NVMe: New driver

This driver is for devices that follow the NVM Express standard Signed-off-by: Matthew Wilcox <matthew.r.wilcox@intel.com>

NVMe: New driver
This driver is for devices that follow the NVM Express standard Signed-off-by: Matthew Wilcox <matthew.r.wilcox@intel.com>
Matthew Wilcox
1 parent 0b934ccd70
Showing 5 changed files with 1399 additions and 0 deletions Side-by-side Diff
Documentation/ioctl/ioctl-number.txt
drivers/block/Kconfig
drivers/block/Makefile
drivers/block/nvme.c
include/linux/nvme.h
@@ -149,6 +149,7 @@
 'M'	01-03	drivers/scsi/megaraid/megaraid_sas.h
 'M'	00-0F	drivers/video/fsl-diu-fb.h	conflict!
 'N'	00-1F	drivers/usb/scanner.h
+'N'	40-7F	drivers/block/nvme.c
 'O'     00-06   mtd/ubi-user.h		UBI
 'P'	all	linux/soundcard.h	conflict!
 'P'	60-6F	sound/sscape_ioctl.h	conflict!
@@ -315,6 +315,17 @@
  
 	  If unsure, say N.
  
+config BLK_DEV_NVME
+	tristate "NVM Express block device"
+	depends on PCI
+	---help---
+	  The NVM Express driver is for solid state drives directly
+	  connected to the PCI or PCI Express bus.  If you know you
+	  don't have one of these, it is safe to answer N.
+
+	  To compile this driver as a module, choose M here: the
+	  module will be called nvme.
+
 config BLK_DEV_OSD
 	tristate "OSD object-as-blkdev support"
 	depends on SCSI_OSD_ULD
@@ -23,6 +23,7 @@
 obj-$(CONFIG_CDROM_PKTCDVD)	+= pktcdvd.o
 obj-$(CONFIG_MG_DISK)		+= mg_disk.o
 obj-$(CONFIG_SUNVDC)		+= sunvdc.o
+obj-$(CONFIG_BLK_DEV_NVME)	+= nvme.o
 obj-$(CONFIG_BLK_DEV_OSD)	+= osdblk.o
  
 obj-$(CONFIG_BLK_DEV_UMEM)	+= umem.o
+/*
+ * NVM Express device driver
+ * Copyright (c) 2011, Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc.,
+ * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ */
+
+#include <linux/nvme.h>
+#include <linux/bio.h>
+#include <linux/blkdev.h>
+#include <linux/errno.h>
+#include <linux/fs.h>
+#include <linux/genhd.h>
+#include <linux/init.h>
+#include <linux/interrupt.h>
+#include <linux/io.h>
+#include <linux/kdev_t.h>
+#include <linux/kernel.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/moduleparam.h>
+#include <linux/pci.h>
+#include <linux/sched.h>
+#include <linux/slab.h>
+#include <linux/types.h>
+#include <linux/version.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
+
+static int nvme_major;
+module_param(nvme_major, int, 0);
+
+/*
+ * 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;
+	int instance;
+	int queue_count;
+	u32 ctrl_config;
+	struct msix_entry *entry;
+	struct nvme_bar __iomem *bar;
+	struct list_head namespaces;
+};
+
+/*
+ * 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).
+ */
+struct nvme_queue {
+	struct device *q_dmadev;
+	spinlock_t q_lock;
+	struct nvme_command *sq_cmds;
+	volatile struct nvme_completion *cqes;
+	dma_addr_t sq_dma_addr;
+	dma_addr_t cq_dma_addr;
+	wait_queue_head_t sq_full;
+	struct bio_list sq_cong;
+	u32 __iomem *q_db;
+	u16 q_depth;
+	u16 cq_vector;
+	u16 sq_head;
+	u16 sq_tail;
+	u16 cq_head;
+	u16 cq_cycle;
+	unsigned long cmdid_data[];
+};
+
+/*
+ * Check we didin't inadvertently grow the command struct
+ */
+static inline void _nvme_check_size(void)
+{
+	BUILD_BUG_ON(sizeof(struct nvme_rw_command) != 64);
+	BUILD_BUG_ON(sizeof(struct nvme_create_cq) != 64);
+	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_command) != 64);
+	BUILD_BUG_ON(sizeof(struct nvme_id_ctrl) != 4096);
+	BUILD_BUG_ON(sizeof(struct nvme_id_ns) != 4096);
+	BUILD_BUG_ON(sizeof(struct nvme_lba_range_type) != 64);
+}
+
+/**
+ * alloc_cmdid - Allocate a Command ID
+ * @param nvmeq The queue that will be used for this command
+ * @param ctx A pointer that will be passed to the handler
+ * @param handler The ID of the handler to call
+ *
+ * Allocate a Command ID for a queue.  The data passed in will
+ * be passed to the completion handler.  This is implemented by using
+ * the bottom two bits of the ctx pointer to store the handler ID.
+ * Passing in a pointer that's not 4-byte aligned will cause a BUG.
+ * We can change this if it becomes a problem.
+ */
+static int alloc_cmdid(struct nvme_queue *nvmeq, void *ctx, int handler)
+{
+	int depth = nvmeq->q_depth;
+	unsigned long data = (unsigned long)ctx | handler;
+	int cmdid;
+
+	BUG_ON((unsigned long)ctx & 3);
+
+	do {
+		cmdid = find_first_zero_bit(nvmeq->cmdid_data, depth);
+		if (cmdid >= depth)
+			return -EBUSY;
+	} while (test_and_set_bit(cmdid, nvmeq->cmdid_data));
+
+	nvmeq->cmdid_data[cmdid + BITS_TO_LONGS(depth)] = data;
+	return cmdid;
+}
+
+static int alloc_cmdid_killable(struct nvme_queue *nvmeq, void *ctx,
+								int handler)
+{
+	int cmdid;
+	wait_event_killable(nvmeq->sq_full,
+			(cmdid = alloc_cmdid(nvmeq, ctx, handler)) >= 0);
+	return (cmdid < 0) ? -EINTR : cmdid;
+}
+
+/* If you need more than four handlers, you'll need to change how
+ * alloc_cmdid and nvme_process_cq work
+ */
+enum {
+	sync_completion_id = 0,
+	bio_completion_id,
+};
+
+static unsigned long free_cmdid(struct nvme_queue *nvmeq, int cmdid)
+{
+	unsigned long data;
+
+	data = nvmeq->cmdid_data[cmdid + BITS_TO_LONGS(nvmeq->q_depth)];
+	clear_bit(cmdid, nvmeq->cmdid_data);
+	wake_up(&nvmeq->sq_full);
+	return data;
+}
+
+static struct nvme_queue *get_nvmeq(struct nvme_ns *ns)
+{
+	return ns->dev->queues[1];
+}
+
+static void put_nvmeq(struct nvme_queue *nvmeq)
+{
+}
+
+/**
+ * nvme_submit_cmd: Copy a command into a queue and ring the doorbell
+ * @nvmeq: The queue to use
+ * @cmd: The command to send
