nvme: LightNVM support

The first generation of Open-Channel SSDs is based on NVMe. The NVMe
driver is extended with support for the LightNVM command set.

Detection is made through PCI IDs. Current supported devices are the
qemu nvme simulator and CNEX Labs Westlake SSD. The qemu nvme enables
support through vendor specific bits in the namespace identification and
the CNEX Labs Westlake SSD implements a LightNVM compatible firmware and
is detected using the same method as qemu.

After detection, vendor specific codes are used to identify the device
and enumerate supported features.
Reviewed-by: Keith Busch <keith.busch@intel.com>
Signed-off-by: Javier González <jg@lightnvm.io>
Signed-off-by: Matias Bjørling <m@bjorling.me>
Signed-off-by: Jens Axboe <axboe@fb.com>

drivers/nvme/host/Makefile

 
 obj-$(CONFIG_BLK_DEV_NVME)     += nvme.o
 
-nvme-y		+= pci.o scsi.o
+nvme-y		+= pci.o scsi.o lightnvm.o

drivers/nvme/host/lightnvm.c

+/*
+ * nvme-lightnvm.c - LightNVM NVMe device
+ *
+ * Copyright (C) 2014-2015 IT University of Copenhagen
+ * Initial release: Matias Bjorling <mb@lightnvm.io>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License version
+ * 2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that 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; see the file COPYING.  If not, write to
+ * the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139,
+ * USA.
+ *
+ */
+
+#include "nvme.h"
+
+#ifdef CONFIG_NVM
+
+#include <linux/nvme.h>
+#include <linux/bitops.h>
+#include <linux/lightnvm.h>
+#include <linux/vmalloc.h>
+
+enum nvme_nvm_admin_opcode {
+	nvme_nvm_admin_identity		= 0xe2,
+	nvme_nvm_admin_get_l2p_tbl	= 0xea,
+	nvme_nvm_admin_get_bb_tbl	= 0xf2,
+	nvme_nvm_admin_set_bb_tbl	= 0xf1,
+};
+
+struct nvme_nvm_hb_rw {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u64			rsvd2;
+	__le64			metadata;
+	__le64			prp1;
+	__le64			prp2;
+	__le64			spba;
+	__le16			length;
+	__le16			control;
+	__le32			dsmgmt;
+	__le64			slba;
+};
+
+struct nvme_nvm_ph_rw {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u64			rsvd2;
+	__le64			metadata;
+	__le64			prp1;
+	__le64			prp2;
+	__le64			spba;
+	__le16			length;
+	__le16			control;
+	__le32			dsmgmt;
+	__le64			resv;
+};
+
+struct nvme_nvm_identity {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u64			rsvd[2];
+	__le64			prp1;
+	__le64			prp2;
+	__le32			chnl_off;
+	__u32			rsvd11[5];
+};
+
+struct nvme_nvm_l2ptbl {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__le32			cdw2[4];
+	__le64			prp1;
+	__le64			prp2;
+	__le64			slba;
+	__le32			nlb;
+	__le16			cdw14[6];
+};
+
