You are not logged in.

#1 2011-06-06 19:37:11

s2
Member
Registered: 2010-03-27
Posts: 7

Building old vanilla kernels

I've encountered regression in kernel:
https://bugs.archlinux.org/task/22223

To hunt this down, I used tested kernel26 coming from existing ISO-installers.
After this I tried building kernel 2.6.33 by hand, just using old default configs (from installers) and make localmodconfig.
The resulting kernel crashed with long stacktrace.

Then I've tried building the package kernel26-2.6.33  (with different build suffix) by using makepkg way (used the same config file which I got from installed old Arch).
Result crashed the same way:

[<c12ce9bd>] panic + 0x4d / 0x100
[<........>] do_exit + 0x6dc / 0x750
[<........>] ? kmsg_dump
             ? printk
             oops_end
             no_context
             __bad_area_nosemaphore
             ? __d_instantiate
             ? d_rehash
             bad_area_nosemaphore
             do_page_fault
             ? _security_d_instantiate
             ? _do_page_fault
             error_code
             ? strcmp
             event_create_dir
             ? module_update_tracepoints
             ? tracepoint_update_probes
             event_trace_init
             do_one_initcall
             ? event_trace_init
             kernel_init
             ? kernel_init
             kernel_thread_helper

Then I've repeated this for kernel26-2.6.32.
The stacktrace of the crash was 2x shorter, but it crashed anyway.


Is there any way of doing kernel code regression testing easier way (possible at all) in Arch Linux?
Without this, almost any git bisect command in kernel tree is useless..

Last edited by s2 (2011-06-06 19:40:23)

Offline

Board footer

Powered by FluxBB