X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Flov%2Flov_pool.c;h=afccd0523c2c97e72e4c02a7739eb9a300b509fd;hb=3c69d46e1766480c0ffd1bef840b4e167b4cf88e;hp=c017ab444aa9a13e5004ab54e58e696ca91ea4ca;hpb=7da33dc0ba361bedd274480890755a8d2691e65b;p=fs%2Flustre-release.git diff --git a/lustre/lov/lov_pool.c b/lustre/lov/lov_pool.c index c017ab4..afccd05 100644 --- a/lustre/lov/lov_pool.c +++ b/lustre/lov/lov_pool.c @@ -14,12 +14,8 @@ * 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 */ @@ -27,11 +23,10 @@ * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved. * Use is subject to license terms. * - * Copyright (c) 2012, Intel Corporation. + * Copyright (c) 2012, 2017, Intel Corporation. */ /* * This file is part of Lustre, http://www.lustre.org/ - * Lustre is a trademark of Sun Microsystems, Inc. * * lustre/lov/lov_pool.c * @@ -44,11 +39,9 @@ #define DEBUG_SUBSYSTEM S_LOV -#ifdef __KERNEL__ #include -#else -#include -#endif +#include +#include #include #include "lov_internal.h" @@ -56,108 +49,50 @@ #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); - cfs_atomic_inc(&pool->pool_refcount); -} - -void lov_pool_putref(struct pool_desc *pool) -{ - CDEBUG(D_INFO, "pool %p\n", pool); - if (cfs_atomic_dec_and_test(&pool->pool_refcount)) { - LASSERT(cfs_hlist_unhashed(&pool->pool_hash)); - LASSERT(cfs_list_empty(&pool->pool_list)); - LASSERT(pool->pool_proc_entry == NULL); - lov_ost_pool_free(&(pool->pool_obds)); - OBD_FREE_PTR(pool); - EXIT; - } -} - -void lov_pool_putref_locked(struct pool_desc *pool) -{ - CDEBUG(D_INFO, "pool %p\n", pool); - LASSERT(cfs_atomic_read(&pool->pool_refcount) > 1); - - cfs_atomic_dec(&pool->pool_refcount); -} - -/* - * hash function using a Rotating Hash algorithm - * Knuth, D. The Art of Computer Programming, - * Volume 3: Sorting and Searching, - * Chapter 6.4. - * Addison Wesley, 1973 - */ -static __u32 pool_hashfn(cfs_hash_t *hash_body, const void *key, unsigned mask) -{ - int i; - __u32 result; - char *poolname; - - result = 0; - poolname = (char *)key; - for (i = 0; i < LOV_MAXPOOLNAME; i++) { - if (poolname[i] == '\0') - break; - result = (result << 4)^(result >> 28) ^ poolname[i]; - } - return (result % mask); -} - -static void *pool_key(cfs_hlist_node_t *hnode) +static u32 pool_hashfh(const void *data, u32 len, u32 seed) { - struct pool_desc *pool; + const char *pool_name = data; - pool = cfs_hlist_entry(hnode, struct pool_desc, pool_hash); - return (pool->pool_name); + return hashlen_hash(cfs_hashlen_string((void *)(unsigned long)seed, + pool_name)); } -static int pool_hashkey_keycmp(const void *key, cfs_hlist_node_t *compared_hnode) +static int pool_cmpfn(struct rhashtable_compare_arg *arg, const void *obj) { - char *pool_name; - struct pool_desc *pool; + const struct pool_desc *pool = obj; + const char *pool_name = arg->key; - pool_name = (char *)key; - pool = cfs_hlist_entry(compared_hnode, struct pool_desc, pool_hash); - return !strncmp(pool_name, pool->pool_name, LOV_MAXPOOLNAME); + return strcmp(pool_name, pool->pool_name); } -static void *pool_hashobject(cfs_hlist_node_t *hnode) -{ - return cfs_hlist_entry(hnode, struct pool_desc, pool_hash); -} +static const struct rhashtable_params pools_hash_params = { + .key_len = 1, /* actually variable */ + .key_offset = offsetof(struct pool_desc, pool_name), + .head_offset = offsetof(struct pool_desc, pool_hash), + .hashfn = pool_hashfh, + .obj_cmpfn = pool_cmpfn, + .automatic_shrinking = true, +}; -static void pool_hashrefcount_get(cfs_hash_t *hs, cfs_hlist_node_t *hnode) +static void lov_pool_getref(struct pool_desc *pool) { - struct pool_desc *pool; - - pool = cfs_hlist_entry(hnode, struct pool_desc, pool_hash); - lov_pool_getref(pool); + CDEBUG(D_INFO, "pool %p\n", pool); + atomic_inc(&pool->pool_refcount); } -static void pool_hashrefcount_put_locked(cfs_hash_t *hs, - cfs_hlist_node_t *hnode) +void lov_pool_putref(struct pool_desc *pool) { - struct