[PATCH] mwifiex: Fix divide error in mwifiex_usb_dnld_fw

This patch try to fix bug reported by syzkaller:
divide error: 0000 [#1] SMP KASAN
CPU: 1 PID: 17 Comm: kworker/1:0 Not tainted 5.15.0-rc5-syzkaller #0
Hardware name: Google Compute Engine, BIOS Google 01/01/2011
Workqueue: events request_firmware_work_func
RIP: 0010:mwifiex_write_data_sync drivers/net/wireless/marvell/mwifiex/usb.c:696 [inline]
RIP: 0010:mwifiex_prog_fw_w_helper drivers/net/wireless/marvell/mwifiex/usb.c:1437 [inline]
RIP: 0010:mwifiex_usb_dnld_fw+0xabd/0x11a0 drivers/net/wireless/marvell/mwifiex/usb.c:1518
Call Trace:
 _mwifiex_fw_dpc+0x181/0x10a0 drivers/net/wireless/marvell/mwifiex/main.c:542
 request_firmware_work_func+0x12c/0x230 drivers/base/firmware_loader/main.c:1081
 process_one_work+0x9bf/0x1620 kernel/workqueue.c:2297
 worker_thread+0x658/0x11f0 kernel/workqueue.c:2444
 kthread+0x3c2/0x4a0 kernel/kthread.c:319
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:295

Link: https://syzkaller.appspot.com/bug?extid=4e7b6c94d22f4bfca9a0
Reported-and-tested-by: syzbot+4e7b6c94d22f4bfca9a0@syzkaller.appspotmail.com
Signed-off-by: Wan Jiabing <redacted>
---
 drivers/net/wireless/marvell/mwifiex/usb.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/wireless/marvell/mwifiex/usb.c b/drivers/net/wireless/marvell/mwifiex/usb.c
index 426e39d4ccf0..c24ec27d4057 100644
--- a/drivers/net/wireless/marvell/mwifiex/usb.c
+++ b/drivers/net/wireless/marvell/mwifiex/usb.c

@@ -693,7 +693,7 @@ static int mwifiex_write_data_sync(struct mwifiex_adapter *adapter, u8 *pbuf,
 	struct usb_card_rec *card = adapter->card;
 	int actual_length, ret;
 
-	if (!(*len % card->bulk_out_maxpktsize))
+	if (card->bulk_out_maxpktsize && !(*len % card->bulk_out_maxpktsize))
 		(*len)++;
 
 	/* Send the data block */

-- 
2.20.1

On 10/18/21 8:31 PM, Wan Jiabing wrote:

This patch try to fix bug reported by syzkaller:
divide error: 0000 [#1] SMP KASAN
CPU: 1 PID: 17 Comm: kworker/1:0 Not tainted 5.15.0-rc5-syzkaller #0
Hardware name: Google Compute Engine, BIOS Google 01/01/2011
Workqueue: events request_firmware_work_func
RIP: 0010:mwifiex_write_data_sync drivers/net/wireless/marvell/mwifiex/usb.c:696 [inline]
RIP: 0010:mwifiex_prog_fw_w_helper drivers/net/wireless/marvell/mwifiex/usb.c:1437 [inline]
RIP: 0010:mwifiex_usb_dnld_fw+0xabd/0x11a0 drivers/net/wireless/marvell/mwifiex/usb.c:1518
Call Trace:
 _mwifiex_fw_dpc+0x181/0x10a0 drivers/net/wireless/marvell/mwifiex/main.c:542
 request_firmware_work_func+0x12c/0x230 drivers/base/firmware_loader/main.c:1081
 process_one_work+0x9bf/0x1620 kernel/workqueue.c:2297
 worker_thread+0x658/0x11f0 kernel/workqueue.c:2444
 kthread+0x3c2/0x4a0 kernel/kthread.c:319
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:295

Link: https://syzkaller.appspot.com/bug?extid=4e7b6c94d22f4bfca9a0

Please provide a Fixes: tag

Are you sure this fix is not working around the real bug ?

In which cases bulk_out_maxpktsize would be zero ?

If this is a valid case, this needs to be explained in the changelog.

 		(*len)++;
 
 	/* Send the data block */

On Mon, Oct 18, 2021 at 9:04 PM Eric Dumazet [off-list ref] wrote:

On 10/18/21 8:31 PM, Wan Jiabing wrote:

quoted

--- a/drivers/net/wireless/marvell/mwifiex/usb.c
+++ b/drivers/net/wireless/marvell/mwifiex/usb.c

@@ -693,7 +693,7 @@ static int mwifiex_write_data_sync(struct mwifiex_adapter *adapter, u8 *pbuf,
      struct usb_card_rec *card = adapter->card;
      int actual_length, ret;

-     if (!(*len % card->bulk_out_maxpktsize))
+     if (card->bulk_out_maxpktsize && !(*len % card->bulk_out_maxpktsize))

Are you sure this fix is not working around the real bug ?

In which cases bulk_out_maxpktsize would be zero ?

If this is a valid case, this needs to be explained in the changelog.

I'm with Eric here. This was a bug reported by a fuzzer, which throws
invalid input at the driver. The right answer is likely that we should
reject such invalid input when we receive it -- i.e., we should fail
to probe() the device if it has invalid capabilities. In particular,
we should fail to probe if wMaxPacketSize==0.

I was thinking of sending such a patch myself, but I don't have any
USB mwifiex hardware to test, so I deferred. It's probably pretty low
risk anyway, though.

Brian