xhci: fix runtime PM imbalance in USB2 resume
Subsystems (2)
PATCH 1/1 xhci: fix runtime PM imbalance in USB2 resume
USB2 resume starts with usb_hcd_start_port_resume() in port status change handling for RESUME link state. usb_hcd_end_port_resume() call is needed to keep runtime PM balance.
--- drivers/usb/host/xhci-hub.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/usb/host/xhci-hub.c +2 -0
@@ -1090,6 +1090,8 @@ static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status,if (link_state ==XDEV_U0 ) { bus_state->resume_done[portnum] =0 ;clear_bit (portnum, &bus_state->resuming_ports); +usb_hcd_end_port_resume (&port->rhub->hcd->self, + portnum);if (bus_state->suspended_ports & (1 << portnum)) { bus_state->suspended_ports &= ~(1 << portnum); bus_state->port_c_suspend |=1 << portnum;
Signature · 1 line
2.17.1
Greg Kroah-Hartman
maintainerQuoted 8 lines · Henry Lin
USB2 resume starts with
usb_hcd_start_port_resume()in port status change handling for RESUME link state.usb_hcd_end_port_resume()call is needed to keep runtime PM balance.Signed-off-by Henry Lin <[email protected]>--- drivers/usb/host/xhci-hub.c | 2 ++ 1 file changed, 2 insertions(+)
What commit does this fix?
It fixes side effect from below change.
commit a231ec41e6f6433daf4c693f169f6c5cfda8cb9d
Author: Mathias Nyman <[email protected]> Date: Fri Dec 7 16:19:35 2018 +0200
xhci: refactor U0 link state handling in get_port_status
Move U0 link state handing to USB3 and USB2 specific functions
Note that
bus_state->resuming_ports:
bus_state->resume_done[]:
are only used for USB2, and don't need to cleared for USB3 ports
No functional changes
--nvpublic
________________________________________
From: Greg Kroah-Hartman <[email protected]>
Sent: Friday, February 25, 2022 2:23 PM
To: Henry Lin
Cc: [email protected]; [email protected]; [email protected]
Subject: Re: [PATCH] xhci: fix runtime PM imbalance in USB2 resume
External email: Use caution opening links or attachments
Quoted 8 lines · Henry Lin
USB2 resume starts with
usb_hcd_start_port_resume()in port status change handling for RESUME link state.usb_hcd_end_port_resume()call is needed to keep runtime PM balance.Signed-off-by Henry Lin <[email protected]>--- drivers/usb/host/xhci-hub.c | 2 ++ 1 file changed, 2 insertions(+)
What commit does this fix?
Greg Kroah-Hartman
maintainerA: http://en.wikipedia.org/wiki/Top_post
Q: Were do I find info about this thing called top-posting?
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing in e-mail?
A: No.
Q: Should I include quotations after my reply?
http://daringfireball.net/2007/07/on_top
Quoted 15 lines · Henry Lin
It fixes side effect from below change.
commit
a231ec41e6f6433daf4c693f169f6c5cfda8cb9dAuthor: Mathias Nyman <[email protected]> Date: Fri Dec 7 16:19:35 2018 +0200xhci: refactor U0 link state handling in
get_port_statusMove U0 link state handing to USB3 and USB2 specific functions
Note that
bus_state->resuming_ports:
bus_state->resume_done[]:
are only used for USB2, and don't need to cleared for USB3 ports
Quoted 4 lines
No functional changes
Signed-off-by Mathias Nyman <[email protected]>Signed-off-by Greg Kroah-Hartman <[email protected]>
Great, please add a "Fixes:" tag to the patch when resubmitting it.
And the "no functional changes" seems not to have been true, right? Does this need to go into the stable kernels?
thanks,
greg k-h
Quoted 1 line
Great, please add a "Fixes:" tag to the patch when resubmitting it.
I will resubmit soon.
Quoted 1 line
And the "no functional changes" seems not to have been true, right?
Yes.
Quoted 1 line
Does this need to go into the stable kernels?
Yes, this need to go into stable kernels.
