Re: [PATCH 7/7] dm: optimize clone_rq() when track_peak_rq_based_ios is disabled

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



This is good idea.

Mikulas


On Fri, 13 Sep 2013, Jun'ichi Nomura wrote:

> Other approach might be putting the info on tracepoints.
> Then the tracepoint framework will take care of the static_key thing.
> 
> Since 'the number of bios in a request' is a block-layer generic info,
> you could extend the block layer events instead of having your own.
> 
> Attached patch adds it to block_rq_remap event.
> You could do something like this to see requests with many bios:
>   echo 'nr_bios > 32' > /sys/kernel/debug/tracing/events/block/block_rq_remap/filter
>   echo 1 > /sys/kernel/debug/tracing/events/block/block_rq_remap/enable 
> 
> For example,
>      kworker/1:1H-308   [001] d..1   276.242448: block_rq_remap: 8,64 W 1967760 + 512 <- (253,1) 1967760 64
>      kworker/1:1H-308   [001] d..1   276.242482: block_rq_remap: 8,160 W 1968272 + 512 <- (253,1) 1968272 64
>      kworker/1:1H-308   [001] d..1   276.242515: block_rq_remap: 65,0 W 1968784 + 512 <- (253,1) 1968784 64
> # the last item in line is the number of bios. ("64" in this case)
> 
> ---
> Jun'ichi Nomura, NEC Corporation
> 
> 
> diff --git a/include/linux/blkdev.h b/include/linux/blkdev.h
> index 2fdb4a4..0e6f765 100644
> --- a/include/linux/blkdev.h
> +++ b/include/linux/blkdev.h
> @@ -862,6 +862,17 @@ static inline unsigned int blk_rq_get_max_sectors(struct request *rq)
>  	return blk_queue_get_max_sectors(q, rq->cmd_flags);
>  }
>  
> +static inline unsigned int blk_rq_count_bios(struct request *rq)
> +{
> +	unsigned int nr_bios = 0;
> +	struct bio *bio;
> +
> +	__rq_for_each_bio(bio, rq)
> +		nr_bios++;
> +
> +	return nr_bios;
> +}
> +
>  /*
>   * Request issue related functions.
>   */
> diff --git a/include/trace/events/block.h b/include/trace/events/block.h
> index 60ae7c3..4c2301d 100644
> --- a/include/trace/events/block.h
> +++ b/include/trace/events/block.h
> @@ -618,6 +618,7 @@ TRACE_EVENT(block_rq_remap,
>  		__field( unsigned int,	nr_sector	)
>  		__field( dev_t,		old_dev		)
>  		__field( sector_t,	old_sector	)
> +		__field( unsigned int,	nr_bios		)
>  		__array( char,		rwbs,	RWBS_LEN)
>  	),
>  
> @@ -627,15 +628,16 @@ TRACE_EVENT(block_rq_remap,
>  		__entry->nr_sector	= blk_rq_sectors(rq);
>  		__entry->old_dev	= dev;
>  		__entry->old_sector	= from;
> +		__entry->nr_bios	= blk_rq_count_bios(rq);
>  		blk_fill_rwbs(__entry->rwbs, rq->cmd_flags, blk_rq_bytes(rq));
>  	),
>  
> -	TP_printk("%d,%d %s %llu + %u <- (%d,%d) %llu",
> +	TP_printk("%d,%d %s %llu + %u <- (%d,%d) %llu %u",
>  		  MAJOR(__entry->dev), MINOR(__entry->dev), __entry->rwbs,
>  		  (unsigned long long)__entry->sector,
>  		  __entry->nr_sector,
>  		  MAJOR(__entry->old_dev), MINOR(__entry->old_dev),
> -		  (unsigned long long)__entry->old_sector)
> +		  (unsigned long long)__entry->old_sector, __entry->nr_bios)
>  );
>  
>  #endif /* _TRACE_BLOCK_H */
> 

--
dm-devel mailing list
dm-devel@xxxxxxxxxx
https://www.redhat.com/mailman/listinfo/dm-devel




[Index of Archives]     [DM Crypt]     [Fedora Desktop]     [ATA RAID]     [Fedora Marketing]     [Fedora Packaging]     [Fedora SELinux]     [Yosemite Discussion]     [KDE Users]     [Fedora Docs]

  Powered by Linux