X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Flov%2Flov_pool.c;h=a8d3f18615d74e507a771b06deb0983390b21981;hb=7107e54983b0a701c5c02a1e3c521302e8f79810;hp=764a494d0bb77e3db5950bf2dd52cbaf5dcdd0a9;hpb=ad5c8ce3c3b561c3913cfe28703652329ec6c4dc;p=fs%2Flustre-release.git diff --git a/lustre/lov/lov_pool.c b/lustre/lov/lov_pool.c index 764a494..a8d3f18 100644 --- a/lustre/lov/lov_pool.c +++ b/lustre/lov/lov_pool.c @@ -1,6 +1,4 @@ -/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- - * vim:expandtab:shiftwidth=8:tabstop=8: - * +/* * GPL HEADER START * * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. @@ -16,18 +14,16 @@ * in the LICENSE file that accompanied this code). * * You should have received a copy of the GNU General Public License - * version 2 along with this program; If not, see [sun.com URL with a - * copy of GPLv2]. - * - * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara, - * CA 95054 USA or visit www.sun.com if you need additional information or - * have any questions. + * version 2 along with this program; If not, see + * http://www.gnu.org/licenses/gpl-2.0.html * * GPL HEADER END */ /* - * Copyright 2008 Sun Microsystems, Inc. All rights reserved + * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. + * + * Copyright (c) 2012, 2014, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ @@ -38,31 +34,46 @@ * OST pool methods * * Author: Jacques-Charles LAFOUCRIERE + * Author: Alex Lyashkov + * Author: Nathaniel Rutman */ #define DEBUG_SUBSYSTEM S_LOV -#ifdef __KERNEL__ #include -#else -#include -#endif #include #include "lov_internal.h" -static void lov_pool_getref(struct pool_desc *pool) { - atomic_inc(&pool->pool_refcount); +#define pool_tgt(_p, _i) \ + _p->pool_lobd->u.lov.lov_tgts[_p->pool_obds.op_array[_i]] + +static void lov_pool_getref(struct pool_desc *pool) +{ + CDEBUG(D_INFO, "pool %p\n", pool); + atomic_inc(&pool->pool_refcount); } -static void lov_pool_putref(struct pool_desc *pool) { - if (atomic_dec_and_test(&pool->pool_refcount)) { - lov_ost_pool_free(&(pool->pool_rr.lqr_pool)); - lov_ost_pool_free(&(pool->pool_obds)); - OBD_FREE_PTR(pool); - } +static void lov_pool_putref(struct pool_desc *pool) +{ + CDEBUG(D_INFO, "pool %p\n", pool); + if (atomic_dec_and_test(&pool->pool_refcount)) { + LASSERT(hlist_unhashed(&pool->pool_hash)); + LASSERT(list_empty(&pool->pool_list)); + LASSERT(pool->pool_proc_entry == NULL); + lov_ost_pool_free(&(pool->pool_obds)); + OBD_FREE_PTR(pool); + EXIT; + } } +static void lov_pool_putref_locked(struct pool_desc *pool) +{ + CDEBUG(D_INFO, "pool %p\n", pool); + LASSERT(atomic_read(&pool->pool_refcount) > 1); + + atomic_dec(&pool->pool_refcount); +} /* * hash function using a Rotating Hash algorithm @@ -71,7 +82,8 @@ static void lov_pool_putref(struct pool_desc *pool) { * Chapter 6.4. * Addison Wesley, 1973 */ -static __u32 pool_hashfn(lustre_hash_t *hash_body, void *key, unsigned mask) +static __u32 pool_hashfn(struct cfs_hash *hash_body, const void *key, + unsigned mask) { int i; __u32 