In the while loop of technisat_usb2_get_ir, it scans through
a fix-sized buffer read from the device side, the termination
condition of the loop is `*b == 0xff`. If no `0xff` byte is read
from the device side, OOB access happens.
This patch fixes the bug by adding an upper bound in the while loop.
Reported-by: Hui Peng <[email protected]>
Reported-by: Mathias Payer <[email protected]>
Signed-off-by: Hui Peng <[email protected]>
---
drivers/media/usb/dvb-usb/technisat-usb2.c | 5 +++++
1 file changed, 5 insertions(+)
diff --git a/drivers/media/usb/dvb-usb/technisat-usb2.c b/drivers/media/usb/dvb-usb/technisat-usb2.c
index c659e18b358b..181f5f97af45 100644
--- a/drivers/media/usb/dvb-usb/technisat-usb2.c
+++ b/drivers/media/usb/dvb-usb/technisat-usb2.c
@@ -612,6 +612,7 @@ static int technisat_usb2_get_ir(struct dvb_usb_device *d)
u8 *b;
int ret;
struct ir_raw_event ev;
+ int i = 0;
buf[0] = GET_IR_DATA_VENDOR_REQUEST;
buf[1] = 0x08;
@@ -656,11 +657,15 @@ static int technisat_usb2_get_ir(struct dvb_usb_device *d)
ev.pulse = 0;
while (1) {
+ // only `ret` bytes are read from the device side
+ if (i >= ret)
+ break;
ev.pulse = !ev.pulse;
ev.duration = (*b * FIRMWARE_CLOCK_DIVISOR * FIRMWARE_CLOCK_TICK) / 1000;
ir_raw_event_store(d->rc_dev, &ev);
b++;
+ i++;
if (*b == 0xff) {
ev.pulse = 0;
ev.duration = 888888*2;
--
2.23.0
The following is the kasan report. This bug was found in v4.20-rc2, but
it is present in the latest version.
BUG: KASAN: slab-out-of-bounds in technisat_usb2_get_ir
drivers/media/usb/dvb-usb/technisat-usb2.c:664 [inline]
BUG: KASAN: slab-out-of-bounds in technisat_usb2_rc_query+0x598/0x5f0
drivers/media/usb/dvb-usb/technisat-usb2.c:679
Read of size 1 at addr ffff88805ee3d3a8 by task kworker/2:3/8681
CPU: 2 PID: 8681 Comm: kworker/2:3 Not tainted 4.20.0-rc2 #1
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS
Ubuntu-1.8.2-1ubuntu1 04/01/2014
Workqueue: events dvb_usb_read_remote_control
Call Trace:
__dump_stack lib/dump_stack.c:77 [inline]
dump_stack+0xd2/0x148 lib/dump_stack.c:113
print_address_description+0x71/0x239 mm/kasan/report.c:256
kasan_report_error mm/kasan/report.c:354 [inline]
kasan_report.cold.5+0x242/0x30b mm/kasan/report.c:412
__asan_report_load1_noabort+0x19/0x20 mm/kasan/report.c:430
technisat_usb2_get_ir drivers/media/usb/dvb-usb/technisat-usb2.c:664
[inline]
technisat_usb2_rc_query+0x598/0x5f0
drivers/media/usb/dvb-usb/technisat-usb2.c:679
dvb_usb_read_remote_control+0xbd/0x150
drivers/media/usb/dvb-usb/dvb-usb-remote.c:261
process_one_work+0x816/0x14d0 kernel/workqueue.c:2153
worker_thread+0x9b/0xce0 kernel/workqueue.c:2296
kthread+0x33d/0x400 kernel/kthread.c:246
ret_from_fork+0x3a/0x50 arch/x86/entry/entry_64.S:352
On 8/20/19 2:19 PM, Hui Peng wrote:
> In the while loop of technisat_usb2_get_ir, it scans through
> a fix-sized buffer read from the device side, the termination
> condition of the loop is `*b == 0xff`. If no `0xff` byte is read
> from the device side, OOB access happens.
