Skip to contentSkip to conversation

iext handling fixes and cleanup

PATCH In review Christoph Hellwig 2 participants 6 messages Apr 30, 2024 – May 1, 2024 XFS filesystem

lore
Share & tools
lore
#0 lore raw

Hi all,

this series fixes two unlikely to hit iext handling bugs in xfs_reflink_end_cow and then cleanups up the iext count updgrade handling.

It has been split from a larger series that needs more work.

Diffstat:

 libxfs/xfs_attr.c       |    5 ----
 libxfs/xfs_bmap.c       |    5 ----
 libxfs/xfs_inode_fork.c |   57 +++++++++++++++++++++---------------------------
 libxfs/xfs_inode_fork.h |    6 +----
 xfs_bmap_item.c         |    4 ---
 xfs_bmap_util.c         |   24 ++++----------------
 xfs_dquot.c             |    5 ----
 xfs_iomap.c             |    9 +------
 xfs_reflink.c           |   23 +++++--------------
 xfs_rtalloc.c           |    5 ----
 10 files changed, 45 insertions(+), 98 deletions(-)
#1 lore raw

PATCH 1/3 xfs: upgrade the extent counters in xfs_reflink_end_cow_extent later

Defer the extent counter size upgrade until we know we're going to modify the extent mapping. This also defers dirtying the transaction and will allow us safely back out later in the function in later changes.

Fixes 4f86bb4b66c9 ("xfs: Conditionally upgrade existing inodes to use large extent counters")
---
 fs/xfs/xfs_reflink.c | 16 ++++++++--------
 1 file changed, 8 insertions(+), 8 deletions(-)
fs/xfs/xfs_reflink.c +8 -8
index 7da0e8f961d351..9ce37d366534c3 100644
@@ -751,14 +751,6 @@ xfs_reflink_end_cow_extent(
 	xfs_ilock(ip, XFS_ILOCK_EXCL);
 	xfs_trans_ijoin(tp, ip, 0);
 
-	error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK,
-			XFS_IEXT_REFLINK_END_COW_CNT);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, ip,
-				XFS_IEXT_REFLINK_END_COW_CNT);
-	if (error)
-		goto out_cancel;
-
 	/*
 	 * In case of racing, overlapping AIO writes no COW extents might be
 	 * left by the time I/O completes for the loser of the race.  In that
@@ -787,6 +779,14 @@ xfs_reflink_end_cow_extent(
 	del = got;
 	xfs_trim_extent(&del, *offset_fsb, end_fsb - *offset_fsb);
 
+	error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK,
+			XFS_IEXT_REFLINK_END_COW_CNT);
+	if (error == -EFBIG)
+		error = xfs_iext_count_upgrade(tp, ip,
+				XFS_IEXT_REFLINK_END_COW_CNT);
+	if (error)
+		goto out_cancel;
+
 	/* Grab the corresponding mapping in the data fork. */
 	nmaps = 1;
 	error = xfs_bmapi_read(ip, del.br_startoff, del.br_blockcount, &data,
Signature · 1 line
2.39.2
#2 lore raw

PATCH 2/3 xfs: remove a racy if_bytes check in xfs_reflink_end_cow_extent

In reply to Christoph Hellwig

Accessing if_bytes without the ilock is racy. Remove the initial if_bytes == 0 check in xfs_reflink_end_cow_extent and let ext_iext_lookup_extent fail for this case after we've taken the ilock.

---
 fs/xfs/xfs_reflink.c | 6 ------
 1 file changed, 6 deletions(-)
fs/xfs/xfs_reflink.c +0 -6
index 9ce37d366534c3..0ab2ef5b58f6c4 100644
@@ -731,12 +731,6 @@ xfs_reflink_end_cow_extent(
 	int			nmaps;
 	int			error;
 
-	/* No COW extents?  That's easy! */
-	if (ifp->if_bytes == 0) {
-		*offset_fsb = end_fsb;
-		return 0;
-	}
-
 	resblks = XFS_EXTENTADD_SPACE_RES(mp, XFS_DATA_FORK);
 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_write, resblks, 0,
 			XFS_TRANS_RESERVE, &tp);
Signature · 1 line
2.39.2
#3 lore raw

PATCH 3/3 xfs: simplify iext overflow checking and upgrade

In reply to Christoph Hellwig

Currently the calls to xfs_iext_count_may_overflow and xfs_iext_count_upgrade are always paired. Merge them into a single function to simplify the callers and the actual check and upgrade logic itself.

---
 fs/xfs/libxfs/xfs_attr.c       |  5 +--
 fs/xfs/libxfs/xfs_bmap.c       |  5 +--
 fs/xfs/libxfs/xfs_inode_fork.c | 57 +++++++++++++++-------------------
 fs/xfs/libxfs/xfs_inode_fork.h |  6 ++--
 fs/xfs/xfs_bmap_item.c         |  4 +--
 fs/xfs/xfs_bmap_util.c         | 24 +++-----------
 fs/xfs/xfs_dquot.c             |  5 +--
 fs/xfs/xfs_iomap.c             |  9 ++----
 fs/xfs/xfs_reflink.c           |  9 ++----
 fs/xfs/xfs_rtalloc.c           |  5 +--
 10 files changed, 41 insertions(+), 88 deletions(-)
fs/xfs/libxfs/xfs_attr.c +1 -4
index 1c2a27fce08a9d..ded92ccefe9f6d 100644
@@ -1050,11 +1050,8 @@ xfs_attr_set(
 		return error;
 
 	if (op != XFS_ATTRUPDATE_REMOVE || xfs_inode_hasattr(dp)) {
-		error = xfs_iext_count_may_overflow(dp, XFS_ATTR_FORK,
+		error = xfs_iext_count_ensure(args->trans, dp, XFS_ATTR_FORK,
 				XFS_IEXT_ATTR_MANIP_CNT(rmt_blks));
-		if (error == -EFBIG)
-			error = xfs_iext_count_upgrade(args->trans, dp,
-					XFS_IEXT_ATTR_MANIP_CNT(rmt_blks));
 		if (error)
 			goto out_trans_cancel;
 	}
fs/xfs/libxfs/xfs_bmap.c +1 -4
index 6053f5e5c71eec..3debd0d561b812 100644
@@ -4621,11 +4621,8 @@ xfs_bmapi_convert_delalloc(
 	xfs_ilock(ip, XFS_ILOCK_EXCL);
 	xfs_trans_ijoin(tp, ip, 0);
 
-	error = xfs_iext_count_may_overflow(ip, whichfork,
+	error = xfs_iext_count_ensure(tp, ip, whichfork,
 			XFS_IEXT_ADD_NOSPLIT_CNT);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, ip,
-				XFS_IEXT_ADD_NOSPLIT_CNT);
 	if (error)
 		goto out_trans_cancel;
 
fs/xfs/libxfs/xfs_inode_fork.c +25 -32

78 lines not on this page · raw message

fs/xfs/libxfs/xfs_inode_fork.h +2 -4
index bd53eb951b6515..9e1456f5cc2c85 100644
@@ -256,10 +256,8 @@ extern void xfs_ifork_init_cow(struct xfs_inode *ip);
 
 int xfs_ifork_verify_local_data(struct xfs_inode *ip);
 int xfs_ifork_verify_local_attr(struct xfs_inode *ip);
-int xfs_iext_count_may_overflow(struct xfs_inode *ip, int whichfork,
-		int nr_to_add);
-int xfs_iext_count_upgrade(struct xfs_trans *tp, struct xfs_inode *ip,
-		uint nr_to_add);
+int xfs_iext_count_ensure(struct xfs_trans *tp, struct xfs_inode *ip,
+		int whichfork, uint nr_to_add);
 bool xfs_ifork_is_realtime(struct xfs_inode *ip, int whichfork);
 
 /* returns true if the fork has extents but they are not read in yet. */
fs/xfs/xfs_bmap_item.c +1 -3
index d27859a684aa69..38067d02ee3ca7 100644
@@ -524,9 +524,7 @@ xfs_bmap_recover_work(
 	else
 		iext_delta = XFS_IEXT_PUNCH_HOLE_CNT;
 
-	error = xfs_iext_count_may_overflow(ip, work->bi_whichfork, iext_delta);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, ip, iext_delta);
+	error = xfs_iext_count_ensure(tp, ip, work->bi_whichfork, iext_delta);
 	if (error)
 		goto err_cancel;
 
fs/xfs/xfs_bmap_util.c +5 -19
index 53aa90a0ee3a85..11062bddba0e5e 100644
@@ -713,11 +713,8 @@ xfs_alloc_file_space(
 		if (error)
 			break;
 
-		error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK,
+		error = xfs_iext_count_ensure(tp, ip, XFS_DATA_FORK,
 				XFS_IEXT_ADD_NOSPLIT_CNT);
-		if (error == -EFBIG)
-			error = xfs_iext_count_upgrade(tp, ip,
-					XFS_IEXT_ADD_NOSPLIT_CNT);
 		if (error)
 			goto error;
 
@@ -775,10 +772,8 @@ xfs_unmap_extent(
 	if (error)
 		return error;
 
-	error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK,
+	error = xfs_iext_count_ensure(tp, ip, XFS_DATA_FORK,
 			XFS_IEXT_PUNCH_HOLE_CNT);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, ip, XFS_IEXT_PUNCH_HOLE_CNT);
 	if (error)
 		goto out_trans_cancel;
 
