mm/page_owner: Don't access uninitialized memmaps when reading /proc/pagetypeinfo

PATCH Applied David Hildenbrand 3 participants 4 messages Oct 11, 2019 – Oct 14, 2019

Subsystems (2)

Memory management Page allocator

lore

Share & tools

lore

David Hildenbrand

author

2019-10-11 14:06 UTC #0 lore raw

PATCH v2 1/1 mm/page_owner: Don't access uninitialized memmaps when reading /proc/pagetypeinfo

From: Qian Cai <[email protected]>

Uninitialized memmaps contain garbage and in the worst case trigger kernel BUGs, especially with CONFIG_PAGE_POISONING. They should not get touched.

For example, when not onlining a memory block that is spanned by a zone and reading /proc/pagetypeinfo with CONFIG_DEBUG_VM_PGFLAGS and CONFIG_PAGE_POISONING, we can trigger a kernel BUG:

:/# echo 1 > /sys/devices/system/memory/memory40/online
:/# echo 1 > /sys/devices/system/memory/memory42/online
:/# cat /proc/pagetypeinfo > test.file
  [   42.489856] page:fffff2c585200000 is uninitialized and poisoned
  [   42.489861] raw: ffffffffffffffff ffffffffffffffff ffffffffffffffff ffffffffffffffff
  [   42.492235] raw: ffffffffffffffff ffffffffffffffff ffffffffffffffff ffffffffffffffff
  [   42.493501] page dumped because: VM_BUG_ON_PAGE(PagePoisoned(p))
  [   42.494533] There is not page extension available.
  [   42.495358] ------------[ cut here ]------------
  [   42.496163] kernel BUG at include/linux/mm.h:1107!
  [   42.497069] invalid opcode: 0000 [#1] SMP NOPTI

Please not that this change does not affect ZONE_DEVICE, because pagetypeinfo_showmixedcount_print() is called from mm/vmstat.c:pagetypeinfo_showmixedcount() only for populated zones, and ZONE_DEVICE is never populated (zone->present_pages always 0).

Fixes f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online") # visible after d0dc12e86b319

Signed-off-by Qian Cai <[email protected]>

Cc Andrew Morton <[email protected]>

Cc Vlastimil Babka <[email protected]>

Cc Michal Hocko <[email protected]>

Cc Thomas Gleixner <[email protected]>

Cc "Peter Zijlstra (Intel)" <[email protected]>

Cc Miles Chen <[email protected]>

Cc Mike Rapoport <[email protected]>

Cc Qian Cai <[email protected]>

Cc Greg Kroah-Hartman <[email protected]>

[ move check to outer loop, add comment, rephrase description ]

Signed-off-by David Hildenbrand <[email protected]>

---

Cai asked me to follow up on:
[PATCH] mm/page_owner: fix a crash after memory offline

---
 mm/page_owner.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

mm/page_owner.c +3 -2

index dee931184788..7d149211f6be 100644
@@ -284,7 +284,8 @@ void pagetypeinfo_showmixedcount_print(struct seq_file *m,
 	 * not matter as the mixed block count will still be correct
 	 */
 	for (; pfn < end_pfn; ) {
-		if (!pfn_valid(pfn)) {
+		page = pfn_to_online_page(pfn);
+		if (!page) {
 			pfn = ALIGN(pfn + 1, MAX_ORDER_NR_PAGES);
 			continue;
 		}

@@ -292,13 +293,13 @@ void pagetypeinfo_showmixedcount_print(struct seq_file *m,
 		block_end_pfn = ALIGN(pfn + 1, pageblock_nr_pages);
 		block_end_pfn = min(block_end_pfn, end_pfn);
 
-		page = pfn_to_page(pfn);
 		pageblock_mt = get_pageblock_migratetype(page);
 
 		for (; pfn < block_end_pfn; pfn++) {
 			if (!pfn_valid_within(pfn))
 				continue;
 
+			/* The pageblock is online, no need to recheck. */
 			page = pfn_to_page(pfn);
 
 			if (page_zone(page) != zone)

Signature · 1 line

2.21.0

Vlastimil Babka

Acked-by on v2 1/1

2019-10-14 08:31 UTC #1 lore raw

Quoted 27 lines · David Hildenbrand

From: Qian Cai <[email protected]>
Uninitialized memmaps contain garbage and in the worst case trigger kernel BUGs, especially with CONFIG_PAGE_POISONING. They should not get touched.
For example, when not onlining a memory block that is spanned by a zone and reading /proc/pagetypeinfo with CONFIG_DEBUG_VM_PGFLAGS and CONFIG_PAGE_POISONING, we can trigger a kernel BUG:
:/# echo 1 > /sys/devices/system/memory/memory40/online
:/# echo 1 > /sys/devices/system/memory/memory42/online
:/# cat /proc/pagetypeinfo > test.file
  [   42.489856] page:fffff2c585200000 is uninitialized and poisoned
  [   42.489861] raw: ffffffffffffffff ffffffffffffffff ffffffffffffffff ffffffffffffffff
  [   42.492235] raw: ffffffffffffffff ffffffffffffffff ffffffffffffffff ffffffffffffffff
  [   42.493501] page dumped because: VM_BUG_ON_PAGE(PagePoisoned(p))
  [   42.494533] There is not page extension available.
  [   42.495358] ------------[ cut here ]------------
  [   42.496163] kernel BUG at include/linux/mm.h:1107!
  [   42.497069] invalid opcode: 0000 [#1] SMP NOPTI
Please not that this change does not affect ZONE_DEVICE, because pagetypeinfo_showmixedcount_print() is called from mm/vmstat.c:pagetypeinfo_showmixedcount() only for populated zones, and ZONE_DEVICE is never populated (zone->present_pages always 0).

Quoted 13 lines

Fixes f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online") # visible after d0dc12e86b319
Signed-off-by Qian Cai <[email protected]>
Cc Andrew Morton <[email protected]>
Cc Vlastimil Babka <[email protected]>
Cc Michal Hocko <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc "Peter Zijlstra (Intel)" <[email protected]>
Cc Miles Chen <[email protected]>
Cc Mike Rapoport <[email protected]>
Cc Qian Cai <[email protected]>
Cc Greg Kroah-Hartman <[email protected]>
[ move check to outer loop, add comment, rephrase description ]
Signed-off-by David Hildenbrand <[email protected]>

Acked-by Vlastimil Babka <[email protected]>

Quoted 39 lines

---

Cai asked me to follow up on:
[PATCH] mm/page_owner: fix a crash after memory offline

---
 mm/page_owner.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

mm/page_owner.c +3 -2

index dee931184788..7d149211f6be 100644
@@ -284,7 +284,8 @@ void pagetypeinfo_showmixedcount_print(struct seq_file *m,
 	 * not matter as the mixed block count will still be correct
 	 */
 	for (; pfn < end_pfn; ) {
-		if (!pfn_valid(pfn)) {
+		page = pfn_to_online_page(pfn);
+		if (!page) {
 			pfn = ALIGN(pfn + 1, MAX_ORDER_NR_PAGES);
 			continue;
 		}

@@ -292,13 +293,13 @@ void pagetypeinfo_showmixedcount_print(struct seq_file *m,
 		block_end_pfn = ALIGN(pfn + 1, pageblock_nr_pages);
 		block_end_pfn = min(block_end_pfn, end_pfn);
 
-		page = pfn_to_page(pfn);
 		pageblock_mt = get_pageblock_migratetype(page);
 
 		for (; pfn < block_end_pfn; pfn++) {
 			if (!pfn_valid_within(pfn))
 				continue;
 
+			/* The pageblock is online, no need to recheck. */
 			page = pfn_to_page(pfn);
 
 			if (page_zone(page) != zone)

