[PATCH] usb: xhci: prevent spurious root hub resume during suspend

[PATCH] usb: xhci: prevent spurious root hub resume during suspend

[Pei Xiao]

handle_port_status() unconditionally resumes the root hub when
hcd->state is HC_STATE_SUSPENDED, regardless of whether a device is
actually attached to the port.  This can cause system suspend to fail
with -EBUSY if a spurious port change event arrives on an empty port
after its bus has entered suspend.

Fix this by checking PORT_CONNECT in handle_port_status() before
resuming the root hub.  If no device is actually present on the port,
the event is spurious and should be silently ignored rather than
aborting suspend.

Logs:
handle_port_status:1700: xhci_hcd 0000:04:00.0: Port change event, 3-2, id 6, portsc: 0x202a0
handle_port_status:1706: xhci_hcd 0000:04:00.0: resume root hub
handle_port_status:1821: xhci_hcd 0000:04:00.0: handle_port_status: starting port polling.
PM: pci_pm_suspend(): hcd_pci_suspend+0x0/0x38 returns -16
xhci_suspend:1017: xhci_hcd 0000:03:00.0: xhci_suspend: stopping port polling.
PM: dpm_run_callback(): pci_pm_suspend+0x0/0x190 returns -16

Signed-off-by: Pei Xiao <xiaopei01@kylinos.cn>
---
 drivers/usb/host/xhci-ring.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
index 4f98d8269625..3bfe49e788ed 100644
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -2042,8 +2042,13 @@ static void handle_port_status(struct xhci_hcd *xhci, union xhci_trb *event)
 	trace_xhci_handle_port_status(port, portsc);
 
 	if (hcd->state == HC_STATE_SUSPENDED) {
-		xhci_dbg(xhci, "resume root hub\n");
-		usb_hcd_resume_root_hub(hcd);
+		if (portsc & PORT_CONNECT) {
+			xhci_dbg(xhci, "resume root hub\n");
+			usb_hcd_resume_root_hub(hcd);
+		} else {
+			xhci_dbg(xhci, "spurious port change on empty port %d-%d, ignored\n",
+				 hcd->self.busnum, hcd_portnum + 1);
+		}
 	}
 
 	if (vdev && (portsc & PORT_PLS_MASK) == XDEV_INACTIVE) {
-- 
2.25.1

Re: [PATCH] usb: xhci: prevent spurious root hub resume during suspend

[Mathias Nyman]

On 6/30/26 10:36, Pei Xiao wrote:
> handle_port_status() unconditionally resumes the root hub when
> hcd->state is HC_STATE_SUSPENDED, regardless of whether a device is
> actually attached to the port.  This can cause system suspend to fail
> with -EBUSY if a spurious port change event arrives on an empty port
> after its bus has entered suspend.
> 
> Fix this by checking PORT_CONNECT in handle_port_status() before
> resuming the root hub.  If no device is actually present on the port,
> the event is spurious and should be silently ignored rather than
> aborting suspend.

Doesn't this change prevent handling device disconnects on suspended hosts?

Thanks
Mathias

Re: [PATCH] usb: xhci: prevent spurious root hub resume during suspend

[Pei Xiao]

在 2026/6/30 16:10, Mathias Nyman 写道:
> On 6/30/26 10:36, Pei Xiao wrote:
>> handle_port_status() unconditionally resumes the root hub when
>> hcd->state is HC_STATE_SUSPENDED, regardless of whether a device is
>> actually attached to the port.  This can cause system suspend to fail
>> with -EBUSY if a spurious port change event arrives on an empty port
>> after its bus has entered suspend.
>>
>> Fix this by checking PORT_CONNECT in handle_port_status() before
>> resuming the root hub.  If no device is actually present on the port,
>> the event is spurious and should be silently ignored rather than
>> aborting suspend.
>
> Doesn't this change prevent handling device disconnects on suspended
> hosts? 
>
*Thank you for your reply. I’m not familiar with xHCI or these registers.
 Can’t we tell the difference between a real unplug and a situation where
 the device is physically still attached but portsc reports a disconnect?

Thanks！
Pei*
> Thanks
> Mathias

Re: [PATCH] usb: xhci: prevent spurious root hub resume during suspend

[Mathias Nyman]

On 6/30/26 11:43, Pei Xiao wrote:
> 
> 
> 在 2026/6/30 16:10, Mathias Nyman 写道:
>> On 6/30/26 10:36, Pei Xiao wrote:
>>> handle_port_status() unconditionally resumes the root hub when
>>> hcd->state is HC_STATE_SUSPENDED, regardless of whether a device is
>>> actually attached to the port.  This can cause system suspend to fail
>>> with -EBUSY if a spurious port change event arrives on an empty port
>>> after its bus has entered suspend.
>>>
>>> Fix this by checking PORT_CONNECT in handle_port_status() before
>>> resuming the root hub.  If no device is actually present on the port,
>>> the event is spurious and should be silently ignored rather than
>>> aborting suspend.
>>
>> Doesn't this change prevent handling device disconnects on suspended
>> hosts?
>>
> *Thank you for your reply. I’m not familiar with xHCI or these registers.
>   Can’t we tell the difference between a real unplug and a situation where
>   the device is physically still attached but portsc reports a disconnect?

Not really. In your case it really looks like a disconnect.

Port Status: 0x202a0
	Disconnected
	Disabled
	Link: Rx Detect
	Powered
	Unknown port speed
	Connect Status Change

Is it only this specific device that disconnects on suspend, or can this be
reproduced with any usb device?

Thanks
Mathias

Re: [PATCH] usb: xhci: prevent spurious root hub resume during suspend

[Pei Xiao]

在 2026/6/30 18:30, Mathias Nyman 写道:
> On 6/30/26 11:43, Pei Xiao wrote:
>>
>>
>> 在 2026/6/30 16:10, Mathias Nyman 写道:
>>> On 6/30/26 10:36, Pei Xiao wrote:
>>>> handle_port_status() unconditionally resumes the root hub when
>>>> hcd->state is HC_STATE_SUSPENDED, regardless of whether a device is
>>>> actually attached to the port.  This can cause system suspend to fail
>>>> with -EBUSY if a spurious port change event arrives on an empty port
>>>> after its bus has entered suspend.
>>>>
>>>> Fix this by checking PORT_CONNECT in handle_port_status() before
>>>> resuming the root hub.  If no device is actually present on the port,
>>>> the event is spurious and should be silently ignored rather than
>>>> aborting suspend.
>>>
>>> Doesn't this change prevent handling device disconnects on suspended
>>> hosts?
>>>
>> *Thank you for your reply. I’m not familiar with xHCI or these
>> registers.
>>   Can’t we tell the difference between a real unplug and a situation
>> where
>>   the device is physically still attached but portsc reports a
>> disconnect?
>
> Not really. In your case it really looks like a disconnect.
>
> Port Status: 0x202a0
>     Disconnected
>     Disabled
>     Link: Rx Detect
>     Powered
>     Unknown port speed
>     Connect Status Change
>
> Is it only this specific device that disconnects on suspend, or can
> this be
> reproduced with any usb device? 
>
It's specific devices—like USB flash drives and external hard drives.
 A USB keyboard doesn't have the problem. And for the device that does 
have the issue, if I plug it into a USB hub and then connect that hub to
the host, 
the issue doesn't happen either.

Thanks
Pei.
> Thanks
> Mathias