+ *
+ * Safe to use from interrupt context
+ */
+static int nvme_submit_cmd(struct nvme_queue *nvmeq, struct nvme_command *cmd)
+{
+	unsigned long flags;
+	u16 tail;
+	/* XXX: Need to check tail isn't going to overrun head */
+	spin_lock_irqsave(&nvmeq->q_lock, flags);
+	tail = nvmeq->sq_tail;
+	memcpy(&nvmeq->sq_cmds[tail], cmd, sizeof(*cmd));
+	writel(tail, nvmeq->q_db);
+	if (++tail == nvmeq->q_depth)
+		tail = 0;
+	nvmeq->sq_tail = tail;
+	spin_unlock_irqrestore(&nvmeq->q_lock, flags);
+
+	return 0;
+}
+
+struct nvme_req_info {
+	struct bio *bio;
+	int nents;
+	struct scatterlist sg[0];
+};
+
+/* XXX: use a mempool */
+static struct nvme_req_info *alloc_info(unsigned nseg, gfp_t gfp)
+{
+	return kmalloc(sizeof(struct nvme_req_info) +
+			sizeof(struct scatterlist) * nseg, gfp);
+}
+
+static void free_info(struct nvme_req_info *info)
+{
+	kfree(info);
+}
+
+static void bio_completion(struct nvme_queue *nvmeq, void *ctx,
+						struct nvme_completion *cqe)
+{
+	struct nvme_req_info *info = ctx;
+	struct bio *bio = info->bio;
+	u16 status = le16_to_cpup(&cqe->status) >> 1;
+
+	dma_unmap_sg(nvmeq->q_dmadev, info->sg, info->nents,
+			bio_data_dir(bio) ? DMA_TO_DEVICE : DMA_FROM_DEVICE);
+	free_info(info);
+	bio_endio(bio, status ? -EIO : 0);
+}
+
+static int nvme_map_bio(struct device *dev, struct nvme_req_info *info,
+		struct bio *bio, enum dma_data_direction dma_dir, int psegs)
+{
+	struct bio_vec *bvec;
+	struct scatterlist *sg = info->sg;
+	int i, nsegs;
+
+	sg_init_table(sg, psegs);
+	bio_for_each_segment(bvec, bio, i) {
+		sg_set_page(sg, bvec->bv_page, bvec->bv_len, bvec->bv_offset);
+		/* XXX: handle non-mergable here */
+		nsegs++;
+	}
+	info->nents = nsegs;
+
+	return dma_map_sg(dev, info->sg, info->nents, dma_dir);
+}
+
+static int nvme_submit_bio_queue(struct nvme_queue *nvmeq, struct nvme_ns *ns,
+								struct bio *bio)
+{
+	struct nvme_rw_command *cmnd;
+	struct nvme_req_info *info;
+	enum dma_data_direction dma_dir;
+	int cmdid;
+	u16 control;
+	u32 dsmgmt;
+	unsigned long flags;
+	int psegs = bio_phys_segments(ns->queue, bio);
+
+	info = alloc_info(psegs, GFP_NOIO);
+	if (!info)
+		goto congestion;
+	info->bio = bio;
+
+	cmdid = alloc_cmdid(nvmeq, info, bio_completion_id);
+	if (unlikely(cmdid < 0))
+		goto free_info;
+
+	control = 0;
+	if (bio->bi_rw & REQ_FUA)
+		control |= NVME_RW_FUA;
+	if (bio->bi_rw & (REQ_FAILFAST_DEV | REQ_RAHEAD))
+		control |= NVME_RW_LR;
+
+	dsmgmt = 0;
+	if (bio->bi_rw & REQ_RAHEAD)
+		dsmgmt |= NVME_RW_DSM_FREQ_PREFETCH;
+
+	spin_lock_irqsave(&nvmeq->q_lock, flags);
+	cmnd = &nvmeq->sq_cmds[nvmeq->sq_tail].rw;
+
+	if (bio_data_dir(bio)) {
+		cmnd->opcode = nvme_cmd_write;
+		dma_dir = DMA_TO_DEVICE;
+	} else {
+		cmnd->opcode = nvme_cmd_read;
+		dma_dir = DMA_FROM_DEVICE;
+	}
+
+	nvme_map_bio(nvmeq->q_dmadev, info, bio, dma_dir, psegs);
+
+	cmnd->flags = 1;
+	cmnd->command_id = cmdid;
+	cmnd->nsid = cpu_to_le32(ns->ns_id);
+	cmnd->prp1 = cpu_to_le64(sg_phys(info->sg));
+	/* XXX: Support more than one PRP */
+	cmnd->slba = cpu_to_le64(bio->bi_sector >> (ns->lba_shift - 9));
+	cmnd->length = cpu_to_le16((bio->bi_size >> ns->lba_shift) - 1);
+	cmnd->control = cpu_to_le16(control);
+	cmnd->dsmgmt = cpu_to_le32(dsmgmt);
+
+	writel(nvmeq->sq_tail, nvmeq->q_db);
+	if (++nvmeq->sq_tail == nvmeq->q_depth)
+		nvmeq->sq_tail = 0;
+
+	spin_unlock_irqrestore(&nvmeq->q_lock, flags);
+
+	return 0;
+
+ free_info:
+	free_info(info);
+ congestion:
+	return -EBUSY;
+}
+
+/*
+ * NB: return value of non-zero would mean that we were a stacking driver.
+ * make_request must always succeed.
+ */
+static int nvme_make_request(struct request_queue *q, struct bio *bio)
+{
+	struct nvme_ns *ns = q->queuedata;
+	struct nvme_queue *nvmeq = get_nvmeq(ns);
+
+	if (nvme_submit_bio_queue(nvmeq, ns, bio)) {
+		blk_set_queue_congested(q, rw_is_sync(bio->bi_rw));
+		bio_list_add(&nvmeq->sq_cong, bio);
+	}
+	put_nvmeq(nvmeq);
+
+	return 0;
+}
+
+struct sync_cmd_info {
+	struct task_struct *task;
+	u32 result;
+	int status;
+};
+
+static void sync_completion(struct nvme_queue *nvmeq, void *ctx,
+						struct nvme_completion *cqe)
+{
+	struct sync_cmd_info *cmdinfo = ctx;
+	cmdinfo->result = le32_to_cpup(&cqe->result);
+	cmdinfo->status = le16_to_cpup(&cqe->status) >> 1;
+	wake_up_process(cmdinfo->task);
+}
+
+typedef void (*completion_fn)(struct nvme_queue *, void *,
+						struct nvme_completion *);
+
+static irqreturn_t nvme_process_cq(struct nvme_queue *nvmeq)
+{
+	u16 head, cycle;
+
+	static const completion_fn completions[4] = {
+		[sync_completion_id] = sync_completion,
+		[bio_completion_id]  = bio_completion,
+	};
+
+	head = nvmeq->cq_head;
+	cycle = nvmeq->cq_cycle;
+
+	for (;;) {
+		unsigned long data;
+		void *ptr;
+		unsigned char handler;
+		struct nvme_completion cqe = nvmeq->cqes[head];
+		if ((le16_to_cpu(cqe.status) & 1) != cycle)
+			break;
+		nvmeq->sq_head = le16_to_cpu(cqe.sq_head);
+		if (++head == nvmeq->q_depth) {
+			head = 0;
+			cycle = !cycle;
+		}
+
+		data = free_cmdid(nvmeq, cqe.command_id);
+		handler = data & 3;
+		ptr = (void *)(data & ~3UL);
+		completions[handler](nvmeq, ptr, &cqe);
+	}
+
+	/* If the controller ignores the cq head doorbell and continuously
+	 * writes to the queue, it is theoretically possible to wrap around
+	 * the queue twice and mistakenly return IRQ_NONE.  Linux only
+	 * requires that 0.1% of your interrupts are handled, so this isn't
+	 * a big problem.
