On 1/4/21 8:56 AM, Lauri Kasanen wrote: > On Mon, 4 Jan 2021 08:49:55 -0700 > Jens Axboe <axboe@xxxxxxxxx> wrote: > >> On 1/4/21 8:43 AM, Lauri Kasanen wrote: >>> On Mon, 4 Jan 2021 08:40:10 -0700 >>> Jens Axboe <axboe@xxxxxxxxx> wrote: >>>> It should definitely get reviewed first. One easy question - there's no >>>> commit message in this one. It's a new driver, it should include some >>>> documentation on what it is, etc. >>> >>> It's already had one (rfc) round on linux-mips. Or do you mean by >>> others (who?) >> >> I mean the actual block driver from the block point of view. From >> a quick look, looks like there's no irq or anything, so it's >> strictly polled mode? Probably would make sense to push this out >> of the app path then, by using BLK_MQ_F_BLOCKING. > > It's very unclear to me what that flag does. I read all the docs > available, but nothing made it clear (same applies for almost all > BLK_MQ_*). It pushes the actual processing to a helper thread instead of doing it inline on submit. Generally used if the driver needs to block to submit IO, as per the name, but also useful for really slow hardware like this one. -- Jens Axboe
