From 7a0f475513fa573bc8e072021960313da32f0ee3 Mon Sep 17 00:00:00 2001 From: Klaus-Dieter Wacker Date: Fri, 10 Oct 2008 21:33:18 +0200 Subject: [S390] qdio enhanced SIGA (iqdio) support. Add support for z10 HiperSockets multiwrite SBALs on output queues. This is used on LPAR with EDDP enabled devices. Signed-off-by: Klaus-Dieter Wacker Signed-off-by: Martin Schwidefsky --- arch/s390/include/asm/qdio.h | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/s390/include/asm/qdio.h b/arch/s390/include/asm/qdio.h index 6813772171f2..4734c3f05354 100644 --- a/arch/s390/include/asm/qdio.h +++ b/arch/s390/include/asm/qdio.h @@ -299,7 +299,13 @@ struct qdio_ssqd_desc { u8 mbccnt; u16 qdioac2; u64 sch_token; - u64:64; + u8 mro; + u8 mri; + u8:8; + u8 sbalic; + u16:16; + u8:8; + u8 mmwc; } __attribute__ ((packed)); /* params are: ccw_device, qdio_error, queue_number, -- cgit v1.2.3-59-g8ed1b From d86730bb9597b02bff59a3a5a01c0094d71a265f Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Fri, 10 Oct 2008 21:33:19 +0200 Subject: [S390] s390: use sys_pause for 31bit pause entry point sys32_pause is a useless copy of the generic sys_pause. (and it's certainly not there for old sparc32 binaries..) Signed-off-by: Christoph Hellwig Signed-off-by: Martin Schwidefsky --- arch/s390/kernel/compat_linux.c | 8 -------- arch/s390/kernel/compat_linux.h | 1 - arch/s390/kernel/compat_wrapper.S | 2 -- arch/s390/kernel/syscalls.S | 2 +- 4 files changed, 1 insertion(+), 12 deletions(-) (limited to 'arch') diff --git a/arch/s390/kernel/compat_linux.c b/arch/s390/kernel/compat_linux.c index d7f22226fc4e..98e246dc0233 100644 --- a/arch/s390/kernel/compat_linux.c +++ b/arch/s390/kernel/compat_linux.c @@ -608,14 +608,6 @@ asmlinkage long sys32_settimeofday(struct compat_timeval __user *tv, struct time return do_sys_settimeofday(tv ? &kts : NULL, tz ? &ktz : NULL); } -/* These are here just in case some old sparc32 binary calls it. */ -asmlinkage long sys32_pause(void) -{ - current->state = TASK_INTERRUPTIBLE; - schedule(); - return -ERESTARTNOHAND; -} - asmlinkage long sys32_pread64(unsigned int fd, char __user *ubuf, size_t count, u32 poshi, u32 poslo) { diff --git a/arch/s390/kernel/compat_linux.h b/arch/s390/kernel/compat_linux.h index 20723a062017..05f8516366ab 100644 --- a/arch/s390/kernel/compat_linux.h +++ b/arch/s390/kernel/compat_linux.h @@ -206,7 +206,6 @@ long sys32_gettimeofday(struct compat_timeval __user *tv, struct timezone __user *tz); long sys32_settimeofday(struct compat_timeval __user *tv, struct timezone __user *tz); -long sys32_pause(void); long sys32_pread64(unsigned int fd, char __user *ubuf, size_t count, u32 poshi, u32 poslo); long sys32_pwrite64(unsigned int fd, const char __user *ubuf, diff --git a/arch/s390/kernel/compat_wrapper.S b/arch/s390/kernel/compat_wrapper.S index 328a20e880b5..ee51ca9e23b5 100644 --- a/arch/s390/kernel/compat_wrapper.S +++ b/arch/s390/kernel/compat_wrapper.S @@ -128,8 +128,6 @@ sys32_alarm_wrapper: llgfr %r2,%r2 # unsigned int jg sys_alarm # branch to system call -#sys32_pause_wrapper # void - .globl compat_sys_utime_wrapper compat_sys_utime_wrapper: llgtr %r2,%r2 # char * diff --git a/arch/s390/kernel/syscalls.S b/arch/s390/kernel/syscalls.S index c66d35e55142..3ae303914b42 100644 --- a/arch/s390/kernel/syscalls.S +++ b/arch/s390/kernel/syscalls.S @@ -37,7 +37,7 @@ SYSCALL(sys_stime,sys_ni_syscall,sys32_stime_wrapper) /* 25 old stime syscall * SYSCALL(sys_ptrace,sys_ptrace,sys32_ptrace_wrapper) SYSCALL(sys_alarm,sys_alarm,sys32_alarm_wrapper) NI_SYSCALL /* old fstat syscall */ -SYSCALL(sys_pause,sys_pause,sys32_pause) +SYSCALL(sys_pause,sys_pause,sys_pause) SYSCALL(sys_utime,sys_utime,compat_sys_utime_wrapper) /* 30 */ NI_SYSCALL /* old stty syscall */ NI_SYSCALL /* old gtty syscall */ -- cgit v1.2.3-59-g8ed1b From 753c4dd6a2fa2af81f5d809d610d29f2d9dd9bc1 Mon Sep 17 00:00:00 2001 From: Martin Schwidefsky Date: Fri, 10 Oct 2008 21:33:20 +0200 Subject: [S390] ptrace changes * System call parameter and result access functions * Add tracehook calls * Split syscall_trace into two functions do_syscall_trace_enter and do_syscall_trace_exit Signed-off-by: Martin Schwidefsky --- arch/s390/Kconfig | 1 + arch/s390/include/asm/ptrace.h | 1 + arch/s390/include/asm/syscall.h | 80 +++++++++++++++++++++++++++++++++++++ arch/s390/include/asm/thread_info.h | 2 + arch/s390/kernel/entry.S | 50 ++++++++++++++++++----- arch/s390/kernel/entry64.S | 42 ++++++++++++++----- arch/s390/kernel/ptrace.c | 61 +++++++++++++++------------- arch/s390/kernel/signal.c | 13 ++++++ 8 files changed, 202 insertions(+), 48 deletions(-) create mode 100644 arch/s390/include/asm/syscall.h (limited to 'arch') diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig index 8d41908e2513..4c03049e7db9 100644 --- a/arch/s390/Kconfig +++ b/arch/s390/Kconfig @@ -74,6 +74,7 @@ config S390 select HAVE_KPROBES select HAVE_KRETPROBES select HAVE_KVM if 64BIT + select HAVE_ARCH_TRACEHOOK source "init/Kconfig" diff --git a/arch/s390/include/asm/ptrace.h b/arch/s390/include/asm/ptrace.h index af2c9ac28a07..a7226f8143fb 100644 --- a/arch/s390/include/asm/ptrace.h +++ b/arch/s390/include/asm/ptrace.h @@ -490,6 +490,7 @@ extern void user_disable_single_step(struct task_struct *); #define user_mode(regs) (((regs)->psw.mask & PSW_MASK_PSTATE) != 0) #define instruction_pointer(regs) ((regs)->psw.addr & PSW_ADDR_INSN) +#define user_stack_pointer(regs)((regs)->gprs[15]) #define regs_return_value(regs)((regs)->gprs[2]) #define profile_pc(regs) instruction_pointer(regs) extern void show_regs(struct pt_regs * regs); diff --git a/arch/s390/include/asm/syscall.h b/arch/s390/include/asm/syscall.h new file mode 100644 index 000000000000..6e623971fbb9 --- /dev/null +++ b/arch/s390/include/asm/syscall.h @@ -0,0 +1,80 @@ +/* + * Access to user system call parameters and results + * + * Copyright IBM Corp. 2008 + * Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com) + * + * 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. + */ + +#ifndef _ASM_SYSCALL_H +#define _ASM_SYSCALL_H 1 + +#include + +static inline long syscall_get_nr(struct task_struct *task, + struct pt_regs *regs) +{ + if (regs->trap != __LC_SVC_OLD_PSW) + return -1; + return regs->gprs[2]; +} + +static inline void syscall_rollback(struct task_struct *task, + struct pt_regs *regs) +{ + regs->gprs[2] = regs->orig_gpr2; +} + +static inline long syscall_get_error(struct task_struct *task, + struct pt_regs *regs) +{ + return (regs->gprs[2] >= -4096UL) ? -regs->gprs[2] : 0; +} + +static inline long syscall_get_return_value(struct task_struct *task, + struct pt_regs *regs) +{ + return regs->gprs[2]; +} + +static inline void syscall_set_return_value(struct task_struct *task, + struct pt_regs *regs, + int error, long val) +{ + regs->gprs[2] = error ? -error : val; +} + +static inline void syscall_get_arguments(struct task_struct *task, + struct pt_regs *regs, + unsigned int i, unsigned int n, + unsigned long *args) +{ + BUG_ON(i + n > 6); +#ifdef CONFIG_COMPAT + if (test_tsk_thread_flag(task, TIF_31BIT)) { + if (i + n == 6) + args[--n] = (u32) regs->args[0]; + while (n-- > 0) + args[n] = (u32) regs->gprs[2 + i + n]; + } +#endif + if (i + n == 6) + args[--n] = regs->args[0]; + memcpy(args, ®s->gprs[2 + i], n * sizeof(args[0])); +} + +static inline void syscall_set_arguments(struct task_struct *task, + struct pt_regs *regs, + unsigned int i, unsigned int n, + const unsigned long *args) +{ + BUG_ON(i + n > 6); + if (i + n == 6) + regs->args[0] = args[--n]; + memcpy(®s->gprs[2 + i], args, n * sizeof(args[0])); +} + +#endif /* _ASM_SYSCALL_H */ diff --git a/arch/s390/include/asm/thread_info.h b/arch/s390/include/asm/thread_info.h index 91a8f93ad355..ea40a9d690fc 100644 --- a/arch/s390/include/asm/thread_info.h +++ b/arch/s390/include/asm/thread_info.h @@ -86,6 +86,7 @@ static inline struct thread_info *current_thread_info(void) * thread information flags bit numbers */ #define TIF_SYSCALL_TRACE 0 /* syscall trace active */ +#define TIF_NOTIFY_RESUME 1 /* callback before returning to user */ #define TIF_SIGPENDING 2 /* signal pending */ #define TIF_NEED_RESCHED 3 /* rescheduling necessary */ #define TIF_RESTART_SVC 4 /* restart svc with new svc number */ @@ -100,6 +101,7 @@ static inline struct thread_info *current_thread_info(void) #define TIF_RESTORE_SIGMASK 20 /* restore signal mask in do_signal() */ #define _TIF_SYSCALL_TRACE (1< #include #include +#include #include #include @@ -639,40 +640,44 @@ long compat_arch_ptrace(struct task_struct *child, compat_long_t request, } #endif -asmlinkage void -syscall_trace(struct pt_regs *regs, int entryexit) +asmlinkage long do_syscall_trace_enter(struct pt_regs *regs) { - if (unlikely(current->audit_context) && entryexit) - audit_syscall_exit(AUDITSC_RESULT(regs->gprs[2]), regs->gprs[2]); - - if (!test_thread_flag(TIF_SYSCALL_TRACE)) - goto out; - if (!(current->ptrace & PT_PTRACED)) - goto out; - ptrace_notify(SIGTRAP | ((current->ptrace & PT_TRACESYSGOOD) - ? 0x80 : 0)); + long ret; /* - * If the debuffer has set an invalid system call number, - * we prepare to skip the system call restart handling. + * The sysc_tracesys code in entry.S stored the system + * call number to gprs[2]. */ - if (!entryexit && regs->gprs[2] >= NR_syscalls) + ret = regs->gprs[2]; + if (test_thread_flag(TIF_SYSCALL_TRACE) && + (tracehook_report_syscall_entry(regs) || + regs->gprs[2] >= NR_syscalls)) { + /* + * Tracing decided this syscall should not happen or the + * debugger stored an invalid system call number. Skip + * the system call and the system call restart handling. + */ regs->trap = -1; - - /* - * this isn't the same as continuing with a signal, but it will do - * for normal use. strace only continues with a signal if the - * stopping signal is not SIGTRAP. -brl - */ - if (current->exit_code) { - send_sig(current->exit_code, current, 1); - current->exit_code = 0; + ret = -1; } - out: - if (unlikely(current->audit_context) && !entryexit) - audit_syscall_entry(test_thread_flag(TIF_31BIT)?AUDIT_ARCH_S390:AUDIT_ARCH_S390X, - regs->gprs[2], regs->orig_gpr2, regs->gprs[3], - regs->gprs[4], regs->gprs[5]); + + if (unlikely(current->audit_context)) + audit_syscall_entry(test_thread_flag(TIF_31BIT) ? + AUDIT_ARCH_S390 : AUDIT_ARCH_S390X, + regs->gprs[2], regs->orig_gpr2, + regs->gprs[3], regs->gprs[4], + regs->gprs[5]); + return ret; +} + +asmlinkage void do_syscall_trace_exit(struct pt_regs *regs) +{ + if (unlikely(current->audit_context)) + audit_syscall_exit(AUDITSC_RESULT(regs->gprs[2]), + regs->gprs[2]); + + if (test_thread_flag(TIF_SYSCALL_TRACE)) + tracehook_report_syscall_exit(regs, 0); } /* diff --git a/arch/s390/kernel/signal.c b/arch/s390/kernel/signal.c index b97682040215..4f7fc3059a8e 100644 --- a/arch/s390/kernel/signal.c +++ b/arch/s390/kernel/signal.c @@ -24,6 +24,7 @@ #include #include #include +#include #include #include #include @@ -507,6 +508,12 @@ void do_signal(struct pt_regs *regs) */ if (current->thread.per_info.single_step) set_thread_flag(TIF_SINGLE_STEP); + + /* + * Let tracing know that we've done the handler setup. + */ + tracehook_signal_handler(signr, &info, &ka, regs, + test_thread_flag(TIF_SINGLE_STEP)); } return; } @@ -526,3 +533,9 @@ void do_signal(struct pt_regs *regs) set_thread_flag(TIF_RESTART_SVC); } } + +void do_notify_resume(struct pt_regs *regs) +{ + clear_thread_flag(TIF_NOTIFY_RESUME); + tracehook_notify_resume(regs); +} -- cgit v1.2.3-59-g8ed1b From b2300b9efe1b8174833e17f37e975c9da00c388a Mon Sep 17 00:00:00 2001 From: Hongjie Yang Date: Fri, 10 Oct 2008 21:33:21 +0200 Subject: [S390] dcssblk: add >2G DCSSs support and stacked contiguous DCSSs support. The DCSS block device driver is modified to add >2G DCSSs support and allow a DCSS block device to map to a set of contiguous DCSSs. The extmem code is also modified to use new Diagnose x'64' subcodes for >2G DCSSs. Signed-off-by: Hongjie Yang Signed-off-by: Martin Schwidefsky --- arch/s390/mm/extmem.c | 251 +++++++++++++++++---- drivers/s390/block/dcssblk.c | 515 ++++++++++++++++++++++++++++++++----------- 2 files changed, 596 insertions(+), 170 deletions(-) (limited to 'arch') diff --git a/arch/s390/mm/extmem.c b/arch/s390/mm/extmem.c index f231f5ec74b6..580fc64cc735 100644 --- a/arch/s390/mm/extmem.c +++ b/arch/s390/mm/extmem.c @@ -43,20 +43,40 @@ #define DCSS_FINDSEG 0x0c #define DCSS_LOADNOLY 0x10 #define DCSS_SEGEXT 0x18 +#define DCSS_LOADSHRX 0x20 +#define DCSS_LOADNSRX 0x24 +#define DCSS_FINDSEGX 0x2c +#define DCSS_SEGEXTX 0x38 #define DCSS_FINDSEGA 0x0c struct qrange { - unsigned int start; // 3byte start address, 1 byte type - unsigned int end; // 3byte end address, 1 byte reserved + unsigned long start; /* last byte type */ + unsigned long end; /* last byte reserved */ }; struct qout64 { + unsigned long segstart; + unsigned long segend; + int segcnt; + int segrcnt; + struct qrange range[6]; +}; + +#ifdef CONFIG_64BIT +struct qrange_old { + unsigned int start; /* last byte type */ + unsigned int end; /* last byte reserved */ +}; + +/* output area format for the Diag x'64' old subcode x'18' */ +struct qout64_old { int segstart; int segend; int segcnt; int segrcnt; - struct qrange range[6]; + struct qrange_old range[6]; }; +#endif struct qin64 { char qopcode; @@ -86,6 +106,55 @@ static DEFINE_MUTEX(dcss_lock); static LIST_HEAD(dcss_list); static char *segtype_string[] = { "SW", "EW", "SR", "ER", "SN", "EN", "SC", "EW/EN-MIXED" }; +static int loadshr_scode, loadnsr_scode, findseg_scode; +static int segext_scode, purgeseg_scode; +static int scode_set; + +/* set correct Diag x'64' subcodes. */ +static int +dcss_set_subcodes(void) +{ +#ifdef CONFIG_64BIT + char *name = kmalloc(8 * sizeof(char), GFP_DMA); + unsigned long rx, ry; + int rc; + + if (name == NULL) + return -ENOMEM; + + rx = (unsigned long) name; + ry = DCSS_FINDSEGX; + + strcpy(name, "dummy"); + asm volatile( + " diag %0,%1,0x64\n" + "0: ipm %2\n" + " srl %2,28\n" + " j 2f\n" + "1: la %2,3\n" + "2:\n" + EX_TABLE(0b, 1b) + : "+d" (rx), "+d" (ry), "=d" (rc) : : "cc"); + + kfree(name); + /* Diag x'64' new subcodes are supported, set to new subcodes */ + if (rc != 3) { + loadshr_scode = DCSS_LOADSHRX; + loadnsr_scode = DCSS_LOADNSRX; + purgeseg_scode = DCSS_PURGESEG; + findseg_scode = DCSS_FINDSEGX; + segext_scode = DCSS_SEGEXTX; + return 0; + } +#endif + /* Diag x'64' new subcodes are not supported, set to old subcodes */ + loadshr_scode = DCSS_LOADNOLY; + loadnsr_scode = DCSS_LOADNSR; + purgeseg_scode = DCSS_PURGESEG; + findseg_scode = DCSS_FINDSEG; + segext_scode = DCSS_SEGEXT; + return 0; +} /* * Create the 8 bytes, ebcdic VM segment name from @@ -135,25 +204,45 @@ segment_by_name (char *name) * Perform a function on a dcss segment. */ static inline int -dcss_diag (__u8 func, void *parameter, +dcss_diag(int *func, void *parameter, unsigned long *ret1, unsigned long *ret2) { unsigned long rx, ry; int rc; + if (scode_set == 0) { + rc = dcss_set_subcodes(); + if (rc < 0) + return rc; + scode_set = 1; + } rx = (unsigned long) parameter; - ry = (unsigned long) func; - asm volatile( + ry = (unsigned long) *func; + #ifdef CONFIG_64BIT - " sam31\n" - " diag %0,%1,0x64\n" - " sam64\n" + /* 64-bit Diag x'64' new subcode, keep in 64-bit addressing mode */ + if (*func > DCSS_SEGEXT) + asm volatile( + " diag %0,%1,0x64\n" + " ipm %2\n" + " srl %2,28\n" + : "+d" (rx), "+d" (ry), "=d" (rc) : : "cc"); + /* 31-bit Diag x'64' old subcode, switch to 31-bit addressing mode */ + else + asm volatile( + " sam31\n" + " diag %0,%1,0x64\n" + " sam64\n" + " ipm %2\n" + " srl %2,28\n" + : "+d" (rx), "+d" (ry), "=d" (rc) : : "cc"); #else + asm volatile( " diag %0,%1,0x64\n" -#endif " ipm %2\n" " srl %2,28\n" : "+d" (rx), "+d" (ry), "=d" (rc) : : "cc"); +#endif *ret1 = rx; *ret2 = ry; return rc; @@ -190,14 +279,45 @@ query_segment_type (struct dcss_segment *seg) qin->qoutlen = sizeof(struct qout64); memcpy (qin->qname, seg->dcss_name, 8); - diag_cc = dcss_diag (DCSS_SEGEXT, qin, &dummy, &vmrc); + diag_cc = dcss_diag(&segext_scode, qin, &dummy, &vmrc); + if (diag_cc < 0) { + rc = diag_cc; + goto out_free; + } if (diag_cc > 1) { PRINT_WARN ("segment_type: diag returned error %ld\n", vmrc); rc = dcss_diag_translate_rc (vmrc); goto out_free; } +#ifdef CONFIG_64BIT + /* Only old format of output area of Diagnose x'64' is supported, + copy data for the new format. */ + if (segext_scode == DCSS_SEGEXT) { + struct qout64_old *qout_old; + qout_old = kzalloc(sizeof(struct qout64_old), GFP_DMA); + if (qout_old == NULL) { + rc = -ENOMEM; + goto out_free; + } + memcpy(qout_old, qout, sizeof(struct qout64_old)); + qout->segstart = (unsigned long) qout_old->segstart; + qout->segend = (unsigned long) qout_old->segend; + qout->segcnt = qout_old->segcnt; + qout->segrcnt = qout_old->segrcnt; + + if (qout->segcnt > 6) + qout->segrcnt = 6; + for (i = 0; i < qout->segrcnt; i++) { + qout->range[i].start = + (unsigned long) qout_old->range[i].start; + qout->range[i].end = + (unsigned long) qout_old->range[i].end; + } + kfree(qout_old); + } +#endif if (qout->segcnt > 6) { rc = -ENOTSUPP; goto out_free; @@ -268,6 +388,30 @@ segment_type (char* name) return seg.vm_segtype; } +/* + * check if segment collides with other segments that are currently loaded + * returns 1 if this is the case, 0 if no collision was found + */ +static int +segment_overlaps_others (struct dcss_segment *seg) +{ + struct list_head *l; + struct dcss_segment *tmp; + + BUG_ON(!mutex_is_locked(&dcss_lock)); + list_for_each(l, &dcss_list) { + tmp = list_entry(l, struct dcss_segment, list); + if ((tmp->start_addr >> 20) > (seg->end >> 20)) + continue; + if ((tmp->end >> 20) < (seg->start_addr >> 20)) + continue; + if (seg == tmp) + continue; + return 1; + } + return 0; +} + /* * real segment loading function, called from segment_load */ @@ -276,7 +420,8 @@ __segment_load (char *name, int do_nonshared, unsigned long *addr, unsigned long { struct dcss_segment *seg = kmalloc(sizeof(struct dcss_segment), GFP_DMA); - int dcss_command, rc, diag_cc; + int rc, diag_cc; + unsigned long start_addr, end_addr, dummy; if (seg == NULL) { rc = -ENOMEM; @@ -287,6 +432,13 @@ __segment_load (char *name, int do_nonshared, unsigned long *addr, unsigned long if (rc < 0) goto out_free; + if (loadshr_scode == DCSS_LOADSHRX) { + if (segment_overlaps_others(seg)) { + rc = -EBUSY; + goto out_free; + } + } + rc = vmem_add_mapping(seg->start_addr, seg->end - seg->start_addr + 1); if (rc) @@ -316,20 +468,28 @@ __segment_load (char *name, int do_nonshared, unsigned long *addr, unsigned long } if (do_nonshared) - dcss_command = DCSS_LOADNSR; + diag_cc = dcss_diag(&loadnsr_scode, seg->dcss_name, + &start_addr, &end_addr); else - dcss_command = DCSS_LOADNOLY; - - diag_cc = dcss_diag(dcss_command, seg->dcss_name, - &seg->start_addr, &seg->end); + diag_cc = dcss_diag(&loadshr_scode, seg->dcss_name, + &start_addr, &end_addr); + if (diag_cc < 0) { + dcss_diag(&purgeseg_scode, seg->dcss_name, + &dummy, &dummy); + rc = diag_cc; + goto out_resource; + } if (diag_cc > 1) { PRINT_WARN ("segment_load: could not load segment %s - " - "diag returned error (%ld)\n",name,seg->end); - rc = dcss_diag_translate_rc (seg->end); - dcss_diag(DCSS_PURGESEG, seg->dcss_name, - &seg->start_addr, &seg->end); + "diag returned error (%ld)\n", + name, end_addr); + rc = dcss_diag_translate_rc(end_addr); + dcss_diag(&purgeseg_scode, seg->dcss_name, + &dummy, &dummy); goto out_resource; } + seg->start_addr = start_addr; + seg->end = end_addr; seg->do_nonshared = do_nonshared; atomic_set(&seg->ref_count, 1); list_add(&seg->list, &dcss_list); @@ -423,8 +583,8 @@ int segment_modify_shared (char *name, int do_nonshared) { struct dcss_segment *seg; - unsigned long dummy; - int dcss_command, rc, diag_cc; + unsigned long start_addr, end_addr, dummy; + int rc, diag_cc; mutex_lock(&dcss_lock); seg = segment_by_name (name); @@ -445,38 +605,51 @@ segment_modify_shared (char *name, int do_nonshared) goto out_unlock; } release_resource(seg->res); - if (do_nonshared) { - dcss_command = DCSS_LOADNSR; + if (do_nonshared) seg->res->flags &= ~IORESOURCE_READONLY; - } else { - dcss_command = DCSS_LOADNOLY; + else if (seg->vm_segtype == SEG_TYPE_SR || seg->vm_segtype == SEG_TYPE_ER) seg->res->flags |= IORESOURCE_READONLY; - } + if (request_resource(&iomem_resource, seg->res)) { PRINT_WARN("segment_modify_shared: could not reload segment %s" " - overlapping resources\n", name); rc = -EBUSY; kfree(seg->res); - goto out_del; + goto out_del_mem; + } + + dcss_diag(&purgeseg_scode, seg->dcss_name, &dummy, &dummy); + if (do_nonshared) + diag_cc = dcss_diag(&loadnsr_scode, seg->dcss_name, + &start_addr, &end_addr); + else + diag_cc = dcss_diag(&loadshr_scode, seg->dcss_name, + &start_addr, &end_addr); + if (diag_cc < 0) { + rc = diag_cc; + goto out_del_res; } - dcss_diag(DCSS_PURGESEG, seg->dcss_name, &dummy, &dummy); - diag_cc = dcss_diag(dcss_command, seg->dcss_name, - &seg->start_addr, &seg->end); if (diag_cc > 1) { PRINT_WARN ("segment_modify_shared: could not reload segment %s" - " - diag returned error (%ld)\n",name,seg->end); - rc = dcss_diag_translate_rc (seg->end); - goto out_del; + " - diag returned error (%ld)\n", + name, end_addr); + rc = dcss_diag_translate_rc(end_addr); + goto out_del_res; } + seg->start_addr = start_addr; + seg->end = end_addr; seg->do_nonshared = do_nonshared; rc = 0; goto out_unlock; - out_del: + out_del_res: + release_resource(seg->res); + kfree(seg->res); + out_del_mem: vmem_remove_mapping(seg->start_addr, seg->end - seg->start_addr + 1); list_del(&seg->list); - dcss_diag(DCSS_PURGESEG, seg->dcss_name, &dummy, &dummy); + dcss_diag(&purgeseg_scode, seg->dcss_name, &dummy, &dummy); kfree(seg); out_unlock: mutex_unlock(&dcss_lock); @@ -510,7 +683,7 @@ segment_unload(char *name) kfree(seg->res); vmem_remove_mapping(seg->start_addr, seg->end - seg->start_addr + 1); list_del(&seg->list); - dcss_diag(DCSS_PURGESEG, seg->dcss_name, &dummy, &dummy); + dcss_diag(&purgeseg_scode, seg->dcss_name, &dummy, &dummy); kfree(seg); out_unlock: mutex_unlock(&dcss_lock); @@ -545,7 +718,7 @@ segment_save(char *name) endpfn = (seg->end) >> PAGE_SHIFT; sprintf(cmd1, "DEFSEG %s", name); for (i=0; isegcnt; i++) { - sprintf(cmd1+strlen(cmd1), " %X-%X %s", + sprintf(cmd1+strlen(cmd1), " %lX-%lX %s", seg->range[i].start >> PAGE_SHIFT, seg->range[i].end >> PAGE_SHIFT, segtype_string[seg->range[i].start & 0xff]); diff --git a/drivers/s390/block/dcssblk.c b/drivers/s390/block/dcssblk.c index ea4272c8c677..a7ff167d5b81 100644 --- a/drivers/s390/block/dcssblk.c +++ b/drivers/s390/block/dcssblk.c @@ -31,7 +31,6 @@ #define PRINT_WARN(x...) printk(KERN_WARNING DCSSBLK_NAME " warning: " x) #define PRINT_ERR(x...) printk(KERN_ERR DCSSBLK_NAME " error: " x) - static int dcssblk_open(struct inode *inode, struct file *filp); static int dcssblk_release(struct inode *inode, struct file *filp); static int dcssblk_make_request(struct request_queue *q, struct bio *bio); @@ -48,6 +47,30 @@ static struct block_device_operations dcssblk_devops = { .direct_access = dcssblk_direct_access, }; +struct dcssblk_dev_info { + struct list_head lh; + struct device dev; + char segment_name[BUS_ID_SIZE]; + atomic_t use_count; + struct gendisk *gd; + unsigned long start; + unsigned long end; + int segment_type; + unsigned char save_pending; + unsigned char is_shared; + struct request_queue *dcssblk_queue; + int num_of_segments; + struct list_head seg_list; +}; + +struct segment_info { + struct list_head lh; + char segment_name[BUS_ID_SIZE]; + unsigned long start; + unsigned long end; + int segment_type; +}; + static ssize_t dcssblk_add_store(struct device * dev, struct device_attribute *attr, const char * buf, size_t count); static ssize_t dcssblk_remove_store(struct device * dev, struct device_attribute *attr, const char * buf, @@ -58,30 +81,20 @@ static ssize_t dcssblk_save_show(struct device *dev, struct device_attribute *at static ssize_t dcssblk_shared_store(struct device * dev, struct device_attribute *attr, const char * buf, size_t count); static ssize_t dcssblk_shared_show(struct device *dev, struct device_attribute *attr, char *buf); +static ssize_t dcssblk_seglist_show(struct device *dev, + struct device_attribute *attr, + char *buf); static DEVICE_ATTR(add, S_IWUSR, NULL, dcssblk_add_store); static DEVICE_ATTR(remove, S_IWUSR, NULL, dcssblk_remove_store); -static DEVICE_ATTR(save, S_IWUSR | S_IRUGO, dcssblk_save_show, +static DEVICE_ATTR(save, S_IWUSR | S_IRUSR, dcssblk_save_show, dcssblk_save_store); -static DEVICE_ATTR(shared, S_IWUSR | S_IRUGO, dcssblk_shared_show, +static DEVICE_ATTR(shared, S_IWUSR | S_IRUSR, dcssblk_shared_show, dcssblk_shared_store); +static DEVICE_ATTR(seglist, S_IRUSR, dcssblk_seglist_show, NULL); static struct device *dcssblk_root_dev; -struct dcssblk_dev_info { - struct list_head lh; - struct device dev; - char segment_name[BUS_ID_SIZE]; - atomic_t use_count; - struct gendisk *gd; - unsigned long start; - unsigned long end; - int segment_type; - unsigned char save_pending; - unsigned char is_shared; - struct request_queue *dcssblk_queue; -}; - static LIST_HEAD(dcssblk_devices); static struct rw_semaphore dcssblk_devices_sem; @@ -91,8 +104,15 @@ static struct rw_semaphore dcssblk_devices_sem; static void dcssblk_release_segment(struct device *dev) { - PRINT_DEBUG("segment release fn called for %s\n", dev_name(dev)); - kfree(container_of(dev, struct dcssblk_dev_info, dev)); + struct dcssblk_dev_info *dev_info; + struct segment_info *entry, *temp; + + dev_info = container_of(dev, struct dcssblk_dev_info, dev); + list_for_each_entry_safe(entry, temp, &dev_info->seg_list, lh) { + list_del(&entry->lh); + kfree(entry); + } + kfree(dev_info); module_put(THIS_MODULE); } @@ -142,6 +162,169 @@ dcssblk_get_device_by_name(char *name) return NULL; } +/* + * get the struct segment_info from seg_list + * for the given name. + * down_read(&dcssblk_devices_sem) must be held. + */ +static struct segment_info * +dcssblk_get_segment_by_name(char *name) +{ + struct dcssblk_dev_info *dev_info; + struct segment_info *entry; + + list_for_each_entry(dev_info, &dcssblk_devices, lh) { + list_for_each_entry(entry, &dev_info->seg_list, lh) { + if (!strcmp(name, entry->segment_name)) + return entry; + } + } + return NULL; +} + +/* + * get the highest address of the multi-segment block. + */ +static unsigned long +dcssblk_find_highest_addr(struct dcssblk_dev_info *dev_info) +{ + unsigned long highest_addr; + struct segment_info *entry; + + highest_addr = 0; + list_for_each_entry(entry, &dev_info->seg_list, lh) { + if (highest_addr < entry->end) + highest_addr = entry->end; + } + return highest_addr; +} + +/* + * get the lowest address of the multi-segment block. + */ +static unsigned long +dcssblk_find_lowest_addr(struct dcssblk_dev_info *dev_info) +{ + int set_first; + unsigned long lowest_addr; + struct segment_info *entry; + + set_first = 0; + lowest_addr = 0; + list_for_each_entry(entry, &dev_info->seg_list, lh) { + if (set_first == 0) { + lowest_addr = entry->start; + set_first = 1; + } else { + if (lowest_addr > entry->start) + lowest_addr = entry->start; + } + } + return lowest_addr; +} + +/* + * Check continuity of segments. + */ +static int +dcssblk_is_continuous(struct dcssblk_dev_info *dev_info) +{ + int i, j, rc; + struct segment_info *sort_list, *entry, temp; + + if (dev_info->num_of_segments <= 1) + return 0; + + sort_list = kzalloc( + sizeof(struct segment_info) * dev_info->num_of_segments, + GFP_KERNEL); + if (sort_list == NULL) + return -ENOMEM; + i = 0; + list_for_each_entry(entry, &dev_info->seg_list, lh) { + memcpy(&sort_list[i], entry, sizeof(struct segment_info)); + i++; + } + + /* sort segments */ + for (i = 0; i < dev_info->num_of_segments; i++) + for (j = 0; j < dev_info->num_of_segments; j++) + if (sort_list[j].start > sort_list[i].start) { + memcpy(&temp, &sort_list[i], + sizeof(struct segment_info)); + memcpy(&sort_list[i], &sort_list[j], + sizeof(struct segment_info)); + memcpy(&sort_list[j], &temp, + sizeof(struct segment_info)); + } + + /* check continuity */ + for (i = 0; i < dev_info->num_of_segments - 1; i++) { + if ((sort_list[i].end + 1) != sort_list[i+1].start) { + PRINT_ERR("Segment %s is not contiguous with " + "segment %s\n", + sort_list[i].segment_name, + sort_list[i+1].segment_name); + rc = -EINVAL; + goto out; + } + /* EN and EW are allowed in a block device */ + if (sort_list[i].segment_type != sort_list[i+1].segment_type) { + if (!(sort_list[i].segment_type & SEGMENT_EXCLUSIVE) || + (sort_list[i].segment_type == SEG_TYPE_ER) || + !