From: "Darrick J. Wong" <darrick.wong@oracle.com>
To: Eric Sandeen <sandeen@sandeen.net>
Cc: linux-xfs@vger.kernel.org
Subject: Re: [PATCH 05/14] xfs: refactor quota expiration timer modification
Date: Wed, 12 Feb 2020 21:33:18 -0800 [thread overview]
Message-ID: <20200213053318.GR6874@magnolia> (raw)
In-Reply-To: <e8d6702c-9e39-6e5e-e910-739b9b4558a3@sandeen.net>
On Wed, Feb 12, 2020 at 09:32:33PM -0600, Eric Sandeen wrote:
>
>
> On 2/12/20 9:27 PM, Eric Sandeen wrote:
> > On 2/12/20 7:46 PM, Darrick J. Wong wrote:
> >> On Wed, Feb 12, 2020 at 05:57:24PM -0600, Eric Sandeen wrote:
> >>> On 12/31/19 7:11 PM, Darrick J. Wong wrote:
> >>>> From: Darrick J. Wong <darrick.wong@oracle.com>
> >
> > ...
> >
> >>>> } else {
> >>>> if (!over) {
> >>>> - d->d_btimer = 0;
> >>>> + xfs_quota_clear_timer(&d->d_btimer);
> >>>
> >>> yeah that's a very fancy way to say "= 0" ;)
> >>>
> >>> ...
> >>
> >> Yes, that's a fancy way to assign zero. However, consider that for
> >> bigtime support, I had to add an incore quota timer so that timers more
> >> or less fire when they're supposed to, and now there's a function to
> >> convert the incore timespec64 value to whatever is ondisk:
> >>
> >> /* Clear a quota grace period expiration timer. */
> >> static inline void
> >> xfs_quota_clear_timer(
> >> struct xfs_disk_dquot *ddq,
> >> time64_t *itimer,
> >> __be32 *dtimer)
> >> {
> >> struct timespec64 tv = { 0 };
> >>
> >> *itimer = tv.tv_sec;
> >> xfs_dquot_to_disk_timestamp(ddq, dtimer, &tv);
> >> }
> >>
> >> It was at *that* point in the patchset that it seemed easier to call a
> >> small function three times than to open-code this three times.
> >
> > +void
> > +xfs_dquot_to_disk_timestamp(
> > + __be32 *dtimer,
> > + const struct timespec64 *tv)
> > +{
> > + *dtimer = cpu_to_be32(tv->tv_sec);
> > +}
> >
> > static inline void
> > xfs_quota_clear_timer(
> > + time64_t *itimer,
> > __be32 *dtimer)
> > {
> > - *dtimer = cpu_to_be32(0);
> > + struct timespec64 tv = { 0 };
> > +
> > + *itimer = tv.tv_sec;
> > + xfs_dquot_to_disk_timestamp(dtimer, &tv);
> > }
> >
> > xfs_quota_clear_timer(&dqp->q_btimer, &d->d_btimer);
> >
> > That's still a very fancy way of saying:
> >
> > dqp->q_btimer = d->d_btimer = 0;
> >
> > I think? Can't really see what value the timespec64 adds here.
> >
> > -Eric
> >
>
> Actually,
>
> xfs_quota_set_timer(
> + time64_t *itimer,
> __be32 *dtimer,
> time_t limit)
> {
> - time64_t new_timeout;
> + struct timespec64 tv = { 0 };
>
> - new_timeout = xfs_dquot_clamp_timer(get_seconds() + limit);
> - *dtimer = cpu_to_be32(new_timeout);
> + tv.tv_sec = xfs_dquot_clamp_timer(ktime_get_real_seconds() + limit);
> + *itimer = tv.tv_sec;
> + xfs_dquot_to_disk_timestamp(dtimer, &tv);
> }
>
> I'm not sure why there's a timespec64 here either. Isn't everything
> we're dealing with on timers in seconds, using only tv_sec, and time64_t would
> suffice instead of using a timespec64 just to carry around a seconds value?
Yes, the grace periods recorded in the timer fields in dquot 0 are
intervals measured in seconds.
However, for dquot != 0, the timer fields store the time of the
expiration, so I settled on timespec64 as the incore structure so that
XFS consistently uses struct timespec64 to represent specific points in
time.
(That and time64_t doesn't exist in userspace.)
--D
> -Eric
next prev parent reply other threads:[~2020-02-13 5:33 UTC|newest]
Thread overview: 50+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-01-01 1:11 [PATCH 00/14] xfs: widen timestamps to deal with y2038 Darrick J. Wong
2020-01-01 1:11 ` [PATCH 01/14] xfs: explicitly define inode timestamp range Darrick J. Wong
2020-02-12 23:00 ` Eric Sandeen
2020-02-13 1:26 ` Darrick J. Wong
2020-02-13 1:50 ` Eric Sandeen
2020-02-13 1:53 ` Darrick J. Wong
2020-01-01 1:11 ` [PATCH 02/14] xfs: preserve default grace interval during quotacheck Darrick J. Wong
2020-02-12 23:35 ` Eric Sandeen
2020-02-19 4:55 ` Eric Sandeen
2020-03-03 3:03 ` Eric Sandeen
2020-03-03 15:48 ` Darrick J. Wong
2020-03-03 15:52 ` Eric Sandeen
2020-01-01 1:11 ` [PATCH 03/14] xfs: refactor quota exceeded test Darrick J. Wong
2020-02-12 23:51 ` Eric Sandeen
2020-02-13 1:41 ` Darrick J. Wong
2020-02-13 1:52 ` Eric Sandeen
2020-02-13 1:59 ` Darrick J. Wong
2020-05-31 14:04 ` Amir Goldstein
2020-01-01 1:11 ` [PATCH 04/14] xfs: fix quota timer inactivation Darrick J. Wong
2020-05-31 15:04 ` Amir Goldstein
2020-06-01 23:56 ` Darrick J. Wong
2020-01-01 1:11 ` [PATCH 05/14] xfs: refactor quota expiration timer modification Darrick J. Wong
2020-02-12 23:57 ` Eric Sandeen
2020-02-13 1:46 ` Darrick J. Wong
2020-02-13 3:27 ` Eric Sandeen
2020-02-13 3:32 ` Eric Sandeen
2020-02-13 5:33 ` Darrick J. Wong [this message]
2020-01-01 1:11 ` [PATCH 06/14] xfs: refactor default quota grace period setting code Darrick J. Wong
2020-02-13 0:15 ` Eric Sandeen
2020-02-13 1:53 ` Darrick J. Wong
2020-02-13 2:03 ` Darrick J. Wong
2020-01-01 1:11 ` [PATCH 07/14] xfs: remove xfs_timestamp_t Darrick J. Wong
2020-01-01 1:11 ` [PATCH 08/14] xfs: move xfs_log_dinode_to_disk to the log code Darrick J. Wong
2020-01-01 1:11 ` [PATCH 09/14] xfs: refactor timestamp coding Darrick J. Wong
2020-01-01 1:12 ` [PATCH 10/14] xfs: convert struct xfs_timestamp to union Darrick J. Wong
2020-01-01 1:12 ` [PATCH 11/14] xfs: widen ondisk timestamps to deal with y2038 problem Darrick J. Wong
2020-05-31 12:30 ` Amir Goldstein
2020-06-01 23:17 ` Darrick J. Wong
2020-06-02 4:26 ` Amir Goldstein
2020-01-01 1:12 ` [PATCH 12/14] xfs: cache quota grace period expiration times incore Darrick J. Wong
2020-01-01 1:12 ` [PATCH 13/14] xfs: enable bigtime for quota timers Darrick J. Wong
2020-05-31 17:07 ` Amir Goldstein
2020-06-02 0:09 ` Darrick J. Wong
2020-06-02 4:04 ` Amir Goldstein
2020-01-01 1:12 ` [PATCH 14/14] xfs: enable big timestamps Darrick J. Wong
2020-05-26 9:20 ` [PATCH 00/14] xfs: widen timestamps to deal with y2038 Amir Goldstein
2020-05-26 15:57 ` Darrick J. Wong
2020-05-26 16:42 ` Amir Goldstein
2020-05-31 17:31 ` Amir Goldstein
2020-06-02 0:09 ` Darrick J. Wong
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20200213053318.GR6874@magnolia \
--to=darrick.wong@oracle.com \
--cc=linux-xfs@vger.kernel.org \
--cc=sandeen@sandeen.net \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).