Re: [PATCH] NFSD: Put the reference of nfs4_file when freeing stid

["J. Bruce Fields" <bfields@fieldses.org>]

On Tue, Aug 05, 2014 at 09:44:30AM -0400, Jeff Layton wrote:
> On Tue, 05 Aug 2014 21:20:27 +0800
> Kinglong Mee <kinglongmee@gmail.com> wrote:
> 
> > After testing nfs4 lock, I restart the nfsd service, got messages as,
> > 
> > [ 5677.403419] nfsd: last server has exited, flushing export cache
> > [ 5677.463728] =============================================================================
> > [ 5677.463942] BUG nfsd4_files (Tainted: G    B      OE): Objects remaining in nfsd4_files on kmem_cache_close()
> > [ 5677.464055] -----------------------------------------------------------------------------
> > 
> > [ 5677.464203] INFO: Slab 0xffffea0000233400 objects=28 used=1 fp=0xffff880008cd3d98 flags=0x3ffc0000004080
> > [ 5677.464318] CPU: 0 PID: 3772 Comm: rmmod Tainted: G    B      OE 3.16.0-rc2+ #29
> > [ 5677.464420] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 07/31/2013
> > [ 5677.464538]  0000000000000000 0000000036af2c9f ffff88000ce97d68 ffffffff816eacfa
> > [ 5677.464643]  ffffea0000233400 ffff88000ce97e40 ffffffff811cda44 ffffffff00000020
> > [ 5677.464774]  ffff88000ce97e50 ffff88000ce97e00 656a624f00000008 616d657220737463
> > [ 5677.464875] Call Trace:
> > [ 5677.464925]  [<ffffffff816eacfa>] dump_stack+0x45/0x56
> > [ 5677.464983]  [<ffffffff811cda44>] slab_err+0xb4/0xe0
> > [ 5677.465040]  [<ffffffff811d0457>] ? __kmalloc+0x117/0x290
> > [ 5677.465099]  [<ffffffff81100eec>] ? on_each_cpu_cond+0xac/0xf0
> > [ 5677.465158]  [<ffffffff811d1bc0>] ? kmem_cache_close+0x110/0x2e0
> > [ 5677.465218]  [<ffffffff811d1be0>] kmem_cache_close+0x130/0x2e0
> > [ 5677.465279]  [<ffffffff8135a0c1>] ? kobject_cleanup+0x91/0x1b0
> > [ 5677.465338]  [<ffffffff811d22be>] __kmem_cache_shutdown+0xe/0x10
> > [ 5677.465399]  [<ffffffff8119bd28>] kmem_cache_destroy+0x48/0x100
> > [ 5677.465466]  [<ffffffffa05ef78d>] nfsd4_free_slabs+0x2d/0x50 [nfsd]
> > [ 5677.465530]  [<ffffffffa05fa987>] exit_nfsd+0x34/0x6ad [nfsd]
> > [ 5677.465589]  [<ffffffff81104ac2>] SyS_delete_module+0x162/0x200
> > [ 5677.465649]  [<ffffffff81013b69>] ? do_notify_resume+0x59/0x90
> > [ 5677.465759]  [<ffffffff816f2369>] system_call_fastpath+0x16/0x1b
> > [ 5677.465822] INFO: Object 0xffff880008cd0000 @offset=0
> > [ 5677.465882] INFO: Allocated in nfsd4_process_open1+0x61/0x350 [nfsd] age=7599 cpu=0 pid=3253
> > [ 5677.466115]  __slab_alloc+0x3b0/0x4b1
> > [ 5677.466166]  kmem_cache_alloc+0x1e4/0x240
> > [ 5677.466220]  nfsd4_process_open1+0x61/0x350 [nfsd]
> > [ 5677.466276]  nfsd4_open+0xee/0x860 [nfsd]
> > [ 5677.466329]  nfsd4_proc_compound+0x4d7/0x7f0 [nfsd]
> > [ 5677.466384]  nfsd_dispatch+0xbb/0x200 [nfsd]
> > [ 5677.466447]  svc_process_common+0x453/0x6f0 [sunrpc]
> > [ 5677.466506]  svc_process+0x103/0x170 [sunrpc]
> > [ 5677.466559]  nfsd+0x117/0x190 [nfsd]
> > [ 5677.466609]  kthread+0xd8/0xf0
> > [ 5677.466656]  ret_from_fork+0x7c/0xb0
> > [ 5677.466775] kmem_cache_destroy nfsd4_files: Slab cache still has objects
> > [ 5677.466839] CPU: 0 PID: 3772 Comm: rmmod Tainted: G    B      OE 3.16.0-rc2+ #29
> > [ 5677.466937] Hardware name: VMware, Inc. VMware Virtual Platform/440BX Desktop Reference Platform, BIOS 6.00 07/31/2013
> > [ 5677.467049]  0000000000000000 0000000036af2c9f ffff88000ce97eb0 ffffffff816eacfa
> > [ 5677.467150]  ffff880020bb2d00 ffff88000ce97ed0 ffffffff8119bdd9 0000000000000000
> > [ 5677.467250]  ffffffffa06065c0 ffff88000ce97ee0 ffffffffa05ef78d ffff88000ce97ef0
> > [ 5677.467351] Call Trace:
> > [ 5677.467397]  [<ffffffff816eacfa>] dump_stack+0x45/0x56
> > [ 5677.467454]  [<ffffffff8119bdd9>] kmem_cache_destroy+0xf9/0x100
> > [ 5677.467516]  [<ffffffffa05ef78d>] nfsd4_free_slabs+0x2d/0x50 [nfsd]
> > [ 5677.467579]  [<ffffffffa05fa987>] exit_nfsd+0x34/0x6ad [nfsd]
> > [ 5677.467639]  [<ffffffff81104ac2>] SyS_delete_module+0x162/0x200
> > [ 5677.467765]  [<ffffffff81013b69>] ? do_notify_resume+0x59/0x90
> > [ 5677.467826]  [<ffffffff816f2369>] system_call_fastpath+0x16/0x1b
> > 
> > Signed-off-by: Kinglong Mee <kinglongmee@gmail.com>
> > ---
> >  fs/nfsd/nfs4state.c | 4 ++++
> >  1 file changed, 4 insertions(+)
> > 
> > diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c
> > index 56999cb..e87005d 100644
> > --- a/fs/nfsd/nfs4state.c
> > +++ b/fs/nfsd/nfs4state.c
> > @@ -1037,6 +1037,7 @@ static void
> >  free_ol_stateid_reaplist(struct list_head *reaplist)
> >  {
> >  	struct nfs4_ol_stateid *stp;
> > +	struct nfs4_file *fp;
> >  
> >  	might_sleep();
> >  
> > @@ -1044,7 +1045,10 @@ free_ol_stateid_reaplist(struct list_head *reaplist)
> >  		stp = list_first_entry(reaplist, struct nfs4_ol_stateid,
> >  				       st_locks);
> >  		list_del(&stp->st_locks);
> > +		fp = stp->st_stid.sc_file;
> >  		stp->st_stid.sc_free(&stp->st_stid);
> > +		if (fp)
> > +			put_nfs4_file(fp);
> >  	}
> >  }
> >  
> 
> Well spotted. I missed that when moved the nfs4_file handling out of
> sc_free.
> 
> Reviewed-by: Jeff Layton <jlayton@primarydata.com>

Thanks, Kinglong Mee!

I'll also add a note that this fixes 11b9164adad7 "nfsd: Add a struct
nfs4_file field to struct nfs4_stid".

--b.

--b.