Re: [REGRESSION] AX88772 card booted without cable can't receive

From: Oleksij Rempel
Date: Fri Jun 24 2022 - 04:17:16 EST


Hi Anton,

On Thu, Jun 23, 2022 at 08:36:49AM +0200, Oleksij Rempel wrote:
> Hi Anton,
>
> Thank you for your report! I'll take a look on it ASAP.
>
> Regards,
> Oleksij
>
> On Wed, Jun 22, 2022 at 04:16:38PM +0200, Anton Lundin wrote:
> > Hi.
> >
> > I've found a issue with a Dlink usb ether adapter, that can't receive
> > anything until it self transmits if it's plugged in while booting, and
> > doesn't have link.
> >
> > Later when a cable is attached, link is detected but nothing is received
> > either by daemons listening to ip address on that interface, or seen
> > with tcpdump.
> >
> > The dongle is a:
> > D-Link Corp. DUB-E100 Fast Ethernet Adapter(rev.C1) [ASIX AX88772]
> >
> > And it's detected at boot as:
> > libphy: Asix MDIO Bus: probed
> > Asix Electronics AX88772C usb-003:004:10: attached PHY driver (mii_bus:phy_addr=usb-003:004:10, irq=POLL)
> > asix 3-10.4:1.0 eth1: register 'asix' at usb-0000:00:14.0-10.4, ASIX AX88772 USB 2.0 Ethernet, <masked-mac>
> > usbcore: registered new interface driver asix
> >
> >
> > While in this state, the hardware starts sending pause frames to the
> > network when it has recived a couple of frames, and they look like:
> > 0000 01 80 c2 00 00 01 00 00 00 00 00 00 88 08 00 01
> > 0010 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> > 0020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> > 0030 00 00 00 00 00 00 00 00 00 00 00 00
> >
> > 0000 01 80 c2 00 00 01 00 00 00 00 00 00 88 08 00 01
> > 0010 ff ff 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> > 0020 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
> > 0030 00 00 00 00 00 00 00 00 00 00 00 00
> >
> > And these two frames are repeated every couple of seconds.
> >
> > The card wakes up when something triggers a transmit on that card, and
> > then starts receiving traffic as normal.
> >
> > I've bisected this issue down to:
> > "net: usb: asix: ax88772: add phylib support" (e532a096be0e)
> >
> >
> > Reverting that makes the interface work as normal, even if the machine
> > boots without a cable plugged in.
> >
> > Another issue found with exactly the same patch is that if it's loaded
> > as a module, then unloaded and loaded again, it fails to initialize the
> > card with:
> >
> > sysfs: cannot create duplicate filename '/devices/virtual/mdio_bus/usb-003:004'
> > CPU: 0 PID: 3733 Comm: modprobe Tainted: G O 5.15.10-core_64_preempt #3
> > Hardware name: <masked-hardware-name>
> > Call Trace:
> > <TASK>
> > ? dump_stack_lvl+0x34/0x44
> > ? sysfs_warn_dup.cold+0x17/0x24
> > ? sysfs_create_dir_ns+0xbc/0xd0
> > ? kobject_add_internal+0xa6/0x260
> > ? kobject_add+0x7e/0xb0
> > ? preempt_count_add+0x68/0xa0
> > ? device_add+0x10f/0x8d0
> > ? dev_set_name+0x53/0x70
> > ? __mdiobus_register+0xc2/0x350
> > ? __devm_mdiobus_register+0x64/0xb0
> > ? ax88772_bind+0x22a/0x340 [asix]
> > ? usbnet_probe+0x346/0x870
> > ? usb_match_dynamic_id+0x8f/0xa0
> > ? usb_probe_interface+0x9b/0x150
> > ? really_probe.part.0+0x237/0x280
> > ? __driver_probe_device+0x8c/0xd0
> > ? driver_probe_device+0x1e/0xe0
> > ? __driver_attach+0xa8/0x170
> > ? __device_attach_driver+0xe0/0xe0
> > ? bus_for_each_dev+0x77/0xc0
> > ? bus_add_driver+0x10b/0x1c0
> > ? driver_register+0x8b/0xe0
> > ? usb_register_driver+0x84/0x120
> > ? 0xffffffffc06e4000
> > ? do_one_initcall+0x41/0x1f0
> > ? kmem_cache_alloc_trace+0x3f/0x1b0
> > ? do_init_module+0x5c/0x260
> > ? __do_sys_finit_module+0xa0/0xe0
> > ? do_syscall_64+0x35/0x80
> > ? entry_SYSCALL_64_after_hwframe+0x44/0xae
> > </TASK>
> > kobject_add_internal failed for usb-003:004 with -EEXIST, don't try to register things with the same name in the same directory.
> > libphy: mii_bus usb-003:004 failed to register
> > asix: probe of 3-10.4:1.0 failed with error -22
> > usbcore: registered new interface driver asix
> >
> >
> > Both these issues with "net: usb: asix: ax88772: add phylib support"
> > (e532a096be0e) can be reproduced all the way from when it was introduced
> > to linus current tree.
> >
> >
> > I'm sorry to say that I don't know enough about either libphy or asix to
> > figure out what cause the issues can be.

It looks like we have here 3 different bugs:
- no rx before tx
addresses by this patch:
https://lore.kernel.org/all/20220624075139.3139300-1-o.rempel@xxxxxxxxxxxxxx/
- pause frames flood without advertising pause frames support
addresses by this patch:
https://lore.kernel.org/all/20220624075139.3139300-2-o.rempel@xxxxxxxxxxxxxx/

Can you please test this patches.

- not everything is properly cleaned after module unload.

The last one i as bit more complicated. Lukas Wunner spend last months to solve
this issues:
https://lore.kernel.org/all/d1c87ebe9fc502bffcd1576e238d685ad08321e4.1655987888.git.lukas@xxxxxxxxx/

Not sure if this work can be easily backported to v5.15.

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 |