yeahuh / rpms / qemu-kvm

Forked from rpms/qemu-kvm 2 years ago
Clone

Blame SOURCES/kvm-spapr-Fix-ibm-max-associativity-domains-property-num.patch

b38b0f
From 4da5757f8ad715c203e2ef9320c49432e8259ee8 Mon Sep 17 00:00:00 2001
b38b0f
From: David Gibson <dgibson@redhat.com>
b38b0f
Date: Thu, 30 May 2019 04:37:23 +0100
b38b0f
Subject: [PATCH 2/8] spapr: Fix ibm, max-associativity-domains property number
b38b0f
 of nodes
b38b0f
b38b0f
RH-Author: David Gibson <dgibson@redhat.com>
b38b0f
Message-id: <20190530043728.32575-2-dgibson@redhat.com>
b38b0f
Patchwork-id: 88418
b38b0f
O-Subject: [RHEL-8.1 qemu-kvm PATCH 1/6] spapr: Fix ibm, max-associativity-domains property number of nodes
b38b0f
Bugzilla: 1710662
b38b0f
RH-Acked-by: Laurent Vivier <lvivier@redhat.com>
b38b0f
RH-Acked-by: Auger Eric <eric.auger@redhat.com>
b38b0f
RH-Acked-by: Cornelia Huck <cohuck@redhat.com>
b38b0f
b38b0f
From: Serhii Popovych <spopovyc@redhat.com>
b38b0f
b38b0f
Laurent Vivier reported off by one with maximum number of NUMA nodes
b38b0f
provided by qemu-kvm being less by one than required according to
b38b0f
description of "ibm,max-associativity-domains" property in LoPAPR.
b38b0f
b38b0f
It appears that I incorrectly treated LoPAPR description of this
b38b0f
property assuming it provides last valid domain (NUMA node here)
b38b0f
instead of maximum number of domains.
b38b0f
b38b0f
  ### Before hot-add
b38b0f
b38b0f
  (qemu) info numa
b38b0f
  3 nodes
b38b0f
  node 0 cpus: 0
b38b0f
  node 0 size: 0 MB
b38b0f
  node 0 plugged: 0 MB
b38b0f
  node 1 cpus:
b38b0f
  node 1 size: 1024 MB
b38b0f
  node 1 plugged: 0 MB
b38b0f
  node 2 cpus:
b38b0f
  node 2 size: 0 MB
b38b0f
  node 2 plugged: 0 MB
b38b0f
b38b0f
  $ numactl -H
b38b0f
  available: 2 nodes (0-1)
b38b0f
  node 0 cpus: 0
b38b0f
  node 0 size: 0 MB
b38b0f
  node 0 free: 0 MB
b38b0f
  node 1 cpus:
b38b0f
  node 1 size: 999 MB
b38b0f
  node 1 free: 658 MB
b38b0f
  node distances:
b38b0f
  node   0   1
b38b0f
    0:  10  40
b38b0f
    1:  40  10
b38b0f
b38b0f
  ### Hot-add
b38b0f
b38b0f
  (qemu) object_add memory-backend-ram,id=mem0,size=1G
b38b0f
  (qemu) device_add pc-dimm,id=dimm1,memdev=mem0,node=2
b38b0f
  (qemu) [   87.704898] pseries-hotplug-mem: Attempting to hot-add 4 ...
b38b0f
  <there is no "Initmem setup node 2 [mem 0xHEX-0xHEX]">
b38b0f
  [   87.705128] lpar: Attempting to resize HPT to shift 21
b38b0f
  ... <HPT resize messages>
b38b0f
b38b0f
  ### After hot-add
b38b0f
b38b0f
  (qemu) info numa
b38b0f
  3 nodes
b38b0f
  node 0 cpus: 0
b38b0f
  node 0 size: 0 MB
b38b0f
  node 0 plugged: 0 MB
b38b0f
  node 1 cpus:
b38b0f
  node 1 size: 1024 MB
b38b0f
  node 1 plugged: 0 MB
b38b0f
  node 2 cpus:
b38b0f
  node 2 size: 1024 MB
b38b0f
  node 2 plugged: 1024 MB
b38b0f
b38b0f
  $ numactl -H
b38b0f
  available: 2 nodes (0-1)
b38b0f
  ^^^^^^^^^^^^^^^^^^^^^^^^
b38b0f
             Still only two nodes (and memory hot-added to node 0 below)
b38b0f
  node 0 cpus: 0
b38b0f
  node 0 size: 1024 MB
b38b0f
  node 0 free: 1021 MB
b38b0f
  node 1 cpus:
b38b0f
  node 1 size: 999 MB
b38b0f
  node 1 free: 658 MB
b38b0f
  node distances:
b38b0f
  node   0   1
b38b0f
    0:  10  40
b38b0f
    1:  40  10
b38b0f
b38b0f
After fix applied numactl(8) reports 3 nodes available and memory
b38b0f
plugged into node 2 as expected.
b38b0f
b38b0f
>From David Gibson:
b38b0f
------------------
b38b0f
  Qemu makes a distinction between "non NUMA" (nb_numa_nodes == 0) and
b38b0f
  "NUMA with one node" (nb_numa_nodes == 1).  But from a PAPR guests's
b38b0f
  point of view these are equivalent.  I don't want to present two
b38b0f
  different cases to the guest when we don't need to, so even though the
b38b0f
  guest can handle it, I'd prefer we put a '1' here for both the
b38b0f
  nb_numa_nodes == 0 and nb_numa_nodes == 1 case.
b38b0f
b38b0f
This consolidates everything discussed previously on mailing list.
b38b0f
b38b0f
Fixes: da9f80fbad21 ("spapr: Add ibm,max-associativity-domains property")
b38b0f
Reported-by: Laurent Vivier <lvivier@redhat.com>
b38b0f
Signed-off-by: Serhii Popovych <spopovyc@redhat.com>
b38b0f
b38b0f
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
b38b0f
Reviewed-by: Greg Kurz <groug@kaod.org>
b38b0f
Reviewed-by: Laurent Vivier <lvivier@redhat.com>
b38b0f
(cherry picked from commit 3908a24fcb83913079d315de0ca6d598e8616dbb)
b38b0f
b38b0f
Bugzilla: https://bugzilla.redhat.com/show_bug.cgi?id=1710662
b38b0f
b38b0f
Signed-off-by: David Gibson <dgibson@redhat.com>
b38b0f
Signed-off-by: Danilo C. L. de Paula <ddepaula@redhat.com>
b38b0f
---
b38b0f
 hw/ppc/spapr.c | 2 +-
b38b0f
 1 file changed, 1 insertion(+), 1 deletion(-)
b38b0f
b38b0f
diff --git a/hw/ppc/spapr.c b/hw/ppc/spapr.c
b38b0f
index ea72782..b57c0be 100644
b38b0f
--- a/hw/ppc/spapr.c
b38b0f
+++ b/hw/ppc/spapr.c
b38b0f
@@ -915,7 +915,7 @@ static void spapr_dt_rtas(sPAPRMachineState *spapr, void *fdt)
b38b0f
         cpu_to_be32(0),
b38b0f
         cpu_to_be32(0),
b38b0f
         cpu_to_be32(0),
b38b0f
-        cpu_to_be32(nb_numa_nodes ? nb_numa_nodes - 1 : 0),
b38b0f
+        cpu_to_be32(nb_numa_nodes ? nb_numa_nodes : 1),
b38b0f
     };
b38b0f
 
b38b0f
     _FDT(rtas = fdt_add_subnode(fdt, 0, "rtas"));
b38b0f
-- 
b38b0f
1.8.3.1
b38b0f