From: John Garry <john.garry@huawei.com>
To: <axboe@kernel.dk>, <jejb@linux.ibm.com>,
<martin.petersen@oracle.com>, <don.brace@microsemi.com>,
<kashyap.desai@broadcom.com>, <ming.lei@redhat.com>,
<bvanassche@acm.org>, <dgilbert@interlog.com>,
<paolo.valente@linaro.org>, <hare@suse.de>, <hch@lst.de>
Cc: <sumit.saxena@broadcom.com>, <linux-block@vger.kernel.org>,
<linux-kernel@vger.kernel.org>, <linux-scsi@vger.kernel.org>,
<esc.storagedev@microsemi.com>, <megaraidlinux.pdl@broadcom.com>,
<chenxiang66@hisilicon.com>, <luojiaxing@huawei.com>,
John Garry <john.garry@huawei.com>
Subject: [PATCH v8 06/18] blk-mq: Facilitate a shared sbitmap per tagset
Date: Wed, 19 Aug 2020 23:20:24 +0800 [thread overview]
Message-ID: <1597850436-116171-7-git-send-email-john.garry@huawei.com> (raw)
In-Reply-To: <1597850436-116171-1-git-send-email-john.garry@huawei.com>
Some SCSI HBAs (such as HPSA, megaraid, mpt3sas, hisi_sas_v3 ..) support
multiple reply queues with single hostwide tags.
In addition, these drivers want to use interrupt assignment in
pci_alloc_irq_vectors(PCI_IRQ_AFFINITY). However, as discussed in [0],
CPU hotplug may cause in-flight IO completion to not be serviced when an
interrupt is shutdown. That problem is solved in commit bf0beec0607d
("blk-mq: drain I/O when all CPUs in a hctx are offline").
However, to take advantage of that blk-mq feature, the HBA HW queuess are
required to be mapped to that of the blk-mq hctx's; to do that, the HBA HW
queues need to be exposed to the upper layer.
In making that transition, the per-SCSI command request tags are no
longer unique per Scsi host - they are just unique per hctx. As such, the
HBA LLDD would have to generate this tag internally, which has a certain
performance overhead.
However another problem is that blk-mq assumes the host may accept
(Scsi_host.can_queue * #hw queue) commands. In commit 6eb045e092ef ("scsi:
core: avoid host-wide host_busy counter for scsi_mq"), the Scsi host busy
counter was removed, which would stop the LLDD being sent more than
.can_queue commands; however, it should still be ensured that the block
layer does not issue more than .can_queue commands to the Scsi host.
To solve this problem, introduce a shared sbitmap per blk_mq_tag_set,
which may be requested at init time.
New flag BLK_MQ_F_TAG_HCTX_SHARED should be set when requesting the
tagset to indicate whether the shared sbitmap should be used.
Even when BLK_MQ_F_TAG_HCTX_SHARED is set, a full set of tags and requests
are still allocated per hctx; the reason for this is that if tags and
requests were only allocated for a single hctx - like hctx0 - it may break
block drivers which expect a request be associated with a specific hctx,
i.e. not always hctx0. This will introduce extra memory usage.
This change is based on work originally from Ming Lei in [1] and from
Bart's suggestion in [2].
