RPi compute errors - signal 11

Message boards : Number crunching : RPi compute errors - signal 11
Message board moderation

To post messages, you must log in.

AuthorMessage
hartacus
 

Send message
Joined: 2 May 14
Posts: 1
Credit: 183,908
RAC: 0
Message 3403 - Posted: 10 Nov 2015, 1:10:06 UTC

Hi,

I've freshly re-installed Raspbian on a Raspberry Pi that's been running WUProp for a good few months without trouble. I've also recompiled BOINC v7.5.0, which has been the most stable I've found so far for the RPi. I've got basically the same setup on an RPi2 that's running WUProp 4.18 just fine. However, on the newly installed RPi, WUProp tasks get a compute error immediately after download.

Example: http://wuprop.boinc-af.org/result.php?resultid=53390258
Stderr gives the message "process got signal 11" which I gather is a segmentation fault. There's plenty of RAM free and swap is all free, so it should have access to some memory, but seems to be having difficulty straight up. The same RPi is currently running Einstein and Collatz tasks happily (and, as previously mentioned, ran WUProp OK in the past) so it doesn't seem to be a hardware fault.

Also, this doesn't seem to be a new error - some guys asked about this in the Questions and Answers section a couple of years ago: http://wuprop.boinc-af.org/forum_thread.php?id=177

Any ideas?
ID: 3403 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Steve Hawker*
       

Send message
Joined: 31 Oct 12
Posts: 125
Credit: 655,634
RAC: 0
Message 3427 - Posted: 15 Nov 2015, 17:40:40 UTC

I've got the same problem.

WUs from this project used to run on my pi without issue. Then I stopped using the pi for a while but eventually returned to it. At that point i followed the instructions to get bitcoin utopia to run, which it does and has never hiccuped once.

But since then, all WUProp tasks fail with signal 11.

Trying to figure it all out, there is this evidence:

1. WUProp applications were updated during my pi's idle period. Since then most of my Android tasks also fail by running forever.
2. Other projects run on the pi without incident.
3. The installation for BU required a bunch of libraries, maybe its missing a 32bit library?

I reported the Android issue seven months ago. I'm not the only one with it. We've had no response at all from Sebastien and I honestly doubt we will hear anything about this one either.
ID: 3427 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Moises Cardona
   

Send message
Joined: 17 Sep 10
Posts: 14
Credit: 1,546,399
RAC: 0
Message 3500 - Posted: 9 Dec 2015, 16:47:53 UTC

I'm having the same signal 11 error in my Pi, but I have the Pi 1 not the 2 running Raspbian (latest version available). BU runs fine without the need for extra libraries in that distro so I don't think that could be the issue. I'm running BU and Enigma@Home without issues in it. the only issue is WUProp which immediately gets a Computational Error ad when reported it says Signal 11... I'm using 7.4.23 which is the package available in the repository.
ID: 3500 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile BilBg
Avatar

Send message
Joined: 20 Jun 12
Posts: 63
Credit: 94,685
RAC: 0
Message 3502 - Posted: 10 Dec 2015, 8:47:42 UTC

I hope I identified the computers properly - I used the combination of ARM + Linux
(When you have many computers it's good to give link to the problem computer. Nobody but you can see your computer names so don't use that to identify the computers)

hartacus - uses BOINC version 7.5.0 - problem seems fixed but hartacus should state why/how (what did he change?)
http://wuprop.boinc-af.org/show_host_detail.php?hostid=82662

Steve Hawker*
http://wuprop.boinc-af.org/show_host_detail.php?hostid=73233

Moises Cardona
http://wuprop.boinc-af.org/show_host_detail.php?hostid=88601



- ALF - "Find out what you don't do well ..... then don't do it!" :)
ID: 3502 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Moises Cardona
   

Send message
Joined: 17 Sep 10
Posts: 14
Credit: 1,546,399
RAC: 0
Message 3504 - Posted: 10 Dec 2015, 13:21:19 UTC

Interestingly, there seems to be 3 flavors of the Linux for ARM app. Could it be the Raspberry Pi is picking up the incorrect app? I can check this if I could know how to make the app_info file to force it to use a specific app?
ID: 3504 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile BilBg
Avatar

Send message
Joined: 20 Jun 12
Posts: 63
Credit: 94,685
RAC: 0
Message 3509 - Posted: 11 Dec 2015, 0:39:40 UTC - in response to Message 3504.  

All the apps are here:
http://wuprop.boinc-af.org/download/

Probably the 3 apps you need are:
http://wuprop.boinc-af.org/download/data_collect_v4_418_arm-unknown-linux-gnueabi__nci
http://wuprop.boinc-af.org/download/data_collect_v4_418_arm-unknown-linux-gnueabihf__nci
http://wuprop.boinc-af.org/download/data_collect_v4_418_armv6l-unknown-linux-gnueabihf__nci

(I just found: The last 2 files are identical so no need to check both)

You may need to make files Executable by:
chmod +x

I made two variants of app_info.xml for you - I'm not sure if the additional 3 lines are needed or not.
("user_friendly_name" can be anything - change it if you like)
app_info.xml should be ANSI encoded (Not Unicode nor UTF8)

This site seems to use buggy version of the forum code - [pre][/pre] and [code][/code] tags don't work properly
I needed to Copy/Paste app_info.xml as plain text.

So use [Quote] under my post and then Copy/Paste from the text-box
(this way the app_info.xml text will be nicely formatted for easy reading)



<app_info>

<app>
<name>data_collect_v4</name>
<user_friendly_name>Data collect version 4</user_friendly_name>
</app>

<file_info>
<name>data_collect_v4_418_arm-unknown-linux-gnueabi__nci</name>
<executable/>
</file_info>

<app_version>
<app_name>data_collect_v4</app_name>
<version_num>418</version_num>
<file_ref>
<file_name>data_collect_v4_418_arm-unknown-linux-gnueabi__nci</file_name>
<main_program/>
</file_ref>
</app_version>

</app_info>





<app_info>

<app>
<name>data_collect_v4</name>
<user_friendly_name>Data collect version 4</user_friendly_name>
</app>

<file_info>
<name>data_collect_v4_418_arm-unknown-linux-gnueabi__nci</name>
<executable/>
</file_info>

<app_version>
<app_name>data_collect_v4</app_name>
<version_num>418</version_num>
<avg_ncpus>0.010000</avg_ncpus>
<max_ncpus>0.010000</max_ncpus>
<plan_class>nci</plan_class>
<file_ref>
<file_name>data_collect_v4_418_arm-unknown-linux-gnueabi__nci</file_name>
<main_program/>
</file_ref>
</app_version>

</app_info>



 



- ALF - "Find out what you don't do well ..... then don't do it!" :)
ID: 3509 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
sorcrosc
     

Send message
Joined: 20 Jan 13
Posts: 9
Credit: 476,419
RAC: 0
Message 3518 - Posted: 13 Dec 2015, 19:57:42 UTC
Last modified: 13 Dec 2015, 20:05:03 UTC

I also have a raspberri pi 1 and i never have been able to run wuprop

data_collect_v4_418_arm-unknown-linux-gnueabihf__nci do not run at all. I tried to run it standalone in terminal and it gives segmentation error imediately.

data_collect_v4_418_arm-unknown-linux-gnueabi__nci seems to work but it freezes the system after short time
http://wuprop.boinc-af.org/result.php?resultid=54213796
I recompiled boinc 7.6.19 to try (7.4.23 in repos). With this version i can see some more info in the output
http://wuprop.boinc-af.org/result.php?resultid=54239895
ID: 3518 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile BilBg
Avatar

Send message
Joined: 20 Jun 12
Posts: 63
Credit: 94,685
RAC: 0
Message 3519 - Posted: 14 Dec 2015, 5:33:55 UTC - in response to Message 3518.  

Which version of app_info.xml did you use?
i.e. are these 3 lines needed?:

<avg_ncpus>0.010000</avg_ncpus>
<max_ncpus>0.010000</max_ncpus>
<plan_class>nci</plan_class>


Task will vanish in a few days so to preserve Stderr output I will post it:

Stderr output

<core_client_version>7.6.19</core_client_version>
<![CDATA[
<message>
process exited with code 193 (0xc1, -63)
</message>
<stderr_txt>
20:17:21 (562): initialisation
SIGSEGV: segmentation violation
Stack trace (3 frames):
../../projects/wuprop.boinc-af.org/data_collect_v4_418_arm-unknown-linux-gnueabi__nci[0x5b15c]
/lib/arm-linux-gnueabihf/libc.so.6(__default_sa_restorer_v2+0x0)[0xb6c46160]
/lib/arm-linux-gnueabihf/libc.so.6(+0x32284)[0xb6c49284]

Exiting...

