[PVFS2-developers] pvfs2_backing_dev_info struct

[Dean Hildebrand]

It seems that somewhere between 2.6.6 and 2.6.9 the backing_dev_info
struct has lost the memory_backed element so we get the following compile
error:
  CC [M]
/mnt/auto/home/dhildeb/pnfs/pvfs2/pvfs2-1.1.0-x86_64/src/kernel/linux-2.6/inode.o
/mnt/auto/home/dhildeb/pnfs/pvfs2/pvfs2-1.1.0-x86_64/src/kernel/linux-2.6/inode.c:261:
error: unknown field `memory_backed' specified in initializer
make[3]: ***
[/mnt/auto/home/dhildeb/pnfs/pvfs2/pvfs2-1.1.0-x86_64/src/kernel/linux-2.6/inode.o]
Error 1
make[2]: ***
[_module_/mnt/auto/home/dhildeb/pnfs/pvfs2/pvfs2-1.1.0-x86_64/src/kernel/linux-2.6]
Error 2
make[1]: *** [default] Error 2
make: *** [kmod] Error 2

I've spent a little time investigating what happened, and, well, I'm not
sure.  Currently I just comment it out, but I don't know if that is the
right thing to do.

Dean

On Tue, 14 Jun 2005, Rob Ross wrote:

> Definitely use both the new release of MPICH2 and the new release of
> PVFS2.  Should work, let us know if (when? :)) it doesn't...
>
> Regards,
>
> Rob
>
> Robert Latham wrote:
> > On Mon, Jun 13, 2005 at 01:16:21PM -0400, Dean Hildebrand wrote:
> >
> >>The real question seems to be whether the following work together:
> >>OS: 2.6.X (2.6.12 in my case)
> >>Platform: x86_64
> >>MPI: mpich2-1.0.1
> >
> >
> > the MPI guys released a new version of mpich2 last week.  You should
> > use that if you can
> >
> >
> >>HDF5: hdf5-1.6.4
> >
> >
> >>Anyone know? Thanks,
> >
> >
> > We don't have an opteron cluster.  Just one small opteron box we can
> > run simple tests on.  Things *should* work, since Pete's done a lot of
> > work to make pvfs2 run well on ia64, and Brian's having better success
> > now that we fixed at least one of the opteron bugs he was seeing.
> >
> > I'd say give it a shot.  There's a good chance everything will just
> > work once you've got the software stack and benchmark built, but if it
> > doesn't, we'd love to know.
> >
> > ==rob
> >
>