io_uring/net: disable partial retries for recvmsg with cmsg
Jens Axboe
maintainerPATCH 1/1 io_uring/net: disable partial retries for recvmsg with cmsg
We cannot sanely handle partial retries for recvmsg if we have cmsg attached. If we don't, then we'd just be overwriting the initial cmsg header on retries. Alternatively we could increment and handle this appropriately, but it doesn't seem worth the complication.
---
io_uring/net.c +2 -1
@@ -788,7 +788,8 @@ int io_recvmsg(struct io_kiocb *req, unsigned int issue_flags) flags = sr->msg_flags;if (force_nonblock) flags |=MSG_DONTWAIT ; -if (flags &MSG_WAITALL ) +/* disable partial retry for recvmsg with cmsg attached */ +if (flags &MSG_WAITALL && !kmsg->controllen) min_ret =iov_iter_count (&kmsg->msg.msg_iter); kmsg->msg.msg_get_inq =1 ;
Signature · 1 line
Jens Axboe
Hi Jens,
Quoted 17 lines
We cannot sanely handle partial retries for recvmsg if we have cmsg attached. If we don't, then we'd just be overwriting the initial cmsg header on retries. Alternatively we could increment and handle this appropriately, but it doesn't seem worth the complication.
Cc [email protected] # 5.10+Reported-by Stefan Metzmacher <[email protected]>Signed-off-by Jens Axboe <[email protected]>---
io_uring/net.c +0
-0
io_uring/net.c int io_recvmsg(struct io_kiocb *req, unsigned int issue_flags) · line 793
Quoted 7 lines
io_uring/net.c +2
-1@@ -788,7 +788,8 @@ int io_recvmsg(struct io_kiocb *req, unsigned int issue_flags) flags = sr->msg_flags;if (force_nonblock) flags |=MSG_DONTWAIT ; -if (flags &MSG_WAITALL ) +/* disable partial retry for recvmsg with cmsg attached */ +if (flags &MSG_WAITALL && !kmsg->controllen) min_ret =iov_iter_count (&kmsg->msg.msg_iter);
Isn't kmsg->controllen only used for REQ_F_APOLL_MULTISHOT?
I guess the correct value would be kmsg->msg.msg_controllen?
Maybe the safest change would be something like this (completely untested!):
io_uring/net.c +2 -2
@@ -781,14 +781,14 @@ int io_recvmsg(struct io_kiocb *req, unsigned int issue_flags) flags = sr->msg_flags;if (force_nonblock) flags |=MSG_DONTWAIT ; -if (flags &MSG_WAITALL ) - min_ret =iov_iter_count (&kmsg->msg.msg_iter); kmsg->msg.msg_get_inq =1 ;if (req->flags &REQ_F_APOLL_MULTISHOT ) ret =io_recvmsg_multishot (sock, sr, kmsg, flags, &mshot_finished);else +if (flags &MSG_WAITALL && !kmsg->msg.msg_controllen) + min_ret =iov_iter_count (&kmsg->msg.msg_iter); ret =__sys_recvmsg_sock (sock, &kmsg->msg, sr->umsg, kmsg->uaddr, flags);
metze
Jens Axboe
maintainerQuoted 29 lines · Stefan Metzmacher
Hi Jens,
Quoted 17 lines
We cannot sanely handle partial retries for recvmsg if we have cmsg attached. If we don't, then we'd just be overwriting the initial cmsg header on retries. Alternatively we could increment and handle this appropriately, but it doesn't seem worth the complication.
Cc [email protected] # 5.10+Reported-by Stefan Metzmacher <[email protected]>Signed-off-by Jens Axboe <[email protected]>---
io_uring/net.c +0
-0io_uring/net.c int io_recvmsg(struct io_kiocb *req, unsigned int issue_flags) · line 793
Quoted 7 lines
io_uring/net.c +2
-1@@ -788,7 +788,8 @@ int io_recvmsg(struct io_kiocb *req, unsigned int issue_flags) flags = sr->msg_flags;if (force_nonblock) flags |=MSG_DONTWAIT ; -if (flags &MSG_WAITALL ) +/* disable partial retry for recvmsg with cmsg attached */ +if (flags &MSG_WAITALL && !kmsg->controllen) min_ret =iov_iter_count (&kmsg->msg.msg_iter);Isn't
kmsg->controllenonly used forREQ_F_APOLL_MULTISHOT?
Quoted 1 line
I guess the correct value would be
kmsg->msg.msg_controllen?
Gah, yes indeed it should!
Quoted 12 lines
Maybe the safest change would be something like this (completely untested!):
io_uring/net.c +0
-2@@ -781,14 +781,14 @@ int io_recvmsg(struct io_kiocb *req, unsigned int issue_flags) flags = sr->msg_flags;if (force_nonblock) flags |=MSG_DONTWAIT ; -if (flags &MSG_WAITALL ) - min_ret =iov_iter_count (&kmsg->msg.msg_iter);
io_uring/net.c int io_recvmsg(struct io_kiocb *req, unsigned int issue_flags) · line 793
Quoted 10 lines
io_uring/net.c +2
-0kmsg->msg.msg_get_inq =1 ;if (req->flags &REQ_F_APOLL_MULTISHOT ) ret =io_recvmsg_multishot (sock, sr, kmsg, flags, &mshot_finished);else +if (flags &MSG_WAITALL && !kmsg->msg.msg_controllen) + min_ret =iov_iter_count (&kmsg->msg.msg_iter); ret =__sys_recvmsg_sock (sock, &kmsg->msg, sr->umsg, kmsg->uaddr, flags);
I like putting it there, as MSG_WAITALL is explicitly disallowed for multishot anyway. The check belongs in that branch rather than as a whole.
Signature · 1 line
Jens Axboe