From 5538eee216a194d47ae67fbcff80bfbcd95f3d69 Mon Sep 17 00:00:00 2001 From: Jinshan Xiong Date: Mon, 8 Aug 2011 19:39:22 -0700 Subject: [PATCH] LU-580: update mgc llog process code This patch does the following work: 1. fix a bug about setting mgc lock callback data; 2. define log instance types; 3. simplify and cleanup mgc_requeue_thread code Signed-off-by: Jinshan Xiong Change-Id: Ic8d25eda9f0a7884b4a8cbeccc89c1e66c19f0ff Reviewed-on: http://review.whamcloud.com/1198 Tested-by: Hudson Tested-by: Maloo Reviewed-by: Oleg Drokin --- lustre/include/lustre_disk.h | 2 +- lustre/include/obd_class.h | 12 +- lustre/mgc/mgc_internal.h | 5 + lustre/mgc/mgc_request.c | 435 +++++++++++++++++++++++-------------------- lustre/mgs/mgs_handler.c | 2 +- 5 files changed, 253 insertions(+), 203 deletions(-) diff --git a/lustre/include/lustre_disk.h b/lustre/include/lustre_disk.h index 10be502..6fb49d8 100644 --- a/lustre/include/lustre_disk.h +++ b/lustre/include/lustre_disk.h @@ -486,7 +486,7 @@ struct mgs_target_info; int server_mti_print(char *title, struct mgs_target_info *mti); /* mgc_request.c */ -int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id); +int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id, int type); #endif diff --git a/lustre/include/obd_class.h b/lustre/include/obd_class.h index 257ab2b..54d0a65 100644 --- a/lustre/include/obd_class.h +++ b/lustre/include/obd_class.h @@ -157,9 +157,14 @@ int class_config_parse_llog(struct llog_ctxt *ctxt, char *name, int class_config_dump_llog(struct llog_ctxt *ctxt, char *name, struct config_llog_instance *cfg); +enum { + CONFIG_T_CONFIG = 0, + CONFIG_T_SPTLRPC = 1, + CONFIG_T_MAX = 2 +}; + /* list of active configuration logs */ struct config_llog_data { - char *cld_logname; struct ldlm_res_id cld_resid; struct config_llog_instance cld_cfg; cfs_list_t cld_list_chain; @@ -167,10 +172,11 @@ struct config_llog_data { struct config_llog_data *cld_sptlrpc;/* depended sptlrpc log */ struct obd_export *cld_mgcexp; cfs_mutex_t cld_lock; + int cld_type; unsigned int cld_stopping:1, /* we were told to stop * watching */ - cld_lostlock:1, /* lock not requeued */ - cld_is_sptlrpc:1; + cld_lostlock:1; /* lock not requeued */ + char cld_logname[0]; }; struct lustre_profile { diff --git a/lustre/mgc/mgc_internal.h b/lustre/mgc/mgc_internal.h index 61ffe55..eb5f403 100644 --- a/lustre/mgc/mgc_internal.h +++ b/lustre/mgc/mgc_internal.h @@ -55,4 +55,9 @@ static void lprocfs_mgc_init_vars(struct lprocfs_static_vars *lvars) int mgc_process_log(struct obd_device *mgc, struct config_llog_data *cld); +static inline int cld_is_sptlrpc(struct config_llog_data *cld) +{ + return cld->cld_type == CONFIG_T_SPTLRPC; +} + #endif /* _MGC_INTERNAL_H */ diff --git a/lustre/mgc/mgc_request.c b/lustre/mgc/mgc_request.c index fd120ae..295b6c7 100644 --- a/lustre/mgc/mgc_request.c +++ b/lustre/mgc/mgc_request.c @@ -64,7 +64,8 @@ #include #include "mgc_internal.h" -static int mgc_name2resid(char *name, int len, struct ldlm_res_id *res_id) +static int mgc_name2resid(char *name, int len, struct ldlm_res_id *res_id, + int type) { __u64 resname = 0; @@ -78,24 +79,33 @@ static int mgc_name2resid(char *name, int len, struct ldlm_res_id *res_id) } memcpy(&resname, name, len); - memset(res_id, 0, sizeof(*res_id)); - /* Always use the same endianness for the resid */ + memset(res_id, 0, sizeof(*res_id)); res_id->name[0] = cpu_to_le64(resname); + /* XXX: unfortunately, sptlprc and config llog share one