[PATCH net-next 0/6] sfc: introduce eth, ipv4 and ipv6 pedit offloads

[PATCH net-next 0/6] sfc: introduce eth, ipv4 and ipv6 pedit offloads

[Pieter Jansen van Vuuren]

This set introduces mac source and destination pedit set action offloads.
It also adds offload for ipv4 ttl and ipv6 hop limit pedit set action as
well pedit add actions that would result in the same semantics as
decrementing the ttl and hop limit.

Pieter Jansen van Vuuren (6):
  sfc: introduce ethernet pedit set action infrastructure
  sfc: add mac source and destination pedit action offload
  sfc: add decrement ttl by offloading set ipv4 ttl actions
  sfc: add decrement ipv6 hop limit by offloading set hop limit actions
  sfc: introduce pedit add actions on the ipv4 ttl field
  sfc: extend pedit add action to handle decrement ipv6 hop limit

 drivers/net/ethernet/sfc/mae.c |  89 ++++++-
 drivers/net/ethernet/sfc/mae.h |   4 +
 drivers/net/ethernet/sfc/tc.c  | 474 +++++++++++++++++++++++++++++++++
 drivers/net/ethernet/sfc/tc.h  |  58 +++-
 4 files changed, 612 insertions(+), 13 deletions(-)

-- 
2.17.1

[PATCH net-next 1/6] sfc: introduce ethernet pedit set action infrastructure

[Pieter Jansen van Vuuren]

Introduce the initial ethernet pedit set action infrastructure in
preparation for adding mac src and dst pedit action offloads.

Co-developed-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
---
 drivers/net/ethernet/sfc/mae.c | 83 ++++++++++++++++++++++++++++++++--
 drivers/net/ethernet/sfc/mae.h |  4 ++
 drivers/net/ethernet/sfc/tc.c  | 70 ++++++++++++++++++++++++++++
 drivers/net/ethernet/sfc/tc.h  | 56 ++++++++++++++++++++---
 4 files changed, 202 insertions(+), 11 deletions(-)

diff --git a/drivers/net/ethernet/sfc/mae.c b/drivers/net/ethernet/sfc/mae.c
index 3b8780c76b6e..a7ad7ab8c5f4 100644
--- a/drivers/net/ethernet/sfc/mae.c
+++ b/drivers/net/ethernet/sfc/mae.c
@@ -1219,6 +1219,71 @@ int efx_mae_enumerate_mports(struct efx_nic *efx)
 	return rc;
 }
 
+/**
+ * efx_mae_allocate_pedit_mac() - allocate pedit MAC address in HW.
+ * @efx:	NIC we're installing a pedit MAC address on
+ * @ped:	pedit MAC action to be installed
+ *
+ * Attempts to install @ped in HW and populates its id with an index of this
+ * entry in the firmware MAC address table on success.
+ *
+ * Return: negative value on error, 0 in success.
+ */
+int efx_mae_allocate_pedit_mac(struct efx_nic *efx,
+			       struct efx_tc_mac_pedit_action *ped)
+{
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_MAE_MAC_ADDR_ALLOC_OUT_LEN);
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_MAE_MAC_ADDR_ALLOC_IN_LEN);
+	size_t outlen;
+	int rc;
+
+	BUILD_BUG_ON(MC_CMD_MAE_MAC_ADDR_ALLOC_IN_MAC_ADDR_LEN !=
+		     sizeof(ped->h_addr));
+	memcpy(MCDI_PTR(inbuf, MAE_MAC_ADDR_ALLOC_IN_MAC_ADDR), ped->h_addr,
+	       sizeof(ped->h_addr));
+	rc = efx_mcdi_rpc(efx, MC_CMD_MAE_MAC_ADDR_ALLOC, inbuf, sizeof(inbuf),
+			  outbuf, sizeof(outbuf), &outlen);
+	if (rc)
+		return rc;
+	if (outlen < sizeof(outbuf))
+		return -EIO;
+	ped->fw_id = MCDI_DWORD(outbuf, MAE_MAC_ADDR_ALLOC_OUT_MAC_ID);
+	return 0;
+}
+
+/**
+ * efx_mae_free_pedit_mac() - free pedit MAC address in HW.
+ * @efx:	NIC we're installing a pedit MAC address on
+ * @ped:	pedit MAC action that needs to be freed
+ *
+ * Frees @ped in HW, check that firmware did not free a different one and clears
+ * the id (which denotes the index of the entry in the MAC address table).
+ */
+void efx_mae_free_pedit_mac(struct efx_nic *efx,
+			    struct efx_tc_mac_pedit_action *ped)
+{
+	MCDI_DECLARE_BUF(outbuf, MC_CMD_MAE_MAC_ADDR_FREE_OUT_LEN(1));
+	MCDI_DECLARE_BUF(inbuf, MC_CMD_MAE_MAC_ADDR_FREE_IN_LEN(1));
+	size_t outlen;
+	int rc;
+
+	MCDI_SET_DWORD(inbuf, MAE_MAC_ADDR_FREE_IN_MAC_ID, ped->fw_id);
+	rc = efx_mcdi_rpc(efx, MC_CMD_MAE_MAC_ADDR_FREE, inbuf,
+			  sizeof(inbuf), outbuf, sizeof(outbuf), &outlen);
+	if (rc || outlen < sizeof(outbuf))
+		return;
+	/* FW freed a different ID than we asked for, should also never happen.
+	 * Warn because it means we've now got a different idea to the FW of
+	 * what MAC addresses exist, which could cause mayhem later.
+	 */
+	if (WARN_ON(MCDI_DWORD(outbuf, MAE_MAC_ADDR_FREE_OUT_FREED_MAC_ID) != ped->fw_id))
+		return;
+	/* We're probably about to free @ped, but let's just make sure its
+	 * fw_id is blatted so that it won't look valid if it leaks out.
+	 */
+	ped->fw_id = MC_CMD_MAE_MAC_ADDR_ALLOC_OUT_MAC_ID_NULL;
+}
+
 int efx_mae_alloc_action_set(struct efx_nic *efx, struct efx_tc_action_set *act)
 {
 	MCDI_DECLARE_BUF(outbuf, MC_CMD_MAE_ACTION_SET_ALLOC_OUT_LEN);
@@ -1231,10 +1296,20 @@ int efx_mae_alloc_action_set(struct efx_nic *efx, struct efx_tc_action_set *act)
 			      MAE_ACTION_SET_ALLOC_IN_VLAN_POP, act->vlan_pop,
 			      MAE_ACTION_SET_ALLOC_IN_DECAP, act->decap);
 
-	MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_SRC_MAC_ID,
-		       MC_CMD_MAE_MAC_ADDR_ALLOC_OUT_MAC_ID_NULL);
-	MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_DST_MAC_ID,
-		       MC_CMD_MAE_MAC_ADDR_ALLOC_OUT_MAC_ID_NULL);
+	if (act->src_mac)
+		MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_SRC_MAC_ID,
+			       act->src_mac->fw_id);
+	else
+		MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_SRC_MAC_ID,
+			       MC_CMD_MAE_MAC_ADDR_ALLOC_OUT_MAC_ID_NULL);
+
+	if (act->dst_mac)
+		MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_DST_MAC_ID,
+			       act->dst_mac->fw_id);
+	else
+		MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_DST_MAC_ID,
+			       MC_CMD_MAE_MAC_ADDR_ALLOC_OUT_MAC_ID_NULL);
+
 	if (act->count && !WARN_ON(!act->count->cnt))
 		MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_COUNTER_ID,
 			       act->count->cnt->fw_id);
diff --git a/drivers/net/ethernet/sfc/mae.h b/drivers/net/ethernet/sfc/mae.h
index e88e80574f15..8df30bc4f3ba 100644
--- a/drivers/net/ethernet/sfc/mae.h
+++ b/drivers/net/ethernet/sfc/mae.h
@@ -103,6 +103,10 @@ int efx_mae_update_encap_md(struct efx_nic *efx,
 int efx_mae_free_encap_md(struct efx_nic *efx,
 			  struct efx_tc_encap_action *encap);
 
