[Fixed] Chromium segfaults with glibc 2.15?

sleekslush · 2012-01-18 22:21:30

Ditto on the 64-bit. Chromium *sometimes* crashes. If I delete my local ~/.config/chromium it works again (obviously a terrible solution).

Here's the last output from strace:

stat("/home/craig/.cache/chromium/Default", {st_mode=S_IFDIR|0700, st_size=4096, ...}) = 0
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
write(45, "\0", 1)                      = 1
futex(0x7f7df6270fa0, FUTEX_WAKE_PRIVATE, 1) = 1
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid()                                = 25904
gettid( <unfinished ...>
+++ killed by SIGSEGV +++

Last edited by sleekslush (2012-01-18 22:29:08)

DarkSniper · 2012-01-18 23:29:50

Tried a clean profile as sleekslush mentioned, however, it still segfaulted:

[dark@Night ~]$ mv ~/.config/chromium ~/.config/chromium.backup && mv ~/.cache/chromium ~/.cache/chromium.backup
[dark@Night ~]$ chromium
[dark@Night ~]$ chromium
Segmentation fault
[dark@Night ~]$ dmesg | tail -n 1
[68291.251752] chromium[8530]: segfault at 17 ip 00007fe160d513ae sp 00007fe166c323e0 error 6 in libc-2.15.so[7fe160c3c000+199000]

sleekslush · 2012-01-18 23:32:48

DarkSniper wrote:

Tried a clean profile as sleekslush mentioned, however, it still segfaulted:
[dark@Night ~]$ chromium
[dark@Night ~]$ chromium
Segmentation fault

I'm glad you posted that because it's basically what is happening to me now too. After clearing, Chrome runs fine the first time. After I close and start up again I get a segfault.

skunktrader · 2012-01-18 23:44:13

Thanks to this post, after setting up and starting nscd chromium doesn't segfault for me anymore at startup.

sleekslush · 2012-01-18 23:52:36

skunktrader wrote:

Thanks to this post, after setting up and starting nscd chromium doesn't segfault for me anymore at startup.

I'm doing this now and so far so good. Will report back if I see failures. Would also be good to note why this might be required if it really is the solution to the issue

DarkSniper · 2012-01-18 23:57:32

Just setup nscd with the default options, and so far, no segfaults here.
Trying a loop to start and kill chromium over and over again to see if any segfaults happen.

Edit: Was able to open/close it 20 times in a row without segfaulting, I think it's safe to say that this works.

Edit2:
After reading the nscd.conf manpage, and some experimentation, I've come up with a minimal nscd.conf to only enable the cache that fixes the segfault.

#
# /etc/nscd.conf
#

# General Settings
	threads			1
	max-threads		4
	debug-level		0
	reload-count		5
	paranoia		no

# Disable Caches that aren't needed for Chromium
	enable-cache		passwd		no
	enable-cache		group		no
	enable-cache		services	no
	enable-cache		netgroup	no

# This one fixes the segfaulting issue
	enable-cache		hosts		yes
	positive-time-to-live	hosts		900
	negative-time-to-live	hosts		30
	suggested-size		hosts		211
	check-files		hosts		yes
	persistent		hosts		yes
	shared			hosts		yes
	max-db-size		hosts		33554432

Last edited by DarkSniper (2012-01-19 00:23:13)

jryan · 2012-01-19 00:07:36

sleekslush wrote:

skunktrader wrote:
Thanks to this post, after setting up and starting nscd chromium doesn't segfault for me anymore at startup.
I'm doing this now and so far so good. Will report back if I see failures. Would also be good to note why this might be required if it really is the solution to the issue

It must have something to do with glibc's namespace lookup functions. Is anyone else besides me running behind an http_proxy?

sleekslush · 2012-01-19 00:27:46

DarkSniper wrote:

Edit2:
After reading the nscd.conf manpage, and some experimentation, I've come up with a minimal nscd.conf to only enable the cache that fixes the segfault.

#
# /etc/nscd.conf
#

# General Settings
	threads			1
	max-threads		4
	debug-level		0
	reload-count		5
	paranoia		no

# Disable Caches that aren't needed for Chromium
	enable-cache		passwd		no
	enable-cache		group		no
	enable-cache		services	no
	enable-cache		netgroup	no

# This one fixes the segfaulting issue
	enable-cache		hosts		yes
	positive-time-to-live	hosts		900
	negative-time-to-live	hosts		30
	suggested-size		hosts		211
	check-files		hosts		yes
	persistent		hosts		yes
	shared			hosts		yes
	max-db-size		hosts		33554432

These settings are also working well for me.

derloy · 2012-01-22 16:39:24

unexist wrote:

Same problem here, chromium usually refuses to start after boot. When I try to start chromium several times it works after e.g. the eighth try. Reinstalling chromium with pacman works as well. 64bit here as well.
Is there any bugreport?

http://sourceware.org/bugzilla/show_bug.cgi?id=13594

Not much we can do besides using nscd until this gets fixed.

Allan · 2012-01-22 20:56:39

You could file a bug report in the Arch bug tracker... the commit causing this can easily be reverted but bug do not exist when only reported on the forums.

ahadzic · 2012-01-27 21:17:33

Hmm. I'm running nscd but I'm still having the same issue.

Backtrace:

#0  0xb6f6ff95 in __nscd_get_mapping () from /lib/libc.so.6
#1  0xb6f701b3 in __nscd_get_map_ref () from /lib/libc.so.6
#2  0xb6f6d590 in nscd_gethst_r () from /lib/libc.so.6
#3  0xb6f6de73 in __nscd_gethostbyname2_r () from /lib/libc.so.6
#4  0xb6f528e0 in gethostbyname2_r@@GLIBC_2.1.2 () from /lib/libc.so.6
#5  0xb6f224e9 in gaih_inet () from /lib/libc.so.6
#6  0xb6f2373d in getaddrinfo () from /lib/libc.so.6
#7  0x80f60683 in ?? ()
#8  0x80f59fc0 in ?? ()
#9  0x80f52f3f in ?? ()
#10 0x80cba803 in ?? ()
#11 0x80c86372 in ?? ()
#12 0x82784c13 in ?? ()
#13 0x80cbb10c in ?? ()
#14 0xb7403d4c in start_thread () from /lib/libpthread.so.0
#15 0xb6f3b03e in clone () from /lib/libc.so.6

Any workarounds or should I file a bug on the tracker ?

Last edited by ahadzic (2012-01-27 21:18:18)

sleekslush · 2012-01-27 21:24:43

Running nscd has been 100% successful for me, so it's possible that you've encountered a different issue?

ahadzic · 2012-01-27 21:29:21

It's possible. I have confirmed that I'm in fact running nscd. It's on the list of running processes. Is there anything else that I might have missed ? Some configuration properties for nscd or something like that?

sleekslush · 2012-01-27 22:10:42

I used whatever the default settings were. Literally all I did was put @nscd into my DAEMONS list

skunktrader · 2012-01-27 23:08:39

ahadzic wrote:

It's possible. I have confirmed that I'm in fact running nscd. It's on the list of running processes. Is there anything else that I might have missed ? Some configuration properties for nscd or something like that?

Did you create /var/run/nscd and /var/db/nscd

Allan · 2012-01-28 00:18:24

ahadzic wrote:

Any workarounds or should I file a bug on the tracker ?

Allan wrote:

You could file a bug report in the Arch bug tracker... the commit causing this can easily be reverted but bug do not exist when only reported on the forums.

DarkSniper · 2012-02-07 18:26:48

This is fixed in glibc 2.15-5, Chromium no longer segfaults without nscd (hooray for requiring one less daemon!):

upgpkg: glibc 2.15-5
fix x86_64 ifunc relocations, fix avx detection, fix gb18030 charmap, fix crashes in __nscd_get_mapping if nscd not running, fix realloc usage in vfscanf

Allan · 2012-02-07 19:00:07

And i would have fixed it a few eeks earlier if someone actually filed a bug report...

ajaxas · 2012-02-10 16:34:46

It seems like this stopped with glibc-2.15-5.
Anybody to confirm this?

Oops. Stupid me, sorry

Last edited by ajaxas (2012-02-10 16:36:08)

Arch Linux

#26 2012-01-18 22:21:30

Re: [Fixed] Chromium segfaults with glibc 2.15?

#27 2012-01-18 23:29:50

Re: [Fixed] Chromium segfaults with glibc 2.15?

#28 2012-01-18 23:32:48

Re: [Fixed] Chromium segfaults with glibc 2.15?

#29 2012-01-18 23:44:13

Re: [Fixed] Chromium segfaults with glibc 2.15?

#30 2012-01-18 23:52:36

Re: [Fixed] Chromium segfaults with glibc 2.15?

#31 2012-01-18 23:57:32

Re: [Fixed] Chromium segfaults with glibc 2.15?

#32 2012-01-19 00:07:36

Re: [Fixed] Chromium segfaults with glibc 2.15?

#33 2012-01-19 00:27:46

Re: [Fixed] Chromium segfaults with glibc 2.15?

#34 2012-01-22 16:39:24

Re: [Fixed] Chromium segfaults with glibc 2.15?

#35 2012-01-22 20:56:39

Re: [Fixed] Chromium segfaults with glibc 2.15?

#36 2012-01-27 21:17:33

Re: [Fixed] Chromium segfaults with glibc 2.15?

#37 2012-01-27 21:24:43

Re: [Fixed] Chromium segfaults with glibc 2.15?

#38 2012-01-27 21:29:21

Re: [Fixed] Chromium segfaults with glibc 2.15?

#39 2012-01-27 22:10:42

Re: [Fixed] Chromium segfaults with glibc 2.15?

#40 2012-01-27 23:08:39

Re: [Fixed] Chromium segfaults with glibc 2.15?

#41 2012-01-28 00:18:24

Re: [Fixed] Chromium segfaults with glibc 2.15?

#42 2012-02-07 18:26:48

Re: [Fixed] Chromium segfaults with glibc 2.15?

#43 2012-02-07 19:00:07

Re: [Fixed] Chromium segfaults with glibc 2.15?

#44 2012-02-10 16:34:46

Re: [Fixed] Chromium segfaults with glibc 2.15?

Board footer