[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Xen-devel] [PATCH v10 5/9] libxl: vnuma types declararion

On Wed, 2014-09-03 at 00:24 -0400, Elena Ufimtseva wrote:
> Adds vnuma topology types declarations to libxl_domain_build_info
> structure.

So, as mentioned I was becoming increasingly confused about what all
these fields represent. So lets stop and take stock.

> Signed-off-by: Elena Ufimtseva <ufimtseva@xxxxxxxxx>
> ---
>  tools/libxl/libxl_types.idl |    8 +++++++-
>  tools/libxl/libxl_vnuma.h   |   16 ++++++++++++++++
>  2 files changed, 23 insertions(+), 1 deletion(-)
>  create mode 100644 tools/libxl/libxl_vnuma.h
> diff --git a/tools/libxl/libxl_types.idl b/tools/libxl/libxl_types.idl
> index 08a7927..ea8bac0 100644
> --- a/tools/libxl/libxl_types.idl
> +++ b/tools/libxl/libxl_types.idl
> @@ -333,7 +333,13 @@ libxl_domain_build_info = Struct("domain_build_info",[
>      ("disable_migrate", libxl_defbool),
>      ("cpuid",           libxl_cpuid_policy_list),
>      ("blkdev_start",    string),
> -    
> +    ("vnodes",          uint32),

I suppose this is the number of NUMA nodes the guest should have?

> +    ("vmemranges",      uint32),

and I suppose this is a number of memory ranges? Does this differ ever
from vnodes?

> +    ("vnuma_mem",       Array(uint64, "num_vnuma_mem")),

The size of each memory node?

When/how can num_vnuma_mem differ from vmemranges?

> +    ("vnuma_vcpumap",   Array(uint32, "num_vnuma_vcpumap")),

Is this supposed to map numa node to a bitmap of vcpus? If yes:

When/how can num_vnuma_vcpumap differ from vnodes?

What happens if a guest has 33 vcpus?

Or maybe the output of this is a single vcpu number?

> +    ("vdistance",       Array(uint32, "num_vdistance")),

nodes to distances? Can num_vdistance ever differ from vnodes?

I thought distances were between two nodes, in which case doesn't this
need to be multidimensional?

> +    ("vnuma_vnodemap",  Array(uint32, "num_vnuma_vnondemap")),

I've run out of guesses for what this might be. vnode to pnode perhaps?
You appear to unconditionally overwrite whatever the users might have
put here.

> +    ("vnuma_autoplacement",  libxl_defbool),

Why/how does this differ from the existing numa placement option?


Xen-devel mailing list



Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.