Add hint for using IOCB_NOWAIT to handle loop aio command for avoiding to cause write(especially randwrite) perf regression on sparse file. Try IOCB_NOWAIT in the following situations: - backing file is block device - READ aio command - there isn't queued aio non-NOWAIT WRITE, since retry of NOWAIT won't cause contention on WRITE and non-NOWAIT WRITE often implies exclusive lock. With this simple policy, perf regression of randwrte/write on sparse backing file is fixed. Meantime this way addresses perf problem[1] in case of stable FS block mapping via NOWAIT. Link: https://lore.kernel.org/dm-devel/7d6ae2c9-df8e-50d0-7ad6-b787cb3cfab4@xxxxxxxxxx/ Signed-off-by: Ming Lei <ming.lei@xxxxxxxxxx> --- drivers/block/loop.c | 61 ++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 61 insertions(+) diff --git a/drivers/block/loop.c b/drivers/block/loop.c index 542b1fe938a7..5bf14549cf8e 100644 --- a/drivers/block/loop.c +++ b/drivers/block/loop.c @@ -70,6 +70,7 @@ struct loop_device { struct rb_root worker_tree; struct timer_list timer; bool sysfs_inited; + unsigned queued_wait_write; struct request_queue *lo_queue; struct blk_mq_tag_set tag_set; @@ -522,6 +523,30 @@ static int lo_rw_aio(struct loop_device *lo, struct loop_cmd *cmd) return 0; } +static inline bool lo_aio_need_try_nowait(struct loop_device *lo, + struct loop_cmd *cmd) +{ + struct file *file = lo->lo_backing_file; + struct inode *inode = file->f_mapping->host; + struct request *rq = blk_mq_rq_from_pdu(cmd); + + /* NOWAIT works fine for backing block device */ + if (S_ISBLK(inode->i_mode)) + return true; + + /* NOWAIT is supposed to be fine for READ */ + if (req_op(rq) == REQ_OP_READ) + return true; + + /* + * If there is any queued non-NOWAIT async WRITE , don't try new + * NOWAIT WRITE for avoiding contention + * + * Here we focus on handling stable FS block mapping via NOWAIT + */ + return READ_ONCE(lo->queued_wait_write) == 0; +} + static blk_status_t lo_rw_aio_nowait(struct loop_device *lo, struct loop_cmd *cmd) { @@ -531,6 +556,9 @@ static blk_status_t lo_rw_aio_nowait(struct loop_device *lo, if (unlikely(ret < 0)) return BLK_STS_IOERR; + if (!lo_aio_need_try_nowait(lo, cmd)) + return BLK_STS_AGAIN; + cmd->iocb.ki_flags |= IOCB_NOWAIT; ret = lo_submit_rw_aio(lo, cmd, nr_bvec); if (ret == -EAGAIN) @@ -821,12 +849,18 @@ static ssize_t loop_attr_dio_show(struct loop_device *lo, char *buf) return sysfs_emit(buf, "%s\n", dio ? "1" : "0"); } +static ssize_t loop_attr_nr_wait_write_show(struct loop_device *lo, char *buf) +{ + return sysfs_emit(buf, "%u\n", lo->queued_wait_write); +} + LOOP_ATTR_RO(backing_file); LOOP_ATTR_RO(offset); LOOP_ATTR_RO(sizelimit); LOOP_ATTR_RO(autoclear); LOOP_ATTR_RO(partscan); LOOP_ATTR_RO(dio); +LOOP_ATTR_RO(nr_wait_write); static struct attribute *loop_attrs[] = { &loop_attr_backing_file.attr, @@ -835,6 +869,7 @@ static struct attribute *loop_attrs[] = { &loop_attr_autoclear.attr, &loop_attr_partscan.attr, &loop_attr_dio.attr, + &loop_attr_nr_wait_write.attr, NULL, }; @@ -910,6 +945,30 @@ static inline int queue_on_root_worker(struct cgroup_subsys_state *css) } #endif +static inline void loop_inc_wait_write(struct loop_device *lo, struct loop_cmd *cmd) +{ + lockdep_assert_held(&lo->lo_mutex); + + if (cmd->use_aio){ + struct request *rq = blk_mq_rq_from_pdu(cmd); + + if (req_op(rq) == REQ_OP_WRITE) + lo->queued_wait_write += 1; + } +} + +static inline void loop_dec_wait_write(struct loop_device *lo, struct loop_cmd *cmd) +{ + lockdep_assert_held(&lo->lo_mutex); + + if (cmd->use_aio){ + struct request *rq = blk_mq_rq_from_pdu(cmd); + + if (req_op(rq) == REQ_OP_WRITE) + lo->queued_wait_write -= 1; + } +} + static void loop_queue_work(struct loop_device *lo, struct loop_cmd *cmd) { struct request __maybe_unused *rq = blk_mq_rq_from_pdu(cmd); @@ -992,6 +1051,7 @@ static void loop_queue_work(struct loop_device *lo, struct loop_cmd *cmd) work = &lo->rootcg_work; cmd_list = &lo->rootcg_cmd_list; } + loop_inc_wait_write(lo, cmd); list_add_tail(&cmd->list_entry, cmd_list); queue_work(lo->workqueue, work); spin_unlock_irq(&lo->lo_work_lock); @@ -2051,6 +2111,7 @@ static void loop_process_work(struct loop_worker *worker, cond_resched(); spin_lock_irq(&lo->lo_work_lock); + loop_dec_wait_write(lo, cmd); } /* -- 2.47.0
