Re: [Xen-devel] [PATCH v2 4/6] x86: track when in #NMI context

[Roger Pau Monné <roger.pau@xxxxxxxxxx>]

On Tue, Feb 18, 2020 at 10:40:02AM +0000, Andrew Cooper wrote:
> On 17/02/2020 18:43, Roger Pau Monne wrote:
> > Add helpers to track when running in #MC context. This is modeled
> > after the in_irq helpers, but does not support reentry.
> >
> > Note that there are no users of in_mc() introduced by the change,
> > further users will be added by followup changes.
> >
> > Signed-off-by: Roger Pau Monné <roger.pau@xxxxxxxxxx>
> 
> You probably mean s/mc/nmi/ throughout the commit message, but I'm
> afraid these are rather problematic.

Er, yes, sorry, c&p from the previous commit and I failed to adjust
it.

> 
> NMIs can be recursively entered, especially on corner cases in the crash
> path.  Asserting that the crash path is not recursive can lead to never
> entering the crash kernel.

Is this specific to how Xen handles #NMI?

Intel SDM states that #NMI is not reentrant, as further #NMIs are
blocked until the execution of the iret instruction:

"While an NMI interrupt handler is executing, the processor blocks
delivery of subsequent NMIs until the next execu- tion of the IRET
instruction. This blocking of NMIs prevents nested execution of the
NMI handler. It is recommended that the NMI interrupt handler be
accessed through an interrupt gate to disable maskable hardware
interrupts (see Section 6.8.1, “Masking Maskable Hardware
Interrupts”)."

AFAICT there's no iret in do_nmi until it has finished execution.

Roger.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/mailman/listinfo/xen-devel