Re: [RFC PATCH v2] implement orangefs_readahead

[Mike Marshall <hubcap@omnibond.com>]

I added the "Four fixes for ITER_XARRAY" patch and got things
running again.

I ran the tests I had regressions on by themselves, and I still fail
generic/075, generic/112, generic/127 and generic/263.

generic/438 passes now.

I was analyzing what test 075 was doing when it was failing, and I
found it to be doing this:

/home/hubcap/xfstests-dev/ltp/fsx -d -N 1000 -S 0 -P
/home/hubcap/xfstests-dev /pvfsmnt/whatever

The above used to fail every time... now it works every time.

Progress :-).

I'm about to launch the whole suite of tests, I'll report back
on what happens later...

-Mike

On Mon, Apr 26, 2021 at 10:53 AM Mike Marshall <hubcap@omnibond.com> wrote:
>
> >> Is it easy to set up an orangefs client and server?
>
> I think it is easy to set up a test system on a single VM,
> but I do it all the time. I souped up the build details in
> Documentation/filesystems/orangefs.rst not too long
> ago, I hope it is useful.
>
> Your VM would need to be a "developer" VM,
> with all the autotools stuff and such if you build
> from source. I also worked on the configure stuff
> so that you would learn about any packages you
> lack at configure time, I hope that is also still good.
>
> I read your message about trying again with the
> "Four fixes for ITER_XARRAY" patch, I'll report
> on how that goes...
>
> -Mike
>
> On Sun, Apr 25, 2021 at 3:49 AM David Howells <dhowells@redhat.com> wrote:
> >
> > Mike Marshall <hubcap@omnibond.com> wrote:
> >
> > > >> I wonder if you should use iov_length(&iter)
> > >
> > > iov_length has two arguments. The first one would maybe be iter.iov and
> > > the second one would be... ?
> >
> > Sorry, I meant iov_iter_count(&iter).
> >
> > I'll look at the other things next week.  Is it easy to set up an orangefs
> > client and server?
> >
> > David
> >