On Tue, Jan 31, 2023 at 10:02 AM David Matlack <dmatlack@xxxxxxxxxx> wrote: > > On Tue, Jan 31, 2023 at 9:46 AM Oliver Upton <oliver.upton@xxxxxxxxx> wrote: > > > > On Tue, Jan 31, 2023 at 01:18:15AM +0000, Sean Christopherson wrote: > > > On Mon, Jan 30, 2023, Oliver Upton wrote: > > > > I think that Marc's suggestion of having userspace configure this is > > > > sound. After all, userspace _should_ know the granularity of the backing > > > > source it chose for guest memory. > > > > > > > > We could also interpret a cache size of 0 to signal that userspace wants > > > > to disable eager page split for a VM altogether. It is entirely possible that > > > > the user will want a differing QoS between slice-of-hardware and > > > > overcommitted VMs. > > > > > > Maybe. It's also entirely possible that QoS is never factored in, e.g. if QoS > > > guarantees for all VMs on a system are better met by enabling eager splitting > > > across the board. > > > > > > There are other reasons to use module/kernel params beyond what Marc listed, e.g. > > > to let the user opt out even when something is on by default. x86's TDP MMU has > > > benefited greatly from downstream users being able to do A/B performance testing > > > this way. I suspect x86's eager_page_split knob was added largely for this > > > reason, e.g. to easily see how a specific workload is affected by eager splitting. > > > That seems like a reasonable fit on the ARM side as well. > > > > There's a rather important distinction here in that we'd allow userspace > > to select the page split cache size, which should be correctly sized for > > the backing memory source. Considering the break-before-make rules of > > the architecture, the only way eager split is performant on arm64 is by > > replacing a block entry with a fully populated table hierarchy in one > > operation. > > I don't see how this can be true if we are able to tolerate splitting > 2M pages. Splitting 2M pages inherently means 512 Break-Before-Make > operations per 1GiB region of guest physical memory. > > If we had a cache size of 1 and were splitting a 1GiB region, we would > then need to do 512+1 BBM per 1GiB region. If we can tolerate 512 per > 1GiB, why not 513? > > It seems more like the 513 cache size is more to optimize splitting > 1GiB pages. That's correct. Although there's also a slight benefit for the 2M huge-pages case as the 512 huge-pages can be split without having to walk down the tree from the root every time. > I agree it can turn those 513 int 1, but future versions > of the architecture also elide BBM requirements which is another way > to optimize 1GiB pages. There's also the CPU cost needed to walk down the tree from the root 513 times (same as above). > > > > AFAICT, you don't have this problem on x86, as the > > architecture generally permits a direct valid->valid transformation > > without an intermediate invalidation. Well, ignoring iTLB multihit :) > > > > So, the largest transformation we need to do right now is on a PUD w/ > > PAGE_SIZE=4K, leading to 513 pages as proposed in the series. Exposing > > that configuration option in a module parameter is presumptive that all > > VMs on a host use the exact same memory configuration, which doesn't > > feel right to me. > > > > -- > > Thanks, > > Oliver