lock */ + switch(type) { + case CONFIG_T_CONFIG: + case CONFIG_T_SPTLRPC: + resname = 0; + break; + default: + LBUG(); + } + res_id->name[1] = cpu_to_le64(resname); CDEBUG(D_MGC, "log %s to resid "LPX64"/"LPX64" (%.8s)\n", name, res_id->name[0], res_id->name[1], (char *)&res_id->name[0]); return 0; } -int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id) +int mgc_fsname2resid(char *fsname, struct ldlm_res_id *res_id, int type) { /* fsname is at most 8 chars long, maybe contain "-". * e.g. "lustre", "SUN-000" */ - return mgc_name2resid(fsname, strlen(fsname), res_id); + return mgc_name2resid(fsname, strlen(fsname), res_id, type); } EXPORT_SYMBOL(mgc_fsname2resid); -int mgc_logname2resid(char *logname, struct ldlm_res_id *res_id) +int mgc_logname2resid(char *logname, struct ldlm_res_id *res_id, int type) { char *name_end; int len; @@ -105,19 +115,17 @@ int mgc_logname2resid(char *logname, struct ldlm_res_id *res_id) name_end = strrchr(logname, '-'); LASSERT(name_end); len = name_end - logname; - return mgc_name2resid(logname, len, res_id); + return mgc_name2resid(logname, len, res_id, type); } /********************** config llog list **********************/ static CFS_LIST_HEAD(config_llog_list); -static cfs_spinlock_t config_list_lock = CFS_SPIN_LOCK_UNLOCKED; +static cfs_spinlock_t config_list_lock = CFS_SPIN_LOCK_UNLOCKED; /* Take a reference to a config log */ static int config_log_get(struct config_llog_data *cld) { ENTRY; - if (cld->cld_stopping) - RETURN(1); cfs_atomic_inc(&cld->cld_refcount); CDEBUG(D_INFO, "log %s refs %d\n", cld->cld_logname, cfs_atomic_read(&cld->cld_refcount)); @@ -135,8 +143,7 @@ static void config_log_put(struct config_llog_data *cld) LASSERT(cfs_atomic_read(&cld->cld_refcount) > 0); /* spinlock to make sure no item with 0 refcount in the list */ - cfs_spin_lock(&config_list_lock); - if (unlikely(cfs_atomic_dec_and_test(&cld->cld_refcount))) { + if (cfs_atomic_dec_and_lock(&cld->cld_refcount, &config_list_lock)) { cfs_list_del(&cld->cld_list_chain); cfs_spin_unlock(&config_list_lock); @@ -144,14 +151,11 @@ static void config_log_put(struct config_llog_data *cld) if (cld->cld_sptlrpc) config_log_put(cld->cld_sptlrpc); - if (cld->cld_is_sptlrpc) + if (cld_is_sptlrpc(cld)) sptlrpc_conf_log_stop(cld->cld_logname); class_export_put(cld->cld_mgcexp); - OBD_FREE(cld->cld_logname, strlen(cld->cld_logname) + 1); - OBD_FREE(cld, sizeof(*cld)); - } else { - cfs_spin_unlock(&config_list_lock); + OBD_FREE(cld, sizeof(*cld) + strlen(cld->cld_logname) + 1); } EXIT; @@ -184,7 +188,7 @@ struct config_llog_data *config_log_find(char *logname, } if (found) { cfs_atomic_inc(&found->cld_refcount); - LASSERT(found->cld_stopping == 0 || found->cld_is_sptlrpc == 0); + LASSERT(found->cld_stopping == 0 || cld_is_sptlrpc(found) == 0); } cfs_spin_unlock(&config_list_lock); RETURN(found); @@ -193,7 +197,7 @@ struct config_llog_data *config_log_find(char *logname, static struct config_llog_data *do_config_log_add(struct obd_device *obd, char *logname, - unsigned int is_sptlrpc, + int type, struct config_llog_instance *cfg, struct super_block *sb) { @@ -204,14 +208,10 @@ struct config_llog_data *do_config_log_add(struct obd_device *obd, CDEBUG(D_MGC, "do adding config log %s:%p\n", logname, cfg ? cfg->cfg_instance : 0); - OBD_ALLOC(cld, sizeof(*cld)); + OBD_ALLOC(cld, sizeof(*cld) + strlen(logname) + 1); if (!cld) RETURN(ERR_PTR(-ENOMEM)); - OBD_ALLOC(cld->cld_logname, strlen(logname) + 1); - if (!cld->cld_logname) { - OBD_FREE(cld, sizeof(*cld)); - RETURN(ERR_PTR(-ENOMEM)); - } + strcpy(cld->cld_logname, logname); if (cfg) cld->cld_cfg = *cfg; @@ -219,18 +219,18 @@ struct config_llog_data *do_config_log_add(struct obd_device *obd, cld->cld_cfg.cfg_last_idx = 0; cld->cld_cfg.cfg_flags = 0; cld->cld_cfg.cfg_sb = sb; - cld->cld_is_sptlrpc = is_sptlrpc; + cld->cld_type = type; cfs_atomic_set(&cld->cld_refcount, 1); /* Keep the mgc around until we are done */ cld->cld_mgcexp = class_export_get(obd->obd_self_export); - if (is_sptlrpc) { + if (cld_is_sptlrpc(cld)) { sptlrpc_conf_log_start(logname); cld->cld_cfg.cfg_obdname = obd->obd_name; } - rc = mgc_logname2resid(logname, &cld->cld_resid); + rc = mgc_logname2resid(logname, &cld->cld_resid, type); cfs_spin_lock(&config_list_lock); cfs_list_add(&cld->cld_list_chain, &config_llog_list); @@ -241,7 +241,7 @@ struct config_llog_data *do_config_log_add(struct obd_device *obd, RETURN(ERR_PTR(rc)); } - if (is_sptlrpc) { + if (cld_is_sptlrpc(cld)) { rc = mgc_process_log(obd, cld); if (rc) CERROR("failed processing sptlrpc log: %d\n", rc); @@ -259,8 +259,9 @@ static int config_log_add(struct obd_device *obd, char *logname, struct config_llog_instance *cfg, struct super_block *sb) { - struct config_llog_data *cld, *sptlrpc_cld; - char seclogname[20]; + struct config_llog_data *cld; + struct config_llog_data *sptlrpc_cld; + char seclogname[32]; char *ptr; ENTRY; @@ -281,14 +282,15 @@ static int config_log_add(struct obd_device *obd, char *logname, sptlrpc_cld = config_log_find(seclogname, NULL); if (sptlrpc_cld == NULL) { - sptlrpc_cld = do_config_log_add(obd, seclogname, 1, NULL, NULL); + sptlrpc_cld = do_config_log_add(obd, seclogname, + CONFIG_T_SPTLRPC, NULL, NULL); if (IS_ERR(sptlrpc_cld)) { CERROR("can't create sptlrpc log: %s\n", seclogname); RETURN(PTR_ERR(sptlrpc_cld)); } } - cld = do_config_log_add(obd, logname, 0, cfg, sb); + cld = do_config_log_add(obd, logname, CONFIG_T_CONFIG, cfg, sb); if (IS_ERR(cld)) { CERROR("can't create log: %s\n", logname); config_log_put(sptlrpc_cld); @@ -355,13 +357,13 @@ static int config_log_end(char *logname, struct config_llog_instance *cfg) #define RQ_NOW 0x2 #define RQ_LATER 0x4 #define RQ_STOP 0x8 -static int rq_state = 0; -static cfs_waitq_t rq_waitq; - -static int mgc_requeue_add(struct config_llog_data *cld, int later); +static int rq_state = 0; +static cfs_waitq_t rq_waitq; +static CFS_DECLARE_COMPLETION(rq_exit); static void do_requeue(struct config_llog_data *cld) { + ENTRY; LASSERT(cfs_atomic_read(&cld->cld_refcount) > 0); /* Do not run mgc_process_log on a disconnected export or an @@ -377,15 +379,18 @@ static void do_requeue(struct config_llog_data *cld) } cfs_up_read(&cld->cld_mgcexp->exp_obd->u.cli.cl_sem); - /* Whether we enqueued again or not in mgc_process_log, we're done - * with the ref from the old enqueue */ - config_log_put(cld); + EXIT; } +/* this timeout represents how many seconds MGC should wait before + * requeue config and recover lock to the MGS. We need to randomize this + * in order to not flood the MGS. + */ +#define MGC_TIMEOUT_MIN_SECONDS 5 +#define MGC_TIMEOUT_RAND_CENTISEC 0x1ff /* ~500 */ + static int mgc_requeue_thread(void *data) { - struct l_wait_info lwi_now, lwi_later; - struct config_llog_data *cld, *cld_next, *cld_prev; char name[] = "ll_cfg_requeue"; int rc = 0; ENTRY; @@ -394,12 +399,16 @@ static int mgc_requeue_thread(void *data) CDEBUG(D_MGC, "Starting requeue thread\n"); - lwi_later = LWI_TIMEOUT(60 * CFS_HZ, NULL, NULL); - l_wait_event(rq_waitq, rq_state & (RQ_NOW | RQ_STOP), &lwi_later); - /* Keep trying failed locks periodically */ cfs_spin_lock(&config_list_lock); - while (rq_state & (RQ_NOW | RQ_LATER)) { + rq_state |= RQ_RUNNING; + while (1) { + struct l_wait_info lwi; + struct config_llog_data *cld, *cld_prev; + int rand = cfs_rand() & MGC_TIMEOUT_RAND_CENTISEC; + int stopped = !!(rq_state & RQ_STOP); + int to; + /* Any new or requeued lostlocks will change the state */ rq_state &= ~(RQ_NOW | RQ_LATER); cfs_spin_unlock(&config_list_lock); @@ -407,10 +416,10 @@ static int mgc_requeue_thread(void *data) /* Always wait a few seconds to allow the server who caused the lock revocation to finish its setup, plus some random so everyone doesn't try to reconnect at once. */ - lwi_now = LWI_TIMEOUT(3 * CFS_HZ + (cfs_rand() & 0xff) * \ - (CFS_HZ / 100), - NULL, NULL); - l_wait_event(rq_waitq, rq_state & RQ_STOP, &lwi_now); + to = MGC_TIMEOUT_MIN_SECONDS * CFS_HZ; + to += rand * CFS_HZ / 100; /* rand is centi-seconds */ + lwi = LWI_TIMEOUT(to, NULL, NULL); + l_wait_event(rq_waitq, rq_state & RQ_STOP, &lwi); /* * iterate & processing through the list. for each cld, process @@ -419,106 +428,91 @@ static int mgc_requeue_thread(void *data) * it's guaranteed any item in the list must have * reference > 0; and if cld_lostlock is set, at * least one reference is taken by the previous enqueue. - * - * Note: releasing a cld might lead to itself and its depended - * sptlrpc cld be unlinked from the list. to safely iterate - * we need to take a reference on next cld before processing. */ cld_prev = NULL; cfs_spin_lock(&config_list_lock); - cfs_list_for_each_entry_safe(cld, cld_next, &config_llog_list, - cld_list_chain) { - if (cld->cld_list_chain.next != &config_llog_list) - cfs_atomic_inc(&cld_next->cld_refcount); + cfs_list_for_each_entry(cld, &config_llog_list, + cld_list_chain) { + if (!cld->cld_lostlock) + continue; - if (cld->cld_lostlock) { - if (cld->cld_sptlrpc && - cld->cld_sptlrpc->cld_lostlock) { - cld->cld_sptlrpc->cld_lostlock = 0; + cfs_spin_unlock(&config_list_lock); - cfs_spin_unlock(&config_list_lock); - do_requeue(cld->cld_sptlrpc); - cfs_spin_lock(&config_list_lock); - LASSERT(cld->cld_lostlock); - } + LASSERT(cfs_atomic_read(&cld->cld_refcount) > 0); - cld->cld_lostlock = 0; + /* Whether we enqueued again or not in mgc_process_log, + * we're done with the ref from the old enqueue */ + if (cld_prev) + config_log_put(cld_prev); + cld_prev = cld; - cfs_spin_unlock(&config_list_lock); + cld->cld_lostlock = 0; + if (likely(!stopped)) do_requeue(cld); - cfs_spin_lock(&config_list_lock); - } - - if (cld_prev) { - cfs_spin_unlock(&config_list_lock); - config_log_put(cld_prev); - cfs_spin_lock(&config_list_lock); - } - - cld_prev = cld_next; + cfs_spin_lock(&config_list_lock); } cfs_spin_unlock(&config_list_lock); + if (cld_prev) + config_log_put(cld_prev); + + /* break after scanning the list so that we can drop + * refcount to losing lock clds */ + if (unlikely(stopped)) { + cfs_spin_lock(&config_list_lock); + break; + } /* Wait a bit to see if anyone else needs a requeue */ + lwi = (struct l_wait_info) { 0 }; l_wait_event(rq_waitq, rq_state & (RQ_NOW | RQ_STOP), - &lwi_later); + &lwi); cfs_spin_lock(&config_list_lock); } /* spinlock and while guarantee RQ_NOW and RQ_LATER are not set */ rq_state &= ~RQ_RUNNING; cfs_spin_unlock(&config_list_lock); + cfs_complete(&rq_exit); + CDEBUG(D_MGC, "Ending