From 29cc6679076a00a6ce193004dcf2d14ae7c428a5 Mon Sep 17 00:00:00 2001 From: Amir Vadai Date: Thu, 14 Jul 2016 10:32:37 +0300 Subject: net/mlx5: Store counters in rbtree instead of list In order to use bulk counters, we need to have counters sorted by id. Signed-off-by: Amir Vadai Reviewed-by: Or Gerlitz Signed-off-by: Saeed Mahameed Signed-off-by: David S. Miller --- include/linux/mlx5/driver.h | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'include') diff --git a/include/linux/mlx5/driver.h b/include/linux/mlx5/driver.h index 81e8396574f4..a041b99fceac 100644 --- a/include/linux/mlx5/driver.h +++ b/include/linux/mlx5/driver.h @@ -469,7 +469,7 @@ struct mlx5_irq_info { }; struct mlx5_fc_stats { - struct list_head list; + struct rb_root counters; struct list_head addlist; /* protect addlist add/splice operations */ spinlock_t addlist_lock; -- cgit v1.2.3 From a351a1b03bf169f77891060be30036ef71cbe618 Mon Sep 17 00:00:00 2001 From: Amir Vadai Date: Thu, 14 Jul 2016 10:32:38 +0300 Subject: net/mlx5: Introduce bulk reading of flow counters This commit utilize the ability of ConnectX-4 to bulk read flow counters. Few bulk counter queries could be done instead of issuing thousands of firmware commands per second to get statistics of all flows set to HW, such as those programmed when we offload tc filters. Counters are stored sorted by hardware id, and queried in blocks (id + number of counters). Due to hardware requirement, start of block and number of counters in a block must be four aligned. Reviewed-by: Or Gerlitz Signed-off-by: Amir Vadai Signed-off-by: Saeed Mahameed Signed-off-by: David S. Miller --- drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c | 67 +++++++++++++++++ drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.h | 12 ++++ .../net/ethernet/mellanox/mlx5/core/fs_counters.c | 83 ++++++++++++++++------ include/linux/mlx5/mlx5_ifc.h | 8 ++- 4 files changed, 146 insertions(+), 24 deletions(-) (limited to 'include') diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c index a5bb6b695242..9134010e2921 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.c @@ -413,3 +413,70 @@ int mlx5_cmd_fc_query(struct mlx5_core_dev *dev, u16 id, return 0; } + +struct mlx5_cmd_fc_bulk { + u16 id; + int num; + int outlen; + u32 out[0]; +}; + +struct mlx5_cmd_fc_bulk * +mlx5_cmd_fc_bulk_alloc(struct mlx5_core_dev *dev, u16 id, int num) +{ + struct mlx5_cmd_fc_bulk *b; + int outlen = sizeof(*b) + + MLX5_ST_SZ_BYTES(query_flow_counter_out) + + MLX5_ST_SZ_BYTES(traffic_counter) * num; + + b = kzalloc(outlen, GFP_KERNEL); + if (!b) + return NULL; + + b->id = id; + b->num = num; + b->outlen = outlen; + + return b; +} + +void mlx5_cmd_fc_bulk_free(struct mlx5_cmd_fc_bulk *b) +{ + kfree(b); +} + +int +mlx5_cmd_fc_bulk_query(struct mlx5_core_dev *dev, struct mlx5_cmd_fc_bulk *b) +{ + u32 in[MLX5_ST_SZ_DW(query_flow_counter_in)]; + + memset(in, 0, sizeof(in)); + + MLX5_SET(query_flow_counter_in, in, opcode, + MLX5_CMD_OP_QUERY_FLOW_COUNTER); + MLX5_SET(query_flow_counter_in, in, op_mod, 0); + MLX5_SET(query_flow_counter_in, in, flow_counter_id, b->id); + MLX5_SET(query_flow_counter_in, in, num_of_counters, b->num); + + return mlx5_cmd_exec_check_status(dev, in, sizeof(in), + b->out, b->outlen); +} + +void mlx5_cmd_fc_bulk_get(struct mlx5_core_dev *dev, + struct mlx5_cmd_fc_bulk *b, u16 id, + u64 *packets, u64 *bytes) +{ + int index = id - b->id; + void *stats; + + if (index < 0 || index >= b->num) { + mlx5_core_warn(dev, "Flow counter id (0x%x) out of range (0x%x..0x%x). Counter ignored.