Re: [PATCH V2] usb: xhci: add support for performing fake doorbell
On Mon, Jan 16, 2017 at 2:32 AM, Rafał Miłecki wrote: > On 21 November 2016 at 16:31, Mathias Nyman > wrote: >> On 21.11.2016 09:57, Rafał Miłecki wrote: >>> >>> Hi Mathias, >>> >>> On 17 October 2016 at 22:30, Rafał Miłecki wrote: From: Rafał Miłecki Broadcom's Northstar XHCI controllers seem to need a special start procedure to work correctly. There isn't any official documentation of this, the problem is that controller doesn't detect any connected devices with default setup. Moreover connecting USB device to controller that doesn't run properly can cause SoC's watchdog issues. A workaround that was successfully tested on multiple devices is to perform a fake doorbell. This patch adds code for doing this and enables it on BCM4708 family. Signed-off-by: Rafał Miłecki --- V2: Enable quirk for brcm,bcm4708 machines instead of adding separated binding for it. Thanks Rob for your comment on this. >>> >>> >>> Do you think you can pick & push this one? V2 follows Rob's suggestion >>> and he has some DT knowledge for sure, so I guess it should be OK. >>> -- >> >> >> Is there some more background information on this? >> >> I don't have any contacts to Broadcom myself, adding the BMC Kernel Feedback >> list to CC. >> Maybe someone over there has an errata, documentation or just general >> feedback. >> >> How was this workaround even figured out? ringing the doorbell for the first >> device doesn't seem like something found by trial and error, especially >> when >> xhci specs state that: >> >> "Software shall not write the Doorbell of an endpoint until after it has >> issued a >> Configure Endpoint Command for the endpoint and received a successful >> Command >> Completion Event." >> >> The whole workaround is a bit intrusive, allocating a fake device, ring a >> doorbell for a >> fake device in the wrong state, clearing off HSE (host system error) which >> should only be set >> when things really go bad, some random usleeps, and possible calling >> xhci_start() twice. >> >> I can't take this as is without some more info. > > Hi (ping) Broadcom guys, could you help us with this USB workaround, please? Investigating internally. I'll let you know if I can find anything. Thanks, Jon > > -- > Rafał
Re: [PATCH V2] usb: xhci: add support for performing fake doorbell
On 21 November 2016 at 16:31, Mathias Nyman wrote: > On 21.11.2016 09:57, Rafał Miłecki wrote: >> >> Hi Mathias, >> >> On 17 October 2016 at 22:30, Rafał Miłecki wrote: >>> >>> From: Rafał Miłecki >>> >>> Broadcom's Northstar XHCI controllers seem to need a special start >>> procedure to work correctly. There isn't any official documentation of >>> this, the problem is that controller doesn't detect any connected >>> devices with default setup. Moreover connecting USB device to controller >>> that doesn't run properly can cause SoC's watchdog issues. >>> >>> A workaround that was successfully tested on multiple devices is to >>> perform a fake doorbell. This patch adds code for doing this and enables >>> it on BCM4708 family. >>> >>> Signed-off-by: Rafał Miłecki >>> --- >>> V2: Enable quirk for brcm,bcm4708 machines instead of adding separated >>> binding >>> for it. Thanks Rob for your comment on this. >> >> >> Do you think you can pick & push this one? V2 follows Rob's suggestion >> and he has some DT knowledge for sure, so I guess it should be OK. >> -- > > > Is there some more background information on this? > > I don't have any contacts to Broadcom myself, adding the BMC Kernel Feedback > list to CC. > Maybe someone over there has an errata, documentation or just general > feedback. > > How was this workaround even figured out? ringing the doorbell for the first > device doesn't seem like something found by trial and error, especially > when > xhci specs state that: > > "Software shall not write the Doorbell of an endpoint until after it has > issued a > Configure Endpoint Command for the endpoint and received a successful > Command > Completion Event." > > The whole workaround is a bit intrusive, allocating a fake device, ring a > doorbell for a > fake device in the wrong state, clearing off HSE (host system error) which > should only be set > when things really go bad, some random usleeps, and possible calling > xhci_start() twice. > > I can't take this as is without some more info. Hi (ping) Broadcom guys, could you help us with this USB workaround, please? -- Rafał
Re: [PATCH V2] usb: xhci: add support for performing fake doorbell
On 21.11.2016 09:57, Rafał Miłecki wrote: Hi Mathias, On 17 October 2016 at 22:30, Rafał Miłecki wrote: From: Rafał Miłecki Broadcom's Northstar XHCI controllers seem to need a special start procedure to work correctly. There isn't any official documentation of this, the problem is that controller doesn't detect any connected devices with default setup. Moreover connecting USB device to controller that doesn't run properly can cause SoC's watchdog issues. A workaround that was successfully tested on multiple devices is to perform a fake doorbell. This patch adds code for doing this and enables it on BCM4708 family. Signed-off-by: Rafał Miłecki --- V2: Enable quirk for brcm,bcm4708 machines instead of adding separated binding for it. Thanks Rob for your comment on this. Do you think you can pick & push this one? V2 follows Rob's suggestion and he has some DT knowledge for sure, so I guess it should be OK. -- Is there some more background information on this? I don't have any contacts to Broadcom myself, adding the BMC Kernel Feedback list to CC. Maybe someone over there has an errata, documentation or just general feedback. How was this workaround even figured out? ringing the doorbell for the first device doesn't seem like something found by trial and error, especially when xhci specs state that: "Software shall not write the Doorbell of an endpoint until after it has issued a Configure Endpoint Command for the endpoint and received a successful Command Completion Event." The whole workaround is a bit intrusive, allocating a fake device, ring a doorbell for a fake device in the wrong state, clearing off HSE (host system error) which should only be set when things really go bad, some random usleeps, and possible calling xhci_start() twice. I can't take this as is without some more info. -Mathias
Re: [PATCH V2] usb: xhci: add support for performing fake doorbell
Hi Mathias, On 17 October 2016 at 22:30, Rafał Miłecki wrote: > From: Rafał Miłecki > > Broadcom's Northstar XHCI controllers seem to need a special start > procedure to work correctly. There isn't any official documentation of > this, the problem is that controller doesn't detect any connected > devices with default setup. Moreover connecting USB device to controller > that doesn't run properly can cause SoC's watchdog issues. > > A workaround that was successfully tested on multiple devices is to > perform a fake doorbell. This patch adds code for doing this and enables > it on BCM4708 family. > > Signed-off-by: Rafał Miłecki > --- > V2: Enable quirk for brcm,bcm4708 machines instead of adding separated binding > for it. Thanks Rob for your comment on this. Do you think you can pick & push this one? V2 follows Rob's suggestion and he has some DT knowledge for sure, so I guess it should be OK.
