Re: [PATCH 7/8] libxfs: add function to free all bufferse in bcache

["Darrick J. Wong" <darrick.wong@oracle.com>]

On Thu, Jan 11, 2018 at 02:11:09PM -0600, Eric Sandeen wrote:
> On 1/11/18 2:05 PM, Darrick J. Wong wrote:
> > On Thu, Jan 11, 2018 at 01:37:37PM -0600, Eric Sandeen wrote:
> >> Subject: libxfs: add function to free all bufferse in bcache
> > 
> > s/bufferse/buffers/
> 
> That's just my flair ;)
> 
> >> libxfs_bcache_purge simply moves all "free" buffers
> >> onto the xfs_buf_freelist mru list; add a new function to
> >> actually free them when we tear everything down, so leak
> >> checkers don't go nuts about lots of unfreed xfs_bufs
> >> at exit.
> >>
> >> Signed-off-by: Eric Sandeen <sandeen@redhat.com>
> >> Signed-off-by: Eric Sandeen <sandeen@sandeen.net>
> >> ---
> >>  libxfs/init.c      |  2 ++
> >>  libxfs/libxfs_io.h |  1 +
> >>  libxfs/rdwr.c      | 16 ++++++++++++++++
> >>  3 files changed, 19 insertions(+)
> >>
> >> diff --git a/libxfs/init.c b/libxfs/init.c
> >> index 7bde8b7..aea308b 100644
> >> --- a/libxfs/init.c
> >> +++ b/libxfs/init.c
> >> @@ -889,6 +889,8 @@ void
> >>  libxfs_destroy(void)
> >>  {
> >>  	manage_zones(1);
> >> +	libxfs_bcache_purge();
> >> +	libxfs_bcache_free();
> >>  	cache_destroy(libxfs_bcache);
> >>  }
> >>  
> >> diff --git a/libxfs/libxfs_io.h b/libxfs/libxfs_io.h
> >> index 2fce04d..81d2804 100644
> >> --- a/libxfs/libxfs_io.h
> >> +++ b/libxfs/libxfs_io.h
> >> @@ -191,6 +191,7 @@ extern void	libxfs_readbuf_verify(struct xfs_buf *bp,
> >>  			const struct xfs_buf_ops *ops);
> >>  extern xfs_buf_t *libxfs_getsb(struct xfs_mount *, int);
> >>  extern void	libxfs_bcache_purge(void);
> >> +extern void	libxfs_bcache_free(void);
> >>  extern void	libxfs_bcache_flush(void);
> >>  extern void	libxfs_purgebuf(xfs_buf_t *);
> >>  extern int	libxfs_bcache_overflowed(void);
> >> diff --git a/libxfs/rdwr.c b/libxfs/rdwr.c
> >> index c5ffd4d..1dcabdd 100644
> >> --- a/libxfs/rdwr.c
> >> +++ b/libxfs/rdwr.c
> >> @@ -1278,6 +1278,22 @@ libxfs_bulkrelse(
> >>  }
> >>  
> >>  /*
> >> + * Free everything from the xfs_buf_freelist MRU, used at final teardown
> >> + */
> >> +void
> >> +libxfs_bcache_free(void)
> >> +{
> >> +	struct list_head	*cm_list;
> >> +	xfs_buf_t		*bp, *next;
> >> +
> >> + 	cm_list = &xfs_buf_freelist.cm_list;
> >> +	list_for_each_entry_safe(bp, next, cm_list, b_node.cn_mru) {
> >> +		free(bp->b_addr);
> > 
> > Hm, do we need to free bp->b_maps here?  Just looking at
> > __libxfs_getbufr (aka the other function that can tear down a
> > bp->b_addr).
> 
> Oh yeah, probably so.
> >> +		free(bp);
> > 
> > kmem_zone_free, since we got the bp from kmem_zone_alloc.
> 
> Well remember I faked up the zone allocation count for buffers in and out of
> cache. If I zone_free here, count ll go way negative (but nobody will check at
> that point).  I'm not sure if this is all too strange to live...?

Yep, and I (just replied to the other patch) think the allocation counts
faking should go away, unless there's some reason why we need to have
xfs_buf_zone->allocated == 0 when there are buffers still hanging out on
xfs_buf_freelist.

> > Do we need to reinit xfs_buf_freelist.cm_list, since ->next now points
> > to freed list items?
> 
> Probably so.  Nobody uses it at this point anyway but still, best to leave it
> in a sane state I guess.

<nod>

--D

> -Eric
> --
> To unsubscribe from this list: send the line "unsubscribe linux-xfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html