On Fri, Jun 28, 2024 at 5:01 PM kernel test robot <lkp@xxxxxxxxx> wrote: > > tree: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master > head: 1eb586a9782cde8e5091b9de74603e0a8386b09e > commit: ee3fe497bdf0c7cb6fbec346f36aa1f9b1dfe6dd [9137/9689] mm: gup: do not call try_grab_folio() in slow path > config: powerpc-randconfig-r005-20221108 > compiler: powerpc-linux-gcc (GCC) 13.2.0 > reproduce (this is a W=1 build): > > If you fix the issue in a separate patch/commit (i.e. not just a new version of > the same patch/commit), kindly add following tags > | Reported-by: kernel test robot <lkp@xxxxxxxxx> > | Closes: https://lore.kernel.org/oe-kbuild-all/202406290729.21FzkVlI-lkp@xxxxxxxxx/ > > All error/warnings (new ones prefixed by >>): > > mm/gup.c: In function 'gup_hugepte': > >> mm/gup.c:476:25: error: implicit declaration of function 'try_grab_folio_fast'; did you mean 'try_grab_folio'? [-Werror=implicit-function-declaration] The definition of try_grab_folio_fast() should be protected by CONFIG_ARCH_HAS_HUGEPD || CONFIG_HAVE_GUP_FAST. Andrew, Please apply the attached fix on top of it. The reproducer requires x86 machine since the cross build tool is on x86, but I don't have a x86 machine by hand right now. But I think the patch should fix it. > 476 | folio = try_grab_folio_fast(page, refs, flags); > | ^~~~~~~~~~~~~~~~~~~ > | try_grab_folio > >> mm/gup.c:476:23: warning: assignment to 'struct folio *' from 'int' makes pointer from integer without a cast [-Wint-conversion] > 476 | folio = try_grab_folio_fast(page, refs, flags); > | ^ > mm/gup.c: At top level: > >> mm/gup.c:2748:22: error: conflicting types for 'try_grab_folio_fast'; have 'struct folio *(struct page *, int, unsigned int)' > 2748 | static struct folio *try_grab_folio_fast(struct page *page, int refs, > | ^~~~~~~~~~~~~~~~~~~ > mm/gup.c:476:25: note: previous implicit declaration of 'try_grab_folio_fast' with type 'int()' > 476 | folio = try_grab_folio_fast(page, refs, flags); > | ^~~~~~~~~~~~~~~~~~~ > cc1: some warnings being treated as errors > > > vim +476 mm/gup.c > > 441 > 442 /* > 443 * Returns 1 if succeeded, 0 if failed, -EMLINK if unshare needed. > 444 * > 445 * NOTE: for the same entry, gup-fast and gup-slow can return different > 446 * results (0 v.s. -EMLINK) depending on whether vma is available. This is > 447 * the expected behavior, where we simply want gup-fast to fallback to > 448 * gup-slow to take the vma reference first. > 449 */ > 450 static int gup_hugepte(struct vm_area_struct *vma, pte_t *ptep, unsigned long sz, > 451 unsigned long addr, unsigned long end, unsigned int flags, > 452 struct page **pages, int *nr, bool fast) > 453 { > 454 unsigned long pte_end; > 455 struct page *page; > 456 struct folio *folio; > 457 pte_t pte; > 458 int refs; > 459 > 460 pte_end = (addr + sz) & ~(sz-1); > 461 if (pte_end < end) > 462 end = pte_end; > 463 > 464 pte = huge_ptep_get(ptep); > 465 > 466 if (!pte_access_permitted(pte, flags & FOLL_WRITE)) > 467 return 0; > 468 > 469 /* hugepages are never "special" */ > 470 VM_BUG_ON(!pfn_valid(pte_pfn(pte))); > 471 > 472 page = pte_page(pte); > 473 refs = record_subpages(page, sz, addr, end, pages + *nr); > 474 > 475 if (fast) { > > 476 folio = try_grab_folio_fast(page, refs, flags); > 477 if (!folio) > 478 return 0; > 479 } else { > 480 folio = page_folio(page); > 481 if (try_grab_folio(folio, refs, flags)) > 482 return 0; > 483 } > 484 > 485 if (unlikely(pte_val(pte) != pte_val(ptep_get(ptep)))) { > 486 gup_put_folio(folio, refs, flags); > 487 return 0; > 488 } > 489 > 490 if (!pte_write(pte) && gup_must_unshare(vma, flags, &folio->page)) { > 491 gup_put_folio(folio, refs, flags); > 492 return -EMLINK; > 493 } > 494 > 495 *nr += refs; > 496 folio_set_referenced(folio); > 497 return 1; > 498 } > 499 > > -- > 0-DAY CI Kernel Test Service > https://github.com/intel/lkp-tests/wiki
Attachment:
mm-gup-stop-abusing-try_grab_folio-fix.patch
Description: Binary data