result; @@ -91,49 +103,54 @@ static void *pool_key(struct hlist_node *hnode) { struct pool_desc *pool; - pool = hlist_entry(hnode, struct pool_desc, pool_hash); + pool = hlist_entry(hnode, struct pool_desc, pool_hash); return (pool->pool_name); } -static int pool_hashkey_compare(void *key, struct hlist_node *compared_hnode) +static int +pool_hashkey_keycmp(const void *key, struct hlist_node *compared_hnode) { char *pool_name; struct pool_desc *pool; - int rc; pool_name = (char *)key; - pool = hlist_entry(compared_hnode, struct pool_desc, pool_hash); - rc = strncmp(pool_name, pool->pool_name, LOV_MAXPOOLNAME); - return (!rc); + pool = hlist_entry(compared_hnode, struct pool_desc, pool_hash); + return !strncmp(pool_name, pool->pool_name, LOV_MAXPOOLNAME); +} + +static void *pool_hashobject(struct hlist_node *hnode) +{ + return hlist_entry(hnode, struct pool_desc, pool_hash); } -static void *pool_hashrefcount_get(struct hlist_node *hnode) +static void pool_hashrefcount_get(struct cfs_hash *hs, struct hlist_node *hnode) { struct pool_desc *pool; - pool = hlist_entry(hnode, struct pool_desc, pool_hash); + pool = hlist_entry(hnode, struct pool_desc, pool_hash); lov_pool_getref(pool); - return (pool); } -static void *pool_hashrefcount_put(struct hlist_node *hnode) +static void pool_hashrefcount_put_locked(struct cfs_hash *hs, + struct hlist_node *hnode) { struct pool_desc *pool; - pool = hlist_entry(hnode, struct pool_desc, pool_hash); - lov_pool_putref(pool); - return (pool); + pool = hlist_entry(hnode, struct pool_desc, pool_hash); + lov_pool_putref_locked(pool); } -lustre_hash_ops_t pool_hash_operations = { - .lh_hash = pool_hashfn, - .lh_key = pool_key, - .lh_compare = pool_hashkey_compare, - .lh_get = pool_hashrefcount_get, - .lh_put = pool_hashrefcount_put, +struct cfs_hash_ops pool_hash_operations = { + .hs_hash = pool_hashfn, + .hs_key = pool_key, + .hs_keycmp = pool_hashkey_keycmp, + .hs_object = pool_hashobject, + .hs_get = pool_hashrefcount_get, + .hs_put_locked = pool_hashrefcount_put_locked, + }; -#ifdef LPROCFS +#ifdef CONFIG_PROC_FS /* ifdef needed for liblustre support */ /* * pool /proc seq_file methods @@ -156,7 +173,7 @@ static void *pool_proc_next(struct seq_file *s, void *v, loff_t *pos) struct pool_iterator *iter = (struct pool_iterator *)s->private; int prev_idx; - LASSERTF(iter->magic == POOL_IT_MAGIC, "%08X", iter->magic); + LASSERTF(iter->magic == POOL_IT_MAGIC, "%08X\n", iter->magic); /* test if end of file */ if (*pos >= pool_tgt_count(iter->pool)) @@ -164,14 +181,14 @@ static void *pool_proc_next(struct seq_file *s, void *v, loff_t *pos) /* iterate to find a non empty entry */ prev_idx = iter->idx; - down_read(&pool_tgt_rw_sem(iter->pool)); + down_read(&pool_tgt_rw_sem(iter->pool)); iter->idx++; if (iter->idx == pool_tgt_count(iter->pool)) { iter->idx = prev_idx; /* we stay on the last entry */ - up_read(&pool_tgt_rw_sem(iter->pool)); + up_read(&pool_tgt_rw_sem(iter->pool)); return NULL; } - up_read(&pool_tgt_rw_sem(iter->pool)); + up_read(&pool_tgt_rw_sem(iter->pool)); (*pos)++; /* return != NULL to continue */ return iter; @@ -237,13 +254,13 @@ static int pool_proc_show(struct seq_file *s, void *v) struct pool_iterator *iter = (struct pool_iterator *)v; struct lov_tgt_desc *tgt; - LASSERTF(iter->magic == POOL_IT_MAGIC, "%08X", iter->magic); - LASSERT(iter->pool != NULL); - LASSERT(iter->idx <= pool_tgt_count(iter->pool)); + LASSERTF(iter->magic == POOL_IT_MAGIC, "%08X\n", iter->magic); + LASSERT(iter->pool != NULL); + LASSERT(iter->idx <= pool_tgt_count(iter->pool)); - down_read(&pool_tgt_rw_sem(iter->pool)); + down_read(&pool_tgt_rw_sem(iter->pool)); tgt = pool_tgt(iter->pool, iter->idx); - up_read(&pool_tgt_rw_sem(iter->pool)); + up_read(&pool_tgt_rw_sem(iter->pool)); if (tgt) seq_printf(s, "%s\n", obd_uuid2str(&(tgt->ltd_uuid))); @@ -264,7 +281,7 @@ static int pool_proc_open(struct inode *inode, struct file *file) rc = seq_open(file, &pool_proc_ops); if (!rc) { struct seq_file *s = file->private_data; - s->private = PROC_I(inode)->pde->data; + s->private = PDE_DATA(inode); } return rc; } @@ -275,7 +292,7 @@ static struct file_operations pool_proc_operations = { .llseek = seq_lseek, .release = seq_release, }; -#endif /* LPROCFS */ +#endif /* CONFIG_PROC_FS */ void lov_dump_pool(int level, struct pool_desc *pool) { @@ -285,7 +302,7 @@ void lov_dump_pool(int level, struct pool_desc *pool) CDEBUG(level, "pool "LOV_POOLNAMEF" has %d members\n", pool->pool_name, pool->pool_obds.op_count); - down_read(&pool_tgt_rw_sem(pool)); + down_read(&pool_tgt_rw_sem(pool)); for (i = 0; i < pool_tgt_count(pool) ; i++) { if (!pool_tgt(pool, i) || !(pool_tgt(pool, i))->ltd_exp) @@ -295,24 +312,27 @@ void lov_dump_pool(int level, struct pool_desc *pool) obd_uuid2str(&((pool_tgt(pool, i))->ltd_uuid))); } - up_read(&pool_tgt_rw_sem(pool)); + up_read(&pool_tgt_rw_sem(pool)); lov_pool_putref(pool); } #define LOV_POOL_INIT_COUNT 2 int lov_ost_pool_init(struct ost_pool *op, unsigned int count) { + ENTRY; + if (count == 0) count = LOV_POOL_INIT_COUNT; op->op_array = NULL; op->op_count = 0; - init_rwsem(&op->op_rw_sem); + init_rwsem(&op->op_rw_sem); op->op_size = count; OBD_ALLOC(op->op_array, op->op_size * sizeof(op->op_array[0])); if (op->op_array == NULL) { op->op_size = 0; - return -ENOMEM; + RETURN(-ENOMEM); } + EXIT; return 0; } @@ -345,7 +365,7 @@ int lov_ost_pool_add(struct ost_pool *op, __u32 idx, unsigned int min_count) int rc = 0, i; ENTRY; - down_write(&op->op_rw_sem); + down_write(&op->op_rw_sem); rc = lov_ost_pool_extend(op, min_count); if (rc) @@ -359,45 +379,50 @@ int lov_ost_pool_add(struct ost_pool *op, __u32 idx, unsigned int min_count) /* ost not found we add it */ op->op_array[op->op_count] = idx; op->op_count++; + EXIT; out: - up_write(&op->op_rw_sem); + up_write(&op->op_rw_sem); return rc; } int lov_ost_pool_remove(struct ost_pool *op, __u32 idx) { int i; + ENTRY; - down_write(&op->op_rw_sem); + down_write(&op->op_rw_sem); for (i = 0; i < op->op_count; i++) { if (op->op_array[i] == idx) { memmove(&op->op_array[i], &op->op_array[i + 1], (op->op_count - i - 1) * sizeof(op->op_array[0])); op->op_count--; - up_write(&op->op_rw_sem); + up_write(&op->op_rw_sem); + EXIT; return 0; } } - up_write(&op->op_rw_sem); - return -EINVAL; + up_write(&op->op_rw_sem); + RETURN(-EINVAL); } int lov_ost_pool_free(struct ost_pool *op) { + ENTRY; + if (op->op_size == 0) - return 0; + RETURN(0); - down_write(&op->op_rw_sem); + down_write(&op->op_rw_sem); OBD_FREE(op->op_array, op->op_size * sizeof(op->op_array[0])); op->op_array = NULL; op->op_count = 0; op->op_size = 0; - up_write(&op->op_rw_sem); - return 0; + up_write(&op->op_rw_sem); + RETURN(0); } @@ -417,63 +442,59 @@ int lov_pool_new(struct obd_device *obd, char *poolname) if (new_pool == NULL) RETURN(-ENOMEM); - strncpy(new_pool->pool_name, poolname, LOV_MAXPOOLNAME); - new_pool->pool_name[LOV_MAXPOOLNAME] = '\0'; - new_pool->pool_lov = lov; - /* ref count init to 1 because when created a pool is always used - * up to deletion - */ - atomic_set(&new_pool->pool_refcount, 1); - rc = lov_ost_pool_init(&new_pool->pool_obds, 0); - if (rc) - GOTO(out_err, rc); + strlcpy(new_pool->pool_name, poolname, sizeof(new_pool->pool_name)); + new_pool->pool_lobd = obd; + /* ref count init to 1 because when created a pool is always used + * up to deletion + */ + atomic_set(&new_pool->pool_refcount, 1); + rc = lov_ost_pool_init(&new_pool->pool_obds, 0); + if (rc) + GOTO(out_err, rc); - memset(&(new_pool->pool_rr), 0, sizeof(struct lov_qos_rr)); - rc = lov_ost_pool_init(&new_pool->pool_rr.lqr_pool, 0); - if (rc) { - lov_ost_pool_free(&new_pool->pool_obds); - GOTO(out_err, rc); - } + INIT_HLIST_NODE(&new_pool->pool_hash); - INIT_HLIST_NODE(&new_pool->pool_hash); - rc = lustre_hash_add_unique(lov->lov_pools_hash_body, poolname, - &new_pool->pool_hash); - if (rc) { - lov_ost_pool_free(&new_pool->pool_rr.lqr_pool); - lov_ost_pool_free(&new_pool->pool_obds); - GOTO(out_err, rc = -EEXIST); - } +#ifdef CONFIG_PROC_FS + /* get ref for /proc file */ + lov_pool_getref(new_pool); + new_pool->pool_proc_entry = lprocfs_add_simple(lov->lov_pool_proc_entry, + poolname, new_pool, + &pool_proc_operations); + if (IS_ERR(new_pool->pool_proc_entry)) { + CWARN("Cannot add proc pool entry "LOV_POOLNAMEF"\n", poolname); + new_pool->pool_proc_entry = NULL; + lov_pool_putref(new_pool); + } + CDEBUG(D_INFO, "pool %p - proc %p\n", + new_pool, new_pool->pool_proc_entry); +#endif - spin_lock(&obd->obd_dev_lock); - list_add_tail(&new_pool->pool_list, &lov->lov_pool_list); - lov->lov_pool_count++; + spin_lock(&obd->obd_dev_lock); + list_add_tail(&new_pool->pool_list, &lov->lov_pool_list); + lov->lov_pool_count++; + spin_unlock(&obd->obd_dev_lock); - spin_unlock(&obd->obd_dev_lock); + /* add to find only when it fully ready */ + rc = cfs_hash_add_unique(lov->lov_pools_hash_body, poolname, + &new_pool->pool_hash); + if (rc) + GOTO(out_err, rc = -EEXIST); CDEBUG(D_CONFIG, LOV_POOLNAMEF" is pool #%d\n", poolname, lov->lov_pool_count); -#ifdef LPROCFS - /* ifdef needed for