On Fri, Aug 18, 2017 at 09:35:01AM -0600, Jens Axboe wrote: > On 08/18/2017 09:13 AM, Shaohua Li wrote: > > discard request usually is very big and easily use all bandwidth budget > > of a cgroup. discard request size doesn't really mean the size of data > > written, so it doesn't make sense to account it into bandwidth budget. > > This patch ignores discard requests size. It makes sense to account > > discard request into iops budget though. > > Some (most) devices to touch media for a discard operation, but the > cost tends to be fairly constant and independent of discard size. > Would it make sense to just treat it as a constant cost? Zero > cost seems wrong. that would be hard to find the cost. Would this make sense? min_t(unsigned int, bio->bi_iter.bi_size, queue_max_sectors(q) << 9)
