Re: [PATCHv7 10/16] x86/tdx: Convert shared memory back to private on kexec

[Kirill A. Shutemov]

On Sun, Feb 25, 2024 at 04:58:46PM +0200, Kirill A. Shutemov wrote:
> On Fri, Feb 23, 2024 at 11:39:07AM -0800, Dave Hansen wrote:
> > On 2/12/24 02:44, Kirill A. Shutemov wrote:
> > > +static void tdx_kexec_stop_conversion(bool crash)
> > > +{
> > > +	/* Stop new private<->shared conversions */
> > > +	conversion_allowed = false;
> > > +
> > > +	/*
> > > +	 * Make sure conversion_allowed is cleared before checking
> > > +	 * conversions_in_progress.
> > > +	 */
> > > +	barrier();
> > > +
> > > +	/*
> > > +	 * Crash kernel reaches here with interrupts disabled: can't wait for
> > > +	 * conversions to finish.
> > > +	 *
> > > +	 * If race happened, just report and proceed.
> > > +	 */
> > > +	if (!crash) {
> > > +		unsigned long timeout;
> > > +
> > > +		/*
> > > +		 * Wait for in-flight conversions to complete.
> > > +		 *
> > > +		 * Do not wait more than 30 seconds.
> > > +		 */
> > > +		timeout = 30 * USEC_PER_SEC;
> > > +		while (atomic_read(&conversions_in_progress) && timeout--)
> > > +			udelay(1);
> > > +	}
> > > +
> > > +	if (atomic_read(&conversions_in_progress))
> > > +		pr_warn("Failed to finish shared<->private conversions\n");
> > > +}
> > 
> > I'd really prefer we find a way to do this with actual locks, especially
> > 'conversion_allowed'.
> > 
> > This is _awfully_ close to being able to be handled by a rwsem where the
> > readers are the converters and tdx_kexec_stop_conversion() takes a write.
> 
> Okay, here's what I come up with. It needs more testing.

I don't see a problem during testing.

#include <linux/delay.h> has to be dropped, but otherwise the patch is
fine to me.

Any feedback?

-- 
  Kiryl Shutsemau / Kirill A. Shutemov