Re: [PATCH] ring-buffer: use atomic_[set|clear]_mask in place of a cmpxchg loop

[Patrick Palka]

On Wed, Oct 15, 2014 at 3:08 PM, Patrick Palka <patrick@xxxxxxxxxxxx> wrote:
> On Wed, Oct 15, 2014 at 9:41 AM, Patrick Palka <patrick@xxxxxxxxxxxx> wrote:
>> Instead of open-coding the equivalent cmpxchg loop we can just use
>> atomic_set_mask and atomic_clear_mask to atomically OR or AND
>> the value in &buffer->record_disabled.
>>
>> Cc: Steven Rostedt <rostedt@xxxxxxxxxxx>
>> Cc: Ingo Molnar <mingo@xxxxxxxxxx>
>> Signed-off-by: Patrick Palka <patrick@xxxxxxxxxxxx>
>> ---
>>
>> NB: I have only tested this patch on x86_64 by booting with
>> CONFIG_RING_BUFFER_STARTUP_TEST=y.
>>
>>  kernel/trace/ring_buffer.c | 16 ++--------------
>>  1 file changed, 2 insertions(+), 14 deletions(-)
>>
>> diff --git a/kernel/trace/ring_buffer.c b/kernel/trace/ring_buffer.c
>> index 2d75c94..e15ce2a 100644
>> --- a/kernel/trace/ring_buffer.c
>> +++ b/kernel/trace/ring_buffer.c
>> @@ -3050,13 +3050,7 @@ EXPORT_SYMBOL_GPL(ring_buffer_record_enable);
>>   */
>>  void ring_buffer_record_off(struct ring_buffer *buffer)
>>  {
>> -       unsigned int rd;
>> -       unsigned int new_rd;
>> -
>> -       do {
>> -               rd = atomic_read(&buffer->record_disabled);
>> -               new_rd = rd | RB_BUFFER_OFF;
>> -       } while (atomic_cmpxchg(&buffer->record_disabled, rd, new_rd) != rd);
>> +       atomic_set_mask(RB_BUFFER_OFF, &buffer->record_disabled);
>>  }
>>  EXPORT_SYMBOL_GPL(ring_buffer_record_off);
>>
>> @@ -3073,13 +3067,7 @@ EXPORT_SYMBOL_GPL(ring_buffer_record_off);
>>   */
>>  void ring_buffer_record_on(struct ring_buffer *buffer)
>>  {
>> -       unsigned int rd;
>> -       unsigned int new_rd;
>> -
>> -       do {
>> -               rd = atomic_read(&buffer->record_disabled);
>> -               new_rd = rd & ~RB_BUFFER_OFF;
>> -       } while (atomic_cmpxchg(&buffer->record_disabled, rd, new_rd) != rd);
>> +       atomic_clear_mask(RB_BUFFER_OFF, &buffer->record_disabled);
>>  }
>>  EXPORT_SYMBOL_GPL(ring_buffer_record_on);
>>
>> --
>> 2.1.2.443.g670a3c1
>>
>
> It just occurred to me that atomic_clear_mask and atomic_set_mask
> expect to take an int *, not an atomic_t *, so this patch is bogus...
> Sorry for the noise.

Uh, actually, the generic implementations of these functions (defined
in include/asm-generic/atomic.h) _do_ take an atomic_t *. It's the
specialized x86 definitions of these functions (defined in
arch/x86/include/asm/atomic.h) that take (expect) an int *. So this
patch is not bogus; the x86 specializations for atomic_clear_mask and
atomic_set_mask are. I will try to fix this in a separate patch.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/