Re: [PATCH v1 05/12] x86/mce: Allow instrumentation during task work queueing

[Peter Zijlstra <peterz@infradead.org>]

On Wed, Dec 08, 2021 at 12:13:36PM +0100, Borislav Petkov wrote:
> From: Borislav Petkov <bp@suse.de>
> 
> Fixes
> 
>   vmlinux.o: warning: objtool: do_machine_check()+0xdb1: call to queue_task_work() leaves .noinstr.text section
> 
> Signed-off-by: Borislav Petkov <bp@suse.de>
> ---
>  arch/x86/kernel/cpu/mce/core.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/arch/x86/kernel/cpu/mce/core.c b/arch/x86/kernel/cpu/mce/core.c
> index f61f14faa532..d7fa444d6282 100644
> --- a/arch/x86/kernel/cpu/mce/core.c
> +++ b/arch/x86/kernel/cpu/mce/core.c
> @@ -1451,6 +1451,14 @@ noinstr void do_machine_check(struct pt_regs *regs)
>  	if (worst != MCE_AR_SEVERITY && !kill_current_task)
>  		goto out;
>  
> +	/*
> +	 * Enable instrumentation around the external facilities like
> +	 * task_work_add() (via queue_task_work()), fixup_exception() etc.
> +	 * For now, that is. Fixing this properly would need a lot more involved
> +	 * reorganization.

The only reason we want to fix that is to limit the amount of code MCE
runs, or is there additional concerns, like perhaps getting #DB traps
from !noinstr code?

> +	 */
> +	instrumentation_begin();
> +
>  	/* Fault was in user mode and we need to take some action */
>  	if ((m.cs & 3) == 3) {
>  		/* If this triggers there is no way to recover. Die hard. */
> @@ -1479,6 +1487,9 @@ noinstr void do_machine_check(struct pt_regs *regs)
>  		if (m.kflags & MCE_IN_KERNEL_COPYIN)
>  			queue_task_work(&m, msg, kill_me_never);
>  	}
> +
> +	instrumentation_end();
> +
>  out:
>  	mce_wrmsrl(MSR_IA32_MCG_STATUS, 0);
>  }
> -- 
> 2.29.2
>