USB_NET_AX8817X dependency on AX88796B_PHY

USB_NET_AX8817X dependency on AX88796B_PHY

[Maciej Żenczykowski]

I've received reports that an ethernet usb dongle doesn't work (google
internal bug 304028301)...

Investigation shows that we have 5.10 (GKI) with USB_NET_AX8817X=y and
AX88796B_PHY not set.
I *think* this configuration combination makes no sense?
[note: I'm unsure how many different phy's this driver supports...]

Obviously, we could simply turn it on 'manually'... but:

commit dde25846925765a88df8964080098174495c1f10
Author: Oleksij Rempel <o.rempel@pengutronix.de>
Date:   Mon Jun 7 10:27:22 2021 +0200

    net: usb/phy: asix: add support for ax88772A/C PHYs

    Add support for build-in x88772A/C PHYs

    Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
    Reviewed-by: Andrew Lunn <andrew@lunn.ch>
    Signed-off-by: David S. Miller <davem@davemloft.net>

includes (as a side effect):

drivers/net/usb/Kconfig
@@ -164,6 +164,7 @@ config USB_NET_AX8817X
        depends on USB_USBNET
        select CRC32
        select PHYLIB
+       select AX88796B_PHY
        default y

which presumably makes this (particular problem) a non issue on 5.15+

I'm guessing the above fix (ie. commit dde25846925765a88df8964080098174495c1f10)
could (should?) simply be backported to older stable kernels?

I've verified it cherrypicks cleanly and builds (on x86_64 5.10 gki),
ie.
$ git checkout android/kernel/common/android13-5.10
$ git cherry-pick -x dde25846925765a88df8964080098174495c1f10
$ make ARCH=x86_64 gki_defconfig
$ egrep -i ax88796b < .config
CONFIG_AX88796B_PHY=y
$ make -j50
./drivers/net/phy/ax88796b.o gets built

I've sourced a copy of the problematic hardware, but I'm hitting
problems where on at least two (both my chromebook and 1 of the 2
usb-c ports on my lenovo laptop) totally different usb
controllers/ports it doesn't even usb enumerate (ie. nothing in dmesg,
no show on lsusb), which is making testing difficult (unsure if I just
got a bad sample)...

- Maciej

Re: USB_NET_AX8817X dependency on AX88796B_PHY

[Maciej Żenczykowski]

On Sun, Oct 15, 2023 at 2:55 PM Maciej Żenczykowski <maze@google.com> wrote:
>
> I've received reports that an ethernet usb dongle doesn't work (google
> internal bug 304028301)...
>
> Investigation shows that we have 5.10 (GKI) with USB_NET_AX8817X=y and
> AX88796B_PHY not set.
> I *think* this configuration combination makes no sense?
> [note: I'm unsure how many different phy's this driver supports...]
>
> Obviously, we could simply turn it on 'manually'... but:
>
> commit dde25846925765a88df8964080098174495c1f10
> Author: Oleksij Rempel <o.rempel@pengutronix.de>
> Date:   Mon Jun 7 10:27:22 2021 +0200
>
>     net: usb/phy: asix: add support for ax88772A/C PHYs
>
>     Add support for build-in x88772A/C PHYs
>
>     Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
>     Reviewed-by: Andrew Lunn <andrew@lunn.ch>
>     Signed-off-by: David S. Miller <davem@davemloft.net>
>
> includes (as a side effect):
>
> drivers/net/usb/Kconfig
> @@ -164,6 +164,7 @@ config USB_NET_AX8817X
>         depends on USB_USBNET
>         select CRC32
>         select PHYLIB
> +       select AX88796B_PHY
>         default y
>
> which presumably makes this (particular problem) a non issue on 5.15+
>
> I'm guessing the above fix (ie. commit dde25846925765a88df8964080098174495c1f10)
> could (should?) simply be backported to older stable kernels?
>
> I've verified it cherrypicks cleanly and builds (on x86_64 5.10 gki),
> ie.
> $ git checkout android/kernel/common/android13-5.10
> $ git cherry-pick -x dde25846925765a88df8964080098174495c1f10
> $ make ARCH=x86_64 gki_defconfig
> $ egrep -i ax88796b < .config
> CONFIG_AX88796B_PHY=y
> $ make -j50
> ./drivers/net/phy/ax88796b.o gets built
>
> I've sourced a copy of the problematic hardware, but I'm hitting
> problems where on at least two (both my chromebook and 1 of the 2
> usb-c ports on my lenovo laptop) totally different usb
> controllers/ports it doesn't even usb enumerate (ie. nothing in dmesg,
> no show on lsusb), which is making testing difficult (unsure if I just
> got a bad sample)...

