Gentoo Websites Logo
Go to: Gentoo Home Documentation Forums Lists Bugs Planet Store Wiki Get Gentoo!
Bug 409905 - <x11-drivers/nvidia-drivers-295-33 - segmentation faults in libGL.so
Summary: <x11-drivers/nvidia-drivers-295-33 - segmentation faults in libGL.so
Status: RESOLVED FIXED
Alias: None
Product: Gentoo Linux
Classification: Unclassified
Component: [OLD] Core system (show other bugs)
Hardware: All Linux
: Normal normal (vote)
Assignee: Doug Goldstein (RETIRED)
URL: http://sourceware.org/git/?p=glibc.gi...
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2012-03-27 18:03 UTC by Alexander Holler
Modified: 2012-04-13 00:34 UTC (History)
3 users (show)

See Also:
Package list:
Runtime testing required: ---


Attachments

Note You need to log in before you can comment on or make changes to this bug.
Description Alexander Holler 2012-03-27 18:03:12 UTC
If nvidia-drivers < 295-33 are installed, almost everything which uses libGL.so segfaults.

There are two bug descriptions:

https://bugzilla.redhat.com/show_bug.cgi?id=737223

https://bugs.launchpad.net/ubuntu/+source/nvidia-graphics-drivers/+bug/929384

The latter one says it will be fixed with a newer version of nvidia-drivers.

Here it works (x86) with 295.33 but segfaults with 295.20.


Reproducible: Always
Comment 1 Jeroen Roovers (RETIRED) gentoo-dev 2012-03-27 18:12:28 UTC

*** This bug has been marked as a duplicate of bug 334325 ***
Comment 2 Jeroen Roovers (RETIRED) gentoo-dev 2012-03-27 18:18:10 UTC
Please post your `emerge --info' output in a comment.
Comment 3 Alexander Holler 2012-03-27 18:27:02 UTC
295.33 does NOT, I repeat NOT segfault.

Everything <295.33 segfaults with glibc-2.15.

So the changed summary is totally wrong and misleading.

Here is the summary I've entered:

sys-libs/glibc-2.15 should block nvidia-drivers < 295-33

I will not change it back, because I don't like to play ping-pong.
Comment 4 Jeroen Roovers (RETIRED) gentoo-dev 2012-03-27 22:57:36 UTC
(In reply to comment #3)
> Everything <295.33 segfaults with glibc-2.15.

Sorry, that bit slipped.

> Here is the summary I've entered:
> 
> sys-libs/glibc-2.15 should block nvidia-drivers < 295-33

It's ridiculous.

> I will not change it back, because I don't like to play ping-pong.

Thanks.

Now please answer comment #2.
Comment 5 Jeroen Roovers (RETIRED) gentoo-dev 2012-03-28 00:24:59 UTC
Also, please post a gdb backtrace of a program segfaulting like that.
Comment 6 Alexander Holler 2012-03-29 15:40:11 UTC
If the summary was ridiculus, change the logic and place blockers for >=glibc-2.15 in all nvidia-drivers < 295.33.

And add glibc-2.15 to the topic.
Comment 7 Doug Goldstein (RETIRED) gentoo-dev 2012-03-30 03:02:13 UTC
No need to attach further information to this bug. I'm pretty familiar with the issue. The question is, for users of older cards do we watch to apply the same patch that other distros use for glib 2.15?
Comment 8 Doug Goldstein (RETIRED) gentoo-dev 2012-03-30 03:02:53 UTC
FWIW, I've attached the link to the patch applied, Mike. Take a peek and give me a yay or nay.
Comment 9 SpanKY gentoo-dev 2012-03-30 03:17:24 UTC
i wasn't interested in adding the hack for broken binary code.  someone else pointed it out and i passed on it.

drop the broken drivers and be done.
Comment 10 devt 2012-04-06 19:12:23 UTC
English
--
Aw mans... Sorry for may bad english, I'm from Ukraine. 
Baad, very bad, that you have ignored the Alexander Holler's message. Jeroen Roovers, did You mean, that problem in the nvidia-drivers? Ohoho... No. Problem in GLIBC-2.15. 
I've created an experiment.
When I reinstalled Gentoo fully, from stage3, with glibc-2.15, I've got a message from syslog:
First tty: 'tail -f /var/log/messages'
Second tty: 'eselect opengl set nvidia'. 
I've tried: '/usr/bin/gnome-session'
Message from first tty:
Mar 31 21:15:40 devt kernel: [ 7154.423076] gnome-session[15127]: segfault at ffffe0b4 ip b6668406 sp bff45fd0 error 4 in libGL.so.295.20[b65e8000+aa000][3~

I've tried: '/usr/bin/gnome-shell'
Message from first tty:
 Mar 31 21:16:28 devt kernel: [ 7202.522155] gnome-shell[15140]: segfault at ffffe0b4 ip b5fe9406 sp bf95d6c0 error 4 in libGL.so.295.20[b5f69000+aa000]

I've tried: '/usr/bin/gnome-panel'
Message from first tty:
Mar 31 21:16:53 devt kernel: [ 7227.394335] gnome-panel[15157]: segfault at ffffe0b4 ip b648b406 sp bfd77490 error 4 in libGL.so.295.20[b640b000+aa000]

But when I've tried to emerge nvidia-drivers-295.33, I have no problems with it too, but... This version of the drivers has been masked. But SpanKY, why do You recommend masked drivers with the message in 'package.use' file from portage profile path:
# Doug Goldstein <cardoe@gentoo.org> (23 Mar 2012)
# Fixing patch issue
?

Okay. Did you know, that nvidia-drivers with glibc-2.15 caused the kernel panic? No one reports has not been saved in any time of kernel panic, no one logs, but I know: when I've send 'startx' or '/etc/init.d/xdm start' - it's (and glibc) - reason for the kernel panic.

First (repeat):
When I've tried to emerge nvidia-drivers-295.33, I have no problems with it.
Second:
When I've tried 'eselect opengl set xorg-x11' - I have no problems with it, but I have no 3D acceleration.
Third:
When I've tried reinstall gentoo (FOUR lost days of re-emerging system and world with glibc-2.14 - system with glibc-2.15 has been removed!) with glibc-2.14 - I have no problems with all nvidia-drivers: 275.43, 290.10, 295.20 and unmasked 295.33.

 What's the conclusion?
 If everything with drivers <295.33 segfaults with glibc-2.15, but all drivers works with glibc-2.14, problem not in nvidia-drivers, right? I know, You're can say me, why I've unmasked glibc-2.15. My answer: I like tests. Many tests. But I'm know english very bad - I can't to report all bugs, because 50% of my bugs will be ignored. Then what sense to send reports?

Also I'm unable to see too. Why bug #334325 is a duplicate of this one. Really, why?
Do not think this message for brutality. I've lost many nerves for searching reasons of this bug. You all can understand me and my situation.

Русский
--
Эх, ребята... Простите за мой кривой английский, я с Украины.
Плохо, очень плохо, что вы игнорировали сообщения от Alexander Holler. Jeroen Roovers, вы думали, что проблема в драйверах от nvidia? Нет, проблема в GLIBC-2.15.
Я провёл эксперимент.
Когда я полностью переустановил Gentoo c третьей стадии, я получал сообщения от syslog.
Первый полноэкранный терминал tty: 'tail -f /var/log/messages'
Второй полноэкранный терминал tty: '/usr/bin/gnome-session'
Сообщения с первого терминала:
Mar 31 21:15:40 devt kernel: [ 7154.423076] gnome-session[15127]: segfault at ffffe0b4 ip b6668406 sp bff45fd0 error 4 in libGL.so.295.20[b65e8000+aa000][3~

Я попробовал '/usr/bin/gnome-shell':
Сообщения снова:
 Mar 31 21:16:28 devt kernel: [ 7202.522155] gnome-shell[15140]: segfault at ffffe0b4 ip b5fe9406 sp bf95d6c0 error 4 in libGL.so.295.20[b5f69000+aa000]

Я попробовал '/usr/bin/gnome-panel'
Сообщения:
Mar 31 21:16:53 devt kernel: [ 7227.394335] gnome-panel[15157]: segfault at ffffe0b4 ip b648b406 sp bfd77490 error 4 in libGL.so.295.20[b640b000+aa000]

Но когда я попробовал установить nvidia-drivers-295.33, эти проблемы исчезли, но... Эта версия драйверов замаскирована. SpanKY, тогда почему Вы рекомендуете установить замаскированные драйверы с описанной причиной в файле 'package.use' с профиля portage? Вот сообщение оттуда:
# Doug Goldstein <cardoe@gentoo.org> (23 Mar 2012)
# Fixing patch issue

Хорошо. А Вы знаете, что драйверы с установленной библиотекой glibc-2.15 вызывали панику ядра? При этом ни один отчет не сохраняется, никакие логи тоже, но я знаю точно: когда я отправлял команду на исполнение - 'startx' или '/etc/init.d/xdm start' - это (и glibc) - причина паники ядра.

Первое (повторяю):
Когда я пробовал установить драйверы версии 295.33 - проблема исчезала.
Второе: когда я пробовал 'eselect opengl set xorg-x11' - проблема исчезала.
Третье: когда я пробовал переустанавливать gentoo (ЧЕТЫРЕ потерянных дней и нервов на переустановку системы с glibc-2.14, система с установленной библиотекой glibc-2.15 была полностью удалена!) с библиотекой glibc-2.14 - проблемы исчезали с любой версией драйверов, начиная от 275.43 и заканчивая 295.33

Следовательно, какие выводы?
Если "сегфолтится" что угодно с драйверами версией ниже 295.33 c glibc-2.15, но всё начинает работать с glibc-2.14 - то проблема явно не в драйверах, верно? Я знаю, Вы можете сказать мне: "зачем Вы размаскировали glibc-2.15?". Мой ответ: я люблю тесты. Много тестов. Но я очень плохо знаю английский, поэтому я не могу отправлять все встречающиеся мной баги, ибо 50% из них точно будет проигнорированы. Тогда какой мне смысл вообще отправлять отчеты?

Кстати: я тоже не могу понять, почему баг под номером 334325 - дубликат этого бага. Действительно, почему? Не сочтите данное сообщение за грубость. Я потерял много нервов в поисках причины данного бага.
Comment 11 Doug Goldstein (RETIRED) gentoo-dev 2012-04-11 17:52:38 UTC
I've fixed the depends on glib appropriately and since we're not going to back port the glibc fix like Fedora and Ubuntu, there's nothing further we can do.
Comment 12 Alexander Holler 2012-04-13 00:34:06 UTC
Thats fine and I think it's a better way than to add a "Horrible workaround for horribly broken software". ;)

Thanks a lot.