on 27/09/2009 15:17 Andriy Gapon said the following: > Another idea of working around this: > 1) in pci fixup code disable USB SMI for these chipsets > 2) (optional) in ohci code skip takeover step > Sounds messy. BTW, just for the sake of experiment I did exactly what I suggested. I've got the following messages: kernel: ohci_controller_init:195: SMM active, request owner change kernel: usbus0: SMM does not respond, resetting kernel: ohci_controller_init:195: SMM active, request owner change kernel: usbus1: SMM does not respond, resetting kernel: ohci_controller_init:195: SMM active, request owner change kernel: usbus3: SMM does not respond, resetting kernel: ohci_controller_init:195: SMM active, request owner change kernel: usbus4: SMM does not respond, resetting kernel: ohci_controller_init:195: SMM active, request owner change kernel: usbus6: SMM does not respond, resetting And the register value stayed intact after initial programming, so no re-programming was needed. Here is the (dirty) hack: diff --git a/sys/dev/pci/fixup_pci.c b/sys/dev/pci/fixup_pci.c index 566e503..1463c24 100644 --- a/sys/dev/pci/fixup_pci.c +++ b/sys/dev/pci/fixup_pci.c _at__at_ -53,6 +53,7 _at__at_ static int fixup_pci_probe(device_t dev); static void fixwsc_natoma(device_t dev); static void fixc1_nforce2(device_t dev); static void fixrtc_piix4(device_t dev); +static void fixsmi_usb(device_t dev); static device_method_t fixup_pci_methods[] = { /* Device interface */ _at__at_ -84,6 +85,9 _at__at_ fixup_pci_probe(device_t dev) case 0x01e010de: /* nVidia nForce2 */ fixc1_nforce2(dev); break; + case 0x96001022: /* AMD SB700 */ + fixsmi_usb(dev); + break; } return(ENXIO); } _at__at_ -124,6 +128,21 _at__at_ } +/* Disable USB SMI */ +static void +fixsmi_usb(device_t dev) +{ + uint32_t features; + + dev = pci_find_device(0x1002, 0x4385); + features = pci_read_config(dev, 0x64, 4); + if (features & (1 << 15)) { + printf("Disabling USB SMI on SB7xx\n"); + features &= ~(1 << 15); + pci_write_config(dev, 0x64, features, 4); + } +} + /* * Set the SYSTEM_IDLE_TIMEOUT to 80 ns on nForce2 systems to work * around a hang that is triggered when the CPU generates a very fast -- Andriy GaponReceived on Sun Sep 27 2009 - 11:10:28 UTC
This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:39:56 UTC