On 07/26/2017 10:06 PM, Sitsofe Wheeler wrote: > Hi, > > On 27 July 2017 at 00:05, <sbates@xxxxxxxxxxxx> wrote: >> From: Stephen Bates <sbates@xxxxxxxxxxxx> >> >> Allow the user to set the probability of a pvsync2 IO being high >> priority. This allows the user to investigate performance when there >> is a mix of higher and lower priority IO. > > I'll just note this seems to be a feature that people are reluctant to > send upstream. I'm aware of at least one fio fork > (https://github.com/Seagate/fio/commits/master ) that adds this > feature and it's been asked about on the mailing list > (https://www.spinics.net/lists/fio/msg05487.html ) so even if it takes > a few rounds of polishing for you to get it in I wish you the best of > luck! Those are different, that's for passing in NCQ priority info. The pvsync2 hipri option sets RWF_HIPRI, which translates into polled IO for direct=1 in the kernel. That goes for both the github repo and the email on the list. It's pointless to add this to fio, before a kernel API exists for passing in priority information. RWF_HIPRI means polling now. Not that this is the right list, but I would suggest that someone wanting to pursue this on the kernel side would add it through the new fcntl F_SET_RW_HINT or F_SET_FILE_RW_HINT interface that was added in 4.13: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=c75b1d9421f80f4143e389d2d50ddfc8a28c8c35 Once that is done, then support can be added to fio. -- 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
