Re: Backport using Microsoft GUID for s2idle on AMD

[Greg KH <gregkh@linuxfoundation.org>]

On Thu, Jan 05, 2023 at 09:09:12AM -0600, Limonciello, Mario wrote:
> On 1/5/2023 01:45, Greg KH wrote:
> > On Wed, Jan 04, 2023 at 07:46:58PM +0000, Limonciello, Mario wrote:
> > > [Public]
> > > 
> > > Hi,
> > > 
> > > At *least* 9 models of laptops across manufacturers have problems with suspend that are root caused
> > > to the firmware not properly implementing an AMD specific codepath, but that did implement a
> > > Microsoft specific one properly. To fix the suspend issues on Linux, a number of commits have been
> > > worked out over the last few kernel releases.
> > > 
> > > We have eventually landed at we're going to just use the Microsoft codepath in Linux.
> > 
> > That is the correct solution as that is the only codepath that vendors
> > test.  And it's what we do for the rest of ACPI, and have done, for
> > decades now.  Odd that it wasn't the way this was done originally.
> 
> Yup :/
> 
> > 
> > > All the patches to accomplish this are now landed in 6.2-rc1, and also in 6.1.3.
> > > 
> > > Now that have this all working satisfactorily I'd like to bring those fixes into 6.0.y and 5.15.y as well.
> > > 
> > > Here is the series of commits for 6.0.y:
> > > 
> > > 100a57379380 ACPI: x86: s2idle: Move _HID handling for AMD systems into structures
> > > fd894f05cf30 ACPI: x86: s2idle: If a new AMD _HID is missing assume Rembrandt
> > > a0bc002393d4 ACPI: x86: s2idle: Add module parameter to prefer Microsoft GUID
> > > d0f61e89f08d ACPI: x86: s2idle: Add a quirk for ASUS TUF Gaming A17 FA707RE
> > > ddeea2c3cb88 ACPI: x86: s2idle: Add a quirk for ASUS ROG Zephyrus G14
> > > 888ca9c7955e ACPI: x86: s2idle: Add a quirk for Lenovo Slim 7 Pro 14ARH7
> > > 631b54519e8e ACPI: x86: s2idle: Add a quirk for ASUSTeK COMPUTER INC. ROG Flow X13
> > > 39f81776c680 ACPI: x86: s2idle: Fix a NULL pointer dereference
> > > 54bd1e548701 ACPI: x86: s2idle: Add another ID to s2idle_dmi_table
> > > 577821f756cf ACPI: x86: s2idle: Force AMD GUID/_REV 2 on HP Elitebook 865
> > > e6d180a35bc0 ACPI: x86: s2idle: Stop using AMD specific codepath for Rembrandt+
> > 
> > 6.0 is about to go end-of-life in a few days, so this isn't needed.
> 
> Got it; thanks.
> 
> > 
> > > Here is the series of commits for 5.15.y:
> > > 
> > > ed470febf837 ACPI: PM: s2idle: Add support for upcoming AMD uPEP HID AMDI008
> > > 1a2dcab517cb ACPI: PM: s2idle: Use LPS0 idle if ACPI_FADT_LOW_POWER_S0 is unset
> > > 100a57379380 ACPI: x86: s2idle: Move _HID handling for AMD systems into structures
> > > fd894f05cf30 ACPI: x86: s2idle: If a new AMD _HID is missing assume Rembrandt
> > > a0bc002393d4 ACPI: x86: s2idle: Add module parameter to prefer Microsoft GUID
> > > d0f61e89f08d ACPI: x86: s2idle: Add a quirk for ASUS TUF Gaming A17 FA707RE
> > > ddeea2c3cb88 ACPI: x86: s2idle: Add a quirk for ASUS ROG Zephyrus G14
> > > 888ca9c7955e ACPI: x86: s2idle: Add a quirk for Lenovo Slim 7 Pro 14ARH7
> > > 631b54519e8e ACPI: x86: s2idle: Add a quirk for ASUSTeK COMPUTER INC. ROG Flow X13
> > > 39f81776c680 ACPI: x86: s2idle: Fix a NULL pointer dereference
> > > 54bd1e548701 ACPI: x86: s2idle: Add another ID to s2idle_dmi_table
> > > 577821f756cf ACPI: x86: s2idle: Force AMD GUID/_REV 2 on HP Elitebook 865
> > > e6d180a35bc0 ACPI: x86: s2idle: Stop using AMD specific codepath for Rembrandt+
> > > 
> > > Can you please backport these for a future stable release?
> > 
> > That's adding new features for an old kernel, what's keeping people with
> > this hardware from using the 6.1 kernel instead?  What distros are
> > insisting that this needs to be in 5.15.y?
> > 
> > thanks,
> > 
> > greg k-h
> 
> I don't believe it's a new feature..
> 
> The Microsoft GUID was already available and usable in 5.15.y, but the
> policy was set for AMDI0007 (which is used for Rembrandt) to pick the AMD
> GUID.  This helps Rembrandt laptops that otherwise work well with 5.15.y.
> 
> Ubuntu 22.04 and Ubuntu 20.04 both track 5.15.y still.
> 
> But you prompted me to realize a MUCH easier solution for 5.15.y.
> Just revert this commit:
> 
> f0c6225531e4 ("ACPI: PM: Add support for upcoming AMD uPEP HID AMDI007")
> 
> That will effectively line up Rembrandt with what is in 6.1.y and 6.2
> already without needing that bigger list of patches.

That works for me!  Please submit the revert so we can get a full reason
and documentation for why it's being reverted here only and not in
Linus's tree.

thanks

greg k-h