Re: [5.12 regression] DSDT overriding from initrd no longer works

["Rafael J. Wysocki" <rafael@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 2280 bytes --]

On Mon, Apr 12, 2021 at 8:53 PM Hans de Goede <hdegoede@redhat.com> wrote:
>
> Hi,
>
> On 4/12/21 8:11 PM, Rafael J. Wysocki wrote:
> > On Mon, Apr 12, 2021 at 8:01 PM Rafael J. Wysocki <rafael@kernel.org> wrote:
> >>
> >> On Mon, Apr 12, 2021 at 7:38 PM Hans de Goede <hdegoede@redhat.com> wrote:
> >>>
> >>> Hi Rafael,
> >>>
> >>> Sorry about the timing of reporting this regression.
> >>
> >> Oh well.
> >>
> >>> I just noticed that overriding the DSDT (*) from the initrd will not work in 5.12,
> >>> this is caused by:
> >>>
> >>> commit 1a1c130ab757 ("ACPI: tables: x86: Reserve memory occupied by ACPI tables")
> >>>
> >>> This makes the initial acpi_locate_initial_tables() call happen earlier
> >>> then before, but the acpi_table_upgrade) call in arch/x86/kernel/setup.c is
> >>> not moved up, so the tables in the initrd are now only parsed and saved
> >>> after the initial ACPI table scanning has already been done.
> >>>
> >>> I guess fixing this might be as easy as moving the acpi_table_upgrade) call
> >>> higher in arch/x86/kernel/setup.c but I'm not sure if that is save to do.
> >>
> >> Why do you think it may not be safe?
> >
> > OK, so it won't work in some cases, because acpi_table_upgrade() needs
> > to be called after reserve_initrd(),
>
> Right I notice it was sitting right after reserve_initrd() which made me think
> that it probably needed to be after that. Sorry I should have mentioned that
> in my original email.
>
> > so I guess the commit above will
> > need to be reverted.
>
> One possible solution which I was wondering about is to modify
> acpi_table_initrd_scan() to have it call acpi_tb_override_table()
> instead of acpi_install_table() for existing tables using the matching
> logic from acpi_table_initrd_override(). But I'm not sure when the
> parsing of the DSDT is done. If acpi_table_initrd_scan() runs before
> the first parsing of the DSDT is done then I think that that should work.
>
> This might be more 5.13 material though and for 5.12 a revert is
> probably best.

The attached change should make it work again, though.  Can you please verify?

> I also just remembered that at least the Intel audio folks rely on
> DSDT overrides to get some (prototype) boards in their CI to work.

But they haven't complained so far.

[-- Attachment #2: acpi-x86-ordering.patch --]
[-- Type: text/x-patch, Size: 756 bytes --]

---
 arch/x86/kernel/setup.c |    5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

Index: linux-pm/arch/x86/kernel/setup.c
===================================================================
--- linux-pm.orig/arch/x86/kernel/setup.c
+++ linux-pm/arch/x86/kernel/setup.c
@@ -1045,9 +1045,6 @@ void __init setup_arch(char **cmdline_p)
 
 	cleanup_highmap();
 
-	/* Look for ACPI tables and reserve memory occupied by them. */
-	acpi_boot_table_init();
-
 	memblock_set_current_limit(ISA_END_ADDRESS);
 	e820__memblock_setup();
 
@@ -1132,6 +1129,8 @@ void __init setup_arch(char **cmdline_p)
 	reserve_initrd();
 
 	acpi_table_upgrade();
+	/* Look for ACPI tables and reserve memory occupied by them. */
+	acpi_boot_table_init();
 
 	vsmp_init();