search

draios / sysdig

Linux system exploration and troubleshooting tool with first class support for containers
http://www.sysdig.com/
Other
7.72k stars 726 forks source link

kernel panic #378

Closed pmbauer closed 9 years ago

pmbauer commented 9 years ago

I can consistently induce a kernel panic when running sysdig, forcing an REISUB This usually happens within a few seconds of running sysdig. It happens when running from inside X or directly from ttyX.

sysdig:    version 0.1.101
System:    Host: saffron Kernel: 3.19.0-21-generic x86_64 (64 bit) Desktop: Unity 7.3.2
           Distro: Ubuntu 15.04 vivid
Machine:   System: LENOVO product: 20BXCTO1WW v: ThinkPad T450s
           Mobo: LENOVO model: 20BXCTO1WW v: SDK0E50512 STD Bios: LENOVO v: JBET46WW (1.11 ) date: 03/02/2015
CPU:       Dual core Intel Core i7-5600U (-HT-MCP-) speed/max: 858/3200 MHz
Graphics:  Card: Intel Broadwell-U Integrated Graphics
           Display Server: X.Org 1.17.1 drivers: intel (unloaded: fbdev,vesa)
           Resolution: 1920x1080@60.0hz, 2560x1600@60.0hz
           GLX Renderer: Mesa DRI Intel HD Graphics 5500 (Broadwell GT2) GLX Version: 3.0 Mesa 10.5.2
Network:   Card-1: Intel Ethernet Connection (3) I218-LM driver: e1000e
           Card-2: Intel Wireless 7265 driver: iwlwifi
Drives:    HDD Total Size: 512.1GB (17.6% used) ID-1: model: SAMSUNG_MZ7LN512
Info:      Processes: 269 Uptime: 7 min Memory: 4511.8/11895.2MB Client: Shell (bash) inxi: 2.2.16

(picture of panic) https://www.dropbox.com/s/2tc0fgzyz9pvbib/2015-06-16%2016.49.11.jpg?dl=0

gianlucaborello commented 9 years ago

Bad indeed.

Decoding the address of the kernel panic, it seems to come from here:

(gdb) info line *(ppm_ioctl)+0x3da
Line 764 of "/root/sysdig/driver/main.c" starts at address 0x1b77 <ppm_ioctl+967> and ends at 0x1b96 <ppm_ioctl+998>.

Which is:

#ifdef for_each_process_thread
        for_each_process_thread(p, t) {
#else
        for_each_process(p) {

And it's highly likely, since it's part of some new code introduced with 0.1.101.

We'll do our best to replicate this and ask for help if we're unable to reproduce.

ldegio commented 9 years ago

@pmbauer, I believe the fix that solves https://github.com/draios/sysdig/issues/391 should solve this one as well. Could you try if the problem disappears when compiling from the dev branch?

pmbauer commented 9 years ago

@ldegio Wonderful. Swamped with work today but will try to get some bandwidth for this later this weekend. Thank you!

ldegio commented 9 years ago

@pmbauer did you get a chance to test? Can we close this issue?

pmbauer commented 9 years ago

No. I'm moving. I spent 20 minutes here and don't have the time to finish building on a clean vm and test this; I'm sorry I indicated otherwise last Friday. If there is a more-recent build of sysdig available with the bug fix I can test it, but don't have time to build it all from scratch.

gianlucaborello commented 9 years ago

We believe this was fixed on 0.1.102 that was just released today. Could you please try and report your feedback?

lilydjwg commented 9 years ago

I got this error in the past too. After updating to 0.1.102, I don't see it any more (to be exact, it doesn't happen on first few runs).

gianlucaborello commented 9 years ago

That's good news, I'm going to close this for the moment then.