Raspberry Pi breakage


Message boards : Problems and bug reports : Raspberry Pi breakage

Message board moderation

To post messages, you must log in.
AuthorMessage
chillywillycd

Send message
Joined: 18 Jul 14
Posts: 2
Credit: 1,399,387
RAC: 0
Message 3706 - Posted: 19 Oct 2014, 13:48:30 UTC
I've been running asteroids@home on my raspberry pi using the default boinc client provided in the raspbian repositories
boinc --version
7.0.27 arm-unknown-linux-gnueabihf


my raspberry pi has been crashing, and this morning i captured the info it spit out when dumping me into the kernel debugger.

i'd be willing to help test/recompile to help out. but having very limited experience with C/assembly that i assume the asteroids@home period search application is written in, i'm not sure how helpful my java development skillz would be


kdb output:
[ 4834.860982] BUG: unsupported FP instruction in kernel mode
[ 4834.868258] Internal error: Oops - undefined instruction: 0 [#1] PREEMPT ARM

Entering kdb (current=0xdbe1d780, pid 2334) Oops: (null)
due to oops @ 0xc00099c4

dCPU: 0 PID: 2334 Comm: period_search_1 Not tainted 3.12.28+ #709
dtask: dbe1d780 ti: dd2dc000 task.ti: dd2dc000
PC is at vfp_save_state+0x0/0x28
LR is at vfp_sync_hwstate+0x70/0x7c
pc : [<c00099c4>]    lr : [<c0009588>]    psr: 60000113
sp : dd2ddeb0  ip : 00000018  fp : 00000037
r10: 01351168  r9 : dbe1db10  r8 : 00000000
r7 : dd2ddfb0  r6 : befe9fe0  r5 : c0000780  r4 : dd2dc030
r3 : dd2dc0f8  r2 : c061e920  r1 : c0000780  r0 : dd2dc0f8
Flags: nZCv  IRQs on  FIQs on  Mode SVC_32  ISA ARM  Segment user
Control: 00c5387d  Table: 1bde8008  DAC: 00000015
dCPU: 0 PID: 2334 Comm: period_search_1 Not tainted 3.12.28+ #709
[<c001444c>] (unwind_backtrace+0x0/0xec) from [<c0011730>] (show_stack+0x10/0x14)
[<c0011730>] (show_stack+0x10/0x14) from [<c00820a4>] (kdb_dumpregs+0x28/0x50)
[<c00820a4>] (kdb_dumpregs+0x28/0x50) from [<c0085180>] (kdb_main_loop+0x370/0x6bc)  
[<c0085180>] (kdb_main_loop+0x370/0x6bc) from [<c0087980>] (kdb_stub+0x164/0x380)
more> �������
Only 'q' or 'Q' are processed at more prompt, input ignored
[<c0087980>] (kdb_stub+0x164/0x380) from [<c007e984>] (kgdb_handle_exception+0x380/0x748)
[<c007e984>] (kgdb_handle_exception+0x380/0x748) from [<c0013bbc>] (kgdb_notify+0x24/0x50)
[<c0013bbc>] (kgdb_notify+0x24/0x50) from [<c0446728>] (notifier_call_chain+0x44/0x84)
[<c0446728>] (notifier_call_chain+0x44/0x84) from [<c04467e8>] (atomic_notifier_call_chain+0x34/0x48)
[<c04467e8>] (atomic_notifier_call_chain+0x34/0x48) from [<c0446834>] (notify_die+0x38/0x44)
[<c0446834>] (notify_die+0x38/0x44) from [<c00117fc>] (die+0xc8/0x390)
[<c00117fc>] (die+0xc8/0x390) from [<c0008334>] (do_undefinstr+0x1b0/0x1dc)
[<c0008334>] (do_undefinstr+0x1b0/0x1dc) from [<c0444d6c>] (__und_svc_finish+0x0/0x34)
Exception stack(0xdd2dde28 to 0xdd2dde70)
de20:                   dd2dc0f8 c0000780 c061e920 dd2dc0f8 dd2dc030 c0000780
de40: befe9fe0 dd2ddfb0 00000000 dbe1db10 01351168 00000037 00000018 dd2ddeb0
de60: c0009588 c00099c4 60000113 ffffffff
[<c0444d6c>] (__und_svc_finish+0x0/0x34) from [<c00099c4>] (vfp_save_state+0x0/0x28) 
[<c00099c4>] (vfp_save_state+0x0/0x28) from [<c0009588>] (vfp_sync_hwstate+0x70/0x7c)
Only 'q' or 'Q' are processed at more prompt, input ignored
[<c0009588>] (vfp_sync_hwstate+0x70/0x7c) from [<c0009794>] (vfp_preserve_user_clear_hwstate+0x20/0x8c)
[<c0009794>] (vfp_preserve_user_clear_hwstate+0x20/0x8c) from [<c0010a48>] (setup_sigframe+0x190/0x1a0)
[<c0010a48>] (setup_sigframe+0x190/0x1a0) from [<c0010ea4>] (do_signal+0x2e8/0x440)
[<c0010ea4>] (do_signal+0x2e8/0x440) from [<c0011194>] (do_work_pending+0xa4/0xb4)
[<c0011194>] (do_work_pending+0xa4/0xb4) from [<c000df00>] (work_pending+0xc/0x20)
ID: 3706 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
MarkJ
Avatar

Send message
Joined: 27 Jun 12
Posts: 129
Credit: 62,716,984
RAC: 41
Message 3708 - Posted: 21 Oct 2014, 10:05:59 UTC

Last modified: 21 Oct 2014, 10:11:09 UTC
Have you picked up a kernel update? From the dump it looks like you've got #709. We went through a bunch of issues when they updated the kernel a while back although this one has been stable for me for a while.

~ $ uname -a
Linux xxx 3.12.28+ #709 PREEMPT Mon Sep 8 15:28:00 BST 2014 armv6l GNU/Linux
~ $ boinccmd --version
boinccmd, built from BOINC 7.4.22


I'm running Jessie and its also got the current alpha-test BOINC client in the repo.

I gave up running Asteroids on mine as they take too long. I did try the boinc-app-seti from the repo but that crashed the Pi too.
BOINC blog
ID: 3708 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
Profile mikey
Avatar

Send message
Joined: 1 Jan 14
Posts: 300
Credit: 32,327,253
RAC: 6,461
Message 3709 - Posted: 21 Oct 2014, 10:46:17 UTC
Not getting into the good or bad points, nor do I have a Pi, but Bitcoin Utopia now has a Pi application if you need to try something else. It is a Boinc project that is helping fun other Boinc Projects thru Bitcoin mining. They have applications for the cpu, gpu, pi and dedicated asic miners too.
ID: 3709 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote
chillywillycd

Send message
Joined: 18 Jul 14
Posts: 2
Credit: 1,399,387
RAC: 0
Message 3775 - Posted: 12 Nov 2014, 17:57:26 UTC
@MarkJ i seem to running the same kernel as you, but i'm still running Wheezy
 ~ $ uname -a
Linux raspi512 3.12.28+ #709 PREEMPT Mon Sep 8 15:28:00 BST 2014 armv6l GNU/Linux

i've ended up just working around the problem.


I turned off the default action of launching KDB by changing
/boot/cmdline.txt
to remove
kgdboc=ttyAMA0,115200


so, now instead of breaking into KDB when an unsupported instruction is encountered the OS now ignores it and turns the period_search process into a zombie.

i set up a cron job to check for zombie processes being run by the boinc user every 15 minutes. this job parses boinccmd --get_tasks for the tasks name and executes
boinccmd --task <asteroids_url> $name abort
boinccmd --project <asteroids_url> update


the raspberry pi seems to be able to complete some work units, and not others...
ID: 3775 · Rating: 0 · rate: Rate + / Rate - Report as offensive     Reply Quote

Message boards : Problems and bug reports : Raspberry Pi breakage