Merge tag 'gfs2-4.14.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2

Pull GFS2 updates from Bob Peterson:
 "We've got a whopping 29 GFS2 patches for this merge window, mainly
  because we held some back from the previous merge window until we
  could get them perfected and well tested. We have a couple patch sets,
  including my patch set for protecting glock gl_object and Andreas
  Gruenbacher's patch set to fix the long-standing shrink- slab hang,
  plus a bunch of assorted bugs and cleanups.

  Summary:

   - I fixed a bug whereby an IO error would lead to a double-brelse.

   - Andreas Gruenbacher made a minor cleanup to call his relatively new
     function, gfs2_holder_initialized, rather than doing it manually.
     This was just missed by a previous patch set.

   - Jan Kara fixed a bug whereby the SGID was being cleared when
     inheriting ACLs.

   - Andreas found a bug and fixed it in his previous patch, "Get rid of
     flush_delayed_work in gfs2_evict_inode". A call to
     flush_delayed_work was deleted from *gfs2_inode_lookup and added to
     gfs2_create_inode.

   - Wang Xibo found and fixed a list_add call in inode_go_lock that
     specified the parameters in the wrong order.

   - Coly Li submitted a patch to add the REQ_PRIO to some of GFS2's
     metadata reads that were accidentally missing them.

   - I submitted a 4-patch set to protect the glock gl_object field.
     GFS2 was setting and checking gl_object with no locking mechanism,
     so the value was occasionally stomped on, which caused file system
     corruption.

   - I submitted a small cleanup to function gfs2_clear_rgrpd. It was
     needlessly adding rgrp glocks to the lru list, then pulling them
     back off immediately. The rgrp glocks don't use the lru list
     anyway, so doing so was just a waste of time.

   - I submitted a patch that checks the GLOF_LRU flag on a glock before
     trying to remove it from the lru_list. This avoids a lot of
     unnecessary spin_lock contention.

   - I submitted a patch to delete GFS2's debugfs files only after we
     evict all the glocks. Before this patch, GFS2 would delete the
     debugfs files, and if unmount hung waiting for a glock, there was
     no way to debug the problem. Now, if a hang occurs during umount,
     we can examine the debugfs files to figure out why it's hung.

   - Andreas Gruenbacher submitted a patch to fix some trivial typos.

   - Andreas also submitted a five-part patch set to fix the
     longstanding hang involving the slab shrinker: dlm requires memory,
     calls the inode shrinker, which calls gfs2's evict, which calls
     back into DLM before it can evict an inode.

   - Abhi Das submitted a patch to forcibly flush the active items list
     to relieve memory pressure. This fixes a long-standing bug whereby
     GFS2 was getting hung permanently in balance_dirty_pages.

   - Thomas Tai submitted a patch to fix a slab corruption problem due
     to a residual pointer left in the lock_dlm lockstruct.

   - I submitted a patch to withdraw the file system if IO errors are
     encountered while writing to the journals or statfs system file
     which were previously not being sent back up. Before, some IO
     errors were sometimes not be detected for several hours, and at
     recovery time, the journal errors made journal replay impossible.

   - Andreas has a patch to fix an annoying format-truncation compiler
     warning so GFS2 compiles cleanly.

   - I have a patch that fixes a handful of sparse compiler warnings.

   - Andreas fixed up an useless gl_object warning caused by an earlier
     patch.

   - Arvind Yadav added a patch to properly constify our rhashtable
     params declare.

   - I added a patch to fix a regression caused by the non-recursive
     delete and truncate patch that caused file system blocks to not be
     properly freed.

   - Ernesto A. Fernández added a patch to fix a place where GFS2 would
     send back the wrong return code setting extended attributes.

   - Ernesto also added a patch to fix a case in which GFS2 was
     improperly setting an inode's i_mode, potentially granting access
     to the wrong users"

* tag 'gfs2-4.14.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/gfs2/linux-gfs2: (29 commits)
  gfs2: preserve i_mode if __gfs2_set_acl() fails
  gfs2: don't return ENODATA in __gfs2_xattr_set unless replacing
  GFS2: Fix non-recursive truncate bug
  gfs2: constify rhashtable_params
  GFS2: Fix gl_object warnings
  GFS2: Fix up some sparse warnings
  gfs2: Silence gcc format-truncation warning
  GFS2: Withdraw for IO errors writing to the journal or statfs
  gfs2: fix slab corruption during mounting and umounting gfs file system
  gfs2: forcibly flush ail to relieve memory pressure
  gfs2: Clean up waiting on glocks
  gfs2: Defer deleting inodes under memory pressure
  gfs2: gfs2_evict_inode: Put glocks asynchronously
  gfs2: Get rid of gfs2_set_nlink
  gfs2: gfs2_glock_get: Wait on freeing glocks
  gfs2: Fix trivial typos
  GFS2: Delete debugfs files only after we evict the glocks
  GFS2: Don't waste time locking lru_lock for non-lru glocks
  GFS2: Don't bother trying to add rgrps to the lru list
  GFS2: Clear gl_object when deleting an inode in gfs2_delete_inode
  ...

fs/gfs2/acl.c

@@ -86,19 +86,6 @@ int __gfs2_set_acl(struct inode *inode, struct posix_acl *acl, int type)
 	char *data;
 	const char *name = gfs2_acl_name(type);
 
-	if (acl && acl->a_count > GFS2_ACL_MAX_ENTRIES(GFS2_SB(inode)))
-		return -E2BIG;
-
-	if (type == ACL_TYPE_ACCESS) {
-		umode_t mode = inode->i_mode;
-
-		error = posix_acl_update_mode(inode, &inode->i_mode, &acl);
-		if (error)
-			return error;
-		if (mode != inode->i_mode)
-			mark_inode_dirty(inode);
-	}
-
 	if (acl) {
 		len = posix_acl_to_xattr(&init_user_ns, acl, NULL, 0);
 		if (len == 0)
@@ -129,6 +116,10 @@ int gfs2_set_acl(struct inode *inode, struct posix_acl *acl, int type)
 	struct gfs2_holder gh;
 	bool need_unlock = false;
 	int ret;
+	umode_t mode;
+
+	if (acl && acl->a_count > GFS2_ACL_MAX_ENTRIES(GFS2_SB(inode)))
+		return -E2BIG;
 
 	ret = gfs2_rsqa_alloc(ip);
 	if (ret)
@@ -140,7 +131,20 @@ int gfs2_set_acl(struct inode *inode, struct posix_acl *acl, int type)
 			return ret;
 		need_unlock = true;
 	}
+
+	mode = inode->i_mode;
+	if (type == ACL_TYPE_ACCESS && acl) {
+		ret = posix_acl_update_mode(inode, &mode, &acl);
+		if (ret)
+			goto unlock;
+	}
+
 	ret = __gfs2_set_acl(inode, acl, type);
+	if (!ret && mode != inode->i_mode) {
+		inode->i_mode = mode;
+		mark_inode_dirty(inode);
+	}
+unlock:
 	if (need_unlock)
 		gfs2_glock_dq_uninit(&gh);
 	return ret;

fs/gfs2/aops.c

@@ -234,7 +234,19 @@ static int gfs2_jdata_writepage(struct page *page, struct writeback_control *wbc
 static int gfs2_writepages(struct address_space *mapping,
 			   struct writeback_control *wbc)
 {
-	return mpage_writepages(mapping, wbc, gfs2_get_block_noalloc);
+	struct gfs2_sbd *sdp = gfs2_mapping2sbd(mapping);
+	int ret = mpage_writepages(mapping, wbc, gfs2_get_block_noalloc);
+
+	/*
+	 * Even if we didn't write any pages here, we might still be holding
+	 * dirty pages in the ail. We forcibly flush the ail because we don't
+	 * want balance_dirty_pages() to loop indefinitely trying to write out
+	 * pages held in the ail that it can't find.
+	 */
+	if (ret == 0)
+		set_bit(SDF_FORCE_AIL_FLUSH, &sdp->sd_flags);
+
+	return ret;
 }
 
 /**

fs/gfs2/bmap.c

@@ -291,8 +291,9 @@ static void gfs2_metapath_ra(struct gfs2_glock *gl,
 		if (trylock_buffer(rabh)) {
 			if (!buffer_uptodate(rabh)) {
 				rabh->b_end_io = end_buffer_read_sync;
-				submit_bh(REQ_OP_READ, REQ_RAHEAD | REQ_META,
-						rabh);
+				submit_bh(REQ_OP_READ,
+					  REQ_RAHEAD | REQ_META | REQ_PRIO,
+					  rabh);
 				continue;
 			}
 			unlock_buffer(rabh);
@@ -1103,8 +1104,15 @@ static bool find_nonnull_ptr(struct gfs2_sbd *sdp, struct metapath *mp,
 
 	while (true) {
 		ptr = metapointer(h, mp);
-		if (*ptr) /* if we have a non-null pointer */
+		if (*ptr) { /* if we have a non-null pointer */
+			/* Now zero the metapath after the current height. */
+			h++;
+			if (h < GFS2_MAX_META_HEIGHT)
+				memset(&mp->mp_list[h], 0,
+				       (GFS2_MAX_META_HEIGHT - h) *
+				       sizeof(mp->mp_list[0]));
 			return true;
+		}
 
 		if (mp->mp_list[h] < ptrs)
 			mp->mp_list[h]++;
@@ -1120,6 +1128,13 @@ enum dealloc_states {
 	DEALLOC_DONE = 3,       /* process complete */
 };
 
+static bool mp_eq_to_hgt(struct metapath *mp, __u16 *nbof, unsigned int h)
+{
+	if (memcmp(mp->mp_list, nbof, h * sizeof(mp->mp_list[0])))
+		return false;
+	return true;
+}
+
 /**
  * trunc_dealloc - truncate a file down to a desired size
  * @ip: inode to truncate
@@ -1197,8 +1212,7 @@ static int trunc_dealloc(struct gfs2_inode *ip, u64 newsize)
 			/* If we're truncating to a non-zero size and the mp is
 			   at the beginning of file for the strip height, we
 			   need to preserve the first metadata pointer. */
-			preserve1 = (newsize &&
-				     (mp.mp_list[mp_h] == nbof[mp_h]));
+			preserve1 = (newsize && mp_eq_to_hgt(&mp, nbof, mp_h));
 			bh = mp.mp_bh[mp_h];
 			gfs2_assert_withdraw(sdp, bh);
 			if (gfs2_assert_withdraw(sdp,

fs/gfs2/dir.c

@@ -1513,7 +1513,9 @@ static void gfs2_dir_readahead(struct inode *inode, unsigned hsize, u32 index,
 				continue;
 			}
 			bh->b_end_io = end_buffer_read_sync;
-			submit_bh(REQ_OP_READ, REQ_RAHEAD | REQ_META, bh);
+			submit_bh(REQ_OP_READ,
+				  REQ_RAHEAD | REQ_META | REQ_PRIO,
+				  bh);
 			continue;
 		}
 		brelse(bh);

fs/gfs2/file.c

@@ -1030,8 +1030,7 @@ static int do_flock(struct file *file, int cmd, struct file_lock *fl)
 
 	mutex_lock(&fp->f_fl_mutex);
 
