On Tue, Feb 18, 2025, Maxim Levitsky wrote: > On Tue, 2025-02-04 at 00:40 +0000, James Houghton wrote: > > By aging sptes locklessly with the TDP MMU and the shadow MMU, neither > > vCPUs nor reclaim (mmu_notifier_invalidate_range*) will get stuck > > waiting for aging. This contention reduction improves guest performance > > and saves a significant amount of Google Cloud's CPU usage, and it has > > valuable improvements for ChromeOS, as Yu has mentioned previously[1]. > > > > Please see v8[8] for some performance results using > > access_tracking_perf_test patched to use MGLRU. > > > > Neither access_tracking_perf_test nor mmu_stress_test trigger any > > splats (with CONFIG_LOCKDEP=y) with the TDP MMU and with the shadow MMU. > > > Hi, I have a question about this patch series and about the > access_tracking_perf_test: > > Some time ago, I investigated a failure in access_tracking_perf_test which > shows up in our CI. > > The root cause was that 'folio_clear_idle' doesn't clear the idle bit when > MGLRU is enabled, and overall I got the impression that MGLRU is not > compatible with idle page tracking. > > I thought that this patch series and the 'mm: multi-gen LRU: Have secondary > MMUs participate in MM_WALK' patch series could address this but the test > still fails. > > > For the reference the exact problem is: > > 1. Idle bits for guest memory under test are set via /sys/kernel/mm/page_idle/bitmap > > 2. Guest dirties memory, which leads to A/D bits being set in the secondary mappings. > > 3. A NUMA autobalance code write protects the guest memory. KVM in response > evicts the SPTE mappings with A/D bit set, and while doing so tells mm > that pages were accessed using 'folio_mark_accessed' (via kvm_set_page_accessed (*) ) > but due to MLGRU the call doesn't clear the idle bit and thus all the traces > of the guest access disappear and the kernel thinks that the page is still idle. > > I can say that the root cause of this is that folio_mark_accessed doesn't do > what it supposed to do. > > Calling 'folio_clear_idle(folio);' in MLGRU case in folio_mark_accessed() > will probably fix this but I don't have enough confidence to say if this is > all that is needed to fix this. If this is the case I can send a patch. My understanding is that the behavior is deliberate. Per Yu[1], page_idle/bitmap effectively isn't supported by MGLRU. [1] https://lore.kernel.org/all/CAOUHufZeADNp_y=Ng+acmMMgnTR=ZGFZ7z-m6O47O=CmJauWjw@xxxxxxxxxxxxxx > This patch makes the test pass (but only on 6.12 kernel and below, see below): > > diff --git a/mm/swap.c b/mm/swap.c > index 59f30a981c6f..2013e1f4d572 100644 > --- a/mm/swap.c > +++ b/mm/swap.c > @@ -460,7 +460,7 @@ void folio_mark_accessed(struct folio *folio) > { > if (lru_gen_enabled()) { > folio_inc_refs(folio); > - return; > + goto clear_idle_bit; > } > > if (!folio_test_referenced(folio)) { > @@ -485,6 +485,7 @@ void folio_mark_accessed(struct folio *folio) > folio_clear_referenced(folio); > workingset_activation(folio); > } > +clear_idle_bit: > if (folio_test_idle(folio)) > folio_clear_idle(folio); > } > > > To always reproduce this, it is best to use a patch to make the test run in a > loop, like below (although the test fails without this as well). .. > With the above patch applied, you will notice after 4-6 iterations that the > number of still idle pages soars: > > Populating memory : 0.798882357s ... > vCPU0: idle pages: 132558 out of 262144, failed to mark idle: 0 no pfn: 0 > Mark memory idle : 2.711946690s > Writing to idle memory : 0.302882502s > > ... > > (*) Turns out that since kernel 6.13, this code that sets accessed bit in the > primary paging structure, when the secondary was zapped was *removed*. I > bisected this to commit: > > 66bc627e7fee KVM: x86/mmu: Don't mark "struct page" accessed when zapping SPTEs > > So now the access_tracking_test is broken regardless of MGLRU. Just to confirm, do you see failures on 6.13 with MGLRU disabled? > Any ideas on how to fix all this mess? The easy answer is to skip the test if MGLRU is in use, or if NUMA balancing is enabled. In a real-world scenario, if the guest is actually accessing the pages that get PROT_NONE'd by NUMA balancing, they will be marked accessed when they're faulted back in. There's a window where page_idle/bitmap could be read between making the VMA PROT_NONE and re-accessing the page from the guest, but IMO that's one of the many flaws of NUMA balancing. That said, one thing is quite odd. In the failing case, *half* of the guest pages are still idle. That's quite insane. Aha! I wonder if in the failing case, the vCPU gets migrated to a pCPU on a different node, and that causes NUMA balancing to go crazy and zap pretty much all of guest memory. If that's what's happening, then a better solution for the NUMA balancing issue would be to affine the vCPU to a single NUMA node (or hard pin it to a single pCPU?).
