Optimized Forwarding Header by JaiOCP · Pull Request #2297 · opencomputeproject/SAI

JaiOCP · 2026-06-17T21:00:22Z

New PR is opened.
Old one where most of the review comments are there is closed
#2285

Signed-off-by: JaiOCP <jai.kumar@broadcom.com>

JaiOCP · 2026-06-18T15:26:52Z

    SAI_NATIVE_HASH_FIELD_RDMA_BTH_DEST_QP = 0x00000023,

+    /** Native hash field RDMA packet BTH destination queue pair */
+    SAI_NATIVE_HASH_FIELD_OFH_FLOW_LABEL = 0x00000024,


Fix the comments cut and paste error.

JaiOCP · 2026-06-18T15:30:13Z

+     * @flags CREATE_ONLY
+     * @default false
+     */
+    SAI_ACL_TABLE_ATTR_FIELD_ACL_OFH_TYPE = SAI_ACL_TABLE_ATTR_FIELD_START + 0x167,


Please add match for version field

BYGX-wcr · 2026-06-18T15:31:56Z

It will be good to draw an abstract SAI pipeline to illustrate the forwarding model of OFH in the documentation.

JaiOCP · 2026-06-18T15:34:27Z

+     * @default false
+     */
+    SAI_ACL_TABLE_ATTR_FIELD_OFH_COS = SAI_ACL_TABLE_ATTR_FIELD_START + 0x168,
+


Please add TTL decrement for action as well as match on TTL

skannan-CLS · 2026-06-18T15:48:21Z

     * @brief End of Rule Actions
     */
-    SAI_ACL_ENTRY_ATTR_ACTION_END = SAI_ACL_ENTRY_ATTR_ACTION_TAM_OBJECT,
+    SAI_ACL_ENTRY_ATTR_ACTION_END = SAI_ACL_ENTRY_ATTR_ACTION_SET_OFH_COS,


SAI_ACL_ACTION_TYPE_SET_ECN = 0x00000017,

Set ECN also supported for normal IP packet.

rdasari-upscaleai · 2026-06-23T07:16:06Z

+    /**
+     * @brief OFH header max size
+     *
+     * @type sai_int8_t


Should it be sai_uint8_t ?

rdasari-upscaleai · 2026-06-23T07:23:25Z

+
+    SAI_OFH_ADDR_FAMILY_AFH,
+
+    SAI_OFH_ADDR_FAMILY_UFH,


Please add doxygen comments

rdasari-upscaleai · 2026-06-23T07:44:11Z


    /** @validonly meta->attrvaluetype == SAI_ATTR_VALUE_TYPE_UINT64_RANGE_LIST */
    sai_u64_range_list_t u64rangelist;



Should the corresponding sai_ofh_addr_and_mask_t to SAI_ATTR_VALUE_TYPE_OFH_ADDR_AND_MASK type be added here?

rdasari-upscaleai · 2026-06-23T07:51:00Z

+/**
+ * @brief Set OFH attribute value(s).
+ *
+ * @param[in] ofh_id TAM id


Please correct TAM to OFH

rdasari-upscaleai · 2026-06-23T07:56:32Z

+ * @param[out] object_statuses List of status for every object. Caller needs to
+ * allocate the buffer
+ *
+ * @return #SAI_STATUS_SUCCESS on success when all objects are removed or


Appear to be cut and paste error from route remove entry.
sairoute.h appear to have same cut and paste error.

rdasari-upscaleai · 2026-06-23T07:57:17Z

+ * @param[out] object_statuses List of status for every object. Caller needs to
+ * allocate the buffer
+ *
+ * @return #SAI_STATUS_SUCCESS on success when all objects are removed or


Appear to be cut and paste error from route remove entry.
sairoute.h appear to have same cut and paste error.

rdasari-upscaleai · 2026-06-23T08:06:31Z

+     *
+     * @type sai_uint8_t
+     * @flags CREATE_AND_SET
+     * @default 6


while other *_WIDTH fields are default to '0', This seems to be special.

hzheng5 · 2026-06-25T02:42:31Z

Hi @JaiOCP , thanks for the OFH contribution. It aligns well with UEC Scale-Up goals.

