All of lore.kernel.org
 help / color / mirror / Atom feed
From: Vinod Koul <vkoul@kernel.org>
To: Emil Velikov <emil.l.velikov@gmail.com>
Cc: Daniel Vetter <daniel@ffwll.ch>, David Airlie <airlied@linux.ie>,
	Rob Clark <robdclark@gmail.com>,
	devicetree <devicetree@vger.kernel.org>,
	Jernej Skrabec <jernej.skrabec@siol.net>,
	Neil Armstrong <narmstrong@baylibre.com>,
	linux-arm-msm <linux-arm-msm@vger.kernel.org>,
	Jonas Karlman <jonas@kwiboo.se>,
	"Linux-Kernel@Vger. Kernel. Org" <linux-kernel@vger.kernel.org>,
	ML dri-devel <dri-devel@lists.freedesktop.org>,
	Bjorn Andersson <bjorn.andersson@linaro.org>,
	Andrzej Hajda <a.hajda@samsung.com>,
	Rob Herring <robh+dt@kernel.org>,
	Laurent Pinchart <Laurent.pinchart@ideasonboard.com>
Subject: Re: [PATCH 3/3] drm/bridge: Introduce LT9611 DSI to HDMI bridge
Date: Thu, 14 May 2020 12:36:50 +0530	[thread overview]
Message-ID: <20200514070650.GO14092@vkoul-mobl> (raw)
In-Reply-To: <CACvgo502+8YroB5QtnGYFeSu92s_vpe_M3cPFeuC77u9xpanXQ@mail.gmail.com>

Hello Emil,

Thanks for the comments.

On 13-05-20, 20:20, Emil Velikov wrote:
> Hi Vinod,
> 
> Few high-level comments:
>  - handful of functions always return 0 and the return value is never
> checked - switch to return void

Sure makes sense, will do

>  - annotate all (nearly) arrays as static const

Will do

>  - consistently use multi_reg_write - in some cases non-const array
> will be fine, overwriting a few entries as needed

Okay that makes sense

>  - there is very partial comments about the registers/values - missing docs or?

yeah am not a big fan either, problem is documentation.

Well the spec I have doesn't have register names and few registers are
missing :( I have few name created but naming registers turned nasty
super quick.. Do let me know if you have suggestions, I will give it one
more shot though

> Personally I'm in favour of using symbolic names, instead of
> hex+comment. Considering how partial the comments are, current
> approach is perfectly fine.
> 
> On Wed, 13 May 2020 at 11:06, Vinod Koul <vkoul@kernel.org> wrote:
> >
> > Lontium Lt9611 is a DSI to HDMI bridge which supports two DSI ports and
> > I2S port as an input and HDMI port as output
> >
> > Co-developed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
> > Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
> > Signed-off-by: Vinod Koul <vkoul@kernel.org>
> 
> > --- /dev/null
> > +++ b/drivers/gpu/drm/bridge/lt9611.c
> 
> Please add a vendor prefix to the filename.

Okay

> > +struct lt9611_mode {
> > +       u16 hdisplay;
> > +       u16 vdisplay;
> > +       u8 fps;
> We all enjoy the odd fps game, but let's use vrefresh here.

Sure will change

> > +static int lt9611_mipi_input_digital(struct lt9611 *lt9611,
> > +                                    const struct drm_display_mode *mode)
> > +{
> > +       regmap_write(lt9611->regmap, 0x8300, LT9611_4LANES);
> > +
> > +       if (mode->hdisplay == 3840)
> > +               regmap_write(lt9611->regmap, 0x830a, 0x03);
> > +       else
> > +               regmap_write(lt9611->regmap, 0x830a, 0x00);
> > +
> > +       regmap_write(lt9611->regmap, 0x824f, 0x80);
> > +       regmap_write(lt9611->regmap, 0x8250, 0x10);
> > +       regmap_write(lt9611->regmap, 0x8302, 0x0a);
> > +       regmap_write(lt9611->regmap, 0x8306, 0x0a);
> Create an (non-const) array, overwriting the [1] entry for 3840 mode?

So array is the recommendation, I dont have much liking for them but I
can see they would be useful here, so will change this and other
instances and we can use regmap_multi_reg_write() while taking care of
static const for non modified arrays

> 
> > +
> > +       return 0;
> Kill return type.

