Hi, On Fri, Sep 26, 2014 at 11:18:48PM +0000, Paul Zimmerman wrote: > > From: Felipe Balbi [mailto:balbi@xxxxxx] > > Sent: Friday, September 26, 2014 2:40 PM > > > > On Fri, Sep 26, 2014 at 08:57:19PM +0000, Paul Zimmerman wrote: > > > > > diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c > > > > > index 0fcc0a3..8277065 100644 > > > > > --- a/drivers/usb/dwc3/gadget.c > > > > > +++ b/drivers/usb/dwc3/gadget.c > > > > > @@ -2635,6 +2635,7 @@ static irqreturn_t dwc3_interrupt(int irq, void *_dwc) > > > > > */ > > > > > int dwc3_gadget_init(struct dwc3 *dwc) > > > > > { > > > > > + u32 reg; > > > > > int ret; > > > > > > > > > > dwc->ctrl_req = dma_alloc_coherent(dwc->dev, sizeof(*dwc->ctrl_req), > > > > > @@ -2689,6 +2690,13 @@ int dwc3_gadget_init(struct dwc3 *dwc) > > > > > if (ret) > > > > > goto err4; > > > > > > > > > > + if (dwc->quirks & DWC3_AMD_NL_PLAT) { > > > > > + reg = dwc3_readl(dwc->regs, DWC3_DCTL); > > > > > + reg |= DWC3_DCTL_LPM_ERRATA(0xf); > > > > > > > > weird, why would Synopsys put this here ? It seems like this is only > > > > useful when LPM Errata is enabled and that's, apparently, a host-side > > > > thing. > > > > > > > > Paul, can you comment ? > > > > > > These bits contribute to how the device responds to an LPM transaction > > > from the host. If DCFG.LPMCap=1, they set the BESL value above which > > > the core will send a NYET. If DCFG.LPMCap=0, they have no effect. So > > > it's definitely a device-side thing, but only if the core is configured > > > with LPM Errata support enabled. > > > > right, and how can SW detect if LPM Errata was enabled ? From the host > > point of view, we can check bit 20 of xHCI capability register. What > > about device ? I can't seem to find anything :-s > > I just talked to one of our RTL designers. You're right, there is no > way to tell from the Device registers alone whether the controller is > configured with LPM Errata support or not. > > You can tell for sure if it is *not* enabled, by checking GSNPSID, and > if the version is earlier than 2.40a, the feature wasn't available. alright, we can use this for something like: WARN(rev < 2.40a && (flags & LPM_ERRATA || of_property_read_bool("lpm-errata")), "LPM Errata not available on dwc3 revisions <= 2.40a\n"); > So for Device-mode only controllers, I guess you will need a DT property > for this. right, DT property and platform_data feature flag, or something. I don't wanna call it a quirk though, although it _is_ an erratum WRT LPM support. Dunno. Any strong feelings ? -- balbi
Attachment:
signature.asc
Description: Digital signature
