nvmet: add a generic NVMe target

This patch introduces a implementation of NVMe subsystems,
controllers and discovery service which allows to export
NVMe namespaces across fabrics such as Ethernet, FC etc.

The implementation conforms to the NVMe 1.2.1 specification
and interoperates with NVMe over fabrics host implementations.

Configuration works using configfs, and is best performed using
the nvmetcli tool from http://git.infradead.org/users/hch/nvmetcli.git,
which also has a detailed explanation of the required steps in the
README file.

Signed-off-by: Armen Baloyan <armenx.baloyan@intel.com>
Signed-off-by: Anthony Knapp <anthony.j.knapp@intel.com>
Signed-off-by: Jay Freyensee <james.p.freyensee@intel.com>
Signed-off-by: Ming Lin <ming.l@ssi.samsung.com>
Signed-off-by: Sagi Grimberg <sagi@grimberg.me>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Steve Wise <swise@opengridcomputing.com>
Signed-off-by: Jens Axboe <axboe@fb.com>

MAINTAINERS

@@ -8153,6 +8153,13 @@ S:	Supported
 F:	drivers/nvme/host/
 F:	include/linux/nvme.h
 
+NVM EXPRESS TARGET DRIVER
+M:	Christoph Hellwig <hch@lst.de>
+M:	Sagi Grimberg <sagi@grimberg.me>
+L:	linux-nvme@lists.infradead.org
+S:	Supported
+F:	drivers/nvme/target/
+
 NVMEM FRAMEWORK
 M:	Srinivas Kandagatla <srinivas.kandagatla@linaro.org>
 M:	Maxime Ripard <maxime.ripard@free-electrons.com>

drivers/nvme/Kconfig

@@ -1 +1,2 @@
 source "drivers/nvme/host/Kconfig"
+source "drivers/nvme/target/Kconfig"

drivers/nvme/Makefile

@@ -1,2 +1,3 @@
 
 obj-y		+= host/
+obj-y		+= target/

drivers/nvme/target/Kconfig

@@ -0,0 +1,16 @@
+
+config NVME_TARGET
+	tristate "NVMe Target support"
+	depends on BLOCK
+	depends on CONFIGFS_FS
+	help
+	  This enabled target side support for the NVMe protocol, that is
+	  it allows the Linux kernel to implement NVMe subsystems and
+	  controllers and export Linux block devices as NVMe namespaces.
+	  You need to select at least one of the transports below to make this
+	  functionality useful.
+
+	  To configure the NVMe target you probably want to use the nvmetcli
+	  tool from http://git.infradead.org/users/hch/nvmetcli.git.
+
+	  If unsure, say N.

drivers/nvme/target/Makefile

@@ -0,0 +1,5 @@
+
+obj-$(CONFIG_NVME_TARGET)		+= nvmet.o
+
+nvmet-y		+= core.o configfs.o admin-cmd.o io-cmd.o fabrics-cmd.o \
+			discovery.o

drivers/nvme/target/admin-cmd.c

@@ -0,0 +1,465 @@
+/*
+ * NVMe admin command implementation.
+ * Copyright (c) 2015-2016 HGST, a Western Digital Company.
+ *
+ * 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.
+ */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+#include <linux/module.h>
+#include <linux/random.h>
+#include <generated/utsrelease.h>
+#include "nvmet.h"
+
+u32 nvmet_get_log_page_len(struct nvme_command *cmd)
+{
+	u32 len = le16_to_cpu(cmd->get_log_page.numdu);
+
+	len <<= 16;
+	len += le16_to_cpu(cmd->get_log_page.numdl);
+	/* NUMD is a 0's based value */
+	len += 1;
+	len *= sizeof(u32);
+
+	return len;
+}
+
+static void nvmet_execute_get_log_page(struct nvmet_req *req)
+{
+	size_t data_len = nvmet_get_log_page_len(req->cmd);
+	void *buf;
+	u16 status = 0;
+
+	buf = kzalloc(data_len, GFP_KERNEL);
+	if (!buf) {
+		status = NVME_SC_INTERNAL;
+		goto out;
+	}
+
+	switch (req->cmd->get_log_page.lid) {
+	case 0x01:
+		/*
+		 * We currently never set the More bit in the status field,
+		 * so all error log entries are invalid and can be zeroed out.
+		 * This is called a minum viable implementation (TM) of this
+		 * mandatory log page.
+		 */
+		break;
+	case 0x02:
+		/*
+		 * XXX: fill out actual smart log
+		 *
+		 * We might have a hard time coming up with useful values for
+		 * many of the fields, and even when we have useful data
+		 * available (e.g. units or commands read/written) those aren't
+		 * persistent over power loss.
+		 */
+		break;
+	case 0x03:
+		/*
+		 * We only support a single firmware slot which always is
+		 * active, so we can zero out the whole firmware slot log and
+		 * still claim to fully implement this mandatory log page.
+		 */
+		break;
+	default:
+		BUG();
+	}
+
+	status = nvmet_copy_to_sgl(req, 0, buf, data_len);
+
+	kfree(buf);
+out:
+	nvmet_req_complete(req, status);
+}
+
+static void nvmet_execute_identify_ctrl(struct nvmet_req *req)
+{
+	struct nvmet_ctrl *ctrl = req->sq->ctrl;
+	struct nvme_id_ctrl *id;
+	u64 serial;
+	u16 status = 0;
+
+	id = kzalloc(sizeof(*id), GFP_KERNEL);
+	if (!id) {
+		status = NVME_SC_INTERNAL;
+		goto out;
+	}
+
+	/* XXX: figure out how to assign real vendors IDs. */
+	id->vid = 0;
+	id->ssvid = 0;
+
+	/* generate a random serial number as our controllers are ephemeral: */
+	get_random_bytes(&serial, sizeof(serial));
+	memset(id->sn, ' ', sizeof(id->sn));
+	snprintf(id->sn, sizeof(id->sn), "%llx", serial);
+
+	memset(id->mn, ' ', sizeof(id->mn));
+	strncpy((char *)id->mn, "Linux", sizeof(id->mn));
+
+	memset(id->fr, ' ', sizeof(id->fr));
+	strncpy((char *)id->fr, UTS_RELEASE, sizeof(id->fr));
+
+	id->rab = 6;
+
+	/*
+	 * XXX: figure out how we can assign a IEEE OUI, but until then
+	 * the safest is to leave it as zeroes.
+	 */
+
+	/* we support multiple ports and multiples hosts: */
+	id->mic = (1 << 0) | (1 << 1);
+
+	/* no limit on data transfer sizes for now */
+	id->mdts = 0;
+	id->cntlid = cpu_to_le16(ctrl->cntlid);
+	id->ver = cpu_to_le32(ctrl->subsys->ver);
+
+	/* XXX: figure out what to do about RTD3R/RTD3 */
+	id->oaes = cpu_to_le32(1 << 8);
+	id->ctratt = cpu_to_le32(1 << 0);
+
+	id->oacs = 0;
+
+	/*
+	 * We don't really have a practical limit on the number of abort
+	 * comands.  But we don't do anything useful for abort either, so
+	 * no point in allowing more abort commands than the spec requires.
+	 */
+	id->acl = 3;
+
+	id->aerl = NVMET_ASYNC_EVENTS - 1;
+
+	/* first slot is read-only, only one slot supported */
+	id->frmw = (1 << 0) | (1 << 1);
+	id->lpa = (1 << 0) | (1 << 2);
+	id->elpe = NVMET_ERROR_LOG_SLOTS - 1;
+	id->npss = 0;
+
+	/* We support keep-alive timeout in granularity of seconds */
+	id->kas = cpu_to_le16(NVMET_KAS);
+
+	id->sqes = (0x6 << 4) | 0x6;
+	id->cqes = (0x4 << 4) | 0x4;
+
+	/* no enforcement soft-limit for maxcmd - pick arbitrary high value */
+	id->maxcmd = cpu_to_le16(NVMET_MAX_CMD);
+
+	id->nn = cpu_to_le32(ctrl->subsys->max_nsid);
+	id->oncs = cpu_to_le16(NVME_CTRL_ONCS_DSM);
+
+	/* XXX: don't report vwc if the underlying device is write through */
+	id->vwc = NVME_CTRL_VWC_PRESENT;
+
+	/*
+	 * We can't support atomic writes bigger than a LBA without support
+	 * from the backend device.
+	 */
+	id->awun = 0;
+	id->awupf = 0;
+
+	id->sgls = cpu_to_le32(1 << 0);	/* we always support SGLs */
+	if (ctrl->ops->has_keyed_sgls)
+		id->sgls |= cpu_to_le32(1 << 2);
+	if (ctrl->ops->sqe_inline_size)
+		id->sgls |= cpu_to_le32(1 << 20);
+
+	strcpy(id->subnqn, ctrl->subsys->subsysnqn);
+
+	/* Max command capsule size is sqe + single page of in-capsule data */
+	id->ioccsz = cpu_to_le32((sizeof(struct nvme_command) +
+				  ctrl->ops->sqe_inline_size) / 16);
+	/* Max response capsule size is cqe */
+	id->iorcsz = cpu_to_le32(sizeof(struct nvme_completion) / 16);
+
+	id->msdbd = ctrl->ops->msdbd;
+
+	/*
+	 * Meh, we don't really support any power state.  Fake up the same
+	 * values that qemu does.
+	 */
+	id->psd[0].max_power = cpu_to_le16(0x9c4);
+	id->psd[0].entry_lat = cpu_to_le32(0x10);
+	id->psd[0].exit_lat = cpu_to_le32(0x4);
+
+	status = nvmet_copy_to_sgl(req, 0, id, sizeof(*id));
+
+	kfree(id);
+out:
+	nvmet_req_complete(req, status);
+}
+
+static void nvmet_execute_identify_ns(struct nvmet_req *req)
+{
+	struct nvmet_ns *ns;
+	struct nvme_id_ns *id;
+	u16 status = 0;
+
+	ns = nvmet_find_namespace(req->sq->ctrl, req->cmd->identify.nsid);
+	if (!ns) {
+		status = NVME_SC_INVALID_NS | NVME_SC_DNR;
+		goto out;
+	}
+
+	id = kzalloc(sizeof(*id), GFP_KERNEL);
+	if (!id) {
+		status = NVME_SC_INTERNAL;
+		goto out_put_ns;
+	}
+
+	/*
+	 * nuse = ncap = nsze isn't aways true, but we have no way to find
+	 * that out from the underlying device.
+	 */
+	id->ncap = id->nuse = id->nsze =
+		cpu_to_le64(ns->size >> ns->blksize_shift);
+
+	/*
+	 * We just provide a single LBA format that matches what the
+	 * underlying device reports.
+	 */
+	id->nlbaf = 0;
+	id->flbas = 0;
+
+	/*
+	 * Our namespace might always be shared.  Not just with other
+	 * controllers, but also with any other user of the block device.
+	 */
+	id->nmic = (1 << 0);
+
+	memcpy(&id->nguid, &ns->nguid, sizeof(uuid_le));
+
+	id->lbaf[0].ds = ns->blksize_shift;
+
+	status = nvmet_copy_to_sgl(req, 0, id, sizeof(*id));
+
+	kfree(id);
+out_put_ns:
+	nvmet_put_namespace(ns);
+out:
+	nvmet_req_complete(req, status);
+}
+
+static void nvmet_execute_identify_nslist(struct nvmet_req *req)
+{
+	static const int buf_size = 4096;
+	struct nvmet_ctrl *ctrl = req->sq->ctrl;
+	struct nvmet_ns *ns;
+	u32 min_nsid = le32_to_cpu(req->cmd->identify.nsid);
+	__le32 *list;
+	u16 status = 0;
+	int i = 0;
+
+	list = kzalloc(buf_size, GFP_KERNEL);
+	if (!list) {
+		status = NVME_SC_INTERNAL;
+		goto out;
+	}
+
+	rcu_read_lock();
+	list_for_each_entry_rcu(ns, &ctrl->subsys->namespaces, dev_link) {
+		if (ns->nsid <= min_nsid)
+			continue;
+		list[i++] = cpu_to_le32(ns->nsid);
+		if (i == buf_size / sizeof(__le32))
+			break;
+	}
+	rcu_read_unlock();
+
+	status = nvmet_copy_to_sgl(req, 0, list, buf_size);
+
+	kfree(list);
+out:
+	nvmet_req_complete(req, status);
+}
+
+/*
+ * A "mimimum viable" abort implementation: the command is mandatory in the
+ * spec, but we are not required to do any useful work.  We couldn't really
+ * do a useful abort, so don't bother even with waiting for the command
+ * to be exectuted and return immediately telling the command to abort
+ * wasn't found.
+ */
+static void nvmet_execute_abort(struct nvmet_req *req)
+{
+	nvmet_set_result(req, 1);
+	nvmet_req_complete(req, 0);
+}
+
+static void nvmet_execute_set_features(struct nvmet_req *req)
+{
+	struct nvmet_subsys *subsys = req->sq->ctrl->subsys;
+	u32 cdw10 = le32_to_cpu(req->cmd->common.cdw10[0]);
+	u64 val;
+	u32 val32;
+	u16 status = 0;
+
+	switch (cdw10 & 0xf) {
+	case NVME_FEAT_NUM_QUEUES:
+		nvmet_set_result(req,
+			(subsys->max_qid - 1) | ((subsys->max_qid - 1) << 16));
+		break;
+	case NVME_FEAT_KATO:
+		val = le64_to_cpu(req->cmd->prop_set.value);
+		val32 = val & 0xffff;
+		req->sq->ctrl->kato = DIV_ROUND_UP(val32, 1000);
+		nvmet_set_result(req, req->sq->ctrl->kato);
+		break;
+	default:
+		status = NVME_SC_INVALID_FIELD | NVME_SC_DNR;
+		break;
+	}
+
+	nvmet_req_complete(req, status);
+}
+
+static void nvmet_execute_get_features(struct nvmet_req *req)
+{
+	struct nvmet_subsys *subsys = req->sq->ctrl->subsys;
+	u32 cdw10 = le32_to_cpu(req->cmd->common.cdw10[0]);
+	u16 status = 0;
+
+	switch (cdw10 & 0xf) {
+	/*
+	 * These features are mandatory in the spec, but we don't
+	 * have a useful way to implement them.  We'll eventually
+	 * need to come up with some fake values for these.
+	 */
+#if 0
+	case NVME_FEAT_ARBITRATION:
+		break;
+	case NVME_FEAT_POWER_MGMT:
+		break;
+	case NVME_FEAT_TEMP_THRESH:
+		break;
+	case NVME_FEAT_ERR_RECOVERY:
+		break;
+	case NVME_FEAT_IRQ_COALESCE:
+		break;
+	case NVME_FEAT_IRQ_CONFIG:
+		break;
+	case NVME_FEAT_WRITE_ATOMIC:
+		break;
+	case NVME_FEAT_ASYNC_EVENT:
+		break;
+#endif
+	case NVME_FEAT_VOLATILE_WC:
+		nvmet_set_result(req, 1);
+		break;
+	case NVME_FEAT_NUM_QUEUES:
+		nvmet_set_result(req,
+			(subsys->max_qid-1) | ((subsys->max_qid-1) << 16));
+		break;
+	case NVME_FEAT_KATO:
+		nvmet_set_result(req, req->sq->ctrl->kato * 1000);
+		break;
+	default:
+		status = NVME_SC_INVALID_FIELD | NVME_SC_DNR;
+		break;
+	}
+
+	nvmet_req_complete(req, status);
+}
+
+static void nvmet_execute_async_event(struct nvmet_req *req)
+{
+	struct nvmet_ctrl *ctrl = req->sq->ctrl;
+
+	mutex_lock(&ctrl->lock);
+	if (ctrl->nr_async_event_cmds >= NVMET_ASYNC_EVENTS) {
+		mutex_unlock(&ctrl->lock);
+		nvmet_req_complete(req, NVME_SC_ASYNC_LIMIT | NVME_SC_DNR);
+		return;
+	}
+	ctrl->async_event_cmds[ctrl->nr_async_event_cmds++] = req;
+	mutex_unlock(&ctrl->lock);
+
+	schedule_work(&ctrl->async_event_work);
+}
+
+static void nvmet_execute_keep_alive(struct nvmet_req *req)
+{
+	struct nvmet_ctrl *ctrl = req->sq->ctrl;
+
+	pr_debug("ctrl %d update keep-alive timer for %d secs\n",
+		ctrl->cntlid, ctrl->kato);
+
+	mod_delayed_work(system_wq, &ctrl->ka_work, ctrl->kato * HZ);
+	nvmet_req_complete(req, 0);
+}
+
+int nvmet_parse_admin_cmd(struct nvmet_req *req)
+{
+	struct nvme_command *cmd = req->cmd;
+
+	req->ns = NULL;
+
+	if (unlikely(!(req->sq->ctrl->cc & NVME_CC_ENABLE))) {
+		pr_err("nvmet: got admin cmd %d while CC.EN == 0\n",
+				cmd->common.opcode);
+		return NVME_SC_CMD_SEQ_ERROR | NVME_SC_DNR;
+	}
+	if (unlikely(!(req->sq->ctrl->csts & NVME_CSTS_RDY))) {
+		pr_err("nvmet: got admin cmd %d while CSTS.RDY == 0\n",
+				cmd->common.opcode);
+		return NVME_SC_CMD_SEQ_ERROR | NVME_SC_DNR;
+	}
+
+	switch (cmd->common.opcode) {
+	case nvme_admin_get_log_page:
+		req->data_len = nvmet_get_log_page_len(cmd);
+
+		switch (cmd->get_log_page.lid) {
+		case 0x01:
+		case 0x02:
+		case 0x03:
+			req->execute = nvmet_execute_get_log_page;
+			return 0;
+		}
+		break;
+	case nvme_admin_identify:
+		req->data_len = 4096;
+		switch (le32_to_cpu(cmd->identify.cns)) {
+		case 0x00:
+			req->execute = nvmet_execute_identify_ns;
+			return 0;
+		case 0x01:
+			req->execute = nvmet_execute_identify_ctrl;
+			return 0;
+		case 0x02:
+			req->execute = nvmet_execute_identify_nslist;
+			return 0;
+		}
+		break;
+	case nvme_admin_abort_cmd:
+		req->execute = nvmet_execute_abort;
+		req->data_len = 0;
+		return 0;
+	case nvme_admin_set_features:
+		req->execute = nvmet_execute_set_features;
+		req->data_len = 0;
+		return 0;
+	case nvme_admin_get_features:
+		req->execute = nvmet_execute_get_features;
+		req->data_len = 0;
+		return 0;
+	case nvme_admin_async_event:
+		req->execute = nvmet_execute_async_event;
+		req->data_len = 0;
+		return 0;
+	case nvme_admin_keep_alive:
+		req->execute = nvmet_execute_keep_alive;
+		req->data_len = 0;
+		return 0;
+	}
+
+	pr_err("nvmet: unhandled cmd %d\n", cmd->common.opcode);
+	return NVME_SC_INVALID_OPCODE | NVME_SC_DNR;
+}

drivers/nvme/target/configfs.c

@@ -0,0 +1,917 @@
+/*
+ * Configfs interface for the NVMe target.
+ * Copyright (c) 2015-2016 HGST, a Western Digital Company.
+ *
+ * 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.
+ */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/stat.h>
+#include <linux/ctype.h>
+
+#include "nvmet.h"
+
+static struct config_item_type nvmet_host_type;
+static struct config_item_type nvmet_subsys_type;
+
+/*
+ * nvmet_port Generic ConfigFS definitions.
+ * Used in any place in the ConfigFS tree that refers to an address.
+ */
+static ssize_t nvmet_addr_adrfam_show(struct config_item *item,
+		char *page)
+{
+	switch (to_nvmet_port(item)->disc_addr.adrfam) {
+	case NVMF_ADDR_FAMILY_IP4:
+		return sprintf(page, "ipv4\n");
+	case NVMF_ADDR_FAMILY_IP6:
+		return sprintf(page, "ipv6\n");
+	case NVMF_ADDR_FAMILY_IB:
+		return sprintf(page, "ib\n");
+	default:
+		return sprintf(page, "\n");
+	}
+}
+
+static ssize_t nvmet_addr_adrfam_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+
+	if (port->enabled) {
+		pr_err("Cannot modify address while enabled\n");
+		pr_err("Disable the address before modifying\n");
+		return -EACCES;
+	}
+
+	if (sysfs_streq(page, "ipv4")) {
+		port->disc_addr.adrfam = NVMF_ADDR_FAMILY_IP4;
+	} else if (sysfs_streq(page, "ipv6")) {
+		port->disc_addr.adrfam = NVMF_ADDR_FAMILY_IP6;
+	} else if (sysfs_streq(page, "ib")) {
+		port->disc_addr.adrfam = NVMF_ADDR_FAMILY_IB;
+	} else {
+		pr_err("Invalid value '%s' for adrfam\n", page);
+		return -EINVAL;
+	}
+
+	return count;
+}
+
+CONFIGFS_ATTR(nvmet_, addr_adrfam);
+
+static ssize_t nvmet_addr_portid_show(struct config_item *item,
+		char *page)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+
+	return snprintf(page, PAGE_SIZE, "%d\n",
+			le16_to_cpu(port->disc_addr.portid));
+}
+
+static ssize_t nvmet_addr_portid_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+	u16 portid = 0;
+
+	if (kstrtou16(page, 0, &portid)) {
+		pr_err("Invalid value '%s' for portid\n", page);
+		return -EINVAL;
+	}
+
+	if (port->enabled) {
+		pr_err("Cannot modify address while enabled\n");
+		pr_err("Disable the address before modifying\n");
+		return -EACCES;
+	}
+	port->disc_addr.portid = cpu_to_le16(portid);
+	return count;
+}
+
+CONFIGFS_ATTR(nvmet_, addr_portid);
+
+static ssize_t nvmet_addr_traddr_show(struct config_item *item,
+		char *page)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+
+	return snprintf(page, PAGE_SIZE, "%s\n",
+			port->disc_addr.traddr);
+}
+
+static ssize_t nvmet_addr_traddr_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+
+	if (count > NVMF_TRADDR_SIZE) {
+		pr_err("Invalid value '%s' for traddr\n", page);
+		return -EINVAL;
+	}
+
+	if (port->enabled) {
+		pr_err("Cannot modify address while enabled\n");
+		pr_err("Disable the address before modifying\n");
+		return -EACCES;
+	}
+	return snprintf(port->disc_addr.traddr,
+			sizeof(port->disc_addr.traddr), "%s", page);
+}
+
+CONFIGFS_ATTR(nvmet_, addr_traddr);
+
+static ssize_t nvmet_addr_treq_show(struct config_item *item,
+		char *page)
+{
+	switch (to_nvmet_port(item)->disc_addr.treq) {
+	case NVMF_TREQ_NOT_SPECIFIED:
+		return sprintf(page, "not specified\n");
+	case NVMF_TREQ_REQUIRED:
+		return sprintf(page, "required\n");
+	case NVMF_TREQ_NOT_REQUIRED:
+		return sprintf(page, "not required\n");
+	default:
+		return sprintf(page, "\n");
+	}
+}
+
+static ssize_t nvmet_addr_treq_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+
+	if (port->enabled) {
+		pr_err("Cannot modify address while enabled\n");
+		pr_err("Disable the address before modifying\n");
+		return -EACCES;
+	}
+
+	if (sysfs_streq(page, "not specified")) {
+		port->disc_addr.treq = NVMF_TREQ_NOT_SPECIFIED;
+	} else if (sysfs_streq(page, "required")) {
+		port->disc_addr.treq = NVMF_TREQ_REQUIRED;
+	} else if (sysfs_streq(page, "not required")) {
+		port->disc_addr.treq = NVMF_TREQ_NOT_REQUIRED;
+	} else {
+		pr_err("Invalid value '%s' for treq\n", page);
+		return -EINVAL;
+	}
+
+	return count;
+}
+
+CONFIGFS_ATTR(nvmet_, addr_treq);
+
+static ssize_t nvmet_addr_trsvcid_show(struct config_item *item,
+		char *page)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+
+	return snprintf(page, PAGE_SIZE, "%s\n",
+			port->disc_addr.trsvcid);
+}
+
+static ssize_t nvmet_addr_trsvcid_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+
+	if (count > NVMF_TRSVCID_SIZE) {
+		pr_err("Invalid value '%s' for trsvcid\n", page);
+		return -EINVAL;
+	}
+	if (port->enabled) {
+		pr_err("Cannot modify address while enabled\n");
+		pr_err("Disable the address before modifying\n");
+		return -EACCES;
+	}
+	return snprintf(port->disc_addr.trsvcid,
+			sizeof(port->disc_addr.trsvcid), "%s", page);
+}
+
+CONFIGFS_ATTR(nvmet_, addr_trsvcid);
+
+static ssize_t nvmet_addr_trtype_show(struct config_item *item,
+		char *page)
+{
+	switch (to_nvmet_port(item)->disc_addr.trtype) {
+	case NVMF_TRTYPE_RDMA:
+		return sprintf(page, "rdma\n");
+	case NVMF_TRTYPE_LOOP:
+		return sprintf(page, "loop\n");
+	default:
+		return sprintf(page, "\n");
+	}
+}
+
+static void nvmet_port_init_tsas_rdma(struct nvmet_port *port)
+{
+	port->disc_addr.trtype = NVMF_TRTYPE_RDMA;
+	memset(&port->disc_addr.tsas.rdma, 0, NVMF_TSAS_SIZE);
+	port->disc_addr.tsas.rdma.qptype = NVMF_RDMA_QPTYPE_CONNECTED;
+	port->disc_addr.tsas.rdma.prtype = NVMF_RDMA_PRTYPE_NOT_SPECIFIED;
+	port->disc_addr.tsas.rdma.cms = NVMF_RDMA_CMS_RDMA_CM;
+}
+
+static void nvmet_port_init_tsas_loop(struct nvmet_port *port)
+{
+	port->disc_addr.trtype = NVMF_TRTYPE_LOOP;
+	memset(&port->disc_addr.tsas, 0, NVMF_TSAS_SIZE);
+}
+
+static ssize_t nvmet_addr_trtype_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+
+	if (port->enabled) {
+		pr_err("Cannot modify address while enabled\n");
+		pr_err("Disable the address before modifying\n");
+		return -EACCES;
+	}
+
+	if (sysfs_streq(page, "rdma")) {
+		nvmet_port_init_tsas_rdma(port);
+	} else if (sysfs_streq(page, "loop")) {
+		nvmet_port_init_tsas_loop(port);
+	} else {
+		pr_err("Invalid value '%s' for trtype\n", page);
+		return -EINVAL;
+	}
+
+	return count;
+}
+
+CONFIGFS_ATTR(nvmet_, addr_trtype);
+
+/*
+ * Namespace structures & file operation functions below
+ */
+static ssize_t nvmet_ns_device_path_show(struct config_item *item, char *page)
+{
+	return sprintf(page, "%s\n", to_nvmet_ns(item)->device_path);
+}
+
+static ssize_t nvmet_ns_device_path_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_ns *ns = to_nvmet_ns(item);
+	struct nvmet_subsys *subsys = ns->subsys;
+	int ret;
+
+	mutex_lock(&subsys->lock);
+	ret = -EBUSY;
+	if (nvmet_ns_enabled(ns))
+		goto out_unlock;
+
+	kfree(ns->device_path);
+
+	ret = -ENOMEM;
+	ns->device_path = kstrdup(page, GFP_KERNEL);
+	if (!ns->device_path)
+		goto out_unlock;
+
+	mutex_unlock(&subsys->lock);
+	return count;
+
+out_unlock:
+	mutex_unlock(&subsys->lock);
+	return ret;
+}
+
+CONFIGFS_ATTR(nvmet_ns_, device_path);
+
+static ssize_t nvmet_ns_device_nguid_show(struct config_item *item, char *page)
+{
+	return sprintf(page, "%pUb\n", &to_nvmet_ns(item)->nguid);
+}
+
+static ssize_t nvmet_ns_device_nguid_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_ns *ns = to_nvmet_ns(item);
+	struct nvmet_subsys *subsys = ns->subsys;
+	u8 nguid[16];
+	const char *p = page;
+	int i;
+	int ret = 0;
+
+	mutex_lock(&subsys->lock);
+	if (nvmet_ns_enabled(ns)) {
+		ret = -EBUSY;
+		goto out_unlock;
+	}
+
+	for (i = 0; i < 16; i++) {
+		if (p + 2 > page + count) {
+			ret = -EINVAL;
+			goto out_unlock;
+		}
+		if (!isxdigit(p[0]) || !isxdigit(p[1])) {
+			ret = -EINVAL;
+			goto out_unlock;
+		}
+
+		nguid[i] = (hex_to_bin(p[0]) << 4) | hex_to_bin(p[1]);
+		p += 2;
+
+		if (*p == '-' || *p == ':')
+			p++;
+	}
+
+	memcpy(&ns->nguid, nguid, sizeof(nguid));
+out_unlock:
+	mutex_unlock(&subsys->lock);
+	return ret ? ret : count;
+}
+
+CONFIGFS_ATTR(nvmet_ns_, device_nguid);
+
+static ssize_t nvmet_ns_enable_show(struct config_item *item, char *page)
+{
+	return sprintf(page, "%d\n", nvmet_ns_enabled(to_nvmet_ns(item)));
+}
+
+static ssize_t nvmet_ns_enable_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_ns *ns = to_nvmet_ns(item);
+	bool enable;
+	int ret = 0;
+
+	if (strtobool(page, &enable))
+		return -EINVAL;
+
+	if (enable)
+		ret = nvmet_ns_enable(ns);
+	else
+		nvmet_ns_disable(ns);
+
+	return ret ? ret : count;
+}
+
+CONFIGFS_ATTR(nvmet_ns_, enable);
+
+static struct configfs_attribute *nvmet_ns_attrs[] = {
+	&nvmet_ns_attr_device_path,
+	&nvmet_ns_attr_device_nguid,
+	&nvmet_ns_attr_enable,
+	NULL,
+};
+
+static void nvmet_ns_release(struct config_item *item)
+{
+	struct nvmet_ns *ns = to_nvmet_ns(item);
+
+	nvmet_ns_free(ns);
+}
+
+static struct configfs_item_operations nvmet_ns_item_ops = {
+	.release		= nvmet_ns_release,
+};
+
+static struct config_item_type nvmet_ns_type = {
+	.ct_item_ops		= &nvmet_ns_item_ops,
+	.ct_attrs		= nvmet_ns_attrs,
+	.ct_owner		= THIS_MODULE,
+};
+
+static struct config_group *nvmet_ns_make(struct config_group *group,
+		const char *name)
+{
+	struct nvmet_subsys *subsys = namespaces_to_subsys(&group->cg_item);
+	struct nvmet_ns *ns;
+	int ret;
+	u32 nsid;
+
+	ret = kstrtou32(name, 0, &nsid);
+	if (ret)
+		goto out;
+
+	ret = -EINVAL;
+	if (nsid == 0 || nsid == 0xffffffff)
+		goto out;
+
+	ret = -ENOMEM;
+	ns = nvmet_ns_alloc(subsys, nsid);
+	if (!ns)
+		goto out;
+	config_group_init_type_name(&ns->group, name, &nvmet_ns_type);
+
+	pr_info("adding nsid %d to subsystem %s\n", nsid, subsys->subsysnqn);
+
+	return &ns->group;
+out:
+	return ERR_PTR(ret);
+}
+
+static struct configfs_group_operations nvmet_namespaces_group_ops = {
+	.make_group		= nvmet_ns_make,
+};
+
+static struct config_item_type nvmet_namespaces_type = {
+	.ct_group_ops		= &nvmet_namespaces_group_ops,
+	.ct_owner		= THIS_MODULE,
+};
+
+static int nvmet_port_subsys_allow_link(struct config_item *parent,
+		struct config_item *target)
+{
+	struct nvmet_port *port = to_nvmet_port(parent->ci_parent);
+	struct nvmet_subsys *subsys;
+	struct nvmet_subsys_link *link, *p;
+	int ret;
+
+	if (target->ci_type != &nvmet_subsys_type) {
+		pr_err("can only link subsystems into the subsystems dir.!\n");
+		return -EINVAL;
+	}
+	subsys = to_subsys(target);
+	link = kmalloc(sizeof(*link), GFP_KERNEL);
+	if (!link)
+		return -ENOMEM;
+	link->subsys = subsys;
+
+	down_write(&nvmet_config_sem);
+	ret = -EEXIST;
+	list_for_each_entry(p, &port->subsystems, entry) {
+		if (p->subsys == subsys)
+			goto out_free_link;
+	}
+
+	if (list_empty(&port->subsystems)) {
+		ret = nvmet_enable_port(port);
+		if (ret)
+			goto out_free_link;
+	}
+
+	list_add_tail(&link->entry, &port->subsystems);
+	nvmet_genctr++;
+	up_write(&nvmet_config_sem);
+	return 0;
+
+out_free_link:
+	up_write(&nvmet_config_sem);
+	kfree(link);
+	return ret;
+}
+
+static int nvmet_port_subsys_drop_link(struct config_item *parent,
+		struct config_item *target)
+{
+	struct nvmet_port *port = to_nvmet_port(parent->ci_parent);
+	struct nvmet_subsys *subsys = to_subsys(target);
+	struct nvmet_subsys_link *p;
+
+	down_write(&nvmet_config_sem);
+	list_for_each_entry(p, &port->subsystems, entry) {
+		if (p->subsys == subsys)
+			goto found;
+	}
+	up_write(&nvmet_config_sem);
+	return -EINVAL;
+
+found:
+	list_del(&p->entry);
+	nvmet_genctr++;
+	if (list_empty(&port->subsystems))
+		nvmet_disable_port(port);
+	up_write(&nvmet_config_sem);
+	kfree(p);
+	return 0;
+}
+
+static struct configfs_item_operations nvmet_port_subsys_item_ops = {
+	.allow_link		= nvmet_port_subsys_allow_link,
+	.drop_link		= nvmet_port_subsys_drop_link,
+};
+
+static struct config_item_type nvmet_port_subsys_type = {
+	.ct_item_ops		= &nvmet_port_subsys_item_ops,
+	.ct_owner		= THIS_MODULE,
+};
+
+static int nvmet_allowed_hosts_allow_link(struct config_item *parent,
+		struct config_item *target)
+{
+	struct nvmet_subsys *subsys = to_subsys(parent->ci_parent);
+	struct nvmet_host *host;
+	struct nvmet_host_link *link, *p;
+	int ret;
+
+	if (target->ci_type != &nvmet_host_type) {
+		pr_err("can only link hosts into the allowed_hosts directory!\n");
+		return -EINVAL;
+	}
+
+	host = to_host(target);
+	link = kmalloc(sizeof(*link), GFP_KERNEL);
+	if (!link)
+		return -ENOMEM;
+	link->host = host;
+
+	down_write(&nvmet_config_sem);
+	ret = -EINVAL;
+	if (subsys->allow_any_host) {
+		pr_err("can't add hosts when allow_any_host is set!\n");
+		goto out_free_link;
+	}
+
+	ret = -EEXIST;
+	list_for_each_entry(p, &subsys->hosts, entry) {
+		if (!strcmp(nvmet_host_name(p->host), nvmet_host_name(host)))
+			goto out_free_link;
+	}
+	list_add_tail(&link->entry, &subsys->hosts);
+	nvmet_genctr++;
+	up_write(&nvmet_config_sem);
+	return 0;
+out_free_link:
+	up_write(&nvmet_config_sem);
+	kfree(link);
+	return ret;
+}
+
+static int nvmet_allowed_hosts_drop_link(struct config_item *parent,
+		struct config_item *target)
+{
+	struct nvmet_subsys *subsys = to_subsys(parent->ci_parent);
+	struct nvmet_host *host = to_host(target);
+	struct nvmet_host_link *p;
+
+	down_write(&nvmet_config_sem);
+	list_for_each_entry(p, &subsys->hosts, entry) {
+		if (!strcmp(nvmet_host_name(p->host), nvmet_host_name(host)))
+			goto found;
+	}
+	up_write(&nvmet_config_sem);
+	return -EINVAL;
+
+found:
+	list_del(&p->entry);
+	nvmet_genctr++;
+	up_write(&nvmet_config_sem);
+	kfree(p);
+	return 0;
+}
+
+static struct configfs_item_operations nvmet_allowed_hosts_item_ops = {
+	.allow_link		= nvmet_allowed_hosts_allow_link,
+	.drop_link		= nvmet_allowed_hosts_drop_link,
+};
+
+static struct config_item_type nvmet_allowed_hosts_type = {
+	.ct_item_ops		= &nvmet_allowed_hosts_item_ops,
+	.ct_owner		= THIS_MODULE,
+};
+
+static ssize_t nvmet_subsys_attr_allow_any_host_show(struct config_item *item,
+		char *page)
+{
+	return snprintf(page, PAGE_SIZE, "%d\n",
+		to_subsys(item)->allow_any_host);
+}
+
+static ssize_t nvmet_subsys_attr_allow_any_host_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_subsys *subsys = to_subsys(item);
+	bool allow_any_host;
+	int ret = 0;
+
+	if (strtobool(page, &allow_any_host))
+		return -EINVAL;
+
+	down_write(&nvmet_config_sem);
+	if (allow_any_host && !list_empty(&subsys->hosts)) {
+		pr_err("Can't set allow_any_host when explicit hosts are set!\n");
+		ret = -EINVAL;
+		goto out_unlock;
+	}
+
+	subsys->allow_any_host = allow_any_host;
+out_unlock:
+	up_write(&nvmet_config_sem);
+	return ret ? ret : count;
+}
+
+CONFIGFS_ATTR(nvmet_subsys_, attr_allow_any_host);
+
+static struct configfs_attribute *nvmet_subsys_attrs[] = {
+	&nvmet_subsys_attr_attr_allow_any_host,
+	NULL,
+};
+
+/*
+ * Subsystem structures & folder operation functions below
+ */
+static void nvmet_subsys_release(struct config_item *item)
+{
+	struct nvmet_subsys *subsys = to_subsys(item);
+
+	nvmet_subsys_put(subsys);
+}
+
+static struct configfs_item_operations nvmet_subsys_item_ops = {
+	.release		= nvmet_subsys_release,
+};
+
+static struct config_item_type nvmet_subsys_type = {
+	.ct_item_ops		= &nvmet_subsys_item_ops,
+	.ct_attrs		= nvmet_subsys_attrs,
+	.ct_owner		= THIS_MODULE,
+};
+
+static struct config_group *nvmet_subsys_make(struct config_group *group,
+		const char *name)
+{
+	struct nvmet_subsys *subsys;
+
+	if (sysfs_streq(name, NVME_DISC_SUBSYS_NAME)) {
+		pr_err("can't create discovery subsystem through configfs\n");
+		return ERR_PTR(-EINVAL);
+	}
+
+	subsys = nvmet_subsys_alloc(name, NVME_NQN_NVME);
+	if (!subsys)
+		return ERR_PTR(-ENOMEM);
+
+	config_group_init_type_name(&subsys->group, name, &nvmet_subsys_type);
+
+	config_group_init_type_name(&subsys->namespaces_group,
+			"namespaces", &nvmet_namespaces_type);
+	configfs_add_default_group(&subsys->namespaces_group, &subsys->group);
+
+	config_group_init_type_name(&subsys->allowed_hosts_group,
+			"allowed_hosts", &nvmet_allowed_hosts_type);
+	configfs_add_default_group(&subsys->allowed_hosts_group,
+			&subsys->group);
+
+	return &subsys->group;
+}
+
+static struct configfs_group_operations nvmet_subsystems_group_ops = {
+	.make_group		= nvmet_subsys_make,
+};
+
+static struct config_item_type nvmet_subsystems_type = {
+	.ct_group_ops		= &nvmet_subsystems_group_ops,
+	.ct_owner		= THIS_MODULE,
+};
+
+static ssize_t nvmet_referral_enable_show(struct config_item *item,
+		char *page)
+{
+	return snprintf(page, PAGE_SIZE, "%d\n", to_nvmet_port(item)->enabled);
+}
+
+static ssize_t nvmet_referral_enable_store(struct config_item *item,
+		const char *page, size_t count)
+{
+	struct nvmet_port *parent = to_nvmet_port(item->ci_parent->ci_parent);
+	struct nvmet_port *port = to_nvmet_port(item);
+	bool enable;
+
+	if (strtobool(page, &enable))
+		goto inval;
+
+	if (enable)
+		nvmet_referral_enable(parent, port);
+	else
+		nvmet_referral_disable(port);
+
+	return count;
+inval:
+	pr_err("Invalid value '%s' for enable\n", page);
+	return -EINVAL;
+}
+
+CONFIGFS_ATTR(nvmet_referral_, enable);
+
+/*
+ * Discovery Service subsystem definitions
+ */
+static struct configfs_attribute *nvmet_referral_attrs[] = {
+	&nvmet_attr_addr_adrfam,
+	&nvmet_attr_addr_portid,
+	&nvmet_attr_addr_treq,
+	&nvmet_attr_addr_traddr,
+	&nvmet_attr_addr_trsvcid,
+	&nvmet_attr_addr_trtype,
+	&nvmet_referral_attr_enable,
+	NULL,
+};
+
+static void nvmet_referral_release(struct config_item *item)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+
+	nvmet_referral_disable(port);
+	kfree(port);
+}
+
+static struct configfs_item_operations nvmet_referral_item_ops = {
+	.release	= nvmet_referral_release,
+};
+
+static struct config_item_type nvmet_referral_type = {
+	.ct_owner	= THIS_MODULE,
+	.ct_attrs	= nvmet_referral_attrs,
+	.ct_item_ops	= &nvmet_referral_item_ops,
+};
+
+static struct config_group *nvmet_referral_make(
+		struct config_group *group, const char *name)
+{
+	struct nvmet_port *port;
+
+	port = kzalloc(sizeof(*port), GFP_KERNEL);
+	if (!port)
+		return ERR_CAST(port);
+
+	INIT_LIST_HEAD(&port->entry);
+	config_group_init_type_name(&port->group, name, &nvmet_referral_type);
+
+	return &port->group;
+}
+
+static struct configfs_group_operations nvmet_referral_group_ops = {
+	.make_group		= nvmet_referral_make,
+};
+
+static struct config_item_type nvmet_referrals_type = {
+	.ct_owner	= THIS_MODULE,
+	.ct_group_ops	= &nvmet_referral_group_ops,
+};
+
+/*
+ * Ports definitions.
+ */
+static void nvmet_port_release(struct config_item *item)
+{
+	struct nvmet_port *port = to_nvmet_port(item);
+
+	kfree(port);
+}
+
+static struct configfs_attribute *nvmet_port_attrs[] = {
+	&nvmet_attr_addr_adrfam,
+	&nvmet_attr_addr_treq,
+	&nvmet_attr_addr_traddr,
+	&nvmet_attr_addr_trsvcid,
+	&nvmet_attr_addr_trtype,
+	NULL,
+};
+
+static struct configfs_item_operations nvmet_port_item_ops = {
+	.release		= nvmet_port_release,
+};
+
+static struct config_item_type nvmet_port_type = {
+	.ct_attrs		= nvmet_port_attrs,
+	.ct_item_ops		= &nvmet_port_item_ops,
+	.ct_owner		= THIS_MODULE,
+};
+
+static struct config_group *nvmet_ports_make(struct config_group *group,
+		const char *name)
+{
+	struct nvmet_port *port;
+	u16 portid;
+
+	if (kstrtou16(name, 0, &portid))
+		return ERR_PTR(-EINVAL);
+
+	port = kzalloc(sizeof(*port), GFP_KERNEL);
+	if (!port)
+		return ERR_CAST(port);
+
+	INIT_LIST_HEAD(&port->entry);
+	INIT_LIST_HEAD(&port->subsystems);
+	INIT_LIST_HEAD(&port->referrals);
+
+	port->disc_addr.portid = cpu_to_le16(portid);
+	config_group_init_type_name(&port->group, name, &nvmet_port_type);
+
+	config_group_init_type_name(&port->subsys_group,
+			"subsystems", &nvmet_port_subsys_type);
+	configfs_add_default_group(&port->subsys_group, &port->group);
+
+	config_group_init_type_name(&port->referrals_group,
+			"referrals", &nvmet_referrals_type);
+	configfs_add_default_group(&port->referrals_group, &port->group);
+
+	return &port->group;
+}
+
+static struct configfs_group_operations nvmet_ports_group_ops = {
+	.make_group		= nvmet_ports_make,
+};
+
+static struct config_item_type nvmet_ports_type = {
+	.ct_group_ops		= &nvmet_ports_group_ops,
+	.ct_owner		= THIS_MODULE,
+};
+
+static struct config_group nvmet_subsystems_group;
+static struct config_group nvmet_ports_group;
+
+static void nvmet_host_release(struct config_item *item)
+{
+	struct nvmet_host *host = to_host(item);
+
+	kfree(host);
+}
+
+static struct configfs_item_operations nvmet_host_item_ops = {
+	.release		= nvmet_host_release,
+};
+
+static struct config_item_type nvmet_host_type = {
+	.ct_item_ops		= &nvmet_host_item_ops,
+	.ct_owner		= THIS_MODULE,
+};
+
+static struct config_group *nvmet_hosts_make_group(struct config_group *group,
+		const char *name)
+{
+	struct nvmet_host *host;
+
+	host = kzalloc(sizeof(*host), GFP_KERNEL);
+	if (!host)
+		return ERR_PTR(-ENOMEM);
+
+	config_group_init_type_name(&host->group, name, &nvmet_host_type);
+
+	return &host->group;
+}
+
+static struct configfs_group_operations nvmet_hosts_group_ops = {
+	.make_group		= nvmet_hosts_make_group,
+};
+
+static struct config_item_type nvmet_hosts_type = {
+	.ct_group_ops		= &nvmet_hosts_group_ops,
+	.ct_owner		= THIS_MODULE,
+};
+
+static struct config_group nvmet_hosts_group;
+
+static struct config_item_type nvmet_root_type = {
+	.ct_owner		= THIS_MODULE,
+};
+
+static struct configfs_subsystem nvmet_configfs_subsystem = {
+	.su_group = {
+		.cg_item = {
+			.ci_namebuf	= "nvmet",
+			.ci_type	= &nvmet_root_type,
+		},
+	},
+};
+
+int __init nvmet_init_configfs(void)
+{
+	int ret;
+
+	config_group_init(&nvmet_configfs_subsystem.su_group);
+	mutex_init(&nvmet_configfs_subsystem.su_mutex);
+
+	config_group_init_type_name(&nvmet_subsystems_group,
+			"subsystems", &nvmet_subsystems_type);
+	configfs_add_default_group(&nvmet_subsystems_group,
+			&nvmet_configfs_subsystem.su_group);
+
+	config_group_init_type_name(&nvmet_ports_group,
+			"ports", &nvmet_ports_type);
+	configfs_add_default_group(&nvmet_ports_group,
+			&nvmet_configfs_subsystem.su_group);
+
+	config_group_init_type_name(&nvmet_hosts_group,
+			"hosts", &nvmet_hosts_type);
+	configfs_add_default_group(&nvmet_hosts_group,
+			&nvmet_configfs_subsystem.su_group);
+
+	ret = configfs_register_subsystem(&nvmet_configfs_subsystem);
+	if (ret) {
+		pr_err("configfs_register_subsystem: %d\n", ret);
+		return ret;
+	}
+
+	return 0;
+}
+
+void __exit nvmet_exit_configfs(void)
+{
+	configfs_unregister_subsystem(&nvmet_configfs_subsystem);
+}

