Maneesh Soni <[EMAIL PROTECTED]> writes: > Following was observed by David Wilder while testig kexec on panic on > a system with Pentium 4 with HT. Looks like the problem might have > occured due to CPU0 holding the lock and then it might have > received the NMI and it was stopped with runq lock held. CPU1, > enabled the interrupts after stopping other CPU0. Somehow it received > the local apic timer interrupt and started spinning on runq lock. > > Appended is the possible fix for this race condition, which avoids > re-enabling irqs on panic shutdown
Grumble. We should not be disabling apics in the first place Grumble. It is probably better to save and restore the irq state in lapic_shutdown(). Than to open code it. Eric
_______________________________________________ fastboot mailing list [email protected] https://lists.osdl.org/mailman/listinfo/fastboot
