efi: Change grub_efi_boolean_t from char to enum
PATCH 1/1 efi: Change grub_efi_boolean_t from char to enum
This allows using GRUB_EFI_TRUE and GRUB_EFI_FALSE with proper type and value checking. The UEFI 2.10 specification, in section 2.3.1, table 2.3, says the size of the boolean is 1 byte and may only contain the values 0 or
1. In order to have the enum be 1-byte in size instead of the default int-sized, add the packed attribute to the enum.
--- include/grub/efi/api.h | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)
include/grub/efi/api.h +7 -1
@@ -552,7 +552,13 @@ enum grub_efi_reset_typetypedef enum grub_efi_reset_type grub_efi_reset_type_t ;/* Types. */ -typedef char grub_efi_boolean_t ; +enum GRUB_PACKED grub_efi_boolean + { +GRUB_EFI_FALSE , + GRUB_EFI_TRUE + }; +typedef enum grub_efi_boolean grub_efi_boolean_t ; +#if GRUB_CPU_SIZEOF_VOID_P ==8 typedef grub_int64_t grub_efi_intn_t ;typedef grub_uint64_t grub_efi_uintn_t ;
Signature · 7 lines
2.34.1 _______________________________________________ Grub-devel mailing list [email protected] https://lists.gnu.org/mailman/listinfo/grub-devel
Quoted 5 lines · Glenn Washburn
This allows using
GRUB_EFI_TRUEandGRUB_EFI_FALSEwith proper type and value checking. The UEFI 2.10 specification, in section 2.3.1, table 2.3, says the size of the boolean is 1 byte and may only contain the values 0 or
1. In order to have the enum be 1-byte in size instead of the default int-sized, add the packed attribute to the enum.
Hmmm... Could you point out us an official doc saying "packed" attribute does that?
And I hope you tested that change because we use grub_efi_boolean_t type quite often...
Quoted 10 lines
Signed-off-by Glenn Washburn <[email protected]>--- include/grub/efi/api.h | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)include/grub/efi/api.h +0
-0
include/grub/efi/api.h enum grub_efi_reset_type · line 557
Quoted 7 lines
include/grub/efi/api.h +3
-1@@ -552,7 +552,13 @@ enum grub_efi_reset_typetypedef enum grub_efi_reset_type grub_efi_reset_type_t ;/* Types. */ -typedef char grub_efi_boolean_t ; +enum GRUB_PACKED grub_efi_boolean + { + GRUB_EFI_FALSE,
I would explicitly do
GRUB_EFI_FALSE = 0,
include/grub/efi/api.h enum grub_efi_reset_type · line 559
Quoted 2 lines
include/grub/efi/api.h +2
-0+ GRUB_EFI_TRUE + };
I would move GRUB_PACKED here to be inline with its other uses.
Daniel
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel
I see little value in using enum here and too much compiler variance about how it's handled. Compiler is pretty much allowed to choose any type for enum. If we go this route at all we should at very least do a compile time assert. Ideally I would keep it as-is. C unlike C++ does nothing to enforce enum. Feel free to put GRUB_EFI_TRUE and FALSE into unnamed enum
Le lun. 14 août 2023, 23:39, Glenn Washburn <[email protected]> a écrit :
Quoted 34 lines
This allows using
GRUB_EFI_TRUEandGRUB_EFI_FALSEwith proper type and value checking. The UEFI 2.10 specification, in section 2.3.1, table 2.3, says the size of the boolean is 1 byte and may only contain the values 0 or
1. In order to have the enum be 1-byte in size instead of the default int-sized, add the packed attribute to the enum.Signed-off-by Glenn Washburn <[email protected]>--- include/grub/efi/api.h | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)include/grub/efi/api.h +7
-2@@ -552,7 +552,13 @@ enum grub_efi_reset_typetypedef enum grub_efi_reset_type grub_efi_reset_type_t ;/* Types. */ -typedef char grub_efi_boolean_t ; +enum GRUB_PACKED grub_efi_boolean + { +GRUB_EFI_FALSE , + GRUB_EFI_TRUE + }; +typedef enum grub_efi_boolean grub_efi_boolean_t ; +#if GRUB_CPU_SIZEOF_VOID_P ==8 typedef grub_int64_t grub_efi_intn_t ;typedef grub_uint64_t grub_efi_uintn_t ; --2.34.1
Quoted 5 lines
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel
Quoted 9 lines · On Thu, 31 Aug 2023 20:05:37 +0200 Daniel Kiper <[email protected]> wrote:
Quoted 5 lines · Glenn Washburn
This allows using
GRUB_EFI_TRUEandGRUB_EFI_FALSEwith proper type and value checking. The UEFI 2.10 specification, in section 2.3.1, table 2.3, says the size of the boolean is 1 byte and may only contain the values 0 or
1. In order to have the enum be 1-byte in size instead of the default int-sized, add the packed attribute to the enum.Hmmm... Could you point out us an official doc saying "packed" attribute does that?
