Message-ID: <ZWr5my7SVKE2HPTZ@tom-HP-ZBook-Fury-15-G7-Mobile-Workstation>
Date:   Sat, 2 Dec 2023 10:32:11 +0100
From:   Tommaso Merciai <tomm.merciai@...il.com>
To:     Sakari Ailus <sakari.ailus@...ux.intel.com>
Cc:     laurent.pinchart@...asonboard.com, martin.hecht@...et.eu,
        michael.roeder@...et.eu, linuxfancy@...glegroups.com,
        mhecht73@...il.com, christophe.jaillet@...adoo.fr,
        Mauro Carvalho Chehab <mchehab@...nel.org>,
        Liam Girdwood <lgirdwood@...il.com>,
        Mark Brown <broonie@...nel.org>,
        Hans de Goede <hdegoede@...hat.com>,
        Hans Verkuil <hverkuil-cisco@...all.nl>,
        Tomi Valkeinen <tomi.valkeinen@...asonboard.com>,
        Gerald Loacker <gerald.loacker@...fvision.net>,
        Nicholas Roth <nicholas@...hemail.net>,
        Andy Shevchenko <andy.shevchenko@...il.com>,
        Daniel Scally <djrscally@...il.com>,
        Bingbu Cao <bingbu.cao@...el.com>,
        Linus Walleij <linus.walleij@...aro.org>,
        linux-kernel@...r.kernel.org, linux-media@...r.kernel.org
Subject: Re: [PATCH v14 3/3] media: i2c: Add support for alvium camera

Hi Sakari,
Thanks for your comments.

On Fri, Dec 01, 2023 at 09:00:33PM +0000, Sakari Ailus wrote:
> Hi Tommaso,
> 
> A few more comments below...
> 
> On Fri, Nov 24, 2023 at 10:30:07AM +0100, Tommaso Merciai wrote:
> 
> ...
> 
> > +static int alvium_get_bcrm_vers(struct alvium_dev *alvium)
> > +{
> > +	struct device *dev = &alvium->i2c_client->dev;
> > +	struct alvium_bcrm_vers *v;
> > +	u64 val;
> > +	int ret;
> > +
> > +	ret = alvium_read(alvium, REG_BCRM_VERSION_R, &val, NULL);
> > +	if (ret)
> > +		return ret;
> > +
> > +	v = (struct alvium_bcrm_vers *)&val;
> 
> You're still reading the entire struct using a single read. :-( This won't
> work on a BE machine as while the struct fields are in the same memory
> locations, the respective data in a single 64-bit value is not.

What about splitting REG_BCRM_VERSION_R in:

#define REG_BCRM_MINOR_VERSION_R			CCI_REG16(0x0000)
#define REG_BCRM_MAJOR_VERSION_R			CCI_REG16(0x0002)

and REG_BCRM_DEVICE_FIRMWARE_VERSION_R in:

#define REG_BCRM_DEVICE_FW_SPEC_VERSION_R		REG_BCRM_V4L2_8BIT(0x0010)
#define REG_BCRM_DEVICE_FW_MAJOR_VERSION_R		REG_BCRM_V4L2_8BIT(0x0011)
#define REG_BCRM_DEVICE_FW_MINOR_VERSION_R		REG_BCRM_V4L2_16BIT(0x0012)
#define REG_BCRM_DEVICE_FW_PATCH_VERSION_R		REG_BCRM_V4L2_32BIT(0x0014)


