Throttler steals bio before allocating requests for them, thus throttled writeback never reaches congestion. This adds bit WB_write_throttled into per-cgroup bdi congestion control. It's set when write bandwidth limit is exceeded and throttler has at least one bio inside and cleared when last throttled bio is gone. Signed-off-by: Konstantin Khlebnikov <khlebnikov@xxxxxxxxxxxxxx> --- block/blk-throttle.c | 8 ++++++++ include/linux/backing-dev-defs.h | 19 +++++++++++++++++++ include/linux/backing-dev.h | 6 ++++-- 3 files changed, 31 insertions(+), 2 deletions(-) diff --git a/block/blk-throttle.c b/block/blk-throttle.c index 0fea76aa0f3f..7ec0aaf9efa7 100644 --- a/block/blk-throttle.c +++ b/block/blk-throttle.c @@ -1145,6 +1145,10 @@ static void tg_dispatch_one_bio(struct throtl_grp *tg, bool rw) bio = throtl_pop_queued(&sq->queued[rw], &tg_to_put); sq->nr_queued[rw]--; + /* Last throttled @bio is gone, clear congestion bit. */ + if (rw && !sq->nr_queued[rw]) + clear_wb_throttled(tg_to_blkg(tg)->wb_congested); + throtl_charge_bio(tg, bio); /* @@ -2215,6 +2219,10 @@ bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, * its @tg's disptime is not in the future. */ if (tg->flags & THROTL_TG_WAS_EMPTY) { + /* Next write will be throttled, set congestion bit. */ + if (rw && blkg) + set_wb_throttled(blkg->wb_congested); + tg_update_disptime(tg); throtl_schedule_next_dispatch(tg->service_queue.parent_sq, true); } diff --git a/include/linux/backing-dev-defs.h b/include/linux/backing-dev-defs.h index 866c433e7d32..343107fd4eff 100644 --- a/include/linux/backing-dev-defs.h +++ b/include/linux/backing-dev-defs.h @@ -29,6 +29,7 @@ enum wb_state { enum wb_congested_state { WB_async_congested, /* The async (write) queue is getting full */ WB_sync_congested, /* The sync queue is getting full */ + WB_write_throttled, /* Associated blkcg is throttled */ }; typedef int (congested_fn)(void *, int); @@ -200,6 +201,16 @@ static inline void set_bdi_congested(struct backing_dev_info *bdi, int sync) #ifdef CONFIG_CGROUP_WRITEBACK +static inline void set_wb_throttled(struct bdi_writeback_congested *congested) +{ + set_bit(WB_write_throttled, &congested->state); +} + +static inline void clear_wb_throttled(struct bdi_writeback_congested *congested) +{ + clear_bit(WB_write_throttled, &congested->state); +} + /** * wb_tryget - try to increment a wb's refcount * @wb: bdi_writeback to get @@ -244,6 +255,14 @@ static inline bool wb_dying(struct bdi_writeback *wb) #else /* CONFIG_CGROUP_WRITEBACK */ +static inline void set_wb_throttled(struct bdi_writeback_congested *congested) +{ +} + +static inline void clear_wb_throttled(struct bdi_writeback_congested *congested) +{ +} + static inline bool wb_tryget(struct bdi_writeback *wb) { return true; diff --git a/include/linux/backing-dev.h b/include/linux/backing-dev.h index 854e1bdd0b2a..3619641b5ec5 100644 --- a/include/linux/backing-dev.h +++ b/include/linux/backing-dev.h @@ -460,13 +460,15 @@ static inline int inode_read_congested(struct inode *inode) static inline int inode_write_congested(struct inode *inode) { - return inode_congested(inode, 1 << WB_async_congested); + return inode_congested(inode, (1 << WB_async_congested) | + (1 << WB_write_throttled)); } static inline int inode_rw_congested(struct inode *inode) { return inode_congested(inode, (1 << WB_sync_congested) | - (1 << WB_async_congested)); + (1 << WB_async_congested) | + (1 << WB_write_throttled)); } static inline int bdi_congested(struct backing_dev_info *bdi, int cong_bits) -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx"> email@xxxxxxxxx </a>