add mira220 image sensor #6717

fiepfiep · 2025-03-12T17:42:45Z

add driver / device tree support for ams mira220

needed changes also done for libcamera.

6by9

I've started to review this, but I'm noticing a lot of my comments from #6552 haven't been addressed.

If embedded data isn't required, then do not include it.

arch/arm/boot/dts/overlays/mira220-overlay.dts

6by9 · 2025-03-12T17:52:57Z

arch/arm/boot/dts/overlays/mira220.dtsi

+			data-lanes = <1 2>;
+			// clock-noncontinuous;
+			link-frequencies =
+				/bits/ 64 <456000000>;


Is this the correct link frequency for this sensor?

currently operating at 1.5 Gbit/s per lane

solved, set to 750, as datarate is 1500Mbit/s

6by9 · 2025-03-12T18:03:21Z

drivers/media/i2c/mira220.c

+
+ /* Embedded metadata stream structure */
+ #define MIRA220_EMBEDDED_LINE_WIDTH 16384
+ #define MIRA220_NUM_EMBEDDED_LINES 1


Does this sensor actually have useful embedded metadata? It's disabled in libcamera.

I'd made the same comment in #6552 (comment) that this isn't how mainline are implementing metadata, so it's simpler to not implement it in the driver if not needed.

I have removed the embedded data

6by9 · 2025-03-12T18:03:37Z

drivers/media/i2c/mira220.c

+ // 1600_1400_30fps_12b_2lanes
+
+ static const struct mira220_reg full_1600_1400_1500_12b_2lanes_reg[] = {
+	{0x1003,0x2}, //,Initial Upload


Comment isn't useful.

6by9 · 2025-03-12T18:12:11Z

drivers/media/i2c/mira220.c

+
+ // 1600_1400_30fps_12b_2lanes
+
+ static const struct mira220_reg full_1600_1400_1500_12b_2lanes_reg[] = {


You have about 18 registers that differ between these 3 huge tables. Break out all the common stuff into a common table, and only send the bits that change per mode

thanks, will do.

simplified, reduced driver to 1 mode only.

6by9 · 2025-03-12T18:12:59Z

drivers/media/i2c/mira220.c

+ #define MIRA220_PIXEL_ARRAY_WIDTH		1600U
+ #define MIRA220_PIXEL_ARRAY_HEIGHT		1400U
+
+ #define MIRA220_ANALOG_GAIN_REG			0x400A


Comment made at #6552 (comment) that the V4L2_CCI is preferred rather than bare register defines as it encodes the size of each register.

I converted everything to the new style cci writes

6by9 · 2025-03-12T18:14:43Z

drivers/media/i2c/mira220.c

+ static const char * const mira220_supply_name[] = {
+	 // TODO(jalv): Check supply names
+	 /* Supplies can be enabled in any order */
+	 "VANA",  /* Analog (2.8V) supply */


Regulators should be named in lower case.

pelwell · 2025-04-02T19:40:34Z

I'll warn you now that, once the content of these commits has been approved, the end result should ideally be 3 commits - driver changes, config changes, and overlay/README changes.

pelwell · 2025-04-02T19:49:18Z

arch/arm/boot/dts/overlays/mira220-overlay.dts

@@ -58,28 +58,6 @@
 		};
 	};

-	fragment@102 {


Did you delete this fragment by mistake? The media-controller parameter refers to it.

Yes that is a mistake.

I put it back. thanks for finding out.

fiepfiep · 2025-04-02T19:58:39Z

I'll warn you now that, once the content of these commits has been approved, the end result should ideally be 3 commits - driver changes, config changes, and overlay/README changes.

I will keep it in mind. Once all is approved by you I will refactor it to 3 commits.

fiepfiep · 2025-04-02T21:48:44Z

I cleaned up the driver a bit, implemented the new CCI_REG style writes, and took care of 6by9's comments.
Please let me know if you find other issues before we can do the merge.

pelwell

Apart from the extra blank lines in the README, the overlay looks good.

Checkpatch has lots of complaints about the driver which you should really fix. Nobody using printk directly anymore, for example.

The only additional change I would like (apart from refactoring down to 3 commits as discussed) is the addition of CONFIG_VIDEO_MIRA220 to the new PREEMPT_RT config file arm64/bcm2711_rt_defconfig.

pelwell · 2025-04-03T08:43:03Z

arch/arm/boot/dts/overlays/README

@@ -3632,6 +3649,7 @@ Params: rotation                Mounting rotation of the camera sensor (0 or
                                Appears not to work on Pi3.


