Re: [Kgdb-bugreport] [PATCH 09/37] kgdb,blackfin: Add in kgdb_arch_set_pc for blackfin

[Jason Wessel <jason.wessel@windriver.com>]

Mike Frysinger wrote:
> On Wed, Jan 6, 2010 at 14:43, Jason Wessel wrote:
>   
>> I pulled in this patch, and made the changes the Mike talked about with
>> respect to cleaning up the arch specific kgdb.c for blackfin.  See the
>> attached patch.
>>
>> Your patch and the one here will go into kgdb-next and onto linux-next.
>>     
>
> the Blackfin versions tail into the common probe_kernel_{read,write}
>   

Got it.

> funcs, so your patch causes an infinite recursion in the common case.
> as i hinted earlier, i'd like to see mm/access.c to have something
> like:
> -long probe_kernel_read(void *dst, void *src, size_t size)
> +long __probe_kernel_read(void *dst, void *src, size_t size)
> ...
> +long __weak probe_kernel_read(void *dst, void *src, size_t size)
>     __attribute__((alias("__probe_kernel_read")));
> that way the arch-specific probe_kernel_read can still fall back to
> the common __probe_kernel_read function.  same goes for the write
> function.
>
>   

I don't see a problem with this.  Ingo was the originator of these
functions, which are mainly used for the kernel debugger.  The s390
already made one of them a weak function, so that arch specific code
might have to get looked at as well.

> also, i see you added "notrace" to the Blackfin probe_kernel_write ...
> was that intentional ?  if so, this should probably go into
> include/linux/uaccess.h instead
>   

The notrace was just a cut and paste of the original in mm/maccess.c

Jason.