[PATCH] OMAP3: OPP: Test for IVA subsystem before attempting to add IVA OPP

[Nishanth Menon]

On 10:26-20120316, Maximilian Schwerin wrote:
[...]
> > >>> +
> > >>> +             if ((strcmp(opp_def->hwmod_name,"iva") == 
> > 0) && !omap3_has_iva())
> > >>> +                     continue;
> > >>> +
> > >>>               oh = omap_hwmod_lookup(opp_def->hwmod_name);
> > >>>               if (!oh || !oh->od) {
> > >>>                       pr_warn("%s: no hwmod or odev for %s, [%d] "
> > >>
> > >> Wouldn't the one-liner below do the same thing?
> > >>
> > >> Actually, your patch makes it less noisy at boot time, avoiding the
> > >> pr_warn(), so is probably better.
> > >
> > > The only issue i have with current patch is that it 
> > focusses to solve
> > > a specific device IVA.
> > > we could have similar variances if we had SGX/AESS device etc
> > > registered in the common
> > > table. a generic solution might be preferable - could we reduce the
> > > severity of pr_warn to pr_debug and do a continue instead?
> > 
> > I agree, that would be a better generic solution.
> > 
> > Kevin
> >
> 
> This is a pragmatic and simple solution for a well understood problem with no side effects. Why not fix the problem now and do the generic solution later on? 
> 
> I'm not a fulltime kernel dev. I have about two weeks to get my new board out to my customer... Every time I set up a new board, I have to fix problems using known patches that are sometimes years old. Every patch I have to find costs me hours of time I really don't have. 
> 
> Just my two cents (euro cents of course :-), Maximilian
ok, so lets fix it generically - here is the patch for it. Let us know
if this works for you


>From 5275d09c9f1a16c8f0814745e1c313c6cca049f6 Mon Sep 17 00:00:00 2001
From: Nishanth Menon <nm at ti.com>
Date: Fri, 16 Mar 2012 09:13:24 -0500
Subject: [PATCH] OMAP2+: OPP: allow OPP enumeration to continue if device is not present

On platforms such as OMAP3, certain variants may not have IVA, SGX
or some specific component. We currently have a check to aid fixing
wrong population of OPP entries for issues such as typos. This however
causes a conflict with valid requirement where the SoC variant does
not actually have the module present.

So, reduce the severity of the print to a debug statement and skip
registering that specific OPP, but continue down the list.

Reported-by: Steve Sakoman <steve at sakoman.com>
Reported-by: Maximilian Schwerin <mvs at tigris.de>
Signed-off-by: Nishanth Menon <nm at ti.com>
---
 arch/arm/mach-omap2/opp.c |    4 ++--
 1 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/arch/arm/mach-omap2/opp.c b/arch/arm/mach-omap2/opp.c
index 9262a6b..de6d464 100644
--- a/arch/arm/mach-omap2/opp.c
+++ b/arch/arm/mach-omap2/opp.c
@@ -64,10 +64,10 @@ int __init omap_init_opp_table(struct omap_opp_def *opp_def,
 		}
 		oh = omap_hwmod_lookup(opp_def->hwmod_name);
 		if (!oh || !oh->od) {
-			pr_warn("%s: no hwmod or odev for %s, [%d] "
+			pr_debug("%s: no hwmod or odev for %s, [%d] "
 				"cannot add OPPs.\n", __func__,
 				opp_def->hwmod_name, i);
-			return -EINVAL;
+			continue;
 		}
 		dev = &oh->od->pdev->dev;
 
-- 
1.7.0.4


-- 
Regards,
Nishanth Menon