com> + */ + +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> +#include "tracefile.h" + +struct lc_watchdog { + spinlock_t lcw_lock; /* check or change lcw_list */ + int lcw_refcount; /* must hold lcw_pending_timers_lock */ + timer_list_t lcw_timer; /* kernel timer */ + struct list_head lcw_list; /* chain on pending list */ + cfs_time_t lcw_last_touched; /* last touched stamp */ + task_t *lcw_task; /* owner task */ + void (*lcw_callback)(pid_t, void *); + void *lcw_data; + + pid_t lcw_pid; + + enum { + LC_WATCHDOG_DISABLED, + LC_WATCHDOG_ENABLED, + LC_WATCHDOG_EXPIRED + } lcw_state; +}; + +#ifdef WITH_WATCHDOG +/* + * The dispatcher will complete lcw_start_completion when it starts, + * and lcw_stop_completion when it exits. + * Wake lcw_event_waitq to signal timer callback dispatches. + */ +static struct completion lcw_start_completion; +static struct completion lcw_stop_completion; +static wait_queue_head_t lcw_event_waitq; + +/* + * Set this and wake lcw_event_waitq to stop the dispatcher. + */ +enum { + LCW_FLAG_STOP = 0 +}; +static unsigned long lcw_flags = 0; + +/* + * Number of outstanding watchdogs. + * When it hits 1, we start the dispatcher. + * When it hits 0, we stop the dispatcher. + */ +static __u32 lcw_refcount = 0; +static DEFINE_MUTEX(lcw_refcount_mutex); + +/* + * List of timers that have fired that need their callbacks run by the + * dispatcher. + */ +/* BH lock! */ +static DEFINE_SPINLOCK(lcw_pending_timers_lock); +static struct list_head lcw_pending_timers = LIST_HEAD_INIT(lcw_pending_timers); + +/* Last time a watchdog expired */ +static cfs_time_t lcw_last_watchdog_time; +static int lcw_recent_watchdog_count; + +static void +lcw_dump(struct lc_watchdog *lcw) +{ + ENTRY; + rcu_read_lock(); + if (lcw->lcw_task == NULL) { + LCONSOLE_WARN("Process " LPPID " was not found in the task " + "list; watchdog callback may be incomplete\n", + (int)lcw->lcw_pid); + } else { + libcfs_debug_dumpstack(lcw->lcw_task); + } + + rcu_read_unlock(); + EXIT; +} + +static void lcw_cb(ulong_ptr_t data) +{ + struct lc_watchdog *lcw = (struct lc_watchdog *)data; + ENTRY; + + if (lcw->lcw_state != LC_WATCHDOG_ENABLED) { + EXIT; + return; + } + + lcw->lcw_state = LC_WATCHDOG_EXPIRED; + + spin_lock_bh(&lcw->lcw_lock); + LASSERT(list_empty(&lcw->lcw_list)); + + spin_lock_bh(&lcw_pending_timers_lock); + lcw->lcw_refcount++; /* +1 for pending list */ + list_add(&lcw->lcw_list, &lcw_pending_timers); + wake_up(&lcw_event_waitq); + + spin_unlock_bh(&lcw_pending_timers_lock); + spin_unlock_bh(&lcw->lcw_lock); + EXIT; +} + +static int is_watchdog_fired(void) +{ + int rc; + + if (test_bit(LCW_FLAG_STOP, &lcw_flags)) + return 1; + + spin_lock_bh(&lcw_pending_timers_lock); + rc = !list_empty(&lcw_pending_timers); + spin_unlock_bh(&lcw_pending_timers_lock); + return rc; +} + +static void lcw_dump_stack(struct lc_watchdog *lcw) +{ + cfs_time_t current_time; + cfs_duration_t delta_time; + struct timeval timediff; + + current_time = cfs_time_current(); + delta_time = cfs_time_sub(current_time, lcw->lcw_last_touched); + cfs_duration_usec(delta_time, &timediff); + + /* + * Check to see if we should throttle the watchdog timer to avoid + * too many dumps going to the console thus triggering an NMI. + */ + delta_time = cfs_duration_sec(cfs_time_sub(current_time, + lcw_last_watchdog_time)); + + if (delta_time < libcfs_watchdog_ratelimit && + lcw_recent_watchdog_count > 3) { + LCONSOLE_WARN("Service thread pid %u was inactive for " + "%lu.