You are not logged in.

#1 2013-10-14 23:42:18

bnolsen
Member
Registered: 2008-12-10
Posts: 64

hard crash: can only boot single user

My desktop is a dual core core2quad based xeon.  Last time I rebooted was with linux 3.11.2   I just rebooted with 3.15.1 (corrected 3.11.5) and also with latest lts.  With normal boot, both get to a full prompt then start dumping sequences of kernel messages regarding core problems then ultimately locks hard with flashing keyboard lights.

I did discover that booting single user gets me into the system.

I almost suspect that some utility was compiled wth core-i7 cpu support and not generic x86_64 support.

At the moment I'm turning on things one at a time from within single user mode to find out what's causing everything to go south.  Thankfully I can open up the network.

I'm working in the system currently.  I went from single user mode and su - into my normal user and am running the system this way with no problems.  Well, systemd is plain idiotic in this state.

Any way to try to debug what packages went south?

Last edited by bnolsen (2013-10-15 00:07:31)

Offline

#2 2013-10-14 23:48:41

Trilby
Inspector Parrot
Registered: 2011-11-29
Posts: 29,525
Website

Re: hard crash: can only boot single user

bnolsen wrote:

Last time I rebooted was with linux 3.11.2   I just rebooted with 3.15.1 ...

I'm not sure if it will be relevant, but I assume that is a typo.  There is no 3.15.1 yet, not even upstream (most recent is a 3.12 rc).  Do you mean 3.11.5?


"UNIX is simple and coherent..." - Dennis Ritchie, "GNU's Not UNIX" -  Richard Stallman

Offline

#3 2013-10-14 23:50:02

bnolsen
Member
Registered: 2008-12-10
Posts: 64

Re: hard crash: can only boot single user

sorry yes, 3.11.5  Attempting to boot LTS also results in the same hard lock behavior so IMHO that seems to point to something in userspace triggering the crash.

More speculation, bad (unsupported) instructions may lock a cpu and the progressive kernel dump messages ending in hard lockup are a result of the whatever it is locking up each cpu in turn, the scheduler moving the process to a new victim cpu each time until there's no cpus left.

message native_smp_send_reschedule pops up in the dump

Last edited by bnolsen (2013-10-15 00:01:06)

Offline

Board footer

Powered by FluxBB