lkml.org 
[lkml]   [2024]   [May]   [7]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
SubjectRe: [PATCH net-next 4/5] gve: Add flow steering adminq commands
From
On 2024-05-07 15:59, Ziwei Xiao wrote:
> From: Jeroen de Borst <jeroendb@google.com>
>
> Adding new adminq commands for the driver to configure and query flow
> rules that are stored in the device. Flow steering rules are assigned
> with a location that determines the relative order of the rules.
>
> Flow rules can run up to an order of millions. In such cases, storing
> a full copy of the rules in the driver to prepare for the ethtool query
> is infeasible while querying them from the device is better. That needs
> to be optimized too so that we don't send a lot of adminq commands. The
> solution here is to store a limited number of rules/rule ids in the
> driver in a cache. This patch uses dma_pool to allocate 4k bytes which
> lets device write at most 46 flow rules(4k/88) or 1k rule ids(4096/4)
> at a time.
>
> For configuring flow rules, there are 3 sub-commands:
> - ADD which adds a rule at the location supplied
> - DEL which deletes the rule at the location supplied
> - RESET which clears all currently active rules in the device
>
> For querying flow rules, there are also 3 sub-commands:
> - QUERY_RULES corresponds to ETHTOOL_GRXCLSRULE. It fills the rules in
> the allocated cache after querying the device
> - QUERY_RULES_IDS corresponds to ETHTOOL_GRXCLSRLALL. It fills the
> rule_ids in the allocated cache after querying the device
> - QUERY_RULES_STATS corresponds to ETHTOOL_GRXCLSRLCNT. It queries the
> device's current flow rule number and the supported max flow rule
> limit
>
> Signed-off-by: Jeroen de Borst <jeroendb@google.com>
> Co-developed-by: Ziwei Xiao <ziweixiao@google.com>
> Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
> Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
> Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
> Reviewed-by: Willem de Bruijn <willemb@google.com>
> ---
> drivers/net/ethernet/google/gve/gve.h | 42 ++++++
> drivers/net/ethernet/google/gve/gve_adminq.c | 133 ++++++++++++++++++
> drivers/net/ethernet/google/gve/gve_adminq.h | 75 ++++++++++
> drivers/net/ethernet/google/gve/gve_ethtool.c | 5 +-
> drivers/net/ethernet/google/gve/gve_main.c | 54 ++++++-
> 5 files changed, 307 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/net/ethernet/google/gve/gve.h b/drivers/net/ethernet/google/gve/gve.h
> index 58213c15e084..355ae797eacf 100644
> --- a/drivers/net/ethernet/google/gve/gve.h
> +++ b/drivers/net/ethernet/google/gve/gve.h
> @@ -60,6 +60,10 @@
>
> #define GVE_DEFAULT_RX_BUFFER_OFFSET 2048
>
> +#define GVE_FLOW_RULES_CACHE_SIZE (GVE_ADMINQ_BUFFER_SIZE / sizeof(struct gve_flow_rule))
> +#define GVE_FLOW_RULE_IDS_CACHE_SIZE \
> + (GVE_ADMINQ_BUFFER_SIZE / sizeof(((struct gve_flow_rule *)0)->location))
> +
> #define GVE_XDP_ACTIONS 5
>
> #define GVE_GQ_TX_MIN_PKT_DESC_BYTES 182
> @@ -678,6 +682,39 @@ enum gve_queue_format {
> GVE_DQO_QPL_FORMAT = 0x4,
> };
>
> +struct gve_flow_spec {
> + __be32 src_ip[4];
> + __be32 dst_ip[4];
> + union {
> + struct {
> + __be16 src_port;
> + __be16 dst_port;
> + };
> + __be32 spi;
> + };
> + union {
> + u8 tos;
> + u8 tclass;
> + };
> +};
> +
> +struct gve_flow_rule {
> + u32 location;
> + u16 flow_type;
> + u16 action;
> + struct gve_flow_spec key;
> + struct gve_flow_spec mask;
> +};
> +
> +struct gve_flow_rules_cache {
> + bool rules_cache_synced; /* False if the driver's rules_cache is outdated */
> + struct gve_flow_rule *rules_cache;
> + u32 *rule_ids_cache;
> + /* The total number of queried rules that stored in the caches */
> + u32 rules_cache_num;
> + u32 rule_ids_cache_num;
> +};
> +
> struct gve_priv {
> struct net_device *dev;
> struct gve_tx_ring *tx; /* array of tx_cfg.num_queues */
> @@ -744,6 +781,8 @@ struct gve_priv {
> u32 adminq_report_link_speed_cnt;
> u32 adminq_get_ptype_map_cnt;
> u32 adminq_verify_driver_compatibility_cnt;
