Werner Almesberger wrote:
Sean McNeil wrote:
Android has moved to expecting what is now a standard interface to RF
devices for switching them on and off. Attached is a patch to
stable-tracking to add support of turning on/off bluetooth via. rfkill.
Hmm, with gta02-moredrivers-defconfig the kernel cheerfully announces
[21474539.295000] ------------[ cut here ]------------
[21474539.300000] WARNING: at net/rfkill/rfkill.c:752
rfkill_register+0x60/0x218()
[21474539.305000] rfkill: attempt to register a badly initialized rfkill struct
[...]
[21474539.340000] [<c035b1a4>] (rfkill_register+0x0/0x218) from [<c0012bf8>]
(gta01_bt_probe+0x94/0xc0)
[21474539.350000] r7:00000000 r6:c046a138 r5:c046a130 r4:c799a800
[21474539.355000] [<c0012b64>] (gta01_bt_probe+0x0/0xc0) from [<c01c81b0>]
(platform_drv_probe+0x20/0x24)
[21474539.365000] r6:c0470078 r5:c046a138 r4:00000000
[...]
Yikes! I didn't see this as my debug board is being shipped by boat and
my dmesg buffer is flooded with jffs2 complaints. Thanks for pointing it
out.
I guess it doesn't like the
rfkill->state = -1;
Should this be RFKILL_STATE_SOFT_BLOCKED ?
It should be RFKILL_STATE_OFF.
By the way, while you're at it, the initialization logic in
gta01_bt_probe is almost the same as what would happen if you
bt_rfkill_toggle_radio(pdev->dev, RFKILL_STATE_OFF);
Close, but in the GTA02 case we wouldn't set the regulator to 0 as the
enable pin would probably already be off.
- Werner
