diff options
Diffstat (limited to 'fs/gfs2/glock.c')
| -rw-r--r-- | fs/gfs2/glock.c | 2399 |
1 files changed, 1538 insertions, 861 deletions
diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 9435384562a2..92e029104d8a 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -1,25 +1,25 @@ +// SPDX-License-Identifier: GPL-2.0-only /* * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved. * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved. - * - * This copyrighted material is made available to anyone wishing to use, - * modify, copy, or redistribute it subject to the terms and conditions - * of the GNU General Public License version 2. */ +#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt + #include <linux/sched.h> #include <linux/slab.h> #include <linux/spinlock.h> #include <linux/buffer_head.h> #include <linux/delay.h> #include <linux/sort.h> +#include <linux/hash.h> #include <linux/jhash.h> #include <linux/kallsyms.h> #include <linux/gfs2_ondisk.h> #include <linux/list.h> #include <linux/wait.h> #include <linux/module.h> -#include <asm/uaccess.h> +#include <linux/uaccess.h> #include <linux/seq_file.h> #include <linux/debugfs.h> #include <linux/kthread.h> @@ -31,6 +31,11 @@ #include <linux/bit_spinlock.h> #include <linux/percpu.h> #include <linux/list_sort.h> +#include <linux/lockref.h> +#include <linux/rhashtable.h> +#include <linux/pid_namespace.h> +#include <linux/file.h> +#include <linux/random.h> #include "gfs2.h" #include "incore.h" @@ -47,9 +52,8 @@ #include "trace_gfs2.h" struct gfs2_glock_iter { - int hash; /* hash bucket index */ - unsigned nhash; /* Index within current bucket */ struct gfs2_sbd *sdp; /* incore superblock */ + struct rhashtable_iter hti; /* rhashtable iterator */ struct gfs2_glock *gl; /* current glock struct */ loff_t last_pos; /* last position */ }; @@ -57,147 +61,207 @@ struct gfs2_glock_iter { typedef void (*glock_examiner) (struct gfs2_glock * gl); static void do_xmote(struct gfs2_glock *gl, struct gfs2_holder *gh, unsigned int target); +static void request_demote(struct gfs2_glock *gl, unsigned int state, + unsigned long delay, bool remote); static struct dentry *gfs2_root; -static struct workqueue_struct *glock_workqueue; -struct workqueue_struct *gfs2_delete_workqueue; static LIST_HEAD(lru_list); static atomic_t lru_count = ATOMIC_INIT(0); static DEFINE_SPINLOCK(lru_lock); #define GFS2_GL_HASH_SHIFT 15 -#define GFS2_GL_HASH_SIZE (1 << GFS2_GL_HASH_SHIFT) -#define GFS2_GL_HASH_MASK (GFS2_GL_HASH_SIZE - 1) +#define GFS2_GL_HASH_SIZE BIT(GFS2_GL_HASH_SHIFT) -static struct hlist_bl_head gl_hash_table[GFS2_GL_HASH_SIZE]; -static struct dentry *gfs2_root; +static const struct rhashtable_params ht_parms = { + .nelem_hint = GFS2_GL_HASH_SIZE * 3 / 4, + .key_len = offsetofend(struct lm_lockname, ln_type), + .key_offset = offsetof(struct gfs2_glock, gl_name), + .head_offset = offsetof(struct gfs2_glock, gl_node), +}; -/** - * gl_hash() - Turn glock number into hash bucket number - * @lock: The glock number - * - * Returns: The number of the corresponding hash bucket - */ +static struct rhashtable gl_hash_table; -static unsigned int gl_hash(const struct gfs2_sbd *sdp, - const struct lm_lockname *name) -{ - unsigned int h; +#define GLOCK_WAIT_TABLE_BITS 12 +#define GLOCK_WAIT_TABLE_SIZE (1 << GLOCK_WAIT_TABLE_BITS) +static wait_queue_head_t glock_wait_table[GLOCK_WAIT_TABLE_SIZE] __cacheline_aligned; - h = jhash(&name->ln_number, sizeof(u64), 0); - h = jhash(&name->ln_type, sizeof(unsigned int), h); - h = jhash(&sdp, sizeof(struct gfs2_sbd *), h); - h &= GFS2_GL_HASH_MASK; +struct wait_glock_queue { + struct lm_lockname *name; + wait_queue_entry_t wait; +}; + +static int glock_wake_function(wait_queue_entry_t *wait, unsigned int mode, + int sync, void *key) +{ + struct wait_glock_queue *wait_glock = + container_of(wait, struct wait_glock_queue, wait); + struct lm_lockname *wait_name = wait_glock->name; + struct lm_lockname *wake_name = key; - return h; + if (wake_name->ln_sbd != wait_name->ln_sbd || + wake_name->ln_number != wait_name->ln_number || + wake_name->ln_type != wait_name->ln_type) + return 0; + return autoremove_wake_function(wait, mode, sync, key); } -static inline void spin_lock_bucket(unsigned int hash) +static wait_queue_head_t *glock_waitqueue(struct lm_lockname *name) { - hlist_bl_lock(&gl_hash_table[hash]); + u32 hash = jhash2((u32 *)name, ht_parms.key_len / 4, 0); + + return glock_wait_table + hash_32(hash, GLOCK_WAIT_TABLE_BITS); } -static inline void spin_unlock_bucket(unsigned int hash) +/** + * wake_up_glock - Wake up waiters on a glock + * @gl: the glock + */ +static void wake_up_glock(struct gfs2_glock *gl) { - hlist_bl_unlock(&gl_hash_table[hash]); + wait_queue_head_t *wq = glock_waitqueue(&gl->gl_name); + + if (waitqueue_active(wq)) + __wake_up(wq, TASK_NORMAL, 1, &gl->gl_name); } static void gfs2_glock_dealloc(struct rcu_head *rcu) { struct gfs2_glock *gl = container_of(rcu, struct gfs2_glock, gl_rcu); + kfree(gl->gl_lksb.sb_lvbptr); if (gl->gl_ops->go_flags & GLOF_ASPACE) { - kmem_cache_free(gfs2_glock_aspace_cachep, gl); - } else { - kfree(gl->gl_lksb.sb_lvbptr); + struct gfs2_glock_aspace *gla = + container_of(gl, struct gfs2_glock_aspace, glock); + kmem_cache_free(gfs2_glock_aspace_cachep, gla); + } else kmem_cache_free(gfs2_glock_cachep, gl); - } } -void gfs2_glock_free(struct gfs2_glock *gl) +static void __gfs2_glock_free(struct gfs2_glock *gl) { - struct gfs2_sbd *sdp = gl->gl_sbd; - + rhashtable_remove_fast(&gl_hash_table, &gl->gl_node, ht_parms); + smp_mb(); + wake_up_glock(gl); call_rcu(&gl->gl_rcu, gfs2_glock_dealloc); +} + +void gfs2_glock_free(struct gfs2_glock *gl) { + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + + __gfs2_glock_free(gl); if (atomic_dec_and_test(&sdp->sd_glock_disposal)) - wake_up(&sdp->sd_glock_wait); + wake_up(&sdp->sd_kill_wait); } -/** - * gfs2_glock_hold() - increment reference count on glock - * @gl: The glock to hold - * - */ +void gfs2_glock_free_later(struct gfs2_glock *gl) { + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + + spin_lock(&lru_lock); + list_add(&gl->gl_lru, &sdp->sd_dead_glocks); + spin_unlock(&lru_lock); + if (atomic_dec_and_test(&sdp->sd_glock_disposal)) + wake_up(&sdp->sd_kill_wait); +} -void gfs2_glock_hold(struct gfs2_glock *gl) +static void gfs2_free_dead_glocks(struct gfs2_sbd *sdp) { - GLOCK_BUG_ON(gl, atomic_read(&gl->gl_ref) == 0); - atomic_inc(&gl->gl_ref); + struct list_head *list = &sdp->sd_dead_glocks; + + while(!list_empty(list)) { + struct gfs2_glock *gl; + + gl = list_first_entry(list, struct gfs2_glock, gl_lru); + list_del_init(&gl->gl_lru); + __gfs2_glock_free(gl); + } } /** - * demote_ok - Check to see if it's ok to unlock a glock - * @gl: the glock + * gfs2_glock_hold() - increment reference count on glock + * @gl: The glock to hold * - * Returns: 1 if it's ok */ -static int demote_ok(const struct gfs2_glock *gl) +struct gfs2_glock *gfs2_glock_hold(struct gfs2_glock *gl) { - const struct gfs2_glock_operations *glops = gl->gl_ops; - - if (gl->gl_state == LM_ST_UNLOCKED) - return 0; - if (!list_empty(&gl->gl_holders)) - return 0; - if (glops->go_demote_ok) - return glops->go_demote_ok(gl); - return 1; + GLOCK_BUG_ON(gl, __lockref_is_dead(&gl->gl_lockref)); + lockref_get(&gl->gl_lockref); + return gl; } - -void gfs2_glock_add_to_lru(struct gfs2_glock *gl) +static void gfs2_glock_add_to_lru(struct gfs2_glock *gl) { spin_lock(&lru_lock); + list_move_tail(&gl->gl_lru, &lru_list); - if (!list_empty(&gl->gl_lru)) - list_del_init(&gl->gl_lru); - else + if (!test_bit(GLF_LRU, &gl->gl_flags)) { + set_bit(GLF_LRU, &gl->gl_flags); atomic_inc(&lru_count); + } - list_add_tail(&gl->gl_lru, &lru_list); - set_bit(GLF_LRU, &gl->gl_flags); spin_unlock(&lru_lock); } -static void __gfs2_glock_remove_from_lru(struct gfs2_glock *gl) +static void gfs2_glock_remove_from_lru(struct gfs2_glock *gl) { - if (!list_empty(&gl->gl_lru)) { + spin_lock(&lru_lock); + if (test_bit(GLF_LRU, &gl->gl_flags)) { list_del_init(&gl->gl_lru); atomic_dec(&lru_count); clear_bit(GLF_LRU, &gl->gl_flags); } -} - -static void gfs2_glock_remove_from_lru(struct gfs2_glock *gl) -{ - spin_lock(&lru_lock); - __gfs2_glock_remove_from_lru(gl); spin_unlock(&lru_lock); } -/** - * gfs2_glock_put_nolock() - Decrement reference count on glock - * @gl: The glock to put - * - * This function should only be used if the caller has its own reference - * to the glock, in addition to the one it is dropping. +/* + * Enqueue the glock on the work queue. Passes one glock reference on to the + * work queue. */ +static void gfs2_glock_queue_work(struct gfs2_glock *gl, unsigned long delay) { + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + + if (!queue_delayed_work(sdp->sd_glock_wq, &gl->gl_work, delay)) { + /* + * We are holding the lockref spinlock, and the work was still + * queued above. The queued work (glock_work_func) takes that + * spinlock before dropping its glock reference(s), so it + * cannot have dropped them in the meantime. + */ + GLOCK_BUG_ON(gl, gl->gl_lockref.count < 2); + gl->gl_lockref.count--; + } +} -void gfs2_glock_put_nolock(struct gfs2_glock *gl) +static void __gfs2_glock_put(struct gfs2_glock *gl) { - if (atomic_dec_and_test(&gl->gl_ref)) - GLOCK_BUG_ON(gl, 1); + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + struct address_space *mapping = gfs2_glock2aspace(gl); + + lockref_mark_dead(&gl->gl_lockref); + spin_unlock(&gl->gl_lockref.lock); + gfs2_glock_remove_from_lru(gl); + GLOCK_BUG_ON(gl, !list_empty(&gl->gl_holders)); + if (mapping) { + truncate_inode_pages_final(mapping); + if (!gfs2_withdrawn(sdp)) + GLOCK_BUG_ON(gl, !mapping_empty(mapping)); + } + trace_gfs2_glock_put(gl); + sdp->sd_lockstruct.ls_ops->lm_put_lock(gl); +} + +static bool __gfs2_glock_put_or_lock(struct gfs2_glock *gl) +{ + if (lockref_put_or_lock(&gl->gl_lockref)) + return true; + GLOCK_BUG_ON(gl, gl->gl_lockref.count != 1); + if (gl->gl_state != LM_ST_UNLOCKED) { + gl->gl_lockref.count--; + gfs2_glock_add_to_lru(gl); + spin_unlock(&gl->gl_lockref.lock); + return true; + } + return false; } /** @@ -208,91 +272,103 @@ void gfs2_glock_put_nolock(struct gfs2_glock *gl) void gfs2_glock_put(struct gfs2_glock *gl) { - struct gfs2_sbd *sdp = gl->gl_sbd; - struct address_space *mapping = gfs2_glock2aspace(gl); + if (__gfs2_glock_put_or_lock(gl)) + return; - if (atomic_dec_and_lock(&gl->gl_ref, &lru_lock)) { - __gfs2_glock_remove_from_lru(gl); - spin_unlock(&lru_lock); - spin_lock_bucket(gl->gl_hash); - hlist_bl_del_rcu(&gl->gl_list); - spin_unlock_bucket(gl->gl_hash); - GLOCK_BUG_ON(gl, !list_empty(&gl->gl_holders)); - GLOCK_BUG_ON(gl, mapping && mapping->nrpages); - trace_gfs2_glock_put(gl); - sdp->sd_lockstruct.ls_ops->lm_put_lock(gl); - } + __gfs2_glock_put(gl); } -/** - * search_bucket() - Find struct gfs2_glock by lock number - * @bucket: the bucket to search - * @name: The lock name +/* + * gfs2_glock_put_async - Decrement reference count without sleeping + * @gl: The glock to put * - * Returns: NULL, or the struct gfs2_glock with the requested number + * Decrement the reference count on glock immediately unless it is the last + * reference. Defer putting the last reference to work queue context. */ - -static struct gfs2_glock *search_bucket(unsigned int hash, - const struct gfs2_sbd *sdp, - const struct lm_lockname *name) +void gfs2_glock_put_async(struct gfs2_glock *gl) { - struct gfs2_glock *gl; - struct hlist_bl_node *h; - - hlist_bl_for_each_entry_rcu(gl, h, &gl_hash_table[hash], gl_list) { - if (!lm_name_equal(&gl->gl_name, name)) - continue; - if (gl->gl_sbd != sdp) - continue; - if (atomic_inc_not_zero(&gl->gl_ref)) - return gl; - } + if (__gfs2_glock_put_or_lock(gl)) + return; - return NULL; + gfs2_glock_queue_work(gl, 