summaryrefslogtreecommitdiff
path: root/drivers
diff options
context:
space:
mode:
authorSarah Sharp <sarah.a.sharp@linux.intel.com>2010-10-14 18:22:54 +0400
committerGreg Kroah-Hartman <gregkh@suse.de>2010-10-22 21:22:12 +0400
commit85f0ff46960c2853fd1436a56798260eae91db86 (patch)
tree10d68d56372ab2c23963c53ee8239ff9092455bc /drivers
parentc8d4af8e2af12cd4835ba5c4b54bdafe9deda71a (diff)
downloadlinux-85f0ff46960c2853fd1436a56798260eae91db86.tar.xz
usb: Fix issue with USB 3.0 devices after system resume
When the system suspends and a host controller's power is lost, the USB core attempts to revive any USB devices that had the persist_enabled flag set. For non-SuperSpeed devices, it will disable the port, and then set the udev->reset_resume flag. This will cause the USB core to reset the device, verify the device descriptors to make sure it's the same device, and re-install any non-default configurations or alternate interface settings. However, we can't disable SuperSpeed root hub ports because that turns off SuperSpeed terminations, which will inhibit any devices connecting at USB 3.0 speeds. (Plus external hubs don't allow SuperSpeed ports to be disabled.) Because of this logic in hub_activate(): /* We can forget about a "removed" device when there's a * physical disconnect or the connect status changes. */ if (!(portstatus & USB_PORT_STAT_CONNECTION) || (portchange & USB_PORT_STAT_C_CONNECTION)) clear_bit(port1, hub->removed_bits); if (!udev || udev->state == USB_STATE_NOTATTACHED) { /* Tell khubd to disconnect the device or * check for a new connection */ if (udev || (portstatus & USB_PORT_STAT_CONNECTION)) set_bit(port1, hub->change_bits); } else if (portstatus & USB_PORT_STAT_ENABLE) { /* The power session apparently survived the resume. * If there was an overcurrent or suspend change * (i.e., remote wakeup request), have khubd * take care of it. */ if (portchange) set_bit(port1, hub->change_bits); } else if (udev->persist_enabled) { udev->reset_resume = 1; set_bit(port1, hub->change_bits); } else { /* The power session is gone; tell khubd */ usb_set_device_state(udev, USB_STATE_NOTATTACHED); set_bit(port1, hub->change_bits); } a SuperSpeed device after a resume with a loss of power will never get the reset_resume flag set. Instead the core will assume the power session survived and that the device still has the same address, configuration, and alternate interface settings. The xHCI host controller will have no knowledge of the device (since all xhci_virt_devices were destroyed when power loss was discovered, and xhci_discover_or_reset_device() has not been called), and all URBs to the device will fail. If the device driver responds by resetting the device, everything will continue smoothly. However, if lsusb is used before the device driver resets the device (or there is no driver), then all lsusb descriptor fetches will fail. The quick fix is to pretend the port is disabled in hub_activate(), by clearing the local variable. But I'm not sure what other parts of the hub driver need to be changed because they have assumptions about when ports will be disabled. Signed-off-by: Sarah Sharp <sarah.a.sharp@linux.intel.com> Signed-off-by: Andiry Xu <andiry.xu@amd.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Diffstat (limited to 'drivers')
-rw-r--r--drivers/usb/core/hub.c3
1 files changed, 3 insertions, 0 deletions
diff --git a/drivers/usb/core/hub.c b/drivers/usb/core/hub.c
index f07ab71859dd..7f82c48a0bae 100644
--- a/drivers/usb/core/hub.c
+++ b/drivers/usb/core/hub.c
@@ -758,6 +758,9 @@ static void hub_activate(struct usb_hub *hub, enum hub_activation_type type)
clear_port_feature(hdev, port1,
USB_PORT_FEAT_ENABLE);
portstatus &= ~USB_PORT_STAT_ENABLE;
+ } else {
+ /* Pretend that power was lost for USB3 devs */
+ portstatus &= ~USB_PORT_STAT_ENABLE;
}
}