summaryrefslogtreecommitdiff
path: root/drivers/nvme/target
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2020-06-02 15:37:03 -0700
committerLinus Torvalds <torvalds@linux-foundation.org>2020-06-02 15:37:03 -0700
commitbce159d734091fe31340976081577333f52a85e4 (patch)
tree8396be51e6703797a60aefb4992e729f327d27c2 /drivers/nvme/target
parent750a02ab8d3c49ca7d23102be90d3d1db19e2827 (diff)
parent0c8d3fceade2ab1bbac68bca013e62bfdb851d19 (diff)
Merge tag 'for-5.8/drivers-2020-06-01' of git://git.kernel.dk/linux-block
Pull block driver updates from Jens Axboe: "On top of the core changes, here are the block driver changes for this merge window: - NVMe changes: - NVMe over Fibre Channel protocol updates, which also reach over to drivers/scsi/lpfc (James Smart) - namespace revalidation support on the target (Anthony Iliopoulos) - gcc zero length array fix (Arnd Bergmann) - nvmet cleanups (Chaitanya Kulkarni) - misc cleanups and fixes (me, Keith Busch, Sagi Grimberg) - use a SRQ per completion vector (Max Gurtovoy) - fix handling of runtime changes to the queue count (Weiping Zhang) - t10 protection information support for nvme-rdma and nvmet-rdma (Israel Rukshin and Max Gurtovoy) - target side AEN improvements (Chaitanya Kulkarni) - various fixes and minor improvements all over, icluding the nvme part of the lpfc driver" - Floppy code cleanup series (Willy, Denis) - Floppy contention fix (Jiri) - Loop CONFIGURE support (Martijn) - bcache fixes/improvements (Coly, Joe, Colin) - q->queuedata cleanups (Christoph) - Get rid of ioctl_by_bdev (Christoph, Stefan) - md/raid5 allocation fixes (Coly) - zero length array fixes (Gustavo) - swim3 task state fix (Xu)" * tag 'for-5.8/drivers-2020-06-01' of git://git.kernel.dk/linux-block: (166 commits) bcache: configure the asynchronous registertion to be experimental bcache: asynchronous devices registration bcache: fix refcount underflow in bcache_device_free() bcache: Convert pr_<level> uses to a more typical style bcache: remove redundant variables i and n lpfc: Fix return value in __lpfc_nvme_ls_abort lpfc: fix axchg pointer reference after free and double frees lpfc: Fix pointer checks and comments in LS receive refactoring nvme: set dma alignment to qword nvmet: cleanups the loop in nvmet_async_events_process nvmet: fix memory leak when removing namespaces and controllers concurrently nvmet-rdma: add metadata/T10-PI support nvmet: add metadata support for block devices nvmet: add metadata/T10-PI support nvme: add Metadata Capabilities enumerations nvmet: rename nvmet_check_data_len to nvmet_check_transfer_len nvmet: rename nvmet_rw_len to nvmet_rw_data_len nvmet: add metadata characteristics for a namespace nvme-rdma: add metadata/T10-PI support nvme-rdma: introduce nvme_rdma_sgl structure ...
Diffstat (limited to 'drivers/nvme/target')
-rw-r--r--drivers/nvme/target/Kconfig1
-rw-r--r--drivers/nvme/target/admin-cmd.c42
-rw-r--r--drivers/nvme/target/configfs.c272
-rw-r--r--drivers/nvme/target/core.c166
-rw-r--r--drivers/nvme/target/discovery.c8
-rw-r--r--drivers/nvme/target/fabrics-cmd.c15
-rw-r--r--drivers/nvme/target/fc.c805
-rw-r--r--drivers/nvme/target/fcloop.c155
-rw-r--r--drivers/nvme/target/io-cmd-bdev.c118
-rw-r--r--drivers/nvme/target/io-cmd-file.c23
-rw-r--r--drivers/nvme/target/nvmet.h36
-rw-r--r--drivers/nvme/target/rdma.c416
-rw-r--r--drivers/nvme/target/tcp.c53
-rw-r--r--drivers/nvme/target/trace.h28
14 files changed, 1627 insertions, 511 deletions
diff --git a/drivers/nvme/target/Kconfig b/drivers/nvme/target/Kconfig
index d7f48c0fb311..4474952d64c6 100644
--- a/drivers/nvme/target/Kconfig
+++ b/drivers/nvme/target/Kconfig
@@ -4,6 +4,7 @@ config NVME_TARGET
tristate "NVMe Target support"
depends on BLOCK
depends on CONFIGFS_FS
+ select BLK_DEV_INTEGRITY_T10 if BLK_DEV_INTEGRITY
select SGL_ALLOC
help
This enabled target side support for the NVMe protocol, that is
diff --git a/drivers/nvme/target/admin-cmd.c b/drivers/nvme/target/admin-cmd.c
index 9d6f75cfa77c..1db8c0498668 100644
--- a/drivers/nvme/target/admin-cmd.c
+++ b/drivers/nvme/target/admin-cmd.c
@@ -295,7 +295,7 @@ out:
static void nvmet_execute_get_log_page(struct nvmet_req *req)
{
- if (!nvmet_check_data_len(req, nvmet_get_log_page_len(req->cmd)))
+ if (!nvmet_check_transfer_len(req, nvmet_get_log_page_len(req->cmd)))
return;
switch (req->cmd->get_log_page.lid) {
@@ -341,6 +341,7 @@ static void nvmet_execute_identify_ctrl(struct nvmet_req *req)
{
struct nvmet_ctrl *ctrl = req->sq->ctrl;
struct nvme_id_ctrl *id;
+ u32 cmd_capsule_size;
u16 status = 0;
id = kzalloc(sizeof(*id), GFP_KERNEL);
@@ -433,9 +434,15 @@ static void nvmet_execute_identify_ctrl(struct nvmet_req *req)
strlcpy(id->subnqn, ctrl->subsys->subsysnqn, sizeof(id->subnqn));
- /* Max command capsule size is sqe + single page of in-capsule data */
- id->ioccsz = cpu_to_le32((sizeof(struct nvme_command) +
- req->port->inline_data_size) / 16);
+ /*
+ * Max command capsule size is sqe + in-capsule data size.
+ * Disable in-capsule data for Metadata capable controllers.
+ */
+ cmd_capsule_size = sizeof(struct nvme_command);
+ if (!ctrl->pi_support)
+ cmd_capsule_size += req->port->inline_data_size;
+ id->ioccsz = cpu_to_le32(cmd_capsule_size / 16);
+
/* Max response capsule size is cqe */
id->iorcsz = cpu_to_le32(sizeof(struct nvme_completion) / 16);
@@ -465,6 +472,7 @@ out:
static void nvmet_execute_identify_ns(struct nvmet_req *req)
{
+ struct nvmet_ctrl *ctrl = req->sq->ctrl;
struct nvmet_ns *ns;
struct nvme_id_ns *id;
u16 status = 0;
@@ -482,10 +490,12 @@ static void nvmet_execute_identify_ns(struct nvmet_req *req)
}
/* return an all zeroed buffer if we can't find an active namespace */
- ns = nvmet_find_namespace(req->sq->ctrl, req->cmd->identify.nsid);
+ ns = nvmet_find_namespace(ctrl, req->cmd->identify.nsid);
if (!ns)
goto done;
+ nvmet_ns_revalidate(ns);
+
/*
* nuse = ncap = nsze isn't always true, but we have no way to find
* that out from the underlying device.
@@ -521,6 +531,16 @@ static void nvmet_execute_identify_ns(struct nvmet_req *req)
id->lbaf[0].ds = ns->blksize_shift;
+ if (ctrl->pi_support && nvmet_ns_has_pi(ns)) {
+ id->dpc = NVME_NS_DPC_PI_FIRST | NVME_NS_DPC_PI_LAST |
+ NVME_NS_DPC_PI_TYPE1 | NVME_NS_DPC_PI_TYPE2 |
+ NVME_NS_DPC_PI_TYPE3;
+ id->mc = NVME_MC_EXTENDED_LBA;
+ id->dps = ns->pi_type;
+ id->flbas = NVME_NS_FLBAS_META_EXT;
+ id->lbaf[0].ms = cpu_to_le16(ns->metadata_size);
+ }
+
if (ns->readonly)
id->nsattr |= (1 << 0);
nvmet_put_namespace(ns);
@@ -625,7 +645,7 @@ out:
static void nvmet_execute_identify(struct nvmet_req *req)
{
- if (!nvmet_check_data_len(req, NVME_IDENTIFY_DATA_SIZE))
+ if (!nvmet_check_transfer_len(req, NVME_IDENTIFY_DATA_SIZE))
return;
switch (req->cmd->identify.cns) {
@@ -654,7 +674,7 @@ static void nvmet_execute_identify(struct nvmet_req *req)
*/
static void nvmet_execute_abort(struct nvmet_req *req)
{
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
nvmet_set_result(req, 1);
nvmet_req_complete(req, 0);
@@ -743,7 +763,7 @@ static void nvmet_execute_set_features(struct nvmet_req *req)
u16 nsqr;
u16 ncqr;
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
switch (cdw10 & 0xff) {
@@ -815,7 +835,7 @@ static void nvmet_execute_get_features(struct nvmet_req *req)
u32 cdw10 = le32_to_cpu(req->cmd->common.cdw10);
u16 status = 0;
- if (!nvmet_check_data_len(req, nvmet_feat_data_len(req, cdw10)))
+ if (!nvmet_check_transfer_len(req, nvmet_feat_data_len(req, cdw10)))
return;
switch (cdw10 & 0xff) {
@@ -882,7 +902,7 @@ void nvmet_execute_async_event(struct nvmet_req *req)
{
struct nvmet_ctrl *ctrl = req->sq->ctrl;
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
mutex_lock(&ctrl->lock);
@@ -901,7 +921,7 @@ void nvmet_execute_keep_alive(struct nvmet_req *req)
{
struct nvmet_ctrl *ctrl = req->sq->ctrl;
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
pr_debug("ctrl %d update keep-alive timer for %d secs\n",
diff --git a/drivers/nvme/target/configfs.c b/drivers/nvme/target/configfs.c
index 58cabd7b6fc5..419e0d4ce79b 100644
--- a/drivers/nvme/target/configfs.c
+++ b/drivers/nvme/target/configfs.c
@@ -20,61 +20,71 @@ static const struct config_item_type nvmet_subsys_type;
static LIST_HEAD(nvmet_ports_list);
struct list_head *nvmet_ports = &nvmet_ports_list;
-static const struct nvmet_transport_name {
+struct nvmet_type_name_map {
u8 type;
const char *name;
-} nvmet_transport_names[] = {
+};
+
+static struct nvmet_type_name_map nvmet_transport[] = {
{ NVMF_TRTYPE_RDMA, "rdma" },
{ NVMF_TRTYPE_FC, "fc" },
{ NVMF_TRTYPE_TCP, "tcp" },
{ NVMF_TRTYPE_LOOP, "loop" },
};
+static const struct nvmet_type_name_map nvmet_addr_family[] = {
+ { NVMF_ADDR_FAMILY_PCI, "pcie" },
+ { NVMF_ADDR_FAMILY_IP4, "ipv4" },
+ { NVMF_ADDR_FAMILY_IP6, "ipv6" },
+ { NVMF_ADDR_FAMILY_IB, "ib" },
+ { NVMF_ADDR_FAMILY_FC, "fc" },
+ { NVMF_ADDR_FAMILY_LOOP, "loop" },
+};
+
+static bool nvmet_is_port_enabled(struct nvmet_port *p, const char *caller)
+{
+ if (p->enabled)
+ pr_err("Disable port '%u' before changing attribute in %s\n",
+ le16_to_cpu(p->disc_addr.portid), caller);
+ return p->enabled;
+}
+
/*
* 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)
+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");
- case NVMF_ADDR_FAMILY_FC:
- return sprintf(page, "fc\n");
- default:
- return sprintf(page, "\n");
+ u8 adrfam = to_nvmet_port(item)->disc_addr.adrfam;
+ int i;
+
+ for (i = 1; i < ARRAY_SIZE(nvmet_addr_family); i++) {
+ if (nvmet_addr_family[i].type == adrfam)
+ return sprintf(page, "%s\n", nvmet_addr_family[i].name);
}
+
+ 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);
+ int i;
- if (port->enabled) {
- pr_err("Cannot modify address while enabled\n");
- pr_err("Disable the address before modifying\n");
+ if (nvmet_is_port_enabled(port, __func__))
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 if (sysfs_streq(page, "fc")) {
- port->disc_addr.adrfam = NVMF_ADDR_FAMILY_FC;
- } else {
- pr_err("Invalid value '%s' for adrfam\n", page);
- return -EINVAL;
+ for (i = 1; i < ARRAY_SIZE(nvmet_addr_family); i++) {
+ if (sysfs_streq(page, nvmet_addr_family[i].name))
+ goto found;
}
+ pr_err("Invalid value '%s' for adrfam\n", page);
+ return -EINVAL;
+
+found:
+ port->disc_addr.adrfam = nvmet_addr_family[i].type;
return count;
}
@@ -100,11 +110,9 @@ static ssize_t nvmet_addr_portid_store(struct config_item *item,
return -EINVAL;
}
- if (port->enabled) {
- pr_err("Cannot modify address while enabled\n");
- pr_err("Disable the address before modifying\n");
+ if (nvmet_is_port_enabled(port, __func__))
return -EACCES;
- }
+
port->disc_addr.portid = cpu_to_le16(portid);
return count;
}
@@ -130,11 +138,8 @@ static ssize_t nvmet_addr_traddr_store(struct config_item *item,
return -EINVAL;
}
- if (port->enabled) {
- pr_err("Cannot modify address while enabled\n");
- pr_err("Disable the address before modifying\n");
+ if (nvmet_is_port_enabled(port, __func__))
return -EACCES;
- }
if (sscanf(page, "%s\n", port->disc_addr.traddr) != 1)
return -EINVAL;
@@ -143,20 +148,24 @@ static ssize_t nvmet_addr_traddr_store(struct config_item *item,
CONFIGFS_ATTR(nvmet_, addr_traddr);
-static ssize_t nvmet_addr_treq_show(struct config_item *item,
- char *page)
+static const struct nvmet_type_name_map nvmet_addr_treq[] = {
+ { NVMF_TREQ_NOT_SPECIFIED, "not specified" },
+ { NVMF_TREQ_REQUIRED, "required" },
+ { NVMF_TREQ_NOT_REQUIRED, "not required" },
+};
+
+static ssize_t nvmet_addr_treq_show(struct config_item *item, char *page)
{
- switch (to_nvmet_port(item)->disc_addr.treq &
- NVME_TREQ_SECURE_CHANNEL_MASK) {
- 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");
+ u8 treq = to_nvmet_port(item)->disc_addr.treq &
+ NVME_TREQ_SECURE_CHANNEL_MASK;
+ int i;
+
+ for (i = 0; i < ARRAY_SIZE(nvmet_addr_treq); i++) {
+ if (treq == nvmet_addr_treq[i].type)
+ return sprintf(page, "%s\n", nvmet_addr_treq[i].name);
}
+
+ return sprintf(page, "\n");
}
static ssize_t nvmet_addr_treq_store(struct config_item *item,
@@ -164,25 +173,22 @@ static ssize_t nvmet_addr_treq_store(struct config_item *item,
{
struct nvmet_port *port = to_nvmet_port(item);
u8 treq = port->disc_addr.treq & ~NVME_TREQ_SECURE_CHANNEL_MASK;
+ int i;
- if (port->enabled) {
- pr_err("Cannot modify address while enabled\n");
- pr_err("Disable the address before modifying\n");
+ if (nvmet_is_port_enabled(port, __func__))
return -EACCES;
- }
- if (sysfs_streq(page, "not specified")) {
- treq |= NVMF_TREQ_NOT_SPECIFIED;
- } else if (sysfs_streq(page, "required")) {
- treq |= NVMF_TREQ_REQUIRED;
- } else if (sysfs_streq(page, "not required")) {
- treq |= NVMF_TREQ_NOT_REQUIRED;
- } else {
- pr_err("Invalid value '%s' for treq\n", page);
- return -EINVAL;
+ for (i = 0; i < ARRAY_SIZE(nvmet_addr_treq); i++) {
+ if (sysfs_streq(page, nvmet_addr_treq[i].name))
+ goto found;
}
- port->disc_addr.treq = treq;
+ pr_err("Invalid value '%s' for treq\n", page);
+ return -EINVAL;
+
+found:
+ treq |= nvmet_addr_treq[i].type;
+ port->disc_addr.treq = treq;
return count;
}
@@ -206,11 +212,8 @@ static ssize_t nvmet_addr_trsvcid_store(struct config_item *item,
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");
+ if (nvmet_is_port_enabled(port, __func__))
return -EACCES;
- }
if (sscanf(page, "%s\n", port->disc_addr.trsvcid) != 1)
return -EINVAL;
@@ -233,11 +236,8 @@ static ssize_t nvmet_param_inline_data_size_store(struct config_item *item,
struct nvmet_port *port = to_nvmet_port(item);
int ret;
- if (port->enabled) {
- pr_err("Cannot modify inline_data_size while port enabled\n");
- pr_err("Disable the port before modifying\n");
+ if (nvmet_is_port_enabled(port, __func__))
return -EACCES;
- }
ret = kstrtoint(page, 0, &port->inline_data_size);
if (ret) {
pr_err("Invalid value '%s' for inline_data_size\n", page);
@@ -248,16 +248,45 @@ static ssize_t nvmet_param_inline_data_size_store(struct config_item *item,
CONFIGFS_ATTR(nvmet_, param_inline_data_size);
+#ifdef CONFIG_BLK_DEV_INTEGRITY
+static ssize_t nvmet_param_pi_enable_show(struct config_item *item,
+ char *page)
+{
+ struct nvmet_port *port = to_nvmet_port(item);
+
+ return snprintf(page, PAGE_SIZE, "%d\n", port->pi_enable);
+}
+
+static ssize_t nvmet_param_pi_enable_store(struct config_item *item,
+ const char *page, size_t count)
+{
+ struct nvmet_port *port = to_nvmet_port(item);
+ bool val;
+
+ if (strtobool(page, &val))
+ return -EINVAL;
+
+ if (port->enabled) {
