Re: [Qemu-devel] [PATCH v3 6/6] spapr_pci: populate ibm,loc-code

[David Gibson]

On Wed, May 06, 2015 at 11:44:59AM +0530, Nikunj A Dadhania wrote:
> Thomas Huth <address@hidden> writes:
> 
> > On Tue,  5 May 2015 14:23:56 +0530
> > Nikunj A Dadhania <address@hidden> wrote:
> >
> >> Each hardware instance has a platform unique location code.  The OF
> >> device tree that describes a part of a hardware entity must include
> >> the “ibm,loc-code” property with a value that represents the location
> >> code for that hardware entity.
> >> 
> >> Populate ibm,loc-code.
> >> 1) PCI passthru devices need to identify with its own ibm,loc-code
> >>    available on the host.
> >> 2) Emulated devices encode as following:
> >>    qemu_<name>:<phb-index>:<slot>.<fn>
> >> 
> >> Signed-off-by: Nikunj A Dadhania <address@hidden>
> >> ---
> >>  hw/ppc/spapr_pci.c | 86 
> >> +++++++++++++++++++++++++++++++++++++++++++++++-------
> >>  1 file changed, 75 insertions(+), 11 deletions(-)
> >> 
> >> diff --git a/hw/ppc/spapr_pci.c b/hw/ppc/spapr_pci.c
> >> index cbd5661..eacf0bd 100644
> >> --- a/hw/ppc/spapr_pci.c
> >> +++ b/hw/ppc/spapr_pci.c
> >> @@ -744,6 +744,70 @@ static AddressSpace *spapr_pci_dma_iommu(PCIBus *bus, 
> >> void *opaque, int devfn)
> >>      return &phb->iommu_as;
> >>  }
> >>  
> >> +static bool spapr_phb_vfio_get_devspec_value(PCIDevice *pdev, char 
> >> **value)
> >> +{
> >> +    char *host;
> >> +    char path[PATH_MAX];
> >> +
> >> +    host = object_property_get_str(OBJECT(pdev), "host", NULL);
> >> +    if (!host) {
> >> +        return false;
> >> +    }
> >> +
> >> +    snprintf(path, sizeof(path), "/sys/bus/pci/devices/%s/devspec", host);
> >> +    g_free(host);
> >> +
> >> +    return g_file_get_contents(path, value, NULL, NULL);
> >> +}
> >> +
> >> +static char *spapr_phb_vfio_get_loc_code(sPAPRPHBState *sphb,  PCIDevice 
> >> *pdev)
> >> +{
> >> +    char path[PATH_MAX], *buf = NULL;
> >> +
> >> +    /* We have a vfio host bridge lets get the path. */
> >> +    if (!spapr_phb_vfio_get_devspec_value(pdev, &buf)) {
> >> +        return NULL;
> >> +    }
> >> +
> >> +    snprintf(path, sizeof(path), "/proc/device-tree%s/ibm,loc-code", buf);
> >> +    g_free(buf);
> >> +
> >> +    if (g_file_get_contents(path, &buf, NULL, NULL)) {
> >> +        return buf;
> >> +    } else {
> >> +        return NULL;
> >> +    }
> >
> > Minor idea for an optimization: g_file_get_contents() should set buf to
> > NULL in case of errors anyway, so you could omit the "if" and
> > "return NULLL" and simply always "return buf" here.
> 
> Sure
> 
> >
> >> +}
> >> +
> >> +static char *spapr_phb_get_loc_code(sPAPRPHBState *sphb,  PCIDevice *pdev)
> >> +{
> >> +    char *path = g_malloc(PATH_MAX);
> >> +
> >> +    if (!path) {
> >> +        return NULL;
> >> +    }
> >> +
> >> +    /*
> >> +     * For non-vfio devices and failures make up the location code out
> >
> > This comment ("and failures") ...
> 
> Oh right, one option is to drop that...
> 
> >
> >> +     * of the name, slot and function.
> >> +     *
> >> +     *       qemu_<name>:<phb-index>:<slot>.<fn>
> >> +     */
> >> +    snprintf(path, PATH_MAX, "qemu_%s:%02d:%02d.%1d", pdev->name,
> >> +             sphb->index, PCI_SLOT(pdev->devfn), PCI_FUNC(pdev->devfn));
> >> +    return path;
> >> +}
> >> +
> >> +
> >> +static char *spapr_ibm_get_loc_code(sPAPRPHBState *sphb, PCIDevice *pdev)
> >> +{
> >> +    if (object_dynamic_cast(OBJECT(pdev), "vfio-pci") != NULL) {
> >> +        return spapr_phb_vfio_get_loc_code(sphb, pdev);
> >> +    } else {
> >> +        return spapr_phb_get_loc_code(sphb, pdev);
> >> +    }
> >
> > ... does not match quite the behavior here, as far as I can see. I
> > guess you also wanted to fall back to spapr_phb_get_loc_code() in case 
> > spapr_phb_vfio_get_loc_code() did not work as expected (eg. when the
> > access to the /sys or /proc filesystem failed)?
> 
> ... In current case, if getting vfio device loc_code fails, we would not
>  add any ibm,loc_code info. I thought that would be the proper
>  behaviour. As "qemu_" would indicate it as an emulated device, which is
>  not true. Or encode it as follows:
> 
> "vfio_<name>:<phb-index>:<slot>.<fn>"
> 
> So more checks would be needed here in that case:
> 
> if (vfio) {
>    buf = vfio_get_loc_code()
>    if (!buf) 
>       buf = cook_vfio_get_loc_code()
>    return buf;
> } else { ... }

If you can't determine the loc code, I think just leaving it out is
probably the best option for now.  We can add fancier fallbacks in
future if they seem like they make sense.

-- 
David Gibson                    | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au  | minimalist, thank you.  NOT _the_ _other_
                                | _way_ _around_!
http://www.ozlabs.org/~dgibson

pgpK0uow6nHtl.pgp
Description: PGP signature