Hi Leif, It isn't required for correctness? i.e. To be consistent with other places? i.e. To be like TranslationRegimeIsDual in UefiCpuPkg/Library/ArmMmuLib/AArch64/ArmMmuLibCore.c? https://github.com/tianocore/edk2/blob/2cfd432ac92f58a14331fec2d2b885c795f684db/UefiCpuPkg/Library/ArmMmuLib/AArch64/ArmMmuLibCore.c#L42
Something like this: ARM: TT_XN_MASK vs. TT_UXN_MASK | TT_PXN_MASK for E2H by jaykrell · Pull Request #12775 · tianocore/edk2<https://github.com/tianocore/edk2/pull/12775> ? * Jay ________________________________ From: Leif Lindholm <[email protected]> Sent: Friday, July 3, 2026 8:14 AM To: [email protected] <[email protected]>; Jay Krell <[email protected]> Subject: [EXTERNAL] Re: [edk2-devel] ArmReadCurrentEL () == AARCH64_EL2 Hi Jay, On Tue, 30 Jun 2026 at 03:48, Jay Krell via groups.io <[email protected]> wrote: > I have a multi-part question. > > 1. ArmPkg/Drivers/CpuDxe/AArch64/Mmu.c(378): > if (ArmReadCurrentEL () == AARCH64_EL2) { > > This should also check E2H, right? Umm, why? Best Regards, Leif > 2. More broadly: > What about all other EL checks, primarily in EL1_OR_EL2 macro. > Mainly with mrs use, but also tlbi. > > 2a Leave them generally alone, let them trap, etc. and let hardware or > hypervisor virtualize (i.e. if E2H; they are already correct in preexisting > contexts). > 2b Change the macro to also check E2H, leaving the name asis? > 2c Make a new macro and gradually/rapidly convert many/all uses? > EL1LIKE_OR_EL2? > EL1orE2H_OR_EL2? > 2d Other? > > If decision is to have it check also E2H, people building physical-only (or > E2H==0 only) might object to the code size growth? So have the definition of > EL1_OR_EL2 itself guided by another ifdef? > > e.g. > #ifdef SUPPORT_E2H > #defi ne EL1_OR_EL2 ... new way > #else > #define EL1_OR_EL2 ... old way > #endif > > And then, there *might* be cases that need old way unconditionally? > #define EL1_OR_EL2_IGNORE_E2H ? > > Even if EL1_OR_EL2 is left alone to be trapped/emulated, maybe the multiple > (2?) places in C that check EL==2 should share a "helper" function? > ArmCurrentELBehavior() == 1 or 2? > ArmEL1OrE2H()? > Or even, redefine ArmReadCurrentEL to return AARCH64_EL1 for E2H? > Maybe changing ArmReadCurrentEL is too heavy handed and confusing. > You'll need some other function to read the "real" EL, ignoring E2H. > > Anyway, surely there are many options, but the main question is to do > something at all, or leave it mostly asis. > > Thank you, > - Jay > > > > -=-=-=-=-=-=-=-=-=-=-=- Groups.io Links: You receive all messages sent to this group. View/Reply Online (#122034): https://edk2.groups.io/g/devel/message/122034 Mute This Topic: https://groups.io/mt/120041798/21656 Group Owner: [email protected] Unsubscribe: https://edk2.groups.io/g/devel/unsub [[email protected]] -=-=-=-=-=-=-=-=-=-=-=-
