Re: [PATCH] efi/bgrt: Accept BGRT tables with a version of 0 on Lenovo laptops

[Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx>]

On Sun, 26 Jan 2020 at 16:02, Hans de Goede <hdegoede@xxxxxxxxxx> wrote:
>
> Some (somewhat older) Lenovo laptops have a correct BGRT table, except
> that the version field is 0 instead of 1.
>
> Quickly after finding this out, even before submitting a first version of
> this patch upstream, the list of DMI matches for affected models grew to
> 3 models (all Ivy Bridge based).
>
> So rather then maintaining an ever growing list with DMI matches for
> affected Lenovo models, this commit simply checks if the vendor is Lenovo
> when the version is 0 and in that case accepts the out of spec version
> working around the Lenovo firmware bug.
>
> Cc: stable@xxxxxxxxxxxxxxx
> BugLink: https://bugzilla.redhat.com/show_bug.cgi?id=1791273
> Signed-off-by: Hans de Goede <hdegoede@xxxxxxxxxx>
> ---
>  drivers/firmware/efi/efi-bgrt.c | 8 +++++++-
>  1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/drivers/firmware/efi/efi-bgrt.c b/drivers/firmware/efi/efi-bgrt.c
> index b07c17643210..3a2d6d3a590b 100644
> --- a/drivers/firmware/efi/efi-bgrt.c
> +++ b/drivers/firmware/efi/efi-bgrt.c
> @@ -15,6 +15,7 @@
>  #include <linux/acpi.h>
>  #include <linux/efi.h>
>  #include <linux/efi-bgrt.h>
> +#include <linux/dmi.h>
>
>  struct acpi_table_bgrt bgrt_tab;
>  size_t bgrt_image_size;
> @@ -42,7 +43,12 @@ void __init efi_bgrt_init(struct acpi_table_header *table)
>                 return;
>         }
>         *bgrt = *(struct acpi_table_bgrt *)table;
> -       if (bgrt->version != 1) {
> +       /*
> +        * Some older Lenovo laptops have a correct BGRT table, except that
> +        * the version field is 0 instead of 1.
> +        */
> +       if (bgrt->version != 1 &&
> +           !(bgrt->version == 0 && dmi_name_in_vendors("LENOVO"))) {
>                 pr_notice("Ignoring BGRT: invalid version %u (expected 1)\n",
>                        bgrt->version);
>                 goto out;

Hi Hans,

Given that the ACPI spec only defines a single version for this table,
which is version #1, wouldn't it be simpler to just assume that
version #0 means version #1 in all cases, rather than using DMI
matches for that? There is no risk of misidentifying another table
version, since none exist ...