+int efx_mae_allocate_pedit_mac(struct efx_nic *efx,
+			       struct efx_tc_mac_pedit_action *ped);
+void efx_mae_free_pedit_mac(struct efx_nic *efx,
+			    struct efx_tc_mac_pedit_action *ped);
 int efx_mae_alloc_action_set(struct efx_nic *efx, struct efx_tc_action_set *act);
 int efx_mae_free_action_set(struct efx_nic *efx, u32 fw_id);
 
diff --git a/drivers/net/ethernet/sfc/tc.c b/drivers/net/ethernet/sfc/tc.c
index 039180c61c83..8a9fc2f47514 100644
--- a/drivers/net/ethernet/sfc/tc.c
+++ b/drivers/net/ethernet/sfc/tc.c
@@ -86,6 +86,12 @@ s64 efx_tc_flower_external_mport(struct efx_nic *efx, struct efx_rep *efv)
 	return mport;
 }
 
+static const struct rhashtable_params efx_tc_mac_ht_params = {
+	.key_len	= offsetofend(struct efx_tc_mac_pedit_action, h_addr),
+	.key_offset	= 0,
+	.head_offset	= offsetof(struct efx_tc_mac_pedit_action, linkage),
+};
+
 static const struct rhashtable_params efx_tc_encap_match_ht_params = {
 	.key_len	= offsetof(struct efx_tc_encap_match, linkage),
 	.key_offset	= 0,
@@ -110,6 +116,56 @@ static const struct rhashtable_params efx_tc_recirc_ht_params = {
 	.head_offset	= offsetof(struct efx_tc_recirc_id, linkage),
 };
 
+static struct efx_tc_mac_pedit_action __maybe_unused *efx_tc_flower_get_mac(struct efx_nic *efx,
+							     unsigned char h_addr[ETH_ALEN],
+							     struct netlink_ext_ack *extack)
+{
+	struct efx_tc_mac_pedit_action *ped, *old;
+	int rc;
+
+	ped = kzalloc(sizeof(*ped), GFP_USER);
+	if (!ped)
+		return ERR_PTR(-ENOMEM);
+	memcpy(ped->h_addr, h_addr, ETH_ALEN);
+	old = rhashtable_lookup_get_insert_fast(&efx->tc->mac_ht,
+						&ped->linkage,
+						efx_tc_mac_ht_params);
+	if (old) {
+		/* don't need our new entry */
+		kfree(ped);
+		if (!refcount_inc_not_zero(&old->ref))
+			return ERR_PTR(-EAGAIN);
+		/* existing entry found, ref taken */
+		return old;
+	}
+
+	rc = efx_mae_allocate_pedit_mac(efx, ped);
+	if (rc < 0) {
+		NL_SET_ERR_MSG_MOD(extack, "Failed to store pedit MAC address in hw");
+		goto out_remove;
+	}
+
+	/* ref and return */
+	refcount_set(&ped->ref, 1);
+	return ped;
+out_remove:
+	rhashtable_remove_fast(&efx->tc->mac_ht, &ped->linkage,
+			       efx_tc_mac_ht_params);
+	kfree(ped);
+	return ERR_PTR(rc);
+}
+
+static void __maybe_unused efx_tc_flower_put_mac(struct efx_nic *efx,
+				  struct efx_tc_mac_pedit_action *ped)
+{
+	if (!refcount_dec_and_test(&ped->ref))
+		return; /* still in use */
+	rhashtable_remove_fast(&efx->tc->mac_ht, &ped->linkage,
+			       efx_tc_mac_ht_params);
+	efx_mae_free_pedit_mac(efx, ped);
+	kfree(ped);
+}
+
 static void efx_tc_free_action_set(struct efx_nic *efx,
 				   struct efx_tc_action_set *act, bool in_hw)
 {
@@ -2156,6 +2212,14 @@ static void efx_tc_lhs_free(void *ptr, void *arg)
 	kfree(rule);
 }
 
+static void efx_tc_mac_free(void *ptr, void *__unused)
+{
+	struct efx_tc_mac_pedit_action *ped = ptr;
+
+	WARN_ON(refcount_read(&ped->ref));
+	kfree(ped);
+}
+
 static void efx_tc_flow_free(void *ptr, void *arg)
 {
 	struct efx_tc_flow_rule *rule = ptr;
@@ -2196,6 +2260,9 @@ int efx_init_struct_tc(struct efx_nic *efx)
 	rc = efx_tc_init_counters(efx);
 	if (rc < 0)
 		goto fail_counters;
+	rc = rhashtable_init(&efx->tc->mac_ht, &efx_tc_mac_ht_params);
+	if (rc < 0)
+		goto fail_mac_ht;
 	rc = rhashtable_init(&efx->tc->encap_match_ht, &efx_tc_encap_match_ht_params);
 	if (rc < 0)
 		goto fail_encap_match_ht;
@@ -2233,6 +2300,8 @@ int efx_init_struct_tc(struct efx_nic *efx)
 fail_match_action_ht:
 	rhashtable_destroy(&efx->tc->encap_match_ht);
 fail_encap_match_ht:
+	rhashtable_destroy(&efx->tc->mac_ht);
+fail_mac_ht:
 	efx_tc_destroy_counters(efx);
 fail_counters:
 	efx_tc_destroy_encap_actions(efx);
@@ -2268,6 +2337,7 @@ void efx_fini_struct_tc(struct efx_nic *efx)
 	rhashtable_free_and_destroy(&efx->tc->recirc_ht, efx_tc_recirc_free, efx);
 	WARN_ON(!ida_is_empty(&efx->tc->recirc_ida));
 	ida_destroy(&efx->tc->recirc_ida);
+	rhashtable_free_and_destroy(&efx->tc->mac_ht, efx_tc_mac_free, NULL);
 	efx_tc_fini_counters(efx);
 	efx_tc_fini_encap_actions(efx);
 	mutex_unlock(&efx->tc->mutex);
diff --git a/drivers/net/ethernet/sfc/tc.h b/drivers/net/ethernet/sfc/tc.h
index 40d2c803fca8..91705411e5c5 100644
--- a/drivers/net/ethernet/sfc/tc.h
+++ b/drivers/net/ethernet/sfc/tc.h
@@ -18,6 +18,23 @@
 
 #define IS_ALL_ONES(v)	(!(typeof (v))~(v))
 
+/**
+ * struct efx_tc_mac_pedit_action - mac pedit action fields
+ *
+ * @h_addr: mac address field of ethernet header
+ * @linkage: rhashtable reference
+ * @ref: reference count
+ * @fw_id: index of this entry in firmware MAC address table
+ *
+ * MAC address edits are indirected through a table in the hardware
+ */
+struct efx_tc_mac_pedit_action {
+	u8 h_addr[ETH_ALEN];
+	struct rhash_head linkage;
+	refcount_t ref;
+	u32 fw_id; /* index of this entry in firmware MAC address table */
+};
+
 static inline bool efx_ipv6_addr_all_ones(struct in6_addr *addr)
 {
 	return !memchr_inv(addr, 0xff, sizeof(*addr));
@@ -25,20 +42,43 @@ static inline bool efx_ipv6_addr_all_ones(struct in6_addr *addr)
 
 struct efx_tc_encap_action; /* see tc_encap_actions.h */
 
+/**
+ * struct efx_tc_action_set - collection of tc action fields
+ *
+ * @vlan_push: the number of vlan headers to push
+ * @vlan_pop: the number of vlan headers to pop
+ * @decap: used to indicate a tunnel header decapsulation should take place
+ * @deliver: used to indicate a deliver action should take place
+ * @vlan_tci: tci fields for vlan push actions
+ * @vlan_proto: ethernet types for vlan push actions
+ * @count: counter mapping
+ * @encap_md: encap entry in tc_encap_ht table
+ * @encap_user: linked list of encap users (encap_md->users)
+ * @user: owning action-set-list. Only populated if @encap_md is; used by efx_tc_update_encap() fallback handling
+ * @count_user: linked list of counter users (counter->users)
+ * @dest_mport: destination mport
+ * @src_mac: source mac entry in tc_mac_ht table
+ * @dst_mac: destination mac entry in tc_mac_ht table
+ * @fw_id: index of this entry in firmware actions table
+ * @list: linked list of tc actions
+ *
+ */
 struct efx_tc_action_set {
 	u16 vlan_push:2;
 	u16 vlan_pop:2;
 	u16 decap:1;
 	u16 deliver:1;
-	__be16 vlan_tci[2]; /* TCIs for vlan_push */
-	__be16 vlan_proto[2]; /* Ethertypes for vlan_push */
+	__be16 vlan_tci[2];
+	__be16 vlan_proto[2];
 	struct efx_tc_counter_index *count;
-	struct efx_tc_encap_action *encap_md; /* entry in tc_encap_ht table */
-	struct list_head encap_user; /* entry on encap_md->users list */
-	struct efx_tc_action_set_list *user; /* Only populated if encap_md */
-	struct list_head count_user; /* entry on counter->users list, if encap */
+	struct efx_tc_encap_action *encap_md;
+	struct list_head encap_user;
+	struct efx_tc_action_set_list *user;
+	struct list_head count_user;
 	u32 dest_mport;
-	u32 fw_id; /* index of this entry in firmware actions table */
+	struct efx_tc_mac_pedit_action *src_mac;
+	struct efx_tc_mac_pedit_action *dst_mac;
+	u32 fw_id;
 	struct list_head list;
 };
 
