mac80211: try to workaround some brcmfmac bug with handling interfaces

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>

SVN-Revision: 46278
This commit is contained in:
Rafał Miłecki 2015-07-09 06:09:50 +00:00
parent f60082f548
commit 4587e39fe7

View File

@ -0,0 +1,49 @@
From: =?UTF-8?q?Rafa=C5=82=20Mi=C5=82ecki?= <zajec5@gmail.com>
Date: Thu, 9 Jul 2015 00:07:59 +0200
Subject: [PATCH] brcmfmac: workaround bug with some inconsistent BSSes state
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
---
--- a/drivers/net/wireless/brcm80211/brcmfmac/cfg80211.c
+++ b/drivers/net/wireless/brcm80211/brcmfmac/cfg80211.c
@@ -662,8 +662,36 @@ static struct wireless_dev *brcmf_cfg80211_add_iface(struct wiphy *wiphy,
u32 *flags,
struct vif_params *params)
{
+ struct net_device *dev;
struct wireless_dev *wdev;
+ /*
+ * There is a bug with in-firmware BSS management. When adding virtual
+ * interface brcmfmac first tells firmware to create new BSS and then
+ * it creates new struct net_device.
+ *
+ * If creating/registering netdev(ice) fails, BSS remains in some bugged
+ * state. It conflicts with existing BSSes by overtaking their auth
+ * requests.
+ *
+ * It results in one BSS (addresss X) sending beacons and another BSS
+ * (address Y) replying to authentication requests. This makes interface
+ * unusable as AP.
+ *
+ * To workaround this bug we may try to guess if register_netdev(ice)
+ * will fail. The most obvious case is using interface name that already
+ * exists. This is actually quite likely with brcmfmac & some user space
+ * scripts as brcmfmac doesn't allow deleting virtual interfaces.
+ * So this bug can be triggered even by something trivial like:
+ * iw dev wlan0 delete
+ * iw phy phy0 interface add wlan0 type __ap
+ */
+ dev = dev_get_by_name(&init_net, name);
+ if (dev) {
+ dev_put(dev);
+ return ERR_PTR(-EEXIST);
+ }
+
brcmf_dbg(TRACE, "enter: %s type %d\n", name, type);
switch (type) {
case NL80211_IFTYPE_ADHOC: