[PATCH net 0/3] Fixes for nfp pre_tunnel code

[PATCH net 0/3] Fixes for nfp pre_tunnel code

[Simon Horman]

Louis Peens says:

The following set of patches fixes up a few bugs in the pre_tun
decap code paths which has been hiding for a while.

Louis Peens (3):
  nfp: flower: fix unsupported pre_tunnel flows
  nfp: flower: add ipv6 bit to pre_tunnel control message
  nfp: flower: fix pre_tun mask id allocation

 .../ethernet/netronome/nfp/flower/metadata.c  | 24 +++++++++++++------
 .../ethernet/netronome/nfp/flower/offload.c   | 18 ++++++++++++++
 .../netronome/nfp/flower/tunnel_conf.c        | 15 ++++++++++--
 3 files changed, 48 insertions(+), 9 deletions(-)

-- 
2.20.1

[PATCH net 1/3] nfp: flower: fix unsupported pre_tunnel flows

[Simon Horman]

From: Louis Peens <louis.peens@corigine.com>

There are some pre_tunnel flows combinations which are incorrectly being
offloaded without proper support, fix these.

- Matching on MPLS is not supported for pre_tun.
- Match on IPv4/IPv6 layer must be present.
- Destination MAC address must match pre_tun.dev MAC

Fixes: 120ffd84a9ec ("nfp: flower: verify pre-tunnel rules")
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
---
 .../ethernet/netronome/nfp/flower/offload.c    | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)

diff --git a/drivers/net/ethernet/netronome/nfp/flower/offload.c b/drivers/net/ethernet/netronome/nfp/flower/offload.c
index 1c59aff2163c..d72225d64a75 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/offload.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/offload.c
@@ -1142,6 +1142,12 @@ nfp_flower_validate_pre_tun_rule(struct nfp_app *app,
 		return -EOPNOTSUPP;
 	}
 
+	if (!(key_layer & NFP_FLOWER_LAYER_IPV4) &&
+	    !(key_layer & NFP_FLOWER_LAYER_IPV6)) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: match on ipv4/ipv6 eth_type must be present");
+		return -EOPNOTSUPP;
+	}
+
 	/* Skip fields known to exist. */
 	mask += sizeof(struct nfp_flower_meta_tci);
 	ext += sizeof(struct nfp_flower_meta_tci);
@@ -1152,6 +1158,13 @@ nfp_flower_validate_pre_tun_rule(struct nfp_app *app,
 	mask += sizeof(struct nfp_flower_in_port);
 	ext += sizeof(struct nfp_flower_in_port);
 
+	/* Ensure destination MAC address matches pre_tun_dev. */
+	mac = (struct nfp_flower_mac_mpls *)ext;
+	if (memcmp(&mac->mac_dst[0], flow->pre_tun_rule.dev->dev_addr, 6)) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: dest MAC must match output dev MAC");
+		return -EOPNOTSUPP;
+	}
+
 	/* Ensure destination MAC address is fully matched. */
 	mac = (struct nfp_flower_mac_mpls *)mask;
 	if (!is_broadcast_ether_addr(&mac->mac_dst[0])) {
@@ -1159,6 +1172,11 @@ nfp_flower_validate_pre_tun_rule(struct nfp_app *app,
 		return -EOPNOTSUPP;
 	}
 
+	if (mac->mpls_lse) {
+		NL_SET_ERR_MSG_MOD(extack, "unsupported pre-tunnel rule: MPLS not supported");
+		return -EOPNOTSUPP;
+	}
+
 	mask += sizeof(struct nfp_flower_mac_mpls);
 	ext += sizeof(struct nfp_flower_mac_mpls);
 	if (key_layer & NFP_FLOWER_LAYER_IPV4 ||
-- 
2.20.1

[PATCH net 2/3] nfp: flower: add ipv6 bit to pre_tunnel control message

[Simon Horman]

From: Louis Peens <louis.peens@corigine.com>

Differentiate between ipv4 and ipv6 flows when configuring the pre_tunnel
table to prevent them trampling each other in the table.

