On Wed, Jul 31, 2024 at 11:06:38AM +0200, Krzysztof Kozlowski wrote:
On 31/07/2024 11:02, Kieran Bingham wrote:
Quoting Umang Jain (2024-07-31 06:41:35)
On 30/07/24 2:40 pm, Laurent Pinchart wrote:
On Tue, Jul 30, 2024 at 10:42:01AM +0200, Krzysztof Kozlowski wrote:
On 30/07/2024 10:24, Sakari Ailus wrote:
Hi Krzysztof,
On Mon, Jul 29, 2024 at 04:09:39PM +0200, Krzysztof Kozlowski wrote: > On 29/07/2024 13:04, Umang Jain wrote: >> Rectify the logical value of reset-gpio so that it is set to >> 0 (disabled) during power-on and to 1 (enabled) during power-off. >> >> Meanwhile at it, set the reset-gpio to GPIO_OUT_HIGH at initialization >> time to make sure it starts off in reset. >> >> Fixes: 45d19b5fb9ae ("media: i2c: Add imx335 camera sensor driver") >> Signed-off-by: Umang Jain umang.jain@ideasonboard.com >> --- >> drivers/media/i2c/imx335.c | 8 ++++---- >> 1 file changed, 4 insertions(+), 4 deletions(-) >> > This will break all the users, so no. At least not without mentioning > ABI break and some sort of investigating how customers or users are > affected.
I know the original authors aren't using the driver anymore and it took quite a bit of time until others started to contribute to it so I suspect the driver hasn't been in use for that long. There are no instances of the device in the in-kernel DTS either.
Any DTS author should have also noticed the issue but of course there's a risk someone could have just changed the polarity and not bothered to chech what it was supposed to be.
I agree the commit message should be more vocal about the effects on existing DTS.
I can imagine that all users (out of tree, in this case) inverted polarity in DTS based on what's implemented. You could go with some trivial hack, like I did for one of codecs - see 738455858a2d ("ASoC: codecs: wsa881x: Use proper shutdown GPIO polarity"), but I remember Mark Brown rejected similar commit for newer drivers.
I don't think there's any out-of-tree user, because when we started using the recently driver, it required lots of fixes to even work at all. I'll let Kieran and Umang comment on that, I haven't follow the development in details.
indeed, initially we had to put up fixes like :
14a60786d72e ("media: imx335: Set reserved register to default value") 81495a59baeba ("media: imx335: Fix active area height discrepency")
to make the sensor work properly on our platforms. Only after that we had a base to support more capabilities on the sensor (multiple lanes support, flips, TPG etc.)
I would also add that we had to provide control for the regulators to be able to power the device as well in fea91ee73b7c ("media: i2c: imx335: Enable regulator supplies").
Hm? That's not a proof of anything. Supplies are often turned on by default.
Given the driver was posted from Intel, I would have anticipated perhaps the driver was in fact only actually tested by Intel on ACPI platforms - yet with no ACPI table registered in the driver - even that could likely be considered broken.
Nope, that does not work like that. Their platforms and such sensors are often used on DT based boards.
What DT-platforms would that be ?
Not mentioning even PRP0001.
I don't think PRP0001 is used by Intel for camera sensors.
Sakari, do you have any information about all this ? Do you think there's a risk that the driver is currently used by anyone with a mainline kernel ?
Based on that I have a high confidence that there are no current users of this driver (except us).
Nope, wrong conclusions, not that many arguments.