+


This looks like an unnecessary change.

pelwell · 2025-04-03T08:44:47Z

arch/arm/boot/dts/overlays/README

@@ -3511,6 +3511,23 @@ Params:
                                (default 16).


+


2 blank lines is enough.

arch/arm/boot/dts/overlays/README

6by9

Still a number of my previous review comments haven't been addressed.

Please remove the references to the embedded data pad.
Please remove all commented out code.
Please remove all redundant comments.
Please rationalise the logging messages. Very few should be needed at all.
Regulator names should be consistently in lower case.

6by9 · 2025-04-03T17:22:53Z

arch/arm/boot/dts/overlays/mira220-overlay.dts

+		       <&csi_frag>, "target:0=",<&csi0>,
+		       <&clk_frag>, "target:0=",<&cam0_clk>,
+		       <&cam_node>, "clocks:0=",<&cam0_clk>,
+		       <&cam_node>, "VANA-supply:0=",<&cam0_reg>;


Your supply is now vana-supply, so this won't match.

6by9 · 2025-04-03T17:24:47Z

drivers/media/i2c/Kconfig

@@ -149,6 +149,20 @@ config VIDEO_HI847
          To compile this driver as a module, choose M here: the
          module will be called hi847.

+config VIDEO_MIRA220


These are normally in alphabetical order. (I see the Arducam drivers have got out of position during forward porting).

Annoyingly this will also move the position in the defconfigs, so perhaps I don't care enough.

6by9 · 2025-04-03T17:25:11Z

drivers/media/i2c/Makefile

@@ -49,6 +49,7 @@ obj-$(CONFIG_VIDEO_HI556) += hi556.o
 obj-$(CONFIG_VIDEO_HI846) += hi846.o
 obj-$(CONFIG_VIDEO_HI847) += hi847.o
 obj-$(CONFIG_VIDEO_I2C) += video-i2c.o
+obj-$(CONFIG_VIDEO_MIRA220)	+= mira220.o


Ditto - alphabetical order please.

6by9 · 2025-04-03T17:27:10Z

drivers/media/i2c/mira220.c

+
+#define MIRA220_ANALOG_GAIN_MIN 1
+#define MIRA220_ANALOG_GAIN_MAX \
+	1 /* Fixed analog gain to 1, to avoid unexpected behavior. */


Weird line breaking. That does seem quite a limitation though if you have no analogue gain.

6by9 · 2025-04-03T17:28:24Z

drivers/media/i2c/mira220.c

+// Outdated. See below.
+// pixel_rate = link_freq * 2 * nr_of_lanes / bits_per_sample
+// 1.0Gb/s * 2 * 2 / 12 = 357913941
+// #define MIRA220_PIXEL_RATE		(357913941)


If it's outdated, then please remove it.

6by9 · 2025-04-03T17:41:00Z

drivers/media/i2c/mira220.c

+
+	/*test cci write*/
+	ret = cci_read(mira220->regmap, MIRA220_EXP_TIME_REG, &readval, &ret);
+	otherval = readval & 0xFFFF;


Why read the value before writing it? You do nothing with the result.

6by9 · 2025-04-03T17:44:01Z

drivers/media/i2c/mira220.c

+	if (mira220_check_hwcfg(dev))
+		return -EINVAL;
+
+	/* Parse device tree to check if dtoverlay has param skip-reg-upload=1 */


6by9 · 2025-04-03T17:44:49Z

drivers/media/i2c/mira220.c

+}
+
+static const struct dev_pm_ops mira220_pm_ops = {
+	SET_SYSTEM_SLEEP_PM_OPS(mira220_suspend, mira220_resume)


You shouldn't be implementing SLEEP_OPS.

6by9 · 2025-04-03T17:46:38Z

drivers/media/i2c/mira220.c

+		SET_RUNTIME_PM_OPS(mira220_power_off, mira220_power_on, NULL)
+};
+
+static const struct of_device_id mira220_dt_ids[] = { { .compatible =


Please use the normal formatting for dt_ids.

static const struct of_device_id mira220_dt_ids[] = { { .compatible = "ams,mira220" }, { /* sentinel */ } };

6by9 · 2025-04-03T17:47:37Z

drivers/media/i2c/mira220.c

+MODULE_DEVICE_TABLE(of, mira220_dt_ids);
+
+static const struct i2c_device_id mira220_ids[] = { { "mira220", 0 }, {} };
+MODULE_DEVICE_TABLE(i2c, mira220_ids);


You shouldn't need I2C ids. This driver needs config data, so you can't load via sysfs.

