From cee7593d81dfb5a3fdb66e101d7a768ce81f405c Mon Sep 17 00:00:00 2001
From: pizza <pizza>
Date: Thu, 31 Jan 2008 14:37:27 +0000
Subject: [PATCH] Apply patch from Karl Relton that improves the reliability of
 USB device initialization and operation.

---
 CHANGES                         |  3 ++
 src/prism2/driver/hfa384x_usb.c | 65 +++++++++++++++++++++++++++------
 2 files changed, 56 insertions(+), 12 deletions(-)

diff --git a/CHANGES b/CHANGES
index 67bd8ee..9573d9c 100644
--- a/CHANGES
+++ b/CHANGES
@@ -41,6 +41,9 @@
 * Intersil Corporation as part of PRISM(R) chipset product development.
 *
 * --------------------------------------------------------------------
+ - Apply a patch from Karl Relton that improves USB reliability.
+ - When in wlan_wext_write=1 mode, auto-transition to ifstate_enable 
+   and bypass userspace firmware load altogether.
  - Fix bug in the GIWENCODE handler (Richard Kennedy & Tormod Volden)
  - Support 2.6.24 kernels (modified patch from Tormod Volden)
  - Support 2.6.23 kernels. (With apologies to Martin Dauskardt for a 
diff --git a/src/prism2/driver/hfa384x_usb.c b/src/prism2/driver/hfa384x_usb.c
index 02fc393..f6abbea 100644
--- a/src/prism2/driver/hfa384x_usb.c
+++ b/src/prism2/driver/hfa384x_usb.c
@@ -3269,17 +3269,35 @@ int hfa384x_drvr_setconfig(hfa384x_t *hw, UINT16 rid, void *buf, UINT16 len)
 ----------------------------------------------------------------*/
 int hfa384x_drvr_start(hfa384x_t *hw)
 {
-	int		result;
+	int		result, result1, result2;
+	u16		status;
 	DBFENTER;
 
 	might_sleep();
-
-	if (usb_clear_halt(hw->usb, hw->endp_in)) {
+    
+	/* Clear endpoint stalls - but only do this if the endpoint
+	 * is showing a stall status. Some prism2 cards seem to behave
+	 * badly if a clear_halt is called when the endpoint is already
+	 * ok
+	 */
+	result = usb_get_status(hw->usb, USB_RECIP_ENDPOINT, hw->endp_in, &status);
+	if (result < 0) {
+		WLAN_LOG_ERROR(
+			"Cannot get bulk in endpoint status.\n");
+		goto done;
+	}
+	if ((status == 1) && usb_clear_halt(hw->usb, hw->endp_in)) {
 		WLAN_LOG_ERROR(
 			"Failed to reset bulk in endpoint.\n");
 	}
 
-	if (usb_clear_halt(hw->usb, hw->endp_out)) {
+	result = usb_get_status(hw->usb, USB_RECIP_ENDPOINT, hw->endp_out, &status);
+	if (result < 0) {
+		WLAN_LOG_ERROR(
+			"Cannot get bulk out endpoint status.\n");
+		goto done;
+	}
+	if ((status == 1) && usb_clear_halt(hw->usb, hw->endp_out)) {
 		WLAN_LOG_ERROR(
 			"Failed to reset bulk out endpoint.\n");
 	}
@@ -3296,14 +3314,37 @@ int hfa384x_drvr_start(hfa384x_t *hw)
 		goto done;
 	}
 
-	/* call initialize */
-	result = hfa384x_cmd_initialize(hw);
-	if (result != 0) {
-		usb_kill_urb(&hw->rx_urb);
-		WLAN_LOG_ERROR(
-			"cmd_initialize() failed, result=%d\n",
-			result);
-		goto done;
+	/* Call initialize twice, with a 1 second sleep in between.
+	 * This is a nasty work-around since many prism2 cards seem to
+	 * need time to settle after an init from cold. The second
+	 * call to initialize in theory is not necessary - but we call
+	 * it anyway as a double insurance policy:
+	 * 1) If the first init should fail, the second may well succeed
+	 *    and the card can still be used
+	 * 2) It helps ensures all is well with the card after the first
+	 *    init and settle time.
+	 */
+	result1 = hfa384x_cmd_initialize(hw);
+	msleep(1000);
+	result = result2 = hfa384x_cmd_initialize(hw);
+	if (result1 != 0) {
+		if (result2 != 0) {
+			WLAN_LOG_ERROR(
+				"cmd_initialize() failed on two attempts, results %d and %d\n",
+				result1, result2);
+			usb_kill_urb(&hw->rx_urb);
+			goto done;
+		} else {
+			WLAN_LOG_DEBUG(0, "First cmd_initialize() failed (result %d),\n",
+				result1);
+			WLAN_LOG_DEBUG(0, "but second attempt succeeded. All should be ok\n");
+		}
+	} else if (result2 != 0) {
+		WLAN_LOG_WARNING(
+			"First cmd_initialize() succeeded, but second attempt failed (result=%d)\n",
+			result2);
+		WLAN_LOG_WARNING("Most likely the card will be functional\n");
+			goto done;
 	}
 
 	hw->state = HFA384x_STATE_RUNNING;