mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Jonathan Cameron <jic23@kernel.org>
To: Arnav Kapoor <kapoorarnav43@gmail.com>
Cc: "Sai Krishna Potthuri" <sai.krishna.potthuri@amd.com>,
	"Conall O'Griofa" <conall.ogriofa@amd.com>,
	"David Lechner" <dlechner@baylibre.com>,
	"Nuno Sá" <nuno.sa@analog.com>,
	"Andy Shevchenko" <andy@kernel.org>,
	"Michal Simek" <michal.simek@amd.com>,
	"Andy Shevchenko" <andriy.shevchenko@intel.com>,
	stable@vger.kernel.org, linux-arm-kernel@lists.infradead.org,
	linux-iio@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH v2] iio: adc: xilinx-ams: fix out-of-bounds accesses when parsing channels
Date: Sun, 1 Nov 2026 23:52:59 +0000	[thread overview]
Message-ID: <20261101235259.27b3b43c@jic23-hlaptop> (raw)
In-Reply-To: <20261023185652.195025-1-kapoorarnav43@gmail.com>

On Sat, 24 Oct 2026 00:26:48 +0530
Arnav Kapoor <kapoorarnav43@gmail.com> wrote:

> ams_parse_firmware() allocates room for ARRAY_SIZE(ams_ps_channels) +
> ARRAY_SIZE(ams_pl_channels) + ARRAY_SIZE(ams_ctrl_channels) = 51
> channel specs and lets ams_init_module() fill them for the AMS node and
> each of its children, without telling it how much room is left.
> 
> For the PL-SYSMON node, ams_get_ext_chan() appends one channel per
> "channel@N" child after the 10 fixed PL channels. The binding allows
> reg values 20..50, so a PL node can have up to 31 such children, i.e.
> up to 41 PL channels, while only 31 are budgeted for it. Together with
> the 13 PS and 7 AMS control channels, this writes past the end of the
> buffer. Since the modules are filled in device tree order, the fixed
> channel blocks copied after the PL node can overflow as well.
> 
> ams_get_ext_chan() also only checks the upper bound of reg. ext_chan is
> unsigned, so a reg below 20 makes it wrap and ams_pl_channels[] is read
> far out of bounds.
> 
> Pass the remaining capacity down to ams_init_module() and
> ams_get_ext_chan() and fail with -ENOSPC when a module does not fit,
> instead of writing past the buffer. Also reject reg values below 20, as
> the binding requires.
> 
> Fixes: d5c70627a794 ("iio: adc: Add Xilinx AMS driver")
> Cc: stable@vger.kernel.org
> Signed-off-by: Arnav Kapoor <kapoorarnav43@gmail.com>
AMD Xilinx folk, please take a look at this one.

Thanks,

Jonathan

> ---
> Changes in v2:
> - Fix the truncated subject line (Andy)
> - Drop the paragraph about an earlier report, I could not find that
>   thread again to reference it properly
> - Pass the remaining capacity as unsigned int and drop the casts (Andy)
> - Keep the ams_get_ext_chan() call on one line (Andy)
> - Rebase on fixes-togreg
> - Drop the bouncing @xilinx.com and @huawei.com addresses from Cc
> 
> v1: https://lore.kernel.org/all/20261021225356.625288-1-kapoorarnav43@gmail.com/
> 
> Compile-tested only (W=1); I don't have ZynqMP hardware.
> 
>  drivers/iio/adc/xilinx-ams.c | 54 ++++++++++++++++++++++++++++--------
>  1 file changed, 43 insertions(+), 11 deletions(-)
> 
> diff --git a/drivers/iio/adc/xilinx-ams.c b/drivers/iio/adc/xilinx-ams.c
> index 158e6133abf5..7f0f7778d6aa 100644
> --- a/drivers/iio/adc/xilinx-ams.c
> +++ b/drivers/iio/adc/xilinx-ams.c
> @@ -144,7 +144,14 @@
>  #define AMS_PL_CSTS_ACCESS_MASK		BIT(1)
>  
>  #define AMS_PL_MAX_FIXED_CHANNEL	10
> -#define AMS_PL_MAX_EXT_CHANNEL		20
> +/*
> + * The external PL channels are addressed by their channel number. The
> + * binding allows the range from AMS_PL_EXT_CHANNEL_MIN to
> + * AMS_PL_EXT_CHANNEL_MAX. AMS_PL_EXT_CHANNEL_MIN maps onto the first
> + * ams_pl_channels[] entry, AMS_PL_EXT_CHANNEL_MAX onto the last one.
> + */
> +#define AMS_PL_EXT_CHANNEL_MIN		20
> +#define AMS_PL_EXT_CHANNEL_MAX		50
>  
>  #define AMS_INIT_POLL_TIME_US		200
>  #define AMS_INIT_TIMEOUT_US		10000
> @@ -1142,20 +1149,31 @@ static const struct iio_chan_spec ams_ctrl_channels[] = {
>  };
>  
>  static int ams_get_ext_chan(struct fwnode_handle *chan_node,
> -			    struct iio_chan_spec *channels, int num_channels)
> +			    struct iio_chan_spec *channels, int num_channels,
> +			    unsigned int max_channels)
>  {
>  	struct iio_chan_spec *chan;
> -	struct fwnode_handle *child;
>  	unsigned int reg, ext_chan;
>  	int ret;
>  
> -	fwnode_for_each_child_node(chan_node, child) {
> +	fwnode_for_each_child_node_scoped(chan_node, child) {
>  		ret = fwnode_property_read_u32(child, "reg", &reg);
> -		if (ret || reg > AMS_PL_MAX_EXT_CHANNEL + 30)
> +		if (ret)
> +			continue;
> +
> +		/*
> +		 * Anything outside of the range of channel numbers allowed
> +		 * by the binding is not a valid channel.
> +		 */
> +		if (reg < AMS_PL_EXT_CHANNEL_MIN ||
> +		    reg > AMS_PL_EXT_CHANNEL_MAX)
>  			continue;
>  
> +		if (num_channels >= max_channels)
> +			return -ENOSPC;
> +
>  		chan = &channels[num_channels];
> -		ext_chan = reg + AMS_PL_MAX_FIXED_CHANNEL - 30;
> +		ext_chan = reg - AMS_PL_EXT_CHANNEL_MIN;
>  		memcpy(chan, &ams_pl_channels[ext_chan], sizeof(*channels));
>  
>  		if (fwnode_property_read_bool(child, "xlnx,bipolar"))
> @@ -1183,7 +1201,8 @@ static void ams_iounmap_pl(void *data)
>  
>  static int ams_init_module(struct iio_dev *indio_dev,
>  			   struct fwnode_handle *fwnode,
> -			   struct iio_chan_spec *channels)
> +			   struct iio_chan_spec *channels,
> +			   unsigned int max_channels)
>  {
>  	struct device *dev = indio_dev->dev.parent;
>  	struct ams *ams = iio_priv(indio_dev);
> @@ -1191,6 +1210,9 @@ static int ams_init_module(struct iio_dev *indio_dev,
>  	int ret;
>  
>  	if (fwnode_device_is_compatible(fwnode, "xlnx,zynqmp-ams-ps")) {
> +		if (max_channels < ARRAY_SIZE(ams_ps_channels))
> +			return -ENOSPC;
> +
>  		ams->ps_base = fwnode_iomap(fwnode, 0);
>  		if (!ams->ps_base)
>  			return -ENXIO;
> @@ -1202,6 +1224,9 @@ static int ams_init_module(struct iio_dev *indio_dev,
>  		memcpy(channels, ams_ps_channels, sizeof(ams_ps_channels));
>  		num_channels = ARRAY_SIZE(ams_ps_channels);
>  	} else if (fwnode_device_is_compatible(fwnode, "xlnx,zynqmp-ams-pl")) {
> +		if (max_channels < AMS_PL_MAX_FIXED_CHANNEL)
> +			return -ENOSPC;
> +
>  		ams->pl_base = fwnode_iomap(fwnode, 0);
>  		if (!ams->pl_base)
>  			return -ENXIO;
> @@ -1213,9 +1238,14 @@ static int ams_init_module(struct iio_dev *indio_dev,
>  		/* Copy only first 10 fix channels */
>  		memcpy(channels, ams_pl_channels, AMS_PL_MAX_FIXED_CHANNEL * sizeof(*channels));
>  		num_channels += AMS_PL_MAX_FIXED_CHANNEL;
> -		num_channels = ams_get_ext_chan(fwnode, channels,
> -						num_channels);
> +		ret = ams_get_ext_chan(fwnode, channels, num_channels, max_channels);
> +		if (ret < 0)
> +			return ret;
> +		num_channels = ret;
>  	} else if (fwnode_device_is_compatible(fwnode, "xlnx,zynqmp-ams")) {
> +		if (max_channels < ARRAY_SIZE(ams_ctrl_channels))
> +			return -ENOSPC;
> +
>  		/* add AMS channels to iio device channels */
>  		memcpy(channels, ams_ctrl_channels, sizeof(ams_ctrl_channels));
>  		num_channels += ARRAY_SIZE(ams_ctrl_channels);
> @@ -1245,7 +1275,7 @@ static int ams_parse_firmware(struct iio_dev *indio_dev)
>  		return -ENOMEM;
>  
>  	if (fwnode_device_is_available(fwnode)) {
> -		ret = ams_init_module(indio_dev, fwnode, ams_channels);
> +		ret = ams_init_module(indio_dev, fwnode, ams_channels, ams_size);
>  		if (ret < 0)
>  			return ret;
>  
> @@ -1253,7 +1283,9 @@ static int ams_parse_firmware(struct iio_dev *indio_dev)
>  	}
>  
>  	device_for_each_child_node_scoped(dev, child) {
> -		ret = ams_init_module(indio_dev, child, ams_channels + num_channels);
> +		ret = ams_init_module(indio_dev, child,
> +				      ams_channels + num_channels,
> +				      ams_size - num_channels);
>  		if (ret < 0)
>  			return ret;
>  
> 
> base-commit: 2ef6f5c4ee56e90c237c5f7687f09e635f7d30a9


  reply	other threads:[~2026-11-01 23:53 UTC|newest]

Thread overview: 6+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-10-23 18:56 Arnav Kapoor
2026-11-01 23:52 ` Jonathan Cameron [this message]
2026-11-02  7:11 ` Sai Krishna Potthuri
2026-11-02  9:14   ` Arnav Kapoor
  -- strict thread matches above, loose matches on Subject: below --
2026-10-22 14:06 Arnav Kapoor
2026-10-23 19:29 ` Arnav Kapoor

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=20261101235259.27b3b43c@jic23-hlaptop \
    --to=jic23@kernel.org \
    --cc=andriy.shevchenko@intel.com \
    --cc=andy@kernel.org \
    --cc=conall.ogriofa@amd.com \
    --cc=dlechner@baylibre.com \
    --cc=kapoorarnav43@gmail.com \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-iio@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=michal.simek@amd.com \
    --cc=nuno.sa@analog.com \
    --cc=sai.krishna.potthuri@amd.com \
    --cc=stable@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

all inboxes | Powered by JetHome®