MT7621 ethernet does not get probed on net-next branch after 5.15 merge

Hi,

After the merge of 5.15.y into net-next, MT7621 ethernet
(mtk_eth_soc.c) does not get probed at all.

Kernel log before 5.15 merge:
...
libphy: Fixed MDIO Bus: probed
libphy: mdio: probed
mt7530 mdio-bus:1f: MT7530 adapts as multi-chip module
mtk_soc_eth 1e100000.ethernet eth0: mediatek frame engine at 0xbe100000, irq 20
mt7621-pci 1e140000.pcie: host bridge /pcie@1e140000 ranges:
...

Kernel log after 5.15 merge:
...
libphy: Fixed MDIO Bus: probed
mt7621-pci 1e140000.pcie: host bridge /pcie@1e140000 ranges:
...


I tried adding debug prints into the .mtk_probe function, but it did
not execute.
There are no dts changes for MT7621 between 5.14 and 5.15, so I
believe it should be something else.

Any ideas?

Hi Qingfang,

On Fri, Oct 15, 2021 at 4:23 PM DENG Qingfang [off-list ref] wrote:

Hi,

After the merge of 5.15.y into net-next, MT7621 ethernet
(mtk_eth_soc.c) does not get probed at all.

Kernel log before 5.15 merge:
...
libphy: Fixed MDIO Bus: probed
libphy: mdio: probed
mt7530 mdio-bus:1f: MT7530 adapts as multi-chip module
mtk_soc_eth 1e100000.ethernet eth0: mediatek frame engine at 0xbe100000, irq 20
mt7621-pci 1e140000.pcie: host bridge /pcie@1e140000 ranges:
...

Kernel log after 5.15 merge:
...
libphy: Fixed MDIO Bus: probed
mt7621-pci 1e140000.pcie: host bridge /pcie@1e140000 ranges:
...


I tried adding debug prints into the .mtk_probe function, but it did
not execute.
There are no dts changes for MT7621 between 5.14 and 5.15, so I
believe it should be something else.

Any ideas?

Maybe you have a chance to try git bisect to see what is the commit
that introduces that regression and report here?

Best regards,
    Sergio Paracuellos

Hi Qingfang,

On Fri, Oct 15, 2021 at 4:23 PM DENG Qingfang [off-list ref] wrote:

Hi,

After the merge of 5.15.y into net-next, MT7621 ethernet
(mtk_eth_soc.c) does not get probed at all.

Kernel log before 5.15 merge:
...
libphy: Fixed MDIO Bus: probed
libphy: mdio: probed
mt7530 mdio-bus:1f: MT7530 adapts as multi-chip module
mtk_soc_eth 1e100000.ethernet eth0: mediatek frame engine at 0xbe100000, irq 20
mt7621-pci 1e140000.pcie: host bridge /pcie@1e140000 ranges:
...

Kernel log after 5.15 merge:
...
libphy: Fixed MDIO Bus: probed
mt7621-pci 1e140000.pcie: host bridge /pcie@1e140000 ranges:
...


I tried adding debug prints into the .mtk_probe function, but it did
not execute.
There are no dts changes for MT7621 between 5.14 and 5.15, so I
believe it should be something else.

Any ideas?

I had time to create a new image for my gnubee board using kernel 5.15
and this problem does not exist on my side. Since no more mails have
come for a while I guess this was a problem from your configuration,
but just in case I preferred to answer to let you know. I am currently
using v5.15.7 from linux-stable with some other patches that will be
for 5.16. Just in case, you can check the kernel tree [0] I am
currently using.

Best regards,
     Sergio Paracuellos

[0]: https://github.com/paraka/linux/tree/gbpc1-5.15

In gmane.comp.embedded.openwrt.devel Sergio Paracuellos [off-list ref] wrote:

Hi Qingfang,

On Fri, Oct 15, 2021 at 4:23 PM DENG Qingfang [off-list ref] wrote:

quoted

Hi,

After the merge of 5.15.y into net-next, MT7621 ethernet
(mtk_eth_soc.c) does not get probed at all.

Kernel log before 5.15 merge:
...
libphy: Fixed MDIO Bus: probed
libphy: mdio: probed
mt7530 mdio-bus:1f: MT7530 adapts as multi-chip module
mtk_soc_eth 1e100000.ethernet eth0: mediatek frame engine at 0xbe100000, irq 20
mt7621-pci 1e140000.pcie: host bridge /pcie@1e140000 ranges:
...

Kernel log after 5.15 merge:
...
libphy: Fixed MDIO Bus: probed
mt7621-pci 1e140000.pcie: host bridge /pcie@1e140000 ranges:
...


I tried adding debug prints into the .mtk_probe function, but it did
not execute.
There are no dts changes for MT7621 between 5.14 and 5.15, so I
believe it should be something else.

Any ideas?

I had time to create a new image for my gnubee board using kernel 5.15
and this problem does not exist on my side. Since no more mails have
come for a while I guess this was a problem from your configuration,
but just in case I preferred to answer to let you know. I am currently
using v5.15.7 from linux-stable with some other patches that will be
for 5.16. Just in case, you can check the kernel tree [0] I am
currently using.

There is problem with reset controller and devlink commutication. reset
controller is abent in mainline, devlink defer all drivers which use reset
lines until reset-controller become available - so no drivers probed.
I'm create for myself this patch:
https://drive.google.com/file/d/1AiKlfvIgtrBsxtI-2XFBvaxGoE0S-s9d/view?usp=sharing

This is fixed with the patch series (yet to be applied upstream) in the 
link below.

https://lore.kernel.org/all/20220110114930.1406665-1-sergio.paracuellos@gmail.com/

On 15/10/2021 17:23, DENG Qingfang wrote:

Hi,

After the merge of 5.15.y into net-next, MT7621 ethernet
(mtk_eth_soc.c) does not get probed at all.

Kernel log before 5.15 merge:
...
libphy: Fixed MDIO Bus: probed
libphy: mdio: probed
mt7530 mdio-bus:1f: MT7530 adapts as multi-chip module
mtk_soc_eth 1e100000.ethernet eth0: mediatek frame engine at 0xbe100000, irq 20
mt7621-pci 1e140000.pcie: host bridge /pcie@1e140000 ranges:
...

Kernel log after 5.15 merge:
...
libphy: Fixed MDIO Bus: probed
mt7621-pci 1e140000.pcie: host bridge /pcie@1e140000 ranges:
...


I tried adding debug prints into the .mtk_probe function, but it did
not execute.
There are no dts changes for MT7621 between 5.14 and 5.15, so I
believe it should be something else.

Any ideas?