From: Jaegeuk Kim <jaegeuk@kernel.org>
To: Chao Yu <chao@kernel.org>
Cc: linux-f2fs-devel@lists.sourceforge.net,
linux-kernel@vger.kernel.org, Yangtao Li <frank.li@vivo.com>
Subject: Re: [f2fs-dev] [PATCH v2] f2fs: add support for counting time of submit discard cmd
Date: Mon, 12 Dec 2022 14:47:15 -0800 [thread overview]
Message-ID: <Y5evc3D8eZlOjTjI@google.com> (raw)
In-Reply-To: <b76428f8-fd9e-7c4e-52ea-0eeb1ea443af@kernel.org>
On 12/12, Chao Yu wrote:
> On 2022/12/12 20:51, Yangtao Li wrote:
> > This patch adds support for counting the average time and
> > peak time of submit discard command, and we can see its
> > value in debugfs.
> >
> > It is not sure whether the block layer has recorded these
> > data, and these data are allowed to be accessed by fs,
> > or they are only exported to user space.
> >
> > On the one hand, I added these data to better understand
> > the current device operating status, and to further control
> > the discard process in a more detailed manner based on the
> > discard submit time in the future.
>
> Again, w'd better to consider this functionality only when DEBUG_FS is
> enabled.
BTW, why can't we use iostat to get the discard latencies?
>
> >
> > Signed-off-by: Yangtao Li <frank.li@vivo.com>
> > ---
> > fs/f2fs/debug.c | 10 +++++++---
> > fs/f2fs/f2fs.h | 6 ++++++
> > fs/f2fs/segment.c | 21 +++++++++++++++++++--
> > 3 files changed, 32 insertions(+), 5 deletions(-)
> >
> > diff --git a/fs/f2fs/debug.c b/fs/f2fs/debug.c
> > index 32af4f0c5735..142c256b89d9 100644
> > --- a/fs/f2fs/debug.c
> > +++ b/fs/f2fs/debug.c
> > @@ -120,6 +120,10 @@ static void update_general_status(struct f2fs_sb_info *sbi)
> > llist_empty(&SM_I(sbi)->fcc_info->issue_list);
> > }
> > if (SM_I(sbi)->dcc_info) {
> > + struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
> > +
> > + si->discard_avg = dcc->discard_time_avg;
> > + si->discard_peak = dcc->discard_time_peak;
> > si->nr_discarded =
> > atomic_read(&SM_I(sbi)->dcc_info->issued_discard);
> > si->nr_discarding =
> > @@ -545,9 +549,9 @@ static int stat_show(struct seq_file *s, void *v)
> > si->nr_wb_cp_data, si->nr_wb_data,
> > si->nr_flushing, si->nr_flushed,
> > si->flush_list_empty);
> > - seq_printf(s, "Discard: (%4d %4d)) cmd: %4d undiscard:%4u\n",
> > - si->nr_discarding, si->nr_discarded,
> > - si->nr_discard_cmd, si->undiscard_blks);
> > + seq_printf(s, "Discard: (%4d %4d, avg:%4lldns, peak:%4lldns)) cmd: %4d undiscard:%4u\n",
> > + si->nr_discarding, si->nr_discarded, ktime_to_us(si->discard_avg),
> > + ktime_to_us(si->discard_peak), si->nr_discard_cmd, si->undiscard_blks);
> > seq_printf(s, " - atomic IO: %4d (Max. %4d)\n",
> > si->aw_cnt, si->max_aw_cnt);
> > seq_printf(s, " - compress: %4d, hit:%8d\n", si->compress_pages, si->compress_page_hit);
> > diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> > index e8953c3dc81a..2cd55cb981ff 100644
> > --- a/fs/f2fs/f2fs.h
> > +++ b/fs/f2fs/f2fs.h
> > @@ -371,6 +371,8 @@ struct discard_cmd {