This has been around since at least gcc 3.3, but this is link for a more current gcc[1] (search for "packed").
Quoted 2 lines
And I hope you tested that change because we use
grub_efi_boolean_ttype quite often...
I've been using this change for a couple months now with no issues. Also, no tests are failing because of this change, although this might not be a great recommendation. I want to do some more testing, just to be sure.
Quoted 21 lines
Quoted 11 lines
Signed-off-by Glenn Washburn <[email protected]>--- include/grub/efi/api.h | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)include/grub/efi/api.h +0
-0include/grub/efi/api.h enum grub_efi_reset_type · line 557
Quoted 7 lines
include/grub/efi/api.h +3
-1@@ -552,7 +552,13 @@ enum grub_efi_reset_typetypedef enum grub_efi_reset_type grub_efi_reset_type_t ;/* Types. */ -typedef char grub_efi_boolean_t ; +enum GRUB_PACKED grub_efi_boolean + { + GRUB_EFI_FALSE,I would explicitly do
Quoted 1 line
GRUB_EFI_FALSE= 0,
Good idea.
Quoted 5 lines
include/grub/efi/api.h enum grub_efi_reset_type · line 559
Quoted 3 lines
include/grub/efi/api.h +2
-0+ GRUB_EFI_TRUE + };I would move
GRUB_PACKEDhere to be inline with its other uses.
The following from the GCC manual has me questioning why we do this:
"You can also place [attributes] just past the closing curly brace of the definition, but this is less preferred because logically the type should be fully defined at the closing brace."[2] I'll change this in the next iteration for consistency, but perhaps we should reconsider the policy?
Glenn
[1] https://gcc.gnu.org/onlinedocs/gcc/Common-Type-Attributes.html
[2] https://gcc.gnu.org/onlinedocs/gcc/Type-Attributes.html
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel
Quoted 10 lines · Glenn Washburn
Quoted 7 lines · On Thu, 31 Aug 2023 20:05:37 +0200 Daniel Kiper <[email protected]> wrote:
Quoted 5 lines · Glenn Washburn
This allows using
GRUB_EFI_TRUEandGRUB_EFI_FALSEwith proper type and value checking. The UEFI 2.10 specification, in section 2.3.1, table 2.3, says the size of the boolean is 1 byte and may only contain the values 0 or
1. In order to have the enum be 1-byte in size instead of the default int-sized, add the packed attribute to the enum.
Quoted 5 lines
Quoted 2 lines
Hmmm... Could you point out us an official doc saying "packed" attribute does that?
This has been around since at least gcc 3.3, but this is link for a more current gcc[1] (search for "packed").
Does it work on clang? I, as Valdimir, am afraid we are entering into shakey grounds...
Quoted 36 lines
Quoted 2 lines
And I hope you tested that change because we use
grub_efi_boolean_ttype quite often...I've been using this change for a couple months now with no issues. Also, no tests are failing because of this change, although this might not be a great recommendation. I want to do some more testing, just to be sure.
Quoted 20 lines
Quoted 10 lines
Signed-off-by Glenn Washburn <[email protected]>--- include/grub/efi/api.h | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)include/grub/efi/api.h +0
-0include/grub/efi/api.h enum grub_efi_reset_type · line 557
Quoted 7 lines
include/grub/efi/api.h +3
-1@@ -552,7 +552,13 @@ enum grub_efi_reset_typetypedef enum grub_efi_reset_type grub_efi_reset_type_t ;/* Types. */ -typedef char grub_efi_boolean_t ; +enum GRUB_PACKED grub_efi_boolean + { + GRUB_EFI_FALSE,I would explicitly do
Quoted 1 line
GRUB_EFI_FALSE= 0,Good idea.
Quoted 4 lines
Quoted 3 lines
include/grub/efi/api.h +2
-0+ GRUB_EFI_TRUE + };
Quoted 8 lines
Quoted 1 line
I would move
GRUB_PACKEDhere to be inline with its other uses.The following from the GCC manual has me questioning why we do this:
"You can also place [attributes] just past the closing curly brace of the definition, but this is less preferred because logically the type should be fully defined at the closing brace."[2] I'll change this in the next iteration for consistency, but perhaps we should reconsider the policy?
I am OK with doing this cleanup but after the release...
Daniel
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel
Quoted 2 lines · On Thu, 31 Aug 2023 20:36:43 +0200 "Vladimir 'phcoder' Serbinenko"…
I see little value in using enum here and too much compiler variance about how it's handled.
Are you specifically talking about GCC, or the diversity of compilers that GRUB might be compiled with?
Quoted 1 line
Compiler is pretty much allowed to choose any type for enum.
If you're talking about GCC, then with the packed attribute this statement seems false[1].
Quoted 2 lines
If we go this route at all we should at very least do a compile time assert.
Sounds like a good idea.
Quoted 2 lines
Ideally I would keep it as-is. C unlike C++ does nothing to enforce enum.
