[PATCH] bpf: Unshare cloned skb before devmap egress XDP program

[PATCH] bpf: Unshare cloned skb before devmap egress XDP program

[Sun Jian]

dev_map_redirect_clone() uses skb_clone() when redirecting a generic XDP
skb to multiple devmap destinations. The cloned skb can share packet data
with other clones.

If the destination devmap entry has an egress XDP program, that program
can modify packet data. Such modifications can then be observed by other
clones sharing the same packet data.

This can be reproduced by strengthening xdp_veth_egress to configure a
different source MAC for each egress device and checking that store_mac_1/2
observe the MAC configured for their own egress devices. Without the fix,
the SKB_MODE subtest observes store_mac_1 receiving the MAC configured for
the next egress device.

Fix this by unsharing the cloned skb before running the devmap egress XDP
program. Limit the extra copy to destinations with an attached egress
program.

Tested with:
  ./test_progs -t xdp_veth_egress
  ./test_progs -t xdp_veth
  ./test_progs -t xdp

Fixes: e624d4ed4aa8 ("xdp: Extend xdp_redirect_map with broadcast support")
Signed-off-by: Sun Jian <sun.jian.kdev@gmail.com>
---
 kernel/bpf/devmap.c                                 |  6 ++++++
 .../selftests/bpf/prog_tests/test_xdp_veth.c        | 13 ++++++++++---
 2 files changed, 16 insertions(+), 3 deletions(-)

diff --git a/kernel/bpf/devmap.c b/kernel/bpf/devmap.c
index cc0a43ebab6b..4ae65d44f9d6 100644
--- a/kernel/bpf/devmap.c
+++ b/kernel/bpf/devmap.c
@@ -730,6 +730,12 @@ static int dev_map_redirect_clone(struct bpf_dtab_netdev *dst,
 	if (!nskb)
 		return -ENOMEM;
 
+	if (dst->xdp_prog) {
+		nskb = skb_unshare(nskb, GFP_ATOMIC);
+		if (!nskb)
+			return -ENOMEM;
+	}
+
 	err = dev_map_generic_redirect(dst, nskb, xdp_prog);
 	if (unlikely(err)) {
 		consume_skb(nskb);
diff --git a/tools/testing/selftests/bpf/prog_tests/test_xdp_veth.c b/tools/testing/selftests/bpf/prog_tests/test_xdp_veth.c
index 3e98a1665936..52d79d5c5629 100644
--- a/tools/testing/selftests/bpf/prog_tests/test_xdp_veth.c
+++ b/tools/testing/selftests/bpf/prog_tests/test_xdp_veth.c
@@ -456,7 +456,11 @@ static void xdp_veth_egress(u32 flags)
 			.remote_flags = flags,
 		}
 	};
-	const char magic_mac[6] = { 0xAA, 0xBB, 0xCC, 0xDD, 0xEE, 0xFF};
+	const unsigned char egress_macs[VETH_PAIRS_COUNT][ETH_ALEN] = {
+		{ 0xAA, 0xBB, 0xCC, 0xDD, 0xEE, 0x01 },
+		{ 0xAA, 0xBB, 0xCC, 0xDD, 0xEE, 0x02 },
+		{ 0xAA, 0xBB, 0xCC, 0xDD, 0xEE, 0x03 },
+	};
 	struct xdp_redirect_multi_kern *xdp_redirect_multi_kern;
 	struct bpf_object *bpf_objs[VETH_EGRESS_SKEL_NB];
 	struct xdp_redirect_map *xdp_redirect_map;
@@ -512,7 +516,7 @@ static void xdp_veth_egress(u32 flags)
 						 &net_config, prog_cfg, i))
 			goto destroy_xdp_redirect_map;
 
-		err = bpf_map_update_elem(mac_map, &ifindex, magic_mac, 0);
+		err = bpf_map_update_elem(mac_map, &ifindex, egress_macs[i], 0);
 		if (!ASSERT_OK(err, "bpf_map_update_elem"))
 			goto destroy_xdp_redirect_map;
 
