2018-02-19 19:07:22

by Kai-Heng Feng

[permalink] [raw]
Subject: [PATCH v2] xhci: Fix front USB ports on ASUS PRIME B350M-A

When a USB device gets plugged on ASUS PRIME B350M-A's front ports, the
xHC stops working:
[ 549.114587] xhci_hcd 0000:02:00.0: WARN: xHC CMD_RUN timeout
[ 549.114608] suspend_common(): xhci_pci_suspend+0x0/0xc0 returns -110
[ 549.114638] xhci_hcd 0000:02:00.0: can't suspend (hcd_pci_runtime_suspend returned -110)

Delay before running xHC command CMD_RUN can workaround the issue.

Use a new quirk to make the delay only targets to the affected xHC.

Signed-off-by: Kai-Heng Feng <[email protected]>
---
v2: Instead of doing xHC reset and disabling D3cold, a simple delay can
workaround the issue. Now both high-speed and super-speed devices work
fine with the v2 quirk.

drivers/usb/host/xhci-pci.c | 3 +++
drivers/usb/host/xhci.c | 3 +++
drivers/usb/host/xhci.h | 1 +
3 files changed, 7 insertions(+)

diff --git a/drivers/usb/host/xhci-pci.c b/drivers/usb/host/xhci-pci.c
index 6c79037876db..9a820b3cae56 100644
--- a/drivers/usb/host/xhci-pci.c
+++ b/drivers/usb/host/xhci-pci.c
@@ -122,6 +122,9 @@ static void xhci_pci_quirks(struct device *dev, struct xhci_hcd *xhci)
if (pdev->vendor == PCI_VENDOR_ID_AMD && usb_amd_find_chipset_info())
xhci->quirks |= XHCI_AMD_PLL_FIX;

+ if (pdev->vendor == PCI_VENDOR_ID_AMD && pdev->device == 0x43bb)
+ xhci->quirks |= XHCI_SUSPEND_DELAY;
+
if (pdev->vendor == PCI_VENDOR_ID_AMD)
xhci->quirks |= XHCI_TRUST_TX_LENGTH;

diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c
index 1eeb3396300f..d554bbd694d3 100644
--- a/drivers/usb/host/xhci.c
+++ b/drivers/usb/host/xhci.c
@@ -878,6 +878,9 @@ int xhci_suspend(struct xhci_hcd *xhci, bool do_wakeup)
clear_bit(HCD_FLAG_POLL_RH, &xhci->shared_hcd->flags);
del_timer_sync(&xhci->shared_hcd->rh_timer);

+ if (xhci->quirks & XHCI_SUSPEND_DELAY)
+ usleep_range(1000, 1500);
+
spin_lock_irq(&xhci->lock);
clear_bit(HCD_FLAG_HW_ACCESSIBLE, &hcd->flags);
clear_bit(HCD_FLAG_HW_ACCESSIBLE, &xhci->shared_hcd->flags);
diff --git a/drivers/usb/host/xhci.h b/drivers/usb/host/xhci.h
index 96099a245c69..6f1f52f3cb40 100644
--- a/drivers/usb/host/xhci.h
+++ b/drivers/usb/host/xhci.h
@@ -1825,6 +1825,7 @@ struct xhci_hcd {
/* Reserved. It was XHCI_U2_DISABLE_WAKE */
#define XHCI_ASMEDIA_MODIFY_FLOWCONTROL (1 << 28)
#define XHCI_HW_LPM_DISABLE (1 << 29)
+#define XHCI_SUSPEND_DELAY (1 << 30)

unsigned int num_active_eps;
unsigned int limit_active_eps;
--
2.15.1



2018-02-20 08:22:30

by Mathias Nyman

[permalink] [raw]
Subject: Re: [PATCH v2] xhci: Fix front USB ports on ASUS PRIME B350M-A

Hi

On 19.02.2018 21:06, Kai-Heng Feng wrote:
> When a USB device gets plugged on ASUS PRIME B350M-A's front ports, the
> xHC stops working:
> [ 549.114587] xhci_hcd 0000:02:00.0: WARN: xHC CMD_RUN timeout
> [ 549.114608] suspend_common(): xhci_pci_suspend+0x0/0xc0 returns -110
> [ 549.114638] xhci_hcd 0000:02:00.0: can't suspend (hcd_pci_runtime_suspend returned -110)

Is the xhci runtime suspend something that you expected here?
I mean was the plugged device first enumerated normally, driver bound, inactive and runtime suspended
first?

