Date: Sat, 8 Sep 2012 19:13:48 +0300 From: Konstantin Belousov <kostikbel@gmail.com> To: Gleb Kurtsou <gleb.kurtsou@gmail.com> Cc: pho@freebsd.org, fs@freebsd.org Subject: Re: Nullfs shared lookup Message-ID: <20120908161348.GG33100@deviant.kiev.zoral.com.ua> In-Reply-To: <20120908045921.GA1419@reks> References: <20120905091854.GD33100@deviant.kiev.zoral.com.ua> <20120908045921.GA1419@reks>
next in thread | previous in thread | raw e-mail | index | archive | help
--pIqiiB6HxdjnOZqe Content-Type: text/plain; charset=us-ascii Content-Disposition: inline Content-Transfer-Encoding: quoted-printable On Fri, Sep 07, 2012 at 09:59:21PM -0700, Gleb Kurtsou wrote: > On (05/09/2012 12:18), Konstantin Belousov wrote: > > I, together with Peter Holm, developed a patch to enable shared lookups > > on nullfs mounts when lower filesystem allows the shared lookups. The l= ack > > of shared lookup support for nullfs is quite visible on any VFS-intensi= ve > > workloads which utilize path translations. In particular, it was a comp= lain > > on $dayjob which started me thinking about this issue. > >=20 > > There are two problems which prevent direct translation of shared > > lookup bit into nullfs upper mount bit: > >=20 > > 1. When vfs_lookup() calls VOP_LOOKUP() for nullfs, which passes lookup > > operation to lower fs, resulting vnode is often only shared-locked. Then > > null_nodeget() cannot instantiate covering vnode for lower vnode, since > > insmntque1() and null_hashins() require exclusive lock on the lower. > >=20 > > The solution is straightforward, if null hash failed to find pre-existi= ng > > nullfs vnode for lower vnode, the lower vnode lock is upgraded. > >=20 > > 2. (More serious). Nullfs reclaims its vnodes on deactivation. The cause > > is due to nullfs inability to detect reclamation of the lower vnode. > > Reclamation of a nullfs vnode at deactivation time prevents a reference > > to the lower vnode to become stale. > >=20 > > Unfortunately, this means that all lookups on nullfs need exclusive lock > > to instantiate upper vnode, which is never cached. > >=20 > > Solution which we propose is to add VFS notification to the upper > > filesystem about reclamation of the vnode in the lower filesystem. Now, > > vgone() calls new VFS op vfs_reclaim_lowervp() with an argument lowervp > > which is reclaimed. It is possible to register several reclamation event > > listeners, to correctly handle the case of several nullfs mounts over > > the same directory. > >=20 > > For the filesystem not having nullfs mounts over it, the overhead added= is > > a single mount interlock lock/unlock in the vnode reclamation path. > >=20 > > Benchmarks consisting of up 1K threads doing parallel stat(2) on the > > same file demonstate almost constant execution time, not depending of > > number of running threads. While without the patch, exec time between > > single-threaded run and run with 1024 threads performing the same total > > count of stat(2), differ in 6 times. > >=20 > > Somewhat problematic detail, IMO, is that nullfs reclamation procedure > > calls vput() on the lowervp vnode, temporary unlocking the vnode being > > reclaimed. This seems to be fine for MPSAFE filesystems, but not-MPSAFE > > code often put partially initialized vnode on some globally visible > > list, and later can decide that half-constructed vnode is not needed. > > If nullfs mount is created above such filesystem, then other threads > > might catch such not properly initialized vnode. Instead of trying > > to overcome this case, e.g. by recursing the lower vnode lock in > > null_reclaim_lowervp(), I decided to rely on nearby extermination of > > non-MPSAFE filesystems support. > >=20 > > I think that unionfs can also benefit from this mechanism, but I did not > > even looked at unionfs. > >=20 > > Patch is available at > > http://people.freebsd.org/~kib/misc/nullfs_shared_lookup.1.patch > > It survived stress2 torturing. > >=20 > > Comments ? >=20 > I only had a glance look at the patch, sorry it I missed something > obvious. How do we achieve propagation of rename/rm/rmdir to upper > level name cache? We don't. If you look at the nullfs vnode op table, you should note that nullfs provides a bypass function for vop_lookup. To use the name cache, filesytem shall set vop_lookup to vfs_cache_lookup, and implement vop_cachedlookup. See for instance UFS. The cache avoidance was the reason why working null_vptocnp() was a high-priority item (together with the weird locking protocol for the vop). --pIqiiB6HxdjnOZqe Content-Type: application/pgp-signature Content-Disposition: inline -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.12 (FreeBSD) iEYEARECAAYFAlBLbrwACgkQC3+MBN1Mb4iUBQCg2NbQ8sAXz9ETnQlgRJSXrhgt kxcAn274Vd6vBO92HnO+YFJM8Y8ryQcx =4yUR -----END PGP SIGNATURE----- --pIqiiB6HxdjnOZqe--
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20120908161348.GG33100>