Re: [PATCH 1/3] ceph: add some lockdep assertions around snaprealm handling

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Mon, Jun 21, 2021 at 10:51 PM Jeff Layton <jlayton@xxxxxxxxxx> wrote:
>
> On Mon, 2021-06-21 at 22:05 +0200, Ilya Dryomov wrote:
> > On Thu, Jun 3, 2021 at 6:52 PM Jeff Layton <jlayton@xxxxxxxxxx> wrote:
> > >
> > > Turn some comments into lockdep asserts.
> > >
> > > Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
> > > ---
> > >  fs/ceph/snap.c | 16 ++++++++++++++++
> > >  1 file changed, 16 insertions(+)
> > >
> > > diff --git a/fs/ceph/snap.c b/fs/ceph/snap.c
> > > index 2a63fb37778b..bc6c33d485e6 100644
> > > --- a/fs/ceph/snap.c
> > > +++ b/fs/ceph/snap.c
> > > @@ -65,6 +65,8 @@
> > >  void ceph_get_snap_realm(struct ceph_mds_client *mdsc,
> > >                          struct ceph_snap_realm *realm)
> > >  {
> > > +       lockdep_assert_held_write(&mdsc->snap_rwsem);
> > > +
> > >         dout("get_realm %p %d -> %d\n", realm,
> > >              atomic_read(&realm->nref), atomic_read(&realm->nref)+1);
> > >         /*
> > > @@ -113,6 +115,8 @@ static struct ceph_snap_realm *ceph_create_snap_realm(
> > >  {
> > >         struct ceph_snap_realm *realm;
> > >
> > > +       lockdep_assert_held_write(&mdsc->snap_rwsem);
> > > +
> > >         realm = kzalloc(sizeof(*realm), GFP_NOFS);
> > >         if (!realm)
> > >                 return ERR_PTR(-ENOMEM);
> > > @@ -143,6 +147,8 @@ static struct ceph_snap_realm *__lookup_snap_realm(struct ceph_mds_client *mdsc,
> > >         struct rb_node *n = mdsc->snap_realms.rb_node;
> > >         struct ceph_snap_realm *r;
> > >
> > > +       lockdep_assert_held_write(&mdsc->snap_rwsem);
> > > +
> > >         while (n) {
> > >                 r = rb_entry(n, struct ceph_snap_realm, node);
> > >                 if (ino < r->ino)
> > > @@ -176,6 +182,8 @@ static void __put_snap_realm(struct ceph_mds_client *mdsc,
> > >  static void __destroy_snap_realm(struct ceph_mds_client *mdsc,
> > >                                  struct ceph_snap_realm *realm)
> > >  {
> > > +       lockdep_assert_held_write(&mdsc->snap_rwsem);
> > > +
> > >         dout("__destroy_snap_realm %p %llx\n", realm, realm->ino);
> > >
> > >         rb_erase(&realm->node, &mdsc->snap_realms);
> > > @@ -198,6 +206,8 @@ static void __destroy_snap_realm(struct ceph_mds_client *mdsc,
> > >  static void __put_snap_realm(struct ceph_mds_client *mdsc,
> > >                              struct ceph_snap_realm *realm)
> > >  {
> > > +       lockdep_assert_held_write(&mdsc->snap_rwsem);
> >
> > This one appears to be redundant since the only caller is
> > __destroy_snap_realm().
> >
> > > +
> > >         dout("__put_snap_realm %llx %p %d -> %d\n", realm->ino, realm,
> > >              atomic_read(&realm->nref), atomic_read(&realm->nref)-1);
> > >         if (atomic_dec_and_test(&realm->nref))
> > > @@ -236,6 +246,8 @@ static void __cleanup_empty_realms(struct ceph_mds_client *mdsc)
> > >  {
> > >         struct ceph_snap_realm *realm;
> > >
> > > +       lockdep_assert_held_write(&mdsc->snap_rwsem);
> >
> > This too since it boils down to calling __destroy_snap_realm().
> >
> > > +
> > >         spin_lock(&mdsc->snap_empty_lock);
> > >         while (!list_empty(&mdsc->snap_empty)) {
> > >                 realm = list_first_entry(&mdsc->snap_empty,
> > > @@ -269,6 +281,8 @@ static int adjust_snap_realm_parent(struct ceph_mds_client *mdsc,
> > >  {
> > >         struct ceph_snap_realm *parent;
> > >
> > > +       lockdep_assert_held_write(&mdsc->snap_rwsem);
> >
> > And this since ceph_lookup_snap_realm() is called right away.
> >
> > > +
> > >         if (realm->parent_ino == parentino)
> > >                 return 0;
> > >
> > > @@ -696,6 +710,8 @@ int ceph_update_snap_trace(struct ceph_mds_client *mdsc,
> > >         int err = -ENOMEM;
> > >         LIST_HEAD(dirty_realms);
> > >
> > > +       lockdep_assert_held_write(&mdsc->snap_rwsem);
> >
> > Ditto.
> >
> > Thanks,
> >
> >                 Ilya
>
> Some of these calls may be redundant, but I'd still like to keep them.
>
> The locking in this code is a mess, and this is the most reliable way
> I've found to approach cleaning it up. These all compile out unless
> you're running with lockdep enabled anyway.

Reviewed-by: Ilya Dryomov <idryomov@xxxxxxxxx>

Thanks,

                Ilya



[Index of Archives]     [CEPH Users]     [Ceph Large]     [Ceph Dev]     [Information on CEPH]     [Linux BTRFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux