The EFI spec states that the data size of an EFI variable is 64 bits.
"unsigned long", on the other hand, isn't on IA32.
diff --git a/drivers/firmware/efivars.c b/drivers/firmware/efivars.c
index bda5bce..488c24c 100644
--- a/drivers/firmware/efivars.c
+++ b/drivers/firmware/efivars.c
@@ -110,7 +110,7 @@ static LIST_HEAD(efivar_list);
struct efi_variable {
efi_char16_t VariableName[1024/sizeof(efi_char16_t)];
efi_guid_t VendorGuid;
- unsigned long DataSize;
+ __u64 DataSize;
__u8 Data[1024];
efi_status_t Status;
__u32 Attributes;
diff --git a/include/linux/efi.h b/include/linux/efi.h
index 9e97bc2..3f0a179 100644
--- a/include/linux/efi.h
+++ b/include/linux/efi.h
@@ -163,7 +163,7 @@ typedef efi_status_t efi_get_wakeup_time
efi_time_t *tm);
typedef efi_status_t efi_set_wakeup_time_t (efi_bool_t enabled, efi_time_t *tm);
typedef efi_status_t efi_get_variable_t (efi_char16_t *name, efi_guid_t *vendor, u32 *attr,
- unsigned long *data_size, void *data);
+ __u64 *data_size, void *data);
typedef efi_status_t efi_get_next_variable_t (unsigned long *name_size, efi_char16_t *name,
efi_guid_t *vendor);
typedef efi_status_t efi_set_variable_t (efi_char16_t *name, efi_guid_t *vendor,
--
Matthew Garrett | [email protected]
On Sun, Mar 19, 2006 at 06:43:25PM +0000, Matthew Garrett wrote:
> The EFI spec states that the data size of an EFI variable is 64 bits.
> "unsigned long", on the other hand, isn't on IA32.
>
> diff --git a/drivers/firmware/efivars.c b/drivers/firmware/efivars.c
> index bda5bce..488c24c 100644
> --- a/drivers/firmware/efivars.c
> +++ b/drivers/firmware/efivars.c
> @@ -110,7 +110,7 @@ static LIST_HEAD(efivar_list);
> struct efi_variable {
> efi_char16_t VariableName[1024/sizeof(efi_char16_t)];
> efi_guid_t VendorGuid;
> - unsigned long DataSize;
> + __u64 DataSize;
> __u8 Data[1024];
> efi_status_t Status;
> __u32 Attributes;
> diff --git a/include/linux/efi.h b/include/linux/efi.h
> index 9e97bc2..3f0a179 100644
> --- a/include/linux/efi.h
> +++ b/include/linux/efi.h
> @@ -163,7 +163,7 @@ typedef efi_status_t efi_get_wakeup_time
> efi_time_t *tm);
> typedef efi_status_t efi_set_wakeup_time_t (efi_bool_t enabled, efi_time_t *tm);
> typedef efi_status_t efi_get_variable_t (efi_char16_t *name, efi_guid_t *vendor, u32 *attr,
> - unsigned long *data_size, void *data);
> + __u64 *data_size, void *data);
> typedef efi_status_t efi_get_next_variable_t (unsigned long *name_size, efi_char16_t *name,
> efi_guid_t *vendor);
> typedef efi_status_t efi_set_variable_t (efi_char16_t *name, efi_guid_t *vendor,
NAK. efibootmgr, the main userspace consumer of this struct, also
thinks this is an "unsigned long". This wasn't specified up through
EFI 1.10, which I complained about (it, and efi_status_t being defined
as an unsigned long) both force userspace consumers of this struct to
match size of the kernel (i.e. compile a 32-bit binary for 32-bit kernel,
and compile a 64-bit binary for 64-bit kernel).
-Matt
--
Matt Domsch
Software Architect
Dell Linux Solutions linux.dell.com & http://www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com
On Sun, Mar 19, 2006 at 03:29:01PM -0600, Matt Domsch wrote:
> NAK. efibootmgr, the main userspace consumer of this struct, also
> thinks this is an "unsigned long".
Hm. My copy of efibootmgr has:
typedef struct _efi_variable_t {
efi_char16_t VariableName[1024/sizeof(efi_char16_t)];
efi_guid_t VendorGuid;
uint64_t DataSize;
uint8_t Data[1024];
efi_status_t Status;
uint32_t Attributes;
} __attribute__((packed)) efi_variable_t;
which certainly makes it look like it's expecting a 64-bit value. But
checking the spec does seem to suggest that datasize is a native value,
so presumably it's an efibootmgr bug rather than a kernel one? In that
case, this ought to be dropped.
--
Matthew Garrett | [email protected]
On Sun, Mar 19, 2006 at 09:33:01PM +0000, Matthew Garrett wrote:
> On Sun, Mar 19, 2006 at 03:29:01PM -0600, Matt Domsch wrote:
>
> > NAK. efibootmgr, the main userspace consumer of this struct, also
> > thinks this is an "unsigned long".
>
> Hm. My copy of efibootmgr has:
>
> typedef struct _efi_variable_t {
> efi_char16_t VariableName[1024/sizeof(efi_char16_t)];
> efi_guid_t VendorGuid;
> uint64_t DataSize;
> uint8_t Data[1024];
> efi_status_t Status;
> uint32_t Attributes;
> } __attribute__((packed)) efi_variable_t;
>
> which certainly makes it look like it's expecting a 64-bit value. But
> checking the spec does seem to suggest that datasize is a native value,
> so presumably it's an efibootmgr bug rather than a kernel one? In that
> case, this ought to be dropped.
Sorry, I did mean to post a link to the latest efibootmgr that has
this fixed:
Home page: http://linux.dell.com/efibootmgr/
For distros that care about 32-bit EFI, please upgrade to efibootmgr 0.5.3.
http://linux.dell.com/efibootmgr/efibootmgr-0.5.3.tar.gz
http://linux.dell.com/efibootmgr/efibootmgr-0.5.3.tar.gz.sign
http://linux.dell.com/efibootmgr/ChangeLog
Thanks,
Matt
--
Matt Domsch
Software Architect
Dell Linux Solutions linux.dell.com & http://www.dell.com/linux
Linux on Dell mailing lists @ http://lists.us.dell.com
On Sun, Mar 19, 2006 at 04:49:40PM -0600, Matt Domsch wrote:
> Sorry, I did mean to post a link to the latest efibootmgr that has
> this fixed:
Thanks - the copy I had was rather older than I thought. Sorry about the
confusion.
--
Matthew Garrett | [email protected]
On Ne 19-03-06 18:43:25, Matthew Garrett wrote:
> The EFI spec states that the data size of an EFI variable is 64 bits.
> "unsigned long", on the other hand, isn't on IA32.
>
> diff --git a/drivers/firmware/efivars.c b/drivers/firmware/efivars.c
> index bda5bce..488c24c 100644
> --- a/drivers/firmware/efivars.c
> +++ b/drivers/firmware/efivars.c
> @@ -110,7 +110,7 @@ static LIST_HEAD(efivar_list);
> struct efi_variable {
> efi_char16_t VariableName[1024/sizeof(efi_char16_t)];
> efi_guid_t VendorGuid;
> - unsigned long DataSize;
> + __u64 DataSize;
> __u8 Data[1024];
> efi_status_t Status;
> __u32 Attributes;
Please, use u64 (not __u64).
Pavel
--
86: {