[Public]
-----Original Message----- From: Rafael J. Wysocki rafael@kernel.org Sent: Tuesday, July 5, 2022 13:24 To: Chuanhong Guo gch981213@gmail.com Cc: Rafael J. Wysocki rafael@kernel.org; Limonciello, Mario Mario.Limonciello@amd.com; ACPI Devel Maling List <linux- acpi@vger.kernel.org>; Stable stable@vger.kernel.org; Tighe Donnelly tighe.donnelly@protonmail.com; Kent Hou Man knthmn0@gmail.com; Len Brown lenb@kernel.org; open list linux-kernel@vger.kernel.org Subject: Re: [PATCH v5] ACPI: skip IRQ1 override on 3 Ryzen 6000 laptops
On Fri, Jul 1, 2022 at 2:45 PM Chuanhong Guo gch981213@gmail.com wrote:
On Fri, Jul 1, 2022 at 4:12 AM Limonciello, Mario mario.limonciello@amd.com wrote:
However I do want to point out that Windows doesn't care about legacy format or not. This bug where keyboard doesn't work only popped up on Linux.
Given the number of systems with the bug is appearing to grow I wonder if the right answer is actually a new heuristic that doesn't apply the kernel override for polarity inversion anymore. Maybe if the system is 2022 or newer? Or on the ACPI version?
The previous attempt to limit the scope of IRQ override ends up breaking some other buggy devices:
https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fpatc hwork.kernel.org%2Fproject%2Flinux- acpi%2Fpatch%2F20210728151958.15205-1- hui.wang%40canonical.com%2F&data=05%7C01%7Cmario.limonciello%4 0amd.com%7C106955e4611344d3bc3808da5eb3971d%7C3dd8961fe4884e608 e11a82d994e183d%7C0%7C0%7C637926422673112765%7CUnknown%7CTWF pbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXV CI6Mn0%3D%7C3000%7C%7C%7C&sdata=xOaRbkCv9EMhpLO%2BGAP mDjEhQ78xjYFBvehLZdg1k1I%3D&reserved=0
It's unfortunate that the original author of this IRQ override doesn't limit the scope to their exact devices.
Hi, Rafael! What do you think? should we skip this IRQ override one-by-one or add a different matching logic to check the bios date instead?
It would be better to find something precise enough to identify the machines in question without pulling in the others and use that for skipping the override instead of listing them all one by one in the blocklist.
How about using the CPU family/model in this case?