public inbox for linux-i2c@vger.kernel.org
 help / color / mirror / Atom feed
From: Yixun Lan <dlan@gentoo.org>
To: Troy Mitchell <troy.mitchell@linux.spacemit.com>
Cc: Andi Shyti <andi.shyti@kernel.org>,
	Alex Elder <elder@riscstar.com>,
	Troy Mitchell <troymitchell988@gmail.com>,
	linux-i2c@vger.kernel.org, linux-kernel@vger.kernel.org,
	linux-riscv@lists.infradead.org, spacemit@lists.linux.dev
Subject: Re: [PATCH v3] i2c: spacemit: introduce pio for k1
Date: Tue, 30 Sep 2025 09:27:02 +0800	[thread overview]
Message-ID: <20250930012702-GYA1357001@gentoo.org> (raw)
In-Reply-To: <20250929-k1-i2c-atomic-v3-1-f7e660c138b6@linux.spacemit.com>

Hi Troy and I2C maintainer,

one thing I'd raise and seek I2C maintainer's suggestion, should this
driver introduce lock to protect multi context - non-atomic vs atomic vs
interrupt context, more than one i2c clients contend single controller?

On 14:07 Mon 29 Sep     , Troy Mitchell wrote:
> This patch introduces I2C PIO functionality for the Spacemit K1 SoC,
> enabling the use of I2C in atomic context.
> 
> Signed-off-by: Troy Mitchell <troy.mitchell@linux.spacemit.com>
> ---
> Changes in v3:
> - drop 1-5 patches (have been merged)
> - modify commit message
> - use readl_poll_timeout_atomic() in wait_pio_xfer()
> - use msecs_to_jiffies() when get PIO mode timeout value
> - factor out transfer state handling into spacemit_i2c_handle_state().
> - do not disable/enable the controller IRQ around PIO transfers.
> - consolidate spacemit_i2c_init() interrupt setup
> - rename is_pio -> use_pio
> - rename spacemit_i2c_xfer() -> spacemit_i2c_xfer_common()
> - rename spacemit_i2c_int_xfer() -> spacemit_i2c_xfer()
> - rename spacemit_i2c_pio_xfer() -> spacemit_i2c_pio_xfer_atomic()
> - call spacemit_i2c_err_check() in wait_pio_xfer() when write last byte
> - Link to v2: https://lore.kernel.org/r/20250925-k1-i2c-atomic-v2-0-46dc13311cda@linux.spacemit.com
> 
> Changes in v2:
> - Patch 1/6:
>   Patch 3/6:
>   Patch 4/6:
>     - nothing
> - Patch 2/6:
>   - remove is_pio judgement in irq_handler()
> - Patch 5/6:
>   - fix wrong comment
>   - In spacemit_i2c_conditionally_reset_bus(), once the condition is met inside the loop, it should
>     return directly instead of using break.
> - Patch 6/6:
>   - add is_pio judgement in irq_handler()
>   - use a fixed timeout value when PIO
>   - use readl_poll_timeout() in  spacemit_i2c_wait_bus_idle() when PIO
> 
> - Link to v1: https://lore.kernel.org/r/20250827-k1-i2c-atomic-v1-0-e59bea02d680@linux.spacemit.com
> ---
>  drivers/i2c/busses/i2c-k1.c | 206 ++++++++++++++++++++++++++++++++++----------
>  1 file changed, 161 insertions(+), 45 deletions(-)
> 
> diff --git a/drivers/i2c/busses/i2c-k1.c b/drivers/i2c/busses/i2c-k1.c
> index 6b918770e612e098b8ad17418f420d87c94df166..21c5dd56e71f5e59434767329fb84d50d5c04178 100644
> --- a/drivers/i2c/busses/i2c-k1.c
> +++ b/drivers/i2c/busses/i2c-k1.c
> @@ -97,6 +97,9 @@
>  
>  #define SPACEMIT_BUS_RESET_CLK_CNT_MAX		9
>  
[snip]..

