This patch hopefully finishes the SoftLED code:
- It adds two more LEDs (rx and radio).
(the FW claims it can support up to 16 LEDs,
but I doubt that any vendor put more than 4 on a board)
- update the LEDs in a _delayed_ workqueue.
No one reported any more crashes.
(see: "PATCH] p54: fix race condition in memory management")
So we can stop burning the mm code.
Signed-off-by: Christian Lamparter <[email protected]>
---
diff --git a/drivers/net/wireless/p54/p54.h b/drivers/net/wireless/p54/p54.h
index 04d95a1..0954044 100644
--- a/drivers/net/wireless/p54/p54.h
+++ b/drivers/net/wireless/p54/p54.h
@@ -125,6 +125,7 @@ struct p54_led_dev {
struct led_classdev led_dev;
char name[P54_LED_MAX_NAME_LEN + 1];
+ unsigned int toggled;
unsigned int index;
unsigned int registered;
};
@@ -187,8 +188,8 @@ struct p54_common {
/* LED management */
#ifdef CONFIG_P54_LEDS
- struct p54_led_dev assoc_led;
- struct p54_led_dev tx_led;
+ struct p54_led_dev leds[4];
+ struct delayed_work led_work;
#endif /* CONFIG_P54_LEDS */
u16 softled_state; /* bit field of glowing LEDs */
diff --git a/drivers/net/wireless/p54/p54common.c b/drivers/net/wireless/p54/p54common.c
index 0c1b057..3766d33 100644
--- a/drivers/net/wireless/p54/p54common.c
+++ b/drivers/net/wireless/p54/p54common.c
@@ -2088,6 +2088,9 @@ static void p54_stop(struct ieee80211_hw *dev)
priv->softled_state = 0;
p54_set_leds(dev);
+#ifdef CONFIG_P54_LEDS
+ cancel_delayed_work_sync(&priv->led_work);
+#endif /* CONFIG_P54_LEDS */
cancel_delayed_work_sync(&priv->work);
if (priv->cached_beacon)
p54_tx_cancel(dev, priv->cached_beacon);
@@ -2421,6 +2424,43 @@ static int p54_set_key(struct ieee80211_hw *dev, enum set_key_cmd cmd,
}
#ifdef CONFIG_P54_LEDS
+static void p54_update_leds(struct work_struct *work)
+{
+ struct p54_common *priv = container_of(work, struct p54_common,
+ led_work.work);
+ int err, i, tmp, blink_delay = 400;
+ bool rerun = false;
+
+ /* Don't toggle the LED, when the device is down. */
+ if (priv->mode == NL80211_IFTYPE_UNSPECIFIED)
+ return ;
+
+ for (i = 0; i < ARRAY_SIZE(priv->leds); i++)
+ if (priv->leds[i].toggled) {
+ priv->softled_state |= BIT(i);
+
+ tmp = 70 + 200 / (priv->leds[i].toggled);
+ if (tmp < blink_delay)
+ blink_delay = tmp;
+
+ if (priv->leds[i].led_dev.brightness == LED_OFF)
+ rerun = true;
+
+ priv->leds[i].toggled =
+ !!priv->leds[i].led_dev.brightness;
+ } else
+ priv->softled_state &= ~BIT(i);
+
+ err = p54_set_leds(priv->hw);
+ if (err && net_ratelimit())
+ printk(KERN_ERR "%s: failed to update LEDs.\n",
+ wiphy_name(priv->hw->wiphy));
+
+ if (rerun)
+ queue_delayed_work(priv->hw->workqueue, &priv->led_work,
+ msecs_to_jiffies(blink_delay));
+}
+
static void p54_led_brightness_set(struct led_classdev *led_dev,
enum led_brightness brightness)
{
@@ -2428,28 +2468,23 @@ static void p54_led_brightness_set(struct led_classdev *led_dev,
led_dev);
struct ieee80211_hw *dev = led->hw_dev;
struct p54_common *priv = dev->priv;
- int err;
- /* Don't toggle the LED, when the device is down. */
if (priv->mode == NL80211_IFTYPE_UNSPECIFIED)
return ;
- if (brightness != LED_OFF)
- priv->softled_state |= BIT(led->index);
- else
- priv->softled_state &= ~BIT(led->index);
-
- err = p54_set_leds(dev);
- if (err && net_ratelimit())
- printk(KERN_ERR "%s: failed to update %s LED.\n",
- wiphy_name(dev->wiphy), led_dev->name);
+ if (brightness) {
+ led->toggled++;
+ queue_delayed_work(priv->hw->workqueue, &priv->led_work,
+ HZ/10);
+ }
}
static int p54_register_led(struct ieee80211_hw *dev,
- struct p54_led_dev *led,
unsigned int led_index,
char *name, char *trigger)
{
+ struct p54_common *priv = dev->priv;
+ struct p54_led_dev *led = &priv->leds[led_index];
int err;
if (led->registered)
@@ -2482,19 +2517,30 @@ static int p54_init_leds(struct ieee80211_hw *dev)
* TODO:
* Figure out if the EEPROM contains some hints about the number
* of available/programmable LEDs of the device.
- * But for now, we can assume that we have two programmable LEDs.
*/
- err = p54_register_led(dev, &priv->assoc_led, 0, "assoc",
+ INIT_DELAYED_WORK(&priv->led_work, p54_update_leds);
+
+ err = p54_register_led(dev, 0, "assoc",
ieee80211_get_assoc_led_name(dev));
if (err)
return err;
- err = p54_register_led(dev, &priv->tx_led, 1, "tx",
+ err = p54_register_led(dev, 1, "tx",
ieee80211_get_tx_led_name(dev));
if (err)
return err;
+ err = p54_register_led(dev, 2, "rx",
+ ieee80211_get_rx_led_name(dev));
+ if (err)
+ return err;
+
+ err = p54_register_led(dev, 3, "radio",
+ ieee80211_get_radio_led_name(dev));
+ if (err)
+ return err;
+
err = p54_set_leds(dev);
return err;
}
@@ -2502,11 +2548,11 @@ static int p54_init_leds(struct ieee80211_hw *dev)
static void p54_unregister_leds(struct ieee80211_hw *dev)
{
struct p54_common *priv = dev->priv;
+ int i;
- if (priv->tx_led.registered)
- led_classdev_unregister(&priv->tx_led.led_dev);
- if (priv->assoc_led.registered)
- led_classdev_unregister(&priv->assoc_led.led_dev);
+ for (i = 0; i < ARRAY_SIZE(priv->leds); i++)
+ if (priv->leds[i].registered)
+ led_classdev_unregister(&priv->leds[i].led_dev);
}
#endif /* CONFIG_P54_LEDS */
Christian Lamparter wrote:
> This patch hopefully finishes the SoftLED code:
> - It adds two more LEDs (rx and radio).
> (the FW claims it can support up to 16 LEDs,
> but I doubt that any vendor put more than 4 on a board)
> - update the LEDs in a _delayed_ workqueue.
> No one reported any more crashes.
> (see: "PATCH] p54: fix race condition in memory management")
> So we can stop burning the mm code.
>
> Signed-off-by: Christian Lamparter <[email protected]>
> ---
My device has 2 LED's - one for power and one blinks on TX(?). No messages in logs.
Acked-by: Larry Finger <[email protected]>
On Tuesday 14 April 2009 17:12:33 Fabio Rossi wrote:
> On Wednesday 25 March 2009, Christian Lamparter wrote:
>
> > This patch hopefully finishes the SoftLED code:
> > - It adds two more LEDs (rx and radio).
> > (the FW claims it can support up to 16 LEDs,
> > but I doubt that any vendor put more than 4 on a board)
> > - update the LEDs in a _delayed_ workqueue.
> > No one reported any more crashes.
> > (see: "PATCH] p54: fix race condition in memory management")
> > So we can stop burning the mm code.
>
> My device has 2 LEDs and I can see 4 registrations (assoc, tx, rx and radio).
> Does it make sense to register LEDs which are not available?
Well, trouble is that you can look (with your eyes) at your device and see the
number of LEDs... This is something we cannot do with the driver.
(Unless we find a EEPROM field which stores the # of LEDs and
maybe even what they are used for?)
4 is just a "save" guess which should cover most - if not all - customer devices.
Regards,
Chr
On Wednesday 25 March 2009, Christian Lamparter wrote:
> This patch hopefully finishes the SoftLED code:
> - It adds two more LEDs (rx and radio).
> (the FW claims it can support up to 16 LEDs,
> but I doubt that any vendor put more than 4 on a board)
> - update the LEDs in a _delayed_ workqueue.
> No one reported any more crashes.
> (see: "PATCH] p54: fix race condition in memory management")
> So we can stop burning the mm code.
My device has 2 LEDs and I can see 4 registrations (assoc, tx, rx and radio).
Does it make sense to register LEDs which are not available?
Fabio