Server Linux server crashed - SIGSEGV (segmentation fault)

Killing · Mar 14, 2006

Same here, doesnt look like this was tested very well

Code:

Signal: SIGSEGV [segmentation fault]
Aborting.
Crash information will be saved to your logfile.
Developer Backtrace:
[ 1]  ./ucc-bin-real [0x86783ba]
[ 2]  /lib/libpthread.so.0 [0x18a039ad]
[ 3]  [0xbfbfffbf]
[ 4]  ./libsteamvalidateuseridtickets_i486.so(__is_pointer__FPv+0x43) [0x18b6b903]
[ 5]  ./libsteamvalidateuseridtickets_i486.so(__cp_pop_exception+0x83) [0x18b6aa43]
[ 6]  ./libsteamvalidateuseridtickets_i486.so(InternalSteamProcessOngoingUserIDTicketValidation__61_GLOBAL_.N..._Engine_Src_SteamValidateUserIDTickets.cppRtesSc
PvP18TSteamGlobalUserIDPUiPUcUiT3+0x54c9) [0x18b563b9]
[ 7]  ./libsteamvalidateuseridtickets_i486.so(SteamProcessOngoingUserIDTicketValidation+0x36) [0x18b50ede]
[ 8]  ./steamclient_i486.so(_ZN21CSteam2GameServerAuth8RunFrameEv+0x306) [0x18f48ec6]
[ 9]  ./steamclient_i486.so(_ZN21CScheduledFunctionMgr13BFrameFuncRunER11CLimitTimer+0x281) [0x18fbbc11]
[10]  ./steamclient_i486.so(_ZN14CFrameFunctionI21CScheduledFunctionMgrE4BRunER11CLimitTimer+0x27) [0x18fc0607]
[11]  ./steamclient_i486.so(_ZN17CFrameFunctionMgr4BRunE10EFrameTypeP11CCycleCountb+0x135) [0x18f8b9f5]
[12]  ./steamclient_i486.so(_ZN12CSteamEngine9BMainLoopEP11CCycleCountS1_+0xce)[0x18f5e6be]
[13]  ./steamclient_i486.so(_ZN12CSteamEngine15Thread_MainLoopEv+0xe0) [0x18f64cf0]
[14]  ./steamclient_i486.so(_ZN12CSteamEngine13CThreadClient3RunEv+0x1e) [0x18f5e04e]
[15]  ./tier0_s_i486.so(_ZN7CThread10ThreadProcEPv+0x51) [0x18a5ceb1]
[16]  /lib/libpthread.so.0 [0x189fdaea]
[17]  /lib/libc.so.6(__clone+0x57) [0x193ed017]
Unreal Call Stack: appSleep <- EnforceTickRate <- UServerCommandlet::Main
Exiting.

LouieLouie · Mar 14, 2006

If I had to guess I think that we're all having libc.so.6 dependency problems. I'm trying to find it, yum can't.

Case · Mar 14, 2006

You're having threading problems. The clone call is causing a segfault. The dump tells you what happened:

Assertion Failed: Thread cannot be joined with self

EDIT: libc.so.6 should be in /lib/

GS_Schimpf · Mar 14, 2006

So any idea how to fix this threading prob?

SirReal · Mar 14, 2006

In the redorchestra/system directory, try renaming these files:

Code:

mv libgcc_s.so.1 libgcc_s.so.1.OLD  
mv libstdc++.so.6 libstdc++.so.6.OLD

GS_Schimpf · Mar 14, 2006

Running without libstdc++.so.6 gets me an error.

Starting only without libgcc_s.so.1 seems to work, at least the server starts.

Case · Mar 14, 2006

From page 1:

Case said:
Just a hunch here...try setting LD_ASSUME_KERNEL=2.4.19 in your environment before starting the server.
bash users:
export LD_ASSUME_KERNEL=2.4.19
tcsh users:
setenv LD_ASSUME_KERNEL 2.4.19

This will select older threading support....

SirReal · Mar 14, 2006

GS_Schimpf said:
Running without libstdc++.so.6 gets me an error.

Starting only without libgcc_s.so.1 seems to work, at least the server starts.

Well, you do need both. You can only rename them if your system can provide it's own versions of them.

LouieLouie · Mar 14, 2006

Case, I tried the export cmd and it didn't work. it looked to me like the system tried to access some i686 libc files and it choked and puked on them.

LouieLouie · Mar 14, 2006

I tried renaming both to .OLD, the server started but crashed when a client attempts to connect. Also tried only renaming the .so.1 file to OLD and got the same thing, ro crashed upon connection of client.

this is on FC4 x86_64.