pelwell · 2025-04-06T20:56:15Z

FYI the build checks explicitly set CONFIG_WERROR=y, overriding the defconfigs:

scripts/config --file ${{github.workspace}}/build/.config --set-val CONFIG_WERROR y

fiepfiep · 2025-04-15T13:26:45Z

@pelwell could you run your checks again? I split my changes into 3 commits as requested.

pelwell · 2025-04-15T13:52:45Z

It looks like somebody has started them. Please mark the PR as "Ready for review" when you think it is. However, there is no excuse for some of these (https://github.com/raspberrypi/linux/actions/runs/14470568709/job/40583365956?pr=6717#step:6:80) checkpatch errors such as trailing whitespace.

fiepfiep · 2025-04-16T09:53:54Z

I still get this error in checkpatch, what does it mean?
@pelwell

 WARNING: From:/Signed-off-by: email name mismatch: 'From: philippe baetens <[email protected]>' != 'Signed-off-by: Philippe Baetens <[email protected]>'
Error: WARNING: From:/Signed-off-by: email name mismatch: 'From: philippe baetens <[email protected]>' != 'Signed-off-by: Philippe Baetens <[email protected]>'

pelwell · 2025-04-16T09:58:17Z

Is it possibly just that your name appears with and without uppercase letters? I'm not bothered about that, but it makes sense to fix it to avoid such errors in the future.

fiepfiep · 2025-04-16T12:51:46Z

Is it possibly just that your name appears with and without uppercase letters? I'm not bothered about that, but it makes sense to fix it to avoid such errors in the future.

that seems to be the case indeed. i modified the commit signoff message + some of the checkpatch remarks. Hope it is ok now.

pelwell · 2025-04-16T12:55:12Z

That's fixed it. I'll give these builds time to complete, in which time 6by9 may want to chip in, but I've been taking his silence for consent.

6by9 · 2025-04-16T12:59:01Z

That's fixed it. I'll give these builds time to complete, in which time 6by9 may want to chip in, but I've been taking his silence for consent.

No, just hadn't had a chance to look through.
It looks like several of my previous comments haven't been addressed, eg the overlay override cam0 still setting VANA-supply and changing vana-supply.

fiepfiep · 2025-04-16T13:05:31Z

So - what are the next steps to get this driver into the raspberry linux kernel?
At the moment I'm only merging into the 6.12 kernel. What about older or newer kernel versions?
How does this driver propagate to official bookworm builds?

thanks.

6by9

In reading through to see if my previous comments had been addressed, add a number more.

There are a fair number of commented out blocks of code. They should be reviewed and removed where redundant.

6by9 · 2025-04-16T13:00:21Z

arch/arm/configs/bcm2709_defconfig

@@ -934,6 +934,7 @@ CONFIG_VIDEO_BCM2835_UNICAM_LEGACY=m
 CONFIG_VIDEO_BCM2835_UNICAM=m
 CONFIG_VIDEO_ARDUCAM_64MP=m
 CONFIG_VIDEO_ARDUCAM_PIVARIETY=m
+CONFIG_VIDEO_MIRA220=m


With the move within Kconfig to keep everything in alphabetical order, I'm expecting this to be the wrong location in the defconfig.

right. will move to the line below the IMX sensors.

6by9 · 2025-04-16T13:02:02Z

drivers/media/i2c/mira220.c

+	//MEDIA_BUS_FMT_Y8_1X8,
+	//MEDIA_BUS_FMT_Y10_1X10,
+	//MEDIA_BUS_FMT_Y12_1X12,
+	MEDIA_BUS_FMT_SGRBG8_1X8,


The comment just before this says that the sensor is monochrome, and yet you've used the GRBG Bayer pattern code. So which is it?

sensor comes both as RGB and mono. For now, I'm only doing the RGB.

6by9 · 2025-04-16T13:03:12Z

drivers/media/i2c/mira220.c

+	reg_list = &mira220->mode->reg_list;
+	ret = cci_multi_reg_write(mira220->regmap, reg_list->regs,
+				  reg_list->num_of_regs, NULL);
+	// ret = mira220_write_regs(mira220, reg_list->regs, reg_list->num_of_regs);


Remove the now redundant line.

6by9 · 2025-04-16T13:03:47Z

drivers/media/i2c/mira220.c

+	int ret = 0;
+
+	mutex_lock(&mira220->mutex);
+	if (mira220->streaming == enable) {


This should never be possible as the framework blocks it.

do you mean all mutex statements are not needed?

mira220->streaming is not required. The framework should never call set_stream(..., 1) if the stream is already enabled, or set_stream(... , 0) if there wasn't also a previous call to set_stream(..., 1) that returned 0.

If that means you don't need the mutex, then that's a bonus.

ah, clear. Will fix that too.

6by9 · 2025-04-16T13:04:31Z

drivers/media/i2c/mira220.c

+	struct mira220 *mira220 = to_mira220(sd);
+	int ret;
+
+	if (mira220->streaming) {


resume shouldn't start streaming directly as the whole pipeline needs to be started correctly.

resume got deleted (unused)

6by9 · 2025-04-16T13:05:27Z

drivers/media/i2c/mira220.c

+
+	// Make the vblank control read only. This could be changed to allow changing framerate in
+	// runtime, but would require adapting other settings
+	// mira220->vblank->flags |= V4L2_CTRL_FLAG_READ_ONLY;


This has changed in that you have VBLANK control, so remove the comment and the commented out code.

6by9 · 2025-04-16T13:08:17Z

drivers/media/i2c/mira220.c

+	(void)mira220;
+
+	regulator_bulk_disable(MIRA220_NUM_SUPPLIES, mira220->supplies);
+	clk_disable_unprepare(mira220->xclk);


Clock and regulators are disabled in the opposite order to being prepared.

ok, will swap order.

6by9 · 2025-04-16T13:09:06Z

drivers/media/i2c/mira220.c

+	return ret;
+}
+
+static int __maybe_unused mira220_suspend(struct device *dev)


Both mira220_suspend and mira220_resume are totally unused. Remove them.

will remove them.

6by9 · 2025-04-16T13:16:48Z

So - what are the next steps to get this driver into the raspberry linux kernel? At the moment I'm only merging into the 6.12 kernel. What about older or newer kernel versions? How does this driver propagate to official bookworm builds?

Older kernels won't get updated.

Later kernels will get a forward port, as long as there aren't any significant build changes required by the every-changing kernel APIs. Simple changes are likely to get fixed up. Significant changes without obvious fixups will be reported back to you to fix - we don't undertake the full support burden of 3rd party drivers, and they may get dropped from the defconfigs if not fixed and/or get the BROKEN tag added to the Kconfig entry if the situation persists for a significant period.

rpi-update gets kernel updates normally several time a week.
Raspberry Pi OS gets updates every few months based on when significant things occur, providing no significant issues are known about at that point.

Adds defconfig for the Mira220 1600x1400 global shutter image sensor Signed-off-by: philippe baetens <[email protected]>

Adds an overlay for the Mira220 1600x1400 global shutter image sensor Signed-off-by: philippe baetens <[email protected]>

Adds a driver for the NIR-enhanced Mira220 1600x1400 global shutter image sensor. Signed-off-by: philippe baetens <[email protected]>

pelwell · 2025-04-22T15:12:59Z

Some of @6by9's comments have yet to be addressed, but the overlay and general structure look okay to me.

fiepfiep · 2025-04-22T19:22:09Z

Some of @6by9's comments have yet to be addressed, but the overlay and general structure look okay to me.

Thanks for reviewing. I believe I addressed all of @6by9's comments, but I might have missed something.. Can you be a bit more specific? thanks!

Also: each time I just 'force push' 3 new commits (overlay, conf, i2c driver) - not sure if this is your preferred way, or if you like to see my full commit history?

pelwell · 2025-04-22T19:30:41Z

arch/arm/boot/dts/overlays/mira220.dtsi

+// Definitions for MIRA220 camera module on VC I2C bus
+
+
+//#include <dt-bindings/gpio/gpio.h>


Is this commented out section left here for a reason? If so, explain why in a comment, otherwise delete.

pelwell · 2025-04-22T19:32:30Z

On this page (https://github.com/raspberrypi/linux/pull/6717/files), scroll down and expand mira.c, then look at the comments that remain, e.g.: https://github.com/raspberrypi/linux/pull/6717/files#r2027459784 and https://github.com/raspberrypi/linux/pull/6717/files#r2027467385

Small fixes, mostly cosmetic. Removed unneeded code. Differentiate between physical array and active array. Signed-off-by: philippe baetens <[email protected]>

Remove commented out code. Signed-off-by: philippe baetens <[email protected]>

6by9 requested changes Mar 12, 2025

View reviewed changes

popcornmix force-pushed the rpi-6.12.y branch from 28d1e43 to de1afd2 Compare March 13, 2025 14:14

popcornmix force-pushed the rpi-6.12.y branch from 617b6a8 to 8d53a07 Compare March 25, 2025 12:36

fiepfiep closed this Mar 27, 2025

fiepfiep force-pushed the rpi-6.12.y branch from 5321fb8 to cd231d4 Compare March 27, 2025 08:32

fiepfiep reopened this Apr 2, 2025

pelwell reviewed Apr 2, 2025

View reviewed changes

fiepfiep marked this pull request as draft April 2, 2025 19:57

fiepfiep force-pushed the rpi-6.12.y branch from 1ef53d1 to b9efe09 Compare April 2, 2025 21:38

fiepfiep requested a review from 6by9 April 2, 2025 21:43

fiepfiep marked this pull request as ready for review April 2, 2025 21:43

fiepfiep marked this pull request as draft April 2, 2025 21:44

fiepfiep force-pushed the rpi-6.12.y branch from 4181345 to e3242fc Compare April 2, 2025 22:15

pelwell reviewed Apr 3, 2025

View reviewed changes

6by9 reviewed Apr 3, 2025

View reviewed changes

6by9 mentioned this pull request Apr 15, 2025

Camera Stream error raspberrypi/libcamera#255

Open

fiepfiep force-pushed the rpi-6.12.y branch 2 times, most recently from 6c62034 to a0c3d9c Compare April 15, 2025 13:23

fiepfiep requested a review from 6by9 April 15, 2025 13:25

fiepfiep force-pushed the rpi-6.12.y branch from a0c3d9c to 59de12a Compare April 15, 2025 15:40

fiepfiep marked this pull request as ready for review April 15, 2025 15:41

fiepfiep force-pushed the rpi-6.12.y branch from 59de12a to 3ee1958 Compare April 16, 2025 12:40

6by9 requested changes Apr 16, 2025

View reviewed changes

philippe baetens added 3 commits April 16, 2025 18:18

defconfigs: Add ams Mira220 image sensor to Pi defconfigs

9e928d3

Adds defconfig for the Mira220 1600x1400 global shutter image sensor Signed-off-by: philippe baetens <[email protected]>

dtoverlays: Add overlay for ams Mira220 image sensor

7887667

Adds an overlay for the Mira220 1600x1400 global shutter image sensor Signed-off-by: philippe baetens <[email protected]>

media/i2c: Add driver for ams Mira220

88be0ee

Adds a driver for the NIR-enhanced Mira220 1600x1400 global shutter image sensor. Signed-off-by: philippe baetens <[email protected]>

fiepfiep force-pushed the rpi-6.12.y branch from aee8a04 to 88be0ee Compare April 16, 2025 16:31

fiepfiep requested a review from 6by9 April 22, 2025 09:07

pelwell reviewed Apr 22, 2025

View reviewed changes

philippe baetens added 2 commits April 23, 2025 23:58

media/i2c: changes to Mira220 driver:

94bcda1

Small fixes, mostly cosmetic. Removed unneeded code. Differentiate between physical array and active array. Signed-off-by: philippe baetens <[email protected]>

dtoverlays: clean up mira220 overlay

2f06530

Remove commented out code. Signed-off-by: philippe baetens <[email protected]>


		// 1600_1400_30fps_12b_2lanes

		static const struct mira220_reg full_1600_1400_1500_12b_2lanes_reg[] = {

		@@ -3632,6 +3649,7 @@ Params: rotation Mounting rotation of the camera sensor (0 or
		Appears not to work on Pi3.

		// Definitions for MIRA220 camera module on VC I2C bus


		//#include <dt-bindings/gpio/gpio.h>

add mira220 image sensor #6717

Are you sure you want to change the base?

add mira220 image sensor #6717

Conversation

fiepfiep commented Mar 12, 2025 • edited Loading

6by9 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pelwell commented Apr 2, 2025

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

fiepfiep commented Apr 2, 2025

fiepfiep commented Apr 2, 2025 • edited Loading

pelwell left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

6by9 left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

pelwell commented Apr 6, 2025

fiepfiep commented Apr 15, 2025

pelwell commented Apr 15, 2025

fiepfiep commented Apr 16, 2025 • edited Loading

pelwell commented Apr 16, 2025

fiepfiep commented Apr 16, 2025

pelwell commented Apr 16, 2025

6by9 commented Apr 16, 2025

fiepfiep commented Apr 16, 2025

6by9 left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

6by9 commented Apr 16, 2025

pelwell commented Apr 22, 2025

fiepfiep commented Apr 22, 2025 • edited Loading

Choose a reason for hiding this comment

pelwell commented Apr 22, 2025

fiepfiep commented Mar 12, 2025 •

edited

Loading

fiepfiep commented Apr 2, 2025 •

edited

Loading

6by9 left a comment •

edited

Loading

fiepfiep commented Apr 16, 2025 •

edited

Loading

fiepfiep commented Apr 22, 2025 •

edited

Loading