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 for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CAJfuBxymeGJ7QtcoN+eO3CcN9+UmnuNBPf1CTAyXT6Brym02MQ@mail.gmail.com>
Date: Sun, 16 Mar 2025 15:07:51 -0600
From: jim.cromie@...il.com
To: Jim Cromie <jim.cromie@...il.com>, linux-kernel@...r.kernel.org, jbaron@...mai.com, 
	gregkh@...uxfoundation.org, ukaszb@...omium.org, 
	intel-gfx-trybot@...ts.freedesktop.org, dri-devel@...ts.freedesktop.org, 
	amd-gfx@...ts.freedesktop.org, intel-gvt-dev@...ts.freedesktop.org, 
	intel-gfx@...ts.freedesktop.org, daniel.vetter@...ll.ch, 
	tvrtko.ursulin@...ux.intel.com, jani.nikula@...el.com, 
	ville.syrjala@...ux.intel.com
Subject: Re: [PATCH 23/63] dyndbg: fail modprobe on ddebug_class_range_overlap()

Im dropping the fail-on-modprobe.

theres some unrelated test-mod touches I'll examine and separate if theyre good

On Tue, Feb 25, 2025 at 7:27 AM Louis Chauvet <louis.chauvet@...tlin.com> wrote:
>
>
>
> Le 25/01/2025 à 07:45, Jim Cromie a écrit :
> > 1. All classes used by a module (declared DYNDBG_CLASSMAP_{DEFINE,USE}
> > by module code) must share 0..62 class-id space; ie their respective
> > base,+length reservations shouldn't overlap.  Overlaps would lead to
> > unintended changes in ddebug enablements.
> >
> > Detecting these class-id range overlaps at compile-time would be ideal
> > but is not obvious how; failing at modprobe at least insures that the
> > developer sees and fixes the conflict.
> >
> > ddebug_class_range_overlap() implements the range check, accumulating
> > the reserved-ids as it examines each class.  It probably should use
> > bitmaps.
> >
> > A previous commit reworked the modprobe callchain to allow failure,
> > now call ddebug_class_range_overlap() to check when classid conflicts
> > happen, and signal that failure.
> >
> > NB: this can only happen when a module defines+uses several classmaps,
> >
> > TBD: failing modprobe is kinda harsh, maybe warn and proceed ?
> >
> > test_dynamic_debug*.ko:
> >
> > If built with -DFORCE_CLASSID_CONFLICT_MODPROBE, the modules get 2 bad
> > DYNDBG_CLASS_DEFINE declarations, into parent and the _submod.  These
> > conflict with one of the good ones in the parent (D2_CORE..etc),
> > causing the modprobe(s) to fail.  TODO: do in submod only, since fail
> > of parent prevents submod from ever trying.
> >
> > Signed-off-by: Jim Cromie <jim.cromie@...il.com>
> > ---
> > ---
> >   lib/dynamic_debug.c      | 30 ++++++++++++++++++++++++------
> >   lib/test_dynamic_debug.c | 11 ++++++++++-
> >   2 files changed, 34 insertions(+), 7 deletions(-)
> >
> > diff --git a/lib/dynamic_debug.c b/lib/dynamic_debug.c
> > index 8afcd4111531..8e1e087e07c3 100644
> > --- a/lib/dynamic_debug.c
> > +++ b/lib/dynamic_debug.c
> > @@ -1211,6 +1211,21 @@ static void ddebug_apply_params(const struct ddebug_class_map *cm, const char *m
> >       }
> >   }
> >
> > +static int ddebug_class_range_overlap(struct ddebug_class_map *cm,
> > +                                   u64 *reserved_ids)
> > +{
> > +     u64 range = (((1ULL << cm->length) - 1) << cm->base);
> > +
> > +     if (range & *reserved_ids) {
> > +             pr_err("[%d..%d] on %s conflicts with %llx\n", cm->base,
> > +                    cm->base + cm->length - 1, cm->class_names[0],
> > +                    *reserved_ids);
> > +             return -EINVAL;
> > +     }
> > +     *reserved_ids |= range;
> > +     return 0;
> > +}
> > +
> >   /*
