Skip to contentSkip to conversation

mm/swap: fix SWP_PFN_BITS with CONFIG_PHYS_ADDR_T_64BIT on 32bit

PATCH In review David Hildenbrand 3 participants 5 messages Dec 6, 2022 – Dec 8, 2022

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH v2 1/1 mm/swap: fix SWP_PFN_BITS with CONFIG_PHYS_ADDR_T_64BIT on 32bit

We use "unsigned long" to store a PFN in the kernel and phys_addr_t to store a physical address.

On a 64bit system, both are 64bit wide. However, on a 32bit system, the latter might be 64bit wide. This is, for example, the case on x86 with
PAE: phys_addr_t and PTEs are 64bit wide, while "unsigned long" only spans 32bit.

The current definition of SWP_PFN_BITS without MAX_PHYSMEM_BITS misses that case, and assumes that the maximum PFN is limited by an 32bit phys_addr_t. This implies, that SWP_PFN_BITS will currently only be able to cover 4 GiB - 1 on any 32bit system with 4k page size, which is wrong.

Let's rely on the number of bits in phys_addr_t instead, but make sure to not exceed the maximum swap offset, to not make the BUILD_BUG_ON() in is_pfn_swap_entry() unhappy. Note that swp_entry_t is effectively an unsigned long and the maximum swap offset shares that value with the swap type.

For example, on an 8 GiB x86 PAE system with a kernel config based on Debian 11.5 (-> CONFIG_FLATMEM=y, CONFIG_X86_PAE=y), we will currently fail removing migration entries (remove_migration_ptes()), because mm/page_vma_mapped.c:check_pte() will fail to identify a PFN match as swp_offset_pfn() wrongly masks off PFN bits. For example, split_huge_page_to_list()->...->remap_page() will leave migration entries in place and continue to unlock the page.

Later, when we stumble over these migration entries (e.g., via /proc/self/pagemap), pfn_swap_entry_to_page() will BUG_ON() because these migration entries shouldn't exist anymore and the page was unlocked.

[   33.067591] kernel BUG at include/linux/swapops.h:497!
[   33.067597] invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
[   33.067602] CPU: 3 PID: 742 Comm: cow Tainted: G            E      6.1.0-rc8+ #16
[   33.067605] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-1.fc36 04/01/2014
[   33.067606] EIP: pagemap_pmd_range+0x644/0x650
[   33.067612] Code: 00 00 00 00 66 90 89 ce b9 00 f0 ff ff e9 ff fb ff ff 89 d8 31 db e8 48 c6 52 00 e9 23 fb ff ff e8 61 83 56 00 e9 b6 fe ff ff <0f> 0b bf 00 f0 ff ff e9 38 fa ff ff 3e 8d 74 26 00 55 89 e5 57 31
[   33.067615] EAX: ee394000 EBX: 00000002 ECX: ee394000 EDX: 00000000
[   33.067617] ESI: c1b0ded4 EDI: 00024a00 EBP: c1b0ddb4 ESP: c1b0dd68
[   33.067619] DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068 EFLAGS: 00010246
[   33.067624] CR0: 80050033 CR2: b7a00000 CR3: 01bbbd20 CR4: 00350ef0
[   33.067625] Call Trace:
[   33.067628]  ? madvise_free_pte_range+0x720/0x720
[   33.067632]  ? smaps_pte_range+0x4b0/0x4b0
[   33.067634]  walk_pgd_range+0x325/0x720
[   33.067637]  ? mt_find+0x1d6/0x3a0
[   33.067641]  ? mt_find+0x1d6/0x3a0
[   33.067643]  __walk_page_range+0x164/0x170
[   33.067646]  walk_page_range+0xf9/0x170
[   33.067648]  ? __kmem_cache_alloc_node+0x2a8/0x340
[   33.067653]  pagemap_read+0x124/0x280
[   33.067658]  ? default_llseek+0x101/0x160
[   33.067662]  ? smaps_account+0x1d0/0x1d0
[   33.067664]  vfs_read+0x90/0x290
[   33.067667]  ? do_madvise.part.0+0x24b/0x390
[   33.067669]  ? debug_smp_processor_id+0x12/0x20
[   33.067673]  ksys_pread64+0x58/0x90
[   33.067675]  __ia32_sys_ia32_pread64+0x1b/0x20
[   33.067680]  __do_fast_syscall_32+0x4c/0xc0
[   33.067683]  do_fast_syscall_32+0x29/0x60
[   33.067686]  do_SYSENTER_32+0x15/0x20
[   33.067689]  entry_SYSENTER_32+0x98/0xf1

Decrease the indentation level of SWP_PFN_BITS and SWP_PFN_MASK to keep it readable and consistent.

