On Wed, May 11, 2022 at 6:11 PM Constantine Gavrilov <constantine.gavrilov@xxxxxxxxx> wrote: > > On Wed, May 11, 2022 at 5:56 PM Constantine Gavrilov > <constantine.gavrilov@xxxxxxxxx> wrote: > > > > On Wed, May 4, 2022 at 6:55 PM Jens Axboe <axboe@xxxxxxxxx> wrote: > > > > > > On 5/4/22 9:28 AM, Jens Axboe wrote: > > > > On 5/4/22 9:21 AM, Constantine Gavrilov wrote: > > > >> On Wed, May 4, 2022 at 4:54 PM Jens Axboe <axboe@xxxxxxxxx> wrote: > > > >>> > > > >>> On 5/3/22 5:05 PM, Constantine Gavrilov wrote: > > > >>>> Jens: > > > >>>> > > > >>>> This is related to the previous thread "Fix MSG_WAITALL for > > > >>>> IORING_OP_RECV/RECVMSG". > > > >>>> > > > >>>> We have a similar issue with TCP socket sends. I see short sends > > > >>>> regarding of the method (I tried write, writev, send, and sendmsg > > > >>>> opcodes, while using MSG_WAITALL for send and sendmsg). It does not > > > >>>> make a difference. > > > >>>> > > > >>>> Most of the time, sends are not short, and I never saw short sends > > > >>>> with loopback and my app. But on real network media, I see short > > > >>>> sends. > > > >>>> > > > >>>> This is a real problem, since because of this it is not possible to > > > >>>> implement queue size of > 1 on a TCP socket, which limits the benefit > > > >>>> of IORING. When we have a short send, the next send in queue will > > > >>>> "corrupt" the stream. > > > >>>> > > > >>>> Can we have complete send before it completes, unless the socket is > > > >>>> disconnected? > > > >>> > > > >>> I'm guessing that this happens because we get a task_work item queued > > > >>> after we've processed some of the send, but not all. What kernel are you > > > >>> using? > > > >>> > > > >>> This: > > > >>> > > > >>> https://git.kernel.dk/cgit/linux-block/commit/?h=for-5.19/io_uring&id=4c3c09439c08b03d9503df0ca4c7619c5842892e > > > >>> > > > >>> is queued up for 5.19, would be worth trying. > > > >>> > > > >>> -- > > > >>> Jens Axboe > > > >>> > > > >> > > > >> Jens: > > > >> > > > >> Thank you for your reply. > > > >> > > > >> The kernel is 5.17.4-200.fc35.x86_64. I have looked at the patch. With > > > >> the solution in place, I am wondering whether it will be possible to > > > >> use multiple uring send IOs on the same socket. I expect that Linux > > > >> TCP will serialize multiple send operations on the same socket. I am > > > >> not sure it happens with uring (meaning that socket is blocked for > > > >> processing a new IO until the pending IO completes). Do I need > > > >> IOSQE_IO_DRAIN / IOSQE_IO_LINK for this to work? Would not be optimal > > > >> because of multiple different sockets in the same uring. While I > > > >> already have a workaround in the form of a "software" queue for > > > >> streaming data on TCP sockets, I would rather have kernel to do > > > >> "native" queueing in sockets layer, and have exrtra CPU cycles > > > >> available to the application. > > > > > > > > The patch above will mess with ordering potentially. If the cause is as > > > > I suspect, task_work causing it to think it's signaled, then the better > > > > approach may indeed be to just flush that work and retry without > > > > re-queueing the current one. I can try a patch against 5.18 if you are > > > > willing and able to test? > > > > > > You can try something like this, if you run my for-5.19/io_uring branch. > > > I'd be curious to know if this solves the short send issue for you. > > > > > > > > > diff --git a/fs/io_uring.c b/fs/io_uring.c > > > index f6b6db216478..b835e80be1fa 100644 > > > --- a/fs/io_uring.c > > > +++ b/fs/io_uring.c > > > @@ -5684,6 +5684,7 @@ static int io_sendmsg(struct io_kiocb *req, unsigned int issue_flags) > > > if (flags & MSG_WAITALL) > > > min_ret = iov_iter_count(&kmsg->msg.msg_iter); > > > > > > +retry: > > > ret = __sys_sendmsg_sock(sock, &kmsg->msg, flags); > > > > > > if (ret < min_ret) { > > > @@ -5694,6 +5695,8 @@ static int io_sendmsg(struct io_kiocb *req, unsigned int issue_flags) > > > if (ret > 0 && io_net_retry(sock, flags)) { > > > sr->done_io += ret; > > > req->flags |= REQ_F_PARTIAL_IO; > > > + if (io_run_task_work()) > > > + goto retry; > > > return io_setup_async_msg(req, kmsg); > > > } > > > req_set_fail(req); > > > @@ -5744,6 +5747,7 @@ static int io_send(struct io_kiocb *req, unsigned int issue_flags) > > > min_ret = iov_iter_count(&msg.msg_iter); > > > > > > msg.msg_flags = flags; > > > +retry: > > > ret = sock_sendmsg(sock, &msg); > > > if (ret < min_ret) { > > > if (ret == -EAGAIN && (issue_flags & IO_URING_F_NONBLOCK)) > > > @@ -5755,6 +5759,8 @@ static int io_send(struct io_kiocb *req, unsigned int issue_flags) > > > sr->buf += ret; > > > sr->done_io += ret; > > > req->flags |= REQ_F_PARTIAL_IO; > > > + if (io_run_task_work()) > > > + goto retry; > > > return -EAGAIN; > > > } > > > req_set_fail(req); > > > > > > -- > > > Jens Axboe > > > > > > > Jens: > > > > I was able to test the first change on the top of Linus kernel git (5.18.0-rc6). > > > > I do not get short sends anymore, but I get corruption in sent > > packets (corruption is detected by the receiver). It looks like short > > sends handled by the patch intermix data from multiple send SQEs in > > the stream, so ordering of multiple SQEs in URING becomes broken. > > > > To test it, I had two implementations of the send functions: > > 1. Uses SEND opcode, asserts on short sends. No asserts but data corruption. > > 2. Uses TCP send queue implementation (internally uses SEND and > > SENDMSG opcodes in URING, only one pending send at a time, and tail of > > the short sends is resent until all data is sent). This always works. > > > > I would like to test the second patch now. Is it on the top of the > > first patch or by itself? Do I really need your tree for that? If yes, > > can you send me the git pull info, please? > > > > -- > > ---------------------------------------- > > Constantine Gavrilov > > Storage Architect > > Master Inventor > > Tel-Aviv IBM Storage Lab > > 1 Azrieli Center, Tel-Aviv > > ---------------------------------------- > > Jens: for git branch, is it under > git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux-block.git? > > -- > ---------------------------------------- > Constantine Gavrilov > Storage Architect > Master Inventor > Tel-Aviv IBM Storage Lab > 1 Azrieli Center, Tel-Aviv > ---------------------------------------- Jens: checked out yout branch, the first patch is already in, applied the second. Will build and test now. -- ---------------------------------------- Constantine Gavrilov Storage Architect Master Inventor Tel-Aviv IBM Storage Lab 1 Azrieli Center, Tel-Aviv ----------------------------------------
