X-Git-Url: https://git.openpandora.org/cgi-bin/gitweb.cgi?p=pandora-kernel.git;a=blobdiff_plain;f=fs%2Fgfs2%2Fglock.c;h=438146904b5839727ac2c378c5df87f574cfedaa;hp=b6edf7c0923bc8fa045835dec36cf1d0e02bccf2;hb=1c1afa3c053d4ccdf44e5a4e159005cdfd48bfc6;hpb=ad73c67e792c752ddc99f2b0587abae05255dd6d diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index b6edf7c0923b..438146904b58 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -4,7 +4,7 @@ * * 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 v.2. + * of the GNU General Public License version 2. */ #include @@ -15,13 +15,13 @@ #include #include #include -#include #include #include +#include +#include #include #include "gfs2.h" -#include "lm_interface.h" #include "incore.h" #include "glock.h" #include "glops.h" @@ -33,21 +33,74 @@ #include "super.h" #include "util.h" -/* Must be kept in sync with the beginning of struct gfs2_glock */ -struct glock_plug { - struct list_head gl_list; - unsigned long gl_flags; -}; - struct greedy { struct gfs2_holder gr_gh; - struct work_struct gr_work; + struct delayed_work gr_work; +}; + +struct gfs2_gl_hash_bucket { + struct hlist_head hb_list; }; typedef void (*glock_examiner) (struct gfs2_glock * gl); static int gfs2_dump_lockstate(struct gfs2_sbd *sdp); static int dump_glock(struct gfs2_glock *gl); +static int dump_inode(struct gfs2_inode *ip); + +#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) + +static struct gfs2_gl_hash_bucket gl_hash_table[GFS2_GL_HASH_SIZE]; + +/* + * Despite what you might think, the numbers below are not arbitrary :-) + * They are taken from the ipv4 routing hash code, which is well tested + * and thus should be nearly optimal. Later on we might tweek the numbers + * but for now this should be fine. + * + * The reason for putting the locks in a separate array from the list heads + * is that we can have fewer locks than list heads and save memory. We use + * the same hash function for both, but with a different hash mask. + */ +#if defined(CONFIG_SMP) || defined(CONFIG_DEBUG_SPINLOCK) || \ + defined(CONFIG_PROVE_LOCKING) + +#ifdef CONFIG_LOCKDEP +# define GL_HASH_LOCK_SZ 256 +#else +# if NR_CPUS >= 32 +# define GL_HASH_LOCK_SZ 4096 +# elif NR_CPUS >= 16 +# define GL_HASH_LOCK_SZ 2048 +# elif NR_CPUS >= 8 +# define GL_HASH_LOCK_SZ 1024 +# elif NR_CPUS >= 4 +# define GL_HASH_LOCK_SZ 512 +# else +# define GL_HASH_LOCK_SZ 256 +# endif +#endif + +/* We never want more locks than chains */ +#if GFS2_GL_HASH_SIZE < GL_HASH_LOCK_SZ +# undef GL_HASH_LOCK_SZ +# define GL_HASH_LOCK_SZ GFS2_GL_HASH_SIZE +#endif + +static rwlock_t gl_hash_locks[GL_HASH_LOCK_SZ]; + +static inline rwlock_t *gl_lock_addr(unsigned int x) +{ + return &gl_hash_locks[x & (GL_HASH_LOCK_SZ-1)]; +} +#else /* not SMP, so no spinlocks required */ +static inline rwlock_t *gl_lock_addr(unsigned int x) +{ + return NULL; +} +#endif /** * relaxed_state_ok - is a requested lock compatible with the current lock mode? @@ -83,12 +136,14 @@ static inline int relaxed_state_ok(unsigned int actual, unsigned requested, * Returns: The number of the corresponding hash bucket */ -static unsigned int gl_hash(struct lm_lockname *name) +static unsigned int gl_hash(const struct gfs2_sbd *sdp, + const struct lm_lockname *name) { unsigned int h; - h = jhash(&name->ln_number, sizeof(uint64_t), 0); + 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; return h; @@ -123,23 +178,7 @@ static void glock_free(struct gfs2_glock *gl) void gfs2_glock_hold(struct gfs2_glock *gl) { - kref_get(&gl->gl_ref); -} - -/* All work is done after the return from kref_put() so we - can release the write_lock before the free. */ - -static void kill_glock(struct kref *kref) -{ - struct gfs2_glock *gl = container_of(kref, struct gfs2_glock, gl_ref); - struct gfs2_sbd *sdp = gl->gl_sbd; - - gfs2_assert(sdp, gl->gl_state == LM_ST_UNLOCKED); - gfs2_assert(sdp, list_empty(&gl->gl_reclaim)); - gfs2_assert(sdp, list_empty(&gl->gl_holders)); - gfs2_assert(sdp, list_empty(&gl->gl_waiters1)); - gfs2_assert(sdp, list_empty(&gl->gl_waiters2)); - gfs2_assert(sdp, list_empty(&gl->gl_waiters3)); + atomic_inc(&gl->gl_ref); } /** @@ -150,24 +189,26 @@ static void kill_glock(struct kref *kref) int gfs2_glock_put(struct gfs2_glock *gl) { - struct gfs2_sbd *sdp = gl->gl_sbd; - struct gfs2_gl_hash_bucket *bucket = gl->gl_bucket; int rv = 0; + struct gfs2_sbd *sdp = gl->gl_sbd; - mutex_lock(&sdp->sd_invalidate_inodes_mutex); - - write_lock(&bucket->hb_lock); - if (kref_put(&gl->gl_ref, kill_glock)) { - list_del_init(&gl->gl_list); - write_unlock(&bucket->hb_lock); + write_lock(gl_lock_addr(gl->gl_hash)); + if (atomic_dec_and_test(&gl->gl_ref)) { + hlist_del(&gl->gl_list); + write_unlock(gl_lock_addr(gl->gl_hash)); BUG_ON(spin_is_locked(&gl->gl_spin)); + gfs2_assert(sdp, gl->gl_state == LM_ST_UNLOCKED); + gfs2_assert(sdp, list_empty(&gl->gl_reclaim)); + gfs2_assert(sdp, list_empty(&gl->gl_holders)); + gfs2_assert(sdp, list_empty(&gl->gl_waiters1)); + gfs2_assert(sdp, list_empty(&gl->gl_waiters2)); + gfs2_assert(sdp, list_empty(&gl->gl_waiters3)); glock_free(gl); rv = 1; goto out; } - write_unlock(&bucket->hb_lock); - out: - mutex_unlock(&sdp->sd_invalidate_inodes_mutex); + write_unlock(gl_lock_addr(gl->gl_hash)); +out: return rv; } @@ -203,18 +244,20 @@ static inline int queue_empty(struct gfs2_glock *gl, struct list_head *head) * Returns: NULL, or the struct gfs2_glock with the requested number */ -static struct gfs2_glock *search_bucket(struct gfs2_gl_hash_bucket *bucket, - struct lm_lockname *name) +static struct gfs2_glock *search_bucket(unsigned int hash, + const struct gfs2_sbd *sdp, + const struct lm_lockname *name) { struct gfs2_glock *gl; + struct hlist_node *h; - list_for_each_entry(gl, &bucket->hb_list, gl_list) { - if (test_bit(GLF_PLUG, &gl->gl_flags)) - continue; + hlist_for_each_entry(gl, h, &gl_hash_table[hash].hb_list, gl_list) { if (!lm_name_equal(&gl->gl_name, name)) continue; + if (gl->gl_sbd != sdp) + continue; - kref_get(&gl->gl_ref); + atomic_inc(&gl->gl_ref); return gl; } @@ -230,15 +273,15 @@ static struct gfs2_glock *search_bucket(struct gfs2_gl_hash_bucket *bucket, * Returns: NULL, or the struct gfs2_glock with the requested number */ -static struct gfs2_glock *gfs2_glock_find(struct gfs2_sbd *sdp, - struct lm_lockname *name) +static struct gfs2_glock *gfs2_glock_find(const struct gfs2_sbd *sdp, + const struct lm_lockname *name) { - struct gfs2_gl_hash_bucket *bucket = &sdp->sd_gl_hash[gl_hash(name)]; + unsigned int hash = gl_hash(sdp, name); struct gfs2_glock *gl; - read_lock(&bucket->hb_lock); - gl = search_bucket(bucket, name); - read_unlock(&bucket->hb_lock); + read_lock(gl_lock_addr(hash)); + gl = search_bucket(hash, sdp, name); + read_unlock(gl_lock_addr(hash)); return gl; } @@ -256,22 +299,18 @@ static struct gfs2_glock *gfs2_glock_find(struct gfs2_sbd *sdp, * Returns: errno */ -int gfs2_glock_get(struct gfs2_sbd *sdp, uint64_t number, - struct gfs2_glock_operations *glops, int create, +int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number, + const struct gfs2_glock_operations *glops, int create, struct gfs2_glock **glp) { - struct lm_lockname name; + struct lm_lockname name = { .ln_number = number, .ln_type = glops->go_type }; struct gfs2_glock *gl, *tmp; - struct gfs2_gl_hash_bucket *bucket; + unsigned int hash = gl_hash(sdp, &name); int error; - name.ln_number = number; - name.ln_type = glops->go_type; - bucket = &sdp->sd_gl_hash[gl_hash(&name)]; - - read_lock(&bucket->hb_lock); - gl = search_bucket(bucket, &name); - read_unlock(&bucket->hb_lock); + read_lock(gl_lock_addr(hash)); + gl = search_bucket(hash, sdp, &name); + read_unlock(gl_lock_addr(hash)); if (gl || !create) { *glp = gl; @@ -282,36 +321,26 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, uint64_t number, if (!gl) return -ENOMEM; - memset(gl, 0, sizeof(struct gfs2_glock)); - - INIT_LIST_HEAD(&gl->gl_list); + gl->gl_flags = 0; gl->gl_name = name; - kref_init(&gl->gl_ref); - - spin_lock_init(&gl->gl_spin); - + atomic_set(&gl->gl_ref, 1); gl->gl_state = LM_ST_UNLOCKED; + gl->gl_hash = hash; gl->gl_owner = NULL; gl->gl_ip = 0; - INIT_LIST_HEAD(&gl->gl_holders); - INIT_LIST_HEAD(&gl->gl_waiters1); - INIT_LIST_HEAD(&gl->gl_waiters2); - INIT_LIST_HEAD(&gl->gl_waiters3); - gl->gl_ops = glops; - - gl->gl_bucket = bucket; - INIT_LIST_HEAD(&gl->gl_reclaim); - + gl->gl_req_gh = NULL; + gl->gl_req_bh = NULL; + gl->gl_vn = 0; + gl->gl_stamp = jiffies; + gl->gl_object = NULL; gl->gl_sbd = sdp; - + gl->gl_aspace = NULL; lops_init_le(&gl->gl_le, &gfs2_glock_lops); - INIT_LIST_HEAD(&gl->gl_ail_list); /* If this glock protects actual on-disk data or metadata blocks, create a VFS inode to manage the pages/buffers holding them. */ - if (glops == &gfs2_inode_glops || - glops == &gfs2_rgrp_glops) { + if (glops == &gfs2_inode_glops || glops == &gfs2_rgrp_glops) { gl->gl_aspace = gfs2_aspace_get(sdp); if (!gl->gl_aspace) { error = -ENOMEM; @@ -323,28 +352,26 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, uint64_t number, if (error) goto fail_aspace; - write_lock(&bucket->hb_lock); - tmp = search_bucket(bucket, &name); + write_lock(gl_lock_addr(hash)); + tmp = search_bucket(hash, sdp, &name); if (tmp) { - write_unlock(&bucket->hb_lock); + write_unlock(gl_lock_addr(hash)); glock_free(gl); gl = tmp; } else { - list_add_tail(&gl->gl_list, &bucket->hb_list); - write_unlock(&bucket->hb_lock); + hlist_add_head(&gl->gl_list, &gl_hash_table[hash].hb_list); + write_unlock(gl_lock_addr(hash)); } *glp = gl; return 0; - fail_aspace: +fail_aspace: if (gl->gl_aspace) gfs2_aspace_put(gl->gl_aspace); - - fail: - kmem_cache_free(gfs2_glock_cachep, gl); - +fail: + kmem_cache_free(gfs2_glock_cachep, gl); return error; } @@ -484,7 +511,7 @@ static int rq_promote(struct gfs2_holder *gh) { struct gfs2_glock *gl = gh->gh_gl; struct gfs2_sbd *sdp = gl->gl_sbd; - struct gfs2_glock_operations *glops = gl->gl_ops; + const struct gfs2_glock_operations *glops = gl->gl_ops; if (!relaxed_state_ok(gl->gl_state, gh->gh_state, gh->gh_flags)) { if (list_empty(&gl->gl_holders)) { @@ -499,9 +526,7 @@ static int rq_promote(struct gfs2_holder *gh) gfs2_reclaim_glock(sdp); } - glops->go_xmote_th(gl, gh->gh_state, - gh->gh_flags); - + glops->go_xmote_th(gl, gh->gh_state, gh->gh_flags); spin_lock(&gl->gl_spin); } return 1; @@ -539,7 +564,7 @@ static int rq_promote(struct gfs2_holder *gh) static int rq_demote(struct gfs2_holder *gh) { struct gfs2_glock *gl = gh->gh_gl; - struct gfs2_glock_operations *glops = gl->gl_ops; + const struct gfs2_glock_operations *glops = gl->gl_ops; if (!list_empty(&gl->gl_holders)) return 1; @@ -589,7 +614,7 @@ static int rq_greedy(struct gfs2_holder *gh) gfs2_holder_uninit(gh); kfree(container_of(gh, struct greedy, gr_gh)); - spin_lock(&gl->gl_spin); + spin_lock(&gl->gl_spin); return 0; } @@ -661,9 +686,9 @@ static void gfs2_glmutex_lock(struct gfs2_glock *gl) set_bit(HIF_MUTEX, &gh.gh_iflags); spin_lock(&gl->gl_spin); - if (test_and_set_bit(GLF_LOCK, &gl->gl_flags)) + if (test_and_set_bit(GLF_LOCK, &gl->gl_flags)) { list_add_tail(&gh.gh_list, &gl->gl_waiters1); - else { + } else { gl->gl_owner = current; gl->gl_ip = (unsigned long)__builtin_return_address(0); complete(&gh.gh_wait); @@ -686,9 +711,9 @@ static int gfs2_glmutex_trylock(struct gfs2_glock *gl) int acquired = 1; spin_lock(&gl->gl_spin); - if (test_and_set_bit(GLF_LOCK, &gl->gl_flags)) + if (test_and_set_bit(GLF_LOCK, &gl->gl_flags)) { acquired = 0; - else { + } else { gl->gl_owner = current; gl->gl_ip = (unsigned long)__builtin_return_address(0); } @@ -744,7 +769,7 @@ restart: } else { spin_unlock(&gl->gl_spin); - new_gh = gfs2_holder_get(gl, state, LM_FLAG_TRY, GFP_KERNEL); + new_gh = gfs2_holder_get(gl, state, LM_FLAG_TRY, GFP_NOFS); if (!new_gh) return; set_bit(HIF_DEMOTE, &new_gh->gh_iflags); @@ -760,21 +785,6 @@ out: gfs2_holder_put(new_gh); } -void gfs2_glock_inode_squish(struct inode *inode) -{ - struct gfs2_holder gh; - struct gfs2_glock *gl = GFS2_I(inode)->i_gl; - gfs2_holder_init(gl, LM_ST_UNLOCKED, 