%.02lus. Watchdog stack traces are limited " + "to 3 per %d seconds, skipping this one.\n", + (int)lcw->lcw_pid, + timediff.tv_sec, + timediff.tv_usec / 10000, + libcfs_watchdog_ratelimit); + } else { + if (delta_time < libcfs_watchdog_ratelimit) { + lcw_recent_watchdog_count++; + } else { + memcpy(&lcw_last_watchdog_time, ¤t_time, + sizeof(current_time)); + lcw_recent_watchdog_count = 0; + } + + LCONSOLE_WARN("Service thread pid %u was inactive for " + "%lu.%.02lus. The thread might be hung, or it " + "might only be slow and will resume later. " + "Dumping the stack trace for debugging purposes:" + "\n", + (int)lcw->lcw_pid, + timediff.tv_sec, + timediff.tv_usec / 10000); + lcw_dump(lcw); + } +} + +static int lcw_dispatch_main(void *data) +{ + int rc = 0; + struct lc_watchdog *lcw; + LIST_HEAD (zombies); + + ENTRY; + + complete(&lcw_start_completion); + + while (1) { + int dumplog = 1; + + cfs_wait_event_interruptible(lcw_event_waitq, + is_watchdog_fired(), rc); + CDEBUG(D_INFO, "Watchdog got woken up...\n"); + if (test_bit(LCW_FLAG_STOP, &lcw_flags)) { + CDEBUG(D_INFO, "LCW_FLAG_STOP set, shutting down...\n"); + + spin_lock_bh(&lcw_pending_timers_lock); + rc = !list_empty(&lcw_pending_timers); + spin_unlock_bh(&lcw_pending_timers_lock); + if (rc) { + CERROR("pending timers list was not empty at " + "time of watchdog dispatch shutdown\n"); + } + break; + } + + spin_lock_bh(&lcw_pending_timers_lock); + while (!list_empty(&lcw_pending_timers)) { + int is_dumplog; + + lcw = list_entry(lcw_pending_timers.next, + struct lc_watchdog, lcw_list); + /* +1 ref for callback to make sure lwc wouldn't be + * deleted after releasing lcw_pending_timers_lock */ + lcw->lcw_refcount++; + spin_unlock_bh(&lcw_pending_timers_lock); + + /* lock ordering */ + spin_lock_bh(&lcw->lcw_lock); + spin_lock_bh(&lcw_pending_timers_lock); + + if (list_empty(&lcw->lcw_list)) { + /* already removed from pending list */ + lcw->lcw_refcount--; /* -1 ref for callback */ + if (lcw->lcw_refcount == 0) + list_add(&lcw->lcw_list, &zombies); + spin_unlock_bh(&lcw->lcw_lock); + /* still hold lcw_pending_timers_lock */ + continue; + } + + list_del_init(&lcw->lcw_list); + lcw->lcw_refcount--; /* -1 ref for pending list */ + + spin_unlock_bh(&lcw_pending_timers_lock); + spin_unlock_bh(&lcw->lcw_lock); + + CDEBUG(D_INFO, "found lcw for pid " LPPID "\n", + lcw->lcw_pid); + lcw_dump_stack(lcw); + + is_dumplog = lcw->lcw_callback == lc_watchdog_dumplog; + if (lcw->lcw_state != LC_WATCHDOG_DISABLED && + (dumplog || !is_dumplog)) { + lcw->lcw_callback(lcw->lcw_pid, lcw->lcw_data); + if (dumplog && is_dumplog) + dumplog = 0; + } + + spin_lock_bh(&lcw_pending_timers_lock); + lcw->lcw_refcount--; /* -1 ref for callback */ + if (lcw->lcw_refcount == 0) + list_add(&lcw->lcw_list, &zombies); + } + spin_unlock_bh(&lcw_pending_timers_lock); + + while (!list_empty(&zombies)) { + lcw = list_entry(lcw_pending_timers.next, + struct lc_watchdog, lcw_list); + list_del(&lcw->lcw_list); + LIBCFS_FREE(lcw, sizeof(*lcw)); + } + } + + complete(&lcw_stop_completion); + + RETURN(rc); +} + +static void lcw_dispatch_start(void) +{ + task_t *task; + + ENTRY; + LASSERT(lcw_refcount == 1); + + init_completion(&lcw_stop_completion); + init_completion(&lcw_start_completion); + init_waitqueue_head(&lcw_event_waitq); + + CDEBUG(D_INFO, "starting dispatch thread\n"); + task = kthread_run(lcw_dispatch_main, NULL, "lc_watchdogd"); + if (IS_ERR(task)) { + CERROR("error spawning watchdog dispatch thread: %ld\n", + PTR_ERR(task)); + EXIT; + return; + } + wait_for_completion(&lcw_start_completion); + CDEBUG(D_INFO, "watchdog dispatcher initialization complete.