usb/core: usb_alloc_dev(): fix setting of ->portnum
authorNicolai Stange <nicstange@gmail.com>
Thu, 17 Mar 2016 22:53:02 +0000 (23:53 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 18 Mar 2016 16:19:02 +0000 (09:19 -0700)
commit7222c832254a75dcd67d683df75753d4a4e125bb
treea37d0e805899b70c18f99323c21dd32830bad0f7
parent4ec0ef3a82125efc36173062a50624550a900ae0
usb/core: usb_alloc_dev(): fix setting of ->portnum

With commit 69bec7259853 ("USB: core: let USB device know device node"),
the port1 argument of usb_alloc_dev() gets overwritten as follows:

  ... usb_alloc_dev(..., unsigned port1)
  {
    ...
    if (!parent->parent) {
      port1 = usb_hcd_find_raw_port_number(..., port1);
    }
    ...
  }

Later on, this now overwritten port1 gets assigned to ->portnum:

  dev->portnum = port1;

However, since xhci_find_raw_port_number() isn't idempotent, the
aforementioned commit causes a number of KASAN splats like the following:

  BUG: KASAN: slab-out-of-bounds in xhci_find_raw_port_number+0x98/0x170
                                       at addr ffff8801d9311670
  Read of size 8 by task kworker/2:1/87
  [...]
  Workqueue: usb_hub_wq hub_event
   0000000000000188 000000005814b877 ffff8800cba17588 ffffffff8191447e
   0000000041b58ab3 ffffffff82a03209 ffffffff819143a2 ffffffff82a252f4
   ffff8801d93115e0 0000000000000188 ffff8801d9311628 ffff8800cba17588
  Call Trace:
   [<ffffffff8191447e>] dump_stack+0xdc/0x15e
   [<ffffffff819143a2>] ? _atomic_dec_and_lock+0xa2/0xa2
   [<ffffffff814e2cd1>] ? print_section+0x61/0xb0
   [<ffffffff814e4939>] print_trailer+0x179/0x2c0
   [<ffffffff814f0d84>] object_err+0x34/0x40
   [<ffffffff814f4388>] kasan_report_error+0x2f8/0x8b0
   [<ffffffff814eb91e>] ? __slab_alloc+0x5e/0x90
   [<ffffffff812178c0>] ? __lock_is_held+0x90/0x130
   [<ffffffff814f5091>] kasan_report+0x71/0xa0
   [<ffffffff814ec082>] ? kmem_cache_alloc_trace+0x212/0x560
   [<ffffffff81d99468>] ? xhci_find_raw_port_number+0x98/0x170
   [<ffffffff814f33d4>] __asan_load8+0x64/0x70
   [<ffffffff81d99468>] xhci_find_raw_port_number+0x98/0x170
   [<ffffffff81db0105>] xhci_setup_addressable_virt_dev+0x235/0xa10
   [<ffffffff81d9ea51>] xhci_setup_device+0x3c1/0x1430
   [<ffffffff8121cddd>] ? trace_hardirqs_on+0xd/0x10
   [<ffffffff81d9fac0>] ? xhci_setup_device+0x1430/0x1430
   [<ffffffff81d9fad3>] xhci_address_device+0x13/0x20
   [<ffffffff81d2081a>] hub_port_init+0x55a/0x1550
   [<ffffffff81d28705>] hub_event+0xef5/0x24d0
   [<ffffffff81d27810>] ? hub_port_debounce+0x2f0/0x2f0
   [<ffffffff8195e1ee>] ? debug_object_deactivate+0x1be/0x270
   [<ffffffff81210203>] ? print_rt_rq+0x53/0x2d0
   [<ffffffff8121657d>] ? trace_hardirqs_off+0xd/0x10
   [<ffffffff8226acfb>] ? _raw_spin_unlock_irqrestore+0x5b/0x60
   [<ffffffff81250000>] ? irq_domain_set_hwirq_and_chip+0x30/0xb0
   [<ffffffff81256339>] ? debug_lockdep_rcu_enabled+0x39/0x40
   [<ffffffff812178c0>] ? __lock_is_held+0x90/0x130
   [<ffffffff81196877>] process_one_work+0x567/0xec0
  [...]

Afterwards, xhci reports some functional errors:

  xhci_hcd 0000:00:14.0: ERROR: unexpected setup address command completion
                                code 0x11.
  xhci_hcd 0000:00:14.0: ERROR: unexpected setup address command completion
                                code 0x11.
  usb 4-3: device not accepting address 2, error -22

Fix this by not overwriting the port1 argument in usb_alloc_dev(), but
storing the raw port number as required by OF in an additional variable,
raw_port.

Fixes: 69bec7259853 ("USB: core: let USB device know device node")
Signed-off-by: Nicolai Stange <nicstange@gmail.com>
Acked-by: Alan Stern <stern@rowland.harvard.edu>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/usb/core/usb.c