1 files changed, 98 insertions, 41 deletions

diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
index 7990c3f22ecf..5f2e3ad2cc52 100644
--- a/drivers/nvme/host/pci.c
+++ b/drivers/nvme/host/pci.c
@@ -141,6 +141,7 @@ struct nvme_dev {
 	struct nvme_ctrl ctrl;
 	u32 last_ps;
 	bool hmb;
+	struct sg_table *hmb_sgt;
 
 	mempool_t *iod_mempool;
 
@@ -153,6 +154,7 @@ struct nvme_dev {
 	/* host memory buffer support: */
 	u64 host_mem_size;
 	u32 nr_host_mem_descs;
+	u32 host_mem_descs_size;
 	dma_addr_t host_mem_descs_dma;
 	struct nvme_host_mem_buf_desc *host_mem_descs;
 	void **host_mem_desc_bufs;
@@ -902,11 +904,12 @@ static blk_status_t nvme_queue_rq(struct blk_mq_hw_ctx *hctx,
 	return BLK_STS_OK;
 }
 
-static void nvme_submit_cmds(struct nvme_queue *nvmeq, struct request **rqlist)
+static void nvme_submit_cmds(struct nvme_queue *nvmeq, struct rq_list *rqlist)
 {
+	struct request *req;
+
 	spin_lock(&nvmeq->sq_lock);
-	while (!rq_list_empty(*rqlist)) {
-		struct request *req = rq_list_pop(rqlist);
+	while ((req = rq_list_pop(rqlist))) {
 		struct nvme_iod *iod = blk_mq_rq_to_pdu(req);
 
 		nvme_sq_copy_cmd(nvmeq, &iod->cmd);
@@ -929,33 +932,26 @@ static bool nvme_prep_rq_batch(struct nvme_queue *nvmeq, struct request *req)
 	return nvme_prep_rq(nvmeq->dev, req) == BLK_STS_OK;
 }
 
-static void nvme_queue_rqs(struct request **rqlist)
+static void nvme_queue_rqs(struct rq_list *rqlist)
 {
-	struct request *req, *next, *prev = NULL;
-	struct request *requeue_list = NULL;
-
-	rq_list_for_each_safe(rqlist, req, next) {
-		struct nvme_queue *nvmeq = req->mq_hctx->driver_data;
+	struct rq_list submit_list = { };
+	struct rq_list requeue_list = { };
+	struct nvme_queue *nvmeq = NULL;
+	struct request *req;
 
-		if (!nvme_prep_rq_batch(nvmeq, req)) {
-			/* detach 'req' and add to remainder list */
-			rq_list_move(rqlist, &requeue_list, req, prev);
+	while ((req = rq_list_pop(rqlist))) {
+		if (nvmeq && nvmeq != req->mq_hctx->driver_data)
+			nvme_submit_cmds(nvmeq, &submit_list);
+		nvmeq = req->mq_hctx->driver_data;
 
-			req = prev;
-			if (!req)
-				continue;
-		}
-
-		if (!next || req->mq_hctx != next->mq_hctx) {
-			/* detach rest of list, and submit */
-			req->rq_next = NULL;
-			nvme_submit_cmds(nvmeq, rqlist);
-			*rqlist = next;
-			prev = NULL;
-		} else
-			prev = req;
+		if (nvme_prep_rq_batch(nvmeq, req))
+			rq_list_add_tail(&submit_list, req);
+		else
+			rq_list_add_tail(&requeue_list, req);
 	}
 
+	if (nvmeq)
+		nvme_submit_cmds(nvmeq, &submit_list);
 	*rqlist = requeue_list;
 }
 
@@ -1083,7 +1079,7 @@ static irqreturn_t nvme_irq(int irq, void *data)
 	DEFINE_IO_COMP_BATCH(iob);
 
 	if (nvme_poll_cq(nvmeq, &iob)) {
-		if (!rq_list_empty(iob.req_list))
+		if (!rq_list_empty(&iob.req_list))
 			nvme_pci_complete_batch(&iob);
 		return IRQ_HANDLED;
 	}
@@ -1951,7 +1947,7 @@ static int nvme_set_host_mem(struct nvme_dev *dev, u32 bits)
 	return ret;
 }
 
