Re: drivers/usb/usbip/stub_rx.c:505 stub_recv_cmd_submit() error: uninitialized symbol 'nents'.

From: shuah
Date: Thu Oct 24 2019 - 18:53:15 EST


On 10/24/19 1:45 PM, Dan Carpenter wrote:
On Wed, Oct 23, 2019 at 04:11:20PM +0900, Suwan Kim wrote:
On Tue, Oct 22, 2019 at 12:28:39PM +0300, Dan Carpenter wrote:
tree: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
head: 7d194c2100ad2a6dded545887d02754948ca5241
commit: ea44d190764b4422af4d1c29eaeb9e69e353b406 usbip: Implement SG support to vhci-hcd and stub driver
date: 7 weeks ago

If you fix the issue, kindly add following tag
Reported-by: kbuild test robot <lkp@xxxxxxxxx>
Reported-by: Dan Carpenter <dan.carpenter@xxxxxxxxxx>

New smatch warnings:
drivers/usb/usbip/stub_rx.c:505 stub_recv_cmd_submit() error: uninitialized symbol 'nents'.

Old smatch warnings:
drivers/usb/usbip/stub_rx.c:450 stub_recv_xbuff() error: uninitialized symbol 'ret'.

# https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=ea44d190764b4422af4d1c29eaeb9e69e353b406
git remote add linus https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
git remote update linus
git checkout ea44d190764b4422af4d1c29eaeb9e69e353b406
vim +/nents +505 drivers/usb/usbip/stub_rx.c

4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 453 static void stub_recv_cmd_submit(struct stub_device *sdev,
4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 454 struct usbip_header *pdu)
4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 455 {
4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 456 struct stub_priv *priv;
4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 457 struct usbip_device *ud = &sdev->ud;
2d8f4595d1f275 drivers/staging/usbip/stub_rx.c Max Vozeler 2011-01-12 458 struct usb_device *udev = sdev->udev;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 459 struct scatterlist *sgl = NULL, *sg;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 460 void *buffer = NULL;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 461 unsigned long long buf_len;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 462 int nents;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 463 int num_urbs = 1;
c6688ef9f29762 drivers/usb/usbip/stub_rx.c Shuah Khan 2017-12-07 464 int pipe = get_pipe(sdev, pdu);
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 465 int use_sg = pdu->u.cmd_submit.transfer_flags & URB_DMA_MAP_SG;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 466 int support_sg = 1;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 467 int np = 0;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 468 int ret, i;
4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 469
635f545a7e8be7 drivers/usb/usbip/stub_rx.c Shuah Khan 2017-12-07 470 if (pipe == -1)
635f545a7e8be7 drivers/usb/usbip/stub_rx.c Shuah Khan 2017-12-07 471 return;
4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 472
4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 473 priv = stub_priv_alloc(sdev, pdu);
4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 474 if (!priv)
4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 475 return;
4d7b5c7f8ad49b drivers/staging/usbip/stub_rx.c Takahiro Hirofuchi 2008-07-09 476
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 477 buf_len = (unsigned long long)pdu->u.cmd_submit.transfer_buffer_length;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 478
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 479 /* allocate urb transfer buffer, if needed */
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 480 if (buf_len) {
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 481 if (use_sg) {
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 482 sgl = sgl_alloc(buf_len, GFP_KERNEL, &nents);
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 483 if (!sgl)
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 484 goto err_malloc;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 485 } else {
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 486 buffer = kzalloc(buf_len, GFP_KERNEL);
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 487 if (!buffer)
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 488 goto err_malloc;
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 489 }
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 490 }
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 491
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 492 /* Check if the server's HCD supports SG */
ea44d190764b44 drivers/usb/usbip/stub_rx.c Suwan Kim 2019-08-28 493 if (use_sg && !udev->bus->sg_tablesize) {

Smatch thinks "use_sg" can be true when "buf_len" is zero. It's hard
to tell if Smatch is right or wrong without more context...

This is a bit strange. The meaning of "use_sg" is that client will
use scatter-gather and client's urb->num_sgs is not zero. And buffer
length should not be zero.

usb_sg and buf_len are both client-dependent variables, so I think
if they have wrong value in the server side, the client must have
sent use_sg and buf_len with incorrect values.

Did this error occur when compiling?

Smatch is doing static analysis, yes.

If then, Did Smatch also consider vhci tx side?

I'm not really sure... I can't reproduce the warning because on my
system Smatch doesn't parse usbip_recv() correctly so it ends up
silencing that warning. :/


Hi Suwan,

This is a problem that needs fixing. nents

/* allocate urb transfer buffer, if needed */
if (buf_len) {
if (use_sg) {
sgl = sgl_alloc(buf_len, GFP_KERNEL, &nents);

nents gets initialized here by sgl_alloc()

if (!sgl)
goto err_malloc;
} else {
buffer = kzalloc(buf_len, GFP_KERNEL);
if (!buffer)
goto err_malloc;
}
}

/* Check if the server's HCD supports SG */
if (use_sg && !udev->bus->sg_tablesize) {
/*
* If the server's HCD doesn't support SG, break a single SG
* request into several URBs and map each SG list entry to
* corresponding URB buffer. The previously allocated SG
* list is stored in priv->sgl (If the server's HCD support SG,
* SG list is stored only in urb->sg) and it is used as an
* indicator that the server split single SG request into
* several URBs. Later, priv->sgl is used by stub_complete() and
* stub_send_ret_submit() to reassemble the divied URBs.
*/
support_sg = 0;
num_urbs = nents;

I think nents will be valid here. Is there need for this additional
check here? You can fold this into the previous use_sg check, right
after the sg_alloc() success, I would think.

priv->completed_urbs = 0;
pdu->u.cmd_submit.transfer_flags &= ~URB_DMA_MAP_SG;
}


thanks,
-- Shuah