iio:kfifo_buf: check for uint overflow
authorMartin Kelly <mkelly@xevo.com>
Mon, 26 Mar 2018 21:27:52 +0000 (14:27 -0700)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 5 Jun 2018 09:41:58 +0000 (11:41 +0200)
commit838f25e3d9f2ad11e941649c56f8d205a80b01f2
tree1ee0ebb3e591ee1ed98d93d95e4baa2e14cf7e40
parent30ab9366f763928176793469fb5c3e0ebe89aeb5
iio:kfifo_buf: check for uint overflow

commit 3d13de4b027d5f6276c0f9d3a264f518747d83f2 upstream.

Currently, the following causes a kernel OOPS in memcpy:

echo 1073741825 > buffer/length
echo 1 > buffer/enable

Note that using 1073741824 instead of 1073741825 causes "write error:
Cannot allocate memory" but no OOPS.

This is because 1073741824 == 2^30 and 1073741825 == 2^30+1. Since kfifo
rounds up to the nearest power of 2, it will actually call kmalloc with
roundup_pow_of_two(length) * bytes_per_datum.

Using length == 1073741825 and bytes_per_datum == 2, we get:

kmalloc(roundup_pow_of_two(1073741825) * 2
or kmalloc(2147483648 * 2)
or kmalloc(4294967296)
or kmalloc(UINT_MAX + 1)

so this overflows to 0, causing kmalloc to return ZERO_SIZE_PTR and
subsequent memcpy to fail once the device is enabled.

Fix this by checking for overflow prior to allocating a kfifo. With this
check added, the above code returns -EINVAL when enabling the buffer,
rather than causing an OOPS.

Signed-off-by: Martin Kelly <mkelly@xevo.com>
cc: <Stable@vger.kernel.org>
Signed-off-by: Jonathan Cameron <Jonathan.Cameron@huawei.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/iio/buffer/kfifo_buf.c