On Fri, Sep 06 2013, Juan Casse wrote: > On Fri, Sep 6, 2013 at 7:46 AM, Jens Axboe <axboe@xxxxxxxxx> wrote: > ... > > Please take a look at how this is done for the experimental_verify. The > > writes are skipped much sooner and generically, there's no need to > > enforce any sort of specific IO engine or verify type. > > From your reply I understood that you want to skip do_io() all > together and call a function that will somehow replicate the workload > and compute the numberio for each block. Is that what you meant? Yes, you want the logic to be in the realm of do_verify(). See the experimental verify. It basically rewinds the various random/lfsr dials and replays, skips writes, etc. > My patch uses do_io() to initialize the vhdr_meta data structures. I > can try to replicate that in do_verify() instead, but that seems more > fragile. Honestly, at this point, not sure what change you are trying to accomplish there... But if you send your suggested change, I'll take a look and comment. If not, please step back a bit and explain to me exactly which logic you are looking for. I think that will help both of us and will get us to a better place on the implementation and changes. -- Jens Axboe -- To unsubscribe from this list: send the line "unsubscribe fio" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