pool_desc *pool; - - pool = cfs_hlist_entry(hnode, struct pool_desc, pool_hash); - lov_pool_putref_locked(pool); + CDEBUG(D_INFO, "pool %p\n", pool); + if (atomic_dec_and_test(&pool->pool_refcount)) { + LASSERT(list_empty(&pool->pool_list)); + LASSERT(pool->pool_proc_entry == NULL); + lu_tgt_pool_free(&(pool->pool_obds)); + kfree_rcu(pool, pool_rcu); + EXIT; + } } -cfs_hash_ops_t 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 needed for liblustre support */ +#ifdef CONFIG_PROC_FS /* * pool /proc seq_file methods */ @@ -176,28 +111,25 @@ struct pool_iterator { 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); - - /* test if end of file */ - if (*pos >= pool_tgt_count(iter->pool)) - return NULL; - - /* iterate to find a non empty entry */ - prev_idx = iter->idx; - 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)); - return NULL; - } - up_read(&pool_tgt_rw_sem(iter->pool)); - (*pos)++; - /* return != NULL to continue */ - return iter; + struct pool_iterator *iter = (struct pool_iterator *)s->private; + int prev_idx; + + LASSERTF(iter->magic == POOL_IT_MAGIC, "%08X\n", iter->magic); + + (*pos)++; + /* test if end of file */ + if (*pos > pool_tgt_count(iter->pool)) + return NULL; + + /* iterate to find a non empty entry */ + prev_idx = iter->idx; + iter->idx++; + if (iter->idx >= pool_tgt_count(iter->pool)) { + iter->idx = prev_idx; /* we stay on the last entry */ + return NULL; + } + /* return != NULL to continue */ + return iter; } static void *pool_proc_start(struct seq_file *s, loff_t *pos) @@ -225,6 +157,7 @@ static void *pool_proc_start(struct seq_file *s, loff_t *pos) * we can free it at stop() */ /* /!\ do not forget to restore it to pool before freeing it */ s->private = iter; + down_read(&pool_tgt_rw_sem(pool)); if (*pos > 0) { loff_t i; void *ptr; @@ -246,13 +179,13 @@ static void pool_proc_stop(struct seq_file *s, void *v) * calling start() method (see seq_read() from fs/seq_file.c) * we have to free only if s->private is an iterator */ if ((iter) && (iter->magic == POOL_IT_MAGIC)) { + up_read(&pool_tgt_rw_sem(iter->pool)); /* we restore s->private so next call to pool_proc_start() * will work */ s->private = iter->pool; lov_pool_putref(iter->pool); OBD_FREE_PTR(iter); } - return; } static int pool_proc_show(struct seq_file *s, void *v) @@ -260,24 +193,22 @@ 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)); tgt = pool_tgt(iter->pool, iter->idx); - up_read(&pool_tgt_rw_sem(iter->pool)); if (tgt) seq_printf(s, "%s\n", obd_uuid2str(&(tgt->ltd_uuid))); return 0; } -static struct seq_operations pool_proc_ops = { - .start = pool_proc_start, - .next = pool_proc_next, - .stop = pool_proc_stop, - .show = pool_proc_show, +static const struct seq_operations pool_proc_ops = { + .start = pool_proc_start, + .next = pool_proc_next, + .stop = pool_proc_stop, + .show = pool_proc_show, }; static int pool_proc_open(struct inode *inode, struct file *file) @@ -287,151 +218,36 @@ 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 = PDE_DATA(inode); + s->private = pde_data(inode); } return rc; } -static struct file_operations pool_proc_operations = { - .open = pool_proc_open, - .read = seq_read, - .llseek = seq_lseek, - .release = seq_release, +const static struct proc_ops pool_proc_operations = { + .proc_open = pool_proc_open, + .proc_read = seq_read, + .proc_lseek = seq_lseek, + .proc_release = seq_release, }; -#endif /* LPROCFS */ - -void lov_dump_pool(int level, struct pool_desc *pool) -{ - int i; - - lov_pool_getref(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)); - - for (i = 0; i < pool_tgt_count(pool) ; i++) { - if (!pool_tgt(pool, i) || !(pool_tgt(pool, i))->ltd_exp) - continue; - CDEBUG(level, "pool "LOV_POOLNAMEF"[%d] = %s\n", - pool->pool_name, i, - obd_uuid2str(&((pool_tgt(pool, i))->ltd_uuid))); - } - - 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); - 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); - } - EXIT; - return 0; -} - -/* Caller must hold write op_rwlock */ -int lov_ost_pool_extend(struct ost_pool *op, unsigned int min_count) -{ - __u32 *new; - int new_size; - - LASSERT(min_count != 0); - - if (op->op_count < op->op_size) - return 0; +#endif /* CONFIG_PROC_FS */ - new_size = max(min_count, 2 * op->op_size); - OBD_ALLOC(new, new_size * sizeof(op->op_array[0])); - if (new == NULL) - return -ENOMEM; - - /* copy old array to new one */ - memcpy(new, op->op_array, op->op_size * sizeof(op->op_array[0])); - OBD_FREE(op->op_array, op->op_size * sizeof(op->op_array[0])); - op->op_array = new; - op->op_size = new_size; - return 0; -} - -int lov_ost_pool_add(struct ost_pool *op, __u32 idx, unsigned int min_count) +static void pools_hash_exit(void *vpool, void *data) { - int rc = 0, i; - ENTRY; - - down_write(&op->op_rw_sem); - - rc = lov_ost_pool_extend(op, min_count); - if (rc) - GOTO(out, rc); + struct pool_desc *pool = vpool; - /* search ost in pool array */ - for (i = 0; i < op->op_count; i++) { - if (op->op_array[i] == idx) - GOTO(out, rc = -EEXIST); - } - /* ost not found we add it */ - op->op_array[op->op_count] = idx; - op->op_count++; - EXIT; -out: - up_write(&op->op_rw_sem); - return rc; + lov_pool_putref(pool); } -int lov_ost_pool_remove(struct ost_pool *op, __u32 idx) +int lov_pool_hash_init(struct rhashtable *tbl) { - int i; - ENTRY; - - 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); - EXIT; - return 0; - } - } - - up_write(&op->op_rw_sem); - RETURN(-EINVAL); + return rhashtable_init(tbl, &pools_hash_params); } -int lov_ost_pool_free(struct ost_pool *op) +void lov_pool_hash_destroy(struct rhashtable *tbl) { - ENTRY; - - if (op->op_size == 0) - RETURN(0); - - 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); + rhashtable_free_and_destroy(tbl, pools_hash_exit, NULL); } - int lov_pool_new(struct obd_device *obd, char *poolname) { struct lov_obd *lov; @@ -444,52 +260,54 @@ int lov_pool_new(struct obd_device *obd, char *poolname) if (strlen(poolname) > LOV_MAXPOOLNAME) RETURN(-ENAMETOOLONG); - OBD_ALLOC_PTR(new_pool); + /* OBD_ALLOC doesn't work with direct use of kfree_rcu */ + new_pool = kmalloc(sizeof(*new_pool), GFP_KERNEL); if (new_pool == NULL) RETURN(-ENOMEM); - strncpy(new_pool->pool_name, poolname, LOV_MAXPOOLNAME); - new_pool->pool_name[LOV_MAXPOOLNAME] = '\0'; + 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 - */ - cfs_atomic_set(&new_pool->pool_refcount, 1); - rc = lov_ost_pool_init(&new_pool->pool_obds, 0); - if (rc) - GOTO(out_err, rc); - - CFS_INIT_HLIST_NODE(&new_pool->pool_hash); - -#ifdef LPROCFS - /* we need this assert seq_file is not implementated for liblustre */ - /* get ref for /proc file */ + /* ref count init to 1 because when created a pool is always used + * up to deletion + */ + atomic_set(&new_pool->pool_refcount, 1); + rc = lu_tgt_pool_init(&new_pool->pool_obds, 0); + if (rc) + GOTO(out_free_pool, rc); + +#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, -#ifndef HAVE_ONLY_PROCFS_SEQ - NULL, NULL, -#endif - 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); + 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); - cfs_list_add_tail(&new_pool->pool_list, &lov->lov_pool_list); + list_add_tail(&new_pool->pool_list, &lov->lov_pool_list); lov->lov_pool_count++; 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); + /* Add to hash table only when it is fully ready. */ + rc = rhashtable_lookup_insert_fast(&lov->lov_pools_hash_body, + &new_pool->pool_hash, + pools_hash_params); + if (rc) { + if (rc != -EEXIST) + /* + * Hide -E2BIG and -EBUSY which + * are not helpful. + */ + rc = -ENOMEM; + GOTO(out_err, rc); + } CDEBUG(D_CONFIG, LOV_POOLNAMEF" is pool #%d\n", poolname, lov->lov_pool_count); @@ -498,16 +316,33 @@ int lov_pool_new(struct obd_device *obd, char *poolname) out_err: spin_lock(&obd->obd_dev_lock); - cfs_list_del_init(&new_pool->pool_list); + 