Fixes 0d206b5d2e0d ("mm/swap: add swp_offset_pfn() to fetch PFN from swap entry")
Cc Andrew Morton <[email protected]>
Cc Peter Xu <[email protected]>
Cc Yang Shi <[email protected]>
Cc Hugh Dickins <[email protected]>
Cc Andrea Arcangeli <[email protected]>

---

v1 -> v2:
* Rely on sizeof(phys_addr_t) and min_t() instead.
* Survives my various cross compilations and testing on x86 PAE.

---
 include/linux/swapops.h | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
include/linux/swapops.h +5 -3
index 86b95ccb81bb..424d586ed1ca 100644
@@ -33,11 +33,13 @@
  * can use the extra bits to store other information besides PFN.
  */
 #ifdef MAX_PHYSMEM_BITS
-#define SWP_PFN_BITS			(MAX_PHYSMEM_BITS - PAGE_SHIFT)
+#define SWP_PFN_BITS		(MAX_PHYSMEM_BITS - PAGE_SHIFT)
 #else  /* MAX_PHYSMEM_BITS */
-#define SWP_PFN_BITS			(BITS_PER_LONG - PAGE_SHIFT)
+#define SWP_PFN_BITS		min_t(phys_addr_t, \
+				      sizeof(phys_addr_t) * 8 - PAGE_SHIFT, \
+				      SWP_TYPE_SHIFT)
 #endif	/* MAX_PHYSMEM_BITS */
-#define SWP_PFN_MASK			(BIT(SWP_PFN_BITS) - 1)
+#define SWP_PFN_MASK		(BIT(SWP_PFN_BITS) - 1)
 
 /**
  * Migration swap entry specific bitfield definitions.  Layout:

base-commit: 76dcd734eca23168cb008912c0f69ff408905235

Signature · 1 line
2.38.1
Acked-by on v2 1/1
#1 lore raw
Quoted 74 lines · David Hildenbrand

We use "unsigned long" to store a PFN in the kernel and phys_addr_t to store a physical address.

On a 64bit system, both are 64bit wide. However, on a 32bit system, the latter might be 64bit wide. This is, for example, the case on x86 with
PAE: phys_addr_t and PTEs are 64bit wide, while "unsigned long" only spans 32bit.

The current definition of SWP_PFN_BITS without MAX_PHYSMEM_BITS misses that case, and assumes that the maximum PFN is limited by an 32bit phys_addr_t. This implies, that SWP_PFN_BITS will currently only be able to cover 4 GiB - 1 on any 32bit system with 4k page size, which is wrong.

Let's rely on the number of bits in phys_addr_t instead, but make sure to not exceed the maximum swap offset, to not make the BUILD_BUG_ON() in is_pfn_swap_entry() unhappy. Note that swp_entry_t is effectively an unsigned long and the maximum swap offset shares that value with the swap type.

For example, on an 8 GiB x86 PAE system with a kernel config based on Debian 11.5 (-> CONFIG_FLATMEM=y, CONFIG_X86_PAE=y), we will currently fail removing migration entries (remove_migration_ptes()), because mm/page_vma_mapped.c:check_pte() will fail to identify a PFN match as swp_offset_pfn() wrongly masks off PFN bits. For example, split_huge_page_to_list()->...->remap_page() will leave migration entries in place and continue to unlock the page.

Later, when we stumble over these migration entries (e.g., via /proc/self/pagemap), pfn_swap_entry_to_page() will BUG_ON() because these migration entries shouldn't exist anymore and the page was unlocked.

[   33.067591] kernel BUG at include/linux/swapops.h:497!
[   33.067597] invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
[   33.067602] CPU: 3 PID: 742 Comm: cow Tainted: G            E      6.1.0-rc8+ #16
[   33.067605] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-1.fc36 04/01/2014
[   33.067606] EIP: pagemap_pmd_range+0x644/0x650
[   33.067612] Code: 00 00 00 00 66 90 89 ce b9 00 f0 ff ff e9 ff fb ff ff 89 d8 31 db e8 48 c6 52 00 e9 23 fb ff ff e8 61 83 56 00 e9 b6 fe ff ff <0f> 0b bf 00 f0 ff ff e9 38 fa ff ff 3e 8d 74 26 00 55 89 e5 57 31
[   33.067615] EAX: ee394000 EBX: 00000002 ECX: ee394000 EDX: 00000000
[   33.067617] ESI: c1b0ded4 EDI: 00024a00 EBP: c1b0ddb4 ESP: c1b0dd68
[   33.067619] DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068 EFLAGS: 00010246
[   33.067624] CR0: 80050033 CR2: b7a00000 CR3: 01bbbd20 CR4: 00350ef0
[   33.067625] Call Trace:
[   33.067628]  ? madvise_free_pte_range+0x720/0x720
[   33.067632]  ? smaps_pte_range+0x4b0/0x4b0
[   33.067634]  walk_pgd_range+0x325/0x720
[   33.067637]  ? mt_find+0x1d6/0x3a0
[   33.067641]  ? mt_find+0x1d6/0x3a0
[   33.067643]  __walk_page_range+0x164/0x170
[   33.067646]  walk_page_range+0xf9/0x170
[   33.067648]  ? __kmem_cache_alloc_node+0x2a8/0x340
[   33.067653]  pagemap_read+0x124/0x280
[   33.067658]  ? default_llseek+0x101/0x160
[   33.067662]  ? smaps_account+0x1d0/0x1d0
[   33.067664]  vfs_read+0x90/0x290
[   33.067667]  ? do_madvise.part.0+0x24b/0x390
[   33.067669]  ? debug_smp_processor_id+0x12/0x20
[   33.067673]  ksys_pread64+0x58/0x90
[   33.067675]  __ia32_sys_ia32_pread64+0x1b/0x20
[   33.067680]  __do_fast_syscall_32+0x4c/0xc0
[   33.067683]  do_fast_syscall_32+0x29/0x60
[   33.067686]  do_SYSENTER_32+0x15/0x20
[   33.067689]  entry_SYSENTER_32+0x98/0xf1

Decrease the indentation level of SWP_PFN_BITS and SWP_PFN_MASK to keep it readable and consistent.

Fixes 0d206b5d2e0d ("mm/swap: add swp_offset_pfn() to fetch PFN from swap entry")
Cc Andrew Morton <[email protected]>
Cc Peter Xu <[email protected]>
Cc Yang Shi <[email protected]>
Cc Hugh Dickins <[email protected]>
Cc Andrea Arcangeli <[email protected]>
Quoted 5 lines

---

v1 -> v2:
* Rely on sizeof(phys_addr_t) and min_t() instead.
* Survives my various cross compilations and testing on x86 PAE.

Good to know it works, thanks a lot.

Signature · 1 line
Peter Xu
#2 lore raw
Quoted 77 lines · Peter Xu
Quoted 76 lines · David Hildenbrand

We use "unsigned long" to store a PFN in the kernel and phys_addr_t to store a physical address.

On a 64bit system, both are 64bit wide. However, on a 32bit system, the latter might be 64bit wide. This is, for example, the case on x86 with
PAE: phys_addr_t and PTEs are 64bit wide, while "unsigned long" only spans 32bit.

The current definition of SWP_PFN_BITS without MAX_PHYSMEM_BITS misses that case, and assumes that the maximum PFN is limited by an 32bit phys_addr_t. This implies, that SWP_PFN_BITS will currently only be able to cover 4 GiB - 1 on any 32bit system with 4k page size, which is wrong.

Let's rely on the number of bits in phys_addr_t instead, but make sure to not exceed the maximum swap offset, to not make the BUILD_BUG_ON() in is_pfn_swap_entry() unhappy. Note that swp_entry_t is effectively an unsigned long and the maximum swap offset shares that value with the swap type.

For example, on an 8 GiB x86 PAE system with a kernel config based on Debian 11.5 (-> CONFIG_FLATMEM=y, CONFIG_X86_PAE=y), we will currently fail removing migration entries (remove_migration_ptes()), because mm/page_vma_mapped.c:check_pte() will fail to identify a PFN match as swp_offset_pfn() wrongly masks off PFN bits. For example, split_huge_page_to_list()->...->remap_page() will leave migration entries in place and continue to unlock the page.