(sort_list[i+1].segment_type & + SEGMENT_EXCLUSIVE) || + (sort_list[i+1].segment_type == SEG_TYPE_ER)) { + PRINT_ERR("Segment %s has different type from " + "segment %s\n", + sort_list[i].segment_name, + sort_list[i+1].segment_name); + rc = -EINVAL; + goto out; + } + } + } + rc = 0; +out: + kfree(sort_list); + return rc; +} + +/* + * Load a segment + */ +static int +dcssblk_load_segment(char *name, struct segment_info **seg_info) +{ + int rc; + + /* already loaded? */ + down_read(&dcssblk_devices_sem); + *seg_info = dcssblk_get_segment_by_name(name); + up_read(&dcssblk_devices_sem); + if (*seg_info != NULL) + return -EEXIST; + + /* get a struct segment_info */ + *seg_info = kzalloc(sizeof(struct segment_info), GFP_KERNEL); + if (*seg_info == NULL) + return -ENOMEM; + + strcpy((*seg_info)->segment_name, name); + + /* load the segment */ + rc = segment_load(name, SEGMENT_SHARED, + &(*seg_info)->start, &(*seg_info)->end); + if (rc < 0) { + segment_warning(rc, (*seg_info)->segment_name); + kfree(*seg_info); + } else { + INIT_LIST_HEAD(&(*seg_info)->lh); + (*seg_info)->segment_type = rc; + } + return rc; +} + static void dcssblk_unregister_callback(struct device *dev) { device_unregister(dev); @@ -165,6 +348,7 @@ static ssize_t dcssblk_shared_store(struct device *dev, struct device_attribute *attr, const char *inbuf, size_t count) { struct dcssblk_dev_info *dev_info; + struct segment_info *entry, *temp; int rc; if ((count > 1) && (inbuf[1] != '\n') && (inbuf[1] != '\0')) @@ -172,46 +356,46 @@ dcssblk_shared_store(struct device *dev, struct device_attribute *attr, const ch down_write(&dcssblk_devices_sem); dev_info = container_of(dev, struct dcssblk_dev_info, dev); if (atomic_read(&dev_info->use_count)) { - PRINT_ERR("share: segment %s is busy!\n", - dev_info->segment_name); rc = -EBUSY; goto out; } if (inbuf[0] == '1') { - // reload segment in shared mode - rc = segment_modify_shared(dev_info->segment_name, - SEGMENT_SHARED); - if (rc < 0) { - BUG_ON(rc == -EINVAL); - if (rc != -EAGAIN) - goto removeseg; - } else { - dev_info->is_shared = 1; - switch (dev_info->segment_type) { - case SEG_TYPE_SR: - case SEG_TYPE_ER: - case SEG_TYPE_SC: - set_disk_ro(dev_info->gd,1); + /* reload segments in shared mode */ + list_for_each_entry(entry, &dev_info->seg_list, lh) { + rc = segment_modify_shared(entry->segment_name, + SEGMENT_SHARED); + if (rc < 0) { + BUG_ON(rc == -EINVAL); + if (rc != -EAGAIN) + goto removeseg; } } + dev_info->is_shared = 1; + switch (dev_info->segment_type) { + case SEG_TYPE_SR: + case SEG_TYPE_ER: + case SEG_TYPE_SC: + set_disk_ro(dev_info->gd, 1); + } } else if (inbuf[0] == '0') { - // reload segment in exclusive mode + /* reload segments in exclusive mode */ if (dev_info->segment_type == SEG_TYPE_SC) { PRINT_ERR("Segment type SC (%s) cannot be loaded in " - "non-shared mode\n", dev_info->segment_name); + "non-shared mode\n", dev_info->segment_name); rc = -EINVAL; goto out; } - rc = segment_modify_shared(dev_info->segment_name, - SEGMENT_EXCLUSIVE); - if (rc < 0) { - BUG_ON(rc == -EINVAL); - if (rc != -EAGAIN) - goto removeseg; - } else { - dev_info->is_shared = 0; - set_disk_ro(dev_info->gd, 0); + list_for_each_entry(entry, &dev_info->seg_list, lh) { + rc = segment_modify_shared(entry->segment_name, + SEGMENT_EXCLUSIVE); + if (rc < 0) { + BUG_ON(rc == -EINVAL); + if (rc != -EAGAIN) + goto removeseg; + } } + dev_info->is_shared = 0; + set_disk_ro(dev_info->gd, 0); } else { rc = -EINVAL; goto out; @@ -220,8 +404,14 @@ dcssblk_shared_store(struct device *dev, struct device_attribute *attr, const ch goto out; removeseg: - PRINT_ERR("Could not reload segment %s, removing it now!\n", - dev_info->segment_name); + PRINT_ERR("Could not reload segment(s) of the device %s, removing " + "segment(s) now!\n", + dev_info->segment_name); + temp = entry; + list_for_each_entry(entry, &dev_info->seg_list, lh) { + if (entry != temp) + segment_unload(entry->segment_name); + } list_del(&dev_info->lh); del_gendisk(dev_info->gd); @@ -254,6 +444,7 @@ static ssize_t dcssblk_save_store(struct device *dev, struct device_attribute *attr, const char *inbuf, size_t count) { struct dcssblk_dev_info *dev_info; + struct segment_info *entry; if ((count > 1) && (inbuf[1] != '\n') && (inbuf[1] != '\0')) return -EINVAL; @@ -263,14 +454,16 @@ dcssblk_save_store(struct device *dev, struct device_attribute *attr, const char if (inbuf[0] == '1') { if (atomic_read(&dev_info->use_count) == 0) { // device is idle => we save immediately - PRINT_INFO("Saving segment %s\n", + PRINT_INFO("Saving segment(s) of the device %s\n", dev_info->segment_name); - segment_save(dev_info->segment_name); + list_for_each_entry(entry, &dev_info->seg_list, lh) { + segment_save(entry->segment_name); + } } else { // device is busy => we save it when it becomes // idle in dcssblk_release - PRINT_INFO("Segment %s is currently busy, it will " - "be saved when it becomes idle...\n", + PRINT_INFO("Device %s is currently busy, segment(s) " + "will be saved when it becomes idle...\n", dev_info->segment_name); dev_info->save_pending = 1; } @@ -279,7 +472,8 @@ dcssblk_save_store(struct device *dev, struct device_attribute *attr, const char // device is busy & the user wants to undo his save // request dev_info->save_pending = 0; - PRINT_INFO("Pending save for segment %s deactivated\n", + PRINT_INFO("Pending save for segment(s) of the device " + "%s deactivated\n", dev_info->segment_name); } } else { @@ -290,67 +484,124 @@ dcssblk_save_store(struct device *dev, struct device_attribute *attr, const char return count; } +/* + * device attribute for showing all segments in a device + */ +static ssize_t +dcssblk_seglist_show(struct device *dev, struct device_attribute *attr, + char *buf) +{ + int i; + + struct dcssblk_dev_info *dev_info; + struct segment_info *entry; + + down_read(&dcssblk_devices_sem); + dev_info = container_of(dev, struct dcssblk_dev_info, dev); + i = 0; + buf[0] = '\0'; + list_for_each_entry(entry, &dev_info->seg_list, lh) { + strcpy(&buf[i], entry->segment_name); + i += strlen(entry->segment_name); + buf[i] = '\n'; + i++; + } + up_read(&dcssblk_devices_sem); + return i; +} + /* * device attribute for adding devices */ static ssize_t dcssblk_add_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) { - int rc, i; + int rc, i, j, num_of_segments; struct dcssblk_dev_info *dev_info; + struct segment_info *seg_info, *temp; char *local_buf; unsigned long seg_byte_size; dev_info = NULL; + seg_info = NULL; if (dev != dcssblk_root_dev) { rc = -EINVAL; goto out_nobuf; } + if ((count < 1) || (buf[0] == '\0') || (buf[0] == '\n')) { + rc = -ENAMETOOLONG; + goto out_nobuf; + } + local_buf = kmalloc(count + 1, GFP_KERNEL); if (local_buf == NULL) { rc = -ENOMEM; goto out_nobuf; } + /* * parse input */ + num_of_segments = 0; for (i = 0; ((buf[i] != '\0') && (buf[i] != '\n') && i < count); i++) { - local_buf[i] = toupper(buf[i]); + for (j = i; (buf[j] != ':') && + (buf[j] != '\0') && + (buf[j] != '\n') && + j < count; j++) { + local_buf[j-i] = toupper(buf[j]); + } + local_buf[j-i] = '\0'; + if (((j - i) == 0) || ((j - i) > 8)) { + rc = -ENAMETOOLONG; + goto seg_list_del; + } + + rc = dcssblk_load_segment(local_buf, &seg_info); + if (rc < 0) + goto seg_list_del; + /* + * get a struct dcssblk_dev_info + */ + if (num_of_segments == 0) { + dev_info = kzalloc(sizeof(struct dcssblk_dev_info), + GFP_KERNEL); + if (dev_info == NULL) { + rc = -ENOMEM; + goto out; + } + strcpy(dev_info->segment_name, local_buf); + dev_info->segment_type = seg_info->segment_type; + INIT_LIST_HEAD(&dev_info->seg_list); + } + list_add_tail(&seg_info->lh, &dev_info->seg_list); + num_of_segments++; + i = j; + + if ((buf[j] == '\0') || (buf[j] == '\n')) + break; } - local_buf[i] = '\0'; - if ((i == 0) || (i > 8)) { + + /* no trailing colon at the end of the input */ + if ((i > 0) && (buf[i-1] == ':')) { rc = -ENAMETOOLONG; - goto out; - } - /* - * already loaded? - */ - down_read(&dcssblk_devices_sem); - dev_info = dcssblk_get_device_by_name(local_buf); - up_read(&dcssblk_devices_sem); - if (dev_info != NULL) { - PRINT_WARN("Segment %s already loaded!