Re: [PATCH] HID: hyperv: make pointer arithmetics understandable for FORTIFY_SOURCE

[Benjamin Tissoires <bentiss@kernel.org>]

On Aug 21 2026, Jiri Kosina wrote:
> From: Jiri Kosina <jkosina@suse.com>
> 
> Commit 83df7b5fa6735b5084ecd2 ("HID: hyperv: add KUnit coverage for device info
> bounds") introduced this piece of code
> 
> 	report = ((u8 *)&info->hid_descriptor) + info->hid_descriptor.bLength;
> 	memset(report, 0x42, 4);
> 
> to populate the report descriptor, making use of the fact that the report.
> &info->hid_descriptor points to a struct hid_descriptor (which is a fixed-size struct)
> GCC's FORTIFY_SOURCE infer the object size from that specific struct field rather than the
> outer dynamically allocated info buffer. As a result, writing past sizeof(struct hid_descriptor)
> triggers the __write_overflow_field warning.
> 
> Calculate the pointer offset using info directly, so the compiler evaluates the
> memory bounds against the allocated flexible layout of struct
> synthhid_device_info instead of the nested struct.
> 
> Fixes: 83df7b5fa6735b5084ecd2 ("HID: hyperv: add KUnit coverage for device info bounds")
> Reported-by: Jürgen Groß <jgross@suse.com>
> Signed-off-by: Jiri Kosina <jkosina@suse.com>
> ---
>  drivers/hid/hid-hyperv.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/hid/hid-hyperv.c b/drivers/hid/hid-hyperv.c
> index 6579bd19da13..4a10b24da9fa 100644
> --- a/drivers/hid/hid-hyperv.c
> +++ b/drivers/hid/hid-hyperv.c
> @@ -687,7 +687,7 @@ static void mousevsc_device_info_valid_descriptor(struct kunit *test)
>  
>  	info->hid_descriptor.bLength = sizeof(struct hid_descriptor);
>  	info->hid_descriptor.rpt_desc.wDescriptorLength = cpu_to_le16(4);
> -	report = ((u8 *)&info->hid_descriptor) + info->hid_descriptor.bLength;
> +	report = (u8 *)info + offsetof(struct synthhid_device_info, hid_descriptor) + info->hid_descriptor.bLength;

Isn't that overcomplicated?:

Above we have:
struct synthhid_device_info {
	struct synthhid_msg_hdr header;
	struct hv_input_dev_info hid_dev_info;
	struct hid_descriptor hid_descriptor;
};
  ...
	struct synthhid_device_info *info;
	...
	info = kunit_kzalloc(test, sizeof(*info) + 4, GFP_KERNEL);

so info is 0 allocated with the sizeof(struct synthhid_device_info) plus
4 for the report (immediately after).

with the bLength being set a couple of lines above, we basically have:
+ report = (u8 *)info + offsetof(struct synthhid_device_info, hid_descriptor) + sizeof(struct hid_descriptor);

So pointer address + offset of the last field + size of the last field.

Isn't that equivalent to:
  report = (u8 *)info + sizeof(*info);

or even: `(u8 *)(info + 1)`?

>  	memset(report, 0x42, 4);
>  
>  	mousevsc_on_receive_device_info(input_dev, info, sizeof(*info) + 4);
> @@ -713,7 +713,7 @@ static void mousevsc_device_info_report_desc_oob(struct kunit *test)
>  
>  	info->hid_descriptor.bLength = sizeof(struct hid_descriptor);
>  	info->hid_descriptor.rpt_desc.wDescriptorLength = cpu_to_le16(64);
> -	report = ((u8 *)&info->hid_descriptor) + info->hid_descriptor.bLength;
> +	report = (u8 *)info + offsetof(struct synthhid_device_info, hid_descriptor) + info->hid_descriptor.bLength;

Same overcomplexification applies here.

Unless of course there is something I missed in the string
fortification.

Cheers,
Benjamin

>  	memset(report, 0x42, 8);
>  
>  	mousevsc_on_receive_device_info(input_dev, info, sizeof(*info) + 8);
> 
> -- 
> Jiri Kosina
> SUSE Labs
> 
>