diff options
Diffstat (limited to 'mm')
-rw-r--r-- | mm/bootmem.c | 6 | ||||
-rw-r--r-- | mm/hugetlb.c | 2 | ||||
-rw-r--r-- | mm/memory.c | 69 | ||||
-rw-r--r-- | mm/migrate.c | 10 | ||||
-rw-r--r-- | mm/mmap.c | 12 | ||||
-rw-r--r-- | mm/nommu.c | 21 | ||||
-rw-r--r-- | mm/page_alloc.c | 42 | ||||
-rw-r--r-- | mm/pagewalk.c | 42 | ||||
-rw-r--r-- | mm/slab.c | 5 | ||||
-rw-r--r-- | mm/slob.c | 5 | ||||
-rw-r--r-- | mm/slub.c | 5 | ||||
-rw-r--r-- | mm/swap.c | 4 | ||||
-rw-r--r-- | mm/vmscan.c | 2 |
13 files changed, 142 insertions, 83 deletions
diff --git a/mm/bootmem.c b/mm/bootmem.c index e8fb927392b9..8d9f60e06f62 100644 --- a/mm/bootmem.c +++ b/mm/bootmem.c | |||
@@ -442,15 +442,17 @@ unsigned long __init init_bootmem_node(pg_data_t *pgdat, unsigned long freepfn, | |||
442 | return init_bootmem_core(pgdat, freepfn, startpfn, endpfn); | 442 | return init_bootmem_core(pgdat, freepfn, startpfn, endpfn); |
443 | } | 443 | } |
444 | 444 | ||
445 | void __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, | 445 | int __init reserve_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, |
446 | unsigned long size, int flags) | 446 | unsigned long size, int flags) |
447 | { | 447 | { |
448 | int ret; | 448 | int ret; |
449 | 449 | ||
450 | ret = can_reserve_bootmem_core(pgdat->bdata, physaddr, size, flags); | 450 | ret = can_reserve_bootmem_core(pgdat->bdata, physaddr, size, flags); |
451 | if (ret < 0) | 451 | if (ret < 0) |
452 | return; | 452 | return -ENOMEM; |
453 | reserve_bootmem_core(pgdat->bdata, physaddr, size, flags); | 453 | reserve_bootmem_core(pgdat->bdata, physaddr, size, flags); |
454 | |||
455 | return 0; | ||
454 | } | 456 | } |
455 | 457 | ||
456 | void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, | 458 | void __init free_bootmem_node(pg_data_t *pgdat, unsigned long physaddr, |
diff --git a/mm/hugetlb.c b/mm/hugetlb.c index bbf953eeb58b..ab171274ef21 100644 --- a/mm/hugetlb.c +++ b/mm/hugetlb.c | |||
@@ -785,7 +785,7 @@ int copy_hugetlb_page_range(struct mm_struct *dst, struct mm_struct *src, | |||
785 | continue; | 785 | continue; |
786 | 786 | ||
787 | spin_lock(&dst->page_table_lock); | 787 | spin_lock(&dst->page_table_lock); |
788 | spin_lock(&src->page_table_lock); | 788 | spin_lock_nested(&src->page_table_lock, SINGLE_DEPTH_NESTING); |
789 | if (!huge_pte_none(huge_ptep_get(src_pte))) { | 789 | if (!huge_pte_none(huge_ptep_get(src_pte))) { |
790 | if (cow) | 790 | if (cow) |
791 | huge_ptep_set_wrprotect(src, addr, src_pte); | 791 | huge_ptep_set_wrprotect(src, addr, src_pte); |
diff --git a/mm/memory.c b/mm/memory.c index fb5608a120ed..d14b251a25a6 100644 --- a/mm/memory.c +++ b/mm/memory.c | |||
@@ -999,17 +999,15 @@ struct page *follow_page(struct vm_area_struct *vma, unsigned long address, | |||
999 | goto no_page_table; | 999 | goto no_page_table; |
1000 | 1000 | ||
1001 | ptep = pte_offset_map_lock(mm, pmd, address, &ptl); | 1001 | ptep = pte_offset_map_lock(mm, pmd, address, &ptl); |
1002 | if (!ptep) | ||
1003 | goto out; | ||
1004 | 1002 | ||
1005 | pte = *ptep; | 1003 | pte = *ptep; |
1006 | if (!pte_present(pte)) | 1004 | if (!pte_present(pte)) |
1007 | goto unlock; | 1005 | goto no_page; |
1008 | if ((flags & FOLL_WRITE) && !pte_write(pte)) | 1006 | if ((flags & FOLL_WRITE) && !pte_write(pte)) |
1009 | goto unlock; | 1007 | goto unlock; |
1010 | page = vm_normal_page(vma, address, pte); | 1008 | page = vm_normal_page(vma, address, pte); |
1011 | if (unlikely(!page)) | 1009 | if (unlikely(!page)) |
1012 | goto unlock; | 1010 | goto bad_page; |
1013 | 1011 | ||
1014 | if (flags & FOLL_GET) | 1012 | if (flags & FOLL_GET) |
