On Thu, Jul 11, 2013 at 11:50 PM, Maxim Patlasov <MPatlasov@xxxxxxxxxxxxx> wrote: > On Thu, Jul 11, 2013 at 10:51 PM, Jan Kara <jack@xxxxxxx> wrote: [snipped] >> If I'm right in the above, then removing NR_WRITEBACK_TEMP would be a nice >> followup patch. > > I'd rather introduce the notion of trusted fuse filesystem. If system > administrator believe given fuse fs "trusted", it works w/o > strictlimit, but fuse daemon is supposed to notify the kernel > explicitly about threads related to processing writeback. The kernel > would raise a per-task flag for those threads. And, calculating > nr_dirty in balance_dirty_pages, we'd add NR_WRITEBACK_TEMP for all, > excepting tasks with the flag set. This is very simple and will work > perfectly. Yes, doing a trusted mode for fuse is a good idea, I think. And it should have a new filesystem type (can't think of a good name though, "fusetrusted" is a bit too long). Thanks, Miklos -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>