Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 1ab0cd69 authored by Jens Axboe's avatar Jens Axboe Committed by Keith Busch
Browse files

nvme-pci: split the nvme queue lock into submission and completion locks



This is now feasible. We protect the submission queue ring with
->sq_lock, and the completion side with ->cq_lock.

Reviewed-by: default avatarChristoph Hellwig <hch@lst.de>
Signed-off-by: default avatarJens Axboe <axboe@kernel.dk>
Signed-off-by: default avatarChristoph Hellwig <hch@lst.de>
parent 5cb525c8
Loading
Loading
Loading
Loading
+23 −21
Original line number Original line Diff line number Diff line
@@ -147,9 +147,10 @@ static inline struct nvme_dev *to_nvme_dev(struct nvme_ctrl *ctrl)
struct nvme_queue {
struct nvme_queue {
	struct device *q_dmadev;
	struct device *q_dmadev;
	struct nvme_dev *dev;
	struct nvme_dev *dev;
	spinlock_t q_lock;
	spinlock_t sq_lock;
	struct nvme_command *sq_cmds;
	struct nvme_command *sq_cmds;
	struct nvme_command __iomem *sq_cmds_io;
	struct nvme_command __iomem *sq_cmds_io;
	spinlock_t cq_lock ____cacheline_aligned_in_smp;
	volatile struct nvme_completion *cqes;
	volatile struct nvme_completion *cqes;
	struct blk_mq_tags **tags;
	struct blk_mq_tags **tags;
	dma_addr_t sq_dma_addr;
	dma_addr_t sq_dma_addr;
@@ -894,9 +895,9 @@ static blk_status_t nvme_queue_rq(struct blk_mq_hw_ctx *hctx,


	blk_mq_start_request(req);
	blk_mq_start_request(req);


	spin_lock_irq(&nvmeq->q_lock);
	spin_lock_irq(&nvmeq->sq_lock);
	__nvme_submit_cmd(nvmeq, &cmnd);
	__nvme_submit_cmd(nvmeq, &cmnd);
	spin_unlock_irq(&nvmeq->q_lock);
	spin_unlock_irq(&nvmeq->sq_lock);
	return BLK_STS_OK;
	return BLK_STS_OK;
out_cleanup_iod:
out_cleanup_iod:
	nvme_free_iod(dev, req);
	nvme_free_iod(dev, req);
@@ -1000,9 +1001,9 @@ static irqreturn_t nvme_irq(int irq, void *data)
	struct nvme_queue *nvmeq = data;
	struct nvme_queue *nvmeq = data;
	u16 start, end;
	u16 start, end;


	spin_lock(&nvmeq->q_lock);
	spin_lock(&nvmeq->cq_lock);
	nvme_process_cq(nvmeq, &start, &end, -1);
	nvme_process_cq(nvmeq, &start, &end, -1);
	spin_unlock(&nvmeq->q_lock);
	spin_unlock(&nvmeq->cq_lock);


	if (start == end)
	if (start == end)
		return IRQ_NONE;
		return IRQ_NONE;
@@ -1026,9 +1027,9 @@ static int __nvme_poll(struct nvme_queue *nvmeq, unsigned int tag)
	if (!nvme_cqe_pending(nvmeq))
	if (!nvme_cqe_pending(nvmeq))
		return 0;
		return 0;


	spin_lock_irq(&nvmeq->q_lock);
	spin_lock_irq(&nvmeq->cq_lock);
	found = nvme_process_cq(nvmeq, &start, &end, tag);
	found = nvme_process_cq(nvmeq, &start, &end, tag);
	spin_unlock_irq(&nvmeq->q_lock);
	spin_unlock_irq(&nvmeq->cq_lock);


	nvme_complete_cqes(nvmeq, start, end);
	nvme_complete_cqes(nvmeq, start, end);
	return found;
	return found;
@@ -1051,9 +1052,9 @@ static void nvme_pci_submit_async_event(struct nvme_ctrl *ctrl)
	c.common.opcode = nvme_admin_async_event;
	c.common.opcode = nvme_admin_async_event;
	c.common.command_id = NVME_AQ_BLK_MQ_DEPTH;
	c.common.command_id = NVME_AQ_BLK_MQ_DEPTH;


	spin_lock_irq(&nvmeq->q_lock);
	spin_lock_irq(&nvmeq->sq_lock);
	__nvme_submit_cmd(nvmeq, &c);
	__nvme_submit_cmd(nvmeq, &c);
	spin_unlock_irq(&nvmeq->q_lock);
	spin_unlock_irq(&nvmeq->sq_lock);
}
}


static int adapter_delete_queue(struct nvme_dev *dev, u8 opcode, u16 id)
static int adapter_delete_queue(struct nvme_dev *dev, u8 opcode, u16 id)
@@ -1310,15 +1311,15 @@ static int nvme_suspend_queue(struct nvme_queue *nvmeq)
{
{
	int vector;
	int vector;


	spin_lock_irq(&nvmeq->q_lock);
	spin_lock_irq(&nvmeq->cq_lock);
	if (nvmeq->cq_vector == -1) {
	if (nvmeq->cq_vector == -1) {
		spin_unlock_irq(&nvmeq->q_lock);
		spin_unlock_irq(&nvmeq->cq_lock);
		return 1;
		return 1;
	}
	}
	vector = nvmeq->cq_vector;
	vector = nvmeq->cq_vector;
	nvmeq->dev->online_queues--;
	nvmeq->dev->online_queues--;
	nvmeq->cq_vector = -1;
	nvmeq->cq_vector = -1;
	spin_unlock_irq(&nvmeq->q_lock);
	spin_unlock_irq(&nvmeq->cq_lock);


