Date: Thu, 20 Oct 2011 13:44:33 +0200 From: =?iso-8859-1?Q?Marcus_M=FCller?= <znek@mulle-kybernetik.com> To: freebsd-usb@freebsd.org Subject: Re: bad performance of aue device when used with nVidia nForce MCP79 USB Message-ID: <F1524373-5589-4223-A462-85449A3212D5@mulle-kybernetik.com> In-Reply-To: <201110200911.21936.hselasky@c2i.net> References: <A54E761C-0326-4118-95B6-5386B1B729BC@mulle-kybernetik.com> <201110200911.21936.hselasky@c2i.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 20.10.2011, at 09:11, Hans Petter Selasky wrote:
> Also see "sysctl hw.usb.ehci".
root@creutzfeld:(/home/znek)# sysctl hw.usb.ehci
sysctl: unknown oid 'hw.usb.ehci'
Please note that im on 8.2-STABLE… I suspect hw.usb.ehci is not available here (or is this another bug?)
root@creutzfeld:(/home/znek)# sysctl hw.usb
hw.usb.no_boot_wait: 0
hw.usb.debug: 0
hw.usb.usb_lang_mask: 255
hw.usb.usb_lang_id: 9
hw.usb.template: 0
hw.usb.power_timeout: 30
hw.usb.no_pf: 0
hw.usb.no_cs_fail: 0
> The Nvidia chipsets are quirked due to what looks like hardware issues. See
> sys/dev/usb/controller/ehci_pci.c. This affects performance.
I commented out the specific quirks section in ehci_pci.c and now performance is as expected!
I didn't test any other USB devices on this board, yet. But the network device appears to be stable, thus I suspect the quirk setting based on vendor NVIDIA and NVIDIA2 is a bit too far fetched?
For reference, here is my patch:
Index: ehci_pci.c
===================================================================
--- ehci_pci.c (revision 226546)
+++ ehci_pci.c (working copy)
@@ -459,6 +459,7 @@
}
/* Doorbell feature workaround */
+/*
switch (pci_get_vendor(self)) {
case PCI_EHCI_VENDORID_NVIDIA:
case PCI_EHCI_VENDORID_NVIDIA2:
@@ -470,7 +471,7 @@
default:
break;
}
-
+*/
err = ehci_init(sc);
if (!err) {
err = device_probe_and_attach(sc->sc_bus.bdev);
Thanks,
Marcus
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?F1524373-5589-4223-A462-85449A3212D5>