Skip to contentSkip to conversation

mm: Prevent derefencing NULL ptr in pfn_section_valid()

PATCH Applied Waiman Long 2 participants 3 messages Jun 26, 2024 – Jul 2, 2024

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 mm: Prevent derefencing NULL ptr in pfn_section_valid()

Commit 5ec8e8ea8b77 ("mm/sparsemem: fix race in accessing memory_section->usage") changed pfn_section_valid() to add a READ_ONCE() call around "ms->usage" to fix a race with section_deactivate() where ms->usage can be cleared. The READ_ONCE() call, by itself, is not enough to prevent NULL pointer dereference. We need to check its value before dereferencing it.

Fixes 5ec8e8ea8b77 ("mm/sparsemem: fix race in accessing memory_section->usage")
---
 include/linux/mmzone.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
include/linux/mmzone.h +2 -1
index 8f9c9590a42c..b1dcf6ddb406 100644
@@ -1980,8 +1980,9 @@ static inline int subsection_map_index(unsigned long pfn)
 static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn)
 {
 	int idx = subsection_map_index(pfn);
+	struct mem_section_usage *usage = READ_ONCE(ms->usage);
 
-	return test_bit(idx, READ_ONCE(ms->usage)->subsection_map);
+	return usage ? test_bit(idx, usage->subsection_map) : 0;
 }
 #else
 static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn)
Signature · 1 line
2.39.3
#1 lore raw

Hi Waiman,

Quoted 6 lines · Waiman Long

Commit 5ec8e8ea8b77 ("mm/sparsemem: fix race in accessing memory_section->usage") changed pfn_section_valid() to add a READ_ONCE() call around "ms->usage" to fix a race with section_deactivate() where ms->usage can be cleared. The READ_ONCE() call, by itself, is not enough to prevent NULL pointer dereference. We need to check its value before dereferencing it.

I am unable to see a scenario where ms->usage will be NULL when pfn_section_valid() is called:

1) In pfn_valid, valid_section() check ensures that pfn_section_valid() is not called as the section is marked as invalid.

2) In pfn_to_online_page, online_section() check ensures that pfn_section_valid() is not called.

and in the update path, we do:

     kfree_rcu(ms->usage, rcu);
     WRITE_ONCE(ms->usage, NULL);

Could you help me in understanding about what I am missing here, please?

Quoted 22 lines
Fixes 5ec8e8ea8b77 ("mm/sparsemem: fix race in accessing memory_section->usage")
---
 include/linux/mmzone.h | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
include/linux/mmzone.h +2 -1
index 8f9c9590a42c..b1dcf6ddb406 100644
@@ -1980,8 +1980,9 @@ static inline int subsection_map_index(unsigned long pfn)
 static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn)
 {
 	int idx = subsection_map_index(pfn);
+	struct mem_section_usage *usage = READ_ONCE(ms->usage);
 
-	return test_bit(idx, READ_ONCE(ms->usage)->subsection_map);
+	return usage ? test_bit(idx, usage->subsection_map) : 0;
 }
 #else
 static inline int pfn_section_valid(struct mem_section *ms, unsigned long pfn)
#2 lore raw
Quoted 15 lines · Charan Teja Kalla

Hi Waiman,

Quoted 6 lines · Waiman Long

Commit 5ec8e8ea8b77 ("mm/sparsemem: fix race in accessing memory_section->usage") changed pfn_section_valid() to add a READ_ONCE() call around "ms->usage" to fix a race with section_deactivate() where ms->usage can be cleared. The READ_ONCE() call, by itself, is not enough to prevent NULL pointer dereference. We need to check its value before dereferencing it.

I am unable to see a scenario where ms->usage will be NULL when pfn_section_valid() is called:

1) In pfn_valid, valid_section() check ensures that pfn_section_valid() is not called as the section is marked as invalid.

Quoted 9 lines

2) In pfn_to_online_page, online_section() check ensures that pfn_section_valid() is not called.

and in the update path, we do:

      kfree_rcu(ms->usage, rcu);
      WRITE_ONCE(ms->usage, NULL);

Could you help me in understanding about what I am missing here, please?

With the below timing sequence:

     CPU 0                                      CPU 1 -----                                      ----- if (!valid_section(ms)) return 0; ms->section_mem_map &=
    ~SECTION_HAS_MEM_MAP                <interrupt>

 WRITE_ONCE(ms->usage, NULL); READ_ONCE(ms->usage)->subsection_map

In the time gap between valid_section() check and accessing ms->usage, it may have been cleared leading to dereferencing a NULL pointer. That is why it will be prudent to do a NULL check first.

Regards,
Longman