Skip to contentSkip to conversation

NVMeTCP Offload ULP

RFC PATCH Superseded Shai Malin 1 participant 8 messages Dec 13, 2020 NVMe

v2 of 7 Superseded Latest is v7 →

lore
Share & tools
lore

Shai Malin

author
#0 lore raw

v2 cover NVMeTCP Offload ULP

With the goal of enabling a generic infrastructure that allows NVMe/TCP offload devices like NICs to seamlessly plug into the NVMe-oF stack, this patch series introduces the nvme-tcp-offload ULP host layer, which will be a new transport type called "tcp-offload" and will serve as an abstraction layer to work with vendor specific nvme-tcp offload drivers.

The nvme-tcp-offload transport can co-exist with the existing tcp and other transports. The tcp offload was designed so that stack changes are kept to a bare minimum: only registering new transports. All other APIs, ops etc. are identical to the regular tcp transport. Representing the TCP offload as a new transport allows clear and manageable differentiation between the connections which should use the offload path and those that are not offloaded (even on the same device).

Queue Initialization:
=====================
The nvme-tcp-offload ULP module shall register with the existing nvmf_transport_ops (.name = "tcp_offload"), nvme_ctrl_ops and blk_mq_ops. The nvme-tcp-offload vendor driver shall register to nvme-tcp-offload ULP with the following ops:
- claim_dev() - in order to resolve the route to the target according to the net_dev.
- create_queue() - in order to create offloaded nvme-tcp queue.

The nvme-tcp-offload ULP module shall manage all the controller level functionalities, call claim_dev and based on the return values shall call the relevant module create_queue in order to create the admin queue and the IO queues.

IO-path:
========
The nvme-tcp-offload shall work at the IO-level - the nvme-tcp-offload ULP module shall pass the request (the IO) to the nvme-tcp-offload vendor driver and later, the nvme-tcp-offload vendor driver return the request completion (the IO completion). No additional handling is needed in between; this design will reduce the CPU utilization as we will describe below.

The nvme-tcp-offload vendor driver shall register to nvme-tcp-offload ULP with the following IO-path ops:
- init_req()
- map_sg() - in order to map the request sg (similar to nvme_rdma_map_data() ).
- send_req() - in order to pass the request to the handling of the offload driver that shall pass it to the vendor specific device.
- poll_queue()

Once the IO completes, the nvme-tcp-offload vendor driver shall call command.done() that will invoke the nvme-tcp-offload ULP layer to complete the request.

TCP events:
===========
The Marvell FastLinQ NIC HW engine handle all the TCP re-transmissions and OOO events.

Teardown and errors:
====================
In case of NVMeTCP queue error the nvme-tcp-offload vendor driver shall call the nvme_tcp_ofld_report_queue_err. The nvme-tcp-offload vendor driver shall register to nvme-tcp-offload ULP with the following teardown ops:
- drain_queue()
- destroy_queue()

The Marvell FastLinQ NIC HW engine:
====================================
The Marvell NIC HW engine is capable of offloading the entire TCP/IP stack and managing up to 64K connections, already implemented and upstream use cases for this include iWARP (by the Marvell qedr driver) and iSCSI (by the Marvell qedi driver). In addition, the Marvell NIC HW engine offloads the NVMeTCP queue layer and is able to manage the IO level also in case of TCP re-transmissions and OOO events.
The HW engine enables direct data placement (including the data digest CRC calculation and validation) and direct data transmission (including data digest CRC calculation).

The series patches:
===================

Patch 1-2       Add the nvme-tcp-offload ULP module, including the upper and lower API.
Patches 3-5     nvme-tcp-offload ULP controller level functionalities.
Patch 6         nvme-tcp-offload ULP queue level functionalities.
Patch 7         nvme-tcp-offload ULP IO level functionalities.

Performance:
============
With this implementation on top of the Marvell qedn driver (using the Marvell FastLinQ NIC), we were able to demonstrate x3 CPU utilization improvement for 4K queued read/write IOs and up to x20 in case of 512K read/write IOs. In addition, we were able to demonstrate latency improvement, and specifically 99.99% tail latency improvement of up to x2-5 (depends on the queue-depth).

Future work:
============
- The Marvell FastLinQ NIC nvme-tcp-offload "qedn" host driver. This driver will interact with the qed core module, in a similar fashion to the existing ethernet (qede), rdma (qedr), iscsi (qedi) and fcoe (qedf) drivers of the same product-line.
- The nvme-tcp-offload ULP target abstraction layer.
- The Marvell nvme-tcp-offload "qednt" target driver.

Changes since RFC v1:
=========================================
- Fix nvme_tcp_ofld_ops return values.
- Remove NVMF_TRTYPE_TCP_OFFLOAD.
- Add nvme_tcp_ofld_poll() implementation.
- Fix nvme_tcp_ofld_queue_rq() to check map_sg() and send_req() return values.

Arie Gershberg (3):
nvme-fabrics: Move NVMF_ALLOWED_OPTS and NVMF_REQUIRED_OPTS
definitions
nvme-tcp-offload: Add controller level implementation
nvme-tcp-offload: Add controller level error recovery implementation

Dean Balandin (3):
nvme-tcp-offload: Add device scan implementation
nvme-tcp-offload: Add queue level implementation
nvme-tcp-offload: Add IO level implementation

Shai Malin (1):
nvme-tcp-offload: Add nvme-tcp-offload - NVMeTCP HW offload ULP

 drivers/nvme/host/Kconfig       |   16 +
 drivers/nvme/host/Makefile      |    3 +
 drivers/nvme/host/fabrics.c     |    7 -
 drivers/nvme/host/fabrics.h     |    7 +
 drivers/nvme/host/tcp-offload.c | 1106 +++++++++++++++++++++++++++++++
 drivers/nvme/host/tcp-offload.h |  184 +++++
 6 files changed, 1316 insertions(+), 7 deletions(-)
 create mode 100644 drivers/nvme/host/tcp-offload.c
 create mode 100644 drivers/nvme/host/tcp-offload.h
Signature · 7 lines
2.22.0


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#1 lore raw

RFC PATCH v2 1/7 nvme-tcp-offload: Add nvme-tcp-offload - NVMeTCP HW offload ULP

This patch will present the structure for the NVMeTCP offload common layer driver. This module is added under "drivers/nvme/host/" and future offload drivers which will register to it will be placed under "drivers/nvme/hw".
This new driver will be enabled by the Kconfig "NVM Express over Fabrics TCP offload commmon layer".
In order to support the new transport type, for host mode, no change is needed.

Each new vendor-specific offload driver will register to this ULP during its probe function, by filling out the nvme_tcp_ofld_dev->ops and nvme_tcp_ofld_dev->private_data and calling nvme_tcp_ofld_register_dev with the initialized struct.

The internal implementation:
- tcp-offload.h:
Includes all common structs and ops to be used and shared by offload drivers.

- tcp-offload.c:
Includes the init function which registers as a NVMf transport just like any other transport.

---
 drivers/nvme/host/Kconfig       |  16 +++
 drivers/nvme/host/Makefile      |   3 +
 drivers/nvme/host/tcp-offload.c | 124 +++++++++++++++++++++++
 drivers/nvme/host/tcp-offload.h | 173 ++++++++++++++++++++++++++++++++
 4 files changed, 316 insertions(+)
 create mode 100644 drivers/nvme/host/tcp-offload.c
 create mode 100644 drivers/nvme/host/tcp-offload.h
drivers/nvme/host/Kconfig +16 -0
index a44d49d63968..6e869e94e67f 100644
@@ -84,3 +84,19 @@ config NVME_TCP
 	  from https://github.com/linux-nvme/nvme-cli.
 
 	  If unsure, say N.
+
+config NVME_TCP_OFFLOAD
+	tristate "NVM Express over Fabrics TCP offload common layer"
+	default m
+	depends on INET
+	depends on BLK_DEV_NVME
+	select NVME_FABRICS
+	help
+	  This provides support for the NVMe over Fabrics protocol using
+	  the TCP offload transport. This allows you to use remote block devices
+	  exported using the NVMe protocol set.
+
+	  To configure a NVMe over Fabrics controller use the nvme-cli tool
+	  from https://github.com/linux-nvme/nvme-cli.
+
+	  If unsure, say N.
drivers/nvme/host/Makefile +3 -0
index d7f6a87687b8..0e7ef044cf29 100644
@@ -8,6 +8,7 @@ obj-$(CONFIG_NVME_FABRICS)		+= nvme-fabrics.o
 obj-$(CONFIG_NVME_RDMA)			+= nvme-rdma.o
 obj-$(CONFIG_NVME_FC)			+= nvme-fc.o
 obj-$(CONFIG_NVME_TCP)			+= nvme-tcp.o
+obj-$(CONFIG_NVME_TCP_OFFLOAD)	+= nvme-tcp-offload.o
 
 nvme-core-y				:= core.o
 nvme-core-$(CONFIG_TRACING)		+= trace.o
@@ -26,3 +27,5 @@ nvme-rdma-y				+= rdma.o
 nvme-fc-y				+= fc.o
 
 nvme-tcp-y				+= tcp.o
+
+nvme-tcp-offload-y		+= tcp-offload.o
drivers/nvme/host/tcp-offload.c new file +124 -0

124 lines not on this page · raw message

drivers/nvme/host/tcp-offload.h new file +173 -0

173 lines not on this page · raw message

Signature · 7 lines
2.22.0


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#2 lore raw

RFC PATCH v2 2/7 nvme-fabrics: Move NVMF_ALLOWED_OPTS and NVMF_REQUIRED_OPTS definitions

In reply to Shai Malin

From: Arie Gershberg <[email protected]>

Move NVMF_ALLOWED_OPTS and NVMF_REQUIRED_OPTS definitions to header file, so it can be used by transport modules.

---
 drivers/nvme/host/fabrics.c | 7 -------
 drivers/nvme/host/fabrics.h | 7 +++++++
 2 files changed, 7 insertions(+), 7 deletions(-)
drivers/nvme/host/fabrics.c +0 -7
index 72ac00173500..ccc626cbf6d0 100644
@@ -1002,13 +1002,6 @@ void nvmf_free_options(struct nvmf_ctrl_options *opts)
 }
 EXPORT_SYMBOL_GPL(nvmf_free_options);
 
-#define NVMF_REQUIRED_OPTS	(NVMF_OPT_TRANSPORT | NVMF_OPT_NQN)
-#define NVMF_ALLOWED_OPTS	(NVMF_OPT_QUEUE_SIZE | NVMF_OPT_NR_IO_QUEUES | \
-				 NVMF_OPT_KATO | NVMF_OPT_HOSTNQN | \
-				 NVMF_OPT_HOST_ID | NVMF_OPT_DUP_CONNECT |\
-				 NVMF_OPT_DISABLE_SQFLOW |\
-				 NVMF_OPT_FAIL_FAST_TMO)
-
 static struct nvme_ctrl *
 nvmf_create_ctrl(struct device *dev, const char *buf)
 {
drivers/nvme/host/fabrics.h +7 -0
index 733010d2eafd..e44955ed6fa9 100644
@@ -61,6 +61,13 @@ enum {
 	NVMF_OPT_FAIL_FAST_TMO	= 1 << 20,
 };
 
+#define NVMF_REQUIRED_OPTS	(NVMF_OPT_TRANSPORT | NVMF_OPT_NQN)
+#define NVMF_ALLOWED_OPTS	(NVMF_OPT_QUEUE_SIZE | NVMF_OPT_NR_IO_QUEUES | \
+				 NVMF_OPT_KATO | NVMF_OPT_HOSTNQN | \
+				 NVMF_OPT_HOST_ID | NVMF_OPT_DUP_CONNECT |\
+				 NVMF_OPT_DISABLE_SQFLOW |\
+				 NVMF_OPT_FAIL_FAST_TMO)
+
 /**
  * struct nvmf_ctrl_options - Used to hold the options specified
  *			      with the parsing opts enum.
Signature · 7 lines
2.22.0


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#3 lore raw

RFC PATCH v2 3/7 nvme-tcp-offload: Add device scan implementation

In reply to Shai Malin

From: Dean Balandin <[email protected]>

This patch implements the create_ctrl skeleton, specifcially to demonstrate how the claim_dev op will be used.

The driver scans the registered devices and calls the claim_dev op on each of them, to find the first devices that matches the connection params. Once the correct devices is found (claim_dev returns true), we raise the refcnt of that device and return that device as the device to be used for ctrl currently being created.

---
 drivers/nvme/host/tcp-offload.c | 105 ++++++++++++++++++++++++++++++++
 1 file changed, 105 insertions(+)
drivers/nvme/host/tcp-offload.c +105 -0
index 5e3e99b1273c..6285f82cb948 100644
@@ -13,6 +13,12 @@
 static LIST_HEAD(nvme_tcp_ofld_devices);
 static DECLARE_RWSEM(nvme_tcp_ofld_devices_rwsem);
 
+static inline struct nvme_tcp_ofld_ctrl *
+to_tcp_ofld_ctrl(struct nvme_ctrl *nctrl)
+{
+	return container_of(nctrl, struct nvme_tcp_ofld_ctrl, nctrl);
+}
+
 /**
  * nvme_tcp_ofld_register_dev() - NVMeTCP Offload Library registration
  * function.
@@ -96,6 +102,104 @@ nvme_tcp_ofld_req_done(struct nvme_tcp_ofld_req *req,
 	/* Placeholder - complete request with/without error */
 }
 
+struct nvme_tcp_ofld_dev *
+nvme_tcp_ofld_lookup_dev(struct nvme_tcp_ofld_ctrl *ctrl)
+{
+	struct nvme_tcp_ofld_dev *dev;
+
+	down_read(&nvme_tcp_ofld_devices_rwsem);
+	list_for_each_entry(dev, &nvme_tcp_ofld_devices, entry) {
+		if (dev->ops->claim_dev(dev, &ctrl->conn_params)) {
+			/* Increase driver refcnt */
+			if (!try_module_get(dev->ops->module)) {
+				pr_err("try_module_get failed");
+				dev = NULL;
+			}
+
+			goto out;
+		}
+	}
+
+	dev = NULL;
+out:
+	up_read(&nvme_tcp_ofld_devices_rwsem);
+
+	return dev;
+}
+
+static int nvme_tcp_ofld_setup_ctrl(struct nvme_ctrl *nctrl, bool new)
+{
+	/* Placeholder - validates inputs and creates admin and IO queues */
+
+	return 0;
+}
+
+static struct nvme_ctrl *
+nvme_tcp_ofld_create_ctrl(struct device *ndev, struct nvmf_ctrl_options *opts)
+{
+	struct nvme_tcp_ofld_queue *queue;
+	struct nvme_tcp_ofld_ctrl *ctrl;
+	struct nvme_tcp_ofld_dev *dev;
+	struct nvme_ctrl *nctrl;
+	int i, rc = 0;
+
+	ctrl = kzalloc(sizeof(*ctrl), GFP_KERNEL);
+	if (!ctrl)
+		return ERR_PTR(-ENOMEM);
+
+	/*
+	 * Init nvme_tcp_ofld_ctrl and nvme_ctrl params based on received
+	 * opts
+	 */
+
+	/* Find device that can reach the dest addr */
+	dev = nvme_tcp_ofld_lookup_dev(ctrl);
+	if (!dev) {
+		pr_info("no device found for addr %s:%s.\n",
+			opts->traddr, opts->trsvcid);
+		rc = -EINVAL;
+		goto out_free_ctrl;
+	}
+
+	ctrl->dev = dev;
+	ctrl->queues = kcalloc(nctrl->queue_count,
+			       sizeof(struct nvme_tcp_ofld_queue),
+			       GFP_KERNEL);
+	if (!ctrl->queues) {
+		rc = -ENOMEM;
+		goto out_module_put;
+	}
+
+	for (i = 0; i < nctrl->queue_count; ++i) {
+		queue = &ctrl->queues[i];
+		queue->ctrl = ctrl;
+		queue->dev = dev;
+		queue->report_err = nvme_tcp_ofld_report_queue_err;
+	}
+
+	/* Call nvme_init_ctrl */
+
+	rc = nvme_tcp_ofld_setup_ctrl(nctrl, true);
+	if (rc)
+		goto out_uninit_ctrl;
+
+	return nctrl;
+
+out_uninit_ctrl:
+	nvme_uninit_ctrl(nctrl);
+	nvme_put_ctrl(nctrl);
+	if (rc > 0)
+		rc = -EIO;
+
+	return ERR_PTR(rc);
+out_module_put:
+	module_put(dev->ops->module);
+out_free_ctrl:
+	kfree(ctrl);
+
+	return ERR_PTR(rc);
+}
+
 static struct nvmf_transport_ops nvme_tcp_ofld_transport = {
 	.name		= "tcp_offload",
 	.module		= THIS_MODULE,
@@ -105,6 +209,7 @@ static struct nvmf_transport_ops nvme_tcp_ofld_transport = {
 			  NVMF_OPT_CTRL_LOSS_TMO | NVMF_OPT_RECONNECT_DELAY |
 			  NVMF_OPT_HDR_DIGEST | NVMF_OPT_DATA_DIGEST |
 			  NVMF_OPT_NR_POLL_QUEUES | NVMF_OPT_TOS,
+	.create_ctrl	= nvme_tcp_ofld_create_ctrl,
 };
 
 static int __init nvme_tcp_ofld_init_module(void)
Signature · 7 lines
2.22.0


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#4 lore raw

RFC PATCH v2 4/7 nvme-tcp-offload: Add controller level implementation

In reply to Shai Malin

From: Arie Gershberg <[email protected]>

In this patch we implement controller level functionality including:
- create_ctrl.
- delete_ctrl.
- free_ctrl.

The implementation is similar to other nvme fabrics modules, the main difference being that the nvme-tcp-offload ULP calls the vendor specific claim_dev() op with the given TCP/IP parameters to determine which device will be used for this controller.
Once found, the vendor specific device and controller will be paired and kept in a controller list managed by the ULP.

---
 drivers/nvme/host/tcp-offload.c | 437 +++++++++++++++++++++++++++++++-
 drivers/nvme/host/tcp-offload.h |   1 +
 2 files changed, 432 insertions(+), 6 deletions(-)
drivers/nvme/host/tcp-offload.c +431 -6
index 6285f82cb948..0d25903d6832 100644
@@ -12,6 +12,10 @@
 
 static LIST_HEAD(nvme_tcp_ofld_devices);
 static DECLARE_RWSEM(nvme_tcp_ofld_devices_rwsem);
+static LIST_HEAD(nvme_tcp_ofld_ctrl_list);
+static DECLARE_RWSEM(nvme_tcp_ofld_ctrl_rwsem);
+static struct blk_mq_ops nvme_tcp_ofld_admin_mq_ops;
+static struct blk_mq_ops nvme_tcp_ofld_mq_ops;
 
 static inline struct nvme_tcp_ofld_ctrl *
 to_tcp_ofld_ctrl(struct nvme_ctrl *nctrl)
@@ -127,13 +131,367 @@ nvme_tcp_ofld_lookup_dev(struct nvme_tcp_ofld_ctrl *ctrl)
 	return dev;
 }
 
+static struct blk_mq_tag_set *
+nvme_tcp_ofld_alloc_tagset(struct nvme_ctrl *nctrl, bool admin)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl = to_tcp_ofld_ctrl(nctrl);
+	struct blk_mq_tag_set *set;
+	int rc;
+
+	if (admin) {
+		set = &ctrl->admin_tag_set;
+		memset(set, 0, sizeof(*set));
+		set->ops = &nvme_tcp_ofld_admin_mq_ops;
+		set->queue_depth = NVME_AQ_MQ_TAG_DEPTH;
+		set->reserved_tags = 2; /* connect + keep-alive */
+		set->numa_node = nctrl->numa_node;
+		set->flags = BLK_MQ_F_BLOCKING;
+		set->cmd_size = sizeof(struct nvme_tcp_ofld_req);
+		set->driver_data = ctrl;
+		set->nr_hw_queues = 1;
+		set->timeout = NVME_ADMIN_TIMEOUT;
+	} else {
+		set = &ctrl->tag_set;
+		memset(set, 0, sizeof(*set));
+		set->ops = &nvme_tcp_ofld_mq_ops;
+		set->queue_depth = nctrl->sqsize + 1;
+		set->reserved_tags = 1; /* fabric connect */
+		set->numa_node = nctrl->numa_node;
+		set->flags = BLK_MQ_F_SHOULD_MERGE | BLK_MQ_F_BLOCKING;
+		set->cmd_size = sizeof(struct nvme_tcp_ofld_req);
+		set->driver_data = ctrl;
+		set->nr_hw_queues = nctrl->queue_count - 1;
+		set->timeout = NVME_IO_TIMEOUT;
+		set->nr_maps = nctrl->opts->nr_poll_queues ?
+							HCTX_MAX_TYPES : 2;
+	}
+
+	rc = blk_mq_alloc_tag_set(set);
+	if (rc)
+		return ERR_PTR(rc);
+
+	return set;
+}
+
+static int nvme_tcp_ofld_start_queue(struct nvme_ctrl *nctrl, int qid)
+{
+	/* Placeholder - start_queue */
+	return 0;
+}
+
+static int nvme_tcp_ofld_configure_admin_queue(struct nvme_ctrl *nctrl,
+					       bool new)
+{
+	int rc;
+
+	/* Placeholder - alloc_admin_queue */
+	if (new) {
+		nctrl->admin_tagset =
+				nvme_tcp_ofld_alloc_tagset(nctrl, true);
+		if (IS_ERR(nctrl->admin_tagset)) {
+			rc = PTR_ERR(nctrl->admin_tagset);
+			nctrl->admin_tagset = NULL;
+			goto out_free_queue;
+		}
+
+		nctrl->fabrics_q = blk_mq_init_queue(nctrl->admin_tagset);
+		if (IS_ERR(nctrl->fabrics_q)) {
+			rc = PTR_ERR(nctrl->fabrics_q);
+			nctrl->fabrics_q = NULL;
+			goto out_free_tagset;
+		}
+
+		nctrl->admin_q = blk_mq_init_queue(nctrl->admin_tagset);
+		if (IS_ERR(nctrl->admin_q)) {
+			rc = PTR_ERR(nctrl->admin_q);
+			nctrl->admin_q = NULL;
+			goto out_cleanup_fabrics_q;
+		}
+	}
+
+	rc = nvme_tcp_ofld_start_queue(nctrl, 0);
+	if (rc)
+		goto out_cleanup_queue;
+
+	rc = nvme_enable_ctrl(nctrl);
+	if (rc)
+		goto out_stop_queue;
+
+	blk_mq_unquiesce_queue(nctrl->admin_q);
+
+	rc = nvme_init_identify(nctrl);
+	if (rc)
+		goto out_stop_queue;
+
+	return 0;
+
+out_stop_queue:
+	/* Placeholder - stop admin queue */
+out_cleanup_queue:
+	if (new)
+		blk_cleanup_queue(nctrl->admin_q);
+out_cleanup_fabrics_q:
+	if (new)
+		blk_cleanup_queue(nctrl->fabrics_q);
+out_free_tagset:
+	if (new)
+		blk_mq_free_tag_set(nctrl->admin_tagset);
+out_free_queue:
+	/* Placeholder - free admin queue */
+
+	return rc;
+}
+
+static int
+nvme_tcp_ofld_configure_io_queues(struct nvme_ctrl *nctrl, bool new)
+{
+	int rc;
+
+	/* Placeholder - alloc_io_queues */
+
+	if (new) {
+		nctrl->tagset = nvme_tcp_ofld_alloc_tagset(nctrl, false);
+		if (IS_ERR(nctrl->tagset)) {
+			rc = PTR_ERR(nctrl->tagset);
+			nctrl->tagset = NULL;
+			goto out_free_io_queues;
+		}
+
+		nctrl->connect_q = blk_mq_init_queue(nctrl->tagset);
+		if (IS_ERR(nctrl->connect_q)) {
+			rc = PTR_ERR(nctrl->connect_q);
+			nctrl->connect_q = NULL;
+			goto out_free_tag_set;
+		}
+	}
+
+	/* Placeholder - start_io_queues */
+
+	if (!new) {
+		nvme_start_queues(nctrl);
+		nvme_wait_freeze(nctrl);
+		blk_mq_update_nr_hw_queues(nctrl->tagset,
+					   nctrl->queue_count - 1);
+		nvme_unfreeze(nctrl);
+	}
+
+	return 0;
+
+out_free_tag_set:
+	if (new)
+		blk_mq_free_tag_set(nctrl->tagset);
+out_free_io_queues:
+	/* Placeholder - free_io_queues */
+
+	return rc;
+}
+
 static int nvme_tcp_ofld_setup_ctrl(struct nvme_ctrl *nctrl, bool new)
 {
-	/* Placeholder - validates inputs and creates admin and IO queues */
+	struct nvmf_ctrl_options *opts = nctrl->opts;
+	int rc;
+
+	rc = nvme_tcp_ofld_configure_admin_queue(nctrl, new);
+	if (rc)
+		goto destroy_admin;
+
+	if (nctrl->icdoff) {
+		dev_err(nctrl->device, "icdoff is not supported!\n");
+		rc = -EINVAL;
+		goto destroy_admin;
+	}
+
+	if (opts->queue_size > nctrl->sqsize + 1)
+		dev_warn(nctrl->device,
+			 "queue_size %zu > ctrl sqsize %u, clamping down\n",
+			 opts->queue_size, nctrl->sqsize + 1);
+
+	if (nctrl->sqsize + 1 > nctrl->maxcmd) {
+		dev_warn(nctrl->device,
+			 "sqsize %u > ctrl maxcmd %u, clamping down\n",
+			 nctrl->sqsize + 1, nctrl->maxcmd);
+		nctrl->sqsize = nctrl->maxcmd - 1;
+	}
+
+	if (nctrl->queue_count > 1) {
+		rc = nvme_tcp_ofld_configure_io_queues(nctrl, new);
+		if (rc)
+			goto destroy_admin;
+	}
+
+	if (!nvme_change_ctrl_state(nctrl, NVME_CTRL_LIVE)) {
+		/*
+		 * state change failure is ok if we started ctrl delete,
+		 * unless we're during creation of a new controller to
+		 * avoid races with teardown flow.
+		 */
+		WARN_ON_ONCE(nctrl->state != NVME_CTRL_DELETING &&
+			     nctrl->state != NVME_CTRL_DELETING_NOIO);
+		WARN_ON_ONCE(new);
+		rc = -EINVAL;
+		goto destroy_io;
+	}
+
+	nvme_start_ctrl(nctrl);
+	return 0;
+
+destroy_io:
+	/* Placeholder - stop and destroy io queues*/
+destroy_admin:
+	/* Placeholder - stop and destroy admin queue*/
+
+	return rc;
+}
+
+static int
+nvme_tcp_ofld_check_dev_opts(struct nvmf_ctrl_options *opts,
+			     struct nvme_tcp_ofld_ops *ofld_ops)
+{
+	unsigned int nvme_tcp_ofld_opt_mask = NVMF_ALLOWED_OPTS |
+			ofld_ops->allowed_opts | ofld_ops->required_opts;
+	if (opts->mask & ~nvme_tcp_ofld_opt_mask) {
+		pr_warn("One or more of the nvmf options isn't supported by %s.\n",
+			ofld_ops->name);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static void nvme_tcp_ofld_free_ctrl(struct nvme_ctrl *nctrl)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl = to_tcp_ofld_ctrl(nctrl);
+	struct nvme_tcp_ofld_dev *dev = ctrl->dev;
+
+	if (list_empty(&ctrl->list))
+		goto free_ctrl;
+
+	down_write(&nvme_tcp_ofld_ctrl_rwsem);
+	list_del(&ctrl->list);
+	up_write(&nvme_tcp_ofld_ctrl_rwsem);
+
+	nvmf_free_options(nctrl->opts);
+free_ctrl:
+	module_put(dev->ops->module);
+	kfree(ctrl->queues);
+	kfree(ctrl);
+}
+
+static void nvme_tcp_ofld_submit_async_event(struct nvme_ctrl *arg)
+{
+	/* Placeholder - submit_async_event */
+}
+
+static void
+nvme_tcp_ofld_teardown_admin_queue(struct nvme_ctrl *ctrl, bool remove)
+{
+	/* Placeholder - teardown_admin_queue */
+}
+
+static void
+nvme_tcp_ofld_teardown_io_queues(struct nvme_ctrl *nctrl, bool remove)
+{
+	/* Placeholder - teardown_io_queues */
+}
+
+static void
+nvme_tcp_ofld_teardown_ctrl(struct nvme_ctrl *nctrl, bool shutdown)
+{
+	/* Placeholder - err_work and connect_work */
+	nvme_tcp_ofld_teardown_io_queues(nctrl, shutdown);
+	blk_mq_quiesce_queue(nctrl->admin_q);
+	if (shutdown)
+		nvme_shutdown_ctrl(nctrl);
+	else
+		nvme_disable_ctrl(nctrl);
+	nvme_tcp_ofld_teardown_admin_queue(nctrl, shutdown);
+}
+
+static void nvme_tcp_ofld_delete_ctrl(struct nvme_ctrl *nctrl)
+{
+	nvme_tcp_ofld_teardown_ctrl(nctrl, true);
+}
+
+static int
+nvme_tcp_ofld_init_request(struct blk_mq_tag_set *set,
+			   struct request *rq,
+			   unsigned int hctx_idx,
+			   unsigned int numa_node)
+{
+	struct nvme_tcp_ofld_req *req = blk_mq_rq_to_pdu(rq);
+	struct nvme_tcp_ofld_ctrl *ctrl = set->driver_data;
+
+	/* Placeholder - init request */
+
+	req->done = nvme_tcp_ofld_req_done;
+	ctrl->dev->ops->init_req(req);
 
 	return 0;
 }
 
+static blk_status_t
+nvme_tcp_ofld_queue_rq(struct blk_mq_hw_ctx *hctx,
+		       const struct blk_mq_queue_data *bd)
+{
+	struct nvme_tcp_ofld_req *req = blk_mq_rq_to_pdu(bd->rq);
+	struct nvme_tcp_ofld_queue *queue = hctx->driver_data;
+	struct nvme_tcp_ofld_ops *ops = queue->dev->ops;
+
+	/* Call nvme_setup_cmd(...) */
+
+	/* Call ops->map_sg(...) */
+
+	return BLK_STS_OK;
+}
+
+static struct blk_mq_ops nvme_tcp_ofld_mq_ops = {
+	.queue_rq	= nvme_tcp_ofld_queue_rq,
+	.init_request	= nvme_tcp_ofld_init_request,
+	/*
+	 * All additional ops will be also implemented and registered similar to
+	 * tcp.c
+	 */
+};
+
+static struct blk_mq_ops nvme_tcp_ofld_admin_mq_ops = {
+	.queue_rq	= nvme_tcp_ofld_queue_rq,
+	.init_request	= nvme_tcp_ofld_init_request,
+	/*
+	 * All additional ops will be also implemented and registered similar to
+	 * tcp.c
+	 */
+};
+
+static const struct nvme_ctrl_ops nvme_tcp_ofld_ctrl_ops = {
+	.name			= "tcp_offload",
+	.module			= THIS_MODULE,
+	.flags			= NVME_F_FABRICS,
+	.reg_read32		= nvmf_reg_read32,
+	.reg_read64		= nvmf_reg_read64,
+	.reg_write32		= nvmf_reg_write32,
+	.free_ctrl		= nvme_tcp_ofld_free_ctrl,
+	.submit_async_event	= nvme_tcp_ofld_submit_async_event,
+	.delete_ctrl		= nvme_tcp_ofld_delete_ctrl,
+	.get_address		= nvmf_get_address,
+};
+
+static bool
+nvme_tcp_ofld_existing_controller(struct nvmf_ctrl_options *opts)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl;
+	bool found = false;
+
+	down_read(&nvme_tcp_ofld_ctrl_rwsem);
+	list_for_each_entry(ctrl, &nvme_tcp_ofld_ctrl_list, list) {
+		found = nvmf_ip_options_match(&ctrl->nctrl, opts);
+		if (found)
+			break;
+	}
+	up_read(&nvme_tcp_ofld_ctrl_rwsem);
+
+	return found;
+}
+
 static struct nvme_ctrl *
 nvme_tcp_ofld_create_ctrl(struct device *ndev, struct nvmf_ctrl_options *opts)
 {
@@ -147,10 +505,48 @@ nvme_tcp_ofld_create_ctrl(struct device *ndev, struct nvmf_ctrl_options *opts)
 	if (!ctrl)
 		return ERR_PTR(-ENOMEM);
 
-	/*
-	 * Init nvme_tcp_ofld_ctrl and nvme_ctrl params based on received
-	 * opts
-	 */
+	INIT_LIST_HEAD(&ctrl->list);
+	nctrl = &ctrl->nctrl;
+	nctrl->opts = opts;
+	nctrl->queue_count = opts->nr_io_queues + opts->nr_write_queues +
+			     opts->nr_poll_queues + 1;
+	nctrl->sqsize = opts->queue_size - 1;
+	nctrl->kato = opts->kato;
+	if (!(opts->mask & NVMF_OPT_TRSVCID)) {
+		opts->trsvcid =
+			kstrdup(__stringify(NVME_TCP_DISC_PORT), GFP_KERNEL);
+		if (!opts->trsvcid) {
+			rc = -ENOMEM;
+			goto out_free_ctrl;
+		}
+		opts->mask |= NVMF_OPT_TRSVCID;
+	}
+
+	rc = inet_pton_with_scope(&init_net, AF_UNSPEC, opts->traddr,
+				  opts->trsvcid,
+				  &ctrl->conn_params.remote_ip_addr);
+	if (rc) {
+		pr_err("malformed address passed: %s:%s\n",
+		       opts->traddr, opts->trsvcid);
+		goto out_free_ctrl;
+	}
+
+	if (opts->mask & NVMF_OPT_HOST_TRADDR) {
+		rc = inet_pton_with_scope(&init_net, AF_UNSPEC,
+					  opts->host_traddr, NULL,
+					  &ctrl->conn_params.local_ip_addr);
+		if (rc) {
+			pr_err("malformed src address passed: %s\n",
+			       opts->host_traddr);
+			goto out_free_ctrl;
+		}
+	}
+
+	if (!opts->duplicate_connect &&
+	    nvme_tcp_ofld_existing_controller(opts)) {
+		rc = -EALREADY;
+		goto out_free_ctrl;
+	}
 
 	/* Find device that can reach the dest addr */
 	dev = nvme_tcp_ofld_lookup_dev(ctrl);
@@ -161,6 +557,10 @@ nvme_tcp_ofld_create_ctrl(struct device *ndev, struct nvmf_ctrl_options *opts)
 		goto out_free_ctrl;
 	}
 
+	rc = nvme_tcp_ofld_check_dev_opts(opts, dev->ops);
+	if (rc)
+		goto out_module_put;
+
 	ctrl->dev = dev;
 	ctrl->queues = kcalloc(nctrl->queue_count,
 			       sizeof(struct nvme_tcp_ofld_queue),
@@ -177,12 +577,27 @@ nvme_tcp_ofld_create_ctrl(struct device *ndev, struct nvmf_ctrl_options *opts)
 		queue->report_err = nvme_tcp_ofld_report_queue_err;
 	}
 
-	/* Call nvme_init_ctrl */
+	rc = nvme_init_ctrl(nctrl, ndev, &nvme_tcp_ofld_ctrl_ops, 0);
+	if (rc)
+		goto out_free_queues;
+
+	if (!nvme_change_ctrl_state(nctrl, NVME_CTRL_CONNECTING)) {
+		WARN_ON_ONCE(1);
+		rc = -EINTR;
+		goto out_uninit_ctrl;
+	}
 
 	rc = nvme_tcp_ofld_setup_ctrl(nctrl, true);
 	if (rc)
 		goto out_uninit_ctrl;
 
+	dev_info(nctrl->device, "new ctrl: NQN \"%s\", addr %pISp\n",
+		 opts->subsysnqn, &ctrl->conn_params.remote_ip_addr);
+
+	down_write(&nvme_tcp_ofld_ctrl_rwsem);
+	list_add_tail(&ctrl->list, &nvme_tcp_ofld_ctrl_list);
+	up_write(&nvme_tcp_ofld_ctrl_rwsem);
+
 	return nctrl;
 
 out_uninit_ctrl:
@@ -192,6 +607,8 @@ nvme_tcp_ofld_create_ctrl(struct device *ndev, struct nvmf_ctrl_options *opts)
 		rc = -EIO;
 
 	return ERR_PTR(rc);
+out_free_queues:
+	kfree(ctrl->queues);
 out_module_put:
 	module_put(dev->ops->module);
 out_free_ctrl:
@@ -221,7 +638,15 @@ static int __init nvme_tcp_ofld_init_module(void)
 
 static void __exit nvme_tcp_ofld_cleanup_module(void)
 {
+	struct nvme_tcp_ofld_ctrl *ctrl;
+
 	nvmf_unregister_transport(&nvme_tcp_ofld_transport);
+
+	down_write(&nvme_tcp_ofld_ctrl_rwsem);
+	list_for_each_entry(ctrl, &nvme_tcp_ofld_ctrl_list, list)
+		nvme_delete_ctrl(&ctrl->nctrl);
+	up_write(&nvme_tcp_ofld_ctrl_rwsem);
+	flush_workqueue(nvme_delete_wq);
 }
 
 module_init(nvme_tcp_ofld_init_module);
drivers/nvme/host/tcp-offload.h +1 -0
index 303e6aab8542..c8a93685fd2a 100644
@@ -76,6 +76,7 @@ struct nvme_tcp_ofld_ctrl_con_params {
 /* Allocated by nvme_tcp_ofld */
 struct nvme_tcp_ofld_ctrl {
 	struct nvme_ctrl nctrl;
+	struct list_head list;
 	struct nvme_tcp_ofld_dev *dev;
 
 	/* admin and IO queues */
Signature · 7 lines
2.22.0


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#5 lore raw

RFC PATCH v2 5/7 nvme-tcp-offload: Add controller level error recovery implementation

In reply to Shai Malin

From: Arie Gershberg <[email protected]>

In this patch, we implement controller level error handling and recovery. Upon an error discovered by the ULP or reset controller initiated by the nvme-core (using reset_ctrl workqueue), the ULP will initiate a controller recovery which includes teardown and re-connect of all queues.

---
 drivers/nvme/host/tcp-offload.c | 121 +++++++++++++++++++++++++++++++-
 drivers/nvme/host/tcp-offload.h |   4 ++
 2 files changed, 124 insertions(+), 1 deletion(-)
drivers/nvme/host/tcp-offload.c +120 -1
index 0d25903d6832..d9da0fc5eb75 100644
@@ -73,6 +73,23 @@ void nvme_tcp_ofld_unregister_dev(struct nvme_tcp_ofld_dev *dev)
 }
 EXPORT_SYMBOL_GPL(nvme_tcp_ofld_unregister_dev);
 
+/**
+ * nvme_tcp_ofld_error_recovery() - NVMeTCP Offload Library error recovery.
+ * function.
+ * @nctrl:	NVMe controller instance to change to resetting.
+ *
+ * API function that change the controller state to resseting.
+ * Part of the overall controller reset sequence.
+ */
+void nvme_tcp_ofld_error_recovery(struct nvme_ctrl *nctrl)
+{
+	if (!nvme_change_ctrl_state(nctrl, NVME_CTRL_RESETTING))
+		return;
+
+	queue_work(nvme_reset_wq, &to_tcp_ofld_ctrl(nctrl)->err_work);
+}
+EXPORT_SYMBOL_GPL(nvme_tcp_ofld_error_recovery);
+
 /**
  * nvme_tcp_ofld_report_queue_err() - NVMeTCP Offload report error event
  * callback function. Pointed to by nvme_tcp_ofld_queue->report_err.
@@ -286,6 +303,27 @@ nvme_tcp_ofld_configure_io_queues(struct nvme_ctrl *nctrl, bool new)
 	return rc;
 }
 
+static void nvme_tcp_ofld_reconnect_or_remove(struct nvme_ctrl *nctrl)
+{
+	/* If we are resetting/deleting then do nothing */
+	if (nctrl->state != NVME_CTRL_CONNECTING) {
+		WARN_ON_ONCE(nctrl->state == NVME_CTRL_NEW ||
+			     nctrl->state == NVME_CTRL_LIVE);
+		return;
+	}
+
+	if (nvmf_should_reconnect(nctrl)) {
+		dev_info(nctrl->device, "Reconnecting in %d seconds...\n",
+			 nctrl->opts->reconnect_delay);
+		queue_delayed_work(nvme_wq,
+				   &to_tcp_ofld_ctrl(nctrl)->connect_work,
+				   nctrl->opts->reconnect_delay * HZ);
+	} else {
+		dev_info(nctrl->device, "Removing controller...\n");
+		nvme_delete_ctrl(nctrl);
+	}
+}
+
 static int nvme_tcp_ofld_setup_ctrl(struct nvme_ctrl *nctrl, bool new)
 {
 	struct nvmf_ctrl_options *opts = nctrl->opts;
@@ -394,10 +432,62 @@ nvme_tcp_ofld_teardown_io_queues(struct nvme_ctrl *nctrl, bool remove)
 	/* Placeholder - teardown_io_queues */
 }
 
+static void nvme_tcp_ofld_reconnect_ctrl_work(struct work_struct *work)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl =
+				container_of(to_delayed_work(work),
+					     struct nvme_tcp_ofld_ctrl,
+					     connect_work);
+	struct nvme_ctrl *nctrl = &ctrl->nctrl;
+
+	++nctrl->nr_reconnects;
+
+	if (nvme_tcp_ofld_setup_ctrl(nctrl, false))
+		goto requeue;
+
+	dev_info(nctrl->device, "Successfully reconnected (%d attempt)\n",
+		 nctrl->nr_reconnects);
+
+	nctrl->nr_reconnects = 0;
+
+	return;
+
+requeue:
+	dev_info(nctrl->device, "Failed reconnect attempt %d\n",
+		 nctrl->nr_reconnects);
+	nvme_tcp_ofld_reconnect_or_remove(nctrl);
+}
+
+static void nvme_tcp_ofld_error_recovery_work(struct work_struct *work)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl =
+		container_of(work, struct nvme_tcp_ofld_ctrl, err_work);
+	struct nvme_ctrl *nctrl = &ctrl->nctrl;
+
+	nvme_stop_keep_alive(nctrl);
+	nvme_tcp_ofld_teardown_io_queues(nctrl, false);
+	/* unquiesce to fail fast pending requests */
+	nvme_start_queues(nctrl);
+	nvme_tcp_ofld_teardown_admin_queue(nctrl, false);
+	blk_mq_unquiesce_queue(nctrl->admin_q);
+
+	if (!nvme_change_ctrl_state(nctrl, NVME_CTRL_CONNECTING)) {
+		/* state change failure is ok if we started nctrl delete */
+		WARN_ON_ONCE(nctrl->state != NVME_CTRL_DELETING &&
+			     nctrl->state != NVME_CTRL_DELETING_NOIO);
+		return;
+	}
+
+	nvme_tcp_ofld_reconnect_or_remove(nctrl);
+}
+
 static void
 nvme_tcp_ofld_teardown_ctrl(struct nvme_ctrl *nctrl, bool shutdown)
 {
-	/* Placeholder - err_work and connect_work */
+	struct nvme_tcp_ofld_ctrl *ctrl = to_tcp_ofld_ctrl(nctrl);
+
+	cancel_work_sync(&ctrl->err_work);
+	cancel_delayed_work_sync(&ctrl->connect_work);
 	nvme_tcp_ofld_teardown_io_queues(nctrl, shutdown);
 	blk_mq_quiesce_queue(nctrl->admin_q);
 	if (shutdown)
@@ -412,6 +502,31 @@ static void nvme_tcp_ofld_delete_ctrl(struct nvme_ctrl *nctrl)
 	nvme_tcp_ofld_teardown_ctrl(nctrl, true);
 }
 
+static void nvme_tcp_ofld_reset_ctrl_work(struct work_struct *work)
+{
+	struct nvme_ctrl *nctrl =
+		container_of(work, struct nvme_ctrl, reset_work);
+
+	nvme_stop_ctrl(nctrl);
+	nvme_tcp_ofld_teardown_ctrl(nctrl, false);
+
+	if (!nvme_change_ctrl_state(nctrl, NVME_CTRL_CONNECTING)) {
+		/* state change failure is ok if we started ctrl delete */
+		WARN_ON_ONCE(nctrl->state != NVME_CTRL_DELETING &&
+			     nctrl->state != NVME_CTRL_DELETING_NOIO);
+		return;
+	}
+
+	if (nvme_tcp_ofld_setup_ctrl(nctrl, false))
+		goto out_fail;
+
+	return;
+
+out_fail:
+	++nctrl->nr_reconnects;
+	nvme_tcp_ofld_reconnect_or_remove(nctrl);
+}
+
 static int
 nvme_tcp_ofld_init_request(struct blk_mq_tag_set *set,
 			   struct request *rq,
@@ -512,6 +627,10 @@ nvme_tcp_ofld_create_ctrl(struct device *ndev, struct nvmf_ctrl_options *opts)
 			     opts->nr_poll_queues + 1;
 	nctrl->sqsize = opts->queue_size - 1;
 	nctrl->kato = opts->kato;
+	INIT_DELAYED_WORK(&ctrl->connect_work,
+			  nvme_tcp_ofld_reconnect_ctrl_work);
+	INIT_WORK(&ctrl->err_work, nvme_tcp_ofld_error_recovery_work);
+	INIT_WORK(&nctrl->reset_work, nvme_tcp_ofld_reset_ctrl_work);
 	if (!(opts->mask & NVMF_OPT_TRSVCID)) {
 		opts->trsvcid =
 			kstrdup(__stringify(NVME_TCP_DISC_PORT), GFP_KERNEL);
drivers/nvme/host/tcp-offload.h +4 -0
index c8a93685fd2a..c3f74f700c6a 100644
@@ -84,6 +84,9 @@ struct nvme_tcp_ofld_ctrl {
 	struct blk_mq_tag_set admin_tag_set;
 	struct nvme_tcp_ofld_queue *queues;
 
