powerpc/mm/numa: skip NUMA_NO_NODE onlining in parse_numa_properties()
authorDaniel Henrique Barboza <danielhb413@gmail.com>
Thu, 24 Feb 2022 18:23:12 +0000 (15:23 -0300)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Fri, 8 Apr 2022 12:40:19 +0000 (14:40 +0200)
commit05abd49972e1965d6ee88b9b7ad55865b4e7e73e
treeeb9218079476bf2e0d9e1d81e83853821021bda3
parent3e04a837dba71593dc3c21ecbe142bb465e48a16
powerpc/mm/numa: skip NUMA_NO_NODE onlining in parse_numa_properties()

[ Upstream commit 749ed4a20657bcea66a6e082ca3dc0d228cbec80 ]

Executing node_set_online() when nid = NUMA_NO_NODE results in an
undefined behavior. node_set_online() will call node_set_state(), into
__node_set(), into set_bit(), and since NUMA_NO_NODE is -1 we'll end up
doing a negative shift operation inside
arch/powerpc/include/asm/bitops.h. This potential UB was detected
running a kernel with CONFIG_UBSAN.

The behavior was introduced by commit 10f78fd0dabb ("powerpc/numa: Fix a
regression on memoryless node 0"), where the check for nid > 0 was
removed to fix a problem that was happening with nid = 0, but the result
is that now we're trying to online NUMA_NO_NODE nids as well.

Checking for nid >= 0 will allow node 0 to be onlined while avoiding
this UB with NUMA_NO_NODE.

Fixes: 10f78fd0dabb ("powerpc/numa: Fix a regression on memoryless node 0")
Reported-by: Ping Fang <pifang@redhat.com>
Signed-off-by: Daniel Henrique Barboza <danielhb413@gmail.com>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://lore.kernel.org/r/20220224182312.1012527-1-danielhb413@gmail.com
Signed-off-by: Sasha Levin <sashal@kernel.org>
arch/powerpc/mm/numa.c