Message-ID: <20140914073803.GA51401@jaegeuk-mac02.hsd1.ca.comcast.net>
Date:	Sun, 14 Sep 2014 00:38:03 -0700
From:	Jaegeuk Kim <jaegeuk@...nel.org>
To:	Huang Ying <ying.huang@...el.com>
Cc:	huang ying <huang.ying.caritas@...il.com>,
	Changman Lee <cm224.lee@...sung.com>,
	linux-f2fs-devel@...ts.sourceforge.net,
	LKML <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH] f2fs: Fix recover when nid of non-inode dnode < nid of
 inode

On Fri, Sep 12, 2014 at 03:34:48PM +0800, Huang Ying wrote:
> On Thu, 2014-09-11 at 22:13 -0700, Jaegeuk Kim wrote:
> > On Thu, Sep 11, 2014 at 08:25:17PM +0800, Huang Ying wrote:
> > > 
> > > On Wed, 2014-09-10 at 22:37 -0700, Jaegeuk Kim wrote:
> > > > On Wed, Sep 10, 2014 at 07:08:32PM +0800, huang ying wrote:
> > > > > On Wed, Sep 10, 2014 at 3:21 PM, Jaegeuk Kim <jaegeuk@...nel.org> wrote:
> > > > > 
> > > > > > On Tue, Sep 09, 2014 at 07:31:49PM +0800, huang ying wrote:
> > > > > > > On Tue, Sep 9, 2014 at 3:09 PM, Jaegeuk Kim <jaegeuk@...nel.org> wrote:
> > > > > > >
> > > > > > > > Hi,
> > > > > > > >
> > > > > > > > On Tue, Sep 09, 2014 at 01:39:30PM +0800, Huang Ying wrote:
> > > > > > > > > On Mon, 2014-09-08 at 22:23 -0700, Jaegeuk Kim wrote:
> > > > > > > > > > Hi Huang,
> > > > > > > > > >
> > > > > > > > > > On Mon, Sep 08, 2014 at 07:38:26PM +0800, Huang Ying wrote:
> > > > > > > > > > > For fsync, if the nid of a non-inode dnode < nid of inode and the
> > > > > > > > > > > inode is not checkpointed.  The non-inode dnode may be written
> > > > > > before
> > > > > > > > > > > inode.  So in find_fsync_dnodes, f2fs_iget will fail, cause the
> > > > > > > > > > > recovery fail.
> > > > > > > > > > >
> > > > > > > > > > > Usually, inode will be allocated before non-inode dnode, so the
> > > > > > nid
> > > > > > > > of
> > > > > > > > > > > inode < nid of non-inode dnode.  But it is possible for the
> > > > > > reverse.
> > > > > > > > > > > For example, because of alloc_nid_failed.
> > > > > > > > > > >
> > > > > > > > > > > This is fixed via ignoring non-inode dnode before inode dnode in
> > > > > > > > > > > find_fsync_dnodes.
> > > > > > > > > > >
> > > > > > > > > > > The patch was tested via allocating nid reversely via a debugging
> > > > > > > > > > > patch, that is, from big number to small number.
> > > > > > > > > > >
> > > > > > > > > > > Signed-off-by: Huang, Ying <ying.huang@...el.com>
> > > > > > > > > > > ---
> > > > > > > > > > >  fs/f2fs/recovery.c |    7 ++++---
> > > > > > > > > > >  1 file changed, 4 insertions(+), 3 deletions(-)
> > > > > > > > > > >
> > > > > > > > > > > --- a/fs/f2fs/recovery.c
> > > > > > > > > > > +++ b/fs/f2fs/recovery.c
> > > > > > > > > > > @@ -172,8 +172,8 @@ static int find_fsync_dnodes(struct f2fs
> > > > > > > > > > >                   if (IS_INODE(page) && is_dent_dnode(page))
> > > > > > > > > > >                           set_inode_flag(F2FS_I(entry->inode),
> > > > > > > > > > >                                                   FI_INC_LINK);
> > > > > > > > > > > -         } else {
> > > > > > > > > > > -                 if (IS_INODE(page) && is_dent_dnode(page)) {
> > > > > > > > > >
> > > > > > > > > > If this is not inode block, we should add this inode to recover its
> > > > > > > > data blocks.