+	 */
+	if (head == nvmeq->cq_head && cycle == nvmeq->cq_cycle)
+		return IRQ_NONE;
+
+	writel(head, nvmeq->q_db + 1);
+	nvmeq->cq_head = head;
+	nvmeq->cq_cycle = cycle;
+
+	return IRQ_HANDLED;
+}
+
+static irqreturn_t nvme_irq(int irq, void *data)
+{
+	return nvme_process_cq(data);
+}
+
+/*
+ * 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 *q, struct nvme_command *cmd,
+								u32 *result)
+{
+	int cmdid;
+	struct sync_cmd_info cmdinfo;
+
+	cmdinfo.task = current;
+	cmdinfo.status = -EINTR;
+
+	cmdid = alloc_cmdid_killable(q, &cmdinfo, sync_completion_id);
+	if (cmdid < 0)
+		return cmdid;
+	cmd->common.command_id = cmdid;
+
+	set_current_state(TASK_UNINTERRUPTIBLE);
+	nvme_submit_cmd(q, cmd);
+	schedule();
+
+	if (result)
+		*result = cmdinfo.result;
+
+	return cmdinfo.status;
+}
+
+static 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);
+}
+
+static int adapter_delete_queue(struct nvme_dev *dev, u8 opcode, u16 id)
+{
+	int status;
+	struct nvme_command c;
+
+	memset(&c, 0, sizeof(c));
+	c.delete_queue.opcode = opcode;
+	c.delete_queue.qid = cpu_to_le16(id);
+
+	status = nvme_submit_admin_cmd(dev, &c, NULL);
+	if (status)
+		return -EIO;
+	return 0;
+}
+
+static int adapter_alloc_cq(struct nvme_dev *dev, u16 qid,
+						struct nvme_queue *nvmeq)
+{
+	int status;
+	struct nvme_command c;
+	int flags = NVME_QUEUE_PHYS_CONTIG | NVME_CQ_IRQ_ENABLED;
+
+	memset(&c, 0, sizeof(c));
+	c.create_cq.opcode = nvme_admin_create_cq;
+	c.create_cq.prp1 = cpu_to_le64(nvmeq->cq_dma_addr);
+	c.create_cq.cqid = cpu_to_le16(qid);
+	c.create_cq.qsize = cpu_to_le16(nvmeq->q_depth - 1);
+	c.create_cq.cq_flags = cpu_to_le16(flags);
+	c.create_cq.irq_vector = cpu_to_le16(nvmeq->cq_vector);
+
+	status = nvme_submit_admin_cmd(dev, &c, NULL);
+	if (status)
+		return -EIO;
+	return 0;
+}
+
+static int adapter_alloc_sq(struct nvme_dev *dev, u16 qid,
+						struct nvme_queue *nvmeq)
+{
+	int status;
+	struct nvme_command c;
+	int flags = NVME_QUEUE_PHYS_CONTIG | NVME_SQ_PRIO_MEDIUM;
+
+	memset(&c, 0, sizeof(c));
+	c.create_sq.opcode = nvme_admin_create_sq;
+	c.create_sq.prp1 = cpu_to_le64(nvmeq->sq_dma_addr);
+	c.create_sq.sqid = cpu_to_le16(qid);
+	c.create_sq.qsize = cpu_to_le16(nvmeq->q_depth - 1);
+	c.create_sq.sq_flags = cpu_to_le16(flags);
+	c.create_sq.cqid = cpu_to_le16(qid);
+
+	status = nvme_submit_admin_cmd(dev, &c, NULL);
+	if (status)
+		return -EIO;
+	return 0;
+}
+
+static int adapter_delete_cq(struct nvme_dev *dev, u16 cqid)
+{
+	return adapter_delete_queue(dev, nvme_admin_delete_cq, cqid);
+}
+
+static int adapter_delete_sq(struct nvme_dev *dev, u16 sqid)
+{
+	return adapter_delete_queue(dev, nvme_admin_delete_sq, sqid);
+}
+
+static void nvme_free_queue(struct nvme_dev *dev, int qid)
+{
+	struct nvme_queue *nvmeq = dev->queues[qid];
+
+	free_irq(dev->entry[nvmeq->cq_vector].vector, nvmeq);
+
+	/* Don't tell the adapter to delete the admin queue */
+	if (qid) {
+		adapter_delete_sq(dev, qid);
+		adapter_delete_cq(dev, qid);
+	}
+
+	dma_free_coherent(nvmeq->q_dmadev, CQ_SIZE(nvmeq->q_depth),
+				(void *)nvmeq->cqes, nvmeq->cq_dma_addr);
+	dma_free_coherent(nvmeq->q_dmadev, SQ_SIZE(nvmeq->q_depth),
+					nvmeq->sq_cmds, nvmeq->sq_dma_addr);
+	kfree(nvmeq);
+}
+
+static struct nvme_queue *nvme_alloc_queue(struct nvme_dev *dev, int qid,
+							int depth, int vector)
+{
+	struct device *dmadev = &dev->pci_dev->dev;
+	unsigned extra = (depth + BITS_TO_LONGS(depth)) * sizeof(long);
+	struct nvme_queue *nvmeq = kzalloc(sizeof(*nvmeq) + extra, GFP_KERNEL);
+	if (!nvmeq)
+		return NULL;
+
+	nvmeq->cqes = dma_alloc_coherent(dmadev, CQ_SIZE(depth),
+					&nvmeq->cq_dma_addr, GFP_KERNEL);
+	if (!nvmeq->cqes)
+		goto free_nvmeq;
+	memset((void *)nvmeq->cqes, 0, CQ_SIZE(depth));
+
+	nvmeq->sq_cmds = dma_alloc_coherent(dmadev, SQ_SIZE(depth),
+					&nvmeq->sq_dma_addr, GFP_KERNEL);
+	if (!nvmeq->sq_cmds)
+		goto free_cqdma;
+
+	nvmeq->q_dmadev = dmadev;
+	spin_lock_init(&nvmeq->q_lock);
+	nvmeq->cq_head = 0;
+	nvmeq->cq_cycle = 1;
+	init_waitqueue_head(&nvmeq->sq_full);
+	bio_list_init(&nvmeq->sq_cong);
+	nvmeq->q_db = &dev->dbs[qid * 2];
+	nvmeq->q_depth = depth;
+	nvmeq->cq_vector = vector;
+
+	return nvmeq;
+
+ free_cqdma:
+	dma_free_coherent(dmadev, CQ_SIZE(nvmeq->q_depth), (void *)nvmeq->cqes,
+							nvmeq->cq_dma_addr);
+ free_nvmeq:
+	kfree(nvmeq);
+	return NULL;
+}
+
+static __devinit struct nvme_queue *nvme_create_queue(struct nvme_dev *dev,
+					int qid, int cq_size, int vector)
+{
+	int result;
+	struct nvme_queue *nvmeq = nvme_alloc_queue(dev, qid, cq_size, vector);
+
+	result = adapter_alloc_cq(dev, qid, nvmeq);
+	if (result < 0)
+		goto free_nvmeq;
+
+	result = adapter_alloc_sq(dev, qid, nvmeq);
+	if (result < 0)
+		goto release_cq;
+
+	result = request_irq(dev->entry[vector].vector, nvme_irq,
+				IRQF_DISABLED | IRQF_SHARED, "nvme", nvmeq);
+	if (result < 0)
+		goto release_sq;
+
+	return nvmeq;
+
+ release_sq:
+	adapter_delete_sq(dev, qid);
+ release_cq:
+	adapter_delete_cq(dev, qid);
+ free_nvmeq:
+	dma_free_coherent(nvmeq->q_dmadev, CQ_SIZE(nvmeq->q_depth),
+				(void *)nvmeq->cqes, nvmeq->cq_dma_addr);
+	dma_free_coherent(nvmeq->q_dmadev, SQ_SIZE(nvmeq->q_depth),
+					nvmeq->sq_cmds, nvmeq->sq_dma_addr);
+	kfree(nvmeq);
+	return NULL;
+}
+
+static int __devinit nvme_configure_admin_queue(struct nvme_dev *dev)
+{
+	int result;
+	u32 aqa;
+	struct nvme_queue *nvmeq;
+
+	dev->dbs = ((void __iomem *)dev->bar) + 4096;
+
+	nvmeq = nvme_alloc_queue(dev, 0, 64, 0);
+
+	aqa = nvmeq->q_depth - 1;
+	aqa |= aqa << 16;
+
+	dev->ctrl_config = NVME_CC_ENABLE | NVME_CC_CSS_NVM;
+	dev->ctrl_config |= (PAGE_SHIFT - 12) << NVME_CC_MPS_SHIFT;
+	dev->ctrl_config |= NVME_CC_ARB_RR | NVME_CC_SHN_NONE;
+
+	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);
+
+	while (!(readl(&dev->bar->csts) & NVME_CSTS_RDY)) {
+		msleep(100);
+		if (fatal_signal_pending(current))
+			return -EINTR;
+	}
+
+	result = request_irq(dev->entry[0].vector, nvme_irq,
+			IRQF_DISABLED | IRQF_SHARED, "nvme admin", nvmeq);
+	dev->queues[0] = nvmeq;
+	return result;
+}
+
+static int nvme_identify(struct nvme_ns *ns, void __user *addr, int cns)
+{
+	struct nvme_dev *dev = ns->dev;
+	int status;
+	struct nvme_command c;
+	void *page;
+	dma_addr_t dma_addr;
+
+	page = dma_alloc_coherent(&dev->pci_dev->dev, 4096, &dma_addr,
+								GFP_KERNEL);
+
+	memset(&c, 0, sizeof(c));
+	c.identify.opcode = nvme_admin_identify;
+	c.identify.nsid = cns ? 0 : cpu_to_le32(ns->ns_id);
+	c.identify.prp1 = cpu_to_le64(dma_addr);
+	c.identify.cns = cpu_to_le32(cns);
+
+	status = nvme_submit_admin_cmd(dev, &c, NULL);
+
+	if (status)
+		status = -EIO;
+	else if (copy_to_user(addr, page, 4096))
+		status = -EFAULT;
+
+	dma_free_coherent(&dev->pci_dev->dev, 4096, page, dma_addr);
+
+	return status;
+}
+
+static int nvme_get_range_type(struct nvme_ns *ns, void __user *addr)
+{
+	struct nvme_dev *dev = ns->dev;
+	int status;
+	struct nvme_command c;
+	void *page;
+	dma_addr_t dma_addr;
+
+	page = dma_alloc_coherent(&dev->pci_dev->dev, 4096, &dma_addr,
+								GFP_KERNEL);
+
+	memset(&c, 0, sizeof(c));
+	c.features.opcode = nvme_admin_get_features;
+	c.features.nsid = cpu_to_le32(ns->ns_id);
+	c.features.prp1 = cpu_to_le64(dma_addr);
+	c.features.fid = cpu_to_le32(NVME_FEAT_LBA_RANGE);
+
+	status = nvme_submit_admin_cmd(dev, &c, NULL);
+
+	/* XXX: Assuming first range for now */
+	if (status)
+		status = -EIO;
+	else if (copy_to_user(addr, page, 64))
+		status = -EFAULT;
+
+	dma_free_coherent(&dev->pci_dev->dev, 4096, page, dma_addr);
+
+	return status;
+}
+
+static int nvme_ioctl(struct block_device *bdev, fmode_t mode, unsigned int cmd,
+							unsigned long arg)
+{
+	struct nvme_ns *ns = bdev->bd_disk->private_data;
+
+	switch (cmd) {
+	case NVME_IOCTL_IDENTIFY_NS:
+		return nvme_identify(ns, (void __user *)arg, 0);
+	case NVME_IOCTL_IDENTIFY_CTRL:
+		return nvme_identify(ns, (void __user *)arg, 1);
+	case NVME_IOCTL_GET_RANGE_TYPE:
+		return nvme_get_range_type(ns, (void __user *)arg);
+	default:
+		return -ENOTTY;
+	}
+}
+
+static const struct block_device_operations nvme_fops = {
+	.owner		= THIS_MODULE,
+	.ioctl		= nvme_ioctl,
+};
+
+static struct nvme_ns *nvme_alloc_ns(struct nvme_dev *dev, int index,
+			struct nvme_id_ns *id, struct nvme_lba_range_type *rt)
+{
+	struct nvme_ns *ns;
+	struct gendisk *disk;
+	int lbaf;
+
+	if (rt->attributes & NVME_LBART_ATTRIB_HIDE)
+		return NULL;
+
+	ns = kzalloc(sizeof(*ns), GFP_KERNEL);
+	if (!ns)
+		return NULL;
+	ns->queue = blk_alloc_queue(GFP_KERNEL);
+	if (!ns->queue)
+		goto out_free_ns;
+	ns->queue->queue_flags = QUEUE_FLAG_DEFAULT | QUEUE_FLAG_NOMERGES |
+				QUEUE_FLAG_NONROT | QUEUE_FLAG_DISCARD;
+	blk_queue_make_request(ns->queue, nvme_make_request);
+	ns->dev = dev;
+	ns->queue->queuedata = ns;
+
+	disk = alloc_disk(NVME_MINORS);
+	if (!disk)
+		goto out_free_queue;
+	ns->ns_id = index;
+	ns->disk = disk;
+	lbaf = id->flbas & 0xf;
+	ns->lba_shift = id->lbaf[lbaf].ds;
+
+	disk->major = nvme_major;
+	disk->minors = NVME_MINORS;
+	disk->first_minor = NVME_MINORS * index;
+	disk->fops = &nvme_fops;
+	disk->private_data = ns;
+	disk->queue = ns->queue;
+	sprintf(disk->disk_name, "nvme%dn%d", dev->instance, index);
+	set_capacity(disk, le64_to_cpup(&id->nsze) << (ns->lba_shift - 9));
+
+	return ns;
+
+ out_free_queue:
+	blk_cleanup_queue(ns->queue);
+ out_free_ns:
+	kfree(ns);
+	return NULL;
+}
+
+static void nvme_ns_free(struct nvme_ns *ns)
+{
+	put_disk(ns->disk);
+	blk_cleanup_queue(ns->queue);
+	kfree(ns);
+}
+
+static int set_queue_count(struct nvme_dev *dev, int sq_count, int cq_count)
+{
+	int status;
+	u32 result;
+	struct nvme_command c;
+	u32 q_count = (sq_count - 1) | ((cq_count - 1) << 16);
+
+	memset(&c, 0, sizeof(c));
+	c.features.opcode = nvme_admin_get_features;
+	c.features.fid = cpu_to_le32(NVME_FEAT_NUM_QUEUES);
+	c.features.dword11 = cpu_to_le32(q_count);
+
+	status = nvme_submit_admin_cmd(dev, &c, &result);
+	if (status)
+		return -EIO;
+	return min(result & 0xffff, result >> 16) + 1;
+}
+
+/* XXX: Create per-CPU queues */
+static int __devinit nvme_setup_io_queues(struct nvme_dev *dev)
+{
+	int this_cpu;
+
+	set_queue_count(dev, 1, 1);
+
+	this_cpu = get_cpu();
+	dev->queues[1] = nvme_create_queue(dev, 1, NVME_Q_DEPTH, this_cpu);
+	put_cpu();
+	if (!dev->queues[1])
+		return -ENOMEM;
+	dev->queue_count++;
+
+	return 0;
+}
+
+static void nvme_free_queues(struct nvme_dev *dev)
+{
+	int i;
+
+	for (i = dev->queue_count - 1; i >= 0; i--)
+		nvme_free_queue(dev, i);
+}
+
+static int __devinit nvme_dev_add(struct nvme_dev *dev)
+{
+	int res, nn, i;
+	struct nvme_ns *ns, *next;
+	void *id;
+	dma_addr_t dma_addr;
+	struct nvme_command cid, crt;
+
+	res = nvme_setup_io_queues(dev);
+	if (res)
+		return res;
+
+	/* XXX: Switch to a SG list once prp2 works */
+	id = dma_alloc_coherent(&dev->pci_dev->dev, 8192, &dma_addr,
+								GFP_KERNEL);
+
+	memset(&cid, 0, sizeof(cid));
+	cid.identify.opcode = nvme_admin_identify;
+	cid.identify.nsid = 0;
+	cid.identify.prp1 = cpu_to_le64(dma_addr);
+	cid.identify.cns = cpu_to_le32(1);
+
+	res = nvme_submit_admin_cmd(dev, &cid, NULL);
+	if (res) {
+		res = -EIO;
+		goto out_free;
+	}
+
+	nn = le32_to_cpup(&((struct nvme_id_ctrl *)id)->nn);
+
+	cid.identify.cns = 0;
+	memset(&crt, 0, sizeof(crt));
+	crt.features.opcode = nvme_admin_get_features;
+	crt.features.prp1 = cpu_to_le64(dma_addr + 4096);
+	crt.features.fid = cpu_to_le32(NVME_FEAT_LBA_RANGE);
+
+	for (i = 0; i < nn; i++) {
+		cid.identify.nsid = cpu_to_le32(i);
+		res = nvme_submit_admin_cmd(dev, &cid, NULL);
+		if (res)
+			continue;
+
+		if (((struct nvme_id_ns *)id)->ncap == 0)
+			continue;
+
+		crt.features.nsid = cpu_to_le32(i);
+		res = nvme_submit_admin_cmd(dev, &crt, NULL);
+		if (res)
+			continue;
+
+		ns = nvme_alloc_ns(dev, i, id, id + 4096);
+		if (ns)
