On Tue 08-10-24 22:14:16, Tang Yizhou wrote:
> On Tue, Oct 8, 2024 at 12:23 AM Jan Kara <jack@xxxxxxx> wrote:
> >
> > On Sun 06-10-24 20:41:11, Tang Yizhou wrote:
> > > On Thu, Oct 3, 2024 at 9:01 PM Jan Kara <jack@xxxxxxx> wrote:
> > > >
> > > > On Wed 02-10-24 21:00:02, Tang Yizhou wrote:
> > > > > From: Tang Yizhou <yizhou.tang@xxxxxxxxxx>
> > > > >
> > > > > The name of the BANDWIDTH_INTERVAL macro is misleading, as it is not
> > > > > only used in the bandwidth update functions wb_update_bandwidth() and
> > > > > __wb_update_bandwidth(), but also in the dirty limit update function
> > > > > domain_update_dirty_limit().
> > > > >
> > > > > Rename BANDWIDTH_INTERVAL to UPDATE_INTERVAL to make things clear.
> > > > >
> > > > > This patche doesn't introduce any behavioral changes.
> > > > >
> > > > > Signed-off-by: Tang Yizhou <yizhou.tang@xxxxxxxxxx>
> > > >
> > > > Umm, I agree BANDWIDTH_INTERVAL may be confusing but UPDATE_INTERVAL does
> > > > not seem much better to be honest. I actually have hard time coming up with
> > > > a more descriptive name so what if we settled on updating the comment only
> > > > instead of renaming to something not much better?
> > > >
> > > > Honza
> > >
> > > Thank you for your review. I agree that UPDATE_INTERVAL is not a good
> > > name. How about
> > > renaming it to BW_DIRTYLIMIT_INTERVAL?
> >
> > Maybe WB_STAT_INTERVAL? Because it is interval in which we maintain
> > statistics about writeback behavior.
> >
>
> I don't think this is a good name, as it suggests a relation to enum
> wb_stat_item, but bandwidth and dirty limit are not in wb_stat_item.
OK, so how about keeping BANDWIDTH_INTERVAL as is and adding
DIRTY_LIMIT_INTERVAL with the same value? There's nothing which would
strictly tie them to the same value.
Honza
> > > > > ---
> > > > > mm/page-writeback.c | 16 ++++++++--------
> > > > > 1 file changed, 8 insertions(+), 8 deletions(-)
> > > > >
> > > > > diff --git a/mm/page-writeback.c b/mm/page-writeback.c
> > > > > index fcd4c1439cb9..a848e7f0719d 100644
> > > > > --- a/mm/page-writeback.c
> > > > > +++ b/mm/page-writeback.c
> > > > > @@ -54,9 +54,9 @@
> > > > > #define DIRTY_POLL_THRESH (128 >> (PAGE_SHIFT - 10))
> > > > >
> > > > > /*
> > > > > - * Estimate write bandwidth at 200ms intervals.
> > > > > + * Estimate write bandwidth or update dirty limit at 200ms intervals.
> > > > > */
> > > > > -#define BANDWIDTH_INTERVAL max(HZ/5, 1)
> > > > > +#define UPDATE_INTERVAL max(HZ/5, 1)
> > > > >
> > > > > #define RATELIMIT_CALC_SHIFT 10
> > > > >
> > > > > @@ -1331,11 +1331,11 @@ static void domain_update_dirty_limit(struct dirty_throttle_control *dtc,
> > > > > /*
> > > > > * check locklessly first to optimize away locking for the most time
> > > > > */
> > > > > - if (time_before(now, dom->dirty_limit_tstamp + BANDWIDTH_INTERVAL))
> > > > > + if (time_before(now, dom->dirty_limit_tstamp + UPDATE_INTERVAL))
> > > > > return;
> > > > >
> > > > > spin_lock(&dom->lock);
> > > > > - if (time_after_eq(now, dom->dirty_limit_tstamp + BANDWIDTH_INTERVAL)) {
> > > > > + if (time_after_eq(now, dom->dirty_limit_tstamp + UPDATE_INTERVAL)) {
> > > > > update_dirty_limit(dtc);
> > > > > dom->dirty_limit_tstamp = now;
> > > > > }
> > > > > @@ -1928,7 +1928,7 @@ static int balance_dirty_pages(struct bdi_writeback *wb,
> > > > > wb->dirty_exceeded = gdtc->dirty_exceeded ||
> > > > > (mdtc && mdtc->dirty_exceeded);
> > > > > if (time_is_before_jiffies(READ_ONCE(wb->bw_time_stamp) +
> > > > > - BANDWIDTH_INTERVAL))
> > > > > + UPDATE_INTERVAL))
> > > > > __wb_update_bandwidth(gdtc, mdtc, true);
> > > > >
> > > > > /* throttle according to the chosen dtc */
> > > > > @@ -2705,7 +2705,7 @@ int do_writepages(struct address_space *mapping, struct writeback_control *wbc)
> > > > > * writeback bandwidth is updated once in a while.
> > > > > */
> > > > > if (time_is_before_jiffies(READ_ONCE(wb->bw_time_stamp) +
> > > > > - BANDWIDTH_INTERVAL))
> > > > > + UPDATE_INTERVAL))
> > > > > wb_update_bandwidth(wb);
> > > > > return ret;
> > > > > }
> > > > > @@ -3057,14 +3057,14 @@ static void wb_inode_writeback_end(struct bdi_writeback *wb)
> > > > > atomic_dec(&wb->writeback_inodes);
> > > > > /*
> > > > > * Make sure estimate of writeback throughput gets updated after
> > > > > - * writeback completed. We delay the update by BANDWIDTH_INTERVAL
> > > > > + * writeback completed. We delay the update by UPDATE_INTERVAL
> > > > > * (which is the interval other bandwidth updates use for batching) so
> > > > > * that if multiple inodes end writeback at a similar time, they get
> > > > > * batched into one bandwidth update.
> > > > > */
> > > > > spin_lock_irqsave(&wb->work_lock, flags);
> > > > > if (test_bit(WB_registered, &wb->state))
> > > > > - queue_delayed_work(bdi_wq, &wb->bw_dwork, BANDWIDTH_INTERVAL);
> > > > > + queue_delayed_work(bdi_wq, &wb->bw_dwork, UPDATE_INTERVAL);
> > > > > spin_unlock_irqrestore(&wb->work_lock, flags);
> > > > > }
> > > > >
> > > > > --
> > > > > 2.25.1
> > > > >
> > > > >
> > > > --
> > > > Jan Kara <jack@xxxxxxxx>
> > > > SUSE Labs, CR
> > --
> > Jan Kara <jack@xxxxxxxx>
> > SUSE Labs, CR
--
Jan Kara <jack@xxxxxxxx>
SUSE Labs, CR
