| Message ID | 20230522183845.354920-1-bvanassche@acm.org (mailing list archive) |
|---|---|
| Headers | show |
| Series | Submit zoned writes in order | expand |
On Mon, May 22, 2023 at 11:38:35AM -0700, Bart Van Assche wrote: > - Changed the approach from one requeue list per hctx into preserving one > requeue list per request queue. Can you explain why? The resulting code looks rather odd to me as we now reach out to a global list from the per-hctx run_queue helper, which seems a bit awkware.
On 5/23/23 00:22, Christoph Hellwig wrote: > On Mon, May 22, 2023 at 11:38:35AM -0700, Bart Van Assche wrote: >> - Changed the approach from one requeue list per hctx into preserving one >> requeue list per request queue. > > Can you explain why? The resulting code looks rather odd to me as we > now reach out to a global list from the per-hctx run_queue helper, > which seems a bit awkward. Hi Christoph, This change is based on the assumption that requeuing and flushing are relatively rare events. Do you perhaps want me to change the approach back to one requeue list and one flush list per hctx? Thanks, Bart.
On Tue, May 23, 2023 at 01:04:44PM -0700, Bart Van Assche wrote: >> Can you explain why? The resulting code looks rather odd to me as we >> now reach out to a global list from the per-hctx run_queue helper, >> which seems a bit awkward. > > Hi Christoph, > > This change is based on the assumption that requeuing and flushing are > relatively rare events. The former are, the latter not so much. But more importantly you now look into a global list in the per-hctx dispatch, adding cache line sharing. > Do you perhaps want me to change the approach back > to one requeue list and one flush list per hctx? Unless we have a very good reason to make them global that would be my preference.
Hi Jens, Tests with a zoned UFS prototype have shown that there are plenty of opportunities for reordering in the block layer for zoned writes (REQ_OP_WRITE). The UFS driver is more likely to trigger reordering than other SCSI drivers because it reports BLK_STS_DEV_RESOURCE more often, e.g. during clock scaling. This patch series makes sure that zoned writes are submitted in order without affecting other workloads significantly. Please consider this patch series for the next merge window. Thanks, Bart. Changes compared to v2: - Changed the approach from one requeue list per hctx into preserving one requeue list per request queue. - Rebased on top of Jens' for-next branch. Left out the mq-deadline patches since these are already in the for-next branch. - Modified patch "block: Requeue requests if a CPU is unplugged" such that it always uses the requeue list. - Added a patch that removes blk_mq_kick_requeue_list() and blk_mq_delay_kick_requeue_list(). - Dropped patch "block: mq-deadline: Disable head insertion for zoned writes". - Dropped patch "block: mq-deadline: Introduce a local variable". Changes compared to v1: - Fixed two issues detected by the kernel test robot. Bart Van Assche (7): block: Rename a local variable in blk_mq_requeue_work() block: Send requeued requests to the I/O scheduler block: Requeue requests if a CPU is unplugged block: Make it easier to debug zoned write reordering block: Preserve the order of requeued requests dm: Inline __dm_mq_kick_requeue_list() block: Inline blk_mq_{,delay_}kick_requeue_list() block/blk-flush.c | 4 +- block/blk-mq-debugfs.c | 2 +- block/blk-mq.c | 107 ++++++++++++++++++----------------- drivers/block/ublk_drv.c | 6 +- drivers/block/xen-blkfront.c | 1 - drivers/md/dm-rq.c | 11 +--- drivers/nvme/host/core.c | 2 +- drivers/s390/block/scm_blk.c | 2 +- drivers/scsi/scsi_lib.c | 2 +- include/linux/blk-mq.h | 6 +- include/linux/blkdev.h | 1 - 11 files changed, 69 insertions(+), 75 deletions(-)