Re: [PATCH v2] panic: Make panic_timeout configurable

[Michael Ellerman <mpe@ellerman.id.au>]

On Thu, Nov 21, 2013 at 04:21:30PM -0500, Jason Baron wrote:
> On 11/21/2013 06:16 AM, Michael Ellerman wrote:
> > On Tue, Nov 19, 2013 at 05:04:14PM -0500, Jason Baron wrote:
> >> On 11/19/2013 02:09 AM, Ingo Molnar wrote:
> >>>
> >>> * Jason Baron <jbaron@akamai.com> wrote:
> >>>
> >>>> On 11/18/2013 05:30 PM, Andrew Morton wrote:
> >>>>> On Mon, 18 Nov 2013 21:04:36 +0000 (GMT) Jason Baron <jbaron@akamai.com> wrote:
> >>>>>
> >>>>>> The panic_timeout value can be set via the command line option 'panic=x', or via
> >>>>>> /proc/sys/kernel/panic, however that is not sufficient when the panic occurs
> >>>>>> before we are able to set up these values. Thus, add a CONFIG_PANIC_TIMEOUT
> >>>>>> so that we can set the desired value from the .config.
> >>>>>>
> >>>>>> The default panic_timeout value continues to be 0 - wait forever, 
> >>>>>> except for powerpc and mips, which have been defaulted to 180 and 
> >>>>>> 5 respectively. This is in keeping with the fact that these 
> >>>>>> arches already set panic_timeout in their arch init code. 
> >>>>>> However, I found three exceptions- two in mips and one in powerpc 
> >>>>>> where the settings didn't match these default values. In those 
> >>>>>> cases, I left the arch code so it continues to override, in case 
> >>>>>> the user has not changed from the default. It would nice if these 
> >>>>>> arches had one default value, or if we could determine the 
> >>>>>> correct setting at compile-time.
> > ...
> >>
> >> Sure, I can round up all the related patches in this area that make
> >> sense and re-submit as a series.
> >>
> >> Felipe, would the CONFIG_PANIC_TIMEOUT=xx .config parameter work for your
> >> needs, or would you still like to see the command-line processing moved
> >> up?
> >>
> >> I'd also like to hear from the PowerPC folks about the arch defaults
> >> there. Now, that mips is ok with CONFIG_PANIC_TIMEOUT, PowerPC is the
> >> only arch doing specific initialization of 'panic_timeout'.
> > 
> > Hi Jason,
> > 
> > I think we'd like to choose the value at runtime, as we do now. The
> > powerpc arch supports a wide spread of different hardware, so it's nice
> > to be able to customise the value based on the platform. Also we build a
> > single kernel that boots on many platforms, and so we can't pick the
> > value at compile time.
> 
> Hi,
> 
> Ok, So powerpc sets the timeout to '180' during setup_arch(), but then
> overrides the value to '10' only for pSeries. 

Yep.

> The patch proposed in this thread, sets the default built-in value for
> powerpc to 180 and then continues to override in pSeries, if its
> still 180 (IE the user hasn't requested another value). This allows
> the panic_timeout value to have an effect before we reach the
> arch_init() code. Are you ok with this? That is, are you ok with the
> proposed powerpc bits?

Yes that looks OK to me.

It means we get a non-zero value during early boot, which is nice, the
user can override the value if they wish, and otherwise the existing
behaviour is unchanged.

cheers