Re: [PATCH] drm/amd/display: Pass app_tf by value rather than by reference

From: Nathan Chancellor
Date: Tue Dec 11 2018 - 17:24:38 EST


On Tue, Dec 11, 2018 at 02:07:31PM -0800, Nick Desaulniers wrote:
> On Tue, Dec 11, 2018 at 1:42 PM Nathan Chancellor
> <natechancellor@xxxxxxxxx> wrote:
> >
> > On Tue, Dec 11, 2018 at 01:25:00PM -0800, Nick Desaulniers wrote:
> > > On Mon, Dec 10, 2018 at 3:42 PM Nathan Chancellor
> > > <natechancellor@xxxxxxxxx> wrote:
> > > >
> > > > Clang warns when an expression that equals zero is used as a null
> > > > pointer constant (in lieu of NULL):
> > > >
> > > > drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm.c:4435:3:
> > > > warning: expression which evaluates to zero treated as a null pointer
> > > > constant of type 'const enum color_transfer_func *'
> > > > [-Wnon-literal-null-conversion]
> > > > TRANSFER_FUNC_UNKNOWN,
> > > > ^~~~~~~~~~~~~~~~~~~~~
> > > > 1 warning generated.
> > > >
> > > > This warning is caused by commit bb47de736661 ("drm/amdgpu: Set FreeSync
> > > > state using drm VRR properties") and it could be solved by using NULL
> > > > instead of TRANSFER_FUNC_UNKNOWN or casting TRANSFER_FUNC_UNKNOWN as a
> > > > pointer. However, after looking into it, there doesn't appear to be a
> > > > good reason to pass app_tf by reference as it is never mutated along the
> > > > way. This is the only code path in which app_tf is used:
> > > >
> > > > mod_freesync_build_vrr_infopacket ->
> > > > build_vrr_infopacket_v2 ->
> > > > build_vrr_infopacket_fs2_data
> > > >
> > > > Neither mod_freesync_build_vrr_infopacket or build_vrr_infopacket_v2
> > > > modify app_tf's value and build_vrr_infopacket_fs2_data expects just
> > > > the value so we can avoid dereferencing anything by just passing in
> > > > app_tf's value to mod_freesync_build_vrr_infopacket and
> > > > build_vrr_infopacket_v2.
> > > >
> > > > There is no functional change because build_vrr_infopacket_fs2_data
> > > > doesn't do anything if TRANSFER_FUNC_UNKNOWN is passed to it, the same
> > > > as not calling build_vrr_infopacket_fs2_data at all like before this
> > > > change when NULL was used for app_tf.
> > >
> > > Nathan,
> > > Thanks for sending this patch. I was hoping to provide review sooner,
> > > but have been quite busy lately.
> > >
> >
> > Late review is better than no review, I appeciate you taking the time to
> > do this!
> >
> > > Yeah, especially for LP64 targets, the pointer to enum is larger than
> > > just the enum, and if it's not being updated ("in/out paramter")
> > > there's no need to pass by pointer.
> > >
> >
> > Thanks for confirming!
> >
> > > >
> > > > Signed-off-by: Nathan Chancellor <natechancellor@xxxxxxxxx>
> > > > ---
> > > > drivers/gpu/drm/amd/display/modules/freesync/freesync.c | 7 +++----
> > > > drivers/gpu/drm/amd/display/modules/inc/mod_freesync.h | 2 +-
> > > > 2 files changed, 4 insertions(+), 5 deletions(-)
> > > >
> > > > diff --git a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c
> > > > index 620a171620ee..520665a9d81a 100644
> > > > --- a/drivers/gpu/drm/amd/display/modules/freesync/freesync.c
> > > > +++ b/drivers/gpu/drm/amd/display/modules/freesync/freesync.c
> > > > @@ -656,7 +656,7 @@ static void build_vrr_infopacket_v1(enum signal_type signal,
> > > >
> > > > static void build_vrr_infopacket_v2(enum signal_type signal,
> > > > const struct mod_vrr_params *vrr,
> > > > - const enum color_transfer_func *app_tf,
> > > > + enum color_transfer_func app_tf,
> > > > struct dc_info_packet *infopacket)
> > > > {
> > > > unsigned int payload_size = 0;
> > > > @@ -664,8 +664,7 @@ static void build_vrr_infopacket_v2(enum signal_type signal,
> > > > build_vrr_infopacket_header_v2(signal, infopacket, &payload_size);
> > > > build_vrr_infopacket_data(vrr, infopacket);
> > > >
> > > > - if (app_tf != NULL)
> > > > - build_vrr_infopacket_fs2_data(*app_tf, infopacket);
> > > > + build_vrr_infopacket_fs2_data(app_tf, infopacket);
> > > >
> > > > build_vrr_infopacket_checksum(&payload_size, infopacket);
> > > >
> > > > @@ -676,7 +675,7 @@ void mod_freesync_build_vrr_infopacket(struct mod_freesync *mod_freesync,
> > > > const struct dc_stream_state *stream,
> > > > const struct mod_vrr_params *vrr,
> > > > enum vrr_packet_type packet_type,
> > > > - const enum color_transfer_func *app_tf,
> > > > + enum color_transfer_func app_tf,
> > > > struct dc_info_packet *infopacket)
> > > > {
> > > > /* SPD info packet for FreeSync */
> > > > diff --git a/drivers/gpu/drm/amd/display/modules/inc/mod_freesync.h b/drivers/gpu/drm/amd/display/modules/inc/mod_freesync.h
> > > > index 949a8b62aa98..063af6258fd9 100644
> > > > --- a/drivers/gpu/drm/amd/display/modules/inc/mod_freesync.h
> > > > +++ b/drivers/gpu/drm/amd/display/modules/inc/mod_freesync.h
> > > > @@ -145,7 +145,7 @@ void mod_freesync_build_vrr_infopacket(struct mod_freesync *mod_freesync,
> > > > const struct dc_stream_state *stream,
> > > > const struct mod_vrr_params *vrr,
> > > > enum vrr_packet_type packet_type,
> > > > - const enum color_transfer_func *app_tf,
> > > > + enum color_transfer_func app_tf,
> > >
> > > Don't you need to update the callsite of `mod_freesync_build_vrr_infopacket` in
> > >
> > > drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c#4949:
> > >
> > > - NULL,
> > > + transfer_func_unknown,
> > >
> >
> > That change in commit bb47de736661 ("drm/amdgpu: Set FreeSync state
> > using drm VRR properties") in -next is what prompted this patch (the
> > warning in the commit message is not present in mainline):
>
> Ah! Sorry, I was looking at mainline. I should have noticed
> bb47de736661 wasn't there.
>
> Shouldn't that change fail to compile, as transfer_func_unknown is an
> `enum color_transfer_func` value, but
> mod_freesync_build_vrr_infopacket takes a *pointer* to an `enum
> color_transfer_func` value?
> https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tree/drivers/gpu/drm/amd/display/modules/freesync/freesync.c#n675
>

