diff options
author | Or Gerlitz <ogerlitz@voltaire.com> | 2010-05-05 17:31:44 +0300 |
---|---|---|
committer | Roland Dreier <rolandd@cisco.com> | 2010-05-12 09:30:44 -0700 |
commit | 39ff05dbbbdb082bbabf06206c56b3cd4ef73904 (patch) | |
tree | 85466e1e75d632b33a294dea436fad2f3233fe52 /drivers/infiniband/ulp/iser/iser_verbs.c | |
parent | d265b9808272c9f25e1c36d3fb5ddb466efd90e9 (diff) |
IB/iser: Enhance disconnection logic for multi-pathing
The iser connection teardown flow isn't over until the underlying
Connection Manager (e.g the IB CM) delivers a disconnected or timeout
event through the RDMA-CM. When the remote (target) side isn't
reachable, e.g when some HW e.g port/hca/switch isn't functioning or
taken down administratively, the CM timeout flow is used and the event
may be generated only after relatively long time -- on the order of
tens of seconds.
The current iser code exposes this possibly long delay to higher
layers, specifically to the iscsid daemon and iscsi kernel stack. As a
result, the iscsi stack doesn't respond well: this low-level CM delay
is added to the fail-over time under HA schemes such as the one
provided by DM multipath through the multipathd(8) service.
This patch enhances the reference counting scheme on iser's IB
connections so that the disconnect flow initiated by iscsid from user
space (ep_disconnect) doesn't wait for the CM to deliver the
disconnect/timeout event. (The connection teardown isn't done from
iser's view point until the event is delivered)
The iser ib (rdma) connection object is destroyed when its reference
count reaches zero. When this happens on the RDMA-CM callback
context, extra care is taken so that the RDMA-CM does the actual
destroying of the associated ID, since doing it in the callback is
prohibited.
The reference count of iser ib connection normally reaches three,
where the <ref, deref> relations are
1. conn <init, terminate>
2. conn <bind, stop/destroy>
3. cma id <create, disconnect/error/timeout callbacks>
With this patch, multipath fail-over time is about 30 seconds, while
without this patch, multipath fail-over time is about 130 seconds.
Signed-off-by: Or Gerlitz <ogerlitz@voltaire.com>
Signed-off-by: Roland Dreier <rolandd@cisco.com>
Diffstat (limited to 'drivers/infiniband/ulp/iser/iser_verbs.c')
-rw-r--r-- | drivers/infiniband/ulp/iser/iser_verbs.c | 72 |
1 files changed, 40 insertions, 32 deletions
diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index ed7c9013541..78fdecacea3 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -238,7 +238,7 @@ alloc_err: * releases the FMR pool, QP and CMA ID objects, returns 0 on success, * -1 on failure */ -static int iser_free_ib_conn_res(struct iser_conn *ib_conn) +static int iser_free_ib_conn_res(struct iser_conn *ib_conn, int can_destroy_id) { BUG_ON(ib_conn == NULL); @@ -253,7 +253,8 @@ static int iser_free_ib_conn_res(struct iser_conn *ib_conn) if (ib_conn->qp != NULL) rdma_destroy_qp(ib_conn->cma_id); - if (ib_conn->cma_id != NULL) + /* if cma handler context, the caller acts s.t the cma destroy the id */ + if (ib_conn->cma_id != NULL && can_destroy_id) rdma_destroy_id(ib_conn->cma_id); ib_conn->fmr_pool = NULL; @@ -331,7 +332,7 @@ static int iser_conn_state_comp_exch(struct iser_conn *ib_conn, /** * Frees all conn objects and deallocs conn descriptor */ -static void iser_conn_release(struct iser_conn *ib_conn) +static void iser_conn_release(struct iser_conn *ib_conn, int can_destroy_id) { struct iser_device *device = ib_conn->device; @@ -341,7 +342,7 @@ static void iser_conn_release(struct iser_conn *ib_conn) list_del(&ib_conn->conn_list); mutex_unlock(&ig.connlist_mutex); iser_free_rx_descriptors(ib_conn); - iser_free_ib_conn_res(ib_conn); + iser_free_ib_conn_res(ib_conn, can_destroy_id); ib_conn->device = NULL; /* on EVENT_ADDR_ERROR there's no device yet for this conn */ if (device != NULL) @@ -354,10 +355,13 @@ void iser_conn_get(struct iser_conn *ib_conn) atomic_inc(&ib_conn->refcount); } -void iser_conn_put(struct iser_conn *ib_conn) +int iser_conn_put(struct iser_conn *ib_conn, int can_destroy_id) { - if (atomic_dec_and_test(&ib_conn->refcount)) - iser_conn_release(ib_conn); + if (atomic_dec_and_test(&ib_conn->refcount)) { + iser_conn_release(ib_conn, can_destroy_id); + return 1; + } + return 0; } /** @@ -381,19 +385,20 @@ void iser_conn_terminate(struct iser_conn *ib_conn) wait_event_interruptible(ib_conn->wait, ib_conn->state == ISER_CONN_DOWN); - iser_conn_put(ib_conn); + iser_conn_put(ib_conn, 1); /* deref ib conn deallocate */ } -static void iser_connect_error(struct rdma_cm_id *cma_id) +static int iser_connect_error(struct rdma_cm_id *cma_id) { struct