Re: [PATCH net] net: stmmac: dwmac-rk: keep the GMAC powered when the PHY wakes the system

[Igor Velkov <iav@iav.lv>]

On Thu, Oct 01, 2026 at 02:07:41PM +0200, Andrew Lunn wrote:
> Why is the PHY interrupting? What interrupts are left enabled? Should
> the mask be changed to disable everything except wol?

Only the WoL interrupt is left: rtl8211f_suspend() sets INER to PME
alone. The interrupt is the magic-packet wake itself.
resume_device_irqs() re-enables the line and it fires before the MAC
resumes; the handler reads INSR (0x80, PME) over MDIO with the GMAC
clocks off.

So the mask is fine; the problem is that with mac_managed_pm phylib
does not postpone the interrupt until the PHY resumes. Patch: [1].

No new version of this patch: the hang is a phylib issue, fixed by
[1]. With [1] and the GMAC powered down as today: 20/20 suspends (an
instrumented build logged no MDIO access to the powered-down GMAC).
Without it the same setup hung within 1 to 7 cycles in 14 of 16 runs,
some on instrumented builds; one run passed 25 cycles. What remains is that
rk_gmac_powerdown() disables phy-supply while the PHY is the wakeup
source; on Helios64 that regulator is always-on, so I cannot test a
fix for it.

[1] https://lore.kernel.org/r/20261002043548.1302145-1-iav@iav.lv

--
Igor Velkov