Thread (21 messages) flat view 21 messages, 3 authors, 2013-04-04

[PATCH v8 00/14] Tegra114 clockframework

From: Stephen Warren <hidden>
Date: 2013-03-25 22:44:23
Also in: linux-devicetree, linux-tegra, lkml

On 03/22/2013 01:05 PM, Stephen Warren wrote:
On 03/22/2013 06:39 AM, Peter De Schrijver wrote:
quoted
This is the eight version of the Tegra114 clockframework. It is based on the
next-20130320-fixed branch of
git://nv-tegra.nvidia.com/user/swarren/linux-2.6.git,
Hmmm. Basing this on v3.9-rcX would be better, since when I apply it to
the Tegra tree, that's what it will be based on. Luckily, it applies
there with trivial conflicts.
quoted
http://patchwork.ozlabs.org/patch/229972/ and
http://patchwork.ozlabs.org/patch/229978/
The series,

Tested-by: Stephen Warren <redacted>

(based on my personal next-20130318-fixed branch, without
http://patchwork.ozlabs.org/patch/229972/ applied, but instead with the
audio clocks initialized per the patch I posted yesterday to do this).

I tested: serial console, eMMC for the root filesystem, audio playback
using different pll_a rates, reboot, and shutdown.

This version also solves the hard hang I was seeing with V7 during
shutdown or reboot. Now I don't have to rely on my filesystem journal:-)
Hmm. I retract that Tested-by now.

If I apply the series on top of Tegra's for-next branch (which is
currently entirely based on v3.9-rc1) then it works fine. (Similarly, it
works fine on top of next-20130318, as I reported above).

If I merge in v3.9-rc3, it's still fine. If I merge in v3.9-rc4, I see a
hard hang during boot. I believe that "hard hang" here means that some
module's registers are being touched with the clock turned off, which
locks up the CPU/bus. Certainly, sysrq doesn't work once it's hung.

I tracked this down to some interaction with commit f7db706 "ARM:
7674/1: smp: Avoid dummy clockevent being preferred over real hardware
clock-event". So says a "git bisect" between v3.9-rc3 and v3.9-rc4,
applying this CCF V8 series at each point.

If I revert that commit, the problem is solved/hidden. I tested this in
my local dev branch that's based on next-20130325 plus this CCF V8
series, plus a bunch of other cruft.

Can you please investigate what the problem is? Thanks.
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help