Check table version before calling UEFI 2.0+ functions

[nicholasbishop]

The create_event_ex function (in boot services) and query_variable_info (in runtime services) were added to their respective tables in UEFI 2.0. In earlier versions these functions can't be called, so check the table revision before using these function pointers.

[namazso]

What platform is this relevant for? EFI 1.10 (the last EFI) only supports ia32 and ia64. ia64 is not supported by Rust or LLVM, and I don't know of a single 32-bit x86 device that uses EFI 1.10 or older. Also the crate is called uefi and not efi, so I'd assume requiring UEFI is reasonable.

[nicholasbishop]

What platform is this relevant for? EFI 1.10 (the last EFI) only supports ia32 and ia64. ia64 is not supported by Rust or LLVM, and I don't know of a single 32-bit x86 device that uses EFI 1.10 or older. Also the crate is called uefi and not efi, so I'd assume requiring UEFI is reasonable.

Unfortunately some x86_64 machines use the ia32 mode of UEFI (for example some Intel Compute Sticks). The specific machines I had in mind with this change though are old (but not that old) Apple machines, which were stuck on EFI 1.10 for a long time. See torvalds/linux@f5390cd for example.

[namazso]

Unfortunately some x86_64 machines use the ia32 mode of UEFI (for example some Intel Compute Sticks).

I knew about those, but they're all UEFI already, not stuck on EFI

The specific machines I had in mind with this change though are old Apple machines, which were stuck on EFI 1.10 for a long time. See torvalds/linux@f5390cd for example.

I see. The mentioned Linux patch disables using 2.0 features on Apple hardware even when they're reported as supported. Should this be replicated here too then? As from what I understand without this the single platform this patch targets still won't work.

[nicholasbishop]

Oops, I somehow edited your comment instead of replying, sorry about that. I believe I have restored your comment to its original form :)

---

The mentioned Linux patch disables using 2.0 features on Apple hardware even when they're reported as supported. Should this be replicated here too then? As from what I understand without this the single platform this patch targets still won't work.

For older Macs that report 1.10 this change will work fine, but you are right that for the MacBookPro16,1 described here it will still be broken. I don't think we have a way to check for system vendor here currently though.

[GabrielMajeri]

Everything looks fine now 👍