On Mon, 10 Jan 2022, Martin K. Petersen wrote: > > Ok so `git blame` shows that Kent added this to md/raid5.c in > > c78afc6261b (Kent Overstreet 2013-07-11 22:39:53 -0700 7526) > > Yep. > > > Do you think we should leave the md-specific > > raid_partial_stripes_expensive setting and require users of RAID > > controllers to set the bit themselves in bcache---or---remove all > > raid_partial_stripes_expensive code and always treat writes as > > "expensive" when `opt_io` is defined? > > I'd prefer the latter since that was the very intent of exporting the > device topology in an abstract and protocol-independent fashion. > However, I don't know enough about bcache internals to know whether it > is always the right choice, what the trade-offs are, etc. I've not looked in detail, but I don't think its too difficult. Do you know if io_opt can be configured on scsi devices that do not have io_opt provided by the controller? If so, or if it would be easy to add, then configuring io_opt at the scsi layer is probably a better option so subsystems besides bcache would benefit. What do you think? -- Eric Wheeler > > -- > Martin K. Petersen Oracle Linux Engineering >
