Skip to contentSkip to conversation

i2c: cp2615: Validate read length before copying

PATCH In review Triet Hoang 3 participants 5 messages Aug 18 – Aug 21

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 i2c: cp2615: Validate read length before copying

The read_len field comes from the untrusted USB payload, which could potentially exceed the client's originally requested buffer length or MAX_I2C_SIZE, allowing an out-of-bounds read and write.

Limit read_len to the maximum size of the response buffer and return -EPROTO for an invalid response.

---
 drivers/i2c/busses/i2c-cp2615.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)
drivers/i2c/busses/i2c-cp2615.c +7 -2
index 951de6249834..2c79530da306 100644
@@ -136,7 +136,7 @@ cp2615_i2c_send(struct usb_interface *usbif, struct cp2615_i2c_transfer *i2c_w)
 }
 
 static int
-cp2615_i2c_recv(struct usb_interface *usbif, unsigned char tag, void *buf)
+cp2615_i2c_recv(struct usb_interface *usbif, unsigned char tag, void *buf, int len)
 {
 	struct usb_device *usbdev = interface_to_usbdev(usbif);
 	struct cp2615_iop_msg *msg;
@@ -160,6 +160,11 @@ cp2615_i2c_recv(struct usb_interface *usbif, unsigned char tag, void *buf)
 		return -EIO;
 	}
 
+	if (i2c_r->read_len > len || i2c_r->read_len > MAX_I2C_SIZE) {
+		kfree(msg);
+		return -EPROTO;
+	}
+
 	res = cp2615_check_status(i2c_r->status);
 	if (!res)
 		memcpy(buf, &i2c_r->data, i2c_r->read_len);
@@ -236,7 +241,7 @@ cp2615_i2c_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs, int num)
 		ret = cp2615_i2c_send(usbif, &i2c_w);
 		if (ret)
 			break;
-		ret = cp2615_i2c_recv(usbif, i2c_w.tag, msg->buf);
+		ret = cp2615_i2c_recv(usbif, i2c_w.tag, msg->buf, msg->len);
 	}
 	if (ret < 0)
 		return ret;
Signature · 1 line
2.53.0
#1 lore raw

…

Quoted 2 lines

Limit read_len to the maximum size of the response buffer and return -EPROTO for an invalid response.

* How do you think about to add any tags (like “Fixes” and “Cc”) accordingly?
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/submitting-patches.rst?h=v7.2#n145
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/stable-kernel-rules.rst?h=v7.2#n34

* Please avoid duplicate source code also for improved implementations of functions like cp2615_i2c_recv().
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/Documentation/process/coding-style.rst?h=v7.2#n526
https://elixir.bootlin.com/linux/v7.2/source/drivers/i2c/busses/i2c-cp2615.c#L138-L169

* How do you think about to increase the application of scope-based resource management?

Regards,
Markus

maintainer
#2 lore raw
In reply to Triet Hoang

Hi Triet,

Quoted 3 lines · Triet Hoang

The read_len field comes from the untrusted USB payload, which could potentially exceed the client's originally requested buffer length or MAX_I2C_SIZE, allowing an out-of-bounds read and write.

have you experienced any issue yourself or are you just speculating based on the code?

Andi

Quoted 4 lines

Limit read_len to the maximum size of the response buffer and return -EPROTO for an invalid response.

#3 lore raw

On Thu, 20 Aug 2026 10:28:22 -0500, Andi wrote

Quoted 2 lines

have you experienced any issue yourself or are you just speculating based on the code?

Hi Andi,

Thanks for replying. I just speculating based on the code.

BR,
Triet

maintainer
#4 lore raw

Hi Triet,

Quoted 5 lines · Triet Hoang

On Thu, 20 Aug 2026 10:28:22 -0500, Andi wrote

Quoted 2 lines

have you experienced any issue yourself or are you just speculating based on the code?

Thanks for replying. I just speculating based on the code.

I'm sorry, but I'm going to drop this as I don't see it as necessary.

As Markus pointed out, if this failure could actually happen, it would result in a driver failure and the patch should then be treated as a bug fix. However, I don't want to fix hypothetical bugs based on speculation or an AI-generated guess.

Thanks,
Andi