X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fmgs%2Fmgs_nids.c;h=4420b64c7ceccca7d454c66a953c1e2e24ddfc06;hp=4f468814b68db05fc3e6f25ce75e50b8d8937adf;hb=f8b4e51fb85c078ae48c3c14472c520e850a2a2d;hpb=1b2547843817b4b7adbeb87ea9b070d9cac35c90 diff --git a/lustre/mgs/mgs_nids.c b/lustre/mgs/mgs_nids.c index 4f46881..4420b64 100644 --- a/lustre/mgs/mgs_nids.c +++ b/lustre/mgs/mgs_nids.c @@ -27,7 +27,7 @@ * Copyright (c) 2007, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2011, 2012, Intel Corporation. + * Copyright (c) 2011, 2015, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -43,9 +43,8 @@ #define DEBUG_SUBSYSTEM S_MGS #define D_MGS D_CONFIG -#ifdef __KERNEL__ +#include #include -#endif #include #include @@ -57,20 +56,20 @@ static unsigned int ir_timeout; static int nidtbl_is_sane(struct mgs_nidtbl *tbl) { - struct mgs_nidtbl_target *tgt; - int version = 0; + struct mgs_nidtbl_target *tgt; + int version = 0; LASSERT(mutex_is_locked(&tbl->mn_lock)); - cfs_list_for_each_entry(tgt, &tbl->mn_targets, mnt_list) { - if (!tgt->mnt_version) - continue; + list_for_each_entry(tgt, &tbl->mn_targets, mnt_list) { + if (!tgt->mnt_version) + continue; - if (version >= tgt->mnt_version) - return 0; + if (version >= tgt->mnt_version) + return 0; - version = tgt->mnt_version; - } - return 1; + version = tgt->mnt_version; + } + return 1; } /** @@ -113,7 +112,7 @@ static int mgs_nidtbl_read(struct obd_export *exp, struct mgs_nidtbl *tbl, * otherwise, it's for clients, then llog entries for both OSTs and * MDTs will be returned. */ - cfs_list_for_each_entry(tgt, &tbl->mn_targets, mnt_list) { + list_for_each_entry(tgt, &tbl->mn_targets, mnt_list) { int entry_len = sizeof(*entry); if (tgt->mnt_version < version) @@ -136,27 +135,18 @@ static int mgs_nidtbl_read(struct obd_export *exp, struct mgs_nidtbl *tbl, break; } - /* check if we need to consume remaining bytes. */ - if (last_in_unit != NULL && bytes_in_unit) { -#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 6, 50, 0) - /* May need to swab back to update the length.*/ - if (exp->exp_need_mne_swab) - lustre_swab_mgs_nidtbl_entry(last_in_unit); -#endif + /* check if we need to consume remaining bytes. */ + if (last_in_unit != NULL && bytes_in_unit) { last_in_unit->mne_length += bytes_in_unit; -#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 6, 50, 0) - if (exp->exp_need_mne_swab) - lustre_swab_mgs_nidtbl_entry(last_in_unit); -#endif - rc += bytes_in_unit; - buf += bytes_in_unit; - last_in_unit = NULL; - } - LASSERT((rc & (unit_size - 1)) == 0); + rc += bytes_in_unit; + buf += bytes_in_unit; + last_in_unit = NULL; + } + LASSERT((rc & (unit_size - 1)) == 0); if (units_in_page == 0) { /* allocate a new page */ - pages[index] = alloc_page(GFP_IOFS); + pages[index] = alloc_page(GFP_KERNEL); if (pages[index] == NULL) { rc = -ENOMEM; break; @@ -190,19 +180,13 @@ static int mgs_nidtbl_read(struct obd_export *exp, struct mgs_nidtbl *tbl, entry->mne_type = tgt->mnt_type; entry->mne_nid_type = 0; entry->mne_nid_size = sizeof(lnet_nid_t); - entry->mne_nid_count = mti->mti_nid_count; - memcpy(entry->u.nids, mti->mti_nids, - mti->mti_nid_count * sizeof(lnet_nid_t)); - -#if LUSTRE_VERSION_CODE < OBD_OCD_VERSION(2, 6, 50, 0) - /* For LU-1644, swab entry for 2.2 clients. */ - if (exp->exp_need_mne_swab) - lustre_swab_mgs_nidtbl_entry(entry); -#endif + entry->mne_nid_count = mti->mti_nid_count; + memcpy(entry->u.nids, mti->mti_nids, + mti->mti_nid_count * sizeof(lnet_nid_t)); - version = tgt->mnt_version; - rc += entry_len; - buf += entry_len; + version = tgt->mnt_version; + rc += entry_len; + buf += entry_len; bytes_in_unit -= entry_len; last_in_unit = entry; @@ -253,7 +237,7 @@ static int nidtbl_update_version(const struct lu_env *env, GOTO(out_put, rc = PTR_ERR(th)); th->th_sync = 1; /* update table synchronously */ - rc = dt_declare_record_write(env, fsdb, buf.lb_len, off, th); + rc = dt_declare_record_write(env, fsdb, &buf, off, th); if (rc) GOTO(out, rc); @@ -301,7 +285,7 @@ static int nidtbl_read_version(const struct lu_env *env, } fsdb = dt_locate_at(env, mgs->mgs_bottom, &fid, - &mgs->mgs_dt_dev.dd_lu_dev); + &mgs->mgs_dt_dev.dd_lu_dev, NULL); if (IS_ERR(fsdb)) RETURN(PTR_ERR(fsdb)); @@ -332,10 +316,11 @@ static int mgs_nidtbl_write(const struct lu_env *env, struct fs_db *fsdb, type &= ~LDD_F_SV_TYPE_MGS; LASSERT(type != 0); - tbl = &fsdb->fsdb_nidtbl; + tbl = &fsdb->fsdb_nidtbl; mutex_lock(&tbl->mn_lock); - cfs_list_for_each_entry(tgt, &tbl->mn_targets, mnt_list) { - struct mgs_target_info *info = &tgt->mnt_mti; + list_for_each_entry(tgt, &tbl->mn_targets, mnt_list) { + struct mgs_target_info *info = &tgt->mnt_mti; + if (type == tgt->mnt_type && mti->mti_stripe_index == info->mti_stripe_index) { found = true; @@ -347,21 +332,21 @@ static int mgs_nidtbl_write(const struct lu_env *env, struct fs_db *fsdb, if (tgt == NULL) GOTO(out, rc = -ENOMEM); - CFS_INIT_LIST_HEAD(&tgt->mnt_list); - tgt->mnt_fs = tbl; - tgt->mnt_version = 0; /* 0 means invalid */ - tgt->mnt_type = type; + INIT_LIST_HEAD(&tgt->mnt_list); + tgt->mnt_fs = tbl; + tgt->mnt_version = 0; /* 0 means invalid */ + tgt->mnt_type = type; - ++tbl->mn_nr_targets; - } + ++tbl->mn_nr_targets; + } - tgt->mnt_version = ++tbl->mn_version; - tgt->mnt_mti = *mti; + tgt->mnt_version = ++tbl->mn_version; + tgt->mnt_mti = *mti; - cfs_list_move_tail(&tgt->mnt_list, &tbl->mn_targets); + list_move_tail(&tgt->mnt_list, &tbl->mn_targets); rc = nidtbl_update_version(env, fsdb->fsdb_mgs, tbl); - EXIT; + EXIT; out: mutex_unlock(&tbl->mn_lock); @@ -374,17 +359,17 @@ out: static void mgs_nidtbl_fini_fs(struct fs_db *fsdb) { struct mgs_nidtbl *tbl = &fsdb->fsdb_nidtbl; - CFS_LIST_HEAD(head); + struct list_head head = LIST_HEAD_INIT(head); mutex_lock(&tbl->mn_lock); tbl->mn_nr_targets = 0; - cfs_list_splice_init(&tbl->mn_targets, &head); + list_splice_init(&tbl->mn_targets, &head); mutex_unlock(&tbl->mn_lock); - while (!cfs_list_empty(&head)) { + while (!list_empty(&head)) { struct mgs_nidtbl_target *tgt; tgt = list_entry(head.next, struct mgs_nidtbl_target, mnt_list); - cfs_list_del(&tgt->mnt_list); + list_del(&tgt->mnt_list); OBD_FREE_PTR(tgt); } } @@ -394,7 +379,7 @@ static int mgs_nidtbl_init_fs(const struct lu_env *env, struct fs_db *fsdb) struct mgs_nidtbl *tbl = &fsdb->fsdb_nidtbl; int rc; - CFS_INIT_LIST_HEAD(&tbl->mn_targets); + INIT_LIST_HEAD(&tbl->mn_targets); mutex_init(&tbl->mn_lock); tbl->mn_nr_targets = 0; tbl->mn_fsdb = fsdb; @@ -414,10 +399,10 @@ static int mgs_nidtbl_init_fs(const struct lu_env *env, struct fs_db *fsdb) /* --------- Imperative Recovery relies on nidtbl stuff ------- */ void mgs_ir_notify_complete(struct fs_db *fsdb) { - struct timeval tv; - cfs_duration_t delta; + struct timeval tv; + cfs_duration_t delta; - cfs_atomic_set(&fsdb->fsdb_notify_phase, 0); + atomic_set(&fsdb->fsdb_notify_phase, 0); /* do statistic */ fsdb->fsdb_notify_count++; @@ -445,47 +430,47 @@ static int mgs_ir_notify(void *arg) set_user_nice(current, -2); - mgc_fsname2resid(fsdb->fsdb_name, &resid, CONFIG_T_RECOVER); - while (1) { - struct l_wait_info lwi = { 0 }; + mgc_fsname2resid(fsdb->fsdb_name, &resid, CONFIG_T_RECOVER); + while (1) { + struct l_wait_info lwi = { 0 }; - l_wait_event(fsdb->fsdb_notify_waitq, - fsdb->fsdb_notify_stop || - cfs_atomic_read(&fsdb->fsdb_notify_phase), - &lwi); - if (fsdb->fsdb_notify_stop) - break; + l_wait_event(fsdb->fsdb_notify_waitq, + fsdb->fsdb_notify_stop || + atomic_read(&fsdb->fsdb_notify_phase), + &lwi); + if (fsdb->fsdb_notify_stop) + break; - CDEBUG(D_MGS, "%s woken up, phase is %d\n", - name, cfs_atomic_read(&fsdb->fsdb_notify_phase)); + CDEBUG(D_MGS, "%s woken up, phase is %d\n", + name, atomic_read(&fsdb->fsdb_notify_phase)); - fsdb->fsdb_notify_start = cfs_time_current(); + fsdb->fsdb_notify_start = cfs_time_current(); mgs_revoke_lock(fsdb->fsdb_mgs, fsdb, CONFIG_T_RECOVER); - } + } complete(&fsdb->fsdb_notify_comp); - return 0; + return 0; } int mgs_ir_init_fs(const struct lu_env *env, struct mgs_device *mgs, struct fs_db *fsdb) { - cfs_task_t *task; + struct task_struct *task; - if (!ir_timeout) - ir_timeout = OBD_IR_MGS_TIMEOUT; + if (!ir_timeout) + ir_timeout = OBD_IR_MGS_TIMEOUT; - fsdb->fsdb_ir_state = IR_FULL; - if (cfs_time_before(cfs_time_current_sec(), - mgs->mgs_start_time + ir_timeout)) - fsdb->fsdb_ir_state = IR_STARTUP; - fsdb->fsdb_nonir_clients = 0; - CFS_INIT_LIST_HEAD(&fsdb->fsdb_clients); + fsdb->fsdb_ir_state = IR_FULL; + if (cfs_time_before(cfs_time_current_sec(), + mgs->mgs_start_time + ir_timeout)) + fsdb->fsdb_ir_state = IR_STARTUP; + fsdb->fsdb_nonir_clients = 0; + INIT_LIST_HEAD(&fsdb->fsdb_clients); - /* start notify thread */ + /* start notify thread */ fsdb->fsdb_mgs = mgs; - cfs_atomic_set(&fsdb->fsdb_notify_phase, 0); - cfs_waitq_init(&fsdb->fsdb_notify_waitq); + atomic_set(&fsdb->fsdb_notify_phase, 0); + init_waitqueue_head(&fsdb->fsdb_notify_waitq); init_completion(&fsdb->fsdb_notify_comp); task = kthread_run(mgs_ir_notify, fsdb, @@ -496,22 +481,22 @@ int mgs_ir_init_fs(const struct lu_env *env, struct mgs_device *mgs, CERROR("Start notify thread error %ld\n", PTR_ERR(task)); mgs_nidtbl_init_fs(env, fsdb); - return 0; + return 0; } void mgs_ir_fini_fs(struct mgs_device *mgs, struct fs_db *fsdb) { if (test_bit(FSDB_MGS_SELF, &fsdb->fsdb_flags)) - return; + return; - mgs_fsc_cleanup_by_fsdb(fsdb); + mgs_fsc_cleanup_by_fsdb(fsdb); - mgs_nidtbl_fini_fs(fsdb); + mgs_nidtbl_fini_fs(fsdb); - LASSERT(cfs_list_empty(&fsdb->fsdb_clients)); + LASSERT(list_empty(&fsdb->fsdb_clients)); - fsdb->fsdb_notify_stop = 1; - cfs_waitq_signal(&fsdb->fsdb_notify_waitq); + fsdb->fsdb_notify_stop = 1; + wake_up(&fsdb->fsdb_notify_waitq); wait_for_completion(&fsdb->fsdb_notify_comp); } @@ -563,14 +548,14 @@ int mgs_ir_update(const struct lu_env *env, struct mgs_device *mgs, } mutex_unlock(&fsdb->fsdb_mutex); - LASSERT(ergo(mti->mti_flags & LDD_F_IR_CAPABLE, notify)); - if (notify) { - CDEBUG(D_MGS, "Try to revoke recover lock of %s\n", - fsdb->fsdb_name); - cfs_atomic_inc(&fsdb->fsdb_notify_phase); - cfs_waitq_signal(&fsdb->fsdb_notify_waitq); - } - return 0; + LASSERT(ergo(mti->mti_flags & LDD_F_IR_CAPABLE, notify)); + if (notify) { + CDEBUG(D_MGS, "Try to revoke recover lock of %s\n", + fsdb->fsdb_name); + atomic_inc(&fsdb->fsdb_notify_phase); + wake_up(&fsdb->fsdb_notify_waitq); + } + return 0; } /* NID table can be cached by two entities: Clients and MDTs */ @@ -658,10 +643,6 @@ int mgs_get_ir_logs(struct ptlrpc_request *req) if (pages == NULL) RETURN(-ENOMEM); - rc = req_capsule_server_pack(&req->rq_pill); - if (rc) - GOTO(out, rc); - res = req_capsule_server_get(&req->rq_pill, &RMF_MGS_CONFIG_RES); if (res == NULL) GOTO(out, rc = -EINVAL); @@ -677,18 +658,22 @@ int mgs_get_ir_logs(struct ptlrpc_request *req) page_count = (bytes + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT; LASSERT(page_count <= nrpages); desc = ptlrpc_prep_bulk_exp(req, page_count, 1, - BULK_PUT_SOURCE, MGS_BULK_PORTAL); + PTLRPC_BULK_PUT_SOURCE | + PTLRPC_BULK_BUF_KIOV, + MGS_BULK_PORTAL, + &ptlrpc_bulk_kiov_pin_ops); if (desc == NULL) GOTO(out, rc = -ENOMEM); for (i = 0; i < page_count && bytes > 0; i++) { - ptlrpc_prep_bulk_page_pin(desc, pages[i], 0, - min_t(int, bytes, PAGE_CACHE_SIZE)); + desc->bd_frag_ops->add_kiov_frag(desc, pages[i], 0, + min_t(int, bytes, + PAGE_CACHE_SIZE)); bytes -= PAGE_CACHE_SIZE; } rc = target_bulk_io(req->rq_export, desc, &lwi); - ptlrpc_free_bulk_pin(desc); + ptlrpc_free_bulk(desc); out: for (i = 0; i < nrpages; i++) { @@ -752,25 +737,25 @@ static struct lproc_ir_cmd { { "0", 1, lprocfs_ir_clear_stats } }; -int lprocfs_wr_ir_state(struct file *file, const char *buffer, - unsigned long count, void *data) +int lprocfs_wr_ir_state(struct file *file, const char __user *buffer, + size_t count, void *data) { struct fs_db *fsdb = data; char *kbuf; char *ptr; int rc = 0; - if (count > PAGE_CACHE_SIZE) - return -EINVAL; + if (count == 0 || count >= PAGE_CACHE_SIZE) + return -EINVAL; - OBD_ALLOC(kbuf, count + 1); - if (kbuf == NULL) - return -ENOMEM; + OBD_ALLOC(kbuf, count + 1); + if (kbuf == NULL) + return -ENOMEM; - if (copy_from_user(kbuf, buffer, count)) { - OBD_FREE(kbuf, count); - return -EFAULT; - } + if (copy_from_user(kbuf, buffer, count)) { + OBD_FREE(kbuf, count + 1); + return -EFAULT; + } kbuf[count] = 0; /* buffer is supposed to end with 0 */ if (kbuf[count - 1] == '\n') @@ -842,15 +827,14 @@ int lprocfs_rd_ir_state(struct seq_file *seq, void *data) return 0; } -int lprocfs_rd_ir_timeout(char *page, char **start, off_t off, int count, - int *eof, void *data) +int lprocfs_ir_timeout_seq_show(struct seq_file *m, void *data) { - *eof = 1; - return snprintf(page, count, "%d\n", ir_timeout); + return lprocfs_uint_seq_show(m, &ir_timeout); } -int lprocfs_wr_ir_timeout(struct file *file, const char *buffer, - unsigned long count, void *data) +ssize_t lprocfs_ir_timeout_seq_write(struct file *file, + const char __user *buffer, + size_t count, loff_t *off) { return lprocfs_wr_uint(file, buffer, count, &ir_timeout); } @@ -878,8 +862,8 @@ int mgs_fsc_attach(const struct lu_env *env, struct obd_export *exp, if (new_fsc == NULL) RETURN(-ENOMEM); - CFS_INIT_LIST_HEAD(&new_fsc->mfc_export_list); - CFS_INIT_LIST_HEAD(&new_fsc->mfc_fsdb_list); + INIT_LIST_HEAD(&new_fsc->mfc_export_list); + INIT_LIST_HEAD(&new_fsc->mfc_fsdb_list); new_fsc->mfc_fsdb = fsdb; new_fsc->mfc_export = class_export_get(exp); new_fsc->mfc_ir_capable = !!(exp_connect_flags(exp) & @@ -890,7 +874,7 @@ int mgs_fsc_attach(const struct lu_env *env, struct obd_export *exp, /* tend to find it in export list because this list is shorter. */ spin_lock(&data->med_lock); - cfs_list_for_each_entry(fsc, &data->med_clients, mfc_export_list) { + list_for_each_entry(fsc, &data->med_clients, mfc_export_list) { if (strcmp(fsname, fsc->mfc_fsdb->fsdb_name) == 0) { found = true; break; @@ -901,10 +885,10 @@ int mgs_fsc_attach(const struct lu_env *env, struct obd_export *exp, new_fsc = NULL; /* add it into export list. */ - cfs_list_add(&fsc->mfc_export_list, &data->med_clients); + list_add(&fsc->mfc_export_list, &data->med_clients); /* add into fsdb list. */ - cfs_list_add(&fsc->mfc_fsdb_list, &fsdb->fsdb_clients); + list_add(&fsc->mfc_fsdb_list, &fsdb->fsdb_clients); if (!fsc->mfc_ir_capable) { ++fsdb->fsdb_nonir_clients; if (fsdb->fsdb_ir_state == IR_FULL) @@ -926,19 +910,19 @@ void mgs_fsc_cleanup(struct obd_export *exp) { struct mgs_export_data *data = &exp->u.eu_mgs_data; struct mgs_fsc *fsc, *tmp; - CFS_LIST_HEAD(head); + struct list_head head = LIST_HEAD_INIT(head); spin_lock(&data->med_lock); - cfs_list_splice_init(&data->med_clients, &head); + list_splice_init(&data->med_clients, &head); spin_unlock(&data->med_lock); - cfs_list_for_each_entry_safe(fsc, tmp, &head, mfc_export_list) { + list_for_each_entry_safe(fsc, tmp, &head, mfc_export_list) { struct fs_db *fsdb = fsc->mfc_fsdb; LASSERT(fsc->mfc_export == exp); mutex_lock(&fsdb->fsdb_mutex); - cfs_list_del_init(&fsc->mfc_fsdb_list); + list_del_init(&fsc->mfc_fsdb_list); if (fsc->mfc_ir_capable == 0) { --fsdb->fsdb_nonir_clients; LASSERT(fsdb->fsdb_ir_state != IR_FULL); @@ -947,7 +931,7 @@ void mgs_fsc_cleanup(struct obd_export *exp) fsdb->fsdb_ir_state = IR_FULL; } mutex_unlock(&fsdb->fsdb_mutex); - cfs_list_del_init(&fsc->mfc_export_list); + list_del_init(&fsc->mfc_export_list); class_export_put(fsc->mfc_export); OBD_FREE_PTR(fsc); } @@ -958,15 +942,15 @@ void mgs_fsc_cleanup_by_fsdb(struct fs_db *fsdb) { struct mgs_fsc *fsc, *tmp; - cfs_list_for_each_entry_safe(fsc, tmp, &fsdb->fsdb_clients, + list_for_each_entry_safe(fsc, tmp, &fsdb->fsdb_clients, mfc_fsdb_list) { struct mgs_export_data *data = &fsc->mfc_export->u.eu_mgs_data; LASSERT(fsdb == fsc->mfc_fsdb); - cfs_list_del_init(&fsc->mfc_fsdb_list); + list_del_init(&fsc->mfc_fsdb_list); spin_lock(&data->med_lock); - cfs_list_del_init(&fsc->mfc_export_list); + list_del_init(&fsc->mfc_export_list); spin_unlock(&data->med_lock); class_export_put(fsc->mfc_export); OBD_FREE_PTR(fsc);