-	gl = fl_gh->gh_gl;
-	if (gl) {
+	if (gfs2_holder_initialized(fl_gh)) {
 		if (fl_gh->gh_state == state)
 			goto out;
 		locks_lock_file_wait(file,

fs/gfs2/glock.c

@@ -15,6 +15,7 @@
 #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>
@@ -71,7 +72,7 @@ static DEFINE_SPINLOCK(lru_lock);
 #define GFS2_GL_HASH_SHIFT      15
 #define GFS2_GL_HASH_SIZE       BIT(GFS2_GL_HASH_SHIFT)
 
-static struct rhashtable_params ht_parms = {
+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),
@@ -80,6 +81,49 @@ static struct rhashtable_params ht_parms = {
 
 static struct rhashtable gl_hash_table;
 
+#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;
+
+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;
+
+	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 wait_queue_head_t *glock_waitqueue(struct lm_lockname *name)
+{
+	u32 hash = jhash2((u32 *)name, sizeof(*name) / 4, 0);
+
+	return glock_wait_table + hash_32(hash, GLOCK_WAIT_TABLE_BITS);
+}
+
+/**
+ * wake_up_glock  -  Wake up waiters on a glock
+ * @gl: the glock
+ */
+static void wake_up_glock(struct gfs2_glock *gl)
+{
+	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);
@@ -96,6 +140,9 @@ void gfs2_glock_free(struct gfs2_glock *gl)
 {
 	struct gfs2_sbd *sdp = gl->gl_name.ln_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);
 	if (atomic_dec_and_test(&sdp->sd_glock_disposal))
 		wake_up(&sdp->sd_glock_wait);
@@ -107,7 +154,7 @@ void gfs2_glock_free(struct gfs2_glock *gl)
  *
  */
 
-static void gfs2_glock_hold(struct gfs2_glock *gl)
+void gfs2_glock_hold(struct gfs2_glock *gl)
 {
 	GLOCK_BUG_ON(gl, __lockref_is_dead(&gl->gl_lockref));
 	lockref_get(&gl->gl_lockref);
@@ -150,6 +197,9 @@ void gfs2_glock_add_to_lru(struct gfs2_glock *gl)
 
 static void gfs2_glock_remove_from_lru(struct gfs2_glock *gl)
 {
+	if (!(gl->gl_ops->go_flags & GLOF_LRU))
+		return;
+
 	spin_lock(&lru_lock);
 	if (!list_empty(&gl->gl_lru)) {
 		list_del_init(&gl->gl_lru);
@@ -191,13 +241,20 @@ static void __gfs2_glock_put(struct gfs2_glock *gl)
 
 	gfs2_glock_remove_from_lru(gl);
 	spin_unlock(&gl->gl_lockref.lock);
-	rhashtable_remove_fast(&gl_hash_table, &gl->gl_node, ht_parms);
 	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);
 }
 
+/*
+ * Cause the glock to be put in work queue context.
+ */
+void gfs2_glock_queue_put(struct gfs2_glock *gl)
+{
+	gfs2_glock_queue_work(gl, 0);
+}
+
 /**
  * gfs2_glock_put() - Decrement reference count on glock
  * @gl: The glock to put
@@ -676,6 +733,40 @@ static void glock_work_func(struct work_struct *work)
 	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);
+	return gl;
+}
+
 /**
  * gfs2_glock_get() - Get a glock, or create one if one doesn't exist
  * @sdp: The GFS2 superblock
@@ -702,15 +793,11 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 	struct kmem_cache *cachep;
 	int ret = 0;
 
-	rcu_read_lock();
-	gl = rhashtable_lookup_fast(&gl_hash_table, &name, ht_parms);
-	if (gl && !lockref_get_not_dead(&gl->gl_lockref))
-		gl = NULL;
-	rcu_read_unlock();
-
-	*glp = gl;
-	if (gl)
+	gl = find_insert_glock(&name, NULL);
+	if (gl) {
+		*glp = gl;
 		return 0;
+	}
 	if (!create)
 		return -ENOENT;
 
@@ -764,10 +851,7 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 		mapping->writeback_index = 0;
 	}
 
-again:
-	rcu_read_lock();
-	tmp = rhashtable_lookup_get_insert_fast(&gl_hash_table, &gl->gl_node,
-						ht_parms);
+	tmp = find_insert_glock(&name, gl);
 	if (!tmp) {
 		*glp = gl;
 		goto out;
@@ -776,13 +860,7 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 		ret = PTR_ERR(tmp);
 		goto out_free;
 	}
-	if (lockref_get_not_dead(&tmp->gl_lockref)) {
-		*glp = tmp;
-		goto out_free;
-	}
-	rcu_read_unlock();
-	cond_resched();
-	goto again;
+	*glp = tmp;
 
 out_free:
 	kfree(gl->gl_lksb.sb_lvbptr);
@@ -790,7 +868,6 @@ int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 	atomic_dec(&sdp->sd_glock_disposal);
 
 out:
-	rcu_read_unlock();
 	return ret;
 }
 
@@ -1473,14 +1550,15 @@ static void glock_hash_walk(glock_examiner examiner, const struct gfs2_sbd *sdp)
 
 	do {
 		gl = ERR_PTR(rhashtable_walk_start(&iter));
-		if (gl)
-			continue;
+		if (IS_ERR(gl))
+			goto walk_stop;
 
 		while ((gl = rhashtable_walk_next(&iter)) && !IS_ERR(gl))
-			if ((gl->gl_name.ln_sbd == sdp) &&
+			if (gl->gl_name.ln_sbd == sdp &&
 			    lockref_get_not_dead(&gl->gl_lockref))
 				examiner(gl);
 
