Re: [PATCH v4] x86/power: Fix 'nosmt' vs. hibernation triple fault during resume

[Jiri Kosina]

On Fri, 31 May 2019, Josh Poimboeuf wrote:

> Something like this (not yet tested)?  Maybe we could also remove the
> resume_play_dead() hack?

I tried to test this, but the resumed kernel doesn't seem to be healthy 
for reason I don't understand yet. 
Symptoms I've seen so far -- 'dazed and confused NMI', spontaneous reboot, 
userspace segfault.

> diff --git a/arch/x86/include/asm/fixmap.h b/arch/x86/include/asm/fixmap.h
> index 9da8cccdf3fb..1c328624162c 100644
> --- a/arch/x86/include/asm/fixmap.h
> +++ b/arch/x86/include/asm/fixmap.h
> @@ -80,6 +80,7 @@ enum fixed_addresses {
>  #ifdef CONFIG_X86_VSYSCALL_EMULATION
>  	VSYSCALL_PAGE = (FIXADDR_TOP - VSYSCALL_ADDR) >> PAGE_SHIFT,
>  #endif
> +	FIX_MWAIT = (FIXADDR_TOP - VSYSCALL_ADDR - 1) >> PAGE_SHIFT,

Two things to this:

- you don't seem to fix x86_32
- shouldn't it rather be FIXADDR_TOP - VSYSCALL_ADDR + 1 instead?

Thanks,

-- 
Jiri Kosina
SUSE Labs