Currently, only the generic parts of the NumericAddressedBackend are tested with the UnifiedBackendTest (using an ExceptionDummyBackend). This tests the accessors, but not the backends based on the NumericAddressedBackend itself.
To test the PcieBackend with the UnifiedBackendTest, a new feature needs to be introduced in the mtcadummy kernel module, which allows to put the dummy device into an exception state similar to a broken Pcie connection. Ideally, a sysfs interface should be introduced (keep in mind, proc is outdated...), which presents a file "force_exception" file. Several different values written to the file could control when to throw an exception ("never", "read", "write", "open", "always" etc.). The file should be write-accessible by the same Unix group as the device node (might be necessary to change udev rule for this?).
Once this is done, it should be relatively straight forward to implement a UnifiedBackendTest for the PcieBackend.
Currently, only the generic parts of the NumericAddressedBackend are tested with the UnifiedBackendTest (using an ExceptionDummyBackend). This tests the accessors, but not the backends based on the NumericAddressedBackend itself.
To test the PcieBackend with the UnifiedBackendTest, a new feature needs to be introduced in the mtcadummy kernel module, which allows to put the dummy device into an exception state similar to a broken Pcie connection. Ideally, a sysfs interface should be introduced (keep in mind, proc is outdated...), which presents a file "force_exception" file. Several different values written to the file could control when to throw an exception ("never", "read", "write", "open", "always" etc.). The file should be write-accessible by the same Unix group as the device node (might be necessary to change udev rule for this?).
Once this is done, it should be relatively straight forward to implement a UnifiedBackendTest for the PcieBackend.