Thanks,
Henry
PATCH v2 1/1 xhci: fix runtime PM imbalance in USB2 resume
In reply to Henry LinUSB2 resume starts with usb_hcd_start_port_resume() in port status change handling for RESUME link state. usb_hcd_end_port_resume() call is needed to keep runtime PM balance.
--- drivers/usb/host/xhci-hub.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/usb/host/xhci-hub.c +2 -0
@@ -1090,6 +1090,8 @@ static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status,if (link_state ==XDEV_U0 ) { bus_state->resume_done[portnum] =0 ;clear_bit (portnum, &bus_state->resuming_ports); +usb_hcd_end_port_resume (&port->rhub->hcd->self, + portnum);if (bus_state->suspended_ports & (1 << portnum)) { bus_state->suspended_ports &= ~(1 << portnum); bus_state->port_c_suspend |=1 << portnum;
Signature · 1 line
2.17.1
Greg KH
maintainerQuoted 26 lines · Henry Lin
USB2 resume starts with
usb_hcd_start_port_resume()in port status change handling for RESUME link state.usb_hcd_end_port_resume()call is needed to keep runtime PM balance.Fixes a231ec41e6f6 ("xhci: refactor U0 link state handling in get_port_status")Signed-off-by Henry Lin <[email protected]>--- drivers/usb/host/xhci-hub.c | 2 ++ 1 file changed, 2 insertions(+)drivers/usb/host/xhci-hub.c +2
-0@@ -1090,6 +1090,8 @@ static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status,if (link_state ==XDEV_U0 ) { bus_state->resume_done[portnum] =0 ;clear_bit (portnum, &bus_state->resuming_ports); +usb_hcd_end_port_resume (&port->rhub->hcd->self, + portnum);if (bus_state->suspended_ports & (1 << portnum)) { bus_state->suspended_ports &= ~(1 << portnum); bus_state->port_c_suspend |=1 << portnum;Signature · 1 line
2.17.1
Hi,
This is the friendly patch-bot of Greg Kroah-Hartman. You have sent him a patch that has triggered this response. He used to manually respond to these common problems, but in order to save his sanity (he kept writing the same thing over and over, yet to different people), I was created. Hopefully you will not take offence and will fix the problem in your patch and resubmit it so that it can be accepted into the Linux kernel tree.
You are receiving this message because of the following common error(s) as indicated below:
- This looks like a new version of a previously submitted patch, but you did not list below the --- line any changes from the previous version. Please read the section entitled "The canonical patch format" in the kernel file, Documentation/SubmittingPatches for what needs to be done here to properly describe this.
If you wish to discuss this problem further, or you have questions about how to resolve this issue, please feel free to respond to this email and Greg will reply once he has dug out from the pending patches received from other developers.
thanks,
greg k-h's patch email bot
USB2 resume starts with usb_hcd_start_port_resume() in port status change handling for RESUME link state. usb_hcd_end_port_resume() call is needed to keep runtime PM balance.
---
V1 -> V2: Added Fixes tag in changelog
drivers/usb/host/xhci-hub.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/usb/host/xhci-hub.c +2 -0
@@ -1090,6 +1090,8 @@ static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status,if (link_state ==XDEV_U0 ) { bus_state->resume_done[portnum] =0 ;clear_bit (portnum, &bus_state->resuming_ports); +usb_hcd_end_port_resume (&port->rhub->hcd->self, + portnum);if (bus_state->suspended_ports & (1 << portnum)) { bus_state->suspended_ports &= ~(1 << portnum); bus_state->port_c_suspend |=1 << portnum;
--
2.17.1
Review
Quoted 2 lines
(quoted diff) +2
-0+ usb_hcd_end_port_resume(&port->rhub->hcd->self, + portnum);
Tab indent is broken in this version. Will resend this.
________________________________________
From: Henry Lin <[email protected]>
Sent: Sunday, February 27, 2022 12:06 AM
To: Greg KH
Cc: Mathias Nyman; [email protected]; [email protected]
Subject: Re: [PATCH v2] xhci: fix runtime PM imbalance in USB2 resume
USB2 resume starts with usb_hcd_start_port_resume() in port status change handling for RESUME link state. usb_hcd_end_port_resume() call is needed to keep runtime PM balance.