requeue thread\n"); RETURN(rc); } /* Add a cld to the list to requeue. Start the requeue thread if needed. We are responsible for dropping the config log reference from here on out. */ -static int mgc_requeue_add(struct config_llog_data *cld, int later) +static void mgc_requeue_add(struct config_llog_data *cld) { - int rc = 0; + ENTRY; - CDEBUG(D_INFO, "log %s: requeue (l=%d r=%d sp=%d st=%x)\n", - cld->cld_logname, later, cfs_atomic_read(&cld->cld_refcount), + CDEBUG(D_INFO, "log %s: requeue (r=%d sp=%d st=%x)\n", + cld->cld_logname, cfs_atomic_read(&cld->cld_refcount), cld->cld_stopping, rq_state); LASSERT(cfs_atomic_read(&cld->cld_refcount) > 0); - /* Hold lock for rq_state */ - cfs_spin_lock(&config_list_lock); - - if (cld->cld_stopping || (rq_state & RQ_STOP)) { - cld->cld_lostlock = 0; - cfs_spin_unlock(&config_list_lock); - config_log_put(cld); - RETURN(0); + cfs_mutex_lock(&cld->cld_lock); + if (cld->cld_stopping || cld->cld_lostlock) { + cfs_mutex_unlock(&cld->cld_lock); + RETURN_EXIT; } - + /* this refcount will be released in mgc_requeue_thread. */ + config_log_get(cld); cld->cld_lostlock = 1; + cfs_mutex_unlock(&cld->cld_lock); - if (!(rq_state & RQ_RUNNING)) { - LASSERT(rq_state == 0); - rq_state = RQ_RUNNING | (later ? RQ_LATER : RQ_NOW); + /* Hold lock for rq_state */ + cfs_spin_lock(&config_list_lock); + if (rq_state & RQ_STOP) { cfs_spin_unlock(&config_list_lock); - rc = cfs_create_thread(mgc_requeue_thread, NULL, - CFS_DAEMON_FLAGS); - if (rc < 0) { - CERROR("log %s: cannot start requeue thread (%d)," - "no more log updates!\n", cld->cld_logname, rc); - /* Drop the ref, since the rq thread won't */ - cld->cld_lostlock = 0; - config_log_put(cld); - rq_state = 0; - RETURN(rc); - } + cld->cld_lostlock = 0; + config_log_put(cld); } else { - rq_state |= later ? RQ_LATER : RQ_NOW; + rq_state |= RQ_NOW; cfs_spin_unlock(&config_list_lock); cfs_waitq_signal(&rq_waitq); } - - RETURN(0); + EXIT; } /********************** class fns **********************/ @@ -626,12 +620,17 @@ static int mgc_precleanup(struct obd_device *obd, enum obd_cleanup_stage stage) break; case OBD_CLEANUP_EXPORTS: if (cfs_atomic_dec_and_test(&mgc_count)) { - /* Kick the requeue waitq - cld's should all be - stopping */ + int running; + /* stop requeue thread */ cfs_spin_lock(&config_list_lock); - rq_state |= RQ_STOP; + running = rq_state & RQ_RUNNING; + if (running) + rq_state |= RQ_STOP; cfs_spin_unlock(&config_list_lock); - cfs_waitq_signal(&rq_waitq); + if (running) { + cfs_waitq_signal(&rq_waitq); + cfs_wait_for_completion(&rq_exit); + } } obd_cleanup_client_import(obd); rc = obd_llog_finish(obd, 0); @@ -685,13 +684,22 @@ static int mgc_setup(struct obd_device *obd, struct lustre_cfg *lcfg) lprocfs_obd_setup(obd, lvars.obd_vars); sptlrpc_lprocfs_cliobd_attach(obd); - cfs_spin_lock(&config_list_lock); - cfs_atomic_inc(&mgc_count); - if (cfs_atomic_read(&mgc_count) == 1) { - rq_state &= ~RQ_STOP; + if (cfs_atomic_inc_return(&mgc_count) == 1) { + rq_state = 0; cfs_waitq_init(&rq_waitq); + + /* start requeue thread */ + rc = cfs_create_thread(mgc_requeue_thread, NULL, + CFS_DAEMON_FLAGS); + if (rc < 0) { + CERROR("%s: Cannot start requeue thread (%d)," + "no more log updates!