[PATCH 1/2] net: rt8169: WAR for DHCP not getting IP after kernel boot/reboot

[Tom Warren]

-----Original Message-----
From: Stephen Warren <swarren at wwwdotorg.org> 
Sent: Wednesday, March 18, 2020 4:30 PM
To: Tom Warren <TWarren at nvidia.com>
Cc: u-boot at lists.denx.de; Stephen Warren <swarren at nvidia.com>; Thierry Reding <treding at nvidia.com>; Jonathan Hunter <jonathanh at nvidia.com>; tomcwarren3959 at gmail.com
Subject: Re: [PATCH 1/2] net: rt8169: WAR for DHCP not getting IP after kernel boot/reboot

External email: Use caution opening links or attachments


On 3/17/20 2:07 PM, twarren at nvidia.com wrote:
> From: Tom Warren <twarren at nvidia.com>
>
> This is a WAR for DHCP failure after rebooting from the L4T kernel. 
> The r8169.c kernel driver is setting bit 19 of the rt816x HW register 
> 0xF0, which goes by FuncEvent and MISC in various driver source/datasheets.
> That bit is called RxDv_Gated_En in the r8169.c kernel driver. Clear 
> it here at the end of probe to ensure that U-Boot can get an IP 
> assigned via DHCP.

Is there a way to reset the entire chip back to power-on state instead, just to make sure there aren't any other register bits that might cause issues in the future?
[Tom] Maybe, I'll have to look at the RTL spec, it's been awhile since I came up with this WAR.

Does this bit exist in all the different chips that the driver supports?
IIRC at least the Linux r8169 driver supports many different similar chips; I'm not sure if the U-Boot driver does too. Since this driver is shared between many boards in U-Boot, we need to make sure this change doesn't break anyone using a different Realtek chip to the exact one we use on Jetson.
[Tom] I'll look into it after I get Nano patches all in place upstream. If anyone runs into the problem described here w/no IP on net boot, they can use this as a WAR until I try for a more robust fix.

We need to send this patch to the network maintainers so they're aware of it, and they'll probably want to apply it.
[Tom] Will do.

> @@ -1207,6 +1210,19 @@ static int rtl8169_eth_probe(struct udevice 
> *dev)

> +     u32 val = RTL_R32(FuncEvent);
> +     debug("%s: FuncEvent/Misc (0xF0) = 0x%08X\n", __func__, val);
> +     val &= ~RxDv_Gated_En;
> +     RTL_W32(FuncEvent, val);

I vagueley recall that U-Boot coding style wants variables declared at the start of the block, but perhaps that's no longer true. What does scripts/checkpatch.pl say?
[Tom] I don't recall, I'll run it again and see.


-----------------------------------------------------------------------------------
This email message is for the sole use of the intended recipient(s) and may contain
confidential information.  Any unauthorized review, use, disclosure or distribution
is prohibited.  If you are not the intended recipient, please contact the sender by
reply email and destroy all copies of the original message.
-----------------------------------------------------------------------------------