From mboxrd@z Thu Jan 1 00:00:00 1970 From: Steven Rostedt Date: Mon, 27 Oct 2014 16:17:24 -0400 Subject: [Cluster-devel] [PATCH 3/7] dlm: Use seq_is_full - remove seq_printf returns In-Reply-To: <20140930103435.GC2707@dhcp128.suse.cz> References: <20140929124246.3e39dac8@gandalf.local.home> <43590057bcb83846acbbcc1fe641f792b2fb7773.1412031505.git.joe@perches.com> <20140930103435.GC2707@dhcp128.suse.cz> Message-ID: <20141027161724.0ec20325@gandalf.local.home> List-Id: To: cluster-devel.redhat.com MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Note, I've started with Joe's patches and I'm massaging them for something I can work with. On Tue, 30 Sep 2014 12:34:35 +0200 Petr Mladek wrote: > > - rv = seq_printf(s, "\"\nInvalid master %d\n", > > - res->res_nodeid); > > - if (rv) > > + seq_printf(s, "\"\nInvalid master %d\n", res->res_nodeid); > > + if (seq_is_full(s)) > > goto out; > > I would check for seq_overflow() > > Etc. There are needed many more changes if we agree on introducing > seq_is_full() and seq_overflow(). As I'm looking at this code, I'm thinking that we never really care about seq_is_full(). We only really care if seq_overflowed(), in which the contents will be discarded. Rational? Because if we break when seq_is_full(), my new logic wont throw away the result. If we break out of the function when it's full and not when it has overflowed, then we may never print out the rest of the content, as the seq_file code will still use a full buffer that hasn't overflowed. I'm thinking of switching everything to use seq_has_overflowed() and try again. Thoughts? -- Steve