-static void nvme_free_host_mem(struct nvme_dev *dev)
+static void nvme_free_host_mem_multi(struct nvme_dev *dev)
 {
 	int i;
 
@@ -1966,18 +1962,54 @@ static void nvme_free_host_mem(struct nvme_dev *dev)
 
 	kfree(dev->host_mem_desc_bufs);
 	dev->host_mem_desc_bufs = NULL;
-	dma_free_coherent(dev->dev,
-			dev->nr_host_mem_descs * sizeof(*dev->host_mem_descs),
+}
+
+static void nvme_free_host_mem(struct nvme_dev *dev)
+{
+	if (dev->hmb_sgt)
+		dma_free_noncontiguous(dev->dev, dev->host_mem_size,
+				dev->hmb_sgt, DMA_BIDIRECTIONAL);
+	else
+		nvme_free_host_mem_multi(dev);
+
+	dma_free_coherent(dev->dev, dev->host_mem_descs_size,
 			dev->host_mem_descs, dev->host_mem_descs_dma);
 	dev->host_mem_descs = NULL;
+	dev->host_mem_descs_size = 0;
 	dev->nr_host_mem_descs = 0;
 }
 
-static int __nvme_alloc_host_mem(struct nvme_dev *dev, u64 preferred,
+static int nvme_alloc_host_mem_single(struct nvme_dev *dev, u64 size)
+{
+	dev->hmb_sgt = dma_alloc_noncontiguous(dev->dev, size,
+				DMA_BIDIRECTIONAL, GFP_KERNEL, 0);
+	if (!dev->hmb_sgt)
+		return -ENOMEM;
+
+	dev->host_mem_descs = dma_alloc_coherent(dev->dev,
+			sizeof(*dev->host_mem_descs), &dev->host_mem_descs_dma,
+			GFP_KERNEL);
+	if (!dev->host_mem_descs) {
+		dma_free_noncontiguous(dev->dev, dev->host_mem_size,
+				dev->hmb_sgt, DMA_BIDIRECTIONAL);
+		dev->hmb_sgt = NULL;
+		return -ENOMEM;
+	}
+	dev->host_mem_size = size;
+	dev->host_mem_descs_size = sizeof(*dev->host_mem_descs);
+	dev->nr_host_mem_descs = 1;
+
+	dev->host_mem_descs[0].addr =
+		cpu_to_le64(dev->hmb_sgt->sgl->dma_address);
+	dev->host_mem_descs[0].size = cpu_to_le32(size / NVME_CTRL_PAGE_SIZE);
+	return 0;
+}
+
+static int nvme_alloc_host_mem_multi(struct nvme_dev *dev, u64 preferred,
 		u32 chunk_size)
 {
 	struct nvme_host_mem_buf_desc *descs;
-	u32 max_entries, len;
+	u32 max_entries, len, descs_size;
 	dma_addr_t descs_dma;
 	int i = 0;
 	void **bufs;
@@ -1990,8 +2022,9 @@ static int __nvme_alloc_host_mem(struct nvme_dev *dev, u64 preferred,
 	if (dev->ctrl.hmmaxd && dev->ctrl.hmmaxd < max_entries)
 		max_entries = dev->ctrl.hmmaxd;
 
-	descs = dma_alloc_coherent(dev->dev, max_entries * sizeof(*descs),
-				   &descs_dma, GFP_KERNEL);
+	descs_size = max_entries * sizeof(*descs);
+	descs = dma_alloc_coherent(dev->dev, descs_size, &descs_dma,
+			GFP_KERNEL);
 	if (!descs)
 		goto out;
 
@@ -2020,6 +2053,7 @@ static int __nvme_alloc_host_mem(struct nvme_dev *dev, u64 preferred,
 	dev->host_mem_size = size;
 	dev->host_mem_descs = descs;
 	dev->host_mem_descs_dma = descs_dma;
+	dev->host_mem_descs_size = descs_size;
 	dev->host_mem_desc_bufs = bufs;
 	return 0;
 