@@ -220,6 +260,7 @@ struct efx_tc_table_ct { /* TABLE_ID_CONNTRACK_TABLE */
  * @counter_ht: Hashtable of TC counters (FW IDs and counter values)
  * @counter_id_ht: Hashtable mapping TC counter cookies to counters
  * @encap_ht: Hashtable of TC encap actions
+ * @mac_ht: Hashtable of MAC address entries (for pedits)
  * @encap_match_ht: Hashtable of TC encap matches
  * @match_action_ht: Hashtable of TC match-action rules
  * @lhs_rule_ht: Hashtable of TC left-hand (act ct & goto chain) rules
@@ -257,6 +298,7 @@ struct efx_tc_state {
 	struct rhashtable counter_ht;
 	struct rhashtable counter_id_ht;
 	struct rhashtable encap_ht;
+	struct rhashtable mac_ht;
 	struct rhashtable encap_match_ht;
 	struct rhashtable match_action_ht;
 	struct rhashtable lhs_rule_ht;
-- 
2.17.1

[PATCH net-next 2/6] sfc: add mac source and destination pedit action offload

[Pieter Jansen van Vuuren]

Introduce the first pedit set offload functionality for the sfc driver.
In addition to this, add offload functionality for both mac source and
destination pedit set actions.

Co-developed-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
---
 drivers/net/ethernet/sfc/tc.c | 209 +++++++++++++++++++++++++++++++++-
 1 file changed, 207 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/sfc/tc.c b/drivers/net/ethernet/sfc/tc.c
index 8a9fc2f47514..47bf59529a46 100644
--- a/drivers/net/ethernet/sfc/tc.c
+++ b/drivers/net/ethernet/sfc/tc.c
@@ -116,7 +116,7 @@ static const struct rhashtable_params efx_tc_recirc_ht_params = {
 	.head_offset	= offsetof(struct efx_tc_recirc_id, linkage),
 };
 
-static struct efx_tc_mac_pedit_action __maybe_unused *efx_tc_flower_get_mac(struct efx_nic *efx,
+static struct efx_tc_mac_pedit_action *efx_tc_flower_get_mac(struct efx_nic *efx,
 							     unsigned char h_addr[ETH_ALEN],
 							     struct netlink_ext_ack *extack)
 {
@@ -155,7 +155,7 @@ static struct efx_tc_mac_pedit_action __maybe_unused *efx_tc_flower_get_mac(stru
 	return ERR_PTR(rc);
 }
 
-static void __maybe_unused efx_tc_flower_put_mac(struct efx_nic *efx,
+static void efx_tc_flower_put_mac(struct efx_nic *efx,
 				  struct efx_tc_mac_pedit_action *ped)
 {
 	if (!refcount_dec_and_test(&ped->ref))
@@ -191,6 +191,10 @@ static void efx_tc_free_action_set(struct efx_nic *efx,
 		list_del(&act->encap_user);
 		efx_tc_flower_release_encap_md(efx, act->encap_md);
 	}
+	if (act->src_mac)
+		efx_tc_flower_put_mac(efx, act->src_mac);
+	if (act->dst_mac)
+		efx_tc_flower_put_mac(efx, act->dst_mac);
 	kfree(act);
 }
 
@@ -753,6 +757,7 @@ static const char *efx_tc_encap_type_name(enum efx_encap_type typ)
 /* For details of action order constraints refer to SF-123102-TC-1§12.6.1 */
 enum efx_tc_action_order {
 	EFX_TC_AO_DECAP,
+	EFX_TC_AO_PEDIT_MAC_ADDRS,
 	EFX_TC_AO_VLAN_POP,
 	EFX_TC_AO_VLAN_PUSH,
 	EFX_TC_AO_COUNT,
@@ -767,6 +772,11 @@ static bool efx_tc_flower_action_order_ok(const struct efx_tc_action_set *act,
 	case EFX_TC_AO_DECAP:
 		if (act->decap)
 			return false;
+		/* PEDIT_MAC_ADDRS must not happen before DECAP, though it
+		 * can wait until much later
+		 */
+		if (act->dst_mac || act->src_mac)
+			return false;
 		fallthrough;
 	case EFX_TC_AO_VLAN_POP:
 		if (act->vlan_pop >= 2)
@@ -786,6 +796,7 @@ static bool efx_tc_flower_action_order_ok(const struct efx_tc_action_set *act,
 		if (act->count)
 			return false;
 		fallthrough;
+	case EFX_TC_AO_PEDIT_MAC_ADDRS:
 	case EFX_TC_AO_ENCAP:
 		if (act->encap_md)
 			return false;
@@ -956,6 +967,191 @@ static void efx_tc_flower_release_lhs_actions(struct efx_nic *efx,
 		efx_tc_flower_put_counter_index(efx, act->count);
 }
 
