On Tue, Jul 12, 2022 at 01:22:40PM +0530, Viresh Kumar wrote: > On 11-07-22, 18:40, Johan Hovold wrote: > > This break OPP parsing on SC8280XP and hence cpufreq and other things: > > > > [ +0.010890] cpu cpu0: _opp_add_static_v2: opp key field not found > > [ +0.000019] cpu cpu0: _of_add_opp_table_v2: Failed to add OPP, -19 > > [ +0.000060] cpu cpu0: _opp_is_duplicate: duplicate OPPs detected. Existing: freq: 300000000, volt: 576000, enabled: 1. New: freq: 403200000, volt: 576000, enabled: 1 > > [ +0.000030] cpu cpu0: _opp_is_duplicate: duplicate OPPs detected. Existing: freq: 300000000, volt: 576000, enabled: 1. New: freq: 499200000, volt: 576000, enabled: 1 > > ... > > > > I just did a rebase on next-20220708 and hit this. > > > > I've narrowed it down to _read_rate() now returning -ENODEV since > > opp_table->clk_count is zero. > > > > Similar to what was reported for tegra for v1: > > > > https://lore.kernel.org/all/58cc8e3c-74d4-e432-8502-299312a1f15e@xxxxxxxxxxxxx/ > > > > I don't have time to look at this any more today, but it would we nice > > if you could unbreak linux-next. > > > > Perhaps Bjorn or Mani can help with further details, but this doesn't > > look like something that is specific to SC8280XP. > > It is actually. This is yet another corner case, Tegra had one as > well. I literally meant that it does not appear to be SC8280XP specific. Bjorn reported seeing similar problems on multiple Qualcomm SoCs. > I have tried to understand the Qcom code / setup to best of my > abilities, and the problem as per me is that qcom-cpufreq-hw doesn't > provide a clk to the OPP core, which breaks it after the new updates > to the OPP core. I believe following will solve it. Can someone please > try this ? I will then merge it with the right commit. > > diff --git a/drivers/opp/core.c b/drivers/opp/core.c > index 666e1ebf91d1..4f4a285886fa 100644 > --- a/drivers/opp/core.c > +++ b/drivers/opp/core.c > @@ -1384,6 +1384,20 @@ static struct opp_table *_update_opp_table_clk(struct device *dev, > } > > if (ret == -ENOENT) { > + /* > + * There are few platforms which don't want the OPP core to > + * manage device's clock settings. In such cases neither the > + * platform provides the clks explicitly to us, nor the DT > + * contains a valid clk entry. The OPP nodes in DT may still > + * contain "opp-hz" property though, which we need to parse and > + * allow the platform to find an OPP based on freq later on. > + * > + * This is a simple solution to take care of such corner cases, > + * i.e. make the clk_count 1, which lets us allocate space for > + * frequency in opp->rates and also parse the entries in DT. > + */ > + opp_table->clk_count = 1; > + > dev_dbg(dev, "%s: Couldn't find clock: %d\n", __func__, ret); > return opp_table; > } This looks like a hack. And it also triggers a bunch of new warning when opp is trying to create debugfs entries for an entirely different table which now gets clk_count set to 1: [ +0.000979] cx: _update_opp_table_clk: Couldn't find clock: -2 [ +0.000022] debugfs: Directory 'opp:0' with parent 'cx' already present! [ +0.000004] debugfs: Directory 'opp:0' with parent 'cx' already present! [ +0.000004] debugfs: Directory 'opp:0' with parent 'cx' already present! [ +0.000003] debugfs: Directory 'opp:0' with parent 'cx' already present! [ +0.000003] debugfs: Directory 'opp:0' with parent 'cx' already present! [ +0.000003] debugfs: Directory 'opp:0' with parent 'cx' already present! [ +0.000003] debugfs: Directory 'opp:0' with parent 'cx' already present! [ +0.000003] debugfs: Directory 'opp:0' with parent 'cx' already present! [ +0.000003] debugfs: Directory 'opp:0' with parent 'cx' already present! This is for the rpmhpd whose opp table does not have either opp-hz or clocks (just opp-level). The above unbreaks cpufreq though. Johan
