[PATCH] Input: ims-pcu - fix heap-buffer-overflow in ims_pcu_process_data()

[PATCH] Input: ims-pcu - fix heap-buffer-overflow in ims_pcu_process_data()

[pip-izony]

From: Seungjin Bae <eeodqql09@gmail.com>

The `ims_pcu_process_data()` processes incoming URB data byte by byte.
However, it fails to check if the `read_pos` index exceeds
IMS_PCU_BUF_SIZE.

If a malicious USB device sends a packet larger than IMS_PCU_BUF_SIZE,
`read_pos` will increment indefinitely. Moreover, since `read_pos` is
located immediately after `read_buf`, the attacker can overwrite
`read_pos` itself to arbitrarily control the index.

This manipulated `read_pos` is subsequently used in
`ims_pcu_handle_response()` to copy data into `cmd_buf`, leading to a
heap buffer overflow.

Specifically, an attacker can overwrite the `cmd_done.wait.head` located
at offset 136 relative to `cmd_buf` in the `ims_pcu_handle_response()`.
Consequently, when the driver calls `complete(&pcu->cmd_done)`, it
triggers a control flow hijack by using the manipulated pointer.

Fix this by adding a bounds check for `read_pos` before writing to
`read_buf`.

Fixes: 628329d524743 ("Input: add IMS Passenger Control Unit driver")
Signed-off-by: Seungjin Bae <eeodqql09@gmail.com>
---
 drivers/input/misc/ims-pcu.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/input/misc/ims-pcu.c b/drivers/input/misc/ims-pcu.c
index 4581f1c53644..402bf03ca8f6 100644
--- a/drivers/input/misc/ims-pcu.c
+++ b/drivers/input/misc/ims-pcu.c
@@ -451,7 +451,8 @@ static void ims_pcu_process_data(struct ims_pcu *pcu, struct urb *urb)
 
 		if (pcu->have_dle) {
 			pcu->have_dle = false;
-			pcu->read_buf[pcu->read_pos++] = data;
+			if (pcu->read_pos < IMS_PCU_BUF_SIZE)
+				pcu->read_buf[pcu->read_pos++] = data;
 			pcu->check_sum += data;
 			continue;
 		}
@@ -491,7 +492,8 @@ static void ims_pcu_process_data(struct ims_pcu *pcu, struct urb *urb)
 			break;
 
 		default:
-			pcu->read_buf[pcu->read_pos++] = data;
+			if (pcu->read_pos < IMS_PCU_BUF_SIZE)
+				pcu->read_buf[pcu->read_pos++] = data;
 			pcu->check_sum += data;
 			break;
 		}
-- 
2.43.0

Re: [PATCH] Input: ims-pcu - fix heap-buffer-overflow in ims_pcu_process_data()

[Dmitry Torokhov]

Hi Seungjin,

On Fri, Dec 19, 2025 at 05:34:18PM -0500, pip-izony wrote:
> From: Seungjin Bae <eeodqql09@gmail.com>
> 
> The `ims_pcu_process_data()` processes incoming URB data byte by byte.
> However, it fails to check if the `read_pos` index exceeds
> IMS_PCU_BUF_SIZE.
> 
> If a malicious USB device sends a packet larger than IMS_PCU_BUF_SIZE,
> `read_pos` will increment indefinitely. Moreover, since `read_pos` is
> located immediately after `read_buf`, the attacker can overwrite
> `read_pos` itself to arbitrarily control the index.
> 
> This manipulated `read_pos` is subsequently used in
> `ims_pcu_handle_response()` to copy data into `cmd_buf`, leading to a
> heap buffer overflow.
> 
> Specifically, an attacker can overwrite the `cmd_done.wait.head` located
> at offset 136 relative to `cmd_buf` in the `ims_pcu_handle_response()`.
> Consequently, when the driver calls `complete(&pcu->cmd_done)`, it
> triggers a control flow hijack by using the manipulated pointer.
> 
> Fix this by adding a bounds check for `read_pos` before writing to
> `read_buf`.
> 
> Fixes: 628329d524743 ("Input: add IMS Passenger Control Unit driver")
> Signed-off-by: Seungjin Bae <eeodqql09@gmail.com>
> ---
>  drivers/input/misc/ims-pcu.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/input/misc/ims-pcu.c b/drivers/input/misc/ims-pcu.c
> index 4581f1c53644..402bf03ca8f6 100644
> --- a/drivers/input/misc/ims-pcu.c
> +++ b/drivers/input/misc/ims-pcu.c
> @@ -451,7 +451,8 @@ static void ims_pcu_process_data(struct ims_pcu *pcu, struct urb *urb)
>  
>  		if (pcu->have_dle) {
>  			pcu->have_dle = false;
> -			pcu->read_buf[pcu->read_pos++] = data;
> +			if (pcu->read_pos < IMS_PCU_BUF_SIZE)
> +				pcu->read_buf[pcu->read_pos++] = data;

I think we might want a warning if we're overflowing the buffer.

>  			pcu->check_sum += data;

I don't think we should be adding the discarded bytes to the checksum.
Or maybe we should reset the state of the parser, discarding bad (too
large) packet.

>  			continue;
>  		}
> @@ -491,7 +492,8 @@ static void ims_pcu_process_data(struct ims_pcu *pcu, struct urb *urb)
>  			break;
>  
>  		default:
> -			pcu->read_buf[pcu->read_pos++] = data;
> +			if (pcu->read_pos < IMS_PCU_BUF_SIZE)
> +				pcu->read_buf[pcu->read_pos++] = data;
>  			pcu->check_sum += data;
>  			break;
>  		}

Thanks.

-- 
Dmitry

[PATCH v2] Input: ims-pcu - fix heap-buffer-overflow in ims_pcu_process_data()

[pip-izony]

From: Seungjin Bae <eeodqql09@gmail.com>

The `ims_pcu_process_data()` processes incoming URB data byte by byte.
However, it fails to check if the `read_pos` index exceeds
IMS_PCU_BUF_SIZE.

If a malicious USB device sends a packet larger than IMS_PCU_BUF_SIZE,
`read_pos` will increment indefinitely. Moreover, since `read_pos` is
located immediately after `read_buf`, the attacker can overwrite
`read_pos` itself to arbitrarily control the index.

This manipulated `read_pos` is subsequently used in
`ims_pcu_handle_response()` to copy data into `cmd_buf`, leading to a
heap buffer overflow.

Specifically, an attacker can overwrite the `cmd_done.wait.head` located
at offset 136 relative to `cmd_buf` in the `ims_pcu_handle_response()`.
Consequently, when the driver calls `complete(&pcu->cmd_done)`, it
triggers a control flow hijack by using the manipulated pointer.

Fix this by adding a bounds check for `read_pos` before writing to
`read_buf`. If the packet is too long, discard it, log a warining,
and reset the parser state.

Fixes: 628329d524743 ("Input: add IMS Passenger Control Unit driver")
Signed-off-by: Seungjin Bae <eeodqql09@gmail.com>
---
 v1 -> v2: Add warning and reset the state of the parser for bad packet

 drivers/input/misc/ims-pcu.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/drivers/input/misc/ims-pcu.c b/drivers/input/misc/ims-pcu.c
index 4581f1c53644..c98ef71c841e 100644
--- a/drivers/input/misc/ims-pcu.c
+++ b/drivers/input/misc/ims-pcu.c
@@ -450,6 +450,16 @@ static void ims_pcu_process_data(struct ims_pcu *pcu, struct urb *urb)
 			continue;
 
 		if (pcu->have_dle) {
+			if (pcu->read_pos >= IMS_PCU_BUF_SIZE) {
+				dev_warn(pcu->dev,
+					 "Packet too long (%d bytes), discarding\n",
+					 pcu->read_pos);
+				pcu->have_stx = false;
+				pcu->have_dle = false;
+				pcu->read_pos = 0;
+				continue;
+			}
+
 			pcu->have_dle = false;
 			pcu->read_buf[pcu->read_pos++] = data;
 			pcu->check_sum += data;
@@ -491,6 +501,16 @@ static void ims_pcu_process_data(struct ims_pcu *pcu, struct urb *urb)
 			break;
 
 		default:
+			if (pcu->read_pos >= IMS_PCU_BUF_SIZE) {
+				dev_warn(pcu->dev,
+					 "Packet too long (%d bytes), discarding\n",
+					 pcu->read_pos);
+				pcu->have_stx = false;
+				pcu->have_dle = false;
+				pcu->read_pos = 0;
+				continue;
+			}
+
 			pcu->read_buf[pcu->read_pos++] = data;
 			pcu->check_sum += data;
 			break;
-- 
2.43.0

[PATCH v3] Input: ims-pcu - fix heap-buffer-overflow in ims_pcu_process_data()

[pip-izony]

From: Seungjin Bae <eeodqql09@gmail.com>

The `ims_pcu_process_data()` processes incoming URB data byte by byte.
However, it fails to check if the `read_pos` index exceeds
IMS_PCU_BUF_SIZE.

If a malicious USB device sends a packet larger than IMS_PCU_BUF_SIZE,
`read_pos` will increment indefinitely. Moreover, since `read_pos` is
located immediately after `read_buf`, the attacker can overwrite
`read_pos` itself to arbitrarily control the index.

This manipulated `read_pos` is subsequently used in
`ims_pcu_handle_response()` to copy data into `cmd_buf`, leading to a
heap buffer overflow.

Specifically, an attacker can overwrite the `cmd_done.wait.head` located
at offset 136 relative to `cmd_buf` in the `ims_pcu_handle_response()`.
Consequently, when the driver calls `complete(&pcu->cmd_done)`, it
triggers a control flow hijack by using the manipulated pointer.

Fix this by adding a bounds check for `read_pos` before writing to
`read_buf`. If the packet is too long, discard it, log a warning,
and reset the parser state.

