2008-01-26 00:31:58

by [email protected]

[permalink] [raw]
Subject: Re: [PATCH 097/100] NLM: have nlm_shutdown_hosts kill off all NLM RPC tasks

On Fri, Jan 25, 2008 at 06:55:48PM -0500, Jeff Layton wrote:
> On Fri, 25 Jan 2008 18:17:17 -0500
> "J. Bruce Fields" <[email protected]> wrote:
>
> > From: Jeff Layton <[email protected]>
> >
> > If we're shutting down all the nlm_hosts anyway, then it doesn't make
> > sense to allow RPC calls to linger. Allowing them to do so can mean
> > that the RPC calls can outlive the currently running lockd and can
> > lead to a use after free situation.
> >
> > Signed-off-by: Jeff Layton <[email protected]>
> > Reviewed-by: NeilBrown <[email protected]>
> > Signed-off-by: J. Bruce Fields <[email protected]>
> > ---
> > fs/lockd/host.c | 4 +++-
> > 1 files changed, 3 insertions(+), 1 deletions(-)
> >
> > diff --git a/fs/lockd/host.c b/fs/lockd/host.c
> > index ebec009..76e4bf5 100644
> > --- a/fs/lockd/host.c
> > +++ b/fs/lockd/host.c
> > @@ -379,8 +379,10 @@ nlm_shutdown_hosts(void)
> > /* First, make all hosts eligible for gc */
> > dprintk("lockd: nuking all hosts...\n");
> > for (chain = nlm_hosts; chain < nlm_hosts + NLM_HOST_NRHASH;
> > ++chain) {
> > - hlist_for_each_entry(host, pos, chain, h_hash)
> > + hlist_for_each_entry(host, pos, chain, h_hash) {
> > host->h_expires = jiffies - 1;
> > + rpc_killall_tasks(host->h_rpcclnt);
> > + }
> > }
> >
> > /* Then, perform a garbage collection pass */
>
> I was doing some more testing today, and noticed that the original
> problem that this patch is intended to fix resurfaced. I think this
> patch just changes the timing on the race somehow, but I haven't tracked
> it down completely yet.

Hm. So you're getting oopses that look like the rpc code attempting to
make calls with an rpc_client that's been freed?

> There's also another problem -- it's possible for host->h_rpcclnt to be
> NULL, and that has special meaning for rpc_killall_tasks. For now, I
> suggest that we drop this patch until I have a chance to work on it
> further.
>
> The other related patches in this series should be OK, however.

OK! Dropped. Let me know what you find out.

--b.


2008-01-26 01:03:32

by Jeff Layton

[permalink] [raw]
Subject: Re: [PATCH 097/100] NLM: have nlm_shutdown_hosts kill off all NLM RPC tasks

On Fri, 25 Jan 2008 19:31:56 -0500
"J. Bruce Fields" <[email protected]> wrote:

> On Fri, Jan 25, 2008 at 06:55:48PM -0500, Jeff Layton wrote:
> > On Fri, 25 Jan 2008 18:17:17 -0500
> > "J. Bruce Fields" <[email protected]> wrote:
> >
> > > From: Jeff Layton <[email protected]>
> > >
> > > If we're shutting down all the nlm_hosts anyway, then it doesn't make
> > > sense to allow RPC calls to linger. Allowing them to do so can mean
> > > that the RPC calls can outlive the currently running lockd and can
> > > lead to a use after free situation.
> > >
> > > Signed-off-by: Jeff Layton <[email protected]>
> > > Reviewed-by: NeilBrown <[email protected]>
> > > Signed-off-by: J. Bruce Fields <[email protected]>
> > > ---
> > > fs/lockd/host.c | 4 +++-
> > > 1 files changed, 3 insertions(+), 1 deletions(-)
> > >
> > > diff --git a/fs/lockd/host.c b/fs/lockd/host.c
> > > index ebec009..76e4bf5 100644
> > > --- a/fs/lockd/host.c
> > > +++ b/fs/lockd/host.c
> > > @@ -379,8 +379,10 @@ nlm_shutdown_hosts(void)
> > > /* First, make all hosts eligible for gc */
> > > dprintk("lockd: nuking all hosts...\n");
> > > for (chain = nlm_hosts; chain < nlm_hosts + NLM_HOST_NRHASH;
> > > ++chain) {
> > > - hlist_for_each_entry(host, pos, chain, h_hash)
> > > + hlist_for_each_entry(host, pos, chain, h_hash) {
> > > host->h_expires = jiffies - 1;
> > > + rpc_killall_tasks(host->h_rpcclnt);
> > > + }
> > > }
> > >
> > > /* Then, perform a garbage collection pass */
> >
> > I was doing some more testing today, and noticed that the original
> > problem that this patch is intended to fix resurfaced. I think this
> > patch just changes the timing on the race somehow, but I haven't tracked
> > it down completely yet.
>
> Hm. So you're getting oopses that look like the rpc code attempting to
> make calls with an rpc_client that's been freed?
>

Right. The situation to reproduce this is a bit contrived and tricky,
but I've been able to reproduce the same panics that I was originally
seeing. The timing seems to be a bit tighter now, but I can still make
it happen.

> > There's also another problem -- it's possible for host->h_rpcclnt to be
> > NULL, and that has special meaning for rpc_killall_tasks. For now, I
> > suggest that we drop this patch until I have a chance to work on it
> > further.
> >
> > The other related patches in this series should be OK, however.
>
> OK! Dropped. Let me know what you find out.
>

Will do, thanks!

--
Jeff Layton <[email protected]>