svn commit: r295557 - head/sys/dev/uart

[Adrian Chadd]

On 15 February 2016 at 14:28, Bruce Evans <brde at optusnet.com.au> wrote:


>
> BTW, if arm needs the barrier, then how does it work with
> bus_space_barrier() referenced in just 25 files in all of /sys/dev?

likely the same reason it does for mips - everything's cache coherent
on hardware that has pci/pcie slots, or people just didn't notice.

For ppc they put an implicit barrier into the bus calls because too
many drivers don't do barriers (but USB does, yay) and fixing it all
was likely a terrible task. I'm doing it for the MIPS related pieces
that need it, but it's slow going.


-adrian