io_uring/net: fix fast_iov assignment in io_setup_async_msg()

PATCH Applied Stefan Metzmacher 3 participants 3 messages Sep 29, 2022 1 message not in the archive

Subsystems (2)

Stable branch io_uring

lore

Share & tools

lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

Stefan Metzmacher

author

2022-09-29 07:39 UTC #1 lore raw

PATCH 1/1 io_uring/net: fix fast_iov assignment in io_setup_async_msg()

I hit a very bad problem during my tests of SENDMSG_ZC. BUG(); in first_iovec_segment() triggered very easily. The problem was io_setup_async_msg() in the partial retry case, which seems to happen more often with _ZC.

iov_iter_iovec_advance() may change i->iov in order to have i->iov_offset being only relative to the first element.

Which means kmsg->msg.msg_iter.iov is no longer the same as kmsg->fast_iov.

But this would rewind the copy to be the start of
async_msg->fast_iov, which means the internal
state of sync_msg->msg.msg_iter is inconsitent.

I tested with 5 vectors with length like this 4, 0, 64, 20, 8388608 and got a short writes with:
- ret=2675244 min_ret=8388692 => remaining 5713448 sr->done_io=2675244
- ret=-EAGAIN => io_uring_poll_arm
- ret=4911225 min_ret=5713448 => remaining 802223 sr->done_io=7586469
- ret=-EAGAIN => io_uring_poll_arm
- ret=802223 min_ret=802223 => res=8388692

While this was easily triggered with SENDMSG_ZC (queued for 6.1),
it was a potential problem starting with 7ba89d2af17aa879dda30f5d5d3f152e587fc551 in 5.18 for IORING_OP_RECVMSG.
And also with 4c3c09439c08b03d9503df0ca4c7619c5842892e in 5.19
for IORING_OP_SENDMSG.

However 257e84a5377fbbc336ff563833a8712619acce56 introduced the critical code into io_setup_async_msg() in 5.11.

Fixes 7ba89d2af17aa ("io_uring: ensure recv and recvmsg handle MSG_WAITALL correctly")

Fixes 257e84a5377fb ("io_uring: refactor sendmsg/recvmsg iov managing")

Cc [email protected]

Signed-off-by Stefan Metzmacher <[email protected]>

---
 io_uring/net.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

io_uring/net.c +4 -2

index 60e392f7f2dc..a81fccd38ae4 100644
@@ -165,8 +165,10 @@ static int io_setup_async_msg(struct io_kiocb *req,
 	memcpy(async_msg, kmsg, sizeof(*kmsg));
 	async_msg->msg.msg_name = &async_msg->addr;
 	/* if were using fast_iov, set it to the new one */
-	if (!async_msg->free_iov)
-		async_msg->msg.msg_iter.iov = async_msg->fast_iov;
+	if (!kmsg->free_iov) {
+		size_t fast_idx = kmsg->msg.msg_iter.iov - kmsg->fast_iov;
+		async_msg->msg.msg_iter.iov = &async_msg->fast_iov[fast_idx];
+	}
 
 	return -EAGAIN;
 }

Signature · 1 line

2.34.1

Pavel Begunkov

Reviewed-by on v1 1/1

2022-09-29 12:55 UTC #2 lore raw

Quoted 4 lines · Stefan Metzmacher

I hit a very bad problem during my tests of SENDMSG_ZC. BUG(); in first_iovec_segment() triggered very easily. The problem was io_setup_async_msg() in the partial retry case, which seems to happen more often with _ZC.

Reviewed-by Pavel Begunkov <[email protected]>

And tested with liburing patches I sent yesterday

Quoted 52 lines

iov_iter_iovec_advance() may change i->iov in order to have i->iov_offset being only relative to the first element.
Which means kmsg->msg.msg_iter.iov is no longer the same as kmsg->fast_iov.
But this would rewind the copy to be the start of
async_msg->fast_iov, which means the internal
state of sync_msg->msg.msg_iter is inconsitent.
I tested with 5 vectors with length like this 4, 0, 64, 20, 8388608 and got a short writes with:
- ret=2675244 min_ret=8388692 => remaining 5713448 sr->done_io=2675244
- ret=-EAGAIN => io_uring_poll_arm
- ret=4911225 min_ret=5713448 => remaining 802223 sr->done_io=7586469
- ret=-EAGAIN => io_uring_poll_arm
- ret=802223 min_ret=802223 => res=8388692
While this was easily triggered with SENDMSG_ZC (queued for 6.1),
it was a potential problem starting with 7ba89d2af17aa879dda30f5d5d3f152e587fc551 in 5.18 for IORING_OP_RECVMSG.
And also with 4c3c09439c08b03d9503df0ca4c7619c5842892e in 5.19
for IORING_OP_SENDMSG.
However 257e84a5377fbbc336ff563833a8712619acce56 introduced the critical code into io_setup_async_msg() in 5.11.
Fixes 7ba89d2af17aa ("io_uring: ensure recv and recvmsg handle MSG_WAITALL correctly")
Fixes 257e84a5377fb ("io_uring: refactor sendmsg/recvmsg iov managing")
Cc [email protected]
Signed-off-by Stefan Metzmacher <[email protected]>
---
  io_uring/net.c | 6 ++++--
  1 file changed, 4 insertions(+), 2 deletions(-)
io_uring/net.c +4 -2
index 60e392f7f2dc..a81fccd38ae4 100644
@@ -165,8 +165,10 @@ static int io_setup_async_msg(struct io_kiocb *req,
  	memcpy(async_msg, kmsg, sizeof(*kmsg));
  	async_msg->msg.msg_name = &async_msg->addr;
  	/* if were using fast_iov, set it to the new one */
-	if (!async_msg->free_iov)
-		async_msg->msg.msg_iter.iov = async_msg->fast_iov;
+	if (!kmsg->free_iov) {
+		size_t fast_idx = kmsg->msg.msg_iter.iov - kmsg->fast_iov;
+		async_msg->msg.msg_iter.iov = &async_msg->fast_iov[fast_idx];
+	}
 
  	return -EAGAIN;
  }

Signature · 1 line

Pavel Begunkov

Jens Axboe

maintainer Applied

2022-09-29 13:13 UTC #3 lore raw

In reply to Stefan Metzmacher

Quoted 9 lines · Stefan Metzmacher

I hit a very bad problem during my tests of SENDMSG_ZC. BUG(); in first_iovec_segment() triggered very easily. The problem was io_setup_async_msg() in the partial retry case, which seems to happen more often with _ZC.
iov_iter_iovec_advance() may change i->iov in order to have i->iov_offset being only relative to the first element.
[...]

Applied, thanks!

[1/1] io_uring/net: fix fast_iov assignment in io_setup_async_msg()

      commit: 3e4cb6ebbb2bad201c1186bc0b7e8cf41dd7f7e6

Best regards,

Signature · 1 line

Jens Axboe