+/**
+ * struct efx_tc_mangler_state - accumulates 32-bit pedits into fields
+ *
+ * @dst_mac_32: dst_mac[0:3] has been populated
+ * @dst_mac_16: dst_mac[4:5] has been populated
+ * @src_mac_16: src_mac[0:1] has been populated
+ * @src_mac_32: src_mac[2:5] has been populated
+ * @dst_mac: h_dest field of ethhdr
+ * @src_mac: h_source field of ethhdr
+ *
+ * Since FLOW_ACTION_MANGLE comes in 32-bit chunks that do not
+ * necessarily equate to whole fields of the packet header, this
+ * structure is used to hold the cumulative effect of the partial
+ * field pedits that have been processed so far.
+ */
+struct efx_tc_mangler_state {
+	u8 dst_mac_32:1; /* eth->h_dest[0:3] */
+	u8 dst_mac_16:1; /* eth->h_dest[4:5] */
+	u8 src_mac_16:1; /* eth->h_source[0:1] */
+	u8 src_mac_32:1; /* eth->h_source[2:5] */
+	unsigned char dst_mac[ETH_ALEN];
+	unsigned char src_mac[ETH_ALEN];
+};
+
+/** efx_tc_complete_mac_mangle() - pull complete field pedits out of @mung
+ * @efx: NIC we're installing a flow rule on
+ * @act: action set (cursor) to update
+ * @mung:        accumulated partial mangles
+ * @extack:      netlink extended ack for reporting errors
+ *
+ * Check @mung to find any combinations of partial mangles that can be
+ * combined into a complete packet field edit, add that edit to @act,
+ * and consume the partial mangles from @mung.
+ */
+
+static int efx_tc_complete_mac_mangle(struct efx_nic *efx,
+				      struct efx_tc_action_set *act,
+				      struct efx_tc_mangler_state *mung,
+				      struct netlink_ext_ack *extack)
+{
+	struct efx_tc_mac_pedit_action *ped;
+
+	if (mung->dst_mac_32 && mung->dst_mac_16) {
+		ped = efx_tc_flower_get_mac(efx, mung->dst_mac, extack);
+		if (IS_ERR(ped))
+			return PTR_ERR(ped);
+
+		/* Check that we have not already populated dst_mac */
+		if (act->dst_mac)
+			efx_tc_flower_put_mac(efx, act->dst_mac);
+
+		act->dst_mac = ped;
+
+		/* consume the incomplete state */
+		mung->dst_mac_32 = 0;
+		mung->dst_mac_16 = 0;
+	}
+	if (mung->src_mac_16 && mung->src_mac_32) {
+		ped = efx_tc_flower_get_mac(efx, mung->src_mac, extack);
+		if (IS_ERR(ped))
+			return PTR_ERR(ped);
+
+		/* Check that we have not already populated src_mac */
+		if (act->src_mac)
+			efx_tc_flower_put_mac(efx, act->src_mac);
+
+		act->src_mac = ped;
+
+		/* consume the incomplete state */
+		mung->src_mac_32 = 0;
+		mung->src_mac_16 = 0;
+	}
+	return 0;
+}
+
+/**
+ * efx_tc_mangle() - handle a single 32-bit (or less) pedit
+ * @efx: NIC we're installing a flow rule on
+ * @act: action set (cursor) to update
+ * @fa:          FLOW_ACTION_MANGLE action metadata
+ * @mung:        accumulator for partial mangles
+ * @extack:      netlink extended ack for reporting errors
+ *
+ * Identify the fields written by a FLOW_ACTION_MANGLE, and record
+ * the partial mangle state in @mung.  If this mangle completes an
+ * earlier partial mangle, consume and apply to @act by calling
+ * efx_tc_complete_mac_mangle().
+ */
+
+static int efx_tc_mangle(struct efx_nic *efx, struct efx_tc_action_set *act,
+			 const struct flow_action_entry *fa,
+			 struct efx_tc_mangler_state *mung,
+			 struct netlink_ext_ack *extack)
+{
+	__le32 mac32;
+	__le16 mac16;
+
+	switch (fa->mangle.htype) {
+	case FLOW_ACT_MANGLE_HDR_TYPE_ETH:
+		BUILD_BUG_ON(offsetof(struct ethhdr, h_dest) != 0);
+		BUILD_BUG_ON(offsetof(struct ethhdr, h_source) != 6);
+		if (!efx_tc_flower_action_order_ok(act, EFX_TC_AO_PEDIT_MAC_ADDRS)) {
+			NL_SET_ERR_MSG_MOD(extack,
+					   "Pedit mangle mac action violates action order");
+			return -EOPNOTSUPP;
+		}
+		switch (fa->mangle.offset) {
+		case 0:
+			if (fa->mangle.mask) {
+				NL_SET_ERR_MSG_FMT_MOD(extack,
+						       "Unsupported: mask (%#x) of eth.dst32 mangle",
+						       fa->mangle.mask);
+				return -EOPNOTSUPP;
+			}
+			/* Ethernet address is little-endian */
+			mac32 = cpu_to_le32(fa->mangle.val);
+			memcpy(mung->dst_mac, &mac32, sizeof(mac32));
+			mung->dst_mac_32 = 1;
+			return efx_tc_complete_mac_mangle(efx, act, mung, extack);
+		case 4:
+			if (fa->mangle.mask == 0xffff) {
+				mac16 = cpu_to_le16(fa->mangle.val >> 16);
+				memcpy(mung->src_mac, &mac16, sizeof(mac16));
+				mung->src_mac_16 = 1;
+			} else if (fa->mangle.mask == 0xffff0000) {
+				mac16 = cpu_to_le16((u16)fa->mangle.val);
+				memcpy(mung->dst_mac + 4, &mac16, sizeof(mac16));
+				mung->dst_mac_16 = 1;
+			} else {
+				NL_SET_ERR_MSG_FMT_MOD(extack,
+						       "Unsupported: mask (%#x) of eth+4 mangle is not high or low 16b",
+						       fa->mangle.mask);
+				return -EOPNOTSUPP;
+			}
+			return efx_tc_complete_mac_mangle(efx, act, mung, extack);
+		case 8:
+			if (fa->mangle.mask) {
+				NL_SET_ERR_MSG_FMT_MOD(extack,
+						       "Unsupported: mask (%#x) of eth.src32 mangle",
+						       fa->mangle.mask);
+				return -EOPNOTSUPP;
+			}
+			mac32 = cpu_to_le32(fa->mangle.val);
+			memcpy(mung->src_mac + 2, &mac32, sizeof(mac32));
+			mung->src_mac_32 = 1;
+			return efx_tc_complete_mac_mangle(efx, act, mung, extack);
+		default:
+			NL_SET_ERR_MSG_FMT_MOD(extack, "Unsupported: mangle eth+%u %x/%x",
+					       fa->mangle.offset, fa->mangle.val, fa->mangle.mask);
+			return -EOPNOTSUPP;
+		}
+		break;
+	default:
+		NL_SET_ERR_MSG_FMT_MOD(extack, "Unhandled mangle htype %u for action rule",
+				       fa->mangle.htype);
+		return -EOPNOTSUPP;
+	}
+	return 0;
+}
+
+/**
+ * efx_tc_incomplete_mangle() - check for leftover partial pedits
+ * @mung:        accumulator for partial mangles
+ * @extack:      netlink extended ack for reporting errors
+ *
+ * Since the MAE can only overwrite whole fields, any partial
+ * field mangle left over on reaching packet delivery (mirred or
+ * end of TC actions) cannot be offloaded.  Check for any such
+ * and reject them with -%EOPNOTSUPP.
+ */
+
+static int efx_tc_incomplete_mangle(struct efx_tc_mangler_state *mung,
+				    struct netlink_ext_ack *extack)
+{
+	if (mung->dst_mac_32 || mung->dst_mac_16) {
+		NL_SET_ERR_MSG_MOD(extack, "Incomplete pedit of destination MAC address");
+		return -EOPNOTSUPP;
+	}
+	if (mung->src_mac_16 || mung->src_mac_32) {
+		NL_SET_ERR_MSG_MOD(extack, "Incomplete pedit of source MAC address");
+		return -EOPNOTSUPP;
+	}
+	return 0;
+}
+
 static int efx_tc_flower_replace_foreign(struct efx_nic *efx,
 					 struct net_device *net_dev,
 					 struct flow_cls_offload *tc)
@@ -1351,6 +1547,7 @@ static int efx_tc_flower_replace(struct efx_nic *efx,
 	struct netlink_ext_ack *extack = tc->common.extack;
 	const struct ip_tunnel_info *encap_info = NULL;
 	struct efx_tc_flow_rule *rule = NULL, *old;
+	struct efx_tc_mangler_state mung = {};
 	struct efx_tc_action_set *act = NULL;
 	const struct flow_action_entry *fa;
 	struct efx_rep *from_efv, *to_efv;
@@ -1687,6 +1884,11 @@ static int efx_tc_flower_replace(struct efx_nic *efx,
 			act->vlan_proto[act->vlan_push] = fa->vlan.proto;
 			act->vlan_push++;
 			break;
+		case FLOW_ACTION_MANGLE:
+			rc = efx_tc_mangle(efx, act, fa, &mung, extack);
+			if (rc < 0)
+				goto release;
+			break;
 		case FLOW_ACTION_TUNNEL_ENCAP:
 			if (encap_info) {
 				/* Can't specify encap multiple times.
@@ -1726,6 +1928,9 @@ static int efx_tc_flower_replace(struct efx_nic *efx,
 		}
 	}
 
+	rc = efx_tc_incomplete_mangle(&mung, extack);
+	if (rc < 0)
+		goto release;
 	if (act) {
 		/* Not shot/redirected, so deliver to default dest */
 		if (from_efv == EFX_EFV_PF)
-- 
2.17.1

[PATCH net-next 3/6] sfc: add decrement ttl by offloading set ipv4 ttl actions

[Pieter Jansen van Vuuren]

Offload pedit set ipv4 ttl field, where the ttl field has already been
matched and the new value is one less, by translating it to a decrement.