\n", + id, b->id, b->id + b->num - 1); + return; + } + + stats = MLX5_ADDR_OF(query_flow_counter_out, b->out, + flow_statistics[index]); + *packets = MLX5_GET64(traffic_counter, stats, packets); + *bytes = MLX5_GET64(traffic_counter, stats, octets); +} diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.h b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.h index fc4f7b83fe0a..158844cef82b 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.h +++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_cmd.h @@ -76,4 +76,16 @@ int mlx5_cmd_fc_alloc(struct mlx5_core_dev *dev, u16 *id); int mlx5_cmd_fc_free(struct mlx5_core_dev *dev, u16 id); int mlx5_cmd_fc_query(struct mlx5_core_dev *dev, u16 id, u64 *packets, u64 *bytes); + +struct mlx5_cmd_fc_bulk; + +struct mlx5_cmd_fc_bulk * +mlx5_cmd_fc_bulk_alloc(struct mlx5_core_dev *dev, u16 id, int num); +void mlx5_cmd_fc_bulk_free(struct mlx5_cmd_fc_bulk *b); +int +mlx5_cmd_fc_bulk_query(struct mlx5_core_dev *dev, struct mlx5_cmd_fc_bulk *b); +void mlx5_cmd_fc_bulk_get(struct mlx5_core_dev *dev, + struct mlx5_cmd_fc_bulk *b, u16 id, + u64 *packets, u64 *bytes); + #endif diff --git a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c index aaf8fd149cc6..c2877e9de8a1 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/fs_counters.c @@ -90,16 +90,66 @@ static void mlx5_fc_stats_insert(struct rb_root *root, struct mlx5_fc *counter) rb_insert_color(&counter->node, root); } +static struct rb_node *mlx5_fc_stats_query(struct mlx5_core_dev *dev, + struct mlx5_fc *first, + u16 last_id) +{ + struct mlx5_cmd_fc_bulk *b; + struct rb_node *node = NULL; + u16 afirst_id; + int num; + int err; + int max_bulk = 1 << MLX5_CAP_GEN(dev, log_max_flow_counter_bulk); + + /* first id must be aligned to 4 when using bulk query */ + afirst_id = first->id & ~0x3; + + /* number of counters to query inc. the last counter */ + num = ALIGN(last_id - afirst_id + 1, 4); + if (num > max_bulk) { + num = max_bulk; + last_id = afirst_id + num - 1; + } + + b = mlx5_cmd_fc_bulk_alloc(dev, afirst_id, num); + if (!b) { + mlx5_core_err(dev, "Error allocating resources for bulk query\n"); + return NULL; + } + + err = mlx5_cmd_fc_bulk_query(dev, b); + if (err) { + mlx5_core_err(dev, "Error doing bulk query: %d\n", err); + goto out; + } + + for (node = &first->node; node; node = rb_next(node)) { + struct mlx5_fc *counter = rb_entry(node, struct mlx5_fc, node); + struct mlx5_fc_cache *c = &counter->cache; + + if (counter->id > last_id) + break; + + mlx5_cmd_fc_bulk_get(dev, b, + counter->id, &c->packets, &c->bytes); + } + +out: + mlx5_cmd_fc_bulk_free(b); + + return node; +} + static void mlx5_fc_stats_work(struct work_struct *work) { struct mlx5_core_dev *dev = container_of(work, struct mlx5_core_dev, priv.fc_stats.work.work); struct mlx5_fc_stats *fc_stats = &dev->priv.fc_stats; unsigned long now = jiffies; - struct mlx5_fc *counter; + struct mlx5_fc *counter = NULL; + struct mlx5_fc *last = NULL; struct rb_node *node; LIST_HEAD(tmplist); - int err = 0; spin_lock(&fc_stats->addlist_lock); @@ -115,12 +165,7 @@ static void mlx5_fc_stats_work(struct work_struct *work) node = rb_first(&fc_stats->counters); while (node) { - struct mlx5_fc_cache *c; - u64 packets; - u64 bytes; - counter = rb_entry(node, struct mlx5_fc, node); - c = &counter->cache; node = rb_next(node); @@ -133,26 +178,20 @@ static void mlx5_fc_stats_work(struct work_struct *work) continue; } - if (time_before(now, fc_stats->next_query)) - continue; + last = counter; + } - err = mlx5_cmd_fc_query(dev, counter->id, &packets, &bytes); - if (err) { - pr_err("Error querying stats for counter id %d\n", - counter->id); - continue; - } + if (time_before(now, fc_stats->next_query) || !last) + return; - if (packets == c->packets) - continue; + node = rb_first(&fc_stats->counters); + while (node) { + counter = rb_entry(node, struct mlx5_fc, node); - c->lastuse = jiffies; - c->packets = packets; - c->bytes = bytes; + node = mlx5_fc_stats_query(dev, counter, last->id); } - if (time_after_eq(now, fc_stats->next_query)) - fc_stats->next_query = now + MLX5_FC_STATS_PERIOD; + fc_stats->next_query = now + MLX5_FC_STATS_PERIOD; } struct mlx5_fc *mlx5_fc_create(struct mlx5_core_dev *dev, bool aging) diff --git a/include/linux/mlx5/mlx5_ifc.h b/include/linux/mlx5/mlx5_ifc.h index 152421cc6f44..d671e4e8e7db 100644 --- a/include/linux/mlx5/mlx5_ifc.h +++ b/include/linux/mlx5/mlx5_ifc.h @@ -893,7 +893,10 @@ struct mlx5_ifc_cmd_hca_cap_bits { u8 reserved_at_330[0xb]; u8 log_max_xrcd[0x5]; - u8 reserved_at_340[0x20]; + u8 reserved_at_340[0x8]; + u8 log_max_flow_counter_bulk[0x8]; + u8 max_flow_counter[0x10]; + u8 reserved_at_360[0x3]; u8 log_max_rq[0x5]; @@ -980,7 +983,8 @@ struct mlx5_ifc_dest_format_struct_bits { }; struct mlx5_ifc_flow_counter_list_bits { - u8 reserved_at_0[0x10]; + u8 clear[0x1]; + u8 num_of_counters[0xf]; u8 flow_counter_id[0x10]; u8 reserved_at_20[0x20]; -- cgit v1.2.3 From 8438884d4ab423161b974854ebb90c08219dd678 Mon Sep 17 00:00:00 2001 From: Or Gerlitz Date: Thu, 14 Jul 2016 10:32:43 +0300 Subject: net/switchdev: Export the same parent ID service function This helper serves to know if two switchdev port netdevices belong to the same HW ASIC, e.g to figure out if forwarding offload is possible between them. Signed-off-by: Or Gerlitz Signed-off-by: Saeed Mahameed Signed-off-by: David S. Miller --- include/net/switchdev.h | 8 ++++++++ net/switchdev/switchdev.c | 5 +++-- 2 files changed, 11 insertions(+), 2 deletions(-) (limited to 'include') diff --git a/include/net/switchdev.h b/include/net/switchdev.h index 985619a59323..9023e3e3be0b 100644 --- a/include/net/switchdev.h +++ b/include/net/switchdev.h @@ -227,6 +227,8 @@ void switchdev_port_fwd_mark_set(struct net_device *dev, struct net_device *group_dev, bool joining); +bool switchdev_port_same_parent_id(struct net_device *a, + struct net_device *b); #else static inline void switchdev_deferred_process(void) @@ -351,6 +353,12 @@ static inline void switchdev_port_fwd_mark_set(struct net_device *dev, { } +static inline bool switchdev_port_same_parent_id(struct net_device *a, + struct net_device *b) +{ + return false; +} + #endif #endif /* _LINUX_SWITCHDEV_H_ */ diff --git a/net/switchdev/switchdev.c b/net/switchdev/switchdev.c index 59658b2e9cdf..a5fc9dd24aa9 100644 --- a/net/switchdev/switchdev.c +++ b/net/switchdev/switchdev.c @@ -1286,8 +1286,8 @@ void switchdev_fib_ipv4_abort(struct fib_info *fi) } EXPORT_SYMBOL_GPL(switchdev_fib_ipv4_abort); -static bool switchdev_port_same_parent_id(struct net_device *a, - struct net_device *b) +bool switchdev_port_same_parent_id(struct net_device *a, + struct net_device *b) { struct switchdev_attr a_attr = { .orig_dev = a, @@ -1323,6 +1323,7 @@ static u32 switchdev_port_fwd_mark_get(struct net_device *dev, return dev->ifindex; } +EXPORT_SYMBOL_GPL(switchdev_port_same_parent_id); static void switchdev_port_fwd_mark_reset(struct net_device *group_dev, u32 old_mark, u32 *reset_mark) -- cgit v1.2.3