Re: [PATCH 3/3] efi: Clear up misconceptions about a maximum variable name size

[Ard Biesheuvel <ardb@kernel.org>]

On Fri, 15 Mar 2024 at 01:27, Tim Schumacher <timschumi@gmx.de> wrote:
>
> The UEFI specification does not make any mention of a maximum variable
> name size, so the headers and implementation shouldn't claim that one
> exists either.
>
> Comments referring to this limit have been removed or rewritten, as this
> is an implementation detail local to the Linux kernel.
>
> Where appropriate, the magic value of 1024 has been replaced with
> EFI_VAR_NAME_LEN, as this is used for the efi_variable struct
> definition. This in itself does not change any behavior, but should
> serve as points of interest when making future changes in the same area.
>
> A related build-time check has been added to ensure that the special
> 512 byte sized buffer will not overflow with a potentially decreased
> EFI_VAR_NAME_LEN.
>
> Signed-off-by: Tim Schumacher <timschumi@gmx.de>

Shouldn't we switch to 512 everywhere while at it?

> ---
>  drivers/firmware/efi/vars.c | 2 +-
>  fs/efivarfs/vars.c          | 5 +++--
>  include/linux/efi.h         | 9 ++++-----
>  3 files changed, 8 insertions(+), 8 deletions(-)
>
> diff --git a/drivers/firmware/efi/vars.c b/drivers/firmware/efi/vars.c
> index f654e6f6af87..4056ba7f3440 100644
> --- a/drivers/firmware/efi/vars.c
> +++ b/drivers/firmware/efi/vars.c
> @@ -215,7 +215,7 @@ efi_status_t efivar_set_variable_locked(efi_char16_t *name, efi_guid_t *vendor,
>
>         if (data_size > 0) {
>                 status = check_var_size(nonblocking, attr,
> -                                       data_size + ucs2_strsize(name, 1024));
> +                                       data_size + ucs2_strsize(name, EFI_VAR_NAME_LEN));
>                 if (status != EFI_SUCCESS)
>                         return status;
>         }
> diff --git a/fs/efivarfs/vars.c b/fs/efivarfs/vars.c
> index 4d722af1014f..3cc89bb624f0 100644
> --- a/fs/efivarfs/vars.c
> +++ b/fs/efivarfs/vars.c
> @@ -295,9 +295,9 @@ static bool variable_is_present(efi_char16_t *variable_name, efi_guid_t *vendor,
>         unsigned long strsize1, strsize2;
>         bool found = false;
>
> -       strsize1 = ucs2_strsize(variable_name, 1024);
> +       strsize1 = ucs2_strsize(variable_name, EFI_VAR_NAME_LEN);
>         list_for_each_entry_safe(entry, n, head, list) {
> -               strsize2 = ucs2_strsize(entry->var.VariableName, 1024);
> +               strsize2 = ucs2_strsize(entry->var.VariableName, EFI_VAR_NAME_LEN);
>                 if (strsize1 == strsize2 &&
>                         !memcmp(variable_name, &(entry->var.VariableName),
>                                 strsize2) &&
> @@ -396,6 +396,7 @@ int efivar_init(int (*func)(efi_char16_t *, efi_guid_t, unsigned long, void *,
>
>         do {
>                 variable_name_size = 512;
> +               BUILD_BUG_ON(EFI_VAR_NAME_LEN < 512);
>
>                 status = efivar_get_next_variable(&variable_name_size,
>                                                   variable_name,
> diff --git a/include/linux/efi.h b/include/linux/efi.h
> index c74f47711f0b..62f552057b06 100644
> --- a/include/linux/efi.h
> +++ b/include/linux/efi.h
> @@ -1065,12 +1065,11 @@ static inline u64 efivar_reserved_space(void) { return 0; }
>  #endif
>
>  /*
> - * The maximum size of VariableName + Data = 1024
> - * Therefore, it's reasonable to save that much
> - * space in each part of the structure,
> - * and we use a page for reading/writing.
> + * There is no actual upper limit specified for the variable name size.
> + *
> + * This limit exists only for practical purposes, since name conversions
> + * are bounds-checked and name data is occasionally stored in-line.
>   */
> -
>  #define EFI_VAR_NAME_LEN       1024
>
>  int efivars_register(struct efivars *efivars,
> --
> 2.44.0
>