@@ -1054,10 +1049,8 @@ xfs_insert_file_space(
 	xfs_ilock(ip, XFS_ILOCK_EXCL);
 	xfs_trans_ijoin(tp, ip, 0);
 
-	error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK,
+	error = xfs_iext_count_ensure(tp, ip, XFS_DATA_FORK,
 			XFS_IEXT_PUNCH_HOLE_CNT);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, ip, XFS_IEXT_PUNCH_HOLE_CNT);
 	if (error)
 		goto out_trans_cancel;
 
@@ -1283,23 +1276,16 @@ xfs_swap_extent_rmap(
 			trace_xfs_swap_extent_rmap_remap_piece(tip, &uirec);
 
 			if (xfs_bmap_is_real_extent(&uirec)) {
-				error = xfs_iext_count_may_overflow(ip,
-						XFS_DATA_FORK,
+				error = xfs_iext_count_ensure(tp, ip, XFS_DATA_FORK,
 						XFS_IEXT_SWAP_RMAP_CNT);
-				if (error == -EFBIG)
-					error = xfs_iext_count_upgrade(tp, ip,
-							XFS_IEXT_SWAP_RMAP_CNT);
 				if (error)
 					goto out;
 			}
 
 			if (xfs_bmap_is_real_extent(&irec)) {
-				error = xfs_iext_count_may_overflow(tip,
+				error = xfs_iext_count_ensure(tp, tip,
 						XFS_DATA_FORK,
 						XFS_IEXT_SWAP_RMAP_CNT);
-				if (error == -EFBIG)
-					error = xfs_iext_count_upgrade(tp, ip,
-							XFS_IEXT_SWAP_RMAP_CNT);
 				if (error)
 					goto out;
 			}
fs/xfs/xfs_dquot.c +1 -4
index 13aba84bd64afb..c2e66d392399dd 100644
@@ -341,11 +341,8 @@ xfs_dquot_disk_alloc(
 		goto err_cancel;
 	}
 
-	error = xfs_iext_count_may_overflow(quotip, XFS_DATA_FORK,
+	error = xfs_iext_count_ensure(tp, quotip, XFS_DATA_FORK,
 			XFS_IEXT_ADD_NOSPLIT_CNT);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, quotip,
-				XFS_IEXT_ADD_NOSPLIT_CNT);
 	if (error)
 		goto err_cancel;
 
fs/xfs/xfs_iomap.c +2 -7
index 9ce0f6b9df93e6..9dfb8e1a5a55cb 100644
@@ -299,9 +299,7 @@ xfs_iomap_write_direct(
 	if (error)
 		return error;
 
-	error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK, nr_exts);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, ip, nr_exts);
+	error = xfs_iext_count_ensure(tp, ip, XFS_DATA_FORK, nr_exts);
 	if (error)
 		goto out_trans_cancel;
 
@@ -625,11 +623,8 @@ xfs_iomap_write_unwritten(
 		if (error)
 			return error;
 
-		error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK,
+		error = xfs_iext_count_ensure(tp, ip, XFS_DATA_FORK,
 				XFS_IEXT_WRITE_UNWRITTEN_CNT);
-		if (error == -EFBIG)
-			error = xfs_iext_count_upgrade(tp, ip,
-					XFS_IEXT_WRITE_UNWRITTEN_CNT);
 		if (error)
 			goto error_on_bmapi_transaction;
 
fs/xfs/xfs_reflink.c +2 -7
index 0ab2ef5b58f6c4..323d5494149698 100644
@@ -773,11 +773,8 @@ xfs_reflink_end_cow_extent(
 	del = got;
 	xfs_trim_extent(&del, *offset_fsb, end_fsb - *offset_fsb);
 
-	error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK,
+	error = xfs_iext_count_ensure(tp, ip, XFS_DATA_FORK,
 			XFS_IEXT_REFLINK_END_COW_CNT);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, ip,
-				XFS_IEXT_REFLINK_END_COW_CNT);
 	if (error)
 		goto out_cancel;
 
@@ -1277,9 +1274,7 @@ xfs_reflink_remap_extent(
 	if (dmap_written)
 		++iext_delta;
 
-	error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK, iext_delta);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, ip, iext_delta);
+	error = xfs_iext_count_ensure(tp, ip, XFS_DATA_FORK, iext_delta);
 	if (error)
 		goto out_cancel;
 
fs/xfs/xfs_rtalloc.c +1 -4
index b476a876478d93..37edf4c5ce73ad 100644
@@ -695,11 +695,8 @@ xfs_growfs_rt_alloc(
 		xfs_ilock(ip, XFS_ILOCK_EXCL);
 		xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
 
-		error = xfs_iext_count_may_overflow(ip, XFS_DATA_FORK,
+		error = xfs_iext_count_ensure(tp, ip, XFS_DATA_FORK,
 				XFS_IEXT_ADD_NOSPLIT_CNT);
-		if (error == -EFBIG)
-			error = xfs_iext_count_upgrade(tp, ip,
-					XFS_IEXT_ADD_NOSPLIT_CNT);
 		if (error)
 			goto out_trans_cancel;
 
Signature · 1 line
2.39.2
#4 lore raw

Re: [PATCH 3/3] xfs: simplify iext overflow checking and upgrade

Quoted 62 lines · Christoph Hellwig

Currently the calls to xfs_iext_count_may_overflow and xfs_iext_count_upgrade are always paired. Merge them into a single function to simplify the callers and the actual check and upgrade logic itself.

---
 fs/xfs/libxfs/xfs_attr.c       |  5 +--
 fs/xfs/libxfs/xfs_bmap.c       |  5 +--
 fs/xfs/libxfs/xfs_inode_fork.c | 57 +++++++++++++++-------------------
 fs/xfs/libxfs/xfs_inode_fork.h |  6 ++--
 fs/xfs/xfs_bmap_item.c         |  4 +--
 fs/xfs/xfs_bmap_util.c         | 24 +++-----------
 fs/xfs/xfs_dquot.c             |  5 +--
 fs/xfs/xfs_iomap.c             |  9 ++----
 fs/xfs/xfs_reflink.c           |  9 ++----
 fs/xfs/xfs_rtalloc.c           |  5 +--
 10 files changed, 41 insertions(+), 88 deletions(-)
fs/xfs/libxfs/xfs_attr.c +1 -4
index 1c2a27fce08a9d..ded92ccefe9f6d 100644
@@ -1050,11 +1050,8 @@ xfs_attr_set(
 		return error;
 
 	if (op != XFS_ATTRUPDATE_REMOVE || xfs_inode_hasattr(dp)) {
-		error = xfs_iext_count_may_overflow(dp, XFS_ATTR_FORK,
+		error = xfs_iext_count_ensure(args->trans, dp, XFS_ATTR_FORK,
 				XFS_IEXT_ATTR_MANIP_CNT(rmt_blks));
-		if (error == -EFBIG)
-			error = xfs_iext_count_upgrade(args->trans, dp,
-					XFS_IEXT_ATTR_MANIP_CNT(rmt_blks));
 		if (error)
 			goto out_trans_cancel;
 	}
fs/xfs/libxfs/xfs_bmap.c +1 -4
index 6053f5e5c71eec..3debd0d561b812 100644
@@ -4621,11 +4621,8 @@ xfs_bmapi_convert_delalloc(
 	xfs_ilock(ip, XFS_ILOCK_EXCL);
 	xfs_trans_ijoin(tp, ip, 0);
 
-	error = xfs_iext_count_may_overflow(ip, whichfork,
+	error = xfs_iext_count_ensure(tp, ip, whichfork,
 			XFS_IEXT_ADD_NOSPLIT_CNT);
-	if (error == -EFBIG)
-		error = xfs_iext_count_upgrade(tp, ip,
-				XFS_IEXT_ADD_NOSPLIT_CNT);
 	if (error)
 		goto out_trans_cancel;
 
fs/xfs/libxfs/xfs_inode_fork.c +0 -0
index 7d660a9739090a..82e670dd1212c4 100644
@@ -765,53 +765,46 @@ xfs_ifork_verify_local_attr(
 	return 0;
 }
 

fs/xfs/libxfs/xfs_inode_fork.c xfs_ifork_verify_local_attr( · line 776

Quoted 10 lines
fs/xfs/libxfs/xfs_inode_fork.c +8 -1
+/*
+ * Check if the inode fork supports adding nr_to_add more extents.
+ *
+ * If it doesn't but we can upgrade it to large extent counters, do the upgrade.
+ * If we can't upgrade or are already using big counters but still can't fit the
+ * additional extents, return -EFBIG.
+ */
 int
-xfs_iext_count_may_overflow(
+xfs_iext_count_ensure(

Everything looks fine, but the name isn't very good. What, exactly, is this function ensuring about the iext count? The function is extending the iext count if needed, so to me the function name should reflect what it actually does.

xfs_iext_count_extend() seems like a much better name - it tells the reader what the code is actually doing (i.e. we may have to extend the iext count before performing this operation) and it makes it obvious when it is done out of place....

-Dave.

Signature · 2 lines
Dave Chinner
[email protected]
#5 lore raw

Re: [PATCH 3/3] xfs: simplify iext overflow checking and upgrade

Quoted 4 lines · Dave Chinner

xfs_iext_count_extend() seems like a much better name - it tells the reader what the code is actually doing (i.e. we may have to extend the iext count before performing this operation) and it makes it obvious when it is done out of place....

Sure, I can change the name.

Btw, it would be nice to trim your reply a bit more, I had to scroll down a few pages of quotes text yo get here.