---
V1 -> V2: Added Fixes tag in changelog
drivers/usb/host/xhci-hub.c | 2 ++ 1 file changed, 2 insertions(+)
drivers/usb/host/xhci-hub.c +2 -0
@@ -1090,6 +1090,8 @@ static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status,if (link_state ==XDEV_U0 ) { bus_state->resume_done[portnum] =0 ;clear_bit (portnum, &bus_state->resuming_ports); +usb_hcd_end_port_resume (&port->rhub->hcd->self, + portnum);if (bus_state->suspended_ports & (1 << portnum)) { bus_state->suspended_ports &= ~(1 << portnum); bus_state->port_c_suspend |=1 << portnum;
--
2.17.1
Quoted 3 lines · Henry Lin
USB2 resume starts with
usb_hcd_start_port_resume()in port status change handling for RESUME link state.usb_hcd_end_port_resume()call is needed to keep runtime PM balance.
For normal usb2 port resume the usb_hcd_end_port_resume() is called when resume has been signaled for long enough in xhci_handle_usb2_port_link_resume().
This is also where driver directs the port to go from Resume state to U0. Port can't do this without driver directing it.
If there's a failure during resume signaling (disconnect, reset, error) then stale resume variables are detected in xhci_get_port_status() and usb_hcd_end_port_resume() is called.
I do now see a231ec41e6f6 ("xhci: refactor U0 link state handling in get_port_status") does change order of checking and clearing stale resume variables, but this should only happen if the first port state we read is a fully enabled functional U0 state after a failed resume.
Could you expand a bit how this was detected?
Quoted 12 lines
Fixes a231ec41e6f6 ("xhci: refactor U0 link state handling in get_port_status")Signed-off-by Henry Lin <[email protected]>--- drivers/usb/host/xhci-hub.c | 2 ++ 1 file changed, 2 insertions(+)drivers/usb/host/xhci-hub.c +0
-0
drivers/usb/host/xhci-hub.c static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status, · line 1094
Quoted 5 lines
drivers/usb/host/xhci-hub.c +2
-0@@ -1090,6 +1090,8 @@ static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status,if (link_state ==XDEV_U0 ) { bus_state->resume_done[portnum] =0 ;clear_bit (portnum, &bus_state->resuming_ports); +usb_hcd_end_port_resume (&port->rhub->hcd->self, + portnum);
This will call usb_hcd_end_port_resume() every time port is in normal enabled U0 state even if resume was never signaled, or port suspended.
-Mathias
Quoted 14 lines
Quoted 3 lines
USB2 resume starts with
usb_hcd_start_port_resume()in port status change handling for RESUME link state.usb_hcd_end_port_resume()call is needed to keep runtime PM balance.For normal usb2 port resume the
usb_hcd_end_port_resume()is called when resume has been signaled for long enough inxhci_handle_usb2_port_link_resume().This is also where driver directs the port to go from Resume state to U0. Port can't do this without driver directing it.
If there's a failure during resume signaling (disconnect, reset, error) then stale resume variables are detected in
xhci_get_port_status()andusb_hcd_end_port_resume()is called.
Quoted 6 lines
I do now see
a231ec41e6f6("xhci: refactor U0 link state handling inget_port_status") does change order of checking and clearing stale resume variables, but this should only happen if the first port state we read is a fully enabled functional U0 state after a failed resume.Could you expand a bit how this was detected?
We observed the racing issue when usb2 device-initiated resume occurs in system resume. If usb2 host-initiated resume for system resume directs U0 before xhci_get_usb2_port_status() see RESUME state, xhci_get_usb2_port_status() will not finish resume process in xhci_handle_usb2_port_link_resume(). Its scenario is as follows:
1. System resume starts. All driver system resume callbacks get called in order. XHCI controller is resumed by xhci_resume().
2. USB2 root hub is resuming. hcd_bus_resume() is being executed.
3. Before xhci_bus_resume() is finished, XHCI driver receives a port status change event for an USB2 port with RESUME link state in xhci_irq(). XHCI driver starts the process to resume HS port for device-initiated resume.
