Re: [PATCH] libertas: Fix two buffer overflows at parsing bss descriptor

[Nicolai Stange <nstange@suse.de>]

Hi,

the patch queued as e5e884b42639 ("libertas: Fix two buffer overflows at
parsing bss descriptor") at the wireless tree doesn't look completely
correct to me.

This hunk here...

diff --git a/drivers/net/wireless/marvell/libertas/cfg.c b/drivers/net/wireless/marvell/libertas/cfg.c
index 57edfada0665..c9401c121a14 100644
--- a/drivers/net/wireless/marvell/libertas/cfg.c
+++ b/drivers/net/wireless/marvell/libertas/cfg.c
@@ -1775,9 +1782,12 @@ static int lbs_ibss_join_existing(struct lbs_private *priv,
 	if (!rates_eid) {
 		lbs_add_rates(cmd.bss.rates);
 	} else {
-		int hw, i;
-		u8 rates_max = rates_eid[1];
-		u8 *rates = cmd.bss.rates;
+		rates_max = rates_eid[1];
+		if (rates_max > MAX_RATES) {
+			lbs_deb_join("invalid rates");
+			goto out;

... makes the error path skip over the rcu_read_unlock() following later
in the code and leaves the RCU read section unbalanced.

Also, I'm wondering if ret should perhaps get set to some -Exxxx in case
of rates_max > MAX_RATES?

Thanks,

Nicolai


+		}
+		rates = cmd.bss.rates;
 		for (hw = 0; hw < ARRAY_SIZE(lbs_rates); hw++) {
 			u8 hw_rate = lbs_rates[hw].bitrate / 5;
 			for (i = 0; i < rates_max; i++) {

-- 
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg, Germany
(HRB 36809, AG Nürnberg), GF: Felix Imendörffer