+ pr_err("Disable port before setting pi_enable value.\n");
+ return -EACCES;
+ }
+
+ port->pi_enable = val;
+ return count;
+}
+
+CONFIGFS_ATTR(nvmet_, param_pi_enable);
+#endif
+
static ssize_t nvmet_addr_trtype_show(struct config_item *item,
char *page)
{
struct nvmet_port *port = to_nvmet_port(item);
int i;
- for (i = 0; i < ARRAY_SIZE(nvmet_transport_names); i++) {
- if (port->disc_addr.trtype != nvmet_transport_names[i].type)
- continue;
- return sprintf(page, "%s\n", nvmet_transport_names[i].name);
+ for (i = 0; i < ARRAY_SIZE(nvmet_transport); i++) {
+ if (port->disc_addr.trtype == nvmet_transport[i].type)
+ return sprintf(page, "%s\n", nvmet_transport[i].name);
}
return sprintf(page, "\n");
@@ -276,22 +305,20 @@ static ssize_t nvmet_addr_trtype_store(struct config_item *item,
struct nvmet_port *port = to_nvmet_port(item);
int i;
- if (port->enabled) {
- pr_err("Cannot modify address while enabled\n");
- pr_err("Disable the address before modifying\n");
+ if (nvmet_is_port_enabled(port, __func__))
return -EACCES;
- }
- for (i = 0; i < ARRAY_SIZE(nvmet_transport_names); i++) {
- if (sysfs_streq(page, nvmet_transport_names[i].name))
+ for (i = 0; i < ARRAY_SIZE(nvmet_transport); i++) {
+ if (sysfs_streq(page, nvmet_transport[i].name))
goto found;
}
pr_err("Invalid value '%s' for trtype\n", page);
return -EINVAL;
+
found:
memset(&port->disc_addr.tsas, 0, NVMF_TSAS_SIZE);
- port->disc_addr.trtype = nvmet_transport_names[i].type;
+ port->disc_addr.trtype = nvmet_transport[i].type;
if (port->disc_addr.trtype == NVMF_TRTYPE_RDMA)
nvmet_port_init_tsas_rdma(port);
return count;
@@ -327,7 +354,7 @@ static ssize_t nvmet_ns_device_path_store(struct config_item *item,
kfree(ns->device_path);
ret = -ENOMEM;
- ns->device_path = kstrndup(page, len, GFP_KERNEL);
+ ns->device_path = kmemdup_nul(page, len, GFP_KERNEL);
if (!ns->device_path)
goto out_unlock;
@@ -543,6 +570,31 @@ static ssize_t nvmet_ns_buffered_io_store(struct config_item *item,
CONFIGFS_ATTR(nvmet_ns_, buffered_io);
+static ssize_t nvmet_ns_revalidate_size_store(struct config_item *item,
+ const char *page, size_t count)
+{
+ struct nvmet_ns *ns = to_nvmet_ns(item);
+ bool val;
+
+ if (strtobool(page, &val))
+ return -EINVAL;
+
+ if (!val)
+ return -EINVAL;
+
+ mutex_lock(&ns->subsys->lock);
+ if (!ns->enabled) {
+ pr_err("enable ns before revalidate.\n");
+ mutex_unlock(&ns->subsys->lock);
+ return -EINVAL;
+ }
+ nvmet_ns_revalidate(ns);
+ mutex_unlock(&ns->subsys->lock);
+ return count;
+}
+
+CONFIGFS_ATTR_WO(nvmet_ns_, revalidate_size);
+
static struct configfs_attribute *nvmet_ns_attrs[] = {
&nvmet_ns_attr_device_path,
&nvmet_ns_attr_device_nguid,
@@ -550,6 +602,7 @@ static struct configfs_attribute *nvmet_ns_attrs[] = {
&nvmet_ns_attr_ana_grpid,
&nvmet_ns_attr_enable,
&nvmet_ns_attr_buffered_io,
+ &nvmet_ns_attr_revalidate_size,
#ifdef CONFIG_PCI_P2PDMA
&nvmet_ns_attr_p2pmem,
#endif
@@ -963,7 +1016,7 @@ static ssize_t nvmet_subsys_attr_model_store(struct config_item *item,
return -EINVAL;
}
- new_model_number = kstrndup(page, len, GFP_KERNEL);
+ new_model_number = kmemdup_nul(page, len, GFP_KERNEL);
if (!new_model_number)
return -ENOMEM;
@@ -987,6 +1040,28 @@ static ssize_t nvmet_subsys_attr_model_store(struct config_item *item,
}
CONFIGFS_ATTR(nvmet_subsys_, attr_model);
+#ifdef CONFIG_BLK_DEV_INTEGRITY
+static ssize_t nvmet_subsys_attr_pi_enable_show(struct config_item *item,
+ char *page)
+{
+ return snprintf(page, PAGE_SIZE, "%d\n", to_subsys(item)->pi_support);
+}
+
+static ssize_t nvmet_subsys_attr_pi_enable_store(struct config_item *item,
+ const char *page, size_t count)
+{
+ struct nvmet_subsys *subsys = to_subsys(item);
+ bool pi_enable;
+
+ if (strtobool(page, &pi_enable))
+ return -EINVAL;
+
+ subsys->pi_support = pi_enable;
+ return count;
+}
+CONFIGFS_ATTR(nvmet_subsys_, attr_pi_enable);
+#endif
+
static struct configfs_attribute *nvmet_subsys_attrs[] = {
&nvmet_subsys_attr_attr_allow_any_host,
&nvmet_subsys_attr_attr_version,
@@ -994,6 +1069,9 @@ static struct configfs_attribute *nvmet_subsys_attrs[] = {
&nvmet_subsys_attr_attr_cntlid_min,
&nvmet_subsys_attr_attr_cntlid_max,
&nvmet_subsys_attr_attr_model,
+#ifdef CONFIG_BLK_DEV_INTEGRITY
+ &nvmet_subsys_attr_attr_pi_enable,
+#endif
NULL,
};
@@ -1149,10 +1227,7 @@ static const struct config_item_type nvmet_referrals_type = {
.ct_group_ops = &nvmet_referral_group_ops,
};
-static struct {
- enum nvme_ana_state state;
- const char *name;
-} nvmet_ana_state_names[] = {
+static struct nvmet_type_name_map nvmet_ana_state[] = {
{ NVME_ANA_OPTIMIZED, "optimized" },
{ NVME_ANA_NONOPTIMIZED, "non-optimized" },
{ NVME_ANA_INACCESSIBLE, "inaccessible" },
@@ -1167,10 +1242,9 @@ static ssize_t nvmet_ana_group_ana_state_show(struct config_item *item,
enum nvme_ana_state state = grp->port->ana_state[grp->grpid];
int i;
- for (i = 0; i < ARRAY_SIZE(nvmet_ana_state_names); i++) {
- if (state != nvmet_ana_state_names[i].state)
- continue;
- return sprintf(page, "%s\n", nvmet_ana_state_names[i].name);
+ for (i = 0; i < ARRAY_SIZE(nvmet_ana_state); i++) {
+ if (state == nvmet_ana_state[i].type)
+ return sprintf(page, "%s\n", nvmet_ana_state[i].name);
}
return sprintf(page, "\n");
@@ -1180,10 +1254,11 @@ static ssize_t nvmet_ana_group_ana_state_store(struct config_item *item,
const char *page, size_t count)
{
struct nvmet_ana_group *grp = to_ana_group(item);
+ enum nvme_ana_state *ana_state = grp->port->ana_state;
int i;
- for (i = 0; i < ARRAY_SIZE(nvmet_ana_state_names); i++) {
- if (sysfs_streq(page, nvmet_ana_state_names[i].name))
+ for (i = 0; i < ARRAY_SIZE(nvmet_ana_state); i++) {
+ if (sysfs_streq(page, nvmet_ana_state[i].name))
goto found;
}
@@ -1192,10 +1267,9 @@ static ssize_t nvmet_ana_group_ana_state_store(struct config_item *item,
found:
down_write(&nvmet_ana_sem);
- grp->port->ana_state[grp->grpid] = nvmet_ana_state_names[i].state;
+ ana_state[grp->grpid] = (enum nvme_ana_state) nvmet_ana_state[i].type;
nvmet_ana_chgcnt++;
up_write(&nvmet_ana_sem);
-
nvmet_port_send_ana_event(grp->port);
return count;
}
@@ -1297,6 +1371,9 @@ static struct configfs_attribute *nvmet_port_attrs[] = {
&nvmet_attr_addr_trsvcid,
&nvmet_attr_addr_trtype,
&nvmet_attr_param_inline_data_size,
+#ifdef CONFIG_BLK_DEV_INTEGRITY
+ &nvmet_attr_param_pi_enable,
+#endif
NULL,
};
@@ -1346,6 +1423,7 @@ static struct config_group *nvmet_ports_make(struct config_group *group,
port->inline_data_size = -1; /* < 0 == let the transport choose */
port->disc_addr.portid = cpu_to_le16(portid);
+ port->disc_addr.adrfam = NVMF_ADDR_FAMILY_MAX;
port->disc_addr.treq = NVMF_TREQ_DISABLE_SQFLOW;
config_group_init_type_name(&port->group, name, &nvmet_port_type);
diff --git a/drivers/nvme/target/core.c b/drivers/nvme/target/core.c
index b685f99d56a1..6392bcd30bd7 100644
--- a/drivers/nvme/target/core.c
+++ b/drivers/nvme/target/core.c
@@ -134,15 +134,10 @@ static void nvmet_async_events_process(struct nvmet_ctrl *ctrl, u16 status)
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);
- break;
- }
-
+ mutex_lock(&ctrl->lock);
+ while (ctrl->nr_async_event_cmds && !list_empty(&ctrl->async_events)) {
+ aen = list_first_entry(&ctrl->async_events,
+ struct nvmet_async_event, entry);
req = ctrl->async_event_cmds[--ctrl->nr_async_event_cmds];
if (status == 0)
nvmet_set_result(req, nvmet_async_event_result(aen));
@@ -151,20 +146,21 @@ static void nvmet_async_events_process(struct nvmet_ctrl *ctrl, u16 status)
kfree(aen);
mutex_unlock(&ctrl->lock);
+ trace_nvmet_async_event(ctrl, req->cqe->result.u32);
nvmet_req_complete(req, status);
+ mutex_lock(&ctrl->lock);
}
+ mutex_unlock(&ctrl->lock);
}
static void nvmet_async_events_free(struct nvmet_ctrl *ctrl)
{
- struct nvmet_req *req;
+ struct nvmet_async_event *aen, *tmp;
mutex_lock(&ctrl->lock);
- while (ctrl->nr_async_event_cmds) {
- req = ctrl->async_event_cmds[--ctrl->nr_async_event_cmds];
- mutex_unlock(&ctrl->lock);
- nvmet_req_complete(req, NVME_SC_INTERNAL | NVME_SC_DNR);
- mutex_lock(&ctrl->lock);
+ list_for_each_entry_safe(aen, tmp, &ctrl->async_events, entry) {
+ list_del(&aen->entry);
+ kfree(aen);
}
mutex_unlock(&ctrl->lock);
}
@@ -322,12 +318,21 @@ int nvmet_enable_port(struct nvmet_port *port)
if (!try_module_get(ops->owner))
return -EINVAL;
- ret = ops->add_port(port);
- if (ret) {
- module_put(ops->owner);
- return ret;
+ /*
+ * If the user requested PI support and the transport isn't pi capable,
+ * don't enable the port.
+ */
+ if (port->pi_enable && !ops->metadata_support) {
+ pr_err("T10-PI is not supported by transport type %d\n",
+ port->disc_addr.trtype);
+ ret = -EINVAL;
+ goto out_put;
}
+ ret = ops->add_port(port);
+ if (ret)
+ goto out_put;
+
/* If the transport didn't set inline_data_size, then disable it. */
if (port->inline_data_size < 0)
port->inline_data_size = 0;
@@ -335,6 +340,10 @@ int nvmet_enable_port(struct nvmet_port *port)
port->enabled = true;
port->tr_ops = ops;
return 0;
+
+out_put:
+ module_put(ops->owner);
+ return ret;
}
void nvmet_disable_port(struct nvmet_port *port)
@@ -514,6 +523,19 @@ static void nvmet_p2pmem_ns_add_p2p(struct nvmet_ctrl *ctrl,
ns->nsid);
}
+void nvmet_ns_revalidate(struct nvmet_ns *ns)
+{
+ loff_t oldsize = ns->size;
+
+ if (ns->bdev)
+ nvmet_bdev_ns_revalidate(ns);
+ else
+ nvmet_file_ns_revalidate(ns);
+
+ if (oldsize != ns->size)
+ nvmet_ns_changed(ns->subsys, ns->nsid);
+}
+
int nvmet_ns_enable(struct nvmet_ns *ns)
{
struct nvmet_subsys *subsys = ns->subsys;
@@ -764,10 +786,8 @@ 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 (ctrl && ctrl->sqs && ctrl->sqs[0] == sq) {
+ if (ctrl && ctrl->sqs && ctrl->sqs[0] == sq)
nvmet_async_events_process(ctrl, status);
- nvmet_async_events_free(ctrl);
- }
percpu_ref_kill_and_confirm(&sq->ref, nvmet_confirm_sq);
wait_for_completion(&sq->confirm_done);
wait_for_completion(&sq->free_done);
@@ -873,8 +893,11 @@ bool nvmet_req_init(struct nvmet_req *req, struct nvmet_cq *cq,
req->sq = sq;
req->ops = ops;
req->sg = NULL;
+ req->metadata_sg = NULL;
req->sg_cnt = 0;
+ req->metadata_sg_cnt = 0;
req->transfer_len = 0;
+ req->metadata_len = 0;
req->cqe->status = 0;
req->cqe->sq_head = 0;
req->ns = NULL;
@@ -936,9 +959,9 @@ void nvmet_req_uninit(struct nvmet_req *req)
}
EXPORT_SYMBOL_GPL(nvmet_req_uninit);
-bool nvmet_check_data_len(struct nvmet_req *req, size_t data_len)
+bool nvmet_check_transfer_len(struct nvmet_req *req, size_t len)
{
- if (unlikely(data_len != req->transfer_len)) {
+ if (unlikely(len != req->transfer_len)) {
req->error_loc = offsetof(struct nvme_common_command, dptr);
nvmet_req_complete(req, NVME_SC_SGL_INVALID_DATA | NVME_SC_DNR);
return false;
@@ -946,7 +969,7 @@ bool nvmet_check_data_len(struct nvmet_req *req, size_t data_len)
return true;
}
-EXPORT_SYMBOL_GPL(nvmet_check_data_len);
+EXPORT_SYMBOL_GPL(nvmet_check_transfer_len);
bool nvmet_check_data_len_lte(struct nvmet_req *req, size_t data_len)
{
@@ -959,50 +982,90 @@ bool nvmet_check_data_len_lte(struct nvmet_req *req, size_t data_len)
return true;
}
-int nvmet_req_alloc_sgl(struct nvmet_req *req)
+static unsigned int nvmet_data_transfer_len(struct nvmet_req *req)
{
- struct pci_dev *p2p_dev = NULL;
+ return req->transfer_len - req->metadata_len;
+}
- if (IS_ENABLED(CONFIG_PCI_P2PDMA)) {
- if (req->sq->ctrl && req->ns)
- p2p_dev = radix_tree_lookup(&req->sq->ctrl->p2p_ns_map,
- req->ns->nsid);
+static int nvmet_req_alloc_p2pmem_sgls(struct nvmet_req *req)
+{
+ req->sg = pci_p2pmem_alloc_sgl(req->p2p_dev, &req->sg_cnt,
+ nvmet_data_transfer_len(req));
+ if (!req->sg)
+ goto out_err;
- req->p2p_dev = NULL;
- if (req->sq->qid && p2p_dev) {
- req->sg = pci_p2pmem_alloc_sgl(p2p_dev, &req->sg_cnt,
- req->transfer_len);
- if (req->sg) {
- req->p2p_dev = p2p_dev;
- return 0;
- }
- }
+ if (req->metadata_len) {
+ req->metadata_sg = pci_p2pmem_alloc_sgl(req->p2p_dev,
+ &req->metadata_sg_cnt, req->metadata_len);
+ if (!req->metadata_sg)
+ goto out_free_sg;
+ }
+ return 0;
+out_free_sg:
+ pci_p2pmem_free_sgl(req->p2p_dev, req->sg);
+out_err:
+ return -ENOMEM;
+}
- /*
- * If no P2P memory was available we fallback to using
- * regular memory
- */
+static bool nvmet_req_find_p2p_dev(struct nvmet_req *req)
+{
+ if (!IS_ENABLED(CONFIG_PCI_P2PDMA))
+ return false;
+
+ if (req->sq->ctrl && req->sq->qid && req->ns) {
+ req->p2p_dev = radix_tree_lookup(&req->sq->ctrl->p2p_ns_map,
+ req->ns->nsid);
+ if (req->p2p_dev)
+ return true;
}
- req->sg = sgl_alloc(req->transfer_len, GFP_KERNEL, &req->sg_cnt);
+ req->p2p_dev = NULL;
+ return false;
+}
+
+int nvmet_req_alloc_sgls(struct nvmet_req *req)
+{
+ if (nvmet_req_find_p2p_dev(req) && !nvmet_req_alloc_p2pmem_sgls(req))
+ return 0;
+
+ req->sg = sgl_alloc(nvmet_data_transfer_len(req), GFP_KERNEL,
+ &req->sg_cnt);
if (unlikely(!req->sg))
- return -ENOMEM;
+ goto out;
+
+ if (req->metadata_len) {
+ req->metadata_sg = sgl_alloc(req->metadata_len, GFP_KERNEL,
+ &req->metadata_sg_cnt);
+ if (unlikely(!req->metadata_sg))
+ goto out_free;
+ }
return 0;
+out_free:
+ sgl_free(req->sg);
+out:
+ return -ENOMEM;
}
-EXPORT_SYMBOL_GPL(nvmet_req_alloc_sgl);
+EXPORT_SYMBOL_GPL(nvmet_req_alloc_sgls);
-void nvmet_req_free_sgl(struct nvmet_req *req)
+void nvmet_req_free_sgls(struct nvmet_req *req)
{
- if (req->p2p_dev)
+ if (req->p2p_dev) {
pci_p2pmem_free_sgl(req->p2p_dev, req->sg);
- else
+ if (req->metadata_sg)
+ pci_p2pmem_free_sgl(req->p2p_dev, req->metadata_sg);
+ } else {
sgl_free(req->sg);
+ if (req->metadata_sg)
+ sgl_free(req->metadata_sg);
+ }
req->sg = NULL;
+ req->metadata_sg = NULL;
req->sg_cnt = 0;
+ req->metadata_sg_cnt = 0;
}
-EXPORT_SYMBOL_GPL(nvmet_req_free_sgl);
+EXPORT_SYMBOL_GPL(nvmet_req_free_sgls);
static inline bool nvmet_cc_en(u32 cc)
{
@@ -1357,6 +1420,7 @@ static void nvmet_ctrl_free(struct kref *ref)
ida_simple_remove(&cntlid_ida, ctrl->cntlid);
+ nvmet_async_events_free(ctrl);
kfree(ctrl->sqs);
kfree(ctrl->cqs);
kfree(ctrl->changed_ns_list);
diff --git a/drivers/nvme/target/discovery.c b/drivers/nvme/target/discovery.c
index 0c2274b21e15..40cf0b6e6c9d 100644
--- a/drivers/nvme/target/discovery.c
+++ b/drivers/nvme/target/discovery.c
@@ -171,7 +171,7 @@ static void nvmet_execute_disc_get_log_page(struct nvmet_req *req)
u16 status = 0;
void *buffer;
- if (!nvmet_check_data_len(req, data_len))
+ if (!nvmet_check_transfer_len(req, data_len))
return;
if (req->cmd->get_log_page.lid != NVME_LOG_DISC) {
@@ -244,7 +244,7 @@ static void nvmet_execute_disc_identify(struct nvmet_req *req)
const char model[] = "Linux";
u16 status = 0;
- if (!nvmet_check_data_len(req, NVME_IDENTIFY_DATA_SIZE))
+ if (!nvmet_check_transfer_len(req, NVME_IDENTIFY_DATA_SIZE))
return;
if (req->cmd->identify.cns != NVME_ID_CNS_CTRL) {
@@ -298,7 +298,7 @@ static void nvmet_execute_disc_set_features(struct nvmet_req *req)
u32 cdw10 = le32_to_cpu(req->cmd->common.cdw10);
u16 stat;
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
switch (cdw10 & 0xff) {
@@ -324,7 +324,7 @@ static void nvmet_execute_disc_get_features(struct nvmet_req *req)
u32 cdw10 = le32_to_cpu(req->cmd->common.cdw10);
u16 stat = 0;
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
switch (cdw10 & 0xff) {
diff --git a/drivers/nvme/target/fabrics-cmd.c b/drivers/nvme/target/fabrics-cmd.c
index feef15c38ec9..42bd12b8bf00 100644
--- a/drivers/nvme/target/fabrics-cmd.c
+++ b/drivers/nvme/target/fabrics-cmd.c
@@ -12,7 +12,7 @@ static void nvmet_execute_prop_set(struct nvmet_req *req)
u64 val = le64_to_cpu(req->cmd->prop_set.value);
u16 status = 0;
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
if (req->cmd->prop_set.attrib & 1) {
@@ -41,7 +41,7 @@ static void nvmet_execute_prop_get(struct nvmet_req *req)
u16 status = 0;
u64 val = 0;
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
if (req->cmd->prop_get.attrib & 1) {
@@ -156,7 +156,7 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req)
struct nvmet_ctrl *ctrl = NULL;
u16 status = 0;
- if (!nvmet_check_data_len(req, sizeof(struct nvmf_connect_data)))
+ if (!nvmet_check_transfer_len(req, sizeof(struct nvmf_connect_data)))
return;
d = kmalloc(sizeof(*d), GFP_KERNEL);
@@ -197,6 +197,8 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req)
goto out;
}
+ ctrl->pi_support = ctrl->port->pi_enable && ctrl->subsys->pi_support;
+
uuid_copy(&ctrl->hostid, &d->hostid);
status = nvmet_install_queue(ctrl, req);
@@ -205,8 +207,9 @@ static void nvmet_execute_admin_connect(struct nvmet_req *req)
goto out;
}
- pr_info("creating controller %d for subsystem %s for NQN %s.\n",
- ctrl->cntlid, ctrl->subsys->subsysnqn, ctrl->hostnqn);
+ pr_info("creating controller %d for subsystem %s for NQN %s%s.\n",
+ ctrl->cntlid, ctrl->subsys->subsysnqn, ctrl->hostnqn,
+ ctrl->pi_support ? " T10-PI is enabled" : "");
req->cqe->result.u16 = cpu_to_le16(ctrl->cntlid);
out:
@@ -223,7 +226,7 @@ static void nvmet_execute_io_connect(struct nvmet_req *req)
u16 qid = le16_to_cpu(c->qid);
u16 status = 0;
- if (!nvmet_check_data_len(req, sizeof(struct nvmf_connect_data)))
+ if (!nvmet_check_transfer_len(req, sizeof(struct nvmf_connect_data)))
return;
d = kmalloc(sizeof(*d), GFP_KERNEL);
diff --git a/drivers/nvme/target/fc.c b/drivers/nvme/target/fc.c
index a8ceb7721640..27fd3b5aa621 100644
--- a/drivers/nvme/target/fc.c
+++ b/drivers/nvme/target/fc.c
@@ -14,6 +14,7 @@
#include "nvmet.h"
#include <linux/nvme-fc-driver.h>
#include <linux/nvme-fc.h>
+#include "../host/fc.h"
/* *************************** Data Structures/Defines ****************** */
@@ -21,23 +22,21 @@
#define NVMET_LS_CTX_COUNT 256
-/* for this implementation, assume small single frame rqst/rsp */
-#define NVME_FC_MAX_LS_BUFFER_SIZE 2048
-
struct nvmet_fc_tgtport;
struct nvmet_fc_tgt_assoc;
-struct nvmet_fc_ls_iod {
- struct nvmefc_tgt_ls_req *lsreq;
+struct nvmet_fc_ls_iod { /* for an LS RQST RCV */
+ struct nvmefc_ls_rsp *lsrsp;
struct nvmefc_tgt_fcp_req *fcpreq; /* only if RS */
- struct list_head ls_list; /* tgtport->ls_list */
+ struct list_head ls_rcv_list; /* tgtport->ls_rcv_list */
struct nvmet_fc_tgtport *tgtport;
struct nvmet_fc_tgt_assoc *assoc;
+ void *hosthandle;
- u8 *rqstbuf;
- u8 *rspbuf;
+ union nvmefc_ls_requests *rqstbuf;
+ union nvmefc_ls_responses *rspbuf;
u16 rqstdatalen;
dma_addr_t rspdma;
@@ -46,6 +45,18 @@ struct nvmet_fc_ls_iod {
struct work_struct work;
} __aligned(sizeof(unsigned long long));
+struct nvmet_fc_ls_req_op { /* for an LS RQST XMT */
+ struct nvmefc_ls_req ls_req;
+
+ struct nvmet_fc_tgtport *tgtport;
+ void *hosthandle;
+
+ int ls_error;
+ struct list_head lsreq_list; /* tgtport->ls_req_list */
+ bool req_queued;
+};
+
+
/* desired maximum for a single sequence - if sg list allows it */
#define NVMET_FC_MAX_SEQ_LENGTH (256 * 1024)
@@ -83,7 +94,6 @@ struct nvmet_fc_fcp_iod {
};
struct nvmet_fc_tgtport {
-
struct nvmet_fc_target_port fc_target_port;
struct list_head tgt_list; /* nvmet_fc_target_list */
@@ -92,9 +102,11 @@ struct nvmet_fc_tgtport {
struct nvmet_fc_ls_iod *iod;
spinlock_t lock;
- struct list_head ls_list;
+ struct list_head ls_rcv_list;
+ struct list_head ls_req_list;
struct list_head ls_busylist;
struct list_head assoc_list;
+ struct list_head host_list;
struct ida assoc_cnt;
struct nvmet_fc_port_entry *pe;
struct kref ref;
@@ -136,14 +148,26 @@ struct nvmet_fc_tgt_queue {
struct nvmet_fc_fcp_iod fod[]; /* array of fcp_iods */
} __aligned(sizeof(unsigned long long));
+struct nvmet_fc_hostport {
+ struct nvmet_fc_tgtport *tgtport;
+ void *hosthandle;
+ struct list_head host_list;
+ struct kref ref;
+ u8 invalid;
+};
+
struct nvmet_fc_tgt_assoc {
u64 association_id;
u32 a_id;
+ atomic_t terminating;
struct nvmet_fc_tgtport *tgtport;
+ struct nvmet_fc_hostport *hostport;
+ struct nvmet_fc_ls_iod *rcv_disconn;
struct list_head a_list;
struct nvmet_fc_tgt_queue *queues[NVMET_NR_QUEUES + 1];
struct kref ref;
struct work_struct del_work;
+ atomic_t del_work_active;
};
@@ -227,6 +251,8 @@ static int nvmet_fc_tgtport_get(struct nvmet_fc_tgtport *tgtport);
static void nvmet_fc_handle_fcp_rqst(struct nvmet_fc_tgtport *tgtport,
struct nvmet_fc_fcp_iod *fod);
static void nvmet_fc_delete_target_assoc(struct nvmet_fc_tgt_assoc *assoc);
+static void nvmet_fc_xmt_ls_rsp(struct nvmet_fc_tgtport *tgtport,
+ struct nvmet_fc_ls_iod *iod);
/* *********************** FC-NVME DMA Handling **************************** */
@@ -318,6 +344,188 @@ fc_dma_unmap_sg(struct device *dev, struct scatterlist *sg, int nents,
}
+/* ********************** FC-NVME LS XMT Handling ************************* */
+
+
+static void
+__nvmet_fc_finish_ls_req(struct nvmet_fc_ls_req_op *lsop)
+{
+ struct nvmet_fc_tgtport *tgtport = lsop->tgtport;
+ struct nvmefc_ls_req *lsreq = &lsop->ls_req;
+ unsigned long flags;
+
+ spin_lock_irqsave(&tgtport->lock, flags);
+
+ if (!lsop->req_queued) {
+ spin_unlock_irqrestore(&tgtport->lock, flags);
+ return;
+ }
+
+ list_del(&lsop->lsreq_list);
+
+ lsop->req_queued = false;
+
+ spin_unlock_irqrestore(&tgtport->lock, flags);
+
+ fc_dma_unmap_single(tgtport->dev, lsreq->rqstdma,
+ (lsreq->rqstlen + lsreq->rsplen),
+ DMA_BIDIRECTIONAL);
+
+ nvmet_fc_tgtport_put(tgtport);
+}
+
+static int
+__nvmet_fc_send_ls_req(struct nvmet_fc_tgtport *tgtport,
+ struct nvmet_fc_ls_req_op *lsop,
+ void (*done)(struct nvmefc_ls_req *req, int status))
+{
+ struct nvmefc_ls_req *lsreq = &lsop->ls_req;
+ unsigned long flags;
+ int ret = 0;
+
+ if (!tgtport->ops->ls_req)
+ return -EOPNOTSUPP;
+
+ if (!nvmet_fc_tgtport_get(tgtport))
+ return -ESHUTDOWN;
+
+ lsreq->done = done;
+ lsop->req_queued = false;
+ INIT_LIST_HEAD(&lsop->lsreq_list);
+
+ lsreq->rqstdma = fc_dma_map_single(tgtport->dev, lsreq->rqstaddr,
+ lsreq->rqstlen + lsreq->rsplen,
+ DMA_BIDIRECTIONAL);
+ if (fc_dma_mapping_error(tgtport->dev, lsreq->rqstdma)) {
+ ret = -EFAULT;
+ goto out_puttgtport;
+ }
+ lsreq->rspdma = lsreq->rqstdma + lsreq->rqstlen;
+
+ spin_lock_irqsave(&tgtport->lock, flags);
+
+ list_add_tail(&lsop->lsreq_list, &tgtport->ls_req_list);
+
+ lsop->req_queued = true;
+
+ spin_unlock_irqrestore(&tgtport->lock, flags);
+
+ ret = tgtport->ops->ls_req(&tgtport->fc_target_port, lsop->hosthandle,
+ lsreq);
+ if (ret)
+ goto out_unlink;
+
+ return 0;
+
+out_unlink:
+ lsop->ls_error = ret;
+ spin_lock_irqsave(&tgtport->lock, flags);
+ lsop->req_queued = false;
+ list_del(&lsop->lsreq_list);
+ spin_unlock_irqrestore(&tgtport->lock, flags);
+ fc_dma_unmap_single(tgtport->dev, lsreq->rqstdma,
+ (lsreq->rqstlen + lsreq->rsplen),
+ DMA_BIDIRECTIONAL);
+out_puttgtport:
+ nvmet_fc_tgtport_put(tgtport);
+
+ return ret;
+}
+
+static int
+nvmet_fc_send_ls_req_async(struct nvmet_fc_tgtport *tgtport,
+ struct nvmet_fc_ls_req_op *lsop,
+ void (*done)(struct nvmefc_ls_req *req, int status))
+{
+ /* don't wait for completion */
+
+ return __nvmet_fc_send_ls_req(tgtport, lsop, done);
+}
+
+static void
+nvmet_fc_disconnect_assoc_done(struct nvmefc_ls_req *lsreq, int status)
+{
+ struct nvmet_fc_ls_req_op *lsop =
+ container_of(lsreq, struct nvmet_fc_ls_req_op, ls_req);
+
+ __nvmet_fc_finish_ls_req(lsop);
+
+ /* fc-nvme target doesn't care about success or failure of cmd */
+
+ kfree(lsop);
+}
+
+/*
+ * This routine sends a FC-NVME LS to disconnect (aka terminate)
+ * the FC-NVME Association. Terminating the association also
+ * terminates the FC-NVME connections (per queue, both admin and io
+ * queues) that are part of the association. E.g. things are torn
+ * down, and the related FC-NVME Association ID and Connection IDs
+ * become invalid.
+ *
+ * The behavior of the fc-nvme target is such that it's
+ * understanding of the association and connections will implicitly
+ * be torn down. The action is implicit as it may be due to a loss of
+ * connectivity with the fc-nvme host, so the target may never get a
+ * response even if it tried. As such, the action of this routine
+ * is to asynchronously send the LS, ignore any results of the LS, and
+ * continue on with terminating the association. If the fc-nvme host
+ * is present and receives the LS, it too can tear down.
+ */
+static void
+nvmet_fc_xmt_disconnect_assoc(struct nvmet_fc_tgt_assoc *assoc)
+{
+ struct nvmet_fc_tgtport *tgtport = assoc->tgtport;
+ struct fcnvme_ls_disconnect_assoc_rqst *discon_rqst;
+ struct fcnvme_ls_disconnect_assoc_acc *discon_acc;
+ struct nvmet_fc_ls_req_op *lsop;
+ struct nvmefc_ls_req *lsreq;
+ int ret;
+
+ /*
+ * If ls_req is NULL or no hosthandle, it's an older lldd and no
+ * message is normal. Otherwise, send unless the hostport has
+ * already been invalidated by the lldd.
+ */
+ if (!tgtport->ops->ls_req || !assoc->hostport ||
+ assoc->hostport->invalid)
+ return;
+
+ lsop = kzalloc((sizeof(*lsop) +
+ sizeof(*discon_rqst) + sizeof(*discon_acc) +
+ tgtport->ops->lsrqst_priv_sz), GFP_KERNEL);
+ if (!lsop) {
+ dev_info(tgtport->dev,
+ "{%d:%d} send Disconnect Association failed: ENOMEM\n",
+ tgtport->fc_target_port.port_num, assoc->a_id);
+ return;
+ }
+
+ discon_rqst = (struct fcnvme_ls_disconnect_assoc_rqst *)&lsop[1];
+ discon_acc = (struct fcnvme_ls_disconnect_assoc_acc *)&discon_rqst[1];
+ lsreq = &lsop->ls_req;
+ if (tgtport->ops->lsrqst_priv_sz)
+ lsreq->private = (void *)&discon_acc[1];
+ else
+ lsreq->private = NULL;
+
+ lsop->tgtport = tgtport;
+ lsop->hosthandle = assoc->hostport->hosthandle;
+
+ nvmefc_fmt_lsreq_discon_assoc(lsreq, discon_rqst, discon_acc,
+ assoc->association_id);
+
+ ret = nvmet_fc_send_ls_req_async(tgtport, lsop,
+ nvmet_fc_disconnect_assoc_done);
+ if (ret) {
+ dev_info(tgtport->dev,
+ "{%d:%d} XMT Disconnect Association failed: %d\n",
+ tgtport->fc_target_port.port_num, assoc->a_id, ret);
+ kfree(lsop);
+ }
+}
+
+
/* *********************** FC-NVME Port Management ************************ */
@@ -337,17 +545,18 @@ nvmet_fc_alloc_ls_iodlist(struct nvmet_fc_tgtport *tgtport)
for (i = 0; i < NVMET_LS_CTX_COUNT; iod++, i++) {
INIT_WORK(&iod->work, nvmet_fc_handle_ls_rqst_work);
iod->tgtport = tgtport;
- list_add_tail(&iod->ls_list, &tgtport->ls_list);
+ list_add_tail(&iod->ls_rcv_list, &tgtport->ls_rcv_list);
- iod->rqstbuf = kcalloc(2, NVME_FC_MAX_LS_BUFFER_SIZE,
- GFP_KERNEL);
+ iod->rqstbuf = kzalloc(sizeof(union nvmefc_ls_requests) +
+ sizeof(union nvmefc_ls_responses),
+ GFP_KERNEL);
if (!iod->rqstbuf)
goto out_fail;
- iod->rspbuf = iod->rqstbuf + NVME_FC_MAX_LS_BUFFER_SIZE;
+ iod->rspbuf = (union nvmefc_ls_responses *)&iod->rqstbuf[1];
iod->rspdma = fc_dma_map_single(tgtport->dev, iod->rspbuf,
- NVME_FC_MAX_LS_BUFFER_SIZE,
+ sizeof(*iod->rspbuf),
DMA_TO_DEVICE);
if (fc_dma_mapping_error(tgtport->dev, iod->rspdma))
goto out_fail;
@@ -357,12 +566,12 @@ nvmet_fc_alloc_ls_iodlist(struct nvmet_fc_tgtport *tgtport)
out_fail:
kfree(iod->rqstbuf);
- list_del(&iod->ls_list);
+ list_del(&iod->ls_rcv_list);
for (iod--, i--; i >= 0; iod--, i--) {
fc_dma_unmap_single(tgtport->dev, iod->rspdma,
- NVME_FC_MAX_LS_BUFFER_SIZE, DMA_TO_DEVICE);
+ sizeof(*iod->rspbuf), DMA_TO_DEVICE);
kfree(iod->rqstbuf);
- list_del(&iod->ls_list);
+ list_del(&iod->ls_rcv_list);
}
kfree(iod);
@@ -378,10 +587,10 @@ nvmet_fc_free_ls_iodlist(struct nvmet_fc_tgtport *tgtport)
for (i = 0; i < NVMET_LS_CTX_COUNT; iod++, i++) {
fc_dma_unmap_single(tgtport->dev,
- iod->rspdma, NVME_FC_MAX_LS_BUFFER_SIZE,
+ iod->rspdma, sizeof(*iod->rspbuf),
DMA_TO_DEVICE);
kfree(iod->rqstbuf);
- list_del(&iod->ls_list);
+ list_del(&iod->ls_rcv_list);
}
kfree(tgtport->iod);
}
@@ -393,10 +602,10 @@ nvmet_fc_alloc_ls_iod(struct nvmet_fc_tgtport *tgtport)
unsigned long flags;
spin_lock_irqsave(&tgtport->lock, flags);
- iod = list_first_entry_or_null(&tgtport->ls_list,
- struct nvmet_fc_ls_iod, ls_list);
+ iod = list_first_entry_or_null(&tgtport->ls_rcv_list,
+ struct nvmet_fc_ls_iod, ls_rcv_list);
if (iod)
- list_move_tail(&iod->ls_list, &tgtport->ls_busylist);
+ list_move_tail(&iod->ls_rcv_list, &tgtport->ls_busylist);
spin_unlock_irqrestore(&tgtport->lock, flags);
return iod;
}
@@ -409,7 +618,7 @@ nvmet_fc_free_ls_iod(struct nvmet_fc_tgtport *tgtport,
unsigned long flags;
spin_lock_irqsave(&tgtport->lock, flags);
- list_move(&iod->ls_list, &tgtport->ls_list);
+ list_move(&iod->ls_rcv_list, &tgtport->ls_rcv_list);
spin_unlock_irqrestore(&tgtport->lock, flags);
}
@@ -678,31 +887,33 @@ nvmet_fc_delete_target_queue(struct nvmet_fc_tgt_queue *queue)
struct nvmet_fc_fcp_iod *fod = queue->fod;
struct nvmet_fc_defer_fcp_req *deferfcp, *tempptr;
unsigned long flags;
- int i, writedataactive;
+ int i;
bool disconnect;
disconnect = atomic_xchg(&queue->connected, 0);
+ /* if not connected, nothing to do */
+ if (!disconnect)
+ return;
+
spin_lock_irqsave(&queue->qlock, flags);
/* abort outstanding io's */
for (i = 0; i < queue->sqsize; fod++, i++) {
if (fod->active) {
spin_lock(&fod->flock);
fod->abort = true;
- writedataactive = fod->writedataactive;
- spin_unlock(&fod->flock);
/*
* only call lldd abort routine if waiting for
* writedata. other outstanding ops should finish
* on their own.
*/
- if (writedataactive) {
- spin_lock(&fod->flock);
+ if (fod->writedataactive) {
fod->aborted = true;
spin_unlock(&fod->flock);
tgtport->ops->fcp_abort(
&tgtport->fc_target_port, fod->fcpreq);
- }
+ } else
+ spin_unlock(&fod->flock);
}
}
@@ -742,8 +953,7 @@ nvmet_fc_delete_target_queue(struct nvmet_fc_tgt_queue *queue)
flush_workqueue(queue->work_q);
- if (disconnect)
- nvmet_sq_destroy(&queue->nvme_sq);
+ nvmet_sq_destroy(&queue->nvme_sq);
nvmet_fc_tgt_q_put(queue);
}
@@ -778,17 +988,114 @@ nvmet_fc_find_target_queue(struct nvmet_fc_tgtport *tgtport,
}
static void
+nvmet_fc_hostport_free(struct kref *ref)
+{
+ struct nvmet_fc_hostport *hostport =
+ container_of(ref, struct nvmet_fc_hostport, ref);
+ struct nvmet_fc_tgtport *tgtport = hostport->tgtport;
+ unsigned long flags;
+
+ spin_lock_irqsave(&tgtport->lock, flags);
+ list_del(&hostport->host_list);
+ spin_unlock_irqrestore(&tgtport->lock, flags);
+ if (tgtport->ops->host_release && hostport->invalid)
+ tgtport->ops->host_release(hostport->hosthandle);
+ kfree(hostport);
+ nvmet_fc_tgtport_put(tgtport);
+}
+
+static void
+nvmet_fc_hostport_put(struct nvmet_fc_hostport *hostport)
+{
+ kref_put(&hostport->ref, nvmet_fc_hostport_free);
+}
+
+static int
+nvmet_fc_hostport_get(struct nvmet_fc_hostport *hostport)
+{
+ return kref_get_unless_zero(&hostport->ref);
+}
+
+static void
+nvmet_fc_free_hostport(struct nvmet_fc_hostport *hostport)
+{
+ /* if LLDD not implemented, leave as NULL */
+ if (!hostport->hosthandle)
+ return;
+
+ nvmet_fc_hostport_put(hostport);
+}
+
+static struct nvmet_fc_hostport *
+nvmet_fc_alloc_hostport(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
+{
+ struct nvmet_fc_hostport *newhost, *host, *match = NULL;
+ unsigned long flags;
+
+ /* if LLDD not implemented, leave as NULL */
+ if (!hosthandle)
+ return NULL;
+
+ /* take reference for what will be the newly allocated hostport */
+ if (!nvmet_fc_tgtport_get(tgtport))
+ return ERR_PTR(-EINVAL);
+
+ newhost = kzalloc(sizeof(*newhost), GFP_KERNEL);
+ if (!newhost) {
+ spin_lock_irqsave(&tgtport->lock, flags);
+ list_for_each_entry(host, &tgtport->host_list, host_list) {
+ if (host->hosthandle == hosthandle && !host->invalid) {
+ if (nvmet_fc_hostport_get(host)) {
+ match = host;
+ break;
+ }
+ }
+ }
+ spin_unlock_irqrestore(&tgtport->lock, flags);
+ /* no allocation - release reference */
+ nvmet_fc_tgtport_put(tgtport);
+ return (match) ? match : ERR_PTR(-ENOMEM);
+ }
+
+ newhost->tgtport = tgtport;
+ newhost->hosthandle = hosthandle;
+ INIT_LIST_HEAD(&newhost->host_list);
+ kref_init(&newhost->ref);
+
+ spin_lock_irqsave(&tgtport->lock, flags);
+ list_for_each_entry(host, &tgtport->host_list, host_list) {
+ if (host->hosthandle == hosthandle && !host->invalid) {
+ if (nvmet_fc_hostport_get(host)) {
+ match = host;
+ break;
+ }
+ }
+ }
+ if (match) {
+ kfree(newhost);
+ newhost = NULL;
+ /* releasing allocation - release reference */
+ nvmet_fc_tgtport_put(tgtport);
+ } else
+ list_add_tail(&newhost->host_list, &tgtport->host_list);
+ spin_unlock_irqrestore(&tgtport->lock, flags);
+
+ return (match) ? match : newhost;
+}
+
+static void
nvmet_fc_delete_assoc(struct work_struct *work)
{
struct nvmet_fc_tgt_assoc *assoc =
container_of(work, struct nvmet_fc_tgt_assoc, del_work);
nvmet_fc_delete_target_assoc(assoc);
+ atomic_set(&assoc->del_work_active, 0);
nvmet_fc_tgt_a_put(assoc);
}
static struct nvmet_fc_tgt_assoc *
-nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport)
+nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport, void *hosthandle)
{
struct nvmet_fc_tgt_assoc *assoc, *tmpassoc;
unsigned long flags;
@@ -805,13 +1112,19 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport)
goto out_free_assoc;
if (!nvmet_fc_tgtport_get(tgtport))
- goto out_ida_put;
+ goto out_ida;
+
+ assoc->hostport = nvmet_fc_alloc_hostport(tgtport, hosthandle);
+ if (IS_ERR(assoc->hostport))
+ goto out_put;
assoc->tgtport = tgtport;
assoc->a_id = idx;
INIT_LIST_HEAD(&assoc->a_list);
kref_init(&assoc->ref);
INIT_WORK(&assoc->del_work, nvmet_fc_delete_assoc);
+ atomic_set(&assoc->del_work_active, 0);
+ atomic_set(&assoc->terminating, 0);
while (needrandom) {
get_random_bytes(&ran, sizeof(ran) - BYTES_FOR_QID);
@@ -819,11 +1132,12 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport)
spin_lock_irqsave(&tgtport->lock, flags);
needrandom = false;
- list_for_each_entry(tmpassoc, &tgtport->assoc_list, a_list)
+ list_for_each_entry(tmpassoc, &tgtport->assoc_list, a_list) {
if (ran == tmpassoc->association_id) {
needrandom = true;
break;
}
+ }
if (!needrandom) {
assoc->association_id = ran;
list_add_tail(&assoc->a_list, &tgtport->assoc_list);
@@ -833,7 +1147,9 @@ nvmet_fc_alloc_target_assoc(struct nvmet_fc_tgtport *tgtport)
return assoc;
-out_ida_put:
+out_put:
+ nvmet_fc_tgtport_put(tgtport);
+out_ida:
ida_simple_remove(&tgtport->assoc_cnt, idx);
out_free_assoc:
kfree(assoc);
@@ -846,12 +1162,24 @@ nvmet_fc_target_assoc_free(struct kref *ref)
struct nvmet_fc_tgt_assoc *assoc =
container_of(ref, struct nvmet_fc_tgt_assoc, ref);
struct nvmet_fc_tgtport *tgtport = assoc->tgtport;
+ struct nvmet_fc_ls_iod *oldls;
unsigned long flags;
+ /* Send Disconnect now that all i/o has completed */
+ nvmet_fc_xmt_disconnect_assoc(assoc);
+
+ nvmet_fc_free_hostport(assoc->hostport);
spin_lock_irqsave(&tgtport->lock, flags);
list_del(&assoc->a_list);
+ oldls = assoc->rcv_disconn;
spin_unlock_irqrestore(&tgtport->lock, flags);
+ /* if pending Rcv Disconnect Association LS, send rsp now */
+ if (oldls)
+ nvmet_fc_xmt_ls_rsp(tgtport, oldls);
ida_simple_remove(&tgtport->assoc_cnt, assoc->a_id);
+ dev_info(tgtport->dev,
+ "{%d:%d} Association freed\n",
+ tgtport->fc_target_port.port_num, assoc->a_id);
kfree(assoc);
nvmet_fc_tgtport_put(tgtport);
}
@@ -874,7 +1202,13 @@ nvmet_fc_delete_target_assoc(struct nvmet_fc_tgt_assoc *assoc)
struct nvmet_fc_tgtport *tgtport = assoc->tgtport;
struct nvmet_fc_tgt_queue *queue;
unsigned long flags;
- int i;
+ int i, terminating;
+
+ terminating = atomic_xchg(&assoc->terminating, 1);
+
+ /* if already terminating, do nothing */
+ if (terminating)
+ return;
spin_lock_irqsave(&tgtport->lock, flags);
for (i = NVMET_NR_QUEUES; i >= 0; i--) {
@@ -890,6 +1224,10 @@ nvmet_fc_delete_target_assoc(struct nvmet_fc_tgt_assoc *assoc)
}
spin_unlock_irqrestore(&tgtport->lock, flags);
+ dev_info(tgtport->dev,
+ "{%d:%d} Association deleted\n",
+ tgtport->fc_target_port.port_num, assoc->a_id);
+
nvmet_fc_tgt_a_put(assoc);
}
@@ -1048,16 +1386,21 @@ nvmet_fc_register_targetport(struct nvmet_fc_port_info *pinfo,
newrec->fc_target_port.node_name = pinfo->node_name;
newrec->fc_target_port.port_name = pinfo->port_name;
- newrec->fc_target_port.private = &newrec[1];
+ if (template->target_priv_sz)
+ newrec->fc_target_port.private = &newrec[1];
+ else
+ newrec->fc_target_port.private = NULL;
newrec->fc_target_port.port_id = pinfo->port_id;
newrec->fc_target_port.port_num = idx;
INIT_LIST_HEAD(&newrec->tgt_list);
newrec->dev = dev;
newrec->ops = template;
spin_lock_init(&newrec->lock);
- INIT_LIST_HEAD(&newrec->ls_list);
+ INIT_LIST_HEAD(&newrec->ls_rcv_list);
+ INIT_LIST_HEAD(&newrec->ls_req_list);
INIT_LIST_HEAD(&newrec->ls_busylist);
INIT_LIST_HEAD(&newrec->assoc_list);
+ INIT_LIST_HEAD(&newrec->host_list);
kref_init(&newrec->ref);
ida_init(&newrec->assoc_cnt);
newrec->max_sg_cnt = template->max_sgl_segments;
@@ -1134,17 +1477,90 @@ __nvmet_fc_free_assocs(struct nvmet_fc_tgtport *tgtport)
{
struct nvmet_fc_tgt_assoc *assoc, *next;
unsigned long flags;
+ int ret;
+
+ spin_lock_irqsave(&tgtport->lock, flags);
+ list_for_each_entry_safe(assoc, next,
+ &tgtport->assoc_list, a_list) {
+ if (!nvmet_fc_tgt_a_get(assoc))
+ continue;
+ ret = atomic_cmpxchg(&assoc->del_work_active, 0, 1);
+ if (ret == 0) {
+ if (!schedule_work(&assoc->del_work))
+ nvmet_fc_tgt_a_put(assoc);
+ } else {
+ /* already deleting - release local reference */
+ nvmet_fc_tgt_a_put(assoc);
+ }
+ }
+ spin_unlock_irqrestore(&tgtport->lock, flags);
+}
+
+/**
+ * nvmet_fc_invalidate_host - transport entry point called by an LLDD
+ * to remove references to a hosthandle for LS's.
+ *
+ * The nvmet-fc layer ensures that any references to the hosthandle
+ * on the targetport are forgotten (set to NULL). The LLDD will
+ * typically call this when a login with a remote host port has been
+ * lost, thus LS's for the remote host port are no longer possible.
+ *
+ * If an LS request is outstanding to the targetport/hosthandle (or
+ * issued concurrently with the call to invalidate the host), the
+ * LLDD is responsible for terminating/aborting the LS and completing
+ * the LS request. It is recommended that these terminations/aborts
+ * occur after calling to invalidate the host handle to avoid additional
+ * retries by the nvmet-fc transport. The nvmet-fc transport may
+ * continue to reference host handle while it cleans up outstanding
+ * NVME associations. The nvmet-fc transport will call the
+ * ops->host_release() callback to notify the LLDD that all references
+ * are complete and the related host handle can be recovered.
+ * Note: if there are no references, the callback may be called before
+ * the invalidate host call returns.
+ *
+ * @target_port: pointer to the (registered) target port that a prior
+ * LS was received on and which supplied the transport the
+ * hosthandle.
+ * @hosthandle: the handle (pointer) that represents the host port
+ * that no longer has connectivity and that LS's should
+ * no longer be directed to.
+ */
+void
+nvmet_fc_invalidate_host(struct nvmet_fc_target_port *target_port,
+ void *hosthandle)
+{
+ struct nvmet_fc_tgtport *tgtport = targetport_to_tgtport(target_port);
+ struct nvmet_fc_tgt_assoc *assoc, *next;
+ unsigned long flags;
+ bool noassoc = true;
+ int ret;
spin_lock_irqsave(&tgtport->lock, flags);
list_for_each_entry_safe(assoc, next,
&tgtport->assoc_list, a_list) {
+ if (!assoc->hostport ||
+ assoc->hostport->hosthandle != hosthandle)
+ continue;
if (!nvmet_fc_tgt_a_get(assoc))
continue;
- if (!schedule_work(&assoc->del_work))
+ assoc->hostport->invalid = 1;
+ noassoc = false;
+ ret = atomic_cmpxchg(&assoc->del_work_active, 0, 1);
+ if (ret == 0) {
+ if (!schedule_work(&assoc->del_work))
+ nvmet_fc_tgt_a_put(assoc);
+ } else {
+ /* already deleting - release local reference */
nvmet_fc_tgt_a_put(assoc);
+ }
}
spin_unlock_irqrestore(&tgtport->lock, flags);
+
+ /* if there's nothing to wait for - call the callback */
+ if (noassoc && tgtport->ops->host_release)
+ tgtport->ops->host_release(hosthandle);
}
+EXPORT_SYMBOL_GPL(nvmet_fc_invalidate_host);
/*
* nvmet layer has called to terminate an association
@@ -1157,6 +1573,7 @@ nvmet_fc_delete_ctrl(struct nvmet_ctrl *ctrl)
struct nvmet_fc_tgt_queue *queue;
unsigned long flags;
bool found_ctrl = false;
+ int ret;
/* this is a bit ugly, but don't want to make locks layered */
spin_lock_irqsave(&nvmet_fc_tgtlock, flags);
@@ -1180,8 +1597,14 @@ nvmet_fc_delete_ctrl(struct nvmet_ctrl *ctrl)
nvmet_fc_tgtport_put(tgtport);
if (found_ctrl) {
- if (!schedule_work(&assoc->del_work))
+ ret = atomic_cmpxchg(&assoc->del_work_active, 0, 1);
+ if (ret == 0) {
+ if (!schedule_work(&assoc->del_work))
+ nvmet_fc_tgt_a_put(assoc);
+ } else {
+ /* already deleting - release local reference */
nvmet_fc_tgt_a_put(assoc);
+ }
return;
}
@@ -1211,6 +1634,13 @@ nvmet_fc_unregister_targetport(struct nvmet_fc_target_port *target_port)
/* terminate any outstanding associations */
__nvmet_fc_free_assocs(tgtport);
+ /*
+ * should terminate LS's as well. However, LS's will be generated
+ * at the tail end of association termination, so they likely don't
+ * exist yet. And even if they did, it's worthwhile to just let
+ * them finish and targetport ref counting will clean things up.
+ */
+
nvmet_fc_tgtport_put(tgtport);
return 0;
@@ -1218,113 +1648,15 @@ nvmet_fc_unregister_targetport(struct nvmet_fc_target_port *target_port)
EXPORT_SYMBOL_GPL(nvmet_fc_unregister_targetport);
-/* *********************** FC-NVME LS Handling **************************** */
-
-
-static void
-nvmet_fc_format_rsp_hdr(void *buf, u8 ls_cmd, __be32 desc_len, u8 rqst_ls_cmd)
-{
- struct fcnvme_ls_acc_hdr *acc = buf;
-
- acc->w0.ls_cmd = ls_cmd;
- acc->desc_list_len = desc_len;
- acc->rqst.desc_tag = cpu_to_be32(FCNVME_LSDESC_RQST);
- acc->rqst.desc_len =
- fcnvme_lsdesc_len(sizeof(struct fcnvme_lsdesc_rqst));
- acc->rqst.w0.ls_cmd = rqst_ls_cmd;
-}
+/* ********************** FC-NVME LS RCV Handling ************************* */
-static int
-nvmet_fc_format_rjt(void *buf, u16 buflen, u8 ls_cmd,
- u8 reason, u8 explanation, u8 vendor)
-{
- struct fcnvme_ls_rjt *rjt = buf;
-
- nvmet_fc_format_rsp_hdr(buf, FCNVME_LSDESC_RQST,
- fcnvme_lsdesc_len(sizeof(struct fcnvme_ls_rjt)),
- ls_cmd);
- rjt->rjt.desc_tag = cpu_to_be32(FCNVME_LSDESC_RJT);
- rjt->rjt.desc_len = fcnvme_lsdesc_len(sizeof(struct fcnvme_lsdesc_rjt));
- rjt->rjt.reason_code = reason;
- rjt->rjt.reason_explanation = explanation;
- rjt->rjt.vendor = vendor;
-
- return sizeof(struct fcnvme_ls_rjt);
-}
-
-/* Validation Error indexes into the string table below */
-enum {
- VERR_NO_ERROR = 0,
- VERR_CR_ASSOC_LEN = 1,
- VERR_CR_ASSOC_RQST_LEN = 2,
- VERR_CR_ASSOC_CMD = 3,
- VERR_CR_ASSOC_CMD_LEN = 4,
- VERR_ERSP_RATIO = 5,
- VERR_ASSOC_ALLOC_FAIL = 6,
- VERR_QUEUE_ALLOC_FAIL = 7,
- VERR_CR_CONN_LEN = 8,
- VERR_CR_CONN_RQST_LEN = 9,
- VERR_ASSOC_ID = 10,
- VERR_ASSOC_ID_LEN = 11,
- VERR_NO_ASSOC = 12,
- VERR_CONN_ID = 13,
- VERR_CONN_ID_LEN = 14,
- VERR_NO_CONN = 15,
- VERR_CR_CONN_CMD = 16,
- VERR_CR_CONN_CMD_LEN = 17,
- VERR_DISCONN_LEN = 18,
- VERR_DISCONN_RQST_LEN = 19,
- VERR_DISCONN_CMD = 20,
- VERR_DISCONN_CMD_LEN = 21,
- VERR_DISCONN_SCOPE = 22,
- VERR_RS_LEN = 23,
- VERR_RS_RQST_LEN = 24,
- VERR_RS_CMD = 25,
- VERR_RS_CMD_LEN = 26,
- VERR_RS_RCTL = 27,
- VERR_RS_RO = 28,
-};
-
-static char *validation_errors[] = {
- "OK",
- "Bad CR_ASSOC Length",
- "Bad CR_ASSOC Rqst Length",
- "Not CR_ASSOC Cmd",
- "Bad CR_ASSOC Cmd Length",
- "Bad Ersp Ratio",
- "Association Allocation Failed",
- "Queue Allocation Failed",
- "Bad CR_CONN Length",
- "Bad CR_CONN Rqst Length",
- "Not Association ID",
- "Bad Association ID Length",
- "No Association",
- "Not Connection ID",
- "Bad Connection ID Length",
- "No Connection",
- "Not CR_CONN Cmd",
- "Bad CR_CONN Cmd Length",
- "Bad DISCONN Length",
- "Bad DISCONN Rqst Length",
- "Not DISCONN Cmd",
- "Bad DISCONN Cmd Length",
- "Bad Disconnect Scope",
- "Bad RS Length",
- "Bad RS Rqst Length",
- "Not RS Cmd",
- "Bad RS Cmd Length",
- "Bad RS R_CTL",
- "Bad RS Relative Offset",
-};
static void
nvmet_fc_ls_create_association(struct nvmet_fc_tgtport *tgtport,
struct nvmet_fc_ls_iod *iod)
{
- struct fcnvme_ls_cr_assoc_rqst *rqst =
- (struct fcnvme_ls_cr_assoc_rqst *)iod->rqstbuf;
- struct fcnvme_ls_cr_assoc_acc *acc =
- (struct fcnvme_ls_cr_assoc_acc *)iod->rspbuf;
+ struct fcnvme_ls_cr_assoc_rqst *rqst = &iod->rqstbuf->rq_cr_assoc;
+ struct fcnvme_ls_cr_assoc_acc *acc = &iod->rspbuf->rsp_cr_assoc;
struct nvmet_fc_tgt_queue *queue;
int ret = 0;
@@ -1356,7 +1688,8 @@ nvmet_fc_ls_create_association(struct nvmet_fc_tgtport *tgtport,
else {
/* new association w/ admin queue */
- iod->assoc = nvmet_fc_alloc_target_assoc(tgtport);
+ iod->assoc = nvmet_fc_alloc_target_assoc(
+ tgtport, iod->hosthandle);
if (!iod->assoc)
ret = VERR_ASSOC_ALLOC_FAIL;
else {
@@ -1371,8 +1704,8 @@ nvmet_fc_ls_create_association(struct nvmet_fc_tgtport *tgtport,
dev_err(tgtport->dev,
"Create Association LS failed: %s\n",
validation_errors[ret]);
- iod->lsreq->rsplen = nvmet_fc_format_rjt(acc,
- NVME_FC_MAX_LS_BUFFER_SIZE, rqst->w0.ls_cmd,
+ iod->lsrsp->rsplen = nvme_fc_format_rjt(acc,
+ sizeof(*acc), rqst->w0.ls_cmd,
FCNVME_RJT_RC_LOGIC,
FCNVME_RJT_EXP_NONE, 0);
return;
@@ -1382,11 +1715,15 @@ nvmet_fc_ls_create_association(struct nvmet_fc_tgtport *tgtport,
atomic_set(&queue->connected, 1);
queue->sqhd = 0; /* best place to init value */
+ dev_info(tgtport->dev,
+ "{%d:%d} Association created\n",
+ tgtport->fc_target_port.port_num, iod->assoc->a_id);
+
/* format a response */
- iod->lsreq->rsplen = sizeof(*acc);
+ iod->lsrsp->rsplen = sizeof(*acc);
- nvmet_fc_format_rsp_hdr(acc, FCNVME_LS_ACC,
+ nvme_fc_format_rsp_hdr(acc, FCNVME_LS_ACC,
fcnvme_lsdesc_len(
sizeof(struct fcnvme_ls_cr_assoc_acc)),
FCNVME_LS_CREATE_ASSOCIATION);
@@ -1407,10 +1744,8 @@ static void
nvmet_fc_ls_create_connection(struct nvmet_fc_tgtport *tgtport,
struct nvmet_fc_ls_iod *iod)
{
- struct fcnvme_ls_cr_conn_rqst *rqst =
- (struct fcnvme_ls_cr_conn_rqst *)iod->rqstbuf;
- struct fcnvme_ls_cr_conn_acc *acc =
- (struct fcnvme_ls_cr_conn_acc *)iod->rspbuf;
+ struct fcnvme_ls_cr_conn_rqst *rqst = &iod->rqstbuf->rq_cr_conn;
+ struct fcnvme_ls_cr_conn_acc *acc = &iod->rspbuf->rsp_cr_conn;
struct nvmet_fc_tgt_queue *queue;
int ret = 0;
@@ -1462,8 +1797,8 @@ nvmet_fc_ls_create_connection(struct nvmet_fc_tgtport *tgtport,
dev_err(tgtport->dev,
"Create Connection LS failed: %s\n",
validation_errors[ret]);
- iod->lsreq->rsplen = nvmet_fc_format_rjt(acc,
- NVME_FC_MAX_LS_BUFFER_SIZE, rqst->w0.ls_cmd,
+ iod->lsrsp->rsplen = nvme_fc_format_rjt(acc,
+ sizeof(*acc), rqst->w0.ls_cmd,
(ret == VERR_NO_ASSOC) ?