+			list_add_tail(&ns->list, &dev->namespaces);
+	}
+	list_for_each_entry(ns, &dev->namespaces, list)
+		add_disk(ns->disk);
+
+	dma_free_coherent(&dev->pci_dev->dev, 4096, id, dma_addr);
+	return 0;
+
+ out_free:
+	list_for_each_entry_safe(ns, next, &dev->namespaces, list) {
+		list_del(&ns->list);
+		nvme_ns_free(ns);
+	}
+
+	dma_free_coherent(&dev->pci_dev->dev, 4096, id, dma_addr);
+	return res;
+}
+
+static int nvme_dev_remove(struct nvme_dev *dev)
+{
+	struct nvme_ns *ns, *next;
+
+	/* TODO: wait all I/O finished or cancel them */
+
+	list_for_each_entry_safe(ns, next, &dev->namespaces, list) {
+		list_del(&ns->list);
+		del_gendisk(ns->disk);
+		nvme_ns_free(ns);
+	}
+
+	nvme_free_queues(dev);
+
+	return 0;
+}
+
+/* XXX: Use an ida or something to let remove / add work correctly */
+static void nvme_set_instance(struct nvme_dev *dev)
+{
+	static int instance;
+	dev->instance = instance++;
+}
+
+static void nvme_release_instance(struct nvme_dev *dev)
+{
+}
+
+static int __devinit nvme_probe(struct pci_dev *pdev,
+						const struct pci_device_id *id)
+{
+	int result = -ENOMEM;
+	struct nvme_dev *dev;
+
+	dev = kzalloc(sizeof(*dev), GFP_KERNEL);
+	if (!dev)
+		return -ENOMEM;
+	dev->entry = kcalloc(num_possible_cpus(), sizeof(*dev->entry),
+								GFP_KERNEL);
+	if (!dev->entry)
+		goto free;
+	dev->queues = kcalloc(2, sizeof(void *), GFP_KERNEL);
+	if (!dev->queues)
+		goto free;
+
+	INIT_LIST_HEAD(&dev->namespaces);
+	dev->pci_dev = pdev;
+	pci_set_drvdata(pdev, dev);
+	dma_set_mask(&dev->pci_dev->dev, DMA_BIT_MASK(64));
+	nvme_set_instance(dev);
+
+	dev->bar = ioremap(pci_resource_start(pdev, 0), 8192);
+	if (!dev->bar) {
+		result = -ENOMEM;
+		goto disable;
+	}
+
+	result = nvme_configure_admin_queue(dev);
+	if (result)
+		goto unmap;
+	dev->queue_count++;
+
+	result = nvme_dev_add(dev);
+	if (result)
+		goto delete;
+	return 0;
+
+ delete:
+	nvme_free_queues(dev);
+ unmap:
+	iounmap(dev->bar);
+ disable:
+	pci_disable_msix(pdev);
+	nvme_release_instance(dev);
+ free:
+	kfree(dev->queues);
+	kfree(dev->entry);
+	kfree(dev);
+	return result;
+}
+
+static void __devexit nvme_remove(struct pci_dev *pdev)
+{
+	struct nvme_dev *dev = pci_get_drvdata(pdev);
+	nvme_dev_remove(dev);
+	pci_disable_msix(pdev);
+	iounmap(dev->bar);
+	nvme_release_instance(dev);
+	kfree(dev->queues);
+	kfree(dev->entry);
+	kfree(dev);
+}
+
+/* These functions are yet to be implemented */
+#define nvme_error_detected NULL
+#define nvme_dump_registers NULL
+#define nvme_link_reset NULL
+#define nvme_slot_reset NULL
+#define nvme_error_resume NULL
+#define nvme_suspend NULL
+#define nvme_resume NULL
+
+static struct pci_error_handlers nvme_err_handler = {
+	.error_detected	= nvme_error_detected,
+	.mmio_enabled	= nvme_dump_registers,
+	.link_reset	= nvme_link_reset,
+	.slot_reset	= nvme_slot_reset,
+	.resume		= nvme_error_resume,
+};
+
+/* Move to pci_ids.h later */
+#define PCI_CLASS_STORAGE_EXPRESS	0x010802
+
+static DEFINE_PCI_DEVICE_TABLE(nvme_id_table) = {
+	{ PCI_DEVICE_CLASS(PCI_CLASS_STORAGE_EXPRESS, 0xffffff) },
+	{ 0, }
+};
+MODULE_DEVICE_TABLE(pci, nvme_id_table);
+
+static struct pci_driver nvme_driver = {
+	.name		= "nvme",
+	.id_table	= nvme_id_table,
+	.probe		= nvme_probe,
+	.remove		= __devexit_p(nvme_remove),
+	.suspend	= nvme_suspend,
+	.resume		= nvme_resume,
+	.err_handler	= &nvme_err_handler,
+};
+
+static int __init nvme_init(void)
+{
+	int result;
+
+	nvme_major = register_blkdev(nvme_major, "nvme");
+	if (nvme_major <= 0)
+		return -EBUSY;
+
+	result = pci_register_driver(&nvme_driver);
+	if (!result)
+		return 0;
+
+	unregister_blkdev(nvme_major, "nvme");
+	return result;
+}
+
+static void __exit nvme_exit(void)
+{
+	pci_unregister_driver(&nvme_driver);
+	unregister_blkdev(nvme_major, "nvme");
+}
+
+MODULE_AUTHOR("Matthew Wilcox <willy@linux.intel.com>");
+MODULE_LICENSE("GPL");
+MODULE_VERSION("0.1");
+module_init(nvme_init);
+module_exit(nvme_exit);
+/*
+ * Definitions for the NVM Express interface
+ * Copyright (c) 2011, Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify it
+ * under the terms and conditions of the GNU General Public License,
+ * version 2, as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope it will be useful, but WITHOUT
+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
+ * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License for
+ * more details.
+ *
+ * You should have received a copy of the GNU General Public License along with
+ * this program; if not, write to the Free Software Foundation, Inc., 
+ * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
+ */
+
+#ifndef _LINUX_NVME_H
+#define _LINUX_NVME_H
+
+#include <linux/types.h>
+
+struct nvme_bar {
+	__u64			cap;	/* Controller Capabilities */
+	__u32			vs;	/* Version */
+	__u32			ims;	/* Interrupt Mask Set */
+	__u32			imc;	/* Interrupt Mask Clear */
+	__u32			cc;	/* Controller Configuration */
+	__u32			csts;	/* Controller Status */
+	__u32			aqa;	/* Admin Queue Attributes */
+	__u64			asq;	/* Admin SQ Base Address */
+	__u64			acq;	/* Admin CQ Base Address */
+};
+
+enum {
+	NVME_CC_ENABLE		= 1 << 0,
+	NVME_CC_CSS_NVM		= 0 << 4,
+	NVME_CC_MPS_SHIFT	= 7,
+	NVME_CC_ARB_RR		= 0 << 11,
+	NVME_CC_ARB_WRRU	= 1 << 11,
+	NVME_CC_ARB_VS		= 3 << 11,
+	NVME_CC_SHN_NONE	= 0 << 13,
+	NVME_CC_SHN_NORMAL	= 1 << 13,
+	NVME_CC_SHN_ABRUPT	= 2 << 13,
+	NVME_CSTS_RDY		= 1 << 0,
+	NVME_CSTS_CFS		= 1 << 1,
+	NVME_CSTS_SHST_NORMAL	= 0 << 2,
+	NVME_CSTS_SHST_OCCUR	= 1 << 2,
+	NVME_CSTS_SHST_CMPLT	= 2 << 2,
+};
+
+#define NVME_VS(major, minor)	(major << 16 | minor)
+
+struct nvme_id_ctrl {
+	__le16			vid;
+	__le16			ssvid;
+	char			sn[20];
+	char			mn[40];
+	char			fr[8];
+	__le32			nn;
+	__u8			rab;
+	__u8			rsvd77[178];
+	__le16			oacs;
+	__u8			acl;
+	__u8			aerl;
+	__u8			frmw;
+	__u8			lpa;
+	__u8			elpe;
+	__u8			npss;
+	__u8			rsvd264[248];
+	__le64			psd[32];
+	__le16			oncs;
+	__le16			fuses;
+	__u8			fna;
+	__u8			vwc;
