none
NVMe Deallocate Test RRS feed

  • Question

  • I'm running the NVMe Deallocate test from the command line:

    trimperf /drivenumber 2 /logdriveletter c: /devicetype nvme /disksize 4000000000  /scenario compliance /unittest t /rangecount 1 /sizecount 255 /sizeunit sector

    I occasionally get errors like this:

    Start: Random ValidateTrim 1 range 255 Sector on 500ms intervals, TUID=
    Spawning 1 threads.
    2960 - Random ValidateTrim 1 range 255 Sector on 500ms intervals thread started
    Checking for previous data: Mismatch at byte 126975; Expected: 0xF8, Actual: 0x00
    Checking for 0x00         : Mismatch at byte 49151; Expected: 0x00, Actual: 0x82
    Checking for 0xFF         : Mismatch at byte 126975; Expected: 0xFF, Actual: 0x00
    Error: 0x0, Error 0x00000000
            Trim validation check failed at offset 1639333888, length 130560
            File=testsrc\driverstest\storage\wdk\trimperf\testcase.cpp Line=583
    2960 - Stop event requested. Iterations: 60

    The test is supposed to test for 1 of 3 outcomes - if none of them occurs, post a failure. After the trim, the data should be either the previous data, 0 or 0xFF. A shown here, when testing for the 0x00 value, it is looking at a totally different address. Shouldn't all 3 tests be done on data from the same location?

    Friday, January 16, 2015 9:07 PM