Linux-NVME Archive on lore.kernel.org
 help / color / Atom feed
From: Ming Lei <ming.lei@redhat.com>
To: linux-nvme@lists.infradead.org
Cc: Sagi Grimberg <sagi@grimberg.me>, Long Li <longli@microsoft.com>,
	Ming Lei <ming.lei@redhat.com>, Jens Axboe <axboe@fb.com>,
	Nadolski Edmund <edmund.nadolski@intel.com>,
	Keith Busch <kbusch@kernel.org>, Christoph Hellwig <hch@lst.de>
Subject: [PATCH V3 2/2] nvme-pci: check CQ after batch submission for Microsoft device
Date: Thu, 14 Nov 2019 10:59:17 +0800
Message-ID: <20191114025917.24634-3-ming.lei@redhat.com> (raw)
In-Reply-To: <20191114025917.24634-1-ming.lei@redhat.com>

f9dde187fa92("nvme-pci: remove cq check after submission") removes
cq check after submission, this change actually causes performance
regression on Microsoft Corporation device in which aggressive interrupt
coalescing is used, for example, the max single interrupt per second
can reach 40~50K at most.

Checking CQ after submission can improve IO completion for Microsoft
Corporation device:

1) the check is very cheap, and lockless check on cq is enough,
see nvme_cqe_pending(). Especially the check can be done after batch
submission is done. Compared with the big latency added by aggressive
interrupt coalescing, cost of the check can be ignored.

2) when IO completion is observed via checking CQ in submission, the requst
may be completed without interrupt involved, or the interrupt handler
overload can be decreased.

3) when single sw queue is submiting IOs to this hw queue, if IOs completion
is observed via this check, the IO can be completed locally, which is
cheaper than remote completion.

Follows test result done on Azure L80sv2 guest with NVMe drive(
Microsoft Corporation Device b111). This guest has 80 CPUs and 10
numa nodes, and each NVMe drive supports 8 hw queues.

1) test script:
fio --bs=4k --ioengine=libaio --iodepth=64 --filename=/dev/nvme0n1 \
	--iodepth_batch_submit=16 --iodepth_batch_complete_min=16 \
	--direct=1 --runtime=30 --numjobs=1 --rw=randread \
	--name=test --group_reporting --gtod_reduce=1

2) test result:
     | v5.3       | v5.3 with this patchset
-------------------------------------------
IOPS | 130K       | 424K

Given IO is handled more efficiently in this way, the original report
of CPU lockup[1] on Hyper-V can't be observed any more after this patch
is applied, since this patch reduces IO completions much in interrupt
handler.

[1] https://lore.kernel.org/lkml/1566281669-48212-1-git-send-email-longli@linuxonhyperv.com

Cc: Keith Busch <kbusch@kernel.org>
Cc: Jens Axboe <axboe@fb.com>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Sagi Grimberg <sagi@grimberg.me>
Cc: Long Li <longli@microsoft.com>
Fixes: f9dde187fa92("nvme-pci: remove cq check after submission")
Signed-off-by: Ming Lei <ming.lei@redhat.com>
---
 drivers/nvme/host/nvme.h |  6 +++
 drivers/nvme/host/pci.c  | 95 ++++++++++++++++++++++++++++++++++------
 2 files changed, 88 insertions(+), 13 deletions(-)

diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
index 22e8401352c2..5575827016d1 100644
--- a/drivers/nvme/host/nvme.h
+++ b/drivers/nvme/host/nvme.h
@@ -115,6 +115,12 @@ enum nvme_quirks {
 	 * Prevent tag overlap between queues
 	 */
 	NVME_QUIRK_SHARED_TAGS                  = (1 << 13),
+
+	/*
+	 * Check CQ in submission for improving performance since
+	 * the device applies aggressive interrupt coalescing
+	 */
+	NVME_QUIRK_CHECK_CQ			= (1 << 14),
 };
 
 /*
diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index 01728409a7ea..15fd8574ee68 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -83,6 +83,7 @@ struct nvme_queue;
 
 static void nvme_dev_disable(struct nvme_dev *dev, bool shutdown);
 static bool __nvme_disable_io_queues(struct nvme_dev *dev, u8 opcode);
+static void nvme_check_cq(struct nvme_queue *nvmeq);
 
 /*
  * Represents an NVM Express device.  Each nvme_dev is a PCI function.
@@ -164,8 +165,11 @@ struct nvme_queue {
 	struct nvme_dev *dev;
 	spinlock_t sq_lock;
 	void *sq_cmds;
-	 /* only used for poll queues: */
-	spinlock_t cq_poll_lock ____cacheline_aligned_in_smp;
+
+	union {
+		spinlock_t cq_poll_lock; /* only used for poll queues */
+		spinlock_t cq_lock;	 /* only used for irq queues */
+	}____cacheline_aligned_in_smp;
 	volatile struct nvme_completion *cqes;
 	struct blk_mq_tags **tags;
 	dma_addr_t sq_dma_addr;
@@ -185,6 +189,7 @@ struct nvme_queue {
 #define NVMEQ_SQ_CMB		1
 #define NVMEQ_DELETE_ERROR	2
 #define NVMEQ_POLLED		3
+#define NVMEQ_CHECK_CQ		4
 	u32 *dbbuf_sq_db;
 	u32 *dbbuf_cq_db;
 	u32 *dbbuf_sq_ei;
@@ -912,6 +917,10 @@ static blk_status_t nvme_queue_rq(struct blk_mq_hw_ctx *hctx,
 
