[v3,07/33] net/ena: restructure the llq policy setting process

Message ID 20240306122445.4350-8-shaibran@amazon.com (mailing list archive)
State Superseded, archived
Delegated to: Ferruh Yigit
Headers
Series net/ena: v2.9.0 driver release |

Checks

Context Check Description
ci/checkpatch success coding style OK

Commit Message

Brandes, Shai March 6, 2024, 12:24 p.m. UTC
  From: Shai Brandes <shaibran@amazon.com>

The driver will set the size of the LLQ header size according to the
recommendation from the device.
Replaced `enable_llq` and `large_llq_hdr` devargs with
a new devarg `llq_policy` that accepts the following values:
0 - Disable LLQ.
    Use with extreme caution as it leads to a huge performance
    degradation on AWS instances from 6th generation onwards.
1 - Accept device recommended LLQ policy (Default).
    Device can recommend normal or large LLQ policy.
2 - Enforce normal LLQ policy.
3 - Enforce large LLQ policy.
    Required for packets with header that exceed 96 bytes on
    AWS instances prior to 5th generation.

Signed-off-by: Shai Brandes <shaibran@amazon.com>
Reviewed-by: Amit Bernstein <amitbern@amazon.com>
---
 doc/guides/nics/ena.rst                |  21 ++---
 doc/guides/rel_notes/release_24_03.rst |   1 +
 drivers/net/ena/ena_ethdev.c           | 110 +++++++++++++------------
 drivers/net/ena/ena_ethdev.h           |  11 ++-
 4 files changed, 77 insertions(+), 66 deletions(-)
  

Comments

Ferruh Yigit March 8, 2024, 5:24 p.m. UTC | #1
Addressed
On 3/6/2024 12:24 PM, shaibran@amazon.com wrote:
> From: Shai Brandes <shaibran@amazon.com>
> 
> The driver will set the size of the LLQ header size according to the
> recommendation from the device.
> Replaced `enable_llq` and `large_llq_hdr` devargs with
> a new devarg `llq_policy` that accepts the following values:
> 0 - Disable LLQ.
>     Use with extreme caution as it leads to a huge performance
>     degradation on AWS instances from 6th generation onwards.
> 1 - Accept device recommended LLQ policy (Default).
>     Device can recommend normal or large LLQ policy.
> 2 - Enforce normal LLQ policy.
> 3 - Enforce large LLQ policy.
>     Required for packets with header that exceed 96 bytes on
>     AWS instances prior to 5th generation.
> 

We had similar discussion before, although dev_args is not part of the
ABI, it is an user interface, and changes in the devargs will impact
users directly.

What would you think to either keep backward compatilibity in the
devargs (like not remove old one but add new one), or do this change in
24.11 release?
  
Brandes, Shai March 10, 2024, 2:29 p.m. UTC | #2
> -----Original Message-----
> From: Ferruh Yigit <ferruh.yigit@amd.com>
> Sent: Friday, March 8, 2024 7:24 PM
> To: Brandes, Shai <shaibran@amazon.com>
> Cc: dev@dpdk.org
> Subject: RE: [EXTERNAL] [PATCH v3 07/33] net/ena: restructure the llq policy
> setting process
> 
> CAUTION: This email originated from outside of the organization. Do not click
> links or open attachments unless you can confirm the sender and know the
> content is safe.
> 
> 
> 
> On 3/6/2024 12:24 PM, shaibran@amazon.com wrote:
> > From: Shai Brandes <shaibran@amazon.com>
> >
> > The driver will set the size of the LLQ header size according to the
> > recommendation from the device.
> > Replaced `enable_llq` and `large_llq_hdr` devargs with a new devarg
> > `llq_policy` that accepts the following values:
> > 0 - Disable LLQ.
> >     Use with extreme caution as it leads to a huge performance
> >     degradation on AWS instances from 6th generation onwards.
> > 1 - Accept device recommended LLQ policy (Default).
> >     Device can recommend normal or large LLQ policy.
> > 2 - Enforce normal LLQ policy.
> > 3 - Enforce large LLQ policy.
> >     Required for packets with header that exceed 96 bytes on
> >     AWS instances prior to 5th generation.
> >
> 
> We had similar discussion before, although dev_args is not part of the ABI, it
> is an user interface, and changes in the devargs will impact users directly.
> 
> What would you think to either keep backward compatilibity in the devargs
> (like not remove old one but add new one), or do this change in
> 24.11 release?
[Brandes, Shai] understood. 
The new devarg replaced the old ones and added option to enforce normal-llq mode which is critical for our release.
As you suggested, we will keep backward compatibility and add an additional devarg for enforcing  normal-llq policy.
That way, we can easily replace it in future releases with a common devarg without the need to make major logic changes.
  