David Hildenbrand

author

2019-10-14 08:36 UTC #2 lore raw

In reply to David Hildenbrand

Quoted 69 lines · David Hildenbrand

From: Qian Cai <[email protected]>
Uninitialized memmaps contain garbage and in the worst case trigger kernel BUGs, especially with CONFIG_PAGE_POISONING. They should not get touched.
For example, when not onlining a memory block that is spanned by a zone and reading /proc/pagetypeinfo with CONFIG_DEBUG_VM_PGFLAGS and CONFIG_PAGE_POISONING, we can trigger a kernel BUG:
:/# echo 1 > /sys/devices/system/memory/memory40/online
:/# echo 1 > /sys/devices/system/memory/memory42/online
:/# cat /proc/pagetypeinfo > test.file
   [   42.489856] page:fffff2c585200000 is uninitialized and poisoned
   [   42.489861] raw: ffffffffffffffff ffffffffffffffff ffffffffffffffff ffffffffffffffff
   [   42.492235] raw: ffffffffffffffff ffffffffffffffff ffffffffffffffff ffffffffffffffff
   [   42.493501] page dumped because: VM_BUG_ON_PAGE(PagePoisoned(p))
   [   42.494533] There is not page extension available.
   [   42.495358] ------------[ cut here ]------------
   [   42.496163] kernel BUG at include/linux/mm.h:1107!
   [   42.497069] invalid opcode: 0000 [#1] SMP NOPTI
Please not that this change does not affect ZONE_DEVICE, because pagetypeinfo_showmixedcount_print() is called from mm/vmstat.c:pagetypeinfo_showmixedcount() only for populated zones, and ZONE_DEVICE is never populated (zone->present_pages always 0).
Fixes f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online") # visible after d0dc12e86b319
Signed-off-by Qian Cai <[email protected]>
Cc Andrew Morton <[email protected]>
Cc Vlastimil Babka <[email protected]>
Cc Michal Hocko <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc "Peter Zijlstra (Intel)" <[email protected]>
Cc Miles Chen <[email protected]>
Cc Mike Rapoport <[email protected]>
Cc Qian Cai <[email protected]>
Cc Greg Kroah-Hartman <[email protected]>
[ move check to outer loop, add comment, rephrase description ]
Signed-off-by David Hildenbrand <[email protected]>
---
Cai asked me to follow up on:
[PATCH] mm/page_owner: fix a crash after memory offline
---
  mm/page_owner.c | 5 +++--
  1 file changed, 3 insertions(+), 2 deletions(-)
mm/page_owner.c +2 -2
index dee931184788..7d149211f6be 100644
@@ -284,7 +284,8 @@ void pagetypeinfo_showmixedcount_print(struct seq_file *m,
  	 * not matter as the mixed block count will still be correct
  	 */
  	for (; pfn < end_pfn; ) {
-		if (!pfn_valid(pfn)) {
+		page = pfn_to_online_page(pfn);
+		if (!page) {
  			pfn = ALIGN(pfn + 1, MAX_ORDER_NR_PAGES);
  			continue;
  		}
@@ -292,13 +293,13 @@ void pagetypeinfo_showmixedcount_print(struct seq_file *m,
  		block_end_pfn = ALIGN(pfn + 1, pageblock_nr_pages);
  		block_end_pfn = min(block_end_pfn, end_pfn);
 
-		page = pfn_to_page(pfn);

mm/page_owner.c void pagetypeinfo_showmixedcount_print(struct seq_file *m, · line 305

Quoted 10 lines

mm/page_owner.c +1 -0

  		pageblock_mt = get_pageblock_migratetype(page);
 
  		for (; pfn < block_end_pfn; pfn++) {
  			if (!pfn_valid_within(pfn))
  				continue;
 
+			/* The pageblock is online, no need to recheck. */
  			page = pfn_to_page(pfn);
 
  			if (page_zone(page) != zone)

I guess it's best to just

Cc [email protected] # v4.13+

here as well. Can be cheery-picked.

@Andrew, can you add that?

