summaryrefslogtreecommitdiffstats
path: root/main/kamailio/0001-modules_k-uac-fix-from-to-restore-for-small-original.patch
diff options
context:
space:
mode:
authorTimo Teräs <timo.teras@iki.fi>2011-04-06 09:49:23 +0300
committerTimo Teräs <timo.teras@iki.fi>2011-04-06 09:50:39 +0300
commit2bf9bd38eb8685a170d201fa253868968f40d125 (patch)
tree20f353f16618e4f4d07886aab8c6a9a492c2fcb5 /main/kamailio/0001-modules_k-uac-fix-from-to-restore-for-small-original.patch
parent8f192e3b4f2121daad608ebdf7eb4a37514f0a74 (diff)
downloadaports-2bf9bd38eb8685a170d201fa253868968f40d125.tar.bz2
aports-2bf9bd38eb8685a170d201fa253868968f40d125.tar.xz
main/kamailio: fix a from header rewriting bug
Diffstat (limited to 'main/kamailio/0001-modules_k-uac-fix-from-to-restore-for-small-original.patch')
-rw-r--r--main/kamailio/0001-modules_k-uac-fix-from-to-restore-for-small-original.patch64
1 files changed, 64 insertions, 0 deletions
diff --git a/main/kamailio/0001-modules_k-uac-fix-from-to-restore-for-small-original.patch b/main/kamailio/0001-modules_k-uac-fix-from-to-restore-for-small-original.patch
new file mode 100644
index 000000000..1b997719b
--- /dev/null
+++ b/main/kamailio/0001-modules_k-uac-fix-from-to-restore-for-small-original.patch
@@ -0,0 +1,64 @@
+From e22eb2886c73634020c2747d6247df6bcb978850 Mon Sep 17 00:00:00 2001
+From: =?UTF-8?q?Timo=20Ter=C3=A4s?= <timo.teras@iki.fi>
+Date: Wed, 6 Apr 2011 09:33:10 +0300
+Subject: [PATCH] modules_k/uac: fix from/to restore for small original URI
+MIME-Version: 1.0
+Content-Type: text/plain; charset=UTF-8
+Content-Transfer-Encoding: 8bit
+
+Seems that the URI length check is superfluous and fails under
+certain conditions. It does not make sense for the URI to have
+zero bytes, so just use the first seen zero byte as end marker.
+
+I have a reproducible test case where the restore inserts URI
+with multiple zero-bytes to wire. This happens if the original
+URI is smaller than the one we rewrote it to using uac_replace_from.
+
+Signed-off-by: Timo Teräs <timo.teras@iki.fi>
+---
+ modules_k/uac/from.c | 14 ++++++++------
+ 1 files changed, 8 insertions(+), 6 deletions(-)
+
+However, I think the delta encoding used for the RR attribute
+is flawed. Hostile remote server could rewrite the RR attribute
+and/or From/To headers in a way to forge it to something it was not
+in the first place. Additionally the delta-encoded RR attribute
+breaks if the From/To header isn't exact copy of what we sent.
+
+Would it not make more sense to just send the real original
+header (possibly encrypted) but with a checksum? We could then
+verify if someone had clobbered the RR attribute and ignore it.
+And we could always restore the original URI even if the URI
+we are swapping was modified unexpectedly.
+
+diff --git a/modules_k/uac/from.c b/modules_k/uac/from.c
+index 4657e11..50822b6 100644
+--- a/modules_k/uac/from.c
++++ b/modules_k/uac/from.c
+@@ -463,15 +463,17 @@ int restore_from( struct sip_msg *msg, int *is_from )
+ LM_ERR("new URI shorter than old URI\n");
+ goto failed;
+ }
+- for( i=0 ; i<old_uri.len ; i++ )
++ for( i=0 ; i<old_uri.len ; i++ ) {
+ new_uri.s[i] ^= old_uri.s[i];
+- if (new_uri.len==old_uri.len) {
+- for( ; new_uri.len && (new_uri.s[new_uri.len-1]==0) ; new_uri.len-- );
+- if (new_uri.len==0) {
+- LM_ERR("new URI got 0 len\n");
+- goto failed;
++ if (new_uri.s[i] == 0) {
++ new_uri.len = i;
++ break;
+ }
+ }
++ if (new_uri.len==0) {
++ LM_ERR("new URI got 0 len\n");
++ goto failed;
++ }
+
+ LM_DBG("decoded uris are: new=[%.*s] old=[%.*s]\n",
+ new_uri.len, new_uri.s, old_uri.len, old_uri.s);
+--
+1.7.1
+