\n", local_buf); - rc = -EEXIST; - goto out; - } - /* - * get a struct dcssblk_dev_info - */ - dev_info = kzalloc(sizeof(struct dcssblk_dev_info), GFP_KERNEL); - if (dev_info == NULL) { - rc = -ENOMEM; - goto out; + goto seg_list_del; } + strlcpy(local_buf, buf, i + 1); + dev_info->num_of_segments = num_of_segments; + rc = dcssblk_is_continuous(dev_info); + if (rc < 0) + goto seg_list_del; + + dev_info->start = dcssblk_find_lowest_addr(dev_info); + dev_info->end = dcssblk_find_highest_addr(dev_info); - strcpy(dev_info->segment_name, local_buf); - dev_set_name(&dev_info->dev, local_buf); + dev_set_name(&dev_info->dev, dev_info->segment_name); dev_info->dev.release = dcssblk_release_segment; INIT_LIST_HEAD(&dev_info->lh); - dev_info->gd = alloc_disk(DCSSBLK_MINORS_PER_DISK); if (dev_info->gd == NULL) { rc = -ENOMEM; - goto free_dev_info; + goto seg_list_del; } dev_info->gd->major = dcssblk_major; dev_info->gd->fops = &dcssblk_devops; @@ -360,65 +611,52 @@ dcssblk_add_store(struct device *dev, struct device_attribute *attr, const char dev_info->gd->driverfs_dev = &dev_info->dev; blk_queue_make_request(dev_info->dcssblk_queue, dcssblk_make_request); blk_queue_hardsect_size(dev_info->dcssblk_queue, 4096); - /* - * load the segment - */ - rc = segment_load(local_buf, SEGMENT_SHARED, - &dev_info->start, &dev_info->end); - if (rc < 0) { - segment_warning(rc, dev_info->segment_name); - goto dealloc_gendisk; - } + seg_byte_size = (dev_info->end - dev_info->start + 1); set_capacity(dev_info->gd, seg_byte_size >> 9); // size in sectors - PRINT_INFO("Loaded segment %s, size = %lu Byte, " + PRINT_INFO("Loaded segment(s) %s, size = %lu Byte, " "capacity = %lu (512 Byte) sectors\n", local_buf, seg_byte_size, seg_byte_size >> 9); - dev_info->segment_type = rc; dev_info->save_pending = 0; dev_info->is_shared = 1; dev_info->dev.parent = dcssblk_root_dev; /* - * get minor, add to list + *get minor, add to list */ down_write(&dcssblk_devices_sem); - if (dcssblk_get_device_by_name(local_buf)) { - up_write(&dcssblk_devices_sem); + if (dcssblk_get_segment_by_name(local_buf)) { rc = -EEXIST; - goto unload_seg; + goto release_gd; } rc = dcssblk_assign_free_minor(dev_info); - if (rc) { - up_write(&dcssblk_devices_sem); - PRINT_ERR("No free minor number available! " - "Unloading segment...\n"); - goto unload_seg; - } + if (rc) + goto release_gd; sprintf(dev_info->gd->disk_name, "dcssblk%d", MINOR(disk_devt(dev_info->gd))); list_add_tail(&dev_info->lh, &dcssblk_devices); if (!try_module_get(THIS_MODULE)) { rc = -ENODEV; - goto list_del; + goto dev_list_del; } /* * register the device */ rc = device_register(&dev_info->dev); if (rc) { - PRINT_ERR("Segment %s could not be registered RC=%d\n", - local_buf, rc); module_put(THIS_MODULE); - goto list_del; + goto dev_list_del; } get_device(&dev_info->dev); rc = device_create_file(&dev_info->dev, &dev_attr_shared); if (rc) goto unregister_dev; rc = device_create_file(&dev_info->dev, &dev_attr_save); + if (rc) + goto unregister_dev; + rc = device_create_file(&dev_info->dev, &dev_attr_seglist); if (rc) goto unregister_dev; @@ -434,7 +672,6 @@ dcssblk_add_store(struct device *dev, struct device_attribute *attr, const char set_disk_ro(dev_info->gd,0); break; } - PRINT_DEBUG("Segment %s loaded successfully\n", local_buf); up_write(&dcssblk_devices_sem); rc = count; goto out; @@ -445,20 +682,27 @@ unregister_dev: dev_info->gd->queue = NULL; put_disk(dev_info->gd); device_unregister(&dev_info->dev); - segment_unload(dev_info->segment_name); + list_for_each_entry(seg_info, &dev_info->seg_list, lh) { + segment_unload(seg_info->segment_name); + } put_device(&dev_info->dev); up_write(&dcssblk_devices_sem); goto out; -list_del: +dev_list_del: list_del(&dev_info->lh); - up_write(&dcssblk_devices_sem); -unload_seg: - segment_unload(local_buf); -dealloc_gendisk: +release_gd: blk_cleanup_queue(dev_info->dcssblk_queue); dev_info->gd->queue = NULL; put_disk(dev_info->gd); -free_dev_info: + up_write(&dcssblk_devices_sem); +seg_list_del: + if (dev_info == NULL) + goto out; + list_for_each_entry_safe(seg_info, temp, &dev_info->seg_list, lh) { + list_del(&seg_info->lh); + segment_unload(seg_info->segment_name); + kfree(seg_info); + } kfree(dev_info); out: kfree(local_buf); @@ -473,6 +717,7 @@ static ssize_t dcssblk_remove_store(struct device *dev, struct device_attribute *attr, const char *buf, size_t count) { struct dcssblk_dev_info *dev_info; + struct segment_info *entry; int rc, i; char *local_buf; @@ -499,26 +744,28 @@ dcssblk_remove_store(struct device *dev, struct device_attribute *attr, const ch dev_info = dcssblk_get_device_by_name(local_buf); if (dev_info == NULL) { up_write(&dcssblk_devices_sem); - PRINT_WARN("Segment %s is not loaded!\n", local_buf); + PRINT_WARN("Device %s is not loaded!\n", local_buf); rc = -ENODEV; goto out_buf; } if (atomic_read(&dev_info->use_count) != 0) { up_write(&dcssblk_devices_sem); - PRINT_WARN("Segment %s is in use!\n", local_buf); + PRINT_WARN("Device %s is in use!\n", local_buf); rc = -EBUSY; goto out_buf; } - list_del(&dev_info->lh); + list_del(&dev_info->lh); del_gendisk(dev_info->gd); blk_cleanup_queue(dev_info->dcssblk_queue); dev_info->gd->queue = NULL; put_disk(dev_info->gd); device_unregister(&dev_info->dev); - segment_unload(dev_info->segment_name); - PRINT_DEBUG("Segment %s unloaded successfully\n", - dev_info->segment_name); + + /* unload all related segments */ + list_for_each_entry(entry, &dev_info->seg_list, lh) + segment_unload(entry->segment_name); + put_device(&dev_info->dev); up_write(&dcssblk_devices_sem); @@ -550,6 +797,7 @@ static int dcssblk_release(struct inode *inode, struct file *filp) { struct dcssblk_dev_info *dev_info; + struct segment_info *entry; int rc; dev_info = inode->i_bdev->bd_disk->private_data; @@ -560,9 +808,11 @@ dcssblk_release(struct inode *inode, struct file *filp) down_write(&dcssblk_devices_sem); if (atomic_dec_and_test(&dev_info->use_count) && (dev_info->save_pending)) { - PRINT_INFO("Segment %s became idle and is being saved now\n", + PRINT_INFO("Device %s became idle and is being saved now\n", dev_info->segment_name); - segment_save(dev_info->segment_name); + list_for_each_entry(entry, &dev_info->seg_list, lh) { + segment_save(entry->segment_name); + } dev_info->save_pending = 0; } up_write(&dcssblk_devices_sem); @@ -602,7 +852,7 @@ dcssblk_make_request(struct request_queue *q, struct bio *bio) case SEG_TYPE_SC: /* cannot write to these segments */ if (bio_data_dir(bio) == WRITE) { - PRINT_WARN("rejecting write to ro segment %s\n", + PRINT_WARN("rejecting write to ro device %s\n", dev_name(&dev_info->dev)); goto fail; } @@ -658,7 +908,7 @@ static void dcssblk_check_params(void) { int rc, i, j, k; - char buf[9]; + char buf[DCSSBLK_PARM_LEN + 1]; struct dcssblk_dev_info *dev_info; for (i = 0; (i < DCSSBLK_PARM_LEN) && (dcssblk_segments[i] != '\0'); @@ -666,15 +916,16 @@ dcssblk_check_params(void) for (j = i; (dcssblk_segments[j] != ',') && (dcssblk_segments[j] != '\0') && (dcssblk_segments[j] != '(') && - (j - i) < 8; j++) + (j < DCSSBLK_PARM_LEN); j++) { buf[j-i] = dcssblk_segments[j]; } buf[j-i] = '\0'; rc = dcssblk_add_store(dcssblk_root_dev, NULL, buf, j-i); if ((rc >= 0) && (dcssblk_segments[j] == '(')) { - for (k = 0; buf[k] != '\0'; k++) + for (k = 0; (buf[k] != ':') && (buf[k] != '\0'); k++) buf[k] = toupper(buf[k]); + buf[k] = '\0'; if (!strncmp(&dcssblk_segments[j], "(local)", 7)) { down_read(&dcssblk_devices_sem); dev_info = dcssblk_get_device_by_name(buf); @@ -741,10 +992,12 @@ module_exit(dcssblk_exit); module_param_string(segments, dcssblk_segments, DCSSBLK_PARM_LEN, 0444); MODULE_PARM_DESC(segments, "Name of DCSS segment(s) to be loaded, " - "comma-separated list, each name max. 8 chars.\n" - "Adding \"(local)\" to segment name equals echoing 0 to " - "/sys/devices/dcssblk//shared after loading " - "the segment - \n" - "e.g. segments=\"mydcss1,mydcss2,mydcss3(local)\""); + "comma-separated list, names in each set separated " + "by commas are separated by colons, each set contains " + "names of contiguous segments and each name max. 8 chars.\n" + "Adding \"(local)\" to the end of each set equals echoing 0 " + "to /sys/devices/dcssblk//shared after loading " + "the contiguous segments - \n" + "e.g. segments=\"mydcss1,mydcss2:mydcss3,mydcss4(local)\""); MODULE_LICENSE("GPL"); -- cgit v1.2.3-59-g8ed1b From 5a0d0e65379256b4da2c9092e197a2c761f51c01 Mon Sep 17 00:00:00 2001 From: Heiko Carstens Date: Fri, 10 Oct 2008 21:33:22 +0200 Subject: [S390] Move private simple udelay function to arch/s390/lib/delay.c. Move cio's private simple udelay function to lib/delay.c and turn it into something much more readable. So we have all implementations at one place. Signed-off-by: Heiko Carstens Signed-off-by: Martin Schwidefsky --- arch/s390/include/asm/delay.h | 1 + arch/s390/lib/delay.c | 13 +++++++++++++ drivers/s390/cio/cio.c | 17 ++--------------- 3 files changed, 16 insertions(+), 15 deletions(-) (limited to 'arch') diff --git a/arch/s390/include/asm/delay.h b/arch/s390/include/asm/delay.h index 78357314c450..a356c958e260 100644 --- a/arch/s390/include/asm/delay.h +++ b/arch/s390/include/asm/delay.h @@ -15,6 +15,7 @@ #define _S390_DELAY_H extern void __udelay(unsigned long usecs); +extern void udelay_simple(unsigned long usecs); extern void __delay(unsigned long loops); #define udelay(n) __udelay(n) diff --git a/arch/s390/lib/delay.c b/arch/s390/lib/delay.c index 0953cee05efc..6ccb9fab055a 100644 --- a/arch/s390/lib/delay.c +++ b/arch/s390/lib/delay.c @@ -92,3 +92,16 @@ out: local_irq_restore(flags); preempt_enable(); } + +/* + * Simple udelay variant. To be used on startup and reboot + * when the interrupt handler isn't working. + */ +void udelay_simple(unsigned long usecs) +{ + u64 end; + + end = get_clock() + ((u64) usecs << 12); + while (get_clock() < end) + cpu_relax(); +} diff --git a/drivers/s390/cio/cio.c b/drivers/s390/cio/cio.c index c0cb72547256..3db2c386546f 100644 --- a/drivers/s390/cio/cio.c +++ b/drivers/s390/cio/cio.c @@ -859,19 +859,6 @@ __disable_subchannel_easy(struct subchannel_id schid, struct schib *schib) return -EBUSY; /* uhm... */ } -/* we can't use the normal udelay here, since it enables external interrupts */ - -static void udelay_reset(unsigned long usecs) -{ - uint64_t start_cc, end_cc; - - asm volatile ("STCK %0" : "=m" (start_cc)); - do { - cpu_relax(); - asm volatile ("STCK %0" : "=m" (end_cc)); - } while (((end_cc - start_cc)/4096) < usecs); -} - static int __clear_io_subchannel_easy(struct subchannel_id schid) { @@ -887,7 +874,7 @@ __clear_io_subchannel_easy(struct subchannel_id schid) if (schid_equal(&ti.schid, &schid)) return 0; } - udelay_reset(100); + udelay_simple(100); } return -EBUSY; } @@ -895,7 +882,7 @@ __clear_io_subchannel_easy(struct subchannel_id schid) static void __clear_chsc_subchannel_easy(void) { /* It seems we can only wait for a bit here :/ */ - udelay_reset(100); + udelay_simple(100); } static int pgm_check_occured; -- cgit v1.2.3-59-g8ed1b From ab1d848fd6a9151b02c6cbf4bddce6e24707b094 Mon Sep 17 00:00:00 2001 From: Nigel Hislop Date: Fri, 10 Oct 2008 21:33:25 +0200 Subject: [S390] Add ioctl support for EMC Symmetrix Subsystem Control I/O EMC Symmetrix Subsystem Control I/O through CKD dasd requires a specific parameter list sent to the array via a Perform Subsystem Function CCW. The Symmetrix response is retrieved from the array via a Read Subsystem Data CCW. Signed-off-by: Nigel Hislop Signed-off-by: Hannes Reinecke Signed-off-by: Martin Schwidefsky --- arch/s390/include/asm/dasd.h | 13 ++++++ drivers/s390/block/dasd_eckd.c | 101 +++++++++++++++++++++++++++++++++++++++++ 2 files changed, 114 insertions(+) (limited to 'arch') diff --git a/arch/s390/include/asm/dasd.h b/arch/s390/include/asm/dasd.h index 3f002e13d024..55b2b80cdf6e 100644 --- a/arch/s390/include/asm/dasd.h +++ b/arch/s390/include/asm/dasd.h @@ -3,6 +3,8 @@ * Author(s)......: Holger Smolinski * Bugreports.to..: * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000 + * EMC Symmetrix ioctl Copyright EMC Corporation, 2008 + * Author.........: Nigel Hislop * * This file is the interface of the DASD device driver, which is exported to user space * any future changes wrt the API will result in a change of the APIVERSION reported @@ -202,6 +204,16 @@ typedef struct attrib_data_t { #define DASD_SEQ_PRESTAGE 0x4 #define DASD_REC_ACCESS 0x5 +/* + * Perform EMC Symmetrix I/O + */ +typedef struct dasd_symmio_parms { + unsigned char reserved[8]; /* compat with older releases */ + unsigned long long psf_data; /* char * cast to u64 */ + unsigned long long rssd_result; /* char * cast to u64 */ + int psf_data_len; + int rssd_result_len; +} __attribute__ ((packed)) dasd_symmio_parms_t; /******************************************************************************** * SECTION: Definition of IOCTLs @@ -247,6 +259,7 @@ typedef struct attrib_data_t { /* Set Attributes (cache operations) */ #define BIODASDSATTR _IOW(DASD_IOCTL_LETTER,2,attrib_data_t) +#define BIODASDSYMMIO _IOWR(DASD_IOCTL_LETTER, 240, dasd_symmio_parms_t) #endif /* DASD_H */ diff --git a/drivers/s390/block/dasd_eckd.c b/drivers/s390/block/dasd_eckd.c index 8095629bc493..49f9d221e23d 