Merge git://git.infradead.org/users/willy/linux-nvme

Pull NVMe driver update from Matthew Wilcox:
 "Lots of exciting new features in the NVM Express driver this time,
  including support for emulating SCSI commands, discard support and the
  ability to submit per-sector metadata with I/Os.

  It's still mostly bugfixes though!"

* git://git.infradead.org/users/willy/linux-nvme: (27 commits)
  NVMe: Use user defined admin ioctl timeout
  NVMe: Simplify Firmware Activate code slightly
  NVMe: Only clear the enable bit when disabling controller
  NVMe: Wait for device to acknowledge shutdown
  NVMe: Schedule timeout for sync commands
  NVMe: Meta-data support in NVME_IOCTL_SUBMIT_IO
  NVMe: Device specific stripe size handling
  NVMe: Split non-mergeable bio requests
  NVMe: Remove dead code in nvme_dev_add
  NVMe: Check for NULL memory in nvme_dev_add
  NVMe: Fix error clean-up on nvme_alloc_queue
  NVMe: Free admin queue on request_irq error
  NVMe: Add scsi unmap to SG_IO
  NVMe: queue usage fixes in nvme-scsi
  NVMe: Set TASK_INTERRUPTIBLE before processing queues
  NVMe: Add a character device for each nvme device
  NVMe: Fix endian-related problems in user I/O submission path
  NVMe: Fix I/O cancellation status on big-endian machines
  NVMe: Fix sparse warnings in scsi emulation
  NVMe: Don't fail initialisation unnecessarily
  ...

drivers/block/Makefile

@@ -42,4 +42,5 @@ obj-$(CONFIG_BLK_DEV_PCIESSD_MTIP32XX)	+= mtip32xx/
 
 obj-$(CONFIG_BLK_DEV_RSXX) += rsxx/
 
+nvme-y		:= nvme-core.o nvme-scsi.o
 swim_mod-y	:= swim.o swim_asm.o

drivers/block/nvme-core.c

@@ -39,14 +39,13 @@
 #include <linux/sched.h>
 #include <linux/slab.h>
 #include <linux/types.h>
-
+#include <scsi/sg.h>
 #include <asm-generic/io-64-nonatomic-lo-hi.h>
 
 #define NVME_Q_DEPTH 1024
 #define SQ_SIZE(depth)		(depth * sizeof(struct nvme_command))
 #define CQ_SIZE(depth)		(depth * sizeof(struct nvme_completion))
 #define NVME_MINORS 64
-#define NVME_IO_TIMEOUT	(5 * HZ)
 #define ADMIN_TIMEOUT	(60 * HZ)
 
 static int nvme_major;
@@ -59,43 +58,6 @@ static DEFINE_SPINLOCK(dev_list_lock);
 static LIST_HEAD(dev_list);
 static struct task_struct *nvme_thread;
 
-/*
- * Represents an NVM Express device.  Each nvme_dev is a PCI function.
- */
-struct nvme_dev {
-	struct list_head node;
-	struct nvme_queue **queues;
-	u32 __iomem *dbs;
-	struct pci_dev *pci_dev;
-	struct dma_pool *prp_page_pool;
-	struct dma_pool *prp_small_pool;
-	int instance;
-	int queue_count;
-	int db_stride;
-	u32 ctrl_config;
-	struct msix_entry *entry;
-	struct nvme_bar __iomem *bar;
-	struct list_head namespaces;
-	char serial[20];
-	char model[40];
-	char firmware_rev[8];
-	u32 max_hw_sectors;
-};
-
-/*
- * An NVM Express namespace is equivalent to a SCSI LUN
- */
-struct nvme_ns {
-	struct list_head list;
-
-	struct nvme_dev *dev;
-	struct request_queue *queue;
-	struct gendisk *disk;
-
-	int ns_id;
-	int lba_shift;
-};
-
 /*
  * An NVM Express queue.  Each device has at least two (one for admin
  * commands and one for I/O commands).
@@ -131,6 +93,7 @@ static inline void _nvme_check_size(void)
 	BUILD_BUG_ON(sizeof(struct nvme_create_sq) != 64);
 	BUILD_BUG_ON(sizeof(struct nvme_delete_queue) != 64);
 	BUILD_BUG_ON(sizeof(struct nvme_features) != 64);
+	BUILD_BUG_ON(sizeof(struct nvme_format_cmd) != 64);
 	BUILD_BUG_ON(sizeof(struct nvme_command) != 64);
 	BUILD_BUG_ON(sizeof(struct nvme_id_ctrl) != 4096);
 	BUILD_BUG_ON(sizeof(struct nvme_id_ns) != 4096);
@@ -261,12 +224,12 @@ static void *cancel_cmdid(struct nvme_queue *nvmeq, int cmdid,
 	return ctx;
 }
 
-static struct nvme_queue *get_nvmeq(struct nvme_dev *dev)
+struct nvme_queue *get_nvmeq(struct nvme_dev *dev)
 {
 	return dev->queues[get_cpu() + 1];
 }
 
-static void put_nvmeq(struct nvme_queue *nvmeq)
+void put_nvmeq(struct nvme_queue *nvmeq)
 {
 	put_cpu();
 }
@@ -294,22 +257,6 @@ static int nvme_submit_cmd(struct nvme_queue *nvmeq, struct nvme_command *cmd)
 	return 0;
 }
 
-/*
- * The nvme_iod describes the data in an I/O, including the list of PRP
- * entries.  You can't see it in this data structure because C doesn't let
- * me express that.  Use nvme_alloc_iod to ensure there's enough space
- * allocated to store the PRP list.
- */
-struct nvme_iod {
-	void *private;		/* For the use of the submitter of the I/O */
-	int npages;		/* In the PRP list. 0 means small pool in use */
-	int offset;		/* Of PRP list */
-	int nents;		/* Used in scatterlist */
-	int length;		/* Of data, in bytes */
-	dma_addr_t first_dma;
-	struct scatterlist sg[0];
-};
-
 static __le64 **iod_list(struct nvme_iod *iod)
 {
 	return ((void *)iod) + iod->offset;
@@ -343,7 +290,7 @@ nvme_alloc_iod(unsigned nseg, unsigned nbytes, gfp_t gfp)
 	return iod;
 }
 
-static void nvme_free_iod(struct nvme_dev *dev, struct nvme_iod *iod)
+void nvme_free_iod(struct nvme_dev *dev, struct nvme_iod *iod)
 {
 	const int last_prp = PAGE_SIZE / 8 - 1;
 	int i;
@@ -361,16 +308,6 @@ static void nvme_free_iod(struct nvme_dev *dev, struct nvme_iod *iod)
 	kfree(iod);
 }
 
