| Message ID | 20180114222255.GA9036@light.dominikbrodowski.net (mailing list archive) |
|---|---|
| State | Superseded |
| Delegated to: | Johannes Berg |
| Headers | show |
On Sun, 2018-01-14 at 23:22 +0100, Dominik Brodowski wrote: > > + rcu_read_lock(); > ssid_ie = ieee80211_bss_get_ie(&wdev->current_bss->pub, > WLAN_EID_SSID); > if (!ssid_ie) > - break; nit-picking now: that "break" here may have been easier before these changes > + goto nla_rcu_unlock; > if (nla_put(msg, NL80211_ATTR_SSID, ssid_ie[1], ssid_ie + 2)) > - goto nla_put_failure_locked; > + goto nla_put_failure_rcu_locked; > + nla_rcu_unlock: > + rcu_read_unlock(); > break; but after, perhaps it's easier to just do if (ssid_ie && nla_put(...) goto nla_put_failure_rcu_locked; and avoid the extra label (but yeah, it's getting late) johannes
diff --git a/net/wireless/nl80211.c b/net/wireless/nl80211.c index 2b3dbcd40e46..b53bd8db7974 100644 --- a/net/wireless/nl80211.c +++ b/net/wireless/nl80211.c @@ -2618,12 +2618,15 @@ static int nl80211_send_iface(struct sk_buff *msg, u32 portid, u32 seq, int flag const u8 *ssid_ie; if (!wdev->current_bss) break; + rcu_read_lock(); ssid_ie = ieee80211_bss_get_ie(&wdev->current_bss->pub, WLAN_EID_SSID); if (!ssid_ie) - break; + goto nla_rcu_unlock; if (nla_put(msg, NL80211_ATTR_SSID, ssid_ie[1], ssid_ie + 2)) - goto nla_put_failure_locked; + goto nla_put_failure_rcu_locked; + nla_rcu_unlock: + rcu_read_unlock(); break; } default: @@ -2635,6 +2638,8 @@ static int nl80211_send_iface(struct sk_buff *msg, u32 portid, u32 seq, int flag genlmsg_end(msg, hdr); return 0; + nla_put_failure_rcu_locked: + rcu_read_unlock(); nla_put_failure_locked: wdev_unlock(wdev); nla_put_failure:
As ieee80211_bss_get_ie() derefences an RCU, it needs to be called with rcu_read_lock held. Fixes: 44905265bc15 ("nl80211: don't expose wdev->ssid for most interfaces") Signed-off-by: Dominik Brodowski <linux@dominikbrodowski.net> --- > This uses the ssid_ie, so that doesn't really seem right? The > protection should extend beyond the usage. Indeed -- I had misread the code and hadn't thought of ssid_ie also needing the protection during its lifetime. So here's a new version 2 -- which I will only be able to test tomorrow, though... Thanks, Dominik