Re: [PATCH v9 06/15] kpageflags: Use is_page_hwpoison() to set KPF_HWPOISON

["David Hildenbrand (Arm)" <david@kernel.org>]

On 8/13/26 09:25, David Hildenbrand (Arm) wrote:
> On 8/5/26 23:05, Matthew Wilcox (Oracle) wrote:
>> Instead of knowing how hugetlb handles hwpoison, just ask
>> is_page_hwpoison().  This gives us the flexibility to change how hwpoison
>> is handled without updating this function in the future.
>>
>> Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
>> Reviewed-by: Jane Chu <jane.chu@oracle.com>
>> Reviewed-by: Gregory Price (Meta) <gourry@gourry.net>
>> ---
>>  fs/proc/page.c | 8 ++------
>>  1 file changed, 2 insertions(+), 6 deletions(-)
>>
>> diff --git a/fs/proc/page.c b/fs/proc/page.c
>> index 7d9387143435..323e4c049f69 100644
>> --- a/fs/proc/page.c
>> +++ b/fs/proc/page.c
>> @@ -233,12 +233,8 @@ u64 stable_page_flags(const struct page *page)
>>  	u |= kpf_copy_bit(k, KPF_UNEVICTABLE,	PG_unevictable);
>>  	u |= kpf_copy_bit(k, KPF_MLOCKED,	PG_mlocked);
>>  
>> -#ifdef CONFIG_MEMORY_FAILURE
>> -	if (u & (1 << KPF_HUGE))
>> -		u |= kpf_copy_bit(k, KPF_HWPOISON,	PG_hwpoison);
>> -	else
>> -		u |= kpf_copy_bit(ps.page_snapshot.flags.f, KPF_HWPOISON, PG_hwpoison);
>> -#endif
>> +	if (is_page_hwpoison(page))
>> +		u |= 1 << KPF_HWPOISON;

I also just realized, that other users in this function now use BIT_ULL.

> 
> snapshot_page() takes a snapshot of the folio and the page. If you look at
> "page", you start going to the live version again.
> 

If we really cannot use the snapshot_page, one alternative to avoid using "page"
after actually snapshotting it would be:

diff --git a/fs/proc/page.c b/fs/proc/page.c
index 260772b20bd99..f2fc0917f0b6c 100644
--- a/fs/proc/page.c
+++ b/fs/proc/page.c
@@ -223,12 +223,8 @@ u64 stable_page_flags(const struct page *page)
        u |= kpf_copy_bit(k, KPF_UNEVICTABLE,   PG_unevictable);
        u |= kpf_copy_bit(k, KPF_MLOCKED,       PG_mlocked);
 
-#ifdef CONFIG_MEMORY_FAILURE
-       if (u & BIT_ULL(KPF_HUGE))
-               u |= kpf_copy_bit(k, KPF_HWPOISON,      PG_hwpoison);
-       else
-               u |= kpf_copy_bit(ps.page_snapshot.flags.f, KPF_HWPOISON, PG_hwpoison);
-#endif
+       if (ps.flags & PAGE_SNAPSHOT_HWPOISON)
+               u |= BIT_ULL(KPF_HWPOISON);
 
        u |= kpf_copy_bit(k, KPF_RESERVED,      PG_reserved);
        u |= kpf_copy_bit(k, KPF_OWNER_2,       PG_owner_2);
diff --git a/include/linux/mm.h b/include/linux/mm.h
index 29f13cc6b52a2..7e14e58b98c38 100644
--- a/include/linux/mm.h
+++ b/include/linux/mm.h
@@ -5496,6 +5496,7 @@ static inline bool page_pool_page_is_pp(const struct page *page)
 #define PAGE_SNAPSHOT_FAITHFUL (1 << 0)
 #define PAGE_SNAPSHOT_PG_BUDDY (1 << 1)
 #define PAGE_SNAPSHOT_PG_IDLE  (1 << 2)
+#define PAGE_SNAPSHOT_HWPOISON (1 << 3)
 
 struct page_snapshot {
        struct folio folio_snapshot;
diff --git a/mm/util.c b/mm/util.c
index bf0513d1d3d08..06195d0801130 100644
--- a/mm/util.c
+++ b/mm/util.c
@@ -1294,6 +1294,8 @@ static void set_ps_flags(struct page_snapshot *ps, const struct folio *folio,
                ps->flags |= PAGE_SNAPSHOT_PG_BUDDY;
        else if (page_count(page) == 0 && is_free_buddy_page(page))
                ps->flags |= PAGE_SNAPSHOT_PG_BUDDY;
+       if (is_page_hwpoison(page))
+               ps->flags |= PAGE_SNAPSHOT_HWPOISON;
 
        if (folio_test_idle(folio))
                ps->flags |= PAGE_SNAPSHOT_PG_IDLE;


(I'm starting to hate all these different hwpoison helpers)

-- 
Cheers,

David