-static void requeue_bio(struct nvme_dev *dev, struct bio *bio)
-{
-	struct nvme_queue *nvmeq = get_nvmeq(dev);
-	if (bio_list_empty(&nvmeq->sq_cong))
-		add_wait_queue(&nvmeq->sq_full, &nvmeq->sq_cong_wait);
-	bio_list_add(&nvmeq->sq_cong, bio);
-	put_nvmeq(nvmeq);
-	wake_up_process(nvme_thread);
-}
-
 static void bio_completion(struct nvme_dev *dev, void *ctx,
 						struct nvme_completion *cqe)
 {
@@ -382,19 +319,15 @@ static void bio_completion(struct nvme_dev *dev, void *ctx,
 		dma_unmap_sg(&dev->pci_dev->dev, iod->sg, iod->nents,
 			bio_data_dir(bio) ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
 	nvme_free_iod(dev, iod);
-	if (status) {
+	if (status)
 		bio_endio(bio, -EIO);
-	} else if (bio->bi_vcnt > bio->bi_idx) {
+	else
-		requeue_bio(dev, bio);
-	} else {
 		bio_endio(bio, 0);
-	}
 }
 
 /* length is in bytes.  gfp flags indicates whether we may sleep. */
-static int nvme_setup_prps(struct nvme_dev *dev,
+int nvme_setup_prps(struct nvme_dev *dev, struct nvme_common_command *cmd,
-			struct nvme_common_command *cmd, struct nvme_iod *iod,
+			struct nvme_iod *iod, int total_len, gfp_t gfp)
-			int total_len, gfp_t gfp)
 {
 	struct dma_pool *pool;
 	int length = total_len;
@@ -473,43 +406,193 @@ static int nvme_setup_prps(struct nvme_dev *dev,
 	return total_len;
 }
 
+struct nvme_bio_pair {
+	struct bio b1, b2, *parent;
+	struct bio_vec *bv1, *bv2;
+	int err;
+	atomic_t cnt;
+};
+
+static void nvme_bio_pair_endio(struct bio *bio, int err)
+{
+	struct nvme_bio_pair *bp = bio->bi_private;
+
+	if (err)
+		bp->err = err;
+
+	if (atomic_dec_and_test(&bp->cnt)) {
+		bio_endio(bp->parent, bp->err);
+		if (bp->bv1)
+			kfree(bp->bv1);
+		if (bp->bv2)
+			kfree(bp->bv2);
+		kfree(bp);
+	}
+}
+
+static struct nvme_bio_pair *nvme_bio_split(struct bio *bio, int idx,
+							int len, int offset)
+{
+	struct nvme_bio_pair *bp;
+
+	BUG_ON(len > bio->bi_size);
+	BUG_ON(idx > bio->bi_vcnt);
+
+	bp = kmalloc(sizeof(*bp), GFP_ATOMIC);
+	if (!bp)
+		return NULL;
+	bp->err = 0;
+
+	bp->b1 = *bio;
+	bp->b2 = *bio;
+
+	bp->b1.bi_size = len;
+	bp->b2.bi_size -= len;
+	bp->b1.bi_vcnt = idx;
+	bp->b2.bi_idx = idx;
+	bp->b2.bi_sector += len >> 9;
+
+	if (offset) {
+		bp->bv1 = kmalloc(bio->bi_max_vecs * sizeof(struct bio_vec),
+								GFP_ATOMIC);
+		if (!bp->bv1)
+			goto split_fail_1;
+
+		bp->bv2 = kmalloc(bio->bi_max_vecs * sizeof(struct bio_vec),
+								GFP_ATOMIC);
+		if (!bp->bv2)
+			goto split_fail_2;
+
+		memcpy(bp->bv1, bio->bi_io_vec,
+			bio->bi_max_vecs * sizeof(struct bio_vec));
+		memcpy(bp->bv2, bio->bi_io_vec,
+			bio->bi_max_vecs * sizeof(struct bio_vec));
+
+		bp->b1.bi_io_vec = bp->bv1;
+		bp->b2.bi_io_vec = bp->bv2;
+		bp->b2.bi_io_vec[idx].bv_offset += offset;
+		bp->b2.bi_io_vec[idx].bv_len -= offset;
+		bp->b1.bi_io_vec[idx].bv_len = offset;
+		bp->b1.bi_vcnt++;
+	} else
+		bp->bv1 = bp->bv2 = NULL;
+
+	bp->b1.bi_private = bp;
+	bp->b2.bi_private = bp;
+
+	bp->b1.bi_end_io = nvme_bio_pair_endio;
+	bp->b2.bi_end_io = nvme_bio_pair_endio;
+
+	bp->parent = bio;
+	atomic_set(&bp->cnt, 2);
+
+	return bp;
+
+ split_fail_2:
+	kfree(bp->bv1);
+ split_fail_1:
+	kfree(bp);
+	return NULL;
+}
+
+static int nvme_split_and_submit(struct bio *bio, struct nvme_queue *nvmeq,
+						int idx, int len, int offset)
+{
+	struct nvme_bio_pair *bp = nvme_bio_split(bio, idx, len, offset);
+	if (!bp)
+		return -ENOMEM;
+
+	if (bio_list_empty(&nvmeq->sq_cong))
+		add_wait_queue(&nvmeq->sq_full, &nvmeq->sq_cong_wait);
+	bio_list_add(&nvmeq->sq_cong, &bp->b1);
+	bio_list_add(&nvmeq->sq_cong, &bp->b2);
+
+	return 0;
+}
+
 /* NVMe scatterlists require no holes in the virtual address */
 #define BIOVEC_NOT_VIRT_MERGEABLE(vec1, vec2)	((vec2)->bv_offset || \
 			(((vec1)->bv_offset + (vec1)->bv_len) % PAGE_SIZE))
 
-static int nvme_map_bio(struct device *dev, struct nvme_iod *iod,
+static int nvme_map_bio(struct nvme_queue *nvmeq, struct nvme_iod *iod,
 		struct bio *bio, enum dma_data_direction dma_dir, int psegs)
 {
 	struct bio_vec *bvec, *bvprv = NULL;
 	struct scatterlist *sg = NULL;
-	int i, old_idx, length = 0, nsegs = 0;
+	int i, length = 0, nsegs = 0, split_len = bio->bi_size;
+
+	if (nvmeq->dev->stripe_size)
+		split_len = nvmeq->dev->stripe_size -
+			((bio->bi_sector << 9) & (nvmeq->dev->stripe_size - 1));
 
 	sg_init_table(iod->sg, psegs);
-	old_idx = bio->bi_idx;
 	bio_for_each_segment(bvec, bio, i) {
 		if (bvprv && BIOVEC_PHYS_MERGEABLE(bvprv, bvec)) {
 			sg->length += bvec->bv_len;
 		} else {
 			if (bvprv && BIOVEC_NOT_VIRT_MERGEABLE(bvprv, bvec))
-				break;
+				return nvme_split_and_submit(bio, nvmeq, i,
+								length, 0);
+
 			sg = sg ? sg + 1 : iod->sg;
 			sg_set_page(sg, bvec->bv_page, bvec->bv_len,
 							bvec->bv_offset);
 			nsegs++;
 		}
+
+		if (split_len - length < bvec->bv_len)
+			return nvme_split_and_submit(bio, nvmeq, i, split_len,
+							split_len - length);
 		length += bvec->bv_len;
 		bvprv = bvec;
 	}
-	bio->bi_idx = i;
 	iod->nents = nsegs;
 	sg_mark_end(sg);
-	if (dma_map_sg(dev, iod->sg, iod->nents, dma_dir) == 0) {
+	if (dma_map_sg(nvmeq->q_dmadev, iod->sg, iod->nents, dma_dir) == 0)
-		bio->bi_idx = old_idx;
 		return -ENOMEM;
-	}
+
+	BUG_ON(length != bio->bi_size);
 	return length;
 }
 
+/*
+ * We reuse the small pool to allocate the 16-byte range here as it is not
+ * worth having a special pool for these or additional cases to handle freeing
+ * the iod.
+ */
+static int nvme_submit_discard(struct nvme_queue *nvmeq, struct nvme_ns *ns,
+		struct bio *bio, struct nvme_iod *iod, int cmdid)
+{
+	struct nvme_dsm_range *range;
+	struct nvme_command *cmnd = &nvmeq->sq_cmds[nvmeq->sq_tail];
+
+	range = dma_pool_alloc(nvmeq->dev->prp_small_pool, GFP_ATOMIC,
+							&iod->first_dma);
+	if (!range)
+		return -ENOMEM;
+
+	iod_list(iod)[0] = (__le64 *)range;
+	iod->npages = 0;
+
+	range->cattr = cpu_to_le32(0);
+	range->nlb = cpu_to_le32(bio->bi_size >> ns->lba_shift);
+	range->slba = cpu_to_le64(nvme_block_nr(ns, bio->bi_sector));
+
+	memset(cmnd, 0, sizeof(*cmnd));
+	cmnd->dsm.opcode = nvme_cmd_dsm;
+	cmnd->dsm.command_id = cmdid;
+	cmnd->dsm.nsid = cpu_to_le32(ns->ns_id);
+	cmnd->dsm.prp1 = cpu_to_le64(iod->first_dma);
+	cmnd->dsm.nr = 0;
+	cmnd->dsm.attributes = cpu_to_le32(NVME_DSMGMT_AD);
+
+	if (++nvmeq->sq_tail == nvmeq->q_depth)
+		nvmeq->sq_tail = 0;
+	writel(nvmeq->sq_tail, nvmeq->q_db);
+
+	return 0;
+}
+
 static int nvme_submit_flush(struct nvme_queue *nvmeq, struct nvme_ns *ns,
 								int cmdid)
 {
@@ -527,7 +610,7 @@ static int nvme_submit_flush(struct nvme_queue *nvmeq, struct nvme_ns *ns,
 	return 0;
 }
 
-static int nvme_submit_flush_data(struct nvme_queue *nvmeq, struct nvme_ns *ns)
+int nvme_submit_flush_data(struct nvme_queue *nvmeq, struct nvme_ns *ns)
 {
 	int cmdid = alloc_cmdid(nvmeq, (void *)CMD_CTX_FLUSH,
 					special_completion, NVME_IO_TIMEOUT);
@@ -567,6 +650,12 @@ static int nvme_submit_bio_queue(struct nvme_queue *nvmeq, struct nvme_ns *ns,
 	if (unlikely(cmdid < 0))
 		goto free_iod;
 
+	if (bio->bi_rw & REQ_DISCARD) {
+		result = nvme_submit_discard(nvmeq, ns, bio, iod, cmdid);
+		if (result)
+			goto free_cmdid;
+		return result;
+	}
 	if ((bio->bi_rw & REQ_FLUSH) && !psegs)
 		return nvme_submit_flush(nvmeq, ns, cmdid);
 
@@ -591,8 +680,8 @@ static int nvme_submit_bio_queue(struct nvme_queue *nvmeq, struct nvme_ns *ns,
 		dma_dir = DMA_FROM_DEVICE;
 	}
 
-	result = nvme_map_bio(nvmeq->q_dmadev, iod, bio, dma_dir, psegs);
+	result = nvme_map_bio(nvmeq, iod, bio, dma_dir, psegs);
-	if (result < 0)
+	if (result <= 0)
 		goto free_cmdid;
 	length = result;
 
@@ -600,13 +689,11 @@ static int nvme_submit_bio_queue(struct nvme_queue *nvmeq, struct nvme_ns *ns,
 	cmnd->rw.nsid = cpu_to_le32(ns->ns_id);
 	length = nvme_setup_prps(nvmeq->dev, &cmnd->common, iod, length,
 								GFP_ATOMIC);
-	cmnd->rw.slba = cpu_to_le64(bio->bi_sector >> (ns->lba_shift - 9));
+	cmnd->rw.slba = cpu_to_le64(nvme_block_nr(ns, bio->bi_sector));
 	cmnd->rw.length = cpu_to_le16((length >> ns->lba_shift) - 1);
 	cmnd->rw.control = cpu_to_le16(control);
 	cmnd->rw.dsmgmt = cpu_to_le32(dsmgmt);
 
-	bio->bi_sector += length >> 9;
-
 	if (++nvmeq->sq_tail == nvmeq->q_depth)
 		nvmeq->sq_tail = 0;
 	writel(nvmeq->sq_tail, nvmeq->q_db);
@@ -724,8 +811,8 @@ static void sync_completion(struct nvme_dev *dev, void *ctx,
  * Returns 0 on success.  If the result is negative, it's a Linux error code;
  * if the result is positive, it's an NVM Express status code
  */
-static int nvme_submit_sync_cmd(struct nvme_queue *nvmeq,
+int nvme_submit_sync_cmd(struct nvme_queue *nvmeq, struct nvme_command *cmd,
-			struct nvme_command *cmd, u32 *result, unsigned timeout)
+						u32 *result, unsigned timeout)
 {
 	int cmdid;
 	struct sync_cmd_info cmdinfo;
@@ -741,7 +828,7 @@ static int nvme_submit_sync_cmd(struct nvme_queue *nvmeq,
 
 	set_current_state(TASK_KILLABLE);
 	nvme_submit_cmd(nvmeq, cmd);
-	schedule();
+	schedule_timeout(timeout);
 
 	if (cmdinfo.status == -EINTR) {
 		nvme_abort_command(nvmeq, cmdid);
@@ -754,7 +841,7 @@ static int nvme_submit_sync_cmd(struct nvme_queue *nvmeq,
 	return cmdinfo.status;
 }
 
-static int nvme_submit_admin_cmd(struct nvme_dev *dev, struct nvme_command *cmd,
+int nvme_submit_admin_cmd(struct nvme_dev *dev, struct nvme_command *cmd,
 								u32 *result)
 {
 	return nvme_submit_sync_cmd(dev->queues[0], cmd, result, ADMIN_TIMEOUT);
@@ -827,7 +914,7 @@ static int adapter_delete_sq(struct nvme_dev *dev, u16 sqid)
 	return adapter_delete_queue(dev, nvme_admin_delete_sq, sqid);
 }
 
-static int nvme_identify(struct nvme_dev *dev, unsigned nsid, unsigned cns,
+int nvme_identify(struct nvme_dev *dev, unsigned nsid, unsigned cns,
 							dma_addr_t dma_addr)
 {
 	struct nvme_command c;
@@ -841,7 +928,7 @@ static int nvme_identify(struct nvme_dev *dev, unsigned nsid, unsigned cns,
 	return nvme_submit_admin_cmd(dev, &c, NULL);
 }
 
-static int nvme_get_features(struct nvme_dev *dev, unsigned fid, unsigned nsid,
+int nvme_get_features(struct nvme_dev *dev, unsigned fid, unsigned nsid,
 					dma_addr_t dma_addr, u32 *result)
 {
 	struct nvme_command c;
@@ -855,8 +942,8 @@ static int nvme_get_features(struct nvme_dev *dev, unsigned fid, unsigned nsid,
 	return nvme_submit_admin_cmd(dev, &c, result);
 }
 
-static int nvme_set_features(struct nvme_dev *dev, unsigned fid,
+int nvme_set_features(struct nvme_dev *dev, unsigned fid, unsigned dword11,
-			unsigned dword11, dma_addr_t dma_addr, u32 *result)
+					dma_addr_t dma_addr, u32 *result)
 {
 	struct nvme_command c;
 
@@ -885,7 +972,7 @@ static void nvme_cancel_ios(struct nvme_queue *nvmeq, bool timeout)
 		void *ctx;
 		nvme_completion_fn fn;
 		static struct nvme_completion cqe = {
-			.status = cpu_to_le16(NVME_SC_ABORT_REQ) << 1,
+			.status = cpu_to_le16(NVME_SC_ABORT_REQ << 1),
 		};
 
 		if (timeout && !time_after(now, info[cmdid].timeout))
@@ -966,7 +1053,7 @@ static struct nvme_queue *nvme_alloc_queue(struct nvme_dev *dev, int qid,
 	return nvmeq;
 
  free_cqdma:
-	dma_free_coherent(dmadev, CQ_SIZE(nvmeq->q_depth), (void *)nvmeq->cqes,
+	dma_free_coherent(dmadev, CQ_SIZE(depth), (void *)nvmeq->cqes,
 							nvmeq->cq_dma_addr);
  free_nvmeq:
 	kfree(nvmeq);
@@ -1021,15 +1108,60 @@ static struct nvme_queue *nvme_create_queue(struct nvme_dev *dev, int qid,
 	return ERR_PTR(result);
 }
 
+static int nvme_wait_ready(struct nvme_dev *dev, u64 cap, bool enabled)
+{
+	unsigned long timeout;
+	u32 bit = enabled ? NVME_CSTS_RDY : 0;
+
+	timeout = ((NVME_CAP_TIMEOUT(cap) + 1) * HZ / 2) + jiffies;
+
+	while ((readl(&dev->bar->csts) & NVME_CSTS_RDY) != bit) {
+		msleep(100);
+		if (fatal_signal_pending(current))
+			return -EINTR;
+		if (time_after(jiffies, timeout)) {
+			dev_err(&dev->pci_dev->dev,
+				"Device not ready; aborting initialisation\n");
+			return -ENODEV;
+		}
+	}
+
+	return 0;
+}
+
+/*
+ * If the device has been passed off to us in an enabled state, just clear
+ * the enabled bit.  The spec says we should set the 'shutdown notification
+ * bits', but doing so may cause the device to complete commands to the
+ * admin queue ... and we don't know what memory that might be pointing at!
+ */
+static int nvme_disable_ctrl(struct nvme_dev *dev, u64 cap)
+{
+	u32 cc = readl(&dev->bar->cc);
+
+	if (cc & NVME_CC_ENABLE)
+		writel(cc & ~NVME_CC_ENABLE, &dev->bar->cc);
+	return nvme_wait_ready(dev, cap, false);
+}
+
+static int nvme_enable_ctrl(struct nvme_dev *dev, u64 cap)
+{
+	return nvme_wait_ready(dev, cap, true);
+}
+
 static int nvme_configure_admin_queue(struct nvme_dev *dev)
 {
-	int result = 0;
+	int result;
 	u32 aqa;
-	u64 cap;
+	u64 cap = readq(&dev->bar->cap);
-	unsigned long timeout;
 	struct nvme_queue *nvmeq;
 
 	dev->dbs = ((void __iomem *)dev->bar) + 4096;
+	dev->db_stride = NVME_CAP_STRIDE(cap);
+
+	result = nvme_disable_ctrl(dev, cap);
+	if (result < 0)
+		return result;
 
 	nvmeq = nvme_alloc_queue(dev, 0, 64, 0);
 	if (!nvmeq)
@@ -1043,38 +1175,28 @@ static int nvme_configure_admin_queue(struct nvme_dev *dev)
 	dev->ctrl_config |= NVME_CC_ARB_RR | NVME_CC_SHN_NONE;
 	dev->ctrl_config |= NVME_CC_IOSQES | NVME_CC_IOCQES;
 
-	writel(0, &dev->bar->cc);
 	writel(aqa, &dev->bar->aqa);
 	writeq(nvmeq->sq_dma_addr, &dev->bar->asq);
 	writeq(nvmeq->cq_dma_addr, &dev->bar->acq);
 	writel(dev->ctrl_config, &dev->bar->cc);
 
