linux-kernel - Re: [PATCH] PCI: qcom: Disable power management for uPD720201 USB3 controller

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Date:   Thu, 18 Jun 2020 18:20:11 -0500
From:   Bjorn Helgaas <helgaas@...nel.org>
To:     Georgi Djakov <georgi.djakov@...aro.org>
Cc:     svarbanov@...sol.com, bjorn.andersson@...aro.org, vkoul@...nel.org,
        sanm@...eaurora.org, mgautam@...eaurora.org, agross@...nel.org,
        bhelgaas@...gle.com, robh@...nel.org, lorenzo.pieralisi@....com,
        linux-arm-msm@...r.kernel.org, linux-pci@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        "Rafael J. Wysocki" <rjw@...ysocki.net>, linux-pm@...r.kernel.org
Subject: Re: [PATCH] PCI: qcom: Disable power management for uPD720201 USB3
 controller

On Tue, Jun 16, 2020 at 04:17:11PM -0500, Bjorn Helgaas wrote:
> On Mon, Jun 15, 2020 at 09:24:13PM +0300, Georgi Djakov wrote:
> > The uPD720201 USB3 host controller (connected to PCIe) on the Dragonboard
> > 845c is often failing during suspend and resume. The following messages
> > are seen over the console:
> > 
> >   PM: suspend entry (s2idle)
> >   Filesystems sync: 0.000 seconds
> >   Freezing user space processes ... (elapsed 0.001 seconds) done.
> >   OOM killer disabled.
> >   Freezing remaining freezable tasks ... (elapsed 0.001 seconds) done.
> >   printk: Suspending console(s) (use no_console_suspend to debug)
> >   dwc3-qcom a8f8800.usb: HS-PHY not in L2
> >   dwc3-qcom a6f8800.usb: HS-PHY not in L2
> >   xhci_hcd 0000:01:00.0: can't change power state from D3hot to D0 (config
> >   space inaccessible)
> >   xhci_hcd 0000:01:00.0: can't change power state from D3hot to D0 (config
> >   space inaccessible)
> >   xhci_hcd 0000:01:00.0: Controller not ready at resume -19
> >   xhci_hcd 0000:01:00.0: PCI post-resume error -19!
> >   xhci_hcd 0000:01:00.0: HC died; cleaning up
> > 
> > Then the USB devices are not functional anymore. Let's disable the PM of
> > the controller for now, as this will at least keep USB devices working
> > even after suspend and resume.

Georgi, can you collect the complete dmesg log and "sudo lspci
-vvxxxx" output somewhere?  A new report at bugzilla.kernel.org would
be a good spot.

Maybe we're missing a delay here.  The "config space inaccessible"
message means we read 0xffff from PCI_PM_CTRL, which probably means
the device is still in D3cold.  If it were in any other power state,
PCI_PM_CTRL should be readable, and 0xffff is not a valid value.

Could you also insert a dump_stack() right after we print that "config
space inaccessible" message?  I don't know enough about power
management to understand why we see that message twice.

> This seems like we're just covering up a deeper problem here.  I think
> it would be better to fix the underlying problem.
> 
> The quirk you're adding is specific to the Renesas 0x0014 device.  Is
> there some reason to think the problem is specific to that device, or
> might other devices have the same problem?
> 
> Maybe we're missing something in pcie-qcom.c?  Is there any
> suspend/resume support required in that driver?  It doesn't look like
> it has anything except that it calls pm_runtime_enable().
> 
> > Signed-off-by: Georgi Djakov <georgi.djakov@...aro.org>
> > ---
> >  drivers/pci/controller/dwc/pcie-qcom.c | 8 ++++++++
> >  1 file changed, 8 insertions(+)
> > 
> > diff --git a/drivers/pci/controller/dwc/pcie-qcom.c b/drivers/pci/controller/dwc/pcie-qcom.c
> > index 138e1a2d21cc..c1f502682a19 100644
> > --- a/drivers/pci/controller/dwc/pcie-qcom.c
> > +++ b/drivers/pci/controller/dwc/pcie-qcom.c
> > @@ -1439,6 +1439,13 @@ static void qcom_fixup_class(struct pci_dev *dev)
> >  {
> >  	dev->class = PCI_CLASS_BRIDGE_PCI << 8;
> >  }
> > +
> > +static void qcom_fixup_nopm(struct pci_dev *dev)
> > +{
> > +	dev->pm_cap = 0;
> > +	dev_info(&dev->dev, "Disabling PCI power management\n");
> > +}
> > +
> >  DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_QCOM, 0x0101, qcom_fixup_class);
> >  DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_QCOM, 0x0104, qcom_fixup_class);
> >  DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_QCOM, 0x0106, qcom_fixup_class);
> > @@ -1446,6 +1453,7 @@ DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_QCOM, 0x0107, qcom_fixup_class);
> >  DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_QCOM, 0x0302, qcom_fixup_class);
> >  DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_QCOM, 0x1000, qcom_fixup_class);
> >  DECLARE_PCI_FIXUP_EARLY(PCI_VENDOR_ID_QCOM, 0x1001, qcom_fixup_class);
> > +DECLARE_PCI_FIXUP_FINAL(PCI_VENDOR_ID_RENESAS, 0x0014, qcom_fixup_nopm);
> 
> The convention is that DECLARE_PCI_FIXUP_*() comes immediately after
> the quirk function itself, so the whole patch would be a single diff
> hunk.  See drivers/pci/quirks.c for many examples.
> 
> >  static struct platform_driver qcom_pcie_driver = {
> >  	.probe = qcom_pcie_probe,