Re: [PATCH][2.5] fix preempt-issues with smp_call_function()

[Andrew Morton]

Thomas Schlichter <[e-mail]> wrote:
>
> Hello,
> 
> here is a patch to solve all (I hope I missed none) possible problems that 
> could occur on SMP machines running a preemptible kernel when 
> smp_call_function() calls a function which should be also executed on the 
> current processor.
> 

Patch looks pretty good, thanks.  Fixes a real bug.

I worry a little about the s390/s390x change.  smp_ctl_set_bit() and
smp_ctl_clear_bit().  The functions which are being called on local and
remote are fairly different.  I'm sure it's OK, but...  I changed that bit
to open-code the preempt_disable/enable.

The arch/x86_64/kernel/bluesmoke.c change looks right.  Seems that someone
didn't understand the smp_call_function() API in there.