	/*
	/*
	 * Ensure that nvme_queue_rq() sees it ->cq_vector == -1 without
	 * Ensure that nvme_queue_rq() sees it ->cq_vector == -1 without
@@ -1344,9 +1345,9 @@ static void nvme_disable_admin_queue(struct nvme_dev *dev, bool shutdown)
	else
	else
		nvme_disable_ctrl(&dev->ctrl, dev->ctrl.cap);
		nvme_disable_ctrl(&dev->ctrl, dev->ctrl.cap);


	spin_lock_irq(&nvmeq->q_lock);
	spin_lock_irq(&nvmeq->cq_lock);
	nvme_process_cq(nvmeq, &start, &end, -1);
	nvme_process_cq(nvmeq, &start, &end, -1);
	spin_unlock_irq(&nvmeq->q_lock);
	spin_unlock_irq(&nvmeq->cq_lock);


	nvme_complete_cqes(nvmeq, start, end);
	nvme_complete_cqes(nvmeq, start, end);
}
}
@@ -1406,7 +1407,8 @@ static int nvme_alloc_queue(struct nvme_dev *dev, int qid, int depth)


	nvmeq->q_dmadev = dev->dev;
	nvmeq->q_dmadev = dev->dev;
	nvmeq->dev = dev;
	nvmeq->dev = dev;
	spin_lock_init(&nvmeq->q_lock);
	spin_lock_init(&nvmeq->sq_lock);
	spin_lock_init(&nvmeq->cq_lock);
	nvmeq->cq_head = 0;
	nvmeq->cq_head = 0;
	nvmeq->cq_phase = 1;
	nvmeq->cq_phase = 1;
	nvmeq->q_db = &dev->dbs[qid * 2 * dev->db_stride];
	nvmeq->q_db = &dev->dbs[qid * 2 * dev->db_stride];
@@ -1442,7 +1444,7 @@ static void nvme_init_queue(struct nvme_queue *nvmeq, u16 qid)
{
{
	struct nvme_dev *dev = nvmeq->dev;
	struct nvme_dev *dev = nvmeq->dev;


	spin_lock_irq(&nvmeq->q_lock);
	spin_lock_irq(&nvmeq->cq_lock);
	nvmeq->sq_tail = 0;
	nvmeq->sq_tail = 0;
	nvmeq->cq_head = 0;
	nvmeq->cq_head = 0;
	nvmeq->cq_phase = 1;
	nvmeq->cq_phase = 1;
@@ -1450,7 +1452,7 @@ static void nvme_init_queue(struct nvme_queue *nvmeq, u16 qid)
	memset((void *)nvmeq->cqes, 0, CQ_SIZE(nvmeq->q_depth));
	memset((void *)nvmeq->cqes, 0, CQ_SIZE(nvmeq->q_depth));
	nvme_dbbuf_init(dev, nvmeq, qid);
	nvme_dbbuf_init(dev, nvmeq, qid);
	dev->online_queues++;
	dev->online_queues++;
	spin_unlock_irq(&nvmeq->q_lock);
	spin_unlock_irq(&nvmeq->cq_lock);
}
}


static int nvme_create_queue(struct nvme_queue *nvmeq, int qid)
static int nvme_create_queue(struct nvme_queue *nvmeq, int qid)
@@ -2001,14 +2003,14 @@ static void nvme_del_cq_end(struct request *req, blk_status_t error)
		unsigned long flags;
		unsigned long flags;


		/*
		/*
		 * We might be called with the AQ q_lock held
		 * We might be called with the AQ cq_lock held
		 * and the I/O queue q_lock should always
		 * and the I/O queue cq_lock should always
		 * nest inside the AQ one.
		 * nest inside the AQ one.
		 */
		 */
		spin_lock_irqsave_nested(&nvmeq->q_lock, flags,
		spin_lock_irqsave_nested(&nvmeq->cq_lock, flags,
					SINGLE_DEPTH_NESTING);
					SINGLE_DEPTH_NESTING);
		nvme_process_cq(nvmeq, &start, &end, -1);
		nvme_process_cq(nvmeq, &start, &end, -1);
		spin_unlock_irqrestore(&nvmeq->q_lock, flags);
		spin_unlock_irqrestore(&nvmeq->cq_lock, flags);


		nvme_complete_cqes(nvmeq, start, end);
		nvme_complete_cqes(nvmeq, start, end);
	}
	}