[GIT PULL] ARM: mvebu: SoC changes for v3.17

[thomas.petazzoni@free-electrons.com (Thomas Petazzoni)]

Dear Jason Cooper,

(Adding Russell in Cc, to get the status of the "ARM: mm: add support
for HW coherent systems in PL310 cache" patch)

On Sat, 28 Jun 2014 11:07:06 -0400, Jason Cooper wrote:

> > >     - don't apply thermal quirk by default
> > >     - make Aramada 375 external abort logic a quirk
> > >     - update L2/PCIe deadlock workaround after L2CC cleanup
> > 
> > This patch is needed in 3.16, not in 3.17. They really are bug fixes,
> > fixing a complete system deadlock under PCI load.
> 
> crap.  :(  Ok, guys, please ignore this pull request and I'll send a new
> one with this done properly.
> 
> Thomas, you say 'patch', but you quote three.  Should all three be in
> mvebu/fixes?  Or, just the last one?

Yeah, sorry, I wasn't clear. If you look back at what I sent in
https://www.mail-archive.com/devicetree at vger.kernel.org/msg30473.html
([PATCHv5 0/3] ARM: implement workaround for Cortex-A9/PL310/PCIe
deadlock), there were three patches:

 PATCH 1/3, "ARM: mm: add support for HW coherent systems in PL310
 cache", for Russell. It has already been merged by Russell in
 "git-curr", as visible at
 http://www.arm.linux.org.uk/developer/patches/viewpatch.php?id=8076/1.
 I'm hoping this is scheduled for 3.16, because I need this patch for
 3.16 to avoid the PCIe deadlock.

 PATCH 2/3 and 3/3 are related, and the latter is needed to solve the
 PCIe deadlock, by using the new DT property made available in PATCH
 1/3. PATCH 2/3 is ARM: mvebu: move Armada 375 external abort logic as
 a quirk and PATCH 3/3 is ARM: mvebu: update L2/PCIe deadlock
 workaround after L2CC cleanup.

All in all, I'm hoping for the series at
https://www.mail-archive.com/devicetree at vger.kernel.org/msg30473.html
to fully show up in 3.16, because it's really related to fixing a
serious PCIe deadlock (can be triggered in under a minute of PCIe
usage).

Note that even though PCIe was supported in 3.15 for Armada 375/38x,
the bug cannot be triggered because it requires SMP to be visible, and
SMP support for 375 and 38x was only merged in 3.16. So there is no
need for these patches to be marked for 3.15-stable. But clearly they
are needed for 3.16.

Thanks!

Thomas Petazzoni
-- 
Thomas Petazzoni, CTO, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com