> > > > > > > > >
> > > > > > > > > Is it possible that there is only non-inode dnode but no inode when
> > > > > > > > > find_fsync_dnodes checking dnodes?  Per my understanding, any
> > > > > > changes to
> > > > > > > > > file will cause inode page dirty (for example, mtime changed), so
> > > > > > that
> > > > > > > > > we will write inode block.  Is it right?  If so, the solution in this
> > > > > > > > > patch should work too.
> > > > > > > >
> > > > > > > > Your description says that f2fs_iget will fail, which causes the
> > > > > > recovery
> > > > > > > > fail.
> > > > > > > > So, I thought it would be better to handle the f2fs_iget failure
> > > > > > directly.
> > > > > > > >
> > > > > > >
> > > > > > > Yes.  That is another way to fix the issue.
> > > > > > >
> > > > > > >
> > > > > > > > In addition, we cannot guarantee the write order of dnode and inode.
> > > > > > > > For exmaple,
> > > > > > > > 1. the inode is written by flusher or kswapd, then,
> > > > > > > > 2. f2fs_sync_file writes its dnode.
> > > > > > > >
> > > > > > > > In that case, we can get only non-inode dnode in the node chain, since
> > > > > > the
> > > > > > > > inode
> > > > > > > > has not fsync_mark.
> > > > > > > >
> > > > > > >
> > > > > > > I think your solution is better here, but does not fix all scenarios.  If
> > > > > > > the inode is checkpointed, the file can be recovered, although the inode
> > > > > > > information may be not up to date.  But if the inode is not checkpointed,
> > > > > > > f2fs_iget will fail too and recover will fail.
> > > > > >
> > > > > > Ok, let me consider your scenarios.
> > > > > >
> > > > > > Term: F: fsync_mark, D: dentry_mark
> > > > > >
> > > > > > 1. inode(x) | CP | inode(x) | dnode(F)
> > > > > >  -> Lose the latest inode(x). Need to fix.
> > > > > >
> > > > > > 2. inode(x) | CP | dnode(F) | inode(x)
> > > > > >  -> Impossible, but recover latest dnode(F)
> > > > > >
> > > > > > 3. CP | inode(x) | dnode(F)
> > > > > >  -> Need to write inode(DF) in f2fs_sync_file.
> > > > > >
> > > > > > 4. CP | dnode(F) | inode(DF)
> > > > > >  -> If f2fs_iget fails, then goto next.
> > > > > >
> > > > > > 5. CP | dnode(F) | inode(x)
> > > > > >  -> If f2fs_iget fails, then goto next. But, this is an impossible
> > > > > > scenario.
> > > > > >     Drop this dnode(F).
> > > > > >
> > > > > > Indeed, there were some missing scenarios.
> > > > > >
> > > > > > So, how about this patch?
> > > > > >
> > > > > > From 552dc68c5f07a335d7b55c197bab531efb135521 Mon Sep 17 00:00:00 2001
> > > > > > From: Jaegeuk Kim <jaegeuk@...nel.org>
> > > > > > Date: Wed, 10 Sep 2014 00:16:34 -0700
> > > > > > Subject: [PATCH] f2fs: fix roll-forward missing scenarios
> > > > > >
> > > > > > We can summarize the roll forward recovery scenarios as follows.
> > > > > >
> > > > > > [Term] F: fsync_mark, D: dentry_mark
> > > > > >
> > > > > > 1. inode(x) | CP | inode(x) | dnode(F)
> > > > > > -> Update the latest inode(x).
> > > > > >
> > > > > > 2. inode(x) | CP | inode(F) | dnode(F)
> > > > > > -> No problem.
> > > > > >
> > > > > > 3. inode(x) | CP | dnode(F) | inode(x)
> > > > > > -> Impossible, but recover latest dnode(F)
> > > > > >
> > > > > 
> > > > > I think this is possible.  If f2fs_sync_file runs concurrently with
> > > > > writeback. f2fs_sync_file written dnode(F), then writeback written inode(x).