+struct nvme_nvm_bbtbl {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u64			rsvd[2];
+	__le64			prp1;
+	__le64			prp2;
+	__le32			prp1_len;
+	__le32			prp2_len;
+	__le32			lbb;
+	__u32			rsvd11[3];
+};
+
+struct nvme_nvm_erase_blk {
+	__u8			opcode;
+	__u8			flags;
+	__u16			command_id;
+	__le32			nsid;
+	__u64			rsvd[2];
+	__le64			prp1;
+	__le64			prp2;
+	__le64			spba;
+	__le16			length;
+	__le16			control;
+	__le32			dsmgmt;
+	__le64			resv;
+};
+
+struct nvme_nvm_command {
+	union {
+		struct nvme_common_command common;
+		struct nvme_nvm_identity identity;
+		struct nvme_nvm_hb_rw hb_rw;
+		struct nvme_nvm_ph_rw ph_rw;
+		struct nvme_nvm_l2ptbl l2p;
+		struct nvme_nvm_bbtbl get_bb;
+		struct nvme_nvm_bbtbl set_bb;
+		struct nvme_nvm_erase_blk erase;
+	};
+};
+
+struct nvme_nvm_id_group {
+	__u8			mtype;
+	__u8			fmtype;
+	__le16			res16;
+	__u8			num_ch;
+	__u8			num_lun;
+	__u8			num_pln;
+	__le16			num_blk;
+	__le16			num_pg;
+	__le16			fpg_sz;
+	__le16			csecs;
+	__le16			sos;
+	__le32			trdt;
+	__le32			trdm;
+	__le32			tprt;
+	__le32			tprm;
+	__le32			tbet;
+	__le32			tbem;
+	__le32			mpos;
+	__le16			cpar;
+	__u8			reserved[913];
+} __packed;
+
+struct nvme_nvm_addr_format {
+	__u8			ch_offset;
+	__u8			ch_len;
+	__u8			lun_offset;
+	__u8			lun_len;
+	__u8			pln_offset;
+	__u8			pln_len;
+	__u8			blk_offset;
+	__u8			blk_len;
+	__u8			pg_offset;
+	__u8			pg_len;
+	__u8			sect_offset;
+	__u8			sect_len;
+	__u8			res[4];
+} __packed;
+
+struct nvme_nvm_id {
+	__u8			ver_id;
+	__u8			vmnt;
+	__u8			cgrps;
+	__u8			res[5];
+	__le32			cap;
+	__le32			dom;
+	struct nvme_nvm_addr_format ppaf;
+	__u8			ppat;
+	__u8			resv[223];
+	struct nvme_nvm_id_group groups[4];
+} __packed;
+
+/*
+ * Check we didn't inadvertently grow the command struct
+ */
+static inline void _nvme_nvm_check_size(void)
+{
+	BUILD_BUG_ON(sizeof(struct nvme_nvm_identity) != 64);
+	BUILD_BUG_ON(sizeof(struct nvme_nvm_hb_rw) != 64);
+	BUILD_BUG_ON(sizeof(struct nvme_nvm_ph_rw) != 64);
+	BUILD_BUG_ON(sizeof(struct nvme_nvm_bbtbl) != 64);
+	BUILD_BUG_ON(sizeof(struct nvme_nvm_l2ptbl) != 64);
+	BUILD_BUG_ON(sizeof(struct nvme_nvm_erase_blk) != 64);
+	BUILD_BUG_ON(sizeof(struct nvme_nvm_id_group) != 960);
+	BUILD_BUG_ON(sizeof(struct nvme_nvm_addr_format) != 128);
+	BUILD_BUG_ON(sizeof(struct nvme_nvm_id) != 4096);
+}
+
+static int init_grps(struct nvm_id *nvm_id, struct nvme_nvm_id *nvme_nvm_id)
+{
+	struct nvme_nvm_id_group *src;
+	struct nvm_id_group *dst;
+	int i, end;
+
+	end = min_t(u32, 4, nvm_id->cgrps);
+
+	for (i = 0; i < end; i++) {
+		src = &nvme_nvm_id->groups[i];
+		dst = &nvm_id->groups[i];
+
+		dst->mtype = src->mtype;