> >    * scan the named array: @_vec, ref'd from inside @_box, for the
> >    * start,len of the sub-array of elements matching on ->mod_name;
> > @@ -1242,9 +1257,11 @@ static int ddebug_module_apply_class_maps(struct ddebug_table *dt,
> >       struct ddebug_class_map *cm;
> >       int i;
> >
> > -     for_subvec(i, cm, &dt->info, maps)
> > +     for_subvec(i, cm, &dt->info, maps) {
> > +             if (ddebug_class_range_overlap(cm, reserved_ids))
> > +                     return -EINVAL;
> >               ddebug_apply_params(cm, cm->mod_name);
> > -
> > +     }
> >       vpr_info("module:%s attached %d classmaps\n", dt->mod_name, dt->info.maps.len);
> >       return 0;
> >   }
> > @@ -1255,10 +1272,11 @@ static int ddebug_module_apply_class_users(struct ddebug_table *dt,
> >       struct ddebug_class_user *cli;
> >       int i;
> >
> > -     /* now iterate dt */
> > -     for_subvec(i, cli, &dt->info, users)
> > +     for_subvec(i, cli, &dt->info, users) {
> > +             if (ddebug_class_range_overlap(cli->map, reserved_ids))
> > +                     return -EINVAL;
> >               ddebug_apply_params(cli->map, cli->mod_name);
> > -
> > +     }
> >       vpr_info("module:%s attached %d classmap uses\n", dt->mod_name, dt->info.users.len);
> >       return 0;
> >   }
> > @@ -1311,11 +1329,11 @@ static int ddebug_add_module(struct _ddebug_info *di, const char *modname)
> >                       return rc;
> >               }
> >       }
> > +
> >       mutex_lock(&ddebug_lock);
> >       list_add_tail(&dt->link, &ddebug_tables);
> >       mutex_unlock(&ddebug_lock);
> >
> > -
>
> Hi Jim,
>
> Strange line issues, can you squash it with the correct patch?
>
> Reviewed-by: Louis Chauvet <louis.chauvet@...tlin.com>
>
> Thanks,
>
> >       if (dt->info.users.len) {
> >               rc = ddebug_module_apply_class_users(dt, &reserved_ids);
> >               if (rc)
> > diff --git a/lib/test_dynamic_debug.c b/lib/test_dynamic_debug.c
> > index b1555b0a2bb1..74b98adc4ed0 100644
> > --- a/lib/test_dynamic_debug.c
> > +++ b/lib/test_dynamic_debug.c
> > @@ -81,7 +81,7 @@ enum cat_disjoint_bits {
> >       D2_DRMRES };
> >
> >   /* numeric verbosity, V2 > V1 related */
> > -enum cat_level_num { V0 = 14, V1, V2, V3, V4, V5, V6, V7 };
> > +enum cat_level_num { V0 = 16, V1, V2, V3, V4, V5, V6, V7 };
> >
> >   /* recapitulate DRM's parent(drm.ko) <-- _submod(drivers,helpers) */
> >   #if !defined(TEST_DYNAMIC_DEBUG_SUBMOD)
> > @@ -90,6 +90,7 @@ enum cat_level_num { V0 = 14, V1, V2, V3, V4, V5, V6, V7 };
> >    * classmaps on the client enums above, and then declares the PARAMS
> >    * ref'g the classmaps.  Each is exported.
> >    */
> > +
> >   DYNDBG_CLASSMAP_DEFINE(map_disjoint_bits, DD_CLASS_TYPE_DISJOINT_BITS,
> >                      D2_CORE,
> >                      "D2_CORE",
> > @@ -113,6 +114,14 @@ DYNDBG_CLASSMAP_DEFINE(map_level_num, DD_CLASS_TYPE_LEVEL_NUM,
> >   DYNDBG_CLASSMAP_PARAM(disjoint_bits, p);
> >   DYNDBG_CLASSMAP_PARAM(level_num, p);
> >
> > +#ifdef FORCE_CLASSID_CONFLICT_MODPROBE
> > +/*
> > + * Enable with -Dflag on compile to test overlapping class-id range
> > + * detection.  This should break on modprobe.
> > + */
> > +DYNDBG_CLASSMAP_DEFINE(classid_range_conflict, 0, D2_CORE + 1, "D3_CORE");
> > +#endif
> > +
> >   #else /* TEST_DYNAMIC_DEBUG_SUBMOD */
> >
> >   /*
>
> --
> Louis Chauvet, Bootlin
> Embedded Linux and Kernel engineering
> https://bootlin.com
>

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