iser_conn *ib_conn; ib_conn = (struct iser_conn *)cma_id->context; ib_conn->state = ISER_CONN_DOWN; wake_up_interruptible(&ib_conn->wait); + return iser_conn_put(ib_conn, 0); /* deref ib conn's cma id */ } -static void iser_addr_handler(struct rdma_cm_id *cma_id) +static int iser_addr_handler(struct rdma_cm_id *cma_id) { struct iser_device *device; struct iser_conn *ib_conn; @@ -402,8 +407,7 @@ static void iser_addr_handler(struct rdma_cm_id *cma_id) device = iser_device_find_by_ib_device(cma_id); if (!device) { iser_err("device lookup/creation failed\n"); - iser_connect_error(cma_id); - return; + return iser_connect_error(cma_id); } ib_conn = (struct iser_conn *)cma_id->context; @@ -412,11 +416,13 @@ static void iser_addr_handler(struct rdma_cm_id *cma_id) ret = rdma_resolve_route(cma_id, 1000); if (ret) { iser_err("resolve route failed: %d\n", ret); - iser_connect_error(cma_id); + return iser_connect_error(cma_id); } + + return 0; } -static void iser_route_handler(struct rdma_cm_id *cma_id) +static int iser_route_handler(struct rdma_cm_id *cma_id) { struct rdma_conn_param conn_param; int ret; @@ -437,9 +443,9 @@ static void iser_route_handler(struct rdma_cm_id *cma_id) goto failure; } - return; + return 0; failure: - iser_connect_error(cma_id); + return iser_connect_error(cma_id); } static void iser_connected_handler(struct rdma_cm_id *cma_id) @@ -451,12 +457,12 @@ static void iser_connected_handler(struct rdma_cm_id *cma_id) wake_up_interruptible(&ib_conn->wait); } -static void iser_disconnected_handler(struct rdma_cm_id *cma_id) +static int iser_disconnected_handler(struct rdma_cm_id *cma_id) { struct iser_conn *ib_conn; + int ret; ib_conn = (struct iser_conn *)cma_id->context; - ib_conn->disc_evt_flag = 1; /* getting here when the state is UP means that the conn is being * * terminated asynchronously from the iSCSI layer's perspective. */ @@ -471,20 +477,24 @@ static void iser_disconnected_handler(struct rdma_cm_id *cma_id) ib_conn->state = ISER_CONN_DOWN; wake_up_interruptible(&ib_conn->wait); } + + ret = iser_conn_put(ib_conn, 0); /* deref ib conn's cma id */ + return ret; } static int iser_cma_handler(struct rdma_cm_id *cma_id, struct rdma_cm_event *event) { int ret = 0; - iser_err("event %d conn %p id %p\n",event->event,cma_id->context,cma_id); + iser_err("event %d status %d conn %p id %p\n", + event->event, event->status, cma_id->context, cma_id); switch (event->event) { case RDMA_CM_EVENT_ADDR_RESOLVED: - iser_addr_handler(cma_id); + ret = iser_addr_handler(cma_id); break; case RDMA_CM_EVENT_ROUTE_RESOLVED: - iser_route_handler(cma_id); + ret = iser_route_handler(cma_id); break; case RDMA_CM_EVENT_ESTABLISHED: iser_connected_handler(cma_id); @@ -494,13 +504,12 @@ static int iser_cma_handler(struct rdma_cm_id *cma_id, struct rdma_cm_event *eve case RDMA_CM_EVENT_CONNECT_ERROR: case RDMA_CM_EVENT_UNREACHABLE: case RDMA_CM_EVENT_REJECTED: - iser_err("event: %d, error: %d\n", event->event, event->status); - iser_connect_error(cma_id); + ret = iser_connect_error(cma_id); break; case RDMA_CM_EVENT_DISCONNECTED: case RDMA_CM_EVENT_DEVICE_REMOVAL: case RDMA_CM_EVENT_ADDR_CHANGE: - iser_disconnected_handler(cma_id); + ret = iser_disconnected_handler(cma_id); break; default: iser_err("Unexpected RDMA CM event (%d)\n", event->event); @@ -515,7 +524,7 @@ void iser_conn_init(struct iser_conn *ib_conn) init_waitqueue_head(&ib_conn->wait); ib_conn->post_recv_buf_count = 0; atomic_set(&ib_conn->post_send_buf_count, 0); - atomic_set(&ib_conn->refcount, 1); + atomic_set(&ib_conn->refcount, 1); /* ref ib conn allocation */ INIT_LIST_HEAD(&ib_conn->conn_list); spin_lock_init(&ib_conn->lock); } @@ -543,6 +552,7 @@ int iser_connect(struct iser_conn *ib_conn, ib_conn->state = ISER_CONN_PENDING; + iser_conn_get(ib_conn); /* ref ib conn's cma id */ ib_conn->cma_id = rdma_create_id(iser_cma_handler, (void *)ib_conn, RDMA_PS_TCP); @@ -580,7 +590,7 @@ id_failure: addr_failure: ib_conn->state = ISER_CONN_DOWN; connect_failure: - iser_conn_release(ib_conn); + iser_conn_release(ib_conn, 1); return err; } @@ -749,12 +759,10 @@ static void iser_handle_comp_error(struct iser_tx_desc *desc, iscsi_conn_failure(ib_conn->iser_conn->iscsi_conn, ISCSI_ERR_CONN_FAILED); - /* complete the termination process if disconnect event was delivered * - * note there are no more non completed posts to the QP */ - if (ib_conn->disc_evt_flag) { - ib_conn->state = ISER_CONN_DOWN; - wake_up_interruptible(&ib_conn->wait); - } + /* no more non completed posts to the QP, complete the + * termination process w.o worrying on disconnect event */ + ib_conn->state = ISER_CONN_DOWN; + wake_up_interruptible(&ib_conn->wait); } } |