+		dst->fmtype = src->fmtype;
+		dst->num_ch = src->num_ch;
+		dst->num_lun = src->num_lun;
+		dst->num_pln = src->num_pln;
+
+		dst->num_pg = le16_to_cpu(src->num_pg);
+		dst->num_blk = le16_to_cpu(src->num_blk);
+		dst->fpg_sz = le16_to_cpu(src->fpg_sz);
+		dst->csecs = le16_to_cpu(src->csecs);
+		dst->sos = le16_to_cpu(src->sos);
+
+		dst->trdt = le32_to_cpu(src->trdt);
+		dst->trdm = le32_to_cpu(src->trdm);
+		dst->tprt = le32_to_cpu(src->tprt);
+		dst->tprm = le32_to_cpu(src->tprm);
+		dst->tbet = le32_to_cpu(src->tbet);
+		dst->tbem = le32_to_cpu(src->tbem);
+		dst->mpos = le32_to_cpu(src->mpos);
+
+		dst->cpar = le16_to_cpu(src->cpar);
+	}
+
+	return 0;
+}
+
+static int nvme_nvm_identity(struct request_queue *q, struct nvm_id *nvm_id)
+{
+	struct nvme_ns *ns = q->queuedata;
+	struct nvme_nvm_id *nvme_nvm_id;
+	struct nvme_nvm_command c = {};
+	int ret;
+
+	c.identity.opcode = nvme_nvm_admin_identity;
+	c.identity.nsid = cpu_to_le32(ns->ns_id);
+	c.identity.chnl_off = 0;
+
+	nvme_nvm_id = kmalloc(sizeof(struct nvme_nvm_id), GFP_KERNEL);
+	if (!nvme_nvm_id)
+		return -ENOMEM;
+
+	ret = nvme_submit_sync_cmd(q, (struct nvme_command *)&c, nvme_nvm_id,
+						sizeof(struct nvme_nvm_id));
+	if (ret) {
+		ret = -EIO;
+		goto out;
+	}
+
+	nvm_id->ver_id = nvme_nvm_id->ver_id;
+	nvm_id->vmnt = nvme_nvm_id->vmnt;
+	nvm_id->cgrps = nvme_nvm_id->cgrps;
+	nvm_id->cap = le32_to_cpu(nvme_nvm_id->cap);
+	nvm_id->dom = le32_to_cpu(nvme_nvm_id->dom);
+
+	ret = init_grps(nvm_id, nvme_nvm_id);
+out:
+	kfree(nvme_nvm_id);
+	return ret;
+}
+
+static int nvme_nvm_get_l2p_tbl(struct request_queue *q, u64 slba, u32 nlb,
+				nvm_l2p_update_fn *update_l2p, void *priv)
+{
+	struct nvme_ns *ns = q->queuedata;
+	struct nvme_dev *dev = ns->dev;
+	struct nvme_nvm_command c = {};
+	u32 len = queue_max_hw_sectors(q) << 9;
+	u64 nlb_pr_rq = len / sizeof(u64);
+	u64 cmd_slba = slba;
+	void *entries;
+	int ret = 0;
+
+	c.l2p.opcode = nvme_nvm_admin_get_l2p_tbl;
+	c.l2p.nsid = cpu_to_le32(ns->ns_id);
+	entries = kmalloc(len, GFP_KERNEL);
+	if (!entries)
+		return -ENOMEM;
+
+	while (nlb) {
+		u32 cmd_nlb = min_t(u32, nlb_pr_rq, nlb);
+
+		c.l2p.slba = cpu_to_le64(cmd_slba);
+		c.l2p.nlb = cpu_to_le32(cmd_nlb);
+
+		ret = nvme_submit_sync_cmd(q, (struct nvme_command *)&c,
+								entries, len);
+		if (ret) {
+			dev_err(dev->dev, "L2P table transfer failed (%d)\n",
+									ret);
+			ret = -EIO;
+			goto out;
+		}
+
+		if (update_l2p(cmd_slba, cmd_nlb, entries, priv)) {
+			ret = -EINTR;
+			goto out;
+		}
+
+		cmd_slba += cmd_nlb;
+		nlb -= cmd_nlb;
+	}
+
+out:
+	kfree(entries);
+	return ret;
+}
+