-	cap = readq(&dev->bar->cap);
+	result = nvme_enable_ctrl(dev, cap);
-	timeout = ((NVME_CAP_TIMEOUT(cap) + 1) * HZ / 2) + jiffies;
+	if (result)
-	dev->db_stride = NVME_CAP_STRIDE(cap);
+		goto free_q;
-
-	while (!result && !(readl(&dev->bar->csts) & NVME_CSTS_RDY)) {
-		msleep(100);
-		if (fatal_signal_pending(current))
-			result = -EINTR;
-		if (time_after(jiffies, timeout)) {
-			dev_err(&dev->pci_dev->dev,
-				"Device not ready; aborting initialisation\n");
-			result = -ENODEV;
-		}
-	}
-
-	if (result) {
-		nvme_free_queue_mem(nvmeq);
-		return result;
-	}
 
 	result = queue_request_irq(dev, nvmeq, "nvme admin");
+	if (result)
+		goto free_q;
+
 	dev->queues[0] = nvmeq;
 	return result;
+
+ free_q:
+	nvme_free_queue_mem(nvmeq);
+	return result;
 }
 
-static struct nvme_iod *nvme_map_user_pages(struct nvme_dev *dev, int write,
+struct nvme_iod *nvme_map_user_pages(struct nvme_dev *dev, int write,
 				unsigned long addr, unsigned length)
 {
 	int i, err, count, nents, offset;
@@ -1130,7 +1252,7 @@ static struct nvme_iod *nvme_map_user_pages(struct nvme_dev *dev, int write,
 	return ERR_PTR(err);
 }
 
-static void nvme_unmap_user_pages(struct nvme_dev *dev, int write,
+void nvme_unmap_user_pages(struct nvme_dev *dev, int write,
 			struct nvme_iod *iod)
 {
 	int i;
@@ -1148,13 +1270,19 @@ static int nvme_submit_io(struct nvme_ns *ns, struct nvme_user_io __user *uio)
 	struct nvme_queue *nvmeq;
 	struct nvme_user_io io;
 	struct nvme_command c;
-	unsigned length;
+	unsigned length, meta_len;
-	int status;
+	int status, i;
-	struct nvme_iod *iod;
+	struct nvme_iod *iod, *meta_iod = NULL;
+	dma_addr_t meta_dma_addr;
+	void *meta, *uninitialized_var(meta_mem);
 
 	if (copy_from_user(&io, uio, sizeof(io)))
 		return -EFAULT;
 	length = (io.nblocks + 1) << ns->lba_shift;
+	meta_len = (io.nblocks + 1) * ns->ms;
+
+	if (meta_len && ((io.metadata & 3) || !io.metadata))
+		return -EINVAL;
 
 	switch (io.opcode) {
 	case nvme_cmd_write:
@@ -1176,11 +1304,42 @@ static int nvme_submit_io(struct nvme_ns *ns, struct nvme_user_io __user *uio)
 	c.rw.slba = cpu_to_le64(io.slba);
 	c.rw.length = cpu_to_le16(io.nblocks);
 	c.rw.control = cpu_to_le16(io.control);
-	c.rw.dsmgmt = cpu_to_le16(io.dsmgmt);
+	c.rw.dsmgmt = cpu_to_le32(io.dsmgmt);
-	c.rw.reftag = io.reftag;
+	c.rw.reftag = cpu_to_le32(io.reftag);
-	c.rw.apptag = io.apptag;
+	c.rw.apptag = cpu_to_le16(io.apptag);
-	c.rw.appmask = io.appmask;
+	c.rw.appmask = cpu_to_le16(io.appmask);
-	/* XXX: metadata */
+
+	if (meta_len) {
+		meta_iod = nvme_map_user_pages(dev, io.opcode & 1, io.metadata, meta_len);
+		if (IS_ERR(meta_iod)) {
+			status = PTR_ERR(meta_iod);
+			meta_iod = NULL;
+			goto unmap;
+		}
+
+		meta_mem = dma_alloc_coherent(&dev->pci_dev->dev, meta_len,
+						&meta_dma_addr, GFP_KERNEL);
+		if (!meta_mem) {
+			status = -ENOMEM;
+			goto unmap;
+		}
+
+		if (io.opcode & 1) {
+			int meta_offset = 0;
+
+			for (i = 0; i < meta_iod->nents; i++) {
+				meta = kmap_atomic(sg_page(&meta_iod->sg[i])) +
+						meta_iod->sg[i].offset;
+				memcpy(meta_mem + meta_offset, meta,
+						meta_iod->sg[i].length);
+				kunmap_atomic(meta);
+				meta_offset += meta_iod->sg[i].length;
+			}
+		}
+
+		c.rw.metadata = cpu_to_le64(meta_dma_addr);
+	}
+
 	length = nvme_setup_prps(dev, &c.common, iod, length, GFP_KERNEL);
 
 	nvmeq = get_nvmeq(dev);
@@ -1196,8 +1355,33 @@ static int nvme_submit_io(struct nvme_ns *ns, struct nvme_user_io __user *uio)
 	else
 		status = nvme_submit_sync_cmd(nvmeq, &c, NULL, NVME_IO_TIMEOUT);
 
+	if (meta_len) {
+		if (status == NVME_SC_SUCCESS && !(io.opcode & 1)) {
+			int meta_offset = 0;
+
+			for (i = 0; i < meta_iod->nents; i++) {
+				meta = kmap_atomic(sg_page(&meta_iod->sg[i])) +
+						meta_iod->sg[i].offset;
+				memcpy(meta, meta_mem + meta_offset,
+						meta_iod->sg[i].length);
+				kunmap_atomic(meta);
+				meta_offset += meta_iod->sg[i].length;
+			}
+		}
+
+		dma_free_coherent(&dev->pci_dev->dev, meta_len, meta_mem,
+								meta_dma_addr);
+	}
+
+ unmap:
 	nvme_unmap_user_pages(dev, io.opcode & 1, iod);
 	nvme_free_iod(dev, iod);
+
+	if (meta_iod) {
+		nvme_unmap_user_pages(dev, io.opcode & 1, meta_iod);
+		nvme_free_iod(dev, meta_iod);
+	}
+
 	return status;
 }
 
@@ -1208,6 +1392,7 @@ static int nvme_user_admin_cmd(struct nvme_dev *dev,
 	struct nvme_command c;
 	int status, length;
 	struct nvme_iod *uninitialized_var(iod);
+	unsigned timeout;
 
 	if (!capable(CAP_SYS_ADMIN))
 		return -EACCES;
@@ -1237,10 +1422,13 @@ static int nvme_user_admin_cmd(struct nvme_dev *dev,
 								GFP_KERNEL);
 	}
 
+	timeout = cmd.timeout_ms ? msecs_to_jiffies(cmd.timeout_ms) :
+								ADMIN_TIMEOUT;
 	if (length != cmd.data_len)
 		status = -ENOMEM;
 	else
-		status = nvme_submit_admin_cmd(dev, &c, &cmd.result);
+		status = nvme_submit_sync_cmd(dev->queues[0], &c, &cmd.result,
+								timeout);
 
 	if (cmd.data_len) {
 		nvme_unmap_user_pages(dev, cmd.opcode & 1, iod);
@@ -1266,6 +1454,10 @@ static int nvme_ioctl(struct block_device *bdev, fmode_t mode, unsigned int cmd,
 		return nvme_user_admin_cmd(ns->dev, (void __user *)arg);
 	case NVME_IOCTL_SUBMIT_IO:
 		return nvme_submit_io(ns, (void __user *)arg);
+	case SG_GET_VERSION_NUM:
+		return nvme_sg_get_version_num((void __user *)arg);
+	case SG_IO:
+		return nvme_sg_io(ns, (void __user *)arg);
 	default:
 		return -ENOTTY;
 	}
@@ -1282,13 +1474,17 @@ static void nvme_resubmit_bios(struct nvme_queue *nvmeq)
 	while (bio_list_peek(&nvmeq->sq_cong)) {
 		struct bio *bio = bio_list_pop(&nvmeq->sq_cong);
 		struct nvme_ns *ns = bio->bi_bdev->bd_disk->private_data;
-		if (nvme_submit_bio_queue(nvmeq, ns, bio)) {
+
-			bio_list_add_head(&nvmeq->sq_cong, bio);
-			break;
-		}
 		if (bio_list_empty(&nvmeq->sq_cong))
 			remove_wait_queue(&nvmeq->sq_full,
 							&nvmeq->sq_cong_wait);
+		if (nvme_submit_bio_queue(nvmeq, ns, bio)) {
+			if (bio_list_empty(&nvmeq->sq_cong))
+				add_wait_queue(&nvmeq->sq_full,
+							&nvmeq->sq_cong_wait);
+			bio_list_add_head(&nvmeq->sq_cong, bio);
+			break;
+		}
 	}
 }
 
@@ -1297,7 +1493,7 @@ static int nvme_kthread(void *data)
 	struct nvme_dev *dev;
 
 	while (!kthread_should_stop()) {
-		__set_current_state(TASK_RUNNING);
+		set_current_state(TASK_INTERRUPTIBLE);
 		spin_lock(&dev_list_lock);
 		list_for_each_entry(dev, &dev_list, node) {
 			int i;
@@ -1314,8 +1510,7 @@ static int nvme_kthread(void *data)
 			}
 		}
 		spin_unlock(&dev_list_lock);
-		set_current_state(TASK_INTERRUPTIBLE);
+		schedule_timeout(round_jiffies_relative(HZ));
-		schedule_timeout(HZ);
 	}
 	return 0;
 }
@@ -1347,6 +1542,16 @@ static void nvme_put_ns_idx(int index)
 	spin_unlock(&dev_list_lock);
 }
 
+static void nvme_config_discard(struct nvme_ns *ns)
+{
+	u32 logical_block_size = queue_logical_block_size(ns->queue);
+	ns->queue->limits.discard_zeroes_data = 0;
+	ns->queue->limits.discard_alignment = logical_block_size;
+	ns->queue->limits.discard_granularity = logical_block_size;
+	ns->queue->limits.max_discard_sectors = 0xffffffff;
+	queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, ns->queue);
+}
+
 static struct nvme_ns *nvme_alloc_ns(struct nvme_dev *dev, int nsid,
 			struct nvme_id_ns *id, struct nvme_lba_range_type *rt)
 {
@@ -1366,7 +1571,6 @@ static struct nvme_ns *nvme_alloc_ns(struct nvme_dev *dev, int nsid,
 	ns->queue->queue_flags = QUEUE_FLAG_DEFAULT;
 	queue_flag_set_unlocked(QUEUE_FLAG_NOMERGES, ns->queue);
 	queue_flag_set_unlocked(QUEUE_FLAG_NONROT, ns->queue);
-/*	queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, ns->queue); */
 	blk_queue_make_request(ns->queue, nvme_make_request);
 	ns->dev = dev;
 	ns->queue->queuedata = ns;
@@ -1378,6 +1582,7 @@ static struct nvme_ns *nvme_alloc_ns(struct nvme_dev *dev, int nsid,
 	ns->disk = disk;
 	lbaf = id->flbas & 0xf;
 	ns->lba_shift = id->lbaf[lbaf].ds;
+	ns->ms = le16_to_cpu(id->lbaf[lbaf].ms);
 	blk_queue_logical_block_size(ns->queue, 1 << ns->lba_shift);
 	if (dev->max_hw_sectors)
 		blk_queue_max_hw_sectors(ns->queue, dev->max_hw_sectors);
@@ -1392,6 +1597,9 @@ static struct nvme_ns *nvme_alloc_ns(struct nvme_dev *dev, int nsid,
 	sprintf(disk->disk_name, "nvme%dn%d", dev->instance, nsid);
 	set_capacity(disk, le64_to_cpup(&id->nsze) << (ns->lba_shift - 9));
 
+	if (dev->oncs & NVME_CTRL_ONCS_DSM)
+		nvme_config_discard(ns);
+
 	return ns;
 
  out_free_queue:
@@ -1496,14 +1704,21 @@ static void nvme_free_queues(struct nvme_dev *dev)
 		nvme_free_queue(dev, i);
 }
 
+/*
+ * Return: error value if an error occurred setting up the queues or calling
+ * Identify Device.  0 if these succeeded, even if adding some of the
+ * namespaces failed.  At the moment, these failures are silent.  TBD which
+ * failures should be reported.
+ */
 static int nvme_dev_add(struct nvme_dev *dev)
 {
 	int res, nn, i;
-	struct nvme_ns *ns, *next;
+	struct nvme_ns *ns;
 	struct nvme_id_ctrl *ctrl;
 	struct nvme_id_ns *id_ns;
 	void *mem;
 	dma_addr_t dma_addr;
+	int shift = NVME_CAP_MPSMIN(readq(&dev->bar->cap)) + 12;
 
 	res = nvme_setup_io_queues(dev);
 	if (res)
@@ -1511,22 +1726,26 @@ static int nvme_dev_add(struct nvme_dev *dev)
 
 	mem = dma_alloc_coherent(&dev->pci_dev->dev, 8192, &dma_addr,
 								GFP_KERNEL);
+	if (!mem)
+		return -ENOMEM;
 
 	res = nvme_identify(dev, 0, 1, dma_addr);
 	if (res) {
 		res = -EIO;
-		goto out_free;
+		goto out;
 	}
 
 	ctrl = mem;
 	nn = le32_to_cpup(&ctrl->nn);
+	dev->oncs = le16_to_cpup(&ctrl->oncs);
 	memcpy(dev->serial, ctrl->sn, sizeof(ctrl->sn));
 	memcpy(dev->model, ctrl->mn, sizeof(ctrl->mn));
 	memcpy(dev->firmware_rev, ctrl->fr, sizeof(ctrl->fr));
-	if (ctrl->mdts) {
+	if (ctrl->mdts)
-		int shift = NVME_CAP_MPSMIN(readq(&dev->bar->cap)) + 12;
 		dev->max_hw_sectors = 1 << (ctrl->mdts + shift - 9);
-	}
+	if ((dev->pci_dev->vendor == PCI_VENDOR_ID_INTEL) &&
+			(dev->pci_dev->device == 0x0953) && ctrl->vs[3])
+		dev->stripe_size = 1 << (ctrl->vs[3] + shift);
 
 	id_ns = mem;
 	for (i = 1; i <= nn; i++) {
@@ -1548,14 +1767,7 @@ static int nvme_dev_add(struct nvme_dev *dev)
 	}
 	list_for_each_entry(ns, &dev->namespaces, list)
 		add_disk(ns->disk);
-
+	res = 0;
-	goto out;
-
- out_free:
-	list_for_each_entry_safe(ns, next, &dev->namespaces, list) {
-		list_del(&ns->list);
-		nvme_ns_free(ns);
-	}
 
  out:
 	dma_free_coherent(&dev->pci_dev->dev, 8192, mem, dma_addr);
@@ -1634,6 +1846,56 @@ static void nvme_release_instance(struct nvme_dev *dev)
 	spin_unlock(&dev_list_lock);
 }
 
+static void nvme_free_dev(struct kref *kref)
+{
+	struct nvme_dev *dev = container_of(kref, struct nvme_dev, kref);
+	nvme_dev_remove(dev);
+	pci_disable_msix(dev->pci_dev);
+	iounmap(dev->bar);
+	nvme_release_instance(dev);
+	nvme_release_prp_pools(dev);
+	pci_disable_device(dev->pci_dev);
+	pci_release_regions(dev->pci_dev);
+	kfree(dev->queues);
+	kfree(dev->entry);
+	kfree(dev);
+}
+
+static int nvme_dev_open(struct inode *inode, struct file *f)
+{
+	struct nvme_dev *dev = container_of(f->private_data, struct nvme_dev,
+								miscdev);
+	kref_get(&dev->kref);
+	f->private_data = dev;
+	return 0;
+}
+
+static int nvme_dev_release(struct inode *inode, struct file *f)
+{
+	struct nvme_dev *dev = f->private_data;
+	kref_put(&dev->kref, nvme_free_dev);
+	return 0;
+}
+
+static long nvme_dev_ioctl(struct file *f, unsigned int cmd, unsigned long arg)
+{
+	struct nvme_dev *dev = f->private_data;
+	switch (cmd) {
+	case NVME_IOCTL_ADMIN_CMD:
+		return nvme_user_admin_cmd(dev, (void __user *)arg);
+	default:
+		return -ENOTTY;
+	}
+}
+
+static const struct file_operations nvme_dev_fops = {
+	.owner		= THIS_MODULE,
+	.open		= nvme_dev_open,
+	.release	= nvme_dev_release,
+	.unlocked_ioctl	= nvme_dev_ioctl,
+	.compat_ioctl	= nvme_dev_ioctl,
+};
+
 static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 {
 	int bars, result = -ENOMEM;
@@ -1692,8 +1954,20 @@ static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	if (result)
 		goto delete;
 
+	scnprintf(dev->name, sizeof(dev->name), "nvme%d", dev->instance);
+	dev->miscdev.minor = MISC_DYNAMIC_MINOR;
+	dev->miscdev.parent = &pdev->dev;
+	dev->miscdev.name = dev->name;
+	dev->miscdev.fops = &nvme_dev_fops;
+	result = misc_register(&dev->miscdev);
+	if (result)
+		goto remove;
+
+	kref_init(&dev->kref);
 	return 0;
 
+ remove:
+	nvme_dev_remove(dev);
  delete:
 	spin_lock(&dev_list_lock);
 	list_del(&dev->node);
@@ -1719,16 +1993,8 @@ static int nvme_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 static void nvme_remove(struct pci_dev *pdev)
 {
 	struct nvme_dev *dev = pci_get_drvdata(pdev);
-	nvme_dev_remove(dev);
+	misc_deregister(&dev->miscdev);
-	pci_disable_msix(pdev);
+	kref_put(&dev->kref, nvme_free_dev);
-	iounmap(dev->bar);
-	nvme_release_instance(dev);
-	nvme_release_prp_pools(dev);
-	pci_disable_device(pdev);
-	pci_release_regions(pdev);
-	kfree(dev->queues);
-	kfree(dev->entry);
-	kfree(dev);
 }
 
 /* These functions are yet to be implemented */

include/linux/nvme.h

@@ -107,6 +107,12 @@ struct nvme_id_ctrl {
 	__u8			vs[1024];
 };
 
+enum {
+	NVME_CTRL_ONCS_COMPARE			= 1 << 0,
+	NVME_CTRL_ONCS_WRITE_UNCORRECTABLE	= 1 << 1,
+	NVME_CTRL_ONCS_DSM			= 1 << 2,
+};
+
 struct nvme_lbaf {
 	__le16			ms;
 	__u8			ds;
@@ -201,11 +207,11 @@ struct nvme_common_command {
 	__u8			flags;
 	__u16			command_id;
 	__le32			nsid;
-	__u32			cdw2[2];
+	__le32			cdw2[2];
 	__le64			metadata;
 	__le64			prp1;
 	__le64			prp2;
-	__u32			cdw10[6];
+	__le32			cdw10[6];
 };
 
 struct nvme_rw_command {
@@ -246,6 +252,31 @@ enum {
 	NVME_RW_DSM_COMPRESSED		= 1 << 7,
 };
 
+struct nvme_dsm_cmd {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u64			rsvd2[2];
+	__le64			prp1;
+	__le64			prp2;
+	__le32			nr;
+	__le32			attributes;
+	__u32			rsvd12[4];
+};
+
+enum {
+	NVME_DSMGMT_IDR		= 1 << 0,
+	NVME_DSMGMT_IDW		= 1 << 1,
+	NVME_DSMGMT_AD		= 1 << 2,
+};
+
+struct nvme_dsm_range {
+	__le32			cattr;
+	__le32			nlb;
+	__le64			slba;
+};
+
 /* Admin commands */
 
 enum nvme_admin_opcode {
@@ -285,6 +316,9 @@ enum {
 	NVME_FEAT_WRITE_ATOMIC	= 0x0a,
 	NVME_FEAT_ASYNC_EVENT	= 0x0b,
 	NVME_FEAT_SW_PROGRESS	= 0x0c,
+	NVME_FWACT_REPL		= (0 << 3),
+	NVME_FWACT_REPL_ACTV	= (1 << 3),
+	NVME_FWACT_ACTV		= (2 << 3),
 };
 
 struct nvme_identify {
@@ -362,6 +396,16 @@ struct nvme_download_firmware {
 	__u32			rsvd12[4];
 };
 
+struct nvme_format_cmd {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u64			rsvd2[4];
+	__le32			cdw10;
+	__u32			rsvd11[5];
+};
+
 struct nvme_command {
 	union {
 		struct nvme_common_command common;
@@ -372,6 +416,8 @@ struct nvme_command {
 		struct nvme_create_sq create_sq;
 		struct nvme_delete_queue delete_queue;
 		struct nvme_download_firmware dlfw;
+		struct nvme_format_cmd format;
+		struct nvme_dsm_cmd dsm;
 	};
 };
 
@@ -388,6 +434,7 @@ enum {
 	NVME_SC_FUSED_FAIL		= 0x9,
 	NVME_SC_FUSED_MISSING		= 0xa,
 	NVME_SC_INVALID_NS		= 0xb,
+	NVME_SC_CMD_SEQ_ERROR		= 0xc,
 	NVME_SC_LBA_RANGE		= 0x80,
 	NVME_SC_CAP_EXCEEDED		= 0x81,
 	NVME_SC_NS_NOT_READY		= 0x82,
@@ -461,4 +508,111 @@ struct nvme_admin_cmd {
 #define NVME_IOCTL_ADMIN_CMD	_IOWR('N', 0x41, struct nvme_admin_cmd)
 #define NVME_IOCTL_SUBMIT_IO	_IOW('N', 0x42, struct nvme_user_io)
 
+#ifdef __KERNEL__
+#include <linux/pci.h>
+#include <linux/miscdevice.h>
+#include <linux/kref.h>
+
+#define NVME_IO_TIMEOUT	(5 * HZ)
+
+/*
+ * Represents an NVM Express device.  Each nvme_dev is a PCI function.
+ */
+struct nvme_dev {
+	struct list_head node;
+	struct nvme_queue **queues;
+	u32 __iomem *dbs;
+	struct pci_dev *pci_dev;
+	struct dma_pool *prp_page_pool;
+	struct dma_pool *prp_small_pool;
+	int instance;
+	int queue_count;
+	int db_stride;
+	u32 ctrl_config;
+	struct msix_entry *entry;
+	struct nvme_bar __iomem *bar;
+	struct list_head namespaces;
+	struct kref kref;
+	struct miscdevice miscdev;
+	char name[12];
+	char serial[20];
+	char model[40];
+	char firmware_rev[8];
+	u32 max_hw_sectors;
+	u32 stripe_size;
+	u16 oncs;
+};
+
+/*
+ * An NVM Express namespace is equivalent to a SCSI LUN
+ */
+struct nvme_ns {
+	struct list_head list;
+
+	struct nvme_dev *dev;
+	struct request_queue *queue;
+	struct gendisk *disk;
+
+	int ns_id;
+	int lba_shift;
+	int ms;
+	u64 mode_select_num_blocks;
+	u32 mode_select_block_len;
+};
+
+/*
+ * The nvme_iod describes the data in an I/O, including the list of PRP
+ * entries.  You can't see it in this data structure because C doesn't let
+ * me express that.  Use nvme_alloc_iod to ensure there's enough space
+ * allocated to store the PRP list.
+ */
+struct nvme_iod {
+	void *private;		/* For the use of the submitter of the I/O */
+	int npages;		/* In the PRP list. 0 means small pool in use */
+	int offset;		/* Of PRP list */
+	int nents;		/* Used in scatterlist */
+	int length;		/* Of data, in bytes */
+	dma_addr_t first_dma;
+	struct scatterlist sg[0];
+};
+
+static inline u64 nvme_block_nr(struct nvme_ns *ns, sector_t sector)
+{
+	return (sector >> (ns->lba_shift - 9));
+}
+
+/**
+ * nvme_free_iod - frees an nvme_iod
+ * @dev: The device that the I/O was submitted to
+ * @iod: The memory to free
+ */
+void nvme_free_iod(struct nvme_dev *dev, struct nvme_iod *iod);
+
+int nvme_setup_prps(struct nvme_dev *dev, struct nvme_common_command *cmd,
+			struct nvme_iod *iod, int total_len, gfp_t gfp);
+struct nvme_iod *nvme_map_user_pages(struct nvme_dev *dev, int write,
+				unsigned long addr, unsigned length);
+void nvme_unmap_user_pages(struct nvme_dev *dev, int write,
+			struct nvme_iod *iod);
+struct nvme_queue *get_nvmeq(struct nvme_dev *dev);
+void put_nvmeq(struct nvme_queue *nvmeq);
+int nvme_submit_sync_cmd(struct nvme_queue *nvmeq, struct nvme_command *cmd,
+						u32 *result, unsigned timeout);
+int nvme_submit_flush_data(struct nvme_queue *nvmeq, struct nvme_ns *ns);
+int nvme_submit_admin_cmd(struct nvme_dev *, struct nvme_command *,
+							u32 *result);
+int nvme_identify(struct nvme_dev *, unsigned nsid, unsigned cns,
+							dma_addr_t dma_addr);
+int nvme_get_features(struct nvme_dev *dev, unsigned fid, unsigned nsid,
+			dma_addr_t dma_addr, u32 *result);
+int nvme_set_features(struct nvme_dev *dev, unsigned fid, unsigned dword11,
+			dma_addr_t dma_addr, u32 *result);
+
+struct sg_io_hdr;
+
+int nvme_sg_io(struct nvme_ns *ns, struct sg_io_hdr __user *u_hdr);
+int nvme_sg_get_version_num(int __user *ip);
+
+#endif
+
 #endif /* _LINUX_NVME_H */