2006-01-30 16:25:04

by Ben Collins

Subject: [PATCH 2.6.15/2.6.16-git] Fix off-by-one for num_values in uref_multi requests

Found this when working with a HAPP UGCI device. It has a usage with 7
indexes. I could read them all one at a time, but using a multiref it
would only allow me to read the first 6. The patch below fixed it.

Signed-off-by: Ben Collins <[email protected]>

--- a/drivers/usb/input/hiddev.c
+++ b/drivers/usb/input/hiddev.c
@@ -632,7 +632,7 @@ static int hiddev_ioctl(struct inode *in

else if ((cmd == HIDIOCGUSAGES || cmd == HIDIOCSUSAGES) &&
(uref_multi->num_values > HID_MAX_MULTI_USAGES ||
- uref->usage_index + uref_multi->num_values >= field->report_count))
+ uref->usage_index + uref_multi->num_values > field->report_count))
goto inval;
}


--
Ben Collins
Kernel Developer - Ubuntu Linux

2006-01-31 06:32:07

by Dmitry Torokhov

Subject: Re: [PATCH 2.6.15/2.6.16-git] Fix off-by-one for num_values in uref_multi requests

On Monday 30 January 2006 11:24, Ben Collins wrote:
> Found this when working with a HAPP UGCI device. It has a usage with 7
> indexes. I could read them all one at a time, but using a multiref it
> would only allow me to read the first 6. The patch below fixed it.
>
> Signed-off-by: Ben Collins <[email protected]>
>

I applied this to the input tree, thanks!

> --- a/drivers/usb/input/hiddev.c
> +++ b/drivers/usb/input/hiddev.c
> @@ -632,7 +632,7 @@ static int hiddev_ioctl(struct inode *in
>
> else if ((cmd == HIDIOCGUSAGES || cmd == HIDIOCSUSAGES) &&
> (uref_multi->num_values > HID_MAX_MULTI_USAGES ||
> - uref->usage_index + uref_multi->num_values >= field->report_count))
> + uref->usage_index + uref_multi->num_values > field->report_count))
> goto inval;
> }
>
>

--
Dmitry