nvme: add function to reconnect qpairs.

While it is unlikely that a single qpair will be failed, it is important
to make it possible to reconnect a single qpair.

This function is also handy at the application layer when going through
a reconnect workflow. If we get -ENXIO from a qpair when we poll, we
will turn around and call this function. If we get -ENXIO from this
function, then we know the whole controller is failed and we need to do
a reset.

Change-Id: I6a8ea0ce27fce2f5fc0a5b3db05834acd68e6a39
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/471417
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>

include/spdk/nvme.h

@@ -1041,6 +1041,25 @@ struct spdk_nvme_qpair *spdk_nvme_ctrlr_alloc_io_qpair(struct spdk_nvme_ctrlr *c
 		const struct spdk_nvme_io_qpair_opts *opts,
 		size_t opts_size);
 
+/**
+ * Attempt to reconnect the given qpair.
+ *
+ * This function is intended to be called on qpairs that have already been connected,
+ * but have since entered a failed state as indicated by a return value of -ENXIO from
+ * either spdk_nvme_qpair_process_completions or one of the spdk_nvme_ns_cmd_* functions.
+ *
+ * \param qpair The qpair to reconnect.
+ *
+ * \return 0 on success, or if the qpair was already connected.
+ * -EAGAIN if the driver was unable to reconnect during this call,
+ * but the controller is still connected and is either resetting or enabled.
+ * -ENODEV if the controller is removed. In this case, the controller cannot be recovered
+ * and the application will have to destroy it and the associated qpairs.
+ * -ENXIO if the controller is in a failed state but is not yet resetting. In this case,
+ * the application should call spdk_nvme_ctrlr_reset to reset the entire controller.
+ */
+int spdk_nvme_ctrlr_reconnect_io_qpair(struct spdk_nvme_qpair *qpair);
+
 /**
  * Free an I/O queue pair that was allocated by spdk_nvme_ctrlr_alloc_io_qpair().
  *

lib/nvme/nvme_ctrlr.c

@@ -379,6 +379,52 @@ spdk_nvme_ctrlr_alloc_io_qpair(struct spdk_nvme_ctrlr *ctrlr,
 	return qpair;
 }
 
+int
+spdk_nvme_ctrlr_reconnect_io_qpair(struct spdk_nvme_qpair *qpair)
+{
+	struct spdk_nvme_ctrlr *ctrlr;
+	int rc;
+
+	assert(qpair != NULL);
+	assert(nvme_qpair_is_admin_queue(qpair) == false);
+	assert(qpair->ctrlr != NULL);
+
+	ctrlr = qpair->ctrlr;
+	nvme_robust_mutex_lock(&ctrlr->ctrlr_lock);
+
+	if (ctrlr->is_removed) {
+		rc = -ENODEV;
+		goto out;
+	}
+
+	if (ctrlr->is_resetting) {
+		rc = -EAGAIN;
+		goto out;
+	}
+
+	if (ctrlr->is_failed) {
+		rc = -ENXIO;
+		goto out;
+	}
+
+	if (!qpair->transport_qp_is_failed) {
+		rc = 0;
+		goto out;
+	}
+
+	rc = nvme_transport_ctrlr_connect_qpair(ctrlr, qpair);
+	if (rc) {
+		nvme_qpair_set_state(qpair, NVME_QPAIR_DISABLED);
+		rc = -EAGAIN;
+		goto out;
+	}
+	nvme_qpair_set_state(qpair, NVME_QPAIR_CONNECTED);
+
+out:
+	nvme_robust_mutex_unlock(&ctrlr->ctrlr_lock);
+	return rc;
+}
+
 int
 spdk_nvme_ctrlr_free_io_qpair(struct spdk_nvme_qpair *qpair)
 {

test/unit/lib/nvme/nvme_ctrlr.c/nvme_ctrlr_ut.c

@@ -168,12 +168,6 @@ nvme_transport_ctrlr_delete_io_qpair(struct spdk_nvme_ctrlr *ctrlr, struct spdk_
 	return 0;
 }
 
