On Mon, Mar 16, 2009 at 23:23, gyang wrote:
> On Mon, 2009-03-16 at 22:59 -0400, Mike Frysinger wrote:
>> On Mon, Mar 16, 2009 at 22:16,  <[email protected]> wrote:
>> > Disable shm for smp on bf561, because cache coherency issue.
>> >
>> > --- trunk/ipc/shm.c 2009-03-16 12:22:07 UTC (rev 6180)
>> > +++ trunk/ipc/shm.c 2009-03-17 02:16:05 UTC (rev 6181)
>> > @@ -446,6 +446,10 @@ SYSCALL_DEFINE3(shmget, key_t, key,
>> >     struct ipc_ops shm_ops;
>> >     struct ipc_params shm_params;
>> >
>> > +#if defined __ARCH_SYNC_CORE_DCACHE
>> > +   return -EINVAL;
>> > +#endif
>>
>> i dont think mainline will like this.
>
> I'm not going to send this ugly patch to mainline.

all changes to the kernel have to be done with a mind to push to
mainline.  the idea of carrying custom changes forever needs to be
discarded.

>> why dont we modify _sys_shmget
>> in mach-common/entry.S so that it goes to _sys_ni_syscall when SMP is
>> enabled ?
>
> Good idea.

so shmem should work between processes as long as they all bind to the
same processor right ?  how about we just document it that way ?
-mike

_______________________________________________
Linux-kernel-commits mailing list
[email protected]
http://blackfin.uclinux.org/mailman/listinfo/linux-kernel-commits

Reply via email to