[PATCH v2] arm: cmpxchg syscall should data abort if page not write

[Po-Yu Chuang]

Dear Nicolas Pitre,

On Fri, Mar 18, 2011 at 1:57 AM, Nicolas Pitre <nicolas.pitre at linaro.org> wrote:
> On Thu, 17 Mar 2011, Ashwin Chaugule wrote:
>>
>> Just beginning to understand the subtleties involved, so please
>> correct me if I'm wrong.
>> Wont this patch also fix the problem that was brought up with futexes
>> on ARM SMP ?
>
> Nope.  The code being fixed here was suptly broken so it needs fixing.
> However this code is almost never used, if at all, as it is a fall-back
> solution for when all the better alternatives are not available for some
> reasons (and I'm still wondering what those reasons are for Po-Yu Chuang
> to actually use that code).  In practice this bug should have affected
> no one.

We met this problem while porting an v5 SMP processor because kernel
selects NEEDS_SYSCALL_FOR_CMPXCHG by default if CPU_32v5.

After we added our own implementation of __kuser_cmpxchg, this code
is not needed anymore. But since this is actually a bug, it is still a good
idea to submit a patch. :-)

> If you have SMP then this code is definitively not what you should be
> using.

Best regards,
Po-Yu Chuang