Unfortunately, no. GCC does not warn like Clang does when an expression
that equals zero is passed to a function requiring a pointer (it just
silently treats it as NULL, whereas Clang audibly complains).

They both do when the expression is something other than zero.

https://godbolt.org/z/goo91G

Nathan

> >
> > mod_freesync_build_vrr_infopacket(
> > dm->freesync_module,
> > new_stream,
> > &vrr,
> > packet_type_vrr,
> > transfer_func_unknown,
> > &vrr_infopacket);
> >
> > > Maybe at that point the `if (app_tf != NULL)` could be replaced with
> > > `if (app_tf != transfer_func_unknown)` hoisted from
> > > `build_vrr_infopacket_fs2_data`? (There's only one caller of
> > > `build_vrr_infopacket_fs2_data` today, maybe fine to leave the
> > > unconditional call and check).
> > >
> >
> > Hmmm that's not unreasonable I suppose. I guess it depends on if
> > build_vrr_infopacket_fs2_data could ever be called from outside of
> > build_vrr_infopacket_v2; if it can, it makes sense to leave the
> > conditional check for 'app_tf != transfer_func_unknown' in
> > build_vrr_infopacket_fs2_data and leaving the unconditional call
> > to it in build_vrr_infopacket_v2 (since app_tf is no longer a pointer,
> > no need to check against NULL).
> >
> > I'm happy to do a v2 if the maintainers feel strongly about it, thank
> > you for bringing that up.
>
> Don't worry about it, I think it's fine.
> Reviewed-by: Nick Desaulniers <ndesaulniers@xxxxxxxxxx>
>
> --
> Thanks,
> ~Nick Desaulniers