A few points before merge:
• Naming: Standardize to SAI_OFH_ATTR_* for OFH objects and SAI_ROUTING_ATTR_OFH_* for forwarding behavior.
• Capability Query: Add sai_query_ofh_capability() to expose max size, supported address families, and hash limits before config.
• Counters: Define SAI_PORT_STAT_OFH_* counters to enable silicon-level verification.
• OFH Size: Clarify units for @default 6 (bytes vs DW) and add validation range.

LGTM overall.

rck-innovium · 2026-06-22T05:02:41Z

+    /** OFH VLAN drop */
+    SAI_IN_DROP_REASON_OFH_VLAN_DROP,
+
+    /** OFH VLAN to virtual router id drop */


Please explain this drop reason either as comments in the header file or in the markdown file to be created.

Which of the below drops are accounted under this drop reason?

SAI_ROUTER_INTERFACE_ATTR_ADMIN_OFH_STATE is false

SAI_VIRTUAL_ROUTER_ATTR_ADMIN_OFH_STATE is false

rck-innovium · 2026-06-22T05:04:18Z

+    SAI_IN_DROP_REASON_OFH_DISCARD,
+
+    /** OFH VLAN drop */
+    SAI_IN_DROP_REASON_OFH_VLAN_DROP,


If I understand correctly, SAI_IN_DROP_REASON_OFH_VLAN_DROP is meant to account the drops when RIF of type SAI_ROUTER_INTERFACE_TYPE_VLAN is not created for the given ingress port + packet.VID. Ideally, this would be accounted under SAI_IN_DROP_REASON_INGRESS_VLAN_FILTER. We do not have IPv4/IPv6 specific VLAN drops. I am trying to understand as to why we need OFH specific VLAN drops.

rck-innovium · 2026-06-24T05:13:37Z

+     * @type sai_ofh_sub_type_t
+     * @flags CREATE_ONLY
+     * @default SAI_OFH_SUB_TYPE_NONE
+     */


I presume this is to identify UFH1/2, AFH1/2 and is not valid for ESUN. Please add appropriate @validonly

rck-innovium · 2026-06-24T05:31:18Z

+    SAI_OFH_ATTR_SUB_TYPE,
+
+    /**
+     * @brief OFH Version number


@brief Version/Revision number

rck-innovium · 2026-06-24T05:33:57Z

+     * @type sai_ofh_ver_t
+     * @flags CREATE_AND_SET
+     * @default SAI_OFH_VER_1
+     * @validonly SAI_OFH_ATTR_TYPE == SAI_OFH_TYPE_AFH or SAI_OFH_ATTR_TYPE == SAI_OFH_TYPE_ESUN


Version is valid for UHF also. If the idea is to add UHF as a separate PR, then we could remove SAI_OFH_TYPE_UFH from sai_ofh_type_t.

rck-innovium · 2026-06-25T15:00:18Z

    SAI_QOS_MAP_TYPE_QUEUE_TO_VC = 0x00000011,

+    /** QOS Map to set OFH COS to Traffic class */
+    SAI_QOS_MAP_TYPE_OFH_COS_TO_TC = 0x00000012,


Need to extend sai_qos_map_params_t to include a new field: sai_uint8_t ofh_cos.
Note that increases the size of the struct and needs handling in sairedis serialization/deserialization.

Optimized Forwarding Header

b3a920a

Signed-off-by: JaiOCP <jai.kumar@broadcom.com>

JaiOCP mentioned this pull request Jun 17, 2026

Optimized Forwarding Headers #2285

Closed

JaiOCP commented Jun 18, 2026

View reviewed changes

skannan-CLS reviewed Jun 18, 2026

View reviewed changes

rdasari-upscaleai reviewed Jun 23, 2026

View reviewed changes

rck-innovium suggested changes Jun 25, 2026

View reviewed changes


		/** @validonly meta->attrvaluetype == SAI_ATTR_VALUE_TYPE_UINT64_RANGE_LIST */
		sai_u64_range_list_t u64rangelist;

Uh oh!

Conversation

JaiOCP commented Jun 17, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

BYGX-wcr commented Jun 18, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rdasari-upscaleai Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

rdasari-upscaleai Jun 23, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

hzheng5 commented Jun 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

rdasari-upscaleai Jun 23, 2026 •

edited

Loading

rdasari-upscaleai Jun 23, 2026 •

edited

Loading

hzheng5 commented Jun 25, 2026 •

edited

Loading