[GTALUG] Crashes

Giles Orr gilesorr at gmail.com
Tue Jan 31 10:28:05 EST 2017


On 31 January 2017 at 10:03, Alvin Starr via talk <talk at gtalug.org> wrote:
> On 01/31/2017 09:07 AM, Giles Orr via talk wrote:
>> My primary machine is crashing with increasing frequency.  The
>> commonest error I'm seeing in the log looks like this:
>>
>> Jan 29 18:29:39 toshi7 kernel: nouveau 0000:01:00.0: DRM: suspending
>> kernel object tree...
>> Jan 29 18:30:00 toshi7 kernel: NMI watchdog: BUG: soft lockup - CPU#3
>> stuck for 23s! [kscreenlocker_g:19647]
>> Jan 29 18:30:00 toshi7 kernel: Modules linked in: fuse uas usb_storage
>> rfcomm ip6t_rpfilter ip6t_REJECT nf_reject_ipv6 xt_conntrack ip_set
>> nfnetlink ebtable_broute bridge stp llc ebtable_nat ip6table_nat
>> nf_conntrack ...
>>
>> I realize that I'm probably not giving enough information, but pasting
>> large chunks of log files would be just as counterproductive in its
>> own way.  I've seen this one A LOT - and sometimes I get it and the
>> machine goes hours (but not days) before crashing.  So ... is
>> kscreenlocker likely to be the problem here?  When I searched for "BUG
>> soft lockup CPU stuck for" on Google, the top result had exactly the
>> same number of seconds, and said that replacing the power supply fixed
>> the problem.  Which is a step I'd probably be willing to take, but
>> this isn't a desktop, it's a laptop.  So I'd want to be very sure as
>> the power supply is unique to this machine (if it's available at all)
>> and probably quite expensive.
>>
>> The processor:
>>
>> Intel(R) Core(TM) i7-3610QM CPU @ 2.30GHz (4594 bogomips)
>> current speed: 1274MHz, 4 cores, 8 threads
>>
>> While it's not a current gen processor, this is still a good machine
>> and I'd rather fix it than toss it.
>>
>> Got an immediate crash this morning, and to my surprise the error was
>> very different:
>>
>> Jan 31 07:56:35 toshi7 kernel: ------------[ cut here ]------------
>> Jan 31 07:56:35 toshi7 kernel: kernel BUG at lib/radix-tree.c:769!
>> Jan 31 07:56:35 toshi7 kernel: invalid opcode: 0000 [#1] SMP
>> Jan 31 07:56:35 toshi7 kernel: Modules linked in: uas usb_storage
>> rfcomm ip6t_rpfilter ip6t_REJECT nf_reject
>> _ipv6 xt_conntrack ip_set nfnetlink ebtable_nat ebtable_broute bridge
>> stp llc ip6table_nat nf_conntrack_ipv6 ...
>>
>> Finally, I'm also getting this periodically:
>>
>> Jan 28 08:49:52 toshi7 kernel: CPU2: Core temperature above threshold,
>> cpu clock throttled (total events = 1
>> )
>> Jan 28 08:49:52 toshi7 kernel: CPU6: Core temperature above threshold,
>> cpu clock throttled (total events = 1)
> [snip]
>> Jan 28 08:49:52 toshi7 kernel: CPU0: Package temperature/speed normal
>> Jan 28 08:49:52 toshi7 kernel: CPU2: Package temperature/speed normal
>> Jan 28 08:49:52 toshi7 kernel: CPU6: Package temperature/speed normal
>>
>> This suggests that it's overheating, throttling, and recovering pretty
>> much instantaneously: my thought is that it's probably not a problem,
>> but I thought I should check.
>>
>> How should I proceed from here:
>> - the processor is going funny, replace it
>> - junk the laptop, it's toast
>> - debug further (how?)
>> - replace the power supply
>> - uninstall kscreenlocker and see what happens
>>
>
> If the CPU is going over temp then it could start acting unpredictably.
>
> If you have lm_sensors installed then it would be worthwhile checking
> the temp of the CPU during normal operation.
> I would also check the fans because most fans out there are
> "inexpensive" and will start to cease up over time slowing down till
> things start getting hot.
> Another thing that has bitten me in the past was pushing a computer with
> a side vent up against a wall causing the still good fans from working
> almost at all.
>
> Another thing that will cause random problems is memory so if the
> cooling is not the issue then try running a memory test.
> Unless you have ECC and there are no errors being logged.

I should add that I ran memtest86(+?) for a couple hours a month ago,
and it came up error-free.  And I ran the smartctl long test on the
hard drive quite recently, again without error.  I should run the
memory test again (and possibly even the HD one), but it makes me
think that these aren't the problem.  I think the fans are functioning
okay, but that's worth looking at and I'll get lmsensors installed
again.

-- 
Giles
http://www.gilesorr.com/
gilesorr at gmail.com


More information about the talk mailing list