> > int error; /* bio error */
> > spinlock_t lock; /* for state/bio_ref updating */
> > unsigned short bio_ref; /* bio reference count */
> > + struct discard_cmd_control *dcc; /* global discard cmd control */
> > + ktime_t submit_start; /* submit start time */
> > };
> > enum {
> > @@ -415,6 +417,9 @@ struct discard_cmd_control {
> > unsigned int max_ordered_discard; /* maximum discard granularity issued by lba order */
> > unsigned int undiscard_blks; /* # of undiscard blocks */
> > unsigned int next_pos; /* next discard position */
> > + spinlock_t discard_time_lock; /* for discard time statistics */
> > + ktime_t discard_time_avg; /* issued discard cmd avg time */
> > + ktime_t discard_time_peak; /* issued discard cmd peak time */
> > atomic_t issued_discard; /* # of issued discard */
> > atomic_t queued_discard; /* # of queued discard */
> > atomic_t discard_cmd_cnt; /* # of cached cmd count */
> > @@ -3896,6 +3901,7 @@ struct f2fs_stat_info {
> > int nr_dio_read, nr_dio_write;
> > unsigned int io_skip_bggc, other_skip_bggc;
> > int nr_flushing, nr_flushed, flush_list_empty;
> > + ktime_t discard_avg, discard_peak;
> > int nr_discarding, nr_discarded;
> > int nr_discard_cmd;
> > unsigned int undiscard_blks;
> > diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> > index a9099a754dd2..73cd05bb3f4a 100644
> > --- a/fs/f2fs/segment.c
> > +++ b/fs/f2fs/segment.c
> > @@ -937,6 +937,7 @@ static struct discard_cmd *__create_discard_cmd(struct f2fs_sb_info *sbi,
> > list_add_tail(&dc->list, pend_list);
> > spin_lock_init(&dc->lock);
> > dc->bio_ref = 0;
> > + dc->dcc = dcc;
> > atomic_inc(&dcc->discard_cmd_cnt);
> > dcc->undiscard_blks += len;
> > @@ -1006,9 +1007,13 @@ static void __remove_discard_cmd(struct f2fs_sb_info *sbi,
> > static void f2fs_submit_discard_endio(struct bio *bio)
> > {
> > struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;
> > + struct discard_cmd_control *dcc = dc->dcc;
> > unsigned long flags;
> > + ktime_t submit_time;
> > + int nr_discarded;
> > spin_lock_irqsave(&dc->lock, flags);
> > + submit_time = ktime_sub(ktime_get(), dc->submit_start);
> > if (!dc->error)
> > dc->error = blk_status_to_errno(bio->bi_status);
> > dc->bio_ref--;
> > @@ -1018,6 +1023,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)
> > }
> > spin_unlock_irqrestore(&dc->lock, flags);
> > bio_put(bio);
> > +
> > + spin_lock_irqsave(&dcc->discard_time_lock, flags);
> > + nr_discarded = atomic_read(&dcc->issued_discard);
> > + dcc->discard_time_avg = div_u64(ktime_add(nr_discarded * dcc->discard_time_avg,
> > + submit_time),
> > + nr_discarded + 1);
> > + if (dcc->discard_time_avg > dcc->discard_time_peak)
> > + dcc->discard_time_peak = dcc->discard_time_avg;
> > + atomic_inc(&dcc->issued_discard);
> > + spin_unlock_irqrestore(&dcc->discard_time_lock, flags);
>
> Why not calculating average time only in update_general_status()? and here,
> we just need to account total_{discard_time, discard_count} w/o additional
> spinlock.
