Andrew Cooper: > On 13/04/18 19:25, Simon Gaiser wrote: >> Jan Beulich: >>> NMI and #MC can occur at any time after S3 resume, yet the MSR_SPEC_CTRL >>> may become available only once we're reloaded microcode. Make >>> SPEC_CTRL_ENTRY_FROM_INTR_IST and DO_SPEC_CTRL_EXIT_TO_XEN no-ops for >>> the critical period of time. >>> >>> Also set the MSR back to its intended value. >>> >>> Signed-off-by: Jan Beulich >>> >>> --- a/xen/arch/x86/acpi/power.c >>> +++ b/xen/arch/x86/acpi/power.c >>> @@ -28,6 +28,7 @@ >>> #include >>> #include >>> #include >>> +#include >>> #include >>> >>> uint32_t system_reset_counter = 1; >>> @@ -163,6 +164,7 @@ static int enter_state(u32 state) >>> { >>> unsigned long flags; >>> int error; >>> + struct cpu_info *ci; >>> unsigned long cr4; >>> >>> if ( (state <= ACPI_STATE_S0) || (state > ACPI_S_STATES_MAX) ) >>> @@ -210,6 +212,10 @@ static int enter_state(u32 state) >>> else >>> error = 0; >>> >>> + ci = get_cpu_info(); >>> + ci->use_shadow_spec_ctrl = 0; >>> + ci->bti_ist_info = 0; >>> + >>> ACPI_FLUSH_CPU_CACHE(); >>> >>> switch ( state ) >>> @@ -248,6 +254,11 @@ static int enter_state(u32 state) >>> >>> microcode_resume_cpu(0); >>> >>> + ci->bti_ist_info = default_bti_ist_info; >>> + asm volatile (ALTERNATIVE("", "wrmsr", X86_FEATURE_XEN_IBRS_SET) >> This does not compile for me: >> >> power.c: Assembler messages: >> power.c:272: Error: value of 257 too large for field of 1 bytes at 0 >> >> Changing the alternative based on the other "wrmsr" calls fixes it: >> >> asm volatile (ALTERNATIVE(ASM_NOP3, "wrmsr", X86_FEATURE_XEN_IBRS_SET) > > Ah - you're presumably back-porting this to 4.8? > > Jan's code is correct for staging, and your version here is correct for > all currently-released versions of Xen. (I've done quite a lot of > playing with alternatives generation for 4.11.) Yeah, sorry, I should have checked if it works with staging.