Rudici Cazeao
2017-04-21 22:53:09 UTC
Using kexec on a dual-core arm processor (arm7), the non-boot CPU (CPU1) appears to be locking up during kexec.
The SoC is mach-transcede
The kernel version is 3.0.51-rt75
The issue is localized to linux-3.0.51/arch\arm/mach-transcede/hotplug.c.
cpu_enter_lowpower();
platform_do_lowpower(cpu, &spurious);
cpu_leave_lowpower(). / * bring this CPU back into the world of cache coherency, and then restore interrupts */
static inline void platform_do_lowpower(unsigned int cpu, int *spurious)
{
for (;;) { / * there is no power-control hardware on this platform, so all we can do is put the core into WFI; this is safe as the calling code will have already disabled interrupts */
wfe();
if (pen_release == cpu) {
break; / * OK, proper wakeup, we're done */
}
(*spurious)++; / * Getting here, means that we have come out of WFI without having been woken up - Just note it happening - when we're woken, we can report its occurrence.
}
}
When the issue occurs the non-boot CPU (ie CPU 1) is entering the platform_do_lowpower sequence where it should wait until pren_release is set to 1 by CPU1
Please note that This failure is random. It usually happens somewhere between 300 and 5000 kexec loops.
The SoC is mach-transcede
The kernel version is 3.0.51-rt75
The issue is localized to linux-3.0.51/arch\arm/mach-transcede/hotplug.c.
cpu_enter_lowpower();
platform_do_lowpower(cpu, &spurious);
cpu_leave_lowpower(). / * bring this CPU back into the world of cache coherency, and then restore interrupts */
static inline void platform_do_lowpower(unsigned int cpu, int *spurious)
{
for (;;) { / * there is no power-control hardware on this platform, so all we can do is put the core into WFI; this is safe as the calling code will have already disabled interrupts */
wfe();
if (pen_release == cpu) {
break; / * OK, proper wakeup, we're done */
}
(*spurious)++; / * Getting here, means that we have come out of WFI without having been woken up - Just note it happening - when we're woken, we can report its occurrence.
}
}
When the issue occurs the non-boot CPU (ie CPU 1) is entering the platform_do_lowpower sequence where it should wait until pren_release is set to 1 by CPU1
Please note that This failure is random. It usually happens somewhere between 300 and 5000 kexec loops.