linux-kernel - Re: [PATCH 3/7] OPP: Rework _set_required_devs() to manage a single device per call

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAPDyKFqY_mNnaT8j4vCXxYtARkGb_bkvcwKkyXcLPwW+gutO8Q@mail.gmail.com>
Date: Thu, 11 Jul 2024 12:19:53 +0200
From: Ulf Hansson <ulf.hansson@...aro.org>
To: Viresh Kumar <viresh.kumar@...aro.org>
Cc: Viresh Kumar <vireshk@...nel.org>, Nishanth Menon <nm@...com>, Stephen Boyd <sboyd@...nel.org>, 
	Bjorn Andersson <andersson@...nel.org>, Konrad Dybcio <konrad.dybcio@...aro.org>, 
	Nikunj Kela <nkela@...cinc.com>, Prasad Sodagudi <psodagud@...cinc.com>, 
	Thierry Reding <thierry.reding@...il.com>, Jonathan Hunter <jonathanh@...dia.com>, 
	linux-pm@...r.kernel.org, linux-arm-kernel@...ts.infradead.org, 
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 3/7] OPP: Rework _set_required_devs() to manage a single
 device per call

On Wed, 26 Jun 2024 at 08:33, Viresh Kumar <viresh.kumar@...aro.org> wrote:
>
> On 19-06-24, 16:08, Ulf Hansson wrote:
> > @@ -2494,36 +2495,68 @@ static int _opp_set_required_devs(struct opp_table *opp_table,
> >               return -EINVAL;
> >       }
> >
> > -     /* Another device that shares the OPP table has set the required devs ? */
> > -     if (opp_table->required_devs[0])
> > -             return 0;
> > +     /* Genpd core takes care of propagation to parent genpd */
> > +     if (opp_table->is_genpd) {
>
> A genpd can have non-genpd devices in the required OPPs and so this
> isn't sufficient. What we were ignoring earlier was genpd having
> another genpd as required opp.

Unless I am mistaken, I don't think that is a scenario we should care
about here.

_opp_set_required_dev() is being called for a device that is about to
be attached to its corresponding genpd.

Yes, in some cases, we attach a genpd provider's device to its
genpd-parent, but that is not to control the required-opps.

>
> > +             dev_err(dev, "%s: Operation not supported for genpds\n", __func__);
> > +             return -EOPNOTSUPP;
> > +     }
> >
> >       for (i = 0; i < opp_table->required_opp_count; i++) {
> > -             /* Genpd core takes care of propagation to parent genpd */
> > -             if (required_devs[i] && opp_table->is_genpd &&
> > -                 opp_table->required_opp_tables[i]->is_genpd) {
> > -                     dev_err(dev, "%s: Operation not supported for genpds\n", __func__);
> > -                     return -EOPNOTSUPP;
> > -             }
> > +             struct opp_table *table = opp_table->required_opp_tables[i];
> > +
> > +             /*
> > +              * The OPP table should be available at this point. If not, it's
> > +              * not the one we are looking for.
> > +              */
> > +             if (IS_ERR(table))
> > +                     continue;
> > +
> > +             /* Move to the next available index. */
> > +             if (opp_table->required_devs[i])
> > +                     continue;
> >
> > -             opp_table->required_devs[i] = required_devs[i];
> > +             /*
> > +              * We need to compare the nodes for the OPP tables, rather than
> > +              * the OPP tables themselves, as we may have separate instances.
> > +              */
> > +             if (required_opp_table->np == table->np) {
> > +
>
> We don't keep such empty lines in OPP core generally at this place.

Yep, let me drop it!

>
> > +                     /* Cross check the OPP tables and fix it if needed. */
>
> Copy the bigger comment from_opp_attach_genpd() here too. It helps
> understanding why required_opp_tables entry is getting replaced.

Right, makes sense!

>
> > +                     if (required_opp_table != table) {
> > +                             dev_pm_opp_put_opp_table(table);
> > +                             _get_opp_table_kref(required_opp_table);
> > +                             opp_table->required_opp_tables[i] = required_opp_table;
> > +                     }
> > +
> > +                     opp_table->required_devs[i] = required_dev;
> > +
> > +                     /*
> > +                      * Add the required_dev as a user of the OPP table, so
> > +                      * we can call dev_pm_opp_set_opp() on it directly.
> > +                      */
> > +                     if (!_add_opp_dev(required_dev, required_opp_table)) {
> > +                             dev_err(dev, "Failed to add the device to the required OPP table\n");
> > +                             return -ENOMEM;
> > +                     }
> > +
> > +                     return i;
> > +             }
> >       }
>
> --
> viresh

Kind regards
Uffe