>
> Thanks,
>
> > }
> > static void __check_sit_bitmap(struct f2fs_sb_info *sbi,
> > @@ -1166,6 +1181,7 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,
> > * right away
> > */
> > spin_lock_irqsave(&dc->lock, flags);
> > + dc->submit_start = ktime_get();
> > if (last)
> > dc->state = D_SUBMIT;
> > else
> > @@ -1185,8 +1201,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,
> > bio->bi_opf |= flag;
> > submit_bio(bio);
> > - atomic_inc(&dcc->issued_discard);
> > -
> > f2fs_update_iostat(sbi, NULL, FS_DISCARD, len * F2FS_BLKSIZE);
> > lstart += len;
> > @@ -2079,9 +2093,12 @@ static int create_discard_cmd_control(struct f2fs_sb_info *sbi)
> > INIT_LIST_HEAD(&dcc->wait_list);
> > INIT_LIST_HEAD(&dcc->fstrim_list);
> > mutex_init(&dcc->cmd_lock);
> > + spin_lock_init(&dcc->discard_time_lock);
> > atomic_set(&dcc->issued_discard, 0);
> > atomic_set(&dcc->queued_discard, 0);
> > atomic_set(&dcc->discard_cmd_cnt, 0);
> > + dcc->discard_time_avg = 0;
> > + dcc->discard_time_peak = 0;
> > dcc->nr_discards = 0;
> > dcc->max_discards = MAIN_SEGS(sbi) << sbi->log_blocks_per_seg;
> > dcc->max_discard_request = DEF_MAX_DISCARD_REQUEST;
_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
WARNING: multiple messages have this Message-ID (diff)
From: Jaegeuk Kim <jaegeuk@kernel.org>
To: Chao Yu <chao@kernel.org>
Cc: Yangtao Li <frank.li@vivo.com>,
linux-f2fs-devel@lists.sourceforge.net,
linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] f2fs: add support for counting time of submit discard cmd
Date: Mon, 12 Dec 2022 14:47:15 -0800 [thread overview]
Message-ID: <Y5evc3D8eZlOjTjI@google.com> (raw)
In-Reply-To: <b76428f8-fd9e-7c4e-52ea-0eeb1ea443af@kernel.org>
On 12/12, Chao Yu wrote:
> On 2022/12/12 20:51, Yangtao Li wrote:
> > This patch adds support for counting the average time and
> > peak time of submit discard command, and we can see its
> > value in debugfs.
> >
> > It is not sure whether the block layer has recorded these
> > data, and these data are allowed to be accessed by fs,
> > or they are only exported to user space.
> >
> > On the one hand, I added these data to better understand
> > the current device operating status, and to further control
> > the discard process in a more detailed manner based on the
> > discard submit time in the future.
>
> Again, w'd better to consider this functionality only when DEBUG_FS is
> enabled.
BTW, why can't we use iostat to get the discard latencies?
>
> >
> > Signed-off-by: Yangtao Li <frank.li@vivo.com>
> > ---
> > fs/f2fs/debug.c | 10 +++++++---
> > fs/f2fs/f2fs.h | 6 ++++++
> > fs/f2fs/segment.c | 21 +++++++++++++++++++--
> > 3 files changed, 32 insertions(+), 5 deletions(-)
> >
> > diff --git a/fs/f2fs/debug.c b/fs/f2fs/debug.c
> > index 32af4f0c5735..142c256b89d9 100644
> > --- a/fs/f2fs/debug.c
> > +++ b/fs/f2fs/debug.c
> > @@ -120,6 +120,10 @@ static void update_general_status(struct f2fs_sb_info *sbi)
> > llist_empty(&SM_I(sbi)->fcc_info->issue_list);
> > }
> > if (SM_I(sbi)->dcc_info) {
> > + struct discard_cmd_control *dcc = SM_I(sbi)->dcc_info;
> > +
> > + si->discard_avg = dcc->discard_time_avg;
> > + si->discard_peak = dcc->discard_time_peak;
> > si->nr_discarded =
> > atomic_read(&SM_I(sbi)->dcc_info->issued_discard);
> > si->nr_discarding =
> > @@ -545,9 +549,9 @@ static int stat_show(struct seq_file *s, void *v)
> > si->nr_wb_cp_data, si->nr_wb_data,
> > si->nr_flushing, si->nr_flushed,
> > si->flush_list_empty);
> > - seq_printf(s, "Discard: (%4d %4d)) cmd: %4d undiscard:%4u\n",
> > - si->nr_discarding, si->nr_discarded,
> > - si->nr_discard_cmd, si->undiscard_blks);
> > + seq_printf(s, "Discard: (%4d %4d, avg:%4lldns, peak:%4lldns)) cmd: %4d undiscard:%4u\n",
> > + si->nr_discarding, si->nr_discarded, ktime_to_us(si->discard_avg),
> > + ktime_to_us(si->discard_peak), si->nr_discard_cmd, si->undiscard_blks);
> > seq_printf(s, " - atomic IO: %4d (Max. %4d)\n",
> > si->aw_cnt, si->max_aw_cnt);
> > seq_printf(s, " - compress: %4d, hit:%8d\n", si->compress_pages, si->compress_page_hit);
> > diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> > index e8953c3dc81a..2cd55cb981ff 100644
> > --- a/fs/f2fs/f2fs.h
> > +++ b/fs/f2fs/f2fs.h
> > @@ -371,6 +371,8 @@ struct discard_cmd {
> > int error; /* bio error */
> > spinlock_t lock; /* for state/bio_ref updating */
> > unsigned short bio_ref; /* bio reference count */
> > + struct discard_cmd_control *dcc; /* global discard cmd control */
> > + ktime_t submit_start; /* submit start time */
> > };
> > enum {
> > @@ -415,6 +417,9 @@ struct discard_cmd_control {
> > unsigned int max_ordered_discard; /* maximum discard granularity issued by lba order */
> > unsigned int undiscard_blks; /* # of undiscard blocks */
> > unsigned int next_pos; /* next discard position */
> > + spinlock_t discard_time_lock; /* for discard time statistics */
> > + ktime_t discard_time_avg; /* issued discard cmd avg time */
> > + ktime_t discard_time_peak; /* issued discard cmd peak time */
> > atomic_t issued_discard; /* # of issued discard */
> > atomic_t queued_discard; /* # of queued discard */
> > atomic_t discard_cmd_cnt; /* # of cached cmd count */
> > @@ -3896,6 +3901,7 @@ struct f2fs_stat_info {
> > int nr_dio_read, nr_dio_write;
> > unsigned int io_skip_bggc, other_skip_bggc;
> > int nr_flushing, nr_flushed, flush_list_empty;
> > + ktime_t discard_avg, discard_peak;
> > int nr_discarding, nr_discarded;
> > int nr_discard_cmd;
> > unsigned int undiscard_blks;
> > diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> > index a9099a754dd2..73cd05bb3f4a 100644
> > --- a/fs/f2fs/segment.c
> > +++ b/fs/f2fs/segment.c
> > @@ -937,6 +937,7 @@ static struct discard_cmd *__create_discard_cmd(struct f2fs_sb_info *sbi,
> > list_add_tail(&dc->list, pend_list);
> > spin_lock_init(&dc->lock);
> > dc->bio_ref = 0;
> > + dc->dcc = dcc;
> > atomic_inc(&dcc->discard_cmd_cnt);
> > dcc->undiscard_blks += len;
> > @@ -1006,9 +1007,13 @@ static void __remove_discard_cmd(struct f2fs_sb_info *sbi,
> > static void f2fs_submit_discard_endio(struct bio *bio)
> > {
> > struct discard_cmd *dc = (struct discard_cmd *)bio->bi_private;
> > + struct discard_cmd_control *dcc = dc->dcc;
> > unsigned long flags;
> > + ktime_t submit_time;
> > + int nr_discarded;
> > spin_lock_irqsave(&dc->lock, flags);
> > + submit_time = ktime_sub(ktime_get(), dc->submit_start);
> > if (!dc->error)
> > dc->error = blk_status_to_errno(bio->bi_status);
> > dc->bio_ref--;
> > @@ -1018,6 +1023,16 @@ static void f2fs_submit_discard_endio(struct bio *bio)
> > }
> > spin_unlock_irqrestore(&dc->lock, flags);
> > bio_put(bio);
> > +
> > + spin_lock_irqsave(&dcc->discard_time_lock, flags);
> > + nr_discarded = atomic_read(&dcc->issued_discard);
> > + dcc->discard_time_avg = div_u64(ktime_add(nr_discarded * dcc->discard_time_avg,
> > + submit_time),
> > + nr_discarded + 1);
> > + if (dcc->discard_time_avg > dcc->discard_time_peak)
> > + dcc->discard_time_peak = dcc->discard_time_avg;
> > + atomic_inc(&dcc->issued_discard);
> > + spin_unlock_irqrestore(&dcc->discard_time_lock, flags);
>
> Why not calculating average time only in update_general_status()? and here,
> we just need to account total_{discard_time, discard_count} w/o additional
> spinlock.
