Re: [f2fs-dev] [PATCH 2/2] f2fs: fix to release inode correctly

[Jaegeuk Kim <jaegeuk@kernel.org>]

On Tue, Aug 25, 2015 at 02:38:00PM +0800, Chao Yu wrote:
> Hi Jaegeuk,
> 
> > -----Original Message-----
> > From: Jaegeuk Kim [mailto:jaegeuk@kernel.org]
> > Sent: Tuesday, August 25, 2015 6:53 AM
> > To: Chao Yu
> > Cc: linux-kernel@vger.kernel.org; linux-f2fs-devel@lists.sourceforge.net
> > Subject: Re: [f2fs-dev] [PATCH 2/2] f2fs: fix to release inode correctly
> > 
> > Hi Chao,
> > 
> > On Mon, Aug 24, 2015 at 09:54:23AM -0700, Jaegeuk Kim wrote:
> > > On Mon, Aug 24, 2015 at 05:40:45PM +0800, Chao Yu wrote:
> > > > In following call stack, if unfortunately we lose all chances to truncate
> > > > inode page in remove_inode_page, eventually we will add the nid allocated
> > > > previously into free nid cache, this nid is with NID_NEW status and with
> > > > NEW_ADDR in its blkaddr pointer:
> > > >
> > > >  - f2fs_create
> > > >   - f2fs_add_link
> > > >    - __f2fs_add_link
> > > >     - init_inode_metadata
> > > >      - new_inode_page
> > > >       - new_node_page
> > > >        - set_node_addr(, NEW_ADDR)
> > > >      - f2fs_init_acl   failed
> > > >      - remove_inode_page  failed
> > > >   - handle_failed_inode
> > > >    - remove_inode_page  failed
> > > >    - iput
> > > >     - f2fs_evict_inode
> > > >      - remove_inode_page  failed
> > > >      - alloc_nid_failed   cache a nid with valid blkaddr: NEW_ADDR
> > 
> > Unfortunately, this couldn't fix my bug case.
> 
> Another thing I note is that: we do not cover free_nid_list_lock with build_lock,
> so when we are building free nid cache, we can change the status of free nid
> cache, so I guess it is one possible suspect who cause our nid issue.

But, when building the free nids, it doesn't add any candidate if there is
an existing entry in the list which might be set as NID_ALLOC.

> 
> And, could you share me the information for reproducing the nid reallocation
> issue? So I can reproduce in my environment for invistigating.

Sure. I've running ubuntu on virtualbox equipped with an 8GB virtual partition
with 1GB main memory.
The issue occurs during sometimes xfstests, or sometimes fsstress which fills up
the partition to 100%.
With my quick fix, I couldn't meet that bug for two days run.

Thanks,

> 
> > I'm still struggling to find out something tho.
> > Meanwhile, let's stay with both of the patches.
> 
> OK.
> 
> Thanks,