51 lines
1.8 KiB
Diff
51 lines
1.8 KiB
Diff
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
|
|
@@ -609,9 +609,37 @@ static struct wireless_dev *brcmf_cfg802
|
|
u32 *flags,
|
|
struct vif_params *params)
|
|
{
|
|
+ struct net_device *dev;
|
|
struct wireless_dev *wdev;
|
|
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(-EEXIST);
|
|
+ }
|
|
+
|
|
brcmf_dbg(TRACE, "enter: %s type %d\n", name, type);
|
|
err = brcmf_vif_add_validate(wiphy_to_cfg(wiphy), type);
|
|
if (err) {
|