[PATCH 5/5] kasan: suppress recursive reports for HW_TAGS

[Andrey Konovalov]

On Mon, Mar 13, 2023 at 12:20 PM Marco Elver <elver at google.com> wrote:
>
> > + * Hardware Tag-Based KASAN instead relies on:
> > + * For #1: Resetting tags via kasan_reset_tag().
> > + * For #2: Supression of tag checks via CPU, see report_suppress_start/end().
>
> Typo: "Suppression"

Will fix in v2.

> > +static void report_suppress_start(void)
> > +{
> > +#ifdef CONFIG_KASAN_HW_TAGS
> > +       /*
> > +        * Disable migration for the duration of printing a KASAN report, as
> > +        * hw_suppress_tag_checks_start() disables checks on the current CPU.
> > +        */
> > +       migrate_disable();
>
> This still allows this task to be preempted by another task. If the
> other task is scheduled in right after hw_suppress_tag_checks_start()
> then there won't be any tag checking in that task. If HW-tags KASAN is
> used as a mitigation technique, that may unnecessarily weaken KASAN,
> because right after report_suppress_start(), it does
> spin_lock_irqsave() which disables interrupts (and thereby preemption)
> anyway.
>
> Why not just use preempt_disable()?

Ah, yes, I intended to do that but forgot to make the change.

I'll wait for comments from arm64 maintainers on the other patches and
then send v2 with a fix.

Thank you, Marco!