RE: [f2fs-dev] [PATCH 5/5 v3] f2fs: introduce a batched trim

From: Chao Yu
Date: Thu Feb 05 2015 - 04:31:46 EST

Hi Jaegeuk, Changman,

> -----Original Message-----
> From: Jaegeuk Kim [mailto:jaegeuk@xxxxxxxxxx]
> Sent: Wednesday, February 04, 2015 4:11 AM
> To: Changman Lee
> Cc: Chao Yu; linux-fsdevel@xxxxxxxxxxxxxxx; linux-kernel@xxxxxxxxxxxxxxx;
> linux-f2fs-devel@xxxxxxxxxxxxxxxxxxxxx
> Subject: Re: [f2fs-dev] [PATCH 5/5 v3] f2fs: introduce a batched trim
>
> Hi Changman,
>
> Good idea!
>
> Change log from v2:
> o add sysfs to change the # of sections for trimming.

Good idea!

>
> Change log from v1:
> o add description
> o change the # of batched segments suggested by Chao
> o make consistent for # of batched segments
>
> This patch introduces a batched trimming feature, which submits split discard
> commands.
>
> This is to avoid long latency due to huge trim commands.
> If fstrim was triggered ranging from 0 to the end of device, we should lock
> all the checkpoint-related mutexes, resulting in very long latency.
>
> Signed-off-by: Jaegeuk Kim <jaegeuk@xxxxxxxxxx>
> ---
> Documentation/ABI/testing/sysfs-fs-f2fs | 6 ++++++
> Documentation/filesystems/f2fs.txt | 4 ++++
> fs/f2fs/f2fs.h | 7 +++++++
> fs/f2fs/segment.c | 18 +++++++++++++-----
> fs/f2fs/super.c | 2 ++
> 5 files changed, 32 insertions(+), 5 deletions(-)
>
> diff --git a/Documentation/ABI/testing/sysfs-fs-f2fs
> b/Documentation/ABI/testing/sysfs-fs-f2fs
> index 6f9157f..2c4cc42 100644
> --- a/Documentation/ABI/testing/sysfs-fs-f2fs
> +++ b/Documentation/ABI/testing/sysfs-fs-f2fs
> @@ -74,3 +74,9 @@ Date: March 2014
> Contact: "Jaegeuk Kim" <jaegeuk.kim@xxxxxxxxxxx>
> Description:
> Controls the memory footprint used by f2fs.
> +
> +What: /sys/fs/f2fs/<disk>/trim_sections
> +Date: February 2015
> +Contact: "Jaegeuk Kim" <jaegeuk@xxxxxxxxxx>
> +Description:
> + Controls the trimming rate in batch mode.
> diff --git a/Documentation/filesystems/f2fs.txt b/Documentation/filesystems/f2fs.txt
> index 6758aa3..dac11d7 100644
> --- a/Documentation/filesystems/f2fs.txt
> +++ b/Documentation/filesystems/f2fs.txt
> @@ -199,6 +199,10 @@ Files in /sys/fs/f2fs/<devname>
> checkpoint is triggered, and issued during the
> checkpoint. By default, it is disabled with 0.
>
> + trim_sections This parameter controls the number of sections
> + to be trimmed out in batch mode when FITRIM
> + conducts. 32 sections is set by default.
> +
> ipu_policy This parameter controls the policy of in-place
> updates in f2fs. There are five policies:
> 0x01: F2FS_IPU_FORCE, 0x02: F2FS_IPU_SSR,
> diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> index 964c240..6f57da1 100644
> --- a/fs/f2fs/f2fs.h
> +++ b/fs/f2fs/f2fs.h
> @@ -105,6 +105,10 @@ enum {
> CP_DISCARD,
> };
>
> +#define DEF_BATCHED_TRIM_SECTIONS 32
> +#define BATCHED_TRIM_SEGMENTS(sbi) \
> + (SM_I(sbi)->trim_sections * (sbi)->segs_per_sec)
> +
> struct cp_control {
> int reason;
> __u64 trim_start;
> @@ -448,6 +452,9 @@ struct f2fs_sm_info {
> int nr_discards; /* # of discards in the list */
> int max_discards; /* max. discards to be issued */
>
> + /* for batched trimming */
> + int trim_sections; /* # of sections to trim */
> +
> struct list_head sit_entry_set; /* sit entry set list */
>
> unsigned int ipu_policy; /* in-place-update policy */
> diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> index 5ea57ec..c542f63 100644
> --- a/fs/f2fs/segment.c
> +++ b/fs/f2fs/segment.c
> @@ -1066,14 +1066,20 @@ int f2fs_trim_fs(struct f2fs_sb_info *sbi, struct fstrim_range *range)
> end_segno = (end >= MAX_BLKADDR(sbi)) ? MAIN_SEGS(sbi) - 1 :
> GET_SEGNO(sbi, end);
> cpc.reason = CP_DISCARD;
> - cpc.trim_start = start_segno;
> - cpc.trim_end = end_segno;
> cpc.trim_minlen = range->minlen >> sbi->log_blocksize;
>
> /* do checkpoint to issue discard commands safely */
> - mutex_lock(&sbi->gc_mutex);
> - write_checkpoint(sbi, &cpc);
> - mutex_unlock(&sbi->gc_mutex);
> + for (; start_segno <= end_segno;
> + start_segno += BATCHED_TRIM_SEGMENTS(sbi)) {
> + cpc.trim_start = start_segno;
> + cpc.trim_end = min_t(unsigned int,
> + start_segno + BATCHED_TRIM_SEGMENTS (sbi) - 1,

