Very much leaning towards XHCI_NOCSS on all AMD chipsets. I wonder if there are some AMD xhci that behave different, or if past machines with S3 have hidden this problem by taking control of the xhci themselves. But we probably used to suspend without CSS handling on those machines, right?
Dan Jones <[email protected]> wrote: > On Tue, Aug 5, 2025 at 9:02 PM Jonathan Gray <[email protected]> wrote: > > > > On Tue, Aug 05, 2025 at 05:17:32PM -0600, Dan Jones wrote: > > > After updating to a Fri Aug 1 11:00:19 MDT 2025 snapshot which > > > includes the new amdpmc driver, I was able to successfully > > > suspend-to-idle and resume. Unfortunately, after suspend any usb hubs > > > with devices attached fail with a console error "uhub0: device > > > problem, disabling port". I was able to attach the keyboard to a > > > different usb hub and that hub and keyboard was functional. > > > > > > I am appreciative of all the work by Mark Kettenis and the other > > > openbsd devs. Please let me know what other information I can provide > > > or steps to take. > > > > > xhci0 at pci6 dev 0 function 3 "AMD 19h/7xh xHCI" rev 0x00: msix, xHCI > > > 1.20 > > > > try this > > > > Index: sys/dev/pci/xhci_pci.c > > =================================================================== > > RCS file: /cvs/src/sys/dev/pci/xhci_pci.c,v > > diff -u -p -r1.16 xhci_pci.c > > --- sys/dev/pci/xhci_pci.c 19 Jun 2025 14:08:13 -0000 1.16 > > +++ sys/dev/pci/xhci_pci.c 6 Aug 2025 02:58:38 -0000 > > @@ -152,15 +152,7 @@ xhci_pci_attach(struct device *parent, s > > pa->pa_flags &= ~PCI_FLAGS_MSI_ENABLED; > > break; > > case PCI_VENDOR_AMD: > > - if (PCI_PRODUCT(pa->pa_id) == PCI_PRODUCT_AMD_17_1X_XHCI_1 > > || > > - PCI_PRODUCT(pa->pa_id) == PCI_PRODUCT_AMD_17_1X_XHCI_2 > > || > > - PCI_PRODUCT(pa->pa_id) == PCI_PRODUCT_AMD_17_6X_XHCI || > > - PCI_PRODUCT(pa->pa_id) == PCI_PRODUCT_AMD_19_4X_XHCI_1 > > || > > - PCI_PRODUCT(pa->pa_id) == PCI_PRODUCT_AMD_19_4X_XHCI_2 > > || > > - PCI_PRODUCT(pa->pa_id) == PCI_PRODUCT_AMD_19_4X_XHCI_3 > > || > > - PCI_PRODUCT(pa->pa_id) == PCI_PRODUCT_AMD_19_4X_XHCI_4 > > || > > - PCI_PRODUCT(pa->pa_id) == PCI_PRODUCT_AMD_19_4X_XHCI_5) > > - psc->sc.sc_flags |= XHCI_NOCSS; > > + psc->sc.sc_flags |= XHCI_NOCSS; > > break; > > } > > > > Thank you, after applying the patch the system is able to suspend and > resume without any issues. Let me know if you need a dmesg or other > information. >
