linux-kernel - Re: [f2fs-dev] [PATCH 3/4] f2fs: avoid f2fs_lock_op in f2fs_write

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20151224203350.GB43846@jaegeuk.local>
Date:	Thu, 24 Dec 2015 12:33:50 -0800
From:	Jaegeuk Kim <jaegeuk@...nel.org>
To:	Chao Yu <chao2.yu@...sung.com>
Cc:	linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org,
	linux-f2fs-devel@...ts.sourceforge.net
Subject: Re: [f2fs-dev] [PATCH 3/4] f2fs: avoid f2fs_lock_op in
 f2fs_write_begin

Hi Chao,

On Thu, Dec 24, 2015 at 01:50:32PM +0800, Chao Yu wrote:
> Hi Jaegeuk,
> 
> > -----Original Message-----
> > From: Jaegeuk Kim [mailto:jaegeuk@...nel.org]
> > Sent: Thursday, December 24, 2015 10:15 AM
> > To: linux-kernel@...r.kernel.org; linux-fsdevel@...r.kernel.org;
> > linux-f2fs-devel@...ts.sourceforge.net
> > Cc: Jaegeuk Kim
> > Subject: [f2fs-dev] [PATCH 3/4] f2fs: avoid f2fs_lock_op in f2fs_write_begin
> > 
> > If f2fs_write_begin is to update data, we can bypass calling f2fs_lock_op() in
> > order to avoid the checkpoint latency in the write syscall.
> > 
> > Signed-off-by: Jaegeuk Kim <jaegeuk@...nel.org>
> > ---
> >  fs/f2fs/data.c | 43 ++++++++++++++++++++++++++++++++++---------
> >  1 file changed, 34 insertions(+), 9 deletions(-)
> > 
> > diff --git a/fs/f2fs/data.c b/fs/f2fs/data.c
> > index 49092da..d53cf4f 100644
> > --- a/fs/f2fs/data.c
> > +++ b/fs/f2fs/data.c
> > @@ -1418,10 +1418,16 @@ static int prepare_write_begin(struct f2fs_sb_info *sbi,
> >  	pgoff_t index = page->index;
> >  	struct dnode_of_data dn;
> >  	struct page *ipage;
> > +	bool locked = false;
> > +	struct extent_info ei;
> >  	int err = 0;
> > 
> > -	f2fs_lock_op(sbi);
> > -
> > +	if (f2fs_has_inline_data(inode) ||
> > +			(pos & PAGE_CACHE_MASK) >= i_size_read(inode)) {
> > +		f2fs_lock_op(sbi);
> > +		locked = true;
> > +	}
> > +restart:
> >  	/* check inline_data */
> >  	ipage = get_node_page(sbi, inode->i_ino);
> >  	if (IS_ERR(ipage)) {
> > @@ -1436,22 +1442,41 @@ static int prepare_write_begin(struct f2fs_sb_info *sbi,
> >  			read_inline_data(page, ipage);
> >  			set_inode_flag(F2FS_I(inode), FI_DATA_EXIST);
> >  			sync_inode_page(&dn);
> > -			goto done;
> >  		} else {
> >  			err = f2fs_convert_inline_page(&dn, page);
> >  			if (err)
> > -				goto err_out;
> > +				goto out;
> > +			err = f2fs_get_block(&dn, index);
> 
> Seems after f2fs_convert_inline_page, data_blkaddr and node_changed should
> have been set, so f2fs_get_block could be removed here.

I don't think so. The inline_data treats with 0'th index only.
In order to handle non-zero index, we should do f2fs_get_block.
Oh, if you meant zero index case only, we can do that like this.

			if (index)
				err = f2fs_get_block();

> 
> > +		}
> > +	} else if (locked) {
> > +		err = f2fs_get_block(&dn, index);
> > +	} else {
> > +		if (f2fs_lookup_extent_cache(inode, index, &ei)) {
> > +			dn.data_blkaddr = ei.blk + index - ei.fofs;
> > +		} else {
> > +			bool restart = false;
> > +
> > +			/* hole case */
> > +			err = get_dnode_of_data(&dn, index, LOOKUP_NODE);
> 
> Better to handle error case like -ENOMEM or -EIO.

I think we can just give another chance at this moment and let f2fs_get_block()
handle that in the next round.

Thanks,

> 
> Thanks,
> 
> > +			if (err || (!err && dn.data_blkaddr == NULL_ADDR))
> > +				restart = true;
> > +			if (restart) {
> > +				f2fs_put_dnode(&dn);
> > +				f2fs_lock_op(sbi);
> > +				locked = true;
> > +				goto restart;
> > +			}
> >  		}
> >  	}
> > -	err = f2fs_get_block(&dn, index);
> > -done:
> > +
> >  	/* convert_inline_page can make node_changed */
> >  	*blk_addr = dn.data_blkaddr;
> >  	*node_changed = dn.node_changed;
> > -err_out:
> > +out:
> >  	f2fs_put_dnode(&dn);
> >  unlock_out:
> > -	f2fs_unlock_op(sbi);
> > +	if (locked)
> > +		f2fs_unlock_op(sbi);
> >  	return err;
> >  }
> > 
> > @@ -1488,7 +1513,7 @@ repeat:
> > 
> >  	*pagep = page;
> > 
> > -	err = prepare_write_begin(sbi, page, pos + len,
> > +	err = prepare_write_begin(sbi, page, pos,  len,
> >  					&blkaddr, &need_balance);
> >  	if (err)
> >  		goto fail;
> > --
> > 2.5.4 (Apple Git-61)
> > 
> > 
> > ------------------------------------------------------------------------------
> > _______________________________________________
> > Linux-f2fs-devel mailing list
> > Linux-f2fs-devel@...ts.sourceforge.net
> > https://lists.sourceforge.net/lists/listinfo/linux-f2fs-devel
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/