powerpc/numa: Fix a regression on memoryless node 0

Commit e75130f20b1f ("powerpc/numa: Offline memoryless cpuless node 0")
offlines node 0 and expects nodes to be subsequently onlined when CPUs
or nodes are detected.

Commit 6398eaa26816 ("powerpc/numa: Prefer node id queried from vphn")
skips onlining node 0 when CPUs are associated with node 0.

On systems with node 0 having CPUs but no memory, this causes node 0 be
marked offline. This causes issues at boot time when trying to set
memory node for online CPUs while building the zonelist.

0:mon> t
[link register   ] c000000000400354 __build_all_zonelists+0x164/0x280
[c00000000161bda0] c0000000016533c8 node_states+0x20/0xa0 (unreliable)
[c00000000161bdc0] c000000000400384 __build_all_zonelists+0x194/0x280
[c00000000161be30] c000000001041800 build_all_zonelists_init+0x4c/0x118
[c00000000161be80] c0000000004020d0 build_all_zonelists+0x190/0x1b0
[c00000000161bef0] c000000001003cf8 start_kernel+0x18c/0x6a8
[c00000000161bf90] c00000000000adb4 start_here_common+0x1c/0x3e8
0:mon> r
R00 = c000000000400354   R16 = 000000000b57a0e8
R01 = c00000000161bda0   R17 = 000000000b57a6b0
R02 = c00000000161ce00   R18 = 000000000b5afee8
R03 = 0000000000000000   R19 = 000000000b6448a0
R04 = 0000000000000000   R20 = fffffffffffffffd
R05 = 0000000000000000   R21 = 0000000001400000
R06 = 0000000000000000   R22 = 000000001ec00000
R07 = 0000000000000001   R23 = c000000001175580
R08 = 0000000000000000   R24 = c000000001651ed8
R09 = c0000000017e84d8   R25 = c000000001652480
R10 = 0000000000000000   R26 = c000000001175584
R11 = c000000c7fac0d10   R27 = c0000000019568d0
R12 = c000000000400180   R28 = 0000000000000000
R13 = c000000002200000   R29 = c00000000164dd78
R14 = 000000000b579f78   R30 = 0000000000000000
R15 = 000000000b57a2b8   R31 = c000000001175584
pc  = c000000000400194 local_memory_node+0x24/0x80
cfar= c000000000074334 mcount+0xc/0x10
lr  = c000000000400354 __build_all_zonelists+0x164/0x280
msr = 8000000002001033   cr  = 44002284
ctr = c000000000400180   xer = 0000000000000001   trap =  380
dar = 0000000000001388   dsisr = c00000000161bc90
0:mon>

Fix this by setting node to be online while onlining CPUs that belong to
node 0.

Fixes: e75130f20b1f ("powerpc/numa: Offline memoryless cpuless node 0")
Fixes: 6398eaa26816 ("powerpc/numa: Prefer node id queried from vphn")
Reported-by: Milan Mohanty <milmohan@in.ibm.com>
Signed-off-by: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20201127053738.10085-1-srikar@linux.vnet.ibm.com

diff --git a/arch/powerpc/mm/numa.c b/arch/powerpc/mm/numa.c
index 63f61d8..f2bf98b 100644 (file)
--- a/arch/powerpc/mm/numa.c
+++ b/arch/powerpc/mm/numa.c
@@ -742,8 +742,7 @@ static int __init parse_numa_properties(void)
                        of_node_put(cpu);
                }
 
-               if (likely(nid > 0))
-                       node_set_online(nid);
+               node_set_online(nid);
        }
 
        get_n_mem_cells(&n_mem_addr_cells, &n_mem_size_cells);