On Thu, 2011-09-22 at 14:49 +0200, Oliver Neukum wrote:
> Hi,
>
> I see a device that fails during initialization after S3. It doesn't matter
> when S3 happens. After the controller has undergone an S3 cycle
> with or without the device, it becomes unable to cope with the device.
>
> I made logs with Andiry's debug patch cmd_ring.patch from the thread
> "problems with usb 3.0 on clevo p150hm"
> The problem is reproducible on xhci implementations from two vendors.
> It happens at least on 3.0.4, 3.1-rc5 and 3.1-rc7
>
> The logs are
> xhcilog2.txt: taken with device plugged in
> xhcilog.txt: taken with the device plugged in after resumption from S3
>
> The device is problematic even on UHCI:
>
Oliver,
Can you apply the patch attached to see if it helps?
Sarah,
According to the xhcilog2.txt Oliver sent, do you think there is a bug
in the command ring TRB queueing? The last TRB of command ring should be
a link TRB, but it's replaced by a Set TRDP command, and the deq and enq
pointer moves out of the segment. Maybe there is something wrong in the
room_on_ring() check for the command ring?
Thanks,
Andiry
diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
index 70f8da0..7f3814c 100644
--- a/drivers/usb/host/xhci.c
+++ b/drivers/usb/host/xhci.c
@@ -782,6 +782,9 @@ int xhci_resume(struct xhci_hcd *xhci, bool hibernated)
temp = xhci_readl(xhci, &xhci->op_regs->status);
}
+ /* force the host to re-initialize */
+ temp = STS_SRE;
+
/* If restore operation fails, re-initialize the HC during resume */
if ((temp & STS_SRE) || hibernated) {
/* Let the USB core know _both_ roothubs lost power. */
