This method also has the advantage of not relying on lock promotion semantics, which (apparently) will make the Windoze version easier.<div><br></div><div>Thomas<br><br><div class="gmail_quote">On Thu, Aug 25, 2011 at 10:30 AM, David Henningsson <span dir="ltr"><<a href="mailto:david.henningsson@canonical.com">david.henningsson@canonical.com</a>></span> wrote:<br> <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;"><div class="im">On 08/21/2011 04:38 PM, Thomas Bushnell, BSG wrote:<br> <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"> Whoops. They need to repeat the read after obtaining the write lock and<br> only update the file if the contents are still bad in that case.<br> </blockquote> <br></div> Would a good handling of this be:<br> <br> 1) Open the cookie read-only<br> 2) read the cookie<br> 3) close file<br> 4) if we have a correct cookie, do nothing more<br> 5) if we have the wrong cookie, do the old handling unchanged: open with write lock, check the contents (again), and write if something is (still) wrong.<br> <br> <blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div class="im"> <br> On Aug 21, 2011 4:46 AM, "David Henningsson"<br> <<a href="mailto:david.henningsson@canonical.com" target="_blank">david.henningsson@canonical.<u></u>com</a><br></div><div><div></div><div class="h5"> <mailto:<a href="mailto:david.henningsson@canonical.com" target="_blank">david.henningsson@<u></u>canonical.com</a>>> wrote:<br> > On 08/19/2011 08:14 PM, Thomas Bushnell, BSG wrote:<br> >> Sorry for breaking the threading, but I only just subscribed to the list<br> >> so I can't reply properly.<br> >><br> >> I'm the origin of the patch recently posted by David Henningsson which<br> >> alters the way locking works. Maarten Bosmans had some questions I'd<br> >> like to address.<br> >><br> >> The confusing formatting of the diff in core-util.c is just unidiff<br> >> being clever. Basically I created a new function to wrap around fcntl to<br> >> share the common code between pa_lock_fd and pa_read_lock_fd.<br> >><br> >> I have no objection of course to simply defining it unconditionally and<br> >> using it always. I do not know Windows, so I was trying to make the<br> >> minimally disruptive change. I didn't know that Windows has read locks.<br> >><br> >> In Unix, promoting a read lock to a write lock converts the lock--it<br> >> does not add another lock--without releasing the readlock in the middle.<br> >><br> >> I am not wedded at all to the specific details of what the generic<br> >> functions in core-util.c do.<br> >><br> >> The root issue is as David Henningsson explained. By using an exclusive<br> >> lock, pulseaudio creates an unnecessary contention in reading the<br> >> .pulse-cookie file, and because of the less-than-ideal (but quite<br> >> unchangeable) behavior of NFSv3, this forces a thirty-second delay<br> >> anytime two pulse clients try to read the cookie at the same time.<br> >> Switching to a read lock for the read, and only using an exclusive lock<br> >> when the cookie needs to be written (a much rarer operation), avoids<br> >> this problem entirely.<br> ><br> > Hi Thomas and thanks for coming here!<br> ><br> > I have a question about the proposed handling. Assume that the cookie is<br> > wrong, and that two clients both find that out in parallel. Then they<br> > both want to promote their read lock to a write lock. What will happen<br> > in that case?<br> ><br> > --<br> > David Henningsson, Canonical Ltd.<br> > <a href="http://launchpad.net/~diwic" target="_blank">http://launchpad.net/~diwic</a><br> </div></div></blockquote><div class="HOEnZb"><div></div><div class="h5"> <br> <br> <br> -- <br> David Henningsson, Canonical Ltd.<br> <a href="http://launchpad.net/~diwic" target="_blank">http://launchpad.net/~diwic</a><br> </div></div></blockquote></div><br></div>