In hid_submit_ctrl(), the way of calculating the report length doesn't
take into account that report->size can be zero. When running the
syzkaller reproducer, a report of size 0 causes hid_submit_ctrl) to
calculate transfer_buffer_length as 16384. When this urb is passed to
the usb core layer, KMSAN reports an info leak of 16384 bytes.
To fix this, first modify hid_report_len() to account for the zero
report size case by using DIV_ROUND_UP for the division. Then, call it
from hid_submit_ctrl().
Reported-by: [email protected]
Signed-off-by: Anirudh Rayabharam <[email protected]>
---
drivers/hid/usbhid/hid-core.c | 2 +-
include/linux/hid.h | 3 +--
2 files changed, 2 insertions(+), 3 deletions(-)
diff --git a/drivers/hid/usbhid/hid-core.c b/drivers/hid/usbhid/hid-core.c
index 86257ce6d619..4e9077363c96 100644
--- a/drivers/hid/usbhid/hid-core.c
+++ b/drivers/hid/usbhid/hid-core.c
@@ -374,7 +374,7 @@ static int hid_submit_ctrl(struct hid_device *hid)
raw_report = usbhid->ctrl[usbhid->ctrltail].raw_report;
dir = usbhid->ctrl[usbhid->ctrltail].dir;
- len = ((report->size - 1) >> 3) + 1 + (report->id > 0);
+ len = hid_report_len(report);
if (dir == USB_DIR_OUT) {
usbhid->urbctrl->pipe = usb_sndctrlpipe(hid_to_usb_dev(hid), 0);
usbhid->urbctrl->transfer_buffer_length = len;
diff --git a/include/linux/hid.h b/include/linux/hid.h
index 271021e20a3f..10e922cee4eb 100644
--- a/include/linux/hid.h
+++ b/include/linux/hid.h
@@ -1167,8 +1167,7 @@ static inline void hid_hw_wait(struct hid_device *hdev)
*/
static inline u32 hid_report_len(struct hid_report *report)
{
- /* equivalent to DIV_ROUND_UP(report->size, 8) + !!(report->id > 0) */
- return ((report->size - 1) >> 3) + 1 + (report->id > 0);
+ return DIV_ROUND_UP(report->size, 8) + (report->id > 0);
}
int hid_report_raw_event(struct hid_device *hid, int type, u8 *data, u32 size,
--
2.26.2
On Sun, 25 Apr 2021, Anirudh Rayabharam wrote:
> In hid_submit_ctrl(), the way of calculating the report length doesn't
> take into account that report->size can be zero. When running the
> syzkaller reproducer, a report of size 0 causes hid_submit_ctrl) to
> calculate transfer_buffer_length as 16384. When this urb is passed to
> the usb core layer, KMSAN reports an info leak of 16384 bytes.
>
> To fix this, first modify hid_report_len() to account for the zero
> report size case by using DIV_ROUND_UP for the division. Then, call it
> from hid_submit_ctrl().
>
> Reported-by: [email protected]
> Signed-off-by: Anirudh Rayabharam <[email protected]>
Benjamin, could you please run this one through your regression testing
machinery before we send it upstream?
Thanks,
--
Jiri Kosina
SUSE Labs
On Wed, 5 May 2021, Benjamin Tissoires wrote:
> I don't have a reproducer like syzbot has for the exact bug here, as I
> am relying on one real USB device to check if usbhid is not too broken.
> However, the test suite should catch if there is an error implied by the
> hid_report_len() change.
Yes, that was exactly what I wanted to check, sorry for not being verbose
enough :)
> Anyway, I manually started the job and will report when it is done.
Thanks,
--
Jiri Kosina
SUSE Labs
On Wed, May 5, 2021 at 3:28 PM Jiri Kosina <[email protected]> wrote:
>
> On Wed, 5 May 2021, Benjamin Tissoires wrote:
>
> > I don't have a reproducer like syzbot has for the exact bug here, as I
> > am relying on one real USB device to check if usbhid is not too broken.
> > However, the test suite should catch if there is an error implied by the
> > hid_report_len() change.
>
> Yes, that was exactly what I wanted to check, sorry for not being verbose
> enough :)
>
> > Anyway, I manually started the job and will report when it is done.
>
Heh, no problems.
"Job succeeded" \o/
Given that you are on a spree:
Acked-by: Benjamin Tissoires <[email protected]>
Cheers,
Benjamin
On Wed, May 5, 2021 at 2:42 PM Jiri Kosina <[email protected]> wrote:
>
> On Sun, 25 Apr 2021, Anirudh Rayabharam wrote:
>
> > In hid_submit_ctrl(), the way of calculating the report length doesn't
> > take into account that report->size can be zero. When running the
> > syzkaller reproducer, a report of size 0 causes hid_submit_ctrl) to
> > calculate transfer_buffer_length as 16384. When this urb is passed to
> > the usb core layer, KMSAN reports an info leak of 16384 bytes.
> >
> > To fix this, first modify hid_report_len() to account for the zero
> > report size case by using DIV_ROUND_UP for the division. Then, call it
> > from hid_submit_ctrl().
> >
> > Reported-by: [email protected]
> > Signed-off-by: Anirudh Rayabharam <[email protected]>
>
> Benjamin, could you please run this one through your regression testing
> machinery before we send it upstream?
>
I don't have a reproducer like syzbot has for the exact bug here, as I
am relying on one real USB device to check if usbhid is not too
broken.
However, the test suite should catch if there is an error implied by
the hid_report_len() change.
Anyway, I manually started the job and will report when it is done.
Cheers,
Benjamin
On Wed, 5 May 2021, Benjamin Tissoires wrote:
> > > I don't have a reproducer like syzbot has for the exact bug here, as I
> > > am relying on one real USB device to check if usbhid is not too broken.
> > > However, the test suite should catch if there is an error implied by the
> > > hid_report_len() change.
> >
> > Yes, that was exactly what I wanted to check, sorry for not being verbose
> > enough :)
> >
> > > Anyway, I manually started the job and will report when it is done.
> >
>
> Heh, no problems.
>
> "Job succeeded" \o/
>
> Given that you are on a spree:
:-)
> Acked-by: Benjamin Tissoires <[email protected]>
Applied to for-5.13/upstream-fixes. Thanks,
--
Jiri Kosina
SUSE Labs