Re: KASAN: global-out-of-bounds Read in qmi_wwan_probe

[Andrey Konovalov]

On Mon, Jun 24, 2019 at 2:59 PM BjÃrn Mork <bjorn@xxxxxxx> wrote:
>
> syzbot <syzbot+b68605d7fadd21510de1@xxxxxxxxxxxxxxxxxxxxxxxxx> writes:
>
> > Hello,
> >
> > syzbot found the following crash on:
> >
> > HEAD commit:    9939f56e usb-fuzzer: main usb gadget fuzzer driver
> > git tree:       https://github.com/google/kasan.git usb-fuzzer
> > console output: https://syzkaller.appspot.com/x/log.txt?x=1615a669a00000
> > kernel config:  https://syzkaller.appspot.com/x/.config?x=df134eda130bb43a
> > dashboard link: https://syzkaller.appspot.com/bug?extid=b68605d7fadd21510de1
> > compiler:       gcc (GCC) 9.0.0 20181231 (experimental)
> > syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=10630af6a00000
> > C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=1127da69a00000
> >
> > IMPORTANT: if you fix the bug, please add the following tag to the commit:
> > Reported-by: syzbot+b68605d7fadd21510de1@xxxxxxxxxxxxxxxxxxxxxxxxx
> >
> > usb 1-1: new high-speed USB device number 2 using dummy_hcd
> > usb 1-1: Using ep0 maxpacket: 8
> > usb 1-1: New USB device found, idVendor=12d1, idProduct=14f1,
> > bcdDevice=d4.d9
> > usb 1-1: New USB device strings: Mfr=0, Product=0, SerialNumber=0
> > usb 1-1: config 0 descriptor??
> > ==================================================================
> > BUG: KASAN: global-out-of-bounds in qmi_wwan_probe+0x342/0x360
> > drivers/net/usb/qmi_wwan.c:1417
> > Read of size 8 at addr ffffffff8618c140 by task kworker/1:1/22
> >
> > CPU: 1 PID: 22 Comm: kworker/1:1 Not tainted 5.2.0-rc5+ #11
> > Hardware name: Google Google Compute Engine/Google Compute Engine,
> > BIOS Google 01/01/2011
> > Workqueue: usb_hub_wq hub_event
> > Call Trace:
> >  __dump_stack lib/dump_stack.c:77 [inline]
> >  dump_stack+0xca/0x13e lib/dump_stack.c:113
> >  print_address_description+0x67/0x231 mm/kasan/report.c:188
> >  __kasan_report.cold+0x1a/0x32 mm/kasan/report.c:317
> >  kasan_report+0xe/0x20 mm/kasan/common.c:614
> >  qmi_wwan_probe+0x342/0x360 drivers/net/usb/qmi_wwan.c:1417
> >  usb_probe_interface+0x305/0x7a0 drivers/usb/core/driver.c:361
> >  really_probe+0x281/0x660 drivers/base/dd.c:509
> >  driver_probe_device+0x104/0x210 drivers/base/dd.c:670
> >  __device_attach_driver+0x1c2/0x220 drivers/base/dd.c:777
> >  bus_for_each_drv+0x15c/0x1e0 drivers/base/bus.c:454
> >
>
> Hello Kristian!
>
> I need some help understanding this...  IIUC syzbot is claiming an
> out-of-bounds access at line 1417 in v5.2-rc5.  Or whatever - I'm having
> a hard time deciphering what kernel version the bot is actually
> testing. The claimed HEAD is not a kernel commit.  At least not in my
> kernel...

The bot currently tests this tree:
https://github.com/google/kasan/tree/usb-fuzzer, which is essentially
5.2-rc5.

>
>
> But if this is correct, then it points to the info->data access you
> recently added:
>
> 822e44b45eb99 (Kristian Evensen        2019-03-02 13:32:26 +0100 1409)  /* Several Quectel modems supports dynamic interface configuration, so
> 7c5cca3588545 (Kristian Evensen        2018-09-08 13:50:48 +0200 1410)   * we need to match on class/subclass/protocol. These values are
> 7c5cca3588545 (Kristian Evensen        2018-09-08 13:50:48 +0200 1411)   * identical for the diagnostic- and QMI-interface, but bNumEndpoints is
> 7c5cca3588545 (Kristian Evensen        2018-09-08 13:50:48 +0200 1412)   * different. Ignore the current interface if the number of endpoints
> e4bf63482c309 (Kristian Evensen        2019-04-07 15:39:09 +0200 1413)   * equals the number for the diag interface (two).
> 7c5cca3588545 (Kristian Evensen        2018-09-08 13:50:48 +0200 1414)   */
> e4bf63482c309 (Kristian Evensen        2019-04-07 15:39:09 +0200 1415)  info = (void *)&id->driver_info;
> e4bf63482c309 (Kristian Evensen        2019-04-07 15:39:09 +0200 1416)
> e4bf63482c309 (Kristian Evensen        2019-04-07 15:39:09 +0200 1417)  if (info->data & QMI_WWAN_QUIRK_QUECTEL_DYNCFG) {
> e4bf63482c309 (Kristian Evensen        2019-04-07 15:39:09 +0200 1418)          if (desc->bNumEndpoints == 2)
> e4bf63482c309 (Kristian Evensen        2019-04-07 15:39:09 +0200 1419)                  return -ENODEV;
> e4bf63482c309 (Kristian Evensen        2019-04-07 15:39:09 +0200 1420)  }
>
>
> I must be blind. I cannot see how this could end up failing.
> id->driver_info is always set to one of qmi_wwan_info,
> qmi_wwan_info_quirk_dtr or qmi_wwan_info_quirk_quectel_dyncfg at this
> point.  How does that end up out-of-bounds?

I've run the reproducer locally and checked the addresses. The
structures that you mentioned are at:

gef> p &qmi_wwan_info
$1 = (const struct driver_info *) 0xffffffff85d32e80 <qmi_wwan_info>
gef> p &qmi_wwan_info_quirk_dtr
$2 = (const struct driver_info *) 0xffffffff85d32dc0 <qmi_wwan_info_quirk_dtr>
gef> p &qmi_wwan_info_quirk_quectel_dyncfg
$3 = (const struct driver_info *) 0xffffffff85d32d00
<qmi_wwan_info_quirk_quectel_dyncfg>

And the bad access for me happens on address 0xffffffff85d32ce0, so it
seems that driver_info somehow ended up lying below
qmi_wwan_info_quirk_quectel_dyncfg.

gef> x/6gx 0xffffffff85d32ce0
0xffffffff85d32ce0: 0x0000000000000000 0x0000000000000000
0xffffffff85d32cf0: 0x0000000000000000 0x0000000000000000
0xffffffff85d32d00 <qmi_wwan_info_quirk_quectel_dyncfg>:
0xffffffff85d32cc0 0x0000000000000600

>
>
>
> BjÃrn