Re: [Qemu-devel] [PATCH 3/3] tests: enable virtio tests on SPAPR

[Laurent Vivier]

On 30/09/2016 12:18, Greg Kurz wrote:
> On Thu, 29 Sep 2016 19:15:07 +0200
> Laurent Vivier <address@hidden> wrote:
> 
>> but disable MSI-X tests on SPAPR as we can't check the result
>> (the memory region used on PC is not readable on SPAPR).
>>
>> Signed-off-by: Laurent Vivier <address@hidden>
>> ---
>>  tests/Makefile.include    |  3 ++-
>>  tests/libqos/virtio-pci.c | 22 ++++++++++++++++++++--
>>  tests/virtio-9p-test.c    | 11 ++++++++++-
>>  tests/virtio-blk-test.c   | 22 +++++++++++++++++-----
>>  tests/virtio-net-test.c   | 17 +++++++++++++++--
>>  tests/virtio-rng-test.c   |  7 ++++++-
>>  tests/virtio-scsi-test.c  | 10 +++++++++-
>>  7 files changed, 79 insertions(+), 13 deletions(-)
...
>> diff --git a/tests/virtio-9p-test.c b/tests/virtio-9p-test.c
>> index 28d7f5b..a73bccb 100644
>> --- a/tests/virtio-9p-test.c
>> +++ b/tests/virtio-9p-test.c
>> @@ -11,6 +11,7 @@
>>  #include "libqtest.h"
>>  #include "qemu-common.h"
>>  #include "libqos/libqos-pc.h"
>> +#include "libqos/libqos-spapr.h"
>>  #include "libqos/virtio.h"
>>  #include "libqos/virtio-pci.h"
>>  #include "standard-headers/linux/virtio_ids.h"
>> @@ -22,12 +23,20 @@ static char *test_share;
>>  
>>  static QOSState *qvirtio_9p_start(void)
>>  {
>> +    const char *arch = qtest_get_arch();
>> +    QOSState *qs = NULL;
>>      test_share = g_strdup("/tmp/qtest.XXXXXX");
>>      g_assert_nonnull(mkdtemp(test_share));
>>      const char *cmd = "-fsdev local,id=fsdev0,security_model=none,path=%s "
>>                        "-device virtio-9p-pci,fsdev=fsdev0,mount_tag=%s";
>>  
>> -    return qtest_pc_boot(cmd, test_share, mount_tag);
>> +    if (strcmp(arch, "i386") == 0 || strcmp(arch, "x86_64") == 0) {
>> +        qs = qtest_pc_boot(cmd, test_share, mount_tag);
>> +    } else if (strcmp(arch, "ppc64") == 0) {
>> +        qs = qtest_spapr_boot(cmd, test_share, mount_tag);
>> +    }
>> +
> 
> What about introducing a qtest_arch_boot() helper that does ^^ and
> 
> } else {
>     g_printerr("qtest_arch_boot() not supported for arch %s\n",
>                qtest_get_arch());
>     exit(EXIT_FAILURE);
> }
> 

The problem with adding a function like that is it will pull
$(libqos-pc-obj-y) and  $(libqos-spapr-obj-y) for every tests using it,
and for the moment we are pulling pc or spapr objects only if we need
them for the given test.

I think it explains why qtest_pc_boot() calls qtest_vboot() and we don't
have a generic qtest_boot() calling the architecture specific function.

I cc: John Snow as he has written the initial code for this.
("90e5add libqos: add pc specific interface")

Laurent