drivers/nvme/target/core.c

@@ -0,0 +1,964 @@
+/*
+ * Common code for the NVMe target.
+ * Copyright (c) 2015-2016 HGST, a Western Digital Company.
+ *
+ * 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.
+ */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+#include <linux/module.h>
+#include "nvmet.h"
+
+static struct nvmet_fabrics_ops *nvmet_transports[NVMF_TRTYPE_MAX];
+
+/*
+ * This read/write semaphore is used to synchronize access to configuration
+ * information on a target system that will result in discovery log page
+ * information change for at least one host.
+ * The full list of resources to protected by this semaphore is:
+ *
+ *  - subsystems list
+ *  - per-subsystem allowed hosts list
+ *  - allow_any_host subsystem attribute
+ *  - nvmet_genctr
+ *  - the nvmet_transports array
+ *
+ * When updating any of those lists/structures write lock should be obtained,
+ * while when reading (popolating discovery log page or checking host-subsystem
+ * link) read lock is obtained to allow concurrent reads.
+ */
+DECLARE_RWSEM(nvmet_config_sem);
+
+static struct nvmet_subsys *nvmet_find_get_subsys(struct nvmet_port *port,
+		const char *subsysnqn);
+
+u16 nvmet_copy_to_sgl(struct nvmet_req *req, off_t off, const void *buf,
+		size_t len)
+{
+	if (sg_pcopy_from_buffer(req->sg, req->sg_cnt, buf, len, off) != len)
+		return NVME_SC_SGL_INVALID_DATA | NVME_SC_DNR;
+	return 0;
+}
+
+u16 nvmet_copy_from_sgl(struct nvmet_req *req, off_t off, void *buf, size_t len)
+{
+	if (sg_pcopy_to_buffer(req->sg, req->sg_cnt, buf, len, off) != len)
+		return NVME_SC_SGL_INVALID_DATA | NVME_SC_DNR;
+	return 0;
+}
+
+static u32 nvmet_async_event_result(struct nvmet_async_event *aen)
+{
+	return aen->event_type | (aen->event_info << 8) | (aen->log_page << 16);
+}
+
+static void nvmet_async_events_free(struct nvmet_ctrl *ctrl)
+{
+	struct nvmet_req *req;
+
+	while (1) {
+		mutex_lock(&ctrl->lock);
+		if (!ctrl->nr_async_event_cmds) {
+			mutex_unlock(&ctrl->lock);
+			return;
+		}
+
+		req = ctrl->async_event_cmds[--ctrl->nr_async_event_cmds];
+		mutex_unlock(&ctrl->lock);
+		nvmet_req_complete(req, NVME_SC_INTERNAL | NVME_SC_DNR);
+	}
+}
+
+static void nvmet_async_event_work(struct work_struct *work)
+{
+	struct nvmet_ctrl *ctrl =
+		container_of(work, struct nvmet_ctrl, async_event_work);
+	struct nvmet_async_event *aen;
+	struct nvmet_req *req;
+
+	while (1) {
+		mutex_lock(&ctrl->lock);
+		aen = list_first_entry_or_null(&ctrl->async_events,
+				struct nvmet_async_event, entry);
+		if (!aen || !ctrl->nr_async_event_cmds) {
+			mutex_unlock(&ctrl->lock);
+			return;
+		}
+
+		req = ctrl->async_event_cmds[--ctrl->nr_async_event_cmds];
+		nvmet_set_result(req, nvmet_async_event_result(aen));
+
+		list_del(&aen->entry);
+		kfree(aen);
+
+		mutex_unlock(&ctrl->lock);
+		nvmet_req_complete(req, 0);
+	}
+}
+
+static void nvmet_add_async_event(struct nvmet_ctrl *ctrl, u8 event_type,
+		u8 event_info, u8 log_page)
+{
+	struct nvmet_async_event *aen;
+
+	aen = kmalloc(sizeof(*aen), GFP_KERNEL);
+	if (!aen)
+		return;
+
+	aen->event_type = event_type;
+	aen->event_info = event_info;
+	aen->log_page = log_page;
+
+	mutex_lock(&ctrl->lock);
+	list_add_tail(&aen->entry, &ctrl->async_events);
+	mutex_unlock(&ctrl->lock);
+
+	schedule_work(&ctrl->async_event_work);
+}
+
+int nvmet_register_transport(struct nvmet_fabrics_ops *ops)
+{
+	int ret = 0;
+
+	down_write(&nvmet_config_sem);
+	if (nvmet_transports[ops->type])
+		ret = -EINVAL;
+	else
+		nvmet_transports[ops->type] = ops;
+	up_write(&nvmet_config_sem);
+
+	return ret;
+}
+EXPORT_SYMBOL_GPL(nvmet_register_transport);
+
+void nvmet_unregister_transport(struct nvmet_fabrics_ops *ops)
+{
+	down_write(&nvmet_config_sem);
+	nvmet_transports[ops->type] = NULL;
+	up_write(&nvmet_config_sem);
+}
+EXPORT_SYMBOL_GPL(nvmet_unregister_transport);
+
+int nvmet_enable_port(struct nvmet_port *port)
+{
+	struct nvmet_fabrics_ops *ops;
+	int ret;
+
+	lockdep_assert_held(&nvmet_config_sem);
+
+	ops = nvmet_transports[port->disc_addr.trtype];
+	if (!ops) {
+		up_write(&nvmet_config_sem);
+		request_module("nvmet-transport-%d", port->disc_addr.trtype);
+		down_write(&nvmet_config_sem);
+		ops = nvmet_transports[port->disc_addr.trtype];
+		if (!ops) {
+			pr_err("transport type %d not supported\n",
+				port->disc_addr.trtype);
+			return -EINVAL;
+		}
+	}
+
+	if (!try_module_get(ops->owner))
+		return -EINVAL;
+
+	ret = ops->add_port(port);
+	if (ret) {
+		module_put(ops->owner);
+		return ret;
+	}
+
+	port->enabled = true;
+	return 0;
+}
+
+void nvmet_disable_port(struct nvmet_port *port)
+{
+	struct nvmet_fabrics_ops *ops;
+
+	lockdep_assert_held(&nvmet_config_sem);
+
+	port->enabled = false;
+
+	ops = nvmet_transports[port->disc_addr.trtype];
+	ops->remove_port(port);
+	module_put(ops->owner);
+}
+
+static void nvmet_keep_alive_timer(struct work_struct *work)
+{
+	struct nvmet_ctrl *ctrl = container_of(to_delayed_work(work),
+			struct nvmet_ctrl, ka_work);
+
+	pr_err("ctrl %d keep-alive timer (%d seconds) expired!\n",
+		ctrl->cntlid, ctrl->kato);
+
+	ctrl->ops->delete_ctrl(ctrl);
+}
+
+static void nvmet_start_keep_alive_timer(struct nvmet_ctrl *ctrl)
+{
+	pr_debug("ctrl %d start keep-alive timer for %d secs\n",
+		ctrl->cntlid, ctrl->kato);
+
+	INIT_DELAYED_WORK(&ctrl->ka_work, nvmet_keep_alive_timer);
+	schedule_delayed_work(&ctrl->ka_work, ctrl->kato * HZ);
+}
+
+static void nvmet_stop_keep_alive_timer(struct nvmet_ctrl *ctrl)
+{
+	pr_debug("ctrl %d stop keep-alive\n", ctrl->cntlid);
+
+	cancel_delayed_work_sync(&ctrl->ka_work);
+}
+
+static struct nvmet_ns *__nvmet_find_namespace(struct nvmet_ctrl *ctrl,
+		__le32 nsid)
+{
+	struct nvmet_ns *ns;
+
+	list_for_each_entry_rcu(ns, &ctrl->subsys->namespaces, dev_link) {
+		if (ns->nsid == le32_to_cpu(nsid))
+			return ns;
+	}
+
+	return NULL;
+}
+
+struct nvmet_ns *nvmet_find_namespace(struct nvmet_ctrl *ctrl, __le32 nsid)
+{
+	struct nvmet_ns *ns;
+
+	rcu_read_lock();
+	ns = __nvmet_find_namespace(ctrl, nsid);
+	if (ns)
+		percpu_ref_get(&ns->ref);
+	rcu_read_unlock();
+
+	return ns;
+}
+
+static void nvmet_destroy_namespace(struct percpu_ref *ref)
+{
+	struct nvmet_ns *ns = container_of(ref, struct nvmet_ns, ref);
+
+	complete(&ns->disable_done);
+}
+
+void nvmet_put_namespace(struct nvmet_ns *ns)
+{
+	percpu_ref_put(&ns->ref);
+}
+
+int nvmet_ns_enable(struct nvmet_ns *ns)
+{
+	struct nvmet_subsys *subsys = ns->subsys;
+	struct nvmet_ctrl *ctrl;
+	int ret = 0;
+
+	mutex_lock(&subsys->lock);
+	if (!list_empty(&ns->dev_link))
+		goto out_unlock;
+
+	ns->bdev = blkdev_get_by_path(ns->device_path, FMODE_READ | FMODE_WRITE,
+			NULL);
+	if (IS_ERR(ns->bdev)) {
+		pr_err("nvmet: failed to open block device %s: (%ld)\n",
+			ns->device_path, PTR_ERR(ns->bdev));
+		ret = PTR_ERR(ns->bdev);
+		ns->bdev = NULL;
+		goto out_unlock;
+	}
+
+	ns->size = i_size_read(ns->bdev->bd_inode);
+	ns->blksize_shift = blksize_bits(bdev_logical_block_size(ns->bdev));
+
+	ret = percpu_ref_init(&ns->ref, nvmet_destroy_namespace,
+				0, GFP_KERNEL);
+	if (ret)
+		goto out_blkdev_put;
+
+	if (ns->nsid > subsys->max_nsid)
+		subsys->max_nsid = ns->nsid;
+
+	/*
+	 * The namespaces list needs to be sorted to simplify the implementation
+	 * of the Identify Namepace List subcommand.
+	 */
+	if (list_empty(&subsys->namespaces)) {
+		list_add_tail_rcu(&ns->dev_link, &subsys->namespaces);
+	} else {
+		struct nvmet_ns *old;
+
+		list_for_each_entry_rcu(old, &subsys->namespaces, dev_link) {
+			BUG_ON(ns->nsid == old->nsid);
+			if (ns->nsid < old->nsid)
+				break;
+		}
+
+		list_add_tail_rcu(&ns->dev_link, &old->dev_link);
+	}
+
+	list_for_each_entry(ctrl, &subsys->ctrls, subsys_entry)
+		nvmet_add_async_event(ctrl, NVME_AER_TYPE_NOTICE, 0, 0);
+
+	ret = 0;
+out_unlock:
+	mutex_unlock(&subsys->lock);
+	return ret;
+out_blkdev_put:
+	blkdev_put(ns->bdev, FMODE_WRITE|FMODE_READ);
+	ns->bdev = NULL;
+	goto out_unlock;
+}
+
+void nvmet_ns_disable(struct nvmet_ns *ns)
+{
+	struct nvmet_subsys *subsys = ns->subsys;
+	struct nvmet_ctrl *ctrl;
+
+	mutex_lock(&subsys->lock);
+	if (list_empty(&ns->dev_link)) {
+		mutex_unlock(&subsys->lock);
+		return;
+	}
+	list_del_init(&ns->dev_link);
+	mutex_unlock(&subsys->lock);
+
+	/*
+	 * Now that we removed the namespaces from the lookup list, we
+	 * can kill the per_cpu ref and wait for any remaining references
+	 * to be dropped, as well as a RCU grace period for anyone only
+	 * using the namepace under rcu_read_lock().  Note that we can't
+	 * use call_rcu here as we need to ensure the namespaces have
+	 * been fully destroyed before unloading the module.
+	 */
+	percpu_ref_kill(&ns->ref);
+	synchronize_rcu();
+	wait_for_completion(&ns->disable_done);
+	percpu_ref_exit(&ns->ref);
+
+	mutex_lock(&subsys->lock);
+	list_for_each_entry(ctrl, &subsys->ctrls, subsys_entry)
+		nvmet_add_async_event(ctrl, NVME_AER_TYPE_NOTICE, 0, 0);
+
+	if (ns->bdev)
+		blkdev_put(ns->bdev, FMODE_WRITE|FMODE_READ);
+	mutex_unlock(&subsys->lock);
+}
+
+void nvmet_ns_free(struct nvmet_ns *ns)
+{
+	nvmet_ns_disable(ns);
+
+	kfree(ns->device_path);
+	kfree(ns);
+}
+
+struct nvmet_ns *nvmet_ns_alloc(struct nvmet_subsys *subsys, u32 nsid)
+{
+	struct nvmet_ns *ns;
+
+	ns = kzalloc(sizeof(*ns), GFP_KERNEL);
+	if (!ns)
+		return NULL;
+
+	INIT_LIST_HEAD(&ns->dev_link);
+	init_completion(&ns->disable_done);
+
+	ns->nsid = nsid;
+	ns->subsys = subsys;
+
+	return ns;
+}
+
+static void __nvmet_req_complete(struct nvmet_req *req, u16 status)
+{
+	if (status)
+		nvmet_set_status(req, status);
+
+	/* XXX: need to fill in something useful for sq_head */
+	req->rsp->sq_head = 0;
+	if (likely(req->sq)) /* may happen during early failure */
+		req->rsp->sq_id = cpu_to_le16(req->sq->qid);
+	req->rsp->command_id = req->cmd->common.command_id;
+
+	if (req->ns)
+		nvmet_put_namespace(req->ns);
+	req->ops->queue_response(req);
+}
+
+void nvmet_req_complete(struct nvmet_req *req, u16 status)
+{
+	__nvmet_req_complete(req, status);
+	percpu_ref_put(&req->sq->ref);
+}
+EXPORT_SYMBOL_GPL(nvmet_req_complete);
+
+void nvmet_cq_setup(struct nvmet_ctrl *ctrl, struct nvmet_cq *cq,
+		u16 qid, u16 size)
+{
+	cq->qid = qid;
+	cq->size = size;
+
+	ctrl->cqs[qid] = cq;
+}
+
+void nvmet_sq_setup(struct nvmet_ctrl *ctrl, struct nvmet_sq *sq,
+		u16 qid, u16 size)
+{
+	sq->qid = qid;
+	sq->size = size;
+
+	ctrl->sqs[qid] = sq;
+}
+
+void nvmet_sq_destroy(struct nvmet_sq *sq)
+{
+	/*
+	 * If this is the admin queue, complete all AERs so that our
+	 * queue doesn't have outstanding requests on it.
+	 */
+	if (sq->ctrl && sq->ctrl->sqs && sq->ctrl->sqs[0] == sq)
+		nvmet_async_events_free(sq->ctrl);
+	percpu_ref_kill(&sq->ref);
+	wait_for_completion(&sq->free_done);
+	percpu_ref_exit(&sq->ref);
+
+	if (sq->ctrl) {
+		nvmet_ctrl_put(sq->ctrl);
+		sq->ctrl = NULL; /* allows reusing the queue later */
+	}
+}
+EXPORT_SYMBOL_GPL(nvmet_sq_destroy);
+
+static void nvmet_sq_free(struct percpu_ref *ref)
+{
+	struct nvmet_sq *sq = container_of(ref, struct nvmet_sq, ref);
+
+	complete(&sq->free_done);
+}
+
+int nvmet_sq_init(struct nvmet_sq *sq)
+{
+	int ret;
+
+	ret = percpu_ref_init(&sq->ref, nvmet_sq_free, 0, GFP_KERNEL);
+	if (ret) {
+		pr_err("percpu_ref init failed!\n");
+		return ret;
+	}
+	init_completion(&sq->free_done);
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(nvmet_sq_init);
+
+bool nvmet_req_init(struct nvmet_req *req, struct nvmet_cq *cq,
+		struct nvmet_sq *sq, struct nvmet_fabrics_ops *ops)
+{
+	u8 flags = req->cmd->common.flags;
+	u16 status;
+
+	req->cq = cq;
+	req->sq = sq;
+	req->ops = ops;
+	req->sg = NULL;
+	req->sg_cnt = 0;
+	req->rsp->status = 0;
+
+	/* no support for fused commands yet */
+	if (unlikely(flags & (NVME_CMD_FUSE_FIRST | NVME_CMD_FUSE_SECOND))) {
+		status = NVME_SC_INVALID_FIELD | NVME_SC_DNR;
+		goto fail;
+	}
+
+	/* either variant of SGLs is fine, as we don't support metadata */
+	if (unlikely((flags & NVME_CMD_SGL_ALL) != NVME_CMD_SGL_METABUF &&
+		     (flags & NVME_CMD_SGL_ALL) != NVME_CMD_SGL_METASEG)) {
+		status = NVME_SC_INVALID_FIELD | NVME_SC_DNR;
+		goto fail;
+	}
+
+	if (unlikely(!req->sq->ctrl))
+		/* will return an error for any Non-connect command: */
+		status = nvmet_parse_connect_cmd(req);
+	else if (likely(req->sq->qid != 0))
+		status = nvmet_parse_io_cmd(req);
+	else if (req->cmd->common.opcode == nvme_fabrics_command)
+		status = nvmet_parse_fabrics_cmd(req);
+	else if (req->sq->ctrl->subsys->type == NVME_NQN_DISC)
+		status = nvmet_parse_discovery_cmd(req);
+	else
+		status = nvmet_parse_admin_cmd(req);
+
+	if (status)
+		goto fail;
+
+	if (unlikely(!percpu_ref_tryget_live(&sq->ref))) {
+		status = NVME_SC_INVALID_FIELD | NVME_SC_DNR;
+		goto fail;
+	}
+
+	return true;
+
+fail:
+	__nvmet_req_complete(req, status);
+	return false;
+}
+EXPORT_SYMBOL_GPL(nvmet_req_init);
+
+static inline bool nvmet_cc_en(u32 cc)
+{
+	return cc & 0x1;
+}
+
+static inline u8 nvmet_cc_css(u32 cc)
+{
+	return (cc >> 4) & 0x7;
+}
+
+static inline u8 nvmet_cc_mps(u32 cc)
+{
+	return (cc >> 7) & 0xf;
+}
+
+static inline u8 nvmet_cc_ams(u32 cc)
+{
+	return (cc >> 11) & 0x7;
+}
+
+static inline u8 nvmet_cc_shn(u32 cc)
+{
+	return (cc >> 14) & 0x3;
+}
+
+static inline u8 nvmet_cc_iosqes(u32 cc)
+{
+	return (cc >> 16) & 0xf;
+}
+
+static inline u8 nvmet_cc_iocqes(u32 cc)
+{
+	return (cc >> 20) & 0xf;
+}
+
+static void nvmet_start_ctrl(struct nvmet_ctrl *ctrl)
+{
+	lockdep_assert_held(&ctrl->lock);
+
+	if (nvmet_cc_iosqes(ctrl->cc) != NVME_NVM_IOSQES ||
+	    nvmet_cc_iocqes(ctrl->cc) != NVME_NVM_IOCQES ||
+	    nvmet_cc_mps(ctrl->cc) != 0 ||
+	    nvmet_cc_ams(ctrl->cc) != 0 ||
+	    nvmet_cc_css(ctrl->cc) != 0) {
+		ctrl->csts = NVME_CSTS_CFS;
+		return;
+	}
+
+	ctrl->csts = NVME_CSTS_RDY;
+}
+
+static void nvmet_clear_ctrl(struct nvmet_ctrl *ctrl)
+{
+	lockdep_assert_held(&ctrl->lock);
+
+	/* XXX: tear down queues? */
+	ctrl->csts &= ~NVME_CSTS_RDY;
+	ctrl->cc = 0;
+}
+
+void nvmet_update_cc(struct nvmet_ctrl *ctrl, u32 new)
+{
+	u32 old;
+
+	mutex_lock(&ctrl->lock);
+	old = ctrl->cc;
+	ctrl->cc = new;
+
+	if (nvmet_cc_en(new) && !nvmet_cc_en(old))
+		nvmet_start_ctrl(ctrl);
+	if (!nvmet_cc_en(new) && nvmet_cc_en(old))
+		nvmet_clear_ctrl(ctrl);
+	if (nvmet_cc_shn(new) && !nvmet_cc_shn(old)) {
+		nvmet_clear_ctrl(ctrl);
+		ctrl->csts |= NVME_CSTS_SHST_CMPLT;
+	}
+	if (!nvmet_cc_shn(new) && nvmet_cc_shn(old))
+		ctrl->csts &= ~NVME_CSTS_SHST_CMPLT;
+	mutex_unlock(&ctrl->lock);
+}
+
+static void nvmet_init_cap(struct nvmet_ctrl *ctrl)
+{
+	/* command sets supported: NVMe command set: */
+	ctrl->cap = (1ULL << 37);
+	/* CC.EN timeout in 500msec units: */
+	ctrl->cap |= (15ULL << 24);
+	/* maximum queue entries supported: */
+	ctrl->cap |= NVMET_QUEUE_SIZE - 1;
+}
+
+u16 nvmet_ctrl_find_get(const char *subsysnqn, const char *hostnqn, u16 cntlid,
+		struct nvmet_req *req, struct nvmet_ctrl **ret)
+{
+	struct nvmet_subsys *subsys;
+	struct nvmet_ctrl *ctrl;
+	u16 status = 0;
+
+	subsys = nvmet_find_get_subsys(req->port, subsysnqn);
+	if (!subsys) {
+		pr_warn("connect request for invalid subsystem %s!\n",
+			subsysnqn);
+		req->rsp->result = IPO_IATTR_CONNECT_DATA(subsysnqn);
+		return NVME_SC_CONNECT_INVALID_PARAM | NVME_SC_DNR;
+	}
+
+	mutex_lock(&subsys->lock);
+	list_for_each_entry(ctrl, &subsys->ctrls, subsys_entry) {
+		if (ctrl->cntlid == cntlid) {
+			if (strncmp(hostnqn, ctrl->hostnqn, NVMF_NQN_SIZE)) {
+				pr_warn("hostnqn mismatch.\n");
+				continue;
+			}
+			if (!kref_get_unless_zero(&ctrl->ref))
+				continue;
+
+			*ret = ctrl;
+			goto out;
+		}
+	}
+
+	pr_warn("could not find controller %d for subsys %s / host %s\n",
+		cntlid, subsysnqn, hostnqn);
+	req->rsp->result = IPO_IATTR_CONNECT_DATA(cntlid);
+	status = NVME_SC_CONNECT_INVALID_PARAM | NVME_SC_DNR;
+
+out:
+	mutex_unlock(&subsys->lock);
+	nvmet_subsys_put(subsys);
+	return status;
+}
+
+static bool __nvmet_host_allowed(struct nvmet_subsys *subsys,
+		const char *hostnqn)
+{
+	struct nvmet_host_link *p;
+
+	if (subsys->allow_any_host)
+		return true;
+
+	list_for_each_entry(p, &subsys->hosts, entry) {
+		if (!strcmp(nvmet_host_name(p->host), hostnqn))
+			return true;
+	}
+
+	return false;
+}
+
+static bool nvmet_host_discovery_allowed(struct nvmet_req *req,
+		const char *hostnqn)
+{
+	struct nvmet_subsys_link *s;
+
+	list_for_each_entry(s, &req->port->subsystems, entry) {
+		if (__nvmet_host_allowed(s->subsys, hostnqn))
+			return true;
+	}
+
+	return false;
+}
+
+bool nvmet_host_allowed(struct nvmet_req *req, struct nvmet_subsys *subsys,
+		const char *hostnqn)
+{
+	lockdep_assert_held(&nvmet_config_sem);
+
+	if (subsys->type == NVME_NQN_DISC)
+		return nvmet_host_discovery_allowed(req, hostnqn);
+	else
+		return __nvmet_host_allowed(subsys, hostnqn);
+}
+
+u16 nvmet_alloc_ctrl(const char *subsysnqn, const char *hostnqn,
+		struct nvmet_req *req, u32 kato, struct nvmet_ctrl **ctrlp)
+{
+	struct nvmet_subsys *subsys;
+	struct nvmet_ctrl *ctrl;
+	int ret;
+	u16 status;
+
+	status = NVME_SC_CONNECT_INVALID_PARAM | NVME_SC_DNR;
+	subsys = nvmet_find_get_subsys(req->port, subsysnqn);
+	if (!subsys) {
+		pr_warn("connect request for invalid subsystem %s!\n",
+			subsysnqn);
+		req->rsp->result = IPO_IATTR_CONNECT_DATA(subsysnqn);
+		goto out;
+	}
+
+	status = NVME_SC_CONNECT_INVALID_PARAM | NVME_SC_DNR;
+	down_read(&nvmet_config_sem);
+	if (!nvmet_host_allowed(req, subsys, hostnqn)) {
+		pr_info("connect by host %s for subsystem %s not allowed\n",
+			hostnqn, subsysnqn);
+		req->rsp->result = IPO_IATTR_CONNECT_DATA(hostnqn);
+		up_read(&nvmet_config_sem);
+		goto out_put_subsystem;
+	}
+	up_read(&nvmet_config_sem);
+
+	status = NVME_SC_INTERNAL;
+	ctrl = kzalloc(sizeof(*ctrl), GFP_KERNEL);
+	if (!ctrl)
+		goto out_put_subsystem;
+	mutex_init(&ctrl->lock);
+
+	nvmet_init_cap(ctrl);
+
+	INIT_WORK(&ctrl->async_event_work, nvmet_async_event_work);
+	INIT_LIST_HEAD(&ctrl->async_events);
+
+	memcpy(ctrl->subsysnqn, subsysnqn, NVMF_NQN_SIZE);
+	memcpy(ctrl->hostnqn, hostnqn, NVMF_NQN_SIZE);
+
+	kref_init(&ctrl->ref);
+	ctrl->subsys = subsys;
+
+	ctrl->cqs = kcalloc(subsys->max_qid + 1,
+			sizeof(struct nvmet_cq *),
+			GFP_KERNEL);
+	if (!ctrl->cqs)
+		goto out_free_ctrl;
+
+	ctrl->sqs = kcalloc(subsys->max_qid + 1,
+			sizeof(struct nvmet_sq *),
+			GFP_KERNEL);
+	if (!ctrl->sqs)
+		goto out_free_cqs;
+
+	ret = ida_simple_get(&subsys->cntlid_ida,
+			     NVME_CNTLID_MIN, NVME_CNTLID_MAX,
+			     GFP_KERNEL);
+	if (ret < 0) {
+		status = NVME_SC_CONNECT_CTRL_BUSY | NVME_SC_DNR;
+		goto out_free_sqs;
+	}
+	ctrl->cntlid = ret;
+
+	ctrl->ops = req->ops;
+	if (ctrl->subsys->type == NVME_NQN_DISC) {
+		/* Don't accept keep-alive timeout for discovery controllers */
+		if (kato) {
+			status = NVME_SC_INVALID_FIELD | NVME_SC_DNR;
+			goto out_free_sqs;
+		}
+
+		/*
+		 * Discovery controllers use some arbitrary high value in order
+		 * to cleanup stale discovery sessions
+		 *
+		 * From the latest base diff RC:
+		 * "The Keep Alive command is not supported by
+		 * Discovery controllers. A transport may specify a
+		 * fixed Discovery controller activity timeout value
+		 * (e.g., 2 minutes).  If no commands are received
+		 * by a Discovery controller within that time
+		 * period, the controller may perform the
+		 * actions for Keep Alive Timer expiration".
+		 */
+		ctrl->kato = NVMET_DISC_KATO;
+	} else {
+		/* keep-alive timeout in seconds */
+		ctrl->kato = DIV_ROUND_UP(kato, 1000);
+	}
+	nvmet_start_keep_alive_timer(ctrl);
+
+	mutex_lock(&subsys->lock);
+	list_add_tail(&ctrl->subsys_entry, &subsys->ctrls);
+	mutex_unlock(&subsys->lock);
+
+	*ctrlp = ctrl;
+	return 0;
+
+out_free_sqs:
+	kfree(ctrl->sqs);
+out_free_cqs:
+	kfree(ctrl->cqs);
+out_free_ctrl:
+	kfree(ctrl);
+out_put_subsystem:
+	nvmet_subsys_put(subsys);
+out:
+	return status;
+}
+
+static void nvmet_ctrl_free(struct kref *ref)
+{
+	struct nvmet_ctrl *ctrl = container_of(ref, struct nvmet_ctrl, ref);
+	struct nvmet_subsys *subsys = ctrl->subsys;
+
+	nvmet_stop_keep_alive_timer(ctrl);
+
+	mutex_lock(&subsys->lock);
+	list_del(&ctrl->subsys_entry);
+	mutex_unlock(&subsys->lock);
+
+	ida_simple_remove(&subsys->cntlid_ida, ctrl->cntlid);
+	nvmet_subsys_put(subsys);
+
+	kfree(ctrl->sqs);
+	kfree(ctrl->cqs);
+	kfree(ctrl);
+}
+
+void nvmet_ctrl_put(struct nvmet_ctrl *ctrl)
+{
+	kref_put(&ctrl->ref, nvmet_ctrl_free);
+}
+
+static void nvmet_fatal_error_handler(struct work_struct *work)
+{
+	struct nvmet_ctrl *ctrl =
+			container_of(work, struct nvmet_ctrl, fatal_err_work);
+
+	pr_err("ctrl %d fatal error occurred!\n", ctrl->cntlid);
+	ctrl->ops->delete_ctrl(ctrl);
+}
+
+void nvmet_ctrl_fatal_error(struct nvmet_ctrl *ctrl)
+{
+	ctrl->csts |= NVME_CSTS_CFS;
+	INIT_WORK(&ctrl->fatal_err_work, nvmet_fatal_error_handler);
+	schedule_work(&ctrl->fatal_err_work);
+}
+EXPORT_SYMBOL_GPL(nvmet_ctrl_fatal_error);
+
+static struct nvmet_subsys *nvmet_find_get_subsys(struct nvmet_port *port,
+		const char *subsysnqn)
+{
+	struct nvmet_subsys_link *p;
+
+	if (!port)
+		return NULL;
+
+	if (!strncmp(NVME_DISC_SUBSYS_NAME, subsysnqn,
+			NVMF_NQN_SIZE)) {
+		if (!kref_get_unless_zero(&nvmet_disc_subsys->ref))
+			return NULL;
+		return nvmet_disc_subsys;
+	}
+
+	down_read(&nvmet_config_sem);
+	list_for_each_entry(p, &port->subsystems, entry) {
+		if (!strncmp(p->subsys->subsysnqn, subsysnqn,
+				NVMF_NQN_SIZE)) {
+			if (!kref_get_unless_zero(&p->subsys->ref))
+				break;
+			up_read(&nvmet_config_sem);
+			return p->subsys;
+		}
+	}
+	up_read(&nvmet_config_sem);
+	return NULL;
+}
+
+struct nvmet_subsys *nvmet_subsys_alloc(const char *subsysnqn,
+		enum nvme_subsys_type type)
+{
+	struct nvmet_subsys *subsys;
+
+	subsys = kzalloc(sizeof(*subsys), GFP_KERNEL);
+	if (!subsys)
+		return NULL;
+
+	subsys->ver = (1 << 16) | (2 << 8) | 1; /* NVMe 1.2.1 */
+
+	switch (type) {
+	case NVME_NQN_NVME:
+		subsys->max_qid = NVMET_NR_QUEUES;
+		break;
+	case NVME_NQN_DISC:
+		subsys->max_qid = 0;
+		break;
+	default:
+		pr_err("%s: Unknown Subsystem type - %d\n", __func__, type);
+		kfree(subsys);
+		return NULL;
+	}
+	subsys->type = type;
+	subsys->subsysnqn = kstrndup(subsysnqn, NVMF_NQN_SIZE,
+			GFP_KERNEL);
+	if (IS_ERR(subsys->subsysnqn)) {
+		kfree(subsys);
+		return NULL;
+	}
+
+	kref_init(&subsys->ref);
+
+	mutex_init(&subsys->lock);
+	INIT_LIST_HEAD(&subsys->namespaces);
+	INIT_LIST_HEAD(&subsys->ctrls);
+
+	ida_init(&subsys->cntlid_ida);
+
+	INIT_LIST_HEAD(&subsys->hosts);
+
+	return subsys;
+}
+
+static void nvmet_subsys_free(struct kref *ref)
+{
+	struct nvmet_subsys *subsys =
+		container_of(ref, struct nvmet_subsys, ref);
+
+	WARN_ON_ONCE(!list_empty(&subsys->namespaces));
+
+	ida_destroy(&subsys->cntlid_ida);
+	kfree(subsys->subsysnqn);
+	kfree(subsys);
+}
+
+void nvmet_subsys_put(struct nvmet_subsys *subsys)
+{
+	kref_put(&subsys->ref, nvmet_subsys_free);
+}
+
+static int __init nvmet_init(void)
+{
+	int error;
+
+	error = nvmet_init_discovery();
+	if (error)
+		goto out;
+
+	error = nvmet_init_configfs();
+	if (error)
+		goto out_exit_discovery;
+	return 0;
+
+out_exit_discovery:
+	nvmet_exit_discovery();
+out:
+	return error;
+}
+
+static void __exit nvmet_exit(void)
+{
+	nvmet_exit_configfs();
+	nvmet_exit_discovery();
+
+	BUILD_BUG_ON(sizeof(struct nvmf_disc_rsp_page_entry) != 1024);
+	BUILD_BUG_ON(sizeof(struct nvmf_disc_rsp_page_hdr) != 1024);
+}
+
+module_init(nvmet_init);
+module_exit(nvmet_exit);
+
+MODULE_LICENSE("GPL v2");

