From: sebastian.hesselbarth@gmail.com (Sebastian Hesselbarth)
To: linux-arm-kernel@lists.infradead.org
Subject: armada_drm clock selection - try 2
Date: Tue, 02 Jul 2013 09:21:14 +0200 [thread overview]
Message-ID: <51D27F6A.5000503@gmail.com> (raw)
In-Reply-To: <CAMLZHHRrdazrJwujCWrfzo_fcudtd4GUgMEEN2YScOAZeZ-4vA@mail.gmail.com>
On 07/02/13 03:57, Daniel Drake wrote:
> On Mon, Jul 1, 2013 at 3:48 PM, Sebastian Hesselbarth
> <sebastian.hesselbarth@gmail.com> wrote:
>> I prefer not to try to find the best clock (source) at all. Let the
>> user pass the clock name by e.g. platform_data (or DT) and just try to
>> get the requested pixclk or a integer multiple of it. You will never be
>> able to find the best clock for all use cases.
>>
>> Just figure out, if integer division is sufficient to get requested
>> pixclk and clk_set_rate otherwise. This may be useful for current mode
>> running at 148.5MHz (1080p50) and new mode at 74.25MHz (1080i50, 720p).
>
> I am not opposed to this approach, it is nice and simple, but as
> Russell points out we do additionally need to distinguish between
> clocks that are "ours" to play with, vs those that are shared with
> other devices. It would be a bad idea to try call clk_set_rate() on
> the AXI bus clock, for example, changing the clock for a whole bunch
> of other devices. This is what the is_dedicated flag is about. However
> such a flag could be easily added to the DT/platform data definition
> that you propose.
Daniel,
now I got the idea of .is_dedicated. At least for Dove, we could still
implement AXI bus clock as fixed-rate clock, so you cannot mess with it.
I am almost sure, it will hang Dove when you change it, as there are
some devices depending on it (and the correct rate of it). Moreover,
LCDCLK is dedicated to CRTC0/1 so AXICLK is the only clock not
dedicated to LCD controllers.
But I am fine with .is_dedicated - I just think we should not try to
find the best clock source but leave that decision to the user
(=developer, DTS author; not userspace user).
Sebastian
WARNING: multiple messages have this Message-ID (diff)
From: Sebastian Hesselbarth <sebastian.hesselbarth@gmail.com>
To: Daniel Drake <dsd@laptop.org>
Cc: "Jean-François Moine" <moinejf@free.fr>,
"linux@arm.linux.org.uk" <linux@arm.linux.org.uk>,
linux-arm-kernel <linux-arm-kernel@lists.infradead.org>,
dri-devel@lists.freedesktop.org
Subject: Re: armada_drm clock selection - try 2
Date: Tue, 02 Jul 2013 09:21:14 +0200 [thread overview]
Message-ID: <51D27F6A.5000503@gmail.com> (raw)
In-Reply-To: <CAMLZHHRrdazrJwujCWrfzo_fcudtd4GUgMEEN2YScOAZeZ-4vA@mail.gmail.com>
On 07/02/13 03:57, Daniel Drake wrote:
> On Mon, Jul 1, 2013 at 3:48 PM, Sebastian Hesselbarth
> <sebastian.hesselbarth@gmail.com> wrote:
>> I prefer not to try to find the best clock (source) at all. Let the
>> user pass the clock name by e.g. platform_data (or DT) and just try to
>> get the requested pixclk or a integer multiple of it. You will never be
>> able to find the best clock for all use cases.
>>
>> Just figure out, if integer division is sufficient to get requested
>> pixclk and clk_set_rate otherwise. This may be useful for current mode
>> running at 148.5MHz (1080p50) and new mode at 74.25MHz (1080i50, 720p).
>
> I am not opposed to this approach, it is nice and simple, but as
> Russell points out we do additionally need to distinguish between
> clocks that are "ours" to play with, vs those that are shared with
> other devices. It would be a bad idea to try call clk_set_rate() on
> the AXI bus clock, for example, changing the clock for a whole bunch
> of other devices. This is what the is_dedicated flag is about. However
> such a flag could be easily added to the DT/platform data definition
> that you propose.
Daniel,
now I got the idea of .is_dedicated. At least for Dove, we could still
implement AXI bus clock as fixed-rate clock, so you cannot mess with it.
I am almost sure, it will hang Dove when you change it, as there are
some devices depending on it (and the correct rate of it). Moreover,
LCDCLK is dedicated to CRTC0/1 so AXICLK is the only clock not
dedicated to LCD controllers.
But I am fine with .is_dedicated - I just think we should not try to
find the best clock source but leave that decision to the user
(=developer, DTS author; not userspace user).
Sebastian
next prev parent reply other threads:[~2013-07-02 7:21 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2013-07-01 20:30 armada_drm clock selection - try 2 Daniel Drake
2013-07-01 20:30 ` Daniel Drake
2013-07-01 21:48 ` Sebastian Hesselbarth
2013-07-01 21:48 ` Sebastian Hesselbarth
2013-07-02 1:57 ` Daniel Drake
2013-07-02 1:57 ` Daniel Drake
2013-07-02 7:21 ` Sebastian Hesselbarth [this message]
2013-07-02 7:21 ` Sebastian Hesselbarth
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=51D27F6A.5000503@gmail.com \
--to=sebastian.hesselbarth@gmail.com \
--cc=linux-arm-kernel@lists.infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.