Re: [PATCH v4] ata: libata: disable LPM for WDC WD20EFAX-68FB5N0 hard drives

[Niklas Cassel <cassel@kernel.org>]

On Mon, May 05, 2025 at 08:09:44PM +0300, Mikko Juhani Korhonen wrote:
> ma 5.5.2025 klo 18.37 Niklas Cassel (cassel@kernel.org) kirjoitti:
> 
> > On Mon, May 05, 2025 at 10:58:22AM +0300, Mikko Juhani Korhonen wrote:
> > > > > On 5/1/25 05:36, Ioannis Barkas wrote:
> > > > >> It would be better to have more details on this since only the 2TB
> > > > >> model is targeted.
> > > I have two WD20EFAX-68FB5N0 and both behave consistently, unusable
> > > from 6.9.0 on as they just keep resetting the SATA link , but they
> > > have worked flawlessly with 6.8.x for a year and also now with the
> > > quirk applied.
> > > I have also tested a different WDC model with med_power_with_dipm on
> > > the same SATA port and there are no problems.
> > > But after you guys got suspicious I found there is in fact a SATA port
> > > on my motherboard where the WD20EFAX-68FB5N0 works with LPM on
> > This motherboard looks to be AMD chipset, so I assume AMD AHCI controller.
> >
> > Could you send the output of:
> > $ lspci -nn | grep -E "SATA|AHCI"
> 
> mjkorhon@taavi:~$ sudo lspci -nn | grep -E "SATA|AHCI"
> 01:00.1 SATA controller [0106]: Advanced Micro Devices, Inc. [AMD] 500
> Series Chipset SATA Controller [1022:43eb]


Hello Mario,

We are seeing some LPM issues with an AMD AHCI controller when using a
(fairly recent) WD drive (so I would be slightly surprised if LPM wasn't)
implemented correctly.

The AMD AHCI controller is however something that I haven't seen before:
AMD 500 Series Chipset SATA Controller [1022:43eb]

For AMD, I have mostly seen:

PCI_VDEVICE(AMD, 0x7800)
PCI_VDEVICE(AMD, 0x7801)
PCI_VDEVICE(AMD, 0x7900)
PCI_VDEVICE(AMD, 0x7901)


I have never seen any 0x43eb before.
Is this AHCI chip less common than the others?


We did add a quirk that disables LPM (when used with Samsung drives) for
ATI AHCI controllers recently:
https://lore.kernel.org/linux-ide/20250317170348.1748671-2-cassel@kernel.org/


Could perhaps this 0x43eb be a rebranded ATI AHCI controller?


If we had LPM problems with AMD 0x78XX/0x79XX AHCI controllers,
I'm quite sure that we would have seen way more bug reports by now.


Kind regards,
Niklas