[PATCH v7 04/19] netfs: don't hang a read whose retry is abandoned

[Paulo Alcantara <pc@manguebit.org>]

When ->prepare_read() fails while retrying a read, or a subrequest can't
be allocated, netfs_retry_read_subrequests() marks the remaining
subrequests failed without reissuing them.  If none was reissued in that
pass, nothing is in progress any more to wake the collector, which
returned after the retry, so an offloaded read (readahead, async direct
I/O) never completes and its folios stay locked.

Wake the collector after abandoning, so it collects the failed
subrequests and completes the request.

Fixes: e2d46f2ec332 ("netfs: Change the read result collector to only use one work item")
Signed-off-by: Paulo Alcantara <pc@manguebit.org>
Cc: stable@vger.kernel.org
---
 fs/netfs/read_retry.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/fs/netfs/read_retry.c b/fs/netfs/read_retry.c
index 8a722995fa9d..bfaaebea12df 100644
--- a/fs/netfs/read_retry.c
+++ b/fs/netfs/read_retry.c
@@ -269,6 +269,7 @@ static void netfs_retry_read_subrequests(struct netfs_io_request *rreq)
 		__set_bit(NETFS_SREQ_FAILED, &subreq->flags);
 		__clear_bit(NETFS_SREQ_NEED_RETRY, &subreq->flags);
 	}
+	netfs_wake_collector(rreq);
 }
 
 /*
-- 
2.56.0