On Mon, Jun 8, 2020 at 12:35 AM Christoph Hellwig <hch@xxxxxxxxxxxxx> wrote: > > On Fri, Jun 05, 2020 at 03:28:19PM -0700, Harshad Shirwadkar wrote: > > Ext4 before this patch can issue discards without respecting block > > device's discard alignment. Such a discard results in EIO and > > kernel logs. > > No, that is not how discard works. The granularity is a hint and > blk_bio_discard_split already does all the work to align to it. If > you have a make_request based driver that doesn't do that you need > to fix that driver instead. Thanks for the clarification, my commit message was wrong. Fixing the driver instead is the right solution here. But, if doing discard alignment in Ext4 can increase the chances of discards being successful, is there any harm in doing it in Ext4 too?