Yes, this is unfortunate. I made this change because I think it looks nicer. I was thinking that the compiler might do obvious checking to fail if say a literal value outside the enum range is passed as a function argument that is an enum type. But you're right it doesn't do that. Using the enum type then obscures the fact that the boolean type should be 8 bits as specified by the spec. So typedef'ing as a char, as is currently done, seems to be a clearer implementation.
I was also hoping that this change would provide a little more detail for backtraces in GDB, but I've yet to check this.
Quoted 1 line
Feel free to put
GRUB_EFI_TRUEand FALSE into unnamed enum
Yes, I do like this instead of the macros. Now I'm questioning if its worth the change though.
[1] https://gcc.gnu.org/onlinedocs/gcc/Common-Type-Attributes.html
Quoted 37 lines
Le lun. 14 août 2023, 23:39, Glenn Washburn <[email protected]> a écrit :
Quoted 34 lines
This allows using
GRUB_EFI_TRUEandGRUB_EFI_FALSEwith proper type and value checking. The UEFI 2.10 specification, in section 2.3.1, table 2.3, says the size of the boolean is 1 byte and may only contain the values 0 or
1. In order to have the enum be 1-byte in size instead of the default int-sized, add the packed attribute to the enum.Signed-off-by Glenn Washburn <[email protected]>--- include/grub/efi/api.h | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)include/grub/efi/api.h +7
-2@@ -552,7 +552,13 @@ enum grub_efi_reset_typetypedef enum grub_efi_reset_type grub_efi_reset_type_t ;/* Types. */ -typedef char grub_efi_boolean_t ; +enum GRUB_PACKED grub_efi_boolean + { +GRUB_EFI_FALSE , + GRUB_EFI_TRUE + }; +typedef enum grub_efi_boolean grub_efi_boolean_t ; +#if GRUB_CPU_SIZEOF_VOID_P ==8 typedef grub_int64_t grub_efi_intn_t ;typedef grub_uint64_t grub_efi_uintn_t ; --2.34.1
Quoted 5 lines
Quoted 5 lines
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel
Quoted 11 lines · On Sat, 2 Sep 2023 20:45:34 +0200 Daniel Kiper <[email protected]> wrote:
Quoted 10 lines · Glenn Washburn
Quoted 7 lines · On Thu, 31 Aug 2023 20:05:37 +0200 Daniel Kiper <[email protected]> wrote:
Quoted 5 lines · Glenn Washburn
This allows using
GRUB_EFI_TRUEandGRUB_EFI_FALSEwith proper type and value checking. The UEFI 2.10 specification, in section 2.3.1, table 2.3, says the size of the boolean is 1 byte and may only contain the values 0 or
1. In order to have the enum be 1-byte in size instead of the default int-sized, add the packed attribute to the enum.
Quoted 8 lines
Quoted 5 lines
Quoted 2 lines
Hmmm... Could you point out us an official doc saying "packed" attribute does that?
This has been around since at least gcc 3.3, but this is link for a more current gcc[1] (search for "packed").
Does it work on clang? I, as Valdimir, am afraid we are entering into shakey grounds...
I didn't realize that clang was the concern. No I've not tested with clang and not setup to test it right now. So I'll table this for now.
Quoted 38 lines
Quoted 38 lines
Quoted 2 lines
And I hope you tested that change because we use
grub_efi_boolean_ttype quite often...I've been using this change for a couple months now with no issues. Also, no tests are failing because of this change, although this might not be a great recommendation. I want to do some more testing, just to be sure.
Quoted 20 lines
Quoted 10 lines
Signed-off-by Glenn Washburn <[email protected]>--- include/grub/efi/api.h | 8 +++++++- 1 file changed, 7 insertions(+), 1 deletion(-)include/grub/efi/api.h +0
-0include/grub/efi/api.h enum grub_efi_reset_type · line 557
Quoted 7 lines
include/grub/efi/api.h +3
-1@@ -552,7 +552,13 @@ enum grub_efi_reset_typetypedef enum grub_efi_reset_type grub_efi_reset_type_t ;/* Types. */ -typedef char grub_efi_boolean_t ; +enum GRUB_PACKED grub_efi_boolean + { + GRUB_EFI_FALSE,I would explicitly do
Quoted 1 line
GRUB_EFI_FALSE= 0,Good idea.
Quoted 5 lines
include/grub/efi/api.h enum grub_efi_reset_type · line 559
Quoted 3 lines
include/grub/efi/api.h +2
-0+ GRUB_EFI_TRUE + };I would move
GRUB_PACKEDhere to be inline with its other uses.
Quoted 8 lines
Quoted 6 lines
The following from the GCC manual has me questioning why we do this:
"You can also place [attributes] just past the closing curly brace of the definition, but this is less preferred because logically the type should be fully defined at the closing brace."[2] I'll change this in the next iteration for consistency, but perhaps we should reconsider the policy?I am OK with doing this cleanup but after the release...
Sounds good. I think its pretty low priority.
Glenn
_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel