Re: [Qemu-devel] [PATCH v4 0/5] ARM: Add NUMA support for machine virt

[Andrew Jones <drjones@redhat.com>]

On Fri, Jan 29, 2016 at 02:52:35PM +0800, Shannon Zhao wrote:
> 
> 
> On 2016/1/29 14:32, Ashok Kumar wrote:
> > Hi, 
> > 
> > On Sat, Jan 23, 2016 at 07:36:41PM +0800, Shannon Zhao wrote:
> >> > From: Shannon Zhao <shannon.zhao@linaro.org>
> >> > 
> >> > Add NUMA support for machine virt. Tested successfully running a guest
> >> > Linux kernel with the following patch applied:
> >> > 
> >> > - [PATCH v9 0/6] arm64, numa: Add numa support for arm64 platforms
> >> > https://lwn.net/Articles/672329/
> >> > - [PATCH v2 0/4] ACPI based NUMA support for ARM64
> >> > http://www.spinics.net/lists/linux-acpi/msg61795.html
> >> > 
> >> > Changes since v3:
> >> > * based on new kernel driver and device bindings
> >> > * add ACPI part
> >> > 
> >> > Changes since v2:
> >> > * update to use NUMA node property arm,associativity.
> >> > 
> >> > Changes since v1:
> >> > Take into account Peter's comments:
> >> > * rename virt_memory_init to arm_generate_memory_dtb
> >> > * move arm_generate_memory_dtb to boot.c and make it a common func
> >> > * use a struct numa_map to generate numa dtb
> >> > 
> >> > Example qemu command line:
> >> > qemu-system-aarch64 \
> >> >     -enable-kvm -smp 4\
> >> >     -kernel Image \
> >> >     -m 512 -machine virt,kernel_irqchip=on \
> >> >     -initrd guestfs.cpio.gz \
> >> >     -cpu host -nographic \
> >> >     -numa node,mem=256M,cpus=0-1,nodeid=0 \
> >> >     -numa node,mem=256M,cpus=2-3,nodeid=1 \
> >> >     -append "console=ttyAMA0 root=/dev/ram"
> >> > 
> >> > Shannon Zhao (5):
> >> >   ARM: Virt: Add /distance-map node for NUMA
> >> >   ARM: Virt: Set numa-node-id for CPUs
> >> >   ARM: Add numa-node-id for /memory node
> >> >   include/hw/acpi/acpi-defs: Add GICC Affinity Structure
> >> >   hw/arm/virt-acpi-build: Generate SRAT table
> >> > 
> >> >  hw/arm/boot.c               | 29 ++++++++++++++++++++++-
> >> >  hw/arm/virt-acpi-build.c    | 58 +++++++++++++++++++++++++++++++++++++++++++++
> >> >  hw/arm/virt.c               | 37 +++++++++++++++++++++++++++++
> >> >  hw/i386/acpi-build.c        |  2 +-
> >> >  include/hw/acpi/acpi-defs.h | 15 +++++++++++-
> >> >  5 files changed, 138 insertions(+), 3 deletions(-)
> >> > 
> >> > -- 
> >> > 2.0.4
> >> > 
> > Don't we need to populate the NUMA node in the Affinity byte of MPIDR?
> > Linux uses the Affinity information in MPIDR to build topology which
> > might go wrong for the guest in this case. 
> > Maybe a non Linux OS might be impacted more?
> > 
> Ah, yes. It needs to update the MPIDR. But currently QEMU uses the value
> from KVM when using KVM. It needs to call kvm_set_one_reg to set the
> MPIDR and I'm not sure if this will affect KVM by looking at following
> comments:
>     /*
>      * When KVM is in use, PSCI is emulated in-kernel and not by qemu.
>      * Currently KVM has its own idea about MPIDR assignment, so we
>      * override our defaults with what we get from KVM.
>      */
> 
> Peter, do you have any suggestion?
> 
> > distance-map compatible string has been changed from
> > "numa,distance-map-v1" to "numa-distance-map-v1"
> Will update this.

Sigh... I've had the MPIDR rework (to let QEMU dictate it0 on my TODO list
for a loooooong time. I'll pick it up today and hopefully have something to
discuss next week. I'll keep this series in mind too.

Thanks,
drew