 	blk_mq_start_request(req);
 	nvme_submit_cmd(nvmeq, &cmnd, bd->last);
+
+	if (bd->last && test_bit(NVMEQ_CHECK_CQ, &nvmeq->flags))
+		nvme_check_cq(nvmeq);
+
 	return BLK_STS_OK;
 out_unmap_data:
 	nvme_unmap_data(dev, req);
@@ -1017,6 +1026,19 @@ static inline int nvme_process_cq(struct nvme_queue *nvmeq, u16 *start,
 	return found;
 }
 
+static inline irqreturn_t
+nvme_update_cq(struct nvme_queue *nvmeq, u16 *start, u16 *end)
+{
+	irqreturn_t ret = IRQ_NONE;
+
+	if (nvmeq->cq_head != nvmeq->last_cq_head)
+		ret = IRQ_HANDLED;
+	nvme_process_cq(nvmeq, start, end, -1);
+	nvmeq->last_cq_head = nvmeq->cq_head;
+
+	return ret;
+}
+
 static irqreturn_t nvme_irq(int irq, void *data)
 {
 	struct nvme_queue *nvmeq = data;
@@ -1028,10 +1050,7 @@ static irqreturn_t nvme_irq(int irq, void *data)
 	 * the irq handler, even if that was on another CPU.
 	 */
 	rmb();
-	if (nvmeq->cq_head != nvmeq->last_cq_head)
-		ret = IRQ_HANDLED;
-	nvme_process_cq(nvmeq, &start, &end, -1);
-	nvmeq->last_cq_head = nvmeq->cq_head;
+	ret = nvme_update_cq(nvmeq, &start, &end);
 	wmb();
 
 	if (start != end) {
@@ -1042,6 +1061,24 @@ static irqreturn_t nvme_irq(int irq, void *data)
 	return ret;
 }
 
+static irqreturn_t nvme_irq_for_check_cq(int irq, void *data)
+{
+	struct nvme_queue *nvmeq = data;
+	irqreturn_t ret = IRQ_NONE;
+	u16 start, end;
+
+	spin_lock(&nvmeq->cq_lock);
+	ret = nvme_update_cq(nvmeq, &start, &end);
+	spin_unlock(&nvmeq->cq_lock);
+
+	if (start != end) {
+		nvme_complete_cqes(nvmeq, start, end);
+		return IRQ_HANDLED;
+	}
+
+	return ret;
+}
+
 static irqreturn_t nvme_irq_check(int irq, void *data)
 {
 	struct nvme_queue *nvmeq = data;
@@ -1050,6 +1087,22 @@ static irqreturn_t nvme_irq_check(int irq, void *data)
 	return IRQ_NONE;
 }
 
+static void nvme_check_cq(struct nvme_queue *nvmeq)
+{
+	if (!nvme_cqe_pending(nvmeq))
+		return;
+
+	if (spin_trylock_irq(&nvmeq->cq_lock)) {
+		u16 start, end;
+
+		nvme_update_cq(nvmeq, &start, &end);
+		spin_unlock_irq(&nvmeq->cq_lock);
+
+		if (start != end)
+			nvme_complete_cqes(nvmeq, start, end);
+	}
+}
+
 /*
  * Poll for completions any queue, including those not dedicated to polling.
  * Can be called from any context.
@@ -1500,12 +1553,14 @@ static int queue_request_irq(struct nvme_queue *nvmeq)
 {
 	struct pci_dev *pdev = to_pci_dev(nvmeq->dev->dev);
 	int nr = nvmeq->dev->ctrl.instance;
+	irq_handler_t handler = test_bit(NVMEQ_CHECK_CQ, &nvmeq->flags) ?
+		nvme_irq_for_check_cq : nvme_irq;
 
 	if (use_threaded_interrupts) {
 		return pci_request_irq(pdev, nvmeq->cq_vector, nvme_irq_check,
-				nvme_irq, nvmeq, "nvme%dq%d", nr, nvmeq->qid);
+				handler, nvmeq, "nvme%dq%d", nr, nvmeq->qid);
 	} else {
-		return pci_request_irq(pdev, nvmeq->cq_vector, nvme_irq,
+		return pci_request_irq(pdev, nvmeq->cq_vector, handler,
 				NULL, nvmeq, "nvme%dq%d", nr, nvmeq->qid);
 	}
 }
@@ -1515,7 +1570,13 @@ static void nvme_init_queue(struct nvme_queue *nvmeq, u16 qid)
 	struct nvme_dev *dev = nvmeq->dev;
 
 	spin_lock_init(&nvmeq->sq_lock);
-	spin_lock_init(&nvmeq->cq_poll_lock);
+
+	if (test_bit(NVMEQ_CHECK_CQ, &nvmeq->flags)) {
+		WARN_ON_ONCE(test_bit(NVMEQ_POLLED, &nvmeq->flags));
+		spin_lock_init(&nvmeq->cq_lock);
+	} else {
+		spin_lock_init(&nvmeq->cq_poll_lock);
+	}
 
 	nvmeq->sq_tail = 0;
 	nvmeq->last_sq_tail = 0;
@@ -1528,22 +1589,27 @@ static void nvme_init_queue(struct nvme_queue *nvmeq, u16 qid)
 	wmb(); /* ensure the first interrupt sees the initialization */
 }
 
-static int nvme_create_queue(struct nvme_queue *nvmeq, int qid, bool polled)
+static int nvme_create_queue(struct nvme_queue *nvmeq, int qid, bool polled,
+		bool check_cq)
 {
 	struct nvme_dev *dev = nvmeq->dev;
 	int result;
 	u16 vector = 0;
 
 	clear_bit(NVMEQ_DELETE_ERROR, &nvmeq->flags);
+	clear_bit(NVMEQ_CHECK_CQ, &nvmeq->flags);
 
 	/*
 	 * A queue's vector matches the queue identifier unless the controller
 	 * has only one vector available.
 	 */
-	if (!polled)
+	if (!polled) {
 		vector = dev->num_vecs == 1 ? 0 : qid;
-	else
+		if (vector && check_cq)
+			set_bit(NVMEQ_CHECK_CQ, &nvmeq->flags);
+	} else {
 		set_bit(NVMEQ_POLLED, &nvmeq->flags);
+	}
 
 	result = adapter_alloc_cq(dev, qid, nvmeq, vector);
 	if (result)
@@ -1740,7 +1806,8 @@ static int nvme_create_io_queues(struct nvme_dev *dev)
 	for (i = dev->online_queues; i <= max; i++) {
 		bool polled = i > rw_queues;
 
-		ret = nvme_create_queue(&dev->queues[i], i, polled);
+		ret = nvme_create_queue(&dev->queues[i], i, polled,
+				dev->ctrl.quirks & NVME_QUIRK_CHECK_CQ);
 		if (ret)
 			break;
 	}
@@ -3112,6 +3179,8 @@ static const struct pci_device_id nvme_id_table[] = {
 	{ PCI_DEVICE(0x1cc1, 0x8201),   /* ADATA SX8200PNP 512GB */
 		.driver_data = NVME_QUIRK_NO_DEEPEST_PS |
 				NVME_QUIRK_IGNORE_DEV_SUBNQN, },
+	{ PCI_DEVICE(0x1414, 0xb111),   /* Microsoft Corporation device */
+		.driver_data = NVME_QUIRK_CHECK_CQ, },
 	{ PCI_DEVICE_CLASS(PCI_CLASS_STORAGE_EXPRESS, 0xffffff) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_APPLE, 0x2001) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_APPLE, 0x2003) },
-- 
2.20.1


_______________________________________________
Linux-nvme mailing list
Linux-nvme@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-nvme

  parent reply index

Thread overview: 16+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-11-14  2:59 [PATCH V3 0/2] " Ming Lei
2019-11-14  2:59 ` [PATCH V3 1/2] nvme-pci: move sq/cq_poll lock initialization into nvme_init_queue Ming Lei
2019-11-14  2:59 ` Ming Lei [this message]
2019-11-14  4:56   ` [PATCH V3 2/2] nvme-pci: check CQ after batch submission for Microsoft device Keith Busch
2019-11-14  8:56     ` Ming Lei
2019-11-21  3:11 ` [PATCH V3 0/2] " Ming Lei
2019-11-21  6:14   ` Christoph Hellwig
2019-11-21  7:46     ` Ming Lei
2019-11-21 15:45       ` Keith Busch
2019-11-22  9:44         ` Ming Lei
2019-11-22  9:57           ` Christoph Hellwig
2019-11-22 10:25             ` Ming Lei
2019-11-22 14:04               ` Jens Axboe
2019-11-22 21:49                 ` Ming Lei
2019-11-22 21:58                   ` Jens Axboe
2019-11-22 22:30                     ` Ming Lei

Reply instructions:

You may reply publically 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=20191114025917.24634-3-ming.lei@redhat.com \
    --to=ming.lei@redhat.com \
    --cc=axboe@fb.com \
    --cc=edmund.nadolski@intel.com \
    --cc=hch@lst.de \
    --cc=kbusch@kernel.org \
    --cc=linux-nvme@lists.infradead.org \
    --cc=longli@microsoft.com \
    --cc=sagi@grimberg.me \
    /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

Linux-NVME Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-nvme/0 linux-nvme/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-nvme linux-nvme/ https://lore.kernel.org/linux-nvme \
		linux-nvme@lists.infradead.org
	public-inbox-index linux-nvme

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.infradead.lists.linux-nvme


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git