Re: [RFC PATCH V2] Add camera (CSI) driver for MX1

[Darius Augulis <augulis.darius@gmail.com>]

>> + * struct mx1_camera_pdata - i.MX1/i.MXL camera platform data
>> + * @init:	Init board resources
>> + * @exit:	Release board resources
>> + * @mclk_10khz:	master clock frequency in 10kHz units
>> + * @flags:	MX1 camera platform flags
>> + */
>> +struct mx1_camera_pdata {
>> +	int (*init)(struct device *);
>> +	int (*exit)(struct device *);
>>     
>
> I thought the agreement was to avoid these .init() and .exit() hooks in 
> new code...
>   

Should I config board statically during system start-up?


>> +static void mx1_videobuf_queue(struct videobuf_queue *vq,
>> +						struct videobuf_buffer *vb)
>> +{
>> +	struct soc_camera_device *icd = vq->priv_data;
>> +	struct soc_camera_host *ici = to_soc_camera_host(icd->dev.parent);
>> +	struct mx1_camera_dev *pcdev = ici->priv;
>> +	struct mx1_buffer *buf = container_of(vb, struct mx1_buffer, vb);
>> +
>> +	dev_dbg(&icd->dev, "%s (vb=0x%p) 0x%08lx %d\n", __func__,
>> +		vb, vb->baddr, vb->bsize);
>> +
>> +	list_add_tail(&vb->queue, &pcdev->capture);
>>     
>
> No, you had a spinlock here and in DMA ISR in the previous version, and it 
> was correct. Without that lock the above list_add races with 
> list_del_init() in mx1_camera_wakeup().
>   

what can save and help for the spinlock on single-core system? mx3 there 
does not have spinlock.