Re: [B.A.T.M.A.N.] [PATCH] batman-adv: fix null pointer dereference in batadv_gw_election

From: Sven Eckelmann
Date: Fri Nov 30 2018 - 02:49:33 EST


On Friday, 30 November 2018 15:00:02 CET Wen Yang wrote:
> This patch fixes a possible null pointer dereference in
> batadv_gw_election, detected by the semantic patch
> deref_null.cocci, with the following warning:
>
> ./net/batman-adv/gateway_client.c:289:15-24: ERROR: next_gw is NULL but dereferenced.
> ./net/batman-adv/gateway_client.c:290:15-29: ERROR: next_gw is NULL but dereferenced.
> ./net/batman-adv/gateway_client.c:291:15-29: ERROR: next_gw is NULL but dereferenced.
> ./net/batman-adv/gateway_client.c:292:15-27: ERROR: next_gw is NULL but dereferenced.
> ./net/batman-adv/gateway_client.c:293:15-27: ERROR: next_gw is NULL but dereferenced.

This patch is seems to be nonsensical. next_gw cannot be NULL at this point
(let us call this location in the code "4."). Let us go through the code

// 1. when both are NULL then it would jump out of the the function.
if (curr_gw == next_gw)
goto out;

[...]

if (curr_gw && !next_gw) {
[...]
// 2. this handles the only valid case when next_gw is NULL
} else if (!curr_gw && next_gw) {
// 3. here we know that next_gw is not NULL and curr_gw is NULL
// we can therefore infer that
[...]
} else {
// 4. here you try to add an ugly patch to handle a non-existing
// next_gw == NULL case
[...]
}

Let us go through all possible combinations:

curr_gw next_gw
I 0 0
II x 0
III 0 y
IV x y

For I: we would leave the function even at 1. and never reach 4.
For II: would be handled by 2. and thus never reach 4.
For III: would be handled by 3. and thus never reach 4.
For IV: This can be handled by 1. (when x == y). Or otherwise it would be
handled by 4. but is not the next_gw == NULL case.

Please correct me (in case I missed something) but it looks to me that this
patch should be rejected.

Kind regards,
Sven

Attachment: signature.asc
Description: This is a digitally signed message part.