Then reading those values as a single values as you suggest:

static int alvium_get_bcrm_vers(struct alvium_dev *alvium)
{
	struct device *dev = &alvium->i2c_client->dev;
	u64 min, maj;
	int ret = 0;

	ret = alvium_read(alvium, REG_BCRM_MINOR_VERSION_R, &min, &ret);
	ret = alvium_read(alvium, REG_BCRM_MAJOR_VERSION_R, &maj, &ret);
	if (ret)
		return ret;

	dev_info(dev, "bcrm version: %llu.%llu\n", min, maj);

	return 0;
}

static int alvium_get_fw_version(struct alvium_dev *alvium)
{
	struct device *dev = &alvium->i2c_client->dev;
	u64 spec, maj, min, pat;
	int ret = 0;

	ret = alvium_read(alvium, REG_BCRM_DEVICE_FW_SPEC_VERSION_R, &spec, &ret);
	ret = alvium_read(alvium, REG_BCRM_DEVICE_FW_MAJOR_VERSION_R, &maj, &ret);
	ret = alvium_read(alvium, REG_BCRM_DEVICE_FW_MINOR_VERSION_R, &min, &ret);
	ret = alvium_read(alvium, REG_BCRM_DEVICE_FW_PATCH_VERSION_R, &pat, &ret);
	if (ret)
		return ret;

	dev_info(dev, "fw version: %llu.%llu.%llu.%llu\n", spec, maj, min, pat);

	return 0;
}

Then I'm going to remove alvium_bcrm_vers and alvium_fw_vers.
And alvium_is_alive became: 

static int alvium_is_alive(struct alvium_dev *alvium)
{
	u64 bcrm, hbeat;
	int ret = 0;

	alvium_read(alvium, REG_BCRM_MINOR_VERSION_R, &bcrm, &ret);
	alvium_read(alvium, REG_BCRM_HEARTBEAT_RW, &hbeat, &ret);
	if (ret)
		return ret;

	return hbeat;
}

What do you think? Let me know.
(Maybe is this that you are trying to explain me but I haven't catch,
sorry) :)


> 
> > +
> > +	dev_info(dev, "bcrm version: %u.%u\n", v->minor, v->major);
> > +
> > +	return 0;
> > +}
> > +
> > +static int alvium_get_fw_version(struct alvium_dev *alvium)
> > +{
> > +	struct device *dev = &alvium->i2c_client->dev;
> > +	struct alvium_fw_vers *fw_v;
> > +	u64 val;
> > +	int ret;
> > +
> > +	ret = alvium_read(alvium, REG_BCRM_DEVICE_FIRMWARE_VERSION_R, &val, NULL);
> > +	if (ret)
> > +		return ret;
> > +
> > +	fw_v = (struct alvium_fw_vers *)&val;
> 
> Same here.
> 
> > +
> > +	dev_info(dev, "fw version: %u.%u.%u.%u\n", fw_v->special, fw_v->major,
> > +		 fw_v->minor, fw_v->patch);
> > +
> > +	return 0;
> > +}
> > +
> > +static int alvium_get_bcrm_addr(struct alvium_dev *alvium)
> > +{
> > +	u64 val;
> > +	int ret;
> > +
> > +	ret = alvium_read(alvium, REG_BCRM_REG_ADDR_R, &val, NULL);
> > +	if (ret)
> > +		return ret;
> > +
> > +	alvium->bcrm_addr = val;
> > +
> > +	return 0;
> > +}
> 
> ...
> 
> > +static int alvium_setup_mipi_fmt(struct alvium_dev *alvium)
> > +{
> > +	unsigned int avail_fmt_cnt = 0;
> > +	unsigned int fmt = 0;
> > +	size_t sz = 0;
> > +
> > +	alvium->alvium_csi2_fmt = NULL;
> 
> This seems to be unnnecessary: the field is assigned below without using it
> (obviously).

Ok, I will remove this in v15.

> 
> > +
> > +	/* calculate fmt array size */
> > +	for (fmt = 0; fmt < ALVIUM_NUM_SUPP_MIPI_DATA_FMT; fmt++) {
> > +		if (!alvium->is_mipi_fmt_avail[alvium_csi2_fmts[fmt].fmt_av_bit])
> > +			continue;
> > +
> > +		if ((!alvium_csi2_fmts[fmt].is_raw) ||
> > +		    (alvium->is_bay_avail[alvium_csi2_fmts[fmt].bay_av_bit]))
> > +			sz++;
> > +	}
> > +
> > +	/* init alvium_csi2_fmt array */
> > +	alvium->alvium_csi2_fmt_n = sz;
> > +	alvium->alvium_csi2_fmt =
> > +		kmalloc_array(sz, sizeof(struct alvium_pixfmt), GFP_KERNEL);
> > +	if (!alvium->alvium_csi2_fmt)
> > +		return -ENOMEM;
> > +
> > +	/* Create the alvium_csi2 fmt array from formats available */
> > +	for (fmt = 0; fmt < ALVIUM_NUM_SUPP_MIPI_DATA_FMT; fmt++) {
> > +		if (!alvium->is_mipi_fmt_avail[alvium_csi2_fmts[fmt].fmt_av_bit])
> > +			continue;
> > +
> > +		if ((!alvium_csi2_fmts[fmt].is_raw) ||
> > +		    (alvium->is_bay_avail[alvium_csi2_fmts[fmt].bay_av_bit])) {
> > +			alvium->alvium_csi2_fmt[avail_fmt_cnt] = alvium_csi2_fmts[fmt];
> > +			avail_fmt_cnt++;
> > +		}
> > +	}
> > +
> > +	return 0;
> > +}
> 
> ...
> 
> > +static const struct alvium_pixfmt *
> > +alvium_code_to_pixfmt(struct alvium_dev *alvium, u32 code)
> > +{
> > +	const struct alvium_pixfmt *formats = alvium->alvium_csi2_fmt;
> 
> I'd use alvium->alvium_csi2_fmt and not add a local variable. Up to you.

Ok also for me.

> 
> > +	unsigned int i;
> > +
> > +	for (i = 0; formats[i].code; ++i)
> > +		if (formats[i].code == code)
> > +			return &formats[i];
> > +
> > +	return &formats[0];
> > +}
> > +
> > +static int alvium_set_mode(struct alvium_dev *alvium,
> > +			   struct v4l2_subdev_state *state)
> > +{
> > +	struct v4l2_mbus_framefmt *fmt;
> > +	struct v4l2_rect *crop;
> > +	int ret;
> > +
> > +	crop = v4l2_subdev_state_get_crop(state, 0);
> > +	fmt = v4l2_subdev_state_get_format(state, 0);
> > +
> > +	v4l_bound_align_image(&fmt->width, alvium->img_min_width,
> > +			      alvium->img_max_width, 0,
> > +			      &fmt->height, alvium->img_min_height,
> > +			      alvium->img_max_height, 0, 0);
> > +
> > +	/* alvium don't accept negative crop left/top */
> > +	crop->left = clamp((u32)max(0, crop->left), alvium->min_offx,
> > +			   (u32)(alvium->img_max_width - fmt->width));
> > +	crop->top = clamp((u32)max(0, crop->top), alvium->min_offy,
> > +			  (u32)(alvium->img_max_height - fmt->height));
> > +
> > +	ret = alvium_set_img_width(alvium, fmt->width);
> > +	if (ret)
> > +		return ret;
> > +
> > +	ret = alvium_set_img_height(alvium, fmt->height);
> > +	if (ret)
> > +		return ret;
> > +
> > +	ret = alvium_set_img_offx(alvium, crop->left);
> > +	if (ret)
> > +		return ret;
> > +
> > +	ret = alvium_set_img_offy(alvium, crop->top);
> > +	if (ret)
> > +		return ret;
> > +
> > +	return 0;
> > +}

I'm going to rebase v15 on top of your master branch.
My plan is moving alvium_init_cfg now alvium_init_state into
v4l2_subdev_internal_ops.


Thanks & Regards,
Tommaso

> 
> -- 
> Kind regards,
> 
> Sakari Ailus

Powered by blists - more mailing lists