linux-i2c.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Jean Delvare <jdelvare@suse.de>
To: Heiner Kallweit <hkallweit1@gmail.com>
Cc: linux-i2c@vger.kernel.org
Subject: Re: [PATCH 6/8] i2c: i801: add i801_single_transaction(), complementing i801_block_transaction()
Date: Fri, 10 Jun 2022 13:03:24 +0200	[thread overview]
Message-ID: <20220610130324.1ab2725d@endymion.delvare> (raw)
In-Reply-To: <eba39e4d-fb5f-c9d3-0d51-001f8d584d51@gmail.com>

Hi Heiner,

On Fri, 15 Apr 2022 18:58:03 +0200, Heiner Kallweit wrote:
> This patch factors out non-block pre/post processing to a new function
> i801_single_transaction(), complementing existing function
> i801_block_transaction(). This makes i801_access() better readable.

I like the idea, but I have objections about some implementation
details, see below.

> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
> ---
>  drivers/i2c/busses/i2c-i801.c | 95 +++++++++++++++++++++--------------
>  1 file changed, 58 insertions(+), 37 deletions(-)
> 
> diff --git a/drivers/i2c/busses/i2c-i801.c b/drivers/i2c/busses/i2c-i801.c
> index bf77f8640..8c2245f38 100644
> --- a/drivers/i2c/busses/i2c-i801.c
> +++ b/drivers/i2c/busses/i2c-i801.c
> @@ -771,6 +771,62 @@ static int i801_block_transaction(struct i801_priv *priv, union i2c_smbus_data *
>  	return result;
>  }
>  
> +/* Single transaction function */

The term "single transaction" is a bit misleading. Block transactions
are also single transactions, in the sense that there's one start
condition at the beginning and one stop condition at the end. I'd
rather call non-block transactions "single value transactions" or
"simple transactions".

> +static int i801_single_transaction(struct i801_priv *priv, union i2c_smbus_data *data,
> +				   char read_write, int command)
> +{
> +	int xact, ret;
> +
> +	switch (command) {
> +	case I2C_SMBUS_QUICK:
> +		xact = I801_QUICK;
> +		break;
> +	case I2C_SMBUS_BYTE:
> +                xact = I801_BYTE;
> +                break;

Previous 2 lines are indented with spaces instead of tabs.

> +	case I2C_SMBUS_BYTE_DATA:
> +		if (read_write == I2C_SMBUS_WRITE)
> +			outb_p(data->byte, SMBHSTDAT0(priv));
> +		xact = I801_BYTE_DATA;
> +		break;
> +	case I2C_SMBUS_WORD_DATA:
> +		if (read_write == I2C_SMBUS_WRITE) {
> +			outb_p(data->word & 0xff, SMBHSTDAT0(priv));
> +			outb_p((data->word & 0xff00) >> 8, SMBHSTDAT1(priv));
> +		}
> +		xact = I801_WORD_DATA;
> +		break;
> +	case I2C_SMBUS_PROC_CALL:
> +		outb_p(data->word & 0xff, SMBHSTDAT0(priv));
> +		outb_p((data->word & 0xff00) >> 8, SMBHSTDAT1(priv));
> +		xact = I801_PROC_CALL;
> +		break;
> +	default:
> +		return -EOPNOTSUPP;

That's never going to happen.

Generally speaking, I'm worried about having the same switch/case
construct here that we already have in i801_access. Looks to me like we
are doing half of the work here and the other half there and I fail to
see the rationale for splitting the work like that. I mean, I see how
it solves the asymmetry between the block and non-block code paths, but
the result doesn't look appealing. From a performance perspective it's
questionable too.

What prevents us from doing all the work on either side? Maybe we
should move more code into i801_single_transaction (possibly in a
subsequent patch)?

> +	}
> +
> +	ret = i801_transaction(priv, xact);
> +

Traditionally no blank line here.

> +	if (ret || read_write == I2C_SMBUS_WRITE)
> +		return ret;
> +
> +	switch (command) {
> +	case I2C_SMBUS_BYTE:
> +	case I2C_SMBUS_BYTE_DATA:
> +		data->byte = inb_p(SMBHSTDAT0(priv));
> +		break;
> +	case I2C_SMBUS_WORD_DATA:
> +	case I2C_SMBUS_PROC_CALL:
> +		data->word = inb_p(SMBHSTDAT0(priv)) +
> +			     (inb_p(SMBHSTDAT1(priv)) << 8);
> +		break;
> +	default:
> +		break;

Default case is not needed.

> +	}
> +
> +	return 0;
> +}
> +
>  static void i801_set_hstadd(struct i801_priv *priv, u8 addr, char read_write)
>  {
>  	addr <<= 1;
> @@ -784,9 +840,7 @@ static s32 i801_access(struct i2c_adapter *adap, u16 addr,
>  		       unsigned short flags, char read_write, u8 command,
>  		       int size, union i2c_smbus_data *data)
>  {
> -	int hwpec;
> -	int block = 0;
> -	int ret, xact;
> +	int hwpec, ret, block = 0;
>  	struct i801_priv *priv = i2c_get_adapdata(adap);
>  
>  	mutex_lock(&priv->acpi_lock);
> @@ -804,36 +858,23 @@ static s32 i801_access(struct i2c_adapter *adap, u16 addr,
>  	switch (size) {
>  	case I2C_SMBUS_QUICK:
>  		i801_set_hstadd(priv, addr, read_write);
> -		xact = I801_QUICK;
>  		break;
>  	case I2C_SMBUS_BYTE:
>  		i801_set_hstadd(priv, addr, read_write);
>  		if (read_write == I2C_SMBUS_WRITE)
>  			outb_p(command, SMBHSTCMD(priv));
> -		xact = I801_BYTE;
>  		break;
>  	case I2C_SMBUS_BYTE_DATA:
>  		i801_set_hstadd(priv, addr, read_write);
>  		outb_p(command, SMBHSTCMD(priv));
> -		if (read_write == I2C_SMBUS_WRITE)
> -			outb_p(data->byte, SMBHSTDAT0(priv));
> -		xact = I801_BYTE_DATA;
>  		break;
>  	case I2C_SMBUS_WORD_DATA:
>  		i801_set_hstadd(priv, addr, read_write);
>  		outb_p(command, SMBHSTCMD(priv));
> -		if (read_write == I2C_SMBUS_WRITE) {
> -			outb_p(data->word & 0xff, SMBHSTDAT0(priv));
> -			outb_p((data->word & 0xff00) >> 8, SMBHSTDAT1(priv));
> -		}
> -		xact = I801_WORD_DATA;
>  		break;
>  	case I2C_SMBUS_PROC_CALL:
>  		i801_set_hstadd(priv, addr, I2C_SMBUS_WRITE);
>  		outb_p(command, SMBHSTCMD(priv));
> -		outb_p(data->word & 0xff, SMBHSTDAT0(priv));
> -		outb_p((data->word & 0xff00) >> 8, SMBHSTDAT1(priv));
> -		xact = I801_PROC_CALL;
>  		read_write = I2C_SMBUS_READ;
>  		break;
>  	case I2C_SMBUS_BLOCK_DATA:
> @@ -883,7 +924,7 @@ static s32 i801_access(struct i2c_adapter *adap, u16 addr,
>  	if (block)
>  		ret = i801_block_transaction(priv, data, read_write, size);
>  	else
> -		ret = i801_transaction(priv, xact);
> +		ret = i801_single_transaction(priv, data, read_write, size);
>  
>  	/* Some BIOSes don't like it when PEC is enabled at reboot or resume
>  	   time, so we forcibly disable it after every transaction. Turn off
> @@ -891,26 +932,6 @@ static s32 i801_access(struct i2c_adapter *adap, u16 addr,
>  	if (hwpec || block)
>  		outb_p(inb_p(SMBAUXCTL(priv)) &
>  		       ~(SMBAUXCTL_CRC | SMBAUXCTL_E32B), SMBAUXCTL(priv));
> -
> -	if (block)
> -		goto out;
> -	if (ret)
> -		goto out;
> -	if ((read_write == I2C_SMBUS_WRITE) || (xact == I801_QUICK))
> -		goto out;
> -
> -	switch (xact) {
> -	case I801_BYTE:	/* Result put in SMBHSTDAT0 */
> -	case I801_BYTE_DATA:
> -		data->byte = inb_p(SMBHSTDAT0(priv));
> -		break;
> -	case I801_WORD_DATA:
> -	case I801_PROC_CALL:
> -		data->word = inb_p(SMBHSTDAT0(priv)) +
> -			     (inb_p(SMBHSTDAT1(priv)) << 8);
> -		break;
> -	}
> -
>  out:
>  	/*
>  	 * Unlock the SMBus device for use by BIOS/ACPI,


-- 
Jean Delvare
SUSE L3 Support

  reply	other threads:[~2022-06-10 11:03 UTC|newest]

Thread overview: 26+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-04-15 16:53 [PATCH 0/8] i2c: i801: Series with minor improvements Heiner Kallweit
2022-04-15 16:54 ` [PATCH 1/8] i2c: i801: improve interrupt handler Heiner Kallweit
2022-06-07 12:34   ` Jean Delvare
2022-12-15 22:15     ` Heiner Kallweit
2022-04-15 16:55 ` [PATCH 2/8] i2c: i801: make FEATURE_HOST_NOTIFY dependent on FEATURE_IRQ Heiner Kallweit
2022-06-07 12:48   ` Jean Delvare
2022-12-16 20:23     ` Heiner Kallweit
2022-04-15 16:55 ` [PATCH 3/8] i2c: i801: make FEATURE_BLOCK_PROC dependent on FEATURE_BLOCK_BUFFER Heiner Kallweit
2022-06-07 14:13   ` Jean Delvare
2022-12-16 20:57     ` Heiner Kallweit
2022-04-15 16:56 ` [PATCH 4/8] i2c: i801: enable FEATURE_IRQ and FEATURE_I2C_BLOCK_READ on all chip versions Heiner Kallweit
2022-06-07 14:24   ` Jean Delvare
2022-06-13 17:08     ` Jean Delvare
2022-06-14 12:59       ` Jean Delvare
2022-12-16 21:36         ` Heiner Kallweit
2022-04-15 16:57 ` [PATCH 5/8] i2c: i801: add helper i801_set_hstadd() Heiner Kallweit
2022-06-09 13:53   ` Jean Delvare
2022-12-16 21:37     ` Heiner Kallweit
2022-04-15 16:58 ` [PATCH 6/8] i2c: i801: add i801_single_transaction(), complementing i801_block_transaction() Heiner Kallweit
2022-06-10 11:03   ` Jean Delvare [this message]
2022-12-17 17:07     ` Heiner Kallweit
2022-04-15 16:58 ` [PATCH 7/8] i2c: i801: call i801_check_pre() from i801_access() Heiner Kallweit
2022-06-10 13:52   ` Jean Delvare
2022-04-15 16:59 ` [PATCH 8/8] i2c: i801: call i801_check_post() " Heiner Kallweit
2022-06-10 14:31   ` Jean Delvare
2022-12-17 17:21     ` Heiner Kallweit

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=20220610130324.1ab2725d@endymion.delvare \
    --to=jdelvare@suse.de \
    --cc=hkallweit1@gmail.com \
    --cc=linux-i2c@vger.kernel.org \
    /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
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).