+	struct work_struct err_work;
+	struct delayed_work connect_work;
+
 	/*
 	 * Each entry in the array indicates the number of queues of
 	 * corresponding type.
@@ -172,3 +175,4 @@ struct nvme_tcp_ofld_ops {
 /* Exported functions for lower vendor specific offload drivers */
 int nvme_tcp_ofld_register_dev(struct nvme_tcp_ofld_dev *dev);
 void nvme_tcp_ofld_unregister_dev(struct nvme_tcp_ofld_dev *dev);
+void nvme_tcp_ofld_error_recovery(struct nvme_ctrl *nctrl);
Signature · 7 lines
2.22.0


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#6 lore raw

RFC PATCH v2 6/7 nvme-tcp-offload: Add queue level implementation

In reply to Shai Malin

From: Dean Balandin <[email protected]>

In this patch we implement queue level functionality.
The implementation is similar to the nvme-tcp module, the main difference being that we call the vendor specific create_queue op which creates the TCP connection, and NVMeTPC connection including icreq+icresp negotiation.
Once create_queue returns successfully, we can move on to the fabrics connect.

---
 drivers/nvme/host/tcp-offload.c | 307 ++++++++++++++++++++++++++++++--
 drivers/nvme/host/tcp-offload.h |   6 +
 2 files changed, 294 insertions(+), 19 deletions(-)
drivers/nvme/host/tcp-offload.c +288 -19
index d9da0fc5eb75..12f675bdd221 100644
@@ -23,6 +23,11 @@ to_tcp_ofld_ctrl(struct nvme_ctrl *nctrl)
 	return container_of(nctrl, struct nvme_tcp_ofld_ctrl, nctrl);
 }
 
+static inline int nvme_tcp_ofld_qid(struct nvme_tcp_ofld_queue *queue)
+{
+	return queue - queue->ctrl->queues;
+}
+
 /**
  * nvme_tcp_ofld_register_dev() - NVMeTCP Offload Library registration
  * function.
@@ -190,18 +195,98 @@ nvme_tcp_ofld_alloc_tagset(struct nvme_ctrl *nctrl, bool admin)
 	return set;
 }
 
+static bool nvme_tcp_ofld_poll_queue(struct nvme_tcp_ofld_queue *queue)
+{
+	/* Placeholder - implement logic to determine if poll queue */
+
+	return false;
+}
+
+static void __nvme_tcp_ofld_stop_queue(struct nvme_tcp_ofld_queue *queue)
+{
+	queue->dev->ops->drain_queue(queue);
+	queue->dev->ops->destroy_queue(queue);
+
+	/* Placeholder - additional cleanup such as cancel_work_sync io_work */
+	clear_bit(NVME_TCP_OFLD_Q_LIVE, &queue->flags);
+}
+
+static void nvme_tcp_ofld_stop_queue(struct nvme_ctrl *nctrl, int qid)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl = to_tcp_ofld_ctrl(nctrl);
+	struct nvme_tcp_ofld_queue *queue = &ctrl->queues[qid];
+
+	if (!test_and_clear_bit(NVME_TCP_OFLD_Q_LIVE, &queue->flags))
+		return;
+
+	__nvme_tcp_ofld_stop_queue(queue);
+}
+
+static void nvme_tcp_ofld_free_queue(struct nvme_ctrl *nctrl, int qid)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl = to_tcp_ofld_ctrl(nctrl);
+	struct nvme_tcp_ofld_queue *queue = &ctrl->queues[qid];
+
+	if (!test_and_clear_bit(NVME_TCP_OFLD_Q_ALLOCATED, &queue->flags))
+		return;
+
+	/* Placeholder - additional queue cleanup */
+}
+
+static void
+nvme_tcp_ofld_terminate_admin_queue(struct nvme_ctrl *nctrl, bool remove)
+{
+	nvme_tcp_ofld_stop_queue(nctrl, 0);
+	if (remove) {
+		if (nctrl->admin_q && !blk_queue_dead(nctrl->admin_q))
+			blk_cleanup_queue(nctrl->admin_q);
+
+		if (nctrl->fabrics_q)
+			blk_cleanup_queue(nctrl->fabrics_q);
+
+		if (nctrl->admin_tagset)
+			blk_mq_free_tag_set(nctrl->admin_tagset);
+	}
+}
+
 static int nvme_tcp_ofld_start_queue(struct nvme_ctrl *nctrl, int qid)
 {
-	/* Placeholder - start_queue */
-	return 0;
+	struct nvme_tcp_ofld_ctrl *ctrl = to_tcp_ofld_ctrl(nctrl);
+	struct nvme_tcp_ofld_queue *queue = &ctrl->queues[qid];
+	int rc;
+
+	set_bit(NVME_TCP_OFLD_Q_ALLOCATED, &queue->flags);
+	queue = &ctrl->queues[qid];
+	if (qid)
+		rc = nvmf_connect_io_queue(nctrl, qid,
+					   nvme_tcp_ofld_poll_queue(queue));
+	else
+		rc = nvmf_connect_admin_queue(nctrl);
+
+	if (!rc) {
+		set_bit(NVME_TCP_OFLD_Q_LIVE, &queue->flags);
+	} else {
+		if (test_bit(NVME_TCP_OFLD_Q_ALLOCATED, &queue->flags))
+			__nvme_tcp_ofld_stop_queue(queue);
+		dev_err(nctrl->device,
+			"failed to connect queue: %d ret=%d\n", qid, rc);
+	}
+
+	return rc;
 }
 
 static int nvme_tcp_ofld_configure_admin_queue(struct nvme_ctrl *nctrl,
 					       bool new)
 {
+	struct nvme_tcp_ofld_ctrl *ctrl = to_tcp_ofld_ctrl(nctrl);
+	struct nvme_tcp_ofld_queue *queue = &ctrl->queues[0];
 	int rc;
 
-	/* Placeholder - alloc_admin_queue */
+	rc = ctrl->dev->ops->create_queue(queue, 0, NVME_AQ_DEPTH);
+	if (rc)
+		return rc;
+
+	set_bit(NVME_TCP_OFLD_Q_ALLOCATED, &queue->flags);
 	if (new) {
 		nctrl->admin_tagset =
 				nvme_tcp_ofld_alloc_tagset(nctrl, true);
@@ -243,7 +328,7 @@ static int nvme_tcp_ofld_configure_admin_queue(struct nvme_ctrl *nctrl,
 	return 0;
 
 out_stop_queue:
-	/* Placeholder - stop admin queue */
+	nvme_tcp_ofld_stop_queue(nctrl, 0);
 out_cleanup_queue:
 	if (new)
 		blk_cleanup_queue(nctrl->admin_q);
@@ -254,7 +339,115 @@ static int nvme_tcp_ofld_configure_admin_queue(struct nvme_ctrl *nctrl,
 	if (new)
 		blk_mq_free_tag_set(nctrl->admin_tagset);
 out_free_queue:
-	/* Placeholder - free admin queue */
+	nvme_tcp_ofld_free_queue(nctrl, 0);
+
+	return rc;
+}
+
+static unsigned int nvme_tcp_ofld_nr_io_queues(struct nvme_ctrl *nctrl)
+{
+	unsigned int nr_io_queues;
+
+	nr_io_queues = min(nctrl->opts->nr_io_queues, num_online_cpus());
+	nr_io_queues += min(nctrl->opts->nr_write_queues, num_online_cpus());
+	nr_io_queues += min(nctrl->opts->nr_poll_queues, num_online_cpus());
+
+	return nr_io_queues;
+}
+
+static void
+nvme_tcp_ofld_set_io_queues(struct nvme_ctrl *nctrl, unsigned int nr_io_queues)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl = to_tcp_ofld_ctrl(nctrl);
+	struct nvmf_ctrl_options *opts = nctrl->opts;
+
+	if (opts->nr_write_queues && opts->nr_io_queues < nr_io_queues) {
+		/*
+		 * separate read/write queues
+		 * hand out dedicated default queues only after we have
+		 * sufficient read queues.
+		 */
+		ctrl->queue_type_mapping[HCTX_TYPE_READ] = opts->nr_io_queues;
+		nr_io_queues -= ctrl->queue_type_mapping[HCTX_TYPE_READ];
+		ctrl->queue_type_mapping[HCTX_TYPE_DEFAULT] =
+			min(opts->nr_write_queues, nr_io_queues);
+		nr_io_queues -= ctrl->queue_type_mapping[HCTX_TYPE_DEFAULT];
+	} else {
+		/*
+		 * shared read/write queues
+		 * either no write queues were requested, or we don't have
+		 * sufficient queue count to have dedicated default queues.
+		 */
+		ctrl->queue_type_mapping[HCTX_TYPE_DEFAULT] =
+			min(opts->nr_io_queues, nr_io_queues);
+		nr_io_queues -= ctrl->queue_type_mapping[HCTX_TYPE_DEFAULT];
+	}
+}
+
+static void
+nvme_tcp_ofld_terminate_io_queues(struct nvme_ctrl *nctrl, int start_from)
+{
+	int i;
+
+	/* admin-q will be ignored because of the loop condition */
+	for (i = start_from; i >= 1; i--)
+		nvme_tcp_ofld_stop_queue(nctrl, i);
+}
+
+static int nvme_tcp_ofld_create_io_queues(struct nvme_ctrl *nctrl)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl = to_tcp_ofld_ctrl(nctrl);
+	int i, rc;
+
+	for (i = 1; i < nctrl->queue_count; i++) {
+		rc = ctrl->dev->ops->create_queue(&ctrl->queues[i],
+						  i, nctrl->sqsize + 1);
+		if (rc)
+			goto out_free_queues;
+	}
+
+	return 0;
+
+out_free_queues:
+	nvme_tcp_ofld_terminate_io_queues(nctrl, --i);
+
+	return rc;
+}
+
+static int nvme_tcp_ofld_alloc_io_queues(struct nvme_ctrl *nctrl)
+{
+	unsigned int nr_io_queues;
+	int rc;
+
+	nr_io_queues = nvme_tcp_ofld_nr_io_queues(nctrl);
+	rc = nvme_set_queue_count(nctrl, &nr_io_queues);
+	if (rc)
+		return rc;
+
+	nctrl->queue_count = nr_io_queues + 1;
+	if (nctrl->queue_count < 2)
+		return 0;
+
+	dev_info(nctrl->device, "creating %d I/O queues.\n", nr_io_queues);
+	nvme_tcp_ofld_set_io_queues(nctrl, nr_io_queues);
+
+	return nvme_tcp_ofld_create_io_queues(nctrl);
+}
+
+static int nvme_tcp_ofld_start_io_queues(struct nvme_ctrl *nctrl)
+{
+	int i, rc = 0;
+
+	for (i = 1; i < nctrl->queue_count; i++) {
+		rc = nvme_tcp_ofld_start_queue(nctrl, i);
+		if (rc)
+			goto terminate_queues;
+	}
+
+	return 0;
+
+terminate_queues:
+	nvme_tcp_ofld_terminate_io_queues(nctrl, --i);
 
 	return rc;
 }
@@ -262,9 +455,10 @@ static int nvme_tcp_ofld_configure_admin_queue(struct nvme_ctrl *nctrl,
 static int
 nvme_tcp_ofld_configure_io_queues(struct nvme_ctrl *nctrl, bool new)
 {
-	int rc;
+	int rc = nvme_tcp_ofld_alloc_io_queues(nctrl);
 
-	/* Placeholder - alloc_io_queues */
+	if (rc)
+		return rc;
 
 	if (new) {
 		nctrl->tagset = nvme_tcp_ofld_alloc_tagset(nctrl, false);
@@ -282,7 +476,9 @@ nvme_tcp_ofld_configure_io_queues(struct nvme_ctrl *nctrl, bool new)
 		}
 	}
 
-	/* Placeholder - start_io_queues */
+	rc = nvme_tcp_ofld_start_io_queues(nctrl);
+	if (rc)
+		goto out_cleanup_connect_q;
 
 	if (!new) {
 		nvme_start_queues(nctrl);
@@ -294,11 +490,14 @@ nvme_tcp_ofld_configure_io_queues(struct nvme_ctrl *nctrl, bool new)
 
 	return 0;
 
+out_cleanup_connect_q:
+	if (new && nctrl->connect_q)
+		blk_cleanup_queue(nctrl->connect_q);
 out_free_tag_set:
 	if (new)
 		blk_mq_free_tag_set(nctrl->tagset);
 out_free_io_queues:
-	/* Placeholder - free_io_queues */
+	nvme_tcp_ofld_terminate_io_queues(nctrl, nctrl->queue_count);
 
 	return rc;
 }
@@ -374,9 +573,9 @@ static int nvme_tcp_ofld_setup_ctrl(struct nvme_ctrl *nctrl, bool new)
 	return 0;
 
 destroy_io:
-	/* Placeholder - stop and destroy io queues*/
+	nvme_tcp_ofld_terminate_io_queues(nctrl, nctrl->queue_count);
 destroy_admin:
-	/* Placeholder - stop and destroy admin queue*/
+	nvme_tcp_ofld_terminate_admin_queue(nctrl, new);
 
 	return rc;
 }
@@ -559,22 +758,92 @@ nvme_tcp_ofld_queue_rq(struct blk_mq_hw_ctx *hctx,
 	return BLK_STS_OK;
 }
 
+static void
+nvme_tcp_ofld_exit_request(struct blk_mq_tag_set *set,
+			   struct request *rq, unsigned int hctx_idx)
+{
+	/* Placeholder */
+}
+
+static int
+nvme_tcp_ofld_init_hctx(struct blk_mq_hw_ctx *hctx, void *data,
+			unsigned int hctx_idx)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl = data;
+	struct nvme_tcp_ofld_queue *queue = &ctrl->queues[hctx_idx + 1];
+
+	hctx->driver_data = queue;
+	return 0;
+}
+
+static int nvme_tcp_ofld_map_queues(struct blk_mq_tag_set *set)
+{
+	struct nvme_tcp_ofld_ctrl *ctrl = set->driver_data;
+	struct nvmf_ctrl_options *opts = ctrl->nctrl.opts;
+
+	if (opts->nr_write_queues && ctrl->queue_type_mapping[HCTX_TYPE_READ]) {
+		/* separate read/write queues */
+		set->map[HCTX_TYPE_DEFAULT].nr_queues =
+			ctrl->queue_type_mapping[HCTX_TYPE_DEFAULT];
+		set->map[HCTX_TYPE_DEFAULT].queue_offset = 0;
+		set->map[HCTX_TYPE_READ].nr_queues =
+			ctrl->queue_type_mapping[HCTX_TYPE_READ];
+		set->map[HCTX_TYPE_READ].queue_offset =
+			ctrl->queue_type_mapping[HCTX_TYPE_DEFAULT];
+	} else {
+		/* shared read/write queues */
+		set->map[HCTX_TYPE_DEFAULT].nr_queues =
+			ctrl->queue_type_mapping[HCTX_TYPE_DEFAULT];
+		set->map[HCTX_TYPE_DEFAULT].queue_offset = 0;
+		set->map[HCTX_TYPE_READ].nr_queues =
+			ctrl->queue_type_mapping[HCTX_TYPE_DEFAULT];
+		set->map[HCTX_TYPE_READ].queue_offset = 0;
+	}
+	blk_mq_map_queues(&set->map[HCTX_TYPE_DEFAULT]);
+	blk_mq_map_queues(&set->map[HCTX_TYPE_READ]);
+
+	if (opts->nr_poll_queues && ctrl->queue_type_mapping[HCTX_TYPE_POLL]) {
+		/* map dedicated poll queues only if we have queues left */
+		set->map[HCTX_TYPE_POLL].nr_queues =
+				ctrl->queue_type_mapping[HCTX_TYPE_POLL];
+		set->map[HCTX_TYPE_POLL].queue_offset =
+			ctrl->queue_type_mapping[HCTX_TYPE_DEFAULT] +
+			ctrl->queue_type_mapping[HCTX_TYPE_READ];
+		blk_mq_map_queues(&set->map[HCTX_TYPE_POLL]);
+	}
+
+	dev_info(ctrl->nctrl.device,
+		 "mapped %d/%d/%d default/read/poll queues.\n",
+		 ctrl->queue_type_mapping[HCTX_TYPE_DEFAULT],
+		 ctrl->queue_type_mapping[HCTX_TYPE_READ],
+		 ctrl->queue_type_mapping[HCTX_TYPE_POLL]);
+
+	return 0;
+}
+
+static int nvme_tcp_ofld_poll(struct blk_mq_hw_ctx *hctx)
+{
+	/* Placeholder - Implement polling mechanism */
+
+	return 0;
+}
+
 static struct blk_mq_ops nvme_tcp_ofld_mq_ops = {
 	.queue_rq	= nvme_tcp_ofld_queue_rq,
 	.init_request	= nvme_tcp_ofld_init_request,
-	/*
-	 * All additional ops will be also implemented and registered similar to
-	 * tcp.c
-	 */
+	.complete	= nvme_complete_rq,
+	.exit_request	= nvme_tcp_ofld_exit_request,
+	.init_hctx	= nvme_tcp_ofld_init_hctx,
+	.map_queues	= nvme_tcp_ofld_map_queues,
+	.poll		= nvme_tcp_ofld_poll,
 };
 
 static struct blk_mq_ops nvme_tcp_ofld_admin_mq_ops = {
 	.queue_rq	= nvme_tcp_ofld_queue_rq,
 	.init_request	= nvme_tcp_ofld_init_request,
-	/*
-	 * All additional ops will be also implemented and registered similar to
-	 * tcp.c
-	 */
+	.complete	= nvme_complete_rq,
+	.exit_request	= nvme_tcp_ofld_exit_request,
+	.init_hctx	= nvme_tcp_ofld_init_hctx,
 };
 
 static const struct nvme_ctrl_ops nvme_tcp_ofld_ctrl_ops = {
drivers/nvme/host/tcp-offload.h +6 -0
index c3f74f700c6a..02a2592d4dca 100644
@@ -42,11 +42,17 @@ struct nvme_tcp_ofld_req {
 		     __le16 status);
 };
 