drivers/nvme/target/discovery.c

@@ -0,0 +1,221 @@
+/*
+ * Discovery service for the NVMe over Fabrics target.
+ * Copyright (C) 2016 Intel Corporation. All rights reserved.
+ *
+ * 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.
+ */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+#include <linux/slab.h>
+#include <generated/utsrelease.h>
+#include "nvmet.h"
+
+struct nvmet_subsys *nvmet_disc_subsys;
+
+u64 nvmet_genctr;
+
+void nvmet_referral_enable(struct nvmet_port *parent, struct nvmet_port *port)
+{
+	down_write(&nvmet_config_sem);
+	if (list_empty(&port->entry)) {
+		list_add_tail(&port->entry, &parent->referrals);
+		port->enabled = true;
+		nvmet_genctr++;
+	}
+	up_write(&nvmet_config_sem);
+}
+
+void nvmet_referral_disable(struct nvmet_port *port)
+{
+	down_write(&nvmet_config_sem);
+	if (!list_empty(&port->entry)) {
+		port->enabled = false;
+		list_del_init(&port->entry);
+		nvmet_genctr++;
+	}
+	up_write(&nvmet_config_sem);
+}
+
+static void nvmet_format_discovery_entry(struct nvmf_disc_rsp_page_hdr *hdr,
+		struct nvmet_port *port, char *subsys_nqn, u8 type, u32 numrec)
+{
+	struct nvmf_disc_rsp_page_entry *e = &hdr->entries[numrec];
+
+	e->trtype = port->disc_addr.trtype;
+	e->adrfam = port->disc_addr.adrfam;
+	e->treq = port->disc_addr.treq;
+	e->portid = port->disc_addr.portid;
+	/* we support only dynamic controllers */
+	e->cntlid = cpu_to_le16(NVME_CNTLID_DYNAMIC);
+	e->asqsz = cpu_to_le16(NVMF_AQ_DEPTH);
+	e->nqntype = type;
+	memcpy(e->trsvcid, port->disc_addr.trsvcid, NVMF_TRSVCID_SIZE);
+	memcpy(e->traddr, port->disc_addr.traddr, NVMF_TRADDR_SIZE);
+	memcpy(e->tsas.common, port->disc_addr.tsas.common, NVMF_TSAS_SIZE);
+	memcpy(e->subnqn, subsys_nqn, NVMF_NQN_SIZE);
+}
+
+static void nvmet_execute_get_disc_log_page(struct nvmet_req *req)
+{
+	const int entry_size = sizeof(struct nvmf_disc_rsp_page_entry);
+	struct nvmet_ctrl *ctrl = req->sq->ctrl;
+	struct nvmf_disc_rsp_page_hdr *hdr;
+	size_t data_len = nvmet_get_log_page_len(req->cmd);
+	size_t alloc_len = max(data_len, sizeof(*hdr));
+	int residual_len = data_len - sizeof(*hdr);
+	struct nvmet_subsys_link *p;
+	struct nvmet_port *r;
+	u32 numrec = 0;
+	u16 status = 0;
+
+	/*
+	 * Make sure we're passing at least a buffer of response header size.
+	 * If host provided data len is less than the header size, only the
+	 * number of bytes requested by host will be sent to host.
+	 */
+	hdr = kzalloc(alloc_len, GFP_KERNEL);
+	if (!hdr) {
+		status = NVME_SC_INTERNAL;
+		goto out;
+	}
+
+	down_read(&nvmet_config_sem);
+	list_for_each_entry(p, &req->port->subsystems, entry) {
+		if (!nvmet_host_allowed(req, p->subsys, ctrl->hostnqn))
+			continue;
+		if (residual_len >= entry_size) {
+			nvmet_format_discovery_entry(hdr, req->port,
+					p->subsys->subsysnqn,
+					NVME_NQN_NVME, numrec);
+			residual_len -= entry_size;
+		}
+		numrec++;
+	}
+
+	list_for_each_entry(r, &req->port->referrals, entry) {
+		if (residual_len >= entry_size) {
+			nvmet_format_discovery_entry(hdr, r,
+					NVME_DISC_SUBSYS_NAME,
+					NVME_NQN_DISC, numrec);
+			residual_len -= entry_size;
+		}
+		numrec++;
+	}
+
+	hdr->genctr = cpu_to_le64(nvmet_genctr);
+	hdr->numrec = cpu_to_le64(numrec);
+	hdr->recfmt = cpu_to_le16(0);
+
+	up_read(&nvmet_config_sem);
+
+	status = nvmet_copy_to_sgl(req, 0, hdr, data_len);
+	kfree(hdr);
+out:
+	nvmet_req_complete(req, status);
+}
+
+static void nvmet_execute_identify_disc_ctrl(struct nvmet_req *req)
+{
+	struct nvmet_ctrl *ctrl = req->sq->ctrl;
+	struct nvme_id_ctrl *id;
+	u16 status = 0;
+
+	id = kzalloc(sizeof(*id), GFP_KERNEL);
+	if (!id) {
+		status = NVME_SC_INTERNAL;
+		goto out;
+	}
+
+	memset(id->fr, ' ', sizeof(id->fr));
+	strncpy((char *)id->fr, UTS_RELEASE, sizeof(id->fr));
+
+	/* no limit on data transfer sizes for now */
+	id->mdts = 0;
+	id->cntlid = cpu_to_le16(ctrl->cntlid);
+	id->ver = cpu_to_le32(ctrl->subsys->ver);
+	id->lpa = (1 << 2);
+
+	/* no enforcement soft-limit for maxcmd - pick arbitrary high value */
+	id->maxcmd = cpu_to_le16(NVMET_MAX_CMD);
+
+	id->sgls = cpu_to_le32(1 << 0);	/* we always support SGLs */
+	if (ctrl->ops->has_keyed_sgls)
+		id->sgls |= cpu_to_le32(1 << 2);
+	if (ctrl->ops->sqe_inline_size)
+		id->sgls |= cpu_to_le32(1 << 20);
+
+	strcpy(id->subnqn, ctrl->subsys->subsysnqn);
+
+	status = nvmet_copy_to_sgl(req, 0, id, sizeof(*id));
+
+	kfree(id);
+out:
+	nvmet_req_complete(req, status);
+}
+
+int nvmet_parse_discovery_cmd(struct nvmet_req *req)
+{
+	struct nvme_command *cmd = req->cmd;
+
+	req->ns = NULL;
+
+	if (unlikely(!(req->sq->ctrl->csts & NVME_CSTS_RDY))) {
+		pr_err("nvmet: got cmd %d while not ready\n",
+				cmd->common.opcode);
+		return NVME_SC_INVALID_OPCODE | NVME_SC_DNR;
+	}
+
+	switch (cmd->common.opcode) {
+	case nvme_admin_get_log_page:
+		req->data_len = nvmet_get_log_page_len(cmd);
+
+		switch (cmd->get_log_page.lid) {
+		case NVME_LOG_DISC:
+			req->execute = nvmet_execute_get_disc_log_page;
+			return 0;
+		default:
+			pr_err("nvmet: unsupported get_log_page lid %d\n",
+				cmd->get_log_page.lid);
+		return NVME_SC_INVALID_OPCODE | NVME_SC_DNR;
+		}
+	case nvme_admin_identify:
+		req->data_len = 4096;
+		switch (le32_to_cpu(cmd->identify.cns)) {
+		case 0x01:
+			req->execute =
+				nvmet_execute_identify_disc_ctrl;
+			return 0;
+		default:
+			pr_err("nvmet: unsupported identify cns %d\n",
+				le32_to_cpu(cmd->identify.cns));
+			return NVME_SC_INVALID_OPCODE | NVME_SC_DNR;
+		}
+	default:
+		pr_err("nvmet: unsupported cmd %d\n",
+				cmd->common.opcode);
+		return NVME_SC_INVALID_OPCODE | NVME_SC_DNR;
+	}
+
+	pr_err("nvmet: unhandled cmd %d\n", cmd->common.opcode);
+	return NVME_SC_INVALID_OPCODE | NVME_SC_DNR;
+}
+
+int __init nvmet_init_discovery(void)
+{
+	nvmet_disc_subsys =
+		nvmet_subsys_alloc(NVME_DISC_SUBSYS_NAME, NVME_NQN_DISC);
+	if (!nvmet_disc_subsys)
+		return -ENOMEM;
+	return 0;
+}
+
+void nvmet_exit_discovery(void)
+{
+	nvmet_subsys_put(nvmet_disc_subsys);
+}