And actually looking longer at the logs:

usb 1-5: new high-speed USB device number 9 using xhci_hcd
usb 1-5: New USB device found, idVendor=0b95, idProduct=772b, bcdDevice= 0.02
usb 1-5: New USB device strings: Mfr=1, Product=2, SerialNumber=3
usb 1-5: Product: AX88772C
usb 1-5: Manufacturer: ASIX Elec. Corp.
usb 1-5: SerialNumber: 000002
asix 1-5:1.0 (unnamed net_device) (uninitialized): PHY
[usb-001:009:10] driver [Asix Electronics AX88772C] (irq=POLL)
Asix Electronics AX88772C usb-001:009:10: attached PHY driver
(mii_bus:phy_addr=usb-001:009:10, irq=POLL)
asix 1-5:1.0 eth0: register 'asix' at usb-0000:00:14.0-5, ASIX
AX88772B USB 2.0 Ethernet, 14:ae:85:70:44:29
asix 1-5:1.0 enx14ae85704429: renamed from eth0
asix 1-5:1.0 enx14ae85704429: configuring for phy/internal link mode
usb 1-5: USB disconnect, device number 9
asix 1-5:1.0 enx14ae85704429: unregister 'asix' usb-0000:00:14.0-5,
ASIX AX88772B USB 2.0 Ethernet

It looks like the rest of that patch is needed too, since it is a AX88772C phy.

This means even with the option manually enabled, we'd still need to
cherrypick dde25846925765a88df8964080098174495c1f10
"net: usb/phy: asix: add support for ax88772A/C PHYs"
since apparently this is simply new(ish) hardware with built-in x88772C PHY.

Re: USB_NET_AX8817X dependency on AX88796B_PHY

[Oleksij Rempel]

Hi Maciej,

On Sun, Oct 15, 2023 at 02:55:29PM -0700, Maciej Żenczykowski wrote:
> I've received reports that an ethernet usb dongle doesn't work (google
> internal bug 304028301)...
> 
> Investigation shows that we have 5.10 (GKI) with USB_NET_AX8817X=y and
> AX88796B_PHY not set.
> I *think* this configuration combination makes no sense?
> [note: I'm unsure how many different phy's this driver supports...]

After porting part of this driver to PHYlib and then to PHYlink, the driver is
capable of supporting nearly all drivers/net/phy/*

I did this port for following USB Ethernet adapter which needs different PHY:
https://linux-automation.com/en/products/usb-t1l.html

> Obviously, we could simply turn it on 'manually'... but:
> 
> commit dde25846925765a88df8964080098174495c1f10
> Author: Oleksij Rempel <o.rempel@pengutronix.de>
> Date:   Mon Jun 7 10:27:22 2021 +0200
> 
>     net: usb/phy: asix: add support for ax88772A/C PHYs
> 
>     Add support for build-in x88772A/C PHYs
> 
>     Signed-off-by: Oleksij Rempel <o.rempel@pengutronix.de>
>     Reviewed-by: Andrew Lunn <andrew@lunn.ch>
>     Signed-off-by: David S. Miller <davem@davemloft.net>
> 
> includes (as a side effect):
> 
> drivers/net/usb/Kconfig
> @@ -164,6 +164,7 @@ config USB_NET_AX8817X
>         depends on USB_USBNET
>         select CRC32
>         select PHYLIB
> +       select AX88796B_PHY
>         default y
> 
> which presumably makes this (particular problem) a non issue on 5.15+
> 
> I'm guessing the above fix (ie. commit dde25846925765a88df8964080098174495c1f10)
> could (should?) simply be backported to older stable kernels?
> 
> I've verified it cherrypicks cleanly and builds (on x86_64 5.10 gki),
> ie.
> $ git checkout android/kernel/common/android13-5.10
> $ git cherry-pick -x dde25846925765a88df8964080098174495c1f10
> $ make ARCH=x86_64 gki_defconfig
> $ egrep -i ax88796b < .config
> CONFIG_AX88796B_PHY=y
> $ make -j50
> ./drivers/net/phy/ax88796b.o gets built

