From: Bob Peterson Date: Fri, 11 Sep 2020 16:08:30 -0400 (EDT) Subject: [Cluster-devel] [GFS2 PATCH] gfs2: truncate glock address space pages during evict In-Reply-To: References: <454279735.16583317.1599751575623.JavaMail.zimbra@redhat.com> <883057823.16586032.1599751594154.JavaMail.zimbra@redhat.com> Message-ID: <1182259322.16848570.1599854910130.JavaMail.zimbra@redhat.com> List-Id: To: cluster-devel.redhat.com MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit ----- Original Message ----- > Bob, > > > not sure why, but this patch breaks xfstests generic/311 and generic/467. Hm, that's odd. Can you tell me how it misbehaves? fwiw, both 311 and 467 work for me, no hangs, panics, etc. > Also, > don't we want to initialize metamapping earlier and also call > truncate_inode_pages_final on it before we detach the glock? Yes, this is a great concern now you mention it. I've created a 5-patch set that refactors function gfs2_evict_inode to actually make it readable. I'll test and post that next week. > (Mind that truncate_inode_pages_final leaves AS_EXITING set in > metamapping->flags, so we'll also need to do somthing about that.) Yes, that is another concern. I'll wait and see if Johannes W. responds to your query and we'll go from there. Regards, Bob Peterson