Co-developed-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
---
 drivers/net/ethernet/sfc/mae.c |  6 ++-
 drivers/net/ethernet/sfc/tc.c  | 74 +++++++++++++++++++++++++++++++++-
 drivers/net/ethernet/sfc/tc.h  |  2 +
 3 files changed, 78 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/sfc/mae.c b/drivers/net/ethernet/sfc/mae.c
index a7ad7ab8c5f4..c3e2b4a21d10 100644
--- a/drivers/net/ethernet/sfc/mae.c
+++ b/drivers/net/ethernet/sfc/mae.c
@@ -1291,10 +1291,12 @@ int efx_mae_alloc_action_set(struct efx_nic *efx, struct efx_tc_action_set *act)
 	size_t outlen;
 	int rc;
 
-	MCDI_POPULATE_DWORD_3(inbuf, MAE_ACTION_SET_ALLOC_IN_FLAGS,
+	MCDI_POPULATE_DWORD_4(inbuf, MAE_ACTION_SET_ALLOC_IN_FLAGS,
 			      MAE_ACTION_SET_ALLOC_IN_VLAN_PUSH, act->vlan_push,
 			      MAE_ACTION_SET_ALLOC_IN_VLAN_POP, act->vlan_pop,
-			      MAE_ACTION_SET_ALLOC_IN_DECAP, act->decap);
+			      MAE_ACTION_SET_ALLOC_IN_DECAP, act->decap,
+			      MAE_ACTION_SET_ALLOC_IN_DO_DECR_IP_TTL,
+			      act->do_ttl_dec);
 
 	if (act->src_mac)
 		MCDI_SET_DWORD(inbuf, MAE_ACTION_SET_ALLOC_IN_SRC_MAC_ID,
diff --git a/drivers/net/ethernet/sfc/tc.c b/drivers/net/ethernet/sfc/tc.c
index 47bf59529a46..03205789dd6a 100644
--- a/drivers/net/ethernet/sfc/tc.c
+++ b/drivers/net/ethernet/sfc/tc.c
@@ -31,6 +31,7 @@ enum efx_encap_type efx_tc_indr_netdev_type(struct net_device *net_dev)
 	return EFX_ENCAP_TYPE_NONE;
 }
 
+#define EFX_TC_HDR_TYPE_TTL_MASK ((u32)0xff)
 #define EFX_EFV_PF	NULL
 /* Look up the representor information (efv) for a device.
  * May return NULL for the PF (us), or an error pointer for a device that
@@ -757,6 +758,7 @@ static const char *efx_tc_encap_type_name(enum efx_encap_type typ)
 /* For details of action order constraints refer to SF-123102-TC-1§12.6.1 */
 enum efx_tc_action_order {
 	EFX_TC_AO_DECAP,
+	EFX_TC_AO_DEC_TTL,
 	EFX_TC_AO_PEDIT_MAC_ADDRS,
 	EFX_TC_AO_VLAN_POP,
 	EFX_TC_AO_VLAN_PUSH,
@@ -777,6 +779,10 @@ static bool efx_tc_flower_action_order_ok(const struct efx_tc_action_set *act,
 		 */
 		if (act->dst_mac || act->src_mac)
 			return false;
+
+		/* Decrementing ttl must not happen before DECAP */
+		if (act->do_ttl_dec)
+			return false;
 		fallthrough;
 	case EFX_TC_AO_VLAN_POP:
 		if (act->vlan_pop >= 2)
@@ -803,6 +809,10 @@ static bool efx_tc_flower_action_order_ok(const struct efx_tc_action_set *act,
 		fallthrough;
 	case EFX_TC_AO_DELIVER:
 		return !act->deliver;
+	case EFX_TC_AO_DEC_TTL:
+		if (act->encap_md)
+			return false;
+		return !act->do_ttl_dec;
 	default:
 		/* Bad caller.  Whatever they wanted to do, say they can't. */
 		WARN_ON_ONCE(1);
@@ -1059,10 +1069,12 @@ static int efx_tc_complete_mac_mangle(struct efx_nic *efx,
 static int efx_tc_mangle(struct efx_nic *efx, struct efx_tc_action_set *act,
 			 const struct flow_action_entry *fa,
 			 struct efx_tc_mangler_state *mung,
-			 struct netlink_ext_ack *extack)
+			 struct netlink_ext_ack *extack,
+			 struct efx_tc_match *match)
 {
 	__le32 mac32;
 	__le16 mac16;
+	u8 tr_ttl;
 
 	switch (fa->mangle.htype) {
 	case FLOW_ACT_MANGLE_HDR_TYPE_ETH:
@@ -1119,6 +1131,64 @@ static int efx_tc_mangle(struct efx_nic *efx, struct efx_tc_action_set *act,
 			return -EOPNOTSUPP;
 		}
 		break;
+	case FLOW_ACT_MANGLE_HDR_TYPE_IP4:
+		switch (fa->mangle.offset) {
+		case offsetof(struct iphdr, ttl):
+			/* we currently only support pedit IP4 when it applies
+			 * to TTL and then only when it can be achieved with a
+			 * decrement ttl action
+			 */
+
+			/* check that pedit applies to ttl only */
+			if (fa->mangle.mask != ~EFX_TC_HDR_TYPE_TTL_MASK) {
+				NL_SET_ERR_MSG_FMT_MOD(extack,
+						       "Unsupported: mask (%#x) out of range, only support mangle action on ipv4.ttl",
+						       fa->mangle.mask);
+				return -EOPNOTSUPP;
+			}
+
+			/* we can only convert to a dec ttl when we have an
+			 * exact match on the ttl field
+			 */
+			if (match->mask.ip_ttl != U8_MAX) {
+				NL_SET_ERR_MSG_FMT_MOD(extack,
+						       "Unsupported: only support mangle ipv4.ttl when we have an exact match on ttl, mask used for match (%#x)",
+						       match->mask.ip_ttl);
+				return -EOPNOTSUPP;
+			}
+
+			/* check that we don't try to decrement 0, which equates
+			 * to setting the ttl to 0xff
+			 */
+			if (match->value.ip_ttl == 0) {
+				NL_SET_ERR_MSG_MOD(extack,
+						   "Unsupported: we cannot decrement ttl past 0");
+				return -EOPNOTSUPP;
+			}
+
+			/* check that we do not decrement ttl twice */
+			if (!efx_tc_flower_action_order_ok(act,
+							   EFX_TC_AO_DEC_TTL)) {
+				NL_SET_ERR_MSG_MOD(extack,
+						   "Unsupported: multiple dec ttl");
+				return -EOPNOTSUPP;
+			}
+
+			/* check pedit can be achieved with decrement action */
+			tr_ttl = match->value.ip_ttl - 1;
+			if ((fa->mangle.val & EFX_TC_HDR_TYPE_TTL_MASK) == tr_ttl) {
+				act->do_ttl_dec = 1;
+				return 0;
+			}
+
+			fallthrough;
+		default:
+			NL_SET_ERR_MSG_FMT_MOD(extack,
+					       "Unsupported: only support mangle on the ttl field (offset is %u)",
+					       fa->mangle.offset);
+			return -EOPNOTSUPP;
+		}
+		break;
 	default:
 		NL_SET_ERR_MSG_FMT_MOD(extack, "Unhandled mangle htype %u for action rule",
 				       fa->mangle.htype);
@@ -1885,7 +1955,7 @@ static int efx_tc_flower_replace(struct efx_nic *efx,
 			act->vlan_push++;
 			break;
 		case FLOW_ACTION_MANGLE:
-			rc = efx_tc_mangle(efx, act, fa, &mung, extack);
+			rc = efx_tc_mangle(efx, act, fa, &mung, extack, &match);
 			if (rc < 0)
 				goto release;
 			break;
diff --git a/drivers/net/ethernet/sfc/tc.h b/drivers/net/ethernet/sfc/tc.h
index 91705411e5c5..d50f45b444a9 100644
--- a/drivers/net/ethernet/sfc/tc.h
+++ b/drivers/net/ethernet/sfc/tc.h
@@ -48,6 +48,7 @@ struct efx_tc_encap_action; /* see tc_encap_actions.h */
  * @vlan_push: the number of vlan headers to push
  * @vlan_pop: the number of vlan headers to pop
  * @decap: used to indicate a tunnel header decapsulation should take place
+ * @do_ttl_dec: used to indicate IP TTL / Hop Limit should be decremented
  * @deliver: used to indicate a deliver action should take place
  * @vlan_tci: tci fields for vlan push actions
  * @vlan_proto: ethernet types for vlan push actions
@@ -67,6 +68,7 @@ struct efx_tc_action_set {
 	u16 vlan_push:2;
 	u16 vlan_pop:2;
 	u16 decap:1;
+	u16 do_ttl_dec:1;
 	u16 deliver:1;
 	__be16 vlan_tci[2];
 	__be16 vlan_proto[2];
-- 
2.17.1

[PATCH net-next 4/6] sfc: add decrement ipv6 hop limit by offloading set hop limit actions

[Pieter Jansen van Vuuren]

Offload pedit set ipv6 hop limit, where the hop limit has already been
matched and the new value is one less, by translating it to a decrement.

