Merge tag 'fs_for_v5.19-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs

Pull ext2, writeback, and quota fixes and cleanups from Jan Kara:
 "A fix for race in writeback code and two cleanups in quota and ext2"

* tag 'fs_for_v5.19-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
  quota: Prevent memory allocation recursion while holding dq_lock
  writeback: Fix inode->i_io_list not be protected by inode->i_lock error
  fs: Fix syntax errors in comments

4 files changed, 40 insertions, 11 deletions

diff --git a/fs/ext2/inode.c b/fs/ext2/inode.c
index 360ce3604a2d..e6b932219803 100644
--- a/fs/ext2/inode.c
+++ b/fs/ext2/inode.c
@@ -1549,7 +1549,7 @@ static int __ext2_write_inode(struct inode *inode, int do_sync)
 	if (IS_ERR(raw_inode))
  		return -EIO;
 
-	/* For fields not not tracking in the in-memory inode,
+	/* For fields not tracking in the in-memory inode,
 	 * initialise them to zero for new inodes. */
 	if (ei->i_state & EXT2_STATE_NEW)
 		memset(raw_inode, 0, EXT2_SB(sb)->s_inode_size);
diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c
index a21d8f1a56d1..05221366a16d 100644
--- a/fs/fs-writeback.c
+++ b/fs/fs-writeback.c
@@ -120,6 +120,7 @@ static bool inode_io_list_move_locked(struct inode *inode,
 				      struct list_head *head)
 {
 	assert_spin_locked(&wb->list_lock);
+	assert_spin_locked(&inode->i_lock);
 
 	list_move(&inode->i_io_list, head);
 
@@ -1365,9 +1366,9 @@ static int move_expired_inodes(struct list_head *delaying_queue,
 		inode = wb_inode(delaying_queue->prev);
 		if (inode_dirtied_after(inode, dirtied_before))
 			break;
+		spin_lock(&inode->i_lock);
 		list_move(&inode->i_io_list, &tmp);
 		moved++;
-		spin_lock(&inode->i_lock);
 		inode->i_state |= I_SYNC_QUEUED;
 		spin_unlock(&inode->i_lock);
 		if (sb_is_blkdev_sb(inode->i_sb))
@@ -1383,7 +1384,12 @@ static int move_expired_inodes(struct list_head *delaying_queue,
 		goto out;
 	}
 
-	/* Move inodes from one superblock together */
+	/*
+	 * Although inode's i_io_list is moved from 'tmp' to 'dispatch_queue',
+	 * we don't take inode->i_lock here because it is just a pointless overhead.
+	 * Inode is already marked as I_SYNC_QUEUED so writeback list handling is
+	 * fully under our control.
+	 */
 	while (!list_empty(&tmp)) {
 		sb = wb_inode(tmp.prev)->i_sb;
 		list_for_each_prev_safe(pos, node, &tmp) {
@@ -1826,8 +1832,8 @@ static long writeback_sb_inodes(struct super_block *sb,
 			 * We'll have another go at writing back this inode
 			 * when we completed a full scan of b_io.
 			 */
-			spin_unlock(&inode->i_lock);
 			requeue_io(inode, wb);
+			spin_unlock(&inode->i_lock);
 			trace_writeback_sb_inodes_requeue(inode);
 			continue;
 		}
@@ -2358,6 +2364,7 @@ void __mark_inode_dirty(struct inode *inode, int flags)
 {
 	struct super_block *sb = inode->i_sb;
 	int dirtytime = 0;
+	struct bdi_writeback *wb = NULL;
 
 	trace_writeback_mark_inode_dirty(inode, flags);
 
@@ -2410,13 +2417,24 @@ void __mark_inode_dirty(struct inode *inode, int flags)
 		inode->i_state |= flags;
 
 		/*
+		 * Grab inode's wb early because it requires dropping i_lock and we
+		 * need to make sure following checks happen atomically with dirty
+		 * list handling so that we don't move inodes under flush worker's
+		 * hands.
+		 */
+		if (!was_dirty) {
+			wb = locked_inode_to_wb_and_lock_list(inode);
+			spin_lock(&inode->i_lock);
+		}
+
+		/*
 		 * If the inode is queued for writeback by flush worker, just
 		 * update its dirty state. Once the flush worker is done with
 		 * the inode it will place it on the appropriate superblock
 		 * list, based upon its state.
 		 */
 		if (inode->i_state & I_SYNC_QUEUED)
-			goto out_unlock_inode;
+			goto out_unlock;
 
 		/*
 		 * Only add valid (hashed) inodes to the superblock's
@@ -2424,22 +2442,19 @@ void __mark_inode_dirty(struct inode *inode, int flags)
 		 */
 		if (!S_ISBLK(inode->i_mode)) {
 			if (inode_unhashed(inode))
-				goto out_unlock_inode;
+				goto out_unlock;
 		}
 		if (inode->i_state & I_FREEING)
-			goto out_unlock_inode;
+			goto out_unlock;
 