0, &gh); - set_bit(HIF_DEMOTE, &gh.gh_iflags); - spin_lock(&gl->gl_spin); - gfs2_assert(inode->i_sb->s_fs_info, list_empty(&gl->gl_holders)); - list_add_tail(&gh.gh_list, &gl->gl_waiters2); - run_queue(gl); - spin_unlock(&gl->gl_spin); - wait_for_completion(&gh.gh_wait); - gfs2_holder_uninit(&gh); -} - /** * state_change - record that the glock is now in a different state * @gl: the glock @@ -809,7 +819,7 @@ static void state_change(struct gfs2_glock *gl, unsigned int new_state) static void xmote_bh(struct gfs2_glock *gl, unsigned int ret) { struct gfs2_sbd *sdp = gl->gl_sbd; - struct gfs2_glock_operations *glops = gl->gl_ops; + const struct gfs2_glock_operations *glops = gl->gl_ops; struct gfs2_holder *gh = gl->gl_req_gh; int prev_state = gl->gl_state; int op_done = 1; @@ -822,32 +832,30 @@ static void xmote_bh(struct gfs2_glock *gl, unsigned int ret) if (prev_state != LM_ST_UNLOCKED && !(ret & LM_OUT_CACHEABLE)) { if (glops->go_inval) - glops->go_inval(gl, DIO_METADATA | DIO_DATA); + glops->go_inval(gl, DIO_METADATA); } else if (gl->gl_state == LM_ST_DEFERRED) { /* We might not want to do this here. Look at moving to the inode glops. */ if (glops->go_inval) - glops->go_inval(gl, DIO_DATA); + glops->go_inval(gl, 0); } /* Deal with each possible exit condition */ if (!gh) gl->gl_stamp = jiffies; - else if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags))) { spin_lock(&gl->gl_spin); list_del_init(&gh->gh_list); gh->gh_error = -EIO; spin_unlock(&gl->gl_spin); - } else if (test_bit(HIF_DEMOTE, &gh->gh_iflags)) { spin_lock(&gl->gl_spin); list_del_init(&gh->gh_list); if (gl->gl_state == gh->gh_state || - gl->gl_state == LM_ST_UNLOCKED) + gl->gl_state == LM_ST_UNLOCKED) { gh->gh_error = 0; - else { + } else { if (gfs2_assert_warn(sdp, gh->gh_flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB)) == -1) fs_warn(sdp, "ret = 0x%.8X\n", ret); @@ -856,7 +864,7 @@ static void xmote_bh(struct gfs2_glock *gl, unsigned int ret) spin_unlock(&gl->gl_spin); if (ret & LM_OUT_CANCELED) - handle_callback(gl, LM_ST_UNLOCKED); /* Lame */ + handle_callback(gl, LM_ST_UNLOCKED); } else if (ret & LM_OUT_CANCELED) { spin_lock(&gl->gl_spin); @@ -919,7 +927,7 @@ static void xmote_bh(struct gfs2_glock *gl, unsigned int ret) void gfs2_glock_xmote_th(struct gfs2_glock *gl, unsigned int state, int flags) { struct gfs2_sbd *sdp = gl->gl_sbd; - struct gfs2_glock_operations *glops = gl->gl_ops; + const struct gfs2_glock_operations *glops = gl->gl_ops; int lck_flags = flags & (LM_FLAG_TRY | LM_FLAG_TRY_1CB | LM_FLAG_NOEXP | LM_FLAG_ANY | LM_FLAG_PRIORITY); @@ -930,17 +938,13 @@ void gfs2_glock_xmote_th(struct gfs2_glock *gl, unsigned int state, int flags) gfs2_assert_warn(sdp, state != LM_ST_UNLOCKED); gfs2_assert_warn(sdp, state != gl->gl_state); - if (gl->gl_state == LM_ST_EXCLUSIVE) { - if (glops->go_sync) - glops->go_sync(gl, - DIO_METADATA | DIO_DATA | DIO_RELEASE); - } + if (gl->gl_state == LM_ST_EXCLUSIVE && glops->go_sync) + glops->go_sync(gl); gfs2_glock_hold(gl); gl->gl_req_bh = xmote_bh; - lck_ret = gfs2_lm_lock(sdp, gl->gl_lock, gl->gl_state, state, - lck_flags); + lck_ret = gfs2_lm_lock(sdp, gl->gl_lock, gl->gl_state, state, lck_flags); if (gfs2_assert_withdraw(sdp, !(lck_ret & LM_OUT_ERROR))) return; @@ -964,7 +968,7 @@ void gfs2_glock_xmote_th(struct gfs2_glock *gl, unsigned int state, int flags) static void drop_bh(struct gfs2_glock *gl, unsigned int ret) { struct gfs2_sbd *sdp = gl->gl_sbd; - struct gfs2_glock_operations *glops = gl->gl_ops; + const struct gfs2_glock_operations *glops = gl->gl_ops; struct gfs2_holder *gh = gl->gl_req_gh; clear_bit(GLF_PREFETCH, &gl->gl_flags); @@ -976,7 +980,7 @@ static void drop_bh(struct gfs2_glock *gl, unsigned int ret) state_change(gl, LM_ST_UNLOCKED); if (glops->go_inval) - glops->go_inval(gl, DIO_METADATA | DIO_DATA); + glops->go_inval(gl, DIO_METADATA); if (gh) { spin_lock(&gl->gl_spin); @@ -1014,18 +1018,15 @@ static void drop_bh(struct gfs2_glock *gl, unsigned int ret) void gfs2_glock_drop_th(struct gfs2_glock *gl) { struct gfs2_sbd *sdp = gl->gl_sbd; - struct gfs2_glock_operations *glops = gl->gl_ops; + const struct gfs2_glock_operations *glops = gl->gl_ops; unsigned int ret; gfs2_assert_warn(sdp, test_bit(GLF_LOCK, &gl->gl_flags)); gfs2_assert_warn(sdp, queue_empty(gl, &gl->gl_holders)); gfs2_assert_warn(sdp, gl->gl_state != LM_ST_UNLOCKED); - if (gl->gl_state == LM_ST_EXCLUSIVE) { - if (glops->go_sync) - glops->go_sync(gl, - DIO_METADATA | DIO_DATA | DIO_RELEASE); - } + if (gl->gl_state == LM_ST_EXCLUSIVE && glops->go_sync) + glops->go_sync(gl); gfs2_glock_hold(gl); gl->gl_req_bh = drop_bh; @@ -1057,9 +1058,8 @@ static void do_cancels(struct gfs2_holder *gh) while (gl->gl_req_gh != gh && !test_bit(HIF_HOLDER, &gh->gh_iflags) && !list_empty(&gh->gh_list)) { - if (gl->gl_req_bh && - !(gl->gl_req_gh && - (gl->gl_req_gh->gh_flags & GL_NOCANCEL))) { + if (gl->gl_req_bh && !(gl->gl_req_gh && + (gl->gl_req_gh->gh_flags & GL_NOCANCEL))) { spin_unlock(&gl->gl_spin); gfs2_lm_cancel(gl->gl_sbd, gl->gl_lock); msleep(100); @@ -1085,7 +1085,7 @@ static int glock_wait_internal(struct gfs2_holder *gh) { struct gfs2_glock *gl = gh->gh_gl; struct gfs2_sbd *sdp = gl->gl_sbd; - struct gfs2_glock_operations *glops = gl->gl_ops; + const struct gfs2_glock_operations *glops = gl->gl_ops; if (test_bit(HIF_ABORTED, &gh->gh_iflags)) return -EIO; @@ -1113,8 +1113,7 @@ static int glock_wait_internal(struct gfs2_holder *gh) return gh->gh_error; gfs2_assert_withdraw(sdp, test_bit(HIF_HOLDER, &gh->gh_iflags)); - gfs2_assert_withdraw(sdp, relaxed_state_ok(gl->gl_state, - gh->gh_state, + gfs2_assert_withdraw(sdp, relaxed_state_ok(gl->gl_state, gh->gh_state, gh->gh_flags)); if (test_bit(HIF_FIRST, &gh->gh_iflags)) { @@ -1169,7 +1168,13 @@ static void add_to_queue(struct gfs2_holder *gh) existing = find_holder_by_owner(&gl->gl_holders, gh->gh_owner); if (existing) { print_symbol(KERN_WARNING "original: %s\n", existing->gh_ip); + printk(KERN_INFO "pid : %d\n", existing->gh_owner->pid); + printk(KERN_INFO "lock type : %d lock state : %d\n", + existing->gh_gl->gl_name.ln_type, existing->gh_gl->gl_state); print_symbol(KERN_WARNING "new: %s\n", gh->gh_ip); + printk(KERN_INFO "pid : %d\n", gh->gh_owner->pid); + printk(KERN_INFO "lock type : %d lock state : %d\n", + gl->gl_name.ln_type, gl->gl_state); BUG(); } @@ -1183,7 +1188,7 @@ static void add_to_queue(struct gfs2_holder *gh) if (gh->gh_flags & LM_FLAG_PRIORITY) list_add(&gh->gh_list, &gl->gl_waiters3); else - list_add_tail(&gh->gh_list, &gl->gl_waiters3); + list_add_tail(&gh->gh_list, &gl->gl_waiters3); } /** @@ -1224,9 +1229,6 @@ restart: clear_bit(GLF_PREFETCH, &gl->gl_flags); - if (error == GLR_TRYFAILED && (gh->gh_flags & GL_DUMP)) - dump_glock(gl); - return error; } @@ -1292,10 +1294,7 @@ int gfs2_glock_wait(struct gfs2_holder *gh) void gfs2_glock_dq(struct gfs2_holder *gh) { struct gfs2_glock *gl = gh->gh_gl; - struct gfs2_glock_operations *glops = gl->gl_ops; - - if (gh->gh_flags & GL_SYNC) - set_bit(GLF_SYNC, &gl->gl_flags); + const struct gfs2_glock_operations *glops = gl->gl_ops; if (gh->gh_flags & GL_NOCACHE) handle_callback(gl, LM_ST_UNLOCKED); @@ -1311,11 +1310,6 @@ void gfs2_glock_dq(struct gfs2_holder *gh) if (glops->go_unlock) glops->go_unlock(gh); - if (test_bit(GLF_SYNC, &gl->gl_flags)) { - if (glops->go_sync) - glops->go_sync(gl, DIO_METADATA | DIO_DATA); - } - gl->gl_stamp = jiffies; spin_lock(&gl->gl_spin); @@ -1337,14 +1331,12 @@ void gfs2_glock_dq(struct gfs2_holder *gh) static void gfs2_glock_prefetch(struct gfs2_glock *gl, unsigned int state, int flags) { - struct gfs2_glock_operations *glops = gl->gl_ops; + const struct gfs2_glock_operations *glops = gl->gl_ops; spin_lock(&gl->gl_spin); - if (test_bit(GLF_LOCK, &gl->gl_flags) || - !list_empty(&gl->gl_holders) || - !list_empty(&gl->gl_waiters1) || - !list_empty(&gl->gl_waiters2) || + if (test_bit(GLF_LOCK, &gl->gl_flags) || !list_empty(&gl->gl_holders) || + !list_empty(&gl->gl_waiters1) || !list_empty(&gl->gl_waiters2) || !list_empty(&gl->gl_waiters3) || relaxed_state_ok(gl->gl_state, state, flags)) { spin_unlock(&gl->gl_spin); @@ -1358,12 +1350,12 @@ static void gfs2_glock_prefetch(struct gfs2_glock *gl, unsigned int state, glops->go_xmote_th(gl, state, flags); } -static void greedy_work(void *data) +static void greedy_work(struct work_struct *work) { - struct greedy *gr = data; + struct greedy *gr = container_of(work, struct greedy, gr_work.work); struct gfs2_holder *gh = &gr->gr_gh; struct gfs2_glock *gl = gh->gh_gl; - struct gfs2_glock_operations *glops = gl->gl_ops; + const struct gfs2_glock_operations *glops = gl->gl_ops; clear_bit(GLF_SKIP_WAITERS2, &gl->gl_flags); @@ -1412,7 +1404,7 @@ int gfs2_glock_be_greedy(struct gfs2_glock *gl, unsigned int time) gfs2_holder_init(gl, 0, 0, gh); set_bit(HIF_GREEDY, &gh->gh_iflags); - INIT_WORK(&gr->gr_work, greedy_work, gr); + INIT_DELAYED_WORK(&gr->gr_work, greedy_work); set_bit(GLF_SKIP_WAITERS2, &gl->gl_flags); schedule_delayed_work(&gr->gr_work, time); @@ -1444,9 +1436,9 @@ void gfs2_glock_dq_uninit(struct gfs2_holder *gh) * Returns: errno */ -int gfs2_glock_nq_num(struct gfs2_sbd *sdp, uint64_t number, - struct gfs2_glock_operations *glops, unsigned int state, - int flags, 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) { struct gfs2_glock *gl; int error; @@ -1469,26 +1461,20 @@ int gfs2_glock_nq_num(struct gfs2_sbd *sdp, uint64_t number, static int glock_compare(const void *arg_a, const void *arg_b) { - struct gfs2_holder *gh_a = *(struct gfs2_holder **)arg_a; - struct gfs2_holder *gh_b = *(struct gfs2_holder **)arg_b; - struct lm_lockname *a = &gh_a->gh_gl->gl_name; - struct lm_lockname *b = &gh_b->gh_gl->gl_name; - int ret = 0; + const struct gfs2_holder *gh_a = *(const struct gfs2_holder **)arg_a; + const struct gfs2_holder *gh_b = *(const struct gfs2_holder **)arg_b; + const struct lm_lockname *a = &gh_a->gh_gl->gl_name; + const struct lm_lockname *b = &gh_b->gh_gl->gl_name; if (a->ln_number > b->ln_number) - ret = 1; - else if (a->ln_number < b->ln_number) - ret = -1; - else { - if (gh_a->gh_state == LM_ST_SHARED && - gh_b->gh_state == LM_ST_EXCLUSIVE) - ret = 1; - else if (!(gh_a->gh_flags & GL_LOCAL_EXCL) && - (gh_b->gh_flags & GL_LOCAL_EXCL)) - ret = 1; - } - - return ret; + return 1; + if (a->ln_number < b->ln_number) + return -1; + if (gh_a->gh_state == LM_ST_SHARED && gh_b->gh_state == LM_ST_EXCLUSIVE) + return 1; + if (!(gh_a->gh_flags & GL_LOCAL_EXCL) && (gh_b->gh_flags & GL_LOCAL_EXCL)) + return 1; + return 0; } /** @@ -1642,8 +1628,8 @@ void gfs2_glock_dq_uninit_m(unsigned int num_gh, struct gfs2_holder *ghs) * Returns: errno */ -void gfs2_glock_prefetch_num(struct gfs2_sbd *sdp, uint64_t number, - struct gfs2_glock_operations *glops, +void gfs2_glock_prefetch_num(struct gfs2_sbd *sdp, u64 number, + const struct gfs2_glock_operations *glops, unsigned int state, int flags) { struct gfs2_glock *gl; @@ -1708,19 +1694,6 @@ void gfs2_lvb_unhold(struct gfs2_glock *gl) gfs2_glock_put(gl); } -#if 0 -void gfs2_lvb_sync(struct gfs2_glock *gl) -{ - gfs2_glmutex_lock(gl); - - gfs2_assert(gl->gl_sbd, atomic_read(&gl->gl_lvb_count)); - if (!gfs2_assert_warn(gl->gl_sbd, gfs2_glock_is_held_excl(gl))) - gfs2_lm_sync_lvb(gl->gl_sbd, gl->gl_lock, gl->gl_lvb); - - gfs2_glmutex_unlock(gl); -} -#endif /* 0 */ - static void blocking_cb(struct gfs2_sbd *sdp, struct lm_lockname *name, unsigned int state) { @@ -1743,7 +1716,7 @@ static void blocking_cb(struct gfs2_sbd *sdp, struct lm_lockname *name, /** * gfs2_glock_cb - Callback used by locking module - * @fsdata: Pointer to the superblock + * @sdp: Pointer to the superblock * @type: Type of callback * @data: Type dependent data pointer * @@ -1752,9 +1725,9 @@ static void blocking_cb(struct gfs2_sbd *sdp, struct lm_lockname *name, * a journal from another client needs to be recovered. */ -void gfs2_glock_cb(lm_fsdata_t *fsdata, unsigned int type, void *data) +void gfs2_glock_cb(void *cb_data, unsigned int type, void *data) { - struct gfs2_sbd *sdp = (struct gfs2_sbd *)fsdata; + struct gfs2_sbd *sdp = cb_data; switch (type) { case LM_CB_NEED_E: @@ -1799,22 +1772,6 @@ void gfs2_glock_cb(lm_fsdata_t *fsdata, unsigned int type, void *data) } } -/** - * gfs2_iopen_go_callback - Try to kick the inode/vnode associated with an - * iopen glock from memory - * @io_gl: the iopen glock - * @state: the state into which the glock should be put - * - */ - -void gfs2_iopen_go_callback(struct gfs2_glock *io_gl, unsigned int state) -{ - - if (state != LM_ST_UNLOCKED) - return; - /* FIXME: remove this? */ -} - /** * demote_ok - Check to see if it's ok to unlock a glock * @gl: the glock @@ -1825,14 +1782,13 @@ void gfs2_iopen_go_callback(struct gfs2_glock *io_gl, unsigned int state) static int demote_ok(struct gfs2_glock *gl) { struct gfs2_sbd *sdp = gl->gl_sbd; - struct gfs2_glock_operations *glops = gl->gl_ops; + const struct gfs2_glock_operations *glops = gl->gl_ops; int demote = 1; if (test_bit(GLF_STICKY, &gl->gl_flags)) demote = 0; else if (test_bit(GLF_PREFETCH, &gl->gl_flags)) - demote = time_after_eq(jiffies, - gl->gl_stamp + + demote = time_after_eq(jiffies, gl->gl_stamp + gfs2_tune_get(sdp, gt_prefetch_secs) * HZ); else if (glops->go_demote_ok) demote = glops->go_demote_ok(gl); @@ -1890,8 +1846,7 @@ void gfs2_reclaim_glock(struct gfs2_sbd *sdp) if (gfs2_glmutex_trylock(gl)) { if (queue_empty(gl, &gl->gl_holders) && - gl->gl_state != LM_ST_UNLOCKED && - demote_ok(gl)) + gl->gl_state != LM_ST_UNLOCKED && demote_ok(gl)) handle_callback(gl, LM_ST_UNLOCKED); gfs2_glmutex_unlock(gl); } @@ -1909,51 +1864,37 @@ void gfs2_reclaim_glock(struct gfs2_sbd *sdp) */ static int examine_bucket(glock_examiner examiner, struct gfs2_sbd *sdp, - struct gfs2_gl_hash_bucket *bucket) + unsigned int hash) { - struct glock_plug plug; - struct list_head *tmp; - struct gfs2_glock *gl; - int entries; - - /* Add "plug" to end of bucket list, work back up list from there */ - memset(&plug.gl_flags, 0, sizeof(unsigned long)); - set_bit(GLF_PLUG, &plug.gl_flags); - - write_lock(&bucket->hb_lock); - list_add(&plug.gl_list, &bucket->hb_list); - write_unlock(&bucket->hb_lock); - - for (;;) { - write_lock(&bucket->hb_lock); - - for (;;) { - tmp = plug.gl_list.next; - - if (tmp == &bucket->hb_list) { - list_del(&plug.gl_list); - entries = !list_empty(&bucket->hb_list); - write_unlock(&bucket->hb_lock); - return entries; - } - gl = list_entry(tmp, struct gfs2_glock, gl_list); + struct gfs2_glock *gl, *prev = NULL; + int has_entries = 0; + struct hlist_head *head = &gl_hash_table[hash].hb_list; - /* Move plug up list */ - list_move(&plug.gl_list, &gl->gl_list); - - if (test_bit(GLF_PLUG, &gl->gl_flags)) - continue; - - /* examiner() must glock_put() */ + read_lock(gl_lock_addr(hash)); + /* Can't use hlist_for_each_entry - don't want prefetch here */ + if (hlist_empty(head)) + goto out; + gl = list_entry(head->first, struct gfs2_glock, gl_list); + while(1) { + if (gl->gl_sbd == sdp) { gfs2_glock_hold(gl); - - break; + read_unlock(gl_lock_addr(hash)); + if (prev) + gfs2_glock_put(prev); + prev = gl; + examiner(gl); + has_entries = 1; + read_lock(gl_lock_addr(hash)); } - - write_unlock(&bucket->hb_lock); - - examiner(gl); + if (gl->gl_list.next == NULL) + break; + gl = list_entry(gl->gl_list.next, struct gfs2_glock, gl_list); } +out: + read_unlock(gl_lock_addr(hash)); + if (prev) + gfs2_glock_put(prev); + return has_entries; } /** @@ -1964,25 +1905,20 @@ static int examine_bucket(glock_examiner examiner, struct gfs2_sbd *sdp, static void scan_glock(struct gfs2_glock *gl) { + if (gl->gl_ops == &gfs2_inode_glops && gl->gl_object) + return; + if (gfs2_glmutex_trylock(gl)) { - if (gl->gl_ops == &gfs2_inode_glops) - goto out; if (queue_empty(gl, &gl->gl_holders) && - gl->gl_state != LM_ST_UNLOCKED && - demote_ok(gl)) + gl->gl_state != LM_ST_UNLOCKED && demote_ok(gl)) goto out_schedule; -out: gfs2_glmutex_unlock(gl); } - - gfs2_glock_put(gl); - return; out_schedule: gfs2_glmutex_unlock(gl); gfs2_glock_schedule_for_reclaim(gl); - gfs2_glock_put(gl); } /** @@ -1995,10 +1931,8 @@ void gfs2_scand_internal(struct gfs2_sbd *sdp) { unsigned int x; - for (x = 0; x < GFS2_GL_HASH_SIZE; x++) { - examine_bucket(scan_glock, sdp, &sdp->sd_gl_hash[x]); - cond_resched(); - } + for (x = 0; x < GFS2_GL_HASH_SIZE; x++) + examine_bucket(scan_glock, sdp, x); } /** @@ -2027,11 +1961,8 @@ static void clear_glock(struct gfs2_glock *gl) if (queue_empty(gl, &gl->gl_holders) && gl->gl_state != LM_ST_UNLOCKED) handle_callback(gl, LM_ST_UNLOCKED); - gfs2_glmutex_unlock(gl); } - - gfs2_glock_put(gl); } /** @@ -2053,11 +1984,10 @@ void gfs2_gl_hash_clear(struct gfs2_sbd *sdp, int wait) for (;;) { cont = 0; - - for (x = 0; x < GFS2_GL_HASH_SIZE; x++) - if (examine_bucket(clear_glock, sdp, - &sdp->sd_gl_hash[x])) + for (x = 0; x < GFS2_GL_HASH_SIZE; x++) { + if (examine_bucket(clear_glock, sdp, x)) cont = 1; + } if (!wait || !cont) break; @@ -2070,16 +2000,7 @@ void gfs2_gl_hash_clear(struct gfs2_sbd *sdp, int wait) t = jiffies; } - /* invalidate_inodes() requires that the sb inodes list - not change, but an async completion callback for an - unlock can occur which does glock_put() which - can call iput() which will change the sb inodes list. - invalidate_inodes_mutex prevents glock_put()'s during - an invalidate_inodes() */ - - mutex_lock(&sdp->sd_invalidate_inodes_mutex); invalidate_inodes(sdp->sd_vfs); - mutex_unlock(&sdp->sd_invalidate_inodes_mutex); msleep(10); } } @@ -2139,7 +2060,7 @@ static int dump_inode(struct gfs2_inode *ip) printk(KERN_INFO " num = %llu %llu\n", (unsigned long long)ip->i_num.no_formal_ino, (unsigned long long)ip->i_num.no_addr); - printk(KERN_INFO " type = %u\n", IF2DT(ip->i_di.di_mode)); + printk(KERN_INFO " type = %u\n", IF2DT(ip->i_inode.i_mode)); printk(KERN_INFO " i_flags ="); for (x = 0; x < 32; x++) if (test_bit(x, &ip->i_flags)) @@ -2167,14 +2088,15 @@ static int dump_glock(struct gfs2_glock *gl) spin_lock(&gl->gl_spin); - printk(KERN_INFO "Glock (%u, %llu)\n", gl->gl_name.ln_type, + printk(KERN_INFO "Glock 0x%p (%u, %llu)\n", gl, gl->gl_name.ln_type, (unsigned long long)gl->gl_name.ln_number); printk(KERN_INFO " gl_flags ="); - for (x = 0; x < 32; x++) + for (x = 0; x < 32; x++) { if (test_bit(x, &gl->gl_flags)) printk(" %u", x); + } printk(" \n"); - printk(KERN_INFO " gl_ref = %d\n", atomic_read(&gl->gl_ref.refcount)); + printk(KERN_INFO " gl_ref = %d\n", atomic_read(&gl->gl_ref)); printk(KERN_INFO " gl_state = %u\n", gl->gl_state); printk(KERN_INFO " gl_owner = %s\n", gl->gl_owner->comm); print_symbol(KERN_INFO " gl_ip = %s\n", gl->gl_ip); @@ -2187,8 +2109,8 @@ static int dump_glock(struct gfs2_glock *gl) printk(KERN_INFO " reclaim = %s\n", (list_empty(&gl->gl_reclaim)) ? "no" : "yes"); if (gl->gl_aspace) - printk(KERN_INFO " aspace = %lu\n", - gl->gl_aspace->i_mapping->nrpages); + printk(KERN_INFO " aspace = 0x%p nrpages = %lu\n", gl->gl_aspace, + gl->gl_aspace->i_mapping->nrpages); else printk(KERN_INFO " aspace = no\n"); printk(KERN_INFO " ail = %d\n", atomic_read(&gl->gl_ail_count)); @@ -2231,9 +2153,8 @@ static int dump_glock(struct gfs2_glock *gl) error = 0; - out: +out: spin_unlock(&gl->gl_spin); - return error; } @@ -2248,18 +2169,17 @@ static int dump_glock(struct gfs2_glock *gl) static int gfs2_dump_lockstate(struct gfs2_sbd *sdp) { - struct gfs2_gl_hash_bucket *bucket; struct gfs2_glock *gl; + struct hlist_node *h; unsigned int x; int error = 0; for (x = 0; x < GFS2_GL_HASH_SIZE; x++) { - bucket = &sdp->sd_gl_hash[x]; - read_lock(&bucket->hb_lock); + read_lock(gl_lock_addr(x)); - list_for_each_entry(gl, &bucket->hb_list, gl_list) { - if (test_bit(GLF_PLUG, &gl->gl_flags)) + hlist_for_each_entry(gl, h, &gl_hash_table[x].hb_list, gl_list) { + if (gl->gl_sbd != sdp) continue; error = dump_glock(gl); @@ -2267,7 +2187,7 @@ static int gfs2_dump_lockstate(struct gfs2_sbd *sdp) break; } - read_unlock(&bucket->hb_lock); + read_unlock(gl_lock_addr(x)); if (error) break; @@ -2277,3 +2197,17 @@ static int gfs2_dump_lockstate(struct gfs2_sbd *sdp) return error; } +int __init gfs2_glock_init(void) +{ + unsigned i; + for(i = 0; i < GFS2_GL_HASH_SIZE; i++) { + INIT_HLIST_HEAD(&gl_hash_table[i].hb_list); + } +#ifdef GL_HASH_LOCK_SZ + for(i = 0; i < GL_HASH_LOCK_SZ; i++) { + rwlock_init(&gl_hash_locks[i]); + } +#endif + return 0; +} +