Co-developed-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
---
 drivers/net/ethernet/sfc/tc.c | 59 +++++++++++++++++++++++++++++++++++
 1 file changed, 59 insertions(+)

diff --git a/drivers/net/ethernet/sfc/tc.c b/drivers/net/ethernet/sfc/tc.c
index 03205789dd6a..dc6275dee94a 100644
--- a/drivers/net/ethernet/sfc/tc.c
+++ b/drivers/net/ethernet/sfc/tc.c
@@ -32,6 +32,8 @@ enum efx_encap_type efx_tc_indr_netdev_type(struct net_device *net_dev)
 }
 
 #define EFX_TC_HDR_TYPE_TTL_MASK ((u32)0xff)
+/* Hoplimit is stored in the most significant byte in the pedit ipv6 header action */
+#define EFX_TC_HDR_TYPE_HLIMIT_MASK ~((u32)0xff000000)
 #define EFX_EFV_PF	NULL
 /* Look up the representor information (efv) for a device.
  * May return NULL for the PF (us), or an error pointer for a device that
@@ -1189,6 +1191,63 @@ static int efx_tc_mangle(struct efx_nic *efx, struct efx_tc_action_set *act,
 			return -EOPNOTSUPP;
 		}
 		break;
+	case FLOW_ACT_MANGLE_HDR_TYPE_IP6:
+		switch (fa->mangle.offset) {
+		case round_down(offsetof(struct ipv6hdr, hop_limit), 4):
+			/* we currently only support pedit IP6 when it applies
+			 * to the hoplimit and then only when it can be achieved
+			 * with a decrement hoplimit action
+			 */
+
+			/* check that pedit applies to ttl only */
+			if (fa->mangle.mask != EFX_TC_HDR_TYPE_HLIMIT_MASK) {
+				NL_SET_ERR_MSG_FMT_MOD(extack,
+						       "Unsupported: mask (%#x) out of range, only support mangle action on ipv6.hop_limit",
+						       fa->mangle.mask);
+
+				return -EOPNOTSUPP;
+			}
+
+			/* we can only convert to a dec ttl when we have an
+			 * exact match on the ttl field
+			 */
+			if (match->mask.ip_ttl != U8_MAX) {
+				NL_SET_ERR_MSG_FMT_MOD(extack,
+						       "Unsupported: only support mangle ipv6.hop_limit when we have an exact match on ttl, mask used for match (%#x)",
+						       match->mask.ip_ttl);
+				return -EOPNOTSUPP;
+			}
+
+			/* check that we don't try to decrement 0, which equates
+			 * to setting the ttl to 0xff
+			 */
+			if (match->value.ip_ttl == 0) {
+				NL_SET_ERR_MSG_MOD(extack,
+						   "Unsupported: we cannot decrement hop_limit past 0");
+				return -EOPNOTSUPP;
+			}
+
+			/* check that we do not decrement hoplimit twice */
+			if (!efx_tc_flower_action_order_ok(act,
+							   EFX_TC_AO_DEC_TTL)) {
+				NL_SET_ERR_MSG_MOD(extack,
+						   "Unsupported: multiple dec ttl");
+				return -EOPNOTSUPP;
+			}
+
+			/* check pedit can be achieved with decrement action */
+			tr_ttl = match->value.ip_ttl - 1;
+			if ((fa->mangle.val >> 24) == tr_ttl) {
+				act->do_ttl_dec = 1;
+				return 0;
+			}
+
+			fallthrough;
+		default:
+			NL_SET_ERR_MSG_FMT_MOD(extack,
+					       "Unsupported: only support mangle on the hop_limit field");
+			return -EOPNOTSUPP;
+		}
 	default:
 		NL_SET_ERR_MSG_FMT_MOD(extack, "Unhandled mangle htype %u for action rule",
 				       fa->mangle.htype);
-- 
2.17.1

[PATCH net-next 5/6] sfc: introduce pedit add actions on the ipv4 ttl field

[Pieter Jansen van Vuuren]

Introduce pedit add actions and use it to achieve decrement ttl offload.
Decrement ttl can be achieved by adding 0xff to the ttl field.

Co-developed-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
---
 drivers/net/ethernet/sfc/tc.c | 45 +++++++++++++++++++++++++++++++++++
 1 file changed, 45 insertions(+)

diff --git a/drivers/net/ethernet/sfc/tc.c b/drivers/net/ethernet/sfc/tc.c
index dc6275dee94a..8ba5816b51f2 100644
--- a/drivers/net/ethernet/sfc/tc.c
+++ b/drivers/net/ethernet/sfc/tc.c
@@ -1067,6 +1067,46 @@ static int efx_tc_complete_mac_mangle(struct efx_nic *efx,
  * earlier partial mangle, consume and apply to @act by calling
  * efx_tc_complete_mac_mangle().
  */
+static int efx_tc_pedit_add(struct efx_nic *efx, struct efx_tc_action_set *act,
+			    const struct flow_action_entry *fa,
+			    struct netlink_ext_ack *extack)
+{
+	switch (fa->mangle.htype) {
+	case FLOW_ACT_MANGLE_HDR_TYPE_IP4:
+		switch (fa->mangle.offset) {
+		case offsetof(struct iphdr, ttl):
+			/* check that pedit applies to ttl only */
+			if (fa->mangle.mask != ~EFX_TC_HDR_TYPE_TTL_MASK)
+				break;
+
+			/* Adding 0xff is equivalent to decrementing the ttl.
+			 * Other added values are not supported.
+			 */
+			if ((fa->mangle.val & EFX_TC_HDR_TYPE_TTL_MASK) != U8_MAX)
+				break;
+
+			/* check that we do not decrement ttl twice */
+			if (!efx_tc_flower_action_order_ok(act,
+							   EFX_TC_AO_DEC_TTL)) {
+				NL_SET_ERR_MSG_MOD(extack, "Unsupported: multiple dec ttl");
+				return -EOPNOTSUPP;
+			}
+			act->do_ttl_dec = 1;
+			return 0;
+		default:
+			break;
+		}
+		break;
+	default:
+		break;
+	}
+
+	NL_SET_ERR_MSG_FMT_MOD(extack,
+			       "Unsupported: ttl add action type %x %x %x/%x",
+			       fa->mangle.htype, fa->mangle.offset,
+			       fa->mangle.val, fa->mangle.mask);
+	return -EOPNOTSUPP;
+}
 
 static int efx_tc_mangle(struct efx_nic *efx, struct efx_tc_action_set *act,
 			 const struct flow_action_entry *fa,
@@ -2013,6 +2053,11 @@ static int efx_tc_flower_replace(struct efx_nic *efx,
 			act->vlan_proto[act->vlan_push] = fa->vlan.proto;
 			act->vlan_push++;
 			break;
+		case FLOW_ACTION_ADD:
+			rc = efx_tc_pedit_add(efx, act, fa, extack);
+			if (rc < 0)
+				goto release;
+			break;
 		case FLOW_ACTION_MANGLE:
 			rc = efx_tc_mangle(efx, act, fa, &mung, extack, &match);
 			if (rc < 0)
-- 
2.17.1

[PATCH net-next 6/6] sfc: extend pedit add action to handle decrement ipv6 hop limit

[Pieter Jansen van Vuuren]

Extend the pedit add actions to handle this case for ipv6. Similar to ipv4
dec ttl, decrementing ipv6 hop limit can be achieved by adding 0xff to the
hop limit field.