As far as i know, mainline kernel v5.10.198 do not supports ax88772A/C PHYs
within the ax88796b.c driver:
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/tree/drivers/net/phy/ax88796b.c?h=v5.10.198

There is no need to backport this patch to other stable kernels, except
they include porting of drivers/net/usb/asix_devices to PHYlink
framework.

> I've sourced a copy of the problematic hardware, but I'm hitting
> problems where on at least two (both my chromebook and 1 of the 2
> usb-c ports on my lenovo laptop) totally different usb
> controllers/ports it doesn't even usb enumerate (ie. nothing in dmesg,
> no show on lsusb), which is making testing difficult (unsure if I just
> got a bad sample)...

I use ax88772 based HW on different systems including on UCB-C ports (for
example on DELL XPS 13 developer edition)

Regards,
Oleksij
-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

Re: USB_NET_AX8817X dependency on AX88796B_PHY

[Oleksij Rempel]

On Sun, Oct 15, 2023 at 03:13:26PM -0700, Maciej Żenczykowski wrote:
> On Sun, Oct 15, 2023 at 2:55 PM Maciej Żenczykowski <maze@google.com> wrote:
> And actually looking longer at the logs:
> 
> usb 1-5: new high-speed USB device number 9 using xhci_hcd
> usb 1-5: New USB device found, idVendor=0b95, idProduct=772b, bcdDevice= 0.02
> usb 1-5: New USB device strings: Mfr=1, Product=2, SerialNumber=3
> usb 1-5: Product: AX88772C
> usb 1-5: Manufacturer: ASIX Elec. Corp.
> usb 1-5: SerialNumber: 000002
> asix 1-5:1.0 (unnamed net_device) (uninitialized): PHY
> [usb-001:009:10] driver [Asix Electronics AX88772C] (irq=POLL)
> Asix Electronics AX88772C usb-001:009:10: attached PHY driver
> (mii_bus:phy_addr=usb-001:009:10, irq=POLL)
> asix 1-5:1.0 eth0: register 'asix' at usb-0000:00:14.0-5, ASIX
> AX88772B USB 2.0 Ethernet, 14:ae:85:70:44:29
> asix 1-5:1.0 enx14ae85704429: renamed from eth0
> asix 1-5:1.0 enx14ae85704429: configuring for phy/internal link mode
> usb 1-5: USB disconnect, device number 9
> asix 1-5:1.0 enx14ae85704429: unregister 'asix' usb-0000:00:14.0-5,
> ASIX AX88772B USB 2.0 Ethernet
> 
> It looks like the rest of that patch is needed too, since it is a AX88772C phy.
> 
> This means even with the option manually enabled, we'd still need to
> cherrypick dde25846925765a88df8964080098174495c1f10
> "net: usb/phy: asix: add support for ax88772A/C PHYs"
> since apparently this is simply new(ish) hardware with built-in x88772C PHY.


As far as I see, you are not using clean mainline stable kernel. All changes
which make your kernel to need an external PHY driver are _not_ in v5.10.198.