[0] https://lore.kernel.org/linux-block/alpine.DEB.2.21.1904051331270.1802@nanos.tec.linutronix.de/
[1] https://lore.kernel.org/linux-block/20190531022801.10003-1-ming.lei@redhat.com/
[2] https://lore.kernel.org/linux-block/ff77beff-5fd9-9f05-12b6-826922bace1f@huawei.com/T/#m3db0a602f095cbcbff27e9c884d6b4ae826144be
Tested-by: Don Brace<don.brace@microsemi.com> #SCSI resv cmds patches used
Signed-off-by: John Garry <john.garry@huawei.com>
---
block/blk-mq-sched.c | 8 ++++---
block/blk-mq-tag.c | 51 ++++++++++++++++++++++++++++++++++++++----
block/blk-mq-tag.h | 7 ++++++
block/blk-mq.c | 15 +++++++++++++
block/blk-mq.h | 5 +++++
include/linux/blk-mq.h | 6 +++++
6 files changed, 85 insertions(+), 7 deletions(-)
diff --git a/block/blk-mq-sched.c b/block/blk-mq-sched.c
index ee970d848e62..777be967c4f2 100644
--- a/block/blk-mq-sched.c
+++ b/block/blk-mq-sched.c
@@ -607,7 +607,7 @@ static void blk_mq_sched_free_tags(struct blk_mq_tag_set *set,
struct blk_mq_hw_ctx *hctx,
unsigned int hctx_idx)
{
- unsigned int flags = set->flags;
+ unsigned int flags = set->flags & ~BLK_MQ_F_TAG_HCTX_SHARED;
if (hctx->sched_tags) {
blk_mq_free_rqs(set, hctx->sched_tags, hctx_idx);
@@ -621,7 +621,8 @@ static int blk_mq_sched_alloc_tags(struct request_queue *q,
unsigned int hctx_idx)
{
struct blk_mq_tag_set *set = q->tag_set;
- unsigned int flags = set->flags;
+ /* Clear HCTX_SHARED so tags are init'ed */
+ unsigned int flags = set->flags & ~BLK_MQ_F_TAG_HCTX_SHARED;
int ret;
hctx->sched_tags = blk_mq_alloc_rq_map(set, hctx_idx, q->nr_requests,
@@ -643,7 +644,8 @@ static void blk_mq_sched_tags_teardown(struct request_queue *q)
int i;
queue_for_each_hw_ctx(q, hctx, i) {
- unsigned int flags = hctx->flags;
+ /* Clear HCTX_SHARED so tags are freed */
+ unsigned int flags = hctx->flags & ~BLK_MQ_F_TAG_HCTX_SHARED;
if (hctx->sched_tags) {
blk_mq_free_rq_map(hctx->sched_tags, flags);
diff --git a/block/blk-mq-tag.c b/block/blk-mq-tag.c
index 77fbeae05745..c6d7ebc62bdb 100644
--- a/block/blk-mq-tag.c
+++ b/block/blk-mq-tag.c
@@ -197,7 +197,7 @@ static bool bt_iter(struct sbitmap *bitmap, unsigned int bitnr, void *data)
* We can hit rq == NULL here, because the tagging functions
* test and set the bit before assigning ->rqs[].
*/
- if (rq && rq->q == hctx->queue)
+ if (rq && rq->q == hctx->queue && rq->mq_hctx == hctx)
return iter_data->fn(hctx, rq, iter_data->data, reserved);
return true;
}
@@ -450,6 +450,38 @@ static int blk_mq_init_bitmap_tags(struct blk_mq_tags *tags,
return -ENOMEM;
}
+int blk_mq_init_shared_sbitmap(struct blk_mq_tag_set *set, unsigned int flags)
+{
+ unsigned int depth = set->queue_depth - set->reserved_tags;
+ int alloc_policy = BLK_MQ_FLAG_TO_ALLOC_POLICY(set->flags);
+ bool round_robin = alloc_policy == BLK_TAG_ALLOC_RR;
+ int i, node = set->numa_node;
+
+ if (bt_alloc(&set->__bitmap_tags, depth, round_robin, node))
+ return -ENOMEM;
+ if (bt_alloc(&set->__breserved_tags, set->reserved_tags,
+ round_robin, node))
+ goto free_bitmap_tags;
+
+ for (i = 0; i < set->nr_hw_queues; i++) {
+ struct blk_mq_tags *tags = set->tags[i];
+
+ tags->bitmap_tags = &set->__bitmap_tags;
+ tags->breserved_tags = &set->__breserved_tags;
+ }
+
