On Mon, 2017-04-03 at 11:24 +0900, AKASHI Takahiro wrote:
>
> +static void ipi_cpu_crash_stop(unsigned int cpu, struct pt_regs
> *regs)
> +{
> +#ifdef CONFIG_KEXEC_CORE
> +???????crash_save_cpu(regs, cpu);
> +
> +???????atomic_dec(&waiting_for_crash_ipi);
> +
> +???????local_irq_disable();
Shouldn't the above two be the other way round? Stop handling
interrupts *before* we tell the surviving CPU that we're done?
Also, should the surviving CPU be calling __cpu_die() to attempt to
check that the others really have died? Sure, it can't do much more
than print a warning if the call to ->cpu_kill() times out, but that's
still useful information. And perhaps we could allow platforms to
register a method to shoot other CPUs down forcefully if they don't
respond. My board could apparently support that.
It would certainly be useful to allow smp_send_crash_stop() to be
overridden with a platform-specific method instead of normal IPIs. The
platform might be able to do something... less maskable.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/x-pkcs7-signature
Size: 4938 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/kexec/attachments/20170403/694d66c3/attachment-0001.bin>
