> > > There is surely no need to duplicate all that. > > > > Yeah, we could teach generic_writepages() to conditionally not submit > > for io just test/clear pte dirtyness. > > > > Maybe that would be somewhat cleaner, dunno. > > > > Then there are the ram backed filesystems, which don't have dirty > > accounting and radix trees, and for which this pte walking is still > > needed to provide semantics consistent with normal filesystems. > > hm. > > I don't know how important all this is, really - we've had this bug for > ever and presumably we've already trained everyone to work around it. > > What usage scenarios are people actually hurting from? Is there anything > interesting in the mysterious Novell Bugzilla #206431?
That's just a failing LTP testcase, not quite real life ;) But Peter Staubach says a RH custumer has files written thorugh mmap, which are not being backed up. > Perhaps we can get away with doing something half-assed which covers most > requirements... OK. At least I can split the patch into two half asses. The big question is tmpfs and friends. Those won't get any timestamp update without additional page table walking. Miklos - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
