[ofa-general][PATCH] mlx4_core: Synch catastrophic flow with module unload

[Roland Dreier]

 > If it gets sent to netdev, it's for a networking driver, and it says
 > "PATCH" rather than "RFC" or "please review" or "don't apply" you
 > cannot reasonably expect me to not look into applying the thing.

 > And if you're saying that patches for this device should start not
 > going through me, and the tactic to accomplish that is to move the
 > bulk of the driver into some driver/shared area, that's really weird.

Well, first of all, if a driver, networking or not, has an active
maintainer, I would expect you to give that maintainer a chance to look
at any not-totally-trivial patches affecting that driver.

But in this case, mlx4_core (as opposed to mlx4_en from the same
drivers/net/mlx4 directory) really is not a network driver -- it is a
low-level multiplexer for access to hardware that really is more
InfiniBand than ethernet (with a dash of Fibre Channel thrown in).  And
yes, I am saying that making it clearer that mlx4_core is not an
network driver by moving the source to a more appropriate place does
seem to make sense.

 > Anyways I didn't push the patch out to kernel.org yet so it's easy for
 > me to remove it.

Thanks.

 - R.