+ return 0;
+free_bitmap_tags:
+ sbitmap_queue_free(&set->__bitmap_tags);
+ return -ENOMEM;
+}
+
+void blk_mq_exit_shared_sbitmap(struct blk_mq_tag_set *set)
+{
+ sbitmap_queue_free(&set->__bitmap_tags);
+ sbitmap_queue_free(&set->__breserved_tags);
+}
+
struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
unsigned int reserved_tags,
int node, unsigned int flags)
@@ -469,6 +501,9 @@ struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
tags->nr_tags = total_tags;
tags->nr_reserved_tags = reserved_tags;
+ if (flags & BLK_MQ_F_TAG_HCTX_SHARED)
+ return tags;
+
if (blk_mq_init_bitmap_tags(tags, node, alloc_policy) < 0) {
kfree(tags);
return NULL;
@@ -478,8 +513,10 @@ struct blk_mq_tags *blk_mq_init_tags(unsigned int total_tags,
void blk_mq_free_tags(struct blk_mq_tags *tags, unsigned int flags)
{
- sbitmap_queue_free(&tags->__bitmap_tags);
- sbitmap_queue_free(&tags->__breserved_tags);
+ if (!(flags & BLK_MQ_F_TAG_HCTX_SHARED)) {
+ sbitmap_queue_free(tags->bitmap_tags);
+ sbitmap_queue_free(tags->breserved_tags);
+ }
kfree(tags);
}
@@ -498,7 +535,8 @@ int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
*/
if (tdepth > tags->nr_tags) {
struct blk_mq_tag_set *set = hctx->queue->tag_set;
- unsigned int flags = set->flags;
+ /* Only sched tags can grow, so clear HCTX_SHARED flag */
+ unsigned int flags = set->flags & ~BLK_MQ_F_TAG_HCTX_SHARED;
struct blk_mq_tags *new;
bool ret;
@@ -537,6 +575,11 @@ int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
return 0;
}
+void blk_mq_tag_resize_shared_sbitmap(struct blk_mq_tag_set *set, unsigned int size)
+{
+ sbitmap_queue_resize(&set->__bitmap_tags, size - set->reserved_tags);
+}
+
/**
* blk_mq_unique_tag() - return a tag that is unique queue-wide
* @rq: request for which to compute a unique tag
diff --git a/block/blk-mq-tag.h b/block/blk-mq-tag.h
index 5b6f25bd1966..8f4cce52ba2d 100644
--- a/block/blk-mq-tag.h
+++ b/block/blk-mq-tag.h
@@ -29,12 +29,19 @@ extern struct blk_mq_tags *blk_mq_init_tags(unsigned int nr_tags,
int node, unsigned int flags);
extern void blk_mq_free_tags(struct blk_mq_tags *tags, unsigned int flags);
+extern int blk_mq_init_shared_sbitmap(struct blk_mq_tag_set *set,
+ unsigned int flags);
+extern void blk_mq_exit_shared_sbitmap(struct blk_mq_tag_set *set);
+
extern unsigned int blk_mq_get_tag(struct blk_mq_alloc_data *data);
extern void blk_mq_put_tag(struct blk_mq_tags *tags, struct blk_mq_ctx *ctx,
unsigned int tag);
extern int blk_mq_tag_update_depth(struct blk_mq_hw_ctx *hctx,
struct blk_mq_tags **tags,
unsigned int depth, bool can_grow);
+extern void blk_mq_tag_resize_shared_sbitmap(struct blk_mq_tag_set *set,
+ unsigned int size);
+
extern void blk_mq_tag_wakeup_all(struct blk_mq_tags *tags, bool);
void blk_mq_queue_tag_busy_iter(struct request_queue *q, busy_iter_fn *fn,
void *priv);
diff --git a/block/blk-mq.c b/block/blk-mq.c
index ce712ec818e4..a500d1dfa1bd 100644
--- a/block/blk-mq.c
+++ b/block/blk-mq.c
@@ -3430,11 +3430,21 @@ int blk_mq_alloc_tag_set(struct blk_mq_tag_set *set)
if (ret)
goto out_free_mq_map;
+ if (blk_mq_is_sbitmap_shared(set->flags)) {
+ if (blk_mq_init_shared_sbitmap(set, set->flags)) {
+ ret = -ENOMEM;
+ goto out_free_mq_rq_maps;
+ }
+ }
+
mutex_init(&set->tag_list_lock);
INIT_LIST_HEAD(&set->tag_list);
return 0;
+out_free_mq_rq_maps:
+ for (i = 0; i < set->nr_hw_queues; i++)
+ blk_mq_free_map_and_requests(set, i);
out_free_mq_map:
for (i = 0; i < set->nr_maps; i++) {
kfree(set->map[i].mq_map);
@@ -3453,6 +3463,9 @@ void blk_mq_free_tag_set(struct blk_mq_tag_set *set)
for (i = 0; i < set->nr_hw_queues; i++)
blk_mq_free_map_and_requests(set, i);
+ if (blk_mq_is_sbitmap_shared(set->flags))
+ blk_mq_exit_shared_sbitmap(set);
+
for (j = 0; j < set->nr_maps; j++) {
kfree(set->map[j].mq_map);
set->map[j].mq_map = NULL;
@@ -3489,6 +3502,8 @@ int blk_mq_update_nr_requests(struct request_queue *q, unsigned int nr)
if (!hctx->sched_tags) {
ret = blk_mq_tag_update_depth(hctx, &hctx->tags, nr,
false);
+ if (!ret && blk_mq_is_sbitmap_shared(set->flags))
+ blk_mq_tag_resize_shared_sbitmap(set, nr);
} else {
ret = blk_mq_tag_update_depth(hctx, &hctx->sched_tags,
nr, true);
diff --git a/block/blk-mq.h b/block/blk-mq.h
index b86e7fe47789..3acdb56065e1 100644
--- a/block/blk-mq.h
+++ b/block/blk-mq.h
@@ -159,6 +159,11 @@ struct blk_mq_alloc_data {
struct blk_mq_hw_ctx *hctx;
};
+static inline bool blk_mq_is_sbitmap_shared(unsigned int flags)
+{
+ return flags & BLK_MQ_F_TAG_HCTX_SHARED;
+}
+
static inline struct blk_mq_tags *blk_mq_tags_from_data(struct blk_mq_alloc_data *data)
{
if (data->q->elevator)
diff --git a/include/linux/blk-mq.h b/include/linux/blk-mq.h
index 95aca7aa8957..03b1f39d5a72 100644
--- a/include/linux/blk-mq.h
+++ b/include/linux/blk-mq.h
@@ -231,6 +231,9 @@ enum hctx_type {
* @flags: Zero or more BLK_MQ_F_* flags.
* @driver_data: Pointer to data owned by the block driver that created this
* tag set.
+ * @__bitmap_tags: A shared tags sbitmap, used over all hctx's
+ * @__breserved_tags:
+ * A shared reserved tags sbitmap, used over all hctx's
* @tags: Tag sets. One tag set per hardware queue. Has @nr_hw_queues
* elements.
* @tag_list_lock: Serializes tag_list accesses.
@@ -250,6 +253,8 @@ struct blk_mq_tag_set {
unsigned int flags;
void *driver_data;
+ struct sbitmap_queue __bitmap_tags;
+ struct sbitmap_queue __breserved_tags;
struct blk_mq_tags **tags;
struct mutex tag_list_lock;
@@ -384,6 +389,7 @@ enum {
* completing IO:
*/
BLK_MQ_F_STACKING = 1 << 2,
+ BLK_MQ_F_TAG_HCTX_SHARED = 1 << 3,
BLK_MQ_F_BLOCKING = 1 << 5,
BLK_MQ_F_NO_SCHED = 1 << 6,
BLK_MQ_F_ALLOC_POLICY_START_BIT = 8,
--
2.26.2
next prev parent reply other threads:[~2020-08-19 15:26 UTC|newest]
Thread overview: 58+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-08-19 15:20 [PATCH v8 00/18] blk-mq/scsi: Provide hostwide shared tags for SCSI HBAs John Garry
2020-08-19 15:20 ` [PATCH v8 01/18] blk-mq: Rename BLK_MQ_F_TAG_SHARED as BLK_MQ_F_TAG_QUEUE_SHARED John Garry
2020-08-19 15:20 ` [PATCH v8 02/18] blk-mq: Rename blk_mq_update_tag_set_depth() John Garry
2020-08-19 15:20 ` [PATCH v8 03/18] blk-mq: Free tags in blk_mq_init_tags() upon error John Garry
2020-08-19 15:20 ` [PATCH v8 04/18] blk-mq: Pass flags for tag init/free John Garry
2020-08-19 15:20 ` [PATCH v8 05/18] blk-mq: Use pointers for blk_mq_tags bitmap tags John Garry
2020-08-19 15:20 ` John Garry [this message]