Later, when we stumble over these migration entries (e.g., via /proc/self/pagemap), pfn_swap_entry_to_page() will BUG_ON() because these migration entries shouldn't exist anymore and the page was unlocked.

[   33.067591] kernel BUG at include/linux/swapops.h:497!
[   33.067597] invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
[   33.067602] CPU: 3 PID: 742 Comm: cow Tainted: G            E      6.1.0-rc8+ #16
[   33.067605] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-1.fc36 04/01/2014
[   33.067606] EIP: pagemap_pmd_range+0x644/0x650
[   33.067612] Code: 00 00 00 00 66 90 89 ce b9 00 f0 ff ff e9 ff fb ff ff 89 d8 31 db e8 48 c6 52 00 e9 23 fb ff ff e8 61 83 56 00 e9 b6 fe ff ff <0f> 0b bf 00 f0 ff ff e9 38 fa ff ff 3e 8d 74 26 00 55 89 e5 57 31
[   33.067615] EAX: ee394000 EBX: 00000002 ECX: ee394000 EDX: 00000000
[   33.067617] ESI: c1b0ded4 EDI: 00024a00 EBP: c1b0ddb4 ESP: c1b0dd68
[   33.067619] DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068 EFLAGS: 00010246
[   33.067624] CR0: 80050033 CR2: b7a00000 CR3: 01bbbd20 CR4: 00350ef0
[   33.067625] Call Trace:
[   33.067628]  ? madvise_free_pte_range+0x720/0x720
[   33.067632]  ? smaps_pte_range+0x4b0/0x4b0
[   33.067634]  walk_pgd_range+0x325/0x720
[   33.067637]  ? mt_find+0x1d6/0x3a0
[   33.067641]  ? mt_find+0x1d6/0x3a0
[   33.067643]  __walk_page_range+0x164/0x170
[   33.067646]  walk_page_range+0xf9/0x170
[   33.067648]  ? __kmem_cache_alloc_node+0x2a8/0x340
[   33.067653]  pagemap_read+0x124/0x280
[   33.067658]  ? default_llseek+0x101/0x160
[   33.067662]  ? smaps_account+0x1d0/0x1d0
[   33.067664]  vfs_read+0x90/0x290
[   33.067667]  ? do_madvise.part.0+0x24b/0x390
[   33.067669]  ? debug_smp_processor_id+0x12/0x20
[   33.067673]  ksys_pread64+0x58/0x90
[   33.067675]  __ia32_sys_ia32_pread64+0x1b/0x20
[   33.067680]  __do_fast_syscall_32+0x4c/0xc0
[   33.067683]  do_fast_syscall_32+0x29/0x60
[   33.067686]  do_SYSENTER_32+0x15/0x20
[   33.067689]  entry_SYSENTER_32+0x98/0xf1

Decrease the indentation level of SWP_PFN_BITS and SWP_PFN_MASK to keep it readable and consistent.

Fixes 0d206b5d2e0d ("mm/swap: add swp_offset_pfn() to fetch PFN from swap entry")
Cc Andrew Morton <[email protected]>
Cc Peter Xu <[email protected]>
Cc Yang Shi <[email protected]>
Cc Hugh Dickins <[email protected]>
Cc Andrea Arcangeli <[email protected]>

---

Quoted 8 lines
Quoted 3 lines

v1 -> v2:
* Rely on sizeof(phys_addr_t) and min_t() instead.
* Survives my various cross compilations and testing on x86 PAE.

Good to know it works, thanks a lot.

Thanks. I realized that we can simply use

	min_t(int,
instead of
	min_t(phys_addr_t,

because we're only comparing numbers < 64 ...

Might look cleaner and less complicated.

From e6f0c1099de69a11e643e4d4ee6f470147aa174e Mon Sep 17 00:00:00 2001
From: David Hildenbrand <[email protected]>
Date: Wed, 7 Dec 2022 21:15:36 +0100
Subject: [PATCH] Fixup: "mm/swap: fix SWP_PFN_BITS with
CONFIG_PHYS_ADDR_T_64BIT on 32bit"

Let's use "int" for comparison, as we're only comparing numbers < 64.

