On Sat, Jul 11, 2015 at 12:20 PM, Sitsofe Wheeler <sitsofe@xxxxxxxxx> wrote: > > I guess what might be needed to cope with the "stale overwrite / data > from another part of the disk" scenario is something akin to dt's > pattern=iot (see > http://www.scsifaq.org/RMiller_Tools/ftp/dt/dt-UsersGuide.pdf ) so > each block has a different pattern based on LBA. You still need to > change the seed of the LBA based pattern on each pass and you'd need > to know what THAT seed was if you abort partway through a run if you > wanted to verify things later... > > The advantage writing an exact "headerless" pattern has is that it can > verify patterns written by other tools (this makes it easier to verify > the operation of commands like WRITE SAME without having to synthesise > a header). > > -- > Sitsofe | http://sucs.org/~sits/ Sitsofa, I do agree on your point about the header less pattern. Your other point about the LBA dependent pattern has been kind of implemented in the fio by Chromium Dev team. They do write and verify "numberio" and "rand_seed" in each LBA. NumberIO is equal to the number of writes that has been written so far on the device. Obviously this value is dependent to the rand_seed. --Alireza -- 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