+walk_stop:
 		rhashtable_walk_stop(&iter);
 	} while (cond_resched(), gl == ERR_PTR(-EAGAIN));
 
@@ -1803,7 +1881,7 @@ static int gfs2_sbstats_seq_show(struct seq_file *seq, void *iter_ptr)
 
 int __init gfs2_glock_init(void)
 {
-	int ret;
+	int i, ret;
 
 	ret = rhashtable_init(&gl_hash_table, &ht_parms);
 	if (ret < 0)
@@ -1832,6 +1910,9 @@ int __init gfs2_glock_init(void)
 		return ret;
 	}
 
+	for (i = 0; i < GLOCK_WAIT_TABLE_SIZE; i++)
+		init_waitqueue_head(glock_wait_table + i);
+
 	return 0;
 }
 
@@ -1860,6 +1941,7 @@ static void gfs2_glock_iter_next(struct gfs2_glock_iter *gi)
 }
 
 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;
@@ -1892,6 +1974,7 @@ static void *gfs2_glock_seq_next(struct seq_file *seq, void *iter_ptr,
 }
 
 static void gfs2_glock_seq_stop(struct seq_file *seq, void *iter_ptr)
+	__releases(RCU)
 {
 	struct gfs2_glock_iter *gi = seq->private;
 

fs/gfs2/glock.h

@@ -13,6 +13,7 @@
 #include <linux/sched.h>
 #include <linux/parser.h>
 #include "incore.h"
+#include "util.h"
 
 /* Options for hostdata parser */
 
@@ -181,7 +182,9 @@ static inline struct address_space *gfs2_glock2aspace(struct gfs2_glock *gl)
 extern int gfs2_glock_get(struct gfs2_sbd *sdp, u64 number,
 			  const struct gfs2_glock_operations *glops,
 			  int create, struct gfs2_glock **glp);
+extern void gfs2_glock_hold(struct gfs2_glock *gl);
 extern void gfs2_glock_put(struct gfs2_glock *gl);
+extern void gfs2_glock_queue_put(struct gfs2_glock *gl);
 extern void gfs2_holder_init(struct gfs2_glock *gl, unsigned int state,
 			     u16 flags, struct gfs2_holder *gh);
 extern void gfs2_holder_reinit(unsigned int state, u16 flags,
@@ -257,11 +260,44 @@ static inline bool gfs2_holder_initialized(struct gfs2_holder *gh)
 	return gh->gh_gl;
 }
 
+/**
+ * glock_set_object - set the gl_object field of a glock
+ * @gl: the glock
+ * @object: the object
+ */
 static inline void glock_set_object(struct gfs2_glock *gl, void *object)
 {
 	spin_lock(&gl->gl_lockref.lock);
+	if (gfs2_assert_warn(gl->gl_name.ln_sbd, gl->gl_object == NULL))
+		gfs2_dump_glock(NULL, gl);
 	gl->gl_object = object;
 	spin_unlock(&gl->gl_lockref.lock);
 }
 
+/**
+ * glock_clear_object - clear the gl_object field of a glock
+ * @gl: the glock
+ * @object: the object
+ *
+ * I'd love to similarly add this:
+ *	else if (gfs2_assert_warn(gl->gl_sbd, gl->gl_object == object))
+ *		gfs2_dump_glock(NULL, gl);
+ * Unfortunately, that's not possible because as soon as gfs2_delete_inode
+ * frees the block in the rgrp, another process can reassign it for an I_NEW
+ * inode in gfs2_create_inode because that calls new_inode, not gfs2_iget.
+ * That means gfs2_delete_inode may subsequently try to call this function
+ * for a glock that's already pointing to a brand new inode. If we clear the
+ * new inode's gl_object, we'll introduce metadata corruption. Function
+ * gfs2_delete_inode calls clear_inode which calls gfs2_clear_inode which also
+ * tries to clear gl_object, so it's more than just gfs2_delete_inode.
+ *
+ */
+static inline void glock_clear_object(struct gfs2_glock *gl, void *object)
+{
+	spin_lock(&gl->gl_lockref.lock);
+	if (gl->gl_object == object)
+		gl->gl_object = NULL;
+	spin_unlock(&gl->gl_lockref.lock);
+}
+
 #endif /* __GLOCK_DOT_H__ */

fs/gfs2/glops.c

@@ -329,32 +329,6 @@ static int inode_go_demote_ok(const struct gfs2_glock *gl)
 	return 1;
 }
 