> + u32 adminq_query_flow_rules_cnt;
> + u32 adminq_cfg_flow_rule_cnt;
>
> /* Global stats */
> u32 interface_up_cnt; /* count of times interface turned up since last reset */
> @@ -788,6 +827,9 @@ struct gve_priv {
> bool header_split_enabled; /* True if the header split is enabled by the user */
>
> u32 max_flow_rules;
> + u32 num_flow_rules; /* Current number of flow rules registered in the device */
> +
> + struct gve_flow_rules_cache flow_rules_cache;
> };
>
> enum gve_service_task_flags_bit {
> diff --git a/drivers/net/ethernet/google/gve/gve_adminq.c b/drivers/net/ethernet/google/gve/gve_adminq.c
> index 85d0d742ad21..7a929e20cf96 100644
> --- a/drivers/net/ethernet/google/gve/gve_adminq.c
> +++ b/drivers/net/ethernet/google/gve/gve_adminq.c
> @@ -287,6 +287,8 @@ int gve_adminq_alloc(struct device *dev, struct gve_priv *priv)
> priv->adminq_report_stats_cnt = 0;
> priv->adminq_report_link_speed_cnt = 0;
> priv->adminq_get_ptype_map_cnt = 0;
> + priv->adminq_query_flow_rules_cnt = 0;
> + priv->adminq_cfg_flow_rule_cnt = 0;
>
> /* Setup Admin queue with the device */
> if (priv->pdev->revision < 0x1) {
> @@ -526,6 +528,12 @@ static int gve_adminq_issue_cmd(struct gve_priv *priv,
> case GVE_ADMINQ_VERIFY_DRIVER_COMPATIBILITY:
> priv->adminq_verify_driver_compatibility_cnt++;
> break;
> + case GVE_ADMINQ_QUERY_FLOW_RULES:
> + priv->adminq_query_flow_rules_cnt++;
> + break;
> + case GVE_ADMINQ_CONFIGURE_FLOW_RULE:
> + priv->adminq_cfg_flow_rule_cnt++;
> + break;
> default:
> dev_err(&priv->pdev->dev, "unknown AQ command opcode %d\n", opcode);
> }
> @@ -1188,3 +1196,128 @@ int gve_adminq_get_ptype_map_dqo(struct gve_priv *priv,
> ptype_map_bus);
> return err;
> }
> +
> +static int
> +gve_adminq_configure_flow_rule(struct gve_priv *priv,
> + struct gve_adminq_configure_flow_rule *flow_rule_cmd)
> +{
> + int err = gve_adminq_execute_extended_cmd(priv,
> + GVE_ADMINQ_CONFIGURE_FLOW_RULE,
> + sizeof(struct gve_adminq_configure_flow_rule),
> + flow_rule_cmd);
> +
> + if (err) {
> + dev_err(&priv->pdev->dev, "Timeout to configure the flow rule, trigger reset");
> + gve_reset(priv, true);
> + } else {
> + priv->flow_rules_cache.rules_cache_synced = false;
> + }
> +
> + return err;
> +}
> +
> +int gve_adminq_add_flow_rule(struct gve_priv *priv, struct gve_adminq_flow_rule *rule, u32 loc)
> +{
> + struct gve_adminq_configure_flow_rule flow_rule_cmd = {
> + .opcode = cpu_to_be16(GVE_RULE_ADD),
> + .location = cpu_to_be32(loc),
> + .rule = *rule,
> + };
> +
> + return gve_adminq_configure_flow_rule(priv, &flow_rule_cmd);
> +}
> +
> +int gve_adminq_del_flow_rule(struct gve_priv *priv, u32 loc)
> +{
> + struct gve_adminq_configure_flow_rule flow_rule_cmd = {
> + .opcode = cpu_to_be16(GVE_RULE_DEL),
> + .location = cpu_to_be32(loc),
> + };
> +
> + return gve_adminq_configure_flow_rule(priv, &flow_rule_cmd);
> +}
> +
> +int gve_adminq_reset_flow_rules(struct gve_priv *priv)
> +{
> + struct gve_adminq_configure_flow_rule flow_rule_cmd = {
> + .opcode = cpu_to_be16(GVE_RULE_RESET),
> + };
> +
> + return gve_adminq_configure_flow_rule(priv, &flow_rule_cmd);
> +}
> +
> +/* In the dma memory that the driver allocated for the device to query the flow rules, the device
> + * will first write it with a struct of gve_query_flow_rules_descriptor. Next to it, the device
> + * will write an array of rules or rule ids with the count that specified in the descriptor.
> + * For GVE_FLOW_RULE_QUERY_STATS, the device will only write the descriptor.
> + */
> +static int gve_adminq_process_flow_rules_query(struct gve_priv *priv, u16 query_opcode,
> + struct gve_query_flow_rules_descriptor *descriptor)
> +{
> + struct gve_flow_rules_cache *flow_rules_cache = &priv->flow_rules_cache;
> + u32 num_queried_rules, total_memory_len, rule_info_len;
> + void *descriptor_end, *rule_info;
> +
> + total_memory_len = be32_to_cpu(descriptor->total_length);
> + if (total_memory_len > GVE_ADMINQ_BUFFER_SIZE) {
> + dev_err(&priv->dev->dev, "flow rules query is out of memory.\n");