\n"); + + EXIT; +} + +static void lcw_dispatch_stop(void) +{ + ENTRY; + LASSERT(lcw_refcount == 0); + + CDEBUG(D_INFO, "trying to stop watchdog dispatcher.\n"); + + set_bit(LCW_FLAG_STOP, &lcw_flags); + wake_up(&lcw_event_waitq); + + wait_for_completion(&lcw_stop_completion); + + CDEBUG(D_INFO, "watchdog dispatcher has shut down.\n"); + + EXIT; +} + +struct lc_watchdog *lc_watchdog_add(int timeout, + void (*callback)(pid_t, void *), + void *data) +{ + struct lc_watchdog *lcw = NULL; + ENTRY; + + LIBCFS_ALLOC(lcw, sizeof(*lcw)); + if (lcw == NULL) { + CDEBUG(D_INFO, "Could not allocate new lc_watchdog\n"); + RETURN(ERR_PTR(-ENOMEM)); + } + + spin_lock_init(&lcw->lcw_lock); + lcw->lcw_refcount = 1; /* refcount for owner */ + lcw->lcw_task = current; + lcw->lcw_pid = current_pid(); + lcw->lcw_callback = (callback != NULL) ? callback : lc_watchdog_dumplog; + lcw->lcw_data = data; + lcw->lcw_state = LC_WATCHDOG_DISABLED; + + INIT_LIST_HEAD(&lcw->lcw_list); + cfs_timer_init(&lcw->lcw_timer, lcw_cb, lcw); + + mutex_lock(&lcw_refcount_mutex); + if (++lcw_refcount == 1) + lcw_dispatch_start(); + mutex_unlock(&lcw_refcount_mutex); + + /* Keep this working in case we enable them by default */ + if (lcw->lcw_state == LC_WATCHDOG_ENABLED) { + lcw->lcw_last_touched = cfs_time_current(); + cfs_timer_arm(&lcw->lcw_timer, cfs_time_seconds(timeout) + + cfs_time_current()); + } + + RETURN(lcw); +} +EXPORT_SYMBOL(lc_watchdog_add); + +static void lcw_update_time(struct lc_watchdog *lcw, const char *message) +{ + cfs_time_t newtime = cfs_time_current();; + + if (lcw->lcw_state == LC_WATCHDOG_EXPIRED) { + struct timeval timediff; + cfs_time_t delta_time = cfs_time_sub(newtime, + lcw->lcw_last_touched); + cfs_duration_usec(delta_time, &timediff); + + LCONSOLE_WARN("Service thread pid %u %s after %lu.%.02lus. " + "This indicates the system was overloaded (too " + "many service threads, or there were not enough " + "hardware resources).\n", + lcw->lcw_pid, + message, + timediff.tv_sec, + timediff.tv_usec / 10000); + } + lcw->lcw_last_touched = newtime; +} + +static void lc_watchdog_del_pending(struct lc_watchdog *lcw) +{ + spin_lock_bh(&lcw->lcw_lock); + if (unlikely(!list_empty(&lcw->lcw_list))) { + spin_lock_bh(&lcw_pending_timers_lock); + list_del_init(&lcw->lcw_list); + lcw->lcw_refcount--; /* -1 ref for pending list */ + spin_unlock_bh(&lcw_pending_timers_lock); + } + + spin_unlock_bh(&lcw->lcw_lock); +} + +void lc_watchdog_touch(struct lc_watchdog *lcw, int timeout) +{ + ENTRY; + LASSERT(lcw != NULL); + + lc_watchdog_del_pending(lcw); + + lcw_update_time(lcw, "resumed"); + lcw->lcw_state = LC_WATCHDOG_ENABLED; + + cfs_timer_arm(&lcw->lcw_timer, cfs_time_current() + + cfs_time_seconds(timeout)); + + EXIT; +} +EXPORT_SYMBOL(lc_watchdog_touch); + +void lc_watchdog_disable(struct lc_watchdog *lcw) +{ + ENTRY; + LASSERT(lcw != NULL); + + lc_watchdog_del_pending(lcw); + + lcw_update_time(lcw, "completed"); + lcw->lcw_state = LC_WATCHDOG_DISABLED; + + EXIT; +} +EXPORT_SYMBOL(lc_watchdog_disable); + +void lc_watchdog_delete(struct lc_watchdog *lcw) +{ + int dead; + + ENTRY; + LASSERT(lcw != NULL); + + cfs_timer_disarm(&lcw->lcw_timer); + + lcw_update_time(lcw, "stopped"); + + spin_lock_bh(&lcw->lcw_lock); + spin_lock_bh(&lcw_pending_timers_lock); + if (unlikely(!list_empty(&lcw->lcw_list))) { + list_del_init(&lcw->lcw_list); + lcw->lcw_refcount--; /* -1 ref for pending list */ + } + + lcw->lcw_refcount--; /* -1 ref for owner */ + dead = lcw->lcw_refcount == 0; + spin_unlock_bh(&lcw_pending_timers_lock); + spin_unlock_bh(&lcw->lcw_lock); + + if (dead) + LIBCFS_FREE(lcw, sizeof(*lcw)); + + mutex_lock(&lcw_refcount_mutex); + if (--lcw_refcount == 0) + lcw_dispatch_stop(); + mutex_unlock(&lcw_refcount_mutex); + + EXIT; +} +EXPORT_SYMBOL(lc_watchdog_delete); + +/* + * Provided watchdog handlers + */ + +void lc_watchdog_dumplog(pid_t pid, void *data) +{ + libcfs_debug_dumplog_internal((void *)((long_ptr_t)pid)); +} +EXPORT_SYMBOL(lc_watchdog_dumplog); + +#else /* !defined(WITH_WATCHDOG) */ + +struct lc_watchdog *lc_watchdog_add(int timeout, + void (*callback)(pid_t pid, void *), + void *data) +{ + static struct lc_watchdog watchdog; + return &watchdog; +} +EXPORT_SYMBOL(lc_watchdog_add); + +void lc_watchdog_touch(struct lc_watchdog *lcw, int timeout) +{ +} +EXPORT_SYMBOL(lc_watchdog_touch); + +void lc_watchdog_disable(struct lc_watchdog *lcw) +{ +} +EXPORT_SYMBOL(lc_watchdog_disable); + +void lc_watchdog_delete(struct lc_watchdog *lcw) +{ +} +EXPORT_SYMBOL(lc_watchdog_delete); + +#endif diff --git a/drivers/staging/lustre/lustre/libcfs/workitem.c b/drivers/staging/lustre/lustre/libcfs/workitem.c new file mode 100644 index 000000000000..b533666c1900 --- /dev/null +++ b/drivers/staging/lustre/lustre/libcfs/workitem.c @@ -0,0 +1,475 @@ +/* + * GPL HEADER START + * + * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. + * + * This program is free software; you can redistribute it and/or modify + * it under the terms of the GNU General Public License version 2 only, + * as published by the Free Software Foundation. + * + * This program is distributed in the hope that it will be useful, but + * WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + * General Public License version 2 for more details (a copy is included + * in the LICENSE file that accompanied this code). + * + * You should have received a copy of the GNU General Public License + * version 2 along with this program; If not, see + * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf + * + * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, + * CA 95054 USA or visit www.sun.com if you need additional information or + * have any questions. + * + * GPL HEADER END + */ +/* + * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. + * Use is subject to license terms. + * + * Copyright (c) 2011, 2012, Intel Corporation. + */ +/* + * This file is part of Lustre, http://www.lustre.org/ + * Lustre is a trademark of Sun Microsystems, Inc. + * + * libcfs/libcfs/workitem.c + * + * Author: Isaac Huang <isaac@clusterfs.com> + * Liang Zhen <zhen.liang@sun.com> + */ + +#define DEBUG_SUBSYSTEM S_LNET + +#include <linux/libcfs/libcfs.h> + +#define CFS_WS_NAME_LEN 16 + +typedef struct cfs_wi_sched { + struct list_head ws_list; /* chain on global list */ + /** serialised