drivers/nvme/target/fabrics-cmd.c

@@ -0,0 +1,240 @@
+/*
+ * NVMe Fabrics command implementation.
+ * Copyright (c) 2015-2016 HGST, a Western Digital Company.
+ *
+ * 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.
+ */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+#include <linux/blkdev.h>
+#include "nvmet.h"
+
+static void nvmet_execute_prop_set(struct nvmet_req *req)
+{
+	u16 status = 0;
+
+	if (!(req->cmd->prop_set.attrib & 1)) {
+		u64 val = le64_to_cpu(req->cmd->prop_set.value);
+
+		switch (le32_to_cpu(req->cmd->prop_set.offset)) {
+		case NVME_REG_CC:
+			nvmet_update_cc(req->sq->ctrl, val);
+			break;
+		default:
+			status = NVME_SC_INVALID_FIELD | NVME_SC_DNR;
+			break;
+		}
+	} else {
+		status = NVME_SC_INVALID_FIELD | NVME_SC_DNR;
+	}
+
+	nvmet_req_complete(req, status);
+}
+
+static void nvmet_execute_prop_get(struct nvmet_req *req)
+{
+	struct nvmet_ctrl *ctrl = req->sq->ctrl;
+	u16 status = 0;
+	u64 val = 0;
+
+	if (req->cmd->prop_get.attrib & 1) {
+		switch (le32_to_cpu(req->cmd->prop_get.offset)) {
+		case NVME_REG_CAP:
+			val = ctrl->cap;
+			break;
+		default:
+			status = NVME_SC_INVALID_FIELD | NVME_SC_DNR;
+			break;
+		}
+	} else {
+		switch (le32_to_cpu(req->cmd->prop_get.offset)) {
+		case NVME_REG_VS:
+			val = ctrl->subsys->ver;
+			break;
+		case NVME_REG_CC:
+			val = ctrl->cc;
+			break;
+		case NVME_REG_CSTS:
+			val = ctrl->csts;
+			break;
+		default:
+			status = NVME_SC_INVALID_FIELD | NVME_SC_DNR;
+			break;
+		}
+	}
+
+	req->rsp->result64 = cpu_to_le64(val);
+	nvmet_req_complete(req, status);
+}
+
+int nvmet_parse_fabrics_cmd(struct nvmet_req *req)
+{
+	struct nvme_command *cmd = req->cmd;
+
+	req->ns = NULL;
+
+	switch (cmd->fabrics.fctype) {
+	case nvme_fabrics_type_property_set:
+		req->data_len = 0;
+		req->execute = nvmet_execute_prop_set;
+		break;
+	case nvme_fabrics_type_property_get:
+		req->data_len = 0;
+		req->execute = nvmet_execute_prop_get;
+		break;
+	default:
+		pr_err("received unknown capsule type 0x%x\n",
+			cmd->fabrics.fctype);
+		return NVME_SC_INVALID_OPCODE | NVME_SC_DNR;
+	}
+
+	return 0;
+}
+
+static u16 nvmet_install_queue(struct nvmet_ctrl *ctrl, struct nvmet_req *req)
+{
+	struct nvmf_connect_command *c = &req->cmd->connect;
+	u16 qid = le16_to_cpu(c->qid);
+	u16 sqsize = le16_to_cpu(c->sqsize);
+	struct nvmet_ctrl *old;
+
+	old = cmpxchg(&req->sq->ctrl, NULL, ctrl);
+	if (old) {
+		pr_warn("queue already connected!\n");
+		return NVME_SC_CONNECT_CTRL_BUSY | NVME_SC_DNR;
+	}
+
+	nvmet_cq_setup(ctrl, req->cq, qid, sqsize);
+	nvmet_sq_setup(ctrl, req->sq, qid, sqsize);
+	return 0;
+}
+
+static void nvmet_execute_admin_connect(struct nvmet_req *req)
+{
+	struct nvmf_connect_command *c = &req->cmd->connect;
+	struct nvmf_connect_data *d;
+	struct nvmet_ctrl *ctrl = NULL;
+	u16 status = 0;
+
+	d = kmap(sg_page(req->sg)) + req->sg->offset;
+
+	/* zero out initial completion result, assign values as needed */
+	req->rsp->result = 0;
+
+	if (c->recfmt != 0) {
+		pr_warn("invalid connect version (%d).\n",
+			le16_to_cpu(c->recfmt));
+		status = NVME_SC_CONNECT_FORMAT | NVME_SC_DNR;
+		goto out;
+	}
+
+	if (unlikely(d->cntlid != cpu_to_le16(0xffff))) {
+		pr_warn("connect attempt for invalid controller ID %#x\n",
+			d->cntlid);
+		status = NVME_SC_CONNECT_INVALID_PARAM | NVME_SC_DNR;
+		req->rsp->result = IPO_IATTR_CONNECT_DATA(cntlid);
+		goto out;
+	}
+
+	status = nvmet_alloc_ctrl(d->subsysnqn, d->hostnqn, req,
+			le32_to_cpu(c->kato), &ctrl);
+	if (status)
+		goto out;
+
+	status = nvmet_install_queue(ctrl, req);
+	if (status) {
+		nvmet_ctrl_put(ctrl);
+		goto out;
+	}
+
+	pr_info("creating controller %d for NQN %s.\n",
+			ctrl->cntlid, ctrl->hostnqn);
+	req->rsp->result16 = cpu_to_le16(ctrl->cntlid);
+
+out:
+	kunmap(sg_page(req->sg));
+	nvmet_req_complete(req, status);
+}
+
+static void nvmet_execute_io_connect(struct nvmet_req *req)
+{
+	struct nvmf_connect_command *c = &req->cmd->connect;
+	struct nvmf_connect_data *d;
+	struct nvmet_ctrl *ctrl = NULL;
+	u16 qid = le16_to_cpu(c->qid);
+	u16 status = 0;
+
+	d = kmap(sg_page(req->sg)) + req->sg->offset;
+
+	/* zero out initial completion result, assign values as needed */
+	req->rsp->result = 0;
+
+	if (c->recfmt != 0) {
+		pr_warn("invalid connect version (%d).\n",
+			le16_to_cpu(c->recfmt));
+		status = NVME_SC_CONNECT_FORMAT | NVME_SC_DNR;
+		goto out;
+	}
+
+	status = nvmet_ctrl_find_get(d->subsysnqn, d->hostnqn,
+			le16_to_cpu(d->cntlid),
+			req, &ctrl);
+	if (status)
+		goto out;
+
+	if (unlikely(qid > ctrl->subsys->max_qid)) {
+		pr_warn("invalid queue id (%d)\n", qid);
+		status = NVME_SC_CONNECT_INVALID_PARAM | NVME_SC_DNR;
+		req->rsp->result = IPO_IATTR_CONNECT_SQE(qid);
+		goto out_ctrl_put;
+	}
+
+	status = nvmet_install_queue(ctrl, req);
+	if (status) {
+		/* pass back cntlid that had the issue of installing queue */
+		req->rsp->result16 = cpu_to_le16(ctrl->cntlid);
+		goto out_ctrl_put;
+	}
+
+	pr_info("adding queue %d to ctrl %d.\n", qid, ctrl->cntlid);
+
+out:
+	kunmap(sg_page(req->sg));
+	nvmet_req_complete(req, status);
+	return;
+
+out_ctrl_put:
+	nvmet_ctrl_put(ctrl);
+	goto out;
+}
+
+int nvmet_parse_connect_cmd(struct nvmet_req *req)
+{
+	struct nvme_command *cmd = req->cmd;
+
+	req->ns = NULL;
+
+	if (req->cmd->common.opcode != nvme_fabrics_command) {
+		pr_err("invalid command 0x%x on unconnected queue.\n",
+			cmd->fabrics.opcode);
+		return NVME_SC_INVALID_OPCODE | NVME_SC_DNR;
+	}
+	if (cmd->fabrics.fctype != nvme_fabrics_type_connect) {
+		pr_err("invalid capsule type 0x%x on unconnected queue.\n",
+			cmd->fabrics.fctype);
+		return NVME_SC_INVALID_OPCODE | NVME_SC_DNR;
+	}
+
+	req->data_len = sizeof(struct nvmf_connect_data);
+	if (cmd->connect.qid == 0)
+		req->execute = nvmet_execute_admin_connect;
+	else
+		req->execute = nvmet_execute_io_connect;
+	return 0;
+}

drivers/nvme/target/io-cmd.c

@@ -0,0 +1,215 @@
+/*
+ * NVMe I/O command implementation.
+ * Copyright (c) 2015-2016 HGST, a Western Digital Company.
+ *
+ * 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.
+ */
+#define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
+#include <linux/blkdev.h>
+#include <linux/module.h>
+#include "nvmet.h"
+
+static void nvmet_bio_done(struct bio *bio)
+{
+	struct nvmet_req *req = bio->bi_private;
+
+	nvmet_req_complete(req,
+		bio->bi_error ? NVME_SC_INTERNAL | NVME_SC_DNR : 0);
+
+	if (bio != &req->inline_bio)
+		bio_put(bio);
+}
+
+static inline u32 nvmet_rw_len(struct nvmet_req *req)
+{
+	return ((u32)le16_to_cpu(req->cmd->rw.length) + 1) <<
+			req->ns->blksize_shift;
+}
+
+static void nvmet_inline_bio_init(struct nvmet_req *req)
+{
+	struct bio *bio = &req->inline_bio;
+
+	bio_init(bio);
+	bio->bi_max_vecs = NVMET_MAX_INLINE_BIOVEC;
+	bio->bi_io_vec = req->inline_bvec;
+}
+
+static void nvmet_execute_rw(struct nvmet_req *req)
+{
+	int sg_cnt = req->sg_cnt;
+	struct scatterlist *sg;
+	struct bio *bio;
+	sector_t sector;
+	blk_qc_t cookie;
+	int op, op_flags = 0, i;
+
+	if (!req->sg_cnt) {
+		nvmet_req_complete(req, 0);
+		return;
+	}
+
+	if (req->cmd->rw.opcode == nvme_cmd_write) {
+		op = REQ_OP_WRITE;
+		if (req->cmd->rw.control & cpu_to_le16(NVME_RW_FUA))
+			op_flags |= REQ_FUA;
+	} else {
+		op = REQ_OP_READ;
+	}
+
+	sector = le64_to_cpu(req->cmd->rw.slba);
+	sector <<= (req->ns->blksize_shift - 9);
+
+	nvmet_inline_bio_init(req);
+	bio = &req->inline_bio;
+	bio->bi_bdev = req->ns->bdev;
+	bio->bi_iter.bi_sector = sector;
+	bio->bi_private = req;
+	bio->bi_end_io = nvmet_bio_done;
+	bio_set_op_attrs(bio, op, op_flags);
+
+	for_each_sg(req->sg, sg, req->sg_cnt, i) {
+		while (bio_add_page(bio, sg_page(sg), sg->length, sg->offset)
+				!= sg->length) {
+			struct bio *prev = bio;
+
+			bio = bio_alloc(GFP_KERNEL, min(sg_cnt, BIO_MAX_PAGES));
+			bio->bi_bdev = req->ns->bdev;
+			bio->bi_iter.bi_sector = sector;
+			bio_set_op_attrs(bio, op, op_flags);
+
+			bio_chain(bio, prev);
+			cookie = submit_bio(prev);
+		}
+
+		sector += sg->length >> 9;
+		sg_cnt--;
+	}
+
+	cookie = submit_bio(bio);
+
+	blk_poll(bdev_get_queue(req->ns->bdev), cookie);
+}
+
+static void nvmet_execute_flush(struct nvmet_req *req)
+{
+	struct bio *bio;
+
+	nvmet_inline_bio_init(req);
+	bio = &req->inline_bio;
+
+	bio->bi_bdev = req->ns->bdev;
+	bio->bi_private = req;
+	bio->bi_end_io = nvmet_bio_done;
+	bio_set_op_attrs(bio, REQ_OP_WRITE, WRITE_FLUSH);
+
+	submit_bio(bio);
+}
+
+static u16 nvmet_discard_range(struct nvmet_ns *ns,
+		struct nvme_dsm_range *range, struct bio **bio)
+{
+	if (__blkdev_issue_discard(ns->bdev,
+			le64_to_cpu(range->slba) << (ns->blksize_shift - 9),
+			le32_to_cpu(range->nlb) << (ns->blksize_shift - 9),
+			GFP_KERNEL, 0, bio))
+		return NVME_SC_INTERNAL | NVME_SC_DNR;
+	return 0;
+}
+
+static void nvmet_execute_discard(struct nvmet_req *req)
+{
+	struct nvme_dsm_range range;
+	struct bio *bio = NULL;
+	int i;
+	u16 status;
+
+	for (i = 0; i <= le32_to_cpu(req->cmd->dsm.nr); i++) {
+		status = nvmet_copy_from_sgl(req, i * sizeof(range), &range,
+				sizeof(range));
+		if (status)
+			break;
+
+		status = nvmet_discard_range(req->ns, &range, &bio);
+		if (status)
+			break;
+	}
+
+	if (bio) {
+		bio->bi_private = req;
+		bio->bi_end_io = nvmet_bio_done;
+		if (status) {
+			bio->bi_error = -EIO;
+			bio_endio(bio);
+		} else {
+			submit_bio(bio);
+		}
+	} else {
+		nvmet_req_complete(req, status);
+	}
+}
+
+static void nvmet_execute_dsm(struct nvmet_req *req)
+{
+	switch (le32_to_cpu(req->cmd->dsm.attributes)) {
+	case NVME_DSMGMT_AD:
+		nvmet_execute_discard(req);
+		return;
+	case NVME_DSMGMT_IDR:
+	case NVME_DSMGMT_IDW:
+	default:
+		/* Not supported yet */
+		nvmet_req_complete(req, 0);
+		return;
+	}
+}
+
+int nvmet_parse_io_cmd(struct nvmet_req *req)
+{
+	struct nvme_command *cmd = req->cmd;
+
+	if (unlikely(!(req->sq->ctrl->cc & NVME_CC_ENABLE))) {
+		pr_err("nvmet: got io cmd %d while CC.EN == 0\n",
+				cmd->common.opcode);
+		req->ns = NULL;
+		return NVME_SC_CMD_SEQ_ERROR | NVME_SC_DNR;
+	}
+
+	if (unlikely(!(req->sq->ctrl->csts & NVME_CSTS_RDY))) {
+		pr_err("nvmet: got io cmd %d while CSTS.RDY == 0\n",
+				cmd->common.opcode);
+		req->ns = NULL;
+		return NVME_SC_CMD_SEQ_ERROR | NVME_SC_DNR;
+	}
+
+	req->ns = nvmet_find_namespace(req->sq->ctrl, cmd->rw.nsid);
+	if (!req->ns)
+		return NVME_SC_INVALID_NS | NVME_SC_DNR;
+
+	switch (cmd->common.opcode) {
+	case nvme_cmd_read:
+	case nvme_cmd_write:
+		req->execute = nvmet_execute_rw;
+		req->data_len = nvmet_rw_len(req);
+		return 0;
+	case nvme_cmd_flush:
+		req->execute = nvmet_execute_flush;
+		req->data_len = 0;
+		return 0;
+	case nvme_cmd_dsm:
+		req->execute = nvmet_execute_dsm;
+		req->data_len = le32_to_cpu(cmd->dsm.nr) *
+			sizeof(struct nvme_dsm_range);
+		return 0;
+	default:
+		pr_err("nvmet: unhandled cmd %d\n", cmd->common.opcode);
+		return NVME_SC_INVALID_OPCODE | NVME_SC_DNR;
+	}
+}

