linux-kernel - Re: [PATCH v2 5/5] media: mt9p031: Fix corrupted frame after restarting stream

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <b479cfef-6342-107a-1dcf-1b7b32872de0@phytec.de>
Date:   Wed, 14 Oct 2020 09:14:53 +0200
From:   Stefan Riedmüller <s.riedmueller@...tec.de>
To:     Laurent Pinchart <laurent.pinchart@...asonboard.com>
Cc:     Mauro Carvalho Chehab <mchehab@...nel.org>,
        Sakari Ailus <sakari.ailus@...ux.intel.com>,
        linux-media@...r.kernel.org, linux-kernel@...r.kernel.org,
        Dirk Bender <d.bender@...tec.de>
Subject: Re: [PATCH v2 5/5] media: mt9p031: Fix corrupted frame after
 restarting stream

Hi Laurent,

On 05.10.20 15:08, Laurent Pinchart wrote:
> Hi Stefan,
> 
> On Mon, Oct 05, 2020 at 11:28:21AM +0200, Stefan Riedmüller wrote:
>> On 02.10.20 02:05, Laurent Pinchart wrote:
>>> On Wed, Sep 30, 2020 at 12:51:33PM +0200, Stefan Riedmueller wrote:
>>>> From: Dirk Bender <d.bender@...tec.de>
>>>>
>>>> To prevent corrupted frames after starting and stopping the sensor it's
>>>
>>> s/it's/its/
>>
>> thanks, I'll fix that.
>>
>>>> datasheet specifies a specific pause sequence to follow:
>>>>
>>>> Stopping:
>>>> 	Set Pause_Restart Bit -> Set Restart Bit -> Set Chip_Enable Off
>>>>
>>>> Restarting:
>>>> 	Set Chip_Enable On -> Clear Pause_Restart Bit
>>>>
>>>> The Restart Bit is cleared automatically and must not be cleared
>>>> manually as this would cause undefined behavior.
>>>>
>>>> Signed-off-by: Dirk Bender <d.bender@...tec.de>
>>>> Signed-off-by: Stefan Riedmueller <s.riedmueller@...tec.de>
>>>> ---
>>>> No changes in v2
>>>> ---
>>>>    drivers/media/i2c/mt9p031.c | 25 +++++++++++++++++++++++++
>>>>    1 file changed, 25 insertions(+)
>>>>
>>>> diff --git a/drivers/media/i2c/mt9p031.c b/drivers/media/i2c/mt9p031.c
>>>> index d10457361e6c..d59f66e3dcf3 100644
>>>> --- a/drivers/media/i2c/mt9p031.c
>>>> +++ b/drivers/media/i2c/mt9p031.c
>>>> @@ -80,6 +80,8 @@
>>>>    #define		MT9P031_PIXEL_CLOCK_SHIFT(n)		((n) << 8)
>>>>    #define		MT9P031_PIXEL_CLOCK_DIVIDE(n)		((n) << 0)
>>>>    #define MT9P031_FRAME_RESTART				0x0b
>>>> +#define		MT9P031_FRAME_RESTART_SET		(1 << 0)
>>>> +#define		MT9P031_FRAME_PAUSE_RESTART_SET		(1 << 1)
>>>
>>> The fields are named Restart and Pause_Restart, I would drop _SET. Could
>>> you also sort them from MSB to LSB as for the other registers ? Using
>>> BIT() would be good too, although this could be done as an additional
>>> patch to convert all the existing macros.
>>
>> I'll do that. Also I will rename the register to MT9P031_RESTART and the
>> bits to MT9P031_FRAME_RESTART and MT9P031_FRAME_PAUSE_RESTART.
>>
>>>>    #define MT9P031_SHUTTER_DELAY				0x0c
>>>>    #define MT9P031_RST					0x0d
>>>>    #define		MT9P031_RST_ENABLE			1
>>>> @@ -483,9 +485,25 @@ static int mt9p031_set_params(struct mt9p031 *mt9p031)
>>>>    static int mt9p031_s_stream(struct v4l2_subdev *subdev, int enable)
>>>>    {
>>>>    	struct mt9p031 *mt9p031 = to_mt9p031(subdev);
>>>> +	struct i2c_client *client = v4l2_get_subdevdata(subdev);
>>>> +	int val;
>>>>    	int ret;
>>>>    
>>>>    	if (!enable) {
>>>> +		val = mt9p031_read(client, MT9P031_FRAME_RESTART);
>>>
>>> Do you need to read the register ? Can't you write
>>> MT9P031_FRAME_PAUSE_RESTART_SET and then MT9P031_FRAME_PAUSE_RESTART_SET
>>> | MT9P031_FRAME_RESTART_SET ? And actually, can't we just write both
>>> bits in one go, do we need two writes ?
>>
>> I think you're right we don't necessarily need to read the registers. The
>> only other bit is not used by the driver.
>>
>> But I think we do need two separate writes, at least that is what the
>> datasheet states.
>>
>> So I would drop the read but keep both write, ok?
> 
> That's fine with me if required, although I don't see where this is
> indicated in the datasheet, but I may have missed it.

It's in "Standby and Chip Enable". There is a Sequence for entering soft 
standby with two separate writes:

REG = 0x0B, 0x0002
REG = 0x0B, 0x0003

Regards,
Stefan

> 
>>>> +
>>>> +		/* enable pause restart */
>>>> +		val |= MT9P031_FRAME_PAUSE_RESTART_SET;
>>>> +		ret = mt9p031_write(client, MT9P031_FRAME_RESTART, val);
>>>> +		if (ret < 0)
>>>> +			return ret;
>>>> +
>>>> +		/* enable restart + keep pause restart set */
>>>> +		val |= MT9P031_FRAME_RESTART_SET;
>>>> +		ret = mt9p031_write(client, MT9P031_FRAME_RESTART, val);
>>>> +		if (ret < 0)
>>>> +			return ret;
>>>> +
>>>>    		/* Stop sensor readout */
>>>>    		ret = mt9p031_set_output_control(mt9p031,
>>>>    						 MT9P031_OUTPUT_CONTROL_CEN, 0);
>>>> @@ -505,6 +523,13 @@ static int mt9p031_s_stream(struct v4l2_subdev *subdev, int enable)
>>>>    	if (ret < 0)
>>>>    		return ret;
>>>>    
>>>> +	val = mt9p031_read(client, MT9P031_FRAME_RESTART);
>>>> +	/* disable reset + pause restart */
>>>> +	val &= ~MT9P031_FRAME_PAUSE_RESTART_SET;
>>>
>>> Same here, I think you can simply write MT9P031_FRAME_PAUSE_RESTART_SET.
>>
>> I'll drop the read here as well. But I need to make sure, that the Restart
>> Bit is not cleared manually here.
>>
>>>> +	ret = mt9p031_write(client, MT9P031_FRAME_RESTART, val);
>>>> +	if (ret < 0)
>>>> +		return ret;
>>>> +
>>>>    	return mt9p031_pll_enable(mt9p031);
>>>>    }
>>>>    
>