4. In xhci_bus_resume(), host-initiated resume (direct U0) is performed on the same port that is
resuming in step 3 in below loop:
if (bus_state->bus_suspended) {
spin_unlock_irqrestore(&xhci->lock, flags);
msleep(USB_RESUME_TIMEOUT);
spin_lock_irqsave(&xhci->lock, flags);
}
for_each_set_bit(port_index, &bus_state->bus_suspended,
BITS_PER_LONG) {
/* Clear PLC to poll it later for U0 transition */
xhci_test_and_clear_bit(xhci, ports[port_index],
PORT_PLC);
xhci_set_link_state(xhci, ports[port_index], XDEV_U0);
}5. Then, link state of the resuming port is observed as U0 in following xhci_get_usb2_port_status(). xhci_handle_usb2_port_link_resume() has no chance to get called on the resuming port.
Thanks,
Henry
Quoted 46 lines · Henry Lin
Quoted 14 lines
Quoted 3 lines
USB2 resume starts with
usb_hcd_start_port_resume()in port status change handling for RESUME link state.usb_hcd_end_port_resume()call is needed to keep runtime PM balance.For normal usb2 port resume the
usb_hcd_end_port_resume()is called when resume has been signaled for long enough inxhci_handle_usb2_port_link_resume().This is also where driver directs the port to go from Resume state to U0. Port can't do this without driver directing it.
If there's a failure during resume signaling (disconnect, reset, error) then stale resume variables are detected in
xhci_get_port_status()andusb_hcd_end_port_resume()is called.Quoted 6 lines
I do now see
a231ec41e6f6("xhci: refactor U0 link state handling inget_port_status") does change order of checking and clearing stale resume variables, but this should only happen if the first port state we read is a fully enabled functional U0 state after a failed resume.Could you expand a bit how this was detected?
We observed the racing issue when usb2 device-initiated resume occurs in system resume. If usb2 host-initiated resume for system resume directs U0 before
xhci_get_usb2_port_status()see RESUME state,xhci_get_usb2_port_status()will not finish resume process inxhci_handle_usb2_port_link_resume(). Its scenario is as follows:1. System resume starts. All driver system resume callbacks get called in order. XHCI controller is resumed by
xhci_resume().
2. USB2 root hub is resuming.hcd_bus_resume()is being executed.
3. Beforexhci_bus_resume()is finished, XHCI driver receives a port status change event for an USB2 port with RESUME link state inxhci_irq(). XHCI driver starts the process to resume HS port for device-initiated resume.
4. Inxhci_bus_resume(), host-initiated resume (direct U0) is performed on the same port that isresuming in step 3 in below loop: if (bus_state->bus_suspended) { spin_unlock_irqrestore(&xhci->lock, flags); msleep(USB_RESUME_TIMEOUT); spin_lock_irqsave(&xhci->lock, flags); } for_each_set_bit(port_index, &bus_state->bus_suspended, BITS_PER_LONG) { /* Clear PLC to poll it later for U0 transition */ xhci_test_and_clear_bit(xhci, ports[port_index], PORT_PLC); xhci_set_link_state(xhci, ports[port_index], XDEV_U0); }
Quoted 4 lines
5. Then, link state of the resuming port is observed as U0 in following
xhci_get_usb2_port_status().xhci_handle_usb2_port_link_resume()has no chance to get called on the resuming port.
True, thanks for the explanation.
If there's a race between system resume and device-initiated resume, and port is resumed in xhci_bus_resume() then yes I see how this could happen.
Maybe only call usb_hcd_end_port_resume() if xhci_irq() detected the device-initiated resume. i.e. set a value to resume_done[portnum] and called usb_hcd_start_port_resume() something like:
@@ -1088,6 +1088,8 @@ static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status,
if (link_state == XDEV_U2)
*status |= USB_PORT_STAT_L1;
if (link_state == XDEV_U0) {
+ if (bus_state->resume_done[portnum])
+ usb_hcd_end_port_resume(&port->rhub->hcd->self, portnum);
bus_state->resume_done[portnum] = 0;
clear_bit(portnum, &bus_state->resuming_ports);Also xhci_bus_resume() only resumes ports that were forcefully suspended to U3 in xhci_bus_suspend(). Could be worth checking why that device wasn't already in U3 when suspend reached xhci_bus_suspend().
Thanks
Mathias
Review static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status, · line 1094
Quoted 12 lines
Maybe only call
usb_hcd_end_port_resume()ifxhci_irq()detected the device-initiated resume. i.e. set a value to resume_done[portnum] and calledusb_hcd_start_port_resume()something like:(quoted diff) +2
-0@@ -1088,6 +1088,8 @@ static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status, if (link_state == XDEV_U2) *status |= USB_PORT_STAT_L1; if (link_state == XDEV_U0) { + if (bus_state->resume_done[portnum]) + usb_hcd_end_port_resume(&port->rhub->hcd->self, portnum); bus_state->resume_done[portnum] = 0; clear_bit(portnum, &bus_state->resuming_ports);
This looks good. I can submit a new version based on this.
Quoted 2 lines
Also
xhci_bus_resume()only resumes ports that were forcefully suspended to U3 inxhci_bus_suspend(). Could be worth checking why that device wasn't already in U3 when suspend reachedxhci_bus_suspend().
It happens when we trigger USB device-initiated resume to bring system out of system suspend state. For example, we can connect an USB2 external hub on root port and put system into suspend state. The USB2 external hub (the port it connects to) is in U3 after xhci_bus_suspend(). Once we connect a USB device to downstream port of the USB2 external hub, the hub will trigger device initiated resume to wake up the system. During system resume, XHCI controller will report the USB2 external hub is in RESUME state.
Thanks,
Henry
PATCH v3 1/1 xhci: fix runtime PM imbalance in USB2 resume
In reply to Mathias NymanA race between system resume and device-initiated resume may result in runtime PM imbalance on USB2 root hub. If a device-initiated resume starts and system resume xhci_bus_resume() directs U0 before hub driver sees the resuming device in RESUME state, device-initiated resume will not be finished in xhci_handle_usb2_port_link_resume(). In this case, usb_hcd_end_port_resume() call is missing.
This changes calls usb_hcd_end_port_resume() if resuming device reaches U0 to keep runtime PM balance.
---
Changes in v2:
- Add Fixes tag in change log
Changes in v3:
- Revise changelog
- Only call usb_hcd_end_port_resume() if xhci_irq() detected the device-initiated resume
[v2] https://lore.kernel.org/all/[email protected]/
[v1] https://lore.kernel.org/all/[email protected]/
drivers/usb/host/xhci-hub.c | 3 +++ 1 file changed, 3 insertions(+)
drivers/usb/host/xhci-hub.c +3 -0
@@ -1088,6 +1088,9 @@ static void xhci_get_usb2_port_status(struct xhci_port *port, u32 *status,if (link_state ==XDEV_U2 ) *status |=USB_PORT_STAT_L1 ;if (link_state ==XDEV_U0 ) { +if (bus_state->resume_done[portnum]) +usb_hcd_end_port_resume (&port->rhub->hcd->self, + portnum); bus_state->resume_done[portnum] =0 ;clear_bit (portnum, &bus_state->resuming_ports);if (bus_state->suspended_ports & (1 << portnum)) {
Signature · 1 line
2.17.1
Quoted 7 lines · Henry Lin
A race between system resume and device-initiated resume may result in runtime PM imbalance on USB2 root hub. If a device-initiated resume starts and system resume
xhci_bus_resume()directs U0 before hub driver sees the resuming device in RESUME state, device-initiated resume will not be finished inxhci_handle_usb2_port_link_resume(). In this case,usb_hcd_end_port_resume()call is missing.
Quoted 6 lines
This changes calls
usb_hcd_end_port_resume()if resuming device reaches U0 to keep runtime PM balance.Fixes a231ec41e6f6 ("xhci: refactor U0 link state handling in get_port_status")Signed-off-by Henry Lin <[email protected]>---
Thanks, adding to queue
-Mathias