2019-10-27 21:03:33

by Navid Emamdoost

Subject: [PATCH] cfg80211: Fix memory leak in cfg80211_inform_single_bss_frame_data

In the implementation of cfg80211_inform_single_bss_frame_data() the
allocated memory for ies is leaked in case of an error. Release ies if
cfg80211_bss_update() fails.

Fixes: 2a5193119269 ("cfg80211/nl80211: scanning (and mac80211 update to use it)")
Signed-off-by: Navid Emamdoost <[email protected]>
---
net/wireless/scan.c | 4 +++-
1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/net/wireless/scan.c b/net/wireless/scan.c
index aef240fdf8df..fae5af24f668 100644
--- a/net/wireless/scan.c
+++ b/net/wireless/scan.c
@@ -1856,8 +1856,10 @@ cfg80211_inform_single_bss_frame_data(struct wiphy *wiphy,
wiphy->max_adj_channel_rssi_comp;
res = cfg80211_bss_update(wiphy_to_rdev(wiphy), &tmp, signal_valid,
jiffies);
- if (!res)
+ if (!res) {
+ kfree(ies);
return NULL;
+ }

if (channel->band == NL80211_BAND_60GHZ) {
bss_type = res->pub.capability & WLAN_CAPABILITY_DMG_TYPE_MASK;
--
2.17.1

2019-10-28 20:07:38

by Johannes Berg

Subject: Re: [PATCH] cfg80211: Fix memory leak in cfg80211_inform_single_bss_frame_data

On Sun, 2019-10-27 at 13:15 -0500, Navid Emamdoost wrote:
> In the implementation of cfg80211_inform_single_bss_frame_data() the
> allocated memory for ies is leaked in case of an error. Release ies if
> cfg80211_bss_update() fails.

I'm pretty sure it's more complicated than this patch (and the previous
one) - we already do free this at least in the case that "new =
kzalloc(...)" fails in cfg80211_bss_update().

Your Fixes: tag is also wrong, back then we didn't even have the dynamic
allocation of the IEs.

I'm dropping this patch and the other and will make a note to eventually
look at the lifetime issue here you point out, but if you want to work
on it instead let me know.

johannes

2019-10-28 21:06:36

by Navid Emamdoost

Subject: Re: [PATCH] cfg80211: Fix memory leak in cfg80211_inform_single_bss_frame_data

On Mon, Oct 28, 2019 at 7:12 AM Johannes Berg <[email protected]> wrote:
>
> On Sun, 2019-10-27 at 13:15 -0500, Navid Emamdoost wrote:
> > In the implementation of cfg80211_inform_single_bss_frame_data() the
> > allocated memory for ies is leaked in case of an error. Release ies if
> > cfg80211_bss_update() fails.
>
> I'm pretty sure it's more complicated than this patch (and the previous
> one) - we already do free this at least in the case that "new =
> kzalloc(...)" fails in cfg80211_bss_update().

Now I agree, there are two cases of release for ies in cfg80211_bss_update().
But for future reference, I suspect the last two cases of goto drop to
be mishandling ies.


>
> Your Fixes: tag is also wrong, back then we didn't even have the dynamic
> allocation of the IEs.
>
> I'm dropping this patch and the other and will make a note to eventually
> look at the lifetime issue here you point out, but if you want to work
> on it instead let me know.
>
> johannes
>


--
Navid.