SOCK_NONBLOCK not honored

SOCK_NONBLOCK not honored

[Lavrentiev, Anton (NIH/NLM/NCBI) [C] via cygwin]

Hi,

Looks like CYGWIN defines but does not honor the SOCK_NONBLOCK flag when used with socket(2).

(It also defines SOCK_CLOEXEC but I haven't checked whether it is honored -- full disclosure.)

Consider the following code:

$ cat bug.c
#include <errno.h>
#include <fcntl.h>
#include <poll.h>
#include <unistd.h>
#include <stdio.h>
#include <string.h>
#include <netinet/in.h>
#include <sys/socket.h>


#ifdef BUG
#  define SOCK_TYPE  (SOCK_STREAM | SOCK_NONBLOCK)
#else
#  define SOCK_TYPE   SOCK_STREAM
#endif


int main()
{
    char buf[80];
    ssize_t n;
    struct sockaddr_in in;
    int s = socket(AF_INET, SOCK_TYPE, 0);

#ifndef BUG
    fcntl(s, F_SETFL, O_NONBLOCK);
#endif

    memset(&in, 0, sizeof(in));
    in.sin_family      = AF_INET;
    in.sin_addr.s_addr = htonl(INADDR_LOOPBACK);
    in.sin_port        = htons(6666);

    if (connect(s, (struct sockaddr*) &in, sizeof(in)) < 0) {
        int err;
        socklen_t len = sizeof(err);
        struct pollfd pfd;
        if (errno != EINPROGRESS) {
            printf("connect failed immediately: %m\n");
            return 1;
        }
        memset(&pfd, 0, sizeof(pfd));
        pfd.fd = s;
        pfd.events = POLLOUT;
        if (!poll(&pfd, 1, 1000)) {
            printf("poll timed-out\n");
            return 2;
        }
        if (pfd.revents & (POLLERR | POLLHUP)) {
            printf("poll failed 0x%04X\n", pfd.revents);
            return 3;
        }
        if (getsockopt(s, SOL_SOCKET, SO_ERROR, (void*) &err, &len) != 0) {
            printf("getsockopt failed: %m\n");
            return 4;
        }
        if (err != 0) {
            printf("connect failed: %s\n", strerror(err));
            return 5;
        }
    }
    errno = 0;
    n = recv(s, buf, sizeof(buf), 0);
    printf("Got %zd byte(s): %m\n", n);
    return 0;
}

When the program is built on Linux with and without the macro BUG defined, the result is all the same,
the last recv() correctly finishes with -1 and EAGAIN (run "nc -l 6666" as a server).

$ gcc -Wall -o bug bug.c
$ ./bug
Got -1 byte(s): Resource temporarily unavailable
$ gcc -Wall -DBUG -o bug bug.c
$ ./bug
Got -1 byte(s): Resource temporarily unavailable

However, on CYGWIN when the code is compiled with -DBUG, the last recv() becomes blocking, and the program never exits:

$ gcc -Wall -o bug bug.c
$ ./bug
Got -1 byte(s): Resource temporarily unavailable
$ gcc -Wall -DBUG -o bug bug.c
$ ./bug
...

strace confirms that the recv() has blocked the execution.  Also, connect() appears to be blocking, too, but it's harder to see.

Thanks,
Anton Lavrentiev


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple

Re: SOCK_NONBLOCK not honored

[Corinna Vinschen]

[-- Attachment #1: Type: text/plain, Size: 665 bytes --]

On Nov  1 20:56, Lavrentiev, Anton (NIH/NLM/NCBI) [C] via cygwin wrote:
> Hi,
> 
> Looks like CYGWIN defines but does not honor the SOCK_NONBLOCK flag when used with socket(2).
> 
> (It also defines SOCK_CLOEXEC but I haven't checked whether it is honored -- full disclosure.)
> 
> Consider the following code:

Spot on, thanks for the testcase.  Neither SOCK_NONBLOCK, nor
SOCK_CLOEXEC worked as expected.  What was I thinking at the time...?

I pushed a patch and I'm just uploading new developer snapshots to
https://cygwin.com/snapshots/ while I'm typing.  Please give them a try.


Thanks,
Corinna

-- 
Corinna Vinschen
Cygwin Maintainer

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: SOCK_NONBLOCK not honored

[Marco Atzeri]

Am 05.11.2018 um 21:04 schrieb Corinna Vinschen:
> On Nov  1 20:56, Lavrentiev, Anton (NIH/NLM/NCBI) [C] via cygwin wrote:
>> Hi,
>>
>> Looks like CYGWIN defines but does not honor the SOCK_NONBLOCK flag when used with socket(2).
>>
>> (It also defines SOCK_CLOEXEC but I haven't checked whether it is honored -- full disclosure.)
>>
>> Consider the following code:
>
> Spot on, thanks for the testcase.  Neither SOCK_NONBLOCK, nor
> SOCK_CLOEXEC worked as expected.  What was I thinking at the time...?
>
> I pushed a patch and I'm just uploading new developer snapshots to
> https://cygwin.com/snapshots/ while I'm typing.  Please give them a try.
>
>
> Thanks,
> Corinna
>

It reduces the test failures on libuv from 33 to 8  :-))

Thanks
Marco



---
Diese E-Mail wurde von Avast Antivirus-Software auf Viren geprüft.
https://www.avast.com/antivirus


--
Problem reports:       http://cygwin.com/problems.html
FAQ:                   http://cygwin.com/faq/
Documentation:         http://cygwin.com/docs.html
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple