On Thu, 24 Jan 2013, Simon Wood wrote:

quoted

From: simon <simon@simon-virtual-machine.(none)>

This patch to the SRW-S1 driver adds the ability to control all
LEDs simultaneously as testing showed that it was slow (noticably!!)
when seting or clearing all the LEDs in turn.

It adds a 'RPMALL' LED, whose behavoir is asserted to all the LEDs in
the bar graph, individual LEDs can subsequently be turned on/off
individually.

Signed-off-by: Simon Wood <redacted>
tested-by: John Murphy <redacted>

[ ... snip ... ]

quoted

@@ -219,13 +255,34 @@ static int srws1_probe(struct hid_device *hdev,

 	/* register led subsystem */
 	drv_data->led_state = 0;
-	for (i = 0; i < SRWS1_NUMBER_LEDS; i++)
+	for (i = 0; i < SRWS1_NUMBER_LEDS + 1; i++)
 		drv_data->led[i] = NULL;

 	srws1_set_leds(hdev, 0);

-	name_sz = strlen(hdev->uniq) + 15;
+	name_sz = strlen(hdev->uniq) + 16;
+
+	/* 'ALL', for setting all LEDs simultaneously */
+	led = kzalloc(sizeof(struct led_classdev)+name_sz, GFP_KERNEL);

Is this memory ever freed?

Yes. The pointer to it is stored at the end of the drv_data leds[] array
(line 280), and will be free'd on probe fail (line 314) or when device is
removed (line 341).

Simon