| Message ID | 20221107215644.1895162-1-oliver.upton@linux.dev (mailing list archive) |
|---|---|
| Headers | show |
| Series | KVM: arm64: Parallel stage-2 fault handling | expand |
On Mon, 7 Nov 2022 21:56:30 +0000, Oliver Upton wrote: > Presently KVM only takes a read lock for stage 2 faults if it believes > the fault can be fixed by relaxing permissions on a PTE (write unprotect > for dirty logging). Otherwise, stage 2 faults grab the write lock, which > predictably can pile up all the vCPUs in a sufficiently large VM. > > Like the TDP MMU for x86, this series loosens the locking around > manipulations of the stage 2 page tables to allow parallel faults. RCU > and atomics are exploited to safely build/destroy the stage 2 page > tables in light of multiple software observers. > > [...] I've gone over this for quite a while, and while I'm still sh*t scared about it, I've decided to let it simmer in -next for a bit. If anything goes wrong or that someone spots something ugly, it will be easy to simply drop the branch. For simple fixes, they can go on top. [01/14] KVM: arm64: Combine visitor arguments into a context structure commit: dfc7a7769ab7f2a2f629c673717ef1fa7b63aa42 [02/14] KVM: arm64: Stash observed pte value in visitor context commit: 83844a2317ecad935f6735abd854e4bf3f757040 [03/14] KVM: arm64: Pass mm_ops through the visitor context commit: 2a611c7f87f26cca405da63a57f06d0e4dc14240 [04/14] KVM: arm64: Don't pass kvm_pgtable through kvm_pgtable_walk_data commit: fa002e8e79b3f980455ba585c1f47b26680de5b9 [05/14] KVM: arm64: Add a helper to tear down unlinked stage-2 subtrees commit: 8e94e1252cc054bb31fd3e9a15235cd831970ec1 [06/14] KVM: arm64: Use an opaque type for pteps commit: 6b91b8f95cadd3441c056182daf9024475ac4a91 [07/14] KVM: arm64: Tear down unlinked stage-2 subtree after break-before-make commit: 5c359cca1faf6d7671537fe1c240e8668467864d [08/14] KVM: arm64: Protect stage-2 traversal with RCU commit: c3119ae45dfb6038ca458ab5ba7a9fba2810845b [09/14] KVM: arm64: Atomically update stage 2 leaf attributes in parallel walks commit: ca5de2448c3b4c018fe3d6223df8b59068be1cd7 [10/14] KVM: arm64: Split init and set for table PTE commit: 331aa3a0547d1c794587e0df374d13b16645e832 [11/14] KVM: arm64: Make block->table PTE changes parallel-aware commit: 0ab12f3574db6cb432917a667f9392a88e8f0dfc [12/14] KVM: arm64: Make leaf->leaf PTE changes parallel-aware commit: 946fbfdf336b811479e024136c7cabc00157b6b9 [13/14] KVM: arm64: Make table->block changes parallel-aware commit: af87fc03cfdf6893011df419588d27acdfb9c197 [14/14] KVM: arm64: Handle stage-2 faults in parallel commit: 1577cb5823cefdff4416f272a88143ee933d97f5 Fingers crossed, M.
Presently KVM only takes a read lock for stage 2 faults if it believes the fault can be fixed by relaxing permissions on a PTE (write unprotect for dirty logging). Otherwise, stage 2 faults grab the write lock, which predictably can pile up all the vCPUs in a sufficiently large VM. Like the TDP MMU for x86, this series loosens the locking around manipulations of the stage 2 page tables to allow parallel faults. RCU and atomics are exploited to safely build/destroy the stage 2 page tables in light of multiple software observers. Patches 1-4 clean up the context associated with a page table walk / PTE visit. This is helpful for: - Extending the context passed through for a visit - Building page table walkers that operate outside of a kvm_pgtable context (e.g. RCU callback) Patches 5-7 clean up the stage-2 map walkers by calling a helper to tear down removed tables. There is a small improvement here in that a broken PTE is replaced more quickly, as page table teardown happens afterwards. Patch 8 sprinkles in RCU to the page table walkers, punting the teardown of removed tables to an RCU callback. Patches 9-13 implement the meat of this series, extending the 'break-before-make' sequence with atomics to realize locking on PTEs. Effectively a cmpxchg() is used to 'break' a PTE, thereby serializing changes to a given PTE. Finally, patch 14 flips the switch on all the new code and starts grabbing the read side of the MMU lock for stage 2 faults. Applies to 6.1-rc3. Tested with KVM selftests, kvm-unit-tests, and live migrating a 24 vCPU, 96GB VM that was running a Debian install. Confirmed all stage-2 table memory was freed by checking the SecPageTables stat in meminfo. Branch available at: https://github.com/oupton/linux kvm-arm64/parallel_mmu benchmarked with dirty_log_perf_test, scaling from 1 to 48 vCPUs with 4GB of memory per vCPU backed by THP. ./dirty_log_perf_test -s anonymous_thp -m 2 -b 4G -v ${NR_VCPUS} Time to dirty memory: +-------+----------+-------------------+ | vCPUs | 6.1-rc3 | 6.1-rc3 + series | +-------+----------+-------------------+ | 1 | 0.87s | 0.93s | | 2 | 1.11s | 1.16s | | 4 | 2.39s | 1.27s | | 8 | 5.01s | 1.39s | | 16 | 8.89s | 2.07s | | 32 | 19.90s | 4.45s | | 48 | 32.10s | 6.23s | +-------+----------+-------------------+ It is also worth mentioning that the time to populate memory has improved: +-------+----------+-------------------+ | vCPUs | 6.1-rc3 | 6.1-rc3 + series | +-------+----------+-------------------+ | 1 | 0.21s | 0.17s | | 2 | 0.26s | 0.23s | | 4 | 0.39s | 0.31s | | 8 | 0.68s | 0.39s | | 16 | 1.26s | 0.53s | | 32 | 2.51s | 1.04s | | 48 | 3.94s | 1.55s | +-------+----------+-------------------+ v4 -> v5: - Fix an obvious leak of table memory (Ricardo) v3 -> v4: - Fix some type conversion misses caught by sparse (test robot) - Squash RCU locking and RCU callback patches together into one (Sean) - Commit message nits (Sean) - Take a pointer to kvm_s2_mmu in stage2_try_break_pte(), in anticipation of eager page splitting (Ricardo) v3: https://lore.kernel.org/kvmarm/20221027221752.1683510-1-oliver.upton@linux.dev/ v4: https://lore.kernel.org/kvmarm/20221103091140.1040433-1-oliver.upton@linux.dev/ Oliver Upton (14): KVM: arm64: Combine visitor arguments into a context structure KVM: arm64: Stash observed pte value in visitor context KVM: arm64: Pass mm_ops through the visitor context KVM: arm64: Don't pass kvm_pgtable through kvm_pgtable_walk_data KVM: arm64: Add a helper to tear down unlinked stage-2 subtrees KVM: arm64: Use an opaque type for pteps KVM: arm64: Tear down unlinked stage-2 subtree after break-before-make KVM: arm64: Protect stage-2 traversal with RCU KVM: arm64: Atomically update stage 2 leaf attributes in parallel walks KVM: arm64: Split init and set for table PTE KVM: arm64: Make block->table PTE changes parallel-aware KVM: arm64: Make leaf->leaf PTE changes parallel-aware KVM: arm64: Make table->block changes parallel-aware KVM: arm64: Handle stage-2 faults in parallel arch/arm64/include/asm/kvm_pgtable.h | 92 +++- arch/arm64/kvm/hyp/nvhe/mem_protect.c | 21 +- arch/arm64/kvm/hyp/nvhe/setup.c | 22 +- arch/arm64/kvm/hyp/pgtable.c | 628 ++++++++++++++------------ arch/arm64/kvm/mmu.c | 53 ++- 5 files changed, 466 insertions(+), 350 deletions(-) base-commit: 30a0b95b1335e12efef89dd78518ed3e4a71a763