On Tue, Jun 21, 2016 at 09:02:39AM +0200, Alexander Gordeev wrote: > > > I tried pci-testdev against ARM and got the device semi-operational. > > > It is still to investigate, but that could be addressed separately. > > > > I'd like to either have this resolved, or at least understood it > > before we accept the series. If you drop your pci-testdev > > implementation (patch 11/12) into x86/vmexit.c, in order to replace > > the implementation there, does pci-testdev still work? I.e. have > > you checked if x86 works both with its current implementation and > > yours? > > pci-testdev does not work. It does not work for x86/vmexit.c as well, > but unlike my implementation x86/vmexit.c does not check status of > device writes and therefore fails to notice that the writes do not > take place for "wildcard-eventfd" and "datamatch-eventfd" cases. > CONFIG_EVENTFD is set, so I am not sure what is happening here. So I could define PCI_TESTDEV_NUM_TESTS to 1 and hence skip failing tests. But on the other hand these failures are exposing a problem in QEMU and therefore fulfilling kvm-unit-tests purpose :) > pci-testdev also does not implement 16 and 32 bit accesses, although > x86/vmexit.c and pci/pci-testdev.c do have code to test these two. > It is not a bug though. -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
