Repos for: debugging device mapper and multithreaded coreutils

[Luis R. Rodriguez]

I gave a two talks, one a session about debugging device mapper with
crash and another a BoF over multithreading coreutils. Below are repos
for those who may be interested in those notes.

Debugging device mapper
--------------------------------------
Since its a zoo to debug device mapper my talk was focused on a series
of notes / tools to help as well as some changes I wished we'd do
upstream. If its of use to others here is a repo you can clone to jump
right in with a good refresher of things to do:

https://gitlab.com/mcgrof/crash-dbg-helpers

Hannes pointed out the big eye sore:

        struct dm_table {
            int undefined__;
        }

Is now gone, so if you also hated this consider backporting this into
your releases. The respective commit is
9119fedddb5b630eec3f3e39c4e382d9f60cca07 ("dm: remove dummy dm_table
definition") by Mikulas Patocka <mpatocka at redhat.com> merged as of
v4.12. As noted, it was *added* to help RCU but its no longer needed.
Be sure to meet those requirements before removal :)  (if someone has
this list please feel free to share).

Multithreaded coreutils
---------------------------------
I have a BoF on some observations on multithreading a coreutils
binary, in this case I tried rm [0] The example I gave gave results
which the room seem to agree should not be reproducible, given the
removal was all in one directory. The reason multithreading rm in this
case is expected to *not* have helped is we have a shared directory
lock when removing files and so removal should have been effectively
serialized. Any gains I noted is though to have come from other
things, but the results are *not* expected.

During hallways / dinner talks it seems the main point was missed too
though given the exact example I gave but gains on multithreading
*other* coreutils is expected to help. Case in point seems folks would
love a multithreaded recursive cp.

[0] https://gitlab.com/mcgrof/py-coreutils-m

  Luis