On Thu, 23 May 2013, Roger Quadros wrote: > Device tree probed devices don't get dma_mask set. Previously > we were setting the dma_mask pointer only if it was NULL. > However, the address of 'omap_ehci_dma_mask' would change > each time the module is unloaded and loaded back thus causing > the devices dma_mask pointer to be invalid on the next load. > > This will cause page faults if any driver tries to access the > old dma_mask pointer. > > Unconditionally re-setting the dma_mask pointer fixes this problem. > diff --git a/drivers/usb/host/ehci-omap.c b/drivers/usb/host/ehci-omap.c > index 3d1491b..b33e306 100644 > --- a/drivers/usb/host/ehci-omap.c > +++ b/drivers/usb/host/ehci-omap.c > @@ -146,8 +146,7 @@ static int ehci_hcd_omap_probe(struct platform_device *pdev) > * Since shared usb code relies on it, set it here for now. > * Once we have dma capability bindings this can go away. > */ > - if (!pdev->dev.dma_mask) > - pdev->dev.dma_mask = &omap_ehci_dma_mask; > + pdev->dev.dma_mask = &omap_ehci_dma_mask; Is this the solution that people have agreed on? There has been a lot of discussion on this topic. In particular, there has been talk about fixing it in the DT core. This particular approach doesn't seem very robust. What if pdev->dev.dma_mask is already set to a different value for some good reason? Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html