Re: [PATCH v2 2/3] usb: xhci: This reworks ff8cbf250b448aac35589f6075082c3fcad8a8fe

From: Lu, Baolu
Date: Tue Nov 04 2014 - 22:10:45 EST



On 11/5/2014 12:58 AM, Alan Stern wrote:
On Tue, 4 Nov 2014, Lu Baolu wrote:

xhci: clear root port wake on bits if controller isn't wake-up capable

When system is being suspended, if host device is not wakeup capable,
xhci_suspend() needs to clear all root port wake on bits. Otherwise,
some platforms may generate spurious wakeup, even if PCI PME# is dis-
abled.

Signed-off-by: Lu Baolu <baolu.lu@xxxxxxxxxxxxxxx>
---
drivers/usb/host/xhci.c | 42 ++++++++++++++++++++++++++++++++++++++++++
1 file changed, 42 insertions(+)

diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
index 2a5d45b..cd57aae 100644
--- a/drivers/usb/host/xhci.c
+++ b/drivers/usb/host/xhci.c
@@ -35,6 +35,8 @@
#define DRIVER_AUTHOR "Sarah Sharp"
#define DRIVER_DESC "'eXtensible' Host Controller (xHC) Driver"
+#define PORT_WAKE_BITS (PORT_WKOC_E | PORT_WKDISC_E | PORT_WKCONN_E)
+
/* Some 0.95 hardware can't handle the chain bit on a Link TRB being cleared */
static int link_quirk;
module_param(link_quirk, int, S_IRUGO | S_IWUSR);
@@ -851,6 +853,42 @@ static void xhci_clear_command_ring(struct xhci_hcd *xhci)
xhci_set_cmd_ring_deq(xhci);
}
+static void xhci_disable_port_wake_on_bits(struct xhci_hcd *xhci)
+{
+ int port_index;
+ __le32 __iomem **port_array;
+ unsigned long flags;
+ u32 t1, t2;
+
+ spin_lock_irqsave(&xhci->lock, flags);
+
+ /* disble usb3 ports Wake bits*/
+ port_index = xhci->num_usb3_ports;
+ port_array = xhci->usb3_ports;
+ while (port_index--) {
+ t1 = readl(port_array[port_index]);
+ t2 = xhci_port_state_to_neutral(t1);
+ t2 &= ~PORT_WAKE_BITS;
+ t1 = xhci_port_state_to_neutral(t1);
+ if (t1 != t2)
+ writel(t2, port_array[port_index]);
+ }
+
+ /* disble usb2 ports Wake bits*/
+ port_index = xhci->num_usb2_ports;
+ port_array = xhci->usb2_ports;
+ while (port_index--) {
+ t1 = readl(port_array[port_index]);
+ t2 = xhci_port_state_to_neutral(t1);
+ t2 &= ~PORT_WAKE_BITS;
+ t1 = xhci_port_state_to_neutral(t1);
+ if (t1 != t2)
+ writel(t2, port_array[port_index]);
+ }
+
+ spin_unlock_irqrestore(&xhci->lock, flags);
+}
+
/*
* Stop HC (not bus-specific)
*
@@ -868,6 +906,10 @@ int xhci_suspend(struct xhci_hcd *xhci)
xhci->shared_hcd->state != HC_STATE_SUSPENDED)
return -EINVAL;
+ /* Clear root port wake on bits if not wakeup capable. */
+ if (!device_may_wakeup(hcd->self.controller))
+ xhci_disable_port_wake_on_bits(xhci);
+
/* Don't poll the roothubs on bus suspend. */
xhci_dbg(xhci, "%s: stopping port polling.\n", __func__);
clear_bit(HCD_FLAG_POLL_RH, &hcd->flags);
This is better but still wrong. Remember, this same code gets called
for system suspend _and_ for runtime suspend. During runtime suspend,
wakeup is always supposed to be turned on, even if device_may_wakeup()
is false. That's because device_may_wakeup() refers only to system
suspend. What you need to test is the do_wakeup flag, which should be
passed into xhci_suspend() by xhci_pci_suspend() and
xhci_plat_suspend().
Yes, it should cover runtime suspend as well. Thanks for the comments. I will resend the patch.

Another problem is in the patch description and the comments. If
device_may_wakeup() returns false, it doesn't mean the controller isn't
wakeup-capable -- it means the controller isn't _allowed_ to wake up
the system. Those are two different things.

Accept, I will change this in new patch version.

Finally, the code in xhci_disable_port_wake_on_bits() looks a little
peculiar -- I'm not sure about all those calls to
xhci_port_state_to_neutral(). But I'm not an expert on that; Mathias
will have to advise on it.
This part of code was written with reference to code in xhci-hub.c.

Comments of xhci_port_state_to_neutral():
/*
* Given a port state, this function returns a value that would result in the
* port being in the same state, if the value was written to the port status
* control register.
* Save Read Only (RO) bits and save read/write bits where
* writing a 0 clears the bit and writing a 1 sets the bit (RWS).
* For all other types (RW1S, RW1CS, RW, and RZ), writing a '0' has no effect.
*/

This calculation is used to avoid side effect of changing other bit fields.

Thanks,
-baolu


Alan Stern


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