| Message ID | 20200219173727.9882-1-martin.kepplinger@puri.sm (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | usb: host: xhci-plat: add XHCI_MISSING_CAS quirk | expand |
> -----Original Message----- > From: Martin Kepplinger <martin.kepplinger@puri.sm> > Sent: 2020年2月20日 1:37 > To: Jun Li <jun.li@nxp.com>; Peter Chen <peter.chen@nxp.com>; > mathias.nyman@intel.com > Cc: linux-usb@vger.kernel.org; dl-linux-imx <linux-imx@nxp.com>; Anson Huang > <anson.huang@nxp.com>; shawnguo@kernel.org; kernel@pengutronix.de > Subject: [PATCH] usb: host: xhci-plat: add XHCI_MISSING_CAS quirk > > From: Li Jun <jun.li@nxp.com> > > i.MX8MQ USB3 host needs XHCI_MISSING_CAS quirk to warm reset the port to enum the > USB3 device plugged in while system sleep, as the port state is stuck in polling > mode after resume. > > Signed-off-by: Li Jun <jun.li@nxp.com> > Acked-by: Peter Chen <peter.chen@nxp.com> > --- > > Hi, > > Because resume from S3 suspend is broken for me on imx8mq, I stumbled upon this > patch in NXP's linux tree. (Please note that I'm not the author and I've not yet > put my SoB tag under it). This is just a > question: > > This patch (and the docs) clearly is missing in mainline Linux because the imx8mq > devicetree description includes it (which does nothing now). > > I've tested this and this particular addition doesn't fix my problem: > > [ 84.257538] imx8mq-usb-phy 381f0040.usb-phy: bus resume > [ 84.263195] imx8mq-usb-phy 382f0040.usb-phy: bus resume > [ 84.268898] dwc3 38100000.usb: driver resume > > during resume from S3 suspend, here it still hangs. Is your problem a system hang? If yes, this may another issue, where the hang happens? dwc3_resume_common()? The question patch is to give a warm reset for connected USB device if the link state is not connect/CAS after system resume, otherwise host will wait 2s for device appear: [ 44.834831] usb 2-1: Waited 2000ms for CONNECT ... [ 45.055718] PM: resume devices took 3.132 seconds I will post this patch and doc(to be updated) to upstream later. Li Jun > > What else could be missing here? I'm using mainline Linux only of course. > > thanks, > > martin > > > > drivers/usb/host/xhci-plat.c | 4 ++++ > 1 file changed, 4 insertions(+) > > diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c index > d90cd5ec09cf..4655016eaf45 100644 > --- a/drivers/usb/host/xhci-plat.c > +++ b/drivers/usb/host/xhci-plat.c > @@ -291,6 +291,10 @@ static int xhci_plat_probe(struct platform_device *pdev) > > device_property_read_u32(tmpdev, "imod-interval-ns", > &xhci->imod_interval); > + > + if (device_property_read_bool(tmpdev, > + "usb3-resume-missing-cas")) > + xhci->quirks |= XHCI_MISSING_CAS; > } > > hcd->usb_phy = devm_usb_get_phy_by_phandle(sysdev, "usb-phy", 0); > -- > 2.20.1
On 20.02.20 07:31, Jun Li wrote: >> -----Original Message----- >> From: Martin Kepplinger <martin.kepplinger@puri.sm> >> Sent: 2020年2月20日 1:37 >> To: Jun Li <jun.li@nxp.com>; Peter Chen <peter.chen@nxp.com>; >> mathias.nyman@intel.com >> Cc: linux-usb@vger.kernel.org; dl-linux-imx <linux-imx@nxp.com>; Anson Huang >> <anson.huang@nxp.com>; shawnguo@kernel.org; kernel@pengutronix.de >> Subject: [PATCH] usb: host: xhci-plat: add XHCI_MISSING_CAS quirk >> >> From: Li Jun <jun.li@nxp.com> >> >> i.MX8MQ USB3 