On 8/23/19 9:01 PM, Laurent Pinchart wrote:
> (And CC'ing Andrzej Hajda and Neil Armstrong as the new DRM bridge
> maintainers, as well as Boris Brezillon, to make sure they're aware of
> the problem)
>
> I would really appreciate if we could delay merging this series and
> other similar changes until we find a proper solution.
>
> On Fri, Aug 23, 2019 at 09:58:47PM +0300, Laurent Pinchart wrote:
>> Hi Hans,
>>
>> Thank you for the patch.
>>
>> On Fri, Aug 23, 2019 at 01:24:27PM +0200, Hans Verkuil wrote:
>>> Set the connector info to help userspace associate the CEC adapter
>>> with the HDMI connector.
>>>
>>> This required that the cec initialization and unregistering the
>>> CEC adapter takes place in the bridge attach and detach ops.
>>>
>>> Tested on an R-Car Koelsch board.
>>>
>>> Signed-off-by: Hans Verkuil <hverkuil-cisco@xxxxxxxxx>
>>> Tested-by: Hans Verkuil <hverkuil-cisco@xxxxxxxxx>
>>> ---
>>> drivers/gpu/drm/bridge/adv7511/adv7511_cec.c | 7 ++++++-
>>> drivers/gpu/drm/bridge/adv7511/adv7511_drv.c | 22 ++++++++++----------
>>> 2 files changed, 17 insertions(+), 12 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/bridge/adv7511/adv7511_cec.c b/drivers/gpu/drm/bridge/adv7511/adv7511_cec.c
>>> index a20a45c0b353..accf5e232396 100644
>>> --- a/drivers/gpu/drm/bridge/adv7511/adv7511_cec.c
>>> +++ b/drivers/gpu/drm/bridge/adv7511/adv7511_cec.c
>>> @@ -302,6 +302,7 @@ static int adv7511_cec_parse_dt(struct device *dev, struct adv7511 *adv7511)
>>>
>>> int adv7511_cec_init(struct device *dev, struct adv7511 *adv7511)
>>> {
>>> + struct cec_connector_info conn_info;
>>> unsigned int offset = adv7511->type == ADV7533 ?
>>> ADV7533_REG_CEC_OFFSET : 0;
>>> int ret = adv7511_cec_parse_dt(dev, adv7511);
>>> @@ -310,7 +311,8 @@ int adv7511_cec_init(struct device *dev, struct adv7511 *adv7511)
>>> goto err_cec_parse_dt;
>>>
>>> adv7511->cec_adap = cec_allocate_adapter(&adv7511_cec_adap_ops,
>>> - adv7511, dev_name(dev), CEC_CAP_DEFAULTS, ADV7511_MAX_ADDRS);
>>> + adv7511, dev_name(dev),
>>> + CEC_CAP_DEFAULTS | CEC_CAP_CONNECTOR_INFO, ADV7511_MAX_ADDRS);
>>> if (IS_ERR(adv7511->cec_adap)) {
>>> ret = PTR_ERR(adv7511->cec_adap);
>>> goto err_cec_alloc;
>>> @@ -331,6 +333,9 @@ int adv7511_cec_init(struct device *dev, struct adv7511 *adv7511)
>>> ADV7511_REG_CEC_CLK_DIV + offset,
>>> ((adv7511->cec_clk_freq / 750000) - 1) << 2);
>>>
>>> + cec_fill_conn_info_from_drm(&conn_info, &adv7511->connector);
>>> + cec_s_conn_info(adv7511->cec_adap, &conn_info);
>>
>> I'm painfully trying to decouple bridges and connectors, if we keep
>> merging patches that go in the opposite direction, I'll never manage to
>> complete this :-(
>>
>> Bridges are moving to a model where they won't create connectors
>> themselves, so any new access to drm_connector contained in a bridge
>> structure is a no-go I'm afraid (I'm replying to this patch as I know
>> this driver best, but this comment applies to the other two patches in
>> the series as well).
>>
>> Here's what I wrote in a private e-mail, regarding similar changes for
>> the omapdrm driver.
>>
>> --------
>> Please have a look at "[PATCH 00/60] drm/omap: Replace custom display
>> drivers with drm_bridge and drm_panel", available in a new version at
>>
>> git://linuxtv.org/pinchartl/media.git omapdrm/bridge/devel
>>
>> (I will post v2 soon)
>>
>> The patches show the direction the omapdrm driver is taking. The goal is
>> to decouple connectors from bridges, which I'm afraid will have an
>> impact on associating drm_connector with a CEC adapter. This should be
>> implemented through new drm_bridge operations, as bridges, when created,
>> will not create drm_connector anymore.
>>
>> I've solved a similar problem related to associating DRM connectors with
>> an I2C adapter for DDC. Please see the drm_bridge_connector_init()
>> function and how the DDC adapter is handled. Something similar could be
>> done for CEC.
>> --------
>>
>> Since then v2 has been posted ("[PATCH v2 00/50] drm/omap: Replace
>> custom display drivers with drm_bridge and drm_panel") and v3 is in
>> preparation.
>>
>> So, please, let's both go in the right direction and solve the problem
>> properly for CEC.
I don't mind waiting one kernel cycle, but not longer. This CEC feature is
ready to be rolled out, so I am not willing to wait a long time for all
of this to land. I can help test CEC once you are going to roll this out
for all drm drivers (except for the sti SoC, since I don't have any HW).
So do you think this will land for 5.5? If yes, then I'll wait.
Regards,
Hans
>>
>>> +
>>> ret = cec_register_adapter(adv7511->cec_adap, dev);
>>> if (ret)
>>> goto err_cec_register;
>>> diff --git a/drivers/gpu/drm/bridge/adv7511/adv7511_drv.c b/drivers/gpu/drm/bridge/adv7511/adv7511_drv.c
>>> index f6d2681f6927..bbcb996c4d4f 100644
>>> --- a/drivers/gpu/drm/bridge/adv7511/adv7511_drv.c
>>> +++ b/drivers/gpu/drm/bridge/adv7511/adv7511_drv.c
>>> @@ -881,14 +881,24 @@ static int adv7511_bridge_attach(struct drm_bridge *bridge)
>>> regmap_write(adv->regmap, ADV7511_REG_INT_ENABLE(0),
>>> ADV7511_INT0_HPD);
>>>
>>> + if (!ret)
>>> + ret = adv7511_cec_init(&adv->i2c_main->dev, adv);
>>> return ret;
>>> }
>>>
>>> +static void adv7511_bridge_detach(struct drm_bridge *bridge)
>>> +{
>>> + struct adv7511 *adv = bridge_to_adv7511(bridge);
>>> +
>>> + cec_unregister_adapter(adv->cec_adap);
>>> +}
>>> +
>>> static const struct drm_bridge_funcs adv7511_bridge_funcs = {
>>> .enable = adv7511_bridge_enable,
>>> .disable = adv7511_bridge_disable,
>>> .mode_set = adv7511_bridge_mode_set,
>>> .attach = adv7511_bridge_attach,
>>> + .detach = adv7511_bridge_detach,
>>> };
>>>
>>> /* -----------------------------------------------------------------------------
>>> @@ -1202,7 +1212,7 @@ static int adv7511_probe(struct i2c_client *i2c, const struct i2c_device_id *id)
>>> IRQF_ONESHOT, dev_name(dev),
>>> adv7511);
>>> if (ret)
>>> - goto err_unregister_cec;
>>> + goto err_i2c_unregister_packet;
>>> }
>>>
>>> adv7511_power_off(adv7511);
>>> @@ -1212,10 +1222,6 @@ static int adv7511_probe(struct i2c_client *i2c, const struct i2c_device_id *id)
>>> if (adv7511->type == ADV7511)
>>> adv7511_set_link_config(adv7511, &link_config);
>>>
>>> - ret = adv7511_cec_init(dev, adv7511);
>>> - if (ret)
>>> - goto err_unregister_cec;
>>> -
>>> adv7511->bridge.funcs = &adv7511_bridge_funcs;
>>> adv7511->bridge.of_node = dev->of_node;
>>>
>>> @@ -1224,10 +1230,6 @@ static int adv7511_probe(struct i2c_client *i2c, const struct i2c_device_id *id)
>>> adv7511_audio_init(dev, adv7511);
>>> return 0;
>>>
>>> -err_unregister_cec:
>>> - i2c_unregister_device(adv7511->i2c_cec);
>>> - if (adv7511->cec_clk)
>>> - clk_disable_unprepare(adv7511->cec_clk);
>>> err_i2c_unregister_packet:
>>> i2c_unregister_device(adv7511->i2c_packet);
>>> err_i2c_unregister_edid:
>>> @@ -1254,8 +1256,6 @@ static int adv7511_remove(struct i2c_client *i2c)
>>>
>>> adv7511_audio_exit(adv7511);
>>>
>>> - cec_unregister_adapter(adv7511->cec_adap);
>>> -
>>> i2c_unregister_device(adv7511->i2c_packet);
>>> i2c_unregister_device(adv7511->i2c_edid);
>>>
>>
>> --
>> Regards,
>>
>> Laurent Pinchart
>
