[PATCH] HACK: ARM: Fix generic timer broadcast for TWD

[Mark Rutland]

Hi,

On Tue, Feb 12, 2013 at 03:55:16PM +0000, Russell King - ARM Linux wrote:
> On Tue, Feb 12, 2013 at 04:49:54PM +0100, Thierry Reding wrote:
> > Note that I have close to no clue what I'm doing, so the patch might be
> > the completely wrong thing to do. An alternative I had initially thought
> > about was to check for NULL before calling the clock_event_device's
> > .broadcast() function.
> > 
> > The reason why I chose to always assign .broadcast instead is that a
> > previous patch (3d06770: Add generic timer broadcast support) aims at
> > generically implementing broadcast support on ARM and providing a
> > tick_broadcast() implementation for this purpose so it seemed like the
> > right thing to do.
> > 
> > The above-mentioned patch for some reason removed the assignment to the
> > .broadcast() member for apparently no reason, so maybe that was just
> > done by mistake?

Apologies for this. I believe you've hit the same problem as Stephen Warren [1].

The broadcast function is now meant to be set by timer core, but unfortunately
I made a mistake when adding the functionality such that it isn't set for
timers with CLOCK_EVT_FEAT_C3STOP. I posted a fix [2] on Friday, but so far
I've had no reply.

Thomas, are you happy to take the fix at [2] ?

> 
> This is now supposed to be handled by the timer core stuff.  It
> looks to me like 3d06770eef43eaad606e77246bfcc7e82b1d9fb4
> (arm: Add generic timer broadcast support) is slightly busted in
> that it doesn't deal with the #else case you identified below.

That's a minor cosmetic defect. We shouldn't touch smp_timer_broadcast at all
now.

> 
> Certainly, initializing evt->broadcast after the setup is not the
> right solution - it must be done before, but this was removed by
> the above commit.
> 
> So, things to be done here:
> 1. Fix the #else part of the code.
> 2. Fix the reported oops.
> 
> I think this falls to Mark, being the last one to touch this and cause
> this breakage. :)
> 

Hopefully with [2], I've fixed 2. I'll fix 1 shortly. :)

Thanks,
Mark.

[1] http://lists.infradead.org/pipermail/linux-arm-kernel/2013-February/148065.html
[2] http://lists.infradead.org/pipermail/linux-arm-kernel/2013-February/148471.html