Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Javier Martinez Canillas <javierm@redhat.com>
To:     Geert Uytterhoeven <geert@linux-m68k.org>
Cc:     linux-kernel@vger.kernel.org,
        Thomas Zimmermann <tzimmermann@suse.de>,
        Daniel Vetter <daniel.vetter@ffwll.ch>,
        Maxime Ripard <mripard@kernel.org>,
        Daniel Vetter <daniel@ffwll.ch>,
        David Airlie <airlied@gmail.com>,
        dri-devel@lists.freedesktop.org
Subject: Re: [PATCH v5 2/2] drm/ssd130x: Allocate buffer in the plane's
 .atomic_check() callback
In-Reply-To: <CAMuHMdX1ECG44BMOb5zK_N1cZCpjoKmms2p_80+_F72HH0NBEA@mail.gmail.com>
References: <20230726105433.389740-1-javierm@redhat.com>
 <20230726105433.389740-2-javierm@redhat.com>
 <CAMuHMdX1ECG44BMOb5zK_N1cZCpjoKmms2p_80+_F72HH0NBEA@mail.gmail.com>
Date:   Wed, 26 Jul 2023 14:25:15 +0200
Message-ID: <871qgulv10.fsf@minerva.mail-host-address-is-not-set>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Geert Uytterhoeven <geert@linux-m68k.org> writes:

> Hi Javier,
>
> On Wed, Jul 26, 2023 at 12:55=E2=80=AFPM Javier Martinez Canillas
> <javierm@redhat.com> wrote:
>> Drivers are not allowed to fail after drm_atomic_helper_swap_state() has
>> been called and the new atomic state is stored into the current sw state.
>>
>> Since the struct ssd130x_device .data_array is allocated in the encoder's
>> .atomic_enable callback, the operation can fail and this is after the new
>> state has been stored. So it can break an atomic mode settings assumptio=
n.
>>
>> Fix this by having custom helpers to allocate, duplicate and destroy the
>> plane state, that will take care of allocating and freeing these buffers.
>>
>> Suggested-by: Maxime Ripard <mripard@kernel.org>
>> Signed-off-by: Javier Martinez Canillas <javierm@redhat.com>
>> Acked-by: Maxime Ripard <mripard@kernel.org>
>> Tested-by: Geert Uytterhoeven <geert@linux-m68k.org>
>> ---
>>
>> Changes in v5:
>> - Add collected tags from Maxime and Geert.
>> - Update commit message to not mention the kernel oops (Geert Uytterhoev=
en).
>> - Drop Reported-by and Fixes tags (Geert Uytterhoeven).
>> - Update comment about buffer and data_array fields (Geert Uytterhoeven).
>> - Remove superfluous NULL check in ssd130x_fb_blit_rect() (Geert Uytterh=
oeven).
>> - Reset .buffer to NULL if .data_array allocation fails (Geert Uytterhoe=
ven).
>> - Inline buffer alloc/free helper functions (Geert Uytterhoeven).
>
> Thanks for the update!
>

You are welcome, and thanks for the review!

>> --- a/drivers/gpu/drm/solomon/ssd130x.c
>> +++ b/drivers/gpu/drm/solomon/ssd130x.c
>
>> @@ -535,7 +550,8 @@ static int ssd130x_update_rect(struct ssd130x_device=
 *ssd130x, struct drm_rect *
>>         return ret;
>>  }
>>
>> -static void ssd130x_clear_screen(struct ssd130x_device *ssd130x)
>> +static void ssd130x_clear_screen(struct ssd130x_device *ssd130x,
>> +                                struct ssd130x_plane_state *ssd130x_sta=
te)
>>  {
>>         struct drm_rect fullscreen =3D {
>>                 .x1 =3D 0,
>> @@ -544,21 +560,21 @@ static void ssd130x_clear_screen(struct ssd130x_de=
vice *ssd130x)
>>                 .y2 =3D ssd130x->height,
>>         };
>>
>> -       ssd130x_update_rect(ssd130x, &fullscreen);
>> +       ssd130x_update_rect(ssd130x, ssd130x_state, &fullscreen);
>
> I've just realized another issue: since 49d7d581ceaf4cf8 ("drm/ssd130x:
> Don't allocate buffers on each plane update"). this no longer
> clears the screens, but just updates the hardware with the data in
> ssd130x_device.buffer, i.e. with the last image shown.
> So this should at least clear all of ssd130x_device.buffer before
> calling ssd130x_update_rect().
>

Oh, you are right. I missed that.

> As it's a bit pointless to transpose a black image, a better fix would
> be to just clear ssd130x.data_array, and call the low-level hardware
> functions like ssd130x_update_rect() does.
>

Yeah, this is a left over when we used to allocate a buffer here and I
agree with you that calling ssd130x_update_rect() is a pointless.

We can fix this as a separate follow-up patch though if you agree.

--=20
Best regards,

Javier Martinez Canillas
Core Platforms
Red Hat