@@ -531,13 +535,16 @@ static void xdp_veth_egress(u32 flags)
 
 	for (i = 0; i < 2; i++) {
 		u32 key = i;
+		__be64 expected = 0;
 		u64 res;
 
 		err = bpf_map_lookup_elem(res_map, &key, &res);
 		if (!ASSERT_OK(err, "get MAC res"))
 			goto destroy_xdp_redirect_map;
 
-		ASSERT_STRNEQ((const char *)&res, magic_mac, ETH_ALEN, "compare mac");
+		/* store_mac_1/2 run on the second/third remote veths. */
+		memcpy(&expected, egress_macs[i + 1], ETH_ALEN);
+		ASSERT_EQ(res, expected, "compare mac");
 	}
 
 destroy_xdp_redirect_map:
-- 
2.43.0

Re: [PATCH] bpf: Unshare cloned skb before devmap egress XDP program

[Menglong Dong]

On 2026/6/9 18:02 Sun Jian <sun.jian.kdev@gmail.com> write:
> dev_map_redirect_clone() uses skb_clone() when redirecting a generic XDP
> skb to multiple devmap destinations. The cloned skb can share packet data
> with other clones.
> 
> If the destination devmap entry has an egress XDP program, that program
> can modify packet data. Such modifications can then be observed by other
> clones sharing the same packet data.
> 
> This can be reproduced by strengthening xdp_veth_egress to configure a
> different source MAC for each egress device and checking that store_mac_1/2
> observe the MAC configured for their own egress devices. Without the fix,
> the SKB_MODE subtest observes store_mac_1 receiving the MAC configured for
> the next egress device.
> 
> Fix this by unsharing the cloned skb before running the devmap egress XDP
> program. Limit the extra copy to destinations with an attached egress
> program.

Hi, Jian.

This sounds like a good idea in this case. When I have a look at bpf_clone_redirect(),
I found that it use skb_clone() too, which means it has the same problem. The
data can be modified by other xdp prog in the destination NIC if we use
bpf_clone_redirect().

So maybe this is the default logic, and I'm not sure if this patch can break the
existing users :/

Thanks!
Menglong Dong

> 
> Tested with:
>   ./test_progs -t xdp_veth_egress
>   ./test_progs -t xdp_veth
>   ./test_progs -t xdp
[...]
>  
>  destroy_xdp_redirect_map:
> -- 
> 2.43.0
> 
> 
> 

Re: [PATCH] bpf: Unshare cloned skb before devmap egress XDP program

[Emil Tsalapatis]

On Tue Jun 9, 2026 at 7:06 AM EDT, Menglong Dong wrote:
> On 2026/6/9 18:02 Sun Jian <sun.jian.kdev@gmail.com> write:
>> dev_map_redirect_clone() uses skb_clone() when redirecting a generic XDP
>> skb to multiple devmap destinations. The cloned skb can share packet data
>> with other clones.
>> 
>> If the destination devmap entry has an egress XDP program, that program
>> can modify packet data. Such modifications can then be observed by other
>> clones sharing the same packet data.
>> 
>> This can be reproduced by strengthening xdp_veth_egress to configure a
>> different source MAC for each egress device and checking that store_mac_1/2
>> observe the MAC configured for their own egress devices. Without the fix,
>> the SKB_MODE subtest observes store_mac_1 receiving the MAC configured for
>> the next egress device.
>> 
>> Fix this by unsharing the cloned skb before running the devmap egress XDP
>> program. Limit the extra copy to destinations with an attached egress
>> program.
>
> Hi, Jian.
>
> This sounds like a good idea in this case. When I have a look at bpf_clone_redirect(),
> I found that it use skb_clone() too, which means it has the same problem. The
> data can be modified by other xdp prog in the destination NIC if we use
> bpf_clone_redirect().
>
> So maybe this is the default logic, and I'm not sure if this patch can break the
> existing users :/

I think for use cases where we are using bpf_clone_redirect() to use
one clone for inspection this would add an unnecessary copy. Maybe
adding *_copy() variants instead of changing the *_clone() would be
better? That way we wouldn't be changing the behavior for existing
consumers and the naming would be consistent with the skb_* methods.

But more importantly, is there an actual use case for the kind of API
that the modified selftest requires? Nobody until now has considered
the existing behavior to be a problem.

