On Sun 03-11-19 22:09:03, John Hubbard wrote: > On 11/3/19 8:34 PM, Hillf Danton wrote: > > future, we have options for instance that gupers periodically release > > their references and re-pin pages after data sync the same way as the > > current flusher does. > > > > That's one idea. I don't see it as viable, given the behavior of, say, > a compute process running OpenCL jobs on a GPU that is connected via > a network or Infiniband card--the idea of "pause" really looks more like > "tear down the complicated multi-driver connection, writeback, then set it > all up again", I suspect. (And if we could easily interrupt the job, we'd > probably really be running with a page-fault-capable GPU plus and IB card > that does ODP, plus HMM, and we wouldn't need to gup-pin anyway...) > > Anyway, this is not amenable to quick fixes, because the problem is > a couple of missing design pieces. Which we're working on putting in. > But meanwhile, smaller changes such as this one are just going to move > the problems to different places, rather than solving them. So it's best > not to do that. Yeah, fully agreed here. Quick half baked fixes will make the current messy situation even worse... Honza -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
