Linux/WL1807MOD: Driver / Firmware seems to load fine but any wireless action crashes driver with "wlcore: Scan completed due to error."

Yes, this is from the device tree:

/* wlcore sdio */
&usdhc2 {
	pinctrl-names = "default";
	pinctrl-0 = <&pinctrl_usdhc2>;
	bus-width = <4>;
	vmmc-supply = <&wlan_en_reg>; 
	no-1-8-v;
	non-removable;

	status = "okay";

	keep-power-in-suspend;
	enable-sdio-wakeup;
   	non-removable;    


    	cap-power-off-card;

	#address-cells = <1>;
	#size-cells = <0>;

	disable-wp;
	
	
	wlcore: wlcore@2 {
		compatible = "ti,wl1807";
		reg = <2>;
		interrupt-parent = <&gpio6>;
		interrupts = <10 IRQ_TYPE_EDGE_FALLING>; /* Level shift on actual board reverses polarity, changed to falling edge */
		tcxo-clock-frequency = <26000000>; 	
		status = "okay";
	};
};

The Wlan_en does stay asserted the entire time, I have checked this on a scope. It also properly goes low when the interface is set 'down'.

Thanks,

Thomas

Hi Eyal,

After boot:

root@colibri-imx6:~# cat /proc/interrupts | grep wl
251:          0  gpio-mxc  10 Edge      wl18xx

After bringup:

oot@colibri-imx6:/home# ifconfig wlan0 up
[   56.847080] wlcore: using inverted interrupt logic: 2
[   56.923340] wlcore: PHY firmware version: Rev 8.2.0.0.242
[   57.049237] wlcore: firmware booted (Rev 8.9.0.0.79)
[   57.078727] IPv6: ADDRCONF(NETDEV_UP): wlan0: link is not ready
root@colibri-imx6:/home# cat /proc/interrupts | grep wl
251:          2  gpio-mxc  10 Edge      wl18xx

After failed scan:

root@colibri-imx6:/home# cat /proc/interrupts | grep wl
251:          4  gpio-mxc  10 Edge      wl18xx

Note this is the device tree binding for wlcore (in usdhc2):

wlcore: wlcore@2 {
		compatible = "ti,wl1807";
		reg = <2>;
		interrupt-parent = <&gpio6>;
		interrupts = <10 IRQ_TYPE_EDGE_FALLING>; /*tdp change to falling b/c of design on actual board */	
		tcxo-clock-frequency = <26000000>;
		status = "okay";
	};

I have verified that the pin is muxed to gpio as well.

Should the interrupt be rising edge or high level? The datasheet seems to suggest rising-edge, but I noticed most other dts files with 183x in them seem to use high level. If I try low level on our board (since the logic is inverted) it hangs indefinitely on bringing up wlan0, with this error:

root@colibri-imx6:~# ifconfig wlan0 up
[   34.337463] wlcore: using inverted interrupt logic: 8
[   34.414059] wlcore: PHY firmware version: Rev 8.2.0.0.242
[   34.476812] wlcore: firmware booted (Rev 8.9.0.0.79)
[   34.547675] random: crng init done
[   51.043067] INFO: task sd-resolve:236 blocked for more than 10 seconds.
[   51.049712]       Tainted: G           O    4.9.84-2.8.2+gb2a7f2f #25
[   51.056193] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[   51.064047] sd-resolve      D    0   236      1 0x00000000
[   51.069554] Backtrace:
[   51.072039] [<807f93c4>] (__schedule) from [<807f9940>] (schedule+0x44/0xa4)
[   51.079131]  r10:00000000 r9:80b3e86c r8:00000002 r7:80b3e868 r6:863d7900 r5:ffffffff
[   51.086986]  r4:ffffe000
[   51.089535] [<807f98fc>] (schedule) from [<807f9cd0>] (schedule_preempt_disabled+0x10/0x14)
[   51.097915]  r5:ffffffff r4:80b3e864
[   51.101505] [<807f9cc0>] (schedule_preempt_disabled) from [<807fb1b0>] (__mutex_lock_slowpath+0xa4/0x158)
[   51.111100] [<807fb10c>] (__mutex_lock_slowpath) from [<807fb2bc>] (mutex_lock+0x58/0x5c)
[   51.119311]  r9:00000000 r8:87be3e24 r7:86f30e40 r6:00000014 r5:86c1c800 r4:80b3e864
[   51.127110] [<807fb264>] (mutex_lock) from [<80702dc4>] (rtnetlink_rcv+0x1c/0x34)
[   51.134618]  r5:86c1c800 r4:86f30e40
[   51.138209] [<80702da8>] (rtnetlink_rcv) from [<8071656c>] (netlink_unicast+0x18c/0x224)
[   51.146331]  r5:86c1c800 r4:860d8c00
[   51.149918] [<807163e0>] (netlink_unicast) from [<807169cc>] (netlink_sendmsg+0x2f4/0x360)
[   51.158209]  r9:00000014 r8:00000000 r7:00000000 r6:86f30e40 r5:86c1c800 r4:87be3ed8
[   51.165994] [<807166d8>] (netlink_sendmsg) from [<806d2d70>] (sock_sendmsg+0x1c/0x2c)
[   51.173851]  r10:00000000 r9:87be2000 r8:80107ee4 r7:00000122 r6:85322a80 r5:00000000
[   51.181684]  r4:00000000
[   51.184248] [<806d2d54>] (sock_sendmsg) from [<806d3d7c>] (SyS_sendto+0x94/0xf0)
[   51.191659] [<806d3ce8>] (SyS_sendto) from [<80107d20>] (ret_fast_syscall+0x0/0x48)
[   51.199346]  r6:76b09638 r5:0000000c r4:76b0962c

Hi,

This is really dependent on the host processor and the target hardware, type of external pulls etc.

Some hosts are better with level interrupts and with others, edge interrupt is better.

Based on the fact that right after ifconfig wlan0 up, you already see 2 interrupts (I would have expected to see 1) and and after scan command you only see two additional ones, i believe you have not set the wlan_irq correctly. Try changing the pullup/pulldown configuration in your .dts pin muxing . You didn't paste this part.

Best Regards,

Eyal

Hi Eyal,

Thanks for the info. The pad control for the irq pad is :

MX6QDL_PAD_NANDF_RB0__GPIO6_IO10     PAD_CTRL_HYS_PU     /* PAD_CTRL_HYS_PU  is 0x1b0b0  */

Which is a 100k pull up mode. The level shifter for the IRQ is a simple mosfet that has a pull down on the WL1807 side and an (external) 10k pullup on the processor side, which seems to translate the levels fine with a reasonable rise/fall time. I can try an internal pulldown on the irq pin as well but I would think there wouldn't be much effect since the external pullup is 10k.

The SDIO lines are also using 100k internal pullups on the processor side, but the level shifter itself (TXS0108) has 40k internal pullups as well so the processor pullups aren't doing much.

The datasheet seems to suggest that the IRQ line on the 1807 is configurable to different polarities, how is this done? We are assuming it is not changing from default (active high) because we need it active high on the 1807 side and active low on the processor side.