On 01/08/2025, Dmitry Baryshkov wrote:
> On Wed, Jan 08, 2025 at 06:13:51PM +0800, Liu Ying wrote:
>> It's ok to pass atomic check successfully if an atomic commit tries to
>> disable the display pipeline which the connector belongs to. That is,
>> when the crtc or the best_encoder pointers in struct drm_connector_state
>> are NULL, drm_bridge_connector_atomic_check() should return 0 directly.
>> Without the check against the NULL pointers, drm_default_rgb_quant_range()
>> called by drm_atomic_helper_connector_hdmi_check() would dereference
>> the NULL pointer to_match in drm_match_cea_mode().
>>
>
> [..]
>
>> [ 46.823581] pc : drm_default_rgb_quant_range+0x0/0x4c [drm]
>> [ 46.829244] lr : drm_atomic_helper_connector_hdmi_check+0xb0/0x6b0 [drm_display_helper]
>> [ 46.837293] sp : ffff80008364ba00
>
> [..]
>
>> [ 46.911984] Call trace:
>> [ 46.914429] drm_default_rgb_quant_range+0x0/0x4c [drm] (P)
>> [ 46.920106] drm_bridge_connector_atomic_check+0x20/0x2c [drm_display_helper]
>> [ 46.927275] drm_atomic_helper_check_modeset+0x488/0xc78 [drm_kms_helper]
>> [ 46.934111] drm_atomic_helper_check+0x20/0xa4 [drm_kms_helper]
>> [ 46.940063] drm_atomic_check_only+0x4b8/0x984 [drm]
>> [ 46.945136] drm_atomic_commit+0x48/0xc4 [drm]
>> [ 46.949674] drm_framebuffer_remove+0x44c/0x530 [drm]
>> [ 46.954809] drm_mode_rmfb_work_fn+0x7c/0xa0 [drm]
>> [ 46.959687] process_one_work+0x150/0x294
>> [ 46.963700] worker_thread+0x2dc/0x3dc
>> [ 46.967450] kthread+0x130/0x204
>> [ 46.970679] ret_from_fork+0x10/0x20
>> [ 46.974258] Code: d50323bf d65f03c0 52800041 17ffffe6 (b9400001)
>> [ 46.980350] ---[ end trace 0000000000000000 ]---
>
> Please trim the backtrace in a way I trimmed it. Also you can drop the
> timestamps, they don't have useful information.
Ack.
>
>>
>> Fixes: 8ec116ff21a9 ("drm/display: bridge_connector: provide atomic_check for HDMI bridges")
>> Signed-off-by: Liu Ying <victor.liu@xxxxxxx>
>> ---
>> drivers/gpu/drm/display/drm_bridge_connector.c | 5 +++++
>> 1 file changed, 5 insertions(+)
>>
>> diff --git a/drivers/gpu/drm/display/drm_bridge_connector.c b/drivers/gpu/drm/display/drm_bridge_connector.c
>> index 32108307de66..587020a3f3e8 100644
>> --- a/drivers/gpu/drm/display/drm_bridge_connector.c
>> +++ b/drivers/gpu/drm/display/drm_bridge_connector.c
>> @@ -343,6 +343,11 @@ static int drm_bridge_connector_atomic_check(struct drm_connector *connector,
>> {
>> struct drm_bridge_connector *bridge_connector =
>> to_drm_bridge_connector(connector);
>> + struct drm_connector_state *new_conn_state =
>> + drm_atomic_get_new_connector_state(state, connector);
>> +
>> + if (!new_conn_state->crtc || !new_conn_state->best_encoder)
>> + return 0;
>
> Unfortunately, this is not enough. Other drivers (e.g. sun4i) use
> drm_atomic_helper_connector_hdmi_check() at connector's atomic_check()
> function. Please move necessary checks to the helper itself. Also please
Will move the checks to drm_atomic_helper_connector_hdmi_check().
> add corresponding KUnit test: attempt to atomic_check() the unconnected
> HDMI connector, there is a test suite for the HDMI connector functions,
> but the atomic_check() for the disconnected connecor seems to be missing.
Will add a KUnit test case to test HDMI connector disablement.
>
>>
>> if (bridge_connector->bridge_hdmi)
>> return drm_atomic_helper_connector_hdmi_check(connector, state);
>> --
>> 2.34.1
>>
>
--
Regards,
Liu Ying
