[PATCH v8 06/20] dlb: add domain software reset

From: Mike Ximing Chen
Date: Mon Jan 04 2021 - 21:59:14 EST


Add operation to reset a domain's resource's software state when its
reference count reaches zero, and re-inserts those resources in their
respective available-resources linked lists, for use by future scheduling
domains.

Signed-off-by: Gage Eads <gage.eads@xxxxxxxxx>
Signed-off-by: Mike Ximing Chen <mike.ximing.chen@xxxxxxxxx>
Reviewed-by: Magnus Karlsson <magnus.karlsson@xxxxxxxxx>
Reviewed-by: Dan Williams <dan.j.williams@xxxxxxxxx>
---
drivers/misc/dlb/dlb_bitmap.h | 28 ++++
drivers/misc/dlb/dlb_hw_types.h | 6 +
drivers/misc/dlb/dlb_ioctl.c | 10 +-
drivers/misc/dlb/dlb_main.c | 10 +-
drivers/misc/dlb/dlb_main.h | 2 +
drivers/misc/dlb/dlb_pf_ops.c | 7 +
drivers/misc/dlb/dlb_resource.c | 222 ++++++++++++++++++++++++++++++++
drivers/misc/dlb/dlb_resource.h | 5 +
include/uapi/linux/dlb.h | 1 +
9 files changed, 289 insertions(+), 2 deletions(-)

diff --git a/drivers/misc/dlb/dlb_bitmap.h b/drivers/misc/dlb/dlb_bitmap.h
index 5cebf833fab4..332135689dd9 100644
--- a/drivers/misc/dlb/dlb_bitmap.h
+++ b/drivers/misc/dlb/dlb_bitmap.h
@@ -73,6 +73,34 @@ static inline void dlb_bitmap_free(struct dlb_bitmap *bitmap)
kfree(bitmap);
}

+/**
+ * dlb_bitmap_set_range() - set a range of bitmap entries
+ * @bitmap: pointer to dlb_bitmap structure.
+ * @bit: starting bit index.
+ * @len: length of the range.
+ *
+ * Return:
+ * Returns 0 upon success, < 0 otherwise.
+ *
+ * Errors:
+ * EINVAL - bitmap is NULL or is uninitialized, or the range exceeds the bitmap
+ * length.
+ */
+static inline int dlb_bitmap_set_range(struct dlb_bitmap *bitmap,
+ unsigned int bit,
+ unsigned int len)
+{
+ if (!bitmap || !bitmap->map)
+ return -EINVAL;
+
+ if (bitmap->len <= bit)
+ return -EINVAL;
+
+ bitmap_set(bitmap->map, bit, len);
+
+ return 0;
+}
+
/**
* dlb_bitmap_clear_range() - clear a range of bitmap entries
* @bitmap: pointer to dlb_bitmap structure.
diff --git a/drivers/misc/dlb/dlb_hw_types.h b/drivers/misc/dlb/dlb_hw_types.h
index a791f6c98a50..766e471dbcb7 100644
--- a/drivers/misc/dlb/dlb_hw_types.h
+++ b/drivers/misc/dlb/dlb_hw_types.h
@@ -152,6 +152,12 @@ struct dlb_sn_group {
u32 id;
};

+static inline void
+dlb_sn_group_free_slot(struct dlb_sn_group *group, int slot)
+{
+ group->slot_use_bitmap &= ~(BIT(slot));
+}
+
struct dlb_hw_domain {
struct dlb_function_resources *parent_func;
struct list_head func_list;
diff --git a/drivers/misc/dlb/dlb_ioctl.c b/drivers/misc/dlb/dlb_ioctl.c
index 984f0bf89a9f..674c352b50ec 100644
--- a/drivers/misc/dlb/dlb_ioctl.c
+++ b/drivers/misc/dlb/dlb_ioctl.c
@@ -41,13 +41,21 @@ static int dlb_ioctl_create_sched_domain(struct dlb *dlb,
BUILD_BUG_ON(offsetof(typeof(*arg), response) != 0);
resp = &arg->response;