drivers/nvme/target/nvmet.h

@@ -0,0 +1,331 @@
+/*
+ * Copyright (c) 2015-2016 HGST, a Western Digital Company.
+ *
+ * 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.
+ */
+
+#ifndef _NVMET_H
+#define _NVMET_H
+
+#include <linux/dma-mapping.h>
+#include <linux/types.h>
+#include <linux/device.h>
+#include <linux/kref.h>
+#include <linux/percpu-refcount.h>
+#include <linux/list.h>
+#include <linux/mutex.h>
+#include <linux/nvme.h>
+#include <linux/configfs.h>
+#include <linux/rcupdate.h>
+#include <linux/blkdev.h>
+
+#define NVMET_ASYNC_EVENTS		4
+#define NVMET_ERROR_LOG_SLOTS		128
+
+/* Helper Macros when NVMe error is NVME_SC_CONNECT_INVALID_PARAM
+ * The 16 bit shift is to set IATTR bit to 1, which means offending
+ * offset starts in the data section of connect()
+ */
+#define IPO_IATTR_CONNECT_DATA(x)	\
+	(cpu_to_le32((1 << 16) | (offsetof(struct nvmf_connect_data, x))))
+#define IPO_IATTR_CONNECT_SQE(x)	\
+	(cpu_to_le32(offsetof(struct nvmf_connect_command, x)))
+
+struct nvmet_ns {
+	struct list_head	dev_link;
+	struct percpu_ref	ref;
+	struct block_device	*bdev;
+	u32			nsid;
+	u32			blksize_shift;
+	loff_t			size;
+	u8			nguid[16];
+
+	struct nvmet_subsys	*subsys;
+	const char		*device_path;
+
+	struct config_group	device_group;
+	struct config_group	group;
+
+	struct completion	disable_done;
+};
+
+static inline struct nvmet_ns *to_nvmet_ns(struct config_item *item)
+{
+	return container_of(to_config_group(item), struct nvmet_ns, group);
+}
+
+static inline bool nvmet_ns_enabled(struct nvmet_ns *ns)
+{
+	return !list_empty_careful(&ns->dev_link);
+}
+
+struct nvmet_cq {
+	u16			qid;
+	u16			size;
+};
+
+struct nvmet_sq {
+	struct nvmet_ctrl	*ctrl;
+	struct percpu_ref	ref;
+	u16			qid;
+	u16			size;
+	struct completion	free_done;
+};
+
+/**
+ * struct nvmet_port -	Common structure to keep port
+ *				information for the target.
+ * @entry:		List head for holding a list of these elements.
+ * @disc_addr:		Address information is stored in a format defined
+ *				for a discovery log page entry.
+ * @group:		ConfigFS group for this element's folder.
+ * @priv:		Private data for the transport.
+ */
+struct nvmet_port {
+	struct list_head		entry;
+	struct nvmf_disc_rsp_page_entry	disc_addr;
+	struct config_group		group;
+	struct config_group		subsys_group;
+	struct list_head		subsystems;
+	struct config_group		referrals_group;
+	struct list_head		referrals;
+	void				*priv;
+	bool				enabled;
+};
+
+static inline struct nvmet_port *to_nvmet_port(struct config_item *item)
+{
+	return container_of(to_config_group(item), struct nvmet_port,
+			group);
+}
+
+struct nvmet_ctrl {
+	struct nvmet_subsys	*subsys;
+	struct nvmet_cq		**cqs;
+	struct nvmet_sq		**sqs;
+
+	struct mutex		lock;
+	u64			cap;
+	u32			cc;
+	u32			csts;
+
+	u16			cntlid;
+	u32			kato;
+
+	struct nvmet_req	*async_event_cmds[NVMET_ASYNC_EVENTS];
+	unsigned int		nr_async_event_cmds;
+	struct list_head	async_events;
+	struct work_struct	async_event_work;
+
+	struct list_head	subsys_entry;
+	struct kref		ref;
+	struct delayed_work	ka_work;
+	struct work_struct	fatal_err_work;
+
+	struct nvmet_fabrics_ops *ops;
+
+	char			subsysnqn[NVMF_NQN_FIELD_LEN];
+	char			hostnqn[NVMF_NQN_FIELD_LEN];
+};
+
+struct nvmet_subsys {
+	enum nvme_subsys_type	type;
+
+	struct mutex		lock;
+	struct kref		ref;
+
+	struct list_head	namespaces;
+	unsigned int		max_nsid;
+
+	struct list_head	ctrls;
+	struct ida		cntlid_ida;
+
+	struct list_head	hosts;
+	bool			allow_any_host;
+
+	u16			max_qid;
+
+	u64			ver;
+	char			*subsysnqn;
+
+	struct config_group	group;
+
+	struct config_group	namespaces_group;
+	struct config_group	allowed_hosts_group;
+};
+
+static inline struct nvmet_subsys *to_subsys(struct config_item *item)
+{
+	return container_of(to_config_group(item), struct nvmet_subsys, group);
+}
+
+static inline struct nvmet_subsys *namespaces_to_subsys(
+		struct config_item *item)
+{
+	return container_of(to_config_group(item), struct nvmet_subsys,
+			namespaces_group);
+}
+
+struct nvmet_host {
+	struct config_group	group;
+};
+
+static inline struct nvmet_host *to_host(struct config_item *item)
+{
+	return container_of(to_config_group(item), struct nvmet_host, group);
+}
+
+static inline char *nvmet_host_name(struct nvmet_host *host)
+{
+	return config_item_name(&host->group.cg_item);
+}
+
+struct nvmet_host_link {
+	struct list_head	entry;
+	struct nvmet_host	*host;
+};
+
+struct nvmet_subsys_link {
+	struct list_head	entry;
+	struct nvmet_subsys	*subsys;
+};
+
+struct nvmet_req;
+struct nvmet_fabrics_ops {
+	struct module *owner;
+	unsigned int type;
+	unsigned int sqe_inline_size;
+	unsigned int msdbd;
+	bool has_keyed_sgls : 1;
+	void (*queue_response)(struct nvmet_req *req);
+	int (*add_port)(struct nvmet_port *port);
+	void (*remove_port)(struct nvmet_port *port);
+	void (*delete_ctrl)(struct nvmet_ctrl *ctrl);
+};
+
+#define NVMET_MAX_INLINE_BIOVEC	8
+
+struct nvmet_req {
+	struct nvme_command	*cmd;
+	struct nvme_completion	*rsp;
+	struct nvmet_sq		*sq;
+	struct nvmet_cq		*cq;
+	struct nvmet_ns		*ns;
+	struct scatterlist	*sg;
+	struct bio		inline_bio;
+	struct bio_vec		inline_bvec[NVMET_MAX_INLINE_BIOVEC];
+	int			sg_cnt;
+	size_t			data_len;
+
+	struct nvmet_port	*port;
+
+	void (*execute)(struct nvmet_req *req);
+	struct nvmet_fabrics_ops *ops;
+};
+
+static inline void nvmet_set_status(struct nvmet_req *req, u16 status)
+{
+	req->rsp->status = cpu_to_le16(status << 1);
+}
+
+static inline void nvmet_set_result(struct nvmet_req *req, u32 result)
+{
+	req->rsp->result = cpu_to_le32(result);
+}
+
+/*
+ * NVMe command writes actually are DMA reads for us on the target side.
+ */
+static inline enum dma_data_direction
+nvmet_data_dir(struct nvmet_req *req)
+{
+	return nvme_is_write(req->cmd) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
+}
+
+struct nvmet_async_event {
+	struct list_head	entry;
+	u8			event_type;
+	u8			event_info;
+	u8			log_page;
+};
+
+int nvmet_parse_connect_cmd(struct nvmet_req *req);
+int nvmet_parse_io_cmd(struct nvmet_req *req);
+int nvmet_parse_admin_cmd(struct nvmet_req *req);
+int nvmet_parse_discovery_cmd(struct nvmet_req *req);
+int nvmet_parse_fabrics_cmd(struct nvmet_req *req);
+
+bool nvmet_req_init(struct nvmet_req *req, struct nvmet_cq *cq,
+		struct nvmet_sq *sq, struct nvmet_fabrics_ops *ops);
+void nvmet_req_complete(struct nvmet_req *req, u16 status);
+
+void nvmet_cq_setup(struct nvmet_ctrl *ctrl, struct nvmet_cq *cq, u16 qid,
+		u16 size);
+void nvmet_sq_setup(struct nvmet_ctrl *ctrl, struct nvmet_sq *sq, u16 qid,
+		u16 size);
+void nvmet_sq_destroy(struct nvmet_sq *sq);
+int nvmet_sq_init(struct nvmet_sq *sq);
+
+void nvmet_ctrl_fatal_error(struct nvmet_ctrl *ctrl);
+
+void nvmet_update_cc(struct nvmet_ctrl *ctrl, u32 new);
+u16 nvmet_alloc_ctrl(const char *subsysnqn, const char *hostnqn,
+		struct nvmet_req *req, u32 kato, struct nvmet_ctrl **ctrlp);
+u16 nvmet_ctrl_find_get(const char *subsysnqn, const char *hostnqn, u16 cntlid,
+		struct nvmet_req *req, struct nvmet_ctrl **ret);
+void nvmet_ctrl_put(struct nvmet_ctrl *ctrl);
+
+struct nvmet_subsys *nvmet_subsys_alloc(const char *subsysnqn,
+		enum nvme_subsys_type type);
+void nvmet_subsys_put(struct nvmet_subsys *subsys);
+
+struct nvmet_ns *nvmet_find_namespace(struct nvmet_ctrl *ctrl, __le32 nsid);
+void nvmet_put_namespace(struct nvmet_ns *ns);
+int nvmet_ns_enable(struct nvmet_ns *ns);
+void nvmet_ns_disable(struct nvmet_ns *ns);
+struct nvmet_ns *nvmet_ns_alloc(struct nvmet_subsys *subsys, u32 nsid);
+void nvmet_ns_free(struct nvmet_ns *ns);
+
+int nvmet_register_transport(struct nvmet_fabrics_ops *ops);
+void nvmet_unregister_transport(struct nvmet_fabrics_ops *ops);
+
+int nvmet_enable_port(struct nvmet_port *port);
+void nvmet_disable_port(struct nvmet_port *port);
+
+void nvmet_referral_enable(struct nvmet_port *parent, struct nvmet_port *port);
+void nvmet_referral_disable(struct nvmet_port *port);
+
+u16 nvmet_copy_to_sgl(struct nvmet_req *req, off_t off, const void *buf,
+		size_t len);
+u16 nvmet_copy_from_sgl(struct nvmet_req *req, off_t off, void *buf,
+		size_t len);
+
+u32 nvmet_get_log_page_len(struct nvme_command *cmd);
+
+#define NVMET_QUEUE_SIZE	1024
+#define NVMET_NR_QUEUES		64
+#define NVMET_MAX_CMD		NVMET_QUEUE_SIZE
+#define NVMET_KAS		10
+#define NVMET_DISC_KATO		120
+
+int __init nvmet_init_configfs(void);
+void __exit nvmet_exit_configfs(void);
+
+int __init nvmet_init_discovery(void);
+void nvmet_exit_discovery(void);
+
+extern struct nvmet_subsys *nvmet_disc_subsys;
+extern u64 nvmet_genctr;
+extern struct rw_semaphore nvmet_config_sem;
+
+bool nvmet_host_allowed(struct nvmet_req *req, struct nvmet_subsys *subsys,
+		const char *hostnqn);
+
+#endif /* _NVMET_H */