Debian Bug report logs - #629184
fifo translators don't die

Toggle useless messages

Message #5 received at submit@bugs.debian.org (full text, mbox, reply):

From: Gabriele Giacone <1o5g4r8o@gmail.com>

To: Debian Bug Tracking System <submit@bugs.debian.org>

Subject: file_pager_write_page: Assertion `block' failed

Date: Sat, 4 Jun 2011 12:32:34 +0200

Package: hurd
Severity: important

ext2fs: device:hd0s1: warning: FILESYSTEM NOT UNMOUNTED CLEANLY; PLEASE fsck
ext2fs:
/home/buildd/build/chroot-sid/home/buildd/byhand/hurd/./ext2fs/pager.c:399:
file_pager_write_page: Assertion `block' failed

Root filesystem become unresponsive, machine unusable.
Kvm vm with ram 1GB, disk 10GB logical volume, installed with netinst.iso (30-May-2011 17:41 129M).

Reproducible, it happens during every gnash build. Last one during autoconf run.
http://www.gnashdev.org:8010/builders/sid-hurd-i386/builds/20/steps/autogen/logs/stdio

Message #10 received at 629184@bugs.debian.org (full text, mbox, reply):

From: Gabriele Giacone <1o5g4r8o@gmail.com>

To: 629184@bugs.debian.org

Cc: control@bugs.debian.org

Subject: Re: Bug#629184: file_pager_write_page: Assertion `block' failed

Date: Sun, 24 Jul 2011 22:16:16 +0200

retitle 629184 fifo translators don't die
thanks

To reproduce, create and use 1024 names pipes. System will crash with failed
assertion in question or ipc_task_init panic.
e.g. You might reproduce it by rebuilding gnash package with --with-swfdec-testsuite
configure option. It'd create, use and remove many fifos but translators would remain
present until limit is reached.

Workaround (thanks braunr): rebuild gnumach increasing TASK_MAX. 2048 will
suffice to get a complete build. Bigger values won't hurt.

Message #17 received at 629184@bugs.debian.org (full text, mbox, reply):

From: Thomas Schwinge <thomas@schwinge.name>

To: Gabriele Giacone <1o5g4r8o@gmail.com>, 629184@bugs.debian.org

Subject: Re: Bug#629184: file_pager_write_page: Assertion `block' failed

Date: Mon, 25 Jul 2011 10:17:34 +0200

[Message part 1 (text/plain, inline)]

Hi!

On Sun, 24 Jul 2011 22:16:16 +0200, Gabriele Giacone <1o5g4r8o@gmail.com> wrote:
> retitle 629184 fifo translators don't die

> To reproduce, create and use 1024 names pipes. System will crash with failed
> assertion in question or ipc_task_init panic.
> e.g. You might reproduce it by rebuilding gnash package with --with-swfdec-testsuite
> configure option.

Ouch.  But at least you finally figured out what's going on.

> It'd create, use and remove many fifos but translators would remain
> present until limit is reached.

<https://savannah.gnu.org/bugs/?17128>

> Workaround (thanks braunr): rebuild gnumach increasing TASK_MAX. 2048 will
> suffice to get a complete build. Bigger values won't hurt.

Obviously, the artificial limit should go (but that may be nontrivial).
(And it's only one of several.)

And, a fifo process should terminate as soon as it is not ``reachable''
anymore (no senders notification?).


Grüße,
 Thomas

[Message part 2 (application/pgp-signature, inline)]

Message #22 received at 629184@bugs.debian.org (full text, mbox, reply):

From: Samuel Thibault <sthibault@debian.org>

To: Gabriele Giacone <1o5g4r8o@gmail.com>, 629184@bugs.debian.org

Subject: Re: Bug#629184: file_pager_write_page: Assertion `block' failed

Date: Mon, 22 Aug 2011 22:57:37 +0200

Hello,

Gabriele Giacone, le Sun 24 Jul 2011 22:16:16 +0200, a écrit :
> To reproduce, create and use 1024 names pipes. System will crash with failed
> assertion in question or ipc_task_init panic.

I have fixed gnumach into not panicking just for this.

> retitle 629184 fifo translators don't die

The fifo translators not dieing is still an issue.

Samuel

Message #27 received at 629184@bugs.debian.org (full text, mbox, reply):

From: Gabriele Giacone <1o5g4r8o@gmail.com>

To: Samuel Thibault <sthibault@debian.org>, 629184@bugs.debian.org

Subject: Re: Bug#629184: file_pager_write_page: Assertion `block' failed

Date: Sun, 28 Jul 2013 13:18:22 +0200

On Mon, Aug 22, 2011 at 10:57:37PM +0200, Samuel Thibault wrote:
> The fifo translators not dieing is still an issue.

I can't reproduce it anymore. Creating, using and removing fifos don't leave stale process around anymore. It's been fixed somehow.

Send a report that this bug log contains spam.