On Thu, Aug 21, 2025 at 06:35:42PM +0900, Harry Yoo wrote:
KASAN unconditionally references kasan_early_shadow_{p4d,pud}. However, these global variables may not exist depending on the number of page table levels. For example, if CONFIG_PGTABLE_LEVELS=3, both variables do not exist. Although KASAN may refernce non-existent variables, it didn't break builds because calls to {pgd,p4d}_populate() are optimized away at compile time.
However, {pgd,p4d}_populate_kernel() is defined as a function regardless of the number of page table levels, so the compiler may not optimize them away. In this case, the following linker error occurs:
ld.lld: error: undefined symbol: kasan_early_shadow_p4d
referenced by init.c:260 (/home/hyeyoo/mm-new/mm/kasan/init.c:260) mm/kasan/init.o:(kasan_populate_early_shadow) in archive vmlinux.a referenced by init.c:260 (/home/hyeyoo/mm-new/mm/kasan/init.c:260) mm/kasan/init.o:(kasan_populate_early_shadow) in archive vmlinux.a did you mean: kasan_early_shadow_pmd defined in: vmlinux.a(mm/kasan/init.o)
ld.lld: error: undefined symbol: kasan_early_shadow_pud
referenced by init.c:263 (/home/hyeyoo/mm-new/mm/kasan/init.c:263) mm/kasan/init.o:(kasan_populate_early_shadow) in archive vmlinux.a referenced by init.c:263 (/home/hyeyoo/mm-new/mm/kasan/init.c:263) mm/kasan/init.o:(kasan_populate_early_shadow) in archive vmlinux.a referenced by init.c:200 (/home/hyeyoo/mm-new/mm/kasan/init.c:200) mm/kasan/init.o:(zero_p4d_populate) in archive vmlinux.a referenced 1 more times
Therefore, to allow calls to {pgd,p4d}_populate_kernel() to be optimized out at compile time, define {pgd,p4d}_populate_kernel() as macros. This way, when pgd_populate() or p4d_populate() are simply empty macros, the corresponding *_populate_kernel() functions can also be optimized away.
Signed-off-by: Harry Yoo harry.yoo@oracle.com
This looks good, other than the nit below re: a comment, I think when we are doing this kind of thing it's necessary to spell out plainly why exactly we're doing it because it's not obvious at first glance.
Anyway have checked locally and all good and LGTM code-wise so aside from above:
Reviewed-by: Lorenzo Stoakes lorenzo.stoakes@oracle.com
While the description is quite verbose, it is intended to be fold-merged into patch [1] of the page table sync series V5.
[1] https://lore.kernel.org/linux-mm/20250818020206.4517-3-harry.yoo@oracle.com/
include/linux/pgalloc.h | 26 ++++++++++++-------------- 1 file changed, 12 insertions(+), 14 deletions(-)
diff --git a/include/linux/pgalloc.h b/include/linux/pgalloc.h index 290ab864320f..8812f842978f 100644 --- a/include/linux/pgalloc.h +++ b/include/linux/pgalloc.h @@ -5,20 +5,18 @@ #include <linux/pgtable.h> #include <asm/pgalloc.h>
-static inline void pgd_populate_kernel(unsigned long addr, pgd_t *pgd,
p4d_t *p4d)-{
- pgd_populate(&init_mm, pgd, p4d);
- if (ARCH_PAGE_TABLE_SYNC_MASK & PGTBL_PGD_MODIFIED)
arch_sync_kernel_mappings(addr, addr);-} +#define pgd_populate_kernel(addr, pgd, p4d) \
- do { \
pgd_populate(&init_mm, pgd, p4d); \if (ARCH_PAGE_TABLE_SYNC_MASK & PGTBL_PGD_MODIFIED) \arch_sync_kernel_mappings(addr, addr); \- } while (0)
-static inline void p4d_populate_kernel(unsigned long addr, p4d_t *p4d,
pud_t *pud)-{
- p4d_populate(&init_mm, p4d, pud);
- if (ARCH_PAGE_TABLE_SYNC_MASK & PGTBL_P4D_MODIFIED)
arch_sync_kernel_mappings(addr, addr);-} +#define p4d_populate_kernel(addr, p4d, pud) \
- do { \
p4d_populate(&init_mm, p4d, pud); \if (ARCH_PAGE_TABLE_SYNC_MASK & PGTBL_P4D_MODIFIED) \arch_sync_kernel_mappings(addr, addr); \- } while (0)
Can we have a quick comment above these explaining why they have to be macros? Thanks!
#endif /* _LINUX_PGALLOC_H */
2.43.0