Re: [Qemu-devel] [PATCH] pci-testdev: add nodata test

["Michael S. Tsirkin" <mst@redhat.com>]

On Mon, Aug 31, 2015 at 04:01:11PM +0800, Jason Wang wrote:
> 
> 
> On 08/31/2015 03:28 PM, Michael S. Tsirkin wrote:
> > On Mon, Aug 31, 2015 at 11:04:42AM +0800, Jason Wang wrote:
> >>
> >> On 08/27/2015 06:30 PM, Michael S. Tsirkin wrote:
> >>> Add nodata test where data length is ignored.
> >>> Skip it for port IO since kvm does not support it there.
> >>>
> >>> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> >>> ---
> >>>  hw/misc/pci-testdev.c | 15 ++++++++++++---
> >>>  1 file changed, 12 insertions(+), 3 deletions(-)
> >>>
> >>> diff --git a/hw/misc/pci-testdev.c b/hw/misc/pci-testdev.c
> >>> index 26b9b86..b6e11d6 100644
> >>> --- a/hw/misc/pci-testdev.c
> >>> +++ b/hw/misc/pci-testdev.c
> >>> @@ -45,6 +45,7 @@ typedef struct IOTest {
> >>>  
> >>>  #define IOTEST_DATAMATCH 0xFA
> >>>  #define IOTEST_NOMATCH   0xCE
> >>> +#define IOTEST_NODATA  0xAB
> >> This is never used.
> >>
> >>>  
> >>>  #define IOTEST_IOSIZE 128
> >>>  #define IOTEST_MEMSIZE 2048
> >>> @@ -52,7 +53,8 @@ typedef struct IOTest {
> >>>  static const char *iotest_test[] = {
> >>>      "no-eventfd",
> >>>      "wildcard-eventfd",
> >>> -    "datamatch-eventfd"
> >>> +    "datamatch-eventfd",
> >>> +    "nodata-eventfd"
> >> For PIO test, "nodata-eventfd" actually means "nodata-noeventfd".
> > Not really - this test is skipped completely for PIO, as
> > kvm does not support it.
> 
> But I get:
> 
> ...
> portio-no-eventfd:pci-io 5024
> portio-wildcard-eventfd:pci-io 1731
> portio-datamatch-eventfd:pci-io 1731
> portio-nodata-eventfd:pci-io 4967
> ...

Weird - I thought I fixed it, but now I see it too.
I'll add code to drop portio-nodata-eventfd.

> >>>  };
> >>>  
> >>>  static const char *iotest_type[] = {
> >>> @@ -261,12 +263,19 @@ static void pci_testdev_realize(PCIDevice *pci_dev, Error **errp)
> >>>          memcpy(test->hdr->name, name, strlen(name) + 1);
> >>>          g_free(name);
> >>>          test->hdr->offset = cpu_to_le32(IOTEST_SIZE(i) + i * IOTEST_ACCESS_WIDTH);
> >>> -        test->size = IOTEST_ACCESS_WIDTH;
> >>> -        test->match_data = strcmp(IOTEST_TEST(i), "wildcard-eventfd");
> >>> +        test->size = strcmp(IOTEST_TEST(i), "nodata-eventfd") ? IOTEST_ACCESS_WIDTH : 0;
> >>> +
> >>> +        test->match_data = strcmp(IOTEST_TEST(i), "wildcard-eventfd") &&
> >>> +                             strcmp(IOTEST_TEST(i), "nodata-eventfd");
> >>>          test->hdr->test = i;
> >>>          test->hdr->data = test->match_data ? IOTEST_DATAMATCH : IOTEST_NOMATCH;
> >>>          test->hdr->width = IOTEST_ACCESS_WIDTH;
> >>>          test->mr = IOTEST_REGION(d, i);
> >>> +
> >>> +        if (!test->size && !IOTEST_IS_MEM(i)) {
> >>> +            test->hasnotifier = false;
> >>> +            continue;
> >>> +        }
> >>>          if (!strcmp(IOTEST_TEST(i), "no-eventfd")) {
> >>>              test->hasnotifier = false;
> >>>              continue;
> >> Since you're at this, how about introduce a test that mixes datamatch
> >> and wildcard?
> > How do you mean? wildcard means ignore data ...
> >
> 
> I mean for example having both datamatch and wildcard eventfds in a
> single subtest. E.g in "datamatch-eventfd", registering 1 or more
> wildcard eventfd before doing the real tests.

For same address? We can't do that - kvm won't let us.

-- 
MST