+	__le16			awun;
+	__le16			awupf;
+	__u8			rsvd778[246];
+	__u8			cmdset[2048];
+	__u8			vs[1024];
+};
+
+struct nvme_lbaf {
+	__le16			ms;
+	__u8			ds;
+	__u8			rp;
+};
+
+struct nvme_id_ns {
+	__le64			nsze;
+	__le64			ncap;
+	__le64			nuse;
+	__u8			nsfeat;
+	__u8			nlbaf;
+	__u8			flbas;
+	__u8			mc;
+	__u8			dpc;
+	__u8			dps;
+	__u8			rsvd30[98];
+	struct nvme_lbaf	lbaf[16];
+	__u8			rsvd192[192];
+	__u8			vs[3712];
+};
+
+enum {
+	NVME_NS_FEAT_THIN	= 1 << 0,
+	NVME_LBAF_RP_BEST	= 0,
+	NVME_LBAF_RP_BETTER	= 1,
+	NVME_LBAF_RP_GOOD	= 2,
+	NVME_LBAF_RP_DEGRADED	= 3,
+};
+
+struct nvme_lba_range_type {
+	__u8			type;
+	__u8			attributes;
+	__u8			rsvd2[14];
+	__u64			slba;
+	__u64			nlb;
+	__u8			guid[16];
+	__u8			rsvd48[16];
+};
+
+enum {
+	NVME_LBART_TYPE_FS	= 0x01,
+	NVME_LBART_TYPE_RAID	= 0x02,
+	NVME_LBART_TYPE_CACHE	= 0x03,
+	NVME_LBART_TYPE_SWAP	= 0x04,
+
+	NVME_LBART_ATTRIB_TEMP	= 1 << 0,
+	NVME_LBART_ATTRIB_HIDE	= 1 << 1,
+};
+
+/* I/O commands */
+
+enum nvme_opcode {
+	nvme_cmd_flush		= 0x00,
+	nvme_cmd_write		= 0x01,
+	nvme_cmd_read		= 0x02,
+	nvme_cmd_write_uncor	= 0x04,
+	nvme_cmd_compare	= 0x05,
+	nvme_cmd_dsm		= 0x09,
+};
+
+struct nvme_rw_command {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u64			rsvd2;
+	__le64			metadata;
+	__le64			prp1;
+	__le64			prp2;
+	__le64			slba;
+	__le16			length;
+	__le16			control;
+	__le32			dsmgmt;
+	__le32			reftag;
+	__le16			apptag;
+	__le16			appmask;
+};
+
+enum {
+	NVME_RW_LR			= 1 << 15,
+	NVME_RW_FUA			= 1 << 14,
+	NVME_RW_DSM_FREQ_UNSPEC		= 0,
+	NVME_RW_DSM_FREQ_TYPICAL	= 1,
+	NVME_RW_DSM_FREQ_RARE		= 2,
+	NVME_RW_DSM_FREQ_READS		= 3,
+	NVME_RW_DSM_FREQ_WRITES		= 4,
+	NVME_RW_DSM_FREQ_RW		= 5,
+	NVME_RW_DSM_FREQ_ONCE		= 6,
+	NVME_RW_DSM_FREQ_PREFETCH	= 7,
+	NVME_RW_DSM_FREQ_TEMP		= 8,
+	NVME_RW_DSM_LATENCY_NONE	= 0 << 4,
+	NVME_RW_DSM_LATENCY_IDLE	= 1 << 4,
+	NVME_RW_DSM_LATENCY_NORM	= 2 << 4,
+	NVME_RW_DSM_LATENCY_LOW		= 3 << 4,
+	NVME_RW_DSM_SEQ_REQ		= 1 << 6,
+	NVME_RW_DSM_COMPRESSED		= 1 << 7,
+};
+
+/* Admin commands */
+
+enum nvme_admin_opcode {
+	nvme_admin_delete_sq		= 0x00,
+	nvme_admin_create_sq		= 0x01,
+	nvme_admin_get_features		= 0x02,
+	nvme_admin_delete_cq		= 0x04,
+	nvme_admin_create_cq		= 0x05,
+	nvme_admin_identify		= 0x06,
+	nvme_admin_abort_cmd		= 0x08,
+	nvme_admin_set_features		= 0x09,
+	nvme_admin_get_log_page		= 0x0a,
+	nvme_admin_async_event		= 0x0c,
+	nvme_admin_download_fw		= 0x0d,
+	nvme_admin_security_recv	= 0x0e,
+	nvme_admin_format_nvm		= 0x10,
+	nvme_admin_security_send	= 0x11,
+	nvme_admin_activate_fw		= 0x14,
+};
+
+enum {
+	NVME_QUEUE_PHYS_CONTIG	= (1 << 0),
+	NVME_CQ_IRQ_ENABLED	= (1 << 1),
+	NVME_SQ_PRIO_URGENT	= (0 << 1),
+	NVME_SQ_PRIO_HIGH	= (1 << 1),
+	NVME_SQ_PRIO_MEDIUM	= (2 << 1),
+	NVME_SQ_PRIO_LOW	= (3 << 1),
+	NVME_FEAT_ARBITRATION	= 0x01,
+	NVME_FEAT_POWER_MGMT	= 0x02,
+	NVME_FEAT_LBA_RANGE	= 0x03,
+	NVME_FEAT_TEMP_THRESH	= 0x04,
+	NVME_FEAT_ERR_RECOVERY	= 0x05,
+	NVME_FEAT_VOLATILE_WC	= 0x06,
+	NVME_FEAT_NUM_QUEUES	= 0x07,
+	NVME_FEAT_IRQ_COALESCE	= 0x08,
+	NVME_FEAT_IRQ_CONFIG	= 0x09,
+	NVME_FEAT_WRITE_ATOMIC	= 0x0a,
+	NVME_FEAT_ASYNC_EVENT	= 0x0b,
+	NVME_FEAT_SW_PROGRESS	= 0x0c,
+};
+
+struct nvme_identify {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u64			rsvd2[2];
+	__le64			prp1;
+	__le64			prp2;
+	__le32			cns;
+	__u32			rsvd11[5];
+};
+
+struct nvme_features {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u64			rsvd2[2];
+	__le64			prp1;
+	__le64			prp2;
+	__le32			fid;
+	__le32			dword11;
+	__u32			rsvd12[4];
+};
+
+struct nvme_create_cq {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			rsvd1[5];
+	__le64			prp1;
+	__u64			rsvd8;
+	__le16			cqid;
+	__le16			qsize;
+	__le16			cq_flags;
+	__le16			irq_vector;
+	__u32			rsvd12[4];
+};
+
+struct nvme_create_sq {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			rsvd1[5];
+	__le64			prp1;
+	__u64			rsvd8;
+	__le16			sqid;
+	__le16			qsize;
+	__le16			sq_flags;
+	__le16			cqid;
+	__le32			rsvd12[4];
+};
+
+struct nvme_delete_queue {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__u32			rsvd1[9];
+	__le16			qid;
+	__le16			rsvd10;
+	__le32			rsvd11[5];
+};
+
+struct nvme_common_command {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u32			rsvd2[14];
+};
+
+struct nvme_command {
+	union {
+		struct nvme_common_command common;
+		struct nvme_rw_command rw;
+		struct nvme_identify identify;
+		struct nvme_features features;
+		struct nvme_create_cq create_cq;
+		struct nvme_create_sq create_sq;
+		struct nvme_delete_queue delete_queue;
+	};
+};
+
+/* XXX: Sync with spec */
+enum {
+	NVME_SC_SUCCESS			= 0x0,
+	NVME_SC_INVALID_OPCODE		= 0x1,
+	NVME_SC_INVALID_FIELD		= 0x2,
+	NVME_SC_CMDID_CONFLICT		= 0x3,
+	NVME_SC_DATA_XFER_ERROR		= 0x4,
+	NVME_SC_POWER_LOSS		= 0x5,
+	NVME_SC_INTERNAL		= 0x6,
+	NVME_SC_ABORT_REQ		= 0x7,
+	NVME_SC_ABORT_QUEUE		= 0x8,
+	NVME_SC_FUSED_FAIL		= 0x9,
+	NVME_SC_FUSED_MISSING		= 0xa,
+	NVME_SC_LBA_RANGE		= 0x80,
+	NVME_SC_CAP_EXCEEDED		= 0x81,
+	NVME_SC_NS_NOT_READY		= 0x82,
+	NVME_SC_CQ_INVALID		= 0x100,
+	NVME_SC_QID_INVALID		= 0x101,
+	NVME_SC_QUEUE_SIZE		= 0x102,
+	NVME_SC_WRITE_FAULT		= 0x280,
+	NVME_SC_READ_ERROR		= 0x281,
+};
+
+struct nvme_completion {
+	__le32	result;		/* Used by admin commands to return data */
+	__le32	rsvd;
+	__le16	sq_head;	/* how much of this queue may be reclaimed */
+	__le16	sq_id;		/* submission queue that generated this entry */
+	__u16	command_id;	/* of the command which completed */
+	__le16	status;		/* did the command fail, and if so, why? */
+};
+
+#define NVME_IOCTL_IDENTIFY_NS	_IOW('N', 0x40, struct nvme_id_ns)
+#define NVME_IOCTL_IDENTIFY_CTRL _IOW('N', 0x41, struct nvme_id_ctrl)
+#define NVME_IOCTL_GET_RANGE_TYPE _IOW('N', 0x42, struct nvme_lba_range_type)
+
+#endif /* _LINUX_NVME_H */
...	...	@@ -149,6 +149,7 @@
149	149	'M' 01-03 drivers/scsi/megaraid/megaraid_sas.h
150	150	'M' 00-0F drivers/video/fsl-diu-fb.h conflict!
