Devicetree
 help / color / mirror / Atom feed
From: sashiko-bot@kernel.org
To: "Antoni Pokusinski" <apokusinski01@gmail.com>
Cc: devicetree@vger.kernel.org, dri-devel@lists.freedesktop.org,
	robh@kernel.org, conor+dt@kernel.org, media-ci@linuxtv.org
Subject: Re: [PATCH v2 4/4] i2c: omap: add DMA support
Date: Wed, 18 Nov 2026 21:10:12 +0000	[thread overview]
Message-ID: <sashiko-outbox-221973@kernel.org> (raw)
In-Reply-To: <20261118205213.88778-5-apokusinski01@gmail.com>

Thank you for your contribution! Sashiko AI review found 3 potential
issues to consider:
- [High] i2c: omap: DMA wait ordering and cmd_err gating in
  omap_i2c_xfer_msg()
- [Medium] i2c: omap: NULL dereference in omap_i2c_dma_free() on probe
  error
- [Medium] i2c: omap: uninitialized pio_irq_mask and missing DMA cleanup
  on timeout
--

commit 7af8eb0f57b65f48015dfec1a234f795398aca59
Author: Antoni Pokusinski <apokusinski01@gmail.com>

i2c: omap: add DMA support

This commit adds TX and RX DMA transfer support to the OMAP I2C driver for
messages of at least 8 bytes whose length is a multiple of the configured
FIFO threshold.

> diff --git a/drivers/i2c/busses/i2c-omap.c b/drivers/i2c/busses/i2c-omap.c
> index 3f7060d57a869..acff27f3f4385 100644
> --- a/drivers/i2c/busses/i2c-omap.c
> +++ b/drivers/i2c/busses/i2c-omap.c
[ ... ]
> @@ -663,8 +837,8 @@ static int omap_i2c_xfer_msg(struct i2c_adapter *adap,
>  {
>  	struct omap_i2c_dev *omap = i2c_get_adapdata(adap);
>  	unsigned long time_left;
> -	u16 w;
> -	int ret;
> +	u16 w, pio_irq_mask;
> +	int ret, use_dma;
>  
>  	dev_dbg(omap->dev, "addr: 0x%04x, len: %d, flags: 0x%x, stop: %d\n",
>  		msg->addr, msg->len, msg->flags, stop);
[ ... ]
> @@ -692,6 +874,27 @@ static int omap_i2c_xfer_msg(struct i2c_adapter *adap,
>  		reinit_completion(&omap->cmd_complete);
>  	omap->cmd_err = 0;
>  
> +	if (use_dma) {
> +		ret = omap_i2c_dma_xfer(omap, msg);
> +		if (ret) {
> +			omap_i2c_reset(omap);
> +			__omap_i2c_init(omap);
> +			return ret;
> +		}
> +
> +		omap_i2c_dma_enable_req(omap);
> +	} else {
> +		pio_irq_mask = OMAP_I2C_IE_XRDY | OMAP_I2C_IE_RRDY |
> +			       ((omap->fifo_size) ? (OMAP_I2C_IE_RDR |
> +						    OMAP_I2C_IE_XDR) : 0);
> +
> +		if (omap->scheme == OMAP_I2C_SCHEME_0)
> +			omap_i2c_write_reg(omap, OMAP_I2C_IE_REG,
> +					   omap->iestate | pio_irq_mask);
> +		else
> +			omap_i2c_write_reg(omap, OMAP_I2C_IE_REG, pio_irq_mask);
> +	}
[ ... ]
> @@ -727,6 +930,16 @@ static int omap_i2c_xfer_msg(struct i2c_adapter *adap,
>  			if (time_after(jiffies, delay)) {
>  				dev_err(omap->dev, "controller timed out "
>  				"waiting for start condition to finish\n");
> +
> +				if (omap->scheme == OMAP_I2C_SCHEME_0)
> +					omap_i2c_write_reg(omap,
> +							   OMAP_I2C_IE_REG,
> +							   omap->iestate);
> +				else
> +					omap_i2c_write_reg(omap,
> +							   OMAP_I2C_IP_V2_IRQENABLE_CLR,
> +							   pio_irq_mask);
> +
>  				return -ETIMEDOUT;
>  			}
>  			cpu_relax();

[Severity: Medium]
Is pio_irq_mask used uninitialized here when use_dma is true?

pio_irq_mask is only assigned in the !use_dma branch, which triggers a
-Wmaybe-uninitialized warning on this timeout path.  If this timeout is
reached during a DMA transfer, returning directly also leaves the DMA
channel active and leaks the DMA mapping and bounce buffer without calling
dmaengine_terminate_sync(), omap_i2c_dma_cleanup() and
omap_i2c_dma_disable_req().

> @@ -742,8 +955,15 @@ static int omap_i2c_xfer_msg(struct i2c_adapter *adap,
>  	 * into arbitration and we're currently unable to recover from it.
>  	 */
>  	if (!polling) {
> -		time_left = wait_for_completion_timeout(&omap->cmd_complete,
> -							OMAP_I2C_TIMEOUT);
> +		time_left = OMAP_I2C_TIMEOUT;
> +
> +		if (use_dma)
> +			time_left = wait_for_completion_timeout(&omap->dma->complete,
> +								time_left);
> +
> +		if (time_left && !omap->cmd_err)
> +			time_left = wait_for_completion_timeout(&omap->cmd_complete,
> +								time_left);
>  	} else {
>  		do {
>  			omap_i2c_wait(omap);
> @@ -753,7 +973,23 @@ static int omap_i2c_xfer_msg(struct i2c_adapter *adap,
>  		time_left = !ret;
>  	}
>  
> -	if (time_left == 0) {
> +	if (use_dma) {
> +		if (!time_left || omap->cmd_err)
> +			dmaengine_terminate_sync(omap->dma->chan_using);
> +
> +		omap_i2c_dma_cleanup(omap, time_left && !omap->cmd_err);
> +
> +		omap_i2c_dma_disable_req(omap);
> +	} else {
> +		if (omap->scheme == OMAP_I2C_SCHEME_0)
> +			omap_i2c_write_reg(omap, OMAP_I2C_IE_REG,
> +					   omap->iestate);
> +		else
> +			omap_i2c_write_reg(omap, OMAP_I2C_IP_V2_IRQENABLE_CLR,
> +					   pio_irq_mask);
> +	}
> +
> +	if (!time_left && likely(!omap->cmd_err)) {
>  		omap_i2c_reset(omap);
>  		__omap_i2c_init(omap);
>  		return -ETIMEDOUT;

[Severity: High]
Does waiting on omap->dma->complete before omap->cmd_complete stall for
the full OMAP_I2C_TIMEOUT (1 second) whenever a DMA transfer encounters a
NACK or arbitration loss?

When a bus error occurs before all bytes are transferred (for example, on
any RX DMA transfer or multi-burst TX DMA transfer), the controller stops
asserting DMA requests and the IRQ handler signals omap->cmd_complete via
omap_i2c_complete_cmd(), while omap->dma->complete is never signaled.

Additionally, in omap_i2c_xfer_data(), a standalone NACK interrupt sets
omap->cmd_err |= OMAP_I2C_STAT_NACK and returns -EAGAIN to wait for the
subsequent ARDY interrupt before completing omap->cmd_complete:

drivers/i2c/busses/i2c-omap.c:omap_i2c_xfer_data() {
    ...
		if (stat & OMAP_I2C_STAT_NACK) {
			omap->cmd_err |= OMAP_I2C_STAT_NACK;
			omap_i2c_ack_stat(omap, OMAP_I2C_STAT_NACK);

			if (!(stat & ~OMAP_I2C_STAT_NACK)) {
				err = -EAGAIN;
				break;
			}
		}
    ...
}

Guarding wait_for_completion_timeout(&omap->cmd_complete, ...) with
!omap->cmd_err causes omap_i2c_xfer_msg() to skip waiting for ARDY if the
standalone NACK IRQ has already set omap->cmd_err, accessing controller
registers before hardware is ready and leaving a stale completion on
omap->cmd_complete when ARDY later fires.  Similarly, changing the timeout
check to if (!time_left && likely(!omap->cmd_err)) skips
omap_i2c_reset() and fails to return -ETIMEDOUT when waiting for ARDY
times out after a standalone NACK interrupt (and even returns 0 if
I2C_M_IGNORE_NAK is set).

[ ... ]
> @@ -1336,10 +1562,70 @@ static struct i2c_bus_recovery_info omap_i2c_bus_recovery_info = {
>  	.recover_bus		= i2c_generic_scl_recovery,
>  };
>  
[ ... ]
> +static void omap_i2c_dma_free(struct omap_i2c_dev *omap)
> +{
> +	struct omap_i2c_dma *dma = omap->dma;
> +
> +	dma->dma_buf = 0;
> +	dma->dma_len = 0;
> +
> +	dma_release_channel(dma->chan_tx);
> +	dma->chan_tx = NULL;
> +
> +	dma_release_channel(dma->chan_rx);
> +	dma->chan_rx = NULL;
> +
> +	dma->chan_using = NULL;
> +}
[ ... ]
> @@ -1476,6 +1762,22 @@ omap_i2c_probe(struct platform_device *pdev)
>  		goto err_unuse_clocks;
>  	}
>  
> +	if (omap->scheme != OMAP_I2C_SCHEME_0) {
> +		omap->dma = omap_i2c_dma_init(omap, (dma_addr_t)res->start);
> +		if (PTR_ERR(omap->dma) == -EPROBE_DEFER) {
> +			r = -EPROBE_DEFER;
> +			dev_err_probe(&pdev->dev, r,
> +				      "can't get DMA channels\n");
> +			goto err_unuse_clocks;
> +		} else if (PTR_ERR(omap->dma) == -ENODEV) {
> +			dev_dbg(omap->dev, "DMA setup failed\n");
> +			omap->dma = NULL;
> +		} else if (IS_ERR(omap->dma)) {
> +			r = PTR_ERR(omap->dma);
> +			goto err_unuse_clocks;
> +		}
> +	}
> +
>  	adap = &omap->adapter;
>  	i2c_set_adapdata(adap, omap);
>  	adap->owner = THIS_MODULE;
> @@ -1491,7 +1793,7 @@ omap_i2c_probe(struct platform_device *pdev)
>  	adap->nr = pdev->id;
>  	r = i2c_add_numbered_adapter(adap);
>  	if (r)
> -		goto err_unuse_clocks;
> +		goto err_dma_free;
>  
>  	dev_info(omap->dev, "bus %d rev%d.%d at %d kHz\n", adap->nr,
>  		 major, minor, omap->speed);
> @@ -1500,6 +1802,8 @@ omap_i2c_probe(struct platform_device *pdev)
>  
>  	return 0;
>  
> +err_dma_free:
> +	omap_i2c_dma_free(omap);

[Severity: Medium]
Can err_dma_free trigger a NULL pointer dereference in
omap_i2c_dma_free()?

When omap->scheme == OMAP_I2C_SCHEME_0 or omap_i2c_dma_init() returns
ERR_PTR(-ENODEV), omap->dma is NULL.  If i2c_add_numbered_adapter() then
fails, omap_i2c_dma_free(omap) is called without an if (omap->dma) check
and dereferences omap->dma.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20261118205213.88778-1-apokusinski01@gmail.com?part=4

      reply	other threads:[~2026-11-18 21:10 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-11-18 20:52 [PATCH v2 0/4] i2c: omap: add support for DMA Antoni Pokusinski
2026-11-18 20:52 ` [PATCH v2 1/4] dt-bindings: i2c: ti,omap4-i2c: Add DMA properties Antoni Pokusinski
2026-11-18 20:52 ` [PATCH v2 2/4] ARM: dts: am33xx-l4: Add missing DMAs for I2C controllers Antoni Pokusinski
2026-11-18 20:52 ` [PATCH v2 3/4] i2c: omap: remove unused is_rdr and is_xdr function arguments Antoni Pokusinski
2026-11-18 20:52 ` [PATCH v2 4/4] i2c: omap: add DMA support Antoni Pokusinski
2026-11-18 21:10   ` sashiko-bot [this message]

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=sashiko-outbox-221973@kernel.org \
    --to=sashiko-bot@kernel.org \
    --cc=apokusinski01@gmail.com \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=media-ci@linuxtv.org \
    --cc=robh@kernel.org \
    --cc=sashiko-reviews@lists.linux.dev \
    /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