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); + lu_tgt_pool_free(&new_pool->pool_obds); +out_free_pool: OBD_FREE_PTR(new_pool); return rc; } +struct pool_desc *lov_pool_find(struct obd_device *obd, char *poolname) +{ + struct pool_desc *pool; + struct lov_obd *lov = &obd->u.lov; + + rcu_read_lock(); + pool = rhashtable_lookup(&lov->lov_pools_hash_body, + poolname, + pools_hash_params); + if (pool && !atomic_inc_not_zero(&pool->pool_refcount)) + pool = NULL; + rcu_read_unlock(); + + return pool; +} + int lov_pool_del(struct obd_device *obd, char *poolname) { struct lov_obd *lov; @@ -516,10 +351,17 @@ int lov_pool_del(struct obd_device *obd, char *poolname) lov = &(obd->u.lov); - /* lookup and kill hash reference */ - pool = cfs_hash_del_key(lov->lov_pools_hash_body, poolname); - if (pool == NULL) - RETURN(-ENOENT); + /* lookup and kill hash reference */ + rcu_read_lock(); + pool = rhashtable_lookup(&lov->lov_pools_hash_body, poolname, + pools_hash_params); + if (pool && rhashtable_remove_fast(&lov->lov_pools_hash_body, + &pool->pool_hash, + pools_hash_params) != 0) + pool = NULL; + rcu_read_unlock(); + if (!pool) + RETURN(-ENOENT); if (pool->pool_proc_entry != NULL) { CDEBUG(D_INFO, "proc entry %p\n", pool->pool_proc_entry); @@ -528,7 +370,7 @@ int lov_pool_del(struct obd_device *obd, char *poolname) } spin_lock(&obd->obd_dev_lock); - cfs_list_del_init(&pool->pool_list); + list_del_init(&pool->pool_list); lov->lov_pool_count--; spin_unlock(&obd->obd_dev_lock); @@ -550,15 +392,20 @@ int lov_pool_add(struct obd_device *obd, char *poolname, char *ostname) lov = &(obd->u.lov); - pool = cfs_hash_lookup(lov->lov_pools_hash_body, poolname); - if (pool == NULL) - RETURN(-ENOENT); + rcu_read_lock(); + pool = rhashtable_lookup(&lov->lov_pools_hash_body, poolname, + pools_hash_params); + if (pool && !atomic_inc_not_zero(&pool->pool_refcount)) + pool = NULL; + rcu_read_unlock(); + if (!pool) + RETURN(-ENOENT); obd_str2uuid(&ost_uuid, ostname); /* search ost in lov array */ - obd_getref(obd); + lov_tgts_getref(obd); for (lov_idx = 0; lov_idx < lov->desc.ld_tgt_count; lov_idx++) { if (!lov->lov_tgts[lov_idx]) continue; @@ -570,7 +417,7 @@ int lov_pool_add(struct obd_device *obd, char *poolname, char *ostname) if (lov_idx == lov->desc.ld_tgt_count) GOTO(out, rc = -EINVAL); - rc = lov_ost_pool_add(&pool->pool_obds, lov_idx, lov->lov_tgt_size); + rc = lu_tgt_pool_add(&pool->pool_obds, lov_idx, lov->lov_tgt_size); if (rc) GOTO(out, rc); @@ -579,9 +426,10 @@ int lov_pool_add(struct obd_device *obd, char *poolname, char *ostname) EXIT; out: - obd_putref(obd); - lov_pool_putref(pool); - return rc; + lov_tgts_putref(obd); + lov_pool_putref(pool); + + return rc; } int lov_pool_remove(struct obd_device *obd, char *poolname, char *ostname) @@ -595,13 +443,19 @@ int lov_pool_remove(struct obd_device *obd, char *poolname, char *ostname) lov = &(obd->u.lov); - pool = cfs_hash_lookup(lov->lov_pools_hash_body, poolname); - if (pool == NULL) - RETURN(-ENOENT); + /* lookup and kill hash reference */ + rcu_read_lock(); + pool = rhashtable_lookup(&lov->lov_pools_hash_body, poolname, + pools_hash_params); + if (pool && !atomic_inc_not_zero(&pool->pool_refcount)) + pool = NULL; + rcu_read_unlock(); + if (!pool) + RETURN(-ENOENT); obd_str2uuid(&ost_uuid, ostname); - obd_getref(obd); + lov_tgts_getref(obd); /* search ost in lov array, to get index */ for (lov_idx = 0; lov_idx < lov->desc.ld_tgt_count; lov_idx++) { if (!lov->lov_tgts[lov_idx]) @@ -616,14 +470,15 @@ int lov_pool_remove(struct obd_device *obd, char *poolname, char *ostname) if (lov_idx == lov->desc.ld_tgt_count) GOTO(out, rc = -EINVAL); - lov_ost_pool_remove(&pool->pool_obds, lov_idx); + lu_tgt_pool_remove(&pool->pool_obds, lov_idx); CDEBUG(D_CONFIG, "%s removed from "LOV_POOLNAMEF"\n", ostname, poolname); EXIT; out: - obd_putref(obd); - lov_pool_putref(pool); - return rc; + lov_tgts_putref(obd); + lov_pool_putref(pool); + + return rc; }