 		/*
 		 * If the inode was already on b_dirty/b_io/b_more_io, don't
 		 * reposition it (that would break b_dirty time-ordering).
 		 */
 		if (!was_dirty) {
-			struct bdi_writeback *wb;
 			struct list_head *dirty_list;
 			bool wakeup_bdi = false;
 
-			wb = locked_inode_to_wb_and_lock_list(inode);
-
 			inode->dirtied_when = jiffies;
 			if (dirtytime)
 				inode->dirtied_time_when = jiffies;
@@ -2453,6 +2468,7 @@ void __mark_inode_dirty(struct inode *inode, int flags)
 							       dirty_list);
 
 			spin_unlock(&wb->list_lock);
+			spin_unlock(&inode->i_lock);
 			trace_writeback_dirty_inode_enqueue(inode);
 
 			/*
@@ -2467,6 +2483,9 @@ void __mark_inode_dirty(struct inode *inode, int flags)
 			return;
 		}
 	}
+out_unlock:
+	if (wb)
+		spin_unlock(&wb->list_lock);
 out_unlock_inode:
 	spin_unlock(&inode->i_lock);
 }
diff --git a/fs/inode.c b/fs/inode.c
index 9d9b422504d1..bd4da9c5207e 100644
--- a/fs/inode.c
+++ b/fs/inode.c
@@ -27,7 +27,7 @@
  * Inode locking rules:
  *
  * inode->i_lock protects:
- *   inode->i_state, inode->i_hash, __iget()
+ *   inode->i_state, inode->i_hash, __iget(), inode->i_io_list
  * Inode LRU list locks protect:
  *   inode->i_sb->s_inode_lru, inode->i_lru
  * inode->i_sb->s_inode_list_lock protects:
diff --git a/fs/quota/dquot.c b/fs/quota/dquot.c
index a74aef99bd3d..09d1307959d0 100644
--- a/fs/quota/dquot.c
+++ b/fs/quota/dquot.c
@@ -79,6 +79,7 @@
 #include <linux/capability.h>
 #include <linux/quotaops.h>
 #include <linux/blkdev.h>
+#include <linux/sched/mm.h>
 #include "../internal.h" /* ugh */
 
 #include <linux/uaccess.h>
@@ -425,9 +426,11 @@ EXPORT_SYMBOL(mark_info_dirty);
 int dquot_acquire(struct dquot *dquot)
 {
 	int ret = 0, ret2 = 0;
+	unsigned int memalloc;
 	struct quota_info *dqopt = sb_dqopt(dquot->dq_sb);
 
 	mutex_lock(&dquot->dq_lock);
+	memalloc = memalloc_nofs_save();
 	if (!test_bit(DQ_READ_B, &dquot->dq_flags)) {
 		ret = dqopt->ops[dquot->dq_id.type]->read_dqblk(dquot);
 		if (ret < 0)
@@ -458,6 +461,7 @@ int dquot_acquire(struct dquot *dquot)
 	smp_mb__before_atomic();
 	set_bit(DQ_ACTIVE_B, &dquot->dq_flags);
 out_iolock:
+	memalloc_nofs_restore(memalloc);
 	mutex_unlock(&dquot->dq_lock);
 	return ret;
 }
@@ -469,9 +473,11 @@ EXPORT_SYMBOL(dquot_acquire);
 int dquot_commit(struct dquot *dquot)
 {
 	int ret = 0;
+	unsigned int memalloc;
 	struct quota_info *dqopt = sb_dqopt(dquot->dq_sb);
 
 	mutex_lock(&dquot->dq_lock);
+	memalloc = memalloc_nofs_save();
 	if (!clear_dquot_dirty(dquot))
 		goto out_lock;
 	/* Inactive dquot can be only if there was error during read/init
@@ -481,6 +487,7 @@ int dquot_commit(struct dquot *dquot)
 	else
 		ret = -EIO;
 out_lock:
+	memalloc_nofs_restore(memalloc);
 	mutex_unlock(&dquot->dq_lock);
 	return ret;
 }
@@ -492,9 +499,11 @@ EXPORT_SYMBOL(dquot_commit);
 int dquot_release(struct dquot *dquot)
 {
 	int ret = 0, ret2 = 0;
+	unsigned int memalloc;
 	struct quota_info *dqopt = sb_dqopt(dquot->dq_sb);
 
 	mutex_lock(&dquot->dq_lock);
+	memalloc = memalloc_nofs_save();
 	/* Check whether we are not racing with some other dqget() */
 	if (dquot_is_busy(dquot))
 		goto out_dqlock;
@@ -510,6 +519,7 @@ int dquot_release(struct dquot *dquot)
 	}
 	clear_bit(DQ_ACTIVE_B, &dquot->dq_flags);
 out_dqlock:
+	memalloc_nofs_restore(memalloc);
 	mutex_unlock(&dquot->dq_lock);
 	return ret;
 }