>
> Thanks,
>
> > }
> > static void __check_sit_bitmap(struct f2fs_sb_info *sbi,
> > @@ -1166,6 +1181,7 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,
> > * right away
> > */
> > spin_lock_irqsave(&dc->lock, flags);
> > + dc->submit_start = ktime_get();
> > if (last)
> > dc->state = D_SUBMIT;
> > else
> > @@ -1185,8 +1201,6 @@ static int __submit_discard_cmd(struct f2fs_sb_info *sbi,
> > bio->bi_opf |= flag;
> > submit_bio(bio);
> > - atomic_inc(&dcc->issued_discard);
> > -
> > f2fs_update_iostat(sbi, NULL, FS_DISCARD, len * F2FS_BLKSIZE);
> > lstart += len;
> > @@ -2079,9 +2093,12 @@ static int create_discard_cmd_control(struct f2fs_sb_info *sbi)
> > INIT_LIST_HEAD(&dcc->wait_list);
> > INIT_LIST_HEAD(&dcc->fstrim_list);
> > mutex_init(&dcc->cmd_lock);
> > + spin_lock_init(&dcc->discard_time_lock);
> > atomic_set(&dcc->issued_discard, 0);
> > atomic_set(&dcc->queued_discard, 0);
> > atomic_set(&dcc->discard_cmd_cnt, 0);
> > + dcc->discard_time_avg = 0;
> > + dcc->discard_time_peak = 0;
> > dcc->nr_discards = 0;
> > dcc->max_discards = MAIN_SEGS(sbi) << sbi->log_blocks_per_seg;
> > dcc->max_discard_request = DEF_MAX_DISCARD_REQUEST;
next prev parent reply other threads:[~2022-12-12 22:47 UTC|newest]
Thread overview: 20+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-12-12 12:51 [f2fs-dev] [PATCH v2] f2fs: add support for counting time of submit discard cmd Yangtao Li via Linux-f2fs-devel
2022-12-12 12:51 ` Yangtao Li
2022-12-12 13:40 ` [f2fs-dev] " Chao Yu
2022-12-12 13:40 ` Chao Yu
2022-12-12 22:47 ` Jaegeuk Kim [this message]
2022-12-12 22:47 ` Jaegeuk Kim
2022-12-13 1:22 ` [f2fs-dev] " Chao Yu
2022-12-13 1:22 ` Chao Yu
2022-12-13 12:21 ` [f2fs-dev] " Yangtao Li via Linux-f2fs-devel
2022-12-13 12:21 ` Yangtao Li
2022-12-13 19:17 ` [f2fs-dev] " Jaegeuk Kim
2022-12-13 19:17 ` Jaegeuk Kim
2022-12-14 7:28 ` [f2fs-dev] " Yangtao Li via Linux-f2fs-devel
2022-12-14 7:28 ` Yangtao Li
2022-12-14 21:52 ` [f2fs-dev] " Jaegeuk Kim
2022-12-14 21:52 ` Jaegeuk Kim
2022-12-16 1:04 ` [f2fs-dev] " Chao Yu
2022-12-16 1:04 ` Chao Yu
2022-12-16 2:49 ` [f2fs-dev] " Yangtao Li via Linux-f2fs-devel
2022-12-16 2:49 ` Yangtao Li
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=Y5evc3D8eZlOjTjI@google.com \
--to=jaegeuk@kernel.org \
--cc=chao@kernel.org \
--cc=frank.li@vivo.com \
--cc=linux-f2fs-devel@lists.sourceforge.net \
--cc=linux-kernel@vger.kernel.org \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.