The backport of upstream ea5df88aeca1 introduces a regression on 6.6.y stable

The backport of upstream ea5df88aeca1 introduces a regression on 6.6.y stable

[He, Guocai (CN)]

Horatiu.vultur:

Commit 85ede044f43d ("phy: mscc: Fix PTP for VSC8574 and VSC8572"),
which is a backport of upstream ea5df88aeca1, introduces a regression
on 6.6.y stable.

The commit changes the probe function for VSC8574 and VSC8572 from
vsc8574_probe to vsc8584_probe. However, vsc8584_probe on 6.6.y
contains a PHY ID check that rejects anything that is not VSC8584 revB:

    if ((phydev->phy_id & MSCC_DEV_REV_MASK) != VSC8584_REVB) {
        dev_err(..., "Only VSC8584 revB is supported.\n");
        return -ENOTSUPP;
    }

This causes VSC8574/VSC8572 to fail to probe:
logs:
Microsemi GE VSC8574 SyncE ff0d0000.ethernet-ffffffff:08: Only VSC8584 revB is supported.
Microsemi GE VSC8574 SyncE: probe of ff0d0000.ethernet-ffffffff:08 failed with error -524

On mainline, this check was removed by commit 1bc80d673087 ("phy: mscc:
Use PHY_ID_MATCH_EXACT for VSC8584, VSC8582, VSC8575, VSC856X"), which
was patch 1/2 of the same series. However, only patch 2/2 (ea5df88aeca1)
was backported to 6.6.y, without its prerequisite.

Who know why? 


Regards,
Guocai He

Re: The backport of upstream ea5df88aeca1 introduces a regression on 6.6.y stable

[Andrew Lunn]

> On mainline, this check was removed by commit 1bc80d673087 ("phy: mscc:
> Use PHY_ID_MATCH_EXACT for VSC8584, VSC8582, VSC8575, VSC856X"), which
> was patch 1/2 of the same series. However, only patch 2/2 (ea5df88aeca1)
> was backported to 6.6.y, without its prerequisite.
> 
> Who know why? 

1bc80d673087 does not have a Fixes: tag. It is not supper clear it is
a dependency for the next patch.

GregKH, Sasha: Please add:

Fixes: 1bc80d673087 ("phy: mscc: Use PHY_ID_MATCH_EXACT for VSC8584, VSC8582, VSC8575, VSC856X")

to stable.

Thanks
	Andrew

Re: The backport of upstream ea5df88aeca1 introduces a regression on 6.6.y stable

[Sasha Levin]

> ea5df88aeca1 switched VSC8574/VSC8572 to vsc8584_probe() but its
> prerequisite 1bc80d673087 ("phy: mscc: Use PHY_ID_MATCH_EXACT for
> VSC8584, VSC8582, VSC8575, VSC856X") was not backported, so the revB
> guard rejects these PHYs.

Thanks for tracking this down, and thanks to Andrew for confirming the
missing prerequisite.

Queued 1bc80d673087 for 6.18.y, 6.12.y, 6.6.y, 6.1.y and 5.15.y. On
6.1.y and 5.15.y it also needed its own prerequisite 31605c01fb24
("phy: mscc: Use PHY_ID_MATCH_VENDOR to minimize PHY ID table"), which
I've queued there as well.

-- 
Thanks,
Sasha