* [PATCH 1/3] block: Provide icq in request allocation data
2021-07-12 17:27 [PATCH 0/3] bfq: Limit number of allocated scheduler tags per cgroup Jan Kara
@ 2021-07-12 17:27 ` Jan Kara
2021-07-12 17:27 ` [PATCH 2/3] bfq: Track number of allocated requests in bfq_entity Jan Kara
2021-07-12 17:27 ` [PATCH 3/3] bfq: Limit number of requests consumed by each cgroup Jan Kara
2 siblings, 0 replies; 5+ messages in thread
From: Jan Kara @ 2021-07-12 17:27 UTC (permalink / raw)
To: linux-block; +Cc: Paolo Valente, Jens Axboe, mkoutny, Jan Kara
Currently we lookup ICQ only after the request is allocated. However BFQ
will want to decide how many scheduler tags it allows a given bfq queue
(effectively a process) to consume based on cgroup weight. So lookup ICQ
earlier and provide it in struct blk_mq_alloc_data so that BFQ can use
it.
Signed-off-by: Jan Kara <jack@suse.cz>
---
block/blk-mq-sched.c | 18 ++++++++++--------
block/blk-mq-sched.h | 3 ++-
block/blk-mq.c | 7 ++++---
block/blk-mq.h | 1 +
4 files changed, 17 insertions(+), 12 deletions(-)
diff --git a/block/blk-mq-sched.c b/block/blk-mq-sched.c
index c838d81ac058..3e34f5bb24ae 100644
--- a/block/blk-mq-sched.c
+++ b/block/blk-mq-sched.c
@@ -18,9 +18,8 @@
#include "blk-mq-tag.h"
#include "blk-wbt.h"
-void blk_mq_sched_assign_ioc(struct request *rq)
+struct io_cq *blk_mq_sched_lookup_icq(struct request_queue *q)
{
- struct request_queue *q = rq->q;
struct io_context *ioc;
struct io_cq *icq;
@@ -29,17 +28,20 @@ void blk_mq_sched_assign_ioc(struct request *rq)
*/
ioc = current->io_context;
if (!ioc)
- return;
+ return NULL;
spin_lock_irq(&q->queue_lock);
icq = ioc_lookup_icq(ioc, q);
spin_unlock_irq(&q->queue_lock);
+ if (icq)
+ return icq;
+ return ioc_create_icq(ioc, q, GFP_ATOMIC);
+}
- if (!icq) {
- icq = ioc_create_icq(ioc, q, GFP_ATOMIC);
- if (!icq)
- return;
- }
+void blk_mq_sched_assign_ioc(struct request *rq, struct io_cq *icq)
+{
+ if (!icq)
+ return;
get_io_context(icq->ioc);
rq->elv.icq = icq;
}
diff --git a/block/blk-mq-sched.h b/block/blk-mq-sched.h
index 5246ae040704..4529991e55e6 100644
--- a/block/blk-mq-sched.h
+++ b/block/blk-mq-sched.h
@@ -7,7 +7,8 @@
#define MAX_SCHED_RQ (16 * BLKDEV_MAX_RQ)
-void blk_mq_sched_assign_ioc(struct request *rq);
+struct io_cq *blk_mq_sched_lookup_icq(struct request_queue *q);
+void blk_mq_sched_assign_ioc(struct request *rq, struct io_cq *icq);
bool blk_mq_sched_try_merge(struct request_queue *q, struct bio *bio,
unsigned int nr_segs, struct request **merged_request);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index 2c4ac51e54eb..b9d83644158f 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -333,9 +333,7 @@ static struct request *blk_mq_rq_ctx_init(struct blk_mq_alloc_data *data,
rq->elv.icq = NULL;
if (e && e->type->ops.prepare_request) {
- if (e->type->icq_cache)
- blk_mq_sched_assign_ioc(rq);
-
+ blk_mq_sched_assign_ioc(rq, data->icq);
e->type->ops.prepare_request(rq);
rq->rq_flags |= RQF_ELVPRIV;
}
