Re: [PATCH v3 02/22] x86,mce: Delete ist_begin_non_atomic()

[Peter Zijlstra]

On Wed, Feb 19, 2020 at 02:12:13PM -0800, Andy Lutomirski wrote:
> On Wed, Feb 19, 2020 at 9:34 AM Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote:

> > Do you really want to create code that unwinds enough of nmi_enter() to
> > get you to a preemptible context? *shudder*
> 
> Well, there's another way to approach this:
> 
> void notrace nonothing do_machine_check(struct pt_regs *regs)
> {
>   if (user_mode(regs))
>     do_sane_machine_check(regs);
>   else
>     do_awful_machine_check(regs);
> }
> 
> void do_sane_machine_check(regs)
> {
>   nothing special here.  just a regular exception, more or less.
> }
> 
> void do_awful_macine_check(regs)
> {
>   basically an NMI.  No funny business, no recovery possible.
> task_work_add() not allowed.
> }

Right, that looks like major surgery to the current code though; I'd
much prefer someone that knows that code do that.

> I suppose the general consideration I'm trying to get at is: is
> task_work_add() actually useful at all here?  For the case when a
> kernel thread does memcpy_mcsafe() or similar, task work registered
> using task_work_add() will never run.

task_work isn't at all useful when we didn't come from userspace. In
that case irq_work is the best option, but that doesn't provide a
preemptible context.