Re: [PATCH] net: phy: bcm7xxx: preserve __phy_write() error in write_mmd()

[Andrew Lunn]

On Fri, Oct 09, 2026 at 12:42:40AM +0800, Haotian Zhang wrote:
> bcm7xxx_28nm_ephy_write_mmd() jumps to reset_shadow_mode when the write of
> the shadow register address fails and then returns the value of the
> __phy_set_clr_bits() cleanup call, overwriting the negative error code from
> __phy_write(). Since __phy_set_clr_bits() returns the positive register
> value on success, the failure is reported to phy_write_mmd() as a success.
> The write of the shadow register value is also unchecked the same way.
> 
> Keep the negative error code in ret and return it after the shadow mode
> reset, and check the shadow register value write, matching the read_mmd()
> counterpart.

Probably not worth the churn. MDIO operation don't fail. It is a big
shift register which just clocks the bits out on the MDIO line. There
is no checksum, there is no checking if the device responded, nothing.

The only time i've seen an MDIO bus fail is because the driver for it
was fatally broken resulting in all operations failing. And we would
not of got this far if the bus driver was FUBAR.

Please work on fixing issues which bother people, not theoretical
issues which never happen.

    Andrew

---
pw-bot: cr