\n", + obd->obd_name, rc); + GOTO(err_cleanup, rc); + } + /* rc is the pid of mgc_requeue_thread. */ + rc = 0; } - cfs_spin_unlock(&config_list_lock); RETURN(rc); @@ -718,7 +726,7 @@ static int mgc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, ldlm_lock2handle(lock, &lockh); rc = ldlm_cli_cancel(&lockh); break; - case LDLM_CB_CANCELING: { + case LDLM_CB_CANCELING: /* We've given up the lock, prepare ourselves to update. */ LDLM_DEBUG(lock, "MGC cancel CB"); @@ -727,9 +735,12 @@ static int mgc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, (char *)&lock->l_resource->lr_name.name[0]); if (!cld) { - CERROR("missing data, won't requeue\n"); + CDEBUG(D_INFO, "missing data, won't requeue\n"); break; } + + /* held at mgc_process_log(). */ + LASSERT(cfs_atomic_read(&cld->cld_refcount) > 0); /* Are we done with this log? */ if (cld->cld_stopping) { CDEBUG(D_MGC, "log %s: stopping, won't requeue\n", @@ -741,34 +752,20 @@ static int mgc_blocking_ast(struct ldlm_lock *lock, struct ldlm_lock_desc *desc, (we get called from client_disconnect_export) */ if (!lock->l_conn_export || !lock->l_conn_export->exp_obd->u.cli.cl_conn_count) { - CDEBUG(D_MGC, "log %s: disconnecting, won't requeue\n", + CDEBUG(D_MGC, "log %.8s: disconnecting, won't requeue\n", cld->cld_logname); config_log_put(cld); break; } - /* Did we fail to get the lock? */ - if (lock->l_req_mode != lock->l_granted_mode && - !cld->cld_is_sptlrpc) { - CDEBUG(D_MGC, "log %s: original grant failed, will " - "requeue later\n", cld->cld_logname); - /* Try to re-enqueue later */ - rc = mgc_requeue_add(cld, 1); - break; - } + /* Re-enqueue now */ - rc = mgc_requeue_add(cld, 0); + mgc_requeue_add(cld); + config_log_put(cld); break; - } default: LBUG(); } - - if (rc) { - CERROR("%s CB failed %d:\n", flag == LDLM_CB_BLOCKING ? - "blocking" : "cancel", rc); - LDLM_ERROR(lock, "MGC ast"); - } RETURN(rc); } @@ -823,19 +820,15 @@ static int mgc_enqueue(struct obd_export *exp, struct lov_stripe_md *lsm, { struct config_llog_data *cld = (struct config_llog_data *)data; struct ldlm_enqueue_info einfo = { type, mode, mgc_blocking_ast, - ldlm_completion_ast, NULL, NULL, data}; + ldlm_completion_ast, NULL, NULL, NULL }; struct ptlrpc_request *req; - int short_limit = cld->cld_is_sptlrpc; + int short_limit = cld_is_sptlrpc(cld); int rc; ENTRY; CDEBUG(D_MGC, "Enqueue for %s (res "LPX64")\n", cld->cld_logname, cld->cld_resid.name[0]); - /* We can only drop this config log ref when we drop the lock */ - if (config_log_get(cld)) - RETURN(ELDLM_LOCK_ABORTED); - /* We need a callback for every lockholder, so don't try to ldlm_lock_match (see rev 1.1.2.11.2.47) */ req = ptlrpc_request_alloc_pack(class_exp2cliimp(exp), @@ -844,6 +837,7 @@ static int mgc_enqueue(struct obd_export *exp, struct lov_stripe_md *lsm, if (req == NULL) RETURN(-ENOMEM); ptlrpc_request_set_replen(req); + /* check if this is server or client */ if (cld->cld_cfg.cfg_sb) { struct lustre_sb_info *lsi = s2lsi(cld->cld_cfg.cfg_sb); @@ -870,6 +864,17 @@ static int mgc_cancel(struct obd_export *exp, struct lov_stripe_md *md, RETURN(0); } +static void mgc_notify_active(struct obd_device *unused) +{ + /* wakeup mgc_requeue_thread to requeue mgc lock */ + cfs_spin_lock(&config_list_lock); + rq_state |= RQ_NOW; + cfs_spin_unlock(&config_list_lock); + cfs_waitq_signal(&rq_waitq); + + /* TODO: Help the MGS rebuild nidtbl. -jay */ +} + /* Send target_reg message to MGS */ static int mgc_target_register(struct obd_export *exp, struct mgs_target_info *mti) @@ -1045,6 +1050,7 @@ static int mgc_import_event(struct obd_device *obd, LCONSOLE_WARN("%s: Reactivating import\n", obd->obd_name); /* Clearing obd_no_recov allows us to continue pinging */ obd->obd_no_recov = 0; + mgc_notify_active(obd); break; case IMP_EVENT_OCD: break; @@ -1236,71 +1242,45 @@ out: RETURN(rc); } -/** Get a config log from the MGS and process it. - * This func is called for both clients and servers. - * Copy the log locally before parsing it if appropriate (non-MGS server) - */ -int mgc_process_log(struct obd_device *mgc, - struct config_llog_data *cld) +/* local_only means it cannot get remote llogs */ +static int mgc_process_cfg_log(struct obd_device *mgc, + struct config_llog_data *cld, + int local_only) { - struct llog_ctxt *ctxt, *lctxt; - struct lustre_handle lockh; + struct llog_ctxt *ctxt, *lctxt = NULL; struct client_obd *cli = &mgc->u.cli; struct lvfs_run_ctxt *saved_ctxt; struct lustre_sb_info *lsi = NULL; - int rc = 0, rcl, flags = 0, must_pop = 0; + int rc = 0, must_pop = 0; + bool sptlrpc_started = false; + ENTRY; LASSERT(cld); + LASSERT(cfs_mutex_is_locked(&cld->cld_lock)); - /* I don't want multiple processes running process_log at once -- - sounds like badness. It actually might be fine, as long as - we're not trying to update from the same log - simultaneously (in which case we should use a per-log sem.) */ - cfs_mutex_lock(&cld->cld_lock); - - if (cld->cld_stopping) { - cfs_mutex_unlock(&cld->cld_lock); + /* + * local copy of sptlrpc log is controlled elsewhere, don't try to + * read it up here. + */ + if (cld_is_sptlrpc(cld) && local_only) RETURN(0); - } - - OBD_FAIL_TIMEOUT(OBD_FAIL_MGC_PAUSE_PROCESS_LOG, 20); if (cld->cld_cfg.cfg_sb) lsi = s2lsi(cld->cld_cfg.cfg_sb); - CDEBUG(D_MGC, "Process log %s:%p from %d\n", cld->cld_logname, - cld->cld_cfg.cfg_instance, cld->cld_cfg.cfg_last_idx + 1); - ctxt = llog_get_context(mgc, LLOG_CONFIG_REPL_CTXT); if (!ctxt) { CERROR("missing llog context\n"); - cfs_mutex_unlock(&cld->cld_lock); RETURN(-EINVAL); } OBD_ALLOC_PTR(saved_ctxt); - if (saved_ctxt == NULL) { - cfs_mutex_unlock(&cld->cld_lock); + if (saved_ctxt == NULL) RETURN(-ENOMEM); - } - - /* Get the cfg lock on the llog */ - rcl = mgc_enqueue(mgc->u.cli.cl_mgc_mgsexp, NULL, LDLM_PLAIN, NULL, - LCK_CR, &flags, NULL, NULL, NULL, - cld, 0, NULL, &lockh); - if (rcl) - CDEBUG(D_MGC, "Can't get cfg lock: %d\n", rcl); lctxt = llog_get_context(mgc, LLOG_CONFIG_ORIG_CTXT); - /* - * local copy of sptlrpc log is controlled elsewhere, don't try to - * read it up here. - */ - if (rcl && cld->cld_is_sptlrpc) - GOTO(out_pop, rc); - /* Copy the setup log locally if we can. Don't mess around if we're running an MGS though (logs are already local). */ if (lctxt && lsi && (lsi->lsi_flags & LSI_SERVER) && @@ -1308,10 +1288,10 @@ int mgc_process_log(struct obd_device *mgc, !IS_MGS(lsi->lsi_ldd)) { push_ctxt(saved_ctxt, &mgc->obd_lvfs_ctxt, NULL); must_pop++; - if (rcl == 0) + if (!local_only) /* Only try to copy log if we have the lock. */ rc = mgc_copy_llog(mgc, ctxt, lctxt, cld->cld_logname); - if (rcl || rc) { + if (local_only || rc) { if (mgc_llog_is_empty(mgc, lctxt, cld->cld_logname)) { LCONSOLE_ERROR_MSG(0x13a, "Failed to get MGS " "log %s and no local copy." @@ -1327,19 +1307,25 @@ int mgc_process_log(struct obd_device *mgc, log has. */ llog_ctxt_put(ctxt); ctxt = lctxt; + lctxt = NULL; + } else if (local_only) { /* no local log at client side */ + GOTO(out_pop, rc = -EIO); } - if (cld->cld_is_sptlrpc) + if (cld_is_sptlrpc(cld)) { sptlrpc_conf_log_update_begin(cld->cld_logname); + sptlrpc_started = true; + } /* logname and instance info should be the same, so use our copy of the instance for the update. The cfg_last_idx will be updated here. */ - if (rcl == 0 || lctxt == ctxt) - rc = class_config_parse_llog(ctxt, cld->cld_logname, &cld->cld_cfg); + rc = class_config_parse_llog(ctxt, cld->cld_logname, &cld->cld_cfg); + EXIT; + out_pop: llog_ctxt_put(ctxt); - if (ctxt != lctxt) + if (lctxt) llog_ctxt_put(lctxt); if (must_pop) pop_ctxt(saved_ctxt, &mgc->obd_lvfs_ctxt, NULL); @@ -1351,13 +1337,70 @@ out_pop: * in parallel. * the logname must be -sptlrpc */ - if (cld->cld_is_sptlrpc && rcl == 0) { + if (sptlrpc_started) { + LASSERT(cld_is_sptlrpc(cld)); sptlrpc_conf_log_update_end(cld->cld_logname); class_notify_sptlrpc_conf(cld->cld_logname, strlen(cld->cld_logname) - strlen("-sptlrpc")); } + RETURN(rc); +} + +/** Get a config log from the MGS and process it. + * This func is called for both clients and servers. + * Copy the log locally before parsing it if appropriate (non-MGS server) + */ +int mgc_process_log(struct obd_device *mgc, struct config_llog_data *cld) +{ + struct lustre_handle lockh = { 0 }; + int rc = 0, rcl, flags = 0; + ENTRY; + + LASSERT(cld); + + /* I don't want multiple processes running process_log at once -- + sounds like badness. It actually might be fine, as long as + we're not trying to update from the same log + simultaneously (in which case we should use a per-log sem.) */ + cfs_mutex_lock(&cld->cld_lock); + if (cld->cld_stopping) { + cfs_mutex_unlock(&cld->cld_lock); + RETURN(0); + } + + OBD_FAIL_TIMEOUT(OBD_FAIL_MGC_PAUSE_PROCESS_LOG, 20); + + CDEBUG(D_MGC, "Process log %s:%p from %d\n", cld->cld_logname, + cld->cld_cfg.cfg_instance, cld->cld_cfg.cfg_last_idx + 1); + + /* Get the cfg lock on the llog */ + rcl = mgc_enqueue(mgc->u.cli.cl_mgc_mgsexp, NULL, LDLM_PLAIN, NULL, + LCK_CR, &flags, NULL, NULL, NULL, + cld, 0, NULL, &lockh); + if (rcl == 0) { + /* Get the cld, it will be released in mgc_blocking_ast. */ + config_log_get(cld); + rc = ldlm_lock_set_data(&lockh, (void *)cld); + LASSERT(rc == 0); + } else { + CDEBUG(D_MGC, "Can't get cfg lock: %d\n", rcl); + + /* mark cld_lostlock so that it will requeue + * after MGC becomes available. */ + cld->cld_lostlock = 1; + /* Get extra reference, it will be put in requeue thread */ + config_log_get(cld); + } + + rc = mgc_process_cfg_log(mgc, cld, rcl != 0); + + CDEBUG(D_MGC, "%s: configuration from log '%s' %sed (%d).\n", + mgc->obd_name, cld->cld_logname, rc ? "fail" : "succeed", rc); + + cfs_mutex_unlock(&cld->cld_lock); + /* Now drop the lock so MGS can revoke it */ if (!rcl) { rcl = mgc_cancel(mgc->u.cli.cl_mgc_mgsexp, NULL, @@ -1366,14 +1409,10 @@ out_pop: CERROR("Can't drop cfg lock: %d\n", rcl); } - CDEBUG(D_MGC, "%s: configuration from log '%s' %sed (%d).\n", - mgc->obd_name, cld->cld_logname, rc ? "fail" : "succeed", rc); - - cfs_mutex_unlock(&cld->cld_lock); - RETURN(rc); } + /** Called from lustre_process_log. * LCFG_LOG_START gets the config log from the MGS, processes it to start * any services, and adds it to the list logs to watch (follow). diff --git a/lustre/mgs/mgs_handler.c b/lustre/mgs/mgs_handler.c index d1e4b65..4829c1f 100644 --- a/lustre/mgs/mgs_handler.c +++ b/lustre/mgs/mgs_handler.c @@ -332,7 +332,7 @@ static int mgs_get_cfg_lock(struct obd_device *obd, char *fsname, int rc, flags = 0; ENTRY; - rc = mgc_fsname2resid(fsname, &res_id); + rc = mgc_fsname2resid(fsname, &res_id, CONFIG_T_CONFIG); if (!rc) rc = ldlm_cli_enqueue_local(obd->obd_namespace, &res_id, LDLM_PLAIN, NULL, LCK_EX, -- 1.8.3.1