RE: [PATCH v5 00/10] OMAP4 : DSS2 : HDMI support on OMAP4

[Tomi Valkeinen <tomi.valkeinen@ti.com>]

On Mon, 2011-03-14 at 03:35 -0500, Hiremath, Vaibhav wrote:
> > -----Original Message-----
> > From: linux-omap-owner@vger.kernel.org [mailto:linux-omap-
> > owner@vger.kernel.org] On Behalf Of Hiremath, Vaibhav
> > Sent: Friday, March 11, 2011 6:53 PM
> > To: Valkeinen, Tomi; K, Mythri P
> > Cc: Stephan Raue; linux-omap@vger.kernel.org
> > Subject: RE: [PATCH v5 00/10] OMAP4 : DSS2 : HDMI support on OMAP4
> > 
> > > -----Original Message-----
> > > From: linux-omap-owner@vger.kernel.org [mailto:linux-omap-
> > > owner@vger.kernel.org] On Behalf Of Valkeinen, Tomi
> > > Sent: Friday, March 11, 2011 12:55 PM
> > > To: K, Mythri P
> > > Cc: Stephan Raue; linux-omap@vger.kernel.org
> > > Subject: Re: [PATCH v5 00/10] OMAP4 : DSS2 : HDMI support on OMAP4
> > >
> > > On Fri, 2011-03-11 at 00:16 -0600, K, Mythri P wrote:
> > > > Hi Stephan,
> > > >
> > > > On Fri, Mar 11, 2011 at 5:37 AM, Stephan Raue
> > <mailinglists@openelec.tv>
> > > wrote:
> > >
> > > > > thanks, this helps to boot the kernel. but now i get:
> > > > >
> > > >
> > > > <snip>
> > > >
> > > > > see also: http://paste.pocoo.org/show/351648/
> > > > >
> > > >
> > > > I see that you kernel is not booting because of dss clk
> > > >  [    3.335601] PC is at dss_clk_disable_no_ctx+0x0/0xa4
> > > > [    3.335632] LR is at omap_dispc_register_isr+0xa4/0xcc
> > > > Tomi is this related to the clock issue you were mentioning , which
> > > > gets solved by adding a delay ?
> > >
> > > Well, I have a hack patch in my tree which adds a delay of 10us. That
> > > fixed the problem for me, but the 10us is just a random guess. It could
> > > be that it needs to be longer wait. But this could be something else
> > > also.
> > >
> > >  Tomi
> > >
> > [Hiremath, Vaibhav] Tomi, just thought of updating you,
> > 
> > The linux-omap/dss2 OMAP3EVM seems to be broken, I am trying to debug this
> > at the moment and will update about my findings.
> > 
> > Since linux-omap/master is booting up fine, it looks like one of new DSS
> > patch leading to this.
> > 
> [Hiremath, Vaibhav] I think I found the where and why the kernel is crashing but not sure about root-cause - 
> 
> The root-cause turned out to be - 
> 
> void dss_clk_enable(enum dss_clock clks)
> {
> ...
> 
> if (check_ctx && cpu_is_omap34xx() && dss_need_ctx_restore())
>        restore_all_ctx();
> ...
> }
> 
> In this case, restore never happens, if I understand correctly, I am expecting, the context must be restored when all clock (especially interface clock) is disabled. 
> In order to do this, dss_need_ctx_restore must be implemented here, which I think should be ORed with other conditions. 
> 
> if (cpu_is_omap34xx() && (check_ctx || dss_need_ctx_restore()))
> 
> (This results in kernel crash at my end)
> 
> Personally I don't know any platform is implementing this function OR how one should make use of it. I remember last time we had similar discussion and the comment came was, restore only required in case of off mode. I feel, this is not applicable here, since irrespective of retention/inactive/off mode if driver is disabling clock for the peripheral we must restore the context.

I don't know about OMAP4, but on OMAP3 the register contents are only
lost when DSS goes to OFF mode. If you just turn off the clocks and OFF
mode is not enabled in the PM, context restore is not needed. I am not
sure of the current status of OFF mode in the mainline kernel.

However, for some reason DSS works fine on my Overo board. I would
imagine that it would break also if OFF is enabled for all omap
boards...

As for get_last_off_on_transaction_id(), it seems to be called
get_context_loss_count() in the mainline kernel and returns u32, not
int. I haven't tested it, but get_last_off_on_transaction_id pointer in
the dss platform data should be set to get_context_loss_count in the
board file (or in arch/arm/plat-omap/display.c if using the dss2 tree).

 Tomi