</stderr_txt>



- ALF - "Find out what you don't do well ..... then don't do it!" :)
ID: 3519 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
sorcrosc
     

Send message
Joined: 20 Jan 13
Posts: 9
Credit: 476,419
RAC: 0
Message 3520 - Posted: 14 Dec 2015, 20:15:19 UTC

Thank you BilBg

The app_info was without the 3 lines. Adding <plan_class>nci</plan_class> gives the same result

I can still get the non hf version without app_info, supplying arm-unknown-linux-gnueabi as alt_platform in cc_config. Same result
ID: 3520 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Steve Hawker*
       

Send message
Joined: 31 Oct 12
Posts: 125
Credit: 655,634
RAC: 0
Message 3523 - Posted: 15 Dec 2015, 16:50:12 UTC - in response to Message 3518.  
Last modified: 15 Dec 2015, 16:53:12 UTC

I also have a raspberri pi 1 and i never have been able to run wuprop


I have a Pi model A and I have run WUProp. But not since March 2015 when Sebastien added support for PIE/non-PIE on Android. Version 4.18 was produced a few days later and that's the version that doesn't work on our devices. I suspect that some common ARM code is the problem here.

Maybe if we could revert to an earlier version, that might do the trick. Haven't got time right now but perhaps this one:

http://wuprop.boinc-af.org/download/data_collect_v4_4.17_arm-android-linux-gnu__nci

Or better yet, Sebastien returns and fixes things.
ID: 3523 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile BilBg
Avatar

Send message
Joined: 20 Jun 12
Posts: 63
Credit: 94,685
RAC: 0
Message 3526 - Posted: 15 Dec 2015, 19:51:28 UTC - in response to Message 3523.  

No, "arm-android-linux" is for Android, on the Download page Ctrl+F for:
arm-unknown-linux

Looks like Sebastien released new version:
http://wuprop.boinc-af.org/apps.php
Linux running on an ARM compatible CPU 4.19 (nci) 15 Dec 2015

Which is this file:
http://wuprop.boinc-af.org/download/data_collect_v4_4.19_arm-unknown-linux-gnueabihf__nci



- ALF - "Find out what you don't do well ..... then don't do it!" :)
ID: 3526 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
sorcrosc
     

Send message
Joined: 20 Jan 13
Posts: 9
Credit: 476,419
RAC: 0
Message 3527 - Posted: 15 Dec 2015, 19:55:51 UTC

I just saw it, it's runnig. Let's see if it works
ID: 3527 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Steve Hawker*
       

Send message
Joined: 31 Oct 12
Posts: 125
Credit: 655,634
RAC: 0
Message 3529 - Posted: 16 Dec 2015, 1:03:03 UTC - in response to Message 3527.  

I just saw it, it's runnig. Let's see if it works



Worked for me!! Thanks Sebastien!!

Now for 24 hours of BU a day!! woot!
ID: 3529 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile Moises Cardona
   

Send message
Joined: 17 Sep 10
Posts: 14
Credit: 1,546,399
RAC: 0
Message 3531 - Posted: 16 Dec 2015, 11:36:37 UTC
Last modified: 16 Dec 2015, 11:38:59 UTC

Yup, confirmed working here too. Thanks!
ID: 3531 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
sorcrosc
     

Send message
Joined: 20 Jan 13
Posts: 9
Credit: 476,419
RAC: 0
Message 3532 - Posted: 16 Dec 2015, 12:01:42 UTC

Thank you Sebastien. Should we start to see results for raspberry in arm devices now?

Footnote: boinc from repos in Raspbian is compiled with arm-unknown-linux-gnueabihf as platform but if I compile it myself it takes armv6l-unknown-linux-gnueabihf for raspberry pi 1.
I have configured it with the first as alt_platform while compiling so i get the right app but I think you have to stop sending the armv6l app which is not working or make it the same as arm, as it was with 4.18 version
ID: 3532 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Number crunching : RPi compute errors - signal 11

©2024 Sébastien