Re: [libvirt] [PATCH v2] nodedev: Document the udevEventHandleThread
On Fri, Nov 02, 2018 at 01:33:32PM +0100, Andrea Bolognani wrote:
On Fri, 2018-11-02 at 07:46 -0400, John Ferlan wrote:
> > > + * NB: Usage of event based socket algorithm causes some issues with
> > > + * older platforms such as CentOS 6 in which the data socket is opened
> >
> > ^Sounds a bit too generic, as an event based algorithm is not forced to open a
> > socket without O_NONBLOCK - just put something like "older platforms'
libudev
> > opens sockets without the NONBLOCK flag which might cause issues with event
> > based algorithm" or something alike in there.
>
> * NB: Some older distros, such as CentOS 6, libudev opens sockets
> * without the NONBLOCK flag which might cause issues with event
> * based algorithm. Although the issue can be mitigated by resetting
> * priv->dataReady for each event found; however, the scheduler issues
> * would still come into play.
I haven't looked into this at all so I might be missing something
entirely obvious, but I'd just like to point out that we no longer
target RHEL 6 / CentOS 6, so any issue that's specific to those
platforms we no longer have to concern ourselves with.
Right, I pointed that out in my review for the patch which was the impetus for
this patch. But John's right here that it's better we document the reasons for
the algorithm used which might save us from NACKing any further attempts to
change the code to suit old platforms, actually we might benefit from the
description ourselves at some point in the future, so I'm fine with that.
Erik