Re: [PATCH v5 15/15] tests/qtest: Add s390x PCI boot test to cdrom-test.c

[Thomas Huth <thuth@redhat.com>]

On 09/03/2026 17.53, Thomas Huth wrote:
> On 09/03/2026 17.49, Thomas Huth wrote:
>> On 09/03/2026 01.36, jrossi@linux.ibm.com wrote:
>>> From: Jared Rossi <jrossi@linux.ibm.com>
>>>
>>> Add a rudimentary test for s390x IPL to verify that a guest may boot using
>>> virtio-blk-pci device.
>>>
>>> Reviewed-by: Thomas Huth <thuth@redhat.com>
>>> Signed-off-by: Jared Rossi <jrossi@linux.ibm.com>
>>> ---
>>>   tests/qtest/cdrom-test.c | 7 +++++++
>>>   1 file changed, 7 insertions(+)
>>>
>>> diff --git a/tests/qtest/cdrom-test.c b/tests/qtest/cdrom-test.c
>>> index 56e2d283a9..a65854d2bc 100644
>>> --- a/tests/qtest/cdrom-test.c
>>> +++ b/tests/qtest/cdrom-test.c
>>> @@ -246,6 +246,13 @@ static void add_s390x_tests(void)
>>>                               "-drive if=none,id=d2,media=cdrom,file=",
>>>                               test_cdboot);
>>>       }
>>> +    if (qtest_has_device("virtio-blk-pci")) {
>>> +        qtest_add_data_func("cdrom/boot/pci-bus-with-bootindex",
>>> +                            "-device virtio-scsi -device virtio-serial "
>>> +                            "-device virtio-blk-pci,drive=d1,bootindex=1 "
>>> +                            "-drive if=none,id=d1,media=cdrom,file=",
>>> +                            test_cdboot);
>>> +    }
>>>   }
>>
>>   Hi Jared!
>>
>> While testing your series, it seems like the cdrom-test is now triggering 
>> a segmentation fault in QEMU here ... not sure whether I did something 
>> wrong on my side while building the bios, or whether there is a bug 
>> somewhere? Does this qtest work for you with this v5 series?
> 
> Ah, it's working on my s390x LPAR, but it's failing on my x86 laptop ... 
> maybe there is some endianness issue left somewhere?

Yes, it's a missing swap of endianness data, this patch fixes it:

diff --git a/hw/s390x/ipl.c b/hw/s390x/ipl.c
--- a/hw/s390x/ipl.c
+++ b/hw/s390x/ipl.c
@@ -524,9 +524,9 @@ static bool s390_build_iplb(DeviceState *dev_st, 
IplParameterBlock *iplb)

          switch (devtype) {
          case PCI_DEVTYPE_VIRTIO:
-            iplb->len = S390_IPLB_MIN_PCI_LEN;
+            iplb->len = cpu_to_be32(S390_IPLB_MIN_PCI_LEN);
              iplb->pbt = S390_IPL_TYPE_PCI;
-            iplb->pci.fid = pbdev->fid;
+            iplb->pci.fid = cpu_to_be32(pbdev->fid);
              break;
          default:
              return false;

If you agree, I can squash this into patch 13.

  Thomas