Re: [PATCH] PM/reboot: call syscore_shutdown() after disable_nonboot_cpus()

[chenhc@lemote.com]

> On Monday, April 08, 2013 10:51:45 AM chenhc@lemote.com wrote:
>> > On Sunday, April 07, 2013 08:29:32 AM Greg KH wrote:
>> >> On Sun, Apr 07, 2013 at 10:46:00AM +0200, Rafael J. Wysocki wrote:
>> >> > On Sunday, April 07, 2013 10:14:14 AM Huacai Chen wrote:
>> >> > > As commit 40dc166c (PM / Core: Introduce struct syscore_ops for
>> core
>> >> > > subsystems PM) say, syscore_ops operations should be carried with
>> >> one
>> >> > > CPU on-line and interrupts disabled. However, after commit
>> f96972f2d
>> >> > > (kernel/sys.c: call disable_nonboot_cpus() in kernel_restart()),
>> >> > > syscore_shutdown() is called before disable_nonboot_cpus(), so
>> break
>> >> > > the rules. We have a MIPS machine with a 8259A PIC, and there is
>> an
>> >> > > external timer (HPET) linked at 8259A. Since 8259A has been
>> shutdown
>> >> > > too early (by syscore_shutdown()), disable_nonboot_cpus() runs
>> >> without
>> >> > > timer interrupt, so it hangs and reboot fails. This patch call
>> >> > > syscore_shutdown() a little later (after disable_nonboot_cpus())
>> to
>> >> > > avoid reboot failure, this is the same way as poweroff does.
>> >> > >
>> >> > > BTW, add disable_nonboot_cpus() in kernel_halt() for consistency.
>> >> > >
>> >> > > Signed-off-by: Huacai Chen <chenhc@lemote.com>
>> >> > > Cc: <stable@vger.kernel.org>
>> >> >
>> >> > While I agree with the changes, I'm not sure if I'm the right
>> >> maintainer,
>> >> > as this isn't really PM code.
>> >> >
>> >> > Andrew, should I take this?
>> >>
>> >> Andrew is on vacation for a few weeks, so you might need to take this
>> >> through your tree.
>> >
>> > OK
>> >
>> > But, it looks like we should actually disable interrupts on the
>> remaining
>> > CPU after we've called disable_nonboot_cpus() so that the
>> > syscore_shutdown()
>> > assumptions are satisfied which the patch doesn't do.
>> >
>> > Chen (I apologize if that's not the right part of your full name to
>> use
>> > here),
>> > do you think that's not necessary and if so, then for what reason?
>>
>> Reboot and poweroff are both OK after I move syscore_shutdown(), I also
>> don't
>> know whether we should disable interrupts. Since you are the author of
>> commit 40dc166c (PM / Core: Introduce struct syscore_ops for core
>> subsystems
>>  PM), please tell me why syscore_shutdown() need interrupt disabled?
>
> Well, the syscore_ callbacks generally do things that may go wrong when an
> interrupt comes in while they are being done.  At least that's what
> happens
> for the suspend/resume syscore_ callbacks, but I'm not really sure about
> the
> shutdown ones.
>
> I suppose we can leave it as is for the time being.  At least I'm not
> aware of
> any problems related to that.
OK, linux-3.0, 3.2 and 3.4 -stable branch also need this patch because they
also have merged the commit f96972f2d (kernel/sys.c: call
disable_nonboot_cpus() in kernel_restart())

>
> Thanks,
> Rafael
>
>
> --
> I speak only for myself.
> Rafael J. Wysocki, Intel Open Source Technology Center.
>