Received-SPF: pass (google.com: domain of linux-kernel+bounces-188612-linux.lists.archive=gmail.com@vger.kernel.org designates 147.75.199.223 as permitted sender) client-ip=147.75.199.223;
Precedence: bulk
MIME-Version: 1.0
References: <20240516174357.26755-1-jim.cromie@gmail.com> <20240516174357.26755-16-jim.cromie@gmail.com>
 <CALwA+NYXBgennLhnFUvaU+U871hoiQbjNK2uRaPEd6fxPr68Jg@mail.gmail.com>
 <CAJfuBxzNKvFGB3AyJyPRDALkZ9nNYvG33u6yZ4OWEbmNypotQg@mail.gmail.com> <CAJfuBxzJqgnwy1m9f-a7+Ke-g13mwQoA9NrPFAZU_Ems_fMPiw@mail.gmail.com>
In-Reply-To: <CAJfuBxzJqgnwy1m9f-a7+Ke-g13mwQoA9NrPFAZU_Ems_fMPiw@mail.gmail.com>
From: =?UTF-8?Q?=C5=81ukasz_Bartosik?= <ukaszb@chromium.org>
Date: Fri, 24 May 2024 12:38:36 +0200
Message-ID: <CALwA+Na1rg2S0Nh_AtpLWSC6m6pCAZZG=FUo727_jdj8fO__Mw@mail.gmail.com>
Subject: Re: [PATCH v8-RESEND 15/33] dyndbg-API: fix DECLARE_DYNDBG_CLASSMAP
To: jim.cromie@gmail.com
Cc: jbaron@akamai.com, gregkh@linuxfoundation.org, 
	linux-kernel@vger.kernel.org, linux@rasmusvillemoes.dk, joe@perches.com, 
	mcgrof@kernel.org, daniel.vetter@ffwll.ch, tvrtko.ursulin@linux.intel.com, 
	jani.nikula@intel.com, ville.syrjala@linux.intel.com, seanpaul@chromium.org, 
	robdclark@gmail.com, groeck@google.com, yanivt@google.com, bleung@google.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Wed, May 22, 2024 at 8:52=E2=80=AFPM <jim.cromie@gmail.com> wrote:
>
> On Tue, May 21, 2024 at 10:31=E2=80=AFAM <jim.cromie@gmail.com> wrote:
> >
> > On Tue, May 21, 2024 at 5:46=E2=80=AFAM =C5=81ukasz Bartosik <ukaszb@ch=
romium.org> wrote:
> > >
> > > On Thu, May 16, 2024 at 7:44=E2=80=AFPM Jim Cromie <jim.cromie@gmail.=
com> wrote:
> > > >
> > > > DECLARE_DYNDBG_CLASSMAP() has a design error; its usage fails a bas=
ic
> > > > K&R rule: "define once, refer many times".
> > > >
> > > > It is used across DRM core & drivers, each use re-defines the class=
map
> > > > understood by that module; and all must match for the modules to
> > > > respond together when DRM.debug categories are enabled.  This is
> > > > brittle; a maintenance foot-gun.
> > > >
> > > > Worse, it causes the CONFIG_DRM_USE_DYNAMIC_DEBUG=3DY regression; 1=
st
> > > > drm.ko loads, and dyndbg initializes its DRM.debug callsites, then =
a
> > > > drm-driver loads, but too late - it missed the DRM.debug enablement=
.
> > > >
> > > > So replace it with 2 macros:
> > > >   DYNDBG_CLASSMAP_DEFINE - invoked once from core - drm.ko
> > > >   DYNDBG_CLASSMAP_USE    - from all drm drivers and helpers.
> >
> > >
> > > Why does DYNDBG_CLASSMAP_DEFINE take "stringified" enum name-vals ?
> > >
> >
> > short version:
> >
> > real enum vals would be better - misspellings would be compiler errors
> > but doing the stringification inside the macro doesnt work,
> >  __stringify(__VA__ARGS__) produces "DRM_UT_CORE, DRM_UT_DRIVER" etc,
> > not 10 separate stringifications
> >
> > I have a patchset for later that fixes this,
> > but I didnt want to complicate this submission any further.
> > Its already touching 2 sub-systems, revising and adding API
> >
> >
> > > > -         This module registers a tracer callback to count enabled
> > > > -         pr_debugs in a 'do_debugging' function, then alters their
> > > > -         enablements, calls the function, and compares counts.
> > > > +         This module exersizes/demonstrates dyndbg's classmap API,=
 by
> > >
> > > Typo exersizes -> exercises
> > >
> >
> > yes
> >
> > > > +static void ddebug_apply_params(const struct ddebug_class_map *cm,=
 const char *modnm)
> > > > +{
> > > > +       const struct kernel_param *kp;
> > > > +#if IS_ENABLED(CONFIG_MODULES)
> > >
> > > Instead of the above maybe use "if (IS_ENABLED(CONFIG_MODULES)) {" ?
> > > This will allow to get rid of #if and #endif which make code less rea=
dable.
> > >
> >
> > will do.
> >
> > > > +       int i;
> > > > +
>
> meh, turns out the #ifdef was doing more work - hiding the use of
> struct module in the expression.
> and since the cm->mod->kp actually refers thru the struct, it must be def=
ined,
> a forward decl looks insufficient
>
>
> >> lib/dynamic_debug.c:1202:28: error: incomplete definition of type 'str=
uct module'
>     1202 |                         for (i =3D 0, kp =3D cm->mod->kp; i <
> cm->mod->num_kp; i++, kp++)
>          |                                          ~~~~~~~^
>    arch/x86/include/asm/alternative.h:108:8: note: forward declaration
> of 'struct module'
>      108 | struct module;
>          |        ^
> [jimc:dd-classmap-fix-8d 15/36] lib/dynamic_debug.c:1202:28: error:
> incomplete definition of type 'struct module'
>
>
> I'll probably revert, and stick with the #if block,
> anything else feels too subtle by half

Ah, you're right, without CONFIG_MODULES the compilation fails.
Reverting to #if statement sounds good to me then.