0); + spin_unlock(&gl->gl_lockref.lock); } /** - * may_grant - check if its ok to grant a new lock + * may_grant - check if it's ok to grant a new lock * @gl: The glock + * @current_gh: One of the current holders of @gl * @gh: The lock request which we wish to grant * - * Returns: true if its ok to grant the lock + * With our current compatibility rules, if a glock has one or more active + * holders (HIF_HOLDER flag set), any of those holders can be passed in as + * @current_gh; they are all the same as far as compatibility with the new @gh + * goes. + * + * Returns true if it's ok to grant the lock. */ -static inline int may_grant(const struct gfs2_glock *gl, const struct gfs2_holder *gh) -{ - const struct gfs2_holder *gh_head = list_entry(gl->gl_holders.next, const struct gfs2_holder, gh_list); - if ((gh->gh_state == LM_ST_EXCLUSIVE || - gh_head->gh_state == LM_ST_EXCLUSIVE) && gh != gh_head) - return 0; +static inline bool may_grant(struct gfs2_glock *gl, + struct gfs2_holder *current_gh, + struct gfs2_holder *gh) +{ + if (current_gh) { + GLOCK_BUG_ON(gl, !test_bit(HIF_HOLDER, ¤t_gh->gh_iflags)); + + switch(current_gh->gh_state) { + case LM_ST_EXCLUSIVE: + /* + * Here we make a special exception to grant holders + * who agree to share the EX lock with other holders + * who also have the bit set. If the original holder + * has the LM_FLAG_NODE_SCOPE bit set, we grant more + * holders with the bit set. + */ + return gh->gh_state == LM_ST_EXCLUSIVE && + (current_gh->gh_flags & LM_FLAG_NODE_SCOPE) && + (gh->gh_flags & LM_FLAG_NODE_SCOPE); + + case LM_ST_SHARED: + case LM_ST_DEFERRED: + return gh->gh_state == current_gh->gh_state; + + default: + return false; + } + } + if (gl->gl_state == gh->gh_state) - return 1; + return true; if (gh->gh_flags & GL_EXACT) - return 0; + return false; if (gl->gl_state == LM_ST_EXCLUSIVE) { - if (gh->gh_state == LM_ST_SHARED && gh_head->gh_state == LM_ST_SHARED) - return 1; - if (gh->gh_state == LM_ST_DEFERRED && gh_head->gh_state == LM_ST_DEFERRED) - return 1; + return gh->gh_state == LM_ST_SHARED || + gh->gh_state == LM_ST_DEFERRED; } - if (gl->gl_state != LM_ST_UNLOCKED && (gh->gh_flags & LM_FLAG_ANY)) - return 1; - return 0; + if (gh->gh_flags & LM_FLAG_ANY) + return gl->gl_state != LM_ST_UNLOCKED; + return false; } static void gfs2_holder_wake(struct gfs2_holder *gh) { clear_bit(HIF_WAIT, &gh->gh_iflags); - smp_mb__after_clear_bit(); + smp_mb__after_atomic(); wake_up_bit(&gh->gh_iflags, HIF_WAIT); + if (gh->gh_flags & GL_ASYNC) { + struct gfs2_sbd *sdp = gh->gh_gl->gl_name.ln_sbd; + + wake_up(&sdp->sd_async_glock_wait); + } } /** * do_error - Something unexpected has happened during a lock request - * + * @gl: The glock + * @ret: The status from the DLM */ -static inline void do_error(struct gfs2_glock *gl, const int ret) +static void do_error(struct gfs2_glock *gl, const int ret) { struct gfs2_holder *gh, *tmp; @@ -312,57 +388,103 @@ static inline void do_error(struct gfs2_glock *gl, const int ret) } /** - * do_promote - promote as many requests as possible on the current queue - * @gl: The glock - * - * Returns: 1 if there is a blocked holder at the head of the list, or 2 - * if a type specific operation is underway. + * find_first_holder - find the first "holder" gh + * @gl: the glock */ -static int do_promote(struct gfs2_glock *gl) -__releases(&gl->gl_spin) -__acquires(&gl->gl_spin) +static inline struct gfs2_holder *find_first_holder(const struct gfs2_glock *gl) { + struct gfs2_holder *gh; + + if (!list_empty(&gl->gl_holders)) { + gh = list_first_entry(&gl->gl_holders, struct gfs2_holder, + gh_list); + if (test_bit(HIF_HOLDER, &gh->gh_iflags)) + return gh; + } + return NULL; +} + +/* + * gfs2_instantiate - Call the glops instantiate function + * @gh: The glock holder + * + * Returns: 0 if instantiate was successful, or error. + */ +int gfs2_instantiate(struct gfs2_holder *gh) +{ + struct gfs2_glock *gl = gh->gh_gl; const struct gfs2_glock_operations *glops = gl->gl_ops; - struct gfs2_holder *gh, *tmp; int ret; -restart: - list_for_each_entry_safe(gh, tmp, &gl->gl_holders, gh_list) { +again: + if (!test_bit(GLF_INSTANTIATE_NEEDED, &gl->gl_flags)) + goto done; + + /* + * Since we unlock the lockref lock, we set a flag to indicate + * instantiate is in progress. + */ + if (test_and_set_bit(GLF_INSTANTIATE_IN_PROG, &gl->gl_flags)) { + wait_on_bit(&gl->gl_flags, GLF_INSTANTIATE_IN_PROG, + TASK_UNINTERRUPTIBLE); + /* + * Here we just waited for a different instantiate to finish. + * But that may not have been successful, as when a process + * locks an inode glock _before_ it has an actual inode to + * instantiate into. So we check again. This process might + * have an inode to instantiate, so might be successful. + */ + goto again; + } + + ret = glops->go_instantiate(gl); + if (!ret) + clear_bit(GLF_INSTANTIATE_NEEDED, &gl->gl_flags); + clear_and_wake_up_bit(GLF_INSTANTIATE_IN_PROG, &gl->gl_flags); + if (ret) + return ret; + +done: + if (glops->go_held) + return glops->go_held(gh); + return 0; +} + +/** + * do_promote - promote as many requests as possible on the current queue + * @gl: The glock + */ + +static void do_promote(struct gfs2_glock *gl) +{ + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + struct gfs2_holder *gh, *current_gh; + + if (gfs2_withdrawn(sdp)) { + do_error(gl, LM_OUT_ERROR); + return; + } + + current_gh = find_first_holder(gl); + list_for_each_entry(gh, &gl->gl_holders, gh_list) { if (test_bit(HIF_HOLDER, &gh->gh_iflags)) continue; - if (may_grant(gl, gh)) { - if (gh->gh_list.prev == &gl->gl_holders && - glops->go_lock) { - spin_unlock(&gl->gl_spin); - /* FIXME: eliminate this eventually */ - ret = glops->go_lock(gh); - spin_lock(&gl->gl_spin); - if (ret) { - if (ret == 1) - return 2; - gh->gh_error = ret; - list_del_init(&gh->gh_list); - trace_gfs2_glock_queue(gh, 0); - gfs2_holder_wake(gh); - goto restart; - } - set_bit(HIF_HOLDER, &gh->gh_iflags); - trace_gfs2_promote(gh, 1); - gfs2_holder_wake(gh); - goto restart; - } - set_bit(HIF_HOLDER, &gh->gh_iflags); - trace_gfs2_promote(gh, 0); - gfs2_holder_wake(gh); - continue; + if (!may_grant(gl, current_gh, gh)) { + /* + * If we get here, it means we may not grant this + * holder for some reason. + */ + if (current_gh) + do_error(gl, 0); /* Fail queued try locks */ + break; } - if (gh->gh_list.prev == &gl->gl_holders) - return 1; - do_error(gl, 0); - break; + set_bit(HIF_HOLDER, &gh->gh_iflags); + trace_gfs2_promote(gh); + gfs2_holder_wake(gh); + if (!current_gh) + current_gh = gh; } - return 0; } /** @@ -382,28 +504,30 @@ static inline struct gfs2_holder *find_first_waiter(const struct gfs2_glock *gl) } /** - * state_change - record that the glock is now in a different state + * find_last_waiter - find the last gh that's waiting for the glock * @gl: the glock - * @new_state the new state * + * This also is a fast way of finding out if there are any waiters. */ -static void state_change(struct gfs2_glock *gl, unsigned int new_state) +static inline struct gfs2_holder *find_last_waiter(const struct gfs2_glock *gl) { - int held1, held2; + struct gfs2_holder *gh; - held1 = (gl->gl_state != LM_ST_UNLOCKED); - held2 = (new_state != LM_ST_UNLOCKED); + if (list_empty(&gl->gl_holders)) + return NULL; + gh = list_last_entry(&gl->gl_holders, struct gfs2_holder, gh_list); + return test_bit(HIF_HOLDER, &gh->gh_iflags) ? NULL : gh; +} - if (held1 != held2) { - if (held2) - gfs2_glock_hold(gl); - else - gfs2_glock_put_nolock(gl); - } - if (held1 && held2 && list_empty(&gl->gl_holders)) - clear_bit(GLF_QUEUED, &gl->gl_flags); +/** + * state_change - record that the glock is now in a different state + * @gl: the glock + * @new_state: the new state + */ +static void state_change(struct gfs2_glock *gl, unsigned int new_state) +{ if (new_state != gl->gl_target) /* shorten our minimum hold time */ gl->gl_hold_time = max(gl->gl_hold_time - GL_GLOCK_HOLD_DECR, @@ -412,11 +536,20 @@ static void state_change(struct gfs2_glock *gl, unsigned int new_state) gl->gl_tchange = jiffies; } +static void gfs2_set_demote(int nr, struct gfs2_glock *gl) +{ + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + + set_bit(nr, &gl->gl_flags); + smp_mb(); + wake_up(&sdp->sd_async_glock_wait); +} + static void gfs2_demote_wake(struct gfs2_glock *gl) { gl->gl_demote_state = LM_ST_EXCLUSIVE; clear_bit(GLF_DEMOTE, &gl->gl_flags); - smp_mb__after_clear_bit(); + smp_mb__after_atomic(); wake_up_bit(&gl->gl_flags, GLF_DEMOTE); } @@ -430,30 +563,31 @@ static void gfs2_demote_wake(struct gfs2_glock *gl) static void finish_xmote(struct gfs2_glock *gl, unsigned int ret) { const struct gfs2_glock_operations *glops = gl->gl_ops; - struct gfs2_holder *gh; - unsigned state = ret & LM_OUT_ST_MASK; - int rv; - spin_lock(&gl->gl_spin); - trace_gfs2_glock_state_change(gl, state); - state_change(gl, state); - gh = find_first_waiter(gl); + if (!(ret & ~LM_OUT_ST_MASK)) { + unsigned state = ret & LM_OUT_ST_MASK; + + trace_gfs2_glock_state_change(gl, state); + state_change(gl, state); + } /* Demote to UN request arrived during demote to SH or DF */ if (test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags) && - state != LM_ST_UNLOCKED && gl->gl_demote_state == LM_ST_UNLOCKED) + gl->gl_state != LM_ST_UNLOCKED && + gl->gl_demote_state == LM_ST_UNLOCKED) gl->gl_target = LM_ST_UNLOCKED; /* Check for state != intended state */ - if (unlikely(state != gl->gl_target)) { + if (unlikely(gl->gl_state != gl->gl_target)) { + struct gfs2_holder *gh = find_first_waiter(gl); + if (gh && !test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags)) { - /* move to back of queue and try next entry */ if (ret & LM_OUT_CANCELED) { - if ((gh->gh_flags & LM_FLAG_PRIORITY) == 0) - list_move_tail(&gh->gh_list, &gl->gl_holders); - gh = find_first_waiter(gl); - gl->gl_target = gh->gh_state; - goto retry; + list_del_init(&gh->gh_list); + trace_gfs2_glock_queue(gh, 0); + gfs2_holder_wake(gh); + gl->gl_target = gl->gl_state; + goto out; } /* Some error or failed "try lock" - report it */ if ((ret & LM_OUT_ERROR) || @@ -463,10 +597,9 @@ static void finish_xmote(struct gfs2_glock *gl, unsigned int ret) goto out; } } - switch(state) { + switch(gl->gl_state) { /* Unlocked due to conversion deadlock, try again */ case LM_ST_UNLOCKED: -retry: do_xmote(gl, gh, gl->gl_target); break; /* Conversion fails, unlock and try again */ @@ -475,34 +608,37 @@ retry: do_xmote(gl, gh, LM_ST_UNLOCKED); break; default: /* Everything else */ - printk(KERN_ERR "GFS2: wanted %u got %u\n", gl->gl_target, state); + fs_err(gl->gl_name.ln_sbd, + "glock %u:%llu requested=%u ret=%u\n", + gl->gl_name.ln_type, gl->gl_name.ln_number, + gl->gl_req, ret); GLOCK_BUG_ON(gl, 1); } - spin_unlock(&gl->gl_spin); return; } /* Fast path - we got what we asked for */ - if (test_and_clear_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags)) + if (test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags)) { + clear_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags); gfs2_demote_wake(gl); - if (state != LM_ST_UNLOCKED) { + } + if (gl->gl_state != LM_ST_UNLOCKED) { if (glops->go_xmote_bh) { - spin_unlock(&gl->gl_spin); - rv = glops->go_xmote_bh(gl, gh); - spin_lock(&gl->gl_spin); + int rv; + + spin_unlock(&gl->gl_lockref.lock); + rv = glops->go_xmote_bh(gl); + spin_lock(&gl->gl_lockref.lock); if (rv) { do_error(gl, rv); goto out; } } - rv = do_promote(gl); - if (rv == 2) - goto out_locked; + do_promote(gl); } out: - clear_bit(GLF_LOCK, &gl->gl_flags); -out_locked: - spin_unlock(&gl->gl_spin); + if (!test_bit(GLF_CANCELING, &gl->gl_flags)) + clear_bit(GLF_LOCK, &gl->gl_flags); } /** @@ -513,69 +649,93 @@ out_locked: * */ -static void do_xmote(struct gfs2_glock *gl, struct gfs2_holder *gh, unsigned int