On Sat, 2011-07-02 at 12:13 +0300, Tomi Valkeinen wrote:

Hi Benoit,

On Fri, 2011-07-01 at 22:41 +0200, Benoit Cousson wrote:

quoted

From: Tomi Valkeinen <redacted>

Add missing DSS optional clocks to HWMOD data for OMAP4xxx.

Add HWMOD_CONTROL_OPT_CLKS_IN_RESET flag for dispc to fix dispc reset.

Signed-off-by: Tomi Valkeinen <redacted>
[b-cousson at ti.com: Remove a comment and update the subject]
Signed-off-by: Benoit Cousson <redacted>

This patch adds "fck" opt-clocks for DSS. After your patch set
for_3.1/5_auto_fck_clkdev, in which you create "fck" automatically, I
see:

[    0.218231] omap_device: omapdss_dss: fck already exist
[    0.218261] omap_device: omapdss_dss: fck already exist
[    0.218261] omap_device: omapdss_dss: sys_clk already exist
[    0.218261] omap_device: omapdss_dss: tv_clk already exist
[    0.218322] omap_device: omapdss_dss: video_clk already exist
[    0.218627] omap_device: omapdss_dispc: fck already exist
[    0.219329] omap_device: omapdss_venc: fck already exist
[    0.219635] omap_device: omapdss_dsi1: fck already exist
[    0.219970] omap_device: omapdss_dsi2: fck already exist
[    0.220336] omap_device: omapdss_hdmi: fck already exist

So I think the "fck" clocks in this patch should be removed. I only
added them for compatibility, as I didn't have the automatic "fck"
creation patch in my tree.

Even after removing the "fck"s from the hwmod data, I see:

[    0.218200] omap_device: omapdss_dss: fck already exist
[    0.218231] omap_device: omapdss_dss: sys_clk already exist
[    0.218231] omap_device: omapdss_dss: tv_clk already exist
[    0.218292] omap_device: omapdss_dss: video_clk already exist

I think those come from the clock44xxx_data.c. I believe those are more
or less exact duplicates from the driver's point of view? If so, I guess
the ones in clockdata could be removed after adding those to hwmod data?

I also see that the main_clk for all dss hwmods is dss_dss_clk. Is that
correct? In some earlier version the clocks were a bit different, and,
at least according to TRM pictures, correct. For example, rfbi didn't
have a main_clk at all, hdmi's main_clk was dss_48mhz_clk etc. Was that
changed on purpose?

I've tested all the branches from 1 to 7 on OMAP4 SDP, and I'm able to
boot all of them (branch 7 needed merging the mmc branch), and I was
able to start the dss driver. I also merged my latest dss pmruntime
branch, which worked fine.

I still need to test OMAP2/3, and the different outputs on OMAP4 (only
DSI is tested so far).

 Tomi