Re: [PATCH] usb: usbtest: prevent divide by zero in simple_io() with zero length
From: Alan Stern
Date: Fri Oct 16 2026 - 10:51:22 EST
On Fri, Oct 16, 2026 at 03:27:50PM +0800, YuanZhe Liu wrote:
> A divide error oops was observed in simple_io():
>
> usbtest 1-1:1.0: TEST 3: write/1 0..0 bytes 16307 times
> Oops: divide error: 0000 [#1] SMP KASAN
> RIP: 0010:simple_io+0x2c4/0x540 drivers/usb/misc/usbtest.c:497
>
> The variable-length tests (test 3 and test 4) compute the next URB
> size with "len %= max", where max is the original
> urb->transfer_buffer_length. However, when a zero-length transfer is
> requested (param->length == 0), simple_alloc_urb() still returns a valid
> URB with transfer_buffer_length set to 0. On the first vary update,
> "len %= max" divides by zero and the kernel oopses.
>
> Commit 564e69893c63 ("USB: usbtest: prevent a divide by zero bug") added
> the equivalent zero check to alloc_sglist(), which covers the
> scatterlist tests (test 7/8), but the single-URB variable-length path in
> simple_io() was left unprotected.
>
> Fix this by returning -EINVAL early from simple_io() when vary is
> requested but max is zero, so the divide by zero never occurs.
> Zero-length transfers without vary remain valid for the other callers of
> simple_io().
>
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Cc: stable@xxxxxxxxxxxxxxx
> Co-developed-by: Bin Yu <byu@xxxxxxxxxxxxx>
> Signed-off-by: Bin Yu <byu@xxxxxxxxxxxxx>
> Co-developed-by: Cheng Wen <wencheng@xxxxxxxxxxxxx>
> Signed-off-by: Cheng Wen <wencheng@xxxxxxxxxxxxx>
> Signed-off-by: YuanZhe Liu <25031212351@xxxxxxxxxxxxxxxxx>
> ---
Reviewed-by: Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>
> drivers/usb/misc/usbtest.c | 3 +++
> 1 file changed, 3 insertions(+)
>
> diff --git a/drivers/usb/misc/usbtest.c b/drivers/usb/misc/usbtest.c
> index 8759df49be28..e2f5e6be353a 100644
> --- a/drivers/usb/misc/usbtest.c
> +++ b/drivers/usb/misc/usbtest.c
> @@ -466,6 +466,9 @@ static int simple_io(
> int retval = 0;
> unsigned long expire;
>
> + if (vary && !max)
> + return -EINVAL;
> +
> urb->context = &completion;
> while (retval == 0 && iterations-- > 0) {
> init_completion(&completion);
> --
> 2.34.1