Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1584312 > unrolled thread

Re: [PATCH] drm/bridge: analogix_dp: Don't return -EBUSY when msg->size is 0 in aux transaction

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  Re: [PATCH] drm/bridge: analogix_dp: Don't return -EBUSY when  msg->size is 0 in aux transaction Tomasz Figa <tfiga@chromium.org> - 2017-02-20 03:50 +0100
    Re: [PATCH] drm/bridge: analogix_dp: Don't return -EBUSY when  msg->size is 0 in aux transaction Zain Wang <wzz@rock-chips.com> - 2017-02-20 05:10 +0100
      Re: [PATCH] drm/bridge: analogix_dp: Don't return -EBUSY when  msg->size is 0 in aux transaction Tomasz Figa <tfiga@chromium.org> - 2017-02-20 05:30 +0100

#1584312 — Re: [PATCH] drm/bridge: analogix_dp: Don't return -EBUSY when msg->size is 0 in aux transaction

Hi Zain,

On Mon, Feb 13, 2017 at 6:27 PM, zain wang <wzz@rock-chips.com> wrote:
>
> The analogix_dp_transfer() will return -EBUSY if num_transferred is zero.
> But sometimes we will send a bare address packet to start the transaction,
> like drm_dp_i2c_xfer() show:
>         ......
>         /* Send a bare address packet to start the transaction.
>          * Zero sized messages specify an address only (bare
>          * address) transaction.
>          */
>         msg.buffer = NULL;
>         msg.size = 0;
>         err = drm_dp_i2c_do_msg(aux, &msg);
>         ......
>
> In this case, the msg->size is zero, so the num_transferred will be zero too.
> We can't return -EBUSY here, let's we return num_transferred if num_transferred
> equals msg->size.
>

Please see my question inline.

> Signed-off-by: zain wang <wzz@rock-chips.com>
> ---
>  drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c b/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c
> index 303083a..5384aca 100644
> --- a/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c
> +++ b/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c
> @@ -1162,5 +1162,5 @@ ssize_t analogix_dp_transfer(struct analogix_dp_device *dp,
>                  (msg->request & ~DP_AUX_I2C_MOT) == DP_AUX_NATIVE_READ)
>                 msg->reply = DP_AUX_NATIVE_REPLY_ACK;
>
> -       return num_transferred > 0 ? num_transferred : -EBUSY;
> +       return (num_transferred == msg->size) ? num_transferred : -EBUSY;

I might be missing something but, looking at the code, I don't see any
possibility of num_transferred ever being different than msg->size. To
be honest, it doesn't seem to even make any sense keeping the local
variable there, because msg->size can be simply always returned, as
errors are handled by jumping to aux_error label.

Best regards,
Tomasz

[toc] | [next] | [standalone]

#1584336

Hi Tomasz,

在 2017/2/20 10:40, Tomasz Figa 写道:
> Hi Zain,
>
> On Mon, Feb 13, 2017 at 6:27 PM, zain wang <wzz@rock-chips.com> wrote:
>> The analogix_dp_transfer() will return -EBUSY if num_transferred is zero.
>> But sometimes we will send a bare address packet to start the transaction,
>> like drm_dp_i2c_xfer() show:
>>          ......
>>          /* Send a bare address packet to start the transaction.
>>           * Zero sized messages specify an address only (bare
>>           * address) transaction.
>>           */
>>          msg.buffer = NULL;
>>          msg.size = 0;
>>          err = drm_dp_i2c_do_msg(aux, &msg);
>>          ......
>>
>> In this case, the msg->size is zero, so the num_transferred will be zero too.
>> We can't return -EBUSY here, let's we return num_transferred if num_transferred
>> equals msg->size.
>>
> Please see my question inline.
>
>> Signed-off-by: zain wang <wzz@rock-chips.com>
>> ---
>>   drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c b/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c
>> index 303083a..5384aca 100644
>> --- a/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c
>> +++ b/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c
>> @@ -1162,5 +1162,5 @@ ssize_t analogix_dp_transfer(struct analogix_dp_device *dp,
>>                   (msg->request & ~DP_AUX_I2C_MOT) == DP_AUX_NATIVE_READ)
>>                  msg->reply = DP_AUX_NATIVE_REPLY_ACK;
>>
>> -       return num_transferred > 0 ? num_transferred : -EBUSY;
>> +       return (num_transferred == msg->size) ? num_transferred : -EBUSY;
> I might be missing something but, looking at the code, I don't see any
> possibility of num_transferred ever being different than msg->size. To
> be honest, it doesn't seem to even make any sense keeping the local
> variable there, because msg->size can be simply always returned, as
> errors are handled by jumping to aux_error label.
Yeah, I agree with you.
The better way to fix this issue is to revert the changes 
https://patchwork.kernel.org/patch/9411711/
(returning num_transferred directly may be better here)

Maybe we can revert the changes above with some new comment.
@Sean, How do you think about Tomasz's comment?

Thanks
Zain
>
> Best regards,
> Tomasz
>
>
>

[toc] | [prev] | [next] | [standalone]

#1584340

On Mon, Feb 20, 2017 at 1:04 PM, Zain Wang <wzz@rock-chips.com> wrote:
> 在 2017/2/20 10:40, Tomasz Figa 写道:
>> On Mon, Feb 13, 2017 at 6:27 PM, zain wang <wzz@rock-chips.com> wrote:
>>> diff --git a/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c
>>> b/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c
>>> index 303083a..5384aca 100644
>>> --- a/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c
>>> +++ b/drivers/gpu/drm/bridge/analogix/analogix_dp_reg.c
>>> @@ -1162,5 +1162,5 @@ ssize_t analogix_dp_transfer(struct
>>> analogix_dp_device *dp,
>>>                   (msg->request & ~DP_AUX_I2C_MOT) == DP_AUX_NATIVE_READ)
>>>                  msg->reply = DP_AUX_NATIVE_REPLY_ACK;
>>>
>>> -       return num_transferred > 0 ? num_transferred : -EBUSY;
>>> +       return (num_transferred == msg->size) ? num_transferred : -EBUSY;
>>
>> I might be missing something but, looking at the code, I don't see any
>> possibility of num_transferred ever being different than msg->size. To
>> be honest, it doesn't seem to even make any sense keeping the local
>> variable there, because msg->size can be simply always returned, as
>> errors are handled by jumping to aux_error label.
>
> Yeah, I agree with you.
> The better way to fix this issue is to revert the changes
> https://patchwork.kernel.org/patch/9411711/
> (returning num_transferred directly may be better here)

I think it's still not enough to clean up the code completely. It
should be just enough to remove num_transferred completely and simply
return msg->size at the end.

Best regards,
Tomasz

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web