Ferruh Yigit March 13, 2024, 11:21 a.m. UTC | #3
On 3/10/2024 2:29 PM, Brandes, Shai wrote:
> 
> 
>> -----Original Message-----
>> From: Ferruh Yigit <ferruh.yigit@amd.com>
>> Sent: Friday, March 8, 2024 7:24 PM
>> To: Brandes, Shai <shaibran@amazon.com>
>> Cc: dev@dpdk.org
>> Subject: RE: [EXTERNAL] [PATCH v3 07/33] net/ena: restructure the llq policy
>> setting process
>>
>> CAUTION: This email originated from outside of the organization. Do not click
>> links or open attachments unless you can confirm the sender and know the
>> content is safe.
>>
>>
>>
>> On 3/6/2024 12:24 PM, shaibran@amazon.com wrote:
>>> From: Shai Brandes <shaibran@amazon.com>
>>>
>>> The driver will set the size of the LLQ header size according to the
>>> recommendation from the device.
>>> Replaced `enable_llq` and `large_llq_hdr` devargs with a new devarg
>>> `llq_policy` that accepts the following values:
>>> 0 - Disable LLQ.
>>>     Use with extreme caution as it leads to a huge performance
>>>     degradation on AWS instances from 6th generation onwards.
>>> 1 - Accept device recommended LLQ policy (Default).
>>>     Device can recommend normal or large LLQ policy.
>>> 2 - Enforce normal LLQ policy.
>>> 3 - Enforce large LLQ policy.
>>>     Required for packets with header that exceed 96 bytes on
>>>     AWS instances prior to 5th generation.
>>>
>>
>> We had similar discussion before, although dev_args is not part of the ABI, it
>> is an user interface, and changes in the devargs will impact users directly.
>>
>> What would you think to either keep backward compatilibity in the devargs
>> (like not remove old one but add new one), or do this change in
>> 24.11 release?
> [Brandes, Shai] understood. 
> The new devarg replaced the old ones and added option to enforce normal-llq mode which is critical for our release.
> As you suggested, we will keep backward compatibility and add an additional devarg for enforcing  normal-llq policy.
> That way, we can easily replace it in future releases with a common devarg without the need to make major logic changes.
> 

ack.
  

Patch

diff --git a/doc/guides/nics/ena.rst b/doc/guides/nics/ena.rst
index b039e75ead..53c9341859 100644
--- a/doc/guides/nics/ena.rst
+++ b/doc/guides/nics/ena.rst
@@ -107,11 +107,15 @@  Configuration
 Runtime Configuration
 ^^^^^^^^^^^^^^^^^^^^^
 
-   * **large_llq_hdr** (default 0)
+   * **llq_policy** (default 1)
 
-     Enables or disables usage of large LLQ headers. This option will have
-     effect only if the device also supports large LLQ headers. Otherwise, the
-     default value will be used.
+     Controls whether use device recommended header policy or override it.
+     0 - Disable LLQ.
+         **Use with extreme caution as it leads to a huge performance
+         degradation on AWS instances from 6th generation onwards.**
+     1 - Accept device recommended LLQ policy (Default).
+     2 - Enforce normal LLQ policy.
+     3 - Enforce large LLQ policy.
 
