[Crash-utility] [PATCH] ARM SMP

[Dave Anderson]

----- "Mika Westerberg" <ext-mika.1.westerberg at nokia.com> wrote:

> Hi Per,
> 
> On Thu, Sep 30, 2010 at 11:19:28AM +0200, ext Per Fransson wrote:
> > 
> > This patch is an attempt to get the ball rolling on SMP support for ARM.
> 
> I noticed that this patch is line-wrapped so it doesn't apply cleanly (or is it
> our brilliant exchange server which mangled that).
> 
> Few questions, see below.

... [ cut ] ...
 
> 
> > @@ -996,20 +1028,20 @@ arm_get_dumpfile_stack_frame(struct bt_info *bt, ulong *nip, ulong *ksp)
> >         if (!ms->crash_task_regs)
> >                 return FALSE;
> > 
> > -       if (tt->panic_task != bt->task || bt->tc->pid != ms->crash_task_pid)
> > -               return FALSE;
> > -
> 
> Was there a reason to remove the check above? Is it so that when we have SMP
> machine, there is still only one panic'ing task?

Yes, there is still only one panic task.

> Anyway, if this this check is not needed anymore, I guess you can remove the
> whole variable from machspec structure.

Right -- in fact, in the patched arm_get_crash_notes() function, ms->crash_task_pid
gets set to whatever pid was running on the highest cpu  -- whether it was the panic
task or not.  But like you say, with the patch, it's become irrelevant anyway. 

Dave