Merge branch 'wireless'

John W. Linville says:

====================
For now I have dropped the mac80211 tree from this request.
We are developing a little backlog of fixes and I would like to
avoid introducing any more uncertainty to this pull request for the
3.10 stream.  All the other bits are the same as what was in the
2013-06-06 request, including the ath9k fixes intended to address
the problems observed by Linus w/ his Pixel, and a CVE fix for a
potential security issue in the b43 driver.

Regarding the wl12xx bits, Luca says:

"Here are three patches that I'd like to get into 3.10.  Two of them, by
me, are related to the firmware version checks in our driver.  Without
them, the firmwares fail to load.  The other one, by Eliad, fixes a typo
bug in our 5GHz scanning code."

And as for the Bluetooth bits, Gustavo says:

"The following patches are important bug fixes for 3.10, plus the
support for a new device. We do have three fixes from Johan. The first
one is a fix to avoid LE-only devices to rely on the (inexistent)
extended features data. The second patch fixes length checks on
incoming L2CAP signalling PDUs so we can discard PDU whose size
doesn't match the one reported in the header.  The last one fixes
the handling of power on failures, we now report proper errors to
mgmt when hci_dev_open()."

Along with that...

Larry Finger corrects an rtlwifi problem that caused some devices to
refuse to connect to non-WPA2 networks if the device had previously
assocated with a WPA2 network.  He also adds a one-line fix to prevent
false reports from kmemleak.

Mark A. Greer fixes an out of bounds array access in mwifiex.

Felix Fietkau reverts an earlier ath9k initval patch that reduced rx
sensitivity in a number of ath9k devices with no corresponding benefit.

Kees Cook fixes a potential uid-0 to ring-0 escalation in b43
(CVE-2013-2852).

Sujith Manoharan turns-off powersave mode by default for ath9k, and
also defaults ath9k to use the minstrel_ht rate control algorithm.
Both of these are believed to contribute to greater stability/usability
of ath9k in real-world situations.

Yijing Wang fixes an iwlegacy build error for il_pm_ops if CONFIG_PM
is set but CONFIG_PM_SLEEP is not set.
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 13 insertions, 0 deletions

diff --git a/drivers/net/wireless/rtlwifi/usb.c b/drivers/net/wireless/rtlwifi/usb.c
index 76732b0cd22..a3532e07787 100644
--- a/drivers/net/wireless/rtlwifi/usb.c
+++ b/drivers/net/wireless/rtlwifi/usb.c
@@ -824,6 +824,7 @@ static void rtl_usb_stop(struct ieee80211_hw *hw)
 
 	/* should after adapter start and interrupt enable. */
 	set_hal_stop(rtlhal);
+	cancel_work_sync(&rtlpriv->works.fill_h2c_cmd);
 	/* Enable software */
 	SET_USB_STOP(rtlusb);
 	rtl_usb_deinit(hw);
@@ -1026,6 +1027,16 @@ static bool rtl_usb_tx_chk_waitq_insert(struct ieee80211_hw *hw,
 	return false;
 }
 
+static void rtl_fill_h2c_cmd_work_callback(struct work_struct *work)
+{
+	struct rtl_works *rtlworks =
+	    container_of(work, struct rtl_works, fill_h2c_cmd);
+	struct ieee80211_hw *hw = rtlworks->hw;
+	struct rtl_priv *rtlpriv = rtl_priv(hw);
+
+	rtlpriv->cfg->ops->fill_h2c_cmd(hw, H2C_RA_MASK, 5, rtlpriv->rate_mask);
+}
+
 static struct rtl_intf_ops rtl_usb_ops = {
 	.adapter_start = rtl_usb_start,
 	.adapter_stop = rtl_usb_stop,
@@ -1057,6 +1068,8 @@ int rtl_usb_probe(struct usb_interface *intf,
 
 	/* this spin lock must be initialized early */
 	spin_lock_init(&rtlpriv->locks.usb_lock);
+	INIT_WORK(&rtlpriv->works.fill_h2c_cmd,
+		  rtl_fill_h2c_cmd_work_callback);
 
 	rtlpriv->usb_data_index = 0;
 	init_completion(&rtlpriv->firmware_loading_complete);