[RFC] Adding a Rados block driver to bhyve

[Conrad Meyer]

On Tue, Mar 10, 2020 at 9:28 AM Willem Jan Withagen <wjw at digiware.nl> wrote:
> >>         problem that libblock_rbd.so is stripped in such a way that
> >>         the symbol I need is removed.
>
> >     So either I'm doing it the wrong way, like special options on the
> >     symbols oid.
>
> > However, I think there's something wrong with your
> > library, too.  The library should be usable even if it's stripped.

Yes, strip only removes local symbols (.symtab), not exported ones (.dynsym).

> I checked with objdump, and the symbol that I need is definitly not
> present in the stripped version.

How are you defining the symbol intended for export?  Is the symbol a
function or data?  Does the compiler flag -fvisibility=hidden get
used?  Which symbol is missing and what are the symptoms?

> And it does not really matter if I declare it static or not.

Not declaring it "static" is necessary, if not sufficient.

Looking at your code on github, here are some issues:

* In block_if.h, you define an object blocklocal_backend.  This is a
header, and every compilation unit that pulls in the header will get
its own copy of blocklocal_backend.  You probably want 'extern
block_backend_t blocklocal_backend;' instead.

* You SET_DECLARE block_backend_set in block_if.c, but I think it
needs to be in block_if.h.

* There is some weirdness around linker sets being removed by the
linker if they are empty, so you may want to add blockbackend_local to
the linker set in the main program.  (It's unclear to me why
blockbackend_local is treated specially regardless.)

However, I'm not quite sure why DATA_SET() in block_rbd.c is not
creating __start_set_block_backend_set / __stop_set_block_backend_set
exported symbols.  As Alan asked, please provide 'nm -D foo.so'.

Best,
Conrad