151	151	'N' 00-1F drivers/usb/scanner.h
	152	+'N' 40-7F drivers/block/nvme.c
152	153	'O' 00-06 mtd/ubi-user.h UBI
153	154	'P' all linux/soundcard.h conflict!
154	155	'P' 60-6F sound/sscape_ioctl.h conflict!
...	...	@@ -315,6 +315,17 @@
315	315
316	316	If unsure, say N.
317	317
	318	+config BLK_DEV_NVME
	319	+ tristate "NVM Express block device"
	320	+ depends on PCI
	321	+ ---help---
	322	+ The NVM Express driver is for solid state drives directly
	323	+ connected to the PCI or PCI Express bus. If you know you
	324	+ don't have one of these, it is safe to answer N.
	325	+
	326	+ To compile this driver as a module, choose M here: the
	327	+ module will be called nvme.
	328	+
318	329	config BLK_DEV_OSD
319	330	tristate "OSD object-as-blkdev support"
320	331	depends on SCSI_OSD_ULD
...	...	@@ -23,6 +23,7 @@
23	23	obj-$(CONFIG_CDROM_PKTCDVD) += pktcdvd.o
24	24	obj-$(CONFIG_MG_DISK) += mg_disk.o
25	25	obj-$(CONFIG_SUNVDC) += sunvdc.o
	26	+obj-$(CONFIG_BLK_DEV_NVME) += nvme.o
26	27	obj-$(CONFIG_BLK_DEV_OSD) += osdblk.o
27	28
28	29	obj-$(CONFIG_BLK_DEV_UMEM) += umem.o
	1	+/*
	2	+ * Definitions for the NVM Express interface
	3	+ * Copyright (c) 2011, Intel Corporation.
	4	+ *
	5	+ * This program is free software; you can redistribute it and/or modify it
	6	+ * under the terms and conditions of the GNU General Public License,
	7	+ * version 2, as published by the Free Software Foundation.
	8	+ *
	9	+ * This program is distributed in the hope it will be useful, but WITHOUT
	10	+ * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
	11	+ * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
	12	+ * more details.
	13	+ *
	14	+ * You should have received a copy of the GNU General Public License along with
	15	+ * this program; if not, write to the Free Software Foundation, Inc.,
	16	+ * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
	17	+ */
	18	+
	19	+#ifndef _LINUX_NVME_H
	20	+#define _LINUX_NVME_H
	21	+
	22	+#include <linux/types.h>
	23	+
	24	+struct nvme_bar {
	25	+ __u64 cap; /* Controller Capabilities */
	26	+ __u32 vs; /* Version */
	27	+ __u32 ims; /* Interrupt Mask Set */
	28	+ __u32 imc; /* Interrupt Mask Clear */
	29	+ __u32 cc; /* Controller Configuration */
	30	+ __u32 csts; /* Controller Status */
	31	+ __u32 aqa; /* Admin Queue Attributes */
	32	+ __u64 asq; /* Admin SQ Base Address */
	33	+ __u64 acq; /* Admin CQ Base Address */
	34	+};
	35	+
	36	+enum {
	37	+ NVME_CC_ENABLE = 1 << 0,
	38	+ NVME_CC_CSS_NVM = 0 << 4,
	39	+ NVME_CC_MPS_SHIFT = 7,
	40	+ NVME_CC_ARB_RR = 0 << 11,
	41	+ NVME_CC_ARB_WRRU = 1 << 11,
	42	+ NVME_CC_ARB_VS = 3 << 11,
	43	+ NVME_CC_SHN_NONE = 0 << 13,
	44	+ NVME_CC_SHN_NORMAL = 1 << 13,
	45	+ NVME_CC_SHN_ABRUPT = 2 << 13,
	46	+ NVME_CSTS_RDY = 1 << 0,
	47	+ NVME_CSTS_CFS = 1 << 1,
	48	+ NVME_CSTS_SHST_NORMAL = 0 << 2,
	49	+ NVME_CSTS_SHST_OCCUR = 1 << 2,
	50	+ NVME_CSTS_SHST_CMPLT = 2 << 2,
	51	+};
	52	+
	53	+#define NVME_VS(major, minor) (major << 16 \| minor)
	54	+
	55	+struct nvme_id_ctrl {
	56	+ __le16 vid;
	57	+ __le16 ssvid;
	58	+ char sn[20];
	59	+ char mn[40];
	60	+ char fr[8];
	61	+ __le32 nn;
	62	+ __u8 rab;
	63	+ __u8 rsvd77[178];
	64	+ __le16 oacs;
	65	+ __u8 acl;
	66	+ __u8 aerl;
	67	+ __u8 frmw;
	68	+ __u8 lpa;
	69	+ __u8 elpe;
	70	+ __u8 npss;
	71	+ __u8 rsvd264[248];
	72	+ __le64 psd[32];
	73	+ __le16 oncs;
	74	+ __le16 fuses;
	75	+ __u8 fna;
	76	+ __u8 vwc;
	77	+ __le16 awun;
	78	+ __le16 awupf;
	79	+ __u8 rsvd778[246];
	80	+ __u8 cmdset[2048];
	81	+ __u8 vs[1024];
	82	+};
	83	+
	84	+struct nvme_lbaf {
	85	+ __le16 ms;
	86	+ __u8 ds;
	87	+ __u8 rp;
	88	+};
	89	+
	90	+struct nvme_id_ns {
	91	+ __le64 nsze;
	92	+ __le64 ncap;
	93	+ __le64 nuse;
	94	+ __u8 nsfeat;
	95	+ __u8 nlbaf;
	96	+ __u8 flbas;
	97	+ __u8 mc;
	98	+ __u8 dpc;
	99	+ __u8 dps;
	100	+ __u8 rsvd30[98];
	101	+ struct nvme_lbaf lbaf[16];
	102	+ __u8 rsvd192[192];
	103	+ __u8 vs[3712];
	104	+};
	105	+
	106	+enum {
	107	+ NVME_NS_FEAT_THIN = 1 << 0,
	108	+ NVME_LBAF_RP_BEST = 0,
	109	+ NVME_LBAF_RP_BETTER = 1,
	110	+ NVME_LBAF_RP_GOOD = 2,
	111	+ NVME_LBAF_RP_DEGRADED = 3,
	112	+};
	113	+
	114	+struct nvme_lba_range_type {
	115	+ __u8 type;
	116	+ __u8 attributes;
	117	+ __u8 rsvd2[14];
	118	+ __u64 slba;
	119	+ __u64 nlb;
	120	+ __u8 guid[16];
	121	+ __u8 rsvd48[16];
	122	+};
	123	+
	124	+enum {
	125	+ NVME_LBART_TYPE_FS = 0x01,
	126	+ NVME_LBART_TYPE_RAID = 0x02,
	127	+ NVME_LBART_TYPE_CACHE = 0x03,
	128	+ NVME_LBART_TYPE_SWAP = 0x04,
	129	+
	130	+ NVME_LBART_ATTRIB_TEMP = 1 << 0,
	131	+ NVME_LBART_ATTRIB_HIDE = 1 << 1,
	132	+};
	133	+
	134	+/* I/O commands */
	135	+
	136	+enum nvme_opcode {
	137	+ nvme_cmd_flush = 0x00,
	138	+ nvme_cmd_write = 0x01,
	139	+ nvme_cmd_read = 0x02,
	140	+ nvme_cmd_write_uncor = 0x04,
	141	+ nvme_cmd_compare = 0x05,
	142	+ nvme_cmd_dsm = 0x09,
	143	+};
	144	+
	145	+struct nvme_rw_command {
	146	+ __u8 opcode;
	147	+ __u8 flags;
	148	+ __u16 command_id;
	149	+ __le32 nsid;
	150	+ __u64 rsvd2;
	151	+ __le64 metadata;
	152	+ __le64 prp1;
	153	+ __le64 prp2;
	154	+ __le64 slba;
	155	+ __le16 length;
	156	+ __le16 control;
	157	+ __le32 dsmgmt;
	158	+ __le32 reftag;
	159	+ __le16 apptag;
	160	+ __le16 appmask;
	161	+};
	162	+
	163	+enum {
	164	+ NVME_RW_LR = 1 << 15,
	165	+ NVME_RW_FUA = 1 << 14,
	166	+ NVME_RW_DSM_FREQ_UNSPEC = 0,
	167	+ NVME_RW_DSM_FREQ_TYPICAL = 1,
	168	+ NVME_RW_DSM_FREQ_RARE = 2,
	169	+ NVME_RW_DSM_FREQ_READS = 3,
	170	+ NVME_RW_DSM_FREQ_WRITES = 4,
	171	+ NVME_RW_DSM_FREQ_RW = 5,
	172	+ NVME_RW_DSM_FREQ_ONCE = 6,
	173	+ NVME_RW_DSM_FREQ_PREFETCH = 7,
	174	+ NVME_RW_DSM_FREQ_TEMP = 8,
	175	+ NVME_RW_DSM_LATENCY_NONE = 0 << 4,
	176	+ NVME_RW_DSM_LATENCY_IDLE = 1 << 4,
	177	+ NVME_RW_DSM_LATENCY_NORM = 2 << 4,
	178	+ NVME_RW_DSM_LATENCY_LOW = 3 << 4,
	179	+ NVME_RW_DSM_SEQ_REQ = 1 << 6,
	180	+ NVME_RW_DSM_COMPRESSED = 1 << 7,
	181	+};
	182	+
	183	+/* Admin commands */
	184	+
	185	+enum nvme_admin_opcode {
	186	+ nvme_admin_delete_sq = 0x00,
	187	+ nvme_admin_create_sq = 0x01,
	188	+ nvme_admin_get_features = 0x02,
	189	+ nvme_admin_delete_cq = 0x04,
	190	+ nvme_admin_create_cq = 0x05,
	191	+ nvme_admin_identify = 0x06,
	192	+ nvme_admin_abort_cmd = 0x08,
	193	+ nvme_admin_set_features = 0x09,
	194	+ nvme_admin_get_log_page = 0x0a,
	195	+ nvme_admin_async_event = 0x0c,
	196	+ nvme_admin_download_fw = 0x0d,
	197	+ nvme_admin_security_recv = 0x0e,
	198	+ nvme_admin_format_nvm = 0x10,
	199	+ nvme_admin_security_send = 0x11,
	200	+ nvme_admin_activate_fw = 0x14,
	201	+};
	202	+
	203	+enum {
	204	+ NVME_QUEUE_PHYS_CONTIG = (1 << 0),
	205	+ NVME_CQ_IRQ_ENABLED = (1 << 1),
	206	+ NVME_SQ_PRIO_URGENT = (0 << 1),
	207	+ NVME_SQ_PRIO_HIGH = (1 << 1),
	208	+ NVME_SQ_PRIO_MEDIUM = (2 << 1),
	209	+ NVME_SQ_PRIO_LOW = (3 << 1),
	210	+ NVME_FEAT_ARBITRATION = 0x01,
	211	+ NVME_FEAT_POWER_MGMT = 0x02,
	212	+ NVME_FEAT_LBA_RANGE = 0x03,
	213	+ NVME_FEAT_TEMP_THRESH = 0x04,
	214	+ NVME_FEAT_ERR_RECOVERY = 0x05,
	215	+ NVME_FEAT_VOLATILE_WC = 0x06,
	216	+ NVME_FEAT_NUM_QUEUES = 0x07,
	217	+ NVME_FEAT_IRQ_COALESCE = 0x08,
	218	+ NVME_FEAT_IRQ_CONFIG = 0x09,
	219	+ NVME_FEAT_WRITE_ATOMIC = 0x0a,
	220	+ NVME_FEAT_ASYNC_EVENT = 0x0b,
	221	+ NVME_FEAT_SW_PROGRESS = 0x0c,
	222	+};
	223	+
	224	+struct nvme_identify {
	225	+ __u8 opcode;
	226	+ __u8 flags;
	227	+ __u16 command_id;
	228	+ __le32 nsid;
	229	+ __u64 rsvd2[2];
	230	+ __le64 prp1;
	231	+ __le64 prp2;
	232	+ __le32 cns;
	233	+ __u32 rsvd11[5];
	234	+};
	235	+
	236	+struct nvme_features {
	237	+ __u8 opcode;
	238	+ __u8 flags;
	239	+ __u16 command_id;
	240	+ __le32 nsid;
	241	+ __u64 rsvd2[2];
	242	+ __le64 prp1;
	243	+ __le64 prp2;
	244	+ __le32 fid;
	245	+ __le32 dword11;
	246	+ __u32 rsvd12[4];
	247	+};
	248	+
	249	+struct nvme_create_cq {
	250	+ __u8 opcode;
	251	+ __u8 flags;
	252	+ __u16 command_id;
	253	+ __le32 rsvd1[5];
	254	+ __le64 prp1;
	255	+ __u64 rsvd8;
	256	+ __le16 cqid;
	257	+ __le16 qsize;
	258	+ __le16 cq_flags;
	259	+ __le16 irq_vector;
	260	+ __u32 rsvd12[4];
	261	+};
	262	+
	263	+struct nvme_create_sq {
	264	+ __u8 opcode;
	265	+ __u8 flags;
	266	+ __u16 command_id;
	267	+ __le32 rsvd1[5];
	268	+ __le64 prp1;
	269	+ __u64 rsvd8;
	270	+ __le16 sqid;
	271	+ __le16 qsize;
	272	+ __le16 sq_flags;
	273	+ __le16 cqid;
	274	+ __le32 rsvd12[4];
	275	+};
	276	+
	277	+struct nvme_delete_queue {
	278	+ __u8 opcode;
	279	+ __u8 flags;
	280	+ __u16 command_id;
	281	+ __u32 rsvd1[9];
	282	+ __le16 qid;
	283	+ __le16 rsvd10;
	284	+ __le32 rsvd11[5];
	285	+};
	286	+
	287	+struct nvme_common_command {
	288	+ __u8 opcode;
	289	+ __u8 flags;
	290	+ __u16 command_id;
	291	+ __le32 nsid;
	292	+ __u32 rsvd2[14];
	293	+};
	294	+
	295	+struct nvme_command {
	296	+ union {
	297	+ struct nvme_common_command common;
	298	+ struct nvme_rw_command rw;
	299	+ struct nvme_identify identify;
	300	+ struct nvme_features features;
	301	+ struct nvme_create_cq create_cq;
	302	+ struct nvme_create_sq create_sq;
	303	+ struct nvme_delete_queue delete_queue;
	304	+ };
	305	+};
	306	+
	307	+/* XXX: Sync with spec */
	308	+enum {
	309	+ NVME_SC_SUCCESS = 0x0,
	310	+ NVME_SC_INVALID_OPCODE = 0x1,
	311	+ NVME_SC_INVALID_FIELD = 0x2,
	312	+ NVME_SC_CMDID_CONFLICT = 0x3,
	313	+ NVME_SC_DATA_XFER_ERROR = 0x4,
	314	+ NVME_SC_POWER_LOSS = 0x5,
	315	+ NVME_SC_INTERNAL = 0x6,
	316	+ NVME_SC_ABORT_REQ = 0x7,
	317	+ NVME_SC_ABORT_QUEUE = 0x8,
	318	+ NVME_SC_FUSED_FAIL = 0x9,
	319	+ NVME_SC_FUSED_MISSING = 0xa,
	320	+ NVME_SC_LBA_RANGE = 0x80,
	321	+ NVME_SC_CAP_EXCEEDED = 0x81,
	322	+ NVME_SC_NS_NOT_READY = 0x82,
	323	+ NVME_SC_CQ_INVALID = 0x100,
	324	+ NVME_SC_QID_INVALID = 0x101,
	325	+ NVME_SC_QUEUE_SIZE = 0x102,
	326	+ NVME_SC_WRITE_FAULT = 0x280,
	327	+ NVME_SC_READ_ERROR = 0x281,
	328	+};
	329	+
	330	+struct nvme_completion {
	331	+ __le32 result; /* Used by admin commands to return data */
	332	+ __le32 rsvd;
	333	+ __le16 sq_head; /* how much of this queue may be reclaimed */
	334	+ __le16 sq_id; /* submission queue that generated this entry */
	335	+ __u16 command_id; /* of the command which completed */
	336	+ __le16 status; /* did the command fail, and if so, why? */
	337	+};
	338	+
	339	+#define NVME_IOCTL_IDENTIFY_NS _IOW('N', 0x40, struct nvme_id_ns)
	340	+#define NVME_IOCTL_IDENTIFY_CTRL _IOW('N', 0x41, struct nvme_id_ctrl)
	341	+#define NVME_IOCTL_GET_RANGE_TYPE _IOW('N', 0x42, struct nvme_lba_range_type)
	342	+
	343	+#endif /* _LINUX_NVME_H */