Re: [PATCH v2] mmc: Provide tracepoints for request processing

[Baolin Wang]

On 25 March 2016 at 22:10, Jens Axboe <axboe@xxxxxx> wrote:
> On 03/25/2016 02:19 AM, Baolin Wang wrote:
>>
>> diff --git a/drivers/mmc/card/block.c b/drivers/mmc/card/block.c
>> index fe207e5..d372a2d 100644
>> --- a/drivers/mmc/card/block.c
>> +++ b/drivers/mmc/card/block.c
>> @@ -46,6 +46,9 @@
>>
>>   #include "queue.h"
>>
>> +#define CREATE_TRACE_POINTS
>> +#include <trace/events/mmc.h>
>> +
>>   MODULE_ALIAS("mmc:block");
>>   #ifdef MODULE_PARAM_PREFIX
>>   #undef MODULE_PARAM_PREFIX
>> @@ -1709,6 +1712,7 @@ static u8 mmc_blk_prep_packed_list(struct mmc_queue
>> *mq, struct request *req)
>>                 if (phys_segments > max_phys_segs)
>>                         break;
>>
>> +               trace_mmc_block_packed_req(next);
>>                 list_add_tail(&next->queuelist, &mqrq->packed->list);
>>                 cur = next;
>>                 reqs++;
>
>
> This looks like the only valid trace point in the block part.

OK.

>
>> @@ -1870,6 +1874,7 @@ static int mmc_blk_end_packed_req(struct
>> mmc_queue_req *mq_rq)
>>                 }
>>                 list_del_init(&prq->queuelist);
>>                 blk_end_request(prq, 0, blk_rq_bytes(prq));
>> +               trace_mmc_block_req_done(prq);
>
>
> We just logged a complete event through blk_end_request() here.

I'll remove this repeated tracepoint.

>>
>> @@ -1985,6 +1990,7 @@ static int mmc_blk_issue_rw_rq(struct mmc_queue *mq,
>> struct request *rqc)
>>                         } else {
>>                                 ret = blk_end_request(req, 0,
>>                                                 brq->data.bytes_xfered);
>> +                               trace_mmc_block_req_done(req);
>>                         }
>
>
> Ditto

Will remove it too.

>
>> diff --git a/drivers/mmc/card/queue.c b/drivers/mmc/card/queue.c
>> index 6f4323c..d0388cf 100644
>> --- a/drivers/mmc/card/queue.c
>> +++ b/drivers/mmc/card/queue.c
>> @@ -16,6 +16,7 @@
>>   #include <linux/kthread.h>
>>   #include <linux/scatterlist.h>
>>   #include <linux/dma-mapping.h>
>> +#include <trace/events/mmc.h>
>>
>>   #include <linux/mmc/card.h>
>>   #include <linux/mmc/host.h>
>> @@ -64,6 +65,9 @@ static int mmc_queue_thread(void *d)
>>                 mq->mqrq_cur->req = req;
>>                 spin_unlock_irq(q->queue_lock);
>>
>> +               if (req)
>> +                       trace_mmc_queue_fetch(req);
>
>
> Issue event was just logged here, if req != NULL. So also redundant.

Make sense.

>
> Basically only your core mmc events would potentially have merit, the rest
> are all dupes. So nak.

I'll remove the redundant tracepoints. Thanks for your comments.

>
> --
> Jens Axboe
>



-- 
Baolin.wang
Best Regards