On 27/02/2019 10.15, Bjoern Walk wrote:
> With further testing...
>
> Thomas Huth <thuth@xxxxxxxxxx> [2019-02-26, 12:02PM +0100]:
>> When running virt-host-validate on an s390x host, the tool currently
>> warns that it is "Unknown if this platform has IOMMU support".
>> We can use the common check for entries in sys/kernel/iommu_groups here,
>> too, but it only makes sense to check it if there are also PCI devices
>> available. It's also common on s390x that there are no PCI devices
>> assigned to the LPAR, and in that case there is no need for the
>> PCI-related IOMMU, so without PCI devices we should simply skip this
>> test.
>>
>> Signed-off-by: Thomas Huth <thuth@xxxxxxxxxx>
>> ---
>> tools/virt-host-validate-common.c | 11 +++++++++--
>> 1 file changed, 9 insertions(+), 2 deletions(-)
>>
>> diff --git a/tools/virt-host-validate-common.c b/tools/virt-host-validate-common.c
>> index 73e3bdb..75c7cfc 100644
>> --- a/tools/virt-host-validate-common.c
>> +++ b/tools/virt-host-validate-common.c
>> @@ -337,7 +337,7 @@ int virHostValidateIOMMU(const char *hvname,
>> virBitmapPtr flags;
>> struct stat sb;
>> const char *bootarg = NULL;
>> - bool isAMD = false, isIntel = false, isPPC = false;
>> + bool isAMD = false, isIntel = false, isPPC, isS390;
>> flags = virHostValidateGetCPUFlags();
>>
>> if (flags && virBitmapIsBitSet(flags, VIR_HOST_VALIDATE_CPU_FLAG_VMX))
>> @@ -348,6 +348,7 @@ int virHostValidateIOMMU(const char *hvname,
>> virBitmapFree(flags);
>>
>> isPPC = ARCH_IS_PPC64(virArchFromHost());
>> + isS390 = ARCH_IS_S390(virArchFromHost());
>>
>> if (isIntel) {
>> virHostMsgCheck(hvname, "%s", _("for device assignment IOMMU support"));
>> @@ -375,6 +376,12 @@ int virHostValidateIOMMU(const char *hvname,
>> }
>> } else if (isPPC) {
>> /* Empty Block */
>> + } else if (isS390) {
>> + /* On s390x, we skip the IOMMU check if there are no PCI devices
>> + * (which is quite usual on s390x) */
>> + if (stat("/sys/bus/pci/devices", &sb) < 0 || !S_ISDIR(sb.st_mode) ||
>> + sb.st_nlink <= 2)
>
> ... this doesn't seem to work, no? The files in /sys/bus/pci/devices are
> actually soft-links and are not counted by stat, am I wrong? At least on
> my setup with a PCI device and populated sysfs it will still skip the
> check.
Bummer. Stupid thinko. And I don't have a real PCI card for testing
here, so I did not notice. So I guess I have to use opendir() +
readdir() instead to find out if there are entries in this directory?
Thomas
Attachment:
signature.asc
Description: OpenPGP digital signature
-- libvir-list mailing list libvir-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/libvir-list
