| Message ID | 20150202232803.GB34575@jaegeuk-mac02.mot.com (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
Hi Jaegeuk, IMHO, it looks better user could decide the size of trim considering latency of trim. Otherwise, additional checkpoints user doesn't want will occur. Regards, Changman On Mon, Feb 02, 2015 at 03:29:25PM -0800, Jaegeuk Kim wrote: > Change long 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 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@kernel.org> > --- > fs/f2fs/f2fs.h | 2 ++ > fs/f2fs/segment.c | 16 +++++++++++----- > 2 files changed, 13 insertions(+), 5 deletions(-) > > diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h > index 8231a59..ec5e66f 100644 > --- a/fs/f2fs/f2fs.h > +++ b/fs/f2fs/f2fs.h > @@ -105,6 +105,8 @@ enum { > CP_DISCARD, > }; > > +#define BATCHED_TRIM_SEGMENTS(sbi) (((sbi)->segs_per_sec) << 5) > + > struct cp_control { > int reason; > __u64 trim_start; > diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c > index 5ea57ec..b85bb97 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, > + 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; > -- > 2.1.1 > > > ------------------------------------------------------------------------------ > Dive into the World of Parallel Programming. The Go Parallel Website, > sponsored by Intel and developed in partnership with Slashdot Media, is your > hub for all things parallel software development, from weekly thought > leadership blogs to news, videos, case studies, tutorials and more. Take a > look and join the conversation now. http://goparallel.sourceforge.net/ > _______________________________________________ > Linux-f2fs-devel mailing list > Linux-f2fs-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h index 8231a59..ec5e66f 100644 --- a/fs/f2fs/f2fs.h +++ b/fs/f2fs/f2fs.h @@ -105,6 +105,8 @@ enum { CP_DISCARD, }; +#define BATCHED_TRIM_SEGMENTS(sbi) (((sbi)->segs_per_sec) << 5) + struct cp_control { int reason; __u64 trim_start; diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c index 5ea57ec..b85bb97 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, + 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;
Change long 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 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@kernel.org> --- fs/f2fs/f2fs.h | 2 ++ fs/f2fs/segment.c | 16 +++++++++++----- 2 files changed, 13 insertions(+), 5 deletions(-)