host needs XHCI_MISSING_CAS quirk to warm reset the port to enum the >> USB3 device plugged in while system sleep, as the port state is stuck in polling >> mode after resume. >> >> Signed-off-by: Li Jun <jun.li@nxp.com> >> Acked-by: Peter Chen <peter.chen@nxp.com> >> --- >> >> Hi, >> >> Because resume from S3 suspend is broken for me on imx8mq, I stumbled upon this >> patch in NXP's linux tree. (Please note that I'm not the author and I've not yet >> put my SoB tag under it). This is just a >> question: >> >> This patch (and the docs) clearly is missing in mainline Linux because the imx8mq >> devicetree description includes it (which does nothing now). >> >> I've tested this and this particular addition doesn't fix my problem: >> >> [ 84.257538] imx8mq-usb-phy 381f0040.usb-phy: bus resume >> [ 84.263195] imx8mq-usb-phy 382f0040.usb-phy: bus resume >> [ 84.268898] dwc3 38100000.usb: driver resume >> >> during resume from S3 suspend, here it still hangs. > > Is your problem a system hang? If yes, this may another issue, > where the hang happens? dwc3_resume_common()? exactly! I followed to the point it hangs once again and it's dwc3_core_init() called from dwc3_resume_common()'s "OTG" case. Specifically, dwc3_writel() is what I don't get past: https://elixir.bootlin.com/linux/v5.6-rc2/source/drivers/usb/dwc3/core.c#L934 do you have an idea why this writel() hangs? > > The question patch is to give a warm reset for connected USB > device if the link state is not connect/CAS after system resume, > otherwise host will wait 2s for device appear: > > [ 44.834831] usb 2-1: Waited 2000ms for CONNECT > ... > [ 45.055718] PM: resume devices took 3.132 seconds > > I will post this patch and doc(to be updated) to upstream later. > ok, good, thanks, martin
> -----Original Message----- > From: Martin Kepplinger <martin.kepplinger@puri.sm> > Sent: 2020年2月20日 19:44 > To: Jun Li <jun.li@nxp.com>; Peter Chen <peter.chen@nxp.com>; > mathias.nyman@intel.com > Cc: linux-usb@vger.kernel.org; dl-linux-imx <linux-imx@nxp.com>; Anson Huang > <anson.huang@nxp.com>; shawnguo@kernel.org; kernel@pengutronix.de > Subject: Re: [PATCH] usb: host: xhci-plat: add XHCI_MISSING_CAS quirk > > On 20.02.20 07:31, Jun Li wrote: > >> -----Original Message----- > >> From: Martin Kepplinger <martin.kepplinger@puri.sm> > >> Sent: 2020年2月20日 1:37 > >> To: Jun Li <jun.li@nxp.com>; Peter Chen <peter.chen@nxp.com>; > >> mathias.nyman@intel.com > >> Cc: linux-usb@vger.kernel.org; dl-linux-imx <linux-imx@nxp.com>; > >> Anson Huang <anson.huang@nxp.com>; shawnguo@kernel.org; > >> kernel@pengutronix.de > >> Subject: [PATCH] usb: host: xhci-plat: add XHCI_MISSING_CAS quirk > >> > >> From: Li Jun <jun.li@nxp.com> > >> > >> i.MX8MQ USB3 host needs XHCI_MISSING_CAS quirk to warm reset the port > >> to enum the > >> USB3 device plugged in while system sleep, as the port state is stuck > >> in polling mode after resume. > >> > >> Signed-off-by: Li Jun <jun.li@nxp.com> > >> Acked-by: Peter Chen <peter.chen@nxp.com> > >> --- > >> > >> Hi, > >> > >> Because resume from S3 suspend is broken for me on imx8mq, I stumbled > >> upon this patch in NXP's linux tree. (Please note that I'm not the > >> author and I've not yet put my SoB tag under it). This is just a > >> question: > >> > >> This patch (and the docs) clearly is missing in mainline Linux > >> because the imx8mq devicetree description includes it (which does nothing now). > >> > >> I've tested this and this particular addition doesn't fix my problem: > >> > >> [ 84.257538] imx8mq-usb-phy 381f0040.usb-phy: bus resume > >> [ 84.263195] imx8mq-usb-phy 382f0040.usb-phy: bus resume > >> [ 84.268898] dwc3 38100000.usb: driver resume > >> > >> during resume from S3 suspend, here it still hangs. > > > > Is your problem a system hang? If yes, this may another issue, where > > the hang happens? dwc3_resume_common()? > > exactly! I followed to the point it hangs once again and it's > > dwc3_core_init() called from dwc3_resume_common()'s "OTG" case. > > Specifically, dwc3_writel() is what I don't get past: > https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Felixir.bootl > in.com%2Flinux%2Fv5.6-rc2%2Fsource%2Fdrivers%2Fusb%2Fdwc3%2Fcore.c%23L934& > data=02%7C01%7Cjun.li%40nxp.com%7C130cd29875c44792d1a908d7b5fa2516%7C686ea1d3b > c2b4c6fa92cd99c5c301635%7C0%7C0%7C637177958284696041&sdata=mqh9MH6ESLVxKvW > vvMq4vwt2dcTuvNopgGVdXEbbMwk%3D&reserved=0 So while dwc3 resume, the first register access cause hang. Looks like some required clocks or power domain of USB0 is not on. > > do you have an idea why this writel() hangs? I never encounter such hang on my iMX8MQ EVK board using upstream kernel(5.x) + changes of enable USB0 port, but I didn't try latest 5.6 kernel. I will enable the first port based on Linux-next to give a try on my NXP iMX8MQ EVK board, do you think I can reproduce your problem? Li Jun > > > > > The question patch is to give a warm reset for connected USB device if > > the link state is not connect/CAS after system resume, otherwise host > > will wait 2s for device appear: > > > > [ 44.834831] usb 2-1: Waited 2000ms for CONNECT > > ... > > [ 45.055718] PM: resume devices took 3.132 seconds > > > > I will post this patch and doc(to be updated) to upstream later. > > > > ok, good, thanks, > > martin
On 20.02.20 16:37, Jun Li wrote: > > >> -----Original Message----- >> From: Martin Kepplinger <martin.kepplinger@puri.sm> >> Sent: 2020年2月20日 19:44 >> To: Jun Li <jun.li@nxp.com>; Peter Chen <peter.chen@nxp.com>; >> mathias.nyman@intel.com >> Cc: linux-usb@vger.kernel.org; dl-linux-imx <linux-imx@nxp.com>; Anson Huang >> <anson.huang@nxp.com>; shawnguo@kernel.org; kernel@pengutronix.de >> Subject: Re: [PATCH] usb: host: xhci-plat: add XHCI_MISSING_CAS quirk >> >> On 20.02.20 07:31, Jun Li wrote: >>>> -----Original Message----- >>>> From: Martin Kepplinger <martin.kepplinger@puri.sm> >>>> Sent: 2020年2月20日 1:37 >>>> To: Jun Li <jun.li@nxp.com>; Peter Chen <peter.chen@nxp.com>; >>>> mathias.nyman@intel.com >>>> Cc: linux-usb@vger.kernel.org; dl-linux-imx <linux-imx@nxp.com>; >>>> Anson Huang <anson.huang@nxp.com>; shawnguo@kernel.org; >>>> kernel@pengutronix.de >>>> Subject: [PATCH] usb: host: xhci-plat: add XHCI_MISSING_CAS quirk >>>> >>>> From: Li Jun <jun.li@nxp.com> >>>> >>>> i.MX8MQ USB3 host needs XHCI_MISSING_CAS quirk to warm reset the port >>>> to enum the >>>> USB3 device plugged in while system sleep, as the port state is stuck >>>> in polling mode after resume. >>>> >>>> Signed-off-by: Li Jun <jun.li@nxp.com> >>>> Acked-by: Peter Chen <peter.chen@nxp.com> >>>> --- >>>> >>>> Hi, >>>> >>>> Because resume from S3 suspend is broken for me on imx8mq, I stumbled >>>> upon this patch in NXP's linux tree. (Please note that I'm not the >>>> author and I've not yet put my SoB tag under it). This is just a >>>> question: >>>> >>>> This patch (and the docs) clearly is missing in mainline Linux >>>> because the imx8mq devicetree description includes it (which does nothing now). >>>> >>>> I've tested this and this particular addition doesn't fix my problem: >>>> >>>> [ 84.257538] imx8mq-usb-phy 381f0040.usb-phy: bus resume >>>> [ 84.263195] imx8mq-usb-phy 382f0040.usb-phy: bus resume >>>> [ 84.268898] dwc3 38100000.usb: driver resume >>>> >>>> during resume from S3 suspend, here it still hangs. >>> >>> Is your problem a system hang? If yes, this may another issue, where >>> the hang happens? dwc3_resume_common()? >> >> exactly! I followed to the point it hangs once again and it's >> >> dwc3_core_init() called from dwc3_resume_common()'s "OTG" case. >> >> Specifically, dwc3_writel() is what I don't get past: >> https://eur01.safelinks.protection.outlook.com/?url=https%3A%2F%2Felixir.bootl >> in.com%2Flinux%2Fv5.6-rc2%2Fsource%2Fdrivers%2Fusb%2Fdwc3%2Fcore.c%23L934& >> data=02%7C01%7Cjun.li%40nxp.com%7C130cd29875c44792d1a908d7b5fa2516%7C686ea1d3b >> c2b4c6fa92cd99c5c301635%7C0%7C0%7C637177958284696041&sdata=mqh9MH6ESLVxKvW >> vvMq4vwt2dcTuvNopgGVdXEbbMwk%3D&reserved=0 > > So while dwc3 resume, the first register access cause hang. > Looks like some required clocks or power domain of USB0 is not on. which would these be in imx8mq.dtsi ? > >> >> do you have an idea why this writel() hangs? > > I never encounter such hang on my iMX8MQ EVK board > using upstream kernel(5.x) + changes of enable USB0 port, > but I didn't try latest 5.6 kernel. > I will enable the first port based on Linux-next to give a > try on my NXP iMX8MQ EVK board, do you think I can reproduce > your problem? > > Li Jun Hi Li, thanks for taking a look. I would think that you can reproduce that, yes. I'm running v5.6-rc2 with no changes to usb or any relevant changes to clk (I think), but in case you're curious, this is the whole tree: https://source.puri.sm/Librem5/linux-next/tree/imx8-linux-next-librem5 What "changes of enable USB0 port" do you mean? thanks, martin >> >>> >>> The question patch is to give a warm reset for connected USB device if >>> the link state is not connect/CAS after system resume, otherwise host >>> will wait 2s for device appear: >>> >>> [ 44.834831] usb 2-1: Waited 2000ms for CONNECT >>> ... >>> [ 45.055718] PM: resume devices took 3.132 seconds >>> >>> I will post this patch and doc(to be updated) to upstream later. >>> >> >> ok, good, thanks, >> >> martin >
diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c index d90cd5ec09cf..4655016eaf45 100644 --- a/drivers/usb/host/xhci-plat.c +++ b/drivers/usb/host/xhci-plat.c @@ -291,6 +291,10 @@ static int xhci_plat_probe(struct platform_device *pdev) device_property_read_u32(tmpdev, "imod-interval-ns", &xhci->imod_interval); + + if (device_property_read_bool(tmpdev, + "usb3-resume-missing-cas")) + xhci->quirks |= XHCI_MISSING_CAS; } hcd->usb_phy = devm_usb_get_phy_by_phandle(sysdev, "usb-phy", 0);