Re: Re: PCI BAR register space written with garbage in HVM guest.

[Dan Gora <dan.gora@gmail.com>]

On Tue, Mar 16, 2010 at 11:46 PM, Konrad Rzeszutek Wilk
<konrad.wilk@oracle.com> wrote:
> There is a snippet of QEMU that has that (hw/pass-through.c),
> function pt_pci_write_config:
>
> .. snip ..
>      ret = pci_write_block(pci_dev, address, (uint8_t *)&val, len);
>
>        if (!ret)
>            PT_LOG("Error: pci_write_block failed. return value[%d].\n",
> ret);
>    }
>
>    if (pm_state != NULL && pm_state->flags & PT_FLAG_TRANSITING)
>        /* set QEMUTimer */
>        qemu_mod_timer(pm_state->pm_timer,
>            (qemu_get_clock(rt_clock) + pm_state->pm_delay));

I turned on all the debugging in there and I see messages, but none
when the bad writes occur, which made me think that they were not
coming from there...

>> <snip>
>>
>> 5904  21:35:48 [    7f05d53f987b] read(16, "o\0\0\0", 4) = 4
>> 5904  21:35:48 [    7f05d53f97fb] write(16, "o\0\0\0", 4) = 4
>> 5904  21:35:48 [    7f05d53f97fb] write(6,
>> "\377\377\377\377\377\377\0\26>2\325d\10\0E\0\2@\0\354\0\0@\21w\302\0\0\0\0\377\377"...,
>> 590) = 256
>
> I would turn on all of those debug options and see if anything is
> happening.

done.. nothing :(  At least not during the error.

> Actually I would instrument all of the 'pci_write_block' calls.

I'll try that tomorrow...

> Can you attach debug to it? Or maybe run 'objdump' on the qemu-dm and
> see if the address correspond to some code in it?

Yes I can debug it, but really the only thing that I have to hang my
hat on is write().  So I figure that it must be coming from pciutils
since it's messing with /sys/bus/pci, but I don't really know much
about pciutils so wasn't sure what to grep around for.

I'll try and trace the pci_write_block calls and put gdb on it
tomorrow and see if anything pops up.

thanks,
dan