>
> This patch fixes the bug by adding an upper bound in the while loop.
>
> Reported-by: Hui Peng <[email protected]>
> Reported-by: Mathias Payer <[email protected]>
> Signed-off-by: Hui Peng <[email protected]>
> ---
> drivers/media/usb/dvb-usb/technisat-usb2.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/drivers/media/usb/dvb-usb/technisat-usb2.c b/drivers/media/usb/dvb-usb/technisat-usb2.c
> index c659e18b358b..181f5f97af45 100644
> --- a/drivers/media/usb/dvb-usb/technisat-usb2.c
> +++ b/drivers/media/usb/dvb-usb/technisat-usb2.c
> @@ -612,6 +612,7 @@ static int technisat_usb2_get_ir(struct dvb_usb_device *d)
> u8 *b;
> int ret;
> struct ir_raw_event ev;
> + int i = 0;
>
> buf[0] = GET_IR_DATA_VENDOR_REQUEST;
> buf[1] = 0x08;
> @@ -656,11 +657,15 @@ static int technisat_usb2_get_ir(struct dvb_usb_device *d)
>
> ev.pulse = 0;
> while (1) {
> + // only `ret` bytes are read from the device side
> + if (i >= ret)
> + break;
> ev.pulse = !ev.pulse;
> ev.duration = (*b * FIRMWARE_CLOCK_DIVISOR * FIRMWARE_CLOCK_TICK) / 1000;
> ir_raw_event_store(d->rc_dev, &ev);
>
> b++;
> + i++;
> if (*b == 0xff) {
> ev.pulse = 0;
> ev.duration = 888888*2;
On Tue, Aug 20, 2019 at 02:19:16PM -0400, Hui Peng wrote:
> In the while loop of technisat_usb2_get_ir, it scans through
> a fix-sized buffer read from the device side, the termination
> condition of the loop is `*b == 0xff`. If no `0xff` byte is read
> from the device side, OOB access happens.
>
> This patch fixes the bug by adding an upper bound in the while loop.
This issue was fixed in:
https://git.linuxtv.org/media_tree.git/commit/drivers/media/usb/dvb-usb/technisat-usb2.c?id=0c4df39e504bf925ab666132ac3c98d6cbbe380b
Note this limits the loop to the size of the buffer; using the ret
return value might be better.
Sean
>
> Reported-by: Hui Peng <[email protected]>
> Reported-by: Mathias Payer <[email protected]>
> Signed-off-by: Hui Peng <[email protected]>
> ---
> drivers/media/usb/dvb-usb/technisat-usb2.c | 5 +++++
> 1 file changed, 5 insertions(+)
>
> diff --git a/drivers/media/usb/dvb-usb/technisat-usb2.c b/drivers/media/usb/dvb-usb/technisat-usb2.c
> index c659e18b358b..181f5f97af45 100644
> --- a/drivers/media/usb/dvb-usb/technisat-usb2.c
> +++ b/drivers/media/usb/dvb-usb/technisat-usb2.c
> @@ -612,6 +612,7 @@ static int technisat_usb2_get_ir(struct dvb_usb_device *d)
> u8 *b;
> int ret;
> struct ir_raw_event ev;
> + int i = 0;
>
> buf[0] = GET_IR_DATA_VENDOR_REQUEST;
> buf[1] = 0x08;
> @@ -656,11 +657,15 @@ static int technisat_usb2_get_ir(struct dvb_usb_device *d)
>
> ev.pulse = 0;
> while (1) {
> + // only `ret` bytes are read from the device side
> + if (i >= ret)
> + break;
> ev.pulse = !ev.pulse;
> ev.duration = (*b * FIRMWARE_CLOCK_DIVISOR * FIRMWARE_CLOCK_TICK) / 1000;
> ir_raw_event_store(d->rc_dev, &ev);
>
> b++;
> + i++;
> if (*b == 0xff) {
> ev.pulse = 0;
> ev.duration = 888888*2;
> --
> 2.23.0