> +static void spacemit_i2c_handle_state(struct spacemit_i2c_dev *i2c);
> +static void spacemit_i2c_err_check(struct spacemit_i2c_dev *i2c);
please sort to avoid this kind declaration, if it's inevitable, 
you can move to top, before the function, after macro 

> +static int spacemit_i2c_wait_pio_xfer(struct spacemit_i2c_dev *i2c)
> +{
> +	u32 msec = jiffies_to_msecs(i2c->adapt.timeout);
> +	ktime_t timeout = ktime_add_ms(ktime_get(), msec);
> +	int ret;
> +
> +	while (i2c->unprocessed && ktime_compare(ktime_get(), timeout) < 0) {
> +		udelay(10);
> +		i2c->status = readl(i2c->base + SPACEMIT_ISR);
> +
> +		spacemit_i2c_clear_int_status(i2c, i2c->status);
> +
> +		if (!(i2c->status & SPACEMIT_SR_IRF) && !(i2c->status & SPACEMIT_SR_ITE))
this is hard to read
		mask = SPACEMIT_SR_IRF | SPACEMIT_SR_ITE;
		if ((i2c->status & mask) == 0)
		..
> +			continue;
> +
> +		spacemit_i2c_handle_state(i2c);
> +
> +		/*
> +		 * This is the last byte to write of the current message.
> +		 * If we do not wait here, control will proceed directly to start(),
> +		 * which would overwrite the current data.
> +		 */
> +		if (!i2c->read && !i2c->unprocessed) {
> +			ret = readl_poll_timeout_atomic(i2c->base + SPACEMIT_ISR,
> +							i2c->status, i2c->status & SPACEMIT_SR_ITE,
> +							30, 1000);
> +			if (ret)
> +				return 0;
> +			/*
> +			 * In read mode, err_check() in handle_state properly handles what happens
> +			 * after the MSD bit is set. For writes it is different: in interrupt mode,
> +			 * an ITE (write-empty) interrupt is triggered after the last byte, and the
> +			 * MSD-related handling takes place there. In PIO mode, however, we need to
> +			 * explicitly call err_check() to emulate this step, otherwise the next
> +			 * transfer will fail.
> +			 */
> +			spacemit_i2c_err_check(i2c);
> +		}
> +	}
> +
> +	if (i2c->unprocessed)
> +		return 0;
> +
> +	return 1;
> +}
> +
>  static int spacemit_i2c_xfer_msg(struct spacemit_i2c_dev *i2c)
>  {
>  	unsigned long time_left;
> @@ -310,10 +376,16 @@ static int spacemit_i2c_xfer_msg(struct spacemit_i2c_dev *i2c)
>  
>  		reinit_completion(&i2c->complete);
>  
> -		spacemit_i2c_start(i2c);
..
> +		if (i2c->use_pio) {
> +			spacemit_i2c_start(i2c);
> +
> +			time_left = spacemit_i2c_wait_pio_xfer(i2c);
> +		} else {
> +			spacemit_i2c_start(i2c);
> +			time_left = wait_for_completion_timeout(&i2c->complete,
> +								i2c->adapt.timeout);
> +		}
suggest to distill a common function spacemit_i2c_wait_xfer_complete()
and converge above if/elselogic inside, which is more readable

>  
> -		time_left = wait_for_completion_timeout(&i2c->complete,
> -							i2c->adapt.timeout);
>  		if (!time_left) {
>  			dev_err(i2c->dev, "msg completion timeout\n");
..
>  			spacemit_i2c_conditionally_reset_bus(i2c);
btw, not introduced in this patch, spacemit_i2c_reset() followed unconditionally after
spacemit_i2c_conditionally_reset_bus() which is very confusing

please double check the logic


> @@ -341,6 +413,9 @@ static bool spacemit_i2c_is_last_msg(struct spacemit_i2c_dev *i2c)
>  
>  static void spacemit_i2c_handle_write(struct spacemit_i2c_dev *i2c)
>  {
> +	if (!(i2c->status & SPACEMIT_SR_ITE))
> +		return;
> +
>  	/* if transfer completes, SPACEMIT_ISR will handle it */
>  	if (i2c->status & SPACEMIT_SR_MSD)
>  		return;
[snip]..
>  
> -static int spacemit_i2c_xfer(struct i2c_adapter *adapt, struct i2c_msg *msgs, int num)
> +static inline int
> +spacemit_i2c_xfer_common(struct i2c_adapter *adapt, struct i2c_msg *msgs, int num, bool use_pio)
>  {
>  	struct spacemit_i2c_dev *i2c = i2c_get_adapdata(adapt);
>  	int ret;
>  
> +	i2c->use_pio = use_pio;
> +
>  	i2c->msgs = msgs;
>  	i2c->msg_num = num;
>  
..
> -	spacemit_i2c_calc_timeout(i2c);
> +	if (!i2c->use_pio)
> +		spacemit_i2c_calc_timeout(i2c);
> +	else
> +		i2c->adapt.timeout = msecs_to_jiffies(SPACEMIT_WAIT_TIMEOUT);
suggest merge "if else" into spacemit_i2c_calc_timeout(), same as previous suggestion
>  
>  	spacemit_i2c_init(i2c);
>  
>  	spacemit_i2c_enable(i2c);
>  
> +	/* To avoid being affected by the state of the previous transfer */
> +	spacemit_i2c_clear_int_status(i2c, SPACEMIT_I2C_INT_STATUS_MASK);
> +
I've not see my previous comment has been addressed!

>  	ret = spacemit_i2c_wait_bus_idle(i2c);
>  	if (!ret) {
>  		ret = spacemit_i2c_xfer_msg(i2c);
> @@ -506,11 +611,21 @@ static int spacemit_i2c_xfer(struct i2c_adapter *adapt, struct i2c_msg *msgs, in
>  
>  	if (ret == -ETIMEDOUT || ret == -EAGAIN)
>  		dev_err(i2c->dev, "i2c transfer failed, ret %d err 0x%lx\n",
> -			  ret, i2c->status & SPACEMIT_SR_ERR);
> +			ret, i2c->status & SPACEMIT_SR_ERR);
>  
have to say, I personally don't like those unrelated changes, which
makes people review the patch even difficult
>  	return ret < 0 ? ret : num;
>  }
>  
> +static int spacemit_i2c_xfer(struct i2c_adapter *adapt, struct i2c_msg *msgs, int num)
> +{
> +	return spacemit_i2c_xfer_common(adapt, msgs, num, false);
> +}
> +
> +static int spacemit_i2c_pio_xfer_atomic(struct i2c_adapter *adapt, struct i2c_msg *msgs, int num)
> +{
> +	return spacemit_i2c_xfer_common(adapt, msgs, num, true);
> +}
> +

atomic API should not call sleepable function, please check this
spacemit_i2c_pio_xfer_atomic() -> 
  spacemit_i2c_xfer_common() -> 
    spacemit_i2c_xfer_msg() -> 
      spacemit_i2c_conditionally_reset_bus() -> 
        usleep_range()
    spacemit_i2c_check_bus_release() -> 
      usleep_range()

they are in err handle path, but still need to take care with caution


-- 
Yixun Lan (dlan)

  parent reply	other threads:[~2025-09-30  1:27 UTC|newest]

Thread overview: 5+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-09-29  6:07 [PATCH v3] i2c: spacemit: introduce pio for k1 Troy Mitchell
2025-09-29  9:15 ` Troy Mitchell
2025-09-30  1:27 ` Yixun Lan [this message]
2025-10-09  1:32   ` Troy Mitchell
2025-10-09  3:37     ` Troy Mitchell

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=20250930012702-GYA1357001@gentoo.org \
    --to=dlan@gentoo.org \
    --cc=andi.shyti@kernel.org \
    --cc=elder@riscstar.com \
    --cc=linux-i2c@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-riscv@lists.infradead.org \
    --cc=spacemit@lists.linux.dev \
    --cc=troy.mitchell@linux.spacemit.com \
    --cc=troymitchell988@gmail.com \
    /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