[PATCH net v3] net: ipv6: keep room for the mac header in dst_dev_overhead()

[PATCH net v3] net: ipv6: keep room for the mac header in dst_dev_overhead()

[Yuya Kusakabe]

The seg6, ioam6 and rpl lwtunnels size their skb_cow_head() request as
the length they are about to push plus dst_dev_overhead(), then push the
new headers and rebuild the mac header below them with
skb_mac_header_rebuild().  That rebuild needs skb->mac_len of headroom,
but dst_dev_overhead() leaves LL_RESERVED_SPACE() of the egress device,
16 bytes for plain Ethernet.

Where the mac header is longer than that, as it is on ingress through a
VLAN device with reorder_hdr off, the rebuild runs out of room:
skb_set_mac_header(skb, -skb->mac_len) computes a negative offset,
stores it unchecked in the u16 skb->mac_header, and the memmove that
follows writes skb->mac_len bytes about 64 KB past skb->head.
Forwarding plain ping6 traffic through such a device reproduces it on
all five seg6 encapsulation modes and on the rpl and ioam6 inline paths;
skb->mac_header comes back as 65534 on a 704-byte head.

Return the larger of the two.  The helper already returns skb->mac_len
when it has no dst, so this only makes the other branch agree, and it
covers every caller rather than each call site in turn.

Fixes: 40475b63761a ("net: ipv6: seg6_iptunnel: mitigate 2-realloc issue")
Fixes: dce525185bc9 ("net: ipv6: ioam6_iptunnel: mitigate 2-realloc issue")
Fixes: 985ec6f5e623 ("net: ipv6: rpl_iptunnel: mitigate 2-realloc issue")
Suggested-by: Andrea Mayer <andrea.mayer@uniroma2.it>
Assisted-by: LLM
Signed-off-by: Yuya Kusakabe <yuya.kusakabe@gmail.com>
---
Changes in v3:
- Fix dst_dev_overhead() itself rather than each of its callers [Andrea]
- Drop Justin's Reviewed-by from v2, which was given for the per-caller
  version; the patch no longer touches the files he reviewed
- Link to v2: https://lore.kernel.org/r/20260918-seg6-maclen-headroom-v2-1-4d370af55b2e@gmail.com

Changes in v2:
- Use max_t(unsigned int, ...) rather than max() [Justin]
- Make the same change in ioam6_iptunnel and rpl_iptunnel [Justin]
- Reproduce and verify the fix on those two as well: unpatched, rpl
  underflows in 12 of 16 probed configurations and ioam6 in 3 of 13,
  both reaching about 64 KB past a 320-byte head; patched, neither
  underflows in 18
- Rewrap the requests so they no longer exceed 80 columns
- Retitle for net: ipv6, now that the change spans three files
- Link to v1: https://lore.kernel.org/r/20260917-seg6-maclen-headroom-v1-1-02ccec50f096@gmail.com
---
 include/net/dst.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/include/net/dst.h b/include/net/dst.h
index 307073eae7f8..dbedfe72e1fd 100644
--- a/include/net/dst.h
+++ b/include/net/dst.h
@@ -455,7 +455,8 @@ static inline unsigned int dst_dev_overhead(struct dst_entry *dst,
 					    struct sk_buff *skb)
 {
 	if (likely(dst))
-		return LL_RESERVED_SPACE(dst->dev);
+		return max_t(unsigned int, skb->mac_len,
+			     LL_RESERVED_SPACE(dst->dev));
 
 	return skb->mac_len;
 }

---
base-commit: 9a1599eeb8d18a2113e4cabcbd3bec5a8377dbed
change-id: 20260917-seg6-maclen-headroom-53a339c17dac

Best regards,
--  
Yuya Kusakabe <yuya.kusakabe@gmail.com>

Re: [PATCH net v3] net: ipv6: keep room for the mac header in dst_dev_overhead()

[Justin Iurman]

On 9/21/26 22:49, Yuya Kusakabe wrote:
> The seg6, ioam6 and rpl lwtunnels size their skb_cow_head() request
> as the length they are about to push plus dst_dev_overhead(), then
> push the new headers and rebuild the mac header below them with 
> skb_mac_header_rebuild().  That rebuild needs skb->mac_len of
> headroom, but dst_dev_overhead() leaves LL_RESERVED_SPACE() of the
> egress device, 16 bytes for plain Ethernet.
> 
> Where the mac header is longer than that, as it is on ingress
> through a VLAN device with reorder_hdr off, the rebuild runs out of
> room: skb_set_mac_header(skb, -skb->mac_len) computes a negative
> offset, stores it unchecked in the u16 skb->mac_header, and the
> memmove that follows writes skb->mac_len bytes about 64 KB past skb-
> >head. Forwarding plain ping6 traffic through such a device
> reproduces it on all five seg6 encapsulation modes and on the rpl
> and ioam6 inline paths; skb->mac_header comes back as 65534 on a 704-
> byte head.
> 
> Return the larger of the two.  The helper already returns skb-
> >mac_len when it has no dst, so this only makes the other branch
> agree, and it covers every caller rather than each call site in
> turn.
> 
> Fixes: 40475b63761a ("net: ipv6: seg6_iptunnel: mitigate 2-realloc
> issue") Fixes: dce525185bc9 ("net: ipv6: ioam6_iptunnel: mitigate 2-
> realloc issue") Fixes: 985ec6f5e623 ("net: ipv6: rpl_iptunnel:
> mitigate 2-realloc issue") Suggested-by: Andrea Mayer
> <andrea.mayer@uniroma2.it> Assisted-by: LLM Signed-off-by: Yuya
> Kusakabe <yuya.kusakabe@gmail.com> --- Changes in v3: - Fix
> dst_dev_overhead() itself rather than each of its callers [Andrea] -
> Drop Justin's Reviewed-by from v2, which was given for the per-
> caller version; the patch no longer touches the files he reviewed -
> Link to v2: https://lore.kernel.org/r/20260918-seg6-maclen-headroom-
> v2-1-4d370af55b2e@gmail.com
> 
> Changes in v2: - Use max_t(unsigned int, ...) rather than max()
> [Justin] - Make the same change in ioam6_iptunnel and rpl_iptunnel
> [Justin] - Reproduce and verify the fix on those two as well:
> unpatched, rpl underflows in 12 of 16 probed configurations and
> ioam6 in 3 of 13, both reaching about 64 KB past a 320-byte head;
> patched, neither underflows in 18 - Rewrap the requests so they no
> longer exceed 80 columns - Retitle for net: ipv6, now that the
> change spans three files - Link to v1: https://lore.kernel.org/
> r/20260917-seg6-maclen-headroom-v1-1-02ccec50f096@gmail.com --- 
> include/net/dst.h | 3 ++- 1 file changed, 2 insertions(+), 1
> deletion(-)
> 
> diff --git a/include/net/dst.h b/include/net/dst.h index
> 307073eae7f8..dbedfe72e1fd 100644 --- a/include/net/dst.h +++ b/
> include/net/dst.h @@ -455,7 +455,8 @@ static inline unsigned int
> dst_dev_overhead(struct dst_entry *dst, struct sk_buff *skb) { if
> (likely(dst)) -		return LL_RESERVED_SPACE(dst->dev); +		return
> max_t(unsigned int, skb->mac_len, +			     LL_RESERVED_SPACE(dst-
> >dev));
> 
> return skb->mac_len; }
> 

Reviewed-by: Justin Iurman <justin.iurman@gmail.com>

Re: [PATCH net v3] net: ipv6: keep room for the mac header in dst_dev_overhead()

[Gabriel Goller]

On 22.09.2026 05:49, Yuya Kusakabe wrote:
> The seg6, ioam6 and rpl lwtunnels size their skb_cow_head() request as
> the length they are about to push plus dst_dev_overhead(), then push the
> new headers and rebuild the mac header below them with
> skb_mac_header_rebuild().  That rebuild needs skb->mac_len of headroom,
> but dst_dev_overhead() leaves LL_RESERVED_SPACE() of the egress device,
> 16 bytes for plain Ethernet.
> 
> Where the mac header is longer than that, as it is on ingress through a
> VLAN device with reorder_hdr off, the rebuild runs out of room:
> skb_set_mac_header(skb, -skb->mac_len) computes a negative offset,
> stores it unchecked in the u16 skb->mac_header, and the memmove that
> follows writes skb->mac_len bytes about 64 KB past skb->head.
> Forwarding plain ping6 traffic through such a device reproduces it on
> all five seg6 encapsulation modes and on the rpl and ioam6 inline paths;
> skb->mac_header comes back as 65534 on a 704-byte head.
> 
> Return the larger of the two.  The helper already returns skb->mac_len
> when it has no dst, so this only makes the other branch agree, and it
> covers every caller rather than each call site in turn.
> 
> Fixes: 40475b63761a ("net: ipv6: seg6_iptunnel: mitigate 2-realloc issue")
> Fixes: dce525185bc9 ("net: ipv6: ioam6_iptunnel: mitigate 2-realloc issue")
> Fixes: 985ec6f5e623 ("net: ipv6: rpl_iptunnel: mitigate 2-realloc issue")
> Suggested-by: Andrea Mayer <andrea.mayer@uniroma2.it>
> Assisted-by: LLM
> Signed-off-by: Yuya Kusakabe <yuya.kusakabe@gmail.com>

LGTM

Did a quick test and could reproduce the issue:

FAIL seg6 inline nsegs=4
[    3.557169] BUG: KASAN: slab-use-after-free in __seg6_do_srh_inline+0x34b/0x620

Couldn't reproduce it after applying this patch.

Consider:
Reviewed-by: Gabriel Goller <g.goller@proxmox.com>
Tested-by: Gabriel Goller <g.goller@proxmox.com>

Re: [PATCH net v3] net: ipv6: keep room for the mac header in dst_dev_overhead()

[Eric Dumazet]

On Tue, Sep 22, 2026 at 1:24 PM Gabriel Goller <g.goller@proxmox.com> wrote:
>
> On 22.09.2026 05:49, Yuya Kusakabe wrote:
> > The seg6, ioam6 and rpl lwtunnels size their skb_cow_head() request as
> > the length they are about to push plus dst_dev_overhead(), then push the
> > new headers and rebuild the mac header below them with
> > skb_mac_header_rebuild().  That rebuild needs skb->mac_len of headroom,
> > but dst_dev_overhead() leaves LL_RESERVED_SPACE() of the egress device,
> > 16 bytes for plain Ethernet.
> >
> > Where the mac header is longer than that, as it is on ingress through a
> > VLAN device with reorder_hdr off, the rebuild runs out of room:
> > skb_set_mac_header(skb, -skb->mac_len) computes a negative offset,
> > stores it unchecked in the u16 skb->mac_header, and the memmove that
> > follows writes skb->mac_len bytes about 64 KB past skb->head.
> > Forwarding plain ping6 traffic through such a device reproduces it on
> > all five seg6 encapsulation modes and on the rpl and ioam6 inline paths;
> > skb->mac_header comes back as 65534 on a 704-byte head.
> >
> > Return the larger of the two.  The helper already returns skb->mac_len
> > when it has no dst, so this only makes the other branch agree, and it
> > covers every caller rather than each call site in turn.
> >
> > Fixes: 40475b63761a ("net: ipv6: seg6_iptunnel: mitigate 2-realloc issue")
> > Fixes: dce525185bc9 ("net: ipv6: ioam6_iptunnel: mitigate 2-realloc issue")
> > Fixes: 985ec6f5e623 ("net: ipv6: rpl_iptunnel: mitigate 2-realloc issue")
> > Suggested-by: Andrea Mayer <andrea.mayer@uniroma2.it>
> > Assisted-by: LLM
> > Signed-off-by: Yuya Kusakabe <yuya.kusakabe@gmail.com>

Reviewed-by: Eric Dumazet <edumazet@google.com>

Re: [PATCH net v3] net: ipv6: keep room for the mac header in dst_dev_overhead()

[Andrea Mayer]

On Tue, 22 Sep 2026 05:49:56 +0900
Yuya Kusakabe <yuya.kusakabe@gmail.com> wrote:

> The seg6, ioam6 and rpl lwtunnels size their skb_cow_head() request as
> the length they are about to push plus dst_dev_overhead(), then push the
> new headers and rebuild the mac header below them with
> skb_mac_header_rebuild().  That rebuild needs skb->mac_len of headroom,
> but dst_dev_overhead() leaves LL_RESERVED_SPACE() of the egress device,
> 16 bytes for plain Ethernet.
> 
> Where the mac header is longer than that, as it is on ingress through a
> VLAN device with reorder_hdr off, the rebuild runs out of room:
> skb_set_mac_header(skb, -skb->mac_len) computes a negative offset,
> stores it unchecked in the u16 skb->mac_header, and the memmove that
> follows writes skb->mac_len bytes about 64 KB past skb->head.
> Forwarding plain ping6 traffic through such a device reproduces it on
> all five seg6 encapsulation modes and on the rpl and ioam6 inline paths;
> skb->mac_header comes back as 65534 on a 704-byte head.
> 
> Return the larger of the two.  The helper already returns skb->mac_len
> when it has no dst, so this only makes the other branch agree, and it
> covers every caller rather than each call site in turn.
> 
> Fixes: 40475b63761a ("net: ipv6: seg6_iptunnel: mitigate 2-realloc issue")
> Fixes: dce525185bc9 ("net: ipv6: ioam6_iptunnel: mitigate 2-realloc issue")
> Fixes: 985ec6f5e623 ("net: ipv6: rpl_iptunnel: mitigate 2-realloc issue")
> Suggested-by: Andrea Mayer <andrea.mayer@uniroma2.it>
> Assisted-by: LLM
> Signed-off-by: Yuya Kusakabe <yuya.kusakabe@gmail.com>
> ---
> Changes in v3:
> - Fix dst_dev_overhead() itself rather than each of its callers [Andrea]
> - Drop Justin's Reviewed-by from v2, which was given for the per-caller
>   version; the patch no longer touches the files he reviewed
> - Link to v2: https://lore.kernel.org/r/20260918-seg6-maclen-headroom-v2-1-4d370af55b2e@gmail.com
> 
> Changes in v2:
> - Use max_t(unsigned int, ...) rather than max() [Justin]
> - Make the same change in ioam6_iptunnel and rpl_iptunnel [Justin]
> - Reproduce and verify the fix on those two as well: unpatched, rpl
>   underflows in 12 of 16 probed configurations and ioam6 in 3 of 13,
>   both reaching about 64 KB past a 320-byte head; patched, neither
>   underflows in 18
> - Rewrap the requests so they no longer exceed 80 columns
> - Retitle for net: ipv6, now that the change spans three files
> - Link to v1: https://lore.kernel.org/r/20260917-seg6-maclen-headroom-v1-1-02ccec50f096@gmail.com
> ---
>  include/net/dst.h | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)

Thanks for the rework,
Andrea

Reviewed-by: Andrea Mayer <andrea.mayer@uniroma2.it>