Skip to contentSkip to conversation

bpf: zero-initialize the fib lookup flow struct

PATCH Applied Avinash Duduskar 3 participants 3 messages Jun 17 – Jun 22

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 bpf: zero-initialize the fib lookup flow struct

bpf_ipv4_fib_lookup() and bpf_ipv6_fib_lookup() build the flow key on the stack with a bare "struct flowi4 fl4;" / "struct flowi6 fl6;" and fill it field by field, but never set flowi4_l3mdev / flowi6_l3mdev.

On the non-DIRECT path the lookup goes through the fib rules whenever the netns has custom rules, which a VRF installs:

	bpf_ipv4_fib_lookup() -> fib_lookup() -> __fib_lookup()
	  -> l3mdev_update_flow()   reads !fl->flowi_l3mdev
	  -> fib_rules_lookup() -> fib_rule_match()
	       -> l3mdev_fib_rule_match()   uses fl->flowi_l3mdev

l3mdev_update_flow() resolves the l3mdev master from the ingress device only while the field is still zero. Left at a nonzero stack value the resolution is skipped, and l3mdev_fib_rule_match() then tests that value as an ifindex, so the VRF master is not resolved and the rule fails to
match: an ingress enslaved to a VRF can fail to select its table. FIB rules matching on an L3 master device (l3mdev_fib_rule_iif_match()/ _oif_match()) read the same value, so an "ip rule iif/oif <vrf>" mismatches the same way.

Zero-initialize the whole flow struct rather than adding one more field assignment, so any flowi field added later is covered too. ip_route_input_slow() likewise zeroes the field before its input lookup.

CONFIG_INIT_STACK_ALL_ZERO masks this by default, but it depends on compiler support (CC_HAS_AUTO_VAR_INIT_ZERO), so INIT_STACK_NONE builds, including older toolchains that fall back to it, are exposed. Built with INIT_STACK_ALL_PATTERN, a plain bpf_fib_lookup (no VLAN, no DIRECT) over a VRF slave whose destination is routed only in the VRF table returns BPF_FIB_LKUP_RET_NOT_FWDED, and resolves with this patch. On the default config the lookup succeeds either way, so ordinary testing does not catch the bug.

Fixes 40867d74c374 ("net: Add l3mdev index to flow struct and avoid oif reset for port devices")
---
 net/core/filter.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
net/core/filter.c +2 -2
index 9590877b0714..7c58df589826 100644
@@ -6139,7 +6139,7 @@ static int bpf_ipv4_fib_lookup(struct net *net, struct bpf_fib_lookup *params,
 	struct in_device *in_dev;
 	struct net_device *dev;
 	struct fib_result res;
-	struct flowi4 fl4;
+	struct flowi4 fl4 = {};
 	u32 mtu = 0;
 	int err;
 
@@ -6279,7 +6279,7 @@ static int bpf_ipv6_fib_lookup(struct net *net, struct bpf_fib_lookup *params,
 	struct neighbour *neigh;
 	struct net_device *dev;
 	struct inet6_dev *idev;
-	struct flowi6 fl6;
+	struct flowi6 fl6 = {};
 	int strict = 0;
 	int oif, err;
 	u32 mtu = 0;
Signature · 1 line
2.54.0
Reviewed-by on v1 1/1
#1 lore raw
Quoted 34 lines · Avinash Duduskar

bpf_ipv4_fib_lookup() and bpf_ipv6_fib_lookup() build the flow key on the stack with a bare "struct flowi4 fl4;" / "struct flowi6 fl6;" and fill it field by field, but never set flowi4_l3mdev / flowi6_l3mdev.

On the non-DIRECT path the lookup goes through the fib rules whenever the netns has custom rules, which a VRF installs:

	bpf_ipv4_fib_lookup() -> fib_lookup() -> __fib_lookup()
	  -> l3mdev_update_flow()   reads !fl->flowi_l3mdev
	  -> fib_rules_lookup() -> fib_rule_match()
	       -> l3mdev_fib_rule_match()   uses fl->flowi_l3mdev

l3mdev_update_flow() resolves the l3mdev master from the ingress device only while the field is still zero. Left at a nonzero stack value the resolution is skipped, and l3mdev_fib_rule_match() then tests that value as an ifindex, so the VRF master is not resolved and the rule fails to
match: an ingress enslaved to a VRF can fail to select its table. FIB rules matching on an L3 master device (l3mdev_fib_rule_iif_match()/ _oif_match()) read the same value, so an "ip rule iif/oif <vrf>" mismatches the same way.

Zero-initialize the whole flow struct rather than adding one more field assignment, so any flowi field added later is covered too. ip_route_input_slow() likewise zeroes the field before its input lookup.

CONFIG_INIT_STACK_ALL_ZERO masks this by default, but it depends on compiler support (CC_HAS_AUTO_VAR_INIT_ZERO), so INIT_STACK_NONE builds, including older toolchains that fall back to it, are exposed. Built with INIT_STACK_ALL_PATTERN, a plain bpf_fib_lookup (no VLAN, no DIRECT) over a VRF slave whose destination is routed only in the VRF table returns BPF_FIB_LKUP_RET_NOT_FWDED, and resolves with this patch. On the default config the lookup succeeds either way, so ordinary testing does not catch the bug.

Quoted 2 lines
Fixes 40867d74c374 ("net: Add l3mdev index to flow struct and avoid oif reset for port devices")
#2 lore raw
In reply to Avinash Duduskar
This patch was applied to bpf/bpf.git (master) by Alexei Starovoitov <[email protected]>:

Hello:

This patch was applied to bpf/bpf.git (master)
by Alexei Starovoitov <[email protected]>:

Quoted 8 lines · On Thu, 18 Jun 2026 04:17:19 +0530 you wrote:

bpf_ipv4_fib_lookup() and bpf_ipv6_fib_lookup() build the flow key on the stack with a bare "struct flowi4 fl4;" / "struct flowi6 fl6;" and fill it field by field, but never set flowi4_l3mdev / flowi6_l3mdev.

On the non-DIRECT path the lookup goes through the fib rules whenever the netns has custom rules, which a VRF installs:

[...]

Here is the summary with links:
- [bpf] bpf: zero-initialize the fib lookup flow struct

    https://git.kernel.org/bpf/bpf/c/0dfcb68a6a5a

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html