On Thu Aug 20, 2026 at 8:26 PM EDT, Yeoreum Yun wrote: > On Thu, Aug 20, 2026 at 08:09:12PM -0400, Zi Yan wrote: >> On Thu Aug 20, 2026 at 5:26 PM EDT, Yeoreum Yun wrote: >> >> 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(). >> >> I just googled the commit and find that Dev did that. ;) >> >> >> > >> >> > 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/ >> >> In that case, is it possible to use and extend check_large_folios() for >> all check_huge_xxx()? You still need pagemap_scan_get_categories() to >> check PAGE_IS_HUGE to identify huge mappings. Or at least >> check_huge_xxx() in your patch can share most of the code. > > Agree. but TBH, I think check_large_folios() can replace checking > PAGE_IS_HUGE and keep the later part to check wehther PAGE_IS_FILE
IIUC, PAGE_IS_HUGE comes from checking page table entry (see pagemap_thp_category()), whereas check_large_folios() checks folios, namely physical memory. A PTE-mapped order-9 (4KB base page) folio is !PAGE_IS_HUGE but check_large_folios() returns true for PMD_SIZE. check_huge_xxx() are checking PMD-mapped folios, so PAGE_IS_HUGE is functionally equivalent. > and SWAPBACKED according to check_huge_xxx(). > -- Best Regards, Yan, Zi