---
 include/linux/swapops.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
include/linux/swapops.h +1 -1
index a70b5c3a68d7..b982dd614572 100644
@@ -35,7 +35,7 @@
 #ifdef MAX_PHYSMEM_BITS
 #define SWP_PFN_BITS		(MAX_PHYSMEM_BITS - PAGE_SHIFT)
 #else  /* MAX_PHYSMEM_BITS */
-#define SWP_PFN_BITS		min_t(phys_addr_t, \
+#define SWP_PFN_BITS		min_t(int, \
 				      sizeof(phys_addr_t) * 8 - PAGE_SHIFT, \
 				      SWP_TYPE_SHIFT)
 #endif	/* MAX_PHYSMEM_BITS */
Signature · 10 lines
2.38.1



@Andrew, I just cross-compiled and retested with that successfully.

-- 
Thanks,

David / dhildenb
Reviewed-by on v2 1/1
#3 lore raw
In reply to David Hildenbrand
Quoted 9 lines · David Hildenbrand

We use "unsigned long" to store a PFN in the kernel and phys_addr_t to store a physical address.

On a 64bit system, both are 64bit wide. However, on a 32bit system, the latter might be 64bit wide. This is, for example, the case on x86 with
PAE: phys_addr_t and PTEs are 64bit wide, while "unsigned long" only spans 32bit.

Quoted 4 lines

The current definition of SWP_PFN_BITS without MAX_PHYSMEM_BITS misses that case, and assumes that the maximum PFN is limited by an 32bit phys_addr_t. This implies, that SWP_PFN_BITS will currently only be able to cover 4 GiB - 1 on any 32bit system with 4k page size, which is wrong.

Thanks for debugging this. IIUC this means even swap is actually broken on x86_32 + PAE?

Quoted 62 lines

Let's rely on the number of bits in phys_addr_t instead, but make sure to not exceed the maximum swap offset, to not make the BUILD_BUG_ON() in is_pfn_swap_entry() unhappy. Note that swp_entry_t is effectively an unsigned long and the maximum swap offset shares that value with the swap type.

For example, on an 8 GiB x86 PAE system with a kernel config based on Debian 11.5 (-> CONFIG_FLATMEM=y, CONFIG_X86_PAE=y), we will currently fail removing migration entries (remove_migration_ptes()), because mm/page_vma_mapped.c:check_pte() will fail to identify a PFN match as swp_offset_pfn() wrongly masks off PFN bits. For example, split_huge_page_to_list()->...->remap_page() will leave migration entries in place and continue to unlock the page.

Later, when we stumble over these migration entries (e.g., via /proc/self/pagemap), pfn_swap_entry_to_page() will BUG_ON() because these migration entries shouldn't exist anymore and the page was unlocked.

[   33.067591] kernel BUG at include/linux/swapops.h:497!
[   33.067597] invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
[   33.067602] CPU: 3 PID: 742 Comm: cow Tainted: G            E      6.1.0-rc8+ #16
[   33.067605] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.0-1.fc36 04/01/2014
[   33.067606] EIP: pagemap_pmd_range+0x644/0x650
[   33.067612] Code: 00 00 00 00 66 90 89 ce b9 00 f0 ff ff e9 ff fb ff ff 89 d8 31 db e8 48 c6 52 00 e9 23 fb ff ff e8 61 83 56 00 e9 b6 fe ff ff <0f> 0b bf 00 f0 ff ff e9 38 fa ff ff 3e 8d 74 26 00 55 89 e5 57 31
[   33.067615] EAX: ee394000 EBX: 00000002 ECX: ee394000 EDX: 00000000
[   33.067617] ESI: c1b0ded4 EDI: 00024a00 EBP: c1b0ddb4 ESP: c1b0dd68
[   33.067619] DS: 007b ES: 007b FS: 00d8 GS: 0033 SS: 0068 EFLAGS: 00010246
[   33.067624] CR0: 80050033 CR2: b7a00000 CR3: 01bbbd20 CR4: 00350ef0
[   33.067625] Call Trace:
[   33.067628]  ? madvise_free_pte_range+0x720/0x720
[   33.067632]  ? smaps_pte_range+0x4b0/0x4b0
[   33.067634]  walk_pgd_range+0x325/0x720
[   33.067637]  ? mt_find+0x1d6/0x3a0
[   33.067641]  ? mt_find+0x1d6/0x3a0
[   33.067643]  __walk_page_range+0x164/0x170
[   33.067646]  walk_page_range+0xf9/0x170
[   33.067648]  ? __kmem_cache_alloc_node+0x2a8/0x340
[   33.067653]  pagemap_read+0x124/0x280
[   33.067658]  ? default_llseek+0x101/0x160
[   33.067662]  ? smaps_account+0x1d0/0x1d0
[   33.067664]  vfs_read+0x90/0x290
[   33.067667]  ? do_madvise.part.0+0x24b/0x390
[   33.067669]  ? debug_smp_processor_id+0x12/0x20
[   33.067673]  ksys_pread64+0x58/0x90
[   33.067675]  __ia32_sys_ia32_pread64+0x1b/0x20
[   33.067680]  __do_fast_syscall_32+0x4c/0xc0
[   33.067683]  do_fast_syscall_32+0x29/0x60
[   33.067686]  do_SYSENTER_32+0x15/0x20
[   33.067689]  entry_SYSENTER_32+0x98/0xf1

