Re: kern/37826 (Kernel sees only one SATA channel (out of four) on Abit AN52 motherboard (nvidia nf520__single chip))

[David Holland <dholland-bugs%netbsd.org@localhost>]

On Mon, Mar 09, 2009 at 11:46:11PM -0600, Art Lemasters wrote:
 > > It is still attaching as viaide, and therefore still not working.
 > 
 >   The following was done with a fresh amd64 system (netbsd-5-0-RC2
 > only a few days old) with the NetBSD drive on the first channel and
 > a Linux drive plugged in on the third. Is that alright? Is the same
 > work occuring with amd64, as I assume?

The IDE drivers are machine-independent, yes.

I think I see what the problem likely is: there are four PCI IDs for
the MCP65 SATA controller, and only one of them is listed in
ahcisata_pci.c.

Try the following patch (with GENERIC):

Index: pci/ahcisata_pci.c
===================================================================
RCS file: /cvsroot/src/sys/dev/pci/ahcisata_pci.c,v
retrieving revision 1.13
diff -u -p -r1.13 ahcisata_pci.c
--- pci/ahcisata_pci.c  8 Dec 2008 15:35:23 -0000       1.13
+++ pci/ahcisata_pci.c  24 Mar 2009 21:06:33 -0000
@@ -54,6 +54,12 @@ __KERNEL_RCSID(0, "$NetBSD: ahcisata_pci
 static const struct pci_quirkdata ahci_pci_quirks[] = {
        { PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP65_SATA,
            AHCI_PCI_QUIRK_FORCE },
+       { PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP65_SATA2,
+           AHCI_PCI_QUIRK_FORCE },
+       { PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP65_SATA3,
+           AHCI_PCI_QUIRK_FORCE },
+       { PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP65_SATA4,
+           AHCI_PCI_QUIRK_FORCE },
        { PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP67_SATA,
            AHCI_PCI_QUIRK_FORCE },
        { PCI_VENDOR_NVIDIA, PCI_PRODUCT_NVIDIA_MCP73_AHCI_1,

If that doesn't work, what we'll need to know is what the PCI ID you
actually have is, which you can get via lspci or equivalent.

 >   The RC50.2 GENERIC, INSTALL with ACPI, INSTALL no ACPI and custom
 > dmesg files are below. There were some "NoMatch" lines after doing
 > the adjustkernel--some lines that I haven't noticed before. I don't
 > know whether they are of any consequence to any of your work, but
 > just in case,...

I don't think adjustkernel is expected to work on any currently
supported versions.

-- 
David A. Holland
dholland%netbsd.org@localhost