Re: [RFC][PATCH v4 -next 1/4] Move kmsg_dump(KMSG_DUMP_PANIC) below smp_send_stop()

[Chen Gong <gong.chen@linux.intel.com>]

于 2012/1/10 1:59, Seiji Aguchi 写道:
>>
>> I don't know how to prove something is hardened other than not seeing any
>> hangs or false reboots on in that piece of code.
>>
>
> I don't know how we can prove reliability of smp_send_stop() of all archs as well.
>
> Another possible solution is replacing smp_send_stop() with
> machine_crash_shutdown().
> This is a function call stopping other cpus reliably in panic situation
> so that kdump can work .
> But, I don't know whether it works in all archs.

machine_crash_shutdown is defined in only some Archs so it is obvious that it
can't be used in all platforms, BTW, this function is bracketed by CONFIG_KEXEC,
which means it can't be used without this macro. So it is not suitable for
this scenario.

> We need further discussion with Eric for checking this idea is feasible...
>
>
> I would like to go forward step by step rather than doing tough work at once.
> I think just moving kmsg_dump() below smp_send_stop() is a good start for improving pstore.

I agree with you. How about adding macros or something like WARN_ON(XX_ARCH) or
Kconfig to limit its scope?