bpf: Add BPF_FIB_LOOKUP_VLAN flag to bpf_fib_lookup() helper
Subsystems (4)
PATCH 1/1 bpf: Add BPF_FIB_LOOKUP_VLAN flag to bpf_fib_lookup() helper
bpf_fib_lookup() returns the FIB-resolved egress ifindex straight from the fib result. When the egress is a VLAN device, the returned ifindex is the VLAN netdev's, which has no XDP xmit handler; XDP programs that want to forward the frame (e.g. xdp-forward) must instead target the underlying physical device and push the VLAN tag themselves. Today the program has no way to learn either the underlying ifindex or the VLAN tag without maintaining its own VLAN-to-ifindex map in userspace and refreshing it on netlink events.
Add BPF_FIB_LOOKUP_VLAN. When the caller sets this flag and the fib result is a VLAN device, populate the existing output fields params->h_vlan_proto and params->h_vlan_TCI from the VLAN device, and replace params->ifindex with the underlying real device's ifindex. params->h_vlan_TCI carries the VID only, with PCP and DEI bits zero; a consumer wanting to set egress priority writes PCP itself. Only the immediate parent is resolved; stacked VLANs (QinQ) are not walked. When the flag is not set, behaviour is unchanged:h_vlan_proto and h_vlan_TCI are zeroed and ifindex is left at the FIB result.
This lets an XDP redirect target the physical device and learn the tag to push in a single lookup, which xdp-forward's optional VLAN mode (xdp-project/xdp-tools#504) wants from the kernel side.
The change extends bpf_fib_set_fwd_params() to take the egress dev and the lookup flags so the VLAN swap happens in the same place the vlan output fields are zeroed by default. Both IPv4 and IPv6 callers pass through. The helper's input semantics are unchanged. Under !CONFIG_VLAN_8021Q, is_vlan_dev() returns false and the new block is a no-op.
--- include/uapi/linux/bpf.h | 21 ++++++++++++++++++++- net/core/filter.c | 27 +++++++++++++++++++++++---- tools/include/uapi/linux/bpf.h | 21 ++++++++++++++++++++- 3 files changed, 63 insertions(+), 6 deletions(-)
include/uapi/linux/bpf.h +20 -1
@@ -3527,6 +3527,19 @@ union bpf_attr {* Use the mark present in *params*->mark for the fib lookup. * This option should not be used with BPF_FIB_LOOKUP_DIRECT, * as it only has meaning for full lookups. +* **BPF_FIB_LOOKUP_VLAN** +* If the fib lookup resolves to a VLAN device, set +* *params*->h_vlan_proto and *params*->h_vlan_TCI from +* the VLAN device and replace *params*->ifindex with the +* underlying real device's ifindex. This lets XDP +* programs that target the underlying physical device +* (VLAN devices have no XDP xmit) discover both the +* real egress ifindex and the VLAN tag to push in one +* call. *params*->h_vlan_TCI carries the VID only, +* with PCP and DEI bits zero; a consumer wanting to +* set egress priority writes PCP itself. Only the +* immediate parent is resolved; stacked VLANs (QinQ) +* are not walked. * * *ctx* is either **struct xdp_md** for XDP programs or * **struct sk_buff** tc cls_act programs.
@@ -7322,6 +7335,7 @@ enum {BPF_FIB_LOOKUP_TBID = (1U <<3 ),BPF_FIB_LOOKUP_SRC = (1U <<4 ),BPF_FIB_LOOKUP_MARK = (1U <<5 ), + BPF_FIB_LOOKUP_VLAN = (1U << 6), }; enum {
@@ -7388,7 +7402,12 @@ struct bpf_fib_lookup {union {struct { -/* output */ +/* output: only populated with BPF_FIB_LOOKUP_VLAN +* when the resolved egress is a VLAN device, in +* which case *ifindex* is replaced with the +* underlying real device's ifindex. Otherwise +* both fields are zeroed. +*/ __be16 h_vlan_proto;__be16 h_vlan_TCI; };
net/core/filter.c +23 -4
@@ -6119,10 +6119,28 @@ static const struct bpf_func_proto bpf_skb_get_xfrm_state_proto = {#endif #if IS_ENABLED (CONFIG_INET ) ||IS_ENABLED (CONFIG_IPV6 ) -static int bpf_fib_set_fwd_params (struct bpf_fib_lookup *params,u32 mtu) +static int bpf_fib_set_fwd_params (struct net_device *dev, +struct bpf_fib_lookup *params, +u32 flags,u32 mtu) { params->h_vlan_TCI =0 ; params->h_vlan_proto =0 ; + +if ((flags &BPF_FIB_LOOKUP_VLAN ) &&is_vlan_dev (dev)) { +struct net_device *real_dev =vlan_dev_real_dev (dev); + +/* Only the immediate parent is resolved; stacked VLANs +* (QinQ) are not walked, and a NULL real_dev (which +* is_vlan_dev() rules out in practice) keeps the +* original ifindex. +*/ +if (real_dev) { + params->h_vlan_proto =vlan_dev_vlan_proto (dev); + params->h_vlan_TCI =htons (vlan_dev_vlan_id (dev)); + params->ifindex = real_dev->ifindex; + } + } +if (mtu) params->mtu_result = mtu;/* union with tot_len */
@@ -6265,7 +6283,7 @@ static int bpf_ipv4_fib_lookup(struct net *net, struct bpf_fib_lookup *params,memcpy (params->smac, dev->dev_addr,ETH_ALEN ); set_fwd_params: - return bpf_fib_set_fwd_params(params, mtu); + return bpf_fib_set_fwd_params(dev, params, flags, mtu); }#endif
@@ -6404,13 +6422,14 @@ static int bpf_ipv6_fib_lookup(struct net *net, struct bpf_fib_lookup *params,memcpy (params->smac, dev->dev_addr,ETH_ALEN ); set_fwd_params: -return bpf_fib_set_fwd_params (params, mtu); +return bpf_fib_set_fwd_params (dev, params, flags, mtu); }#endif #define BPF_FIB_LOOKUP_MASK (BPF_FIB_LOOKUP_DIRECT | BPF_FIB_LOOKUP_OUTPUT | \ BPF_FIB_LOOKUP_SKIP_NEIGH | BPF_FIB_LOOKUP_TBID | \ - BPF_FIB_LOOKUP_SRC | BPF_FIB_LOOKUP_MARK) + BPF_FIB_LOOKUP_SRC | BPF_FIB_LOOKUP_MARK | \ + BPF_FIB_LOOKUP_VLAN) BPF_CALL_4(bpf_xdp_fib_lookup, struct xdp_buff *, ctx, struct bpf_fib_lookup *, params, int, plen, u32, flags)
tools/include/uapi/linux/bpf.h +20 -1
@@ -3527,6 +3527,19 @@ union bpf_attr {* Use the mark present in *params*->mark for the fib lookup. * This option should not be used with BPF_FIB_LOOKUP_DIRECT, * as it only has meaning for full lookups. +* **BPF_FIB_LOOKUP_VLAN** +* If the fib lookup resolves to a VLAN device, set +* *params*->h_vlan_proto and *params*->h_vlan_TCI from +* the VLAN device and replace *params*->ifindex with the +* underlying real device's ifindex. This lets XDP +* programs that target the underlying physical device +* (VLAN devices have no XDP xmit) discover both the +* real egress ifindex and the VLAN tag to push in one +* call. *params*->h_vlan_TCI carries the VID only, +* with PCP and DEI bits zero; a consumer wanting to +* set egress priority writes PCP itself. Only the +* immediate parent is resolved; stacked VLANs (QinQ) +* are not walked. * * *ctx* is either **struct xdp_md** for XDP programs or * **struct sk_buff** tc cls_act programs.
@@ -7322,6 +7335,7 @@ enum {BPF_FIB_LOOKUP_TBID = (1U <<3 ),BPF_FIB_LOOKUP_SRC = (1U <<4 ),BPF_FIB_LOOKUP_MARK = (1U <<5 ), + BPF_FIB_LOOKUP_VLAN = (1U << 6), }; enum {
@@ -7388,7 +7402,12 @@ struct bpf_fib_lookup {union {struct { -/* output */ +/* output: only populated with BPF_FIB_LOOKUP_VLAN +* when the resolved egress is a VLAN device, in +* which case *ifindex* is replaced with the +* underlying real device's ifindex. Otherwise +* both fields are zeroed. +*/ __be16 h_vlan_proto;__be16 h_vlan_TCI; };
base-commit: f1a660bbd12dd855fce6cf13f144008c4e45e7c7
Signature · 1 line
2.54.0
[email protected]
AI reviewThank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider: - [High] Incorrect traversal of stacked VLANs leads to dropped intermediate VLAN tags due to the use of…
Thank you for your contribution! Sashiko AI review found 1 potential issue(s) to consider:
- [High] Incorrect traversal of stacked VLANs leads to dropped intermediate VLAN tags due to the use of vlan_dev_real_dev(). --
commit 87422c811faeffeb7be03864e938458c390f9a15
Author: Avinash Duduskar <[email protected]>
bpf: Add BPF_FIB_LOOKUP_VLAN flag to bpf_fib_lookup() helper
This commit introduces the BPF_FIB_LOOKUP_VLAN flag for the bpf_fib_lookup() helper. When set, if the egress resolves to a VLAN device, the helper returns the VLAN tag and the underlying device's ifindex to allow XDP programs to push the tag and forward the frame directly.