Yup, here and other places

> > +       regmap_write(lt9611->regmap, 0x82e3, pclk >> 17); /* pclk[19:16] */
> > +       regmap_write(lt9611->regmap, 0x82e4, pclk >> 9);  /* pclk[15:8]  */
> > +       regmap_write(lt9611->regmap, 0x82e5, pclk >> 1);  /* pclk[7:0]   */
> Comment does not match the code.
> We're discarding the LSB, so we cannot realistically be writing
> pclk[7:0]. Similar applies for the other two.

Thanks for pointing, will fix it up

> > +       /* v_act */
> > +       ret = regmap_read(lt9611->regmap, 0x8282, &temp);
> > +       if (ret)
> > +               goto end;
> > +
> > +       v_act = temp << 8;
> > +       ret = regmap_read(lt9611->regmap, 0x8283, &temp);
> > +       if (ret)
> > +               goto end;
> > +       v_act = v_act + temp;
> > +
> Having a helper for the above "result = read(x) << 8 | read(x+1)"
> would be great.
> This way one doesn't have to repeat the pattern 4-5 times.

will add

> > +static int lt9611_read_edid(struct lt9611 *lt9611)
> > +{
> > +       unsigned int temp;
> > +       int ret = 0;
> > +       int i, j;
> > +
> > +       memset(lt9611->edid_buf, 0, EDID_SEG_SIZE);
> How about:
>   memset(lt9611->edid_buf, 0, sizeof(lt9611->edid_buf));
> 
> Then again, do we need the memset()? We are allocating the memory with
> devm_kzalloc()

Yes but lt9611_read_edid() is called multiple times so would make sense
to memset it, will modify this to sizeof.

> > +
> > +       regmap_write(lt9611->regmap, 0x8503, 0xc9);
> > +
> > +       /* 0xA0 is EDID device address */
> > +       regmap_write(lt9611->regmap, 0x8504, 0xa0);
> > +       /* 0x00 is EDID offset address */
> > +       regmap_write(lt9611->regmap, 0x8505, 0x00);
> > +       /* length for read */
> > +       regmap_write(lt9611->regmap, 0x8506, 0x20);
> Is this the same 32 as seen in the loops below? #define and use consistently?

Sure will use defines here and other places

> > +       if (block > 1)
> > +               return -EINVAL;
> > +
> > +       if (block == 0) {
> > +               /* always read 2 edid blocks once */
> Please mention why that's a good idea. From memory - there aren't many
> other drivers that do this.

Okay will find the reason for this and 
> 
> > +               ret = lt9611_read_edid(lt9611);
> > +               if (ret) {
> > +                       dev_err(lt9611->dev, "edid read failed\n");
> > +                       return ret;
> > +               }
> > +       }
> > +
> > +       if (block % 2 == 0)
> > +               memcpy(buf, lt9611->edid_buf, len);
> > +       else
> > +               memcpy(buf, lt9611->edid_buf + 128, len);
> The above can be written as:
>    memcpy(buf, lt9611->edid_buf + (block * 128), len);

correct

> > +       /* Attach secondary DSI, if specified */
> > +       if (lt9611->dsi1_node) {
> > +               lt9611->dsi1 = lt9611_attach_dsi(lt9611, lt9611->dsi1_node);
> > +               if (IS_ERR(lt9611->dsi1)) {
> > +                       ret = PTR_ERR(lt9611->dsi1);
> > +                       goto err_unregister_dsi0;
> > +               }
> > +       }
> > +
> > +       return 0;
> > +
> > +err_unregister_dsi0:
> Missing detach? If possible directly use lt9611_bridge_detach().

will update

> > +static int lt9611_read_device_rev(struct lt9611 *lt9611)
> > +{
> > +       unsigned int rev;
> > +       int ret;
> > +
> > +       regmap_write(lt9611->regmap, 0x80ee, 0x01);
> > +       ret = regmap_read(lt9611->regmap, 0x8002, &rev);
> > +       if (ret)
> > +               dev_err(lt9611->dev, "failed to read revision: %d\n", ret);
> > +
> The "failed" message will be followed by printing random kernel memory.
> Initialize rev to some dummy number or omit the dev_info.

Am printing the 'ret' error code here and not the uninitialized rev, so
I guess this one should be fine

> > +       ret = lt9611_parse_dt(&client->dev, lt9611);
> > +       if (ret) {
> > +               dev_err(dev, "failed to parse device tree\n");
> > +               return ret;
> > +       }
> > +
> > +       ret = lt9611_gpio_init(lt9611);
> > +       if (ret < 0)
> Missing of_node_put() here and for the next few error paths.

Yes this should be replaced by jump to of_node_put below

> > +static const struct of_device_id lt9611_match_table[] = {
> > +       {.compatible = "lontium,lt9611"},
> In the above two - add space after { and before }. Pretty sure

Correct, will fix this.

> ./scripts/checkpatch.pl will complain about those.
> Might want to double-check for other issues reported by said tool.

Somehow that was not the case :( I always run checkpatch.pl with
--strict option. I have 1 warn about 80 char limit for a error message
which I have ignored :)

-- 
~Vinod

WARNING: multiple messages have this Message-ID
From: Vinod Koul <vkoul@kernel.org>
To: Emil Velikov <emil.l.velikov@gmail.com>
Cc: devicetree <devicetree@vger.kernel.org>,
	Jernej Skrabec <jernej.skrabec@siol.net>,
	Neil Armstrong <narmstrong@baylibre.com>,
	David Airlie <airlied@linux.ie>,
	linux-arm-msm <linux-arm-msm@vger.kernel.org>,
	Jonas Karlman <jonas@kwiboo.se>,
	"Linux-Kernel@Vger. Kernel. Org" <linux-kernel@vger.kernel.org>,
	ML dri-devel <dri-devel@lists.freedesktop.org>,
	Bjorn Andersson <bjorn.andersson@linaro.org>,
	Andrzej Hajda <a.hajda@samsung.com>,
	Rob Herring <robh+dt@kernel.org>,
	Laurent Pinchart <Laurent.pinchart@ideasonboard.com>
Subject: Re: [PATCH 3/3] drm/bridge: Introduce LT9611 DSI to HDMI bridge
Date: Thu, 14 May 2020 12:36:50 +0530	[thread overview]
Message-ID: <20200514070650.GO14092@vkoul-mobl> (raw)
In-Reply-To: <CACvgo502+8YroB5QtnGYFeSu92s_vpe_M3cPFeuC77u9xpanXQ@mail.gmail.com>

Hello Emil,

Thanks for the comments.

On 13-05-20, 20:20, Emil Velikov wrote:
> Hi Vinod,
> 
> Few high-level comments:
>  - handful of functions always return 0 and the return value is never
> checked - switch to return void

Sure makes sense, will do

>  - annotate all (nearly) arrays as static const

Will do

>  - consistently use multi_reg_write - in some cases non-const array
> will be fine, overwriting a few entries as needed

Okay that makes sense

>  - there is very partial comments about the registers/values - missing docs or?

yeah am not a big fan either, problem is documentation.

Well the spec I have doesn't have register names and few registers are
missing :( I have few name created but naming registers turned nasty
super quick.. Do let me know if you have suggestions, I will give it one
more shot though

> Personally I'm in favour of using symbolic names, instead of
> hex+comment. Considering how partial the comments are, current
> approach is perfectly fine.
> 
> On Wed, 13 May 2020 at 11:06, Vinod Koul <vkoul@kernel.org> wrote:
> >
> > Lontium Lt9611 is a DSI to HDMI bridge which supports two DSI ports and
> > I2S port as an input and HDMI port as output
> >
> > Co-developed-by: Bjorn Andersson <bjorn.andersson@linaro.org>
> > Signed-off-by: Bjorn Andersson <bjorn.andersson@linaro.org>
> > Signed-off-by: Vinod Koul <vkoul@kernel.org>
> 
> > --- /dev/null
> > +++ b/drivers/gpu/drm/bridge/lt9611.c
> 
> Please add a vendor prefix to the filename.

Okay

> > +struct lt9611_mode {
> > +       u16 hdisplay;
> > +       u16 vdisplay;
> > +       u8 fps;
> We all enjoy the odd fps game, but let's use vrefresh here.

Sure will change

> > +static int lt9611_mipi_input_digital(struct lt9611 *lt9611,
> > +                                    const struct drm_display_mode *mode)
> > +{
> > +       regmap_write(lt9611->regmap, 0x8300, LT9611_4LANES);
> > +
> > +       if (mode->hdisplay == 3840)
> > +               regmap_write(lt9611->regmap, 0x830a, 0x03);
> > +       else
> > +               regmap_write(lt9611->regmap, 0x830a, 0x00);
> > +
> > +       regmap_write(lt9611->regmap, 0x824f, 0x80);
> > +       regmap_write(lt9611->regmap, 0x8250, 0x10);
> > +       regmap_write(lt9611->regmap, 0x8302, 0x0a);
> > +       regmap_write(lt9611->regmap, 0x8306, 0x0a);
> Create an (non-const) array, overwriting the [1] entry for 3840 mode?

