• Alan Stern's avatar
    USB: dummy-hcd: Fix deadlock caused by disconnect detection · ab219221
    Alan Stern authored
    The dummy-hcd driver calls the gadget driver's disconnect callback
    under the wrong conditions.  It should invoke the callback when Vbus
    power is turned off, but instead it does so when the D+ pullup is
    turned off.
    
    This can cause a deadlock in the composite core when a gadget driver
    is unregistered:
    
    [   88.361471] ============================================
    [   88.362014] WARNING: possible recursive locking detected
    [   88.362580] 4.14.0-rc2+ #9 Not tainted
    [   88.363010] --------------------------------------------
    [   88.363561] v4l_id/526 is trying to acquire lock:
    [   88.364062]  (&(&cdev->lock)->rlock){....}, at: [<ffffffffa0547e03>] composite_disconnect+0x43/0x100 [libcomposite]
    [   88.365051]
    [   88.365051] but task is already holding lock:
    [   88.365826]  (&(&cdev->lock)->rlock){....}, at: [<ffffffffa0547b09>] usb_function_deactivate+0x29/0x80 [libcomposite]
    [   88.366858]
    [   88.366858] other info that might help us debug this:
    [   88.368301]  Possible unsafe locking scenario:
    [   88.368301]
    [   88.369304]        CPU0
    [   88.369701]        ----
    [   88.370101]   lock(&(&cdev->lock)->rlock);
    [   88.370623]   lock(&(&cdev->lock)->rlock);
    [   88.371145]
    [   88.371145]  *** DEADLOCK ***
    [   88.371145]
    [   88.372211]  May be due to missing lock nesting notation
    [   88.372211]
    [   88.373191] 2 locks held by v4l_id/526:
    [   88.373715]  #0:  (&(&cdev->lock)->rlock){....}, at: [<ffffffffa0547b09>] usb_function_deactivate+0x29/0x80 [libcomposite]
    [   88.374814]  #1:  (&(&dum_hcd->dum->lock)->rlock){....}, at: [<ffffffffa05bd48d>] dummy_pullup+0x7d/0xf0 [dummy_hcd]
    [   88.376289]
    [   88.376289] stack backtrace:
    [   88.377726] CPU: 0 PID: 526 Comm: v4l_id Not tainted 4.14.0-rc2+ #9
    [   88.378557] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1 04/01/2014
    [   88.379504] Call Trace:
    [   88.380019]  dump_stack+0x86/0xc7
    [   88.380605]  __lock_acquire+0x841/0x1120
    [   88.381252]  lock_acquire+0xd5/0x1c0
    [   88.381865]  ? composite_disconnect+0x43/0x100 [libcomposite]
    [   88.382668]  _raw_spin_lock_irqsave+0x40/0x54
    [   88.383357]  ? composite_disconnect+0x43/0x100 [libcomposite]
    [   88.384290]  composite_disconnect+0x43/0x100 [libcomposite]
    [   88.385490]  set_link_state+0x2d4/0x3c0 [dummy_hcd]
    [   88.386436]  dummy_pullup+0xa7/0xf0 [dummy_hcd]
    [   88.387195]  usb_gadget_disconnect+0xd8/0x160 [udc_core]
    [   88.387990]  usb_gadget_deactivate+0xd3/0x160 [udc_core]
    [   88.388793]  usb_function_deactivate+0x64/0x80 [libcomposite]
    [   88.389628]  uvc_function_disconnect+0x1e/0x40 [usb_f_uvc]
    
    This patch changes the code to test the port-power status bit rather
    than the port-connect status bit when deciding whether to isue the
    callback.
    Signed-off-by: default avatarAlan Stern <stern@rowland.harvard.edu>
    Reported-by: default avatarDavid Tulloh <david@tulloh.id.au>
    CC: <stable@vger.kernel.org>
    Signed-off-by: default avatarFelipe Balbi <felipe.balbi@linux.intel.com>
    ab219221
dummy_hcd.c 72 KB