On 3/21/23 17:00, Ming Lei wrote: >>> But I guess it is hard to maintain bio order, especially md/dm is >>> involved. >> >> It is not that hard once you get rid of plugging, which we did. So far, with >> everything we support, we are not detecting any issues and we test weekly, every rc. > > I see, but I meant current->bio_list, see the following bio order issue: > > https://lore.kernel.org/linux-block/1609233522-25837-1-git-send-email-dannyshih@xxxxxxxxxxxx/ Interesting. We have never seen such issue in practice with the device mappers that support zoned devices. But it seems interesting to try to find a use case that could trigger this. Will look into it, and if everything is fixed, it would still be a good regression test for blocktest. -- Damien Le Moal Western Digital Research