Decrease the indentation level of SWP_PFN_BITS and SWP_PFN_MASK to keep it readable and consistent.

Fixes 0d206b5d2e0d ("mm/swap: add swp_offset_pfn() to fetch PFN from swap entry")
Cc Andrew Morton <[email protected]>
Cc Peter Xu <[email protected]>
Cc Yang Shi <[email protected]>
Cc Hugh Dickins <[email protected]>
Cc Andrea Arcangeli <[email protected]>
Quoted 5 lines

---

v1 -> v2:
* Rely on sizeof(phys_addr_t) and min_t() instead.
* Survives my various cross compilations and testing on x86 PAE.

Quoted 31 lines
---
 include/linux/swapops.h | 8 +++++---
 1 file changed, 5 insertions(+), 3 deletions(-)
include/linux/swapops.h +5 -3
index 86b95ccb81bb..424d586ed1ca 100644
@@ -33,11 +33,13 @@
  * can use the extra bits to store other information besides PFN.
  */
 #ifdef MAX_PHYSMEM_BITS
-#define SWP_PFN_BITS                   (MAX_PHYSMEM_BITS - PAGE_SHIFT)
+#define SWP_PFN_BITS           (MAX_PHYSMEM_BITS - PAGE_SHIFT)
 #else  /* MAX_PHYSMEM_BITS */
-#define SWP_PFN_BITS                   (BITS_PER_LONG - PAGE_SHIFT)
+#define SWP_PFN_BITS           min_t(phys_addr_t, \
+                                     sizeof(phys_addr_t) * 8 - PAGE_SHIFT, \
+                                     SWP_TYPE_SHIFT)
 #endif /* MAX_PHYSMEM_BITS */
-#define SWP_PFN_MASK                   (BIT(SWP_PFN_BITS) - 1)
+#define SWP_PFN_MASK           (BIT(SWP_PFN_BITS) - 1)
 
 /**
  * Migration swap entry specific bitfield definitions.  Layout:

base-commit: 76dcd734eca23168cb008912c0f69ff408905235 --
2.38.1

#4 lore raw
Quoted 10 lines · Yang Shi
Quoted 9 lines · David Hildenbrand

We use "unsigned long" to store a PFN in the kernel and phys_addr_t to store a physical address.

On a 64bit system, both are 64bit wide. However, on a 32bit system, the latter might be 64bit wide. This is, for example, the case on x86 with
PAE: phys_addr_t and PTEs are 64bit wide, while "unsigned long" only spans 32bit.

Quoted 7 lines
Quoted 4 lines

The current definition of SWP_PFN_BITS without MAX_PHYSMEM_BITS misses that case, and assumes that the maximum PFN is limited by an 32bit phys_addr_t. This implies, that SWP_PFN_BITS will currently only be able to cover 4 GiB - 1 on any 32bit system with 4k page size, which is wrong.

Thanks for debugging this. IIUC this means even swap is actually broken on x86_32 + PAE?

I saw all different kinds of issues while testing debugging without this patch, but they might just be a fallout from previous page migration/THP splitting issues.

I think swap should be fine, because SWP_PFN_BITS only affects swp_offset_pfn(): only used when is_pfn_swap_entry()==true.

Thanks!

Signature · 3 lines
Thanks,

David / dhildenb