Re: Deadlock between device_detach() and usbd_do_request_flags()

[Hans Petter Selasky]

On 09/05/16 12:21, Andriy Voskoboinyk wrote:

Mon, 05 Sep 2016 12:10:34 +0300 було написано Hans Petter Selasky
:

Works fine, thanks!

P.S. Reliable test case:
1) ifconfig wlan1 create wlandev 
2) wpa_supplicant -i wlan1 -c /etc/wpa_supplicant.conf &
   * wait for CTRL-EVENT-CONNECTED event
3) usbconfig -d . power_off


Hi,

Can you test the attached patch? Does it solve your issue?

--HPS




FYI:

https://svnweb.freebsd.org/changeset/base/305421

--HPS
___
freebsd-wireless@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-wireless
To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"

Re: Deadlock between device_detach() and usbd_do_request_flags()

[Andriy Voskoboinyk]

Mon, 05 Sep 2016 12:10:34 +0300 було написано Hans Petter Selasky  
:


Works fine, thanks!

P.S. Reliable test case:
1) ifconfig wlan1 create wlandev 
2) wpa_supplicant -i wlan1 -c /etc/wpa_supplicant.conf &
   * wait for CTRL-EVENT-CONNECTED event
3) usbconfig -d . power_off


Hi,

Can you test the attached patch? Does it solve your issue?

--HPS

___
freebsd-wireless@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-wireless
To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"

Re: Deadlock between device_detach() and usbd_do_request_flags()

[Hans Petter Selasky]

Hi,

Can you test the attached patch? Does it solve your issue?

--HPS
Index: sys/dev/usb/usb_device.c
===
--- sys/dev/usb/usb_device.c	(revision 304569)
+++ sys/dev/usb/usb_device.c	(working copy)
@@ -1585,6 +1585,7 @@
 	/* initialise our SX-lock */
 	sx_init_flags(>enum_sx, "USB config SX lock", SX_DUPOK);
 	sx_init_flags(>sr_sx, "USB suspend and resume SX lock", SX_NOWITNESS);
+	sx_init_flags(>ctrl_sx, "USB control transfer SX lock", SX_DUPOK);
 
 	cv_init(>ctrlreq_cv, "WCTRL");
 	cv_init(>ref_cv, "UGONE");
@@ -2195,6 +2196,7 @@
 	
 	sx_destroy(>enum_sx);
 	sx_destroy(>sr_sx);
+	sx_destroy(>ctrl_sx);
 
 	cv_destroy(>ctrlreq_cv);
 	cv_destroy(>ref_cv);
Index: sys/dev/usb/usb_device.h
===
--- sys/dev/usb/usb_device.h	(revision 304569)
+++ sys/dev/usb/usb_device.h	(working copy)
@@ -183,6 +183,7 @@
 	struct usb_udev_msg cs_msg[2];
 	struct sx enum_sx;
 	struct sx sr_sx;
+  	struct sx ctrl_sx;
 	struct mtx device_mtx;
 	struct cv ctrlreq_cv;
 	struct cv ref_cv;
Index: sys/dev/usb/usb_request.c
===
--- sys/dev/usb/usb_request.c	(revision 304569)
+++ sys/dev/usb/usb_request.c	(working copy)
@@ -418,7 +418,6 @@
 	uint16_t length;
 	uint16_t temp;
 	uint16_t acttemp;
-	uint8_t do_unlock;
 
 	if (timeout < 50) {
 		/* timeout is too small */
@@ -460,16 +459,16 @@
 	}
 
 	/*
-	 * Grab the USB device enumeration SX-lock serialization is
-	 * achieved when multiple threads are involved:
+	 * Serialize access to this function:
 	 */
-	do_unlock = usbd_enum_lock(udev);
+	sx_xlock(>ctrl_sx);
 
 	/*
 	 * We need to allow suspend and resume at this point, else the
 	 * control transfer will timeout if the device is suspended!
 	 */
-	usbd_sr_unlock(udev);
+	if (usbd_enum_is_locked(udev))
+		usbd_sr_unlock(udev);
 
 	hr_func = usbd_get_hr_func(udev);
 
@@ -713,10 +712,10 @@
 	USB_XFER_UNLOCK(xfer);
 
 done:
-	usbd_sr_lock(udev);
+	sx_xunlock(>ctrl_sx);
 
-	if (do_unlock)
-		usbd_enum_unlock(udev);
+	if (usbd_enum_is_locked(udev))
+		usbd_sr_lock(udev);
 
 	if ((mtx != NULL) && (mtx != ))
 		mtx_lock(mtx);
___
freebsd-wireless@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-wireless
To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"

Re: Deadlock between device_detach() and usbd_do_request_flags()

[Hans Petter Selasky]

On 09/05/16 09:53, Hans Petter Selasky wrote:


Hi,

I think the right solution is to let usbd_do_request_flags() use its own
SX lock for synchronization, instead of re-using the enumeration SX
lock. What do you think about that?

--HPS



Hi,

Another approach which will work is to setup your own USB control 
endpoint xfer, and use that. I'll have a look and see what can be done.


--HPS

___
freebsd-wireless@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-wireless
To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"

Re: Deadlock between device_detach() and usbd_do_request_flags()

[Hans Petter Selasky]

On 09/04/16 23:20, Andriy Voskoboinyk wrote:

There is a rare, but reproducible deadlock for wlan(4) drivers:

Thread 1:
 * uhub_explore_handle_re_enumerate() (obtains enum_sx lock)
 * usbd_set_config_index()
 * usb_unconfigure()
 * usb_detach_device()
 * usb_detach_device_sub()
 * 
   typically  is executed here (prevents
another possible deadlock?)
 * ieee80211_ifdetach()
 * ieee80211_vap_destroy()
 * ic_vap_delete>
 * ieee80211_vap_detach()
   here it calls ieee80211_stop() and waits for  -> INIT state
   transition

Thread 2 (started from thread 1):
 * ieee80211_newstate_cb()
 * vap->iv_newstate()
   here: if the driver will try to call usbd_do_request_flags()
   (typically via  / ) it will hang
   (because enum_sx lock is already held by thread 1).


Another way: execute some periodical task that will try to access
some registers (urtwn_temp_calib(), rum_ratectl_task(),
run_ratectl_cb()) while thread 1 is running - deadlock is
here too, since  will wait for them indefinitely
(via ieee80211_draintask())

Right now the most obvious (and, probably, wrong) way is to just
detect & release all locks (usbd_enum_unlock()) for
ieee80211_ifdetach() / ieee80211_draintask() and re-acquire them
later (not tested yet).



Hi,

I think the right solution is to let usbd_do_request_flags() use its own 
SX lock for synchronization, instead of re-using the enumeration SX 
lock. What do you think about that?


--HPS

___
freebsd-wireless@freebsd.org mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-wireless
To unsubscribe, send any mail to "freebsd-wireless-unsubscr...@freebsd.org"