[OpenWrt-Devel] [PATCH] ar71xx: fix kernel Oops in at803x_link_change_notify

[Christian Lamparter]

On Friday, July 03, 2015 12:29:32 PM Sven Eckelmann wrote:
> > Sven, I've seen you did quite a bit of work on the OM5P-AN
> > with the same F1E-PHY. I went through the code from Atheros'
> > original driver SDK v17 (which was part of the GPL source
> > for the mynet). Apart from tx delay handling, everything
> > else matches perfectly, so the tx/rx delay code should be
> > enabled by default for this chip on all platforms without
> > needing any platform data overwrites.
> 
> Interesting, I don't have the SDK and I am only using OpenWrt. So I cannot say
> anything about the Atheros driver. But sounds like you are right. At least all
> recent devices I had in my hand required this change.
Oh, I figured you have access to some of Qualcomm Atheros' SDKs
and most importantly: some documentations from Qualcomm Atheros? 
I hoped you could confirm or disconfirm based on the documents.
 
> I also saw your fixup_rgmii_tx_delay delay change. I would ack when you submit
> a change like this. Just don't forget that the pll_1000 has also to be changed
> (especially when you move the default values to at803x.c):
> 
> --- a/target/linux/ar71xx/files/arch/mips/ath79/mach-om5p.c
> +++ b/target/linux/ar71xx/files/arch/mips/ath79/mach-om5p.c
> @@ -154,7 +154,8 @@ static struct i2c_board_info om5pan_i2c_devs[] __initdata = {
>  static struct at803x_platform_data om5p_an_at803x_data = {
>  	.disable_smarteee = 1,
>  	.enable_rgmii_rx_delay = 1,
> -	.enable_rgmii_tx_delay = 1,
> +	.enable_rgmii_tx_delay = 0,
> +	.fixup_rgmii_tx_delay = 1,
>  };
>  
>  static struct mdio_board_info om5p_an_mdio0_info[] = {
> @@ -201,7 +202,7 @@ static void __init om5p_an_setup(void)
>  	ath79_eth0_data.phy_if_mode = PHY_INTERFACE_MODE_RGMII;
>  	ath79_eth0_data.mii_bus_dev = &ath79_mdio0_device.dev;
>  	ath79_eth0_data.phy_mask = BIT(7);
> -	ath79_eth0_pll_data.pll_1000 = 0x02000000;
> +	ath79_eth0_pll_data.pll_1000 = 0x0e000000;
>  	ath79_eth0_pll_data.pll_100 = 0x00000101;
>  	ath79_eth0_pll_data.pll_10 = 0x00001313;
>  	ath79_register_eth(0);
> 
> 
> I don't have the equipment to check the actual signals (as you said in an
> earlier mail) - only some people which complained very loud when their long
> cable setup didn't work. :)
Long cable... That's a good point, I think I never tested the AT8035 with
anything beyond a 5m / 16 feet cable.  

> Maybe there are some non atheros SoC's out there which would have problems
> when you would change the default behavior of the AT8035.
I've been looking around for devices with an AT8035. And I found a few gems.

<http://lists.infradead.org/pipermail/linux-arm-kernel/2013-September/200978.html>
<https://forum.openwrt.org/viewtopic.php?id=42896><https://dev.openwrt.org/ticket/13203>
...
(The AT8035 is also used in some of the new HomePlug AV2 equipment) 

As far as I can tell, the defaults only seem to work for the 100mbps.
This makes sense, since the F1E has different PLLs and rx/tx delay
settings. Fixing this "globally" might actually be a good thing. At
least I'll give it a try.
> > I'm interested to hear from any other devices which have 
> > a AT803x. Also, please let me know where to get the GPL
> > tars for the devices.
> 
> I have forwarded your mail to the person which is handling the actual
> firmware builds of the OM5P-AN. He will contact you later and provide
> the sources.
Oh, that might be helpful yes. I can also post the sources from Western
Digital's S17_SSDK [The Ethernet driver SDK is part of their GPL.tar.gz].

Thanks,
  Christian
_______________________________________________
openwrt-devel mailing list
openwrt-devel at lists.openwrt.org
https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel