Skip to contentSkip to conversation

Patch "netfs: Fix unbuffered write error handling" has been added to the 6.16-stable tree

PATCH [email protected] 1 participant 1 message Aug 21, 2025

Subsystems (2)
lore
Share & tools
lore

[email protected]

author Applied
#0 lore raw

This is a note to let you know that I've just added the patch titled

netfs: Fix unbuffered write error handling

to the 6.16-stable tree which can be found at:
http://www.kernel.org/git/?p=linux/kernel/git/stable/stable-queue.git;a=summary

The filename of the patch is:
netfs-fix-unbuffered-write-error-handling.patch
and it can be found in the queue-6.16 subdirectory.

If you, or anyone else, feels it should not be added to the stable tree, please let <[email protected]> know about it.

From a3de58b12ce074ec05b8741fa28d62ccb1070468 Mon Sep 17 00:00:00 2001
From: David Howells <[email protected]>
Date: Thu, 14 Aug 2025 22:45:50 +0100
Subject: netfs: Fix unbuffered write error handling

From: David Howells <[email protected]>

commit a3de58b12ce074ec05b8741fa28d62ccb1070468 upstream.

If all the subrequests in an unbuffered write stream fail, the subrequest collector doesn't update the stream->transferred value and it retains its initial LONG_MAX value. Unfortunately, if all active streams fail, then we take the smallest value of { LONG_MAX, LONG_MAX, ... } as the value to set in wreq->transferred - which is then returned from ->write_iter().

LONG_MAX was chosen as the initial value so that all the streams can be quickly assessed by taking the smallest value of all stream->transferred - but this only works if we've set any of them.

Fix this by adding a flag to indicate whether the value in stream->transferred is valid and checking that when we integrate the values. stream->transferred can then be initialised to zero.

This was found by running the generic/750 xfstest against cifs with cache=none. It splices data to the target file. Once (if) it has used up all the available scratch space, the writes start failing with ENOSPC. This causes ->write_iter() to fail. However, it was returning wreq->transferred, i.e. LONG_MAX, rather than an error (because it thought the amount transferred was non-zero) and iter_file_splice_write() would then try to clean up that amount of pipe bufferage - leading to an oops when it overran. The kernel log showed:

CIFS: VFS: Send error in write = -28

followed by:

    BUG: kernel NULL pointer dereference, address: 0000000000000008

with:

    RIP: 0010:iter_file_splice_write+0x3a4/0x520
    do_splice+0x197/0x4e0

or:

    RIP: 0010:pipe_buf_release (include/linux/pipe_fs_i.h:282)
    iter_file_splice_write (fs/splice.c:755)

Also put a warning check into splice to announce if ->write_iter() returned that it had written more than it was asked to.

---
 fs/netfs/read_collect.c  |    4 +++-
 fs/netfs/write_collect.c |   10 ++++++++--
 fs/netfs/write_issue.c   |    4 ++--
 fs/splice.c              |    3 +++
 include/linux/netfs.h    |    1 +
 5 files changed, 17 insertions(+), 5 deletions(-)
fs/netfs/read_collect.c +3 -1
@@ -281,8 +281,10 @@ reassess:
 		} else if (test_bit(NETFS_RREQ_SHORT_TRANSFER, &rreq->flags)) {
 			notes |= MADE_PROGRESS;
 		} else {
-			if (!stream->failed)
+			if (!stream->failed) {
 				stream->transferred += transferred;
+				stream->transferred_valid = true;
+			}
 			if (front->transferred < front->len)
 				set_bit(NETFS_RREQ_SHORT_TRANSFER, &rreq->flags);
 			notes |= MADE_PROGRESS;
fs/netfs/write_collect.c +8 -2
@@ -254,6 +254,7 @@ reassess_streams:
 			if (front->start + front->transferred > stream->collected_to) {
 				stream->collected_to = front->start + front->transferred;
 				stream->transferred = stream->collected_to - wreq->start;
+				stream->transferred_valid = true;
 				notes |= MADE_PROGRESS;
 			}
 			if (test_bit(NETFS_SREQ_FAILED, &front->flags)) {
@@ -356,6 +357,7 @@ bool netfs_write_collection(struct netfs
 {
 	struct netfs_inode *ictx = netfs_inode(wreq->inode);
 	size_t transferred;
+	bool transferred_valid = false;
 	int s;
 
 	_enter("R=%x", wreq->debug_id);
@@ -376,12 +378,16 @@ bool netfs_write_collection(struct netfs
 			continue;
 		if (!list_empty(&stream->subrequests))
 			return false;
-		if (stream->transferred < transferred)
+		if (stream->transferred_valid &&
+		    stream->transferred < transferred) {
 			transferred = stream->transferred;
+			transferred_valid = true;
+		}
 	}
 
 	/* Okay, declare that all I/O is complete. */
-	wreq->transferred = transferred;
+	if (transferred_valid)
+		wreq->transferred = transferred;
 	trace_netfs_rreq(wreq, netfs_rreq_trace_write_done);
 
 	if (wreq->io_streams[1].active &&
fs/netfs/write_issue.c +2 -2
@@ -118,12 +118,12 @@ struct netfs_io_request *netfs_create_wr
 	wreq->io_streams[0].prepare_write	= ictx->ops->prepare_write;
 	wreq->io_streams[0].issue_write		= ictx->ops->issue_write;
 	wreq->io_streams[0].collected_to	= start;
-	wreq->io_streams[0].transferred		= LONG_MAX;
+	wreq->io_streams[0].transferred		= 0;
 
 	wreq->io_streams[1].stream_nr		= 1;
 	wreq->io_streams[1].source		= NETFS_WRITE_TO_CACHE;
 	wreq->io_streams[1].collected_to	= start;
-	wreq->io_streams[1].transferred		= LONG_MAX;
+	wreq->io_streams[1].transferred		= 0;
 	if (fscache_resources_valid(&wreq->cache_resources)) {
 		wreq->io_streams[1].avail	= true;
 		wreq->io_streams[1].active	= true;
fs/splice.c +3 -0
@@ -739,6 +739,9 @@ iter_file_splice_write(struct pipe_inode
 		sd.pos = kiocb.ki_pos;
 		if (ret <= 0)
 			break;
+		WARN_ONCE(ret > sd.total_len - left,
+			  "Splice Exceeded! ret=%zd tot=%zu left=%zu\n",
+			  ret, sd.total_len, left);
 
 		sd.num_spliced += ret;
 		sd.total_len -= ret;
include/linux/netfs.h +1 -0
@@ -150,6 +150,7 @@ struct netfs_io_stream {
 	bool			active;		/* T if stream is active */
 	bool			need_retry;	/* T if this stream needs retrying */
 	bool			failed;		/* T if this stream failed */
+	bool			transferred_valid; /* T is ->transferred is valid */
 };
 
 /*

Patches currently in stable-queue which might be from [email protected] are

queue-6.16/netfs-fix-unbuffered-write-error-handling.patch