Merge tag 'nvme-6.16-2025-06-05' of git://git.infradead.org/nvme into block-6.16

axboe · axboe · commit 6f65947a1e68 · 2025-06-05T07:40:38.000-06:00
Pull NVMe updates and fixes from Christoph:

"nvme updates for Linux 6.16

 - TCP error handling fix (Shin'ichiro Kawasaki)
 - TCP I/O stall handling fixes (Hannes Reinecke)
 - fix command limits status code (Keith Busch)
 - support vectored buffers also for passthrough (Pavel Begunkov)
 - spelling fixes (Yi Zhang)"

* tag 'nvme-6.16-2025-06-05' of git://git.infradead.org/nvme:
  nvme: spelling fixes
  nvme-tcp: fix I/O stalls on congested sockets
  nvme-tcp: sanitize request list handling
  nvme-tcp: remove tag set when second admin queue config fails
  nvme: enable vectored registered bufs for passthrough cmds
  nvme: fix implicit bool to flags conversion
  nvme: fix command limits status code
diff --git a/drivers/nvme/common/auth.c b/drivers/nvme/common/auth.c
@@ -471,7 +471,7 @@ EXPORT_SYMBOL_GPL(nvme_auth_generate_key);
  * @c1: Value of challenge C1
  * @c2: Value of challenge C2
  * @hash_len: Hash length of the hash algorithm
- * @ret_psk: Pointer too the resulting generated PSK
+ * @ret_psk: Pointer to the resulting generated PSK
  * @ret_len: length of @ret_psk
  *
  * Generate a PSK for TLS as specified in NVMe base specification, section
@@ -759,8 +759,8 @@ int nvme_auth_derive_tls_psk(int hmac_id, u8 *psk, size_t psk_len,
 		goto out_free_prk;
 
 	/*
-	 * 2 addtional bytes for the length field from HDKF-Expand-Label,
-	 * 2 addtional bytes for the HMAC ID, and one byte for the space
+	 * 2 additional bytes for the length field from HDKF-Expand-Label,
+	 * 2 additional bytes for the HMAC ID, and one byte for the space
 	 * separator.
 	 */
 	info_len = strlen(psk_digest) + strlen(psk_prefix) + 5;
diff --git a/drivers/nvme/host/Kconfig b/drivers/nvme/host/Kconfig
@@ -106,7 +106,7 @@ config NVME_TCP_TLS
 	help
 	  Enables TLS encryption for NVMe TCP using the netlink handshake API.
 
-	  The TLS handshake daemon is availble at
+	  The TLS handshake daemon is available at
 	  https://github.com/oracle/ktls-utils.
 
 	  If unsure, say N.
diff --git a/drivers/nvme/host/constants.c b/drivers/nvme/host/constants.c
@@ -145,7 +145,7 @@ static const char * const nvme_statuses[] = {
 	[NVME_SC_BAD_ATTRIBUTES] = "Conflicting Attributes",
 	[NVME_SC_INVALID_PI] = "Invalid Protection Information",
 	[NVME_SC_READ_ONLY] = "Attempted Write to Read Only Range",
-	[NVME_SC_ONCS_NOT_SUPPORTED] = "ONCS Not Supported",
+	[NVME_SC_CMD_SIZE_LIM_EXCEEDED	] = "Command Size Limits Exceeded",
 	[NVME_SC_ZONE_BOUNDARY_ERROR] = "Zoned Boundary Error",
 	[NVME_SC_ZONE_FULL] = "Zone Is Full",
 	[NVME_SC_ZONE_READ_ONLY] = "Zone Is Read Only",
diff --git a/drivers/nvme/host/core.c b/drivers/nvme/host/core.c
@@ -290,7 +290,6 @@ static blk_status_t nvme_error_status(u16 status)
 	case NVME_SC_NS_NOT_READY:
 		return BLK_STS_TARGET;
 	case NVME_SC_BAD_ATTRIBUTES:
-	case NVME_SC_ONCS_NOT_SUPPORTED:
 	case NVME_SC_INVALID_OPCODE:
 	case NVME_SC_INVALID_FIELD:
 	case NVME_SC_INVALID_NS:
@@ -1027,7 +1026,7 @@ static inline blk_status_t nvme_setup_rw(struct nvme_ns *ns,
 
 	if (ns->head->ms) {
 		/*
-		 * If formated with metadata, the block layer always provides a
+		 * If formatted with metadata, the block layer always provides a
 		 * metadata buffer if CONFIG_BLK_DEV_INTEGRITY is enabled.  Else
 		 * we enable the PRACT bit for protection information or set the
 		 * namespace capacity to zero to prevent any I/O.
diff --git a/drivers/nvme/host/fabrics.c b/drivers/nvme/host/fabrics.c
@@ -582,7 +582,7 @@ EXPORT_SYMBOL_GPL(nvmf_connect_io_queue);
  * Do not retry when:
  *
  * - the DNR bit is set and the specification states no further connect
- *   attempts with the same set of paramenters should be attempted.
+ *   attempts with the same set of parameters should be attempted.
  *
  * - when the authentication attempt fails, because the key was invalid.
  *   This error code is set on the host side.
diff --git a/drivers/nvme/host/fabrics.h b/drivers/nvme/host/fabrics.h
@@ -80,7 +80,7 @@ enum {
  * @transport:	Holds the fabric transport "technology name" (for a lack of
  *		better description) that will be used by an NVMe controller
  *		being added.
- * @subsysnqn:	Hold the fully qualified NQN subystem name (format defined
+ * @subsysnqn:	Hold the fully qualified NQN subsystem name (format defined
  *		in the NVMe specification, "NVMe Qualified Names").
  * @traddr:	The transport-specific TRADDR field for a port on the
  *              subsystem which is adding a controller.
@@ -156,7 +156,7 @@ struct nvmf_ctrl_options {
  * @create_ctrl():	function pointer that points to a non-NVMe
  *			implementation-specific fabric technology
  *			that would go into starting up that fabric
- *			for the purpose of conneciton to an NVMe controller
+ *			for the purpose of connection to an NVMe controller
  *			using that fabric technology.
  *
  * Notes:
@@ -165,7 +165,7 @@ struct nvmf_ctrl_options {
  *	2. create_ctrl() must be defined (even if it does nothing)
  *	3. struct nvmf_transport_ops must be statically allocated in the
  *	   modules .bss section so that a pure module_get on @module
- *	   prevents the memory from beeing freed.
+ *	   prevents the memory from being freed.
  */
 struct nvmf_transport_ops {
 	struct list_head	entry;
diff --git a/drivers/nvme/host/fc.c b/drivers/nvme/host/fc.c
@@ -1955,7 +1955,7 @@ nvme_fc_fcpio_done(struct nvmefc_fcp_req *req)
 	}
 
 	/*
-	 * For the linux implementation, if we have an unsuccesful
+	 * For the linux implementation, if we have an unsucceesful
 	 * status, they blk-mq layer can typically be called with the
 	 * non-zero status and the content of the cqe isn't important.
 	 */
@@ -2479,7 +2479,7 @@ __nvme_fc_abort_outstanding_ios(struct nvme_fc_ctrl *ctrl, bool start_queues)
 	 * writing the registers for shutdown and polling (call
 	 * nvme_disable_ctrl()). Given a bunch of i/o was potentially
 	 * just aborted and we will wait on those contexts, and given
-	 * there was no indication of how live the controlelr is on the
+	 * there was no indication of how live the controller is on the
 	 * link, don't send more io to create more contexts for the
 	 * shutdown. Let the controller fail via keepalive failure if
 	 * its still present.
diff --git a/drivers/nvme/host/ioctl.c b/drivers/nvme/host/ioctl.c
@@ -493,13 +493,15 @@ static int nvme_uring_cmd_io(struct nvme_ctrl *ctrl, struct nvme_ns *ns,
 	d.timeout_ms = READ_ONCE(cmd->timeout_ms);
 
 	if (d.data_len && (ioucmd->flags & IORING_URING_CMD_FIXED)) {
-		/* fixedbufs is only for non-vectored io */
-		if (vec)
-			return -EINVAL;
+		int ddir = nvme_is_write(&c) ? WRITE : READ;
 
-		ret = io_uring_cmd_import_fixed(d.addr, d.data_len,
-			nvme_is_write(&c) ? WRITE : READ, &iter, ioucmd,
-			issue_flags);
+		if (vec)
+			ret = io_uring_cmd_import_fixed_vec(ioucmd,
+					u64_to_user_ptr(d.addr), d.data_len,
+					ddir, &iter, issue_flags);
+		else
+			ret = io_uring_cmd_import_fixed(d.addr, d.data_len,
+					ddir, &iter, ioucmd, issue_flags);
 		if (ret < 0)
 			return ret;
 
@@ -521,7 +523,7 @@ static int nvme_uring_cmd_io(struct nvme_ctrl *ctrl, struct nvme_ns *ns,
 	if (d.data_len) {
 		ret = nvme_map_user_request(req, d.addr, d.data_len,
 			nvme_to_user_ptr(d.metadata), d.metadata_len,
-			map_iter, vec);
+			map_iter, vec ? NVME_IOCTL_VEC : 0);
 		if (ret)
 			goto out_free_req;
 	}
@@ -727,7 +729,7 @@ int nvme_ns_head_ioctl(struct block_device *bdev, blk_mode_t mode,
 
 	/*
 	 * Handle ioctls that apply to the controller instead of the namespace
-	 * seperately and drop the ns SRCU reference early.  This avoids a
+	 * separately and drop the ns SRCU reference early.  This avoids a
 	 * deadlock when deleting namespaces using the passthrough interface.
 	 */
 	if (is_ctrl_ioctl(cmd))
diff --git a/drivers/nvme/host/multipath.c b/drivers/nvme/host/multipath.c
@@ -760,7 +760,7 @@ int nvme_mpath_alloc_disk(struct nvme_ctrl *ctrl, struct nvme_ns_head *head)
 	 * controller's scan_work context. If a path error occurs here, the IO
 	 * will wait until a path becomes available or all paths are torn down,
 	 * but that action also occurs within scan_work, so it would deadlock.
-	 * Defer the partion scan to a different context that does not block
+	 * Defer the partition scan to a different context that does not block
 	 * scan_work.
 	 */
 	set_bit(GD_SUPPRESS_PART_SCAN, &head->disk->state);
diff --git a/drivers/nvme/host/nvme.h b/drivers/nvme/host/nvme.h
@@ -523,7 +523,7 @@ static inline bool nvme_ns_head_multipath(struct nvme_ns_head *head)
 enum nvme_ns_features {
 	NVME_NS_EXT_LBAS = 1 << 0, /* support extended LBA format */
 	NVME_NS_METADATA_SUPPORTED = 1 << 1, /* support getting generated md */
-	NVME_NS_DEAC = 1 << 2,		/* DEAC bit in Write Zeores supported */
+	NVME_NS_DEAC = 1 << 2,		/* DEAC bit in Write Zeroes supported */
 };
 
 struct nvme_ns {
diff --git a/drivers/nvme/host/pci.c b/drivers/nvme/host/pci.c
@@ -3015,7 +3015,7 @@ static void nvme_reset_work(struct work_struct *work)
 		goto out;
 
 	/*
-	 * Freeze and update the number of I/O queues as thos might have
+	 * Freeze and update the number of I/O queues as those might have
 	 * changed.  If there are no I/O queues left after this reset, keep the
 	 * controller around but remove all namespaces.
 	 */
@@ -3186,7 +3186,7 @@ static unsigned long check_vendor_combination_bug(struct pci_dev *pdev)
 		/*
 		 * Exclude some Kingston NV1 and A2000 devices from
 		 * NVME_QUIRK_SIMPLE_SUSPEND. Do a full suspend to save a
-		 * lot fo energy with s2idle sleep on some TUXEDO platforms.
+		 * lot of energy with s2idle sleep on some TUXEDO platforms.
 		 */
 		if (dmi_match(DMI_BOARD_NAME, "NS5X_NS7XAU") ||
 		    dmi_match(DMI_BOARD_NAME, "NS5x_7xAU") ||
diff --git a/drivers/nvme/host/pr.c b/drivers/nvme/host/pr.c
@@ -82,8 +82,6 @@ static int nvme_status_to_pr_err(int status)
 		return PR_STS_SUCCESS;
 	case NVME_SC_RESERVATION_CONFLICT:
 		return PR_STS_RESERVATION_CONFLICT;
-	case NVME_SC_ONCS_NOT_SUPPORTED:
-		return -EOPNOTSUPP;
 	case NVME_SC_BAD_ATTRIBUTES:
 	case NVME_SC_INVALID_OPCODE:
 	case NVME_SC_INVALID_FIELD:
diff --git a/drivers/nvme/host/rdma.c b/drivers/nvme/host/rdma.c
@@ -221,7 +221,7 @@ static struct nvme_rdma_qe *nvme_rdma_alloc_ring(struct ib_device *ibdev,
 
 	/*
 	 * Bind the CQEs (post recv buffers) DMA mapping to the RDMA queue
-	 * lifetime. It's safe, since any chage in the underlying RDMA device
+	 * lifetime. It's safe, since any change in the underlying RDMA device
 	 * will issue error recovery and queue re-creation.
 	 */
 	for (i = 0; i < ib_queue_size; i++) {
@@ -800,7 +800,7 @@ static int nvme_rdma_configure_admin_queue(struct nvme_rdma_ctrl *ctrl,
 
 	/*
 	 * Bind the async event SQE DMA mapping to the admin queue lifetime.
-	 * It's safe, since any chage in the underlying RDMA device will issue
+	 * It's safe, since any change in the underlying RDMA device will issue
 	 * error recovery and queue re-creation.
 	 */
 	error = nvme_rdma_alloc_qe(ctrl->device->dev, &ctrl->async_event_sqe,
diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
@@ -452,7 +452,8 @@ nvme_tcp_fetch_request(struct nvme_tcp_queue *queue)
 			return NULL;
 	}
 
-	list_del(&req->entry);
+	list_del_init(&req->entry);
+	init_llist_node(&req->lentry);
 	return req;
 }
 
@@ -560,6 +561,8 @@ static int nvme_tcp_init_request(struct blk_mq_tag_set *set,
 	req->queue = queue;
 	nvme_req(rq)->ctrl = &ctrl->ctrl;
 	nvme_req(rq)->cmd = &pdu->cmd;
+	init_llist_node(&req->lentry);
+	INIT_LIST_HEAD(&req->entry);
 
 	return 0;
 }
@@ -764,6 +767,14 @@ static int nvme_tcp_handle_r2t(struct nvme_tcp_queue *queue,
 		return -EPROTO;
 	}
 
+	if (llist_on_list(&req->lentry) ||
+	    !list_empty(&req->entry)) {
+		dev_err(queue->ctrl->ctrl.device,
+			"req %d unexpected r2t while processing request\n",
+			rq->tag);
+		return -EPROTO;
+	}
+
 	req->pdu_len = 0;
 	req->h2cdata_left = r2t_length;
 	req->h2cdata_offset = r2t_offset;
@@ -1350,7 +1361,7 @@ static int nvme_tcp_try_recv(struct nvme_tcp_queue *queue)
 	queue->nr_cqe = 0;
 	consumed = sock->ops->read_sock(sk, &rd_desc, nvme_tcp_recv_skb);
 	release_sock(sk);
-	return consumed;
+	return consumed == -EAGAIN ? 0 : consumed;
 }
 
 static void nvme_tcp_io_work(struct work_struct *w)
@@ -1378,6 +1389,11 @@ static void nvme_tcp_io_work(struct work_struct *w)
 		else if (unlikely(result < 0))
 			return;
 
+		/* did we get some space after spending time in recv? */
+		if (nvme_tcp_queue_has_pending(queue) &&
+		    sk_stream_is_writeable(queue->sock->sk))
+			pending = true;
+
 		if (!pending || !queue->rd_enabled)
 			return;
 
@@ -2394,7 +2410,7 @@ static int nvme_tcp_setup_ctrl(struct nvme_ctrl *ctrl, bool new)
 		nvme_tcp_teardown_admin_queue(ctrl, false);
 		ret = nvme_tcp_configure_admin_queue(ctrl, false);
 		if (ret)
-			return ret;
+			goto destroy_admin;
 	}
 
 	if (ctrl->icdoff) {
@@ -2638,6 +2654,8 @@ static void nvme_tcp_submit_async_event(struct nvme_ctrl *arg)
 	ctrl->async_req.offset = 0;
 	ctrl->async_req.curr_bio = NULL;
 	ctrl->async_req.data_len = 0;
+	init_llist_node(&ctrl->async_req.lentry);
+	INIT_LIST_HEAD(&ctrl->async_req.entry);
 
 	nvme_tcp_queue_request(&ctrl->async_req, true);
 }
diff --git a/drivers/nvme/target/admin-cmd.c b/drivers/nvme/target/admin-cmd.c
@@ -1165,7 +1165,7 @@ static void nvmet_execute_identify(struct nvmet_req *req)
  * A "minimum 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
+ * to be executed and return immediately telling the command to abort
  * wasn't found.
  */
 static void nvmet_execute_abort(struct nvmet_req *req)
diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c
@@ -62,14 +62,7 @@ inline u16 errno_to_nvme_status(struct nvmet_req *req, int errno)
 		return  NVME_SC_LBA_RANGE | NVME_STATUS_DNR;
 	case -EOPNOTSUPP:
 		req->error_loc = offsetof(struct nvme_common_command, opcode);
-		switch (req->cmd->common.opcode) {
-		case nvme_cmd_dsm:
-		case nvme_cmd_write_zeroes:
-			return NVME_SC_ONCS_NOT_SUPPORTED | NVME_STATUS_DNR;
-		default:
-			return NVME_SC_INVALID_OPCODE | NVME_STATUS_DNR;
-		}
-		break;
+		return NVME_SC_INVALID_OPCODE | NVME_STATUS_DNR;
 	case -ENODATA:
 		req->error_loc = offsetof(struct nvme_rw_command, nsid);
 		return NVME_SC_ACCESS_DENIED;
@@ -651,7 +644,7 @@ void nvmet_ns_disable(struct nvmet_ns *ns)
 	 * 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
+	 * using the namespace 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.
 	 */
diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
@@ -1339,7 +1339,7 @@ nvmet_fc_portentry_rebind_tgt(struct nvmet_fc_tgtport *tgtport)
 /**
  * nvmet_fc_register_targetport - transport entry point called by an
  *                              LLDD to register the existence of a local
- *                              NVME subystem FC port.
+ *                              NVME subsystem FC port.
  * @pinfo:     pointer to information about the port to be registered
  * @template:  LLDD entrypoints and operational parameters for the port
  * @dev:       physical hardware device node port corresponds to. Will be
diff --git a/drivers/nvme/target/io-cmd-bdev.c b/drivers/nvme/target/io-cmd-bdev.c
@@ -133,7 +133,7 @@ u16 blk_to_nvme_status(struct nvmet_req *req, blk_status_t blk_sts)
 	 * Right now there exists M : 1 mapping between block layer error
 	 * to the NVMe status code (see nvme_error_status()). For consistency,
 	 * when we reverse map we use most appropriate NVMe Status code from
-	 * the group of the NVMe staus codes used in the nvme_error_status().
+	 * the group of the NVMe status codes used in the nvme_error_status().
 	 */
 	switch (blk_sts) {
 	case BLK_STS_NOSPC:
@@ -145,15 +145,8 @@ u16 blk_to_nvme_status(struct nvmet_req *req, blk_status_t blk_sts)
 		req->error_loc = offsetof(struct nvme_rw_command, slba);
 		break;
 	case BLK_STS_NOTSUPP:
+		status = NVME_SC_INVALID_OPCODE | NVME_STATUS_DNR;
 		req->error_loc = offsetof(struct nvme_common_command, opcode);
-		switch (req->cmd->common.opcode) {
-		case nvme_cmd_dsm:
-		case nvme_cmd_write_zeroes:
-			status = NVME_SC_ONCS_NOT_SUPPORTED | NVME_STATUS_DNR;
-			break;
-		default:
-			status = NVME_SC_INVALID_OPCODE | NVME_STATUS_DNR;
-		}
 		break;
 	case BLK_STS_MEDIUM:
 		status = NVME_SC_ACCESS_DENIED;
diff --git a/drivers/nvme/target/passthru.c b/drivers/nvme/target/passthru.c
@@ -99,7 +99,7 @@ static u16 nvmet_passthru_override_id_ctrl(struct nvmet_req *req)
 
 	/*
 	 * The passthru NVMe driver may have a limit on the number of segments
-	 * which depends on the host's memory fragementation. To solve this,
+	 * which depends on the host's memory fragmentation. To solve this,
 	 * ensure mdts is limited to the pages equal to the number of segments.
 	 */
 	max_hw_sectors = min_not_zero(pctrl->max_segments << PAGE_SECTORS_SHIFT,
diff --git a/include/linux/nvme.h b/include/linux/nvme.h
@@ -2171,7 +2171,7 @@ enum {
 	NVME_SC_BAD_ATTRIBUTES		= 0x180,
 	NVME_SC_INVALID_PI		= 0x181,
 	NVME_SC_READ_ONLY		= 0x182,
-	NVME_SC_ONCS_NOT_SUPPORTED	= 0x183,
+	NVME_SC_CMD_SIZE_LIM_EXCEEDED	= 0x183,
 
 	/*
 	 * I/O Command Set Specific - Fabrics commands:

Original file line number	Diff line number	Diff line change
`@@ -582,7 +582,7 @@ EXPORT_SYMBOL_GPL(nvmf_connect_io_queue);`
`582`	`582`	`* Do not retry when:`
`583`	`583`	`*`
`584`	`584`	`* - the DNR bit is set and the specification states no further connect`
`585`		`- * attempts with the same set of paramenters should be attempted.`
	`585`	`+ * attempts with the same set of parameters should be attempted.`
`586`	`586`	`*`
`587`	`587`	`* - when the authentication attempt fails, because the key was invalid.`
`588`	`588`	`* This error code is set on the host side.`