@@ -2034,8 +2068,7 @@ out_free_bufs:
 
 	kfree(bufs);
 out_free_descs:
-	dma_free_coherent(dev->dev, max_entries * sizeof(*descs), descs,
-			descs_dma);
+	dma_free_coherent(dev->dev, descs_size, descs, descs_dma);
 out:
 	dev->host_mem_descs = NULL;
 	return -ENOMEM;
@@ -2047,9 +2080,18 @@ static int nvme_alloc_host_mem(struct nvme_dev *dev, u64 min, u64 preferred)
 	u64 hmminds = max_t(u32, dev->ctrl.hmminds * 4096, PAGE_SIZE * 2);
 	u64 chunk_size;
 
+	/*
+	 * If there is an IOMMU that can merge pages, try a virtually
+	 * non-contiguous allocation for a single segment first.
+	 */
+	if (!(PAGE_SIZE & dma_get_merge_boundary(dev->dev))) {
+		if (!nvme_alloc_host_mem_single(dev, preferred))
+			return 0;
+	}
+
 	/* start big and work our way down */
 	for (chunk_size = min_chunk; chunk_size >= hmminds; chunk_size /= 2) {
-		if (!__nvme_alloc_host_mem(dev, preferred, chunk_size)) {
+		if (!nvme_alloc_host_mem_multi(dev, preferred, chunk_size)) {
 			if (!min || dev->host_mem_size >= min)
 				return 0;
 			nvme_free_host_mem(dev);
@@ -2097,8 +2139,10 @@ static int nvme_setup_host_mem(struct nvme_dev *dev)
 		}
 
 		dev_info(dev->ctrl.device,
-			"allocated %lld MiB host memory buffer.\n",
-			dev->host_mem_size >> ilog2(SZ_1M));
+			"allocated %lld MiB host memory buffer (%u segment%s).\n",
+			dev->host_mem_size >> ilog2(SZ_1M),
+			dev->nr_host_mem_descs,
+			str_plural(dev->nr_host_mem_descs));
 	}
 
 	ret = nvme_set_host_mem(dev, enable_bits);
@@ -2506,17 +2550,29 @@ static unsigned int nvme_pci_nr_maps(struct nvme_dev *dev)
 	return 1;
 }
 
-static void nvme_pci_update_nr_queues(struct nvme_dev *dev)
+static bool nvme_pci_update_nr_queues(struct nvme_dev *dev)
 {
 	if (!dev->ctrl.tagset) {
 		nvme_alloc_io_tag_set(&dev->ctrl, &dev->tagset, &nvme_mq_ops,
 				nvme_pci_nr_maps(dev), sizeof(struct nvme_iod));
-		return;
+		return true;
+	}
+
+	/* Give up if we are racing with nvme_dev_disable() */
+	if (!mutex_trylock(&dev->shutdown_lock))
+		return false;
+
+	/* Check if nvme_dev_disable() has been executed already */
+	if (!dev->online_queues) {
+		mutex_unlock(&dev->shutdown_lock);
+		return false;
 	}
 
 	blk_mq_update_nr_hw_queues(&dev->tagset, dev->online_queues - 1);
 	/* free previously allocated queues that are no longer usable */
 	nvme_free_queues(dev, dev->online_queues);
+	mutex_unlock(&dev->shutdown_lock);
+	return true;
 }
 
 static int nvme_pci_enable(struct nvme_dev *dev)
@@ -2797,7 +2853,8 @@ static void nvme_reset_work(struct work_struct *work)
 		nvme_dbbuf_set(dev);
 		nvme_unquiesce_io_queues(&dev->ctrl);
 		nvme_wait_freeze(&dev->ctrl);
-		nvme_pci_update_nr_queues(dev);
+		if (!nvme_pci_update_nr_queues(dev))
+			goto out;
 		nvme_unfreeze(&dev->ctrl);
 	} else {
 		dev_warn(dev->ctrl.device, "IO queues lost\n");