The error doesn't seem to match the inequality used. Also, how can the
HW write more than GVE_ADMINQ_BUFFER_SIZE?

> + return -ENOMEM;
> + }
> +
> + num_queried_rules = be32_to_cpu(descriptor->num_queried_rules);
> + descriptor_end = (void *)descriptor + total_memory_len;

This isn't being used.

> + rule_info = (void *)(descriptor + 1);
> +
> + switch (query_opcode) {
> + case GVE_FLOW_RULE_QUERY_RULES:
> + rule_info_len = num_queried_rules * sizeof(*flow_rules_cache->rules_cache);

Do you need to verify what the HW has written matches your expectations
i.e. is sizeof(*descriptor) + rule_info_len == total_memory_len?

> +
> + memcpy(flow_rules_cache->rules_cache, rule_info, rule_info_len);
> + flow_rules_cache->rules_cache_num = num_queried_rules;
> + break;
> + case GVE_FLOW_RULE_QUERY_IDS:
> + rule_info_len = num_queried_rules * sizeof(*flow_rules_cache->rule_ids_cache);
> +
> + memcpy(flow_rules_cache->rule_ids_cache, rule_info, rule_info_len);
> + flow_rules_cache->rule_ids_cache_num = num_queried_rules;
> + break;
> + case GVE_FLOW_RULE_QUERY_STATS:
> + priv->num_flow_rules = be32_to_cpu(descriptor->num_flow_rules);
> + priv->max_flow_rules = be32_to_cpu(descriptor->max_flow_rules);
> + return 0;
> + default:
> + return -EINVAL;
> + }
> +
> + return 0;
> +}
> +
> +int gve_adminq_query_flow_rules(struct gve_priv *priv, u16 query_opcode, u32 starting_loc)
> +{
> + struct gve_query_flow_rules_descriptor *descriptor;
> + union gve_adminq_command cmd;
> + dma_addr_t descriptor_bus;
> + int err = 0;
> +
> + memset(&cmd, 0, sizeof(cmd));
> + descriptor = dma_pool_alloc(priv->adminq_pool, GFP_KERNEL, &descriptor_bus);

Why is adminq_pool only used for 2 commands?

