Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20180920070940.14773-1-acelan.kao@canonical.com> <Pine.LNX.4.44L0.1809201037540.2141-100000@iolanthe.rowland.org>
In-Reply-To: <Pine.LNX.4.44L0.1809201037540.2141-100000@iolanthe.rowland.org>
From:   AceLan Kao <acelan.kao@canonical.com>
Date:   Thu, 27 Sep 2018 10:28:37 +0800
Message-ID: <CAFv23Q=xUFxwgXZZ_Q6G4CoHx_-Hs5PJRxcrUmcqutV9Ovrg1g@mail.gmail.com>
Subject: Re: [PATCH] usb: core: disable USB2 LPM when suspending
To:     stern@rowland.harvard.edu
Cc:     felipe.balbi@linux.intel.com, Daniel Drake <drake@endlessm.com>,
        joe@perches.com, mathias.nyman@linux.intel.com,
        linux-usb@vger.kernel.org,
        "Linux-Kernel@Vger. Kernel. Org" <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 8BIT
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Alan Stern <stern@rowland.harvard.edu> 於 2018年9月20日 週四 下午10:43寫道：
>
> On Thu, 20 Sep 2018, AceLan Kao wrote:
>
> > We found a S5 current leakage issue on Dell DW1820 WiFi/BT combo card
> > which uses Qualcomm QCA6174 SoC. It also comes with WiFi and BT failure
> > when encountered current leakage issue.
> >    1. Power on, both WiFi and BT work.
> >    2. Power off and found a current leakage issue(consumes ~0.5W)
> >    3. Power on, no WiFi and BT devices can be found in lspci and lsusb.
> >    4. Power off, there is no current leakage issue at S5.
> >    5. continue to 1.
> >
> > From Qualcomm's report:
> > Based on the USB sniffer log, the difference between Linux and Windows
> > is USB LPM setting(no LPM transaction on Windows) which may leads to
> > the voltage leakage on Linux S5 state.
> >
> > After checked the LPM related code and found, when system is going to
> > enter S5, it resumes the USB devices from runtime suspend and enables
> > USB2 LPM, and then it calls usb_dev_poweroff() -> usb_suspend(), and
> > leave USB2 LPM stays enabled.
>
> But usb_suspend() -> usb_suspend_both() -> usb_suspend_device() ->
> generic_suspend() -> usb_port_suspend() ->
> usb_set_usb2_hardware_lpm(udev, 0).  So why does USB2 LPM stay enabled?
Right, after checking the normal case, it calls
usb_set_usb2_hardware_lpm(udev, 0) eventually.
In the buggy machine, it doesn't reach that part and stops somewhere
after resume.
But unfortunately, I can't duplicate the issue after doing some
experiments anymore, so I can't get more info about that.
I'll dig further and get back to you when I have any progress.
>
> > Disable USB2 LPM in usb_suspend() fixes the issue mentioned above,
> > and try 30 times of s2idle, S3 and S5, the USB devices keep working
> > well. Disable USB2 LPM seems do no harm to the system.
> >
> > Signed-off-by: AceLan Kao <acelan.kao@canonical.com>
> > ---
> >  drivers/usb/core/driver.c | 3 +++
> >  1 file changed, 3 insertions(+)
> >
> > diff --git a/drivers/usb/core/driver.c b/drivers/usb/core/driver.c
> > index e76e95f62f76..ac5e60d7104f 100644
> > --- a/drivers/usb/core/driver.c
> > +++ b/drivers/usb/core/driver.c
> > @@ -1463,6 +1463,9 @@ int usb_suspend(struct device *dev, pm_message_t msg)
> >       struct usb_device       *udev = to_usb_device(dev);
> >       int r;
> >
> > +     if (udev->usb2_hw_lpm_enabled == 1)
> > +             usb_set_usb2_hardware_lpm(udev, 0);
>
> At this point the device may still be in runtime suspend.  Is that
> really okay?
>
> Alan Stern
>
> > +
> >       unbind_no_pm_drivers_interfaces(udev);
> >
> >       /* From now on we are sure all drivers support suspend/resume
> >
>