-/**
- * gfs2_set_nlink - Set the inode's link count based on on-disk info
- * @inode: The inode in question
- * @nlink: The link count
- *
- * If the link count has hit zero, it must never be raised, whatever the
- * on-disk inode might say. When new struct inodes are created the link
- * count is set to 1, so that we can safely use this test even when reading
- * in on disk information for the first time.
- */
-
-static void gfs2_set_nlink(struct inode *inode, u32 nlink)
-{
-	/*
-	 * We will need to review setting the nlink count here in the
-	 * light of the forthcoming ro bind mount work. This is a reminder
-	 * to do that.
-	 */
-	if ((inode->i_nlink != nlink) && (inode->i_nlink != 0)) {
-		if (nlink == 0)
-			clear_nlink(inode);
-		else
-			set_nlink(inode, nlink);
-	}
-}
-
 static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 {
 	const struct gfs2_dinode *str = buf;
@@ -376,7 +350,7 @@ static int gfs2_dinode_in(struct gfs2_inode *ip, const void *buf)
 
 	i_uid_write(&ip->i_inode, be32_to_cpu(str->di_uid));
 	i_gid_write(&ip->i_inode, be32_to_cpu(str->di_gid));
-	gfs2_set_nlink(&ip->i_inode, be32_to_cpu(str->di_nlink));
+	set_nlink(&ip->i_inode, be32_to_cpu(str->di_nlink));
 	i_size_write(&ip->i_inode, be64_to_cpu(str->di_size));
 	gfs2_set_inode_blocks(&ip->i_inode, be64_to_cpu(str->di_blocks));
 	atime.tv_sec = be64_to_cpu(str->di_atime);
@@ -470,7 +444,7 @@ static int inode_go_lock(struct gfs2_holder *gh)
 	    (gh->gh_state == LM_ST_EXCLUSIVE)) {
 		spin_lock(&sdp->sd_trunc_lock);
 		if (list_empty(&ip->i_trunc_list))
-			list_add(&sdp->sd_trunc_list, &ip->i_trunc_list);
+			list_add(&ip->i_trunc_list, &sdp->sd_trunc_list);
 		spin_unlock(&sdp->sd_trunc_lock);
 		wake_up(&sdp->sd_quota_wait);
 		return 1;

fs/gfs2/incore.h

@@ -606,6 +606,7 @@ enum {
 	SDF_NOJOURNALID		= 6,
 	SDF_RORECOVERY		= 7, /* read only recovery */
 	SDF_SKIP_DLM_UNLOCK	= 8,
+	SDF_FORCE_AIL_FLUSH     = 9,
 };
 
 enum gfs2_freeze_state {
@@ -816,6 +817,7 @@ struct gfs2_sbd {
 	atomic_t sd_log_in_flight;
 	struct bio *sd_log_bio;
 	wait_queue_head_t sd_log_flush_wait;
+	int sd_log_error;
 
 	atomic_t sd_reserving_log;
 	wait_queue_head_t sd_reserving_log_wait;
@@ -831,7 +833,7 @@ struct gfs2_sbd {
 	atomic_t sd_freeze_state;
 	struct mutex sd_freeze_mutex;
 
-	char sd_fsname[GFS2_FSNAME_LEN];
+	char sd_fsname[GFS2_FSNAME_LEN + 3 * sizeof(int) + 2];
 	char sd_table_name[GFS2_FSNAME_LEN];
 	char sd_proto_name[GFS2_FSNAME_LEN];
 

fs/gfs2/inode.c

@@ -109,7 +109,7 @@ static void gfs2_set_iop(struct inode *inode)
  * @no_addr: The inode number
  * @no_formal_ino: The inode generation number
  * @blktype: Requested block type (GFS2_BLKST_DINODE or GFS2_BLKST_UNLINKED;
- *           GFS2_BLKST_FREE do indicate not to verify)
+ *           GFS2_BLKST_FREE to indicate not to verify)
  *
  * If @type is DT_UNKNOWN, the inode type is fetched from disk.
  *
@@ -145,7 +145,6 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 		if (unlikely(error))
 			goto fail;
 		flush_delayed_work(&ip->i_gl->gl_work);
-		glock_set_object(ip->i_gl, ip);
 
 		error = gfs2_glock_get(sdp, no_addr, &gfs2_iopen_glops, CREATE, &io_gl);
 		if (unlikely(error))
@@ -170,11 +169,11 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 			}
 		}
 
+		glock_set_object(ip->i_gl, ip);
 		set_bit(GIF_INVALID, &ip->i_flags);
 		error = gfs2_glock_nq_init(io_gl, LM_ST_SHARED, GL_EXACT, &ip->i_iopen_gh);
 		if (unlikely(error))
 			goto fail_put;
-		flush_delayed_work(&ip->i_iopen_gh.gh_gl->gl_work);
 		glock_set_object(ip->i_iopen_gh.gh_gl, ip);
 		gfs2_glock_put(io_gl);
 		io_gl = NULL;
@@ -202,14 +201,14 @@ struct inode *gfs2_inode_lookup(struct super_block *sb, unsigned int type,
 
 fail_refresh:
 	ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
-	glock_set_object(ip->i_iopen_gh.gh_gl, NULL);
+	glock_clear_object(ip->i_iopen_gh.gh_gl, ip);
 	gfs2_glock_dq_uninit(&ip->i_iopen_gh);
 fail_put:
 	if (io_gl)
 		gfs2_glock_put(io_gl);
+	glock_clear_object(ip->i_gl, ip);
 	if (gfs2_holder_initialized(&i_gh))
 		gfs2_glock_dq_uninit(&i_gh);
-	glock_set_object(ip->i_gl, NULL);
 fail:
 	iget_failed(inode);
 	return ERR_PTR(error);
@@ -706,8 +705,9 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	error = gfs2_glock_get(sdp, ip->i_no_addr, &gfs2_inode_glops, CREATE, &ip->i_gl);
 	if (error)
 		goto fail_free_inode;
-
+	flush_delayed_work(&ip->i_gl->gl_work);
 	glock_set_object(ip->i_gl, ip);
+
 	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, GL_SKIP, ghs + 1);
 	if (error)
 		goto fail_free_inode;
@@ -775,14 +775,17 @@ static int gfs2_create_inode(struct inode *dir, struct dentry *dentry,
 	return error;
 