+ if (dlb->domain_reset_failed) {
+ resp->status = DLB_ST_DOMAIN_RESET_FAILED;
+ ret = -EINVAL;
+ goto unlock;
+ }
+
ret = dlb->ops->create_sched_domain(&dlb->hw, arg, resp);
if (ret)
goto unlock;

ret = dlb_init_domain(dlb, resp->id);
- if (ret)
+ if (ret) {
+ dlb->ops->reset_domain(&dlb->hw, resp->id);
goto unlock;
+ }

domain = dlb->sched_domains[resp->id];

diff --git a/drivers/misc/dlb/dlb_main.c b/drivers/misc/dlb/dlb_main.c
index 52cbf3bb7a95..05cc02ff9e4a 100644
--- a/drivers/misc/dlb/dlb_main.c
+++ b/drivers/misc/dlb/dlb_main.c
@@ -127,12 +127,20 @@ int dlb_init_domain(struct dlb *dlb, u32 domain_id)
static int __dlb_free_domain(struct dlb_domain *domain)
{
struct dlb *dlb = domain->dlb;
+ int ret;
+
+ ret = dlb->ops->reset_domain(&dlb->hw, domain->id);
+ if (ret) {
+ dlb->domain_reset_failed = true;
+ dev_err(dlb->dev,
+ "Internal error: Domain reset failed. To recover, reset the device.\n");
+ }

dlb->sched_domains[domain->id] = NULL;

kfree(domain);

- return 0;
+ return ret;
}

void dlb_free_domain(struct kref *kref)
diff --git a/drivers/misc/dlb/dlb_main.h b/drivers/misc/dlb/dlb_main.h
index 4ec04c903f67..5f3c671e466f 100644
--- a/drivers/misc/dlb/dlb_main.h
+++ b/drivers/misc/dlb/dlb_main.h
@@ -44,6 +44,7 @@ struct dlb_device_ops {
struct dlb_cmd_response *resp);
int (*get_num_resources)(struct dlb_hw *hw,
struct dlb_get_num_resources_args *args);
+ int (*reset_domain)(struct dlb_hw *hw, u32 domain_id);
};

extern struct dlb_device_ops dlb_pf_ops;
@@ -70,6 +71,7 @@ struct dlb {
enum dlb_device_type type;
int id;
dev_t dev_number;
+ u8 domain_reset_failed;
};

int dlb_init_domain(struct dlb *dlb, u32 domain_id);
diff --git a/drivers/misc/dlb/dlb_pf_ops.c b/drivers/misc/dlb/dlb_pf_ops.c
index 25ae8f1228d8..7822290ce8a6 100644
--- a/drivers/misc/dlb/dlb_pf_ops.c
+++ b/drivers/misc/dlb/dlb_pf_ops.c
@@ -119,6 +119,12 @@ dlb_pf_get_num_resources(struct dlb_hw *hw,
return dlb_hw_get_num_resources(hw, args, false, 0);
}

+static int
+dlb_pf_reset_domain(struct dlb_hw *hw, u32 id)
+{
+ return dlb_reset_domain(hw, id, false, 0);
+}
+
/********************************/
/****** DLB PF Device Ops ******/
/********************************/
@@ -131,4 +137,5 @@ struct dlb_device_ops dlb_pf_ops = {
.wait_for_device_ready = dlb_pf_wait_for_device_ready,
.create_sched_domain = dlb_pf_create_sched_domain,
.get_num_resources = dlb_pf_get_num_resources,
+ .reset_domain = dlb_pf_reset_domain,
};
diff --git a/drivers/misc/dlb/dlb_resource.c b/drivers/misc/dlb/dlb_resource.c
index e6cb48427b2f..a47e492b8af2 100644
--- a/drivers/misc/dlb/dlb_resource.c
+++ b/drivers/misc/dlb/dlb_resource.c
@@ -201,6 +201,30 @@ int dlb_resource_init(struct dlb_hw *hw)
return ret;
}

+static struct dlb_hw_domain *dlb_get_domain_from_id(struct dlb_hw *hw,
+ u32 id,
+ bool vdev_req,
+ unsigned int vdev_id)
+{
+ struct dlb_function_resources *rsrcs;
+ struct dlb_hw_domain *domain;
+
+ if (id >= DLB_MAX_NUM_DOMAINS)
+ return NULL;
+
+ if (!vdev_req)
+ return &hw->domains[id];
+
+ rsrcs = &hw->vdev[vdev_id];
+
+ list_for_each_entry(domain, &rsrcs->used_domains, func_list) {
+ if (domain->id.virt_id == id)
+ return domain;
+ }
+
+ return NULL;
+}
+
static int dlb_attach_ldb_queues(struct dlb_hw *hw,
struct dlb_function_resources *rsrcs,
struct dlb_hw_domain *domain,
@@ -850,6 +874,204 @@ int dlb_hw_create_sched_domain(struct dlb_hw *hw,
return 0;
}