workitems */ + spinlock_t ws_lock; + /** where schedulers sleep */ + wait_queue_head_t ws_waitq; + /** concurrent workitems */ + struct list_head ws_runq; + /** rescheduled running-workitems, a workitem can be rescheduled + * while running in wi_action(), but we don't to execute it again + * unless it returns from wi_action(), so we put it on ws_rerunq + * while rescheduling, and move it to runq after it returns + * from wi_action() */ + struct list_head ws_rerunq; + /** CPT-table for this scheduler */ + struct cfs_cpt_table *ws_cptab; + /** CPT id for affinity */ + int ws_cpt; + /** number of scheduled workitems */ + int ws_nscheduled; + /** started scheduler thread, protected by cfs_wi_data::wi_glock */ + unsigned int ws_nthreads:30; + /** shutting down, protected by cfs_wi_data::wi_glock */ + unsigned int ws_stopping:1; + /** serialize starting thread, protected by cfs_wi_data::wi_glock */ + unsigned int ws_starting:1; + /** scheduler name */ + char ws_name[CFS_WS_NAME_LEN]; +} cfs_wi_sched_t; + +struct cfs_workitem_data { + /** serialize */ + spinlock_t wi_glock; + /** list of all schedulers */ + struct list_head wi_scheds; + /** WI module is initialized */ + int wi_init; + /** shutting down the whole WI module */ + int wi_stopping; +} cfs_wi_data; + +static inline void +cfs_wi_sched_lock(cfs_wi_sched_t *sched) +{ + spin_lock(&sched->ws_lock); +} + +static inline void +cfs_wi_sched_unlock(cfs_wi_sched_t *sched) +{ + spin_unlock(&sched->ws_lock); +} + +static inline int +cfs_wi_sched_cansleep(cfs_wi_sched_t *sched) +{ + cfs_wi_sched_lock(sched); + if (sched->ws_stopping) { + cfs_wi_sched_unlock(sched); + return 0; + } + + if (!list_empty(&sched->ws_runq)) { + cfs_wi_sched_unlock(sched); + return 0; + } + cfs_wi_sched_unlock(sched); + return 1; +} + + +/* XXX: + * 0. it only works when called from wi->wi_action. + * 1. when it returns no one shall try to schedule the workitem. + */ +void +cfs_wi_exit(struct cfs_wi_sched *sched, cfs_workitem_t *wi) +{ + LASSERT(!in_interrupt()); /* because we use plain spinlock */ + LASSERT(!sched->ws_stopping); + + cfs_wi_sched_lock(sched); + + LASSERT(wi->wi_running); + if (wi->wi_scheduled) { /* cancel pending schedules */ + LASSERT(!list_empty(&wi->wi_list)); + list_del_init(&wi->wi_list); + + LASSERT(sched->ws_nscheduled > 0); + sched->ws_nscheduled--; + } + + LASSERT(list_empty(&wi->wi_list)); + + wi->wi_scheduled = 1; /* LBUG future schedule attempts */ + cfs_wi_sched_unlock(sched); + + return; +} +EXPORT_SYMBOL(cfs_wi_exit); + +/** + * cancel schedule request of workitem \a wi + */ +int +cfs_wi_deschedule(struct cfs_wi_sched *sched, cfs_workitem_t *wi) +{ + int rc; + + LASSERT(!in_interrupt()); /* because we use plain spinlock */ + LASSERT(!sched->ws_stopping); + + /* + * return 0 if it's running already, otherwise return 1, which + * means the workitem will not be scheduled and will not have + * any race with wi_action. + */ + cfs_wi_sched_lock(sched); + + rc = !(wi->wi_running); + + if (wi->wi_scheduled) { /* cancel pending schedules */ + LASSERT(!list_empty(&wi->wi_list)); + list_del_init(&wi->wi_list); + + LASSERT(sched->ws_nscheduled > 0); + sched->ws_nscheduled--; + + wi->wi_scheduled = 0; + } + + LASSERT (list_empty(&wi->wi_list)); + + cfs_wi_sched_unlock(sched); + return rc; +} +EXPORT_SYMBOL(cfs_wi_deschedule); + +/* + * Workitem scheduled with (serial == 1) is strictly serialised not only with + * itself, but also with others scheduled this way. + * + * Now there's only one static serialised queue, but in the future more might + * be added, and even dynamic creation of serialised queues might be supported. + */ +void +cfs_wi_schedule(struct cfs_wi_sched *sched, cfs_workitem_t *wi) +{ + LASSERT(!in_interrupt()); /* because we use plain spinlock */ + LASSERT(!sched->ws_stopping); + + cfs_wi_sched_lock(sched); + + if (!wi->wi_scheduled) { + LASSERT (list_empty(&wi->wi_list)); + + wi->wi_scheduled = 1; + sched->ws_nscheduled++; + if (!wi->wi_running) { + list_add_tail(&wi->wi_list, &sched->ws_runq); + wake_up(&sched->ws_waitq); + } else { + list_add(&wi->wi_list, &sched->ws_rerunq); + } + } + + LASSERT (!list_empty(&wi->wi_list)); + cfs_wi_sched_unlock(sched); + return; +} +EXPORT_SYMBOL(cfs_wi_schedule); + + +static int +cfs_wi_scheduler (void *arg) +{ + struct cfs_wi_sched *sched = (cfs_wi_sched_t *)arg; + + cfs_block_allsigs(); + + /* CPT affinity scheduler? */ + if (sched->ws_cptab != NULL) + cfs_cpt_bind(sched->ws_cptab, sched->ws_cpt); + + spin_lock(&cfs_wi_data.wi_glock); + + LASSERT(sched->ws_starting == 1); + sched->ws_starting--; + sched->ws_nthreads++; + + spin_unlock(&cfs_wi_data.wi_glock); + + cfs_wi_sched_lock(sched); + + while (!sched->ws_stopping) { + int nloops = 0; + int rc; + cfs_workitem_t *wi; + + while (!list_empty(&sched->ws_runq) && + nloops < CFS_WI_RESCHED) { + wi = list_entry(sched->ws_runq.next, + cfs_workitem_t, wi_list); + LASSERT(wi->wi_scheduled && !wi->wi_running); + + list_del_init(&wi->wi_list); + + LASSERT(sched->ws_nscheduled > 0); + sched->ws_nscheduled--; + + wi->wi_running = 1; + wi->wi_scheduled = 0; + + + cfs_wi_sched_unlock(sched); + nloops++; + + rc = (*wi->wi_action) (wi); + + cfs_wi_sched_lock(sched); + if (rc != 0) /* WI should be dead, even be freed! */ + continue; + + wi->wi_running = 0; + if (list_empty(&wi->wi_list)) + continue; + + LASSERT(wi->wi_scheduled); + /* wi is rescheduled, should be on rerunq now, we + * move it to runq so it can run action now */ + list_move_tail(&wi->wi_list, &sched->ws_runq); + } + + if (!list_empty(&sched->ws_runq)) { + cfs_wi_sched_unlock(sched); + /* don't sleep because some workitems still + * expect me to come back soon */ + cond_resched(); + cfs_wi_sched_lock(sched); + continue; + } + + cfs_wi_sched_unlock(sched); + cfs_wait_event_interruptible_exclusive(sched->ws_waitq, + !cfs_wi_sched_cansleep(sched), rc); + cfs_wi_sched_lock(sched); + } + + cfs_wi_sched_unlock(sched); + + spin_lock(&cfs_wi_data.wi_glock); + sched->ws_nthreads--; + spin_unlock(&cfs_wi_data.wi_glock); + + return 0; +} + + +void +cfs_wi_sched_destroy(struct cfs_wi_sched *sched) +{ + int i; + + LASSERT(cfs_wi_data.wi_init); + LASSERT(!cfs_wi_data.wi_stopping); + + spin_lock(&cfs_wi_data.wi_glock); + if (sched->ws_stopping) { + CDEBUG(D_INFO, "%s is in progress of stopping\n", + sched->ws_name); + spin_unlock(&cfs_wi_data.wi_glock); + return; + } + + LASSERT(!list_empty(&sched->ws_list)); + sched->ws_stopping = 1; + + spin_unlock(&cfs_wi_data.wi_glock); + + i = 2; + wake_up_all(&sched->ws_waitq); + + spin_lock(&cfs_wi_data.wi_glock); + while (sched->ws_nthreads > 0) { + CDEBUG(IS_PO2(++i) ? D_WARNING : D_NET, + "waiting for %d threads of WI sched[%s] to