Re: [Qemu-ppc] [PATCH] numa, spapr: align default numa node memory size to 256MB

[Thomas Huth]

On 20.03.2017 12:24, Laurent Vivier wrote:
> Since commit 224245b ("spapr: Add LMB DR connectors"), NUMA node
> memory size must be aligned to 256MB (SPAPR_MEMORY_BLOCK_SIZE).
> 
> But when "-numa" option is provided without "mem" parameter,
> the memory is equally divided between nodes, but 8MB aligned.
> This can be not valid for pseries.
> 
> In that case we can have:
> $ ./ppc64-softmmu/qemu-system-ppc64 -m 4G -numa node -numa node -numa node
> qemu-system-ppc64: Node 0 memory size 0x55000000 is not aligned to 256 MiB
> 
> With this patch, we have:
> (qemu) info numa
> 3 nodes
> node 0 cpus: 0
> node 0 size: 1280 MB
> node 1 cpus:
> node 1 size: 1280 MB
> node 2 cpus:
> node 2 size: 1536 MB
> 
> Signed-off-by: Laurent Vivier <address@hidden>
> ---
>  dtc    |  2 +-
>  numa.c | 14 +++++++++-----
>  2 files changed, 10 insertions(+), 6 deletions(-)
> 
> diff --git a/dtc b/dtc
> index 558cd81..fa8bc7f 160000
> --- a/dtc
> +++ b/dtc
> @@ -1 +1 @@
> -Subproject commit 558cd81bdd432769b59bff01240c44f82cfb1a9d
> +Subproject commit fa8bc7f928ac25f23532afc8beb2073efc8fb063

Accidential change?

> diff --git a/numa.c b/numa.c
> index e01cb54..a911284 100644
> --- a/numa.c
> +++ b/numa.c
> @@ -337,15 +337,19 @@ void parse_numa_opts(MachineClass *mc)
>          }
>          if (i == nb_numa_nodes) {
>              uint64_t usedmem = 0;
> -
> -            /* On Linux, each node's border has to be 8MB aligned,
> -             * the final node gets the rest.
> -             */
> +#if defined(TARGET_PPC64)
> +            /* pseries requests each node's border has to be 256 MB aligned 
> */
> +            const uint64_t numa_mem_align_mask = ~((1 << 28UL) - 1);
> +#else
> +            /* On Linux, each node's border has to be 8MB aligned */
> +            const uint64_t numa_mem_align_mask = ~((1 << 23UL)  - 1);
> +#endif

With that #if, you enable this for all ppc machines, not just for the
pseries machine. So it might be cleaner to add a setting to the
MachineClass instead...?

>              for (i = 0; i < nb_numa_nodes - 1; i++) {
>                  numa_info[i].node_mem = (ram_size / nb_numa_nodes) &
> -                                        ~((1 << 23UL) - 1);
> +                                        numa_mem_align_mask;
>                  usedmem += numa_info[i].node_mem;
>              }
> +            /* the final node gets the rest. */
>              numa_info[i].node_mem = ram_size - usedmem;
>          }

 Thomas