[PATCH 5.7 118/393] io_uring: fix racy overflow count reporting

From: Greg Kroah-Hartman
Date: Mon Aug 17 2020 - 15:04:13 EST


From: Pavel Begunkov <asml.silence@xxxxxxxxx>

[ Upstream commit b2bd1cf99f3e7c8fbf12ea07af2c6998e1209e25 ]

All ->cq_overflow modifications should be under completion_lock,
otherwise it can report a wrong number to the userspace. Fix it in
io_uring_cancel_files().

Signed-off-by: Pavel Begunkov <asml.silence@xxxxxxxxx>
Signed-off-by: Jens Axboe <axboe@xxxxxxxxx>
Signed-off-by: Sasha Levin <sashal@xxxxxxxxxx>
---
fs/io_uring.c | 3 +--
1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index 159338b5f8263..c212af69c15b4 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -7579,10 +7579,9 @@ static void io_uring_cancel_files(struct io_ring_ctx *ctx,
clear_bit(0, &ctx->sq_check_overflow);
clear_bit(0, &ctx->cq_check_overflow);
}
- spin_unlock_irq(&ctx->completion_lock);
-
WRITE_ONCE(ctx->rings->cq_overflow,
atomic_inc_return(&ctx->cached_cq_overflow));
+ spin_unlock_irq(&ctx->completion_lock);

/*
* Put inflight ref and overflow ref. If that's
--
2.25.1