+static int nvme_nvm_get_bb_tbl(struct request_queue *q, int lunid,
+				unsigned int nr_blocks,
+				nvm_bb_update_fn *update_bbtbl, void *priv)
+{
+	struct nvme_ns *ns = q->queuedata;
+	struct nvme_dev *dev = ns->dev;
+	struct nvme_nvm_command c = {};
+	void *bb_bitmap;
+	u16 bb_bitmap_size;
+	int ret = 0;
+
+	c.get_bb.opcode = nvme_nvm_admin_get_bb_tbl;
+	c.get_bb.nsid = cpu_to_le32(ns->ns_id);
+	c.get_bb.lbb = cpu_to_le32(lunid);
+	bb_bitmap_size = ((nr_blocks >> 15) + 1) * PAGE_SIZE;
+	bb_bitmap = kmalloc(bb_bitmap_size, GFP_KERNEL);
+	if (!bb_bitmap)
+		return -ENOMEM;
+
+	bitmap_zero(bb_bitmap, nr_blocks);
+
+	ret = nvme_submit_sync_cmd(q, (struct nvme_command *)&c, bb_bitmap,
+								bb_bitmap_size);
+	if (ret) {
+		dev_err(dev->dev, "get bad block table failed (%d)\n", ret);
+		ret = -EIO;
+		goto out;
+	}
+
+	ret = update_bbtbl(lunid, bb_bitmap, nr_blocks, priv);
+	if (ret) {
+		ret = -EINTR;
+		goto out;
+	}
+
+out:
+	kfree(bb_bitmap);
+	return ret;
+}
+
+static inline void nvme_nvm_rqtocmd(struct request *rq, struct nvm_rq *rqd,
+				struct nvme_ns *ns, struct nvme_nvm_command *c)
+{
+	c->ph_rw.opcode = rqd->opcode;
+	c->ph_rw.nsid = cpu_to_le32(ns->ns_id);
+	c->ph_rw.spba = cpu_to_le64(rqd->ppa_addr.ppa);
+	c->ph_rw.control = cpu_to_le16(rqd->flags);
+	c->ph_rw.length = cpu_to_le16(rqd->nr_pages - 1);
+
+	if (rqd->opcode == NVM_OP_HBWRITE || rqd->opcode == NVM_OP_HBREAD)
+		c->hb_rw.slba = cpu_to_le64(nvme_block_nr(ns,
+						rqd->bio->bi_iter.bi_sector));
+}
+
+static void nvme_nvm_end_io(struct request *rq, int error)
+{
+	struct nvm_rq *rqd = rq->end_io_data;
+	struct nvm_dev *dev = rqd->dev;
+
+	if (dev->mt->end_io(rqd, error))
+		pr_err("nvme: err status: %x result: %lx\n",
+				rq->errors, (unsigned long)rq->special);
+
+	kfree(rq->cmd);
+	blk_mq_free_request(rq);
+}
+
+static int nvme_nvm_submit_io(struct request_queue *q, struct nvm_rq *rqd)
+{
+	struct nvme_ns *ns = q->queuedata;
+	struct request *rq;
+	struct bio *bio = rqd->bio;
+	struct nvme_nvm_command *cmd;
+
+	rq = blk_mq_alloc_request(q, bio_rw(bio), GFP_KERNEL, 0);
+	if (IS_ERR(rq))
+		return -ENOMEM;
+
+	cmd = kzalloc(sizeof(struct nvme_nvm_command), GFP_KERNEL);
+	if (!cmd) {
+		blk_mq_free_request(rq);
+		return -ENOMEM;
+	}
+
+	rq->cmd_type = REQ_TYPE_DRV_PRIV;
+	rq->ioprio = bio_prio(bio);
+
+	if (bio_has_data(bio))
+		rq->nr_phys_segments = bio_phys_segments(q, bio);
+
+	rq->__data_len = bio->bi_iter.bi_size;
+	rq->bio = rq->biotail = bio;
+
+	nvme_nvm_rqtocmd(rq, rqd, ns, cmd);
+
+	rq->cmd = (unsigned char *)cmd;
+	rq->cmd_len = sizeof(struct nvme_nvm_command);
+	rq->special = (void *)0;
+
+	rq->end_io_data = rqd;
+
+	blk_execute_rq_nowait(q, NULL, rq, 0, nvme_nvm_end_io);
+
+	return 0;
+}
+
+static int nvme_nvm_erase_block(struct request_queue *q, struct nvm_rq *rqd)
+{
+	struct nvme_ns *ns = q->queuedata;
+	struct nvme_nvm_command c = {};
+
+	c.erase.opcode = NVM_OP_ERASE;
+	c.erase.nsid = cpu_to_le32(ns->ns_id);
+	c.erase.spba = cpu_to_le64(rqd->ppa_addr.ppa);
+	c.erase.length = cpu_to_le16(rqd->nr_pages - 1);
+
+	return nvme_submit_sync_cmd(q, (struct nvme_command *)&c, NULL, 0);
+}
+
+static void *nvme_nvm_create_dma_pool(struct request_queue *q, char *name)
+{
+	struct nvme_ns *ns = q->queuedata;
+	struct nvme_dev *dev = ns->dev;
+
+	return dma_pool_create(name, dev->dev, PAGE_SIZE, PAGE_SIZE, 0);
+}
+
+static void nvme_nvm_destroy_dma_pool(void *pool)
+{
+	struct dma_pool *dma_pool = pool;
+
+	dma_pool_destroy(dma_pool);
+}
+
+static void *nvme_nvm_dev_dma_alloc(struct request_queue *q, void *pool,
+				    gfp_t mem_flags, dma_addr_t *dma_handler)
+{
+	return dma_pool_alloc(pool, mem_flags, dma_handler);
+}
+
+static void nvme_nvm_dev_dma_free(void *pool, void *ppa_list,
+							dma_addr_t dma_handler)
+{
+	dma_pool_free(pool, ppa_list, dma_handler);
+}
+
+static struct nvm_dev_ops nvme_nvm_dev_ops = {
+	.identity		= nvme_nvm_identity,
+
+	.get_l2p_tbl		= nvme_nvm_get_l2p_tbl,
+
+	.get_bb_tbl		= nvme_nvm_get_bb_tbl,
+
+	.submit_io		= nvme_nvm_submit_io,
+	.erase_block		= nvme_nvm_erase_block,
+
+	.create_dma_pool	= nvme_nvm_create_dma_pool,
+	.destroy_dma_pool	= nvme_nvm_destroy_dma_pool,
+	.dev_dma_alloc		= nvme_nvm_dev_dma_alloc,
+	.dev_dma_free		= nvme_nvm_dev_dma_free,
+
+	.max_phys_sect		= 64,
+};
+
+int nvme_nvm_register(struct request_queue *q, char *disk_name)
+{
+	return nvm_register(q, disk_name, &nvme_nvm_dev_ops);
+}
+
+void nvme_nvm_unregister(struct request_queue *q, char *disk_name)
+{
+	nvm_unregister(disk_name);
+}
+
+int nvme_nvm_ns_supported(struct nvme_ns *ns, struct nvme_id_ns *id)
+{
+	struct nvme_dev *dev = ns->dev;
+	struct pci_dev *pdev = to_pci_dev(dev->dev);
+
+	/* QEMU NVMe simulator - PCI ID + Vendor specific bit */
+	if (pdev->vendor == PCI_VENDOR_ID_INTEL && pdev->device == 0x5845 &&
+							id->vs[0] == 0x1)
+		return 1;
+
+	/* CNEX Labs - PCI ID + Vendor specific bit */
+	if (pdev->vendor == 0x1d1d && pdev->device == 0x2807 &&
+							id->vs[0] == 0x1)
+		return 1;
+
+	return 0;
+}
+#else
+int nvme_nvm_register(struct request_queue *q, char *disk_name)
+{
+	return 0;
+}
+void nvme_nvm_unregister(struct request_queue *q, char *disk_name) {};
+int nvme_nvm_ns_supported(struct nvme_ns *ns, struct nvme_id_ns *id)
+{
+	return 0;
+}
+#endif /* CONFIG_NVM */