+static int dlb_domain_reset_software_state(struct dlb_hw *hw,
+ struct dlb_hw_domain *domain)
+{
+ struct dlb *dlb = container_of(hw, struct dlb, hw);
+ struct dlb_dir_pq_pair *tmp_dir_port;
+ struct dlb_function_resources *rsrcs;
+ struct dlb_ldb_queue *tmp_ldb_queue;
+ struct dlb_ldb_port *tmp_ldb_port;
+ struct dlb_dir_pq_pair *dir_port;
+ struct dlb_ldb_queue *ldb_queue;
+ struct dlb_ldb_port *ldb_port;
+ int ret, i;
+
+ lockdep_assert_held(&dlb->resource_mutex);
+
+ rsrcs = domain->parent_func;
+
+ /* Move the domain's ldb queues to the function's avail list */
+ list_for_each_entry_safe(ldb_queue, tmp_ldb_queue,
+ &domain->used_ldb_queues, domain_list) {
+ if (ldb_queue->sn_cfg_valid) {
+ struct dlb_sn_group *grp;
+
+ grp = &hw->rsrcs.sn_groups[ldb_queue->sn_group];
+
+ dlb_sn_group_free_slot(grp, ldb_queue->sn_slot);
+ ldb_queue->sn_cfg_valid = false;
+ }
+
+ ldb_queue->owned = false;
+ ldb_queue->num_mappings = 0;
+ ldb_queue->num_pending_additions = 0;
+
+ list_del(&ldb_queue->domain_list);
+ list_add(&ldb_queue->func_list, &rsrcs->avail_ldb_queues);
+ rsrcs->num_avail_ldb_queues++;
+ }
+
+ list_for_each_entry_safe(ldb_queue, tmp_ldb_queue,
+ &domain->avail_ldb_queues, domain_list) {
+ ldb_queue->owned = false;
+
+ list_del(&ldb_queue->domain_list);
+ list_add(&ldb_queue->func_list, &rsrcs->avail_ldb_queues);
+ rsrcs->num_avail_ldb_queues++;
+ }
+
+ /* Move the domain's ldb ports to the function's avail list */
+ for (i = 0; i < DLB_NUM_COS_DOMAINS; i++) {
+ list_for_each_entry_safe(ldb_port, tmp_ldb_port,
+ &domain->used_ldb_ports[i], domain_list) {
+ int j;
+
+ ldb_port->owned = false;
+ ldb_port->configured = false;
+ ldb_port->num_pending_removals = 0;
+ ldb_port->num_mappings = 0;
+ ldb_port->init_tkn_cnt = 0;
+ for (j = 0; j < DLB_MAX_NUM_QIDS_PER_LDB_CQ; j++)
+ ldb_port->qid_map[j].state =
+ DLB_QUEUE_UNMAPPED;
+
+ list_del(&ldb_port->domain_list);
+ list_add(&ldb_port->func_list,
+ &rsrcs->avail_ldb_ports[i]);
+ rsrcs->num_avail_ldb_ports[i]++;
+ }
+
+ list_for_each_entry_safe(ldb_port, tmp_ldb_port,
+ &domain->avail_ldb_ports[i], domain_list) {
+ ldb_port->owned = false;
+
+ list_del(&ldb_port->domain_list);
+ list_add(&ldb_port->func_list,
+ &rsrcs->avail_ldb_ports[i]);
+ rsrcs->num_avail_ldb_ports[i]++;
+ }
+ }
+
+ /* Move the domain's dir ports to the function's avail list */
+ list_for_each_entry_safe(dir_port, tmp_dir_port,
+ &domain->used_dir_pq_pairs, domain_list) {
+ dir_port->owned = false;
+ dir_port->port_configured = false;
+ dir_port->init_tkn_cnt = 0;
+
+ list_del(&dir_port->domain_list);
+
+ list_add(&dir_port->func_list, &rsrcs->avail_dir_pq_pairs);
+ rsrcs->num_avail_dir_pq_pairs++;
+ }
+
+ list_for_each_entry_safe(dir_port, tmp_dir_port,
+ &domain->avail_dir_pq_pairs, domain_list) {
+ dir_port->owned = false;
+
+ list_del(&dir_port->domain_list);
+
+ list_add(&dir_port->func_list, &rsrcs->avail_dir_pq_pairs);
+ rsrcs->num_avail_dir_pq_pairs++;
+ }
+
+ /* Return hist list entries to the function */
+ ret = dlb_bitmap_set_range(rsrcs->avail_hist_list_entries,
+ domain->hist_list_entry_base,
+ domain->total_hist_list_entries);
+ if (ret) {
+ DLB_HW_ERR(hw,
+ "[%s()] Internal error: domain hist list base doesn't match the function's bitmap.\n",
+ __func__);
+ return ret;
+ }
+
+ domain->total_hist_list_entries = 0;
+ domain->avail_hist_list_entries = 0;
+ domain->hist_list_entry_base = 0;
+ domain->hist_list_entry_offset = 0;
+
+ rsrcs->num_avail_qed_entries += domain->num_ldb_credits;
+ domain->num_ldb_credits = 0;
+
+ rsrcs->num_avail_dqed_entries += domain->num_dir_credits;
+ domain->num_dir_credits = 0;
+
+ rsrcs->num_avail_aqed_entries += domain->num_avail_aqed_entries;
+ rsrcs->num_avail_aqed_entries += domain->num_used_aqed_entries;
+ domain->num_avail_aqed_entries = 0;
+ domain->num_used_aqed_entries = 0;
+
+ domain->num_pending_removals = 0;
+ domain->num_pending_additions = 0;
+ domain->configured = false;
+ domain->started = false;
+
+ /*
+ * Move the domain out of the used_domains list and back to the
+ * function's avail_domains list.
+ */
+ list_del(&domain->func_list);
+ list_add(&domain->func_list, &rsrcs->avail_domains);
+ rsrcs->num_avail_domains++;
+
+ return 0;
+}
+
+static void dlb_log_reset_domain(struct dlb_hw *hw,
+ u32 domain_id,
+ bool vdev_req,
+ unsigned int vdev_id)
+{
+ DLB_HW_DBG(hw, "DLB reset domain:\n");
+ if (vdev_req)
+ DLB_HW_DBG(hw, "(Request from vdev %d)\n", vdev_id);
+ DLB_HW_DBG(hw, "\tDomain ID: %d\n", domain_id);
+}
+
+/**
+ * dlb_reset_domain() - reset a scheduling domain
+ * @hw: dlb_hw handle for a particular device.
+ * @domain_id: domain ID.
+ * @vdev_req: indicates whether this request came from a vdev.
+ * @vdev_id: If vdev_req is true, this contains the vdev's ID.
+ *
+ * This function resets and frees a DLB 2.0 scheduling domain and its associated
+ * resources.
+ *
+ * Pre-condition: the driver must ensure software has stopped sending QEs
+ * through this domain's producer ports before invoking this function, or
+ * undefined behavior will result.
+ *
+ * A vdev can be either an SR-IOV virtual function or a Scalable IOV virtual
+ * device.
+ *
+ * Return:
+ * Returns 0 upon success, -1 otherwise.
+ *
+ * EINVAL - Invalid domain ID, or the domain is not configured.
+ * EFAULT - Internal error. (Possibly caused if software is the pre-condition
+ * is not met.)
+ * ETIMEDOUT - Hardware component didn't reset in the expected time.
+ */
+int dlb_reset_domain(struct dlb_hw *hw,
+ u32 domain_id,
+ bool vdev_req,
+ unsigned int vdev_id)
+{
+ struct dlb_hw_domain *domain;
+
+ dlb_log_reset_domain(hw, domain_id, vdev_req, vdev_id);
+
+ domain = dlb_get_domain_from_id(hw, domain_id, vdev_req, vdev_id);
+
+ if (!domain || !domain->configured)
+ return -EINVAL;
+
+ return dlb_domain_reset_software_state(hw, domain);
+}
+
/**
* dlb_hw_get_num_resources() - query the PCI function's available resources
* @hw: dlb_hw handle for a particular device.
diff --git a/drivers/misc/dlb/dlb_resource.h b/drivers/misc/dlb/dlb_resource.h
index e2e49f5d9051..1b916b8a6f08 100644
--- a/drivers/misc/dlb/dlb_resource.h
+++ b/drivers/misc/dlb/dlb_resource.h
@@ -20,6 +20,11 @@ int dlb_hw_create_sched_domain(struct dlb_hw *hw,
bool vdev_req,
unsigned int vdev_id);

+int dlb_reset_domain(struct dlb_hw *hw,
+ u32 domain_id,
+ bool vdev_req,
+ unsigned int vdev_id);
+
int dlb_hw_get_num_resources(struct dlb_hw *hw,
struct dlb_get_num_resources_args *arg,
bool vdev_req,
diff --git a/include/uapi/linux/dlb.h b/include/uapi/linux/dlb.h
index 55390a1a2416..83dba8daf6e7 100644
--- a/include/uapi/linux/dlb.h
+++ b/include/uapi/linux/dlb.h
@@ -19,6 +19,7 @@ enum dlb_error {
DLB_ST_ATOMIC_INFLIGHTS_UNAVAILABLE,
DLB_ST_HIST_LIST_ENTRIES_UNAVAILABLE,
DLB_ST_LDB_PORT_REQUIRED_FOR_LDB_QUEUES,
+ DLB_ST_DOMAIN_RESET_FAILED,
};

struct dlb_cmd_response {
--
2.17.1