Actually what I mean is that in each trim we try to align start segment
to first segment of section as much as possible.

How about using:
cpc.trim_end = min_t(unsigned int, rounddown(start_segno +
BATCHED_TRIM_SEGMENTS(sbi),
(sbi)->segs_per_sec) - 1, end_segno);
Thanks,

> + end_segno);
> +
> + mutex_lock(&sbi->gc_mutex);
> + write_checkpoint(sbi, &cpc);
> + mutex_unlock(&sbi->gc_mutex);
> + }
> out:
> range->len = cpc.trimmed << sbi->log_blocksize;
> return 0;
> @@ -2127,6 +2133,8 @@ int build_segment_manager(struct f2fs_sb_info *sbi)
> sm_info->nr_discards = 0;
> sm_info->max_discards = 0;
>
> + sm_info->trim_sections = DEF_BATCHED_TRIM_SECTIONS;
> +
> INIT_LIST_HEAD(&sm_info->sit_entry_set);
>
> if (test_opt(sbi, FLUSH_MERGE) && !f2fs_readonly(sbi->sb)) {
> diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> index 1e92c2e..f2fe666 100644
> --- a/fs/f2fs/super.c
> +++ b/fs/f2fs/super.c
> @@ -195,6 +195,7 @@ F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_no_gc_sleep_time,
> no_gc_sleep_time);
> F2FS_RW_ATTR(GC_THREAD, f2fs_gc_kthread, gc_idle, gc_idle);
> F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, reclaim_segments, rec_prefree_segments);
> F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, max_small_discards, max_discards);
> +F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, batched_trim_sections, trim_sections);
> F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, ipu_policy, ipu_policy);
> F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_ipu_util, min_ipu_util);
> F2FS_RW_ATTR(SM_INFO, f2fs_sm_info, min_fsync_blocks, min_fsync_blocks);
> @@ -210,6 +211,7 @@ static struct attribute *f2fs_attrs[] = {
> ATTR_LIST(gc_idle),
> ATTR_LIST(reclaim_segments),
> ATTR_LIST(max_small_discards),
> + ATTR_LIST(batched_trim_sections),
> ATTR_LIST(ipu_policy),
> ATTR_LIST(min_ipu_util),
> ATTR_LIST(min_fsync_blocks),
> --
> 2.1.1

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/