@@ -360,6 +358,9 @@ static struct request *__blk_mq_alloc_request(struct blk_mq_alloc_data *data)
data->flags |= BLK_MQ_REQ_NOWAIT;
if (e) {
+ if (!op_is_flush(data->cmd_flags) && e->type->icq_cache &&
+ e->type->ops.prepare_request)
+ data->icq = blk_mq_sched_lookup_icq(q);
/*
* Flush/passthrough requests are special and go directly to the
* dispatch list. Don't include reserved tags in the
diff --git a/block/blk-mq.h b/block/blk-mq.h
index d08779f77a26..c502232384c6 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -151,6 +151,7 @@ static inline struct blk_mq_ctx *blk_mq_get_ctx(struct request_queue *q)
struct blk_mq_alloc_data {
/* input parameter */
struct request_queue *q;
+ struct io_cq *icq;
blk_mq_req_flags_t flags;
unsigned int shallow_depth;
unsigned int cmd_flags;
--
2.26.2
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH 2/3] bfq: Track number of allocated requests in bfq_entity
2021-07-12 17:27 [PATCH 0/3] bfq: Limit number of allocated scheduler tags per cgroup Jan Kara
2021-07-12 17:27 ` [PATCH 1/3] block: Provide icq in request allocation data Jan Kara
@ 2021-07-12 17:27 ` Jan Kara
2021-07-12 17:27 ` [PATCH 3/3] bfq: Limit number of requests consumed by each cgroup Jan Kara
2 siblings, 0 replies; 5+ messages in thread
From: Jan Kara @ 2021-07-12 17:27 UTC (permalink / raw)
To: linux-block; +Cc: Paolo Valente, Jens Axboe, mkoutny, Jan Kara
When we want to limit number of requests used by each bfqq and also
cgroup, we need to track also number of requests used by each cgroup.
So track number of allocated requests for each bfq_entity.
Signed-off-by: Jan Kara <jack@suse.cz>
---
block/bfq-iosched.c | 28 ++++++++++++++++++++++------
block/bfq-iosched.h | 5 +++--
2 files changed, 25 insertions(+), 8 deletions(-)
diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
index 727955918563..9ef057dc0028 100644
--- a/block/bfq-iosched.c
+++ b/block/bfq-iosched.c
@@ -1113,7 +1113,8 @@ bfq_bfqq_resume_state(struct bfq_queue *bfqq, struct bfq_data *bfqd,
static int bfqq_process_refs(struct bfq_queue *bfqq)
{
- return bfqq->ref - bfqq->allocated - bfqq->entity.on_st_or_in_serv -
+ return bfqq->ref - bfqq->entity.allocated -
+ bfqq->entity.on_st_or_in_serv -
(bfqq->weight_counter != NULL) - bfqq->stable_ref;
}
@@ -5875,6 +5876,22 @@ static void bfq_rq_enqueued(struct bfq_data *bfqd, struct bfq_queue *bfqq,
}
}
+static void bfqq_request_allocated(struct bfq_queue *bfqq)
+{
+ struct bfq_entity *entity = &bfqq->entity;
+
+ for_each_entity(entity)
+ entity->allocated++;
+}
+
+static void bfqq_request_freed(struct bfq_queue *bfqq)
+{
+ struct bfq_entity *entity = &bfqq->entity;
+
+ for_each_entity(entity)
+ entity->allocated--;
+}
+
/* returns true if it causes the idle timer to be disabled */
static bool __bfq_insert_request(struct bfq_data *bfqd, struct request *rq)
{
@@ -5888,8 +5905,8 @@ static bool __bfq_insert_request(struct bfq_data *bfqd, struct request *rq)
* Release the request's reference to the old bfqq
* and make sure one is taken to the shared queue.
*/
- new_bfqq->allocated++;
- bfqq->allocated--;
+ bfqq_request_allocated(new_bfqq);
+ bfqq_request_freed(bfqq);
new_bfqq->ref++;
/*
* If the bic associated with the process
@@ -6248,8 +6265,7 @@ static void bfq_completed_request(struct bfq_queue *bfqq, struct bfq_data *bfqd)
static void bfq_finish_requeue_request_body(struct bfq_queue *bfqq)
{
- bfqq->allocated--;
-
+ bfqq_request_freed(bfqq);
bfq_put_queue(bfqq);
}
@@ -6669,7 +6685,7 @@ static struct bfq_queue *bfq_init_rq(struct request *rq)
}
}
- bfqq->allocated++;
+ bfqq_request_allocated(bfqq);
bfqq->ref++;
bfq_log_bfqq(bfqd, bfqq, "get_request %p: bfqq %p, %d",
rq, bfqq, bfqq->ref);
diff --git a/block/bfq-iosched.h b/block/bfq-iosched.h
index 99c2a3cb081e..70d4a9b54613 100644
--- a/block/bfq-iosched.h
+++ b/block/bfq-iosched.h
@@ -170,6 +170,9 @@ struct bfq_entity {
/* budget, used also to calculate F_i: F_i = S_i + @budget / @weight */
int budget;
+ /* Number of requests allocated in the subtree of this entity */
+ int allocated;
+
/* device weight, if non-zero, it overrides the default weight of
* bfq_group_data */
int dev_weight;
@@ -266,8 +269,6 @@ struct bfq_queue {
struct request *next_rq;
/* number of sync and async requests queued */
int queued[2];
- /* number of requests currently allocated */
- int allocated;
/* number of pending metadata requests */
int meta_pending;
/* fifo list of requests in sort_list */
--
2.26.2
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH 3/3] bfq: Limit number of requests consumed by each cgroup
2021-07-12 17:27 [PATCH 0/3] bfq: Limit number of allocated scheduler tags per cgroup Jan Kara
2021-07-12 17:27 ` [PATCH 1/3] block: Provide icq in request allocation data Jan Kara
2021-07-12 17:27 ` [PATCH 2/3] bfq: Track number of allocated requests in bfq_entity Jan Kara
@ 2021-07-12 17:27 ` Jan Kara
2 siblings, 0 replies; 5+ messages in thread
From: Jan Kara @ 2021-07-12 17:27 UTC (permalink / raw)
To: linux-block; +Cc: Paolo Valente, Jens Axboe, mkoutny, Jan Kara
When cgroup IO scheduling is used with BFQ it does not really provide
service differentiation if the cgroup drives a big IO depth. That for
example happens with writeback which asynchronously submits lots of IO
but it can happen with AIO as well. The problem is that if we have two
cgroups that submit IO with different weights, the cgroup with higher
weight properly gets more IO time and is able to dispatch more IO.
However this causes lower weight cgroup to accumulate more requests
inside BFQ and eventually lower weight cgroup consumes most of IO
scheduler tags. At that point higher weight cgroup stops getting better
service as it is mostly blocked waiting for a scheduler tag while its
queues inside BFQ are empty and thus lower weight cgroup gets served.
Check how many requests submitting cgroup has allocated in
bfq_limit_depth() and if it consumes more requests than what would
correspond to its weight limit available depth to 1 so that the cgroup
cannot consume many more requests. With this limitation the higher
weight cgroup gets proper service even with writeback.
