libnetfilter_conntrack endian issue

libnetfilter_conntrack endian issue

[Albert Veli]

[-- Attachment #1: Type: text/plain, Size: 747 bytes --]

Hi!

I run contrack-tools on a big-endian arm machine and had a problem:

$ conntrack -L
conntrack v0.9.7: Operation failed: Address family not supported by protocol

The problem was in libnetfilter_conntrack-0.0.96/src/conntrack/api.c.
See attached endian.patch. In nfct_build_query() the *data argument is
converted into a u_int8_t*. This works for little-endian but not for
big-endian. The fix is a one-liner. Replace

 const u_int8_t *family = data;

with

 const u_int32_t *family = data;

After the patch it works:

$ conntrack -L
conntrack v0.9.7: 0 flow entries has been shown.

I also tested the patch on my little-endian laptop and it doesn't seem
to cause any regression there. Just wanted to let you guys know.


Best regards,

Albert

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #2: endian.patch --]
[-- Type: text/x-diff; name=endian.patch, Size: 496 bytes --]

diff -urN libnetfilter_conntrack-0.0.96-org/src/conntrack/api.c libnetfilter_conntrack-0.0.96/src/conntrack/api.c
--- libnetfilter_conntrack-0.0.96-org/src/conntrack/api.c	2008-07-02 15:25:38.000000000 +0200
+++ libnetfilter_conntrack-0.0.96/src/conntrack/api.c	2008-07-29 11:42:43.000000000 +0200
@@ -455,7 +455,7 @@
 		     unsigned int size)
 {
 	struct nfnlhdr *req = buffer;
-	const u_int8_t *family = data;
+	const u_int32_t *family = data;
 
 	assert(ssh != NULL);
 	assert(data != NULL);

Re: libnetfilter_conntrack endian issue

[Pablo Neira Ayuso]

Albert Veli wrote:
> Hi!
> 
> I run contrack-tools on a big-endian arm machine and had a problem:
> 
> $ conntrack -L
> conntrack v0.9.7: Operation failed: Address family not supported by protocol
> 
> The problem was in libnetfilter_conntrack-0.0.96/src/conntrack/api.c.
> See attached endian.patch. In nfct_build_query() the *data argument is
> converted into a u_int8_t*. This works for little-endian but not for
> big-endian. The fix is a one-liner. Replace
> 
>  const u_int8_t *family = data;
> 
> with
> 
>  const u_int32_t *family = data;
> 
> After the patch it works:
> 
> $ conntrack -L
> conntrack v0.9.7: 0 flow entries has been shown.
> 
> I also tested the patch on my little-endian laptop and it doesn't seem
> to cause any regression there. Just wanted to let you guys know.

Applied. Thanks Albert.

-- 
"Los honestos son inadaptados sociales" -- Les Luthiers