> On Thu Aug 20, 2026 at 3:40 PM EDT, Yeoreum Yun wrote: > > Since glibc commit 321e1fc73f (“malloc: Enable 2MB THP by default on > > AArch64”), > > glibc may call madvise(MADV_HUGEPAGE) for sufficiently large allocations > > made by memalign(). > > > > The underlying VMA may start at a different address from the aligned > > address returned by memalign(). Furthermore, a subsequent > > madvise(MADV_HUGEPAGE) call does not split the VMA because the flag is > > already set. > > > > This causes split_huge_page_test to fail because the check_huge_xxx() > > helpers incorrectly require the address returned by memalign() to > > match the VMA start address reported in /proc/self/smaps. > > > > Fix this by using /proc/self/pagemap and /proc/kpageflags instead of > > /proc/self/smaps to detect huge pages and change the meaning of > > check_huge_xxx()'s nr_hpages argument: > > Have you checked Baolin's patches[1] in mm-new? They resue > gather_after_split_folio_orders() to reimplement check_huge_xxx(), also > based on pagemap and kpageflags. Does it fix the issue? > > [1] > https://lore.kernel.org/all/[email protected]/ >
Unfortunately, No. since __check_pmd_huge() in check_huge_xxx() still use /proc/self/smaps [1] for pmd THP, it still has problem though ths patch series applied. [1] https://lore.kernel.org/all/56b16691f605426b33b5cf47319233de6127a6b3.1785985999.git.baolin.w...@linux.alibaba.com/ > > > > - nr_hpages > 0: check all of pages in the range are huge page. > > - nr_hpages < 0: check all of pages in the range are not huge page. > > - nr_hpages == 0: invalid. > > > > Reported-by: David Hildenbrand (Arm) <[email protected]> > > Signed-off-by: Yeoreum Yun <[email protected]> > > --- > > tools/testing/selftests/mm/khugepaged.c | 26 ++--- > > tools/testing/selftests/mm/split_huge_page_test.c | 6 +- > > tools/testing/selftests/mm/vm_util.c | 135 > > ++++++++++++++++++---- > > tools/testing/selftests/mm/vm_util.h | 1 + > > 4 files changed, 131 insertions(+), 37 deletions(-) > > > > > -- > Best Regards, > Yan, Zi > -- Sincerely, Yeoreum Yun
