On 12/18/2014 06:53 PM, David Daney wrote:
> On 12/18/2014 07:09 AM, Markos Chandras wrote:
>> sub $reg, imm is not a real MIPS instruction. The assembler replaces
>> that with 'addi $reg, -imm'.
>
> That is a bug right there. We cannot have faulting instructions like
> this in the kernel.
>
>> However, addi has been removed from R6,
>> so we replace the 'sub' instruction with 'addiu' instead.
>>
>> Signed-off-by: Markos Chandras <markos.chandras@xxxxxxxxxx>
>> ---
>> arch/mips/include/asm/spinlock.h | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/arch/mips/include/asm/spinlock.h
>> b/arch/mips/include/asm/spinlock.h
>> index f63b3543c1a4..c82fc0eefbec 100644
>> --- a/arch/mips/include/asm/spinlock.h
>> +++ b/arch/mips/include/asm/spinlock.h
>> @@ -277,7 +277,7 @@ static inline void arch_read_unlock(arch_rwlock_t
>> *rw)
>> do {
>> __asm__ __volatile__(
>> "1: ll %1, 0(%2)# arch_read_unlock \n"
>> - " sub %1, 1 \n"
>> + " addiu %1, -1 \n"
>
> Instead, how about fixing the real bug by s/sub/subu/.
>
> Would that work for R6 too? If so, just do that.
subu is supported by R6 but i am not sure what the assembler will do in
case of "subu $reg, #imm". It will probably replace that with addiu
--
markos
