On Mon, 15 Oct 2007, David Miller wrote:

quoted text
> I want to help with this, but if I even breath on the kernel the bug
> goes away.  The race just gets harder to trigger, and if we just keep
> adding things it'll make the problem go away but for the absolutely
> wrong reasons.
> 
> The only way we will provably fix this is to make sure EHCI initialize
> fully, first, regardless of kernel config or what userland does.

Unfortunately that simply isn't possible.  No matter what you do, the 
user can always unload ehci-hcd and then load it back in again.

Do you have any idea _where_ in ohci_hub_control the hang still occurs?  
Is it the same unbounded reset loop?

Does the patch below satisfy both Davids?

Alan Stern



Index: usb-2.6/drivers/usb/host/ohci-hub.c
===================================================================
--- usb-2.6.orig/drivers/usb/host/ohci-hub.c
+++ usb-2.6/drivers/usb/host/ohci-hub.c
@@ -560,10 +560,10 @@ static void start_hnp(struct ohci_hcd *o
 /* called from some task, normally khubd */
 static inline int root_port_reset (struct ohci_hcd *ohci, unsigned port)
 {
-	__hc32 __iomem *portstat = &ohci->regs->roothub.portstatus [port];
-	u32	temp;
-	u16	now = ohci_readl(ohci, &ohci->regs->fmnumber);
-	u16	reset_done = now + PORT_RESET_MSEC;
+	__hc32 __iomem	*portstat = &ohci->regs->roothub.portstatus[port];
+	u32		temp;
+	unsigned long	reset_done = jiffies +
+				msecs_to_jiffies(PORT_RESET_MSEC);
 
 	/* build a "continuous enough" reset signal, with up to
 	 * 3msec gap between pulses.  scheduler HZ==100 must work;
@@ -578,6 +578,8 @@ static inline int root_port_reset (struc
 				return -ESHUTDOWN;
 			if (!(temp & RH_PS_PRS))
 				break;
+			if (time_after(jiffies, reset_done))
+				break;
 			udelay (500);
 		}
 
@@ -589,8 +591,7 @@ static inline int root_port_reset (struc
 		/* start the next reset, sleep till it's probably done */
 		ohci_writel (ohci, RH_PS_PRS, portstat);
 		msleep(PORT_RESET_HW_MSEC);
-		now = ohci_readl(ohci, &ohci->regs->fmnumber);
-	} while (tick_before(now, reset_done));
+	} while (time_before_eq(jiffies, reset_done));
 	/* caller synchronizes using PRSC */
 
 	return 0;

-
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/