Signed-off-by: Jan Kara <jack@suse.cz>
---
block/bfq-iosched.c | 54 ++++++++++++++++++++++++++++++---------------
1 file changed, 36 insertions(+), 18 deletions(-)
diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
index 9ef057dc0028..fad54c11c43f 100644
--- a/block/bfq-iosched.c
+++ b/block/bfq-iosched.c
@@ -565,6 +565,22 @@ static struct request *bfq_choose_req(struct bfq_data *bfqd,
}
}
+static bool bfqq_request_over_limit(struct bfq_queue *bfqq, int limit)
+{
+ struct bfq_entity *entity = &bfqq->entity;
+
+ for_each_entity(entity) {
+ if (entity->on_st_or_in_serv &&
+ entity->allocated >= limit * entity->weight /
+ bfq_entity_service_tree(entity)->wsum) {
+ bfq_log_bfqq(bfqq->bfqd, bfqq, "too many requests: allocated %d limit %d weight %d wsum %lu",
+ entity->allocated, limit, entity->weight, bfq_entity_service_tree(entity)->wsum);
+ return true;
+ }
+ }
+ return false;
+}
+
/*
* Async I/O can easily starve sync I/O (both sync reads and sync
* writes), by consuming all tags. Similarly, storms of sync writes,
@@ -575,16 +591,28 @@ static struct request *bfq_choose_req(struct bfq_data *bfqd,
static void bfq_limit_depth(unsigned int op, struct blk_mq_alloc_data *data)
{
struct bfq_data *bfqd = data->q->elevator->elevator_data;
+ struct bfq_io_cq *bic = data->icq ? icq_to_bic(data->icq) : NULL;
+ struct bfq_queue *bfqq = bic ? bic_to_bfqq(bic, op_is_sync(op)) : NULL;
+ int depth;
+ /* Sync reads have full depth available */
if (op_is_sync(op) && !op_is_write(op))
- return;
+ depth = 0;
+ else
+ depth = bfqd->word_depths[!!bfqd->wr_busy_queues][op_is_sync(op)];
- data->shallow_depth =
- bfqd->word_depths[!!bfqd->wr_busy_queues][op_is_sync(op)];
+ /*
+ * Does queue (or any parent entity) exceed number of requests that
+ * should be available to it? Heavily limit depth so that it cannot
+ * consume more available requests and thus starve other entities.
+ */
+ if (bfqq && bfqq_request_over_limit(bfqq, data->q->nr_requests))
+ depth = 1;
bfq_log(bfqd, "[%s] wr_busy %d sync %d depth %u",
- __func__, bfqd->wr_busy_queues, op_is_sync(op),
- data->shallow_depth);
+ __func__, bfqd->wr_busy_queues, op_is_sync(op), depth);
+ if (depth)
+ data->shallow_depth = depth;
}
static struct bfq_queue *
@@ -6848,11 +6876,8 @@ void bfq_put_async_queues(struct bfq_data *bfqd, struct bfq_group *bfqg)
* See the comments on bfq_limit_depth for the purpose of
* the depths set in the function. Return minimum shallow depth we'll use.
*/
-static unsigned int bfq_update_depths(struct bfq_data *bfqd,
- struct sbitmap_queue *bt)
+static void bfq_update_depths(struct bfq_data *bfqd, struct sbitmap_queue *bt)
{
- unsigned int i, j, min_shallow = UINT_MAX;
-
/*
* In-word depths if no bfq_queue is being weight-raised:
* leaving 25% of tags only for sync reads.
@@ -6883,22 +6908,15 @@ static unsigned int bfq_update_depths(struct bfq_data *bfqd,
bfqd->word_depths[1][0] = max(((1U << bt->sb.shift) * 3) >> 4, 1U);
/* no more than ~37% of tags for sync writes (~20% extra tags) */
bfqd->word_depths[1][1] = max(((1U << bt->sb.shift) * 6) >> 4, 1U);
-
- for (i = 0; i < 2; i++)
- for (j = 0; j < 2; j++)
- min_shallow = min(min_shallow, bfqd->word_depths[i][j]);
-
- return min_shallow;
}
static void bfq_depth_updated(struct blk_mq_hw_ctx *hctx)
{
struct bfq_data *bfqd = hctx->queue->elevator->elevator_data;
struct blk_mq_tags *tags = hctx->sched_tags;
- unsigned int min_shallow;
- min_shallow = bfq_update_depths(bfqd, tags->bitmap_tags);
- sbitmap_queue_min_shallow_depth(tags->bitmap_tags, min_shallow);
+ bfq_update_depths(bfqd, tags->bitmap_tags);
+ sbitmap_queue_min_shallow_depth(tags->bitmap_tags, 1);
}
static int bfq_init_hctx(struct blk_mq_hw_ctx *hctx, unsigned int index)
--
2.26.2
^ permalink raw reply related [flat|nested] 5+ messages in thread
* [PATCH 2/3] bfq: Track number of allocated requests in bfq_entity
2021-07-15 13:30 [PATCH 0/3 v2] bfq: Limit number of allocated scheduler tags per cgroup Jan Kara
@ 2021-07-15 13:30 ` Jan Kara
0 siblings, 0 replies; 5+ messages in thread
From: Jan Kara @ 2021-07-15 13:30 UTC (permalink / raw)
To: Jens Axboe; +Cc: linux-block, Paolo Valente, Michal Koutný, Jan Kara
When we want to limit number of requests used by each bfqq and also
cgroup, we need to track also number of requests used by each cgroup.
So track number of allocated requests for each bfq_entity.
Signed-off-by: Jan Kara <jack@suse.cz>
---
block/bfq-iosched.c | 28 ++++++++++++++++++++++------
block/bfq-iosched.h | 5 +++--
2 files changed, 25 insertions(+), 8 deletions(-)
diff --git a/block/bfq-iosched.c b/block/bfq-iosched.c
index 727955918563..9ef057dc0028 100644
--- a/block/bfq-iosched.c
+++ b/block/bfq-iosched.c
@@ -1113,7 +1113,8 @@ bfq_bfqq_resume_state(struct bfq_queue *bfqq, struct bfq_data *bfqd,
static int bfqq_process_refs(struct bfq_queue *bfqq)
{
- return bfqq->ref - bfqq->allocated - bfqq->entity.on_st_or_in_serv -
+ return bfqq->ref - bfqq->entity.allocated -
+ bfqq->entity.on_st_or_in_serv -
(bfqq->weight_counter != NULL) - bfqq->stable_ref;
}
@@ -5875,6 +5876,22 @@ static void bfq_rq_enqueued(struct bfq_data *bfqd, struct bfq_queue *bfqq,
}
}
+static void bfqq_request_allocated(struct bfq_queue *bfqq)
+{
+ struct bfq_entity *entity = &bfqq->entity;
+
+ for_each_entity(entity)
+ entity->allocated++;
+}
+
+static void bfqq_request_freed(struct bfq_queue *bfqq)
+{
+ struct bfq_entity *entity = &bfqq->entity;
+
+ for_each_entity(entity)
+ entity->allocated--;
+}
+
/* returns true if it causes the idle timer to be disabled */
static bool __bfq_insert_request(struct bfq_data *bfqd, struct request *rq)
{
@@ -5888,8 +5905,8 @@ static bool __bfq_insert_request(struct bfq_data *bfqd, struct request *rq)
* Release the request's reference to the old bfqq
* and make sure one is taken to the shared queue.
*/
- new_bfqq->allocated++;
- bfqq->allocated--;
+ bfqq_request_allocated(new_bfqq);
+ bfqq_request_freed(bfqq);
new_bfqq->ref++;
/*
* If the bic associated with the process
@@ -6248,8 +6265,7 @@ static void bfq_completed_request(struct bfq_queue *bfqq, struct bfq_data *bfqd)
static void bfq_finish_requeue_request_body(struct bfq_queue *bfqq)
{
- bfqq->allocated--;
-
+ bfqq_request_freed(bfqq);
bfq_put_queue(bfqq);
}
@@ -6669,7 +6685,7 @@ static struct bfq_queue *bfq_init_rq(struct request *rq)
}
}
- bfqq->allocated++;
+ bfqq_request_allocated(bfqq);
bfqq->ref++;
bfq_log_bfqq(bfqd, bfqq, "get_request %p: bfqq %p, %d",
rq, bfqq, bfqq->ref);
diff --git a/block/bfq-iosched.h b/block/bfq-iosched.h
index 99c2a3cb081e..70d4a9b54613 100644
--- a/block/bfq-iosched.h
+++ b/block/bfq-iosched.h
@@ -170,6 +170,9 @@ struct bfq_entity {
/* budget, used also to calculate F_i: F_i = S_i + @budget / @weight */
int budget;
+ /* Number of requests allocated in the subtree of this entity */
+ int allocated;
+
/* device weight, if non-zero, it overrides the default weight of
* bfq_group_data */
int dev_weight;
@@ -266,8 +269,6 @@ struct bfq_queue {
struct request *next_rq;
/* number of sync and async requests queued */
int queued[2];
- /* number of requests currently allocated */
- int allocated;
/* number of pending metadata requests */
int meta_pending;
/* fifo list of requests in sort_list */
--
2.26.2
^ permalink raw reply related [flat|nested] 5+ messages in thread