UDP socket issue

Question

UDP socket issue

gwaeber opened this issue 6 years ago · comments

We saw that sometime (randomly) OLSRd2 has a problem (OLSRd2 is running but not working) at launch, cf. error message below:

16:02:58.623 WARN(packet_socket) src-plugins/subsystems/oonf_packet_socket.c 241: Cannot send UDP packet to 224.0.0.109:269: Bad file descriptor (9)

This error persist until the instance of OLSRd2 is killed. If we relaunch OLSRd2 with the same configuration, the problem can reappear or not, seems random.
Last version (commit 3a2dd24) of OLSRd2 was used, this issue also appears in previous versions.

Henning Rogge · Answer 1 · Thu Apr 19 2018 13:21:15 GMT+0800 (China Standard Time)

Strange error message, (posix) documentation says it happens if the socket already has been closed, which should not be the case (unless the interface went down without triggering a netlink message to Olsrd2).

What (Linux-)distribution are you using? I assume that this is a corner-case that I don't know about.

Gilles Waeber · Answer 2 · Thu Apr 19 2018 14:17:05 GMT+0800 (China Standard Time)

We are using several Ubuntu 14.04.5 LTS (3.13.0-144) machines on a VMware cluster.

Henning Rogge · Answer 3 · Thu Apr 26 2018 14:31:42 GMT+0800 (China Standard Time)

Hmm... could you try the following patch? (only compile-tested) Henning

On Thu, Apr 19, 2018 at 8:17 AM Gilles Waeber ***@***.***> wrote: We are using a several Ubuntu 14.04.5 LTS (3.13.0-144) machines on a

VMware cluster.

…

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.

Gilles Waeber · Answer 4 · Thu Apr 26 2018 16:38:07 GMT+0800 (China Standard Time)

Are you referring the commits from 24-25 April or something else as "the following patch" ?

Henning Rogge · Answer 5 · Thu Apr 26 2018 16:59:37 GMT+0800 (China Standard Time)

I tried to attach a file to the mail, maybe github removed it... let me see that I publish it on github in a branch. Henning

…

On Thu, Apr 26, 2018 at 10:38 AM Gilles Waeber ***@***.***> wrote: Are you referring the commits from 24-25 April or something else as "the following patch" ? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <#14 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AG8Yt7Mzft7xBp2KxrZGE9TQeLS4BEnMks5tsYdzgaJpZM4TaJNw> .

Henning Rogge · Answer 6 · Thu Apr 26 2018 17:02:03 GMT+0800 (China Standard Time)

Pushed it to the packet_socket_bug branch. Henning

…

On Thu, Apr 26, 2018 at 10:59 AM Henning Rogge ***@***.***> wrote: I tried to attach a file to the mail, maybe github removed it... let me see that I publish it on github in a branch. Henning On Thu, Apr 26, 2018 at 10:38 AM Gilles Waeber ***@***.***> wrote: > Are you referring the commits from 24-25 April or something else as "the > following patch" ? > > — > You are receiving this because you commented. > Reply to this email directly, view it on GitHub > <#14 (comment)>, or mute > the thread > <https://github.com/notifications/unsubscribe-auth/AG8Yt7Mzft7xBp2KxrZGE9TQeLS4BEnMks5tsYdzgaJpZM4TaJNw> > . >

Gilles Waeber · Answer 7 · Thu Apr 26 2018 20:04:18 GMT+0800 (China Standard Time)

With the patch, the issue appears less often but we still have it from time to time.
This time when the error appears, we see one time the warning Cannot send UDP packet to 224.0.0.109:269: Bad file descriptor (9) at the beginning, then the OLSRd2 process dies.

Henning Rogge · Answer 8 · Thu Apr 26 2018 21:23:17 GMT+0800 (China Standard Time)

Hi, the patch does not prevent the problem, it just reopens the socket when it happens. Henning

On Thu, Apr 26, 2018 at 2:04 PM Gilles Waeber ***@***.***> wrote: With the patch, the issue appears less often but we still have it from

time to time. I'll come back to you when we have something more concrete.

…

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or mute the thread.