Date:   Mon, 27 Jul 2020 02:51:12 +0000
From:   Andy Duan <fugang.duan@....com>
To:     Chris Healy <cphealy@...il.com>,
        Laurent Pinchart <laurent.pinchart@...asonboard.com>
CC:     Andrew Lunn <andrew@...n.ch>,
        "David S. Miller" <davem@...emloft.net>,
        netdev <netdev@...r.kernel.org>,
        Martin Fuzzey <martin.fuzzey@...wbird.group>
Subject: RE: [EXT] Re: [RESENT PATCH net--stat 1/1] net: ethernet: fec: Revert
 "net: ethernet: fec: Replace interrupt driven MDIO with polled IO"

From: Chris Healy <cphealy@...il.com> Sent: Monday, July 27, 2020 10:40 AM
> Actually, I was a little quick to say it went from broken to working.
> 
> With net-next, I'm getting CRC errors on 100% of inbound packets.
> With bcf3440c6dd78bfe5836ec0990fe36d7b4bb7d20 reverted, I drop down to a
> 1% error rate.
> 
> This very much feels like a KSZ9031 RGMII timing issue to me...

@Chris/@...rent, I run net-next on imx7d sdb board with BCM54220 switch PHY,
there have no NFS boot issue, no crc error issue by receving 1.1GiB data.

~# ifconfig
eth0      Link encap:Ethernet  HWaddr 00:20:30:40:50:02
          inet addr:10.192.242.202  Bcast:10.192.242.255  Mask:255.255.255.0
          inet6 addr: fe80::220:30ff:fe40:5002/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:853885 errors:0 dropped:0 overruns:0 frame:0
          TX packets:22370 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:1284666734 (1.1 GiB)  TX bytes:3147617 (3.0 MiB)

lo        Link encap:Local Loopback
          inet addr:127.0.0.1  Mask:255.0.0.0
          inet6 addr: ::1/128 Scope:Host
          UP LOOPBACK RUNNING  MTU:65536  Metric:1
          RX packets:322 errors:0 dropped:0 overruns:0 frame:0
          TX packets:322 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:24460 (23.8 KiB)  TX bytes:24460 (23.8 KiB)

~# uname -r
5.8.0-rc6-01914-ga57066b1a019


So, the issue you caught relates to Micrel PHY RGMII timing issue.



Regards,
Fugang

> 
> On Sun, Jul 26, 2020 at 7:35 PM Chris Healy <cphealy@...il.com> wrote:
> >
> > Hi Laurent,
> >
> > I have the exact same copper PHY.  I just reverted a patch specific to
> > this PHY and went from broken to working.  Give this a try:
> >
> > git revert bcf3440c6dd78bfe5836ec0990fe36d7b4bb7d20
> >
> > Regards,
> >
> > Chris
> >
> > On Sun, Jul 26, 2020 at 7:33 PM Laurent Pinchart
> > <laurent.pinchart@...asonboard.com> wrote:
> > >
> > > Hi Andrew,
> > >
> > > On Mon, Jul 27, 2020 at 04:14:32AM +0200, Andrew Lunn wrote:
> > > > On Mon, Jul 27, 2020 at 05:06:31AM +0300, Laurent Pinchart wrote:
> > > > > On Mon, Jul 27, 2020 at 04:24:02AM +0300, Laurent Pinchart wrote:
> > > > > > On Mon, Apr 27, 2020 at 10:08:04PM +0800, Fugang Duan wrote:
> > > > > > > This reverts commit 29ae6bd1b0d8a57d7c00ab12cbb949fc41986eef.
> > > > > > >
> > > > > > > The commit breaks ethernet function on i.MX6SX, i.MX7D,
> > > > > > > i.MX8MM, i.MX8MQ, and i.MX8QXP platforms. Boot yocto system
> > > > > > > by NFS mounting rootfs will be failed with the commit.
> > > > > >
> > > > > > I'm afraid this commit breaks networking on i.MX7D for me :-(
> > > > > > My board is configured to boot over NFS root with IP
> > > > > > autoconfiguration through DHCP. The DHCP request goes out, the
> > > > > > reply it sent back by the server, but never noticed by the fec driver.
> > > > > >
> > > > > > v5.7 works fine. As 29ae6bd1b0d8a57d7c00ab12cbb949fc41986eef
> > > > > > was merged during the v5.8 merge window, I suspect something
> > > > > > else cropped in between
> > > > > > 29ae6bd1b0d8a57d7c00ab12cbb949fc41986eef and this patch that
> > > > > > needs to be reverted too. We're close to v5.8 and it would be
> > > > > > annoying to see this regression ending up in the released
> > > > > > kernel. I can test patches, but I'm not familiar enough with
> > > > > > the driver (or the networking
> > > > > > subsystem) to fix the issue myself.
> > > > >
> > > > > If it can be of any help, I've confirmed that, to get the
> > > > > network back to usable state from v5.8-rc6, I have to revert all
> > > > > patches up to this one. This is the top of my branch, on top of v5.8-rc6:
> > > > >
> > > > > 5bbe80c9efea Revert "net: ethernet: fec: Revert "net: ethernet: fec:
> Replace interrupt driven MDIO with polled IO""
> > > > > 5462896a08c1 Revert "net: ethernet: fec: Replace interrupt driven
> MDIO with polled IO"
> > > > > 824a82e2bdfa Revert "net: ethernet: fec: move GPR register offset and
> bit into DT"
> > > > > bfe330591cab Revert "net: fec: disable correct clk in the err path of
> fec_enet_clk_enable"
> > > > > 109958cad578 Revert "net: ethernet: fec: prevent tx starvation under
> high rx load"
> > > >
> > > > OK.
> > > >
> > > > What PHY are you using? A Micrel?
> > >
> > > KSZ9031RNXIA
> > >
> > > > And which DT file?
> > >
> > > It's out of tree.
> > >
> > > &fec1 {
> > >         pinctrl-names = "default";
> > >         pinctrl-0 = <&pinctrl_enet1>;
> > >         assigned-clocks = <&clks IMX7D_ENET1_TIME_ROOT_SRC>,
> > >                           <&clks IMX7D_ENET1_TIME_ROOT_CLK>;
> > >         assigned-clock-parents = <&clks
> IMX7D_PLL_ENET_MAIN_100M_CLK>;
> > >         assigned-clock-rates = <0>, <100000000>;
> > >         phy-mode = "rgmii";
> > >         phy-handle = <&ethphy0>;
> > >         phy-reset-gpios = <&gpio1 13 GPIO_ACTIVE_LOW>;
> > >         phy-supply = <&reg_3v3_sw>;
> > >         fsl,magic-packet;
> > >         status = "okay";
> > >
> > >         mdio {
> > >                 #address-cells = <1>;
> > >                 #size-cells = <0>;
> > >
> > >                 ethphy0: ethernet-phy@0 {
> > >                         reg = <1>;
> > >                 };
> > >
> > >                 ethphy1: ethernet-phy@1 {
> > >                         reg = <2>;
> > >                 };
> > >         };
> > > };
> > >
> > > I can provide the full DT if needed.
> > >
> > > --
> > > Regards,
> > >
> > > Laurent Pinchart

Powered by blists - more mailing lists