drivers/nvme/host/nvme.h

@@ -22,6 +22,11 @@
 extern unsigned char nvme_io_timeout;
 #define NVME_IO_TIMEOUT	(nvme_io_timeout * HZ)
 
+enum {
+	NVME_NS_LBA		= 0,
+	NVME_NS_LIGHTNVM	= 1,
+};
+
 /*
  * Represents an NVM Express device.  Each nvme_dev is a PCI function.
  */
@@ -84,6 +89,7 @@ struct nvme_ns {
 	u16 ms;
 	bool ext;
 	u8 pi_type;
+	int type;
 	u64 mode_select_num_blocks;
 	u32 mode_select_block_len;
 };
@@ -130,4 +136,8 @@ int nvme_sg_io(struct nvme_ns *ns, struct sg_io_hdr __user *u_hdr);
 int nvme_sg_io32(struct nvme_ns *ns, unsigned long arg);
 int nvme_sg_get_version_num(int __user *ip);
 
+int nvme_nvm_ns_supported(struct nvme_ns *ns, struct nvme_id_ns *id);
+int nvme_nvm_register(struct request_queue *q, char *disk_name);
+void nvme_nvm_unregister(struct request_queue *q, char *disk_name);
+
 #endif /* _NVME_H */

drivers/nvme/host/pci.c

@@ -1948,6 +1948,9 @@ static void nvme_free_ns(struct kref *kref)
 {
 	struct nvme_ns *ns = container_of(kref, struct nvme_ns, kref);
 
+	if (ns->type == NVME_NS_LIGHTNVM)
+		nvme_nvm_unregister(ns->queue, ns->disk->disk_name);
+
 	spin_lock(&dev_list_lock);
 	ns->disk->private_data = NULL;
 	spin_unlock(&dev_list_lock);
@@ -2017,6 +2020,16 @@ static int nvme_revalidate_disk(struct gendisk *disk)
 		return -ENODEV;
 	}
 
+	if (nvme_nvm_ns_supported(ns, id) && ns->type != NVME_NS_LIGHTNVM) {
+		if (nvme_nvm_register(ns->queue, disk->disk_name)) {
+			dev_warn(dev->dev,
+				"%s: LightNVM init failure\n", __func__);
+			kfree(id);
+			return -ENODEV;
+		}
+		ns->type = NVME_NS_LIGHTNVM;
+	}
+
 	old_ms = ns->ms;
 	lbaf = id->flbas & NVME_NS_FLBAS_LBA_MASK;
 	ns->lba_shift = id->lbaf[lbaf].ds;
@@ -2048,7 +2061,9 @@ static int nvme_revalidate_disk(struct gendisk *disk)
 								!ns->ext)
 		nvme_init_integrity(ns);
 
-	if (ns->ms && !(ns->ms == 8 && ns->pi_type) && !blk_get_integrity(disk))
+	if ((ns->ms && !(ns->ms == 8 && ns->pi_type) &&
+						!blk_get_integrity(disk)) ||
+						ns->type == NVME_NS_LIGHTNVM)
 		set_capacity(disk, 0);
 	else
 		set_capacity(disk, le64_to_cpup(&id->nsze) << (ns->lba_shift - 9));
@@ -2171,17 +2186,19 @@ static void nvme_alloc_ns(struct nvme_dev *dev, unsigned nsid)
 		goto out_free_disk;
 
 	kref_get(&dev->kref);
-	add_disk(ns->disk);
-	if (ns->ms) {
-		struct block_device *bd = bdget_disk(ns->disk, 0);
-		if (!bd)
-			return;
-		if (blkdev_get(bd, FMODE_READ, NULL)) {
-			bdput(bd);
-			return;
+	if (ns->type != NVME_NS_LIGHTNVM) {
+		add_disk(ns->disk);
+		if (ns->ms) {
+			struct block_device *bd = bdget_disk(ns->disk, 0);
+			if (!bd)
+				return;
+			if (blkdev_get(bd, FMODE_READ, NULL)) {
+				bdput(bd);
+				return;
+			}
+			blkdev_reread_part(bd);
+			blkdev_put(bd, FMODE_READ);
 		}
-		blkdev_reread_part(bd);
-		blkdev_put(bd, FMODE_READ);
 	}
 	return;
  out_free_disk: