Howdy, On macOS versions/SDKs before 26 you are correct, we had an invalid enum value for HV_SYS_REG_MDCR_EL2 in the API, but the hv_vcpu_get/set_sys_reg APIs work with this value even though the encoding is incorrect. The enum value has been remedied in 26+ (0xe089 as you pointed out) and has been done in a backwards compatible way internally in the framework.
> On Aug 18, 2025, at 8:46 AM, Philippe Mathieu-Daudé <phi...@linaro.org> wrote: > > Cc'ing Apple folks hoping they can forward or directly help :) > > On 18/8/25 15:21, Richard Henderson wrote: >> On 8/18/25 22:41, Philippe Mathieu-Daudé wrote: >>> Rebasing Mohamed's work I'm getting: >>> >>> In file included from ../../target/arm/hvf/hvf.c:413: >>> ../../target/arm/hvf/sysreg.c.inc:156:1: error: static assertion failed due >>> to requirement '!(HV_SYS_REG_MDCR_EL2 != (((((1 << 28) | (19 << 16) | ((3) >>> << 14) | ((4) << 11) | ((1) << 7) | ((1) << 3) | ((1) << 0)) | >>> 13510798882111488ULL)) & 65535))': not expecting: HV_SYS_REG_MDCR_EL2 != >>> KVMID_TO_HVF(KVMID_AA64_SYS_REG64(1, 1, 3, 4, 1)) >>> 156 | DEF_SYSREG(HV_SYS_REG_MDCR_EL2, 1, 1, 3, 4, 1) >>> | ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ >>> HV_SYS_REG_MDCR_EL2 = 0xe019, <---- >> How odd. It should be e089, if it were encoded like all the others. >> I wonder if this is a manual typo that's now baked into the api, or if it's >> an OS bug. >> r~ >