100644 --- a/drivers/s390/block/dasd_eckd.c +++ b/drivers/s390/block/dasd_eckd.c @@ -6,6 +6,8 @@ * Martin Schwidefsky * Bugreports.to..: * (C) IBM Corporation, IBM Deutschland Entwicklung GmbH, 1999,2000 + * EMC Symmetrix ioctl Copyright EMC Corporation, 2008 + * Author.........: Nigel Hislop * */ @@ -2083,6 +2085,103 @@ dasd_eckd_set_attrib(struct dasd_device *device, void __user *argp) return 0; } +/* + * Issue syscall I/O to EMC Symmetrix array. + * CCWs are PSF and RSSD + */ +static int dasd_symm_io(struct dasd_device *device, void __user *argp) +{ + struct dasd_symmio_parms usrparm; + char *psf_data, *rssd_result; + struct dasd_ccw_req *cqr; + struct ccw1 *ccw; + int rc; + + /* Copy parms from caller */ + rc = -EFAULT; + if (copy_from_user(&usrparm, argp, sizeof(usrparm))) + goto out; +#ifndef CONFIG_64BIT + /* Make sure pointers are sane even on 31 bit. */ + if ((usrparm.psf_data >> 32) != 0 || (usrparm.rssd_result >> 32) != 0) { + rc = -EINVAL; + goto out; + } +#endif + /* alloc I/O data area */ + psf_data = kzalloc(usrparm.psf_data_len, GFP_KERNEL | GFP_DMA); + rssd_result = kzalloc(usrparm.rssd_result_len, GFP_KERNEL | GFP_DMA); + if (!psf_data || !rssd_result) { + rc = -ENOMEM; + goto out_free; + } + + /* get syscall header from user space */ + rc = -EFAULT; + if (copy_from_user(psf_data, + (void __user *)(unsigned long) usrparm.psf_data, + usrparm.psf_data_len)) + goto out_free; + + /* sanity check on syscall header */ + if (psf_data[0] != 0x17 && psf_data[1] != 0xce) { + rc = -EINVAL; + goto out_free; + } + + /* setup CCWs for PSF + RSSD */ + cqr = dasd_smalloc_request("ECKD", 2 , 0, device); + if (IS_ERR(cqr)) { + DEV_MESSAGE(KERN_WARNING, device, "%s", + "Could not allocate initialization request"); + rc = PTR_ERR(cqr); + goto out_free; + } + + cqr->startdev = device; + cqr->memdev = device; + cqr->retries = 3; + cqr->expires = 10 * HZ; + cqr->buildclk = get_clock(); + cqr->status = DASD_CQR_FILLED; + + /* Build the ccws */ + ccw = cqr->cpaddr; + + /* PSF ccw */ + ccw->cmd_code = DASD_ECKD_CCW_PSF; + ccw->count = usrparm.psf_data_len; + ccw->flags |= CCW_FLAG_CC; + ccw->cda = (__u32)(addr_t) psf_data; + + ccw++; + + /* RSSD ccw */ + ccw->cmd_code = DASD_ECKD_CCW_RSSD; + ccw->count = usrparm.rssd_result_len; + ccw->flags = CCW_FLAG_SLI ; + ccw->cda = (__u32)(addr_t) rssd_result; + + rc = dasd_sleep_on(cqr); + if (rc) + goto out_sfree; + + rc = -EFAULT; + if (copy_to_user((void __user *)(unsigned long) usrparm.rssd_result, + rssd_result, usrparm.rssd_result_len)) + goto out_sfree; + rc = 0; + +out_sfree: + dasd_sfree_request(cqr, cqr->memdev); +out_free: + kfree(rssd_result); + kfree(psf_data); +out: + DBF_DEV_EVENT(DBF_WARNING, device, "Symmetrix ioctl: rc=%d", rc); + return rc; +} + static int dasd_eckd_ioctl(struct dasd_block *block, unsigned int cmd, void __user *argp) { @@ -2101,6 +2200,8 @@ dasd_eckd_ioctl(struct dasd_block *block, unsigned int cmd, void __user *argp) return dasd_eckd_reserve(device); case BIODASDSLCK: return dasd_eckd_steal_lock(device); + case BIODASDSYMMIO: + return dasd_symm_io(device, argp); default: return -ENOIOCTLCMD; } -- cgit v1.2.3-59-g8ed1b From 15e86b0c752d50e910b2cca6e83ce74c4440d06c Mon Sep 17 00:00:00 2001 From: Florian Funke Date: Fri, 10 Oct 2008 21:33:26 +0200 Subject: [S390] introduce dirty bit for kvm live migration This patch defines a dirty bit in the PGSTE that can be used to implement dirty pages logging for KVM's live migration. The bit is set in the ptep_rcp_copy function, which is called to save dirty and referenced information from the storage key in the PGSTE. The bit can be tested and reset by KVM using the kvm_s390_test_and_clear_page_dirty function that is introduced by this patch. Acked-by: Carsten Otte Signed-off-by: Florian Funke Signed-off-by: Martin Schwidefsky --- arch/s390/include/asm/pgtable.h | 45 +++++++++++++++++++++++++++++++++++++++-- 1 file changed, 43 insertions(+), 2 deletions(-) (limited to 'arch') diff --git a/arch/s390/include/asm/pgtable.h b/arch/s390/include/asm/pgtable.h index 0bdb704ae051..1a928f84afd6 100644 --- a/arch/s390/include/asm/pgtable.h +++ b/arch/s390/include/asm/pgtable.h @@ -281,6 +281,9 @@ extern char empty_zero_page[PAGE_SIZE]; #define RCP_GR_BIT 50 #define RCP_GC_BIT 49 +/* User dirty bit for KVM's migration feature */ +#define KVM_UD_BIT 47 + #ifndef __s390x__ /* Bits in the segment table address-space-control-element */ @@ -575,12 +578,16 @@ static inline void ptep_rcp_copy(pte_t *ptep) unsigned long *pgste = (unsigned long *) (ptep + PTRS_PER_PTE); skey = page_get_storage_key(page_to_phys(page)); - if (skey & _PAGE_CHANGED) + if (skey & _PAGE_CHANGED) { set_bit_simple(RCP_GC_BIT, pgste); + set_bit_simple(KVM_UD_BIT, pgste); + } if (skey & _PAGE_REFERENCED) set_bit_simple(RCP_GR_BIT, pgste); - if (test_and_clear_bit_simple(RCP_HC_BIT, pgste)) + if (test_and_clear_bit_simple(RCP_HC_BIT, pgste)) { SetPageDirty(page); + set_bit_simple(KVM_UD_BIT, pgste); + } if (test_and_clear_bit_simple(RCP_HR_BIT, pgste)) SetPageReferenced(page); #endif @@ -744,6 +751,40 @@ static inline pte_t pte_mkspecial(pte_t pte) return pte; } +#ifdef CONFIG_PGSTE +/* + * Get (and clear) the user dirty bit for a PTE. + */ +static inline int kvm_s390_test_and_clear_page_dirty(struct mm_struct *mm, + pte_t *ptep) +{ + int dirty; + unsigned long *pgste; + struct page *page; + unsigned int skey; + + if (!mm->context.pgstes) + return -EINVAL; + rcp_lock(ptep); + pgste = (unsigned long *) (ptep + PTRS_PER_PTE); + page = virt_to_page(pte_val(*ptep)); + skey = page_get_storage_key(page_to_phys(page)); + if (skey & _PAGE_CHANGED) { + set_bit_simple(RCP_GC_BIT, pgste); + set_bit_simple(KVM_UD_BIT, pgste); + } + if (test_and_clear_bit_simple(RCP_HC_BIT, pgste)) { + SetPageDirty(page); + set_bit_simple(KVM_UD_BIT, pgste); + } + dirty = test_and_clear_bit_simple(KVM_UD_BIT, pgste); + if (skey & _PAGE_CHANGED) + page_clear_dirty(page); + rcp_unlock(ptep); + return dirty; +} +#endif + #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG static inline int ptep_test_and_clear_young(struct vm_area_struct *vma, unsigned long addr, pte_t *ptep) -- cgit v1.2.3-59-g8ed1b From 4a672cfa3a7fcbc6f2adc558f34148be1096c561 Mon Sep 17 00:00:00 2001 From: Martin Schwidefsky Date: Fri, 10 Oct 2008 21:33:29 +0200 Subject: [S390] fix initialization of stp chsc_sstpc returns -EIO on error and 0 on success but stp_reset checks against 1 instead of 0. chsc_sstpc used to return 1 on success, one call location has not been updated .. Signed-off-by: Martin Schwidefsky --- arch/s390/kernel/time.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'arch') diff --git a/arch/s390/kernel/time.c b/arch/s390/kernel/time.c index 06acb1a18bbc..b94e9e3b694a 100644 --- a/arch/s390/kernel/time.c +++ b/arch/s390/kernel/time.c @@ -1356,7 +1356,7 @@ static void __init stp_reset(void) stp_page = alloc_bootmem_pages(PAGE_SIZE); rc = chsc_sstpc(stp_page, STP_OP_CTRL, 0x0000); - if (rc == 1) + if (rc == 0) set_bit(CLOCK_SYNC_HAS_STP, &clock_sync_flags); else if (stp_online) { printk(KERN_WARNING "Running on non STP capable machine.\n"); -- cgit v1.2.3-59-g8ed1b