| Message ID | 20221215191617.1438-1-mario.limonciello@amd.com |
|---|---|
| Headers | show |
| Series | Stop using AMD GUID/_REV 2 by default | expand |
On Thu, Dec 15, 2022 at 8:16 PM Mario Limonciello <mario.limonciello@amd.com> wrote: > > A number of laptops have been showing up where lots of EC controlled > features weren't working after resume. They've varied from KBD > backlight, to fans, brightness control and lots more. > In kernel 6.1 we introduced a module parameter through > commit a0bc002393d4 ("ACPI: x86: s2idle: Add module parameter to > prefer Microsoft GUID") and a series of quirks in follow up commits > for systems that people reported the problems. > > 3 more systems recently reported issues; and so rather than increasing > the list /again/ to add these new systems we took a hard look at the > "why". > > The AMD GUID/_REV 2 path was introduced for vendors to be able to > differentiate from the Microsoft path. Vendors could populate this > with unique code for their designs. Conceptually this was supposed > to help the ecosystem, however in practice we've found that there > are more machines that don't populate it than do. > > The only models that have populated this with unique code for avoiding > a bug specific to their design is the HP Elitebook 835, 845, and 865 G9 > systems. > > To avoid growing the list further this series rips out the module > parameter support, all the quirks and sets the default policy to follow > the Microsoft GUID path for AMD Rembrandt or later. We validated this > on OEM systems and we found this fixes them. > > To avoid regressing the HP systems that use the AMD GUID/_REV 2 > path, let them keep taking it. The reason they take it is believed to > be a bug with WLAN firmware. If this is fixed in the future, we may > consider dropping the HP systems as well and having no quirks. > > Mario Limonciello (2): > ACPI: x86: s2idle: Force AMD GUID/_REV 2 on HP Elitebook 865 > ACPI: x86: s2idle: Stop using AMD specific codepath for Rembrandt+ > > drivers/acpi/x86/s2idle.c | 87 ++++++--------------------------------- > 1 file changed, 13 insertions(+), 74 deletions(-) > > -- Both patches applied as 6.2-rc material, thanks!
A number of laptops have been showing up where lots of EC controlled features weren't working after resume. They've varied from KBD backlight, to fans, brightness control and lots more. In kernel 6.1 we introduced a module parameter through commit a0bc002393d4 ("ACPI: x86: s2idle: Add module parameter to prefer Microsoft GUID") and a series of quirks in follow up commits for systems that people reported the problems. 3 more systems recently reported issues; and so rather than increasing the list /again/ to add these new systems we took a hard look at the "why". The AMD GUID/_REV 2 path was introduced for vendors to be able to differentiate from the Microsoft path. Vendors could populate this with unique code for their designs. Conceptually this was supposed to help the ecosystem, however in practice we've found that there are more machines that don't populate it than do. The only models that have populated this with unique code for avoiding a bug specific to their design is the HP Elitebook 835, 845, and 865 G9 systems. To avoid growing the list further this series rips out the module parameter support, all the quirks and sets the default policy to follow the Microsoft GUID path for AMD Rembrandt or later. We validated this on OEM systems and we found this fixes them. To avoid regressing the HP systems that use the AMD GUID/_REV 2 path, let them keep taking it. The reason they take it is believed to be a bug with WLAN firmware. If this is fixed in the future, we may consider dropping the HP systems as well and having no quirks. Mario Limonciello (2): ACPI: x86: s2idle: Force AMD GUID/_REV 2 on HP Elitebook 865 ACPI: x86: s2idle: Stop using AMD specific codepath for Rembrandt+ drivers/acpi/x86/s2idle.c | 87 ++++++--------------------------------- 1 file changed, 13 insertions(+), 74 deletions(-)