diff options
Diffstat (limited to 'arch/powerpc/platforms/pseries/mobility.c')
| -rw-r--r-- | arch/powerpc/platforms/pseries/mobility.c | 583 |
1 files changed, 491 insertions, 92 deletions
diff --git a/arch/powerpc/platforms/pseries/mobility.c b/arch/powerpc/platforms/pseries/mobility.c index b571285f6c14..634fac5db3f9 100644 --- a/arch/powerpc/platforms/pseries/mobility.c +++ b/arch/powerpc/platforms/pseries/mobility.c @@ -6,12 +6,17 @@ * Copyright (C) 2010 IBM Corporation */ + +#define pr_fmt(fmt) "mobility: " fmt + #include <linux/cpu.h> #include <linux/kernel.h> #include <linux/kobject.h> +#include <linux/nmi.h> #include <linux/sched.h> #include <linux/smp.h> #include <linux/stat.h> +#include <linux/stop_machine.h> #include <linux/completion.h> #include <linux/device.h> #include <linux/delay.h> @@ -21,6 +26,7 @@ #include <asm/machdep.h> #include <asm/rtas.h> #include "pseries.h" +#include "vas.h" /* vas_migration_handler() */ #include "../../kernel/cacheinfo.h" static struct kobject *mobility_kobj; @@ -42,6 +48,39 @@ struct update_props_workarea { #define MIGRATION_SCOPE (1) #define PRRN_SCOPE -2 +#ifdef CONFIG_PPC_WATCHDOG +static unsigned int nmi_wd_lpm_factor = 200; + +#ifdef CONFIG_SYSCTL +static struct ctl_table nmi_wd_lpm_factor_ctl_table[] = { + { + .procname = "nmi_wd_lpm_factor", + .data = &nmi_wd_lpm_factor, + .maxlen = sizeof(int), + .mode = 0644, + .proc_handler = proc_douintvec_minmax, + }, + {} +}; +static struct ctl_table nmi_wd_lpm_factor_sysctl_root[] = { + { + .procname = "kernel", + .mode = 0555, + .child = nmi_wd_lpm_factor_ctl_table, + }, + {} +}; + +static int __init register_nmi_wd_lpm_factor_sysctl(void) +{ + register_sysctl_table(nmi_wd_lpm_factor_sysctl_root); + + return 0; +} +device_initcall(register_nmi_wd_lpm_factor_sysctl); +#endif /* CONFIG_SYSCTL */ +#endif /* CONFIG_PPC_WATCHDOG */ + static int mobility_rtas_call(int token, char *buf, s32 scope) { int rc; @@ -56,16 +95,31 @@ static int mobility_rtas_call(int token, char *buf, s32 scope) return rc; } -static int delete_dt_node(__be32 phandle) +static int delete_dt_node(struct device_node *dn) { - struct device_node *dn; + struct device_node *pdn; + bool is_platfac; - dn = of_find_node_by_phandle(be32_to_cpu(phandle)); - if (!dn) - return -ENOENT; + pdn = of_get_parent(dn); + is_platfac = of_node_is_type(dn, "ibm,platform-facilities") || + of_node_is_type(pdn, "ibm,platform-facilities"); + of_node_put(pdn); + + /* + * The drivers that bind to nodes in the platform-facilities + * hierarchy don't support node removal, and the removal directive + * from firmware is always followed by an add of an equivalent + * node. The capability (e.g. RNG, encryption, compression) + * represented by the node is never interrupted by the migration. + * So ignore changes to this part of the tree. + */ + if (is_platfac) { + pr_notice("ignoring remove operation for %pOFfp\n", dn); + return 0; + } + pr_debug("removing node %pOFfp\n", dn); dlpar_detach_node(dn); - of_node_put(dn); return 0; } @@ -122,6 +176,7 @@ static int update_dt_property(struct device_node *dn, struct property **prop, } if (!more) { + pr_debug("updating node %pOF property %s\n", dn, name); of_update_property(dn, new_prop); *prop = NULL; } @@ -129,10 +184,9 @@ static int update_dt_property(struct device_node *dn, struct property **prop, return 0; } -static int update_dt_node(__be32 phandle, s32 scope) +static int update_dt_node(struct device_node *dn, s32 scope) { struct update_props_workarea *upwa; - struct device_node *dn; struct property *prop = NULL; int i, rc, rtas_rc; char *prop_data; @@ -149,14 +203,8 @@ static int update_dt_node(__be32 phandle, s32 scope) if (!rtas_buf) return -ENOMEM; - dn = of_find_node_by_phandle(be32_to_cpu(phandle)); - if (!dn) { - kfree(rtas_buf); - return -ENOENT; - } - upwa = (struct update_props_workarea *)&rtas_buf[0]; - upwa->phandle = phandle; + upwa->phandle = cpu_to_be32(dn->phandle); do { rtas_rc = mobility_rtas_call(update_properties_token, rtas_buf, @@ -168,7 +216,7 @@ static int update_dt_node(__be32 phandle, s32 scope) nprops = be32_to_cpu(upwa->nprops); /* On the first call to ibm,update-properties for a node the - * the first property value descriptor contains an empty + * first property value descriptor contains an empty * property name, the property value length encoded as u32, * and the property value is the node path being updated. */ @@ -202,11 +250,12 @@ static int update_dt_node(__be32 phandle, s32 scope) rc = update_dt_property(dn, &prop, prop_name, vd, prop_data); if (rc) { - printk(KERN_ERR "Could not update %s" - " property\n", prop_name); + pr_err("updating %s property failed: %d\n", + prop_name, rc); } prop_data += vd; + break; } cond_resched(); @@ -215,59 +264,42 @@ static int update_dt_node(__be32 phandle, s32 scope) cond_resched(); } while (rtas_rc == 1); - of_node_put(dn); kfree(rtas_buf); return 0; } -static int add_dt_node(__be32 parent_phandle, __be32 drc_index) +static int add_dt_node(struct device_node *parent_dn, __be32 drc_index) { struct device_node *dn; - struct device_node *parent_dn; int rc; - parent_dn = of_find_node_by_phandle(be32_to_cpu(parent_phandle)); - if (!parent_dn) - return -ENOENT; - dn = dlpar_configure_connector(drc_index, parent_dn); - if (!dn) { - of_node_put(parent_dn); + if (!dn) return -ENOENT; + + /* + * Since delete_dt_node() ignores this node type, this is the + * necessary counterpart. We also know that a platform-facilities + * node returned from dlpar_configure_connector() has children + * attached, and dlpar_attach_node() only adds the parent, leaking + * the children. So ignore these on the add side for now. + */ + if (of_node_is_type(dn, "ibm,platform-facilities")) { + pr_notice("ignoring add operation for %pOF\n", dn); + dlpar_free_cc_nodes(dn); + return 0; } rc = dlpar_attach_node(dn, parent_dn); if (rc) dlpar_free_cc_nodes(dn); - of_node_put(parent_dn); - return rc; -} + pr_debug("added node %pOFfp\n", dn); -static void prrn_update_node(__be32 phandle) -{ - struct pseries_hp_errorlog hp_elog; - struct device_node *dn; - - /* - * If a node is found from a the given phandle, the phandle does not - * represent the drc index of an LMB and we can ignore. - */ - dn = of_find_node_by_phandle(be32_to_cpu(phandle)); - if (dn) { - of_node_put(dn); - return; - } - - hp_elog.resource = PSERIES_HP_ELOG_RESOURCE_MEM; - hp_elog.action = PSERIES_HP_ELOG_ACTION_READD; - hp_elog.id_type = PSERIES_HP_ELOG_ID_DRC_INDEX; - hp_elog._drc_u.drc_index = phandle; - - handle_dlpar_errorlog(&hp_elog); + return rc; } -int pseries_devicetree_update(s32 scope) +static int pseries_devicetree_update(s32 scope) { char *rtas_buf; __be32 *data; @@ -276,7 +308,7 @@ int pseries_devicetree_update(s32 scope) update_nodes_token = rtas_token("ibm,update-nodes"); if (update_nodes_token == RTAS_UNKNOWN_SERVICE) - return -EINVAL; + return 0; rtas_buf = kzalloc(RTAS_DATA_BUF_SIZE, GFP_KERNEL); if (!rtas_buf) @@ -296,26 +328,31 @@ int pseries_devicetree_update(s32 scope) data++; for (i = 0; i < node_count; i++) { + struct device_node *np; __be32 phandle = *data++; __be32 drc_index; + np = of_find_node_by_phandle(be32_to_cpu(phandle)); + if (!np) { + pr_warn("Failed lookup: phandle 0x%x for action 0x%x\n", + be32_to_cpu(phandle), action); + continue; + } + switch (action) { case DELETE_DT_NODE: - delete_dt_node(phandle); + delete_dt_node(np); break; case UPDATE_DT_NODE: - update_dt_node(phandle, scope); - - if (scope == PRRN_SCOPE) - prrn_update_node(phandle); - + update_dt_node(np, scope); break; case ADD_DT_NODE: drc_index = *data++; - add_dt_node(phandle, drc_index); + add_dt_node(np, drc_index); break; } + of_node_put(np); cond_resched(); } } @@ -330,21 +367,8 @@ int pseries_devicetree_update(s32 scope) void post_mobility_fixup(void) { int rc; - int activate_fw_token; - activate_fw_token = rtas_token("ibm,activate-firmware"); - if (activate_fw_token == RTAS_UNKNOWN_SERVICE) { - printk(KERN_ERR "Could not make post-mobility " - "activate-fw call.\n"); - return; - } - - do { - rc = rtas_call(activate_fw_token, 0, 1, NULL); - } while (rtas_busy_delay(rc)); - - if (rc) - printk(KERN_ERR "Post-mobility activate-fw failed: %d\n", rc); + rtas_activate_firmware(); /* * We don't want CPUs to go online/offline while the device @@ -361,19 +385,405 @@ void post_mobility_fixup(void) rc = pseries_devicetree_update(MIGRATION_SCOPE); if (rc) - printk(KERN_ERR "Post-mobility device tree update " - "failed: %d\n", rc); + pr_err("device tree update failed: %d\n", rc); cacheinfo_rebuild(); cpus_read_unlock(); - /* Possibly switch to a new RFI flush type */ - pseries_setup_rfi_flush(); + /* Possibly switch to a new L1 flush type */ + pseries_setup_security_mitigations(); + + /* Reinitialise system information for hv-24x7 */ + read_24x7_sys_info(); return; } +static int poll_vasi_state(u64 handle, unsigned long *res) +{ + unsigned long retbuf[PLPAR_HCALL_BUFSIZE]; + long hvrc; + int ret; + + hvrc = plpar_hcall(H_VASI_STATE, retbuf, handle); + switch (hvrc) { + case H_SUCCESS: + ret = 0; + *res = retbuf[0]; + break; + case H_PARAMETER: + ret = -EINVAL; + break; + case H_FUNCTION: + ret = -EOPNOTSUPP; + break; + case H_HARDWARE: + default: + pr_err("unexpected H_VASI_STATE result %ld\n", hvrc); + ret = -EIO; + break; + } + return ret; +} + +static int wait_for_vasi_session_suspending(u64 handle) +{ + unsigned long state; + int ret; + + /* + * Wait for transition from H_VASI_ENABLED to + * H_VASI_SUSPENDING. Treat anything else as an error. + */ + while (true) { + ret = poll_vasi_state(handle, &state); + + if (ret != 0 || state == H_VASI_SUSPENDING) { + break; + } else if (state == H_VASI_ENABLED) { + ssleep(1); + } else { + pr_err("unexpected H_VASI_STATE result %lu\n", state); + ret = -EIO; + break; + } + } + + /* + * Proceed even if H_VASI_STATE is unavailable. If H_JOIN or + * ibm,suspend-me are also unimplemented, we'll recover then. + */ + if (ret == -EOPNOTSUPP) + ret = 0; + + return ret; +} + +static void wait_for_vasi_session_completed(u64 handle) +{ + unsigned long state = 0; + int ret; + + pr_info("waiting for memory transfer to complete...\n"); + + /* + * Wait for transition from H_VASI_RESUMED to H_VASI_COMPLETED. + */ + while (true) { + ret = poll_vasi_state(handle, &state); + + /* + * If the memory transfer is already complete and the migration + * has been cleaned up by the hypervisor, H_PARAMETER is return, + * which is translate in EINVAL by poll_vasi_state(). + */ + if (ret == -EINVAL || (!ret && state == H_VASI_COMPLETED)) { + pr_info("memory transfer completed.\n"); + break; + } + + if (ret) { + pr_err("H_VASI_STATE return error (%d)\n", ret); + break; + } + + if (state != H_VASI_RESUMED) { + pr_err("unexpected H_VASI_STATE result %lu\n", state); + break; + } + + msleep(500); + } +} + +static void prod_single(unsigned int target_cpu) +{ + long hvrc; + int hwid; + + hwid = get_hard_smp_processor_id(target_cpu); + hvrc = plpar_hcall_norets(H_PROD, hwid); + if (hvrc == H_SUCCESS) + return; + pr_err_ratelimited("H_PROD of CPU %u (hwid %d) error: %ld\n", + target_cpu, hwid, hvrc); +} + +static void prod_others(void) +{ + unsigned int cpu; + + for_each_online_cpu(cpu) { + if (cpu != smp_processor_id()) + prod_single(cpu); + } +} + +static u16 clamp_slb_size(void) +{ +#ifdef CONFIG_PPC_64S_HASH_MMU + u16 prev = mmu_slb_size; + + slb_set_size(SLB_MIN_SIZE); + + return prev; +#else + return 0; +#endif +} + +static int do_suspend(void) +{ + u16 saved_slb_size; + int status; + int ret; + + pr_info("calling ibm,suspend-me on CPU %i\n", smp_processor_id()); + + /* + * The destination processor model may have fewer SLB entries + * than the source. We reduce mmu_slb_size to a safe minimum + * before suspending in order to minimize the possibility of + * programming non-existent entries on the destination. If + * suspend fails, we restore it before returning. On success + * the OF reconfig path will update it from the new device + * tree after resuming on the destination. + */ + saved_slb_size = clamp_slb_size(); + + ret = rtas_ibm_suspend_me(&status); + if (ret != 0) { + pr_err("ibm,suspend-me error: %d\n", status); + slb_set_size(saved_slb_size); + } + + return ret; +} + +/** + * struct pseries_suspend_info - State shared between CPUs for join/suspend. + * @counter: Threads are to increment this upon resuming from suspend + * or if an error is received from H_JOIN. The thread which performs + * the first increment (i.e. sets it to 1) is responsible for + * waking the other threads. + * @done: False if join/suspend is in progress. True if the operation is + * complete (successful or not). + */ +struct pseries_suspend_info { + atomic_t counter; + bool done; +}; + +static int do_join(void *arg) +{ + struct pseries_suspend_info *info = arg; + atomic_t *counter = &info->counter; + long hvrc; + int ret; + +retry: + /* Must ensure MSR.EE off for H_JOIN. */ + hard_irq_disable(); + hvrc = plpar_hcall_norets(H_JOIN); + + switch (hvrc) { + case H_CONTINUE: + /* + * All other CPUs are offline or in H_JOIN. This CPU + * attempts the suspend. + */ + ret = do_suspend(); + break; + case H_SUCCESS: + /* + * The suspend is complete and this cpu has received a + * prod, or we've received a stray prod from unrelated + * code (e.g. paravirt spinlocks) and we need to join + * again. + * + * This barrier orders the return from H_JOIN above vs + * the load of info->done. It pairs with the barrier + * in the wakeup/prod path below. + */ + smp_mb(); + if (READ_ONCE(info->done) == false) { + pr_info_ratelimited("premature return from H_JOIN on CPU %i, retrying", + smp_processor_id()); + goto retry; + } + ret = 0; + break; + case H_BAD_MODE: + case H_HARDWARE: + default: + ret = -EIO; + pr_err_ratelimited("H_JOIN error %ld on CPU %i\n", + hvrc, smp_processor_id()); + break; + } + + if (atomic_inc_return(counter) == 1) { + pr_info("CPU %u waking all threads\n", smp_processor_id()); + WRITE_ONCE(info->done, true); + /* + * This barrier orders the store to info->done vs subsequent + * H_PRODs to wake the other CPUs. It pairs with the barrier + * in the H_SUCCESS case above. + */ + smp_mb(); + prod_others(); + } + /* + * Execution may have been suspended for several seconds, so + * reset the watchdog. + */ + touch_nmi_watchdog(); + return ret; +} + +/* + * Abort reason code byte 0. We use only the 'Migrating partition' value. + */ +enum vasi_aborting_entity { + ORCHESTRATOR = 1, + VSP_SOURCE = 2, + PARTITION_FIRMWARE = 3, + PLATFORM_FIRMWARE = 4, + VSP_TARGET = 5, + MIGRATING_PARTITION = 6, +}; + +static void pseries_cancel_migration(u64 handle, int err) +{ + u32 reason_code; + u32 detail; + u8 entity; + long hvrc; + + entity = MIGRATING_PARTITION; + detail = abs(err) & 0xffffff; + reason_code = (entity << 24) | detail; + + hvrc = plpar_hcall_norets(H_VASI_SIGNAL, handle, + H_VASI_SIGNAL_CANCEL, reason_code); + if (hvrc) + pr_err("H_VASI_SIGNAL error: %ld\n", hvrc); +} + +static int pseries_suspend(u64 handle) +{ + const unsigned int max_attempts = 5; + unsigned int retry_interval_ms = 1; + unsigned int attempt = 1; + int ret; + + while (true) { + struct pseries_suspend_info info; + unsigned long vasi_state; + int vasi_err; + + info = (struct pseries_suspend_info) { + .counter = ATOMIC_INIT(0), + .done = false, + }; + + ret = stop_machine(do_join, &info, cpu_online_mask); + if (ret == 0) + break; + /* + * Encountered an error. If the VASI stream is still + * in Suspending state, it's likely a transient + * condition related to some device in the partition + * and we can retry in the hope that the cause has + * cleared after some delay. + * + * A better design would allow drivers etc to prepare + * for the suspend and avoid conditions which prevent + * the suspend from succeeding. For now, we have this + * mitigation. + */ + pr_notice("Partition suspend attempt %u of %u error: %d\n", + attempt, max_attempts, ret); + + if (attempt == max_attempts) + break; + + vasi_err = poll_vasi_state(handle, &vasi_state); + if (vasi_err == 0) { + if (vasi_state != H_VASI_SUSPENDING) { + pr_notice("VASI state %lu after failed suspend\n", + vasi_state); + break; + } + } else if (vasi_err != -EOPNOTSUPP) { + pr_err("VASI state poll error: %d", vasi_err); + break; + } + + pr_notice("Will retry partition suspend after %u ms\n", + retry_interval_ms); + + msleep(retry_interval_ms); + retry_interval_ms *= 10; + attempt++; + } + + return ret; +} + +static int pseries_migrate_partition(u64 handle) +{ + int ret; + unsigned int factor = 0; + +#ifdef CONFIG_PPC_WATCHDOG + factor = nmi_wd_lpm_factor; +#endif + /* + * When the migration is initiated, the hypervisor changes VAS + * mappings to prepare before OS gets the notification and + * closes all VAS windows. NX generates continuous faults during + * this time and the user space can not differentiate these + * faults from the migration event. So reduce this time window + * by closing VAS windows at the beginning of this function. + */ + vas_migration_handler(VAS_SUSPEND); + + ret = wait_for_vasi_session_suspending(handle); + if (ret) + goto out; + + if (factor) + watchdog_nmi_set_timeout_pct(factor); + + ret = pseries_suspend(handle); + if (ret == 0) { + post_mobility_fixup(); + /* + * Wait until the memory transfer is complete, so that the user + * space process returns from the syscall after the transfer is + * complete. This allows the user hooks to be executed at the + * right time. + */ + wait_for_vasi_session_completed(handle); + } else + pseries_cancel_migration(handle, ret); + + if (factor) + watchdog_nmi_set_timeout_pct(0); + +out: + vas_migration_handler(VAS_RESUME); + + return ret; +} + +int rtas_syscall_dispatch_ibm_suspend_me(u64 handle) +{ + return pseries_migrate_partition(handle); +} + static ssize_t migration_store(struct class *class, struct class_attribute *attr, const char *buf, size_t count) @@ -385,21 +795,10 @@ static ssize_t migration_store(struct class *class, if (rc) return rc; - stop_topology_update(); - - do { - rc = rtas_ibm_suspend_me(streamid); - if (rc == -EAGAIN) - ssleep(1); - } while (rc == -EAGAIN); - + rc = pseries_migrate_partition(streamid); if (rc) return rc; - post_mobility_fixup(); - - start_topology_update(); - return count; } @@ -424,11 +823,11 @@ static int __init mobility_sysfs_init(void) rc = sysfs_create_file(mobility_kobj, &class_attr_migration.attr); if (rc) - pr_err("mobility: unable to create migration sysfs file (%d)\n", rc); + pr_err("unable to create migration sysfs file (%d)\n", rc); rc = sysfs_create_file(mobility_kobj, &class_attr_api_version.attr.attr); if (rc) - pr_err("mobility: unable to create api_version sysfs file (%d)\n", rc); + pr_err("unable to create api_version sysfs file (%d)\n", rc); return 0; } |