On Thu, Jan 12, 2012 at 09:14:52AM +0800, Alex,Shi wrote: > On Wed, 2012-01-11 at 14:28 -0800, Sarah Sharp wrote: > > On Wed, Jan 11, 2012 at 04:51:00PM +0800, Alex,Shi wrote: > > > On Wed, 2012-01-11 at 16:39 +0800, Andiry Xu wrote: > > > > On 01/11/2012 04:17 PM, Alex,Shi wrote: > > > > >> Since this device ID is not used by anyone, you'd better put it in xhci > > > > >> driver. > > > > > > > > > > how about to move to xhci_pci.c file as following new patch > > > > > > > > This looks better. > > > > > > > > >> > > > > >> And I don't think PCI_DEVICE_ID_INTEL_USB_XHCI is a good enough name - > > > > >> are you sure all the Intel xHCI host controllers will share the same PCI > > > > >> device ID? > > > > > > > > > > I don't have better name for it. :( > > > > > > > > Perhaps you can use the chipset name. For example, panther point? :) > > > > > > Maybe. Let's wait for Sarah's comments. :) > > > > You do need a better name. In fact, it's already there in pci_ids.h: > > > > #define PCI_DEVICE_ID_INTEL_PANTHERPOINT_XHCI 0x1e31 > > Thanks! > > > > However, you've told me that this behavior only occurs on one particular > > Intel BIOS. The BIOS I have on my Panther Point board doesn't fail when > > legacy PCI interrupts are enabled for the xHCI host. So I think you > > need to also somehow check the BIOS or platform name, instead of turning > > on this quirk for all Intel Panther Point xHCI hosts. > > Rui, do you has more info of this BIOS? Please keep BIOS details off list, thanks. Since they're not released yet, we're getting into sticky territory. > :) If we try to enable MSI first for xhci, like others' suggestion. we > may can skip this detailed device ID. Yes, but you're forgetting that this bug fix really needs to go into the stable kernels. You need one small patch (that doesn't change the current USB PCI interrupt allocation for other xHCI hosts) to fix the problem with the Panther Point xHCI host issue, and a second bigger patch to just skip the legacy interrupt PCI allocation all together. > > Let's figure out internally which BIOS versions actually have this > > problem, and how to differentiate between the two. > > Do you mean we can ignore such BIOS issue? > If you can skip such problem in BIOS and make HCD with MSI work like > Windows, why not? And guess there always have BIOS engineers forgetting > do test for Linux. Again, for the stable kernel reasons. I'm not really comfortable with changing the PCI interrupt allocation in the stable kernel trees for other xHCI hosts. If Alan or Greg is, fine, you can do it in one patch, but I want to make sure our platform actually works with the stable kernels. Sarah Sharp -- 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
