* [PATCH 0/1] sctp: check return value of sctp_get_af_specific in sctp_process_init
@ 2026-09-18 3:39 Yuchao Zhang
2026-09-18 3:39 ` [PATCH 1/1] " Yuchao Zhang
0 siblings, 1 reply; 3+ messages in thread
From: Yuchao Zhang @ 2026-09-18 3:39 UTC (permalink / raw)
To: Marcelo Ricardo Leitner, Xin Long
Cc: David S . Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, linux-sctp, netdev, linux-kernel, Yuchao Zhang
This patch fixes a NULL pointer dereference in sctp_process_init()
that a remote peer can trigger on kernels built with CONFIG_IPV6=n.
Problem description:
In sctp_process_init(), when a COOKIE ECHO arrives from a source
address different from the one recorded in its cookie, the embedded
address parameters of the INIT chunk carried in the cookie are walked
to re-validate the source:
af = sctp_get_af_specific(param_type2af(param.p->type));
if (!af->from_addr_param(&addr, param.addr,
chunk->sctp_hdr->source, 0))
continue;
On CONFIG_IPV6=n kernels the IPv6 address family is never registered
and sctp_get_af_specific(AF_INET6) returns NULL. A peer that completes
the INIT/INIT ACK exchange over IPv4 with an SCTP_PARAM_IPV6_ADDRESS
parameter embedded in its INIT, and then sends the COOKIE ECHO from a
different source address (as a multihomed peer legitimately may), makes
the walk dereference NULL, crashing the kernel in softirq context
without any authentication.
Fix:
Add an explicit NULL check for af in sctp_process_init() to skip
unsupported address family parameters, aligning with the other
from_addr_param call sites in net/sctp/input.c and
net/sctp/sm_make_chunk.c.
Tested and verified style-compliant with checkpatch.pl
(0 errors, 0 warnings).
Yuchao Zhang (1):
sctp: check return value of sctp_get_af_specific in sctp_process_init
net/sctp/sm_make_chunk.c | 2 ++
1 file changed, 2 insertions(+)
--
2.53.0
^ permalink raw reply [flat|nested] 3+ messages in thread
* [PATCH 1/1] sctp: check return value of sctp_get_af_specific in sctp_process_init
2026-09-18 3:39 [PATCH 0/1] sctp: check return value of sctp_get_af_specific in sctp_process_init Yuchao Zhang
@ 2026-09-18 3:39 ` Yuchao Zhang
2026-09-18 15:32 ` Xin Long
0 siblings, 1 reply; 3+ messages in thread
From: Yuchao Zhang @ 2026-09-18 3:39 UTC (permalink / raw)
To: Marcelo Ricardo Leitner, Xin Long
Cc: David S . Miller, Eric Dumazet, Jakub Kicinski, Paolo Abeni,
Simon Horman, linux-sctp, netdev, linux-kernel, Yuchao Zhang,
stable
In sctp_process_init(), when matching the source address of a COOKIE
ECHO against the embedded address parameters of the INIT chunk carried
in the cookie, the address family handler is dereferenced without a
NULL check:
af = sctp_get_af_specific(param_type2af(param.p->type));
if (!af->from_addr_param(&addr, param.addr,
chunk->sctp_hdr->source, 0))
continue;
On kernels built without IPv6 (CONFIG_IPV6=n) sctp_af_v6_specific is
never registered and sctp_get_af_specific(AF_INET6) returns NULL. A
remote peer can reach this path by completing a regular INIT/INIT ACK
exchange over IPv4 with an SCTP_PARAM_IPV6_ADDRESS parameter embedded
in its INIT (ignored at INIT time by the address family gate in
sctp_process_param(), but the cookie embeds the whole INIT chunk), and
then sending the COOKIE ECHO from a different source address, e.g. a
second interface of a multihomed peer. sctp_unpack_cookie() does not
compare the packet source address against the peer address stored in
the cookie, so src_match is 0 and the walk dereferences the NULL af.
This crashes the kernel with a NULL pointer dereference in softirq
context, remotely and without authentication.
The other from_addr_param call sites of the same family already check
af (__sctp_rcv_init_lookup() and the ASCONF path in
sctp_process_param(), fixed by commit e40607cbe270 ("net: sctp: fix
NULL pointer dereference in af->from_addr_param on malformed packet")),
which missed this one.
Fixes: 0c5dc070ff3d ("sctp: validate from_addr_param return")
Cc: stable@vger.kernel.org
Signed-off-by: Yuchao Zhang <ndaugoing@gmail.com>
---
net/sctp/sm_make_chunk.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/net/sctp/sm_make_chunk.c b/net/sctp/sm_make_chunk.c
index 84a4c97d0f75..ff1c013e2575 100644
--- a/net/sctp/sm_make_chunk.c
+++ b/net/sctp/sm_make_chunk.c
@@ -2381,6 +2381,8 @@ int sctp_process_init(struct sctp_association *asoc, struct sctp_chunk *chunk,
(param.p->type == SCTP_PARAM_IPV4_ADDRESS ||
param.p->type == SCTP_PARAM_IPV6_ADDRESS)) {
af = sctp_get_af_specific(param_type2af(param.p->type));
+ if (!af)
+ continue;
if (!af->from_addr_param(&addr, param.addr,
chunk->sctp_hdr->source, 0))
continue;
--
2.53.0
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH 1/1] sctp: check return value of sctp_get_af_specific in sctp_process_init
2026-09-18 3:39 ` [PATCH 1/1] " Yuchao Zhang
@ 2026-09-18 15:32 ` Xin Long
0 siblings, 0 replies; 3+ messages in thread
From: Xin Long @ 2026-09-18 15:32 UTC (permalink / raw)
To: Yuchao Zhang
Cc: Marcelo Ricardo Leitner, David S . Miller, Eric Dumazet,
Jakub Kicinski, Paolo Abeni, Simon Horman, linux-sctp, netdev,
linux-kernel, stable
On Thu, Sep 17, 2026 at 11:39 PM Yuchao Zhang <ndaugoing@gmail.com> wrote:
>
> In sctp_process_init(), when matching the source address of a COOKIE
> ECHO against the embedded address parameters of the INIT chunk carried
> in the cookie, the address family handler is dereferenced without a
> NULL check:
>
> af = sctp_get_af_specific(param_type2af(param.p->type));
> if (!af->from_addr_param(&addr, param.addr,
> chunk->sctp_hdr->source, 0))
> continue;
>
> On kernels built without IPv6 (CONFIG_IPV6=n) sctp_af_v6_specific is
> never registered and sctp_get_af_specific(AF_INET6) returns NULL. A
> remote peer can reach this path by completing a regular INIT/INIT ACK
> exchange over IPv4 with an SCTP_PARAM_IPV6_ADDRESS parameter embedded
> in its INIT (ignored at INIT time by the address family gate in
> sctp_process_param(), but the cookie embeds the whole INIT chunk), and
> then sending the COOKIE ECHO from a different source address, e.g. a
> second interface of a multihomed peer. sctp_unpack_cookie() does not
> compare the packet source address against the peer address stored in
> the cookie, so src_match is 0 and the walk dereferences the NULL af.
>
> This crashes the kernel with a NULL pointer dereference in softirq
> context, remotely and without authentication.
>
> The other from_addr_param call sites of the same family already check
> af (__sctp_rcv_init_lookup() and the ASCONF path in
> sctp_process_param(), fixed by commit e40607cbe270 ("net: sctp: fix
> NULL pointer dereference in af->from_addr_param on malformed packet")),
> which missed this one.
>
> Fixes: 0c5dc070ff3d ("sctp: validate from_addr_param return")
> Cc: stable@vger.kernel.org
> Signed-off-by: Yuchao Zhang <ndaugoing@gmail.com>
> ---
> net/sctp/sm_make_chunk.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/net/sctp/sm_make_chunk.c b/net/sctp/sm_make_chunk.c
> index 84a4c97d0f75..ff1c013e2575 100644
> --- a/net/sctp/sm_make_chunk.c
> +++ b/net/sctp/sm_make_chunk.c
> @@ -2381,6 +2381,8 @@ int sctp_process_init(struct sctp_association *asoc, struct sctp_chunk *chunk,
> (param.p->type == SCTP_PARAM_IPV4_ADDRESS ||
> param.p->type == SCTP_PARAM_IPV6_ADDRESS)) {
> af = sctp_get_af_specific(param_type2af(param.p->type));
> + if (!af)
> + continue;
> if (!af->from_addr_param(&addr, param.addr,
> chunk->sctp_hdr->source, 0))
> continue;
> --
> 2.53.0
>
I think this is a dup of
https://lore.kernel.org/netdev/20260917030505.4176635-1-hdthky0@gmail.com/
Thanks.
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-09-18 15:32 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-09-18 3:39 [PATCH 0/1] sctp: check return value of sctp_get_af_specific in sctp_process_init Yuchao Zhang
2026-09-18 3:39 ` [PATCH 1/1] " Yuchao Zhang
2026-09-18 15:32 ` Xin Long
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox
all inboxes | Powered by JetHome®