On Fri, Jan 20, 2017 at 02:08:50PM +0200, Felipe Balbi wrote: > > Hi, > > Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> writes: > > On Fri, Jan 20, 2017 at 01:35:39PM +0200, Mathias Nyman wrote: > >> On 20.01.2017 12:22, Greg KH wrote: > >> > On Fri, Jan 20, 2017 at 11:23:36AM +0200, Mathias Nyman wrote: > >> > > On 19.01.2017 20:48, Greg KH wrote: > >> > > > On Thu, Jan 19, 2017 at 02:21:26PM +0200, Mathias Nyman wrote: > >> > > > > Hi Greg > >> > > > > > >> > > > > This series by Arnd Bergmann was originally six patches, but last two of > >> > > > > them were already taken to 4.10. Without the rest of them there will > >> > > > > be a regression in 4.10. > >> > > > > >> > > > Is it really a regression? I thought this had never worked before in > >> > > > older kernels, right? > >> > > > > >> > > > >> > > Regression when xhci hosts in dwc3 controllers are used. > >> > > >> > So that worked in 4.9? > >> > > >> > > For example patch 5/6 removed setting dma mask for xhci in dwc3 host init: > >> > > > >> > > +++ b/drivers/usb/dwc3/host.c > >> > > @@ -84,11 +84,7 @@ int dwc3_host_init(struct dwc3 *dwc) > >> > > return -ENOMEM; > >> > > } > >> > > - dma_set_coherent_mask(&xhci->dev, dwc->dev->coherent_dma_mask); > >> > > - > >> > > xhci->dev.parent = dwc->dev; > >> > > - xhci->dev.dma_mask = dwc->dev->dma_mask; > >> > > - xhci->dev.dma_parms = dwc->dev->dma_parms; > >> > > > >> > > So now xhci platform driver prints a scary warning because of the missing dma mask: > >> > > > >> > > static int xhci_plat_probe(struct platform_device *pdev) > >> > > /* Try to set 64-bit DMA first */ > >> > > if (WARN_ON(!pdev->dev.dma_mask)) > >> > > /* Platform did not initialize dma_mask */ > >> > > ret = dma_coerce_mask_and_coherent(&pdev->dev, > >> > > DMA_BIT_MASK(64)); > >> > > else > >> > > ... > >> > > This is fixed in the first 4 patches. > >> > > > >> > > There might be other other issues as well caused by having only the dwc3 > >> > > changed applied of this series, but not the core and xhci parts > >> > > >> > Should we just fix the "scary warning" instead, by removing it? :) > >> > > >> > I say all of this because this seems like some very big changes so late > >> > in the -rc cycle. > >> > > >> > >> I guess that would work, or at least get us to the same stage as 4.9. > >> I'll send a patch for it. > > > > Great. > > > >> Gives more time to look at the usb core changes. I'm not really > >> myself running or testing the dwc3 host side. > > > > Me either. Any hints on some hardware that would allow me to do that? > > Intel Edison. Ah nice, if only I knew someone at Intel who could get me one of those... :) > Or any recent TI board (AM437x SK, for instance). Will that run a mainline kernel? > Google Pixel Phone (but good luck running a mainline kernel there ;-) I have a Pixel phone here, but haven't been paying attention to the dwc3 interface. Is dwc3 a block in the SoC on the Pixel? I have another device here with the same SoC as the Pixel (OnePlus 3T) that I can run my own kernel on, but it's a bit older version, due to SoC issues (same one the Pixel has at the moment...) However some of us have a crazy idea to drag that platform up to mainline over the next few months. thanks, greg k-h -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