So array is the recommendation, I dont have much liking for them but I
can see they would be useful here, so will change this and other
instances and we can use regmap_multi_reg_write() while taking care of
static const for non modified arrays

> 
> > +
> > +       return 0;
> Kill return type.

Yup, here and other places

> > +       regmap_write(lt9611->regmap, 0x82e3, pclk >> 17); /* pclk[19:16] */
> > +       regmap_write(lt9611->regmap, 0x82e4, pclk >> 9);  /* pclk[15:8]  */
> > +       regmap_write(lt9611->regmap, 0x82e5, pclk >> 1);  /* pclk[7:0]   */
> Comment does not match the code.
> We're discarding the LSB, so we cannot realistically be writing
> pclk[7:0]. Similar applies for the other two.

Thanks for pointing, will fix it up

> > +       /* v_act */
> > +       ret = regmap_read(lt9611->regmap, 0x8282, &temp);
> > +       if (ret)
> > +               goto end;
> > +
> > +       v_act = temp << 8;
> > +       ret = regmap_read(lt9611->regmap, 0x8283, &temp);
> > +       if (ret)
> > +               goto end;
> > +       v_act = v_act + temp;
> > +
> Having a helper for the above "result = read(x) << 8 | read(x+1)"
> would be great.
> This way one doesn't have to repeat the pattern 4-5 times.

will add

> > +static int lt9611_read_edid(struct lt9611 *lt9611)
> > +{
> > +       unsigned int temp;
> > +       int ret = 0;
> > +       int i, j;
> > +
> > +       memset(lt9611->edid_buf, 0, EDID_SEG_SIZE);
> How about:
>   memset(lt9611->edid_buf, 0, sizeof(lt9611->edid_buf));
> 
> Then again, do we need the memset()? We are allocating the memory with
> devm_kzalloc()

Yes but lt9611_read_edid() is called multiple times so would make sense
to memset it, will modify this to sizeof.

> > +
> > +       regmap_write(lt9611->regmap, 0x8503, 0xc9);
> > +
> > +       /* 0xA0 is EDID device address */
> > +       regmap_write(lt9611->regmap, 0x8504, 0xa0);
> > +       /* 0x00 is EDID offset address */
> > +       regmap_write(lt9611->regmap, 0x8505, 0x00);
> > +       /* length for read */
> > +       regmap_write(lt9611->regmap, 0x8506, 0x20);
> Is this the same 32 as seen in the loops below? #define and use consistently?

Sure will use defines here and other places

> > +       if (block > 1)
> > +               return -EINVAL;
> > +
> > +       if (block == 0) {
> > +               /* always read 2 edid blocks once */
> Please mention why that's a good idea. From memory - there aren't many
> other drivers that do this.

Okay will find the reason for this and 
> 
> > +               ret = lt9611_read_edid(lt9611);
> > +               if (ret) {
> > +                       dev_err(lt9611->dev, "edid read failed\n");
> > +                       return ret;
> > +               }
> > +       }
> > +
> > +       if (block % 2 == 0)
> > +               memcpy(buf, lt9611->edid_buf, len);
> > +       else
> > +               memcpy(buf, lt9611->edid_buf + 128, len);
> The above can be written as:
>    memcpy(buf, lt9611->edid_buf + (block * 128), len);

correct

> > +       /* Attach secondary DSI, if specified */
> > +       if (lt9611->dsi1_node) {
> > +               lt9611->dsi1 = lt9611_attach_dsi(lt9611, lt9611->dsi1_node);
> > +               if (IS_ERR(lt9611->dsi1)) {
> > +                       ret = PTR_ERR(lt9611->dsi1);
> > +                       goto err_unregister_dsi0;
> > +               }
> > +       }
> > +
> > +       return 0;
> > +
> > +err_unregister_dsi0:
> Missing detach? If possible directly use lt9611_bridge_detach().

will update

> > +static int lt9611_read_device_rev(struct lt9611 *lt9611)
> > +{
> > +       unsigned int rev;
> > +       int ret;
> > +
> > +       regmap_write(lt9611->regmap, 0x80ee, 0x01);
> > +       ret = regmap_read(lt9611->regmap, 0x8002, &rev);
> > +       if (ret)
> > +               dev_err(lt9611->dev, "failed to read revision: %d\n", ret);
> > +
> The "failed" message will be followed by printing random kernel memory.
> Initialize rev to some dummy number or omit the dev_info.

Am printing the 'ret' error code here and not the uninitialized rev, so
I guess this one should be fine

> > +       ret = lt9611_parse_dt(&client->dev, lt9611);
> > +       if (ret) {
> > +               dev_err(dev, "failed to parse device tree\n");
> > +               return ret;
> > +       }
> > +
> > +       ret = lt9611_gpio_init(lt9611);
> > +       if (ret < 0)
> Missing of_node_put() here and for the next few error paths.

Yes this should be replaced by jump to of_node_put below

> > +static const struct of_device_id lt9611_match_table[] = {
> > +       {.compatible = "lontium,lt9611"},
> In the above two - add space after { and before }. Pretty sure

Correct, will fix this.

> ./scripts/checkpatch.pl will complain about those.
> Might want to double-check for other issues reported by said tool.

Somehow that was not the case :( I always run checkpatch.pl with
--strict option. I have 1 warn about 80 char limit for a error message
which I have ignored :)

-- 
~Vinod
_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel

  reply	other threads:[~2020-05-14  7:07 UTC|newest]

Thread overview: 32+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-05-13 10:05 [PATCH 0/3] Add " Vinod Koul
2020-05-13 10:05 ` Vinod Koul
2020-05-13 10:05 ` [PATCH 1/3] dt-bindings: vendor-prefixes: Add Lontium vendor prefix Vinod Koul
2020-05-13 10:05   ` Vinod Koul
2020-05-28  1:34   ` Rob Herring
2020-05-28  1:34     ` Rob Herring
2020-05-13 10:05 ` [PATCH 2/3] dt-bindings: display: bridge: Add documentation for LT9611 Vinod Koul
2020-05-13 10:05   ` Vinod Koul
2020-05-28  1:37   ` Rob Herring
2020-05-28  1:37     ` Rob Herring
2020-05-28  1:48   ` Laurent Pinchart
2020-05-28  1:48     ` Laurent Pinchart
2020-06-04  7:18     ` Vinod Koul
2020-06-04  7:18       ` Vinod Koul
2020-06-04  7:29       ` Laurent Pinchart
2020-06-04  7:29         ` Laurent Pinchart
2020-05-13 10:05 ` [PATCH 3/3] drm/bridge: Introduce LT9611 DSI to HDMI bridge Vinod Koul
2020-05-13 10:05   ` Vinod Koul
2020-05-13 16:14   ` kbuild test robot
2020-05-13 16:14     ` kbuild test robot
2020-05-13 16:46   ` kbuild test robot
2020-05-13 16:46     ` kbuild test robot
2020-05-13 19:20   ` Emil Velikov
2020-05-13 19:20     ` Emil Velikov
2020-05-14  7:06     ` Vinod Koul [this message]
2020-05-14  7:06       ` Vinod Koul
2020-05-28  1:52   ` Laurent Pinchart
2020-05-28  1:52     ` Laurent Pinchart
2020-06-04  7:25     ` Vinod Koul
2020-06-04  7:25       ` Vinod Koul
2020-06-04  7:38       ` Laurent Pinchart
2020-06-04  7:38         ` Laurent Pinchart

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20200514070650.GO14092@vkoul-mobl \
    --to=vkoul@kernel.org \
    --cc=Laurent.pinchart@ideasonboard.com \
    --cc=a.hajda@samsung.com \
    --cc=airlied@linux.ie \
    --cc=bjorn.andersson@linaro.org \
    --cc=daniel@ffwll.ch \
    --cc=devicetree@vger.kernel.org \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=emil.l.velikov@gmail.com \
    --cc=jernej.skrabec@siol.net \
    --cc=jonas@kwiboo.se \
    --cc=linux-arm-msm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=narmstrong@baylibre.com \
    --cc=robdclark@gmail.com \
    --cc=robh+dt@kernel.org \
    --subject='Re: [PATCH 3/3] drm/bridge: Introduce LT9611 DSI to HDMI bridge' \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.