    * **miss_txc_to** (default 5)
 
@@ -122,15 +126,6 @@  Runtime Configuration
      timer service. Setting this parameter to 0 disables this feature. Maximum
      allowed value is 60 seconds.
 
-   * **enable_llq** (default 1)
-
-     Determines whenever the driver should use the LLQ (if it's available) or
-     not.
-
-     **NOTE: On the 6th generation AWS instances disabling LLQ may lead to a
-     huge performance degradation. In general disabling LLQ is highly not
-     recommended!**
-
 ENA Configuration Parameters
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
diff --git a/doc/guides/rel_notes/release_24_03.rst b/doc/guides/rel_notes/release_24_03.rst
index 6b73d4fedf..2a22bb07ed 100644
--- a/doc/guides/rel_notes/release_24_03.rst
+++ b/doc/guides/rel_notes/release_24_03.rst
@@ -106,6 +106,7 @@  New Features
   * Removed the reporting of `rx_overruns` errors from xstats and instead updated `imissed` stat with its value.
   * Added support for sub-optimal configuration notifications from the device.
   * Restructured fast release of mbufs when RTE_ETH_TX_OFFLOAD_MBUF_FAST_FREE optimization is enabled.
+  * Replaced `enable_llq` and `large_llq_hdr` devargs with a new devarg `llq_policy`.
 
 * **Updated Atomic Rules' Arkville driver.**
 
diff --git a/drivers/net/ena/ena_ethdev.c b/drivers/net/ena/ena_ethdev.c
index 537ee9f8c3..2414f631c8 100644
--- a/drivers/net/ena/ena_ethdev.c
+++ b/drivers/net/ena/ena_ethdev.c
@@ -40,6 +40,8 @@ 
 
 #define BITS_PER_TYPE(type) (sizeof(type) * BITS_PER_BYTE)
 
+#define DECIMAL_BASE 10
+
 /*
  * We should try to keep ENA_CLEANUP_BUF_SIZE lower than
  * RTE_MEMPOOL_CACHE_MAX_SIZE, so we can fit this in mempool local cache.
@@ -74,17 +76,23 @@  struct ena_stats {
 	ENA_STAT_ENTRY(stat, srd)
 
 /* Device arguments */
-#define ENA_DEVARG_LARGE_LLQ_HDR "large_llq_hdr"
+/* Controls whether to disable LLQ, use device recommended header policy
+ * or overriding the device recommendation.
+ * 0 - Disable LLQ.
+ *     Use with extreme caution as it leads to a huge performance
+ *     degradation on AWS instances from 6th generation onwards.
+ * 1 - Accept device recommended LLQ policy (Default).
+ *     Device can recommend normal or large LLQ policy.
+ * 2 - Enforce normal LLQ policy.
+ * 3 - Enforce large LLQ policy.
+ *     Required for packets with header that exceed 96 bytes on
+ *     AWS instances prior to 5th generation.
+ */
+#define ENA_DEVARG_LLQ_POLICY "llq_policy"
 /* Timeout in seconds after which a single uncompleted Tx packet should be
  * considered as a missing.
  */
 #define ENA_DEVARG_MISS_TXC_TO "miss_txc_to"
-/*
- * Controls whether LLQ should be used (if available). Enabled by default.
- * NOTE: It's highly not recommended to disable the LLQ, as it may lead to a
- * huge performance degradation on 6th generation AWS instances.
- */
-#define ENA_DEVARG_ENABLE_LLQ "enable_llq"
 
