Re: [PATCH v2 2/6] of: dynamic: Refactor action prints to not use "%pOF" inside devtree_lock

[Geert Uytterhoeven]

Hi Rob,

On Fri, Aug 18, 2023 at 6:17 PM Rob Herring <robh@xxxxxxxxxx> wrote:
> On Fri, Aug 18, 2023 at 10:36 AM Geert Uytterhoeven
> <geert@xxxxxxxxxxxxxx> wrote:
> > On Sat, Aug 5, 2023 at 12:42 AM Rob Herring <robh@xxxxxxxxxx> wrote:
> > > While originally it was fine to format strings using "%pOF" while
> > > holding devtree_lock, this now causes a deadlock.  Lockdep reports:
> > >
> > >     of_get_parent from of_fwnode_get_parent+0x18/0x24
> > >     ^^^^^^^^^^^^^
> > >     of_fwnode_get_parent from fwnode_count_parents+0xc/0x28
> > >     fwnode_count_parents from fwnode_full_name_string+0x18/0xac
> > >     fwnode_full_name_string from device_node_string+0x1a0/0x404
> > >     device_node_string from pointer+0x3c0/0x534
> > >     pointer from vsnprintf+0x248/0x36c
> > >     vsnprintf from vprintk_store+0x130/0x3b4
> > >
> > > Fix this by moving the printing in __of_changeset_entry_apply() outside
> > > the lock. As the only difference in the the multiple prints is the
> > > action name, use the existing "action_names" to refactor the prints into
> > > a single print.
> > >
> > > Fixes: a92eb7621b9fb2c2 ("lib/vsprintf: Make use of fwnode API to obtain node names and separators")
> > > Reported-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
> > > Signed-off-by: Rob Herring <robh@xxxxxxxxxx>
> > > ---
> > > v5 (v2 in this series):
> > >  - Move majority of refactoring to separate patch and minimize the fix
> > >    to just moving the print out of the locked section.
> >
> > Thanks for your patch!
> >
> > > --- a/drivers/of/dynamic.c
> > > +++ b/drivers/of/dynamic.c
> >
> > > @@ -648,20 +634,17 @@ static int __of_changeset_entry_apply(struct of_changeset_entry *ce)
> > >                 }
> > >
> > >                 ret = __of_update_property(ce->np, ce->prop, &old_prop);
> > > -               if (ret) {
> > > -                       pr_err("changeset: update_property failed @%pOF/%s\n",
> > > -                               ce->np,
> > > -                               ce->prop->name);
> > > -                       break;
> > > -               }
> > >                 break;
> > >         default:
> > >                 ret = -EINVAL;
> > >         }
> > >         raw_spin_unlock_irqrestore(&devtree_lock, flags);
> > >
> > > -       if (ret)
> > > +       if (ret) {
> > > +               pr_err("changeset: apply failed: cset<%p> %-15s %pOF:%s\n",
> >
> > Printing the cset pointer will (needlessly?) complicate the EXPECT_*()
> > handling in the unit test.
>
> That's added largely because the other prints which I rework later in
> this series had them. Either printing the changeset ptr is useful or
> it isn't. I think people running the unittest and the post-processor
> can easily enough filter this out when looking at the results.
> Honestly, even I probably run it less than once a cycle.

Do you have a use for printing the pointer value?
And by default, it will be an obfuscated cookie anyway.

> > > +                      ce, action_names[ce->action], ce->np, ce->prop->name);
> >
> > This should check ce->action to avoid an out-of-bounds access beyond
> > the end of action_names[].
>
> Indeed.
>
> I think I'll add "invalid" to action_names names and then do something
> like: "(ce->action < FOO) ? ce->action : 0".

OK, zero is invalid.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds