Skip to content

ARM64: Guest performance improvement during dirty

Jing Zhang kvm kvmarm 2 versions – Applied

Status from this message: Applied to next, thanks!

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v2 Latest3 patches 6 repliesApplied
  2. v13 patches 2 repliesSuperseded

Patches in v2 Applied

#Subject and filesSizeRepliesApplied
1/3KVM: arm64: Use read/write spin lock for MMU protectionarch/arm64/include/asm/kvm_host.h arch/arm64/kvm/mmu.c+20 -181 reply
2/3KVM: arm64: Add fast path to handle permission relaxation during dirty loggingarch/arm64/kvm/mmu.c+15 -21 reply
3/3KVM: selftests: Add vgic initialization for dirty log perf test for ARMtools/testing/selftests/kvm/dirty_log_perf_test.c+10 -01 reply

Reviews trailers per patch

All 3 patches have a Reviewed-by or Acked-by.

PatchFuad Tabba
1/3T-b R-b
2/3T-b R-b
3/3T-b R-b

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.

Read the v2 thread

Cover letter [PATCH v2 0/3] ARM64: Guest performance improvement during dirty

This patch is to reduce the performance degradation of guest workload during dirty logging on ARM64. A fast path is added to handle permission relaxation during dirty logging. The MMU lock is replaced with rwlock, by which all permision relaxations on leaf pte can be performed under the read lock. This greatly reduces the MMU lock contention during dirty logging. With this solution, the source guest workload performance degradation can be improved by more than 60%.

Problem:
* A Google internal live migration test shows that the source guest workload performance has >99% degradation for about 105 seconds, >50% degradation for about 112 seconds, >10% degradation for about 112 seconds on ARM64. This shows that most of the time, the guest workload degradtion is above 99%, which obviously needs some improvement compared to the test result on x86 (>99% for 6s, >50% for 9s, >10% for 27s).
* Tested H/W: Ampere Altra 3GHz, #CPU: 64, #Mem: 256GB, PageSize: 4K
* VM spec: #vCPU: 48, #Mem/vCPU: 4GB, PageSize: 4K, 2M hugepage backed

Analysis:
* We enabled CONFIG_LOCK_STAT in kernel and used dirty_log_perf_test to get

    the number of contentions of MMU lock and the "dirty memory time" on
    various VM spec. The "dirty memory time" is the time vCPU threads spent
    in KVM after fault. Higher "dirty memory time" means higher degradation
    to guest workload.
    '-m 2' specifies the mode "PA-bits:48,  VA-bits:48,  4K pages".
    By using test command
    ./dirty_log_perf_test -b 2G -m 2 -i 2 -s anonymous_hugetlb_2mb -v [#vCPU]
    Below are the results:
    +-------+------------------------+-----------------------+
    | #vCPU | dirty memory time (ms) | number of contentions |
    +-------+------------------------+-----------------------+
    | 1     | 926                    | 0                     |
    +-------+------------------------+-----------------------+
    | 2     | 1189                   | 4732558               |
    +-------+------------------------+-----------------------+
    | 4     | 2503                   | 11527185              |
    +-------+------------------------+-----------------------+
    | 8     | 5069                   | 24881677              |
    +-------+------------------------+-----------------------+
    | 16    | 10340                  | 50347956              |
    +-------+------------------------+-----------------------+
    | 32    | 20351                  | 100605720             |
    +-------+------------------------+-----------------------+
    | 64    | 40994                  | 201442478             |
    +-------+------------------------+-----------------------+

* From the test results above, the "dirty memory time" and the number of MMU lock contention scale with the number of vCPUs. That means all the dirty memory operations from all vCPU threads have been serialized by the MMU lock. Further analysis also shows that the permission relaxation during dirty logging is where vCPU threads get serialized.

Solution:
* On ARM64, there is no mechanism as PML (Page Modification Logging) and the dirty-bit solution for dirty logging is much complicated compared to the write-protection solution. The straight way to reduce the guest performance degradation is to enhance the concurrency for the permission fault path during dirty logging.
* In this patch, we only put leaf PTE permission relaxation for dirty

    logging under read lock, all others would go under write lock.
    Below are the results based on the fast path solution:
    +-------+------------------------+
    | #vCPU | dirty memory time (ms) |
    +-------+------------------------+
    | 1     | 965                    |
    +-------+------------------------+
    | 2     | 1006                   |
    +-------+------------------------+
    | 4     | 1128                   |
    +-------+------------------------+
    | 8     | 2005                   |
    +-------+------------------------+
    | 16    | 3903                   |
    +-------+------------------------+
    | 32    | 7595                   |
    +-------+------------------------+
    | 64    | 15783                  |
    +-------+------------------------+

* Furtuer analysis shows that there is another bottleneck caused by the

    setup of the test code itself. The 3rd commit is meant to fix that by
    setting up vgic in the test code. With the test code fix, below are
    the results which show better improvement.
    +-------+------------------------+
    | #vCPU | dirty memory time (ms) |
    +-------+------------------------+
    | 1     | 803                    |
    +-------+------------------------+
    | 2     | 843                    |
    +-------+------------------------+
    | 4     | 942                    |
    +-------+------------------------+
    | 8     | 1458                   |
    +-------+------------------------+
    | 16    | 2853                   |
    +-------+------------------------+
    | 32    | 5886                   |
    +-------+------------------------+
    | 64    | 12190                  |
    +-------+------------------------+
    All "dirty memory time" has been reduced by more than 60% when the
    number of vCPU grows.

* Based on the solution, the test results from the Google internal live migration test also shows more than 60% improvement with >99% for 30s, >50% for 58s and >10% for 76s.

---

* v1 -> v2
- Renamed flag name from use_mmu_readlock to logging_perm_fault.
- Removed unnecessary check for fault_granule to use readlock.
* RFC -> v1
- Rebase to kvm/queue, commit fea31d169094
(KVM: x86/pmu: Fix available_event_types check for REF_CPU_CYCLES event)
- Moved the fast path in user_mem_abort, as suggested by Marc.
- Addressed other comments from Marc.

[v1] https://lore.kernel.org/all/[email protected]
[RFC] https://lore.kernel.org/all/[email protected]

---

Jing Zhang (3):
KVM: arm64: Use read/write spin lock for MMU protection
KVM: arm64: Add fast path to handle permission relaxation during dirty
logging
KVM: selftests: Add vgic initialization for dirty log perf test for
ARM

 arch/arm64/include/asm/kvm_host.h             |  2 +
 arch/arm64/kvm/mmu.c                          | 49 ++++++++++++-------
 .../selftests/kvm/dirty_log_perf_test.c       | 10 ++++
 3 files changed, 43 insertions(+), 18 deletions(-)

base-commit: fea31d1690945e6dd6c3e89ec5591490857bc3d4

Signature · 1 line
2.34.1.703.g22d0c6ccf7-goog