diff options
author | Linus Torvalds <torvalds@linux-foundation.org> | 2023-02-23 01:00:53 +0300 |
---|---|---|
committer | Linus Torvalds <torvalds@linux-foundation.org> | 2023-02-23 01:00:53 +0300 |
commit | b7ee88128242b3460b0016a7e42207c9799f73b7 (patch) | |
tree | 04f4965c3ae015e1f77ebbd321a0b016020e476d | |
parent | 28e335208ce90c2cca9990543983b97ccc66f302 (diff) | |
parent | c1b0c3cfcbad25d2c412863c27638c933f1d911b (diff) | |
download | linux-b7ee88128242b3460b0016a7e42207c9799f73b7.tar.xz |
Merge tag 'gfs2-v6.2-rc5-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2
Pull gfs2 updates from Andreas Gruenbacher:
- Fix a race when disassociating inodes from their glocks after
iget_failed()
- On filesystems with a block size smaller than the page size, make
sure that ->writepages() writes out all buffers of journaled inodes
- Various improvements to the way the delete workqueue is drained to
speed up unmount and prevent leftover inodes. At unmount time, evict
deleted inodes cooperatively across the cluster to avoid unnecessary
timeouts
- Various minor cleanups and fixes
* tag 'gfs2-v6.2-rc5-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2:
gfs2: Convert gfs2_page_add_databufs to folios
gfs2: jdata writepage fix
gfs2: Improve gfs2_make_fs_rw error handling
Revert "GFS2: free disk inode which is deleted by remote node -V2"
gfs2: Evict inodes cooperatively
gfs2: Flush delete work before shrinking inode cache
gfs2: Cease delete work during unmount
gfs2: Add SDF_DEACTIVATING super block flag
gfs2: check gl_object in rgrp glops
gfs2: Split the two kinds of glock "delete" work
gfs2: Move delete workqueue into super block
gfs2: Get rid of GLF_PENDING_DELETE flag
gfs2: Make glock lru list scanning safer
gfs2: Clean up gfs2_scan_glock_lru
gfs2: Improve gfs2_upgrade_iopen_glock comment
gfs2: gl_object races fix
-rw-r--r-- | fs/gfs2/aops.c | 9 | ||||
-rw-r--r-- | fs/gfs2/aops.h | 4 | ||||
-rw-r--r-- | fs/gfs2/bmap.c | 4 | ||||
-rw-r--r-- | fs/gfs2/dentry.c | 18 | ||||
-rw-r--r-- | fs/gfs2/glock.c | 128 | ||||
-rw-r--r-- | fs/gfs2/glock.h | 4 | ||||
-rw-r--r-- | fs/gfs2/glops.c | 21 | ||||
-rw-r--r-- | fs/gfs2/incore.h | 11 | ||||
-rw-r--r-- | fs/gfs2/inode.c | 8 | ||||
-rw-r--r-- | fs/gfs2/ops_fstype.c | 71 | ||||
-rw-r--r-- | fs/gfs2/rgrp.c | 2 | ||||
-rw-r--r-- | fs/gfs2/super.c | 49 | ||||
-rw-r--r-- | fs/gfs2/sys.c | 2 |
13 files changed, 204 insertions, 127 deletions
diff --git a/fs/gfs2/aops.c b/fs/gfs2/aops.c index e782b4f1d104..2748a82de42a 100644 --- a/fs/gfs2/aops.c +++ b/fs/gfs2/aops.c @@ -37,10 +37,10 @@ #include "aops.h" -void gfs2_page_add_databufs(struct gfs2_inode *ip, struct page *page, - unsigned int from, unsigned int len) +void gfs2_trans_add_databufs(struct gfs2_inode *ip, struct folio *folio, + unsigned int from, unsigned int len) { - struct buffer_head *head = page_buffers(page); + struct buffer_head *head = folio_buffers(folio); unsigned int bsize = head->b_size; struct buffer_head *bh; unsigned int to = from + len; @@ -127,7 +127,6 @@ static int __gfs2_jdata_writepage(struct page *page, struct writeback_control *w { struct inode *inode = page->mapping->host; struct gfs2_inode *ip = GFS2_I(inode); - struct gfs2_sbd *sdp = GFS2_SB(inode); if (PageChecked(page)) { ClearPageChecked(page); @@ -135,7 +134,7 @@ static int __gfs2_jdata_writepage(struct page *page, struct writeback_control *w create_empty_buffers(page, inode->i_sb->s_blocksize, BIT(BH_Dirty)|BIT(BH_Uptodate)); } - gfs2_page_add_databufs(ip, page, 0, sdp->sd_vfs->s_blocksize); + gfs2_trans_add_databufs(ip, page_folio(page), 0, PAGE_SIZE); } return gfs2_write_jdata_page(page, wbc); } diff --git a/fs/gfs2/aops.h b/fs/gfs2/aops.h index ff9877a68780..09db1914425e 100644 --- a/fs/gfs2/aops.h +++ b/fs/gfs2/aops.h @@ -9,7 +9,7 @@ #include "incore.h" extern void adjust_fs_space(struct inode *inode); -extern void gfs2_page_add_databufs(struct gfs2_inode *ip, struct page *page, - unsigned int from, unsigned int len); +extern void gfs2_trans_add_databufs(struct gfs2_inode *ip, struct folio *folio, + unsigned int from, unsigned int len); #endif /* __AOPS_DOT_H__ */ diff --git a/fs/gfs2/bmap.c b/fs/gfs2/bmap.c index e191ecfb1fde..eedf6926c652 100644 --- a/fs/gfs2/bmap.c +++ b/fs/gfs2/bmap.c @@ -985,8 +985,8 @@ static void gfs2_iomap_put_folio(struct inode *inode, loff_t pos, struct gfs2_sbd *sdp = GFS2_SB(inode); if (!gfs2_is_stuffed(ip)) - gfs2_page_add_databufs(ip, &folio->page, offset_in_page(pos), - copied); + gfs2_trans_add_databufs(ip, folio, offset_in_folio(folio, pos), + copied); folio_unlock(folio); folio_put(folio); diff --git a/fs/gfs2/dentry.c b/fs/gfs2/dentry.c index 2e215e8c3c88..6fe9ca253b70 100644 --- a/fs/gfs2/dentry.c +++ b/fs/gfs2/dentry.c @@ -83,26 +83,8 @@ static int gfs2_dhash(const struct dentry *dentry, struct qstr *str) return 0; } -static int gfs2_dentry_delete(const struct dentry *dentry) -{ - struct gfs2_inode *ginode; - - if (d_really_is_negative(dentry)) - return 0; - - ginode = GFS2_I(d_inode(dentry)); - if (!gfs2_holder_initialized(&ginode->i_iopen_gh)) - return 0; - - if (test_bit(GLF_DEMOTE, &ginode->i_iopen_gh.gh_gl->gl_flags)) - return 1; - - return 0; -} - const struct dentry_operations gfs2_dops = { .d_revalidate = gfs2_drevalidate, .d_hash = gfs2_dhash, - .d_delete = gfs2_dentry_delete, }; diff --git a/fs/gfs2/glock.c b/fs/gfs2/glock.c index 524f3c96b9a4..5adc7d85dbf3 100644 --- a/fs/gfs2/glock.c +++ b/fs/gfs2/glock.c @@ -67,7 +67,6 @@ static void handle_callback(struct gfs2_glock *gl, unsigned int state, 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); @@ -274,9 +273,8 @@ static void __gfs2_glock_put(struct gfs2_glock *gl) struct address_space *mapping = gfs2_glock2aspace(gl); lockref_mark_dead(&gl->gl_lockref); - - gfs2_glock_remove_from_lru(gl); 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); @@ -883,6 +881,7 @@ void glock_set_object(struct gfs2_glock *gl, void *object) /** * 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) { @@ -892,8 +891,7 @@ void glock_clear_object(struct gfs2_glock *gl, void *object) 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 || prev_object == NULL)) { + if (gfs2_assert_warn(gl->gl_name.ln_sbd, prev_object == object)) { pr_warn("glock=%u/%llx\n", gl->gl_name.ln_type, (unsigned long long)gl->gl_name.ln_number); @@ -977,6 +975,26 @@ static bool gfs2_try_evict(struct gfs2_glock *gl) return evicted; } +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 queue_delayed_work(sdp->sd_delete_wq, + &gl->gl_delete, 0); +} + +static bool gfs2_queue_verify_evict(struct gfs2_glock *gl) +{ + struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; + + if (test_and_set_bit(GLF_VERIFY_EVICT, &gl->gl_flags)) + return false; + return queue_delayed_work(sdp->sd_delete_wq, + &gl->gl_delete, 5 * HZ); +} + static void delete_work_func(struct work_struct *work) { struct delayed_work *dwork = to_delayed_work(work); @@ -985,11 +1003,7 @@ static void delete_work_func(struct work_struct *work) struct inode *inode; u64 no_addr = gl->gl_name.ln_number; - spin_lock(&gl->gl_lockref.lock); - clear_bit(GLF_PENDING_DELETE, &gl->gl_flags); - spin_unlock(&gl->gl_lockref.lock); - - if (test_bit(GLF_DEMOTE, &gl->gl_flags)) { + if (test_and_clear_bit(GLF_TRY_TO_EVICT, &gl->gl_flags)) { /* * If we can evict the inode, give the remote node trying to * delete the inode some time before verifying that the delete @@ -1008,22 +1022,28 @@ static void delete_work_func(struct work_struct *work) * step entirely. */ if (gfs2_try_evict(gl)) { - if (gfs2_queue_delete_work(gl, 5 * HZ)) + if (test_bit(SDF_DEACTIVATING, &sdp->sd_flags)) + goto out; + if (gfs2_queue_verify_evict(gl)) return; } goto out; } - 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 && - (gfs2_queue_delete_work(gl, 5 * HZ))) + if (test_and_clear_bit(GLF_VERIFY_EVICT, &gl->gl_flags)) { + 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_DEACTIVATING, &sdp->sd_flags) && + gfs2_queue_verify_evict(gl)) return; - } else { - d_prune_aliases(inode); - iput(inode); + } else { + d_prune_aliases(inode); + iput(inode); + } } + out: gfs2_glock_put(gl); } @@ -1985,26 +2005,26 @@ add_back_to_lru: static long gfs2_scan_glock_lru(int nr) { - struct gfs2_glock *gl; - LIST_HEAD(skipped); + struct gfs2_glock *gl, *next; LIST_HEAD(dispose); long freed = 0; spin_lock(&lru_lock); - while ((nr-- >= 0) && !list_empty(&lru_list)) { - gl = list_first_entry(&lru_list, struct gfs2_glock, gl_lru); - + list_for_each_entry_safe(gl, next, &lru_list, gl_lru) { + if (nr-- <= 0) + break; /* Test for being demotable */ if (!test_bit(GLF_LOCK, &gl->gl_flags)) { - list_move(&gl->gl_lru, &dispose); - atomic_dec(&lru_count); - freed++; - continue; + if (!spin_trylock(&gl->gl_lockref.lock)) + continue; + if (!gl->gl_lockref.count) { + list_move(&gl->gl_lru, &dispose); + atomic_dec(&lru_count); + freed++; + } + spin_unlock(&gl->gl_lockref.lock); } - - list_move(&gl->gl_lru, &skipped); } - list_splice(&skipped, &lru_list); if (!list_empty(&dispose)) gfs2_dispose_glock_lru(&dispose); spin_unlock(&lru_lock); @@ -2063,37 +2083,21 @@ static void glock_hash_walk(glock_examiner examiner, const struct gfs2_sbd *sdp) rhashtable_walk_exit(&iter); } -bool gfs2_queue_delete_work(struct gfs2_glock *gl, unsigned long delay) -{ - bool queued; - - spin_lock(&gl->gl_lockref.lock); - queued = queue_delayed_work(gfs2_delete_workqueue, - &gl->gl_delete, delay); - if (queued) - set_bit(GLF_PENDING_DELETE, &gl->gl_flags); - spin_unlock(&gl->gl_lockref.lock); - return queued; -} - void gfs2_cancel_delete_work(struct gfs2_glock *gl) { - if (cancel_delayed_work(&gl->gl_delete)) { - clear_bit(GLF_PENDING_DELETE, &gl->gl_flags); + clear_bit(GLF_TRY_TO_EVICT, &gl->gl_flags); + clear_bit(GLF_VERIFY_EVICT, &gl->gl_flags); + if (cancel_delayed_work(&gl->gl_delete)) gfs2_glock_put(gl); - } -} - -bool gfs2_delete_work_queued(const struct gfs2_glock *gl) -{ - return test_bit(GLF_PENDING_DELETE, &gl->gl_flags); } 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(gfs2_delete_workqueue, + queue_delayed_work(sdp->sd_delete_wq, &gl->gl_delete, 0); } } @@ -2102,7 +2106,7 @@ static void flush_delete_work(struct gfs2_glock *gl) void gfs2_flush_delete_work(struct gfs2_sbd *sdp) { glock_hash_walk(flush_delete_work, sdp); - flush_workqueue(gfs2_delete_workqueue); + flush_workqueue(sdp->sd_delete_wq); } /** @@ -2308,14 +2312,16 @@ 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_PENDING_DELETE, gflags)) - *p++ = 'P'; if (test_bit(GLF_FREEING, gflags)) *p++ = 'x'; 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_EVICT, gflags)) + *p++ = 'E'; *p = 0; return buf; } @@ -2465,18 +2471,9 @@ int __init gfs2_glock_init(void) rhashtable_destroy(&gl_hash_table); return -ENOMEM; } - gfs2_delete_workqueue = alloc_workqueue("delete_workqueue", - WQ_MEM_RECLAIM | WQ_FREEZABLE, - 0); - if (!gfs2_delete_workqueue) { - destroy_workqueue(glock_workqueue); - rhashtable_destroy(&gl_hash_table); - return -ENOMEM; - } ret = register_shrinker(&glock_shrinker, "gfs2-glock"); if (ret) { - destroy_workqueue(gfs2_delete_workqueue); destroy_workqueue(glock_workqueue); rhashtable_destroy(&gl_hash_table); return ret; @@ -2493,7 +2490,6 @@ void gfs2_glock_exit(void) unregister_shrinker(&glock_shrinker); rhashtable_destroy(&gl_hash_table); destroy_workqueue(glock_workqueue); - destroy_workqueue(gfs2_delete_workqueue); } static void gfs2_glock_iter_next(struct gfs2_glock_iter *gi, loff_t n) diff --git a/fs/gfs2/glock.h b/fs/gfs2/glock.h index f37ac087e2c1..1f1ba92c15a8 100644 --- a/fs/gfs2/glock.h +++ b/fs/gfs2/glock.h @@ -144,7 +144,6 @@ struct gfs2_glock_aspace { struct address_space mapping; }; -extern struct workqueue_struct *gfs2_delete_workqueue; static inline struct gfs2_holder *gfs2_glock_is_locked_by_me(struct gfs2_glock *gl) { struct gfs2_holder *gh; @@ -268,9 +267,8 @@ static inline int gfs2_glock_nq_init(struct gfs2_glock *gl, extern void gfs2_glock_cb(struct gfs2_glock *gl, unsigned int state); extern void gfs2_glock_complete(struct gfs2_glock *gl, int ret); -extern bool gfs2_queue_delete_work(struct gfs2_glock *gl, unsigned