From mboxrd@z Thu Jan 1 00:00:00 1970 From: Tariq Saeed Date: Fri, 08 Jan 2016 14:44:29 -0800 Subject: [Ocfs2-devel] [PATCH] revert to using ocfs2_acl_chmod to avoid inode cluster lock hang In-Reply-To: <20160108014556.GD819@wotan.suse.de> References: <1452047231-30569-1-git-send-email-tariq.x.saeed@oracle.com> <20160107225507.GB819@wotan.suse.de> <568EF987.9080904@oracle.com> <20160108014556.GD819@wotan.suse.de> Message-ID: <56903BCD.6050806@oracle.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 01/07/2016 05:45 PM, Mark Fasheh wrote: >> 2. ocfs2_reflink -> ocfs2_init_security_and_acl -> ocfs2_iop_set_acl > Could you elaborate for me on the problem you found there? The problem here is dropping dir lock in posix_acl_create after getting default_acl and acl. These two can be changed by another node by the time we get around to using them in ocfs2_init_security_and_acl. The old code in uek3 is mindful of this. In that code, ocfs2_init_security_and_acl gets the dir lock and keeps it until the acls of new node are set. Thanks -Tariq