GS_Schimpf said:
Running without libstdc++.so.6 gets me an error.

Starting only without libgcc_s.so.1 seems to work, at least the server starts.

icculus · Mar 14, 2006

Please connect to debug server...

Please connect to debug server...

We are trying to reproduce this crash. If you could connect to our test server and play a few rounds, I'd appreciate it...mostly we just need a pile of warm bodies to spend some time on the server, and see if it blows up.

If the server stops responding, congratulations, you crashed it.

We have it running under the debugger, and we'll know more if we can reproduce the issue.

The server is in the browser as "IcculusServer", and you can open the console and type "open treefort.icculus.org:7757" to reach it directly.

If it stops responding, try back in a few hours, as we'll put it back up if we need another run to get more information.

Thanks,
--ryan.

icculus · Mar 14, 2006

Also...

Also...

Please don't rename/remove libraries...we shipped them for a reason.

The crash is (probably) not a threading problem...it just happens to happen in a different thread.

--ryan.

dox · Mar 14, 2006

Case said:
Just a hunch here...try setting LD_ASSUME_KERNEL=2.4.19 in your environment before starting the server.
bash users:
export LD_ASSUME_KERNEL=2.4.19
tcsh users:
setenv LD_ASSUME_KERNEL 2.4.19

This will select older threading support....

^^ This (FC4; 2.6.11) actually doesn't make my server crash (where it did almost incessantly before). However strange things happened: 1. Switched teams - didn't die doing so! Stayed where I was. 2. Get "stuck" a LOT, on a lot of things (sandbags, steps, tanks, etc). 3. Invincible! Shoot me in the face, I don't die! -- but the server didn't crash

Slyfox · Mar 15, 2006

The error doesn't cause the server to crash, but the game does. And resonably frequently.

Sly

Amophous · Mar 15, 2006

Possible Fix?

Possible Fix?

I renamed the 2 libgcc/libstdc files *.old and linked them to the equivilent files supplied by ubuntu 5.1 system files with ln -sf

You can use ln -sf /lib/libgcc_s.so.1 libgcc_s.so.1 to do this.

ls -la lib* should then show:

lrwxrwxrwx 1 hlds users 18 2006-03-14 16:27 libgcc_s.so.1 -> /lib/libgcc_s.so.1

lrwxrwxrwx 1 hlds users 23 2006-03-14 16:30 libstdc++.so.6 -> /usr/lib/libstdc++.so.6

After a few hours of moderately loaded testing, no further crashes or strange behaviour.

Hope this helps somebody else out there.

Phill Lemon, Director.
Lucid Communications, LLC

http://www.lucidcomms.com

[SAS]==Colster== · Mar 15, 2006

Yes Phil, that seems to have worked for me, using links to the linux ones (fedora in my case), server has been up for more than half an hour now which is early days to be certain but it was segfaulting within about 5 minutes before.

Thanks,

..for any linux users not sure how to do this:

Naviagate to your RO system folder and put

ln -sf /lib/libgcc_s.so.1 libgcc_s.so.1
ln -sf /lib/libstdc++.so.6 ibstdc++.so.6

I think you should maybe start a separate thread for this fix

icculus · Mar 15, 2006

Again, I think we're going to find that replacing these libs will turn out to be a non-fix, but if it's working for you, hey, that's cool with me.

I'd still appreciate people hammering the server I listed, so I can reproduce the crash.

--ryan.

Walker · Mar 15, 2006

Hi Ryan,

good to see you in here too. I can't join your server right now (at work), but will post it in my clanforum. Hopefully some clan mates are free.

Hope you can reproduce the problem

curlywurly · Mar 15, 2006

Hi quite new linux commands and quite understand what you mean about doing this:

..for any linux users not sure how to do this:

Naviagate to your RO system folder and put

ln -sf /lib/libgcc_s.so.1 libgcc_s.so.1
ln -sf /lib/libstdc++.so.6 ibstdc++.so.6

or do i follow the commands that Amophous has posted??

Koed · Mar 15, 2006

Been on the Icculus server, but it didn't crash like our own server.

I'm starting to think that people are right about the library versions..have you been testing with these versions?

Server Linux server crashed - SIGSEGV (segmentation fault)

Active member

Member

Data Hippo... in the pants

Grizzled Veteran

Member

Grizzled Veteran

Data Hippo... in the pants

Member

Member

Member

Member

Member

Member

Member

FNG / Fresh Meat

Member

Member

Member

Grizzled Veteran

Member