On Fri, Oct 04, 2019 at 06:15:11AM -0700, Michel Lespinasse wrote: > Hi Jason, > > On Thu, Oct 3, 2019 at 5:26 PM Jason Gunthorpe <jgg@xxxxxxxx> wrote: > > Hurm, this is not entirely accurate. Most users do actually want > > overlapping and multiple ranges. I just studied this extensively: > > (Just curious, are you the person we discussed this with after the > Maple Tree talk at LPC 2019 ?) Possibly! > I think we have two separate API problems there: > - overlapping vs non-overlapping intervals (the interval tree API > supports overlapping intervals, but some users are confused about > this) I think we just have a bunch of confused drivers, ie the two drm drivers sure look confused to me. > - closed vs half-open interval definitions I'm not sure why this is a big problem.. We may actually just have bugs in handling the '-1' as it is supposed to be written as start + (size-1) so that start + size == ULONG_MAX+1 works properly. > > hfi1/mmu_rb definitely needs overlapping as it is dealing with > > userspace VA ranges under control of userspace. As do the other > > infiniband users. > > Do you have a handle on what usnic is doing with its intervals ? > usnic_uiom_insert_interval() has some complicated logic to avoid > having overlapping intervals, which is very confusing to me. I don't know why it is so complicated, but I can say that it is storing userspace VA's in that tree. I have some feeling this driver is trying to use the IOMMU to create a mirror of the userspace VA Userspace can request the HW be able to access any set of overlapping regions and so the driver must intersect all the ranges and compute a list of VA pages to IOMMU map. Just guessing. Jason