2020-08-19 15:20 ` [PATCH v8 07/18] blk-mq: Relocate hctx_may_queue() John Garry
2020-08-19 15:20 ` [PATCH v8 08/18] blk-mq: Record nr_active_requests per queue for when using shared sbitmap John Garry
2020-08-19 15:20 ` [PATCH v8 09/18] blk-mq: Record active_queues_shared_sbitmap per tag_set " John Garry
2020-08-19 15:20 ` [PATCH v8 10/18] blk-mq, elevator: Count requests per hctx to improve performance John Garry
2020-08-19 15:20 ` [PATCH v8 11/18] null_blk: Support shared tag bitmap John Garry
2020-08-19 15:20 ` [PATCH v8 12/18] scsi: Add host and host template flag 'host_tagset' John Garry
2020-08-19 15:20 ` [PATCH v8 13/18] scsi: core: Show nr_hw_queues in sysfs John Garry
2020-09-10 8:33 ` John Garry
2020-08-19 15:20 ` [PATCH v8 14/18] scsi: hisi_sas: Switch v3 hw to MQ John Garry
2020-08-19 15:20 ` [PATCH v8 15/18] scsi: scsi_debug: Support host tagset John Garry
2020-08-19 15:20 ` [PATCH v8 16/18] hpsa: enable host_tagset and switch to MQ John Garry
2020-08-19 15:20 ` [PATCH v8 17/18] scsi: megaraid_sas: Added support for shared host tagset for cpuhotplug John Garry
2020-11-02 14:17 ` Qian Cai
2020-11-02 14:31 ` Kashyap Desai
2020-11-02 15:24 ` Qian Cai
2020-11-02 14:51 ` John Garry
2020-11-02 15:18 ` Qian Cai
2020-11-03 10:54 ` John Garry
2020-11-03 13:04 ` Qian Cai
2020-11-04 15:21 ` Qian Cai
2020-11-04 16:07 ` Kashyap Desai
2020-11-04 18:08 ` John Garry
2020-11-06 19:25 ` Sumit Saxena
2020-11-07 0:17 ` Qian Cai
2020-11-09 8:49 ` John Garry
2020-11-09 13:39 ` Qian Cai
2020-11-09 14:05 ` John Garry
2020-11-10 17:42 ` John Garry
2020-11-11 7:27 ` Sumit Saxena
2020-11-11 9:27 ` Ming Lei
2020-11-11 11:36 ` Sumit Saxena
2020-11-11 14:42 ` Qian Cai
2020-11-11 15:04 ` Ming Lei
2020-11-11 11:51 ` John Garry
2020-08-19 15:20 ` [PATCH v8 18/18] smartpqi: enable host tagset John Garry
2020-08-27 8:53 ` [PATCH v8 00/18] blk-mq/scsi: Provide hostwide shared tags for SCSI HBAs John Garry
2020-09-03 19:28 ` Douglas Gilbert
2020-09-03 21:23 ` Jens Axboe
2020-09-04 9:09 ` John Garry
2020-09-04 12:44 ` Martin K. Petersen
2020-09-16 7:21 ` John Garry
2020-09-17 1:10 ` Martin K. Petersen
2020-09-17 6:48 ` John Garry
2020-09-21 21:35 ` Don.Brace
2020-09-21 22:15 ` John Garry
2020-09-22 9:03 ` John Garry
2020-09-28 16:11 ` Kashyap Desai
2020-10-06 14:24 ` John Garry
2020-10-06 14:42 ` Jens Axboe
2020-09-08 12:46 ` Hannes Reinecke
2020-09-08 13:38 ` John Garry
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=1597850436-116171-7-git-send-email-john.garry@huawei.com \
--to=john.garry@huawei.com \
--cc=axboe@kernel.dk \
--cc=bvanassche@acm.org \
--cc=chenxiang66@hisilicon.com \
--cc=dgilbert@interlog.com \
--cc=don.brace@microsemi.com \
--cc=esc.storagedev@microsemi.com \
--cc=hare@suse.de \
--cc=hch@lst.de \
--cc=jejb@linux.ibm.com \
--cc=kashyap.desai@broadcom.com \
--cc=linux-block@vger.kernel.org \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-scsi@vger.kernel.org \
--cc=luojiaxing@huawei.com \
--cc=martin.petersen@oracle.com \
--cc=megaraidlinux.pdl@broadcom.com \
--cc=ming.lei@redhat.com \
--cc=paolo.valente@linaro.org \
--cc=sumit.saxena@broadcom.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox