On 07.02.23 01:32, Mark Brown wrote:

On Fri, Jan 13, 2023 at 06:10:04PM +0100, David Hildenbrand wrote:

quoted

Let's support __HAVE_ARCH_PTE_SWP_EXCLUSIVE by stealing one bit from the
offset. This reduces the maximum swap space per file to 64 GiB (was 128
GiB).

While at it drop the PTE_TYPE_FAULT from __swp_entry_to_pte() which is
defined to be 0 and is rather confusing because we should be dealing
with "Linux PTEs" not "hardware PTEs". Also, properly mask the type in
__swp_entry().

Today's -next (and at least back to Friday, older logs are unclear - I
only noticed -next issues today) fails to NFS boot on an AT91SAM9G20-EK
(an old ARMv5 platform) with multi_v5_defconfig, a bisect appears to
point to this patch (20aae9eff5acd8f5 in today's -next) as the culprit.

It's been in -next for quite a while, thanks for the report!

The failure happens at some point after starting userspace, the kernel
starts spamming the console with messages in the form:

     get_swap_device: Bad swap file entry 10120d20

_swap_info_get() tells us that the swp type seems to be bad.
I assume we're dealing with a migration entry, if swap is disabled, and fail to
detect is_migration_entry() correctly because the type is messed up.

Could you give the following a test?


 From 8c4bdbd9862f85782d5919d044c172b584063e83 Mon Sep 17 00:00:00 2001
From: David Hildenbrand <redacted>
Date: Wed, 8 Feb 2023 15:08:01 +0100
Subject: [PATCH] arm/mm: Fix swp type masking in __swp_entry()

We're masking with the number of type bits instead of the type mask, which
is obviously wrong.

Fixes: 20aae9eff5ac ("arm/mm: support __HAVE_ARCH_PTE_SWP_EXCLUSIVE")
Reported-by: Mark Brown <broonie@kernel.org>
Signed-off-by: David Hildenbrand <redacted>
---
  arch/arm/include/asm/pgtable.h | 2 +-
  1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/arm/include/asm/pgtable.h b/arch/arm/include/asm/pgtable.h
index 2e626e6da9a3..a58ccbb406ad 100644
--- a/arch/arm/include/asm/pgtable.h
+++ b/arch/arm/include/asm/pgtable.h

@@ -292,7 +292,7 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
  
  #define __swp_type(x)		(((x).val >> __SWP_TYPE_SHIFT) & __SWP_TYPE_MASK)
  #define __swp_offset(x)		((x).val >> __SWP_OFFSET_SHIFT)
-#define __swp_entry(type, offset) ((swp_entry_t) { (((type) & __SWP_TYPE_BITS) << __SWP_TYPE_SHIFT) | \
+#define __swp_entry(type, offset) ((swp_entry_t) { (((type) & __SWP_TYPE_MASK) << __SWP_TYPE_SHIFT) | \
  						   ((offset) << __SWP_OFFSET_SHIFT) })
  
  #define __pte_to_swp_entry(pte)	((swp_entry_t) { pte_val(pte) })

-- 
2.39.1


-- 
Thanks,

David / dhildenb