Add validation for report->maxfield and report->field[0] before
dereferencing to prevent NULL pointer dereference.

The HID report descriptor is provided by the USB device firmware via
USB control transfer (GET_DESCRIPTOR). A malicious device can craft
a descriptor that defines an OUTPUT report without any usages
(padding fields). When the HID subsystem parses such a descriptor:

1. hid_add_field() calls hid_register_report() to create the report
   object and stores it in report_id_hash[id]
2. Since parser->local.usage_index is 0, hid_add_field() returns early
   without calling hid_register_field() to add any fields
3. Result: report exists with maxfield=0 and field[0]=NULL

When hidpp_probe() is called for a device matching this driver:
  - hidpp_validate_device() calls hidpp_get_report_length()
  - hidpp_get_report_length() retrieves the report from hash (not NULL)
  - It then dereferences report->field[0]->report_count
  - Since field[0] is NULL, this triggers a kernel NULL pointer
    dereference

Data flow from attacker to crash:
  Malicious USB Device
       |
       v (USB GET_DESCRIPTOR control transfer)
  hid_get_class_descriptor() -- reads HID report descriptor from device
       |
       v
  hid_parse_report() -- stores descriptor in hid->dev_rdesc
       |
       v
  hid_open_report() -> hid_add_field()
       |                    |
       |                    v
       |              hid_register_report() -- creates report, maxfield=0
       |                    |
       |                    v
       |              returns early if usage_index==0, no field added
       |
       v
  hidpp_validate_device() -> hidpp_get_report_length()
       |
       v
  report->field[0]->report_count -- NULL pointer dereference!

This is triggerable by an attacker with physical access using a
malicious USB device (e.g., BadUSB, programmable USB development
boards).

Fixes: d71b18f7c7999 ("HID: logitech-hidpp: do not hardcode very long report length")
Signed-off-by: Kery Qi <qikeyu2017@gmail.com>
---
 drivers/hid/hid-logitech-hidpp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/hid/hid-logitech-hidpp.c b/drivers/hid/hid-logitech-hidpp.c
index d5011a5d0890..02ddbd658e89 100644
--- a/drivers/hid/hid-logitech-hidpp.c
+++ b/drivers/hid/hid-logitech-hidpp.c
@@ -4314,7 +4314,7 @@ static int hidpp_get_report_length(struct hid_device *hdev, int id)
 
 	re = &(hdev->report_enum[HID_OUTPUT_REPORT]);
 	report = re->report_id_hash[id];
-	if (!report)
+	if (!report || report->maxfield < 1 || !report->field[0])
 		return 0;
 
 	return report->field[0]->report_count + 1;
-- 
2.34.1

Hello,

Thanks for the feedback.

Best regards

Kery

Kery Qi <qikeyu2017@gmail.com> 于2026年1月15日周四 22:24写道：
>
> Add validation for report->maxfield and report->field[0] before
> dereferencing to prevent NULL pointer dereference.
>
> The HID report descriptor is provided by the USB device firmware via
> USB control transfer (GET_DESCRIPTOR). A malicious device can craft
> a descriptor that defines an OUTPUT report without any usages
> (padding fields). When the HID subsystem parses such a descriptor:
>
> 1. hid_add_field() calls hid_register_report() to create the report
>    object and stores it in report_id_hash[id]
> 2. Since parser->local.usage_index is 0, hid_add_field() returns early
>    without calling hid_register_field() to add any fields
> 3. Result: report exists with maxfield=0 and field[0]=NULL
>
> When hidpp_probe() is called for a device matching this driver:
>   - hidpp_validate_device() calls hidpp_get_report_length()
>   - hidpp_get_report_length() retrieves the report from hash (not NULL)
>   - It then dereferences report->field[0]->report_count
>   - Since field[0] is NULL, this triggers a kernel NULL pointer
>     dereference
>
> Data flow from attacker to crash:
>   Malicious USB Device
>        |
>        v (USB GET_DESCRIPTOR control transfer)
>   hid_get_class_descriptor() -- reads HID report descriptor from device
>        |
>        v
>   hid_parse_report() -- stores descriptor in hid->dev_rdesc
>        |
>        v
>   hid_open_report() -> hid_add_field()
>        |                    |
>        |                    v
>        |              hid_register_report() -- creates report, maxfield=0
>        |                    |
>        |                    v
>        |              returns early if usage_index==0, no field added
>        |
>        v
>   hidpp_validate_device() -> hidpp_get_report_length()
>        |
>        v
>   report->field[0]->report_count -- NULL pointer dereference!
>
> This is triggerable by an attacker with physical access using a
> malicious USB device (e.g., BadUSB, programmable USB development
> boards).
>
> Fixes: d71b18f7c7999 ("HID: logitech-hidpp: do not hardcode very long report length")
> Signed-off-by: Kery Qi <qikeyu2017@gmail.com>
> ---
>  drivers/hid/hid-logitech-hidpp.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/hid/hid-logitech-hidpp.c b/drivers/hid/hid-logitech-hidpp.c
> index d5011a5d0890..02ddbd658e89 100644
> --- a/drivers/hid/hid-logitech-hidpp.c
> +++ b/drivers/hid/hid-logitech-hidpp.c
> @@ -4314,7 +4314,7 @@ static int hidpp_get_report_length(struct hid_device *hdev, int id)
>
>         re = &(hdev->report_enum[HID_OUTPUT_REPORT]);
>         report = re->report_id_hash[id];
> -       if (!report)
> +       if (!report || report->maxfield < 1 || !report->field[0])
>                 return 0;
>
>         return report->field[0]->report_count + 1;
> --
> 2.34.1
>

On Thu, 2026-01-15 at 22:24 +0800, Kery Qi wrote:
<snip>
> -	if (!report)
> +	if (!report || report->maxfield < 1 || !report->field[0])

A partial fix already exists in the for-next branch:
https://git.kernel.org/pub/scm/linux/kernel/git/hid/hid.git/commit/?h=for-next&id=1547d41f9f19d691c2c9ce4c29f746297baef9e9

You'll probably want to rebase and adapt your fix. See also this review
by GregKH for v1:
https://patchwork.kernel.org/project/linux-input/patch/20260109105912.3141960-2-gnoack@google.com/

Cheers

>  		return 0;
>  
>  	return report->field[0]->report_count + 1;