| Message ID | 20220217031349.98561-1-jianchao.wan9@gmail.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | blk: make blk-rq-qos policies pluggable and modular | expand |
On 2/16/22 8:13 PM, Wang Jianchao (Kuaishou) wrote: > Hi Jens > > blk-rq-qos is a standalone framework out of io-sched and can be used to > control or observe the IO progress in block-layer with hooks. blk-rq-qos > is a great design but right now, it is totally fixed and built-in and shut > out peoples who want to use it with external module. > > This patchset attempts to make blk-rq-qos framework pluggable and modular. > Then we can update the blk-rq-qos policy module w/o stopping the IO workload. > And it is more convenient to introduce new policy on old machines w/o udgrade > kernel. And we can close all of the blk-rq-qos policy if we needn't any of > them. At the moment, the request_queue.rqos list is empty, we needn't to > waste cpu cyles on them. I like this patchset, would be a lot more convenient and helps efficiency. What kind of testing have you done on it?
On 2022/2/17 11:21 上午, Jens Axboe wrote: > On 2/16/22 8:13 PM, Wang Jianchao (Kuaishou) wrote: >> Hi Jens >> >> blk-rq-qos is a standalone framework out of io-sched and can be used to >> control or observe the IO progress in block-layer with hooks. blk-rq-qos >> is a great design but right now, it is totally fixed and built-in and shut >> out peoples who want to use it with external module. >> >> This patchset attempts to make blk-rq-qos framework pluggable and modular. >> Then we can update the blk-rq-qos policy module w/o stopping the IO workload. >> And it is more convenient to introduce new policy on old machines w/o udgrade >> kernel. And we can close all of the blk-rq-qos policy if we needn't any of >> them. At the moment, the request_queue.rqos list is empty, we needn't to >> waste cpu cyles on them. > > I like this patchset, would be a lot more convenient and helps > efficiency. > > What kind of testing have you done on it? > I have run blktests against this patchset in a VM machine along with a new test case for switching rqos policies while running IO. https://marc.info/?l=linux-block&m=164506848620632&w=2 And also there is a bit different version running on v4.18 with more tests Thanks Jianchao
On Thu, Feb 17, 2022 at 11:13:43AM +0800, Wang Jianchao (Kuaishou) wrote: > Hi Jens > > blk-rq-qos is a standalone framework out of io-sched and can be used to > control or observe the IO progress in block-layer with hooks. blk-rq-qos > is a great design but right now, it is totally fixed and built-in and shut > out peoples who want to use it with external module. And that's got. External modules do not matter. That being said this series has a bunch of nice cleanups, but we really do not need the exports and modular build.
On 2022/2/17 4:43 下午, Christoph Hellwig wrote: > On Thu, Feb 17, 2022 at 11:13:43AM +0800, Wang Jianchao (Kuaishou) wrote: >> Hi Jens >> >> blk-rq-qos is a standalone framework out of io-sched and can be used to >> control or observe the IO progress in block-layer with hooks. blk-rq-qos >> is a great design but right now, it is totally fixed and built-in and shut >> out peoples who want to use it with external module. > > And that's got. External modules do not matter. > > That being said this series has a bunch of nice cleanups, but we really > do not need the exports and modular build. Got it. I will get rid of the code for supporting external module in next patch. Thanks Jianchao
Hi Jens blk-rq-qos is a standalone framework out of io-sched and can be used to control or observe the IO progress in block-layer with hooks. blk-rq-qos is a great design but right now, it is totally fixed and built-in and shut out peoples who want to use it with external module. This patchset attempts to make blk-rq-qos framework pluggable and modular. Then we can update the blk-rq-qos policy module w/o stopping the IO workload. And it is more convenient to introduce new policy on old machines w/o udgrade kernel. And we can close all of the blk-rq-qos policy if we needn't any of them. At the moment, the request_queue.rqos list is empty, we needn't to waste cpu cyles on them. Changes since v3: - Fix some code sytle issues - Rename policies to blk-wbt, blk-iolat, blk-iocost, blk-ioprio Changes since v2: Refactor the patches, - patch01 only provide new interfaces and not export the sysfs interface. - patch02 adapt the wbt to new interface, reserve the previous action when write to wbt_lat - export sysfs interface in last patch when all of the blk-rqos policies are preprared well. Changes since v1: - Just make iocost and iolatency pluggable, then we needn't to export those interfaces - Remove the iostat rqos policy - Rename module of blk-ioprio to io-prio to avoid rename ioprio.c file WangJianchao(6): blk: prepare to make blk-rq-qos pluggable and modular blk-wbt: make wbt pluggable blk-iolatency: make iolatency pluggable blk-iocost: make iocost pluggable blk-ioprio: make ioprio pluggable and modular blk: export the sysfs for switching qos block/Kconfig | 2 +- block/Makefile | 3 +- block/blk-cgroup.c | 11 --- block/blk-iocost.c | 49 +++++++----- block/blk-iolatency.c | 33 +++++++-- block/blk-ioprio.c | 50 ++++++++----- block/blk-ioprio.h | 19 ----- block/blk-mq-debugfs.c | 23 ++---- block/blk-rq-qos.c | 303 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++- block/blk-rq-qos.h | 96 +++++++++--------------- block/blk-sysfs.c | 2 + block/blk-wbt.c | 52 +++++++++++-- block/blk-wbt.h | 7 +- block/blk.h | 6 -- include/linux/blkdev.h | 4 + 15 files changed, 488 insertions(+), 172 deletions(-)