>
> Delay before running xHC command CMD_RUN can workaround the issue.
>
> Use a new quirk to make the delay only targets to the affected xHC.

Is this some known issue with this chip, or something figured out by trial and error?

>
> Signed-off-by: Kai-Heng Feng <[email protected]>

Thanks, otherwise workaround seems fine, just curious about the two details above

-Mathias


2018-02-20 11:14:40

by Kai-Heng Feng

[permalink] [raw]
Subject: Re: [PATCH v2] xhci: Fix front USB ports on ASUS PRIME B350M-A



> On 20 Feb 2018, at 4:24 PM, Mathias Nyman <[email protected]> wrote:
>
> Hi
>
> On 19.02.2018 21:06, Kai-Heng Feng wrote:
>> When a USB device gets plugged on ASUS PRIME B350M-A's front ports, the
>> xHC stops working:
>> [ 549.114587] xhci_hcd 0000:02:00.0: WARN: xHC CMD_RUN timeout
>> [ 549.114608] suspend_common(): xhci_pci_suspend+0x0/0xc0 returns -110
>> [ 549.114638] xhci_hcd 0000:02:00.0: can't suspend (hcd_pci_runtime_suspend returned -110)
>
> Is the xhci runtime suspend something that you expected here?
> I mean was the plugged device first enumerated normally, driver bound, inactive and runtime suspended
> first?

I guess this is how it behaves, here’s the log after a device gets plugged:

[ 7535.716578] xhci_hcd 0000:02:00.0: // Setting command ring address to 0xff86c001
[ 7535.720107] xhci_hcd 0000:02:00.0: xhci_resume: starting port polling.
[ 7535.720114] xhci_hcd 0000:02:00.0: xhci_hub_status_data: stopping port polling.
[ 7535.720134] xhci_hcd 0000:02:00.0: xhci_suspend: stopping port polling.
[ 7535.720190] xhci_hcd 0000:02:00.0: Port Status Change Event for port 8
[ 7535.720194] xhci_hcd 0000:02:00.0: resume root hub
[ 7535.720201] xhci_hcd 0000:02:00.0: handle_port_status: starting port polling.
[ 7535.721675] xhci_hcd 0000:02:00.0: // Setting command ring address to 0xff86c001
[ 7535.722106] xhci_hcd 0000:02:00.0: // Setting command ring address to 0xff86c001
[ 7535.725538] xhci_hcd 0000:02:00.0: xhci_resume: starting port polling.

So the xHC resumes and suspends, then resume again.

[ 7535.725543] xhci_hcd 0000:02:00.0: xhci_hub_status_data: stopping port polling.
[ 7535.725587] xhci_hcd 0000:02:00.0: get port status, actual port 0 status = 0x2a0
[ 7535.725590] xhci_hcd 0000:02:00.0: Get port status returned 0x100
[ 7535.725606] xhci_hcd 0000:02:00.0: get port status, actual port 1 status = 0x2a0
[ 7535.725608] xhci_hcd 0000:02:00.0: Get port status returned 0x100
[ 7535.725616] xhci_hcd 0000:02:00.0: get port status, actual port 2 status = 0x2a0
[ 7535.725618] xhci_hcd 0000:02:00.0: Get port status returned 0x100
[ 7535.725625] xhci_hcd 0000:02:00.0: get port status, actual port 3 status = 0x202e1
[ 7535.725627] xhci_hcd 0000:02:00.0: Get port status returned 0x10101
[ 7535.725632] xhci_hcd 0000:02:00.0: Port Status Change Event for port 8
[ 7535.725636] xhci_hcd 0000:02:00.0: handle_port_status: starting port polling.
[ 7535.725656] xhci_hcd 0000:02:00.0: clear port connect change, actual port 3 status = 0x2e1
[ 7535.725663] xhci_hcd 0000:02:00.0: get port status, actual port 4 status = 0x2a0
[ 7535.725665] xhci_hcd 0000:02:00.0: Get port status returned 0x100
[ 7535.725672] xhci_hcd 0000:02:00.0: get port status, actual port 5 status = 0x2a0
[ 7535.725674] xhci_hcd 0000:02:00.0: Get port status returned 0x100
[ 7535.725680] xhci_hcd 0000:02:00.0: get port status, actual port 6 status = 0x2a0
[ 7535.725682] xhci_hcd 0000:02:00.0: Get port status returned 0x100
[ 7535.725689] xhci_hcd 0000:02:00.0: get port status, actual port 7 status = 0x2a0
[ 7535.725691] xhci_hcd 0000:02:00.0: Get port status returned 0x100
[ 7535.725698] xhci_hcd 0000:02:00.0: get port status, actual port 8 status = 0x2a0
[ 7535.725699] xhci_hcd 0000:02:00.0: Get port status returned 0x100
[ 7535.725706] xhci_hcd 0000:02:00.0: get port status, actual port 9 status = 0x2a0
[ 7535.725708] xhci_hcd 0000:02:00.0: Get port status returned 0x100

[ 7535.736165] xhci_hcd 0000:04:00.0: // Setting command ring address to 0xfff58001
[ 7535.740033] xhci_hcd 0000:04:00.0: xhci_resume: starting port polling.
[ 7535.740040] xhci_hcd 0000:04:00.0: xhci_hub_status_data: stopping port polling.
[ 7535.740047] xhci_hcd 0000:04:00.0: xhci_hub_status_data: stopping port polling.
[ 7535.740133] xhci_hcd 0000:04:00.0: xhci_suspend: stopping port polling.
[ 7535.740161] xhci_hcd 0000:04:00.0: // Setting command ring address to 0xfff58001

This is the “other” xHC that I mentioned in patch v1. It provides USB 3.1 capabilities to front USB port.

[ 7535.776003] xhci_hcd 0000:02:00.0: xhci_hub_status_data: stopping port polling.
[ 7535.832028] xhci_hcd 0000:02:00.0: get port status, actual port 3 status = 0x2e1
[ 7535.832031] xhci_hcd 0000:02:00.0: Get port status returned 0x101
[ 7535.832051] xhci_hcd 0000:02:00.0: // Ding dong!
[ 7535.832150] xhci_hcd 0000:02:00.0: Slot 3 output ctx = 0xff125000 (dma)
[ 7535.832153] xhci_hcd 0000:02:00.0: Slot 3 input ctx = 0xff16c000 (dma)
[ 7535.832160] xhci_hcd 0000:02:00.0: Set slot id 3 dcbaa entry 000000000bd2158a to 0xff125000
[ 7535.832189] xhci_hcd 0000:02:00.0: set port reset, actual port 3 status = 0x2e1
[ 7535.835982] xhci_hcd 0000:02:00.0: xhci_hub_status_data: stopping port polling.
[ 7535.892049] xhci_hcd 0000:02:00.0: Port Status Change Event for port 8
[ 7535.892056] xhci_hcd 0000:02:00.0: handle_port_status: starting port polling.
[ 7535.900058] xhci_hcd 0000:02:00.0: get port status, actual port 3 status = 0x200e03
[ 7535.900061] xhci_hcd 0000:02:00.0: Get port status returned 0x100503
[ 7535.900079] xhci_hcd 0000:02:00.0: clear port reset change, actual port 3 status = 0xe03
[ 7535.959990] usb 1-4: new high-speed USB device number 4 using xhci_hcd

>
>> Delay before running xHC command CMD_RUN can workaround the issue.
>> Use a new quirk to make the delay only targets to the affected xHC.
>
> Is this some known issue with this chip, or something figured out by trial and error?

By trial and error.
This might be a more pervasive issue, but I don’t see other bug reports though.

>
>> Signed-off-by: Kai-Heng Feng <[email protected]>
>
> Thanks, otherwise workaround seems fine, just curious about the two details above

Thank you.

Kai-Heng

>
> -Mathias
>


2018-03-07 06:11:11

by Kai-Heng Feng

[permalink] [raw]
Subject: Re: [PATCH v2] xhci: Fix front USB ports on ASUS PRIME B350M-A

Hi Matthias,

Do you have any concern about this patch?

Hopefully this can get merged for v4.16…

Kai-Heng

2018-03-08 13:12:39

by Mathias Nyman

[permalink] [raw]
Subject: Re: [PATCH v2] xhci: Fix front USB ports on ASUS PRIME B350M-A

On 07.03.2018 08:09, Kai Heng Feng wrote:
> Hi Matthias,
>
> Do you have any concern about this patch?
>

Adding patch and sending forward

Thanks
-Mathias