 /*
  * Each rte_memzone should have unique name.
@@ -279,9 +287,9 @@  static int ena_xstats_get_by_id(struct rte_eth_dev *dev,
 				const uint64_t *ids,
 				uint64_t *values,
 				unsigned int n);
-static int ena_process_bool_devarg(const char *key,
-				   const char *value,
-				   void *opaque);
+static int ena_process_llq_policy_devarg(const char *key,
+					 const char *value,
+					 void *opaque);
 static int ena_parse_devargs(struct ena_adapter *adapter,
 			     struct rte_devargs *devargs);
 static void ena_copy_customer_metrics(struct ena_adapter *adapter,
@@ -297,6 +305,7 @@  static int ena_rx_queue_intr_disable(struct rte_eth_dev *dev,
 static int ena_configure_aenq(struct ena_adapter *adapter);
 static int ena_mp_primary_handle(const struct rte_mp_msg *mp_msg,
 				 const void *peer);
+static bool ena_use_large_llq_hdr(struct ena_adapter *adapter, uint8_t recommended_entry_size);
 
 static const struct eth_dev_ops ena_dev_ops = {
 	.dev_configure          = ena_dev_configure,
@@ -1135,6 +1144,7 @@  ena_calc_io_queue_size(struct ena_calc_queue_size_ctx *ctx,
 	ctx->max_tx_queue_size = max_tx_queue_size;
 	ctx->max_rx_queue_size = max_rx_queue_size;
 
+	PMD_DRV_LOG(INFO, "tx queue size %u\n", max_tx_queue_size);
 	return 0;
 }
 
@@ -2034,7 +2044,7 @@  ena_set_queues_placement_policy(struct ena_adapter *adapter,
 	int rc;
 	u32 llq_feature_mask;
 
-	if (!adapter->enable_llq) {
+	if (adapter->llq_header_policy == ENA_LLQ_POLICY_DISABLED) {
 		PMD_DRV_LOG(WARNING,
 			"NOTE: LLQ has been disabled as per user's request. "
 			"This may lead to a huge performance degradation!\n");
@@ -2239,8 +2249,7 @@  static int eth_ena_dev_init(struct rte_eth_dev *eth_dev)
 
 	/* Assign default devargs values */
 	adapter->missing_tx_completion_to = ENA_TX_TIMEOUT;
-	adapter->enable_llq = true;
-	adapter->use_large_llq_hdr = false;
+	adapter->llq_header_policy = ENA_LLQ_POLICY_RECOMMENDED;
 
 	rc = ena_parse_devargs(adapter, pci_dev->device.devargs);
 	if (rc != 0) {
@@ -2264,8 +2273,9 @@  static int eth_ena_dev_init(struct rte_eth_dev *eth_dev)
 	if (!(adapter->all_aenq_groups & BIT(ENA_ADMIN_LINK_CHANGE)))
 		adapter->edev_data->dev_flags &= ~RTE_ETH_DEV_INTR_LSC;
 
-	set_default_llq_configurations(&llq_config, &get_feat_ctx.llq,
-		adapter->use_large_llq_hdr);
+	bool use_large_llq_hdr = ena_use_large_llq_hdr(adapter,
+						       get_feat_ctx.llq.entry_size_recommended);
+	set_default_llq_configurations(&llq_config, &get_feat_ctx.llq, use_large_llq_hdr);
 	rc = ena_set_queues_placement_policy(adapter, ena_dev,
 					     &get_feat_ctx.llq, &llq_config);
 	if (unlikely(rc)) {
@@ -2273,18 +2283,19 @@  static int eth_ena_dev_init(struct rte_eth_dev *eth_dev)
 		return rc;
 	}
 
-	if (ena_dev->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_HOST)
+	if (ena_dev->tx_mem_queue_type == ENA_ADMIN_PLACEMENT_POLICY_HOST) {
 		queue_type_str = "Regular";
-	else
+	} else {
 		queue_type_str = "Low latency";
+		PMD_DRV_LOG(INFO, "LLQ entry size %uB\n", llq_config.llq_ring_entry_size_value);
+	}
 	PMD_DRV_LOG(INFO, "Placement policy: %s\n", queue_type_str);
 
 	calc_queue_ctx.ena_dev = ena_dev;
 	calc_queue_ctx.get_feat_ctx = &get_feat_ctx;
 
 	max_num_io_queues = ena_calc_max_io_queue_num(ena_dev, &get_feat_ctx);
-	rc = ena_calc_io_queue_size(&calc_queue_ctx,
-		adapter->use_large_llq_hdr);
+	rc = ena_calc_io_queue_size(&calc_queue_ctx, use_large_llq_hdr);
 	if (unlikely((rc != 0) || (max_num_io_queues == 0))) {
 		rc = -EFAULT;
 		goto err_device_destroy;
@@ -3632,7 +3643,7 @@  static int ena_process_uint_devarg(const char *key,
 	char *str_end;
 	uint64_t uint_value;
 
-	uint_value = strtoull(value, &str_end, 10);
+	uint_value = strtoull(value, &str_end, DECIMAL_BASE);
 	if (value == str_end) {
 		PMD_INIT_LOG(ERR,
 			"Invalid value for key '%s'. Only uint values are accepted.\n",
@@ -3663,41 +3674,29 @@  static int ena_process_uint_devarg(const char *key,
 	return 0;
 }
 
-static int ena_process_bool_devarg(const char *key,
-				   const char *value,
-				   void *opaque)
+static int ena_process_llq_policy_devarg(const char *key, const char *value, void *opaque)
 {
 	struct ena_adapter *adapter = opaque;
-	bool bool_value;
+	uint32_t policy;
 
-	/* Parse the value. */
-	if (strcmp(value, "1") == 0) {
-		bool_value = true;
-	} else if (strcmp(value, "0") == 0) {
-		bool_value = false;
+	policy = strtoul(value, NULL, DECIMAL_BASE);
+	if (policy < ENA_LLQ_POLICY_LAST) {
+		adapter->llq_header_policy = policy;
 	} else {
-		PMD_INIT_LOG(ERR,
-			"Invalid value: '%s' for key '%s'. Accepted: '0' or '1'\n",
-			value, key);
+		PMD_INIT_LOG(ERR, "Invalid value: '%s' for key '%s'. valid [0-3]\n", value, key);
 		return -EINVAL;
 	}
-
-	/* Now, assign it to the proper adapter field. */
-	if (strcmp(key, ENA_DEVARG_LARGE_LLQ_HDR) == 0)
-		adapter->use_large_llq_hdr = bool_value;
-	else if (strcmp(key, ENA_DEVARG_ENABLE_LLQ) == 0)
-		adapter->enable_llq = bool_value;
-
+	PMD_DRV_LOG(INFO,
+		"LLQ policy is %u [0 - disabled, 1 - device recommended, 2 - normal, 3 - large]\n",
+		adapter->llq_header_policy);
 	return 0;
 }
 
-static int ena_parse_devargs(struct ena_adapter *adapter,
-			     struct rte_devargs *devargs)
+static int ena_parse_devargs(struct ena_adapter *adapter, struct rte_devargs *devargs)
 {
 	static const char * const allowed_args[] = {
-		ENA_DEVARG_LARGE_LLQ_HDR,
+		ENA_DEVARG_LLQ_POLICY,
 		ENA_DEVARG_MISS_TXC_TO,
-		ENA_DEVARG_ENABLE_LLQ,
 		NULL,
 	};
 	struct rte_kvargs *kvlist;
@@ -3708,21 +3707,18 @@  static int ena_parse_devargs(struct ena_adapter *adapter,
 
 	kvlist = rte_kvargs_parse(devargs->args, allowed_args);
 	if (kvlist == NULL) {
-		PMD_INIT_LOG(ERR, "Invalid device arguments: %s\n",
-			devargs->args);
+		PMD_INIT_LOG(ERR, "Invalid device arguments: %s\n", devargs->args);
 		return -EINVAL;
 	}
 
-	rc = rte_kvargs_process(kvlist, ENA_DEVARG_LARGE_LLQ_HDR,
-		ena_process_bool_devarg, adapter);
+	rc = rte_kvargs_process(kvlist, ENA_DEVARG_LLQ_POLICY,
+		ena_process_llq_policy_devarg, adapter);
 	if (rc != 0)
 		goto exit;
 	rc = rte_kvargs_process(kvlist, ENA_DEVARG_MISS_TXC_TO,
 		ena_process_uint_devarg, adapter);
 	if (rc != 0)
 		goto exit;
-	rc = rte_kvargs_process(kvlist, ENA_DEVARG_ENABLE_LLQ,
-		ena_process_bool_devarg, adapter);
 
 exit:
 	rte_kvargs_free(kvlist);
@@ -3942,8 +3938,7 @@  RTE_PMD_REGISTER_PCI(net_ena, rte_ena_pmd);
 RTE_PMD_REGISTER_PCI_TABLE(net_ena, pci_id_ena_map);
 RTE_PMD_REGISTER_KMOD_DEP(net_ena, "* igb_uio | uio_pci_generic | vfio-pci");
 RTE_PMD_REGISTER_PARAM_STRING(net_ena,
-	ENA_DEVARG_LARGE_LLQ_HDR "=<0|1> "
-	ENA_DEVARG_ENABLE_LLQ "=<0|1> "
+	ENA_DEVARG_LLQ_POLICY "=<0|1|2|3> "
 	ENA_DEVARG_MISS_TXC_TO "=<uint>");
 RTE_LOG_REGISTER_SUFFIX(ena_logtype_init, init, NOTICE);
 RTE_LOG_REGISTER_SUFFIX(ena_logtype_driver, driver, NOTICE);
@@ -4129,3 +4124,16 @@  ena_mp_primary_handle(const struct rte_mp_msg *mp_msg, const void *peer)
 	/* Return just IPC processing status */
 	return rte_mp_reply(&mp_rsp, peer);
 }
+
+static bool ena_use_large_llq_hdr(struct ena_adapter *adapter, uint8_t recommended_entry_size)
+{
+	if (adapter->llq_header_policy == ENA_LLQ_POLICY_LARGE) {
+		return true;
+	} else if (adapter->llq_header_policy == ENA_LLQ_POLICY_RECOMMENDED) {
+		PMD_DRV_LOG(INFO, "Recommended device entry size policy %u\n",
+			recommended_entry_size);
+		if (recommended_entry_size == ENA_ADMIN_LIST_ENTRY_SIZE_256B)
+			return true;
+	}
+	return false;
+}
diff --git a/drivers/net/ena/ena_ethdev.h b/drivers/net/ena/ena_ethdev.h
index 20b8307836..6716f01ba5 100644
--- a/drivers/net/ena/ena_ethdev.h
+++ b/drivers/net/ena/ena_ethdev.h
@@ -85,6 +85,14 @@  enum ena_ring_type {
 	ENA_RING_TYPE_TX = 2,
 };
 
+typedef enum ena_llq_policy_t {
+	ENA_LLQ_POLICY_DISABLED    = 0, /* Host queues */
+	ENA_LLQ_POLICY_RECOMMENDED = 1, /* Device recommendation */
+	ENA_LLQ_POLICY_NORMAL      = 2, /* 128B long LLQ entry */
+	ENA_LLQ_POLICY_LARGE       = 3, /* 256B long LLQ entry */
+	ENA_LLQ_POLICY_LAST,
+} ena_llq_policy;
+
 struct ena_tx_buffer {
 	struct rte_mbuf *mbuf;
 	unsigned int tx_descs;
@@ -329,8 +337,7 @@  struct ena_adapter {
 
 	bool trigger_reset;
 
-	bool enable_llq;
-	bool use_large_llq_hdr;
+	ena_llq_policy llq_header_policy;
 
 	uint32_t last_tx_comp_qid;
 	uint64_t missing_tx_completion_to;