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/broadcom/brcm80211/brcmfmac/cfg80211.c
+++ b/drivers/net/wireless/broadcom/brcm80211/brcmfmac/cfg80211.c
@@ -980,8 +980,36 @@ static struct wireless_dev *brcmf_cfg802
 	struct brcmf_cfg80211_info *cfg = wiphy_to_cfg(wiphy);
 	struct brcmf_pub *drvr = cfg->pub;
 	struct wireless_dev *wdev;
+	struct net_device *dev;
 	int err;
 
+	/*
+	 * 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(-ENFILE);
+	}
+
 	brcmf_dbg(TRACE, "enter: %s type %d\n", name, type);
 	err = brcmf_vif_add_validate(wiphy_to_cfg(wiphy), type);
 	if (err) {