Thanks for the Inputs. As there is not kernel crash, crash files are not create here to debug more. Is there any way to force the kernel to create kernel crash files under /var/crash. This helps more If am not wrong. Regards, Sri On Mon, Oct 10, 2011 at 9:20 PM, Mulyadi Santosa <mulyadi.santosa@gmail.com>wrote:
Hi...
On Mon, Oct 10, 2011 at 13:34, sri <bskmohan@gmail.com> wrote:
######################## START OF LOGS ############################ Oct 1 14:16:56 server03 kernel: low stack detected by irq handler Oct 1 14:16:56 server03 kernel: [<c0405e9b>] do_IRQ+0x86/0xc3 Oct 1 14:16:56 server03 kernel: [<c04046e6>] common_interrupt+0x1a/0x20 Oct 1 14:16:56 server03 kernel: [<c05a5c06>] neigh_lookup+0x23/0x7f Oct 1 14:16:56 server03 kernel: [<c05d6800>]
arp_bind_neighbour+0x50/0x71
Oct 1 14:16:56 server03 kernel: [<c05b7b44>] rt_intern_hash+0x372/0x480 Oct 1 14:16:56 server03 kernel: [<c05b8352>] __ip_route_output_key+0x700/
Looks like something is exhausting kernel stack way too fast. IIRC in 4K kernel stack scenario, 4K is assigned for normal usage and another 4K for irq.
So, IMHO, either something is leaking, or code paths are mixing too much and too fast so they all together exhaust the irq stack.
For debugging, I am not sure...maybe ftrace stack usage could help you. Grep your kernel .config and find out everything that contains "stack". Or go to "kernel hacking" section and dig into ftrace section.
Good luck....
-- regards,
Mulyadi Santosa Freelance Linux trainer and consultant
blog: the-hydra.blogspot.com training: mulyaditraining.blogspot.com
-- -- Krishna Mohan B