17e149b8 - dm: add 'use_blk_mq' module param and expose in per-device ro sysfs attr

Wed, 11 Mar 2015 19:01:09 +0000

dm: add 'use_blk_mq' module param and expose in per-device ro sysfs attrRequest-based DM's blk-mq support defaults to off; but a user can easilychange the default using the dm_mod.use_blk_mq module/boot option.Also, you can check what mode a given request-based DM device is usingwith: cat /sys/block/dm-X/dm/use_blk_mqThis change enabled further cleanup and reduced work (e.g. themd->io_pool and md->rq_pool isn't created if using blk-mq).Signed-off-by: Mike Snitzer List of files: /linux-6.15/Documentation/ABI/testing/sysfs-block-dm

0ce65797 - dm: impose configurable deadline for dm_request_fn's merge heuristic

Thu, 26 Feb 2015 05:50:28 +0000

dm: impose configurable deadline for dm_request_fn's merge heuristicOtherwise, for sequential workloads, the dm_request_fn can allowexcessive request merging at the expense of increased service time.Add a per-device sysfs attribute to allow the user to control how long arequest, that is a reasonable merge candidate, can be queued on therequest queue. The resolution of this request dispatch deadline is inmicroseconds (ranging from 1 to 100000 usecs), to set a 20us deadline: echo 20 > /sys/block/dm-7/dm/rq_based_seq_io_merge_deadlineThe dm_request_fn's merge heuristic and associated extra accounting isdisabled by default (rq_based_seq_io_merge_deadline is 0).This sysfs attribute is not applicable to bio-based DM devices so itwill only ever report 0 for them.By allowing a request to remain on the queue it will block othersrequests on the queue. But introducing a short dequeue delay has provenvery effective at enabling certain sequential IO workloads on reallyfast, yet IOPS constrained, devices to build up slightly larger IOs --yielding 90+% throughput improvements. Having precise control over thetime taken to wait for larger requests to build affords control beyondthat of waiting for certain IO sizes to accumulate (which would requirea deadline anyway). This knob will only ever make sense with sequentialIO workloads and the particular value used is storage configurationspecific.Given the expected niche use-case for when this knob is useful it hasbeen deemed acceptable to expose this relatively crude method forcrafting optimal IO on specific storage -- especially given the solutionis simple yet effective. In the context of DM multipath, it isadvisable to tune this sysfs attribute to a value that offers the bestperformance for the common case (e.g. if 4 paths are expected active,tune for that; if paths fail then performance may be slightly reduced).Alternatives were explored to have request-based DM autotune this value(e.g. if/when paths fail) but they were quickly deemed too fragile andcomplex to warrant further design and development time. If this problemproves more common as faster storage emerges we'll have to look atelevating a generic solution into the block core.Tested-by: Shiva Krishna Merla Signed-off-by: Mike Snitzer List of files: /linux-6.15/Documentation/ABI/testing/sysfs-block-dm

1d330ec6 - dm: document sysfs entries

Wed, 28 Mar 2012 17:41:23 +0000

dm: document sysfs entriesDescribe attributes provided by device-mapper in /sys/block.Signed-off-by: Milan Broz Signed-off-by: Alasdair G Kergon List of files: /linux-6.15/Documentation/ABI/testing/sysfs-block-dm

Changes in sysfs-block-dm

17e149b8 - dm: add 'use_blk_mq' module param and expose in per-device ro sysfs attr

0ce65797 - dm: impose configurable deadline for dm_request_fn's merge heuristic

1d330ec6 - dm: document sysfs entries