Hello,

On 6/4/24 10:45, Alan Stern wrote:

On Tue, Jun 04, 2024 at 10:21:15AM -0700, Kees Cook wrote:

quoted

On Tue, Jun 04, 2024 at 10:09:43AM -0700, Nikita Zhandarovich wrote:

quoted

Hi,

On 6/4/24 07:15, Jiri Kosina wrote:

quoted

On Tue, 4 Jun 2024, Kees Cook wrote:

quoted

This isn't the right solution. The problem is that hid_class_descriptor 
is a flexible array but was sized as a single element fake flexible 
array:

struct hid_descriptor {
	   __u8  bLength;
	   __u8  bDescriptorType;
	   __le16 bcdHID;
	   __u8  bCountryCode;
	   __u8  bNumDescriptors;

	   struct hid_class_descriptor desc[1];
} __attribute__ ((packed));

This likely needs to be: 

struct hid_class_descriptor desc[] __counted_by(bNumDescriptors);

And then check for any sizeof() uses of the struct that might have changed.

Alan, I finally got around to preparing a revised version of the
required patch and encountered a few issues. I could use some advice in
this matter...

If we change 'struct hid_descriptor' as you suggested, which does make
sense, most occurrences of that type are easy enough to fix.

1) usbhid_parse() starts working properly if there are more than 1
descriptors, sizeof(struct hid_descriptor) may be turned into something
crude but straightforward like sizeof(struct hid_descriptor) +
sizeof(struct hid_class_descriptor).

2) 'hid_descriptor' in drivers/hid/hid-hyperv.c remains innocuous as
well as only 1 descriptor expected there. My impression is only some
small changes are needed there.

However, the issue that stumps me is the following: static struct
hid_descriptor hidg_desc in drivers/usb/gadget/function/f_hid.c relies
on a static nature of that one descriptor. hidg_desc ends up being used
elsewhere, in other static structures. Basically, using __counted_by
requires a lot of changes, as I see it, out of scope of merely closing
an UBSAN error.

Is this approach still worthy pursuing or should I look into some neater
solution?

Best regards,
Nikita

quoted

quoted

quoted

Ah, you are of course right, not sure what I was thinking. Thanks a lot 
for catching my brainfart.

I am dropping the patch for now; Nikita, will you please send a refreshed 
one?

Thanks for catching my mistake.

I'll gladly send a revised version, hoping to do it very soon.

I spent a little more time looking at this, and I'm not sure I
understand where the actual space for the descriptors comes from?
There's interface->extra that is being parsed, and effectively
hid_descriptor is being mapped into it, but it uses "sizeof(struct
hid_descriptor)" for the limit.

That's a lower limit, not an upper limit.  The hid_descriptor must 
include at least one hid_class_descriptor, but it can include more.
That's what the min_t() calculation of num_descriptors is meant to 
figure out.

quoted

 Is more than 1 descriptor expected to
work correctly?

More than one hid_class_descriptor -- yes.

quoted

 Or is the limit being ignored? I'm a bit confused by
this code...

Does this explain it?

Alan Stern