Commit 296abc0d authored Apr 28, 2022 by Andreas Gruenbacher

gfs2: No short reads or writes upon glock contention

Commit 00bfe02f ("gfs2: Fix mmap + page fault deadlocks for buffered
I/O") changed gfs2_file_read_iter() and gfs2_file_buffered_write() to
allow dropping the inode glock while faulting in user buffers. When the
lock was dropped, a short result was returned to indicate that the
operation was interrupted.

As pointed out by Linus (see the link below), this behavior is broken
and the operations should always re-acquire the inode glock and resume
the operation instead.

Link: https://lore.kernel.org/lkml/CAHk-=whaz-g_nOOoo8RRiWNjnv2R+h6_xk2F1J4TuSRxk1MtLw@mail.gmail.com/

Fixes: 00bfe02f ("gfs2: Fix mmap + page fault deadlocks for buffered I/O")
Signed-off-by: Andreas Gruenbacher <agruenba@redhat.com>

parent 4fad37d5

fs/gfs2/file.c

+0 −4

Original line number	Diff line number	Diff line
		@@ -991,8 +991,6 @@ static ssize_t gfs2_file_read_iter(struct kiocb iocb, struct iov_iter to)
		if (leftover != window_size) {
		if (gfs2_holder_queued(&gh))
		goto retry_under_glock;
		if (written)
		goto out_uninit;
		goto retry;
		}
		}
		@@ -1069,8 +1067,6 @@ static ssize_t gfs2_file_buffered_write(struct kiocb *iocb,
		from->count = min(from->count, window_size - leftover);
		if (gfs2_holder_queued(gh))
		goto retry_under_glock;
		if (read && !(iocb->ki_flags & IOCB_DIRECT))
		goto out_uninit;
		goto retry;
		}
		}