Re: [PATCH v6 01/19] kernel: Standardize vdso_datapage

[Huw Davies]

On Mon, Jun 10, 2019 at 12:07:45PM +0100, Vincenzo Frascino wrote:
> On 10/06/2019 11:31, Huw Davies wrote:
> > On Mon, Jun 10, 2019 at 11:17:48AM +0100, Vincenzo Frascino wrote:
> >> On 10/06/2019 10:27, Huw Davies wrote:
> >>> On Thu, May 30, 2019 at 03:15:13PM +0100, Vincenzo Frascino wrote:
> >>>> --- /dev/null
> >>>> +++ b/include/vdso/datapage.h
> >>>> @@ -0,0 +1,91 @@
> >>>> +/* SPDX-License-Identifier: GPL-2.0 */
> >>>> +#ifndef __VDSO_DATAPAGE_H
> >>>> +#define __VDSO_DATAPAGE_H
> >>>> +
> >>>> +#ifdef __KERNEL__
> >>>> +
> >>>> +#ifndef __ASSEMBLY__
> >>>> +
> >>>> +#include <linux/bits.h>
> >>>> +#include <linux/time.h>
> >>>> +#include <linux/types.h>
> >>>> +
> >>>> +#define VDSO_BASES	(CLOCK_TAI + 1)
> >>>> +#define VDSO_HRES	(BIT(CLOCK_REALTIME)		| \
> >>>> +			 BIT(CLOCK_MONOTONIC)		| \
> >>>> +			 BIT(CLOCK_BOOTTIME)		| \
> >>>> +			 BIT(CLOCK_TAI))
> >>>> +#define VDSO_COARSE	(BIT(CLOCK_REALTIME_COARSE)	| \
> >>>> +			 BIT(CLOCK_MONOTONIC_COARSE))
> >>>> +#define VDSO_RAW	(BIT(CLOCK_MONOTONIC_RAW))
> >>>> +
> >>>> +#define CS_HRES_COARSE	0
> >>>> +#define CS_RAW		1
> >>>
> >>> CS_HRES_COARSE seems like a confusing name choice to me.  What you
> >>> really mean is not RAW.
> >>>
> >>> How about CS_ADJ to indicate that its updated by adjtime?
> >>> CS_XTIME might be another option.
> >>>
> >>
> >> I divided the timers in 3 sets (HRES, COARSE, RAW), CS_HRES_COARSE refers to the
> >> first two and CS_RAW to the third. I will ad a comment to explain the logic in
> >> the next iteration.
> > 
> > I'm thinking ahead about a possible CLOCK_MONOTONIC_RAW_COARSE (which
> > would be useful at least for Wine).  In that case you'd have four clock
> > types non-raw and raw, each with either hres or coarse.
> > 
> 
> Thanks for this, I was not aware of CLOCK_MONOTONIC_RAW_COARSE.
> I tried to find, though, some details, but I could not find any. Could you
> please provide some reference?

It doesn't exist yet ;-)  However it doesn't seem crazy that such a
clock should exist.  I was really using it to illustrate that
raw / non-raw is orthogonal to hres / coarse.

That being said, this really doesn't matter that much.

Huw.