From: David Teigland Date: Tue, 2 Mar 2010 20:54:53 -0600 Subject: [Ocfs2-devel] [PATCH] ocfs2: Ignore BASTs fired after an AST for a drop lock] In-Reply-To: <20100226231245.GE23730@mail.oracle.com> References: <4B720386.7010301@oracle.com> <20100210193540.GA5062@redhat.com> <4B73155C.2000602@oracle.com> <20100210203527.GB5062@redhat.com> <20100223214728.GH31495@redhat.com> <20100226231245.GE23730@mail.oracle.com> Message-ID: <20100303025453.GB14209@redhat.com> List-Id: MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: ocfs2-devel@oss.oracle.com On Fri, Feb 26, 2010 at 03:12:45PM -0800, Joel Becker wrote: > On Tue, Feb 23, 2010 at 01:47:28PM -0800, David Teigland wrote: > > The dlm is indeed sending the bast before the cast. With the fix > > for that (attached), alternate is running fine again. I'll now get > > back to doing some long running tests of make_panic and alternate. > > > > I actually need to split up the fix into a couple different patches > > and will then queue them up for the coming merge window. > > I saw your pull request. Is it safe to say ocfs2 + fs/dlm + > the bast/cast fixes now passes all your tests? It's much improved, but I still seem to get one wrong or missing bast eventually. I'll dig back into that in a couple weeks. My hunch is that it's related to a lock being canceled. Dave