 fail_gunlock3:
+	glock_clear_object(io_gl, ip);
 	gfs2_glock_dq_uninit(&ip->i_iopen_gh);
 	gfs2_glock_put(io_gl);
 fail_gunlock2:
 	if (io_gl)
 		clear_bit(GLF_INODE_CREATING, &io_gl->gl_flags);
 fail_free_inode:
-	if (ip->i_gl)
+	if (ip->i_gl) {
+		glock_clear_object(ip->i_gl, ip);
 		gfs2_glock_put(ip->i_gl);
+	}
 	gfs2_rsqa_delete(ip, NULL);
 fail_free_acls:
 	if (default_acl)

fs/gfs2/lock_dlm.c

@@ -23,8 +23,6 @@
 #include "sys.h"
 #include "trace_gfs2.h"
 
-extern struct workqueue_struct *gfs2_control_wq;
-
 /**
  * gfs2_update_stats - Update time based stats
  * @mv: Pointer to mean/variance structure to update
@@ -1059,6 +1057,7 @@ static void free_recover_size(struct lm_lockstruct *ls)
 	ls->ls_recover_submit = NULL;
 	ls->ls_recover_result = NULL;
 	ls->ls_recover_size = 0;
+	ls->ls_lvb_bits = NULL;
 }
 
 /* dlm calls before it does lock recovery */
@@ -1175,7 +1174,7 @@ static void gdlm_recovery_result(struct gfs2_sbd *sdp, unsigned int jid,
 	spin_unlock(&ls->ls_recover_spin);
 }
 
-const struct dlm_lockspace_ops gdlm_lockspace_ops = {
+static const struct dlm_lockspace_ops gdlm_lockspace_ops = {
 	.recover_prep = gdlm_recover_prep,
 	.recover_slot = gdlm_recover_slot,
 	.recover_done = gdlm_recover_done,

fs/gfs2/log.c

@@ -898,6 +898,10 @@ static inline int gfs2_jrnl_flush_reqd(struct gfs2_sbd *sdp)
 static inline int gfs2_ail_flush_reqd(struct gfs2_sbd *sdp)
 {
 	unsigned int used_blocks = sdp->sd_jdesc->jd_blocks - atomic_read(&sdp->sd_log_blks_free);
+
+	if (test_and_clear_bit(SDF_FORCE_AIL_FLUSH, &sdp->sd_flags))
+		return 1;
+
 	return used_blocks + atomic_read(&sdp->sd_log_blks_needed) >=
 		atomic_read(&sdp->sd_log_thresh2);
 }
@@ -919,6 +923,15 @@ int gfs2_logd(void *data)
 
 	while (!kthread_should_stop()) {
 
+		/* Check for errors writing to the journal */
+		if (sdp->sd_log_error) {
+			gfs2_lm_withdraw(sdp,
+					 "GFS2: fsid=%s: error %d: "
+					 "withdrawing the file system to "
+					 "prevent further damage.\n",
+					 sdp->sd_fsname, sdp->sd_log_error);
+		}
+
 		did_flush = false;
 		if (gfs2_jrnl_flush_reqd(sdp) || t == 0) {
 			gfs2_ail1_empty(sdp);

fs/gfs2/lops.c

@@ -207,8 +207,11 @@ static void gfs2_end_log_write(struct bio *bio)
 	struct page *page;
 	int i;
 
-	if (bio->bi_status)
-		fs_err(sdp, "Error %d writing to log\n", bio->bi_status);
+	if (bio->bi_status) {
+		fs_err(sdp, "Error %d writing to journal, jid=%u\n",
+		       bio->bi_status, sdp->sd_jdesc->jd_jid);
+		wake_up(&sdp->sd_logd_waitq);
+	}
 
 	bio_for_each_segment_all(bvec, bio, i) {
 		page = bvec->bv_page;

fs/gfs2/meta_io.c

@@ -419,8 +419,9 @@ int gfs2_meta_indirect_buffer(struct gfs2_inode *ip, int height, u64 num,
 	if (ret == 0 && gfs2_metatype_check(sdp, bh, mtype)) {
 		brelse(bh);
 		ret = -EIO;
+	} else {
+		*bhp = bh;
 	}
-	*bhp = bh;
 	return ret;
 }
 
@@ -452,7 +453,7 @@ struct buffer_head *gfs2_meta_ra(struct gfs2_glock *gl, u64 dblock, u32 extlen)
 	if (buffer_uptodate(first_bh))
 		goto out;
 	if (!buffer_locked(first_bh))
-		ll_rw_block(REQ_OP_READ, REQ_META, 1, &first_bh);
+		ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &first_bh);
 
 	dblock++;
 	extlen--;
@@ -461,7 +462,9 @@ struct buffer_head *gfs2_meta_ra(struct gfs2_glock *gl, u64 dblock, u32 extlen)
 		bh = gfs2_getbuf(gl, dblock, CREATE);
 
 		if (!buffer_uptodate(bh) && !buffer_locked(bh))
-			ll_rw_block(REQ_OP_READ, REQ_RAHEAD | REQ_META, 1, &bh);
+			ll_rw_block(REQ_OP_READ,
+				    REQ_RAHEAD | REQ_META | REQ_PRIO,
+				    1, &bh);
 		brelse(bh);
 		dblock++;
 		extlen--;

fs/gfs2/ops_fstype.c

@@ -1113,7 +1113,7 @@ static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent
 		return error;
 	}
 
-	snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s", sdp->sd_table_name);
+	snprintf(sdp->sd_fsname, sizeof(sdp->sd_fsname), "%s", sdp->sd_table_name);
 
 	error = gfs2_sys_fs_add(sdp);
 	/*
@@ -1159,10 +1159,10 @@ static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent
 	}
 
 	if (sdp->sd_args.ar_spectator)
-		snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s",
+		snprintf(sdp->sd_fsname, sizeof(sdp->sd_fsname), "%s.s",
 			 sdp->sd_table_name);
 	else
-		snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u",
+		snprintf(sdp->sd_fsname, sizeof(sdp->sd_fsname), "%s.%u",
 			 sdp->sd_table_name, sdp->sd_lockstruct.ls_jid);
 
 	error = init_inodes(sdp, DO);
@@ -1388,7 +1388,6 @@ static void gfs2_kill_sb(struct super_block *sb)
 	sdp->sd_root_dir = NULL;
 	sdp->sd_master_dir = NULL;
 	shrink_dcache_sb(sb);
-	gfs2_delete_debugfs_file(sdp);
 	free_percpu(sdp->sd_lkstats);
 	kill_block_super(sb);
 }

fs/gfs2/quota.c

@@ -730,7 +730,7 @@ static int gfs2_write_buf_to_page(struct gfs2_inode *ip, unsigned long index,
 		if (PageUptodate(page))
 			set_buffer_uptodate(bh);
 		if (!buffer_uptodate(bh)) {
-			ll_rw_block(REQ_OP_READ, REQ_META, 1, &bh);
+			ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &bh);
 			wait_on_buffer(bh);
 			if (!buffer_uptodate(bh))
 				goto unlock_out;
@@ -1474,8 +1474,11 @@ static void quotad_error(struct gfs2_sbd *sdp, const char *msg, int error)
 {
 	if (error == 0 || error == -EROFS)
 		return;
-	if (!test_bit(SDF_SHUTDOWN, &sdp->sd_flags))
+	if (!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) {
 		fs_err(sdp, "gfs2_quotad: %s error %d\n", msg, error);
+		sdp->sd_log_error = error;
+		wake_up(&sdp->sd_logd_waitq);
+	}
 }
 
 static void quotad_check_timeo(struct gfs2_sbd *sdp, const char *msg,

fs/gfs2/rgrp.c

@@ -705,8 +705,7 @@ void gfs2_clear_rgrpd(struct gfs2_sbd *sdp)
 		rb_erase(n, &sdp->sd_rindex_tree);
 
 		if (gl) {
-			glock_set_object(gl, NULL);
-			gfs2_glock_add_to_lru(gl);
+			glock_clear_object(gl, rgd);
 			gfs2_glock_put(gl);
 		}
 

fs/gfs2/super.c

@@ -924,6 +924,7 @@ static void gfs2_put_super(struct super_block *sb)
 	gfs2_jindex_free(sdp);
 	/*  Take apart glock structures and buffer lists  */
 	gfs2_gl_hash_clear(sdp);
+	gfs2_delete_debugfs_file(sdp);
 	/*  Unmount the locking protocol  */
 	gfs2_lm_unmount(sdp);
 
@@ -943,9 +944,9 @@ static int gfs2_sync_fs(struct super_block *sb, int wait)
 	struct gfs2_sbd *sdp = sb->s_fs_info;
 
 	gfs2_quota_sync(sb, -1);
-	if (wait && sdp)
+	if (wait)
 		gfs2_log_flush(sdp, NULL, NORMAL_FLUSH);
-	return 0;
+	return sdp->sd_log_error;
 }
 
 void gfs2_freeze_func(struct work_struct *work)
@@ -1295,7 +1296,7 @@ static int gfs2_remount_fs(struct super_block *sb, int *flags, char *data)
  * gfs2_drop_inode - Drop an inode (test for remote unlink)
  * @inode: The inode to drop
  *
- * If we've received a callback on an iopen lock then its because a
+ * If we've received a callback on an iopen lock then it's because a
  * remote node tried to deallocate the inode but failed due to this node
  * still having the inode open. Here we mark the link count zero
  * since we know that it must have reached zero if the GLF_DEMOTE flag
@@ -1317,6 +1318,23 @@ static int gfs2_drop_inode(struct inode *inode)
 		if (test_bit(GLF_DEMOTE, &gl->gl_flags))
 			clear_nlink(inode);
 	}
+
+	/*
+	 * When under memory pressure when an inode's link count has dropped to
+	 * zero, defer deleting the inode to the delete workqueue.  This avoids
+	 * calling into DLM under memory pressure, which can deadlock.
+	 */
+	if (!inode->i_nlink &&
+	    unlikely(current->flags & PF_MEMALLOC) &&
+	    gfs2_holder_initialized(&ip->i_iopen_gh)) {
+		struct gfs2_glock *gl = ip->i_iopen_gh.gh_gl;
+
+		gfs2_glock_hold(gl);
+		if (queue_work(gfs2_delete_workqueue, &gl->gl_delete) == 0)
+			gfs2_glock_queue_put(gl);
+		return false;
+	}
+
 	return generic_drop_inode(inode);
 }
 
@@ -1500,6 +1518,22 @@ static int gfs2_dinode_dealloc(struct gfs2_inode *ip)
 	return error;
 }
 
+/**
+ * gfs2_glock_put_eventually
+ * @gl:	The glock to put
+ *
+ * When under memory pressure, trigger a deferred glock put to make sure we
+ * won't call into DLM and deadlock.  Otherwise, put the glock directly.
+ */
+
+static void gfs2_glock_put_eventually(struct gfs2_glock *gl)
+{
+	if (current->flags & PF_MEMALLOC)
+		gfs2_glock_queue_put(gl);
+	else
+		gfs2_glock_put(gl);
+}
+
 /**
  * gfs2_evict_inode - Remove an inode from cache
  * @inode: The inode to evict
@@ -1544,9 +1578,14 @@ static void gfs2_evict_inode(struct inode *inode)
 		goto alloc_failed;
 	}
 
+	/* Deletes should never happen under memory pressure anymore.  */
+	if (WARN_ON_ONCE(current->flags & PF_MEMALLOC))
+		goto out;
+
 	/* Must not read inode block until block type has been verified */
 	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, GL_SKIP, &gh);
 	if (unlikely(error)) {
+		glock_clear_object(ip->i_iopen_gh.gh_gl, ip);
 		ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
 		gfs2_glock_dq_uninit(&ip->i_iopen_gh);
 		goto out;
@@ -1562,6 +1601,12 @@ static void gfs2_evict_inode(struct inode *inode)
 			goto out_truncate;
 	}
 
+	/*
+	 * The inode may have been recreated in the meantime.
+	 */
+	if (inode->i_nlink)
+		goto out_truncate;
+
 alloc_failed:
 	if (gfs2_holder_initialized(&ip->i_iopen_gh) &&
 	    test_bit(HIF_HOLDER, &ip->i_iopen_gh.gh_iflags)) {
@@ -1595,6 +1640,11 @@ static void gfs2_evict_inode(struct inode *inode)
 			goto out_unlock;
 	}
 
+	/* We're about to clear the bitmap for the dinode, but as soon as we
+	   do, gfs2_create_inode can create another inode at the same block
+	   location and try to set gl_object again. We clear gl_object here so
+	   that subsequent inode creates don't see an old gl_object. */
+	glock_clear_object(ip->i_gl, ip);
 	error = gfs2_dinode_dealloc(ip);
 	goto out_unlock;
 
@@ -1623,14 +1673,17 @@ static void gfs2_evict_inode(struct inode *inode)
 		gfs2_rs_deltree(&ip->i_res);
 
 	if (gfs2_holder_initialized(&ip->i_iopen_gh)) {
+		glock_clear_object(ip->i_iopen_gh.gh_gl, ip);
 		if (test_bit(HIF_HOLDER, &ip->i_iopen_gh.gh_iflags)) {
 			ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
 			gfs2_glock_dq(&ip->i_iopen_gh);
 		}
 		gfs2_holder_uninit(&ip->i_iopen_gh);
 	}
-	if (gfs2_holder_initialized(&gh))
+	if (gfs2_holder_initialized(&gh)) {
+		glock_clear_object(ip->i_gl, ip);
 		gfs2_glock_dq_uninit(&gh);
+	}
 	if (error && error != GLR_TRYFAILED && error != -EROFS)
 		fs_warn(sdp, "gfs2_evict_inode: %d\n", error);
 out:
@@ -1640,15 +1693,19 @@ static void gfs2_evict_inode(struct inode *inode)
 	gfs2_ordered_del_inode(ip);
 	clear_inode(inode);
 	gfs2_dir_hash_inval(ip);
-	glock_set_object(ip->i_gl, NULL);
+	glock_clear_object(ip->i_gl, ip);
 	wait_on_bit_io(&ip->i_flags, GIF_GLOP_PENDING, TASK_UNINTERRUPTIBLE);
 	gfs2_glock_add_to_lru(ip->i_gl);
-	gfs2_glock_put(ip->i_gl);
+	gfs2_glock_put_eventually(ip->i_gl);
 	ip->i_gl = NULL;
 	if (gfs2_holder_initialized(&ip->i_iopen_gh)) {
-		glock_set_object(ip->i_iopen_gh.gh_gl, NULL);
+		struct gfs2_glock *gl = ip->i_iopen_gh.gh_gl;
+
+		glock_clear_object(gl, ip);
 		ip->i_iopen_gh.gh_flags |= GL_NOCACHE;
+		gfs2_glock_hold(gl);
 		gfs2_glock_dq_uninit(&ip->i_iopen_gh);
+		gfs2_glock_put_eventually(gl);
 	}
 }
 

fs/gfs2/util.h

@@ -151,6 +151,7 @@ extern struct kmem_cache *gfs2_rgrpd_cachep;
 extern struct kmem_cache *gfs2_quotad_cachep;
 extern struct kmem_cache *gfs2_qadata_cachep;
 extern mempool_t *gfs2_page_pool;
+extern struct workqueue_struct *gfs2_control_wq;
 
 static inline unsigned int gfs2_tune_get_i(struct gfs2_tune *gt,
 					   unsigned int *p)

fs/gfs2/xattr.c

@@ -25,6 +25,7 @@
 #include "meta_io.h"
 #include "quota.h"
 #include "rgrp.h"
+#include "super.h"
 #include "trans.h"
 #include "util.h"
 
@@ -1209,8 +1210,12 @@ int __gfs2_xattr_set(struct inode *inode, const char *name,
 	if (namel > GFS2_EA_MAX_NAME_LEN)
 		return -ERANGE;
 
-	if (value == NULL)
-		return gfs2_xattr_remove(ip, type, name);
+	if (value == NULL) {
+		error = gfs2_xattr_remove(ip, type, name);
+		if (error == -ENODATA && !(flags & XATTR_REPLACE))
+			error = 0;
+		return error;
+	}
 
 	if (ea_check_size(sdp, namel, size))
 		return -ERANGE;