+enum nvme_tcp_ofld_queue_flags {
+	NVME_TCP_OFLD_Q_ALLOCATED = 0,
+	NVME_TCP_OFLD_Q_LIVE = 1,
+};
+
 /* Allocated by nvme_tcp_ofld */
 struct nvme_tcp_ofld_queue {
 	/* Offload device associated to this queue */
 	struct nvme_tcp_ofld_dev *dev;
 	struct nvme_tcp_ofld_ctrl *ctrl;
+	unsigned long flags;
 
 	/* Vendor specific driver context */
 	void *private_data;
Signature · 7 lines
2.22.0


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme
author
#7 lore raw

RFC PATCH v2 7/7 nvme-tcp-offload: Add IO level implementation

In reply to Shai Malin

From: Dean Balandin <[email protected]>

In this patch, we present the IO level functionality.
The nvme-tcp-offload shall work on the IO-level, meaning the nvme-tcp-offload ULP module shall pass the request to the nvme-tcp-offload vendor driver and shall expect for the request compilation.
No additional handling is needed in between, this design will reduce the CPU utilization as we will describe below.

The nvme-tcp-offload vendor driver shall register to nvme-tcp-offload ULP with the following IO-path ops:
- init_req
- map_sg - in order to map the request sg (similar to nvme_rdma_map_data).
- send_req - in order to pass the request to the handling of the offload driver that shall pass it to the vendor specific device.
- poll_queue

The vendor driver will manage the context from which the request will be executed and the request aggregations.
Once the IO completed, the nvme-tcp-offload vendor driver shall call command.done() that shall invoke the nvme-tcp-offload ULP layer for completing the request.

---
 drivers/nvme/host/tcp-offload.c | 80 +++++++++++++++++++++++++++++----
 1 file changed, 72 insertions(+), 8 deletions(-)
drivers/nvme/host/tcp-offload.c +72 -8
index 12f675bdd221..37046ed46aea 100644
@@ -113,6 +113,7 @@ int nvme_tcp_ofld_report_queue_err(struct nvme_tcp_ofld_queue *queue)
 /**
  * nvme_tcp_ofld_req_done() - NVMeTCP Offload request done callback
  * function. Pointed to by nvme_tcp_ofld_req->done.
+ * Handles both NVME_TCP_F_DATA_SUCCESS flag and NVMe CQE.
  * @req:	NVMeTCP offload request to complete.
  * @result:     The nvme_result.
  * @status:     The completion status.
@@ -125,7 +126,10 @@ nvme_tcp_ofld_req_done(struct nvme_tcp_ofld_req *req,
 		       union nvme_result *result,
 		       __le16 status)
 {
-	/* Placeholder - complete request with/without error */
+	struct request *rq = blk_mq_rq_from_pdu(req);
+
+	if (!nvme_try_complete_req(rq, cpu_to_le16(status << 1), *result))
+		nvme_complete_rq(rq);
 }
 
 struct nvme_tcp_ofld_dev *
@@ -734,9 +738,10 @@ nvme_tcp_ofld_init_request(struct blk_mq_tag_set *set,
 {
 	struct nvme_tcp_ofld_req *req = blk_mq_rq_to_pdu(rq);
 	struct nvme_tcp_ofld_ctrl *ctrl = set->driver_data;
+	int qid = (set == &ctrl->tag_set) ? hctx_idx + 1 : 0;
 
-	/* Placeholder - init request */
-
+	req->queue = &ctrl->queues[qid];
+	nvme_req(rq)->ctrl = &ctrl->nctrl;
 	req->done = nvme_tcp_ofld_req_done;
 	ctrl->dev->ops->init_req(req);
 
@@ -749,11 +754,30 @@ nvme_tcp_ofld_queue_rq(struct blk_mq_hw_ctx *hctx,
 {
 	struct nvme_tcp_ofld_req *req = blk_mq_rq_to_pdu(bd->rq);
 	struct nvme_tcp_ofld_queue *queue = hctx->driver_data;
-	struct nvme_tcp_ofld_ops *ops = queue->dev->ops;
+	struct nvme_tcp_ofld_ctrl *ctrl = queue->ctrl;
+	struct nvme_ns *ns = hctx->queue->queuedata;
+	struct nvme_tcp_ofld_dev *dev = queue->dev;
+	struct nvme_tcp_ofld_ops *ops = dev->ops;
+	struct request *rq = bd->rq;
+	bool queue_ready;
+	int rc;
+
+	queue_ready = test_bit(NVME_TCP_OFLD_Q_LIVE, &queue->flags);
+	if (!nvmf_check_ready(&ctrl->nctrl, rq, queue_ready))
+		return nvmf_fail_nonready_command(&ctrl->nctrl, rq);
+
+	rc = nvme_setup_cmd(ns, rq, &req->nvme_cmd);
+	if (unlikely(rc))
+		return rc;
 
-	/* Call nvme_setup_cmd(...) */
+	blk_mq_start_request(rq);
+	rc = ops->map_sg(dev, req);
+	if (unlikely(rc))
+		return rc;
 
-	/* Call ops->map_sg(...) */
+	rc = ops->send_req(req);
+	if (unlikely(rc))
+		return rc;
 
 	return BLK_STS_OK;
 }
@@ -823,9 +847,47 @@ static int nvme_tcp_ofld_map_queues(struct blk_mq_tag_set *set)
 
 static int nvme_tcp_ofld_poll(struct blk_mq_hw_ctx *hctx)
 {
-	/* Placeholder - Implement polling mechanism */
+	struct nvme_tcp_ofld_queue *queue = hctx->driver_data;
+	struct nvme_tcp_ofld_dev *dev = queue->dev;
+	struct nvme_tcp_ofld_ops *ops = dev->ops;
 
-	return 0;
+	return ops->poll_queue(queue);
+}
+
+static enum blk_eh_timer_return
+nvme_tcp_ofld_timeout(struct request *rq, bool reserved)
+{
+	struct nvme_tcp_ofld_req *req = blk_mq_rq_to_pdu(rq);
+	struct nvme_tcp_ofld_ctrl *ctrl = req->queue->ctrl;
+
+	/* Restart the timer if a controller reset is already scheduled. Any
+	 * timed out request would be handled before entering the connecting
+	 * state.
+	 */
+	if (ctrl->nctrl.state == NVME_CTRL_RESETTING)
+		return BLK_EH_RESET_TIMER;
+
+	dev_warn(ctrl->nctrl.device,
+		 "queue %d: timeout request %#x type %d\n",
+		 nvme_tcp_ofld_qid(req->queue), rq->tag,
+		 req->nvme_cmd.common.opcode);
+
+	if (ctrl->nctrl.state != NVME_CTRL_LIVE) {
+		/*
+		 * Teardown immediately if controller times out while starting
+		 * or we are already started error recovery. all outstanding
+		 * requests are completed on shutdown, so we return BLK_EH_DONE.
+		 */
+		flush_work(&ctrl->err_work);
+		nvme_tcp_ofld_teardown_io_queues(&ctrl->nctrl, false);
+		nvme_tcp_ofld_teardown_admin_queue(&ctrl->nctrl, false);
+		return BLK_EH_DONE;
+	}
+
+	dev_warn(ctrl->nctrl.device, "starting error recovery\n");
+	nvme_tcp_ofld_error_recovery(&ctrl->nctrl);
+
+	return BLK_EH_RESET_TIMER;
 }
 
 static struct blk_mq_ops nvme_tcp_ofld_mq_ops = {
@@ -835,6 +897,7 @@ static struct blk_mq_ops nvme_tcp_ofld_mq_ops = {
 	.exit_request	= nvme_tcp_ofld_exit_request,
 	.init_hctx	= nvme_tcp_ofld_init_hctx,
 	.map_queues	= nvme_tcp_ofld_map_queues,
+	.timeout	= nvme_tcp_ofld_timeout,
 	.poll		= nvme_tcp_ofld_poll,
 };
 
@@ -844,6 +907,7 @@ static struct blk_mq_ops nvme_tcp_ofld_admin_mq_ops = {
 	.complete	= nvme_complete_rq,
 	.exit_request	= nvme_tcp_ofld_exit_request,
 	.init_hctx	= nvme_tcp_ofld_init_hctx,
+	.timeout	= nvme_tcp_ofld_timeout,
 };
 
 static const struct nvme_ctrl_ops nvme_tcp_ofld_ctrl_ops = {
Signature · 7 lines
2.22.0


_______________________________________________
Linux-nvme mailing list
[email protected]
http://lists.infradead.org/mailman/listinfo/linux-nvme