On Wed, Jun 15, 2016 at 04:21:13AM +0000, Damien Le Moal wrote:
> Jaegeuk,
>
> On 6/15/16 03:45, Jaegeuk Kim wrote:
> > If mkfs.f2fs gives a feature flag for host-managed SMR, we can set mode=lfs
> > by default.
> >
> > Signed-off-by: Jaegeuk Kim <jaeg...@kernel.org>
> > ---
> > fs/f2fs/f2fs.h | 21 +++++++++++++++++++++
> > fs/f2fs/super.c | 14 ++++++++------
> > 2 files changed, 29 insertions(+), 6 deletions(-)
> >
> > diff --git a/fs/f2fs/f2fs.h b/fs/f2fs/f2fs.h
> > index 32be19e..1a19c02 100644
> > --- a/fs/f2fs/f2fs.h
> > +++ b/fs/f2fs/f2fs.h
> > @@ -133,6 +133,7 @@ struct f2fs_mount_info {
> > };
> >
> > #define F2FS_FEATURE_ENCRYPT 0x0001
> > +#define F2FS_FEATURE_HMSMR 0x0002
> >
> > #define F2FS_HAS_FEATURE(sb, mask) \
> > ((F2FS_SB(sb)->raw_super->feature & cpu_to_le32(mask)) != 0)
> > @@ -2354,6 +2355,26 @@ static inline int f2fs_sb_has_crypto(struct
> > super_block *sb)
> > return F2FS_HAS_FEATURE(sb, F2FS_FEATURE_ENCRYPT);
> > }
> >
> > +static inline int f2fs_sb_mounted_hmsmr(struct super_block *sb)
> > +{
> > + return F2FS_HAS_FEATURE(sb, F2FS_FEATURE_HMSMR);
> > +}
> > +
> > +static inline void set_opt_mode(struct f2fs_sb_info *sbi, unsigned int mt)
> > +{
> > + clear_opt(sbi, ADAPTIVE);
> > + clear_opt(sbi, LFS);
> > +
> > + switch (mt) {
> > + case F2FS_MOUNT_ADAPTIVE:
> > + set_opt(sbi, ADAPTIVE);
> > + break;
> > + case F2FS_MOUNT_LFS:
> > + set_opt(sbi, LFS);
> > + break;
> > + }
> > +}
> > +
> > static inline bool f2fs_may_encrypt(struct inode *inode)
> > {
> > #ifdef CONFIG_F2FS_FS_ENCRYPTION
> > diff --git a/fs/f2fs/super.c b/fs/f2fs/super.c
> > index edc736d..71b6066 100644
> > --- a/fs/f2fs/super.c
> > +++ b/fs/f2fs/super.c
> > @@ -515,12 +515,10 @@ static int parse_options(struct super_block *sb, char
> > *options)
> > return -ENOMEM;
> > if (strlen(name) == 8 &&
> > !strncmp(name, "adaptive", 8)) {
> > - set_opt(sbi, ADAPTIVE);
> > - clear_opt(sbi, LFS);
> > + set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
> > } else if (strlen(name) == 3 &&
> > !strncmp(name, "lfs", 3)) {
> > - clear_opt(sbi, ADAPTIVE);
> > - set_opt(sbi, LFS);
> > + set_opt_mode(sbi, F2FS_MOUNT_LFS);
> > } else {
> > kfree(name);
> > return -EINVAL;
> > @@ -980,7 +978,10 @@ static void default_options(struct f2fs_sb_info *sbi)
> > set_opt(sbi, EXTENT_CACHE);
> > sbi->sb->s_flags |= MS_LAZYTIME;
> > set_opt(sbi, FLUSH_MERGE);
> > - set_opt(sbi, ADAPTIVE);
> > + if (f2fs_sb_mounted_hmsmr(sbi->sb))
> > + set_opt_mode(sbi, F2FS_MOUNT_LFS);
> > + else
> > + set_opt_mode(sbi, F2FS_MOUNT_ADAPTIVE);
>
> Strictly speaking, host-aware drives would not require the LFS mode
> as any zone of the disk can be randomly written and so the occasional
> in-place updates of the adaptive mode would be OK. However, randomly
> writing host-aware drives can lead to a lot of background activity on
> the drive side for internal management of the accumulated random writes,
> with a potential performance drop over time (this is highly dependent on
> the drive FW implementation though).
>
> So I think it would be good to also enable the LFS mode by default for
> host-aware devices, unless the user has explicitly specified mount with
> adaptive mode (if the user "knows" that the drive FW handles very
> efficiently random writes).
Agreed. Would HASMR also provide the same geometry, zone information, in mkfs?
Cause I haven't tested it so far.
IIRC, we discussed about HASMR with Marc Lehmann last year, and he gave the
below link which sums up.
http://comments.gmane.org/gmane.linux.file-systems.f2fs/2854
http://blog.schmorp.de/2015-10-08-smr-archive-drives-fast-now.html
> So how about also introducing a F2FS_FEATURE_HASMR feature flag to
> handle these different cases ?
Yes, so once I can retrieve the zone information from the device, surely I'll
add that.
> Also, I think that the DISCARD option must be enabled by default for
> HMSMR disks. Otherwise, zones write pointer will never get reset.
> The same applies to HASMR devices mounted with the LFS mode.
> (In any case, the discard handling does not look like it will always
> align to the device zone size, which will fail on a zoned disk (discard
> granularity is the zone size). I may be missing something though. Still
> checking.)
Yup, will do. BTW, I remember I fixed zone-aligned discard issue before.
Could you please check the latest dev-test branch in f2fs.git?
I'm used to rebase that branch occasionally when I changed the previous patches.
Thanks,
>
> Best regards.
>
> --
> Damien Le Moal, Ph.D.
> Sr. Manager, System Software Group, HGST Research,
> HGST, a Western Digital company
> damien.lem...@hgst.com
> (+81) 0466-98-3593 (ext. 513593)
> 1 kirihara-cho, Fujisawa,
> Kanagawa, 252-0888 Japan
> www.hgst.com
> Western Digital Corporation (and its subsidiaries) E-mail Confidentiality
> Notice & Disclaimer:
>
> This e-mail and any files transmitted with it may contain confidential or
> legally privileged information of WDC and/or its affiliates, and are intended
> solely for the use of the individual or entity to which they are addressed.
> If you are not the intended recipient, any disclosure, copying, distribution
> or any action taken or omitted to be taken in reliance on it, is prohibited.
> If you have received this e-mail in error, please notify the sender
> immediately and delete the e-mail in its entirety from your system.
------------------------------------------------------------------------------
What NetFlow Analyzer can do for you? Monitors network bandwidth and traffic
patterns at an interface-level. Reveals which users, apps, and protocols are
consuming the most bandwidth. Provides multi-vendor support for NetFlow,
J-Flow, sFlow and other flows. Make informed decisions using capacity planning
reports. http://pubads.g.doubleclick.net/gampad/clk?id=1444514421&iu=/41014381
_______________________________________________
Linux-f2fs-devel mailing list
Linux-f2fs-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