liblustre */ - /* get ref for /proc file */ - lov_pool_getref(new_pool); - new_pool->pool_proc_entry = lprocfs_add_simple(lov->lov_pool_proc_entry, - poolname, NULL, NULL, - new_pool, - &pool_proc_operations); -#endif - - if (IS_ERR(new_pool->pool_proc_entry)) { - CWARN("Cannot add proc pool entry "LOV_POOLNAMEF"\n", poolname); - new_pool->pool_proc_entry = NULL; - lov_pool_putref(new_pool); - } - RETURN(0); out_err: - OBD_FREE_PTR(new_pool); - return rc; + spin_lock(&obd->obd_dev_lock); + list_del_init(&new_pool->pool_list); + lov->lov_pool_count--; + spin_unlock(&obd->obd_dev_lock); + lprocfs_remove(&new_pool->pool_proc_entry); + lov_ost_pool_free(&new_pool->pool_obds); + OBD_FREE_PTR(new_pool); + + return rc; } int lov_pool_del(struct obd_device *obd, char *poolname) @@ -484,36 +505,26 @@ int lov_pool_del(struct obd_device *obd, char *poolname) lov = &(obd->u.lov); - spin_lock(&obd->obd_dev_lock); - - pool = lustre_hash_lookup(lov->lov_pools_hash_body, poolname); - if (pool == NULL) { - spin_unlock(&obd->obd_dev_lock); + /* lookup and kill hash reference */ + pool = cfs_hash_del_key(lov->lov_pools_hash_body, poolname); + if (pool == NULL) RETURN(-ENOENT); - } -#ifdef LPROCFS if (pool->pool_proc_entry != NULL) { - remove_proc_entry(pool->pool_proc_entry->name, - pool->pool_proc_entry->parent); - /* remove ref for /proc file */ + CDEBUG(D_INFO, "proc entry %p\n", pool->pool_proc_entry); + lprocfs_remove(&pool->pool_proc_entry); lov_pool_putref(pool); } -#endif - lustre_hash_del_key(lov->lov_pools_hash_body, poolname); - list_del_init(&pool->pool_list); + spin_lock(&obd->obd_dev_lock); + list_del_init(&pool->pool_list); + lov->lov_pool_count--; + spin_unlock(&obd->obd_dev_lock); - lov->lov_pool_count--; - lh_put(lov->lov_pools_hash_body, &pool->pool_hash); - spin_unlock(&obd->obd_dev_lock); + /* release last reference */ + lov_pool_putref(pool); - /* remove ref got when pool was created in memory - * pool will be freed when refount will reach 0 - */ - lov_pool_putref(pool); - - RETURN(0); + RETURN(0); } @@ -522,13 +533,13 @@ int lov_pool_add(struct obd_device *obd, char *poolname, char *ostname) struct obd_uuid ost_uuid; struct lov_obd *lov; struct pool_desc *pool; - unsigned int i, lov_idx; + unsigned int lov_idx; int rc; ENTRY; lov = &(obd->u.lov); - pool = lustre_hash_lookup(lov->lov_pools_hash_body, poolname); + pool = cfs_hash_lookup(lov->lov_pools_hash_body, poolname); if (pool == NULL) RETURN(-ENOENT); @@ -536,35 +547,29 @@ int lov_pool_add(struct obd_device *obd, char *poolname, char *ostname) /* search ost in lov array */ - mutex_down(&lov->lov_lock); - for (i = 0; i < lov->desc.ld_tgt_count; i++) { - if (!lov->lov_tgts[i]) + obd_getref(obd); + for (lov_idx = 0; lov_idx < lov->desc.ld_tgt_count; lov_idx++) { + if (!lov->lov_tgts[lov_idx]) continue; - if (obd_uuid_equals(&ost_uuid, &(lov->lov_tgts[i]->ltd_uuid))) + if (obd_uuid_equals(&ost_uuid, + &(lov->lov_tgts[lov_idx]->ltd_uuid))) break; } - /* test if ost found in lov */ - if (i == lov->desc.ld_tgt_count) { - mutex_up(&lov->lov_lock); + if (lov_idx == lov->desc.ld_tgt_count) GOTO(out, rc = -EINVAL); - } - mutex_up(&lov->lov_lock); - - lov_idx = i; rc = lov_ost_pool_add(&pool->pool_obds, lov_idx, lov->lov_tgt_size); if (rc) GOTO(out, rc); - pool->pool_rr.lqr_dirty = 1; - CDEBUG(D_CONFIG, "Added %s to "LOV_POOLNAMEF" as member %d\n", ostname, poolname, pool_tgt_count(pool)); EXIT; out: - lh_put(lov->lov_pools_hash_body, &pool->pool_hash); + obd_putref(obd); + lov_pool_putref(pool); return rc; } @@ -573,97 +578,41 @@ int lov_pool_remove(struct obd_device *obd, char *poolname, char *ostname) struct obd_uuid ost_uuid; struct lov_obd *lov; struct pool_desc *pool; - unsigned int i, lov_idx; + unsigned int lov_idx; int rc = 0; ENTRY; lov = &(obd->u.lov); - spin_lock(&obd->obd_dev_lock); - pool = lustre_hash_lookup(lov->lov_pools_hash_body, poolname); - if (pool == NULL) { - spin_unlock(&obd->obd_dev_lock); + pool = cfs_hash_lookup(lov->lov_pools_hash_body, poolname); + if (pool == NULL) RETURN(-ENOENT); - } obd_str2uuid(&ost_uuid, ostname); + obd_getref(obd); /* search ost in lov array, to get index */ - for (i = 0; i < lov->desc.ld_tgt_count; i++) { - if (!lov->lov_tgts[i]) + for (lov_idx = 0; lov_idx < lov->desc.ld_tgt_count; lov_idx++) { + if (!lov->lov_tgts[lov_idx]) continue; - if (obd_uuid_equals(&ost_uuid, &(lov->lov_tgts[i]->ltd_uuid))) + if (obd_uuid_equals(&ost_uuid, + &(lov->lov_tgts[lov_idx]->ltd_uuid))) break; } /* test if ost found in lov */ - if (i == lov->desc.ld_tgt_count) { - spin_unlock(&obd->obd_dev_lock); + if (lov_idx == lov->desc.ld_tgt_count) GOTO(out, rc = -EINVAL); - } - - spin_unlock(&obd->obd_dev_lock); - - lov_idx = i; lov_ost_pool_remove(&pool->pool_obds, lov_idx); - pool->pool_rr.lqr_dirty = 1; - CDEBUG(D_CONFIG, "%s removed from "LOV_POOLNAMEF"\n", ostname, poolname); EXIT; out: - lh_put(lov->lov_pools_hash_body, &pool->pool_hash); - return rc; -} - -int lov_check_index_in_pool(__u32 idx, struct pool_desc *pool) -{ - int i, rc; - ENTRY; - - /* caller may no have a ref on pool if it got the pool - * without calling lov_find_pool() (e.g. go through the lov pool - * list) - */ - lov_pool_getref(pool); - - down_read(&pool_tgt_rw_sem(pool)); - - for (i = 0; i < pool_tgt_count(pool); i++) { - if (pool_tgt_array(pool)[i] == idx) - GOTO(out, rc = 0); - } - rc = -ENOENT; - EXIT; -out: - up_read(&pool_tgt_rw_sem(pool)); - + obd_putref(obd); lov_pool_putref(pool); return rc; } - -struct pool_desc *lov_find_pool(struct lov_obd *lov, char *poolname) -{ - struct pool_desc *pool; - - pool = NULL; - if (poolname[0] != '\0') { - pool = lustre_hash_lookup(lov->lov_pools_hash_body, poolname); - if (pool == NULL) - CWARN("Request for an unknown pool ("LOV_POOLNAMEF")\n", - poolname); - if ((pool != NULL) && (pool_tgt_count(pool) == 0)) { - CWARN("Request for an empty pool ("LOV_POOLNAMEF")\n", - poolname); - /* pool is ignored, so we remove ref on it */ - lh_put(lov->lov_pools_hash_body, &pool->pool_hash); - pool = NULL; - } - } - return pool; -} -