target) -__releases(&gl->gl_spin) -__acquires(&gl->gl_spin) +static void do_xmote(struct gfs2_glock *gl, struct gfs2_holder *gh, + unsigned int target) +__releases(&gl->gl_lockref.lock) +__acquires(&gl->gl_lockref.lock) { const struct gfs2_glock_operations *glops = gl->gl_ops; - struct gfs2_sbd *sdp = gl->gl_sbd; - unsigned int lck_flags = gh ? gh->gh_flags : 0; + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + struct lm_lockstruct *ls = &sdp->sd_lockstruct; int ret; - lck_flags &= (LM_FLAG_TRY | LM_FLAG_TRY_1CB | LM_FLAG_NOEXP | - LM_FLAG_PRIORITY); + /* + * When a filesystem is withdrawing, the remaining cluster nodes will + * take care of recovering the withdrawing node's journal. We only + * need to make sure that once we trigger remote recovery, we won't + * write to the shared block device anymore. This means that here, + * + * - no new writes to the filesystem must be triggered (->go_sync()). + * + * - any cached data should be discarded by calling ->go_inval(), dirty + * or not and journaled or unjournaled. + * + * - no more dlm locking operations should be issued (->lm_lock()). + */ + GLOCK_BUG_ON(gl, gl->gl_state == target); GLOCK_BUG_ON(gl, gl->gl_state == gl->gl_target); - if ((target == LM_ST_UNLOCKED || target == LM_ST_DEFERRED) && - glops->go_inval) { - set_bit(GLF_INVALIDATE_IN_PROGRESS, &gl->gl_flags); - do_error(gl, 0); /* Fail queued try locks */ - } - gl->gl_req = target; - set_bit(GLF_BLOCKING, &gl->gl_flags); - if ((gl->gl_req == LM_ST_UNLOCKED) || - (gl->gl_state == LM_ST_EXCLUSIVE) || - (lck_flags & (LM_FLAG_TRY|LM_FLAG_TRY_1CB))) - clear_bit(GLF_BLOCKING, &gl->gl_flags); - spin_unlock(&gl->gl_spin); - if (glops->go_sync) - glops->go_sync(gl); - if (test_bit(GLF_INVALIDATE_IN_PROGRESS, &gl->gl_flags)) - glops->go_inval(gl, target == LM_ST_DEFERRED ? 0 : DIO_METADATA); - clear_bit(GLF_INVALIDATE_IN_PROGRESS, &gl->gl_flags); - gfs2_glock_hold(gl); - if (sdp->sd_lockstruct.ls_ops->lm_lock) { - /* lock_dlm */ - ret = sdp->sd_lockstruct.ls_ops->lm_lock(gl, target, lck_flags); + if (!glops->go_inval || !glops->go_sync) + goto skip_inval; + + spin_unlock(&gl->gl_lockref.lock); + if (!gfs2_withdrawn(sdp)) { + ret = glops->go_sync(gl); if (ret) { - printk(KERN_ERR "GFS2: lm_lock ret %d\n", ret); - GLOCK_BUG_ON(gl, 1); + if (cmpxchg(&sdp->sd_log_error, 0, ret)) { + fs_err(sdp, "Error %d syncing glock\n", ret); + gfs2_dump_glock(NULL, gl, true); + gfs2_withdraw(sdp); + } } - } else { /* lock_nolock */ - finish_xmote(gl, target); - if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) - gfs2_glock_put(gl); } - spin_lock(&gl->gl_spin); -} + if (target == LM_ST_UNLOCKED || target == LM_ST_DEFERRED) + glops->go_inval(gl, target == LM_ST_DEFERRED ? 0 : DIO_METADATA); + spin_lock(&gl->gl_lockref.lock); -/** - * find_first_holder - find the first "holder" gh - * @gl: the glock - */ +skip_inval: + if (gfs2_withdrawn(sdp)) { + if (target != LM_ST_UNLOCKED) + target = LM_OUT_ERROR; + goto out; + } -static inline struct gfs2_holder *find_first_holder(const struct gfs2_glock *gl) -{ - struct gfs2_holder *gh; + if (ls->ls_ops->lm_lock) { + set_bit(GLF_PENDING_REPLY, &gl->gl_flags); + spin_unlock(&gl->gl_lockref.lock); + ret = ls->ls_ops->lm_lock(gl, target, gh ? gh->gh_flags : 0); + spin_lock(&gl->gl_lockref.lock); - if (!list_empty(&gl->gl_holders)) { - gh = list_entry(gl->gl_holders.next, struct gfs2_holder, gh_list); - if (test_bit(HIF_HOLDER, &gh->gh_iflags)) - return gh; + if (!ret) { + /* The operation will be completed asynchronously. */ + gl->gl_lockref.count++; + return; + } + clear_bit(GLF_PENDING_REPLY, &gl->gl_flags); + + if (ret == -ENODEV) { + /* + * The lockspace has been released and the lock has + * been unlocked implicitly. + */ + if (target != LM_ST_UNLOCKED) { + target = LM_OUT_ERROR; + goto out; + } + } else { + fs_err(sdp, "lm_lock ret %d\n", ret); + GLOCK_BUG_ON(gl, !gfs2_withdrawn(sdp)); + return; + } } - return NULL; + +out: + /* Complete the operation now. */ + finish_xmote(gl, target); + gl->gl_lockref.count++; + gfs2_glock_queue_work(gl, 0); } /** @@ -586,19 +746,29 @@ static inline struct gfs2_holder *find_first_holder(const struct gfs2_glock *gl) */ static void run_queue(struct gfs2_glock *gl, const int nonblock) -__releases(&gl->gl_spin) -__acquires(&gl->gl_spin) +__releases(&gl->gl_lockref.lock) +__acquires(&gl->gl_lockref.lock) { - struct gfs2_holder *gh = NULL; - int ret; + struct gfs2_holder *gh; - if (test_and_set_bit(GLF_LOCK, &gl->gl_flags)) + if (test_bit(GLF_LOCK, &gl->gl_flags)) return; - + set_bit(GLF_LOCK, &gl->gl_flags); + + /* + * The GLF_DEMOTE_IN_PROGRESS flag is only set intermittently during + * locking operations. We have just started a locking operation by + * setting the GLF_LOCK flag, so the GLF_DEMOTE_IN_PROGRESS flag must + * be cleared. + */ GLOCK_BUG_ON(gl, test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags)); - if (test_bit(GLF_DEMOTE, &gl->gl_flags) && - gl->gl_demote_state != gl->gl_state) { + if (test_bit(GLF_DEMOTE, &gl->gl_flags)) { + if (gl->gl_demote_state == gl->gl_state) { + gfs2_demote_wake(gl); + goto promote; + } + if (find_first_holder(gl)) goto out_unlock; if (nonblock) @@ -606,56 +776,206 @@ __acquires(&gl->gl_spin) set_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags); GLOCK_BUG_ON(gl, gl->gl_demote_state == LM_ST_EXCLUSIVE); gl->gl_target = gl->gl_demote_state; - } else { - if (test_bit(GLF_DEMOTE, &gl->gl_flags)) - gfs2_demote_wake(gl); - ret = do_promote(gl); - if (ret == 0) - goto out_unlock; - if (ret == 2) - goto out; - gh = find_first_waiter(gl); - gl->gl_target = gh->gh_state; - if (!(gh->gh_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB))) - do_error(gl, 0); /* Fail queued try locks */ + do_xmote(gl, NULL, gl->gl_target); + return; } + +promote: + do_promote(gl); + if (find_first_holder(gl)) + goto out_unlock; + gh = find_first_waiter(gl); + if (!gh) + goto out_unlock; + if (nonblock) + goto out_sched; + gl->gl_target = gh->gh_state; + if (!(gh->gh_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB))) + do_error(gl, 0); /* Fail queued try locks */ do_xmote(gl, gh, gl->gl_target); -out: return; out_sched: clear_bit(GLF_LOCK, &gl->gl_flags); - smp_mb__after_clear_bit(); - gfs2_glock_hold(gl); - if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) - gfs2_glock_put_nolock(gl); + gl->gl_lockref.count++; + gfs2_glock_queue_work(gl, 0); return; out_unlock: clear_bit(GLF_LOCK, &gl->gl_flags); - smp_mb__after_clear_bit(); - return; } -static void delete_work_func(struct work_struct *work) +/** + * glock_set_object - set the gl_object field of a glock + * @gl: the glock + * @object: the object + */ +void glock_set_object(struct gfs2_glock *gl, void *object) +{ + void *prev_object; + + spin_lock(&gl->gl_lockref.lock); + prev_object = gl->gl_object; + gl->gl_object = object; + spin_unlock(&gl->gl_lockref.lock); + if (gfs2_assert_warn(gl->gl_name.ln_sbd, prev_object == NULL)) + gfs2_dump_glock(NULL, gl, true); +} + +/** + * glock_clear_object - clear the gl_object field of a glock + * @gl: the glock + * @object: object the glock currently points at + */ +void glock_clear_object(struct gfs2_glock *gl, void *object) +{ + void *prev_object; + + spin_lock(&gl->gl_lockref.lock); + prev_object = gl->gl_object; + gl->gl_object = NULL; + spin_unlock(&gl->gl_lockref.lock); + if (gfs2_assert_warn(gl->gl_name.ln_sbd, prev_object == object)) + gfs2_dump_glock(NULL, gl, true); +} + +void gfs2_inode_remember_delete(struct gfs2_glock *gl, u64 generation) +{ + struct gfs2_inode_lvb *ri = (void *)gl->gl_lksb.sb_lvbptr; + + if (ri->ri_magic == 0) + ri->ri_magic = cpu_to_be32(GFS2_MAGIC); + if (ri->ri_magic == cpu_to_be32(GFS2_MAGIC)) + ri->ri_generation_deleted = cpu_to_be64(generation); +} + +bool gfs2_inode_already_deleted(struct gfs2_glock *gl, u64 generation) +{ + struct gfs2_inode_lvb *ri = (void *)gl->gl_lksb.sb_lvbptr; + + if (ri->ri_magic != cpu_to_be32(GFS2_MAGIC)) + return false; + return generation <= be64_to_cpu(ri->ri_generation_deleted); +} + +static void gfs2_glock_poke(struct gfs2_glock *gl) +{ + int flags = LM_FLAG_TRY_1CB | LM_FLAG_ANY | GL_SKIP; + struct gfs2_holder gh; + int error; + + __gfs2_holder_init(gl, LM_ST_SHARED, flags, &gh, _RET_IP_); + error = gfs2_glock_nq(&gh); + if (!error) + gfs2_glock_dq(&gh); + gfs2_holder_uninit(&gh); +} + +static struct gfs2_inode *gfs2_grab_existing_inode(struct gfs2_glock *gl) { - struct gfs2_glock *gl = container_of(work, struct gfs2_glock, gl_delete); - struct gfs2_sbd *sdp = gl->gl_sbd; struct gfs2_inode *ip; - struct inode *inode; - u64 no_addr = gl->gl_name.ln_number; + spin_lock(&gl->gl_lockref.lock); ip = gl->gl_object; - /* Note: Unsafe to dereference ip as we don't hold right refs/locks */ + if (ip && !igrab(&ip->i_inode)) + ip = NULL; + spin_unlock(&gl->gl_lockref.lock); + if (ip) { + wait_on_new_inode(&ip->i_inode); + if (is_bad_inode(&ip->i_inode)) { + iput(&ip->i_inode); + ip = NULL; + } + } + return ip; +} - if (ip) - inode = gfs2_ilookup(sdp->sd_vfs, no_addr, 1); - else - inode = gfs2_lookup_by_inum(sdp, no_addr, NULL, GFS2_BLKST_UNLINKED); - if (inode && !IS_ERR(inode)) { - d_prune_aliases(inode); - iput(inode); +static void gfs2_try_to_evict(struct gfs2_glock *gl) +{ + struct gfs2_inode *ip; + + /* + * If there is contention on the iopen glock and we have an inode, try + * to grab and release the inode so that it can be evicted. The + * GLF_DEFER_DELETE flag indicates to gfs2_evict_inode() that the inode + * should not be deleted locally. This will allow the remote node to + * go ahead and delete the inode without us having to do it, which will + * avoid rgrp glock thrashing. + * + * The remote node is likely still holding the corresponding inode + * glock, so it will run before we get to verify that the delete has + * happened below. (Verification is triggered by the call to + * gfs2_queue_verify_delete() in gfs2_evict_inode().) + */ + ip = gfs2_grab_existing_inode(gl); + if (ip) { + set_bit(GLF_DEFER_DELETE, &gl->gl_flags); + d_prune_aliases(&ip->i_inode); + iput(&ip->i_inode); + clear_bit(GLF_DEFER_DELETE, &gl->gl_flags); + + /* If the inode was evicted, gl->gl_object will now be NULL. */ + ip = gfs2_grab_existing_inode(gl); + if (ip) { + gfs2_glock_poke(ip->i_gl); + iput(&ip->i_inode); + } } +} + +bool gfs2_queue_try_to_evict(struct gfs2_glock *gl) +{ + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + + if (test_and_set_bit(GLF_TRY_TO_EVICT, &gl->gl_flags)) + return false; + return !mod_delayed_work(sdp->sd_delete_wq, &gl->gl_delete, 0); +} + +bool gfs2_queue_verify_delete(struct gfs2_glock *gl, bool later) +{ + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + unsigned long delay; + + if (test_and_set_bit(GLF_VERIFY_DELETE, &gl->gl_flags)) + return false; + delay = later ? HZ + get_random_long() % (HZ * 9) : 0; + return queue_delayed_work(sdp->sd_delete_wq, &gl->gl_delete, delay); +} + +static void delete_work_func(struct work_struct *work) +{ + struct delayed_work *dwork = to_delayed_work(work); + struct gfs2_glock *gl = container_of(dwork, struct gfs2_glock, gl_delete); + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + bool verify_delete = test_and_clear_bit(GLF_VERIFY_DELETE, &gl->gl_flags); + + /* + * Check for the GLF_VERIFY_DELETE above: this ensures that we won't + * immediately process GLF_VERIFY_DELETE work that the below call to + * gfs2_try_to_evict() queues. + */ + + if (test_and_clear_bit(GLF_TRY_TO_EVICT, &gl->gl_flags)) + gfs2_try_to_evict(gl); + + if (verify_delete) { + u64 no_addr = gl->gl_name.ln_number; + struct inode *inode; + + inode = gfs2_lookup_by_inum(sdp, no_addr, gl->gl_no_formal_ino, + GFS2_BLKST_UNLINKED); + if (IS_ERR(inode)) { + if (PTR_ERR(inode) == -EAGAIN && + !test_bit(SDF_KILL, &sdp->sd_flags) && + gfs2_queue_verify_delete(gl, true)) + return; + } else { + d_prune_aliases(inode); + iput(inode); + } + } + gfs2_glock_put(gl); } @@ -663,39 +983,84 @@ static void glock_work_func(struct work_struct *work) { unsigned long delay = 0; struct gfs2_glock *gl = container_of(work, struct gfs2_glock, gl_work.work); - int drop_ref = 0; + unsigned int drop_refs = 1; - if (test_and_clear_bit(GLF_REPLY_PENDING, &gl->gl_flags)) { + spin_lock(&gl->gl_lockref.lock); + if (test_bit(GLF_HAVE_REPLY, &gl->gl_flags)) { + clear_bit(GLF_HAVE_REPLY, &gl->gl_flags); finish_xmote(gl, gl->gl_reply); - drop_ref = 1; + drop_refs++; } - spin_lock(&gl->gl_spin); if (test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags) && gl->gl_state != LM_ST_UNLOCKED && gl->gl_demote_state != LM_ST_EXCLUSIVE) { - unsigned long holdtime, now = jiffies; + if (gl->gl_name.ln_type == LM_TYPE_INODE) { + unsigned long holdtime, now = jiffies; - holdtime = gl->gl_tchange + gl->gl_hold_time; - if (time_before(now, holdtime)) - delay = holdtime - now; + holdtime = gl->gl_tchange + gl->gl_hold_time; + if (time_before(now, holdtime)) + delay = holdtime - now; + } if (!delay) { clear_bit(GLF_PENDING_DEMOTE, &gl->gl_flags); - set_bit(GLF_DEMOTE, &gl->gl_flags); + gfs2_set_demote(GLF_DEMOTE, gl); } } run_queue(gl, 0); - spin_unlock(&gl->gl_spin); - if (!delay) - gfs2_glock_put(gl); - else { - if (gl->gl_name.ln_type != LM_TYPE_INODE) - delay = 0; - if (queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0) - gfs2_glock_put(gl); + if (delay) { + /* Keep one glock reference for the work we requeue. */ + drop_refs--; + gfs2_glock_queue_work(gl, delay); } - if (drop_ref) - gfs2_glock_put(gl); + + /* Drop the remaining glock references manually. */ + GLOCK_BUG_ON(gl, gl->gl_lockref.count < drop_refs); + gl->gl_lockref.count -= drop_refs; + if (!gl->gl_lockref.count) { + if (gl->gl_state == LM_ST_UNLOCKED) { + __gfs2_glock_put(gl); + return; + } + gfs2_glock_add_to_lru(gl); + } + spin_unlock(&gl->gl_lockref.lock); +} + +static struct gfs2_glock *find_insert_glock(struct lm_lockname *name, + struct gfs2_glock *new) +{ + struct wait_glock_queue wait; + wait_queue_head_t *wq = glock_waitqueue(name); + struct gfs2_glock *gl; + + wait.name = name; + init_wait(&wait.wait); + wait.wait.func = glock_wake_function; + +again: + prepare_to_wait(wq, &wait.wait, TASK_UNINTERRUPTIBLE); + rcu_read_lock(); + if (new) { + gl = rhashtable_lookup_get_insert_fast(&gl_hash_table, + &new->gl_node, ht_parms); + if (IS_ERR(gl)) + goto out; + } else { + gl = rhashtable_lookup_fast(&gl_hash_table, + name, ht_parms); + } + if (gl && !lockref_get_not_dead(&gl->gl_lockref)) { + rcu_read_unlock(); + schedule(); + goto again; + } +out: + rcu_read_unlock(); + finish_wait(wq, &wait.wait); + if (gl) + gfs2_glock_remove_from_lru(gl); + return gl; } /** @@ -715,52 +1080,52 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number, const struct gfs2_glock_operations *glops, int create, struct gfs2_glock **glp) { - struct super_block *s = sdp->sd_vfs; - struct lm_lockname name = { .ln_number = number, .ln_type = glops->go_type }; + struct lm_lockname name = { .ln_number = number, + .ln_type = glops->go_type, + .ln_sbd = sdp }; struct gfs2_glock *gl, *tmp; - unsigned int hash = gl_hash(sdp, &name); struct address_space *mapping; - struct kmem_cache *cachep; - rcu_read_lock(); - gl = search_bucket(hash, sdp, &name); - rcu_read_unlock(); - - *glp = gl; + gl = find_insert_glock(&name, NULL); if (gl) - return 0; + goto found; if (!create) return -ENOENT; - if (glops->go_flags & GLOF_ASPACE) - cachep = gfs2_glock_aspace_cachep; - else - cachep = gfs2_glock_cachep; - gl = kmem_cache_alloc(cachep, GFP_KERNEL); - if (!gl) - return -ENOMEM; - + if (glops->go_flags & GLOF_ASPACE) { + struct gfs2_glock_aspace *gla = + kmem_cache_alloc(gfs2_glock_aspace_cachep, GFP_NOFS); + if (!gla) + return -ENOMEM; + gl = &gla->glock; + } else { + gl = kmem_cache_alloc(gfs2_glock_cachep, GFP_NOFS); + if (!gl) + return -ENOMEM; + } memset(&gl->gl_lksb, 0, sizeof(struct dlm_lksb)); + gl->gl_ops = glops; if (glops->go_flags & GLOF_LVB) { - gl->gl_lksb.sb_lvbptr = kzalloc(GFS2_MIN_LVB_SIZE, GFP_KERNEL); + gl->gl_lksb.sb_lvbptr = kzalloc(GDLM_LVB_SIZE, GFP_NOFS); if (!gl->gl_lksb.sb_lvbptr) { - kmem_cache_free(cachep, gl); + gfs2_glock_dealloc(&gl->gl_rcu); return -ENOMEM; } } atomic_inc(&sdp->sd_glock_disposal); - gl->gl_sbd = sdp; - gl->gl_flags = 0; + gl->gl_node.next = NULL; + gl->gl_flags = BIT(GLF_INITIAL); + if (glops->go_instantiate) + gl->gl_flags |= BIT(GLF_INSTANTIATE_NEEDED); gl->gl_name = name; - atomic_set(&gl->gl_ref, 1); + lockref_init(&gl->gl_lockref); + lockdep_set_subclass(&gl->gl_lockref.lock, glops->go_subclass); gl->gl_state = LM_ST_UNLOCKED; gl->gl_target = LM_ST_UNLOCKED; gl->gl_demote_state = LM_ST_EXCLUSIVE; - gl->gl_hash = hash; - gl->gl_ops = glops; - gl->gl_dstamp = ktime_set(0, 0); + gl->gl_dstamp = 0; preempt_disable(); /* We use the global stats to estimate the initial per-glock stats */ gl->gl_stats = this_cpu_ptr(sdp->sd_lkstats)->lkstats[glops->go_type]; @@ -771,58 +1136,57 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number, gl->gl_object = NULL; gl->gl_hold_time = GL_GLOCK_DFT_HOLD; INIT_DELAYED_WORK(&gl->gl_work, glock_work_func); - INIT_WORK(&gl->gl_delete, delete_work_func); + if (gl->gl_name.ln_type == LM_TYPE_IOPEN) + INIT_DELAYED_WORK(&gl->gl_delete, delete_work_func); mapping = gfs2_glock2aspace(gl); if (mapping) { + gfp_t gfp_mask; + mapping->a_ops = &gfs2_meta_aops; - mapping->host = s->s_bdev->bd_inode; + mapping->host = sdp->sd_inode; mapping->flags = 0; - mapping_set_gfp_mask(mapping, GFP_NOFS); - mapping->private_data = NULL; - mapping->backing_dev_info = s->s_bdi; + gfp_mask = mapping_gfp_mask(sdp->sd_inode->i_mapping); + mapping_set_gfp_mask(mapping, gfp_mask); + mapping->i_private_data = NULL; mapping->writeback_index = 0; } - spin_lock_bucket(hash); - tmp = search_bucket(hash, sdp, &name); + tmp = find_insert_glock(&name, gl); if (tmp) { - spin_unlock_bucket(hash); - kfree(gl->gl_lksb.sb_lvbptr); - kmem_cache_free(cachep, gl); - atomic_dec(&sdp->sd_glock_disposal); + gfs2_glock_dealloc(&gl->gl_rcu); + if (atomic_dec_and_test(&sdp->sd_glock_disposal)) + wake_up(&sdp->sd_kill_wait); + + if (IS_ERR(tmp)) + return PTR_ERR(tmp); gl = tmp; - } else { - hlist_bl_add_head_rcu(&gl->gl_list, &gl_hash_table[hash]); - spin_unlock_bucket(hash); } +found: *glp = gl; - return 0; } /** - * gfs2_holder_init - initialize a struct gfs2_holder in the default way + * __gfs2_holder_init - initialize a struct gfs2_holder in the default way * @gl: the glock * @state: the state we're requesting * @flags: the modifier flags * @gh: the holder structure - * + * @ip: caller's return address for debugging */ -void gfs2_holder_init(struct gfs2_glock *gl, unsigned int state, unsigned flags, - struct gfs2_holder *gh) +void __gfs2_holder_init(struct gfs2_glock *gl, unsigned int state, u16 flags, + struct gfs2_holder *gh, unsigned long ip) { INIT_LIST_HEAD(&gh->gh_list); - gh->gh_gl = gl; - gh->gh_ip = (unsigned long)__builtin_return_address(0); + gh->gh_gl = gfs2_glock_hold(gl); + gh->gh_ip = ip; gh->gh_owner_pid = get_pid(task_pid(current)); gh->gh_state = state; gh->gh_flags = flags; - gh->gh_error = 0; gh->gh_iflags = 0; - gfs2_glock_hold(gl); } /** @@ -835,14 +1199,13 @@ void gfs2_holder_init(struct gfs2_glock *gl, unsigned int state, unsigned flags, * */ -void gfs2_holder_reinit(unsigned int state, unsigned flags, struct gfs2_holder *gh) +void gfs2_holder_reinit(unsigned int state, u16 flags, struct gfs2_holder *gh) { gh->gh_state = state; gh->gh_flags = flags; gh->gh_iflags = 0; - gh->gh_ip = (unsigned long)__builtin_return_address(0); - if (gh->gh_owner_pid) - put_pid(gh->gh_owner_pid); + gh->gh_ip = _RET_IP_; + put_pid(gh->gh_owner_pid); gh->gh_owner_pid = get_pid(task_pid(current)); } @@ -856,29 +1219,38 @@ void gfs2_holder_uninit(struct gfs2_holder *gh) { put_pid(gh->gh_owner_pid); gfs2_glock_put(gh->gh_gl); - gh->gh_gl = NULL; + gfs2_holder_mark_uninitialized(gh); gh->gh_ip = 0; } +static void gfs2_glock_update_hold_time(struct gfs2_glock *gl, + unsigned long start_time) +{ + /* Have we waited longer that a second? */ + if (time_after(jiffies, start_time + HZ)) { + /* Lengthen the minimum hold time. */ + gl->gl_hold_time = min(gl->gl_hold_time + GL_GLOCK_HOLD_INCR, + GL_GLOCK_MAX_HOLD); + } +} + /** - * gfs2_glock_holder_wait - * @word: unused + * gfs2_glock_holder_ready - holder is ready and its error code can be collected + * @gh: the glock holder * - * This function and gfs2_glock_demote_wait both show up in the WCHAN - * field. Thus I've separated these otherwise identical functions in - * order to be more informative to the user. + * Called when a glock holder no longer needs to be waited for because it is + * now either held (HIF_HOLDER set; gh_error == 0), or acquiring the lock has + * failed (gh_error != 0). */ -static int gfs2_glock_holder_wait(void *word) +int gfs2_glock_holder_ready(struct gfs2_holder *gh) { - schedule(); - return 0; -} - -static int gfs2_glock_demote_wait(void *word) -{ - schedule(); - return 0; + if (gh->gh_error || (gh->gh_flags & GL_SKIP)) + return gh->gh_error; + gh->gh_error = gfs2_instantiate(gh); + if (gh->gh_error) + gfs2_glock_dq(gh); + return gh->gh_error; } /** @@ -890,33 +1262,93 @@ static int gfs2_glock_demote_wait(void *word) int gfs2_glock_wait(struct gfs2_holder *gh) { - unsigned long time1 = jiffies; + unsigned long start_time = jiffies; might_sleep(); - wait_on_bit(&gh->gh_iflags, HIF_WAIT, gfs2_glock_holder_wait, TASK_UNINTERRUPTIBLE); - if (time_after(jiffies, time1 + HZ)) /* have we waited > a second? */ - /* Lengthen the minimum hold time. */ - gh->gh_gl->gl_hold_time = min(gh->gh_gl->gl_hold_time + - GL_GLOCK_HOLD_INCR, - GL_GLOCK_MAX_HOLD); - return gh->gh_error; + wait_on_bit(&gh->gh_iflags, HIF_WAIT, TASK_UNINTERRUPTIBLE); + gfs2_glock_update_hold_time(gh->gh_gl, start_time); + return gfs2_glock_holder_ready(gh); +} + +static int glocks_pending(unsigned int num_gh, struct gfs2_holder *ghs) +{ + int i; + + for (i = 0; i < num_gh; i++) + if (test_bit(HIF_WAIT, &ghs[i].gh_iflags)) + return 1; + return 0; } /** - * handle_callback - process a demote request + * gfs2_glock_async_wait - wait on multiple asynchronous glock acquisitions + * @num_gh: the number of holders in the array + * @ghs: the glock holder array + * + * Returns: 0 on success, meaning all glocks have been granted and are held. + * -ESTALE if the request timed out, meaning all glocks were released, + * and the caller should retry the operation. + */ + +int gfs2_glock_async_wait(unsigned int num_gh, struct gfs2_holder *ghs) +{ + struct gfs2_sbd *sdp = ghs[0].gh_gl->gl_name.ln_sbd; + int i, ret = 0, timeout = 0; + unsigned long start_time = jiffies; + + might_sleep(); + /* + * Total up the (minimum hold time * 2) of all glocks and use that to + * determine the max amount of time we should wait. + */ + for (i = 0; i < num_gh; i++) + timeout += ghs[i].gh_gl->gl_hold_time << 1; + + if (!wait_event_timeout(sdp->sd_async_glock_wait, + !glocks_pending(num_gh, ghs), timeout)) { + ret = -ESTALE; /* request timed out. */ + goto out; + } + + for (i = 0; i < num_gh; i++) { + struct gfs2_holder *gh = &ghs[i]; + int ret2; + + if (test_bit(HIF_HOLDER, &gh->gh_iflags)) { + gfs2_glock_update_hold_time(gh->gh_gl, + start_time); + } + ret2 = gfs2_glock_holder_ready(gh); + if (!ret) + ret = ret2; + } + +out: + if (ret) { + for (i = 0; i < num_gh; i++) { + struct gfs2_holder *gh = &ghs[i]; + + gfs2_glock_dq(gh); + } + } + return ret; +} + +/** + * request_demote - process a demote request * @gl: the glock * @state: the state the caller wants us to change to + * @delay: zero to demote immediately; otherwise pending demote + * @remote: true if this came from a different cluster node * * There are only two requests that we are going to see in actual * practise: LM_ST_SHARED and LM_ST_UNLOCKED */ -static void handle_callback(struct gfs2_glock *gl, unsigned int state, - unsigned long delay, bool remote) +static void request_demote(struct gfs2_glock *gl, unsigned int state, + unsigned long delay, bool remote) { - int bit = delay ? GLF_PENDING_DEMOTE : GLF_DEMOTE; - - set_bit(bit, &gl->gl_flags); + gfs2_set_demote(delay ? GLF_PENDING_DEMOTE : GLF_DEMOTE, gl); if (gl->gl_demote_state == LM_ST_EXCLUSIVE) { gl->gl_demote_state = state; gl->gl_demote_time = jiffies; @@ -942,12 +1374,37 @@ void gfs2_print_dbg(struct seq_file *seq, const char *fmt, ...) vaf.fmt = fmt; vaf.va = &args; - printk(KERN_ERR " %pV", &vaf); + pr_err("%pV", &vaf); } va_end(args); } +static bool gfs2_should_queue_trylock(struct gfs2_glock *gl, + struct gfs2_holder *gh) +{ + struct gfs2_holder *current_gh, *gh2; + + current_gh = find_first_holder(gl); + if (current_gh && !may_grant(gl, current_gh, gh)) + return false; + + list_for_each_entry(gh2, &gl->gl_holders, gh_list) { + if (test_bit(HIF_HOLDER, &gh2->gh_iflags)) + continue; + if (!(gh2->gh_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB))) + return false; + } + return true; +} + +static inline bool pid_is_meaningful(const struct gfs2_holder *gh) +{ + if (!(gh->gh_flags & GL_NOPID)) + return true; + return !test_bit(HIF_HOLDER, &gh->gh_iflags); +} + /** * add_to_queue - Add a holder to the wait queue (but look for recursion) * @gh: the holder structure to add @@ -959,73 +1416,47 @@ void gfs2_print_dbg(struct seq_file *seq, const char *fmt, ...) */ static inline void add_to_queue(struct gfs2_holder *gh) -__releases(&gl->gl_spin) -__acquires(&gl->gl_spin) { struct gfs2_glock *gl = gh->gh_gl; - struct gfs2_sbd *sdp = gl->gl_sbd; - struct list_head *insert_pt = NULL; + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; struct gfs2_holder *gh2; - int try_futile = 0; - BUG_ON(gh->gh_owner_pid == NULL); + GLOCK_BUG_ON(gl, gh->gh_owner_pid == NULL); if (test_and_set_bit(HIF_WAIT, &gh->gh_iflags)) - BUG(); + GLOCK_BUG_ON(gl, true); - if (gh->gh_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB)) { - if (test_bit(GLF_LOCK, &gl->gl_flags)) - try_futile = !may_grant(gl, gh); - if (test_bit(GLF_INVALIDATE_IN_PROGRESS, &gl->gl_flags)) - goto fail; + if ((gh->gh_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB)) && + !gfs2_should_queue_trylock(gl, gh)) { + gh->gh_error = GLR_TRYFAILED; + gfs2_holder_wake(gh); + return; } list_for_each_entry(gh2, &gl->gl_holders, gh_list) { - if (unlikely(gh2->gh_owner_pid == gh->gh_owner_pid && - (gh->gh_gl->gl_ops->go_type != LM_TYPE_FLOCK))) - goto trap_recursive; - if (try_futile && - !(gh2->gh_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB))) { -fail: - gh->gh_error = GLR_TRYFAILED; - gfs2_holder_wake(gh); - return; - } - if (test_bit(HIF_HOLDER, &gh2->gh_iflags)) + if (likely(gh2->gh_owner_pid != gh->gh_owner_pid)) + continue; + if (gh->gh_gl->gl_ops->go_type == LM_TYPE_FLOCK) continue; - if (unlikely((gh->gh_flags & LM_FLAG_PRIORITY) && !insert_pt)) - insert_pt = &gh2->gh_list; + if (!pid_is_meaningful(gh2)) + continue; + goto trap_recursive; } - set_bit(GLF_QUEUED, &gl->gl_flags); trace_gfs2_glock_queue(gh, 1); gfs2_glstats_inc(gl, GFS2_LKS_QCOUNT); gfs2_sbstats_inc(gl, GFS2_LKS_QCOUNT); - if (likely(insert_pt == NULL)) { - list_add_tail(&gh->gh_list, &gl->gl_holders); - if (unlikely(gh->gh_flags & LM_FLAG_PRIORITY)) - goto do_cancel; - return; - } - list_add_tail(&gh->gh_list, insert_pt); -do_cancel: - gh = list_entry(gl->gl_holders.next, struct gfs2_holder, gh_list); - if (!(gh->gh_flags & LM_FLAG_PRIORITY)) { - spin_unlock(&gl->gl_spin); - if (sdp->sd_lockstruct.ls_ops->lm_cancel) - sdp->sd_lockstruct.ls_ops->lm_cancel(gl); - spin_lock(&gl->gl_spin); - } + list_add_tail(&gh->gh_list, &gl->gl_holders); return; trap_recursive: - printk(KERN_ERR "original: %pSR\n", (void *)gh2->gh_ip); - printk(KERN_ERR "pid: %d\n", pid_nr(gh2->gh_owner_pid)); - printk(KERN_ERR "lock type: %d req lock state : %d\n", + fs_err(sdp, "original: %pSR\n", (void *)gh2->gh_ip); + fs_err(sdp, "pid: %d\n", pid_nr(gh2->gh_owner_pid)); + fs_err(sdp, "lock type: %d req lock state : %d\n", gh2->gh_gl->gl_name.ln_type, gh2->gh_state); - printk(KERN_ERR "new: %pSR\n", (void *)gh->gh_ip); - printk(KERN_ERR "pid: %d\n", pid_nr(gh->gh_owner_pid)); - printk(KERN_ERR "lock type: %d req lock state : %d\n", + fs_err(sdp, "new: %pSR\n", (void *)gh->gh_ip); + fs_err(sdp, "pid: %d\n", pid_nr(gh->gh_owner_pid)); + fs_err(sdp, "lock type: %d req lock state : %d\n", gh->gh_gl->gl_name.ln_type, gh->gh_state); - gfs2_dump_glock(NULL, gl); + gfs2_dump_glock(NULL, gl, true); BUG(); } @@ -1041,23 +1472,44 @@ trap_recursive: int gfs2_glock_nq(struct gfs2_holder *gh) { struct gfs2_glock *gl = gh->gh_gl; - struct gfs2_sbd *sdp = gl->gl_sbd; - int error = 0; + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + int error; - if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) + if (gfs2_withdrawn(sdp)) return -EIO; - if (test_bit(GLF_LRU, &gl->gl_flags)) - gfs2_glock_remove_from_lru(gl); + if (gh->gh_flags & GL_NOBLOCK) { + struct gfs2_holder *current_gh; + + error = -ECHILD; + spin_lock(&gl->gl_lockref.lock); + if (find_last_waiter(gl)) + goto unlock; + current_gh = find_first_holder(gl); + if (!may_grant(gl, current_gh, gh)) + goto unlock; + set_bit(HIF_HOLDER, &gh->gh_iflags); + list_add_tail(&gh->gh_list, &gl->gl_holders); + trace_gfs2_promote(gh); + error = 0; +unlock: + spin_unlock(&gl->gl_lockref.lock); + return error; + } - spin_lock(&gl->gl_spin); + gh->gh_error = 0; + spin_lock(&gl->gl_lockref.lock); add_to_queue(gh); - if ((LM_FLAG_NOEXP & gh->gh_flags) && - test_and_clear_bit(GLF_FROZEN, &gl->gl_flags)) - set_bit(GLF_REPLY_PENDING, &gl->gl_flags); + if (unlikely((LM_FLAG_RECOVER & gh->gh_flags) && + test_and_clear_bit(GLF_HAVE_FROZEN_REPLY, &gl->gl_flags))) { + set_bit(GLF_HAVE_REPLY, &gl->gl_flags); + gl->gl_lockref.count++; + gfs2_glock_queue_work(gl, 0); + } run_queue(gl, 1); - spin_unlock(&gl->gl_spin); + spin_unlock(&gl->gl_lockref.lock); + error = 0; if (!(gh->gh_flags & GL_ASYNC)) error = gfs2_glock_wait(gh); @@ -1076,52 +1528,80 @@ int gfs2_glock_poll(struct gfs2_holder *gh) return test_bit(HIF_WAIT, &gh->gh_iflags) ? 0 : 1; } -/** - * gfs2_glock_dq - dequeue a struct gfs2_holder from a glock (release a glock) - * @gh: the glock holder - * - */ - -void gfs2_glock_dq(struct gfs2_holder *gh) +static void __gfs2_glock_dq(struct gfs2_holder *gh) { struct gfs2_glock *gl = gh->gh_gl; - const struct gfs2_glock_operations *glops = gl->gl_ops; unsigned delay = 0; int fast_path = 0; - spin_lock(&gl->gl_spin); + /* + * This holder should not be cached, so mark it for demote. + * Note: this should be done before the glock_needs_demote + * check below. + */ if (gh->gh_flags & GL_NOCACHE) - handle_callback(gl, LM_ST_UNLOCKED, 0, false); + request_demote(gl, LM_ST_UNLOCKED, 0, false); list_del_init(&gh->gh_list); - if (find_first_holder(gl) == NULL) { - if (glops->go_unlock) { - GLOCK_BUG_ON(gl, test_and_set_bit(GLF_LOCK, &gl->gl_flags)); - spin_unlock(&gl->gl_spin); - glops->go_unlock(gh); - spin_lock(&gl->gl_spin); - clear_bit(GLF_LOCK, &gl->gl_flags); - } - if (list_empty(&gl->gl_holders) && - !test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags) && - !test_bit(GLF_DEMOTE, &gl->gl_flags)) + clear_bit(HIF_HOLDER, &gh->gh_iflags); + trace_gfs2_glock_queue(gh, 0); + + /* + * If there hasn't been a demote request we are done. + * (Let the remaining holders, if any, keep holding it.) + */ + if (!glock_needs_demote(gl)) { + if (list_empty(&gl->gl_holders)) fast_path = 1; } - if (!test_bit(GLF_LFLUSH, &gl->gl_flags) && demote_ok(gl)) - gfs2_glock_add_to_lru(gl); - trace_gfs2_glock_queue(gh, 0); - spin_unlock(&gl->gl_spin); - if (likely(fast_path)) - return; + if (unlikely(!fast_path)) { + gl->gl_lockref.count++; + if (test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags) && + !test_bit(GLF_DEMOTE, &gl->gl_flags) && + gl->gl_name.ln_type == LM_TYPE_INODE) + delay = gl->gl_hold_time; + gfs2_glock_queue_work(gl, delay); + } +} - gfs2_glock_hold(gl); - if (test_bit(GLF_PENDING_DEMOTE, &gl->gl_flags) && - !test_bit(GLF_DEMOTE, &gl->gl_flags) && - gl->gl_name.ln_type == LM_TYPE_INODE) - delay = gl->gl_hold_time; - if (queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0) - gfs2_glock_put(gl); +/** + * gfs2_glock_dq - dequeue a struct gfs2_holder from a glock (release a glock) + * @gh: the glock holder + * + */ +void gfs2_glock_dq(struct gfs2_holder *gh) +{ + struct gfs2_glock *gl = gh->gh_gl; + + spin_lock(&gl->gl_lockref.lock); + if (!gfs2_holder_queued(gh)) { + /* + * May have already been dequeued because the locking request + * was GL_ASYNC and it has failed in the meantime. + */ + goto out; + } + + if (list_is_first(&gh->gh_list, &gl->gl_holders) && + !test_bit(HIF_HOLDER, &gh->gh_iflags) && + test_bit(GLF_LOCK, &gl->gl_flags) && + !test_bit(GLF_DEMOTE_IN_PROGRESS, &gl->gl_flags) && + !test_bit(GLF_CANCELING, &gl->gl_flags)) { + set_bit(GLF_CANCELING, &gl->gl_flags); + spin_unlock(&gl->gl_lockref.lock); + gl->gl_name.ln_sbd->sd_lockstruct.ls_ops->lm_cancel(gl); + wait_on_bit(&gh->gh_iflags, HIF_WAIT, TASK_UNINTERRUPTIBLE); + spin_lock(&gl->gl_lockref.lock); + clear_bit(GLF_CANCELING, &gl->gl_flags); + clear_bit(GLF_LOCK, &gl->gl_flags); + if (!gfs2_holder_queued(gh)) + goto out; + } + + __gfs2_glock_dq(gh); +out: + spin_unlock(&gl->gl_lockref.lock); } void gfs2_glock_dq_wait(struct gfs2_holder *gh) @@ -1129,7 +1609,7 @@ void gfs2_glock_dq_wait(struct gfs2_holder *gh) struct gfs2_glock *gl = gh->gh_gl; gfs2_glock_dq(gh); might_sleep(); - wait_on_bit(&gl->gl_flags, GLF_DEMOTE, gfs2_glock_demote_wait, TASK_UNINTERRUPTIBLE); + wait_on_bit(&gl->gl_flags, GLF_DEMOTE, TASK_UNINTERRUPTIBLE); } /** @@ -1158,7 +1638,7 @@ void gfs2_glock_dq_uninit(struct gfs2_holder *gh) int gfs2_glock_nq_num(struct gfs2_sbd *sdp, u64 number, const struct gfs2_glock_operations *glops, - unsigned int state, int flags, struct gfs2_holder *gh) + unsigned int state, u16 flags, struct gfs2_holder *gh) { struct gfs2_glock *gl; int error; @@ -1195,9 +1675,10 @@ static int glock_compare(const void *arg_a, const void *arg_b) } /** - * nq_m_sync - synchonously acquire more than one glock in deadlock free order + * nq_m_sync - synchronously acquire more than one glock in deadlock free order * @num_gh: the number of structures * @ghs: an array of struct gfs2_holder structures + * @p: placeholder for the holder structure to pass back * * Returns: 0 on success (all glocks acquired), * errno on failure (no glocks acquired) @@ -1215,8 +1696,6 @@ static int nq_m_sync(unsigned int num_gh, struct gfs2_holder *ghs, sort(p, num_gh, sizeof(struct gfs2_holder *), glock_compare, NULL); for (x = 0; x < num_gh; x++) { - p[x]->gh_flags &= ~(LM_FLAG_TRY | GL_ASYNC); - error = gfs2_glock_nq(p[x]); if (error) { while (x--) @@ -1233,7 +1712,6 @@ static int nq_m_sync(unsigned int num_gh, struct gfs2_holder *ghs, * @num_gh: the number of structures * @ghs: an array of struct gfs2_holder structures * - * * Returns: 0 on success (all glocks acquired), * errno on failure (no glocks acquired) */ @@ -1248,12 +1726,12 @@ int gfs2_glock_nq_m(unsigned int num_gh, struct gfs2_holder *ghs) case 0: return 0; case 1: - ghs->gh_flags &= ~(LM_FLAG_TRY | GL_ASYNC); return gfs2_glock_nq(ghs); default: if (num_gh <= 4) break; - pph = kmalloc(num_gh * sizeof(struct gfs2_holder *), GFP_NOFS); + pph = kmalloc_array(num_gh, sizeof(struct gfs2_holder *), + GFP_NOFS); if (!pph) return -ENOMEM; } @@ -1282,24 +1760,24 @@ void gfs2_glock_dq_m(unsigned int num_gh, struct gfs2_holder *ghs) void gfs2_glock_cb(struct gfs2_glock *gl, unsigned int state) { unsigned long delay = 0; - unsigned long holdtime; - unsigned long now = jiffies; gfs2_glock_hold(gl); - holdtime = gl->gl_tchange + gl->gl_hold_time; - if (test_bit(GLF_QUEUED, &gl->gl_flags) && + spin_lock(&gl->gl_lockref.lock); + if (!list_empty(&gl->gl_holders) && gl->gl_name.ln_type == LM_TYPE_INODE) { + unsigned long now = jiffies; + unsigned long holdtime; + + holdtime = gl->gl_tchange + gl->gl_hold_time; + if (time_before(now, holdtime)) delay = holdtime - now; - if (test_bit(GLF_REPLY_PENDING, &gl->gl_flags)) + if (test_bit(GLF_HAVE_REPLY, &gl->gl_flags)) delay = gl->gl_hold_time; } - - spin_lock(&gl->gl_spin); - handle_callback(gl, state, delay, true); - spin_unlock(&gl->gl_spin); - if (queue_delayed_work(glock_workqueue, &gl->gl_work, delay) == 0) - gfs2_glock_put(gl); + request_demote(gl, state, delay, true); + gfs2_glock_queue_work(gl, delay); + spin_unlock(&gl->gl_lockref.lock); } /** @@ -1308,7 +1786,7 @@ void gfs2_glock_cb(struct gfs2_glock *gl, unsigned int state) * * Glocks are not frozen if (a) the result of the dlm operation is * an error, (b) the locking operation was an unlock operation or - * (c) if there is a "noexp" flagged request anywhere in the queue + * (c) if there is a "recover" flagged request anywhere in the queue * * Returns: 1 if freezing should occur, 0 otherwise */ @@ -1325,7 +1803,7 @@ static int gfs2_should_freeze(const struct gfs2_glock *gl) list_for_each_entry(gh, &gl->gl_holders, gh_list) { if (test_bit(HIF_HOLDER, &gh->gh_iflags)) continue; - if (LM_FLAG_NOEXP & gh->gh_flags) + if (LM_FLAG_RECOVER & gh->gh_flags) return 0; } @@ -1337,34 +1815,34 @@ static int gfs2_should_freeze(const struct gfs2_glock *gl) * @gl: Pointer to the glock * @ret: The return value from the dlm * - * The gl_reply field is under the gl_spin lock so that it is ok + * The gl_reply field is under the gl_lockref.lock lock so that it is ok * to use a bitfield shared with other glock state fields. */ void gfs2_glock_complete(struct gfs2_glock *gl, int ret) { - struct lm_lockstruct *ls = &gl->gl_sbd->sd_lockstruct; + struct lm_lockstruct *ls = &gl->gl_name.ln_sbd->sd_lockstruct; - spin_lock(&gl->gl_spin); + spin_lock(&gl->gl_lockref.lock); + clear_bit(GLF_PENDING_REPLY, &gl->gl_flags); gl->gl_reply = ret; if (unlikely(test_bit(DFL_BLOCK_LOCKS, &ls->ls_recover_flags))) { if (gfs2_should_freeze(gl)) { - set_bit(GLF_FROZEN, &gl->gl_flags); - spin_unlock(&gl->gl_spin); + set_bit(GLF_HAVE_FROZEN_REPLY, &gl->gl_flags); + spin_unlock(&gl->gl_lockref.lock); return; } } - spin_unlock(&gl->gl_spin); - set_bit(GLF_REPLY_PENDING, &gl->gl_flags); - smp_wmb(); - gfs2_glock_hold(gl); - if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) - gfs2_glock_put(gl); + gl->gl_lockref.count++; + set_bit(GLF_HAVE_REPLY, &gl->gl_flags); + gfs2_glock_queue_work(gl, 0); + spin_unlock(&gl->gl_lockref.lock); } -static int glock_cmp(void *priv, struct list_head *a, struct list_head *b) +static int glock_cmp(void *priv, const struct list_head *a, + const struct list_head *b) { struct gfs2_glock *gla, *glb; @@ -1379,6 +1857,16 @@ static int glock_cmp(void *priv, struct list_head *a, struct list_head *b) return 0; } +static bool can_free_glock(struct gfs2_glock *gl) +{ + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + + return !test_bit(GLF_LOCK, &gl->gl_flags) && + !gl->gl_lockref.count && + (!test_bit(GLF_LFLUSH, &gl->gl_flags) || + test_bit(SDF_KILL, &sdp->sd_flags)); +} + /** * gfs2_dispose_glock_lru - Demote a list of glocks * @list: The list to dispose of @@ -1393,30 +1881,38 @@ static int glock_cmp(void *priv, struct list_head *a, struct list_head *b) * private) */ -static void gfs2_dispose_glock_lru(struct list_head *list) +static unsigned long gfs2_dispose_glock_lru(struct list_head *list) __releases(&lru_lock) __acquires(&lru_lock) { struct gfs2_glock *gl; + unsigned long freed = 0; list_sort(NULL, list, glock_cmp); while(!list_empty(list)) { - gl = list_entry(list->next, struct gfs2_glock, gl_lru); + gl = list_first_entry(list, struct gfs2_glock, gl_lru); + if (!spin_trylock(&gl->gl_lockref.lock)) { +add_back_to_lru: + list_move(&gl->gl_lru, &lru_list); + continue; + } + if (!can_free_glock(gl)) { + spin_unlock(&gl->gl_lockref.lock); + goto add_back_to_lru; + } list_del_init(&gl->gl_lru); + atomic_dec(&lru_count); clear_bit(GLF_LRU, &gl->gl_flags); - gfs2_glock_hold(gl); - spin_unlock(&lru_lock); - spin_lock(&gl->gl_spin); - if (demote_ok(gl)) - handle_callback(gl, LM_ST_UNLOCKED, 0, false); - WARN_ON(!test_and_clear_bit(GLF_LOCK, &gl->gl_flags)); - smp_mb__after_clear_bit(); - if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) - gfs2_glock_put_nolock(gl); - spin_unlock(&gl->gl_spin); - spin_lock(&lru_lock); + freed++; + gl->gl_lockref.count++; + if (gl->gl_state != LM_ST_UNLOCKED) + request_demote(gl, LM_ST_UNLOCKED, 0, false); + gfs2_glock_queue_work(gl, 0); + spin_unlock(&gl->gl_lockref.lock); + cond_resched_lock(&lru_lock); } + return freed; } /** @@ -1428,98 +1924,117 @@ __acquires(&lru_lock) * gfs2_dispose_glock_lru() above. */ -static void gfs2_scan_glock_lru(int nr) +static unsigned long gfs2_scan_glock_lru(unsigned long nr) { - struct gfs2_glock *gl; - LIST_HEAD(skipped); + struct gfs2_glock *gl, *next; LIST_HEAD(dispose); + unsigned long freed = 0; spin_lock(&lru_lock); - while(nr && !list_empty(&lru_list)) { - gl = list_entry(lru_list.next, struct gfs2_glock, gl_lru); - - /* Test for being demotable */ - if (!test_and_set_bit(GLF_LOCK, &gl->gl_flags)) { + list_for_each_entry_safe(gl, next, &lru_list, gl_lru) { + if (!nr--) + break; + if (can_free_glock(gl)) list_move(&gl->gl_lru, &dispose); - atomic_dec(&lru_count); - nr--; - continue; - } - - list_move(&gl->gl_lru, &skipped); } - list_splice(&skipped, &lru_list); if (!list_empty(&dispose)) - gfs2_dispose_glock_lru(&dispose); + freed = gfs2_dispose_glock_lru(&dispose); spin_unlock(&lru_lock); + + return freed; } -static int gfs2_shrink_glock_memory(struct shrinker *shrink, - struct shrink_control *sc) +static unsigned long gfs2_glock_shrink_scan(struct shrinker *shrink, + struct shrink_control *sc) { - if (sc->nr_to_scan) { - if (!(sc->gfp_mask & __GFP_FS)) - return -1; - gfs2_scan_glock_lru(sc->nr_to_scan); - } + if (!(sc->gfp_mask & __GFP_FS)) + return SHRINK_STOP; + return gfs2_scan_glock_lru(sc->nr_to_scan); +} - return (atomic_read(&lru_count) / 100) * sysctl_vfs_cache_pressure; +static unsigned long gfs2_glock_shrink_count(struct shrinker *shrink, + struct shrink_control *sc) +{ + return vfs_pressure_ratio(atomic_read(&lru_count)); } -static struct shrinker glock_shrinker = { - .shrink = gfs2_shrink_glock_memory, - .seeks = DEFAULT_SEEKS, -}; +static struct shrinker *glock_shrinker; /** - * examine_bucket - Call a function for glock in a hash bucket + * glock_hash_walk - Call a function for glock in a hash bucket * @examiner: the function * @sdp: the filesystem - * @bucket: the bucket * + * Note that the function can be called multiple times on the same + * object. So the user must ensure that the function can cope with + * that. */ -static void examine_bucket(glock_examiner examiner, const struct gfs2_sbd *sdp, - unsigned int hash) +static void glock_hash_walk(glock_examiner examiner, const struct gfs2_sbd *sdp) { struct gfs2_glock *gl; - struct hlist_bl_head *head = &gl_hash_table[hash]; - struct hlist_bl_node *pos; + struct rhashtable_iter iter; - rcu_read_lock(); - hlist_bl_for_each_entry_rcu(gl, pos, head, gl_list) { - if ((gl->gl_sbd == sdp) && atomic_read(&gl->gl_ref)) - examiner(gl); - } - rcu_read_unlock(); - cond_resched(); + rhashtable_walk_enter(&gl_hash_table, &iter); + + do { + rhashtable_walk_start(&iter); + + while ((gl = rhashtable_walk_next(&iter)) && !IS_ERR(gl)) { + if (gl->gl_name.ln_sbd == sdp) + examiner(gl); + } + + rhashtable_walk_stop(&iter); + } while (cond_resched(), gl == ERR_PTR(-EAGAIN)); + + rhashtable_walk_exit(&iter); } -static void glock_hash_walk(glock_examiner examiner, const struct gfs2_sbd *sdp) +void gfs2_cancel_delete_work(struct gfs2_glock *gl) { - unsigned x; + clear_bit(GLF_TRY_TO_EVICT, &gl->gl_flags); + clear_bit(GLF_VERIFY_DELETE, &gl->gl_flags); + if (cancel_delayed_work(&gl->gl_delete)) + gfs2_glock_put(gl); +} - for (x = 0; x < GFS2_GL_HASH_SIZE; x++) - examine_bucket(examiner, sdp, x); +static void flush_delete_work(struct gfs2_glock *gl) +{ + if (gl->gl_name.ln_type == LM_TYPE_IOPEN) { + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + + if (cancel_delayed_work(&gl->gl_delete)) { + queue_delayed_work(sdp->sd_delete_wq, + &gl->gl_delete, 0); + } + } } +void gfs2_flush_delete_work(struct gfs2_sbd *sdp) +{ + glock_hash_walk(flush_delete_work, sdp); + flush_workqueue(sdp->sd_delete_wq); +} /** * thaw_glock - thaw out a glock which has an unprocessed reply waiting * @gl: The glock to thaw * - * N.B. When we freeze a glock, we leave a ref to the glock outstanding, - * so this has to result in the ref count being dropped by one. */ static void thaw_glock(struct gfs2_glock *gl) { - if (!test_and_clear_bit(GLF_FROZEN, &gl->gl_flags)) + if (!test_and_clear_bit(GLF_HAVE_FROZEN_REPLY, &gl->gl_flags)) return; - set_bit(GLF_REPLY_PENDING, &gl->gl_flags); - gfs2_glock_hold(gl); - if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) - gfs2_glock_put(gl); + if (!lockref_get_not_dead(&gl->gl_lockref)) + return; + + gfs2_glock_remove_from_lru(gl); + spin_lock(&gl->gl_lockref.lock); + set_bit(GLF_HAVE_REPLY, &gl->gl_flags); + gfs2_glock_queue_work(gl, 0); + spin_unlock(&gl->gl_lockref.lock); } /** @@ -1532,13 +2047,14 @@ static void clear_glock(struct gfs2_glock *gl) { gfs2_glock_remove_from_lru(gl); - spin_lock(&gl->gl_spin); - if (gl->gl_state != LM_ST_UNLOCKED) - handle_callback(gl, LM_ST_UNLOCKED, 0, false); - spin_unlock(&gl->gl_spin); - gfs2_glock_hold(gl); - if (queue_delayed_work(glock_workqueue, &gl->gl_work, 0) == 0) - gfs2_glock_put(gl); + spin_lock(&gl->gl_lockref.lock); + if (!__lockref_is_dead(&gl->gl_lockref)) { + gl->gl_lockref.count++; + if (gl->gl_state != LM_ST_UNLOCKED) + request_demote(gl, LM_ST_UNLOCKED, 0, false); + gfs2_glock_queue_work(gl, 0); + } + spin_unlock(&gl->gl_lockref.lock); } /** @@ -1552,50 +2068,74 @@ void gfs2_glock_thaw(struct gfs2_sbd *sdp) glock_hash_walk(thaw_glock, sdp); } -static int dump_glock(struct seq_file *seq, struct gfs2_glock *gl) +static void dump_glock(struct seq_file *seq, struct gfs2_glock *gl, bool fsid) { - int ret; - spin_lock(&gl->gl_spin); - ret = gfs2_dump_glock(seq, gl); - spin_unlock(&gl->gl_spin); - return ret; + spin_lock(&gl->gl_lockref.lock); + gfs2_dump_glock(seq, gl, fsid); + spin_unlock(&gl->gl_lockref.lock); } static void dump_glock_func(struct gfs2_glock *gl) { - dump_glock(NULL, gl); + dump_glock(NULL, gl, true); +} + +static void withdraw_glock(struct gfs2_glock *gl) +{ + spin_lock(&gl->gl_lockref.lock); + if (!__lockref_is_dead(&gl->gl_lockref)) { + /* + * We don't want to write back any more dirty data. Unlock the + * remaining inode and resource group glocks; this will cause + * their ->go_inval() hooks to toss out all the remaining + * cached data, dirty or not. + */ + if (gl->gl_ops->go_inval && gl->gl_state != LM_ST_UNLOCKED) + request_demote(gl, LM_ST_UNLOCKED, 0, false); + do_error(gl, LM_OUT_ERROR); /* remove pending waiters */ + } + spin_unlock(&gl->gl_lockref.lock); +} + +void gfs2_withdraw_glocks(struct gfs2_sbd *sdp) +{ + glock_hash_walk(withdraw_glock, sdp); } /** * gfs2_gl_hash_clear - Empty out the glock hash table * @sdp: the filesystem - * @wait: wait until it's all gone * * Called when unmounting the filesystem. */ void gfs2_gl_hash_clear(struct gfs2_sbd *sdp) { + unsigned long start = jiffies; + bool timed_out = false; + set_bit(SDF_SKIP_DLM_UNLOCK, &sdp->sd_flags); - flush_workqueue(glock_workqueue); + flush_workqueue(sdp->sd_glock_wq); glock_hash_walk(clear_glock, sdp); - flush_workqueue(glock_workqueue); - wait_event(sdp->sd_glock_wait, atomic_read(&sdp->sd_glock_disposal) == 0); - glock_hash_walk(dump_glock_func, sdp); -} - -void gfs2_glock_finish_truncate(struct gfs2_inode *ip) -{ - struct gfs2_glock *gl = ip->i_gl; - int ret; + flush_workqueue(sdp->sd_glock_wq); - ret = gfs2_truncatei_resume(ip); - gfs2_assert_withdraw(gl->gl_sbd, ret == 0); - - spin_lock(&gl->gl_spin); - clear_bit(GLF_LOCK, &gl->gl_flags); - run_queue(gl, 1); - spin_unlock(&gl->gl_spin); + while (!timed_out) { + wait_event_timeout(sdp->sd_kill_wait, + !atomic_read(&sdp->sd_glock_disposal), + HZ * 60); + if (!atomic_read(&sdp->sd_glock_disposal)) + break; + timed_out = time_after(jiffies, start + (HZ * 600)); + fs_warn(sdp, "%u glocks left after %u seconds%s\n", + atomic_read(&sdp->sd_glock_disposal), + jiffies_to_msecs(jiffies - start) / 1000, + timed_out ? ":" : "; still waiting"); + } + gfs2_lm_unmount(sdp); + gfs2_free_dead_glocks(sdp); + glock_hash_walk(dump_glock_func, sdp); + destroy_workqueue(sdp->sd_glock_wq); + sdp->sd_glock_wq = NULL; } static const char *state2str(unsigned state) @@ -1613,19 +2153,19 @@ static const char *state2str(unsigned state) return "??"; } -static const char *hflags2str(char *buf, unsigned flags, unsigned long iflags) +static const char *hflags2str(char *buf, u16 flags, unsigned long iflags) { char *p = buf; if (flags & LM_FLAG_TRY) *p++ = 't'; if (flags & LM_FLAG_TRY_1CB) *p++ = 'T'; - if (flags & LM_FLAG_NOEXP) + if (flags & LM_FLAG_RECOVER) *p++ = 'e'; if (flags & LM_FLAG_ANY) *p++ = 'A'; - if (flags & LM_FLAG_PRIORITY) - *p++ = 'p'; + if (flags & LM_FLAG_NODE_SCOPE) + *p++ = 'n'; if (flags & GL_ASYNC) *p++ = 'a'; if (flags & GL_EXACT) @@ -1636,8 +2176,8 @@ static const char *hflags2str(char *buf, unsigned flags, unsigned long iflags) *p++ = 'H'; if (test_bit(HIF_WAIT, &iflags)) *p++ = 'W'; - if (test_bit(HIF_FIRST, &iflags)) - *p++ = 'F'; + if (flags & GL_SKIP) + *p++ = 's'; *p = 0; return buf; } @@ -1646,25 +2186,32 @@ static const char *hflags2str(char *buf, unsigned flags, unsigned long iflags) * dump_holder - print information about a glock holder * @seq: the seq_file struct * @gh: the glock holder + * @fs_id_buf: pointer to file system id (if requested) * - * Returns: 0 on success, -ENOBUFS when we run out of space */ -static int dump_holder(struct seq_file *seq, const struct gfs2_holder *gh) +static void dump_holder(struct seq_file *seq, const struct gfs2_holder *gh, + const char *fs_id_buf) { - struct task_struct *gh_owner = NULL; + const char *comm = "(none)"; + pid_t owner_pid = 0; char flags_buf[32]; - if (gh->gh_owner_pid) + rcu_read_lock(); + if (pid_is_meaningful(gh)) { + struct task_struct *gh_owner; + + comm = "(ended)"; + owner_pid = pid_nr(gh->gh_owner_pid); gh_owner = pid_task(gh->gh_owner_pid, PIDTYPE_PID); - gfs2_print_dbg(seq, " H: s:%s f:%s e:%d p:%ld [%s] %pS\n", - state2str(gh->gh_state), + if (gh_owner) + comm = gh_owner->comm; + } + gfs2_print_dbg(seq, "%s H: s:%s f:%s e:%d p:%ld [%s] %pS\n", + fs_id_buf, state2str(gh->gh_state), hflags2str(flags_buf, gh->gh_flags, gh->gh_iflags), - gh->gh_error, - gh->gh_owner_pid ? (long)pid_nr(gh->gh_owner_pid) : -1, - gh_owner ? gh_owner->comm : "(ended)", - (void *)gh->gh_ip); - return 0; + gh->gh_error, (long)owner_pid, comm, (void *)gh->gh_ip); + rcu_read_unlock(); } static const char *gflags2str(char *buf, const struct gfs2_glock *gl) @@ -1684,15 +2231,15 @@ static const char *gflags2str(char *buf, const struct gfs2_glock *gl) *p++ = 'y'; if (test_bit(GLF_LFLUSH, gflags)) *p++ = 'f'; - if (test_bit(GLF_INVALIDATE_IN_PROGRESS, gflags)) - *p++ = 'i'; - if (test_bit(GLF_REPLY_PENDING, gflags)) + if (test_bit(GLF_PENDING_REPLY, gflags)) + *p++ = 'R'; + if (test_bit(GLF_HAVE_REPLY, gflags)) *p++ = 'r'; if (test_bit(GLF_INITIAL, gflags)) - *p++ = 'I'; - if (test_bit(GLF_FROZEN, gflags)) + *p++ = 'a'; + if (test_bit(GLF_HAVE_FROZEN_REPLY, gflags)) *p++ = 'F'; - if (test_bit(GLF_QUEUED, gflags)) + if (!list_empty(&gl->gl_holders)) *p++ = 'q'; if (test_bit(GLF_LRU, gflags)) *p++ = 'L'; @@ -1700,6 +2247,18 @@ static const char *gflags2str(char *buf, const struct gfs2_glock *gl) *p++ = 'o'; if (test_bit(GLF_BLOCKING, gflags)) *p++ = 'b'; + if (test_bit(GLF_INSTANTIATE_NEEDED, gflags)) + *p++ = 'n'; + if (test_bit(GLF_INSTANTIATE_IN_PROG, gflags)) + *p++ = 'N'; + if (test_bit(GLF_TRY_TO_EVICT, gflags)) + *p++ = 'e'; + if (test_bit(GLF_VERIFY_DELETE, gflags)) + *p++ = 'E'; + if (test_bit(GLF_DEFER_DELETE, gflags)) + *p++ = 's'; + if (test_bit(GLF_CANCELING, gflags)) + *p++ = 'C'; *p = 0; return buf; } @@ -1708,6 +2267,7 @@ static const char *gflags2str(char *buf, const struct gfs2_glock *gl) * gfs2_dump_glock - print information about a glock * @seq: The seq_file struct * @gl: the glock + * @fsid: If true, also dump the file system id * * The file format is as follows: * One line per object, capital letters are used to indicate objects @@ -1719,58 +2279,64 @@ static const char *gflags2str(char *buf, const struct gfs2_glock *gl) * example. The field's are n = number (id of the object), f = flags, * t = type, s = state, r = refcount, e = error, p = pid. * - * Returns: 0 on success, -ENOBUFS when we run out of space */ -int gfs2_dump_glock(struct seq_file *seq, const struct gfs2_glock *gl) +void gfs2_dump_glock(struct seq_file *seq, struct gfs2_glock *gl, bool fsid) { const struct gfs2_glock_operations *glops = gl->gl_ops; unsigned long long dtime; const struct gfs2_holder *gh; char gflags_buf[32]; - int error = 0; + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + char fs_id_buf[sizeof(sdp->sd_fsname) + 7]; + unsigned long nrpages = 0; + + if (gl->gl_ops->go_flags & GLOF_ASPACE) { + struct address_space *mapping = gfs2_glock2aspace(gl); + nrpages = mapping->nrpages; + } + memset(fs_id_buf, 0, sizeof(fs_id_buf)); + if (fsid && sdp) /* safety precaution */ + sprintf(fs_id_buf, "fsid=%s: ", sdp->sd_fsname); dtime = jiffies - gl->gl_demote_time; dtime *= 1000000/HZ; /* demote time in uSec */ if (!test_bit(GLF_DEMOTE, &gl->gl_flags)) dtime = 0; - gfs2_print_dbg(seq, "G: s:%s n:%u/%llx f:%s t:%s d:%s/%llu a:%d v:%d r:%d m:%ld\n", - state2str(gl->gl_state), - gl->gl_name.ln_type, - (unsigned long long)gl->gl_name.ln_number, - gflags2str(gflags_buf, gl), - state2str(gl->gl_target), - state2str(gl->gl_demote_state), dtime, - atomic_read(&gl->gl_ail_count), - atomic_read(&gl->gl_revokes), - atomic_read(&gl->gl_ref), gl->gl_hold_time); + gfs2_print_dbg(seq, "%sG: s:%s n:%u/%llx f:%s t:%s d:%s/%llu a:%d " + "v:%d r:%d m:%ld p:%lu\n", + fs_id_buf, state2str(gl->gl_state), + gl->gl_name.ln_type, + (unsigned long long)gl->gl_name.ln_number, + gflags2str(gflags_buf, gl), + state2str(gl->gl_target), + state2str(gl->gl_demote_state), dtime, + atomic_read(&gl->gl_ail_count), + atomic_read(&gl->gl_revokes), + (int)gl->gl_lockref.count, gl->gl_hold_time, nrpages); + + list_for_each_entry(gh, &gl->gl_holders, gh_list) + dump_holder(seq, gh, fs_id_buf); - list_for_each_entry(gh, &gl->gl_holders, gh_list) { - error = dump_holder(seq, gh); - if (error) - goto out; - } if (gl->gl_state != LM_ST_UNLOCKED && glops->go_dump) - error = glops->go_dump(seq, gl); -out: - return error; + glops->go_dump(seq, gl, fs_id_buf); } static int gfs2_glstats_seq_show(struct seq_file *seq, void *iter_ptr) { struct gfs2_glock *gl = iter_ptr; - seq_printf(seq, "G: n:%u/%llx rtt:%lld/%lld rttb:%lld/%lld irt:%lld/%lld dcnt: %lld qcnt: %lld\n", + seq_printf(seq, "G: n:%u/%llx rtt:%llu/%llu rttb:%llu/%llu irt:%llu/%llu dcnt: %llu qcnt: %llu\n", gl->gl_name.ln_type, (unsigned long long)gl->gl_name.ln_number, - (long long)gl->gl_stats.stats[GFS2_LKS_SRTT], - (long long)gl->gl_stats.stats[GFS2_LKS_SRTTVAR], - (long long)gl->gl_stats.stats[GFS2_LKS_SRTTB], - (long long)gl->gl_stats.stats[GFS2_LKS_SRTTVARB], - (long long)gl->gl_stats.stats[GFS2_LKS_SIRT], - (long long)gl->gl_stats.stats[GFS2_LKS_SIRTVAR], - (long long)gl->gl_stats.stats[GFS2_LKS_DCOUNT], - (long long)gl->gl_stats.stats[GFS2_LKS_QCOUNT]); + (unsigned long long)gl->gl_stats.stats[GFS2_LKS_SRTT], + (unsigned long long)gl->gl_stats.stats[GFS2_LKS_SRTTVAR], + (unsigned long long)gl->gl_stats.stats[GFS2_LKS_SRTTB], + (unsigned long long)gl->gl_stats.stats[GFS2_LKS_SRTTVARB], + (unsigned long long)gl->gl_stats.stats[GFS2_LKS_SIRT], + (unsigned long long)gl->gl_stats.stats[GFS2_LKS_SIRTVAR], + (unsigned long long)gl->gl_stats.stats[GFS2_LKS_DCOUNT], + (unsigned long long)gl->gl_stats.stats[GFS2_LKS_QCOUNT]); return 0; } @@ -1803,11 +2369,10 @@ static const char *gfs2_stype[] = { static int gfs2_sbstats_seq_show(struct seq_file *seq, void *iter_ptr) { - struct gfs2_glock_iter *gi = seq->private; - struct gfs2_sbd *sdp = gi->sdp; - unsigned index = gi->hash >> 3; - unsigned subindex = gi->hash & 0x07; - s64 value; + struct gfs2_sbd *sdp = seq->private; + loff_t pos = *(loff_t *)iter_ptr; + unsigned index = pos >> 3; + unsigned subindex = pos & 0x07; int i; if (index == 0 && subindex != 0) @@ -1818,12 +2383,12 @@ static int gfs2_sbstats_seq_show(struct seq_file *seq, void *iter_ptr) for_each_possible_cpu(i) { const struct gfs2_pcpu_lkstats *lkstats = per_cpu_ptr(sdp->sd_lkstats, i); - if (index == 0) { - value = i; - } else { - value = lkstats->lkstats[index - 1].stats[subindex]; - } - seq_printf(seq, " %15lld", (long long)value); + + if (index == 0) + seq_printf(seq, " %15u", i); + else + seq_printf(seq, " %15llu", (unsigned long long)lkstats-> + lkstats[index - 1].stats[subindex]); } seq_putc(seq, '\n'); return 0; @@ -1831,97 +2396,90 @@ static int gfs2_sbstats_seq_show(struct seq_file *seq, void *iter_ptr) int __init gfs2_glock_init(void) { - unsigned i; - for(i = 0; i < GFS2_GL_HASH_SIZE; i++) { - INIT_HLIST_BL_HEAD(&gl_hash_table[i]); - } + int i, ret; + + ret = rhashtable_init(&gl_hash_table, &ht_parms); + if (ret < 0) + return ret; - glock_workqueue = alloc_workqueue("glock_workqueue", WQ_MEM_RECLAIM | - WQ_HIGHPRI | WQ_FREEZABLE, 0); - if (IS_ERR(glock_workqueue)) - return PTR_ERR(glock_workqueue); - gfs2_delete_workqueue = alloc_workqueue("delete_workqueue", - WQ_MEM_RECLAIM | WQ_FREEZABLE, - 0); - if (IS_ERR(gfs2_delete_workqueue)) { - destroy_workqueue(glock_workqueue); - return PTR_ERR(gfs2_delete_workqueue); + glock_shrinker = shrinker_alloc(0, "gfs2-glock"); + if (!glock_shrinker) { + rhashtable_destroy(&gl_hash_table); + return -ENOMEM; } - register_shrinker(&glock_shrinker); + glock_shrinker->count_objects = gfs2_glock_shrink_count; + glock_shrinker->scan_objects = gfs2_glock_shrink_scan; - return 0; -} + shrinker_register(glock_shrinker); -void gfs2_glock_exit(void) -{ - unregister_shrinker(&glock_shrinker); - destroy_workqueue(glock_workqueue); - destroy_workqueue(gfs2_delete_workqueue); -} + for (i = 0; i < GLOCK_WAIT_TABLE_SIZE; i++) + init_waitqueue_head(glock_wait_table + i); -static inline struct gfs2_glock *glock_hash_chain(unsigned hash) -{ - return hlist_bl_entry(hlist_bl_first_rcu(&gl_hash_table[hash]), - struct gfs2_glock, gl_list); + return 0; } -static inline struct gfs2_glock *glock_hash_next(struct gfs2_glock *gl) +void gfs2_glock_exit(void) { - return hlist_bl_entry(rcu_dereference(gl->gl_list.next), - struct gfs2_glock, gl_list); + shrinker_free(glock_shrinker); + rhashtable_destroy(&gl_hash_table); } -static int gfs2_glock_iter_next(struct gfs2_glock_iter *gi) +static void gfs2_glock_iter_next(struct gfs2_glock_iter *gi, loff_t n) { - struct gfs2_glock *gl; + struct gfs2_glock *gl = gi->gl; - do { - gl = gi->gl; - if (gl) { - gi->gl = glock_hash_next(gl); - gi->nhash++; - } else { - if (gi->hash >= GFS2_GL_HASH_SIZE) { - rcu_read_unlock(); - return 1; + if (gl) { + if (n == 0) + return; + gfs2_glock_put_async(gl); + } + for (;;) { + gl = rhashtable_walk_next(&gi->hti); + if (IS_ERR_OR_NULL(gl)) { + if (gl == ERR_PTR(-EAGAIN)) { + n = 1; + continue; } - gi->gl = glock_hash_chain(gi->hash); - gi->nhash = 0; + gl = NULL; + break; } - while (gi->gl == NULL) { - gi->hash++; - if (gi->hash >= GFS2_GL_HASH_SIZE) { - rcu_read_unlock(); - return 1; - } - gi->gl = glock_hash_chain(gi->hash); - gi->nhash = 0; + if (gl->gl_name.ln_sbd != gi->sdp) + continue; + if (n <= 1) { + if (!lockref_get_not_dead(&gl->gl_lockref)) + continue; + break; + } else { + if (__lockref_is_dead(&gl->gl_lockref)) + continue; + n--; } - /* Skip entries for other sb and dead entries */ - } while (gi->sdp != gi->gl->gl_sbd || atomic_read(&gi->gl->gl_ref) == 0); - - return 0; + } + gi->gl = gl; } static void *gfs2_glock_seq_start(struct seq_file *seq, loff_t *pos) + __acquires(RCU) { struct gfs2_glock_iter *gi = seq->private; - loff_t n = *pos; - - if (gi->last_pos <= *pos) - n = gi->nhash + (*pos - gi->last_pos); - else - gi->hash = 0; - - gi->nhash = 0; - rcu_read_lock(); + loff_t n; + + /* + * We can either stay where we are, skip to the next hash table + * entry, or start from the beginning. + */ + if (*pos < gi->last_pos) { + rhashtable_walk_exit(&gi->hti); + rhashtable_walk_enter(&gl_hash_table, &gi->hti); + n = *pos + 1; + } else { + n = *pos - gi->last_pos; + } - do { - if (gfs2_glock_iter_next(gi)) - return NULL; - } while (n--); + rhashtable_walk_start(&gi->hti); + gfs2_glock_iter_next(gi, n); gi->last_pos = *pos; return gi->gl; } @@ -1933,48 +2491,39 @@ static void *gfs2_glock_seq_next(struct seq_file *seq, void *iter_ptr, (*pos)++; gi->last_pos = *pos; - if (gfs2_glock_iter_next(gi)) - return NULL; - + gfs2_glock_iter_next(gi, 1); return gi->gl; } static void gfs2_glock_seq_stop(struct seq_file *seq, void *iter_ptr) + __releases(RCU) { struct gfs2_glock_iter *gi = seq->private; - if (gi->gl) - rcu_read_unlock(); - gi->gl = NULL; + rhashtable_walk_stop(&gi->hti); } static int gfs2_glock_seq_show(struct seq_file *seq, void *iter_ptr) { - return dump_glock(seq, iter_ptr); + dump_glock(seq, iter_ptr, false); + return 0; } static void *gfs2_sbstats_seq_start(struct seq_file *seq, loff_t *pos) { - struct gfs2_glock_iter *gi = seq->private; - - gi->hash = *pos; + preempt_disable(); if (*pos >= GFS2_NR_SBSTATS) return NULL; - preempt_disable(); - return SEQ_START_TOKEN; + return pos; } static void *gfs2_sbstats_seq_next(struct seq_file *seq, void *iter_ptr, loff_t *pos) { - struct gfs2_glock_iter *gi = seq->private; (*pos)++; - gi->hash++; - if (gi->hash >= GFS2_NR_SBSTATS) { - preempt_enable(); + if (*pos >= GFS2_NR_SBSTATS) return NULL; - } - return SEQ_START_TOKEN; + return pos; } static void gfs2_sbstats_seq_stop(struct seq_file *seq, void *iter_ptr) @@ -1996,7 +2545,7 @@ static const struct seq_operations gfs2_glstats_seq_ops = { .show = gfs2_glstats_seq_show, }; -static const struct seq_operations gfs2_sbstats_seq_ops = { +static const struct seq_operations gfs2_sbstats_sops = { .start = gfs2_sbstats_seq_start, .next = gfs2_sbstats_seq_next, .stop = gfs2_sbstats_seq_stop, @@ -2005,46 +2554,48 @@ static const struct seq_operations gfs2_sbstats_seq_ops = { #define GFS2_SEQ_GOODSIZE min(PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER, 65536UL) -static int gfs2_glocks_open(struct inode *inode, struct file *file) +static int __gfs2_glocks_open(struct inode *inode, struct file *file, + const struct seq_operations *ops) { - int ret = seq_open_private(file, &gfs2_glock_seq_ops, - sizeof(struct gfs2_glock_iter)); + int ret = seq_open_private(file, ops, sizeof(struct gfs2_glock_iter)); if (ret == 0) { struct seq_file *seq = file->private_data; struct gfs2_glock_iter *gi = seq->private; + gi->sdp = inode->i_private; seq->buf = kmalloc(GFS2_SEQ_GOODSIZE, GFP_KERNEL | __GFP_NOWARN); if (seq->buf) seq->size = GFS2_SEQ_GOODSIZE; + /* + * Initially, we are "before" the first hash table entry; the + * first call to rhashtable_walk_next gets us the first entry. + */ + gi->last_pos = -1; + gi->gl = NULL; + rhashtable_walk_enter(&gl_hash_table, &gi->hti); } return ret; } -static int gfs2_glstats_open(struct inode *inode, struct file *file) +static int gfs2_glocks_open(struct inode *inode, struct file *file) { - int ret = seq_open_private(file, &gfs2_glstats_seq_ops, - sizeof(struct gfs2_glock_iter)); - if (ret == 0) { - struct seq_file *seq = file->private_data; - struct gfs2_glock_iter *gi = seq->private; - gi->sdp = inode->i_private; - seq->buf = kmalloc(GFS2_SEQ_GOODSIZE, GFP_KERNEL | __GFP_NOWARN); - if (seq->buf) - seq->size = GFS2_SEQ_GOODSIZE; - } - return ret; + return __gfs2_glocks_open(inode, file, &gfs2_glock_seq_ops); } -static int gfs2_sbstats_open(struct inode *inode, struct file *file) +static int gfs2_glocks_release(struct inode *inode, struct file *file) { - int ret = seq_open_private(file, &gfs2_sbstats_seq_ops, - sizeof(struct gfs2_glock_iter)); - if (ret == 0) { - struct seq_file *seq = file->private_data; - struct gfs2_glock_iter *gi = seq->private; - gi->sdp = inode->i_private; - } - return ret; + struct seq_file *seq = file->private_data; + struct gfs2_glock_iter *gi = seq->private; + + if (gi->gl) + gfs2_glock_put(gi->gl); + rhashtable_walk_exit(&gi->hti); + return seq_release_private(inode, file); +} + +static int gfs2_glstats_open(struct inode *inode, struct file *file) +{ + return __gfs2_glocks_open(inode, file, &gfs2_glstats_seq_ops); } static const struct file_operations gfs2_glocks_fops = { @@ -2052,7 +2603,7 @@ static const struct file_operations gfs2_glocks_fops = { .open = gfs2_glocks_open, .read = seq_read, .llseek = seq_lseek, - .release = seq_release_private, + .release = gfs2_glocks_release, }; static const struct file_operations gfs2_glstats_fops = { @@ -2060,73 +2611,199 @@ static const struct file_operations gfs2_glstats_fops = { .open = gfs2_glstats_open, .read = seq_read, .llseek = seq_lseek, - .release = seq_release_private, + .release = gfs2_glocks_release, +}; + +struct gfs2_glockfd_iter { + struct super_block *sb; + unsigned int tgid; + struct task_struct *task; + unsigned int fd; + struct file *file; }; -static const struct file_operations gfs2_sbstats_fops = { +static struct task_struct *gfs2_glockfd_next_task(struct gfs2_glockfd_iter *i) +{ + struct pid_namespace *ns = task_active_pid_ns(current); + struct pid *pid; + + if (i->task) + put_task_struct(i->task); + + rcu_read_lock(); +retry: + i->task = NULL; + pid = find_ge_pid(i->tgid, ns); + if (pid) { + i->tgid = pid_nr_ns(pid, ns); + i->task = pid_task(pid, PIDTYPE_TGID); + if (!i->task) { + i->tgid++; + goto retry; + } + get_task_struct(i->task); + } + rcu_read_unlock(); + return i->task; +} + +static struct file *gfs2_glockfd_next_file(struct gfs2_glockfd_iter *i) +{ + if (i->file) { + fput(i->file); + i->file = NULL; + } + + for(;; i->fd++) { + i->file = fget_task_next(i->task, &i->fd); + if (!i->file) { + i->fd = 0; + break; + } + + if (file_inode(i->file)->i_sb == i->sb) + break; + + fput(i->file); + } + return i->file; +} + +static void *gfs2_glockfd_seq_start(struct seq_file *seq, loff_t *pos) +{ + struct gfs2_glockfd_iter *i = seq->private; + + if (*pos) + return NULL; + while (gfs2_glockfd_next_task(i)) { + if (gfs2_glockfd_next_file(i)) + return i; + i->tgid++; + } + return NULL; +} + +static void *gfs2_glockfd_seq_next(struct seq_file *seq, void *iter_ptr, + loff_t *pos) +{ + struct gfs2_glockfd_iter *i = seq->private; + + (*pos)++; + i->fd++; + do { + if (gfs2_glockfd_next_file(i)) + return i; + i->tgid++; + } while (gfs2_glockfd_next_task(i)); + return NULL; +} + +static void gfs2_glockfd_seq_stop(struct seq_file *seq, void *iter_ptr) +{ + struct gfs2_glockfd_iter *i = seq->private; + + if (i->file) + fput(i->file); + if (i->task) + put_task_struct(i->task); +} + +static void gfs2_glockfd_seq_show_flock(struct seq_file *seq, + struct gfs2_glockfd_iter *i) +{ + struct gfs2_file *fp = i->file->private_data; + struct gfs2_holder *fl_gh = &fp->f_fl_gh; + struct lm_lockname gl_name = { .ln_type = LM_TYPE_RESERVED }; + + if (!READ_ONCE(fl_gh->gh_gl)) + return; + + spin_lock(&i->file->f_lock); + if (gfs2_holder_initialized(fl_gh)) + gl_name = fl_gh->gh_gl->gl_name; + spin_unlock(&i->file->f_lock); + + if (gl_name.ln_type != LM_TYPE_RESERVED) { + seq_printf(seq, "%d %u %u/%llx\n", + i->tgid, i->fd, gl_name.ln_type, + (unsigned long long)gl_name.ln_number); + } +} + +static int gfs2_glockfd_seq_show(struct seq_file *seq, void *iter_ptr) +{ + struct gfs2_glockfd_iter *i = seq->private; + struct inode *inode = file_inode(i->file); + struct gfs2_glock *gl; + + inode_lock_shared(inode); + gl = GFS2_I(inode)->i_iopen_gh.gh_gl; + if (gl) { + seq_printf(seq, "%d %u %u/%llx\n", + i->tgid, i->fd, gl->gl_name.ln_type, + (unsigned long long)gl->gl_name.ln_number); + } + gfs2_glockfd_seq_show_flock(seq, i); + inode_unlock_shared(inode); + return 0; +} + +static const struct seq_operations gfs2_glockfd_seq_ops = { + .start = gfs2_glockfd_seq_start, + .next = gfs2_glockfd_seq_next, + .stop = gfs2_glockfd_seq_stop, + .show = gfs2_glockfd_seq_show, +}; + +static int gfs2_glockfd_open(struct inode *inode, struct file *file) +{ + struct gfs2_glockfd_iter *i; + struct gfs2_sbd *sdp = inode->i_private; + + i = __seq_open_private(file, &gfs2_glockfd_seq_ops, + sizeof(struct gfs2_glockfd_iter)); + if (!i) + return -ENOMEM; + i->sb = sdp->sd_vfs; + return 0; +} + +static const struct file_operations gfs2_glockfd_fops = { .owner = THIS_MODULE, - .open = gfs2_sbstats_open, + .open = gfs2_glockfd_open, .read = seq_read, .llseek = seq_lseek, .release = seq_release_private, }; -int gfs2_create_debugfs_file(struct gfs2_sbd *sdp) +DEFINE_SEQ_ATTRIBUTE(gfs2_sbstats); + +void gfs2_create_debugfs_file(struct gfs2_sbd *sdp) { sdp->debugfs_dir = debugfs_create_dir(sdp->sd_table_name, gfs2_root); - if (!sdp->debugfs_dir) - return -ENOMEM; - sdp->debugfs_dentry_glocks = debugfs_create_file("glocks", - S_IFREG | S_IRUGO, - sdp->debugfs_dir, sdp, - &gfs2_glocks_fops); - if (!sdp->debugfs_dentry_glocks) - goto fail; - - sdp->debugfs_dentry_glstats = debugfs_create_file("glstats", - S_IFREG | S_IRUGO, - sdp->debugfs_dir, sdp, - &gfs2_glstats_fops); - if (!sdp->debugfs_dentry_glstats) - goto fail; - - sdp->debugfs_dentry_sbstats = debugfs_create_file("sbstats", - S_IFREG | S_IRUGO, - sdp->debugfs_dir, sdp, - &gfs2_sbstats_fops); - if (!sdp->debugfs_dentry_sbstats) - goto fail; - return 0; -fail: - gfs2_delete_debugfs_file(sdp); - return -ENOMEM; + debugfs_create_file("glocks", S_IFREG | S_IRUGO, sdp->debugfs_dir, sdp, + &gfs2_glocks_fops); + + debugfs_create_file("glockfd", S_IFREG | S_IRUGO, sdp->debugfs_dir, sdp, + &gfs2_glockfd_fops); + + debugfs_create_file("glstats", S_IFREG | S_IRUGO, sdp->debugfs_dir, sdp, + &gfs2_glstats_fops); + + debugfs_create_file("sbstats", S_IFREG | S_IRUGO, sdp->debugfs_dir, sdp, + &gfs2_sbstats_fops); } void gfs2_delete_debugfs_file(struct gfs2_sbd *sdp) { - if (sdp->debugfs_dir) { - if (sdp->debugfs_dentry_glocks) { - debugfs_remove(sdp->debugfs_dentry_glocks); - sdp->debugfs_dentry_glocks = NULL; - } - if (sdp->debugfs_dentry_glstats) { - debugfs_remove(sdp->debugfs_dentry_glstats); - sdp->debugfs_dentry_glstats = NULL; - } - if (sdp->debugfs_dentry_sbstats) { - debugfs_remove(sdp->debugfs_dentry_sbstats); - sdp->debugfs_dentry_sbstats = NULL; - } - debugfs_remove(sdp->debugfs_dir); - sdp->debugfs_dir = NULL; - } + debugfs_remove_recursive(sdp->debugfs_dir); + sdp->debugfs_dir = NULL; } -int gfs2_register_debugfs(void) +void gfs2_register_debugfs(void) { gfs2_root = debugfs_create_dir("gfs2", NULL); - return gfs2_root ? 0 : -ENOMEM; } void gfs2_unregister_debugfs(void) |