Hi Jaegeuk,
> -----Original Message-----
> From: Jaegeuk Kim [mailto:jaeg...@kernel.org]
> Sent: Friday, January 30, 2015 5:41 AM
> To: Chao Yu
> Cc: linux-ker...@vger.kernel.org; linux-fsde...@vger.kernel.org;
> linux-f2fs-devel@lists.sourceforge.net
> Subject: Re: [PATCH 5/5] f2fs: introduce a batched trim
>
> Hi Chao,
>
> On Thu, Jan 29, 2015 at 08:38:30PM +0800, Chao Yu wrote:
> > Hi Jaegeuk,
> >
> > > -----Original Message-----
> > > From: linux-fsdevel-ow...@vger.kernel.org
> > > [mailto:linux-fsdevel-ow...@vger.kernel.org] On
> > > Behalf Of Jaegeuk Kim
> > > Sent: Wednesday, January 28, 2015 7:32 AM
> > > To: linux-ker...@vger.kernel.org; linux-fsde...@vger.kernel.org;
> > > linux-f2fs-devel@lists.sourceforge.net
> > > Cc: Jaegeuk Kim
> > > Subject: [PATCH 5/5] f2fs: introduce a batched trim
> > >
> > > This patch introduces a batched trimming feature, which submits split
> > > discard
> > > commands.
> >
> > I didn't get it, why we should split discard commands. :(
> >
> > Does smaller discarding for flash shows better performance or effect or
> > safety?
> > Can you please explain more about this patch?
>
> 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.
Ah.. thanks for your explanation, how about adding above description into commit
log?
>
> Thanks,
>
> >
> > Thanks,
> >
> > >
> > > Signed-off-by: Jaegeuk Kim <jaeg...@kernel.org>
> > > ---
> > > fs/f2fs/f2fs.h | 1 +
> > > fs/f2fs/segment.c | 15 ++++++++++-----
> > > 2 files changed, 11 insertions(+), 5 deletions(-)
> > >
> > > diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> > > index c0b83d6..ec4d16b 100644
> > > --- a/fs/f2fs/f2fs.h
> > > +++ b/fs/f2fs/f2fs.h
> > > @@ -104,6 +104,7 @@ enum {
> > > CP_DISCARD,
> > > };
> > >
> > > +#define BATCHED_TRIM_SEGMENTS 10
Our macro here shows one batched trim discard 10 segments, but actually we
discard 11 segments each time in f2fs_trim_fs, how about making them consistent?
One other point is that, I guess gc in FTL will be triggered each time after
discard command was issued, but if our basic size of one batched trim is not
multiple of gc unit size in FTL, efficiency of gc in FTL will be lower.
Now that f2fs has nature advantage that we can align section size to gc unit
size in FTL. Why not trying to op flash with section size as much as possible?
So how about use this:
#define BATCHED_TRIM_SECTIONS_SHIFT 4
our trim size: section size << BATCHED_TRIM_SECTIONS_SHIFT
Thanks,
> > > struct cp_control {
> > > int reason;
> > > __u64 trim_start;
> > > diff --git a/fs/f2fs/segment.c b/fs/f2fs/segment.c
> > > index 31c4e57..6c9c784 100644
> > > --- a/fs/f2fs/segment.c
> > > +++ b/fs/f2fs/segment.c
> > > @@ -1066,14 +1066,19 @@ 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 + 1) {
> > > + cpc.trim_start = start_segno;
> > > + cpc.trim_end = min_t(unsigned int,
> > > + start_segno + BATCHED_TRIM_SEGMENTS, 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
> > >
> > > --
> > > To unsubscribe from this list: send the line "unsubscribe linux-fsdevel"
> > > in
> > > the body of a message to majord...@vger.kernel.org
> > > More majordomo info at http://vger.kernel.org/majordomo-info.html
------------------------------------------------------------------------------
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
