On 2/15/22 18:09, Wang Jianchao wrote:

On 2022/2/16 5:26 上午, Bart Van Assche wrote:

quoted

On 2/15/22 04:37, Wang Jianchao (Kuaishou) wrote:

quoted

diff --git a/block/Makefile b/block/Makefile
index f38eaa612929..f6a3995af285 100644
--- a/block/Makefile
+++ b/block/Makefile

@@ -17,7 +17,8 @@ obj-$(CONFIG_BLK_DEV_BSGLIB)    += bsg-lib.o
   obj-$(CONFIG_BLK_CGROUP)    += blk-cgroup.o
   obj-$(CONFIG_BLK_CGROUP_RWSTAT)    += blk-cgroup-rwstat.o
   obj-$(CONFIG_BLK_DEV_THROTTLING)    += blk-throttle.o
-obj-$(CONFIG_BLK_CGROUP_IOPRIO)    += blk-ioprio.o
+io-prio-y             := blk-ioprio.o
+obj-$(CONFIG_BLK_CGROUP_IOPRIO)    += io-prio.o
   obj-$(CONFIG_BLK_CGROUP_IOLATENCY)    += blk-iolatency.o
   obj-$(CONFIG_BLK_CGROUP_IOCOST)    += blk-iocost.o
   obj-$(CONFIG_MQ_IOSCHED_DEADLINE)    += mq-deadline.o

Is the above change really necessary?

Except for making maintaining easier on a running system, removing a
rqos policy module with cgroup supporting can release a blk-cgroup
policy slots. As BLKCG_MAX_POLS, the max slots number is fixed now.

It seems like my question was not clear? What I meant is that I think 
that the above changes are not necessary to build blk-ioprio as a kernel 
module.

Thanks,

Bart.