kern/73000: UHCI driver fails to detect (certain) interrupts on boot

[Daan Vreeken [PA4DAN]]

>Number:         73000
>Category:       kern
>Synopsis:       UHCI driver fails to detect (certain) interrupts on boot
>Confidential:   no
>Severity:       serious
>Priority:       low
>Responsible:    freebsd-bugs
>State:          open
>Quarter:        
>Keywords:       
>Date-Required:
>Class:          sw-bug
>Submitter-Id:   current-users
>Arrival-Date:   Fri Oct 22 12:50:27 GMT 2004
>Closed-Date:
>Last-Modified:
>Originator:     Daan Vreeken [PA4DAN]
>Release:        FreeBSD 5.1-RELEASE i386
>Organization:
>Environment:
System: FreeBSD Racebeest.Danovitsch.LAN 5.1-RELEASE FreeBSD 5.1-RELEASE #18: Thu Oct 7 11:39:42 CEST 2004 root at Racebeest.Danovitsch.LAN:/usr/src.5.1/sys/i386/compile/Laptop i386


	
>Description:
	During boot the UHCI driver uses the routine uhci_waitintr() to
detect interrupts (since interrupts are disabled at that point). The routine
checks the UHCI interrupt status register and if it detects an interrupt
condition it manually calls uhci_intr1() to handle the interrupt. The fault
is that uhci_waitintr() only checks for UHCI_STS_USBINTR and not all other
interrupt flags (like UHCI_STS_USBEI). 
	This causes interrupts to be ignored (if UHCI_STS_USBINTR isn't
set). Having an axe(4) USB ethernet adapter plugged in during boot causes
the system to freeze (for a very long period of time) during initialisation
of the device. Eventually uhci_waitintr() will timeout, but that can take up
to minutes per request. When using a USB WLAN device with my own driver (
http://vitsch.net/bsd/atuwi ) causes the system to stall for hours at boot,
since the driver does about 100 USB requests during attach.
	
>How-To-Repeat:
	Plugin an axe(4) or atuwi device into a system with a UHCI
controller (not sure if all UHCI controllers exhibit the same behaviour) and
boot the system. Booting should take minutes up to hours.
	
>Fix:
	Apply the following patch (to -current). The patch makes
uhci_waitintr() and uhci_poll() aware of all interrupt flags in the
interrupt status register instead of just UHCI_STS_USBINT.
	

--- uhci-waitintr-2004-10-22.diff begins here ---
--- uhci.c.org	Fri Oct 22 12:16:29 2004
+++ uhci.c	Fri Oct 22 12:21:11 2004
@@ -1552,7 +1552,7 @@
 	for (; timo >= 0; timo--) {
 		usb_delay_ms(&sc->sc_bus, 1);
 		DPRINTFN(20,("uhci_waitintr: 0x%04x\n", UREAD2(sc, UHCI_STS)));
-		if (UREAD2(sc, UHCI_STS) & UHCI_STS_USBINT)
+		if (UREAD2(sc, UHCI_STS) & UHCI_STS_ALLINTRS)
 			uhci_intr1(sc);
 		if (xfer->status != USBD_IN_PROGRESS)
 			return;
@@ -1576,7 +1576,7 @@
 {
 	uhci_softc_t *sc = (uhci_softc_t *)bus;
 
-	if (UREAD2(sc, UHCI_STS) & UHCI_STS_USBINT)
+	if (UREAD2(sc, UHCI_STS) & UHCI_STS_ALLINTRS)
 		uhci_intr1(sc);
 }
 
--- uhci-waitintr-2004-10-22.diff ends here ---

>Release-Note:
>Audit-Trail:
>Unformatted: