Commit 95042f9eb78a8d9a17455e2ef263f2f310ecef15
1 parent
be85bccaa5
Exists in
master
and in
4 other branches
vm: fix mlock() on stack guard page
Commit 53a7706d5ed8 ("mlock: do not hold mmap_sem for extended periods of time") changed mlock() to care about the exact number of pages that __get_user_pages() had brought it. Before, it would only care about errors. And that doesn't work, because we also handled one page specially in __mlock_vma_pages_range(), namely the stack guard page. So when that case was handled, the number of pages that the function returned was off by one. In particular, it could be zero, and then the caller would end up not making any progress at all. Rather than try to fix up that off-by-one error for the mlock case specially, this just moves the logic to handle the stack guard page into__get_user_pages() itself, thus making all the counts come out right automatically. Reported-by: Robert Święcki <robert@swiecki.net> Cc: Hugh Dickins <hughd@google.com> Cc: Oleg Nesterov <oleg@redhat.com> Cc: stable@kernel.org Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Showing 2 changed files with 18 additions and 21 deletions Side-by-side Diff
mm/memory.c
| ... | ... | @@ -1410,6 +1410,13 @@ |
| 1410 | 1410 | return page; |
| 1411 | 1411 | } |
| 1412 | 1412 | |
| 1413 | +static inline int stack_guard_page(struct vm_area_struct *vma, unsigned long addr) | |
| 1414 | +{ | |
| 1415 | + return (vma->vm_flags & VM_GROWSDOWN) && | |
| 1416 | + (vma->vm_start == addr) && | |
| 1417 | + !vma_stack_continue(vma->vm_prev, addr); | |
| 1418 | +} | |
| 1419 | + | |
| 1413 | 1420 | /** |
| 1414 | 1421 | * __get_user_pages() - pin user pages in memory |
| 1415 | 1422 | * @tsk: task_struct of target task |
| ... | ... | @@ -1488,7 +1495,6 @@ |
| 1488 | 1495 | vma = find_extend_vma(mm, start); |
| 1489 | 1496 | if (!vma && in_gate_area(mm, start)) { |
| 1490 | 1497 | unsigned long pg = start & PAGE_MASK; |
| 1491 | - struct vm_area_struct *gate_vma = get_gate_vma(mm); | |
| 1492 | 1498 | pgd_t *pgd; |
| 1493 | 1499 | pud_t *pud; |
| 1494 | 1500 | pmd_t *pmd; |
| 1495 | 1501 | |
| ... | ... | @@ -1513,10 +1519,11 @@ |
| 1513 | 1519 | pte_unmap(pte); |
| 1514 | 1520 | return i ? : -EFAULT; |
| 1515 | 1521 | } |
| 1522 | + vma = get_gate_vma(mm); | |
| 1516 | 1523 | if (pages) { |
| 1517 | 1524 | struct page *page; |
| 1518 | 1525 | |
| 1519 | - page = vm_normal_page(gate_vma, start, *pte); | |
| 1526 | + page = vm_normal_page(vma, start, *pte); | |
| 1520 | 1527 | if (!page) { |
| 1521 | 1528 | if (!(gup_flags & FOLL_DUMP) && |
| 1522 | 1529 | is_zero_pfn(pte_pfn(*pte))) |
| ... | ... | @@ -1530,12 +1537,7 @@ |
| 1530 | 1537 | get_page(page); |
| 1531 | 1538 | } |
| 1532 | 1539 | pte_unmap(pte); |
| 1533 | - if (vmas) | |
| 1534 | - vmas[i] = gate_vma; | |
| 1535 | - i++; | |
| 1536 | - start += PAGE_SIZE; | |
| 1537 | - nr_pages--; | |
| 1538 | - continue; | |
| 1540 | + goto next_page; | |
| 1539 | 1541 | } |
| 1540 | 1542 | |
| 1541 | 1543 | if (!vma || |
| ... | ... | @@ -1549,6 +1551,13 @@ |
| 1549 | 1551 | continue; |
| 1550 | 1552 | } |
| 1551 | 1553 | |
| 1554 | + /* | |
| 1555 | + * If we don't actually want the page itself, | |
| 1556 | + * and it's the stack guard page, just skip it. | |
| 1557 | + */ | |
| 1558 | + if (!pages && stack_guard_page(vma, start)) | |
| 1559 | + goto next_page; | |
| 1560 | + | |
| 1552 | 1561 | do { |
| 1553 | 1562 | struct page *page; |
| 1554 | 1563 | unsigned int foll_flags = gup_flags; |
| ... | ... | @@ -1631,6 +1640,7 @@ |
| 1631 | 1640 | flush_anon_page(vma, page, start); |
| 1632 | 1641 | flush_dcache_page(page); |
| 1633 | 1642 | } |
| 1643 | +next_page: | |
| 1634 | 1644 | if (vmas) |
| 1635 | 1645 | vmas[i] = vma; |
| 1636 | 1646 | i++; |
mm/mlock.c
| ... | ... | @@ -135,13 +135,6 @@ |
| 135 | 135 | } |
| 136 | 136 | } |
| 137 | 137 | |
| 138 | -static inline int stack_guard_page(struct vm_area_struct *vma, unsigned long addr) | |
| 139 | -{ | |
| 140 | - return (vma->vm_flags & VM_GROWSDOWN) && | |
| 141 | - (vma->vm_start == addr) && | |
| 142 | - !vma_stack_continue(vma->vm_prev, addr); | |
| 143 | -} | |
| 144 | - | |
| 145 | 138 | /** |
| 146 | 139 | * __mlock_vma_pages_range() - mlock a range of pages in the vma. |
| 147 | 140 | * @vma: target vma |
| ... | ... | @@ -187,12 +180,6 @@ |
| 187 | 180 | |
| 188 | 181 | if (vma->vm_flags & VM_LOCKED) |
| 189 | 182 | gup_flags |= FOLL_MLOCK; |
| 190 | - | |
| 191 | - /* We don't try to access the guard page of a stack vma */ | |
| 192 | - if (stack_guard_page(vma, start)) { | |
| 193 | - addr += PAGE_SIZE; | |
| 194 | - nr_pages--; | |
| 195 | - } | |
| 196 | 183 | |
| 197 | 184 | return __get_user_pages(current, mm, addr, nr_pages, gup_flags, |
| 198 | 185 | NULL, NULL, nonblocking); |