> > > > 
> > > > If the inode(x) was written, f2fs_sync_file will do write the inode again with
> > > > fsync_mark. So, dnode(F) | inode(x) | inode(F) should be shown.
> > > > 
> > > > In f2fs_sync_file,
> > > > 	...
> > > > 	while (!sync_node_pages(sbi, ino, &wbc)) {
> > > > 		if (fsync_mark_done(sbi, ino))
> > > > 			goto out;
> > > > 		mark_inode_dirty_sync(inode);
> > > > 		ret = f2fs_write_inode(inode, NULL);
> > > > 		if (ret)
> > > > 			goto out;
> > > > 	}
> > > > 	...
> > > 
> > > Is the following situation possible?
> > > 
> > > f2fs_sync_file				<writeback>
> > >   sync_node_pages			  f2fs_write_node_pages
> > >     write dnode(F)			    sync_node_pages
> > > 					      write inode(x) /* clear PAGECACHE_TAG_DIRTY */
> > > 
> > > 
> > > That is, f2fs_sync_file run parallel with <writeback>.  The
> > > sync_node_pages above will return 1, because dnode(F) is written.
> > > inode(x) is written by <writeback> path.  And because
> > > PAGECACHE_TAG_DIRTY is cleared, it is possible that sync_node_pages
> > > called by f2fs_sync_file does not write inode(F).
> > 
> > I think Chao's comment would work.
> > How about this patch?
> > 
> > From 32fe5ff49d2c78d3be4cf3638cc64ae71cf44549 Mon Sep 17 00:00:00 2001
> > From: Jaegeuk Kim <jaegeuk@...nel.org>
> > Date: Wed, 10 Sep 2014 00:16:34 -0700
> > Subject: [PATCH] f2fs: fix roll-forward missing scenarios
> > 
> > We can summarize the roll forward recovery scenarios as follows.
> > 
> > [Term] F: fsync_mark, D: dentry_mark
> > 
> > 1. inode(x) | CP | inode(x) | dnode(F)
> > -> Update the latest inode(x).
> > 
> > 2. inode(x) | CP | inode(F) | dnode(F)
> > -> No problem.
> > 
> > 3. inode(x) | CP | dnode(F) | inode(x)
> > -> Recover to the latest dnode(F), and drop the last inode(x)
> > 
> > 4. inode(x) | CP | dnode(F) | inode(F)
> > -> No problem.
> > 
> > 5. CP | inode(x) | dnode(F)
> > -> The inode(DF) was missing. Should drop this dnode(F).
> > 
> > 6. CP | inode(DF) | dnode(F)
> > -> No problem.
> > 
> > 7. CP | dnode(F) | inode(DF)
> > -> If f2fs_iget fails, then goto next to find inode(DF).
> > 
> > 8. CP | dnode(F) | inode(x)
> > -> If f2fs_iget fails, then goto next to find inode(DF).
> >    But it will fail due to no inode(DF).
> > 
> > So, this patch adds some missing points such as #1, #5, #7, and #8.
> > 
> > Signed-off-by: Jaegeuk Kim <jaegeuk@...nel.org>
> > ---
> >  fs/f2fs/file.c     | 20 ++++++++++++----
> >  fs/f2fs/node.c     | 11 ++++++++-
> >  fs/f2fs/recovery.c | 70 +++++++++++++++++++++++++++++++++++++++++++++---------
> >  3 files changed, 85 insertions(+), 16 deletions(-)
> > 
> > diff --git a/fs/f2fs/file.c b/fs/f2fs/file.c
> > index e7681c3..70f5d4b 100644
> > --- a/fs/f2fs/file.c
> > +++ b/fs/f2fs/file.c
> > @@ -206,15 +206,27 @@ int f2fs_sync_file(struct file *file, loff_t start, loff_t end, int datasync)
> >  			up_write(&fi->i_sem);
> >  		}
> >  	} else {
> > -		/* if there is no written node page, write its inode page */
> > -		while (!sync_node_pages(sbi, ino, &wbc)) {
> > -			if (fsync_mark_done(sbi, ino))
> > -				goto out;
> > +sync_nodes:
> > +		sync_node_pages(sbi, ino, &wbc);
> > +
> > +		/*
> > +		 * inode(x) | CP | inode(x) | dnode(F)
> > +		 *  -> ok
> 
> Is it acceptable that we turn this to:
> 
> inode(x) | CPU | inode (x) | dnode (F) | inode(F)

Why do we have to do writing additonal one block unnecessarily?
