[PATCH netfs v1 1/2] netfs: preserve prepare_write errno on reissue

[Iván Ezequiel Rodriguez]

prepare_write() may call netfs_prepare_write_failed() (FAILED + errno)
before write_retry calls netfs_reissue_write().  Clearing subreq->error
to 0 for a clean I/O reissue (82e8885bd763) then makes
netfs_do_issue_write() terminate with 0.  The collector copies
front->error into stream->error and mapping_set_error(), which ignores
0, so writeback failure is silent.

Skip clearing error when NETFS_SREQ_FAILED is already set;
do_issue_write() already terminates FAILED subrequests.

Fixes: 82e8885bd763 ("netfs: when subreq is marked for retry, do not check if it faced an error")
Cc: Shyam Prasad N <sprasad@xxxxxxxxxxxxx>
Cc: Steve French <stfrench@xxxxxxxxxxxxx>
Signed-off-by: Iván Ezequiel Rodriguez <ivanrwcm25@xxxxxxxxx>
---
 fs/netfs/write_issue.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/fs/netfs/write_issue.c b/fs/netfs/write_issue.c
index b8237e1314d8..f23364116421 100644
--- a/fs/netfs/write_issue.c
+++ b/fs/netfs/write_issue.c
@@ -268,7 +268,9 @@ void netfs_reissue_write(struct netfs_io_stream *stream,
 	iov_iter_truncate(&subreq->io_iter, size);
 
 	subreq->retry_count++;
-	subreq->error = 0;
+	/* Keep prepare_write() errno; do_issue_write() terminates FAILED. */
+	if (!test_bit(NETFS_SREQ_FAILED, &subreq->flags))
+		subreq->error = 0;
 	__clear_bit(NETFS_SREQ_MADE_PROGRESS, &subreq->flags);
 	__set_bit(NETFS_SREQ_IN_PROGRESS, &subreq->flags);
 	netfs_stat(&netfs_n_wh_retry_write_subreq);
-- 
2.43.0