>
> Thanks!
> Menglong Dong
>
>> 
>> Tested with:
>>   ./test_progs -t xdp_veth_egress
>>   ./test_progs -t xdp_veth
>>   ./test_progs -t xdp
> [...]
>>  
>>  destroy_xdp_redirect_map:
>> -- 
>> 2.43.0
>> 
>> 
>> 

Re: [PATCH] bpf: Unshare cloned skb before devmap egress XDP program

[Menglong Dong]

On 2026/6/10 08:06 Emil Tsalapatis <emil@etsalapatis.com> write:
> On Tue Jun 9, 2026 at 7:06 AM EDT, Menglong Dong wrote:
> > On 2026/6/9 18:02 Sun Jian <sun.jian.kdev@gmail.com> write:
> >> dev_map_redirect_clone() uses skb_clone() when redirecting a generic XDP
> >> skb to multiple devmap destinations. The cloned skb can share packet data
> >> with other clones.
> >> 
> >> If the destination devmap entry has an egress XDP program, that program
> >> can modify packet data. Such modifications can then be observed by other
> >> clones sharing the same packet data.
> >> 
> >> This can be reproduced by strengthening xdp_veth_egress to configure a
> >> different source MAC for each egress device and checking that store_mac_1/2
> >> observe the MAC configured for their own egress devices. Without the fix,
> >> the SKB_MODE subtest observes store_mac_1 receiving the MAC configured for
> >> the next egress device.
> >> 
> >> Fix this by unsharing the cloned skb before running the devmap egress XDP
> >> program. Limit the extra copy to destinations with an attached egress
> >> program.
> >
> > Hi, Jian.
> >
> > This sounds like a good idea in this case. When I have a look at bpf_clone_redirect(),
> > I found that it use skb_clone() too, which means it has the same problem. The
> > data can be modified by other xdp prog in the destination NIC if we use
> > bpf_clone_redirect().
> >
> > So maybe this is the default logic, and I'm not sure if this patch can break the
> > existing users :/
> 
> I think for use cases where we are using bpf_clone_redirect() to use
> one clone for inspection this would add an unnecessary copy. Maybe
> adding *_copy() variants instead of changing the *_clone() would be
> better? That way we wouldn't be changing the behavior for existing
> consumers and the naming would be consistent with the skb_* methods.

Agree. It's not a good idea to change the logic of the existing API. Or
maybe we can add a BPF_F_CLONE flag for the existing API.

> 
> But more importantly, is there an actual use case for the kind of API
> that the modified selftest requires? Nobody until now has considered
> the existing behavior to be a problem.

Agree too. Obviously, this is not a bug. For the use case in the commit
log, it's something that can be fixed by the user themself. If we need
modify the MAC, we'd better attach a BPF program for all the egress
device in the devmap.

> 
> >
> > Thanks!
> > Menglong Dong
> >
> >> 
> >> Tested with:
> >>   ./test_progs -t xdp_veth_egress
> >>   ./test_progs -t xdp_veth
> >>   ./test_progs -t xdp
> > [...]
> >>  
> >>  destroy_xdp_redirect_map:
> >> -- 
> >> 2.43.0
> >> 
> >> 
> >> 
> 

Re: [PATCH] bpf: Unshare cloned skb before devmap egress XDP program

[sun jian]