Signature · 3 lines

Thanks,

David / dhildenb

Michal Hocko

Acked-by on v2 1/1

2019-10-14 08:44 UTC #3 lore raw

In reply to David Hildenbrand

Quoted 27 lines · David Hildenbrand

From: Qian Cai <[email protected]>
Uninitialized memmaps contain garbage and in the worst case trigger kernel BUGs, especially with CONFIG_PAGE_POISONING. They should not get touched.
For example, when not onlining a memory block that is spanned by a zone and reading /proc/pagetypeinfo with CONFIG_DEBUG_VM_PGFLAGS and CONFIG_PAGE_POISONING, we can trigger a kernel BUG:
:/# echo 1 > /sys/devices/system/memory/memory40/online
:/# echo 1 > /sys/devices/system/memory/memory42/online
:/# cat /proc/pagetypeinfo > test.file
  [   42.489856] page:fffff2c585200000 is uninitialized and poisoned
  [   42.489861] raw: ffffffffffffffff ffffffffffffffff ffffffffffffffff ffffffffffffffff
  [   42.492235] raw: ffffffffffffffff ffffffffffffffff ffffffffffffffff ffffffffffffffff
  [   42.493501] page dumped because: VM_BUG_ON_PAGE(PagePoisoned(p))
  [   42.494533] There is not page extension available.
  [   42.495358] ------------[ cut here ]------------
  [   42.496163] kernel BUG at include/linux/mm.h:1107!
  [   42.497069] invalid opcode: 0000 [#1] SMP NOPTI
Please not that this change does not affect ZONE_DEVICE, because pagetypeinfo_showmixedcount_print() is called from mm/vmstat.c:pagetypeinfo_showmixedcount() only for populated zones, and ZONE_DEVICE is never populated (zone->present_pages always 0).

Quoted 13 lines

Fixes f1dd2cd13c4b ("mm, memory_hotplug: do not associate hotadded memory to zones until online") # visible after d0dc12e86b319
Signed-off-by Qian Cai <[email protected]>
Cc Andrew Morton <[email protected]>
Cc Vlastimil Babka <[email protected]>
Cc Michal Hocko <[email protected]>
Cc Thomas Gleixner <[email protected]>
Cc "Peter Zijlstra (Intel)" <[email protected]>
Cc Miles Chen <[email protected]>
Cc Mike Rapoport <[email protected]>
Cc Qian Cai <[email protected]>
Cc Greg Kroah-Hartman <[email protected]>
[ move check to outer loop, add comment, rephrase description ]
Signed-off-by David Hildenbrand <[email protected]>

Acked-by Michal Hocko <[email protected]>

Thanks!

Quoted 41 lines

---

Cai asked me to follow up on:
[PATCH] mm/page_owner: fix a crash after memory offline

---
 mm/page_owner.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

mm/page_owner.c +3 -2

index dee931184788..7d149211f6be 100644
@@ -284,7 +284,8 @@ void pagetypeinfo_showmixedcount_print(struct seq_file *m,
 	 * not matter as the mixed block count will still be correct
 	 */
 	for (; pfn < end_pfn; ) {
-		if (!pfn_valid(pfn)) {
+		page = pfn_to_online_page(pfn);
+		if (!page) {
 			pfn = ALIGN(pfn + 1, MAX_ORDER_NR_PAGES);
 			continue;
 		}

@@ -292,13 +293,13 @@ void pagetypeinfo_showmixedcount_print(struct seq_file *m,
 		block_end_pfn = ALIGN(pfn + 1, pageblock_nr_pages);
 		block_end_pfn = min(block_end_pfn, end_pfn);
 
-		page = pfn_to_page(pfn);
 		pageblock_mt = get_pageblock_migratetype(page);
 
 		for (; pfn < block_end_pfn; pfn++) {
 			if (!pfn_valid_within(pfn))
 				continue;
 
+			/* The pageblock is online, no need to recheck. */
 			page = pfn_to_page(pfn);
 
 			if (page_zone(page) != zone)

Signature · 1 line

2.21.0

Signature · 2 lines

Michal Hocko
SUSE Labs