mbox series

[RFC,0/3] KVM/ARM: Fix >256 vcpus

Message ID 20190827160554.30995-1-eric.auger@redhat.com
Headers show
Series KVM/ARM: Fix >256 vcpus | expand

Message

Eric Auger Aug. 27, 2019, 4:05 p.m. UTC
Since 4.18, KVM/ARM exposes a KVM_MAX_VCPUS equal to 512. However it was
reported [1] that a VM with more than 256 vcpus cannot be launched. 5.4
is about to fix the situation with 2 patches:
- one upgrade of the KVM_IRQ_LINE API [2] supporting a vcpu id encoded
  on 12 bits,
- the reduction of KVM IO devices consumed by each GICv3 redistributor [3]

This series uses the new KVM_IRQ_LINE API and also checks the associated
capability (KVM_CAP_ARM_IRQ_LINE_LAYOUT_2) in machvirt.

Without the series, as soon as the -smp arguments exceeds 256, QEMU exits
with "kvm_set_irq: Invalid argument".

Best Regards

Eric

References:
[1] Can we boot a 512U kvm guest?
    https://patchwork.kernel.org/patch/11091501/
[2] [PATCH] KVM: arm/arm64: vgic: Allow more than 256 vcpus for KVM_IRQ_LINE
    https://patchwork.kernel.org/patch/11099609/
[3] [PATCH] KVM: arm/arm64: vgic: Use a single IO device per redistributor
    https://patchwork.kernel.org/patch/11112141/

This series can be found at:
https://github.com/eauger/qemu/tree/v4.1.0-256fix-rfc-v1

Eric Auger (3):
  linux headers: update for KVM_CAP_ARM_IRQ_LINE_LAYOUT_2
  intc/arm_gic: Support PPI injection for more than 256 vpus
  virt: Check KVM_CAP_ARM_IRQ_LINE_LAYOUT_2 for smp_cpus > 256

 hw/arm/virt.c                                |  7 +++++
 hw/intc/arm_gic_kvm.c                        | 10 ++++---
 include/standard-headers/asm-x86/bootparam.h |  2 ++
 include/standard-headers/asm-x86/kvm_para.h  |  1 +
 include/standard-headers/linux/ethtool.h     |  2 ++
 include/standard-headers/linux/pci_regs.h    |  4 +++
 include/standard-headers/linux/virtio_ids.h  |  1 +
 include/standard-headers/linux/virtio_pmem.h |  6 ++---
 linux-headers/asm-arm/kvm.h                  | 16 ++++++++++-
 linux-headers/asm-arm/unistd-common.h        |  2 ++
 linux-headers/asm-arm64/kvm.h                | 21 ++++++++++++++-
 linux-headers/asm-generic/mman-common.h      | 15 ++++++-----
 linux-headers/asm-generic/mman.h             | 10 +++----
 linux-headers/asm-generic/unistd.h           |  8 +++++-
 linux-headers/asm-mips/unistd_n32.h          |  1 +
 linux-headers/asm-mips/unistd_n64.h          |  1 +
 linux-headers/asm-mips/unistd_o32.h          |  1 +
 linux-headers/asm-powerpc/mman.h             |  6 +----
 linux-headers/asm-powerpc/unistd_32.h        |  2 ++
 linux-headers/asm-powerpc/unistd_64.h        |  2 ++
 linux-headers/asm-s390/unistd_32.h           |  2 ++
 linux-headers/asm-s390/unistd_64.h           |  2 ++
 linux-headers/asm-x86/kvm.h                  | 28 +++++++++++++++-----
 linux-headers/asm-x86/unistd_32.h            |  2 ++
 linux-headers/asm-x86/unistd_64.h            |  2 ++
 linux-headers/asm-x86/unistd_x32.h           |  2 ++
 linux-headers/linux/kvm.h                    | 12 ++++++---
 linux-headers/linux/psp-sev.h                |  5 +---
 target/arm/kvm.c                             |  7 +++++
 target/arm/kvm_arm.h                         | 13 +++++++++
 30 files changed, 153 insertions(+), 40 deletions(-)

Comments

Zenghui Yu Aug. 29, 2019, 2:06 a.m. UTC | #1
Hi Eric,

On 2019/8/28 0:05, Eric Auger wrote:
> Since 4.18, KVM/ARM exposes a KVM_MAX_VCPUS equal to 512. However it was
> reported [1] that a VM with more than 256 vcpus cannot be launched. 5.4
> is about to fix the situation with 2 patches:
> - one upgrade of the KVM_IRQ_LINE API [2] supporting a vcpu id encoded
>    on 12 bits,
> - the reduction of KVM IO devices consumed by each GICv3 redistributor [3]
> 
> This series uses the new KVM_IRQ_LINE API and also checks the associated
> capability (KVM_CAP_ARM_IRQ_LINE_LAYOUT_2) in machvirt.
> 
> Without the series, as soon as the -smp arguments exceeds 256, QEMU exits
> with "kvm_set_irq: Invalid argument".

For the series:

Tested-by: Zenghui Yu <yuzenghui@huawei.com>

together with [2], [3].

> 
> Best Regards
> 
> Eric
> 
> References:
> [1] Can we boot a 512U kvm guest?
>      https://patchwork.kernel.org/patch/11091501/
> [2] [PATCH] KVM: arm/arm64: vgic: Allow more than 256 vcpus for KVM_IRQ_LINE
>      https://patchwork.kernel.org/patch/11099609/
> [3] [PATCH] KVM: arm/arm64: vgic: Use a single IO device per redistributor
>      https://patchwork.kernel.org/patch/11112141/
> 
> This series can be found at:
> https://github.com/eauger/qemu/tree/v4.1.0-256fix-rfc-v1
> 
> Eric Auger (3):
>    linux headers: update for KVM_CAP_ARM_IRQ_LINE_LAYOUT_2
>    intc/arm_gic: Support PPI injection for more than 256 vpus
>    virt: Check KVM_CAP_ARM_IRQ_LINE_LAYOUT_2 for smp_cpus > 256

Thanks,
zenghui