FCNVME_RJT_RC_INV_ASSOC :
FCNVME_RJT_RC_LOGIC,
@@ -1477,9 +1812,9 @@ nvmet_fc_ls_create_connection(struct nvmet_fc_tgtport *tgtport,
/* format a response */
- iod->lsreq->rsplen = sizeof(*acc);
+ iod->lsrsp->rsplen = sizeof(*acc);
- nvmet_fc_format_rsp_hdr(acc, FCNVME_LS_ACC,
+ nvme_fc_format_rsp_hdr(acc, FCNVME_LS_ACC,
fcnvme_lsdesc_len(sizeof(struct fcnvme_ls_cr_conn_acc)),
FCNVME_LS_CREATE_CONNECTION);
acc->connectid.desc_tag = cpu_to_be32(FCNVME_LSDESC_CONN_ID);
@@ -1491,46 +1826,28 @@ nvmet_fc_ls_create_connection(struct nvmet_fc_tgtport *tgtport,
be16_to_cpu(rqst->connect_cmd.qid)));
}
-static void
+/*
+ * Returns true if the LS response is to be transmit
+ * Returns false if the LS response is to be delayed
+ */
+static int
nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport,
struct nvmet_fc_ls_iod *iod)
{
struct fcnvme_ls_disconnect_assoc_rqst *rqst =
- (struct fcnvme_ls_disconnect_assoc_rqst *)iod->rqstbuf;
+ &iod->rqstbuf->rq_dis_assoc;
struct fcnvme_ls_disconnect_assoc_acc *acc =
- (struct fcnvme_ls_disconnect_assoc_acc *)iod->rspbuf;
- struct nvmet_fc_tgt_assoc *assoc;
+ &iod->rspbuf->rsp_dis_assoc;
+ struct nvmet_fc_tgt_assoc *assoc = NULL;
+ struct nvmet_fc_ls_iod *oldls = NULL;
+ unsigned long flags;
int ret = 0;
memset(acc, 0, sizeof(*acc));
- if (iod->rqstdatalen < sizeof(struct fcnvme_ls_disconnect_assoc_rqst))
- ret = VERR_DISCONN_LEN;
- else if (rqst->desc_list_len !=
- fcnvme_lsdesc_len(
- sizeof(struct fcnvme_ls_disconnect_assoc_rqst)))
- ret = VERR_DISCONN_RQST_LEN;
- else if (rqst->associd.desc_tag != cpu_to_be32(FCNVME_LSDESC_ASSOC_ID))
- ret = VERR_ASSOC_ID;
- else if (rqst->associd.desc_len !=
- fcnvme_lsdesc_len(
- sizeof(struct fcnvme_lsdesc_assoc_id)))
- ret = VERR_ASSOC_ID_LEN;
- else if (rqst->discon_cmd.desc_tag !=
- cpu_to_be32(FCNVME_LSDESC_DISCONN_CMD))
- ret = VERR_DISCONN_CMD;
- else if (rqst->discon_cmd.desc_len !=
- fcnvme_lsdesc_len(
- sizeof(struct fcnvme_lsdesc_disconn_cmd)))
- ret = VERR_DISCONN_CMD_LEN;
- /*
- * As the standard changed on the LS, check if old format and scope
- * something other than Association (e.g. 0).
- */
- else if (rqst->discon_cmd.rsvd8[0])
- ret = VERR_DISCONN_SCOPE;
- else {
- /* match an active association */
+ ret = nvmefc_vldt_lsreq_discon_assoc(iod->rqstdatalen, rqst);
+ if (!ret) {
+ /* match an active association - takes an assoc ref if !NULL */
assoc = nvmet_fc_find_target_assoc(tgtport,
be64_to_cpu(rqst->associd.association_id));
iod->assoc = assoc;
@@ -1538,34 +1855,63 @@ nvmet_fc_ls_disconnect(struct nvmet_fc_tgtport *tgtport,
ret = VERR_NO_ASSOC;
}
- if (ret) {
+ if (ret || !assoc) {
dev_err(tgtport->dev,
"Disconnect LS failed: %s\n",
validation_errors[ret]);
- iod->lsreq->rsplen = nvmet_fc_format_rjt(acc,
- NVME_FC_MAX_LS_BUFFER_SIZE, rqst->w0.ls_cmd,
+ iod->lsrsp->rsplen = nvme_fc_format_rjt(acc,
+ sizeof(*acc), rqst->w0.ls_cmd,
(ret == VERR_NO_ASSOC) ?
FCNVME_RJT_RC_INV_ASSOC :
- (ret == VERR_NO_CONN) ?
- FCNVME_RJT_RC_INV_CONN :
- FCNVME_RJT_RC_LOGIC,
+ FCNVME_RJT_RC_LOGIC,
FCNVME_RJT_EXP_NONE, 0);
- return;
+ return true;
}
/* format a response */
- iod->lsreq->rsplen = sizeof(*acc);
+ iod->lsrsp->rsplen = sizeof(*acc);
- nvmet_fc_format_rsp_hdr(acc, FCNVME_LS_ACC,
+ nvme_fc_format_rsp_hdr(acc, FCNVME_LS_ACC,
fcnvme_lsdesc_len(
sizeof(struct fcnvme_ls_disconnect_assoc_acc)),
FCNVME_LS_DISCONNECT_ASSOC);
/* release get taken in nvmet_fc_find_target_assoc */
- nvmet_fc_tgt_a_put(iod->assoc);
+ nvmet_fc_tgt_a_put(assoc);
+
+ /*
+ * The rules for LS response says the response cannot
+ * go back until ABTS's have been sent for all outstanding
+ * I/O and a Disconnect Association LS has been sent.
+ * So... save off the Disconnect LS to send the response
+ * later. If there was a prior LS already saved, replace
+ * it with the newer one and send a can't perform reject
+ * on the older one.
+ */
+ spin_lock_irqsave(&tgtport->lock, flags);
+ oldls = assoc->rcv_disconn;
+ assoc->rcv_disconn = iod;
+ spin_unlock_irqrestore(&tgtport->lock, flags);
+
+ nvmet_fc_delete_target_assoc(assoc);
- nvmet_fc_delete_target_assoc(iod->assoc);
+ if (oldls) {
+ dev_info(tgtport->dev,
+ "{%d:%d} Multiple Disconnect Association LS's "
+ "received\n",
+ tgtport->fc_target_port.port_num, assoc->a_id);
+ /* overwrite good response with bogus failure */
+ oldls->lsrsp->rsplen = nvme_fc_format_rjt(oldls->rspbuf,
+ sizeof(*iod->rspbuf),
+ /* ok to use rqst, LS is same */
+ rqst->w0.ls_cmd,
+ FCNVME_RJT_RC_UNAB,
+ FCNVME_RJT_EXP_NONE, 0);
+ nvmet_fc_xmt_ls_rsp(tgtport, oldls);
+ }
+
+ return false;
}
@@ -1577,13 +1923,13 @@ static void nvmet_fc_fcp_nvme_cmd_done(struct nvmet_req *nvme_req);
static const struct nvmet_fabrics_ops nvmet_fc_tgt_fcp_ops;
static void
-nvmet_fc_xmt_ls_rsp_done(struct nvmefc_tgt_ls_req *lsreq)
+nvmet_fc_xmt_ls_rsp_done(struct nvmefc_ls_rsp *lsrsp)
{
- struct nvmet_fc_ls_iod *iod = lsreq->nvmet_fc_private;
+ struct nvmet_fc_ls_iod *iod = lsrsp->nvme_fc_private;
struct nvmet_fc_tgtport *tgtport = iod->tgtport;
fc_dma_sync_single_for_cpu(tgtport->dev, iod->rspdma,
- NVME_FC_MAX_LS_BUFFER_SIZE, DMA_TO_DEVICE);
+ sizeof(*iod->rspbuf), DMA_TO_DEVICE);
nvmet_fc_free_ls_iod(tgtport, iod);
nvmet_fc_tgtport_put(tgtport);
}
@@ -1595,11 +1941,11 @@ nvmet_fc_xmt_ls_rsp(struct nvmet_fc_tgtport *tgtport,
int ret;
fc_dma_sync_single_for_device(tgtport->dev, iod->rspdma,
- NVME_FC_MAX_LS_BUFFER_SIZE, DMA_TO_DEVICE);
+ sizeof(*iod->rspbuf), DMA_TO_DEVICE);
- ret = tgtport->ops->xmt_ls_rsp(&tgtport->fc_target_port, iod->lsreq);
+ ret = tgtport->ops->xmt_ls_rsp(&tgtport->fc_target_port, iod->lsrsp);
if (ret)
- nvmet_fc_xmt_ls_rsp_done(iod->lsreq);
+ nvmet_fc_xmt_ls_rsp_done(iod->lsrsp);
}
/*
@@ -1609,15 +1955,15 @@ static void
nvmet_fc_handle_ls_rqst(struct nvmet_fc_tgtport *tgtport,
struct nvmet_fc_ls_iod *iod)
{
- struct fcnvme_ls_rqst_w0 *w0 =
- (struct fcnvme_ls_rqst_w0 *)iod->rqstbuf;
+ struct fcnvme_ls_rqst_w0 *w0 = &iod->rqstbuf->rq_cr_assoc.w0;
+ bool sendrsp = true;
- iod->lsreq->nvmet_fc_private = iod;
- iod->lsreq->rspbuf = iod->rspbuf;
- iod->lsreq->rspdma = iod->rspdma;
- iod->lsreq->done = nvmet_fc_xmt_ls_rsp_done;
+ iod->lsrsp->nvme_fc_private = iod;
+ iod->lsrsp->rspbuf = iod->rspbuf;
+ iod->lsrsp->rspdma = iod->rspdma;
+ iod->lsrsp->done = nvmet_fc_xmt_ls_rsp_done;
/* Be preventative. handlers will later set to valid length */
- iod->lsreq->rsplen = 0;
+ iod->lsrsp->rsplen = 0;
iod->assoc = NULL;
@@ -1637,15 +1983,16 @@ nvmet_fc_handle_ls_rqst(struct nvmet_fc_tgtport *tgtport,
break;
case FCNVME_LS_DISCONNECT_ASSOC:
/* Terminate a Queue/Connection or the Association */
- nvmet_fc_ls_disconnect(tgtport, iod);
+ sendrsp = nvmet_fc_ls_disconnect(tgtport, iod);
break;
default:
- iod->lsreq->rsplen = nvmet_fc_format_rjt(iod->rspbuf,
- NVME_FC_MAX_LS_BUFFER_SIZE, w0->ls_cmd,
+ iod->lsrsp->rsplen = nvme_fc_format_rjt(iod->rspbuf,
+ sizeof(*iod->rspbuf), w0->ls_cmd,
FCNVME_RJT_RC_INVAL, FCNVME_RJT_EXP_NONE, 0);
}
- nvmet_fc_xmt_ls_rsp(tgtport, iod);
+ if (sendrsp)
+ nvmet_fc_xmt_ls_rsp(tgtport, iod);
}
/*
@@ -1674,35 +2021,53 @@ nvmet_fc_handle_ls_rqst_work(struct work_struct *work)
*
* @target_port: pointer to the (registered) target port the LS was
* received on.
- * @lsreq: pointer to a lsreq request structure to be used to reference
+ * @lsrsp: pointer to a lsrsp structure to be used to reference
* the exchange corresponding to the LS.
* @lsreqbuf: pointer to the buffer containing the LS Request
* @lsreqbuf_len: length, in bytes, of the received LS request
*/
int
nvmet_fc_rcv_ls_req(struct nvmet_fc_target_port *target_port,
- struct nvmefc_tgt_ls_req *lsreq,
+ void *hosthandle,
+ struct nvmefc_ls_rsp *lsrsp,
void *lsreqbuf, u32 lsreqbuf_len)
{
struct nvmet_fc_tgtport *tgtport = targetport_to_tgtport(target_port);
struct nvmet_fc_ls_iod *iod;
-
- if (lsreqbuf_len > NVME_FC_MAX_LS_BUFFER_SIZE)
+ struct fcnvme_ls_rqst_w0 *w0 = (struct fcnvme_ls_rqst_w0 *)lsreqbuf;
+
+ if (lsreqbuf_len > sizeof(union nvmefc_ls_requests)) {
+ dev_info(tgtport->dev,
+ "RCV %s LS failed: payload too large (%d)\n",
+ (w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ?
+ nvmefc_ls_names[w0->ls_cmd] : "",
+ lsreqbuf_len);
return -E2BIG;
+ }
- if (!nvmet_fc_tgtport_get(tgtport))
+ if (!nvmet_fc_tgtport_get(tgtport)) {
+ dev_info(tgtport->dev,
+ "RCV %s LS failed: target deleting\n",
+ (w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ?
+ nvmefc_ls_names[w0->ls_cmd] : "");
return -ESHUTDOWN;
+ }
iod = nvmet_fc_alloc_ls_iod(tgtport);
if (!iod) {
+ dev_info(tgtport->dev,
+ "RCV %s LS failed: context allocation failed\n",
+ (w0->ls_cmd <= NVME_FC_LAST_LS_CMD_VALUE) ?
+ nvmefc_ls_names[w0->ls_cmd] : "");
nvmet_fc_tgtport_put(tgtport);
return -ENOENT;
}
- iod->lsreq = lsreq;
+ iod->lsrsp = lsrsp;
iod->fcpreq = NULL;
memcpy(iod->rqstbuf, lsreqbuf, lsreqbuf_len);
iod->rqstdatalen = lsreqbuf_len;
+ iod->hosthandle = hosthandle;
schedule_work(&iod->work);
diff --git a/drivers/nvme/target/fcloop.c b/drivers/nvme/target/fcloop.c
index f69ce66e2d44..2ff1d1334a03 100644
--- a/drivers/nvme/target/fcloop.c
+++ b/drivers/nvme/target/fcloop.c
@@ -208,10 +208,13 @@ struct fcloop_rport {
};
struct fcloop_tport {
- struct nvmet_fc_target_port *targetport;
- struct nvme_fc_remote_port *remoteport;
- struct fcloop_nport *nport;
- struct fcloop_lport *lport;
+ struct nvmet_fc_target_port *targetport;
+ struct nvme_fc_remote_port *remoteport;
+ struct fcloop_nport *nport;
+ struct fcloop_lport *lport;
+ spinlock_t lock;
+ struct list_head ls_list;
+ struct work_struct ls_work;
};
struct fcloop_nport {
@@ -228,7 +231,8 @@ struct fcloop_nport {
struct fcloop_lsreq {
struct nvmefc_ls_req *lsreq;
- struct nvmefc_tgt_ls_req tgt_ls_req;
+ struct nvmefc_ls_rsp ls_rsp;
+ int lsdir; /* H2T or T2H */
int status;
struct list_head ls_list; /* fcloop_rport->ls_list */
};
@@ -267,9 +271,9 @@ struct fcloop_ini_fcpreq {
};
static inline struct fcloop_lsreq *
-tgt_ls_req_to_lsreq(struct nvmefc_tgt_ls_req *tgt_lsreq)
+ls_rsp_to_lsreq(struct nvmefc_ls_rsp *lsrsp)
{
- return container_of(tgt_lsreq, struct fcloop_lsreq, tgt_ls_req);
+ return container_of(lsrsp, struct fcloop_lsreq, ls_rsp);
}
static inline struct fcloop_fcpreq *
@@ -323,7 +327,7 @@ fcloop_rport_lsrqst_work(struct work_struct *work)
}
static int
-fcloop_ls_req(struct nvme_fc_local_port *localport,
+fcloop_h2t_ls_req(struct nvme_fc_local_port *localport,
struct nvme_fc_remote_port *remoteport,
struct nvmefc_ls_req *lsreq)
{
@@ -344,27 +348,28 @@ fcloop_ls_req(struct nvme_fc_local_port *localport,
}
tls_req->status = 0;
- ret = nvmet_fc_rcv_ls_req(rport->targetport, &tls_req->tgt_ls_req,
- lsreq->rqstaddr, lsreq->rqstlen);
+ ret = nvmet_fc_rcv_ls_req(rport->targetport, rport,
+ &tls_req->ls_rsp,
+ lsreq->rqstaddr, lsreq->rqstlen);
return ret;
}
static int
-fcloop_xmt_ls_rsp(struct nvmet_fc_target_port *targetport,
- struct nvmefc_tgt_ls_req *tgt_lsreq)
+fcloop_h2t_xmt_ls_rsp(struct nvmet_fc_target_port *targetport,
+ struct nvmefc_ls_rsp *lsrsp)
{
- struct fcloop_lsreq *tls_req = tgt_ls_req_to_lsreq(tgt_lsreq);
+ struct fcloop_lsreq *tls_req = ls_rsp_to_lsreq(lsrsp);
struct nvmefc_ls_req *lsreq = tls_req->lsreq;
struct fcloop_tport *tport = targetport->private;
struct nvme_fc_remote_port *remoteport = tport->remoteport;
struct fcloop_rport *rport;
- memcpy(lsreq->rspaddr, tgt_lsreq->rspbuf,
- ((lsreq->rsplen < tgt_lsreq->rsplen) ?
- lsreq->rsplen : tgt_lsreq->rsplen));
+ memcpy(lsreq->rspaddr, lsrsp->rspbuf,
+ ((lsreq->rsplen < lsrsp->rsplen) ?
+ lsreq->rsplen : lsrsp->rsplen));
- tgt_lsreq->done(tgt_lsreq);
+ lsrsp->done(lsrsp);
if (remoteport) {
rport = remoteport->private;
@@ -377,6 +382,99 @@ fcloop_xmt_ls_rsp(struct nvmet_fc_target_port *targetport,
return 0;
}
+static void
+fcloop_tport_lsrqst_work(struct work_struct *work)
+{
+ struct fcloop_tport *tport =
+ container_of(work, struct fcloop_tport, ls_work);
+ struct fcloop_lsreq *tls_req;
+
+ spin_lock(&tport->lock);
+ for (;;) {
+ tls_req = list_first_entry_or_null(&tport->ls_list,
+ struct fcloop_lsreq, ls_list);
+ if (!tls_req)
+ break;
+
+ list_del(&tls_req->ls_list);
+ spin_unlock(&tport->lock);
+
+ tls_req->lsreq->done(tls_req->lsreq, tls_req->status);
+ /*
+ * callee may free memory containing tls_req.
+ * do not reference lsreq after this.
+ */
+
+ spin_lock(&tport->lock);
+ }
+ spin_unlock(&tport->lock);
+}
+
+static int
+fcloop_t2h_ls_req(struct nvmet_fc_target_port *targetport, void *hosthandle,
+ struct nvmefc_ls_req *lsreq)
+{
+ struct fcloop_lsreq *tls_req = lsreq->private;
+ struct fcloop_tport *tport = targetport->private;
+ int ret = 0;
+
+ /*
+ * hosthandle should be the dst.rport value.
+ * hosthandle ignored as fcloop currently is
+ * 1:1 tgtport vs remoteport
+ */
+ tls_req->lsreq = lsreq;
+ INIT_LIST_HEAD(&tls_req->ls_list);
+
+ if (!tport->remoteport) {
+ tls_req->status = -ECONNREFUSED;
+ spin_lock(&tport->lock);
+ list_add_tail(&tport->ls_list, &tls_req->ls_list);
+ spin_unlock(&tport->lock);
+ schedule_work(&tport->ls_work);
+ return ret;
+ }
+
+ tls_req->status = 0;
+ ret = nvme_fc_rcv_ls_req(tport->remoteport, &tls_req->ls_rsp,
+ lsreq->rqstaddr, lsreq->rqstlen);
+
+ return ret;
+}
+
+static int
+fcloop_t2h_xmt_ls_rsp(struct nvme_fc_local_port *localport,
+ struct nvme_fc_remote_port *remoteport,
+ struct nvmefc_ls_rsp *lsrsp)
+{
+ struct fcloop_lsreq *tls_req = ls_rsp_to_lsreq(lsrsp);
+ struct nvmefc_ls_req *lsreq = tls_req->lsreq;
+ struct fcloop_rport *rport = remoteport->private;
+ struct nvmet_fc_target_port *targetport = rport->targetport;
+ struct fcloop_tport *tport;
+
+ memcpy(lsreq->rspaddr, lsrsp->rspbuf,
+ ((lsreq->rsplen < lsrsp->rsplen) ?
+ lsreq->rsplen : lsrsp->rsplen));
+ lsrsp->done(lsrsp);
+
+ if (targetport) {
+ tport = targetport->private;
+ spin_lock(&tport->lock);
+ list_add_tail(&tport->ls_list, &tls_req->ls_list);
+ spin_unlock(&tport->lock);
+ schedule_work(&tport->ls_work);
+ }
+
+ return 0;
+}
+
+static void
+fcloop_t2h_host_release(void *hosthandle)
+{
+ /* host handle ignored for now */
+}
+
/*
* Simulate reception of RSCN and converting it to a initiator transport
* call to rescan a remote port.
@@ -762,13 +860,19 @@ fcloop_fcp_req_release(struct nvmet_fc_target_port *tgtport,
}
static void
-fcloop_ls_abort(struct nvme_fc_local_port *localport,
+fcloop_h2t_ls_abort(struct nvme_fc_local_port *localport,
struct nvme_fc_remote_port *remoteport,
struct nvmefc_ls_req *lsreq)
{
}
static void
+fcloop_t2h_ls_abort(struct nvmet_fc_target_port *targetport,
+ void *hosthandle, struct nvmefc_ls_req *lsreq)
+{
+}
+
+static void
fcloop_fcp_abort(struct nvme_fc_local_port *localport,
struct nvme_fc_remote_port *remoteport,
void *hw_queue_handle,
@@ -867,6 +971,7 @@ fcloop_targetport_delete(struct nvmet_fc_target_port *targetport)
{
struct fcloop_tport *tport = targetport->private;
+ flush_work(&tport->ls_work);
fcloop_nport_put(tport->nport);
}
@@ -879,10 +984,11 @@ static struct nvme_fc_port_template fctemplate = {
.remoteport_delete = fcloop_remoteport_delete,
.create_queue = fcloop_create_queue,
.delete_queue = fcloop_delete_queue,
- .ls_req = fcloop_ls_req,
+ .ls_req = fcloop_h2t_ls_req,
.fcp_io = fcloop_fcp_req,
- .ls_abort = fcloop_ls_abort,
+ .ls_abort = fcloop_h2t_ls_abort,
.fcp_abort = fcloop_fcp_abort,
+ .xmt_ls_rsp = fcloop_t2h_xmt_ls_rsp,
.max_hw_queues = FCLOOP_HW_QUEUES,
.max_sgl_segments = FCLOOP_SGL_SEGS,
.max_dif_sgl_segments = FCLOOP_SGL_SEGS,
@@ -896,11 +1002,14 @@ static struct nvme_fc_port_template fctemplate = {
static struct nvmet_fc_target_template tgttemplate = {
.targetport_delete = fcloop_targetport_delete,
- .xmt_ls_rsp = fcloop_xmt_ls_rsp,
+ .xmt_ls_rsp = fcloop_h2t_xmt_ls_rsp,
.fcp_op = fcloop_fcp_op,
.fcp_abort = fcloop_tgt_fcp_abort,
.fcp_req_release = fcloop_fcp_req_release,
.discovery_event = fcloop_tgt_discovery_evt,
+ .ls_req = fcloop_t2h_ls_req,
+ .ls_abort = fcloop_t2h_ls_abort,
+ .host_release = fcloop_t2h_host_release,
.max_hw_queues = FCLOOP_HW_QUEUES,
.max_sgl_segments = FCLOOP_SGL_SEGS,
.max_dif_sgl_segments = FCLOOP_SGL_SEGS,
@@ -909,6 +1018,7 @@ static struct nvmet_fc_target_template tgttemplate = {
.target_features = 0,
/* sizes of additional private data for data structures */
.target_priv_sz = sizeof(struct fcloop_tport),
+ .lsrqst_priv_sz = sizeof(struct fcloop_lsreq),
};
static ssize_t
@@ -1258,6 +1368,9 @@ fcloop_create_target_port(struct device *dev, struct device_attribute *attr,
tport->nport = nport;
tport->lport = nport->lport;
nport->tport = tport;
+ spin_lock_init(&tport->lock);
+ INIT_WORK(&tport->ls_work, fcloop_tport_lsrqst_work);
+ INIT_LIST_HEAD(&tport->ls_list);
return count;
}
diff --git a/drivers/nvme/target/io-cmd-bdev.c b/drivers/nvme/target/io-cmd-bdev.c
index 26f50c23b82e..3dd6f566a240 100644
--- a/drivers/nvme/target/io-cmd-bdev.c
+++ b/drivers/nvme/target/io-cmd-bdev.c
@@ -47,6 +47,22 @@ void nvmet_bdev_set_limits(struct block_device *bdev, struct nvme_id_ns *id)
id->nows = to0based(ql->io_opt / ql->logical_block_size);
}
+static void nvmet_bdev_ns_enable_integrity(struct nvmet_ns *ns)
+{
+ struct blk_integrity *bi = bdev_get_integrity(ns->bdev);
+
+ if (bi) {
+ ns->metadata_size = bi->tuple_size;
+ if (bi->profile == &t10_pi_type1_crc)
+ ns->pi_type = NVME_NS_DPS_PI_TYPE1;
+ else if (bi->profile == &t10_pi_type3_crc)
+ ns->pi_type = NVME_NS_DPS_PI_TYPE3;
+ else
+ /* Unsupported metadata type */
+ ns->metadata_size = 0;
+ }
+}
+
int nvmet_bdev_ns_enable(struct nvmet_ns *ns)
{
int ret;
@@ -64,6 +80,12 @@ int nvmet_bdev_ns_enable(struct nvmet_ns *ns)
}
ns->size = i_size_read(ns->bdev->bd_inode);
ns->blksize_shift = blksize_bits(bdev_logical_block_size(ns->bdev));
+
+ ns->pi_type = 0;
+ ns->metadata_size = 0;
+ if (IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY_T10))
+ nvmet_bdev_ns_enable_integrity(ns);
+
return 0;
}
@@ -75,6 +97,11 @@ void nvmet_bdev_ns_disable(struct nvmet_ns *ns)
}
}
+void nvmet_bdev_ns_revalidate(struct nvmet_ns *ns)
+{
+ ns->size = i_size_read(ns->bdev->bd_inode);
+}
+
static u16 blk_to_nvme_status(struct nvmet_req *req, blk_status_t blk_sts)
{
u16 status = NVME_SC_SUCCESS;
@@ -142,6 +169,61 @@ static void nvmet_bio_done(struct bio *bio)
bio_put(bio);
}
+#ifdef CONFIG_BLK_DEV_INTEGRITY
+static int nvmet_bdev_alloc_bip(struct nvmet_req *req, struct bio *bio,
+ struct sg_mapping_iter *miter)
+{
+ struct blk_integrity *bi;
+ struct bio_integrity_payload *bip;
+ struct block_device *bdev = req->ns->bdev;
+ int rc;
+ size_t resid, len;
+
+ bi = bdev_get_integrity(bdev);
+ if (unlikely(!bi)) {
+ pr_err("Unable to locate bio_integrity\n");
+ return -ENODEV;
+ }
+
+ bip = bio_integrity_alloc(bio, GFP_NOIO,
+ min_t(unsigned int, req->metadata_sg_cnt, BIO_MAX_PAGES));
+ if (IS_ERR(bip)) {
+ pr_err("Unable to allocate bio_integrity_payload\n");
+ return PTR_ERR(bip);
+ }
+
+ bip->bip_iter.bi_size = bio_integrity_bytes(bi, bio_sectors(bio));
+ /* virtual start sector must be in integrity interval units */
+ bip_set_seed(bip, bio->bi_iter.bi_sector >>
+ (bi->interval_exp - SECTOR_SHIFT));
+
+ resid = bip->bip_iter.bi_size;
+ while (resid > 0 && sg_miter_next(miter)) {
+ len = min_t(size_t, miter->length, resid);
+ rc = bio_integrity_add_page(bio, miter->page, len,
+ offset_in_page(miter->addr));
+ if (unlikely(rc != len)) {
+ pr_err("bio_integrity_add_page() failed; %d\n", rc);
+ sg_miter_stop(miter);
+ return -ENOMEM;
+ }
+
+ resid -= len;
+ if (len < miter->length)
+ miter->consumed -= miter->length - len;
+ }
+ sg_miter_stop(miter);
+
+ return 0;
+}
+#else
+static int nvmet_bdev_alloc_bip(struct nvmet_req *req, struct bio *bio,
+ struct sg_mapping_iter *miter)
+{
+ return -EINVAL;
+}
+#endif /* CONFIG_BLK_DEV_INTEGRITY */
+
static void nvmet_bdev_execute_rw(struct nvmet_req *req)
{
int sg_cnt = req->sg_cnt;
@@ -149,9 +231,12 @@ static void nvmet_bdev_execute_rw(struct nvmet_req *req)
struct scatterlist *sg;
struct blk_plug plug;
sector_t sector;
- int op, i;
+ int op, i, rc;
+ struct sg_mapping_iter prot_miter;
+ unsigned int iter_flags;
+ unsigned int total_len = nvmet_rw_data_len(req) + req->metadata_len;
- if (!nvmet_check_data_len(req, nvmet_rw_len(req)))
+ if (!nvmet_check_transfer_len(req, total_len))
return;
if (!req->sg_cnt) {
@@ -163,8 +248,10 @@ static void nvmet_bdev_execute_rw(struct nvmet_req *req)
op = REQ_OP_WRITE | REQ_SYNC | REQ_IDLE;
if (req->cmd->rw.control & cpu_to_le16(NVME_RW_FUA))
op |= REQ_FUA;
+ iter_flags = SG_MITER_TO_SG;
} else {
op = REQ_OP_READ;
+ iter_flags = SG_MITER_FROM_SG;
}
if (is_pci_p2pdma_page(sg_page(req->sg)))
@@ -186,11 +273,24 @@ static void nvmet_bdev_execute_rw(struct nvmet_req *req)
bio->bi_opf = op;
blk_start_plug(&plug);
+ if (req->metadata_len)
+ sg_miter_start(&prot_miter, req->metadata_sg,
+ req->metadata_sg_cnt, iter_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;
+ if (req->metadata_len) {
+ rc = nvmet_bdev_alloc_bip(req, bio,
+ &prot_miter);
+ if (unlikely(rc)) {
+ bio_io_error(bio);
+ return;
+ }
+ }
+
bio = bio_alloc(GFP_KERNEL, min(sg_cnt, BIO_MAX_PAGES));
bio_set_dev(bio, req->ns->bdev);
bio->bi_iter.bi_sector = sector;
@@ -204,6 +304,14 @@ static void nvmet_bdev_execute_rw(struct nvmet_req *req)
sg_cnt--;
}
+ if (req->metadata_len) {
+ rc = nvmet_bdev_alloc_bip(req, bio, &prot_miter);
+ if (unlikely(rc)) {
+ bio_io_error(bio);
+ return;
+ }
+ }
+
submit_bio(bio);
blk_finish_plug(&plug);
}
@@ -212,7 +320,7 @@ static void nvmet_bdev_execute_flush(struct nvmet_req *req)
{
struct bio *bio = &req->b.inline_bio;
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
bio_init(bio, req->inline_bvec, ARRAY_SIZE(req->inline_bvec));
@@ -304,7 +412,7 @@ static void nvmet_bdev_execute_write_zeroes(struct nvmet_req *req)
sector_t nr_sector;
int ret;
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
sector = le64_to_cpu(write_zeroes->slba) <<
@@ -331,6 +439,8 @@ u16 nvmet_bdev_parse_io_cmd(struct nvmet_req *req)
case nvme_cmd_read:
case nvme_cmd_write:
req->execute = nvmet_bdev_execute_rw;
+ if (req->sq->ctrl->pi_support && nvmet_ns_has_pi(req->ns))
+ req->metadata_len = nvmet_rw_metadata_len(req);
return 0;
case nvme_cmd_flush:
req->execute = nvmet_bdev_execute_flush;
diff --git a/drivers/nvme/target/io-cmd-file.c b/drivers/nvme/target/io-cmd-file.c
index cd5670b83118..0abbefd9925e 100644
--- a/drivers/nvme/target/io-cmd-file.c
+++ b/drivers/nvme/target/io-cmd-file.c
@@ -13,6 +13,18 @@
#define NVMET_MAX_MPOOL_BVEC 16
#define NVMET_MIN_MPOOL_OBJ 16
+int nvmet_file_ns_revalidate(struct nvmet_ns *ns)
+{
+ struct kstat stat;
+ int ret;
+
+ ret = vfs_getattr(&ns->file->f_path, &stat, STATX_SIZE,
+ AT_STATX_FORCE_SYNC);
+ if (!ret)
+ ns->size = stat.size;
+ return ret;
+}
+
void nvmet_file_ns_disable(struct nvmet_ns *ns)
{
if (ns->file) {
@@ -30,7 +42,6 @@ void nvmet_file_ns_disable(struct nvmet_ns *ns)
int nvmet_file_ns_enable(struct nvmet_ns *ns)
{
int flags = O_RDWR | O_LARGEFILE;
- struct kstat stat;
int ret;
if (!ns->buffered_io)
@@ -43,12 +54,10 @@ int nvmet_file_ns_enable(struct nvmet_ns *ns)
return PTR_ERR(ns->file);
}
- ret = vfs_getattr(&ns->file->f_path,
- &stat, STATX_SIZE, AT_STATX_FORCE_SYNC);
+ ret = nvmet_file_ns_revalidate(ns);
if (ret)
goto err;
- ns->size = stat.size;
/*
* i_blkbits can be greater than the universally accepted upper bound,
* so make sure we export a sane namespace lba_shift.
@@ -232,7 +241,7 @@ static void nvmet_file_execute_rw(struct nvmet_req *req)
{
ssize_t nr_bvec = req->sg_cnt;
- if (!nvmet_check_data_len(req, nvmet_rw_len(req)))
+ if (!nvmet_check_transfer_len(req, nvmet_rw_data_len(req)))
return;
if (!req->sg_cnt || !nr_bvec) {
@@ -276,7 +285,7 @@ static void nvmet_file_flush_work(struct work_struct *w)
static void nvmet_file_execute_flush(struct nvmet_req *req)
{
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
INIT_WORK(&req->f.work, nvmet_file_flush_work);
schedule_work(&req->f.work);
@@ -366,7 +375,7 @@ static void nvmet_file_write_zeroes_work(struct work_struct *w)
static void nvmet_file_execute_write_zeroes(struct nvmet_req *req)
{
- if (!nvmet_check_data_len(req, 0))
+ if (!nvmet_check_transfer_len(req, 0))
return;
INIT_WORK(&req->f.work, nvmet_file_write_zeroes_work);
schedule_work(&req->f.work);
diff --git a/drivers/nvme/target/nvmet.h b/drivers/nvme/target/nvmet.h
index 421dff3ea143..809691291e73 100644
--- a/drivers/nvme/target/nvmet.h
+++ b/drivers/nvme/target/nvmet.h
@@ -19,6 +19,7 @@
#include <linux/rcupdate.h>
#include <linux/blkdev.h>
#include <linux/radix-tree.h>
+#include <linux/t10-pi.h>
#define NVMET_ASYNC_EVENTS 4
#define NVMET_ERROR_LOG_SLOTS 128
@@ -77,6 +78,8 @@ struct nvmet_ns {
int use_p2pmem;
struct pci_dev *p2p_dev;
+ int pi_type;
+ int metadata_size;
};
static inline struct nvmet_ns *to_nvmet_ns(struct config_item *item)
@@ -142,6 +145,7 @@ struct nvmet_port {
bool enabled;
int inline_data_size;
const struct nvmet_fabrics_ops *tr_ops;
+ bool pi_enable;
};
static inline struct nvmet_port *to_nvmet_port(struct config_item *item)
@@ -201,6 +205,7 @@ struct nvmet_ctrl {
spinlock_t error_lock;
u64 err_counter;
struct nvme_error_slot slots[NVMET_ERROR_LOG_SLOTS];
+ bool pi_support;
};
struct nvmet_subsys_model {
@@ -230,6 +235,7 @@ struct nvmet_subsys {
u64 ver;
u64 serial;
char *subsysnqn;
+ bool pi_support;
struct config_group group;
@@ -281,6 +287,7 @@ struct nvmet_fabrics_ops {
unsigned int type;
unsigned int msdbd;
bool has_keyed_sgls : 1;
+ bool metadata_support : 1;
void (*queue_response)(struct nvmet_req *req);
int (*add_port)(struct nvmet_port *port);
void (*remove_port)(struct nvmet_port *port);
@@ -302,6 +309,7 @@ struct nvmet_req {
struct nvmet_cq *cq;
struct nvmet_ns *ns;
struct scatterlist *sg;
+ struct scatterlist *metadata_sg;
struct bio_vec inline_bvec[NVMET_MAX_INLINE_BIOVEC];
union {
struct {
@@ -315,8 +323,10 @@ struct nvmet_req {
} f;
};
int sg_cnt;
+ int metadata_sg_cnt;
/* data length as parsed from the SGL descriptor: */
size_t transfer_len;
+ size_t metadata_len;
struct nvmet_port *port;
@@ -384,11 +394,11 @@ u16 nvmet_parse_fabrics_cmd(struct nvmet_req *req);
bool nvmet_req_init(struct nvmet_req *req, struct nvmet_cq *cq,
struct nvmet_sq *sq, const struct nvmet_fabrics_ops *ops);
void nvmet_req_uninit(struct nvmet_req *req);
-bool nvmet_check_data_len(struct nvmet_req *req, size_t data_len);
+bool nvmet_check_transfer_len(struct nvmet_req *req, size_t len);
bool nvmet_check_data_len_lte(struct nvmet_req *req, size_t data_len);
void nvmet_req_complete(struct nvmet_req *req, u16 status);
-int nvmet_req_alloc_sgl(struct nvmet_req *req);
-void nvmet_req_free_sgl(struct nvmet_req *req);
+int nvmet_req_alloc_sgls(struct nvmet_req *req);
+void nvmet_req_free_sgls(struct nvmet_req *req);
void nvmet_execute_keep_alive(struct nvmet_req *req);
@@ -498,13 +508,24 @@ void nvmet_file_ns_disable(struct nvmet_ns *ns);
u16 nvmet_bdev_flush(struct nvmet_req *req);
u16 nvmet_file_flush(struct nvmet_req *req);
void nvmet_ns_changed(struct nvmet_subsys *subsys, u32 nsid);
+void nvmet_bdev_ns_revalidate(struct nvmet_ns *ns);
+int nvmet_file_ns_revalidate(struct nvmet_ns *ns);
+void nvmet_ns_revalidate(struct nvmet_ns *ns);
-static inline u32 nvmet_rw_len(struct nvmet_req *req)
+static inline u32 nvmet_rw_data_len(struct nvmet_req *req)
{
return ((u32)le16_to_cpu(req->cmd->rw.length) + 1) <<
req->ns->blksize_shift;
}
+static inline u32 nvmet_rw_metadata_len(struct nvmet_req *req)
+{
+ if (!IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY))
+ return 0;
+ return ((u32)le16_to_cpu(req->cmd->rw.length) + 1) *
+ req->ns->metadata_size;
+}
+
static inline u32 nvmet_dsm_len(struct nvmet_req *req)
{
return (le32_to_cpu(req->cmd->dsm.nr) + 1) *
@@ -519,4 +540,11 @@ static inline __le16 to0based(u32 a)
return cpu_to_le16(max(1U, min(1U << 16, a)) - 1);
}
+static inline bool nvmet_ns_has_pi(struct nvmet_ns *ns)
+{
+ if (!IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY))
+ return false;
+ return ns->pi_type && ns->metadata_size == sizeof(struct t10_pi_tuple);
+}
+
#endif /* _NVMET_H */
diff --git a/drivers/nvme/target/rdma.c b/drivers/nvme/target/rdma.c
index fd47de0e4e4e..d5141780592e 100644
--- a/drivers/nvme/target/rdma.c
+++ b/drivers/nvme/target/rdma.c
@@ -33,6 +33,9 @@
/* Assume mpsmin == device_page_size == 4KB */
#define NVMET_RDMA_MAX_MDTS 8
+#define NVMET_RDMA_MAX_METADATA_MDTS 5
+
+struct nvmet_rdma_srq;
struct nvmet_rdma_cmd {
struct ib_sge sge[NVMET_RDMA_MAX_INLINE_SGE + 1];
@@ -41,6 +44,7 @@ struct nvmet_rdma_cmd {
struct scatterlist inline_sg[NVMET_RDMA_MAX_INLINE_SGE];
struct nvme_command *nvme_cmd;
struct nvmet_rdma_queue *queue;
+ struct nvmet_rdma_srq *nsrq;
};
enum {
@@ -57,6 +61,7 @@ struct nvmet_rdma_rsp {
struct nvmet_rdma_queue *queue;
struct ib_cqe read_cqe;
+ struct ib_cqe write_cqe;
struct rdma_rw_ctx rw;
struct nvmet_req req;
@@ -83,6 +88,7 @@ struct nvmet_rdma_queue {
struct ib_cq *cq;
atomic_t sq_wr_avail;
struct nvmet_rdma_device *dev;
+ struct nvmet_rdma_srq *nsrq;
spinlock_t state_lock;
enum nvmet_rdma_queue_state state;
struct nvmet_cq nvme_cq;
@@ -100,6 +106,7 @@ struct nvmet_rdma_queue {
int idx;
int host_qid;
+ int comp_vector;
int recv_queue_size;
int send_queue_size;
@@ -113,11 +120,17 @@ struct nvmet_rdma_port {
struct delayed_work repair_work;
};
+struct nvmet_rdma_srq {
+ struct ib_srq *srq;
+ struct nvmet_rdma_cmd *cmds;
+ struct nvmet_rdma_device *ndev;
+};
+
struct nvmet_rdma_device {
struct ib_device *device;
struct ib_pd *pd;
- struct ib_srq *srq;
- struct nvmet_rdma_cmd *srq_cmds;
+ struct nvmet_rdma_srq **srqs;
+ int srq_count;
size_t srq_size;
struct kref ref;
struct list_head entry;
@@ -129,6 +142,16 @@ static bool nvmet_rdma_use_srq;
module_param_named(use_srq, nvmet_rdma_use_srq, bool, 0444);
MODULE_PARM_DESC(use_srq, "Use shared receive queue.");
+static int srq_size_set(const char *val, const struct kernel_param *kp);
+static const struct kernel_param_ops srq_size_ops = {
+ .set = srq_size_set,
+ .get = param_get_int,
+};
+
+static int nvmet_rdma_srq_size = 1024;
+module_param_cb(srq_size, &srq_size_ops, &nvmet_rdma_srq_size, 0644);
+MODULE_PARM_DESC(srq_size, "set Shared Receive Queue (SRQ) size, should >= 256 (default: 1024)");
+
static DEFINE_IDA(nvmet_rdma_queue_ida);
static LIST_HEAD(nvmet_rdma_queue_list);
static DEFINE_MUTEX(nvmet_rdma_queue_mutex);
@@ -140,6 +163,7 @@ static bool nvmet_rdma_execute_command(struct nvmet_rdma_rsp *rsp);
static void nvmet_rdma_send_done(struct ib_cq *cq, struct ib_wc *wc);
static void nvmet_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc);
static void nvmet_rdma_read_data_done(struct ib_cq *cq, struct ib_wc *wc);
+static void nvmet_rdma_write_data_done(struct ib_cq *cq, struct ib_wc *wc);
static void nvmet_rdma_qp_event(struct ib_event *event, void *priv);
static void nvmet_rdma_queue_disconnect(struct nvmet_rdma_queue *queue);
static void nvmet_rdma_free_rsp(struct nvmet_rdma_device *ndev,
@@ -149,6 +173,17 @@ static int nvmet_rdma_alloc_rsp(struct nvmet_rdma_device *ndev,
static const struct nvmet_fabrics_ops nvmet_rdma_ops;
+static int srq_size_set(const char *val, const struct kernel_param *kp)
+{
+ int n = 0, ret;
+
+ ret = kstrtoint(val, 10, &n);
+ if (ret != 0 || n < 256)
+ return -EINVAL;
+
+ return param_set_int(val, kp);
+}
+
static int num_pages(int len)
{
return 1 + (((len - 1) & PAGE_MASK) >> PAGE_SHIFT);
@@ -391,6 +426,9 @@ static int nvmet_rdma_alloc_rsp(struct nvmet_rdma_device *ndev,
/* Data In / RDMA READ */
r->read_cqe.done = nvmet_rdma_read_data_done;
+ /* Data Out / RDMA WRITE */
+ r->write_cqe.done = nvmet_rdma_write_data_done;
+
return 0;
out_free_rsp:
@@ -466,8 +504,8 @@ static int nvmet_rdma_post_recv(struct nvmet_rdma_device *ndev,
cmd->sge[0].addr, cmd->sge[0].length,
DMA_FROM_DEVICE);
- if (ndev->srq)
- ret = ib_post_srq_recv(ndev->srq, &cmd->wr, NULL);
+ if (cmd->nsrq)
+ ret = ib_post_srq_recv(cmd->nsrq->srq, &cmd->wr, NULL);
else
ret = ib_post_recv(cmd->queue->qp, &cmd->wr, NULL);
@@ -500,6 +538,129 @@ static void nvmet_rdma_process_wr_wait_list(struct nvmet_rdma_queue *queue)
spin_unlock(&queue->rsp_wr_wait_lock);
}
+static u16 nvmet_rdma_check_pi_status(struct ib_mr *sig_mr)
+{
+ struct ib_mr_status mr_status;
+ int ret;
+ u16 status = 0;
+
+ ret = ib_check_mr_status(sig_mr, IB_MR_CHECK_SIG_STATUS, &mr_status);
+ if (ret) {
+ pr_err("ib_check_mr_status failed, ret %d\n", ret);
+ return NVME_SC_INVALID_PI;
+ }
+
+ if (mr_status.fail_status & IB_MR_CHECK_SIG_STATUS) {
+ switch (mr_status.sig_err.err_type) {
+ case IB_SIG_BAD_GUARD:
+ status = NVME_SC_GUARD_CHECK;
+ break;
+ case IB_SIG_BAD_REFTAG:
+ status = NVME_SC_REFTAG_CHECK;
+ break;
+ case IB_SIG_BAD_APPTAG:
+ status = NVME_SC_APPTAG_CHECK;
+ break;
+ }
+ pr_err("PI error found type %d expected 0x%x vs actual 0x%x\n",
+ mr_status.sig_err.err_type,
+ mr_status.sig_err.expected,
+ mr_status.sig_err.actual);
+ }
+
+ return status;
+}
+
+static void nvmet_rdma_set_sig_domain(struct blk_integrity *bi,
+ struct nvme_command *cmd, struct ib_sig_domain *domain,
+ u16 control, u8 pi_type)
+{
+ domain->sig_type = IB_SIG_TYPE_T10_DIF;
+ domain->sig.dif.bg_type = IB_T10DIF_CRC;
+ domain->sig.dif.pi_interval = 1 << bi->interval_exp;
+ domain->sig.dif.ref_tag = le32_to_cpu(cmd->rw.reftag);
+ if (control & NVME_RW_PRINFO_PRCHK_REF)
+ domain->sig.dif.ref_remap = true;
+
+ domain->sig.dif.app_tag = le16_to_cpu(cmd->rw.apptag);
+ domain->sig.dif.apptag_check_mask = le16_to_cpu(cmd->rw.appmask);
+ domain->sig.dif.app_escape = true;
+ if (pi_type == NVME_NS_DPS_PI_TYPE3)
+ domain->sig.dif.ref_escape = true;
+}
+
+static void nvmet_rdma_set_sig_attrs(struct nvmet_req *req,
+ struct ib_sig_attrs *sig_attrs)
+{
+ struct nvme_command *cmd = req->cmd;
+ u16 control = le16_to_cpu(cmd->rw.control);
+ u8 pi_type = req->ns->pi_type;
+ struct blk_integrity *bi;
+
+ bi = bdev_get_integrity(req->ns->bdev);
+
+ memset(sig_attrs, 0, sizeof(*sig_attrs));
+
+ if (control & NVME_RW_PRINFO_PRACT) {
+ /* for WRITE_INSERT/READ_STRIP no wire domain */
+ sig_attrs->wire.sig_type = IB_SIG_TYPE_NONE;
+ nvmet_rdma_set_sig_domain(bi, cmd, &sig_attrs->mem, control,
+ pi_type);
+ /* Clear the PRACT bit since HCA will generate/verify the PI */
+ control &= ~NVME_RW_PRINFO_PRACT;
+ cmd->rw.control = cpu_to_le16(control);
+ /* PI is added by the HW */
+ req->transfer_len += req->metadata_len;
+ } else {
+ /* for WRITE_PASS/READ_PASS both wire/memory domains exist */
+ nvmet_rdma_set_sig_domain(bi, cmd, &sig_attrs->wire, control,
+ pi_type);
+ nvmet_rdma_set_sig_domain(bi, cmd, &sig_attrs->mem, control,
+ pi_type);
+ }
+
+ if (control & NVME_RW_PRINFO_PRCHK_REF)
+ sig_attrs->check_mask |= IB_SIG_CHECK_REFTAG;
+ if (control & NVME_RW_PRINFO_PRCHK_GUARD)
+ sig_attrs->check_mask |= IB_SIG_CHECK_GUARD;
+ if (control & NVME_RW_PRINFO_PRCHK_APP)
+ sig_attrs->check_mask |= IB_SIG_CHECK_APPTAG;
+}
+
+static int nvmet_rdma_rw_ctx_init(struct nvmet_rdma_rsp *rsp, u64 addr, u32 key,
+ struct ib_sig_attrs *sig_attrs)
+{
+ struct rdma_cm_id *cm_id = rsp->queue->cm_id;
+ struct nvmet_req *req = &rsp->req;
+ int ret;
+
+ if (req->metadata_len)
+ ret = rdma_rw_ctx_signature_init(&rsp->rw, cm_id->qp,
+ cm_id->port_num, req->sg, req->sg_cnt,
+ req->metadata_sg, req->metadata_sg_cnt, sig_attrs,
+ addr, key, nvmet_data_dir(req));
+ else
+ ret = rdma_rw_ctx_init(&rsp->rw, cm_id->qp, cm_id->port_num,
+ req->sg, req->sg_cnt, 0, addr, key,
+ nvmet_data_dir(req));
+
+ return ret;
+}
+
+static void nvmet_rdma_rw_ctx_destroy(struct nvmet_rdma_rsp *rsp)
+{
+ struct rdma_cm_id *cm_id = rsp->queue->cm_id;
+ struct nvmet_req *req = &rsp->req;
+
+ if (req->metadata_len)
+ rdma_rw_ctx_destroy_signature(&rsp->rw, cm_id->qp,
+ cm_id->port_num, req->sg, req->sg_cnt,
+ req->metadata_sg, req->metadata_sg_cnt,
+ nvmet_data_dir(req));
+ else
+ rdma_rw_ctx_destroy(&rsp->rw, cm_id->qp, cm_id->port_num,
+ req->sg, req->sg_cnt, nvmet_data_dir(req));
+}
static void nvmet_rdma_release_rsp(struct nvmet_rdma_rsp *rsp)
{
@@ -507,14 +668,11 @@ static void nvmet_rdma_release_rsp(struct nvmet_rdma_rsp *rsp)
atomic_add(1 + rsp->n_rdma, &queue->sq_wr_avail);
- if (rsp->n_rdma) {
- rdma_rw_ctx_destroy(&rsp->rw, queue->qp,
- queue->cm_id->port_num, rsp->req.sg,
- rsp->req.sg_cnt, nvmet_data_dir(&rsp->req));
- }
+ if (rsp->n_rdma)
+ nvmet_rdma_rw_ctx_destroy(rsp);
if (rsp->req.sg != rsp->cmd->inline_sg)
- nvmet_req_free_sgl(&rsp->req);
+ nvmet_req_free_sgls(&rsp->req);
if (unlikely(!list_empty_careful(&queue->rsp_wr_wait_list)))
nvmet_rdma_process_wr_wait_list(queue);
@@ -566,11 +724,16 @@ static void nvmet_rdma_queue_response(struct nvmet_req *req)
rsp->send_wr.opcode = IB_WR_SEND;
}
- if (nvmet_rdma_need_data_out(rsp))
- first_wr = rdma_rw_ctx_wrs(&rsp->rw, cm_id->qp,
- cm_id->port_num, NULL, &rsp->send_wr);
- else
+ if (nvmet_rdma_need_data_out(rsp)) {
+ if (rsp->req.metadata_len)
+ first_wr = rdma_rw_ctx_wrs(&rsp->rw, cm_id->qp,
+ cm_id->port_num, &rsp->write_cqe, NULL);
+ else
+ first_wr = rdma_rw_ctx_wrs(&rsp->rw, cm_id->qp,
+ cm_id->port_num, NULL, &rsp->send_wr);
+ } else {
first_wr = &rsp->send_wr;
+ }
nvmet_rdma_post_recv(rsp->queue->dev, rsp->cmd);
@@ -589,15 +752,14 @@ static void nvmet_rdma_read_data_done(struct ib_cq *cq, struct ib_wc *wc)
struct nvmet_rdma_rsp *rsp =
container_of(wc->wr_cqe, struct nvmet_rdma_rsp, read_cqe);
struct nvmet_rdma_queue *queue = cq->cq_context;
+ u16 status = 0;
WARN_ON(rsp->n_rdma <= 0);
atomic_add(rsp->n_rdma, &queue->sq_wr_avail);
- rdma_rw_ctx_destroy(&rsp->rw, queue->qp,
- queue->cm_id->port_num, rsp->req.sg,
- rsp->req.sg_cnt, nvmet_data_dir(&rsp->req));
rsp->n_rdma = 0;
if (unlikely(wc->status != IB_WC_SUCCESS)) {
+ nvmet_rdma_rw_ctx_destroy(rsp);
nvmet_req_uninit(&rsp->req);
nvmet_rdma_release_rsp(rsp);
if (wc->status != IB_WC_WR_FLUSH_ERR) {
@@ -608,7 +770,58 @@ static void nvmet_rdma_read_data_done(struct ib_cq *cq, struct ib_wc *wc)
return;
}
- rsp->req.execute(&rsp->req);
+ if (rsp->req.metadata_len)
+ status = nvmet_rdma_check_pi_status(rsp->rw.reg->mr);
+ nvmet_rdma_rw_ctx_destroy(rsp);
+
+ if (unlikely(status))
+ nvmet_req_complete(&rsp->req, status);
+ else
+ rsp->req.execute(&rsp->req);
+}
+
+static void nvmet_rdma_write_data_done(struct ib_cq *cq, struct ib_wc *wc)
+{
+ struct nvmet_rdma_rsp *rsp =
+ container_of(wc->wr_cqe, struct nvmet_rdma_rsp, write_cqe);
+ struct nvmet_rdma_queue *queue = cq->cq_context;
+ struct rdma_cm_id *cm_id = rsp->queue->cm_id;
+ u16 status;
+
+ if (!IS_ENABLED(CONFIG_BLK_DEV_INTEGRITY))
+ return;
+
+ WARN_ON(rsp->n_rdma <= 0);
+ atomic_add(rsp->n_rdma, &queue->sq_wr_avail);
+ rsp->n_rdma = 0;
+
+ if (unlikely(wc->status != IB_WC_SUCCESS)) {
+ nvmet_rdma_rw_ctx_destroy(rsp);
+ nvmet_req_uninit(&rsp->req);
+ nvmet_rdma_release_rsp(rsp);
+ if (wc->status != IB_WC_WR_FLUSH_ERR) {
+ pr_info("RDMA WRITE for CQE 0x%p failed with status %s (%d).\n",
+ wc->wr_cqe, ib_wc_status_msg(wc->status),
+ wc->status);
+ nvmet_rdma_error_comp(queue);
+ }
+ return;
+ }
+
+ /*
+ * Upon RDMA completion check the signature status
+ * - if succeeded send good NVMe response
+ * - if failed send bad NVMe response with appropriate error
+ */
+ status = nvmet_rdma_check_pi_status(rsp->rw.reg->mr);
+ if (unlikely(status))
+ rsp->req.cqe->status = cpu_to_le16(status << 1);
+ nvmet_rdma_rw_ctx_destroy(rsp);
+
+ if (unlikely(ib_post_send(cm_id->qp, &rsp->send_wr, NULL))) {
+ pr_err("sending cmd response failed\n");
+ nvmet_rdma_release_rsp(rsp);
+ }
}
static void nvmet_rdma_use_inline_sg(struct nvmet_rdma_rsp *rsp, u32 len,
@@ -665,9 +878,9 @@ static u16 nvmet_rdma_map_sgl_inline(struct nvmet_rdma_rsp *rsp)
static u16 nvmet_rdma_map_sgl_keyed(struct nvmet_rdma_rsp *rsp,
struct nvme_keyed_sgl_desc *sgl, bool invalidate)
{
- struct rdma_cm_id *cm_id = rsp->queue->cm_id;
u64 addr = le64_to_cpu(sgl->addr);
u32 key = get_unaligned_le32(sgl->key);
+ struct ib_sig_attrs sig_attrs;
int ret;
rsp->req.transfer_len = get_unaligned_le24(sgl->length);
@@ -676,13 +889,14 @@ static u16 nvmet_rdma_map_sgl_keyed(struct nvmet_rdma_rsp *rsp,
if (!rsp->req.transfer_len)
return 0;
- ret = nvmet_req_alloc_sgl(&rsp->req);
+ if (rsp->req.metadata_len)
+ nvmet_rdma_set_sig_attrs(&rsp->req, &sig_attrs);
+
+ ret = nvmet_req_alloc_sgls(&rsp->req);
if (unlikely(ret < 0))
goto error_out;
- ret = rdma_rw_ctx_init(&rsp->rw, cm_id->qp, cm_id->port_num,
- rsp->req.sg, rsp->req.sg_cnt, 0, addr, key,
- nvmet_data_dir(&rsp->req));
+ ret = nvmet_rdma_rw_ctx_init(rsp, addr, key, &sig_attrs);
if (unlikely(ret < 0))
goto error_out;
rsp->n_rdma += ret;
@@ -845,23 +1059,40 @@ static void nvmet_rdma_recv_done(struct ib_cq *cq, struct ib_wc *wc)
nvmet_rdma_handle_command(queue, rsp);
}
-static void nvmet_rdma_destroy_srq(struct nvmet_rdma_device *ndev)
+static void nvmet_rdma_destroy_srq(struct nvmet_rdma_srq *nsrq)
{
- if (!ndev->srq)
+ nvmet_rdma_free_cmds(nsrq->ndev, nsrq->cmds, nsrq->ndev->srq_size,
+ false);
+ ib_destroy_srq(nsrq->srq);
+
+ kfree(nsrq);
+}
+
+static void nvmet_rdma_destroy_srqs(struct nvmet_rdma_device *ndev)
+{
+ int i;
+
+ if (!ndev->srqs)
return;
- nvmet_rdma_free_cmds(ndev, ndev->srq_cmds, ndev->srq_size, false);
- ib_destroy_srq(ndev->srq);
+ for (i = 0; i < ndev->srq_count; i++)
+ nvmet_rdma_destroy_srq(ndev->srqs[i]);
+
+ kfree(ndev->srqs);
}
-static int nvmet_rdma_init_srq(struct nvmet_rdma_device *ndev)
+static struct nvmet_rdma_srq *
+nvmet_rdma_init_srq(struct nvmet_rdma_device *ndev)
{
struct ib_srq_init_attr srq_attr = { NULL, };
+ size_t srq_size = ndev->srq_size;
+ struct nvmet_rdma_srq *nsrq;
struct ib_srq *srq;
- size_t srq_size;
int ret, i;
- srq_size = 4095; /* XXX: tune */
+ nsrq = kzalloc(sizeof(*nsrq), GFP_KERNEL);
+ if (!nsrq)
+ return ERR_PTR(-ENOMEM);
srq_attr.attr.max_wr = srq_size;
srq_attr.attr.max_sge = 1 + ndev->inline_page_count;
@@ -869,35 +1100,73 @@ static int nvmet_rdma_init_srq(struct nvmet_rdma_device *ndev)
srq_attr.srq_type = IB_SRQT_BASIC;
srq = ib_create_srq(ndev->pd, &srq_attr);
if (IS_ERR(srq)) {
- /*
- * If SRQs aren't supported we just go ahead and use normal
- * non-shared receive queues.