-int
-nvme_transport_ctrlr_connect_qpair(struct spdk_nvme_ctrlr *ctrlr, struct spdk_nvme_qpair *qpair)
-{
-	return 0;
-}
-
 void
 nvme_transport_ctrlr_disconnect_qpair(struct spdk_nvme_ctrlr *ctrlr, struct spdk_nvme_qpair *qpair)
 {
@@ -1424,6 +1418,69 @@ test_alloc_io_qpair_wrr_2(void)
 	cleanup_qpairs(&ctrlr);
 }
 
+bool g_connect_qpair_called = false;
+int g_connect_qpair_return_code = 0;
+int nvme_transport_ctrlr_connect_qpair(struct spdk_nvme_ctrlr *ctrlr, struct spdk_nvme_qpair *qpair)
+{
+	g_connect_qpair_called = true;
+	return g_connect_qpair_return_code;
+}
+
+static void
+test_spdk_nvme_ctrlr_reconnect_io_qpair(void)
+{
+	struct spdk_nvme_ctrlr	ctrlr = {};
+	struct spdk_nvme_qpair	qpair = {};
+	int rc;
+
+	/* Various states of controller disconnect. */
+	qpair.id = 1;
+	qpair.ctrlr = &ctrlr;
+	ctrlr.is_removed = 1;
+	ctrlr.is_failed = 0;
+	ctrlr.is_resetting = 0;
+	rc = spdk_nvme_ctrlr_reconnect_io_qpair(&qpair);
+	CU_ASSERT(rc == -ENODEV)
+
+	ctrlr.is_removed = 0;
+	ctrlr.is_failed = 1;
+	rc = spdk_nvme_ctrlr_reconnect_io_qpair(&qpair);
+	CU_ASSERT(rc == -ENXIO)
+
+	ctrlr.is_failed = 0;
+	ctrlr.is_resetting = 1;
+	rc = spdk_nvme_ctrlr_reconnect_io_qpair(&qpair);
+	CU_ASSERT(rc == -EAGAIN)
+
+	/* Confirm precedence for controller states: removed > resetting > failed */
+	ctrlr.is_removed = 1;
+	ctrlr.is_failed = 1;
+	rc = spdk_nvme_ctrlr_reconnect_io_qpair(&qpair);
+	CU_ASSERT(rc == -ENODEV)
+
+	ctrlr.is_removed = 0;
+	rc = spdk_nvme_ctrlr_reconnect_io_qpair(&qpair);
+	CU_ASSERT(rc == -EAGAIN)
+
+	ctrlr.is_resetting = 0;
+	rc = spdk_nvme_ctrlr_reconnect_io_qpair(&qpair);
+	CU_ASSERT(rc == -ENXIO)
+
+	/* qpair not failed. Make sure we don't call down to the transport */
+	ctrlr.is_failed = 0;
+	qpair.transport_qp_is_failed = false;
+	g_connect_qpair_called = false;
+	rc = spdk_nvme_ctrlr_reconnect_io_qpair(&qpair);
+	CU_ASSERT(g_connect_qpair_called == false);
+	CU_ASSERT(rc == 0)
+
+	/* transport qpair is failed. make sure we call down to the transport */
+	qpair.transport_qp_is_failed = true;
+	rc = spdk_nvme_ctrlr_reconnect_io_qpair(&qpair);
+	CU_ASSERT(g_connect_qpair_called == true);
+	CU_ASSERT(rc == 0)
+}
+
 static void
 test_nvme_ctrlr_fail(void)
 {
@@ -1885,6 +1942,8 @@ int main(int argc, char **argv)
 #endif
 		|| CU_add_test(suite, "test nvme ctrlr function test_nvme_ctrlr_test_active_ns",
 			       test_nvme_ctrlr_test_active_ns) == NULL
+		|| CU_add_test(suite, "test_spdk_nvme_ctrlr_reconnect_io_qpair",
+			       test_spdk_nvme_ctrlr_reconnect_io_qpair) == NULL
 	) {
 		CU_cleanup_registry();
 		return CU_get_error();