[PATCH] USB: EHCI port tweaks

One change may improve some S1 or S3 resume cases, and the other
seems mostly to explain some strange state "lsusb" would show.
Two fixes:

  - On resume, don't think about resuming any unpowered port, or
    resetting any port with OWNER set to the OHCI/UHCI companion.
    This will make some S1 and S3 resume scenarios work better.

  - PORT_CSC was not being cleared correctly in ehci_hub_status_data.
    This was visible at least through current versions of "lsusb",
    and might have caused some other hub related strangeness.

    The fix addresses all three write-to-clear bits, using the same
    approach that UHCI happens to use:  a mask of bits that are
    cleared in most writes to that port status register.

Original patch seems to have been from from William.Morrow@amd.com
and this version (from David) finishes the write-to-clear changes.

Signed-off-by: Jordan Crouse <jordan.crouse@amd.com>
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>

1 files changed, 6 insertions, 2 deletions

diff --git a/drivers/usb/host/ehci-hcd.c b/drivers/usb/host/ehci-hcd.c
index 2507e89..2f7037c 100644
--- a/drivers/usb/host/ehci-hcd.c
+++ b/drivers/usb/host/ehci-hcd.c
@@ -759,12 +759,16 @@ static int ehci_resume (struct usb_hcd *hcd)
 	if (time_before (jiffies, ehci->next_statechange))
 		msleep (100);
 
-	/* If any port is suspended, we know we can/must resume the HC. */
+	/* If any port is suspended (or owned by the companion),
+	 * we know we can/must resume the HC (and mustn't reset it).
+	 */
 	for (port = HCS_N_PORTS (ehci->hcs_params); port > 0; ) {
 		u32	status;
 		port--;
 		status = readl (&ehci->regs->port_status [port]);
-		if (status & PORT_SUSPEND) {
+		if (!(status & PORT_POWER))
+			continue;
+		if (status & (PORT_SUSPEND | PORT_OWNER)) {
 			down (&hcd->self.root_hub->serialize);
 			retval = ehci_hub_resume (hcd);
 			up (&hcd->self.root_hub->serialize);