Hi Vimal... :) On Thu, Sep 9, 2010 at 04:36, Vimal <[email protected]> wrote: > @Mulyadi Santosa: > > Thanks for your suggestion! qemu+gdb was really helpful! > > All the cores were stuck on a spinlock. Attached are the different > backtraces on cores that were spinlocked. We have contacted the > authors and are waiting for a response.
Neat man!!!! Well, I am just showing a method. However, I admit scheduler bugs are sometimes the hardest ones to locate. After reading your attached backware (just briefly, not too deep), I concluded that either it's something wrong with tree of task pointer traversal....or doing lock with wrong assumption. NB: ehm, saw raw_spin_lock in this patch series, shouldn't it be spinlock_irqsave? PS: Glad you find the culprit. You might retry with a kernel image that's fully compiled with -g or -ggdb but without -O2 or -Os. Should give you better backtrace (lesser "value is optimized out"). -- regards, Mulyadi Santosa Freelance Linux trainer and consultant blog: the-hydra.blogspot.com training: mulyaditraining.blogspot.com -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to [email protected] Please read the FAQ at http://kernelnewbies.org/FAQ