Fixes: 783461604f7e ("nfp: flower: update flow merge code to support IPv6 tunnels")
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
---
 .../ethernet/netronome/nfp/flower/tunnel_conf.c   | 15 +++++++++++++--
 1 file changed, 13 insertions(+), 2 deletions(-)

diff --git a/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c b/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c
index 7248d248f604..d19c02e99114 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c
@@ -16,8 +16,9 @@
 #define NFP_FL_MAX_ROUTES               32
 
 #define NFP_TUN_PRE_TUN_RULE_LIMIT	32
-#define NFP_TUN_PRE_TUN_RULE_DEL	0x1
-#define NFP_TUN_PRE_TUN_IDX_BIT		0x8
+#define NFP_TUN_PRE_TUN_RULE_DEL	BIT(0)
+#define NFP_TUN_PRE_TUN_IDX_BIT		BIT(3)
+#define NFP_TUN_PRE_TUN_IPV6_BIT	BIT(7)
 
 /**
  * struct nfp_tun_pre_run_rule - rule matched before decap
@@ -1268,6 +1269,7 @@ int nfp_flower_xmit_pre_tun_flow(struct nfp_app *app,
 {
 	struct nfp_flower_priv *app_priv = app->priv;
 	struct nfp_tun_offloaded_mac *mac_entry;
+	struct nfp_flower_meta_tci *key_meta;
 	struct nfp_tun_pre_tun_rule payload;
 	struct net_device *internal_dev;
 	int err;
@@ -1290,6 +1292,15 @@ int nfp_flower_xmit_pre_tun_flow(struct nfp_app *app,
 	if (!mac_entry)
 		return -ENOENT;
 
+	/* Set/clear IPV6 bit. cpu_to_be16() swap will lead to MSB being
+	 * set/clear for port_idx.
+	 */
+	key_meta = (struct nfp_flower_meta_tci *)flow->unmasked_data;
+	if (key_meta->nfp_flow_key_layer & NFP_FLOWER_LAYER_IPV6)
+		mac_entry->index |= NFP_TUN_PRE_TUN_IPV6_BIT;
+	else
+		mac_entry->index &= ~NFP_TUN_PRE_TUN_IPV6_BIT;
+
 	payload.port_idx = cpu_to_be16(mac_entry->index);
 
 	/* Copy mac id and vlan to flow - dev may not exist at delete time. */
-- 
2.20.1

[PATCH net 3/3] nfp: flower: fix pre_tun mask id allocation

[Simon Horman]

From: Louis Peens <louis.peens@corigine.com>

pre_tun_rule flows does not follow the usual add-flow path, instead
they are used to update the pre_tun table on the firmware. This means
that if the mask-id gets allocated here the firmware will never see the
"NFP_FL_META_FLAG_MANAGE_MASK" flag for the specific mask id, which
triggers the allocation on the firmware side. This leads to the firmware
mask being corrupted and causing all sorts of strange behaviour.

Fixes: f12725d98cbe ("nfp: flower: offload pre-tunnel rules")
Signed-off-by: Louis Peens <louis.peens@corigine.com>
Signed-off-by: Simon Horman <simon.horman@netronome.com>
---
 .../ethernet/netronome/nfp/flower/metadata.c  | 24 +++++++++++++------
 1 file changed, 17 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/netronome/nfp/flower/metadata.c b/drivers/net/ethernet/netronome/nfp/flower/metadata.c
index 5defd31d481c..aa06fcb38f8b 100644
--- a/drivers/net/ethernet/netronome/nfp/flower/metadata.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/metadata.c
@@ -327,8 +327,14 @@ int nfp_compile_flow_metadata(struct nfp_app *app,
 		goto err_free_ctx_entry;
 	}
 
+	/* Do net allocate a mask-id for pre_tun_rules. These flows are used to
+	 * configure the pre_tun table and are never actually send to the
+	 * firmware as an add-flow message. This causes the mask-id allocation
+	 * on the firmware to get out of sync if allocated here.
+	 */
 	new_mask_id = 0;
-	if (!nfp_check_mask_add(app, nfp_flow->mask_data,
+	if (!nfp_flow->pre_tun_rule.dev &&
+	    !nfp_check_mask_add(app, nfp_flow->mask_data,
 				nfp_flow->meta.mask_len,
 				&nfp_flow->meta.flags, &new_mask_id)) {
 		NL_SET_ERR_MSG_MOD(extack, "invalid entry: cannot allocate a new mask id");
@@ -359,7 +365,8 @@ int nfp_compile_flow_metadata(struct nfp_app *app,
 			goto err_remove_mask;
 		}
 
-		if (!nfp_check_mask_remove(app, nfp_flow->mask_data,
+		if (!nfp_flow->pre_tun_rule.dev &&
+		    !nfp_check_mask_remove(app, nfp_flow->mask_data,
 					   nfp_flow->meta.mask_len,
 					   NULL, &new_mask_id)) {
 			NL_SET_ERR_MSG_MOD(extack, "invalid entry: cannot release mask id");
@@ -374,8 +381,10 @@ int nfp_compile_flow_metadata(struct nfp_app *app,
 	return 0;
 
 err_remove_mask:
-	nfp_check_mask_remove(app, nfp_flow->mask_data, nfp_flow->meta.mask_len,
-			      NULL, &new_mask_id);
+	if (!nfp_flow->pre_tun_rule.dev)
+		nfp_check_mask_remove(app, nfp_flow->mask_data,
+				      nfp_flow->meta.mask_len,
+				      NULL, &new_mask_id);
 err_remove_rhash:
 	WARN_ON_ONCE(rhashtable_remove_fast(&priv->stats_ctx_table,
 					    &ctx_entry->ht_node,
@@ -406,9 +415,10 @@ int nfp_modify_flow_metadata(struct nfp_app *app,
 
 	__nfp_modify_flow_metadata(priv, nfp_flow);
 
-	nfp_check_mask_remove(app, nfp_flow->mask_data,
-			      nfp_flow->meta.mask_len, &nfp_flow->meta.flags,
-			      &new_mask_id);
+	if (!nfp_flow->pre_tun_rule.dev)
+		nfp_check_mask_remove(app, nfp_flow->mask_data,
+				      nfp_flow->meta.mask_len, &nfp_flow->meta.flags,
+				      &new_mask_id);
 
 	/* Update flow payload with mask ids. */
 	nfp_flow->unmasked_data[NFP_FL_MASK_ID_LOCATION] = new_mask_id;
-- 
2.20.1

Re: [PATCH net 0/3] Fixes for nfp pre_tunnel code

[Jakub Kicinski]

On Tue, 16 Mar 2021 19:13:07 +0100 Simon Horman wrote:
> Louis Peens says:
> 
> The following set of patches fixes up a few bugs in the pre_tun
> decap code paths which has been hiding for a while.

Acked-by: Jakub Kicinski <kuba@kernel.org>

Re: [PATCH net 0/3] Fixes for nfp pre_tunnel code

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net.git (refs/heads/master):

On Tue, 16 Mar 2021 19:13:07 +0100 you wrote:
> Louis Peens says:
> 
> The following set of patches fixes up a few bugs in the pre_tun
> decap code paths which has been hiding for a while.
> 
> Louis Peens (3):
>   nfp: flower: fix unsupported pre_tunnel flows
>   nfp: flower: add ipv6 bit to pre_tunnel control message
>   nfp: flower: fix pre_tun mask id allocation
> 
> [...]

Here is the summary with links:
  - [net,1/3] nfp: flower: fix unsupported pre_tunnel flows
    https://git.kernel.org/netdev/net/c/982e5ee23d76
  - [net,2/3] nfp: flower: add ipv6 bit to pre_tunnel control message
    https://git.kernel.org/netdev/net/c/5c4f5e19d6a8
  - [net,3/3] nfp: flower: fix pre_tun mask id allocation
    https://git.kernel.org/netdev/net/c/d8ce0275e45e

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html