Quoted 4 lines
net/core/filter.c +0
-0
[ ... ]
Quoted 4 lines
net/core/filter.c +0
-0@@ -6119,10 +6119,28 @@ static const struct bpf_func_proto bpf_skb_get_xfrm_state_proto = {#endif #if IS_ENABLED (CONFIG_INET ) ||IS_ENABLED (CONFIG_IPV6 )
net/core/filter.c static const struct bpf_func_proto bpf_skb_get_xfrm_state_proto = { · line 6130
Quoted 10 lines
net/core/filter.c +6
-1-static int bpf_fib_set_fwd_params (struct bpf_fib_lookup *params,u32 mtu) +static int bpf_fib_set_fwd_params (struct net_device *dev, +struct bpf_fib_lookup *params, +u32 flags,u32 mtu) { params->h_vlan_TCI =0 ; params->h_vlan_proto =0 ; + +if ((flags &BPF_FIB_LOOKUP_VLAN ) &&is_vlan_dev (dev)) { +struct net_device *real_dev =vlan_dev_real_dev (dev);
[Severity: High]
Does this correctly handle stacked VLANs (QinQ)? Both the commit message and the inline comment below mention that stacked VLANs are not walked and only the immediate parent is resolved.
However, looking at the implementation of vlan_dev_real_dev() in net/8021q/vlan_core.c:
struct net_device *vlan_dev_real_dev(const struct net_device *dev)
{
struct net_device *ret = vlan_dev_priv(dev)->real_dev;
while (is_vlan_dev(ret))
ret = vlan_dev_priv(ret)->real_dev;
return ret;
}It appears vlan_dev_real_dev() loops through all stacked VLANs to return the ultimate non-VLAN underlying device. If the configuration is QinQ, wouldn't this set params->ifindex to the base physical device while only returning the topmost VLAN tag in params->h_vlan_TCI?
This could cause an XDP program to push only a single tag and drop the intermediate VLAN tags, bypassing the outer VLAN entirely.
Should this use vlan_dev_priv(dev)->real_dev instead to only get the immediate parent as intended?
Quoted 12 lines
net/core/filter.c +12
-0+ +/* Only the immediate parent is resolved; stacked VLANs +* (QinQ) are not walked, and a NULL real_dev (which +* is_vlan_dev() rules out in practice) keeps the +* original ifindex. +*/ +if (real_dev) { + params->h_vlan_proto =vlan_dev_vlan_proto (dev); + params->h_vlan_TCI =htons (vlan_dev_vlan_id (dev)); + params->ifindex = real_dev->ifindex; + } + }
Signature · 1 line
Sashiko AI review · https://sashiko.dev/#/patchset/[email protected]?part=1
[email protected]
AI reviewDoes this correctly handle stacked VLANs (QinQ)? Both the commit message and the inline comment below mention that stacked VLANs are not walked and only the immediate parent is resolved.
Quoted 4 lines
include/uapi/linux/bpf.h +0
-0
[ ... ]
Quoted 8 lines
net/core/filter.c +0
-0@@ -6119,10 +6119,28 @@ static const struct bpf_func_proto bpf_skb_get_xfrm_state_proto = {#endif #if IS_ENABLED (CONFIG_INET ) ||IS_ENABLED (CONFIG_IPV6 )
net/core/filter.c static const struct bpf_func_proto bpf_skb_get_xfrm_state_proto = { · line 6130
Quoted 10 lines
net/core/filter.c +6
-1-static int bpf_fib_set_fwd_params (struct bpf_fib_lookup *params,u32 mtu) +static int bpf_fib_set_fwd_params (struct net_device *dev, +struct bpf_fib_lookup *params, +u32 flags,u32 mtu) { params->h_vlan_TCI =0 ; params->h_vlan_proto =0 ; + +if ((flags &BPF_FIB_LOOKUP_VLAN ) &&is_vlan_dev (dev)) { +struct net_device *real_dev =vlan_dev_real_dev (dev);
Does this correctly handle stacked VLANs (QinQ)? Both the commit message and the inline comment below mention that stacked VLANs are not walked and only the immediate parent is resolved.
However, looking at the implementation of vlan_dev_real_dev() in net/8021q/vlan_core.c:
const struct net_device *vlan_dev_real_dev(const struct net_device *dev)
{
while (is_vlan_dev(dev))
dev = vlan_dev_priv(dev)->real_dev;
return dev;
}It appears vlan_dev_real_dev() loops through all stacked VLANs to return the ultimate non-VLAN underlying device.
If the configuration is QinQ, wouldn't this set params->ifindex to the base physical device while only returning the topmost VLAN tag in params->h_vlan_TCI? This could cause an XDP program to push only a single tag and drop the intermediate VLAN tags, bypassing the outer VLAN entirely.
Should this use vlan_dev_priv(dev)->real_dev instead to only get the immediate parent as intended?
net/core/filter.c static const struct bpf_func_proto bpf_skb_get_xfrm_state_proto = { · line 6132
Quoted 2 lines
net/core/filter.c +2
-0+ +/* Only the immediate parent is resolved; stacked VLANs
This isn't a bug, but the comment style here differs from BPF subsystem guidelines. The opening /* has comment text on the same line, but the BPF comment style prefers the opening /* on its own line with text starting on the next line.
Quoted 22 lines
net/core/filter.c +12
-1+* (QinQ) are not walked, and a NULL real_dev (which +* is_vlan_dev() rules out in practice) keeps the +* original ifindex. +*/ +if (real_dev) { + params->h_vlan_proto =vlan_dev_vlan_proto (dev); + params->h_vlan_TCI =htons (vlan_dev_vlan_id (dev)); + params->ifindex = real_dev->ifindex; + } + } +if (mtu) params->mtu_result = mtu;/* union with tot_len */ @@ -6265,7 +6283,7 @@ static int bpf_ipv4_fib_lookup(struct net *net, struct bpf_fib_lookup *params,memcpy (params->smac, dev->dev_addr,ETH_ALEN ); set_fwd_params: - return bpf_fib_set_fwd_params(params, mtu); + return bpf_fib_set_fwd_params(dev, params, flags, mtu); }#endif
[ ... ]
tools/include/uapi/linux/bpf.h
Quoted 4 lines
tools/include/uapi/linux/bpf.h +0
-0
[ ... ]
---
AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md
CI run summary: https://github.com/kernel-patches/bpf/actions/runs/27223878961
Quoted 32 lines · Avinash Duduskar
bpf_fib_lookup()returns the FIB-resolved egress ifindex straight from the fib result. When the egress is a VLAN device, the returned ifindex is the VLAN netdev's, which has no XDP xmit handler; XDP programs that want to forward the frame (e.g. xdp-forward) must instead target the underlying physical device and push the VLAN tag themselves. Today the program has no way to learn either the underlying ifindex or the VLAN tag without maintaining its own VLAN-to-ifindex map in userspace and refreshing it on netlink events.Add
BPF_FIB_LOOKUP_VLAN. When the caller sets this flag and the fib result is a VLAN device, populate the existing output fieldsparams->h_vlan_protoandparams->h_vlan_TCIfrom the VLAN device, and replaceparams->ifindexwith the underlying real device's ifindex.params->h_vlan_TCIcarries the VID only, with PCP and DEI bits zero; a consumer wanting to set egress priority writes PCP itself. Only the immediate parent is resolved; stacked VLANs (QinQ) are not walked. When the flag is not set, behaviour is unchanged:h_vlan_protoandh_vlan_TCIare zeroed and ifindex is left at the FIB result.This lets an XDP redirect target the physical device and learn the tag to push in a single lookup, which xdp-forward's optional VLAN mode (xdp-project/xdp-tools#504) wants from the kernel side.
The change extends
bpf_fib_set_fwd_params()to take the egress dev and the lookup flags so the VLAN swap happens in the same place the vlan output fields are zeroed by default. Both IPv4 and IPv6 callers pass through. The helper's input semantics are unchanged. Under !CONFIG_VLAN_8021Q,is_vlan_dev()returns false and the new block is a no-op.
Quoted 2 lines
Suggested-by Toke Høiland-Jørgensen <[email protected]>Signed-off-by Avinash Duduskar <[email protected]>
Other than the bots' comment, I think we should make this bidirectional. I.e., it should also be possible to supply the vlan tag from the packet when doing the lookup.
This requires a second flag, which has to be exclusive with BPF_FIB_LOOKUP_TBID, as the tbid field unfortunately overlaps with the VLAN fields (so they can't be used together as input).
-Toke