- */
- pr_info("SRQ requested but not supported.\n");
- return 0;
+ ret = PTR_ERR(srq);
+ goto out_free;
}
- ndev->srq_cmds = nvmet_rdma_alloc_cmds(ndev, srq_size, false);
- if (IS_ERR(ndev->srq_cmds)) {
- ret = PTR_ERR(ndev->srq_cmds);
+ nsrq->cmds = nvmet_rdma_alloc_cmds(ndev, srq_size, false);
+ if (IS_ERR(nsrq->cmds)) {
+ ret = PTR_ERR(nsrq->cmds);
goto out_destroy_srq;
}
- ndev->srq = srq;
- ndev->srq_size = srq_size;
+ nsrq->srq = srq;
+ nsrq->ndev = ndev;
for (i = 0; i < srq_size; i++) {
- ret = nvmet_rdma_post_recv(ndev, &ndev->srq_cmds[i]);
+ nsrq->cmds[i].nsrq = nsrq;
+ ret = nvmet_rdma_post_recv(ndev, &nsrq->cmds[i]);
if (ret)
goto out_free_cmds;
}
- return 0;
+ return nsrq;
out_free_cmds:
- nvmet_rdma_free_cmds(ndev, ndev->srq_cmds, ndev->srq_size, false);
+ nvmet_rdma_free_cmds(ndev, nsrq->cmds, srq_size, false);
out_destroy_srq:
ib_destroy_srq(srq);
+out_free:
+ kfree(nsrq);
+ return ERR_PTR(ret);
+}
+
+static int nvmet_rdma_init_srqs(struct nvmet_rdma_device *ndev)
+{
+ int i, ret;
+
+ if (!ndev->device->attrs.max_srq_wr || !ndev->device->attrs.max_srq) {
+ /*
+ * If SRQs aren't supported we just go ahead and use normal
+ * non-shared receive queues.
+ */
+ pr_info("SRQ requested but not supported.\n");
+ return 0;
+ }
+
+ ndev->srq_size = min(ndev->device->attrs.max_srq_wr,
+ nvmet_rdma_srq_size);
+ ndev->srq_count = min(ndev->device->num_comp_vectors,
+ ndev->device->attrs.max_srq);
+
+ ndev->srqs = kcalloc(ndev->srq_count, sizeof(*ndev->srqs), GFP_KERNEL);
+ if (!ndev->srqs)
+ return -ENOMEM;
+
+ for (i = 0; i < ndev->srq_count; i++) {
+ ndev->srqs[i] = nvmet_rdma_init_srq(ndev);
+ if (IS_ERR(ndev->srqs[i])) {
+ ret = PTR_ERR(ndev->srqs[i]);
+ goto err_srq;
+ }
+ }
+
+ return 0;
+
+err_srq:
+ while (--i >= 0)
+ nvmet_rdma_destroy_srq(ndev->srqs[i]);
+ kfree(ndev->srqs);
return ret;
}
@@ -910,7 +1179,7 @@ static void nvmet_rdma_free_dev(struct kref *ref)
list_del(&ndev->entry);
mutex_unlock(&device_list_mutex);
- nvmet_rdma_destroy_srq(ndev);
+ nvmet_rdma_destroy_srqs(ndev);
ib_dealloc_pd(ndev->pd);
kfree(ndev);
@@ -957,7 +1226,7 @@ nvmet_rdma_find_get_device(struct rdma_cm_id *cm_id)
goto out_free_dev;
if (nvmet_rdma_use_srq) {
- ret = nvmet_rdma_init_srq(ndev);
+ ret = nvmet_rdma_init_srqs(ndev);
if (ret)
goto out_free_pd;
}
@@ -981,14 +1250,7 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
{
struct ib_qp_init_attr qp_attr;
struct nvmet_rdma_device *ndev = queue->dev;
- int comp_vector, nr_cqe, ret, i, factor;
-
- /*
- * Spread the io queues across completion vectors,
- * but still keep all admin queues on vector 0.
- */
- comp_vector = !queue->host_qid ? 0 :
- queue->idx % ndev->device->num_comp_vectors;
+ int nr_cqe, ret, i, factor;
/*
* Reserve CQ slots for RECV + RDMA_READ/RDMA_WRITE + RDMA_SEND.
@@ -996,7 +1258,7 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
nr_cqe = queue->recv_queue_size + 2 * queue->send_queue_size;
queue->cq = ib_alloc_cq(ndev->device, queue,
- nr_cqe + 1, comp_vector,
+ nr_cqe + 1, queue->comp_vector,
IB_POLL_WORKQUEUE);
if (IS_ERR(queue->cq)) {
ret = PTR_ERR(queue->cq);
@@ -1020,14 +1282,17 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
qp_attr.cap.max_send_sge = max(ndev->device->attrs.max_sge_rd,
ndev->device->attrs.max_send_sge);
- if (ndev->srq) {
- qp_attr.srq = ndev->srq;
+ if (queue->nsrq) {
+ qp_attr.srq = queue->nsrq->srq;
} else {
/* +1 for drain */
qp_attr.cap.max_recv_wr = 1 + queue->recv_queue_size;
qp_attr.cap.max_recv_sge = 1 + ndev->inline_page_count;
}
+ if (queue->port->pi_enable && queue->host_qid)
+ qp_attr.create_flags |= IB_QP_CREATE_INTEGRITY_EN;
+
ret = rdma_create_qp(queue->cm_id, ndev->pd, &qp_attr);
if (ret) {
pr_err("failed to create_qp ret= %d\n", ret);
@@ -1041,7 +1306,7 @@ static int nvmet_rdma_create_queue_ib(struct nvmet_rdma_queue *queue)
__func__, queue->cq->cqe, qp_attr.cap.max_send_sge,
qp_attr.cap.max_send_wr, queue->cm_id);
- if (!ndev->srq) {
+ if (!queue->nsrq) {
for (i = 0; i < queue->recv_queue_size; i++) {
queue->cmds[i].queue = queue;
ret = nvmet_rdma_post_recv(ndev, &queue->cmds[i]);
@@ -1076,7 +1341,7 @@ static void nvmet_rdma_free_queue(struct nvmet_rdma_queue *queue)
nvmet_sq_destroy(&queue->nvme_sq);
nvmet_rdma_destroy_queue_ib(queue);
- if (!queue->dev->srq) {
+ if (!queue->nsrq) {
nvmet_rdma_free_cmds(queue->dev, queue->cmds,
queue->recv_queue_size,
!queue->host_qid);
@@ -1146,6 +1411,7 @@ nvmet_rdma_alloc_queue(struct nvmet_rdma_device *ndev,
struct rdma_cm_id *cm_id,
struct rdma_cm_event *event)
{
+ struct nvmet_rdma_port *port = cm_id->context;
struct nvmet_rdma_queue *queue;
int ret;
@@ -1172,6 +1438,7 @@ nvmet_rdma_alloc_queue(struct nvmet_rdma_device *ndev,
INIT_WORK(&queue->release_work, nvmet_rdma_release_queue_work);
queue->dev = ndev;
queue->cm_id = cm_id;
+ queue->port = port->nport;
spin_lock_init(&queue->state_lock);
queue->state = NVMET_RDMA_Q_CONNECTING;
@@ -1188,13 +1455,23 @@ nvmet_rdma_alloc_queue(struct nvmet_rdma_device *ndev,
goto out_destroy_sq;
}
+ /*
+ * Spread the io queues across completion vectors,
+ * but still keep all admin queues on vector 0.
+ */
+ queue->comp_vector = !queue->host_qid ? 0 :
+ queue->idx % ndev->device->num_comp_vectors;
+
+
ret = nvmet_rdma_alloc_rsps(queue);
if (ret) {
ret = NVME_RDMA_CM_NO_RSC;
goto out_ida_remove;
}
- if (!ndev->srq) {
+ if (ndev->srqs) {
+ queue->nsrq = ndev->srqs[queue->comp_vector % ndev->srq_count];
+ } else {
queue->cmds = nvmet_rdma_alloc_cmds(ndev,
queue->recv_queue_size,
!queue->host_qid);
@@ -1215,7 +1492,7 @@ nvmet_rdma_alloc_queue(struct nvmet_rdma_device *ndev,
return queue;
out_free_cmds:
- if (!ndev->srq) {
+ if (!queue->nsrq) {
nvmet_rdma_free_cmds(queue->dev, queue->cmds,
queue->recv_queue_size,
!queue->host_qid);
@@ -1241,6 +1518,10 @@ static void nvmet_rdma_qp_event(struct ib_event *event, void *priv)
case IB_EVENT_COMM_EST:
rdma_notify(queue->cm_id, event->event);
break;
+ case IB_EVENT_QP_LAST_WQE_REACHED:
+ pr_debug("received last WQE reached event for queue=0x%p\n",
+ queue);
+ break;
default:
pr_err("received IB QP event: %s (%d)\n",
ib_event_msg(event->event), event->event);
@@ -1275,7 +1556,6 @@ static int nvmet_rdma_cm_accept(struct rdma_cm_id *cm_id,
static int nvmet_rdma_queue_connect(struct rdma_cm_id *cm_id,
struct rdma_cm_event *event)
{
- struct nvmet_rdma_port *port = cm_id->context;
struct nvmet_rdma_device *ndev;
struct nvmet_rdma_queue *queue;
int ret = -EINVAL;
@@ -1291,7 +1571,6 @@ static int nvmet_rdma_queue_connect(struct rdma_cm_id *cm_id,
ret = -ENOMEM;
goto put_device;
}
- queue->port = port->nport;
if (queue->host_qid == 0) {
/* Let inflight controller teardown complete */
@@ -1563,6 +1842,14 @@ static int nvmet_rdma_enable_port(struct nvmet_rdma_port *port)
goto out_destroy_id;
}
+ if (port->nport->pi_enable &&
+ !(cm_id->device->attrs.device_cap_flags &
+ IB_DEVICE_INTEGRITY_HANDOVER)) {
+ pr_err("T10-PI is not supported for %pISpcs\n", addr);
+ ret = -EINVAL;
+ goto out_destroy_id;
+ }
+
port->cm_id = cm_id;
return 0;
@@ -1672,6 +1959,8 @@ static void nvmet_rdma_disc_port_addr(struct nvmet_req *req,
static u8 nvmet_rdma_get_mdts(const struct nvmet_ctrl *ctrl)
{
+ if (ctrl->pi_support)
+ return NVMET_RDMA_MAX_METADATA_MDTS;
return NVMET_RDMA_MAX_MDTS;
}
@@ -1680,6 +1969,7 @@ static const struct nvmet_fabrics_ops nvmet_rdma_ops = {
.type = NVMF_TRTYPE_RDMA,
.msdbd = 1,
.has_keyed_sgls = 1,
+ .metadata_support = 1,
.add_port = nvmet_rdma_add_port,
.remove_port = nvmet_rdma_remove_port,
.queue_response = nvmet_rdma_queue_response,
diff --git a/drivers/nvme/target/tcp.c b/drivers/nvme/target/tcp.c
index f0da04e960f4..6f557db0320d 100644
--- a/drivers/nvme/target/tcp.c
+++ b/drivers/nvme/target/tcp.c
@@ -325,6 +325,14 @@ static void nvmet_tcp_fatal_error(struct nvmet_tcp_queue *queue)
kernel_sock_shutdown(queue->sock, SHUT_RDWR);
}
+static void nvmet_tcp_socket_error(struct nvmet_tcp_queue *queue, int status)
+{
+ if (status == -EPIPE || status == -ECONNRESET)
+ kernel_sock_shutdown(queue->sock, SHUT_RDWR);
+ else
+ nvmet_tcp_fatal_error(queue);
+}
+
static int nvmet_tcp_map_data(struct nvmet_tcp_cmd *cmd)
{
struct nvme_sgl_desc *sgl = &cmd->req.cmd->common.dptr.sgl;
@@ -510,7 +518,7 @@ static int nvmet_try_send_data_pdu(struct nvmet_tcp_cmd *cmd)
ret = kernel_sendpage(cmd->queue->sock, virt_to_page(cmd->data_pdu),
offset_in_page(cmd->data_pdu) + cmd->offset,
- left, MSG_DONTWAIT | MSG_MORE);
+ left, MSG_DONTWAIT | MSG_MORE | MSG_SENDPAGE_NOTLAST);
if (ret <= 0)
return ret;
@@ -538,7 +546,7 @@ static int nvmet_try_send_data(struct nvmet_tcp_cmd *cmd, bool last_in_batch)
if ((!last_in_batch && cmd->queue->send_list_len) ||
cmd->wbytes_done + left < cmd->req.transfer_len ||
queue->data_digest || !queue->nvme_sq.sqhd_disabled)
- flags |= MSG_MORE;
+ flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
ret = kernel_sendpage(cmd->queue->sock, page, cmd->offset,
left, flags);
@@ -585,7 +593,7 @@ static int nvmet_try_send_response(struct nvmet_tcp_cmd *cmd,
int ret;
if (!last_in_batch && cmd->queue->send_list_len)
- flags |= MSG_MORE;
+ flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
else
flags |= MSG_EOR;
@@ -614,7 +622,7 @@ static int nvmet_try_send_r2t(struct nvmet_tcp_cmd *cmd, bool last_in_batch)
int ret;
if (!last_in_batch && cmd->queue->send_list_len)
- flags |= MSG_MORE;
+ flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
else
flags |= MSG_EOR;
@@ -644,6 +652,8 @@ static int nvmet_try_send_ddgst(struct nvmet_tcp_cmd *cmd, bool last_in_batch)
if (!last_in_batch && cmd->queue->send_list_len)
msg.msg_flags |= MSG_MORE;
+ else
+ msg.msg_flags |= MSG_EOR;
ret = kernel_sendmsg(queue->sock, &msg, &iov, 1, iov.iov_len);
if (unlikely(ret <= 0))
@@ -716,11 +726,15 @@ static int nvmet_tcp_try_send(struct nvmet_tcp_queue *queue,
for (i = 0; i < budget; i++) {
ret = nvmet_tcp_try_send_one(queue, i == budget - 1);
- if (ret <= 0)
+ if (unlikely(ret < 0)) {
+ nvmet_tcp_socket_error(queue, ret);
+ goto done;
+ } else if (ret == 0) {
break;
+ }
(*sends)++;
}
-
+done:
return ret;
}
@@ -1157,11 +1171,15 @@ static int nvmet_tcp_try_recv(struct nvmet_tcp_queue *queue,
for (i = 0; i < budget; i++) {
ret = nvmet_tcp_try_recv_one(queue);
- if (ret <= 0)
+ if (unlikely(ret < 0)) {
+ nvmet_tcp_socket_error(queue, ret);
+ goto done;
+ } else if (ret == 0) {
break;
+ }
(*recvs)++;
}
-
+done:
return ret;
}
@@ -1186,27 +1204,16 @@ static void nvmet_tcp_io_work(struct work_struct *w)
pending = false;
ret = nvmet_tcp_try_recv(queue, NVMET_TCP_RECV_BUDGET, &ops);
- if (ret > 0) {
+ if (ret > 0)
pending = true;
- } else if (ret < 0) {
- if (ret == -EPIPE || ret == -ECONNRESET)
- kernel_sock_shutdown(queue->sock, SHUT_RDWR);
- else
- nvmet_tcp_fatal_error(queue);
+ else if (ret < 0)
return;
- }
ret = nvmet_tcp_try_send(queue, NVMET_TCP_SEND_BUDGET, &ops);
- if (ret > 0) {
- /* transmitted message/data */
+ if (ret > 0)
pending = true;
- } else if (ret < 0) {
- if (ret == -EPIPE || ret == -ECONNRESET)
- kernel_sock_shutdown(queue->sock, SHUT_RDWR);
- else
- nvmet_tcp_fatal_error(queue);
+ else if (ret < 0)
return;
- }
} while (pending && ops < NVMET_TCP_IO_WORK_BUDGET);
diff --git a/drivers/nvme/target/trace.h b/drivers/nvme/target/trace.h
index e645caa882dd..0458046d6501 100644
--- a/drivers/nvme/target/trace.h
+++ b/drivers/nvme/target/trace.h
@@ -130,6 +130,34 @@ TRACE_EVENT(nvmet_req_complete,
);
+#define aer_name(aer) { aer, #aer }
+
+TRACE_EVENT(nvmet_async_event,
+ TP_PROTO(struct nvmet_ctrl *ctrl, __le32 result),
+ TP_ARGS(ctrl, result),
+ TP_STRUCT__entry(
+ __field(int, ctrl_id)
+ __field(u32, result)
+ ),
+ TP_fast_assign(
+ __entry->ctrl_id = ctrl->cntlid;
+ __entry->result = (le32_to_cpu(result) & 0xff00) >> 8;
+ ),
+ TP_printk("nvmet%d: NVME_AEN=%#08x [%s]",
+ __entry->ctrl_id, __entry->result,
+ __print_symbolic(__entry->result,
+ aer_name(NVME_AER_NOTICE_NS_CHANGED),
+ aer_name(NVME_AER_NOTICE_ANA),
+ aer_name(NVME_AER_NOTICE_FW_ACT_STARTING),
+ aer_name(NVME_AER_NOTICE_DISC_CHANGED),
+ aer_name(NVME_AER_ERROR),
+ aer_name(NVME_AER_SMART),
+ aer_name(NVME_AER_CSS),
+ aer_name(NVME_AER_VS))
+ )
+);
+#undef aer_name
+
#endif /* _TRACE_NVMET_H */
#undef TRACE_INCLUDE_PATH