Date: Mon, 10 Sep 2018 17:28:32 +0200 From: Gary Jennejohn <gljennjohn@gmail.com> To: Tamas Szakaly <sghctoma@gmail.com> Cc: freebsd-hackers@freebsd.org Subject: Re: Attempting to receivce zero-length message with recvmsg Message-ID: <20180910172832.2c0a1357@ernst.home> In-Reply-To: <20180910085833.d4py4ladlyqchjvo@pamparam> References: <20180910085833.d4py4ladlyqchjvo@pamparam>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 10 Sep 2018 10:58:33 +0200
Tamas Szakaly <sghctoma@gmail.com> wrote:
> Hi,
>
> I have a question about the recvmsg syscall. According to POSIX, unless
> O_NONBLOCK is set on the socket fd, recvmsg [1] should block until a message
> arrives. However, recvmsg returns immediately with 0, if we are trying to
> receive a 0-byte message from a SOCK_SEQPACKET AF_UNIX socket. Consider the
> following code:
>
> #include <stdio.h>
> #include <sys/socket.h>
>
> int main(int argc, char** argv) {
> int sock[2];
> socketpair(AF_UNIX, SOCK_SEQPACKET, 0, sock);
>
> struct msghdr msghdr = {0};
> int ret = recvmsg(sock[1], &msghdr, 0);
>
> printf("ret=%d, msghdr.msg_flags=0x%08x\n\n", ret, msghdr.msg_flags);
> }
>
> Running this yields this output:
>
> [0x00 socketstuff]$ cc socketpair.c -o socketpair && ./socketpair
> ret=0, msghdr.msg_flags=0x00000000
>
> You can see that recvmsg returns with 0, even though there were no messages
> sent, and neither of the sockets are closed, so it should block indefinitely.
>
> Is this behavior intentional to match the semantics of read [2] (i.e.
> attempting to read zero bytes should be a no-op)?
>
>
> [1] recvmsg: http://pubs.opengroup.org/onlinepubs/9699919799/functions/recvmsg.html
> [2] read: http://pubs.opengroup.org/onlinepubs/9699919799/functions/read.html
>
You have to initalize msghdr.msg_iov and msghde.iov_len, otherwise
the kernel notices there is no place to put a received message and
simply returns 0.
I did that and recvmsg() did not return.
--
Gary Jennejohn
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20180910172832.2c0a1357>