> + if (!descriptor)
> + return -ENOMEM;
> +
> + cmd.opcode = cpu_to_be32(GVE_ADMINQ_QUERY_FLOW_RULES);
> + cmd.query_flow_rules = (struct gve_adminq_query_flow_rules) {
> + .opcode = cpu_to_be16(query_opcode),
> + .starting_rule_id = cpu_to_be32(starting_loc),
> + .available_length = cpu_to_be64(GVE_ADMINQ_BUFFER_SIZE),
> + .rule_descriptor_addr = cpu_to_be64(descriptor_bus),
> + };
> + err = gve_adminq_execute_cmd(priv, &cmd);
> + if (err)
> + goto out;
> +
> + err = gve_adminq_process_flow_rules_query(priv, query_opcode, descriptor);
> +
> +out:
> + dma_pool_free(priv->adminq_pool, descriptor, descriptor_bus);
> + return err;
> +}
> diff --git a/drivers/net/ethernet/google/gve/gve_adminq.h b/drivers/net/ethernet/google/gve/gve_adminq.h
> index e64a0e72e781..9ddb72f92197 100644
> --- a/drivers/net/ethernet/google/gve/gve_adminq.h
> +++ b/drivers/net/ethernet/google/gve/gve_adminq.h
> @@ -25,11 +25,21 @@ enum gve_adminq_opcodes {
> GVE_ADMINQ_REPORT_LINK_SPEED = 0xD,
> GVE_ADMINQ_GET_PTYPE_MAP = 0xE,
> GVE_ADMINQ_VERIFY_DRIVER_COMPATIBILITY = 0xF,
> + GVE_ADMINQ_QUERY_FLOW_RULES = 0x10,
>
> /* For commands that are larger than 56 bytes */
> GVE_ADMINQ_EXTENDED_COMMAND = 0xFF,
> };
>
> +/* The normal adminq command is restricted to be 56 bytes at maximum. For the
> + * longer adminq command, it is wrapped by GVE_ADMINQ_EXTENDED_COMMAND with
> + * inner opcode of gve_adminq_extended_cmd_opcodes specified. The inner command
> + * is written in the dma memory allocated by GVE_ADMINQ_EXTENDED_COMMAND.
> + */
> +enum gve_adminq_extended_cmd_opcodes {
> + GVE_ADMINQ_CONFIGURE_FLOW_RULE = 0x101,
> +};
> +
> /* Admin queue status codes */
> enum gve_adminq_statuses {
> GVE_ADMINQ_COMMAND_UNSET = 0x0,
> @@ -434,6 +444,66 @@ struct gve_adminq_get_ptype_map {
> __be64 ptype_map_addr;
> };
>
> +/* Flow-steering related definitions */
> +enum gve_adminq_flow_rule_cfg_opcode {
> + GVE_RULE_ADD = 0,
> + GVE_RULE_DEL = 1,
> + GVE_RULE_RESET = 2,
> +};

Could these be more descriptive?

> +
> +enum gve_adminq_flow_rule_query_opcode {
> + GVE_FLOW_RULE_QUERY_RULES = 0,
> + GVE_FLOW_RULE_QUERY_IDS = 1,
> + GVE_FLOW_RULE_QUERY_STATS = 2,
> +};
> +
> +enum gve_adminq_flow_type {
> + GVE_FLOW_TYPE_TCPV4,
> + GVE_FLOW_TYPE_UDPV4,
> + GVE_FLOW_TYPE_SCTPV4,
> + GVE_FLOW_TYPE_AHV4,
> + GVE_FLOW_TYPE_ESPV4,
> + GVE_FLOW_TYPE_TCPV6,
> + GVE_FLOW_TYPE_UDPV6,
> + GVE_FLOW_TYPE_SCTPV6,
> + GVE_FLOW_TYPE_AHV6,
> + GVE_FLOW_TYPE_ESPV6,
> +};
> +
> +/* Flow-steering command */
> +struct gve_adminq_flow_rule {
> + __be16 flow_type;
> + __be16 action; /* RX queue id */
> + struct gve_flow_spec key;
> + struct gve_flow_spec mask;
> +};
> +
> +struct gve_adminq_configure_flow_rule {
> + __be16 opcode;
> + u8 padding[2];
> + struct gve_adminq_flow_rule rule;
> + __be32 location;
> +};
> +
> +static_assert(sizeof(struct gve_adminq_configure_flow_rule) == 92);
> +
> +struct gve_query_flow_rules_descriptor {
> + __be32 num_flow_rules; /* Current rule counts stored in the device */
> + __be32 max_flow_rules;
> + __be32 num_queried_rules;

nit: more comments here are appreciated.

> + __be32 total_length; /* The memory length that the device writes */
> +};
> +
> +struct gve_adminq_query_flow_rules {
> + __be16 opcode;
> + u8 padding[2];
> + __be32 starting_rule_id;
> + __be64 available_length; /* The dma memory length that the driver allocated */
> + __be64 rule_descriptor_addr; /* The dma memory address */
> +};
> +
> +static_assert(sizeof(struct gve_adminq_query_flow_rules) == 24);
> +
> union gve_adminq_command {
> struct {
> __be32 opcode;
> @@ -454,6 +524,7 @@ union gve_adminq_command {
> struct gve_adminq_get_ptype_map get_ptype_map;
> struct gve_adminq_verify_driver_compatibility
> verify_driver_compatibility;
> + struct gve_adminq_query_flow_rules query_flow_rules;
> struct gve_adminq_extended_command extended_command;
> };
> };
> @@ -488,6 +559,10 @@ int gve_adminq_verify_driver_compatibility(struct gve_priv *priv,
> u64 driver_info_len,
> dma_addr_t driver_info_addr);
> int gve_adminq_report_link_speed(struct gve_priv *priv);
> +int gve_adminq_add_flow_rule(struct gve_priv *priv, struct gve_adminq_flow_rule *rule, u32 loc);
> +int gve_adminq_del_flow_rule(struct gve_priv *priv, u32 loc);
> +int gve_adminq_reset_flow_rules(struct gve_priv *priv);
> +int gve_adminq_query_flow_rules(struct gve_priv *priv, u16 query_opcode, u32 starting_loc);
>
> struct gve_ptype_lut;
> int gve_adminq_get_ptype_map_dqo(struct gve_priv *priv,
> diff --git a/drivers/net/ethernet/google/gve/gve_ethtool.c b/drivers/net/ethernet/google/gve/gve_ethtool.c
> index 156b7e128b53..02cee7e0e229 100644
> --- a/drivers/net/ethernet/google/gve/gve_ethtool.c
> +++ b/drivers/net/ethernet/google/gve/gve_ethtool.c
> @@ -74,7 +74,8 @@ static const char gve_gstrings_adminq_stats[][ETH_GSTRING_LEN] = {
> "adminq_create_tx_queue_cnt", "adminq_create_rx_queue_cnt",
> "adminq_destroy_tx_queue_cnt", "adminq_destroy_rx_queue_cnt",
> "adminq_dcfg_device_resources_cnt", "adminq_set_driver_parameter_cnt",
> - "adminq_report_stats_cnt", "adminq_report_link_speed_cnt", "adminq_get_ptype_map_cnt"
> + "adminq_report_stats_cnt", "adminq_report_link_speed_cnt", "adminq_get_ptype_map_cnt",
> + "adminq_query_flow_rules", "adminq_cfg_flow_rule",
> };
>
> static const char gve_gstrings_priv_flags[][ETH_GSTRING_LEN] = {
> @@ -458,6 +459,8 @@ gve_get_ethtool_stats(struct net_device *netdev,
> data[i++] = priv->adminq_report_stats_cnt;
> data[i++] = priv->adminq_report_link_speed_cnt;
> data[i++] = priv->adminq_get_ptype_map_cnt;
> + data[i++] = priv->adminq_query_flow_rules_cnt;
> + data[i++] = priv->adminq_cfg_flow_rule_cnt;
> }
>
> static void gve_get_channels(struct net_device *netdev,
> diff --git a/drivers/net/ethernet/google/gve/gve_main.c b/drivers/net/ethernet/google/gve/gve_main.c
> index cabf7d4bcecb..eb435ccbe98e 100644
> --- a/drivers/net/ethernet/google/gve/gve_main.c
> +++ b/drivers/net/ethernet/google/gve/gve_main.c
> @@ -141,6 +141,52 @@ static void gve_get_stats(struct net_device *dev, struct rtnl_link_stats64 *s)
> }
> }
>
> +static int gve_alloc_flow_rule_caches(struct gve_priv *priv)
> +{
> + struct gve_flow_rules_cache *flow_rules_cache = &priv->flow_rules_cache;
> + int err = 0;
> +
> + if (!priv->max_flow_rules)
> + return 0;
> +
> + flow_rules_cache->rules_cache =
> + kvcalloc(GVE_FLOW_RULES_CACHE_SIZE, sizeof(*flow_rules_cache->rules_cache),
> + GFP_KERNEL);
> + if (!flow_rules_cache->rules_cache) {
> + dev_err(&priv->pdev->dev, "Cannot alloc flow rules cache\n");
> + return -ENOMEM;
> + }
> +
> + flow_rules_cache->rule_ids_cache =
> + kvcalloc(GVE_FLOW_RULE_IDS_CACHE_SIZE, sizeof(*flow_rules_cache->rule_ids_cache),
> + GFP_KERNEL);
> + if (!flow_rules_cache->rule_ids_cache) {
> + dev_err(&priv->pdev->dev, "Cannot alloc flow rule ids cache\n");
> + err = -ENOMEM;
> + goto free_rules_cache;
> + }
> +
> + return 0;
> +
> +free_rules_cache:
> + kvfree(flow_rules_cache->rules_cache);
> + flow_rules_cache->rules_cache = NULL;
> + return err;
> +}
> +
> +static void gve_free_flow_rule_caches(struct gve_priv *priv)
> +{
> + struct gve_flow_rules_cache *flow_rules_cache = &priv->flow_rules_cache;
> +
> + if (!priv->max_flow_rules)
> + return;

Is this needed? Is kernel style just kvfree() w/o checks?

> +
> + kvfree(flow_rules_cache->rule_ids_cache);
> + flow_rules_cache->rule_ids_cache = NULL;
> + kvfree(flow_rules_cache->rules_cache);
> + flow_rules_cache->rules_cache = NULL;
> +}
> +
> static int gve_alloc_counter_array(struct gve_priv *priv)
> {
> priv->counter_array =
> @@ -521,9 +567,12 @@ static int gve_setup_device_resources(struct gve_priv *priv)
> {
> int err;
>
> - err = gve_alloc_counter_array(priv);
> + err = gve_alloc_flow_rule_caches(priv);
> if (err)
> return err;
> + err = gve_alloc_counter_array(priv);
> + if (err)
> + goto abort_with_flow_rule_caches;
> err = gve_alloc_notify_blocks(priv);
> if (err)
> goto abort_with_counter;
> @@ -575,6 +624,8 @@ static int gve_setup_device_resources(struct gve_priv *priv)
> gve_free_notify_blocks(priv);
> abort_with_counter:
> gve_free_counter_array(priv);
> +abort_with_flow_rule_caches:
> + gve_free_flow_rule_caches(priv);
>
> return err;
> }
> @@ -606,6 +657,7 @@ static void gve_teardown_device_resources(struct gve_priv *priv)
> kvfree(priv->ptype_lut_dqo);
> priv->ptype_lut_dqo = NULL;
>
> + gve_free_flow_rule_caches(priv);
> gve_free_counter_array(priv);
> gve_free_notify_blocks(priv);
> gve_free_stats_report(priv);

\
 
 \ /
  Last update: 2024-05-27 18:20    [W:0.147 / U:1.556 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site