[PATCH 1/2] powerpc/mce: mce_init use early_cpu_to_node

cpu_to_node is not available (setup_arch() is called before
setup_per_cpu_areas() by start_kernel()).

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 arch/powerpc/kernel/mce.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/powerpc/kernel/mce.c b/arch/powerpc/kernel/mce.c
index 18173199b79d..6c5d30fba766 100644
--- a/arch/powerpc/kernel/mce.c
+++ b/arch/powerpc/kernel/mce.c

@@ -756,7 +756,7 @@ void __init mce_init(void)
 		mce_info = memblock_alloc_try_nid(sizeof(*mce_info),
 						  __alignof__(*mce_info),
 						  MEMBLOCK_LOW_LIMIT,
-						  limit, cpu_to_node(i));
+						  limit, early_cpu_to_node(i));
 		if (!mce_info)
 			goto err;
 		paca_ptrs[i]->mce_info = mce_info;

-- 
2.35.1

If the boot CPU tries to access per-cpu data of other CPUs before
per cpu areas are set up, it will unexpectedly use offset 0.

Try to catch such accesses by poisoning the __per_cpu_offset array.

Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
---
 arch/powerpc/include/asm/percpu.h | 1 +
 arch/powerpc/kernel/paca.c        | 2 +-
 arch/powerpc/kernel/setup_64.c    | 2 +-
 3 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/powerpc/include/asm/percpu.h b/arch/powerpc/include/asm/percpu.h
index 8e5b7d0b851c..6ca1a9fc5725 100644
--- a/arch/powerpc/include/asm/percpu.h
+++ b/arch/powerpc/include/asm/percpu.h

@@ -7,6 +7,7 @@
  * Same as asm-generic/percpu.h, except that we store the per cpu offset
  * in the paca. Based on the x86-64 implementation.
  */
+#define PER_CPU_OFFSET_POISON 0xfeeeeeeeeeeeeeeeULL
 
 #ifdef CONFIG_SMP
 

diff --git a/arch/powerpc/kernel/paca.c b/arch/powerpc/kernel/paca.c
index ba593fd60124..914d27c8b84a 100644
--- a/arch/powerpc/kernel/paca.c
+++ b/arch/powerpc/kernel/paca.c

@@ -223,7 +223,7 @@ void __init initialise_paca(struct paca_struct *new_paca, int cpu)
 	new_paca->hw_cpu_id = 0xffff;
 	new_paca->kexec_state = KEXEC_STATE_NONE;
 	new_paca->__current = &init_task;
-	new_paca->data_offset = 0xfeeeeeeeeeeeeeeeULL;
+	new_paca->data_offset = PER_CPU_OFFSET_POISON;
 #ifdef CONFIG_PPC_64S_HASH_MMU
 	new_paca->slb_shadow_ptr = NULL;
 #endif

diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c
index 5761f08dae95..60f0d1258526 100644
--- a/arch/powerpc/kernel/setup_64.c
+++ b/arch/powerpc/kernel/setup_64.c

@@ -812,7 +812,7 @@ static __init int pcpu_cpu_to_node(int cpu)
 	return early_cpu_to_node(cpu);
 }
 
-unsigned long __per_cpu_offset[NR_CPUS] __read_mostly;
+unsigned long __per_cpu_offset[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1 ] = PER_CPU_OFFSET_POISON };
 EXPORT_SYMBOL(__per_cpu_offset);
 
 void __init setup_per_cpu_areas(void)

-- 
2.35.1

On Mon, 11 Jul 2022 13:06:52 +1000, Nicholas Piggin wrote:

cpu_to_node is not available (setup_arch() is called before
setup_per_cpu_areas() by start_kernel()).

Patch 1 applied to powerpc/next.

[1/2] powerpc/mce: mce_init use early_cpu_to_node
      https://git.kernel.org/powerpc/c/f57261e69825b332f22480b37a33e03d066c0da2

cheers

Nicholas Piggin [off-list ref] writes:

If the boot CPU tries to access per-cpu data of other CPUs before
per cpu areas are set up, it will unexpectedly use offset 0.

Try to catch such accesses by poisoning the __per_cpu_offset array.

I wasn't sure about this.

On bare metal it's just an instant checkstop which is very user hostile.

I worry it's just going to cause unusual configurations/options to crash
for folks, like eg. booting with page_poison=1 did a while back.

Can we put it behind a debug option? Maybe CONFIG_DEBUG_VM ?

cheers