Allwinner A31 framebuffer fails with new CCU clock bindings

[rah@settrans.net (Bob Ham)]

On Mon, 2017-02-13 at 10:52 +0100, Maxime Ripard wrote:
> On Sat, Feb 11, 2017 at 12:08:40PM +0000, Bob Ham wrote:
> > On IRC, MoeIcenowy requested the contents
> > of /sys/kernel/debug/clk/clk_summary.  I don't know how useful it will
> > be but I've put up two versions, one from the bad commit and another
> > from the previous, good commit in the bisect:
> > 
> > http://settrans.net/~rah/misc/sun6i-a31-bad-fb-clk_summary-good.txt
> > http://settrans.net/~rah/misc/sun6i-a31-bad-fb-clk_summary-bad.txt
> 
> Are you sure the good one works? None of the display clocks are
> enabled.

Definitely.  Here's clk_summary from an even earlier kernel from Debian,
4.7.8-1:

http://settrans.net/~rah/misc/sun6i-a31-bad-fb-clk_summary-good-4.7.8-1.txt

And to make super sure, he's a *screenshot* of a terminal in X running
on that kernel, showing every line with "0" in the second column:

http://settrans.net/~rah/misc/sun6i-a31-bad-fb-clk_summary-good-4.7.8-1.png

Does this imply something odd in this hardware or a problem with the
kernel?


> Could you add "clk_ignore_unused" to the kernel boot parameters and
> see if it works?

That does indeed work.  On the same kernel which produced
sun6i-a31-bad-fb-clk_summary-bad.txt, with clk_ignore_unused added to
the boot parameters, the console continues to show output all through
booting and X runs fine.  Here is the clk_summary:

http://settrans.net/~rah/misc/sun6i-a31-bad-fb-clk_summary-clk_ignore_unused.txt

Can I ask what clk_ignore_unused does that might have fixed this?


Thanks,

Bob

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 181 bytes
Desc: This is a digitally signed message part
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20170213/0eb4a29c/attachment.sig>