You will need to cherrypick at least 28 last patches from this stack:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/net/usb/asix_devices.c?h=v6.6-rc6

and some from here too:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/net/usb/usbnet.c?h=v6.6-rc6

Regards,
Oleksij
-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

Re: USB_NET_AX8817X dependency on AX88796B_PHY

[Maciej Żenczykowski]

On Mon, Oct 16, 2023 at 1:29 AM Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> On Sun, Oct 15, 2023 at 03:13:26PM -0700, Maciej Żenczykowski wrote:
> > On Sun, Oct 15, 2023 at 2:55 PM Maciej Żenczykowski <maze@google.com> wrote:
> > And actually looking longer at the logs:
> >
> > usb 1-5: new high-speed USB device number 9 using xhci_hcd
> > usb 1-5: New USB device found, idVendor=0b95, idProduct=772b, bcdDevice= 0.02
> > usb 1-5: New USB device strings: Mfr=1, Product=2, SerialNumber=3
> > usb 1-5: Product: AX88772C
> > usb 1-5: Manufacturer: ASIX Elec. Corp.
> > usb 1-5: SerialNumber: 000002
> > asix 1-5:1.0 (unnamed net_device) (uninitialized): PHY
> > [usb-001:009:10] driver [Asix Electronics AX88772C] (irq=POLL)
> > Asix Electronics AX88772C usb-001:009:10: attached PHY driver
> > (mii_bus:phy_addr=usb-001:009:10, irq=POLL)
> > asix 1-5:1.0 eth0: register 'asix' at usb-0000:00:14.0-5, ASIX
> > AX88772B USB 2.0 Ethernet, 14:ae:85:70:44:29
> > asix 1-5:1.0 enx14ae85704429: renamed from eth0
> > asix 1-5:1.0 enx14ae85704429: configuring for phy/internal link mode
> > usb 1-5: USB disconnect, device number 9
> > asix 1-5:1.0 enx14ae85704429: unregister 'asix' usb-0000:00:14.0-5,
> > ASIX AX88772B USB 2.0 Ethernet
> >
> > It looks like the rest of that patch is needed too, since it is a AX88772C phy.
> >
> > This means even with the option manually enabled, we'd still need to
> > cherrypick dde25846925765a88df8964080098174495c1f10
> > "net: usb/phy: asix: add support for ax88772A/C PHYs"
> > since apparently this is simply new(ish) hardware with built-in x88772C PHY.
>
>
> As far as I see, you are not using clean mainline stable kernel. All changes
> which make your kernel to need an external PHY driver are _not_ in v5.10.198.

No, the dmesg was actually from a (probably clean-ish 6.4-ish) debian
kernel on my laptop,
where the device enumerates and works in one of the 2 usb-c ports.

As I mentioned the hardware that actually runs 5.10 is having issues
even detecting my test device.
(and while that 5.10 is far from clean mainline, the usb and network
driver portions are more or less untouched)

> You will need to cherrypick at least 28 last patches from this stack:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/net/usb/asix_devices.c?h=v6.6-rc6
>
> and some from here too:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/net/usb/usbnet.c?h=v6.6-rc6

I think this basically means the answer is "if you want this to work
upgrade to a newer kernel".
Which of course won't make any users happy, but oh well...

> Regards,
> Oleksij
> --
> Pengutronix e.K.                           |                             |
> Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
> 31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
> Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

Re: USB_NET_AX8817X dependency on AX88796B_PHY

[Oleksij Rempel]

On Mon, Oct 16, 2023 at 11:27:07AM -0700, Maciej Żenczykowski wrote:
> On Mon, Oct 16, 2023 at 1:29 AM Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> > On Sun, Oct 15, 2023 at 03:13:26PM -0700, Maciej Żenczykowski wrote:
> > > This means even with the option manually enabled, we'd still need to
> > > cherrypick dde25846925765a88df8964080098174495c1f10
> > > "net: usb/phy: asix: add support for ax88772A/C PHYs"
> > > since apparently this is simply new(ish) hardware with built-in x88772C PHY.
> >
> >
> > As far as I see, you are not using clean mainline stable kernel. All changes
> > which make your kernel to need an external PHY driver are _not_ in v5.10.198.
> 
> No, the dmesg was actually from a (probably clean-ish 6.4-ish) debian
> kernel on my laptop,
> where the device enumerates and works in one of the 2 usb-c ports.
> 
> As I mentioned the hardware that actually runs 5.10 is having issues
> even detecting my test device.
> (and while that 5.10 is far from clean mainline, the usb and network
> driver portions are more or less untouched)
> 
> > You will need to cherrypick at least 28 last patches from this stack:
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/net/usb/asix_devices.c?h=v6.6-rc6
> >
> > and some from here too:
> > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/net/usb/usbnet.c?h=v6.6-rc6
> 
> I think this basically means the answer is "if you want this to work
> upgrade to a newer kernel".
> Which of course won't make any users happy, but oh well...

It means - what ever problem there is, it is most probably not related
to the asix driver. In kernel v5.10, there are no external dependencies
to other PHY drivers.

Regards,
Oleksij
-- 
Pengutronix e.K.                           |                             |
Steuerwalder Str. 21                       | http://www.pengutronix.de/  |
31137 Hildesheim, Germany                  | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

Re: USB_NET_AX8817X dependency on AX88796B_PHY

[Maciej Żenczykowski]

On Mon, Oct 16, 2023 at 9:20 PM Oleksij Rempel <o.rempel@pengutronix.de> wrote:
>
> On Mon, Oct 16, 2023 at 11:27:07AM -0700, Maciej Żenczykowski wrote:
> > On Mon, Oct 16, 2023 at 1:29 AM Oleksij Rempel <o.rempel@pengutronix.de> wrote:
> > > On Sun, Oct 15, 2023 at 03:13:26PM -0700, Maciej Żenczykowski wrote:
> > > > This means even with the option manually enabled, we'd still need to
> > > > cherrypick dde25846925765a88df8964080098174495c1f10
> > > > "net: usb/phy: asix: add support for ax88772A/C PHYs"
> > > > since apparently this is simply new(ish) hardware with built-in x88772C PHY.
> > >
> > >
> > > As far as I see, you are not using clean mainline stable kernel. All changes
> > > which make your kernel to need an external PHY driver are _not_ in v5.10.198.
> >
> > No, the dmesg was actually from a (probably clean-ish 6.4-ish) debian
> > kernel on my laptop,
> > where the device enumerates and works in one of the 2 usb-c ports.
> >
> > As I mentioned the hardware that actually runs 5.10 is having issues
> > even detecting my test device.
> > (and while that 5.10 is far from clean mainline, the usb and network
> > driver portions are more or less untouched)
> >
> > > You will need to cherrypick at least 28 last patches from this stack:
> > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/net/usb/asix_devices.c?h=v6.6-rc6
> > >
> > > and some from here too:
> > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/log/drivers/net/usb/usbnet.c?h=v6.6-rc6
> >
> > I think this basically means the answer is "if you want this to work
> > upgrade to a newer kernel".
> > Which of course won't make any users happy, but oh well...
>
> It means - what ever problem there is, it is most probably not related
> to the asix driver. In kernel v5.10, there are no external dependencies
> to other PHY drivers.

I take this to mean you think that the built-in ax88772C PHY should
work out of the box with 5.10,
and if it doesn't then this presumably means there's something wrong
at the usb controller level.
(hopefully I'll get a second unit and be able to confirm this...)

Re: USB_NET_AX8817X dependency on AX88796B_PHY

[Maciej Żenczykowski]

> I take this to mean you think that the built-in ax88772C PHY should
> work out of the box with 5.10,
> and if it doesn't then this presumably means there's something wrong
> at the usb controller level.
> (hopefully I'll get a second unit and be able to confirm this...)

Just to follow up on this.

I got my hands on a second identical adapter (the one that got the bug
filed in the first place).

It behaves (ie. fails) in exactly the same way.
ie. dmesg (on multiple devices, but only on some usb-c ports) doesn't
even show usb connect events, nothing on lsusb (of course), etc.

(The fact that both of these adapters work on exactly the same 1 of 2
usb-c ports on my lenovo thinkpad laptop is particularly weird...
you'd think those would be identical...)

I'm going to blame this on usb controller compatibility issues.

I'll have to track down someone with a usb wire sniffer and a desire
to debug this further.