[SOLVED] system lockups after upgrade to kernel 4.1.6

cgorac · 10-22-2015, 12:39 PM

Hello,

I have Thinkpad W520 machine, and I'm tracking -current (pure 64-bit Slackware install here). Ever since last big -current update back in August, when kernel 4.1.6 get installed, Linux crashes on this machine occasionally (sometimes several times a day). At first, I was thinking that it may have to do with the fact that I'm using NVIDIA binary driver, but in the meantime driver get updated several times, as well as several other suspicious packages (Firefox, Flash player, etc.), and the machine is still crashing. Most of the time, __alloc_pipe_info() would appear in the call trace in console, but for the last crash I succeeded in saving dmesg output to a file (sometimes, X would crash first, and shortly afterwards the system would lock down), and the call trace is somewhat different - I'm attaching it to this post.

So: any ideas on how to approach debugging this issue? Also, does anyone know when we may get a kernel update in -curent? I was hesitant to do something about this as I was hoping all the time for a kernel update, that I would like to try first, but three months passed and -current is still on the same kernel version...

Thanks.

bassmadrigal · 10-22-2015, 01:12 PM

There is no guarantee that the kernel will get updated before the next final release of Slackware. I can't look at the log here (and I'm not even sure I could help by looking at the log), but you should be able to build a newer 4.1.x kernel using the .config from your system with ease. This might help you determine if it is the kernel causing the problem. Many Slackware users run kernels other than the official (me included, since I'm running 3.18.8 on my 14.1 install), and Slackware will work fine with newer kernels.

Aizenmyou · 10-22-2015, 01:28 PM

Looks like the issue started here:

Code:

[14947.407116] task: ffff8800b14e6db0 ti: ffff88007e28c000 task.ti: ffff88007e28c000

Could be an issue with one of the kernel modules right before:

Code:

Modules linked in: hidp rfcomm bnep ipv6 nf_conntrack_ipv4 nf_defrag_ipv4 xt_conntrack nf_conntrack xt_tcpudp iptable_filter ip_tables x_tables nls_iso8859_1 nls_cp437 vfat fat cpufreq_ondemand lp ppdev parport_pc parport fuse uvcvideo videobuf2_vmalloc videobuf2_memops videobuf2_core v4l2_common videodev btusb btbcm btintel bluetooth x86_pkg_temp_thermal intel_powerclamp coretemp intel_rapl iosf_mbi snd_hda_codec_hdmi snd_hda_codec_conexant snd_hda_codec_generic kvm_intel iwldvm kvm mac80211 nvidia(PO) i2c_dev snd_hda_intel crct10dif_pclmul snd_hda_controller crc32_pclmul crc32c_intel snd_hda_codec snd_hda_core snd_hwdep iwlwifi snd_pcm thermal thinkpad_acpi sdhci_pci snd_timer psmouse nvram cfg80211 ghash_clmulni_intel sdhci snd drm e1000e soundcore firewire_ohci video processor mmc_core
[14947.407093]  evdev rfkill i2c_i801 agpgart microcode tpm_tis serio_raw efivars thermal_sys ptp firewire_core xhci_pci lpc_ich mei_me i2c_core tpm hwmon xhci_hcd ehci_pci pps_core wmi ehci_hcd mei battery ac button loop ext4 jbd2 mbcache hid_generic usbhid hid

You're running the most current BIOS for your model too, and I see "CTO" on your machine which means a customized Lenovo.
Did you do a --install-new before --upgrade-all?

Also this cannot be good

Code:

[   14.708383] CPU0: Core temperature above threshold, cpu clock throttled (total events = 1)
[   14.708384] CPU1: Core temperature above threshold, cpu clock throttled (total events = 1)
[ 3039.363925] CPU3: Core temperature above threshold, cpu clock throttled (total events = 1)
[ 3039.363926] CPU2: Core temperature above threshold, cpu clock throttled (total events = 1)

cgorac · 10-22-2015, 01:49 PM

Thanks for your replies.

@bassmadrigal: No problem to use kernel built from scratch. I actually use Slackware for number of years, and during a rather long period I was building kernels from scratch, with minimum possible options turned on to keep it lean, all the time. However, at some point you just get lazy and switch to using pre-built kernels. Main obstacle for trying this route was actually in how to get back to using kernels provided in -current again when I'm done with this testing... I guess I'll have to investigate how to build Slackware kernel packages.

@Aizenmyou: Yes, I'm always doing "slackpkg install-new" before "slackware upgrade-all" (and then "slackpkg clean-system" afterwards).

In the meatime, I've un-installed NVIDIA driver, and enabled Nouveau driver (that was black-listed initially), in order to check will the system keep locking up.

Also, regarding "mce: [Hardware Error]: Machine check events logged" errors that appear in dmesg output sometimes, I was able to run mcelog immediately after spotting one of these, and here is the mcelog output:

--------
mcelog: Unsupported new Family 6 Model 2a CPU: only decoding architectural errors
HARDWARE ERROR. This is *NOT* a software problem!
Please contact your hardware vendor
MCE 15
CPU 5 THERMAL EVENT TSC 556b5758f98
TIME 1445539128 Thu Oct 22 20:38:48 2015
Processor 5 below trip temperature. Throttling disabled
STATUS 88040002 MCGSTATUS 0
MCGCAP c09 APICID 5 SOCKETID 0
CPUID Vendor Intel Family 6 Model 42
--------

So I guess this particular issue is not that serious after all.

bassmadrigal · 10-22-2015, 02:11 PM

Quote:

Originally Posted by cgorac

I guess I'll have to investigate how to build Slackware kernel packages.

There's really no need to worry about packages if you don't want to with the kernel. They're pretty standard to install (and remove, if desired). And if you keep your lilo entries for the current kernel and just add entries for the newer kernel, it's super easy to go back. If you want to completely remove the newer kernel, just remove the source directory, the bzImage file (usually in /boot), your kernel's module directory under /lib/modules/, and finally, remove the entry from your lilo.conf

Forum member ryanpcmcquen even created a script to easily build a kernel and update lilo, if you don't feel like doing it by hand.

cgorac · 10-24-2015, 04:26 PM

Updated to kernel 4.2.4 and no lockups over past two days. So I'm closing the thread, thanks again for replies.