long delay); +extern bool gfs2_queue_try_to_evict(struct gfs2_glock *gl); extern void gfs2_cancel_delete_work(struct gfs2_glock *gl); -extern bool gfs2_delete_work_queued(const struct gfs2_glock *gl); extern void gfs2_flush_delete_work(struct gfs2_sbd *sdp); extern void gfs2_gl_hash_clear(struct gfs2_sbd *sdp); extern void gfs2_gl_dq_holders(struct gfs2_sbd *sdp); diff --git a/fs/gfs2/glops.c b/fs/gfs2/glops.c index d78b61ecc1cd..ad14818a790a 100644 --- a/fs/gfs2/glops.c +++ b/fs/gfs2/glops.c @@ -193,7 +193,7 @@ static int rgrp_go_sync(struct gfs2_glock *gl) struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); int error; - if (!test_and_clear_bit(GLF_DIRTY, &gl->gl_flags)) + if (!rgd || !test_and_clear_bit(GLF_DIRTY, &gl->gl_flags)) return 0; GLOCK_BUG_ON(gl, gl->gl_state != LM_ST_EXCLUSIVE); @@ -222,9 +222,12 @@ static void rgrp_go_inval(struct gfs2_glock *gl, int flags) struct address_space *mapping = &sdp->sd_aspace; struct gfs2_rgrpd *rgd = gfs2_glock2rgrp(gl); const unsigned bsize = sdp->sd_sb.sb_bsize; - loff_t start = (rgd->rd_addr * bsize) & PAGE_MASK; - loff_t end = PAGE_ALIGN((rgd->rd_addr + rgd->rd_length) * bsize) - 1; + loff_t start, end; + if (!rgd) + return; + start = (rgd->rd_addr * bsize) & PAGE_MASK; + end = PAGE_ALIGN((rgd->rd_addr + rgd->rd_length) * bsize) - 1; gfs2_rgrp_brelse(rgd); WARN_ON_ONCE(!(flags & DIO_METADATA)); truncate_inode_pages_range(mapping, start, end); @@ -645,23 +648,18 @@ static void iopen_go_callback(struct gfs2_glock *gl, bool remote) struct gfs2_inode *ip = gl->gl_object; struct gfs2_sbd *sdp = gl->gl_name.ln_sbd; - if (!remote || sb_rdonly(sdp->sd_vfs)) + if (!remote || sb_rdonly(sdp->sd_vfs) || + test_bit(SDF_DEACTIVATING, &sdp->sd_flags)) return; if (gl->gl_demote_state == LM_ST_UNLOCKED && gl->gl_state == LM_ST_SHARED && ip) { gl->gl_lockref.count++; - if (!queue_delayed_work(gfs2_delete_workqueue, - &gl->gl_delete, 0)) + if (!gfs2_queue_try_to_evict(gl)) gl->gl_lockref.count--; } } -static int iopen_go_demote_ok(const struct gfs2_glock *gl) -{ - return !gfs2_delete_work_queued(gl); -} - /** * inode_go_free - wake up anyone waiting for dlm's unlock ast to free it * @gl: glock being freed @@ -767,7 +765,6 @@ const struct gfs2_glock_operations gfs2_iopen_glops = { .go_type = LM_TYPE_IOPEN, .go_callback = iopen_go_callback, .go_dump = inode_go_dump, - .go_demote_ok = iopen_go_demote_ok, .go_flags = GLOF_LRU | GLOF_NONDISK, .go_subclass = 1, }; diff --git a/fs/gfs2/incore.h b/fs/gfs2/incore.h index c26765080f28..79485329118b 100644 --- a/fs/gfs2/incore.h +++ b/fs/gfs2/incore.h @@ -329,8 +329,9 @@ enum { GLF_LRU = 13, GLF_OBJECT = 14, /* Used only for tracing */ GLF_BLOCKING = 15, - GLF_PENDING_DELETE = 17, - GLF_FREEING = 18, /* Wait for glock to be freed */ + GLF_FREEING = 16, /* Wait for glock to be freed */ + GLF_TRY_TO_EVICT = 17, /* iopen glocks only */ + GLF_VERIFY_EVICT = 18, /* iopen glocks only */ }; struct gfs2_glock { @@ -605,6 +606,8 @@ enum { SDF_REMOTE_WITHDRAW = 13, /* Performing remote recovery */ SDF_WITHDRAW_RECOVERY = 14, /* Wait for journal recovery when we are withdrawing */ + SDF_DEACTIVATING = 15, + SDF_EVICTING = 16, }; enum gfs2_freeze_state { @@ -771,6 +774,10 @@ struct gfs2_sbd { struct completion sd_journal_ready; + /* Workqueue stuff */ + + struct workqueue_struct *sd_delete_wq; + /* Daemon stuff */ struct task_struct *sd_logd_process; diff --git a/fs/gfs2/inode.c b/fs/gfs2/inode.c index 713efa3bb732..1291b5ee3584 100644 --- a/fs/gfs2/inode.c +++ b/fs/gfs2/inode.c @@ -225,6 +225,10 @@ fail: gfs2_glock_dq_uninit(&ip->i_iopen_gh); if (gfs2_holder_initialized(&i_gh)) gfs2_glock_dq_uninit(&i_gh); + if (ip->i_gl) { + gfs2_glock_put(ip->i_gl); + ip->i_gl = NULL; + } iget_failed(inode); return ERR_PTR(error); } @@ -816,6 +820,10 @@ fail_gunlock3: fail_gunlock2: gfs2_glock_put(io_gl); fail_free_inode: + if (ip->i_gl) { + gfs2_glock_put(ip->i_gl); + ip->i_gl = NULL; + } gfs2_rs_deltree(&ip->i_res); gfs2_qa_put(ip); fail_free_acls: diff --git a/fs/gfs2/ops_fstype.c b/fs/gfs2/ops_fstype.c index c0cf1d2d0ef5..6de901c3b89b 100644 --- a/fs/gfs2/ops_fstype.c +++ b/fs/gfs2/ops_fstype.c @@ -1197,9 +1197,15 @@ static int gfs2_fill_super(struct super_block *sb, struct fs_context *fc) snprintf(sdp->sd_fsname, sizeof(sdp->sd_fsname), "%s", sdp->sd_table_name); + sdp->sd_delete_wq = alloc_workqueue("gfs2-delete/%s", + WQ_MEM_RECLAIM | WQ_FREEZABLE, 0, sdp->sd_fsname); + error = -ENOMEM; + if (!sdp->sd_delete_wq) + goto fail_free; + error = gfs2_sys_fs_add(sdp); if (error) - goto fail_free; + goto fail_delete_wq; gfs2_create_debugfs_file(sdp); @@ -1309,6 +1315,8 @@ fail_lm: fail_debug: gfs2_delete_debugfs_file(sdp); gfs2_sys_fs_del(sdp); +fail_delete_wq: + destroy_workqueue(sdp->sd_delete_wq); fail_free: free_sbd(sdp); sb->s_fs_info = NULL; @@ -1720,6 +1728,55 @@ static int gfs2_meta_init_fs_context(struct fs_context *fc) return 0; } +/** + * gfs2_evict_inodes - evict inodes cooperatively + * @sb: the superblock + * + * When evicting an inode with a zero link count, we are trying to upgrade the + * inode's iopen glock from SH to EX mode in order to determine if we can + * delete the inode. The other nodes are supposed to evict the inode from + * their caches if they can, and to poke the inode's inode glock if they cannot + * do so. Either behavior allows gfs2_upgrade_iopen_glock() to proceed + * quickly, but if the other nodes are not cooperating, the lock upgrading + * attempt will time out. Since inodes are evicted sequentially, this can add + * up quickly. + * + * Function evict_inodes() tries to keep the s_inode_list_lock list locked over + * a long time, which prevents other inodes from being evicted concurrently. + * This precludes the cooperative behavior we are looking for. This special + * version of evict_inodes() avoids that. + * + * Modeled after drop_pagecache_sb(). + */ +static void gfs2_evict_inodes(struct super_block *sb) +{ + struct inode *inode, *toput_inode = NULL; + struct gfs2_sbd *sdp = sb->s_fs_info; + + set_bit(SDF_EVICTING, &sdp->sd_flags); + + spin_lock(&sb->s_inode_list_lock); + list_for_each_entry(inode, &sb->s_inodes, i_sb_list) { + spin_lock(&inode->i_lock); + if ((inode->i_state & (I_FREEING|I_WILL_FREE|I_NEW)) && + !need_resched()) { + spin_unlock(&inode->i_lock); + continue; + } + atomic_inc(&inode->i_count); + spin_unlock(&inode->i_lock); + spin_unlock(&sb->s_inode_list_lock); + + iput(toput_inode); + toput_inode = inode; + + cond_resched(); + spin_lock(&sb->s_inode_list_lock); + } + spin_unlock(&sb->s_inode_list_lock); + iput(toput_inode); +} + static void gfs2_kill_sb(struct super_block *sb) { struct gfs2_sbd *sdp = sb->s_fs_info; @@ -1735,6 +1792,18 @@ static void gfs2_kill_sb(struct super_block *sb) sdp->sd_root_dir = NULL; sdp->sd_master_dir = NULL; shrink_dcache_sb(sb); + + gfs2_evict_inodes(sb); + + /* + * Flush and then drain the delete workqueue here (via + * destroy_workqueue()) to ensure that any delete work that + * may be running will also see the SDF_DEACTIVATING flag. + */ + set_bit(SDF_DEACTIVATING, &sdp->sd_flags); + gfs2_flush_delete_work(sdp); + destroy_workqueue(sdp->sd_delete_wq); + kill_block_super(sb); } diff --git a/fs/gfs2/rgrp.c b/fs/gfs2/rgrp.c index f602fb844951..3b9b76e980ad 100644 --- a/fs/gfs2/rgrp.c +++ b/fs/gfs2/rgrp.c @@ -1879,7 +1879,7 @@ static void try_rgrp_unlink(struct gfs2_rgrpd *rgd, u64 *last_unlinked, u64 skip */ ip = gl->gl_object; - if (ip || !gfs2_queue_delete_work(gl, 0)) + if (ip || !gfs2_queue_try_to_evict(gl)) gfs2_glock_put(gl); else found++; diff --git a/fs/gfs2/super.c b/fs/gfs2/super.c index 999cc146d708..a83fa62106f0 100644 --- a/fs/gfs2/super.c +++ b/fs/gfs2/super.c @@ -138,8 +138,10 @@ int gfs2_make_fs_rw(struct gfs2_sbd *sdp) return -EIO; error = gfs2_find_jhead(sdp->sd_jdesc, &head, false); - if (error || gfs2_withdrawn(sdp)) + if (error) { + gfs2_consist(sdp); return error; + } if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) { gfs2_consist(sdp); @@ -151,7 +153,9 @@ int gfs2_make_fs_rw(struct gfs2_sbd *sdp) gfs2_log_pointers_init(sdp, head.lh_blkno); error = gfs2_quota_init(sdp); - if (!error && !gfs2_withdrawn(sdp)) + if (!error && gfs2_withdrawn(sdp)) + error = -EIO; + if (!error) set_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags); return error; } @@ -529,7 +533,9 @@ void gfs2_make_fs_ro(struct gfs2_sbd *sdp) { int log_write_allowed = test_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags); - gfs2_flush_delete_work(sdp); + if (!test_bit(SDF_DEACTIVATING, &sdp->sd_flags)) + gfs2_flush_delete_work(sdp); + if (!log_write_allowed && current == sdp->sd_quotad_process) fs_warn(sdp, "The quotad daemon is withdrawing.\n"); else if (sdp->sd_quotad_process) @@ -933,6 +939,7 @@ static int gfs2_statfs(struct dentry *dentry, struct kstatfs *buf) static int gfs2_drop_inode(struct inode *inode) { struct gfs2_inode *ip = GFS2_I(inode); + struct gfs2_sbd *sdp = GFS2_SB(inode); if (inode->i_nlink && gfs2_holder_initialized(&ip->i_iopen_gh)) { @@ -952,11 +959,17 @@ static int gfs2_drop_inode(struct inode *inode) struct gfs2_glock *gl = ip->i_iopen_gh.gh_gl; gfs2_glock_hold(gl); - if (!gfs2_queue_delete_work(gl, 0)) + if (!gfs2_queue_try_to_evict(gl)) gfs2_glock_queue_put(gl); return 0; } + /* + * No longer cache inodes when trying to evict them all. + */ + if (test_bit(SDF_EVICTING, &sdp->sd_flags)) + return 1; + return generic_drop_inode(inode); } @@ -1175,15 +1188,23 @@ static bool gfs2_upgrade_iopen_glock(struct inode *inode) gfs2_glock_dq_wait(gh); /* - * If there are no other lock holders, we'll get the lock immediately. + * If there are no other lock holders, we will immediately get + * exclusive access to the iopen glock here. + * * Otherwise, the other nodes holding the lock will be notified about - * our locking request. If they don't have the inode open, they'll - * evict the cached inode and release the lock. Otherwise, if they - * poke the inode glock, we'll take this as an indication that they - * still need the iopen glock and that they'll take care of deleting - * the inode when they're done. As a last resort, if another node - * keeps holding the iopen glock without showing any activity on the - * inode glock, we'll eventually time out. + * our locking request. If they do not have the inode open, they are + * expected to evict the cached inode and release the lock, allowing us + * to proceed. + * + * Otherwise, if they cannot evict the inode, they are expected to poke + * the inode glock (note: not the iopen glock). We will notice that + * and stop waiting for the iopen glock immediately. The other node(s) + * are then expected to take care of deleting the inode when they no + * longer use it. + * + * As a last resort, if another node keeps holding the iopen glock + * without showing any activity on the inode glock, we will eventually + * time out and fail the iopen glock upgrade. * * Note that we're passing the LM_FLAG_TRY_1CB flag to the first * locking request as an optimization to notify lock holders as soon as @@ -1401,10 +1422,8 @@ static void gfs2_evict_inode(struct inode *inode) if (gfs2_rs_active(&ip->i_res)) gfs2_rs_deltree(&ip->i_res); - if (gfs2_holder_initialized(&gh)) { - glock_clear_object(ip->i_gl, ip); + if (gfs2_holder_initialized(&gh)) gfs2_glock_dq_uninit(&gh); - } if (ret && ret != GLR_TRYFAILED && ret != -EROFS) fs_warn(sdp, "gfs2_evict_inode: %d\n", ret); out: diff --git a/fs/gfs2/sys.c b/fs/gfs2/sys.c index d87ea98cf535..c40118ea4bbc 100644 --- a/fs/gfs2/sys.c +++ b/fs/gfs2/sys.c @@ -87,6 +87,7 @@ static ssize_t status_show(struct gfs2_sbd *sdp, char *buf) "Withdraw In Prog: %d\n" "Remote Withdraw: %d\n" "Withdraw Recovery: %d\n" + "Deactivating: %d\n" "sd_log_error: %d\n" "sd_log_flush_lock: %d\n" "sd_log_num_revoke: %u\n" @@ -115,6 +116,7 @@ static ssize_t status_show(struct gfs2_sbd *sdp, char *buf) test_bit(SDF_WITHDRAW_IN_PROG, &f), test_bit(SDF_REMOTE_WITHDRAW, &f), test_bit(SDF_WITHDRAW_RECOVERY, &f), + test_bit(SDF_DEACTIVATING, &f), sdp->sd_log_error, rwsem_is_locked(&sdp->sd_log_flush_lock), sdp->sd_log_num_revoke, |