Catches outdated/invalid RTATTR_MAX and therefore avoids possible stack
corruption.
Signed-off-by: Thomas Graf <tgraf@suug.ch>
--- linux-2.6.9-rc1-bk15.orig/net/core/rtnetlink.c 2004-09-08 18:33:42.000000000 +0200
+++ linux-2.6.9-rc1-bk15/net/core/rtnetlink.c 2004-09-09 18:18:22.000000000 +0200
@@ -450,6 +450,9 @@
sz_idx = type>>2;
kind = type&3;
+ if (RTATTR_MAX < rta_max[sz_idx])
+ BUG();
+
if (kind != 2 && security_netlink_recv(skb)) {
*errp = -EPERM;
return -1;On Thu, 9 Sep 2004 18:43:46 +0200
Thomas Graf [off-list ref] wrote:
Catches outdated/invalid RTATTR_MAX and therefore avoids possible stack
corruption.
Your test has an off by one error, but more importantly,
it's probably better to do this at compile time with
something like:
extern void rtattr_max_too_small(void);
...
void __init rtnetlink_init(void)
{
if (IFLA_MAX > RTATTR_MAX ||
IFA_MAX > RTATTR_MAX ||
RTA_MAX > RTATTR_MAX ||
NDA_MAX > RTATTR_MAX ||
TCA_MAX > RTATTR_MAX ||
TCAA_MAX > RTATTR_MAX)
rtattr_max_too_small();
I would therefore accept a patch that did things
this way. BUG()'ing at runtime for something like
this is too rude. :)
Thanks.
In article [off-list ref] (at Thu, 9 Sep 2004 18:43:46 +0200), Thomas Graf [off-list ref] says:
kind = type&3;
+ if (RTATTR_MAX < rta_max[sz_idx])
+ BUG();
+
Something like:
if (rta_max[sz_idx] > RTATTR_MAX)
goto err_inval;
--
Hideaki YOSHIFUJI @ USAGI Project [off-list ref]
GPG FP: 9022 65EB 1ECF 3AD1 0BDF 80D8 4807 F894 E062 0EEA