From 14ccb66b3f585b2bc21e7256c96090abed5a512c Mon Sep 17 00:00:00 2001 From: Christoph Hellwig Date: Thu, 6 Jun 2019 12:29:01 +0200 Subject: block: remove the bi_phys_segments field in struct bio We only need the number of segments in the blk-mq submission path. Remove the field from struct bio, and return it from a variant of blk_queue_split instead of that it can passed as an argument to those functions that need the value. This also means we stop recounting segments except for cloning and partial segments. To keep the number of arguments in this how path down remove pointless struct request_queue arguments from any of the functions that had it and grew a nr_segs argument. Signed-off-by: Christoph Hellwig Signed-off-by: Jens Axboe --- include/linux/blk_types.h | 6 ------ 1 file changed, 6 deletions(-) (limited to 'include/linux/blk_types.h') diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h index 95202f80676c..6a53799c3fe2 100644 --- a/include/linux/blk_types.h +++ b/include/linux/blk_types.h @@ -154,11 +154,6 @@ struct bio { blk_status_t bi_status; u8 bi_partno; - /* Number of segments in this BIO after - * physical address coalescing is performed. - */ - unsigned int bi_phys_segments; - struct bvec_iter bi_iter; atomic_t __bi_remaining; @@ -210,7 +205,6 @@ struct bio { */ enum { BIO_NO_PAGE_REF, /* don't put release vec pages */ - BIO_SEG_VALID, /* bi_phys_segments valid */ BIO_CLONED, /* doesn't own data */ BIO_BOUNCED, /* bio is a bounce bio */ BIO_USER_MAPPED, /* contains user pages */ -- cgit v1.3-14-g43fede From d3f77dfdc71835f8db71ca57d272b1fbec9dfc18 Mon Sep 17 00:00:00 2001 From: Tejun Heo Date: Thu, 27 Jun 2019 13:39:52 -0700 Subject: blkcg: implement REQ_CGROUP_PUNT When a shared kthread needs to issue a bio for a cgroup, doing so synchronously can lead to priority inversions as the kthread can be trapped waiting for that cgroup. This patch implements REQ_CGROUP_PUNT flag which makes submit_bio() punt the actual issuing to a dedicated per-blkcg work item to avoid such priority inversions. This will be used to fix priority inversions in btrfs compression and should be generally useful as we grow filesystem support for comprehensive IO control. Cc: Chris Mason Reviewed-by: Josef Bacik Reviewed-by: Jan Kara Signed-off-by: Tejun Heo Signed-off-by: Jens Axboe --- block/blk-cgroup.c | 53 +++++++++++++++++++++++++++++++++++++++++++++ block/blk-core.c | 3 +++ include/linux/backing-dev.h | 1 + include/linux/blk-cgroup.h | 16 +++++++++++++- include/linux/blk_types.h | 10 +++++++++ include/linux/writeback.h | 13 ++++++++--- 6 files changed, 92 insertions(+), 4 deletions(-) (limited to 'include/linux/blk_types.h') diff --git a/block/blk-cgroup.c b/block/blk-cgroup.c index ad7a91dec934..24ed26957367 100644 --- a/block/blk-cgroup.c +++ b/block/blk-cgroup.c @@ -55,6 +55,7 @@ static struct blkcg_policy *blkcg_policy[BLKCG_MAX_POLS]; static LIST_HEAD(all_blkcgs); /* protected by blkcg_pol_mutex */ static bool blkcg_debug_stats = false; +static struct workqueue_struct *blkcg_punt_bio_wq; static bool blkcg_policy_enabled(struct request_queue *q, const struct blkcg_policy *pol) @@ -89,6 +90,8 @@ static void __blkg_release(struct rcu_head *rcu) { struct blkcg_gq *blkg = container_of(rcu, struct blkcg_gq, rcu_head); + WARN_ON(!bio_list_empty(&blkg->async_bios)); + /* release the blkcg and parent blkg refs this blkg has been holding */ css_put(&blkg->blkcg->css); if (blkg->parent) @@ -114,6 +117,23 @@ static void blkg_release(struct percpu_ref *ref) call_rcu(&blkg->rcu_head, __blkg_release); } +static void blkg_async_bio_workfn(struct work_struct *work) +{ + struct blkcg_gq *blkg = container_of(work, struct blkcg_gq, + async_bio_work); + struct bio_list bios = BIO_EMPTY_LIST; + struct bio *bio; + + /* as long as there are pending bios, @blkg can't go away */ + spin_lock_bh(&blkg->async_bio_lock); + bio_list_merge(&bios, &blkg->async_bios); + bio_list_init(&blkg->async_bios); + spin_unlock_bh(&blkg->async_bio_lock); + + while ((bio = bio_list_pop(&bios))) + submit_bio(bio); +} + /** * blkg_alloc - allocate a blkg * @blkcg: block cgroup the new blkg is associated with @@ -142,6 +162,9 @@ static struct blkcg_gq *blkg_alloc(struct blkcg *blkcg, struct request_queue *q, blkg->q = q; INIT_LIST_HEAD(&blkg->q_node); + spin_lock_init(&blkg->async_bio_lock); + bio_list_init(&blkg->async_bios); + INIT_WORK(&blkg->async_bio_work, blkg_async_bio_workfn); blkg->blkcg = blkcg; for (i = 0; i < BLKCG_MAX_POLS; i++) { @@ -1528,6 +1551,25 @@ out_unlock: } EXPORT_SYMBOL_GPL(blkcg_policy_unregister); +bool __blkcg_punt_bio_submit(struct bio *bio) +{ + struct blkcg_gq *blkg = bio->bi_blkg; + + /* consume the flag first */ + bio->bi_opf &= ~REQ_CGROUP_PUNT; + + /* never bounce for the root cgroup */ + if (!blkg->parent) + return false; + + spin_lock_bh(&blkg->async_bio_lock); + bio_list_add(&blkg->async_bios, bio); + spin_unlock_bh(&blkg->async_bio_lock); + + queue_work(blkcg_punt_bio_wq, &blkg->async_bio_work); + return true; +} + /* * Scale the accumulated delay based on how long it has been since we updated * the delay. We only call this when we are adding delay, in case it's been a @@ -1729,5 +1771,16 @@ void blkcg_add_delay(struct blkcg_gq *blkg, u64 now, u64 delta) atomic64_add(delta, &blkg->delay_nsec); } +static int __init blkcg_init(void) +{ + blkcg_punt_bio_wq = alloc_workqueue("blkcg_punt_bio", + WQ_MEM_RECLAIM | WQ_FREEZABLE | + WQ_UNBOUND | WQ_SYSFS, 0); + if (!blkcg_punt_bio_wq) + return -ENOMEM; + return 0; +} +subsys_initcall(blkcg_init); + module_param(blkcg_debug_stats, bool, 0644); MODULE_PARM_DESC(blkcg_debug_stats, "True if you want debug stats, false if not"); diff --git a/block/blk-core.c b/block/blk-core.c index edd009213f5b..260e36a2c343 100644 --- a/block/blk-core.c +++ b/block/blk-core.c @@ -1128,6 +1128,9 @@ EXPORT_SYMBOL_GPL(direct_make_request); */ blk_qc_t submit_bio(struct bio *bio) { + if (blkcg_punt_bio_submit(bio)) + return BLK_QC_T_NONE; + /* * If it's a regular read/write or a barrier with data attached, * go through the normal accounting stuff before submission. diff --git a/include/linux/backing-dev.h b/include/linux/backing-dev.h index f9b029180241..35b31d176f74 100644 --- a/include/linux/backing-dev.h +++ b/include/linux/backing-dev.h @@ -48,6 +48,7 @@ extern spinlock_t bdi_lock; extern struct list_head bdi_list; extern struct workqueue_struct *bdi_wq; +extern struct workqueue_struct *bdi_async_bio_wq; static inline bool wb_has_dirty_io(struct bdi_writeback *wb) { diff --git a/include/linux/blk-cgroup.h b/include/linux/blk-cgroup.h index 33f23a858438..689a58231288 100644 --- a/include/linux/blk-cgroup.h +++ b/include/linux/blk-cgroup.h @@ -132,13 +132,17 @@ struct blkcg_gq { struct blkg_policy_data *pd[BLKCG_MAX_POLS]; - struct rcu_head rcu_head; + spinlock_t async_bio_lock; + struct bio_list async_bios; + struct work_struct async_bio_work; atomic_t use_delay; atomic64_t delay_nsec; atomic64_t delay_start; u64 last_delay; int last_use; + + struct rcu_head rcu_head; }; typedef struct blkcg_policy_data *(blkcg_pol_alloc_cpd_fn)(gfp_t gfp); @@ -701,6 +705,15 @@ static inline bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg struct bio *bio) { return false; } #endif +bool __blkcg_punt_bio_submit(struct bio *bio); + +static inline bool blkcg_punt_bio_submit(struct bio *bio) +{ + if (bio->bi_opf & REQ_CGROUP_PUNT) + return __blkcg_punt_bio_submit(bio); + else + return false; +} static inline void blkcg_bio_issue_init(struct bio *bio) { @@ -848,6 +861,7 @@ static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; } static inline void blkg_get(struct blkcg_gq *blkg) { } static inline void blkg_put(struct blkcg_gq *blkg) { } +static inline bool blkcg_punt_bio_submit(struct bio *bio) { return false; } static inline void blkcg_bio_issue_init(struct bio *bio) { } static inline bool blkcg_bio_issue_check(struct request_queue *q, struct bio *bio) { return true; } diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h index 6a53799c3fe2..feff3fe4467e 100644 --- a/include/linux/blk_types.h +++ b/include/linux/blk_types.h @@ -311,6 +311,14 @@ enum req_flag_bits { __REQ_RAHEAD, /* read ahead, can fail anytime */ __REQ_BACKGROUND, /* background IO */ __REQ_NOWAIT, /* Don't wait if request will block */ + /* + * When a shared kthread needs to issue a bio for a cgroup, doing + * so synchronously can lead to priority inversions as the kthread + * can be trapped waiting for that cgroup. CGROUP_PUNT flag makes + * submit_bio() punt the actual issuing to a dedicated per-blkcg + * work item to avoid such priority inversions. + */ + __REQ_CGROUP_PUNT, /* command specific flags for REQ_OP_WRITE_ZEROES: */ __REQ_NOUNMAP, /* do not free blocks when zeroing */ @@ -337,6 +345,8 @@ enum req_flag_bits { #define REQ_RAHEAD (1ULL << __REQ_RAHEAD) #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND) #define REQ_NOWAIT (1ULL << __REQ_NOWAIT) +#define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT) + #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP) #define REQ_HIPRI (1ULL << __REQ_HIPRI) diff --git a/include/linux/writeback.h b/include/linux/writeback.h index e056a22075cf..8945aac31392 100644 --- a/include/linux/writeback.h +++ b/include/linux/writeback.h @@ -78,6 +78,8 @@ struct writeback_control { */ unsigned no_cgroup_owner:1; + unsigned punt_to_cgroup:1; /* cgrp punting, see __REQ_CGROUP_PUNT */ + #ifdef CONFIG_CGROUP_WRITEBACK struct bdi_writeback *wb; /* wb this writeback is issued under */ struct inode *inode; /* inode being written out */ @@ -94,12 +96,17 @@ struct writeback_control { static inline int wbc_to_write_flags(struct writeback_control *wbc) { + int flags = 0; + + if (wbc->punt_to_cgroup) + flags = REQ_CGROUP_PUNT; + if (wbc->sync_mode == WB_SYNC_ALL) - return REQ_SYNC; + flags |= REQ_SYNC; else if (wbc->for_kupdate || wbc->for_background) - return REQ_BACKGROUND; + flags |= REQ_BACKGROUND; - return 0; + return flags; } static inline struct cgroup_subsys_state * -- cgit v1.3-14-g43fede From 893a1c97205a3ece0cbb3f571a3b972080f3b4c7 Mon Sep 17 00:00:00 2001 From: Jens Axboe Date: Tue, 16 Jul 2019 13:55:23 -0600 Subject: blk-mq: allow REQ_NOWAIT to return an error inline By default, if a caller sets REQ_NOWAIT and we need to block, we'll return -EAGAIN through the bio->bi_end_io() callback. For some use cases, this makes it hard to use. Allow a caller to ask for inline return of errors related to blocking by also setting REQ_NOWAIT_INLINE. Signed-off-by: Jens Axboe --- block/blk-mq.c | 8 ++++++-- include/linux/blk_types.h | 5 ++++- 2 files changed, 10 insertions(+), 3 deletions(-) (limited to 'include/linux/blk_types.h') diff --git a/block/blk-mq.c b/block/blk-mq.c index b038ec680e84..2bc2c0705660 100644 --- a/block/blk-mq.c +++ b/block/blk-mq.c @@ -1960,9 +1960,13 @@ static blk_qc_t blk_mq_make_request(struct request_queue *q, struct bio *bio) rq = blk_mq_get_request(q, bio, &data); if (unlikely(!rq)) { rq_qos_cleanup(q, bio); - if (bio->bi_opf & REQ_NOWAIT) + + cookie = BLK_QC_T_NONE; + if (bio->bi_opf & REQ_NOWAIT_INLINE) + cookie = BLK_QC_T_EAGAIN; + else if (bio->bi_opf & REQ_NOWAIT) bio_wouldblock_error(bio); - return BLK_QC_T_NONE; + return cookie; } trace_block_getrq(q, bio, bio->bi_opf); diff --git a/include/linux/blk_types.h b/include/linux/blk_types.h index feff3fe4467e..1b1fa1557e68 100644 --- a/include/linux/blk_types.h +++ b/include/linux/blk_types.h @@ -311,6 +311,7 @@ enum req_flag_bits { __REQ_RAHEAD, /* read ahead, can fail anytime */ __REQ_BACKGROUND, /* background IO */ __REQ_NOWAIT, /* Don't wait if request will block */ + __REQ_NOWAIT_INLINE, /* Return would-block error inline */ /* * When a shared kthread needs to issue a bio for a cgroup, doing * so synchronously can lead to priority inversions as the kthread @@ -345,6 +346,7 @@ enum req_flag_bits { #define REQ_RAHEAD (1ULL << __REQ_RAHEAD) #define REQ_BACKGROUND (1ULL << __REQ_BACKGROUND) #define REQ_NOWAIT (1ULL << __REQ_NOWAIT) +#define REQ_NOWAIT_INLINE (1ULL << __REQ_NOWAIT_INLINE) #define REQ_CGROUP_PUNT (1ULL << __REQ_CGROUP_PUNT) #define REQ_NOUNMAP (1ULL << __REQ_NOUNMAP) @@ -418,12 +420,13 @@ static inline int op_stat_group(unsigned int op) typedef unsigned int blk_qc_t; #define BLK_QC_T_NONE -1U +#define BLK_QC_T_EAGAIN -2U #define BLK_QC_T_SHIFT 16 #define BLK_QC_T_INTERNAL (1U << 31) static inline bool blk_qc_t_valid(blk_qc_t cookie) { - return cookie != BLK_QC_T_NONE; + return cookie != BLK_QC_T_NONE && cookie != BLK_QC_T_EAGAIN; } static inline unsigned int blk_qc_t_to_queue_num(blk_qc_t cookie) -- cgit v1.3-14-g43fede