Skip to contentSkip to conversation

Revert "netfilter: conntrack: fix bug in for_each_sctp_chunk"

PATCH In review Florian Westphal 2 participants 2 messages Jan 26, 2023

Subsystems (2)
lore
Share & tools
lore

Florian Westphal

author maintainer
#0 lore raw

PATCH 1/1 Revert "netfilter: conntrack: fix bug in for_each_sctp_chunk"

There is no bug. If sch->length == 0, this would result in an infinite loop, but first caller, do_basic_checks(), errors out in this case.

After this change, packets with bogus zero-length chunks are no longer detected as invalid, so revert & add comment wrt. 0 length check.

Fixes 98ee00774525 ("netfilter: conntrack: fix bug in for_each_sctp_chunk")

---
It might be a good idea to merge do_basic_checks and sctp_error to avoid future patches adding for_each_sctp_chunk() earlier in the pipeline.

 net/netfilter/nf_conntrack_proto_sctp.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
net/netfilter/nf_conntrack_proto_sctp.c +3 -2
index 945dd40e7077..2f4459478750 100644
@@ -142,10 +142,11 @@ static void sctp_print_conntrack(struct seq_file *s, struct nf_conn *ct)
 }
 #endif
 
+/* do_basic_checks ensures sch->length > 0, do not use before */
 #define for_each_sctp_chunk(skb, sch, _sch, offset, dataoff, count)	\
 for ((offset) = (dataoff) + sizeof(struct sctphdr), (count) = 0;	\
-	((sch) = skb_header_pointer((skb), (offset), sizeof(_sch), &(_sch))) &&	\
-	(sch)->length;	\
+	(offset) < (skb)->len &&					\
+	((sch) = skb_header_pointer((skb), (offset), sizeof(_sch), &(_sch)));	\
 	(offset) += (ntohs((sch)->length) + 3) & ~3, (count)++)
 
 /* Some validity checks to make sure the chunks are fine */
Signature · 1 line
2.39.1
#1 lore raw
Quoted 9 lines

-----Original Message-----
From: Florian Westphal <[email protected]>
Sent: Thursday, 26 January 2023 02:35
To: [email protected]
Cc: Florian Westphal <[email protected]>
Subject: [PATCH nf] Revert "netfilter: conntrack: fix bug in for_each_sctp_chunk"

Quoted 3 lines

There is no bug. If sch->length == 0, this would result in an infinite loop, but first caller, do_basic_checks(), errors out in this case.

Ah yes, perhaps your comment was for [1], where for_each_sctp_chunk() is used only once in nf_conntrack_sctp_packet(), and I didn't check for (sch)->length. The (sch)->length check in do_basic_checks() was almost invisible to me :(

[1] https://patchwork.ozlabs.org/project/netfilter-devel/patch/[email protected]/

Quoted 34 lines

After this change, packets with bogus zero-length chunks are no longer detected as invalid, so revert & add comment wrt. 0 length check.

Fixes 98ee00774525 ("netfilter: conntrack: fix bug in for_each_sctp_chunk")

---
It might be a good idea to merge do_basic_checks and sctp_error to avoid
future patches adding for_each_sctp_chunk() earlier in the pipeline.

 net/netfilter/nf_conntrack_proto_sctp.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)
net/netfilter/nf_conntrack_proto_sctp.c +0 -0
index 945dd40e7077..2f4459478750 100644
@@ -142,10 +142,11 @@ static void sctp_print_conntrack(struct seq_file *s,

struct nf_conn *ct) } #endif

net/netfilter/nf_conntrack_proto_sctp.c +1 -1
+/* do_basic_checks ensures sch->length > 0, do not use before */
 #define for_each_sctp_chunk(skb, sch, _sch, offset, dataoff, count)	\
 for ((offset) = (dataoff) + sizeof(struct sctphdr), (count) = 0;	\
-	((sch) = skb_header_pointer((skb), (offset), sizeof(_sch), &(_sch))) &&
	\
net/netfilter/nf_conntrack_proto_sctp.c +2 -1
-	(sch)->length;	\
+	(offset) < (skb)->len &&					\
+	((sch) = skb_header_pointer((skb), (offset), sizeof(_sch), &(_sch)));
	\

(offset) += (ntohs((sch)->length) + 3) & ~3, (count)++)

/* Some validity checks to make sure the chunks are fine */
--
2.39.1