[<prev] [next>] [day] [month] [year] [list]
Message-ID: <20241020042212.296781-1-yuzhao@google.com>
Date: Sat, 19 Oct 2024 22:22:12 -0600
From: Yu Zhao <yuzhao@...gle.com>
To: Andrew Morton <akpm@...ux-foundation.org>,
"Matthew Wilcox (Oracle)" <willy@...radead.org>
Cc: Muchun Song <muchun.song@...ux.dev>, linux-mm@...ck.org, linux-kernel@...r.kernel.org,
Yu Zhao <yuzhao@...gle.com>, stable@...r.kernel.org
Subject: [PATCH mm-unstable v1] mm: allow set/clear page_type again
Some page flags (page->flags) were converted to page types
(page->page_types). A recent example is PG_hugetlb.
>From the exclusive writer's perspective, e.g., a thread doing
__folio_set_hugetlb(), there is a difference between the page flag and
type APIs: the former allows the same non-atomic operation to be
repeated whereas the latter does not. For example, calling
__folio_set_hugetlb() twice triggers VM_BUG_ON_FOLIO(), since the
second call expects the type (PG_hugetlb) not to be set previously.
Using add_hugetlb_folio() as an example, it calls
__folio_set_hugetlb() in the following error-handling path. And when
that happens, it triggers the aforementioned VM_BUG_ON_FOLIO().
if (folio_test_hugetlb(folio)) {
rc = hugetlb_vmemmap_restore_folio(h, folio);
if (rc) {
spin_lock_irq(&hugetlb_lock);
add_hugetlb_folio(h, folio, false);
...
It is possible to make hugeTLB comply with the new requirements from
the page type API. However, a straightforward fix would be to just
allow the same page type to be set or cleared again inside the API,
to avoid any changes to its callers.
Fixes: d99e3140a4d3 ("mm: turn folio_test_hugetlb into a PageType")
Signed-off-by: Yu Zhao <yuzhao@...gle.com>
Cc: <stable@...r.kernel.org>
---
include/linux/page-flags.h | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
index ccf3c78faefc..e80665bc51fa 100644
--- a/include/linux/page-flags.h
+++ b/include/linux/page-flags.h
@@ -977,12 +977,16 @@ static __always_inline bool folio_test_##fname(const struct folio *folio) \
} \
static __always_inline void __folio_set_##fname(struct folio *folio) \
{ \
+ if (folio_test_##fname(folio)) \
+ return; \
VM_BUG_ON_FOLIO(data_race(folio->page.page_type) != UINT_MAX, \
folio); \
folio->page.page_type = (unsigned int)PGTY_##lname << 24; \
} \
static __always_inline void __folio_clear_##fname(struct folio *folio) \
{ \
+ if (folio->page.page_type == UINT_MAX) \
+ return; \
VM_BUG_ON_FOLIO(!folio_test_##fname(folio), folio); \
folio->page.page_type = UINT_MAX; \
}
@@ -995,11 +999,15 @@ static __always_inline int Page##uname(const struct page *page) \
} \
static __always_inline void __SetPage##uname(struct page *page) \
{ \
+ if (Page##uname(page)) \
+ return; \
VM_BUG_ON_PAGE(data_race(page->page_type) != UINT_MAX, page); \
page->page_type = (unsigned int)PGTY_##lname << 24; \
} \
static __always_inline void __ClearPage##uname(struct page *page) \
{ \
+ if (page->page_type == UINT_MAX) \
+ return; \
VM_BUG_ON_PAGE(!Page##uname(page), page); \
page->page_type = UINT_MAX; \
}
--
2.47.0.rc1.288.g06298d1525-goog
Powered by blists - more mailing lists