Co-developed-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Edward Cree <ecree.xilinx@gmail.com>
Signed-off-by: Pieter Jansen van Vuuren <pieter.jansen-van-vuuren@amd.com>
---
 drivers/net/ethernet/sfc/tc.c | 25 +++++++++++++++++++++++++
 1 file changed, 25 insertions(+)

diff --git a/drivers/net/ethernet/sfc/tc.c b/drivers/net/ethernet/sfc/tc.c
index 8ba5816b51f2..5be889f1b58f 100644
--- a/drivers/net/ethernet/sfc/tc.c
+++ b/drivers/net/ethernet/sfc/tc.c
@@ -1097,6 +1097,31 @@ static int efx_tc_pedit_add(struct efx_nic *efx, struct efx_tc_action_set *act,
 			break;
 		}
 		break;
+	case FLOW_ACT_MANGLE_HDR_TYPE_IP6:
+		switch (fa->mangle.offset) {
+		case round_down(offsetof(struct ipv6hdr, hop_limit), 4):
+			/* check that pedit applies to hoplimit only */
+			if (fa->mangle.mask != EFX_TC_HDR_TYPE_HLIMIT_MASK)
+				break;
+
+			/* Adding 0xff is equivalent to decrementing the hoplimit.
+			 * Other added values are not supported.
+			 */
+			if ((fa->mangle.val >> 24) != U8_MAX)
+				break;
+
+			/* check that we do not decrement hoplimit twice */
+			if (!efx_tc_flower_action_order_ok(act,
+							   EFX_TC_AO_DEC_TTL)) {
+				NL_SET_ERR_MSG_MOD(extack, "Unsupported: multiple dec ttl");
+				return -EOPNOTSUPP;
+			}
+			act->do_ttl_dec = 1;
+			return 0;
+		default:
+			break;
+		}
+		break;
 	default:
 		break;
 	}
-- 
2.17.1

Re: [PATCH net-next 5/6] sfc: introduce pedit add actions on the ipv4 ttl field

[kernel test robot]

Hi Pieter,

kernel test robot noticed the following build warnings:

[auto build test WARNING on net-next/main]

url:    https://github.com/intel-lab-lkp/linux/commits/Pieter-Jansen-van-Vuuren/sfc-introduce-ethernet-pedit-set-action-infrastructure/20230823-192051
base:   net-next/main
patch link:    https://lore.kernel.org/r/20230823111725.28090-6-pieter.jansen-van-vuuren%40amd.com
patch subject: [PATCH net-next 5/6] sfc: introduce pedit add actions on the ipv4 ttl field
config: alpha-allyesconfig (https://download.01.org/0day-ci/archive/20230823/202308232244.jYYwKnlV-lkp@intel.com/config)
compiler: alpha-linux-gcc (GCC) 13.2.0
reproduce: (https://download.01.org/0day-ci/archive/20230823/202308232244.jYYwKnlV-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202308232244.jYYwKnlV-lkp@intel.com/

All warnings (new ones prefixed by >>):

>> drivers/net/ethernet/sfc/tc.c:1073: warning: expecting prototype for efx_tc_mangle(). Prototype was for efx_tc_pedit_add() instead


vim +1073 drivers/net/ethernet/sfc/tc.c

  1056	
  1057	/**
  1058	 * efx_tc_mangle() - handle a single 32-bit (or less) pedit
  1059	 * @efx: NIC we're installing a flow rule on
  1060	 * @act: action set (cursor) to update
  1061	 * @fa:          FLOW_ACTION_MANGLE action metadata
  1062	 * @mung:        accumulator for partial mangles
  1063	 * @extack:      netlink extended ack for reporting errors
  1064	 *
  1065	 * Identify the fields written by a FLOW_ACTION_MANGLE, and record
  1066	 * the partial mangle state in @mung.  If this mangle completes an
  1067	 * earlier partial mangle, consume and apply to @act by calling
  1068	 * efx_tc_complete_mac_mangle().
  1069	 */
  1070	static int efx_tc_pedit_add(struct efx_nic *efx, struct efx_tc_action_set *act,
  1071				    const struct flow_action_entry *fa,
  1072				    struct netlink_ext_ack *extack)
> 1073	{
  1074		switch (fa->mangle.htype) {
  1075		case FLOW_ACT_MANGLE_HDR_TYPE_IP4:
  1076			switch (fa->mangle.offset) {
  1077			case offsetof(struct iphdr, ttl):
  1078				/* check that pedit applies to ttl only */
  1079				if (fa->mangle.mask != ~EFX_TC_HDR_TYPE_TTL_MASK)
  1080					break;
  1081	
  1082				/* Adding 0xff is equivalent to decrementing the ttl.
  1083				 * Other added values are not supported.
  1084				 */
  1085				if ((fa->mangle.val & EFX_TC_HDR_TYPE_TTL_MASK) != U8_MAX)
  1086					break;
  1087	
  1088				/* check that we do not decrement ttl twice */
  1089				if (!efx_tc_flower_action_order_ok(act,
  1090								   EFX_TC_AO_DEC_TTL)) {
  1091					NL_SET_ERR_MSG_MOD(extack, "Unsupported: multiple dec ttl");
  1092					return -EOPNOTSUPP;
  1093				}
  1094				act->do_ttl_dec = 1;
  1095				return 0;
  1096			default:
  1097				break;
  1098			}
  1099			break;
  1100		default:
  1101			break;
  1102		}
  1103	
  1104		NL_SET_ERR_MSG_FMT_MOD(extack,
  1105				       "Unsupported: ttl add action type %x %x %x/%x",
  1106				       fa->mangle.htype, fa->mangle.offset,
  1107				       fa->mangle.val, fa->mangle.mask);
  1108		return -EOPNOTSUPP;
  1109	}
  1110	

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

Re: [PATCH net-next 3/6] sfc: add decrement ttl by offloading set ipv4 ttl actions

[kernel test robot]

Hi Pieter,

kernel test robot noticed the following build warnings:

[auto build test WARNING on net-next/main]

url:    https://github.com/intel-lab-lkp/linux/commits/Pieter-Jansen-van-Vuuren/sfc-introduce-ethernet-pedit-set-action-infrastructure/20230823-192051
base:   net-next/main
patch link:    https://lore.kernel.org/r/20230823111725.28090-4-pieter.jansen-van-vuuren%40amd.com
patch subject: [PATCH net-next 3/6] sfc: add decrement ttl by offloading set ipv4 ttl actions
config: alpha-allyesconfig (https://download.01.org/0day-ci/archive/20230823/202308232237.WkCLb0PW-lkp@intel.com/config)
compiler: alpha-linux-gcc (GCC) 13.2.0
reproduce: (https://download.01.org/0day-ci/archive/20230823/202308232237.WkCLb0PW-lkp@intel.com/reproduce)

If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202308232237.WkCLb0PW-lkp@intel.com/

All warnings (new ones prefixed by >>):

>> drivers/net/ethernet/sfc/tc.c:1074: warning: Function parameter or member 'match' not described in 'efx_tc_mangle'


vim +1074 drivers/net/ethernet/sfc/tc.c

12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1054  
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1055  /**
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1056   * efx_tc_mangle() - handle a single 32-bit (or less) pedit
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1057   * @efx: NIC we're installing a flow rule on
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1058   * @act: action set (cursor) to update
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1059   * @fa:          FLOW_ACTION_MANGLE action metadata
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1060   * @mung:        accumulator for partial mangles
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1061   * @extack:      netlink extended ack for reporting errors
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1062   *
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1063   * Identify the fields written by a FLOW_ACTION_MANGLE, and record
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1064   * the partial mangle state in @mung.  If this mangle completes an
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1065   * earlier partial mangle, consume and apply to @act by calling
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1066   * efx_tc_complete_mac_mangle().
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1067   */
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1068  
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1069  static int efx_tc_mangle(struct efx_nic *efx, struct efx_tc_action_set *act,
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1070  			 const struct flow_action_entry *fa,
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1071  			 struct efx_tc_mangler_state *mung,
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1072  			 struct netlink_ext_ack *extack,
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1073  			 struct efx_tc_match *match)
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23 @1074  {
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1075  	__le32 mac32;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1076  	__le16 mac16;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1077  	u8 tr_ttl;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1078  
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1079  	switch (fa->mangle.htype) {
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1080  	case FLOW_ACT_MANGLE_HDR_TYPE_ETH:
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1081  		BUILD_BUG_ON(offsetof(struct ethhdr, h_dest) != 0);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1082  		BUILD_BUG_ON(offsetof(struct ethhdr, h_source) != 6);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1083  		if (!efx_tc_flower_action_order_ok(act, EFX_TC_AO_PEDIT_MAC_ADDRS)) {
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1084  			NL_SET_ERR_MSG_MOD(extack,
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1085  					   "Pedit mangle mac action violates action order");
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1086  			return -EOPNOTSUPP;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1087  		}
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1088  		switch (fa->mangle.offset) {
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1089  		case 0:
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1090  			if (fa->mangle.mask) {
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1091  				NL_SET_ERR_MSG_FMT_MOD(extack,
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1092  						       "Unsupported: mask (%#x) of eth.dst32 mangle",
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1093  						       fa->mangle.mask);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1094  				return -EOPNOTSUPP;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1095  			}
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1096  			/* Ethernet address is little-endian */
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1097  			mac32 = cpu_to_le32(fa->mangle.val);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1098  			memcpy(mung->dst_mac, &mac32, sizeof(mac32));
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1099  			mung->dst_mac_32 = 1;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1100  			return efx_tc_complete_mac_mangle(efx, act, mung, extack);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1101  		case 4:
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1102  			if (fa->mangle.mask == 0xffff) {
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1103  				mac16 = cpu_to_le16(fa->mangle.val >> 16);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1104  				memcpy(mung->src_mac, &mac16, sizeof(mac16));
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1105  				mung->src_mac_16 = 1;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1106  			} else if (fa->mangle.mask == 0xffff0000) {
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1107  				mac16 = cpu_to_le16((u16)fa->mangle.val);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1108  				memcpy(mung->dst_mac + 4, &mac16, sizeof(mac16));
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1109  				mung->dst_mac_16 = 1;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1110  			} else {
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1111  				NL_SET_ERR_MSG_FMT_MOD(extack,
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1112  						       "Unsupported: mask (%#x) of eth+4 mangle is not high or low 16b",
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1113  						       fa->mangle.mask);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1114  				return -EOPNOTSUPP;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1115  			}
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1116  			return efx_tc_complete_mac_mangle(efx, act, mung, extack);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1117  		case 8:
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1118  			if (fa->mangle.mask) {
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1119  				NL_SET_ERR_MSG_FMT_MOD(extack,
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1120  						       "Unsupported: mask (%#x) of eth.src32 mangle",
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1121  						       fa->mangle.mask);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1122  				return -EOPNOTSUPP;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1123  			}
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1124  			mac32 = cpu_to_le32(fa->mangle.val);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1125  			memcpy(mung->src_mac + 2, &mac32, sizeof(mac32));
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1126  			mung->src_mac_32 = 1;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1127  			return efx_tc_complete_mac_mangle(efx, act, mung, extack);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1128  		default:
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1129  			NL_SET_ERR_MSG_FMT_MOD(extack, "Unsupported: mangle eth+%u %x/%x",
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1130  					       fa->mangle.offset, fa->mangle.val, fa->mangle.mask);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1131  			return -EOPNOTSUPP;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1132  		}
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1133  		break;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1134  	case FLOW_ACT_MANGLE_HDR_TYPE_IP4:
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1135  		switch (fa->mangle.offset) {
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1136  		case offsetof(struct iphdr, ttl):
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1137  			/* we currently only support pedit IP4 when it applies
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1138  			 * to TTL and then only when it can be achieved with a
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1139  			 * decrement ttl action
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1140  			 */
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1141  
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1142  			/* check that pedit applies to ttl only */
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1143  			if (fa->mangle.mask != ~EFX_TC_HDR_TYPE_TTL_MASK) {
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1144  				NL_SET_ERR_MSG_FMT_MOD(extack,
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1145  						       "Unsupported: mask (%#x) out of range, only support mangle action on ipv4.ttl",
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1146  						       fa->mangle.mask);
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1147  				return -EOPNOTSUPP;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1148  			}
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1149  
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1150  			/* we can only convert to a dec ttl when we have an
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1151  			 * exact match on the ttl field
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1152  			 */
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1153  			if (match->mask.ip_ttl != U8_MAX) {
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1154  				NL_SET_ERR_MSG_FMT_MOD(extack,
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1155  						       "Unsupported: only support mangle ipv4.ttl when we have an exact match on ttl, mask used for match (%#x)",
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1156  						       match->mask.ip_ttl);
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1157  				return -EOPNOTSUPP;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1158  			}
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1159  
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1160  			/* check that we don't try to decrement 0, which equates
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1161  			 * to setting the ttl to 0xff
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1162  			 */
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1163  			if (match->value.ip_ttl == 0) {
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1164  				NL_SET_ERR_MSG_MOD(extack,
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1165  						   "Unsupported: we cannot decrement ttl past 0");
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1166  				return -EOPNOTSUPP;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1167  			}
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1168  
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1169  			/* check that we do not decrement ttl twice */
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1170  			if (!efx_tc_flower_action_order_ok(act,
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1171  							   EFX_TC_AO_DEC_TTL)) {
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1172  				NL_SET_ERR_MSG_MOD(extack,
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1173  						   "Unsupported: multiple dec ttl");
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1174  				return -EOPNOTSUPP;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1175  			}
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1176  
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1177  			/* check pedit can be achieved with decrement action */
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1178  			tr_ttl = match->value.ip_ttl - 1;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1179  			if ((fa->mangle.val & EFX_TC_HDR_TYPE_TTL_MASK) == tr_ttl) {
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1180  				act->do_ttl_dec = 1;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1181  				return 0;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1182  			}
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1183  
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1184  			fallthrough;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1185  		default:
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1186  			NL_SET_ERR_MSG_FMT_MOD(extack,
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1187  					       "Unsupported: only support mangle on the ttl field (offset is %u)",
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1188  					       fa->mangle.offset);
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1189  			return -EOPNOTSUPP;
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1190  		}
b1a793d7986335 Pieter Jansen van Vuuren 2023-08-23  1191  		break;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1192  	default:
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1193  		NL_SET_ERR_MSG_FMT_MOD(extack, "Unhandled mangle htype %u for action rule",
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1194  				       fa->mangle.htype);
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1195  		return -EOPNOTSUPP;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1196  	}
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1197  	return 0;
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1198  }
12322dfaf48dc2 Pieter Jansen van Vuuren 2023-08-23  1199  

-- 
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki

Re: [PATCH net-next 5/6] sfc: introduce pedit add actions on the ipv4 ttl field

[Pieter Jansen van Vuuren]

On 23/08/2023 15:25, kernel test robot wrote:
> Hi Pieter,
> 
> kernel test robot noticed the following build warnings:
> 
> [auto build test WARNING on net-next/main]
> 
> url:    https://github.com/intel-lab-lkp/linux/commits/Pieter-Jansen-van-Vuuren/sfc-introduce-ethernet-pedit-set-action-infrastructure/20230823-192051
> base:   net-next/main
> patch link:    https://lore.kernel.org/r/20230823111725.28090-6-pieter.jansen-van-vuuren%40amd.com
> patch subject: [PATCH net-next 5/6] sfc: introduce pedit add actions on the ipv4 ttl field
> config: alpha-allyesconfig (https://download.01.org/0day-ci/archive/20230823/202308232244.jYYwKnlV-lkp@intel.com/config)
> compiler: alpha-linux-gcc (GCC) 13.2.0
> reproduce: (https://download.01.org/0day-ci/archive/20230823/202308232244.jYYwKnlV-lkp@intel.com/reproduce)
> 
> If you fix the issue in a separate patch/commit (i.e. not just a new version of
> the same patch/commit), kindly add following tags
> | Reported-by: kernel test robot <lkp@intel.com>
> | Closes: https://lore.kernel.org/oe-kbuild-all/202308232244.jYYwKnlV-lkp@intel.com/
> 
> All warnings (new ones prefixed by >>):
> 
>>> drivers/net/ethernet/sfc/tc.c:1073: warning: expecting prototype for efx_tc_mangle(). Prototype was for efx_tc_pedit_add() instead

Sorry, I messed up the kdoc while rebasing here. I will get this fixed in v2.