Source-Changes archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: CVS commit: src/sys/arch/x86/pci



On Feb 4, 2008, at 6:19 PM, David Young wrote:

On Mon, Feb 04, 2008 at 02:01:38PM +0100, Julio M. Merino Vidal wrote:
On Jan 14, 2008, at 7:44 PM, David Young wrote:


Module Name:    src
Committed By:   dyoung
Date:           Mon Jan 14 18:44:18 UTC 2008

Modified Files:
        src/sys/arch/x86/pci: pci_machdep.c

Log Message:
KASSERT() that reads/writes from/to PCI configuration space are
aligned on 32-bit boundaries.

This breaks the mpt(4) driver and does not let NetBSD/amd64 boot
successfully in VMware Fusion.  The machine seems to emulate a
53C1030 chip, and sys/dev/pci/mpt_pci.c has:

        if ((PCI_PRODUCT(pa->pa_id) == PCI_PRODUCT_SYMBIOS_1030) &&
            (PCI_REVISION(pa->pa_class) < 0x08)) {
                aprint_normal("%s: applying 1030 quirk\n",
                    mpt->sc_dev.dv_xname);
                reg = pci_conf_read(pa->pa_pc, pa->pa_tag, 0x6a);
                reg &= 0x8f;
                pci_conf_write(pa->pa_pc, pa->pa_tag, 0x6a, reg);
        }

Note the 0x6a register value in there, which will trigger the
assertion you added.  I've been looking around to see if that value
is correct, but the only thing I found is the Linux code which does
the same as us (maybe we borrowed the fix from there in the first
place, or the other way around).

Any idea on how to resolve this?

There are two ways to resolve it. You can delete that code, since both
the read and the write probably failed.  Or, you can read 0x68, apply
mask 0x008f0000, and write back to 0x68, which seems to be intended.

Hm. 0xbf0000 or 0xbfffff? The fix joerg@ committed used the latter mask.



Home | Main Index | Thread Index | Old Index