Re: [PATCH V2] usb: xhci: add support for performing fake doorbell
On 17 October 2016 at 23:10, Hauke Mehrtens wrote: > On 10/17/2016 10:30 PM, Rafał Miłecki wrote: >> From: Rafał Miłecki >> >> Broadcom's Northstar XHCI controllers seem to need a special start >> procedure to work correctly. There isn't any official documentation of >> this, the problem is that controller doesn't detect any connected >> devices with default setup. Moreover connecting USB device to controller >> that doesn't run properly can cause SoC's watchdog issues. >> >> A workaround that was successfully tested on multiple devices is to >> perform a fake doorbell. This patch adds code for doing this and enables >> it on BCM4708 family. >> >> Signed-off-by: Rafał Miłecki >> --- >> V2: Enable quirk for brcm,bcm4708 machines instead of adding separated >> binding >> for it. Thanks Rob for your comment on this. >> --- >> drivers/usb/host/xhci-plat.c | 6 + >> drivers/usb/host/xhci.c | 63 >> +--- >> drivers/usb/host/xhci.h | 1 + >> 3 files changed, 67 insertions(+), 3 deletions(-) >> >> diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c >> index ed56bf9..b01a3be 100644 >> --- a/drivers/usb/host/xhci-plat.c >> +++ b/drivers/usb/host/xhci-plat.c >> @@ -56,12 +56,18 @@ static int xhci_priv_init_quirk(struct usb_hcd *hcd) >> >> static void xhci_plat_quirks(struct device *dev, struct xhci_hcd *xhci) >> { >> + struct platform_device *pdev = to_platform_device(dev); >> + struct device_node *node = pdev->dev.of_node; >> + >> /* >>* As of now platform drivers don't provide MSI support so we ensure >>* here that the generic code does not try to make a pci_dev from our >>* dev struct in order to setup MSI >>*/ >> xhci->quirks |= XHCI_PLAT; >> + >> + if (node && of_machine_is_compatible("brcm,bcm4708")) >> + xhci->quirks |= XHCI_FAKE_DOORBELL; > > Are you sure only the bcm4708 and similar SoCs are affected? Having here > a list with 3 or more checks would looks strange. I prefer your first patch. > > @Florian do you know if other Broadcom SoC are affected by this problem > or are only Northstar SoCs affected? I also believed usb3-fake-doorbell property looks nicer, but then just followed Rob's suggestion. I don't know about Northstar Plus or Northstar 2, I just know it's not needed for BCM53573. -- Rafał
Re: [PATCH V2] usb: xhci: add support for performing fake doorbell
On 10/17/2016 10:30 PM, Rafał Miłecki wrote: > From: Rafał Miłecki > > Broadcom's Northstar XHCI controllers seem to need a special start > procedure to work correctly. There isn't any official documentation of > this, the problem is that controller doesn't detect any connected > devices with default setup. Moreover connecting USB device to controller > that doesn't run properly can cause SoC's watchdog issues. > > A workaround that was successfully tested on multiple devices is to > perform a fake doorbell. This patch adds code for doing this and enables > it on BCM4708 family. > > Signed-off-by: Rafał Miłecki > --- > V2: Enable quirk for brcm,bcm4708 machines instead of adding separated binding > for it. Thanks Rob for your comment on this. > --- > drivers/usb/host/xhci-plat.c | 6 + > drivers/usb/host/xhci.c | 63 > +--- > drivers/usb/host/xhci.h | 1 + > 3 files changed, 67 insertions(+), 3 deletions(-) > > diff --git a/drivers/usb/host/xhci-plat.c b/drivers/usb/host/xhci-plat.c > index ed56bf9..b01a3be 100644 > --- a/drivers/usb/host/xhci-plat.c > +++ b/drivers/usb/host/xhci-plat.c > @@ -56,12 +56,18 @@ static int xhci_priv_init_quirk(struct usb_hcd *hcd) > > static void xhci_plat_quirks(struct device *dev, struct xhci_hcd *xhci) > { > + struct platform_device *pdev = to_platform_device(dev); > + struct device_node *node = pdev->dev.of_node; > + > /* >* As of now platform drivers don't provide MSI support so we ensure >* here that the generic code does not try to make a pci_dev from our >* dev struct in order to setup MSI >*/ > xhci->quirks |= XHCI_PLAT; > + > + if (node && of_machine_is_compatible("brcm,bcm4708")) > + xhci->quirks |= XHCI_FAKE_DOORBELL; Are you sure only the bcm4708 and similar SoCs are affected? Having here a list with 3 or more checks would looks strange. I prefer your first patch. @Florian do you know if other Broadcom SoC are affected by this problem or are only Northstar SoCs affected? Hauke