On Wed, Jun 10, 2026 at 9:21 AM Menglong Dong <menglong.dong@linux.dev> wrote:
>
> On 2026/6/10 08:06 Emil Tsalapatis <emil@etsalapatis.com> write:
> > On Tue Jun 9, 2026 at 7:06 AM EDT, Menglong Dong wrote:
> > > On 2026/6/9 18:02 Sun Jian <sun.jian.kdev@gmail.com> write:
> > >> dev_map_redirect_clone() uses skb_clone() when redirecting a generic XDP
> > >> skb to multiple devmap destinations. The cloned skb can share packet data
> > >> with other clones.
> > >>
> > >> If the destination devmap entry has an egress XDP program, that program
> > >> can modify packet data. Such modifications can then be observed by other
> > >> clones sharing the same packet data.
> > >>
> > >> This can be reproduced by strengthening xdp_veth_egress to configure a
> > >> different source MAC for each egress device and checking that store_mac_1/2
> > >> observe the MAC configured for their own egress devices. Without the fix,
> > >> the SKB_MODE subtest observes store_mac_1 receiving the MAC configured for
> > >> the next egress device.
> > >>
> > >> Fix this by unsharing the cloned skb before running the devmap egress XDP
> > >> program. Limit the extra copy to destinations with an attached egress
> > >> program.
> > >
> > > Hi, Jian.
> > >
> > > This sounds like a good idea in this case. When I have a look at bpf_clone_redirect(),
> > > I found that it use skb_clone() too, which means it has the same problem. The
> > > data can be modified by other xdp prog in the destination NIC if we use
> > > bpf_clone_redirect().
> > >
> > > So maybe this is the default logic, and I'm not sure if this patch can break the
> > > existing users :/
> >
> > I think for use cases where we are using bpf_clone_redirect() to use
> > one clone for inspection this would add an unnecessary copy. Maybe
> > adding *_copy() variants instead of changing the *_clone() would be
> > better? That way we wouldn't be changing the behavior for existing
> > consumers and the naming would be consistent with the skb_* methods.
>
> Agree. It's not a good idea to change the logic of the existing API. Or
> maybe we can add a BPF_F_CLONE flag for the existing API.
>
> >
> > But more importantly, is there an actual use case for the kind of API
> > that the modified selftest requires? Nobody until now has considered
> > the existing behavior to be a problem.
>
> Agree too. Obviously, this is not a bug. For the use case in the commit
> log, it's something that can be fixed by the user themself. If we need
> modify the MAC, we'd better attach a BPF program for all the egress
> device in the devmap.
>
> >
> > >
> > > Thanks!
> > > Menglong Dong
> > >
> > >>
> > >> Tested with:
> > >>   ./test_progs -t xdp_veth_egress
> > >>   ./test_progs -t xdp_veth
> > >>   ./test_progs -t xdp
> > > [...]
> > >>
> > >>  destroy_xdp_redirect_map:
> > >> --
> > >> 2.43.0
> > >>
> > >>
> > >>
> >
>
>
>
>

Thanks for all the comments.

I agree this should not be treated as a generic skb_clone() issue, and I
understand the concern about changing existing clone/shared-data semantics.

The case I was trying to address is narrower: generic XDP devmap
broadcast/multi redirect with per-devmap-entry egress programs. The use case is
not newly introduced by this patch. The existing xdp_veth_egress test already
attaches xdp_devmap_prog to all egress devmap entries, and that program rewrites
eth->h_source based on ctx->egress_ifindex. My change only strengthens the test
from checking that the observed MAC is not equal to a single magic MAC to
checking that each destination observes the MAC selected for its own egress
ifindex.

So attaching an egress program to all devmap entries is already what this test
does. The SKB_MODE failure happens because the cloned skbs can still share the
packet data, and a later per-egress rewrite can be observed by another
destination.

That said, I see the concern that this may need a clearer semantic boundary
between clone and copy behavior. I will also check the last_dst path pointed out
by Sashiko before deciding whether this should be handled as a bug fix, or
whether it needs a separate explicit copy/isolated redirect semantic instead.

Re: [PATCH] bpf: Unshare cloned skb before devmap egress XDP program

[Jiayuan Chen]

On 6/10/26 9:58 AM, sun jian wrote:
> On Wed, Jun 10, 2026 at 9:21 AM Menglong Dong <menglong.dong@linux.dev> wrote:
>> On 2026/6/10 08:06 Emil Tsalapatis <emil@etsalapatis.com> write:
>>> On Tue Jun 9, 2026 at 7:06 AM EDT, Menglong Dong wrote:
>>>> On 2026/6/9 18:02 Sun Jian <sun.jian.kdev@gmail.com> write:
[...]
>>>> Hi, Jian.
>>>>
>>>> This sounds like a good idea in this case. When I have a look at bpf_clone_redirect(),
>>>> I found that it use skb_clone() too, which means it has the same problem. The
>>>> data can be modified by other xdp prog in the destination NIC if we use
>>>> bpf_clone_redirect().
>>>>
>>>> So maybe this is the default logic, and I'm not sure if this patch can break the
>>>> existing users :/
>>> I think for use cases where we are using bpf_clone_redirect() to use
>>> one clone for inspection this would add an unnecessary copy. Maybe
>>> adding *_copy() variants instead of changing the *_clone() would be
>>> better? That way we wouldn't be changing the behavior for existing
>>> consumers and the naming would be consistent with the skb_* methods.
>> Agree. It's not a good idea to change the logic of the existing API. Or
>> maybe we can add a BPF_F_CLONE flag for the existing API.
>>
>>> But more importantly, is there an actual use case for the kind of API
>>> that the modified selftest requires? Nobody until now has considered
>>> the existing behavior to be a problem.
>> Agree too. Obviously, this is not a bug. For the use case in the commit
>> log, it's something that can be fixed by the user themself. If we need
>> modify the MAC, we'd better attach a BPF program for all the egress
>> device in the devmap.
>>
[...]
>>>>
> Thanks for all the comments.
>
> I agree this should not be treated as a generic skb_clone() issue, and I
> understand the concern about changing existing clone/shared-data semantics.
>
> The case I was trying to address is narrower: generic XDP devmap
> broadcast/multi redirect with per-devmap-entry egress programs. The use case is
> not newly introduced by this patch. The existing xdp_veth_egress test already
> attaches xdp_devmap_prog to all egress devmap entries, and that program rewrites
> eth->h_source based on ctx->egress_ifindex. My change only strengthens the test
> from checking that the observed MAC is not equal to a single magic MAC to
> checking that each destination observes the MAC selected for its own egress
> ifindex.
>
> So attaching an egress program to all devmap entries is already what this test
> does. The SKB_MODE failure happens because the cloned skbs can still share the
> packet data, and a later per-egress rewrite can be observed by another
> destination.
>
> That said, I see the concern that this may need a clearer semantic boundary
> between clone and copy behavior. I will also check the last_dst path pointed out
> by Sashiko before deciding whether this should be handled as a bug fix, or
> whether it needs a separate explicit copy/isolated redirect semantic instead.


I agree there's a real concern here. For context, native XDP already 
makes a full copy
of the frame for every broadcast destination:

   dev_map_enqueue_multi -> dev_map_enqueue_clone -> xdpf_clone

so on the native path each destination gets its own independent buffer.

The generic path is the odd one out, because skb_clone() shares the 
underlying data.
I think this rationale should be spelled out in the commit message.

Regarding performance: generic XDP is already a slow path — the ingress 
side already
linearizes / expands the skb (pskb_expand_head() in 
netif_receive_generic_xdp())
so the extra copy here is not a concern.


However, there is one problem with the patch as it stands.

The last destination is transmitted using the original skb directly, 
without unsharing it.
So with two destinations where the first one has no egress XDP program 
and the second (i.e. the last one) does,
the egress modification on the second destination will corrupt the 
packet sent to the first.

This could be addressed by doing the skb_unshare() inside 
dev_map_generic_redirect() instead ?

For the non-broadcast (existing) path the skb is not cloned, so 
skb_unshare() simply returns it unchanged.

Re: [PATCH] bpf: Unshare cloned skb before devmap egress XDP program

[sun jian]

On Wed, Jun 10, 2026 at 1:13 PM Jiayuan Chen <jiayuan.chen@linux.dev> wrote:
>
>
> On 6/10/26 9:58 AM, sun jian wrote:
> > On Wed, Jun 10, 2026 at 9:21 AM Menglong Dong <menglong.dong@linux.dev> wrote:
> >> On 2026/6/10 08:06 Emil Tsalapatis <emil@etsalapatis.com> write:
> >>> On Tue Jun 9, 2026 at 7:06 AM EDT, Menglong Dong wrote:
> >>>> On 2026/6/9 18:02 Sun Jian <sun.jian.kdev@gmail.com> write:
> [...]
> >>>> Hi, Jian.
> >>>>
> >>>> This sounds like a good idea in this case. When I have a look at bpf_clone_redirect(),
> >>>> I found that it use skb_clone() too, which means it has the same problem. The
> >>>> data can be modified by other xdp prog in the destination NIC if we use
> >>>> bpf_clone_redirect().
> >>>>
> >>>> So maybe this is the default logic, and I'm not sure if this patch can break the
> >>>> existing users :/
> >>> I think for use cases where we are using bpf_clone_redirect() to use
> >>> one clone for inspection this would add an unnecessary copy. Maybe
> >>> adding *_copy() variants instead of changing the *_clone() would be
> >>> better? That way we wouldn't be changing the behavior for existing
> >>> consumers and the naming would be consistent with the skb_* methods.
> >> Agree. It's not a good idea to change the logic of the existing API. Or
> >> maybe we can add a BPF_F_CLONE flag for the existing API.
> >>
> >>> But more importantly, is there an actual use case for the kind of API
> >>> that the modified selftest requires? Nobody until now has considered
> >>> the existing behavior to be a problem.
> >> Agree too. Obviously, this is not a bug. For the use case in the commit
> >> log, it's something that can be fixed by the user themself. If we need
> >> modify the MAC, we'd better attach a BPF program for all the egress
> >> device in the devmap.
> >>
> [...]
> >>>>
> > Thanks for all the comments.
> >
> > I agree this should not be treated as a generic skb_clone() issue, and I
> > understand the concern about changing existing clone/shared-data semantics.
> >
> > The case I was trying to address is narrower: generic XDP devmap
> > broadcast/multi redirect with per-devmap-entry egress programs. The use case is
> > not newly introduced by this patch. The existing xdp_veth_egress test already
> > attaches xdp_devmap_prog to all egress devmap entries, and that program rewrites
> > eth->h_source based on ctx->egress_ifindex. My change only strengthens the test
> > from checking that the observed MAC is not equal to a single magic MAC to
> > checking that each destination observes the MAC selected for its own egress
> > ifindex.
> >
> > So attaching an egress program to all devmap entries is already what this test
> > does. The SKB_MODE failure happens because the cloned skbs can still share the
> > packet data, and a later per-egress rewrite can be observed by another
> > destination.
> >
> > That said, I see the concern that this may need a clearer semantic boundary
> > between clone and copy behavior. I will also check the last_dst path pointed out
> > by Sashiko before deciding whether this should be handled as a bug fix, or
> > whether it needs a separate explicit copy/isolated redirect semantic instead.
>
>
> I agree there's a real concern here. For context, native XDP already
> makes a full copy
> of the frame for every broadcast destination:
>
>    dev_map_enqueue_multi -> dev_map_enqueue_clone -> xdpf_clone
>
> so on the native path each destination gets its own independent buffer.
>
> The generic path is the odd one out, because skb_clone() shares the
> underlying data.
> I think this rationale should be spelled out in the commit message.
>
> Regarding performance: generic XDP is already a slow path — the ingress
> side already
> linearizes / expands the skb (pskb_expand_head() in
> netif_receive_generic_xdp())
> so the extra copy here is not a concern.
>
>
> However, there is one problem with the patch as it stands.
>
> The last destination is transmitted using the original skb directly,
> without unsharing it.
> So with two destinations where the first one has no egress XDP program
> and the second (i.e. the last one) does,
> the egress modification on the second destination will corrupt the
> packet sent to the first.
>
> This could be addressed by doing the skb_unshare() inside
> dev_map_generic_redirect() instead ?
>
> For the non-broadcast (existing) path the skb is not cloned, so
> skb_unshare() simply returns it unchanged.
>
>
>
Hi Jiayuan,

Thanks for the detailed explanation.

I checked the last_dst case and was able to reproduce it locally. In that case,
earlier destinations without devmap egress programs can still observe the packet
data rewritten by the final destination's egress program.

I also tried your suggestion of handling this in dev_map_generic_redirect(),
right before running the devmap egress program. That fixes both the original
case and the last_dst case.

One detail I changed in the implementation is to use skb_copy() + consume_skb()
instead of skb_unshare(), so that allocation failure does not consume the skb
and the existing caller error paths keep their ownership semantics.

With that version, the following tests pass:

./test_progs -t xdp_veth_egress
./test_progs -t xdp_veth
./test_progs -t xdp

I'll send v2 shortly.

Thanks,
sun jian