[PATCH 31/31] perf/core: Allow pte_offset_map() to fail

From: Hugh Dickins
Date: Mon May 22 2023 - 01:29:16 EST


In rare transient cases, not yet made possible, pte_offset_map() and
pte_offet_map_lock() may not find a page table: handle appropriately.

Signed-off-by: Hugh Dickins <hughd@xxxxxxxxxx>
---
This is a perf patch, not an mm patch, and it will want to go in through
the tip tree in due course; but keep it in this series for now, so that
it's not missed, and not submitted before mm review.

kernel/events/core.c | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/kernel/events/core.c b/kernel/events/core.c
index db016e418931..174be710f3b3 100644
--- a/kernel/events/core.c
+++ b/kernel/events/core.c
@@ -7490,6 +7490,7 @@ static u64 perf_get_pgtable_size(struct mm_struct *mm, unsigned long addr)
return pud_leaf_size(pud);

pmdp = pmd_offset_lockless(pudp, pud, addr);
+again:
pmd = pmdp_get_lockless(pmdp);
if (!pmd_present(pmd))
return 0;
@@ -7498,6 +7499,9 @@ static u64 perf_get_pgtable_size(struct mm_struct *mm, unsigned long addr)
return pmd_leaf_size(pmd);

ptep = pte_offset_map(&pmd, addr);
+ if (!ptep)
+ goto again;
+
pte = ptep_get_lockless(ptep);
if (pte_present(pte))
size = pte_leaf_size(pte);
--
2.35.3