terminate\n", + sched->ws_nthreads, sched->ws_name); + + spin_unlock(&cfs_wi_data.wi_glock); + cfs_pause(cfs_time_seconds(1) / 20); + spin_lock(&cfs_wi_data.wi_glock); + } + + list_del(&sched->ws_list); + + spin_unlock(&cfs_wi_data.wi_glock); + LASSERT(sched->ws_nscheduled == 0); + + LIBCFS_FREE(sched, sizeof(*sched)); +} +EXPORT_SYMBOL(cfs_wi_sched_destroy); + +int +cfs_wi_sched_create(char *name, struct cfs_cpt_table *cptab, + int cpt, int nthrs, struct cfs_wi_sched **sched_pp) +{ + struct cfs_wi_sched *sched; + int rc; + + LASSERT(cfs_wi_data.wi_init); + LASSERT(!cfs_wi_data.wi_stopping); + LASSERT(cptab == NULL || cpt == CFS_CPT_ANY || + (cpt >= 0 && cpt < cfs_cpt_number(cptab))); + + LIBCFS_ALLOC(sched, sizeof(*sched)); + if (sched == NULL) + return -ENOMEM; + + strncpy(sched->ws_name, name, CFS_WS_NAME_LEN); + sched->ws_cptab = cptab; + sched->ws_cpt = cpt; + + spin_lock_init(&sched->ws_lock); + init_waitqueue_head(&sched->ws_waitq); + INIT_LIST_HEAD(&sched->ws_runq); + INIT_LIST_HEAD(&sched->ws_rerunq); + INIT_LIST_HEAD(&sched->ws_list); + + rc = 0; + while (nthrs > 0) { + char name[16]; + task_t *task; + spin_lock(&cfs_wi_data.wi_glock); + while (sched->ws_starting > 0) { + spin_unlock(&cfs_wi_data.wi_glock); + schedule(); + spin_lock(&cfs_wi_data.wi_glock); + } + + sched->ws_starting++; + spin_unlock(&cfs_wi_data.wi_glock); + + if (sched->ws_cptab != NULL && sched->ws_cpt >= 0) { + snprintf(name, sizeof(name), "%s_%02d_%02d", + sched->ws_name, sched->ws_cpt, + sched->ws_nthreads); + } else { + snprintf(name, sizeof(name), "%s_%02d", + sched->ws_name, sched->ws_nthreads); + } + + task = kthread_run(cfs_wi_scheduler, sched, name); + if (!IS_ERR(task)) { + nthrs--; + continue; + } + rc = PTR_ERR(task); + + CERROR("Failed to create thread for WI scheduler %s: %d\n", + name, rc); + + spin_lock(&cfs_wi_data.wi_glock); + + /* make up for cfs_wi_sched_destroy */ + list_add(&sched->ws_list, &cfs_wi_data.wi_scheds); + sched->ws_starting--; + + spin_unlock(&cfs_wi_data.wi_glock); + + cfs_wi_sched_destroy(sched); + return rc; + } + spin_lock(&cfs_wi_data.wi_glock); + list_add(&sched->ws_list, &cfs_wi_data.wi_scheds); + spin_unlock(&cfs_wi_data.wi_glock); + + *sched_pp = sched; + return 0; +} +EXPORT_SYMBOL(cfs_wi_sched_create); + +int +cfs_wi_startup(void) +{ + memset(&cfs_wi_data, 0, sizeof(cfs_wi_data)); + + spin_lock_init(&cfs_wi_data.wi_glock); + INIT_LIST_HEAD(&cfs_wi_data.wi_scheds); + cfs_wi_data.wi_init = 1; + + return 0; +} + +void +cfs_wi_shutdown (void) +{ + struct cfs_wi_sched *sched; + + spin_lock(&cfs_wi_data.wi_glock); + cfs_wi_data.wi_stopping = 1; + spin_unlock(&cfs_wi_data.wi_glock); + + /* nobody should contend on this list */ + list_for_each_entry(sched, &cfs_wi_data.wi_scheds, ws_list) { + sched->ws_stopping = 1; + wake_up_all(&sched->ws_waitq); + } + + list_for_each_entry(sched, &cfs_wi_data.wi_scheds, ws_list) { + spin_lock(&cfs_wi_data.wi_glock); + + while (sched->ws_nthreads != 0) { + spin_unlock(&cfs_wi_data.wi_glock); + cfs_pause(cfs_time_seconds(1) / 20); + spin_lock(&cfs_wi_data.wi_glock); + } + spin_unlock(&cfs_wi_data.wi_glock); + } + while (!list_empty(&cfs_wi_data.wi_scheds)) { + sched = list_entry(cfs_wi_data.wi_scheds.next, + struct cfs_wi_sched, ws_list); + list_del(&sched->ws_list); + LIBCFS_FREE(sched, sizeof(*sched)); + } + + cfs_wi_data.wi_stopping = 0; + cfs_wi_data.wi_init = 0; +} |