We got inode(x) for the recovery.

> 
> > +		 * inode(x) | CP | dnode(F) | inode(x)
> > +		 *  -> inode(x) | CP | dnode(F) | inode(x) | inode(F)
> > +		 * CP | inode(x) | dnode(F)
> > +		 *  -> CP | inode(x) | dnode(F) | inode(DF)
> > +		 * CP | dnode(F) | inode(x)
> > +		 *  -> CP | dnode(F) | inode(x) | inode(DF)
> > +		 */
> > +		if (!fsync_mark_done(sbi, ino)) {
> >  			mark_inode_dirty_sync(inode);
> >  			ret = f2fs_write_inode(inode, NULL);
> >  			if (ret)
> >  				goto out;
> > +			goto sync_nodes;
> >  		}
> > +
> >  		ret = wait_on_node_pages_writeback(sbi, ino);
> >  		if (ret)
> >  			goto out;
> > diff --git a/fs/f2fs/node.c b/fs/f2fs/node.c
> > index b32eb56..653aa71 100644
> > --- a/fs/f2fs/node.c
> > +++ b/fs/f2fs/node.c
> > @@ -248,8 +248,17 @@ retry:
> >  
> >  	/* update fsync_mark if its inode nat entry is still alive */
> >  	e = __lookup_nat_cache(nm_i, ni->ino);
> > -	if (e)
> > +	if (e) {
> > +		/*
> > +		 * CP | inode(x) | dnode(F)
> > +		 *  -> CP | inode(x) | dnode(F) | inode(DF)
> > +		 */
> > +		if (!e->checkpointed && !e->fsync_done &&
> > +				ni->ino != ni->nid && fsync_done)
> > +			goto skip;
> >  		e->fsync_done = fsync_done;
> > +	}
> > +skip:
> >  	write_unlock(&nm_i->nat_tree_lock);
> >  }
> 
> I don't understand why we need so complex logic?  Why not just let
> e->fsync_done reflect just latest is_fsync_dnode(page)?
> 
> It appears that in f2fs_sync_file, what we need is just whether inode
> page has fsync mark or not.
> 
> Best Regards,
> Huang, Ying
>  
> > diff --git a/fs/f2fs/recovery.c b/fs/f2fs/recovery.c
> > index 6c5a74a..3736728 100644
> > --- a/fs/f2fs/recovery.c
> > +++ b/fs/f2fs/recovery.c
> > @@ -14,6 +14,36 @@
> >  #include "node.h"
> >  #include "segment.h"
> >  
> > +/*
> > + * Roll forward recovery scenarios.
> > + *
> > + * [Term] F: fsync_mark, D: dentry_mark
> > + *
> > + * 1. inode(x) | CP | inode(x) | dnode(F)
> > + * -> Update the latest inode(x).
> > + *
> > + * 2. inode(x) | CP | inode(F) | dnode(F)
> > + * -> No problem.
> > + *
> > + * 3. inode(x) | CP | dnode(F) | inode(x)
> > + * -> Recover to the latest dnode(F), and drop the last inode(x)
> > + *
> > + * 4. inode(x) | CP | dnode(F) | inode(F)
> > + * -> No problem.
> > + *
> > + * 5. CP | inode(x) | dnode(F)
> > + * -> The inode(DF) was missing. Should drop this dnode(F).
> > + *
> > + * 6. CP | inode(DF) | dnode(F)
> > + * -> No problem.
> > + *
> > + * 7. CP | dnode(F) | inode(DF)
> > + * -> If f2fs_iget fails, then goto next to find inode(DF).
> > + *
> > + * 8. CP | dnode(F) | inode(x)
> > + * -> If f2fs_iget fails, then goto next to find inode(DF).
> > + *    But it will fail due to no inode(DF).
> > + */
> >  static struct kmem_cache *fsync_entry_slab;
> >  
> >  bool space_for_roll_forward(struct f2fs_sb_info *sbi)
> > @@ -110,27 +140,32 @@ out:
> >  	return err;
> >  }
> >  
> > -static int recover_inode(struct inode *inode, struct page *node_page)
> > +static void __recover_inode(struct inode *inode, struct page *page)
> >  {
> > -	struct f2fs_inode *raw_inode = F2FS_INODE(node_page);
> > +	struct f2fs_inode *raw = F2FS_INODE(page);
> > +
> > +	inode->i_mode = le16_to_cpu(raw->i_mode);
> > +	i_size_write(inode, le64_to_cpu(raw->i_size));
> > +	inode->i_atime.tv_sec = le64_to_cpu(raw->i_mtime);
> > +	inode->i_ctime.tv_sec = le64_to_cpu(raw->i_ctime);
> > +	inode->i_mtime.tv_sec = le64_to_cpu(raw->i_mtime);
> > +	inode->i_atime.tv_nsec = le32_to_cpu(raw->i_mtime_nsec);
> > +	inode->i_ctime.tv_nsec = le32_to_cpu(raw->i_ctime_nsec);
> > +	inode->i_mtime.tv_nsec = le32_to_cpu(raw->i_mtime_nsec);
> > +}
> >  
> > +static int recover_inode(struct inode *inode, struct page *node_page)
> > +{
> >  	if (!IS_INODE(node_page))
> >  		return 0;
> >  
> > -	inode->i_mode = le16_to_cpu(raw_inode->i_mode);
> > -	i_size_write(inode, le64_to_cpu(raw_inode->i_size));
> > -	inode->i_atime.tv_sec = le64_to_cpu(raw_inode->i_mtime);
> > -	inode->i_ctime.tv_sec = le64_to_cpu(raw_inode->i_ctime);
> > -	inode->i_mtime.tv_sec = le64_to_cpu(raw_inode->i_mtime);
> > -	inode->i_atime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec);
> > -	inode->i_ctime.tv_nsec = le32_to_cpu(raw_inode->i_ctime_nsec);
> > -	inode->i_mtime.tv_nsec = le32_to_cpu(raw_inode->i_mtime_nsec);
> > +	__recover_inode(inode, node_page);
> >  
> >  	if (is_dent_dnode(node_page))
> >  		return recover_dentry(node_page, inode);
> >  
> >  	f2fs_msg(inode->i_sb, KERN_NOTICE, "recover_inode: ino = %x, name = %s",
> > -			ino_of_node(node_page), raw_inode->i_name);
> > +			ino_of_node(node_page), F2FS_INODE(node_page)->i_name);
> >  	return 0;
> >  }
> >  
> > @@ -186,10 +221,16 @@ static int find_fsync_dnodes(struct f2fs_sb_info *sbi, struct list_head *head)
> >  				break;
> >  			}
> >  
> > +			/*
> > +			 * CP | dnode(F) | inode(DF)
> > +			 * For this case, we should not give up now.
> > +			 */
> >  			entry->inode = f2fs_iget(sbi->sb, ino_of_node(page));
> >  			if (IS_ERR(entry->inode)) {
> >  				err = PTR_ERR(entry->inode);
> >  				kmem_cache_free(fsync_entry_slab, entry);
> > +				if (err == -ENOENT)
> > +					goto next;
> >  				break;
> >  			}
> >  			list_add_tail(&entry->list, head);
> > @@ -416,6 +457,13 @@ static int recover_data(struct f2fs_sb_info *sbi,
> >  		entry = get_fsync_inode(head, ino_of_node(page));
> >  		if (!entry)
> >  			goto next;
> > +		/*
> > +		 * inode(x) | CP | inode(x) | dnode(F)
> > +		 * In this case, we can lose the latest inode(x).
> > +		 * So, call __recover_inode for the inode update.
> > +		 */
> > +		if (IS_INODE(page))
> > +			__recover_inode(entry->inode, page);
> >  
> >  		err = do_recover_data(sbi, entry->inode, page, blkaddr);
> >  		if (err)
--
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/

Powered by blists - more mailing lists