From: Dave Wysochanski Date: Tue, 03 Feb 2009 21:28:54 -0500 Subject: [PATCH] A different implementation of --ignorelockingfailure. In-Reply-To: <871vuva9yn.fsf@eriador.mornfall.net> References: <871vuva9yn.fsf@eriador.mornfall.net> Message-ID: <1233714534.16492.10.camel@localhost.localdomain> List-Id: To: lvm-devel@redhat.com MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Nice attempt at cleanup. Not sure about this patch but I do have comments. 1. Why call it 'boottime_locking' - why not readonly_locking which is clearer? 2. If you make this change you'll also need to modify lvchange.c to take a READ lock if the request is for an availability change (vgchange has a conditional).* Otherwise, lvchange will fail when before it would succeed. 3. Introducing boottime/readonly as a "type" of locking seems out of place with the current set of locking types - file operations, cluster operations. * Then again, lvchange and vgchange seem to have different code for changing availability of an LV (?) so there may be a subtle reason lvchange always takes a WRITE lock. On Thu, 2009-01-22 at 11:53 +0100, Petr Rockai wrote: > Hi, > > two ends are addressed by this patch: getting rid of some global state and also > restricting --ignorelockingfailure to only allow what it is supposed to allow > (instead of a few hacked-in checks at some places and our belief that the users > will be well-behaved). > > Yours, > Petr. > > -- > lvm-devel mailing list > lvm-devel at redhat.com > https://www.redhat.com/mailman/listinfo/lvm-devel