Fixes: 628329d524743 ("Input: add IMS Passenger Control Unit driver")
Co-developed-by: Sanghoon Choi <csh0052@gmail.com>
Signed-off-by: Sanghoon Choi <csh0052@gmail.com>
Signed-off-by: Seungjin Bae <eeodqql09@gmail.com>
---
 v1 -> v2: Add warning and reset the state of the parser for bad packet
 v2 -> v3: Add co-author information

 drivers/input/misc/ims-pcu.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/drivers/input/misc/ims-pcu.c b/drivers/input/misc/ims-pcu.c
index 4581f1c53644..c98ef71c841e 100644
--- a/drivers/input/misc/ims-pcu.c
+++ b/drivers/input/misc/ims-pcu.c
@@ -450,6 +450,16 @@ static void ims_pcu_process_data(struct ims_pcu *pcu, struct urb *urb)
 			continue;
 
 		if (pcu->have_dle) {
+			if (pcu->read_pos >= IMS_PCU_BUF_SIZE) {
+				dev_warn(pcu->dev,
+					 "Packet too long (%d bytes), discarding\n",
+					 pcu->read_pos);
+				pcu->have_stx = false;
+				pcu->have_dle = false;
+				pcu->read_pos = 0;
+				continue;
+			}
+
 			pcu->have_dle = false;
 			pcu->read_buf[pcu->read_pos++] = data;
 			pcu->check_sum += data;
@@ -491,6 +501,16 @@ static void ims_pcu_process_data(struct ims_pcu *pcu, struct urb *urb)
 			break;
 
 		default:
+			if (pcu->read_pos >= IMS_PCU_BUF_SIZE) {
+				dev_warn(pcu->dev,
+					 "Packet too long (%d bytes), discarding\n",
+					 pcu->read_pos);
+				pcu->have_stx = false;
+				pcu->have_dle = false;
+				pcu->read_pos = 0;
+				continue;
+			}
+
 			pcu->read_buf[pcu->read_pos++] = data;
 			pcu->check_sum += data;
 			break;
-- 
2.43.0

Re: [PATCH v3] Input: ims-pcu - fix heap-buffer-overflow in ims_pcu_process_data()

[Dmitry Torokhov]

Hi,

On Sun, Dec 21, 2025 at 04:14:42PM -0500, pip-izony wrote:
> From: Seungjin Bae <eeodqql09@gmail.com>
> 
> The `ims_pcu_process_data()` processes incoming URB data byte by byte.
> However, it fails to check if the `read_pos` index exceeds
> IMS_PCU_BUF_SIZE.
> 
> If a malicious USB device sends a packet larger than IMS_PCU_BUF_SIZE,
> `read_pos` will increment indefinitely. Moreover, since `read_pos` is
> located immediately after `read_buf`, the attacker can overwrite
> `read_pos` itself to arbitrarily control the index.
> 
> This manipulated `read_pos` is subsequently used in
> `ims_pcu_handle_response()` to copy data into `cmd_buf`, leading to a
> heap buffer overflow.
> 
> Specifically, an attacker can overwrite the `cmd_done.wait.head` located
> at offset 136 relative to `cmd_buf` in the `ims_pcu_handle_response()`.
> Consequently, when the driver calls `complete(&pcu->cmd_done)`, it
> triggers a control flow hijack by using the manipulated pointer.
> 
> Fix this by adding a bounds check for `read_pos` before writing to
> `read_buf`. If the packet is too long, discard it, log a warning,
> and reset the parser state.
> 
> Fixes: 628329d524743 ("Input: add IMS Passenger Control Unit driver")
> Co-developed-by: Sanghoon Choi <csh0052@gmail.com>
> Signed-off-by: Sanghoon Choi <csh0052@gmail.com>
> Signed-off-by: Seungjin Bae <eeodqql09@gmail.com>
> ---
>  v1 -> v2: Add warning and reset the state of the parser for bad packet
>  v2 -> v3: Add co-author information
> 
>  drivers/input/misc/ims-pcu.c | 20 ++++++++++++++++++++
>  1 file changed, 20 insertions(+)
> 
> diff --git a/drivers/input/misc/ims-pcu.c b/drivers/input/misc/ims-pcu.c
> index 4581f1c53644..c98ef71c841e 100644
> --- a/drivers/input/misc/ims-pcu.c
> +++ b/drivers/input/misc/ims-pcu.c
> @@ -450,6 +450,16 @@ static void ims_pcu_process_data(struct ims_pcu *pcu, struct urb *urb)
>  			continue;
>  
>  		if (pcu->have_dle) {
> +			if (pcu->read_pos >= IMS_PCU_BUF_SIZE) {
> +				dev_warn(pcu->dev,
> +					 "Packet too long (%d bytes), discarding\n",
> +					 pcu->read_pos);
> +				pcu->have_stx = false;
> +				pcu->have_dle = false;
> +				pcu->read_pos = 0;

Here and in the other place we also need to reset checksum, otherwise a
valid packet might get rejected.

I factored out the code resetting packet state and applied, thanks.

-- 
Dmitry