1015 | get_page(page); | 1013 | get_page(page); |
@@ -1024,6 +1022,15 @@ unlock: | |||
1024 | out: | 1022 | out: |
1025 | return page; | 1023 | return page; |
1026 | 1024 | ||
1025 | bad_page: | ||
1026 | pte_unmap_unlock(ptep, ptl); | ||
1027 | return ERR_PTR(-EFAULT); | ||
1028 | |||
1029 | no_page: | ||
1030 | pte_unmap_unlock(ptep, ptl); | ||
1031 | if (!pte_none(pte)) | ||
1032 | return page; | ||
1033 | /* Fall through to ZERO_PAGE handling */ | ||
1027 | no_page_table: | 1034 | no_page_table: |
1028 | /* | 1035 | /* |
1029 | * When core dumping an enormous anonymous area that nobody | 1036 | * When core dumping an enormous anonymous area that nobody |
@@ -1038,6 +1045,26 @@ no_page_table: | |||
1038 | return page; | 1045 | return page; |
1039 | } | 1046 | } |
1040 | 1047 | ||
1048 | /* Can we do the FOLL_ANON optimization? */ | ||
1049 | static inline int use_zero_page(struct vm_area_struct *vma) | ||
1050 | { | ||
1051 | /* | ||
1052 | * We don't want to optimize FOLL_ANON for make_pages_present() | ||
1053 | * when it tries to page in a VM_LOCKED region. As to VM_SHARED, | ||
1054 | * we want to get the page from the page tables to make sure | ||
1055 | * that we serialize and update with any other user of that | ||
1056 | * mapping. | ||
1057 | */ | ||
1058 | if (vma->vm_flags & (VM_LOCKED | VM_SHARED)) | ||
1059 | return 0; | ||
1060 | /* | ||
1061 | * And if we have a fault or a nopfn routine, it's not an | ||
1062 | * anonymous region. | ||
1063 | */ | ||
1064 | return !vma->vm_ops || | ||
1065 | (!vma->vm_ops->fault && !vma->vm_ops->nopfn); | ||
1066 | } | ||
1067 | |||
1041 | int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, | 1068 | int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, |
1042 | unsigned long start, int len, int write, int force, | 1069 | unsigned long start, int len, int write, int force, |
1043 | struct page **pages, struct vm_area_struct **vmas) | 1070 | struct page **pages, struct vm_area_struct **vmas) |
@@ -1112,8 +1139,7 @@ int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, | |||
1112 | foll_flags = FOLL_TOUCH; | 1139 | foll_flags = FOLL_TOUCH; |
1113 | if (pages) | 1140 | if (pages) |
1114 | foll_flags |= FOLL_GET; | 1141 | foll_flags |= FOLL_GET; |
1115 | if (!write && !(vma->vm_flags & VM_LOCKED) && | 1142 | if (!write && use_zero_page(vma)) |
1116 | (!vma->vm_ops || !vma->vm_ops->fault)) | ||
1117 | foll_flags |= FOLL_ANON; | 1143 | foll_flags |= FOLL_ANON; |
1118 | 1144 | ||
1119 | do { | 1145 | do { |
@@ -1159,6 +1185,8 @@ int get_user_pages(struct task_struct *tsk, struct mm_struct *mm, | |||
1159 | 1185 | ||
1160 | cond_resched(); | 1186 | cond_resched(); |
1161 | } | 1187 | } |
1188 | if (IS_ERR(page)) | ||
1189 | return i ? i : PTR_ERR(page); | ||
1162 | if (pages) { | 1190 | if (pages) { |
1163 | pages[i] = page; | 1191 | pages[i] = page; |
1164 | 1192 | ||
@@ -1757,7 +1785,6 @@ gotten: | |||
1757 | page_table = pte_offset_map_lock(mm, pmd, address, &ptl); | 1785 | page_table = pte_offset_map_lock(mm, pmd, address, &ptl); |
1758 | if (likely(pte_same(*page_table, orig_pte))) { | 1786 | if (likely(pte_same(*page_table, orig_pte))) { |
1759 | if (old_page) { | 1787 | if (old_page) { |
1760 | page_remove_rmap(old_page, vma); | ||
1761 | if (!PageAnon(old_page)) { | 1788 | if (!PageAnon(old_page)) { |
1762 | dec_mm_counter(mm, file_rss); | 1789 | dec_mm_counter(mm, file_rss); |
1763 | inc_mm_counter(mm, anon_rss); | 1790 | inc_mm_counter(mm, anon_rss); |
@@ -1779,6 +1806,32 @@ gotten: | |||
1779 | lru_cache_add_active(new_page); | 1806 | lru_cache_add_active(new_page); |
1780 | page_add_new_anon_rmap(new_page, vma, address); | 1807 | page_add_new_anon_rmap(new_page, vma, address); |
1781 | 1808 | ||
1809 | if (old_page) { | ||
1810 | /* | ||
1811 | * Only after switching the pte to the new page may | ||
1812 | * we remove the mapcount here. Otherwise another | ||
1813 | * process may come and find the rmap count decremented | ||
1814 | * before the pte is switched to the new page, and | ||
1815 | * "reuse" the old page writing into it while our pte | ||
1816 | * here still points into it and can be read by other | ||
1817 | * threads. | ||
1818 | * | ||
1819 | * The critical issue is to order this | ||
1820 | * page_remove_rmap with the ptp_clear_flush above. | ||
1821 | * Those stores are ordered by (if nothing else,) | ||
1822 | * the barrier present in the atomic_add_negative | ||
1823 | * in page_remove_rmap. | ||
1824 | * | ||
1825 | * Then the TLB flush in ptep_clear_flush ensures that | ||
1826 | * no process can access the old page before the | ||
1827 | * decremented mapcount is visible. And the old page | ||
1828 | * cannot be reused until after the decremented | ||
1829 | * mapcount is visible. So transitively, TLBs to | ||
1830 | * old page will be flushed before it can be reused. | ||
1831 | */ | ||
1832 | page_remove_rmap(old_page, vma); | ||
1833 | } | ||
1834 | |||
1782 | /* Free the old page.. */ | 1835 | /* Free the old page.. */ |
1783 | new_page = old_page; | 1836 | new_page = old_page; |
1784 | ret |= VM_FAULT_WRITE; | 1837 | ret |= VM_FAULT_WRITE; |
@@ -2295,8 +2348,6 @@ static int __do_fault(struct mm_struct *mm, struct vm_area_struct *vma, | |||
2295 | vmf.flags = flags; | 2348 | vmf.flags = flags; |
2296 | vmf.page = NULL; | 2349 | vmf.page = NULL; |
2297 | 2350 | ||
2298 | BUG_ON(vma->vm_flags & VM_PFNMAP); | ||
2299 | |||
2300 | ret = vma->vm_ops->fault(vma, &vmf); | 2351 | ret = vma->vm_ops->fault(vma, &vmf); |
2301 | if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) | 2352 | if (unlikely(ret & (VM_FAULT_ERROR | VM_FAULT_NOPAGE))) |
2302 | return ret; | 2353 | return ret; |
diff --git a/mm/migrate.c b/mm/migrate.c index 449d77d409f5..112bcaeaa104 100644 --- a/mm/migrate.c +++ b/mm/migrate.c | |||
@@ -865,6 +865,11 @@ static int do_move_pages(struct mm_struct *mm, struct page_to_node *pm, | |||
865 | goto set_status; | 865 | goto set_status; |
866 | 866 | ||
867 | page = follow_page(vma, pp->addr, FOLL_GET); | 867 | page = follow_page(vma, pp->addr, FOLL_GET); |
868 | |||
869 | err = PTR_ERR(page); | ||
870 | if (IS_ERR(page)) | ||
871 | goto set_status; | ||
872 | |||
868 | err = -ENOENT; | 873 | err = -ENOENT; |
869 | if (!page) | 874 | if (!page) |
870 | goto set_status; | 875 | goto set_status; |
@@ -928,6 +933,11 @@ static int do_pages_stat(struct mm_struct *mm, struct page_to_node *pm) | |||
928 | goto set_status; | 933 | goto set_status; |
929 | 934 | ||
930 | page = follow_page(vma, pm->addr, 0); | 935 | page = follow_page(vma, pm->addr, 0); |
936 | |||
937 | err = PTR_ERR(page); | ||
938 | if (IS_ERR(page)) | ||
939 | goto set_status; | ||
940 | |||
931 | err = -ENOENT; | 941 | err = -ENOENT; |
932 | /* Use PageReserved to check for zero page */ | 942 | /* Use PageReserved to check for zero page */ |
933 | if (!page || PageReserved(page)) | 943 | if (!page || PageReserved(page)) |
@@ -80,7 +80,7 @@ EXPORT_SYMBOL(vm_get_page_prot); | |||
80 | int sysctl_overcommit_memory = OVERCOMMIT_GUESS; /* heuristic overcommit */ | 80 | int sysctl_overcommit_memory = OVERCOMMIT_GUESS; /* heuristic overcommit */ |
81 | int sysctl_overcommit_ratio = 50; /* default is 50% */ | 81 | int sysctl_overcommit_ratio = 50; /* default is 50% */ |
82 | int sysctl_max_map_count __read_mostly = DEFAULT_MAX_MAP_COUNT; | 82 | int sysctl_max_map_count __read_mostly = DEFAULT_MAX_MAP_COUNT; |
83 | atomic_t vm_committed_space = ATOMIC_INIT(0); | 83 | atomic_long_t vm_committed_space = ATOMIC_LONG_INIT(0); |
84 | 84 | ||
85 | /* | 85 | /* |
86 | * Check that a process has enough memory to allocate a new virtual | 86 | * Check that a process has enough memory to allocate a new virtual |
@@ -177,7 +177,7 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) | |||
177 | * cast `allowed' as a signed long because vm_committed_space | 177 | * cast `allowed' as a signed long because vm_committed_space |
178 | * sometimes has a negative value | 178 | * sometimes has a negative value |
179 | */ | 179 | */ |
180 | if (atomic_read(&vm_committed_space) < (long)allowed) | 180 | if (atomic_long_read(&vm_committed_space) < (long)allowed) |
181 | return 0; | 181 | return 0; |
182 | error: | 182 | error: |
183 | vm_unacct_memory(pages); | 183 | vm_unacct_memory(pages); |
@@ -245,10 +245,16 @@ asmlinkage unsigned long sys_brk(unsigned long brk) | |||
245 | unsigned long rlim, retval; | 245 | unsigned long rlim, retval; |
246 | unsigned long newbrk, oldbrk; | 246 | unsigned long newbrk, oldbrk; |
247 | struct mm_struct *mm = current->mm; | 247 | struct mm_struct *mm = current->mm; |
248 | unsigned long min_brk; | ||
248 | 249 | ||
249 | down_write(&mm->mmap_sem); | 250 | down_write(&mm->mmap_sem); |
250 | 251 | ||
251 | if (brk < mm->start_brk) | 252 | #ifdef CONFIG_COMPAT_BRK |
253 | min_brk = mm->end_code; | ||
254 | #else | ||
255 | min_brk = mm->start_brk; | ||
256 | #endif | ||
257 | if (brk < min_brk) | ||
252 | goto out; | 258 | goto out; |
253 | 259 | ||
254 | /* | 260 | /* |
diff --git a/mm/nommu.c b/mm/nommu.c index ef8c62cec697..4462b6a3fcb9 100644 --- a/mm/nommu.c +++ b/mm/nommu.c | |||
@@ -39,7 +39,7 @@ struct page *mem_map; | |||
39 | unsigned long max_mapnr; | 39 | unsigned long max_mapnr; |
40 | unsigned long num_physpages; | 40 | unsigned long num_physpages; |
41 | unsigned long askedalloc, realalloc; | 41 | unsigned long askedalloc, realalloc; |
42 | atomic_t vm_committed_space = ATOMIC_INIT(0); | 42 | atomic_long_t vm_committed_space = ATOMIC_LONG_INIT(0); |
43 | int sysctl_overcommit_memory = OVERCOMMIT_GUESS; /* heuristic overcommit */ | 43 | int sysctl_overcommit_memory = OVERCOMMIT_GUESS; /* heuristic overcommit */ |
44 | int sysctl_overcommit_ratio = 50; /* default is 50% */ | 44 | int sysctl_overcommit_ratio = 50; /* default is 50% */ |
45 | int sysctl_max_map_count = DEFAULT_MAX_MAP_COUNT; | 45 | int sysctl_max_map_count = DEFAULT_MAX_MAP_COUNT; |
@@ -109,16 +109,23 @@ unsigned int kobjsize(const void *objp) | |||
109 | * If the object we have should not have ksize performed on it, | 109 | * If the object we have should not have ksize performed on it, |
110 | * return size of 0 | 110 | * return size of 0 |
111 | */ | 111 | */ |
112 | if (!objp || (unsigned long)objp >= memory_end || !((page = virt_to_page(objp)))) | 112 | if (!objp || !virt_addr_valid(objp)) |
113 | return 0; | 113 | return 0; |
114 | 114 | ||
115 | page = virt_to_head_page(objp); | ||
116 | |||
117 | /* | ||
118 | * If the allocator sets PageSlab, we know the pointer came from | ||
119 | * kmalloc(). | ||
120 | */ | ||
115 | if (PageSlab(page)) | 121 | if (PageSlab(page)) |
116 | return ksize(objp); | 122 | return ksize(objp); |
117 | 123 | ||
118 | BUG_ON(page->index < 0); | 124 | /* |
119 | BUG_ON(page->index >= MAX_ORDER); | 125 | * The ksize() function is only guaranteed to work for pointers |
120 | 126 | * returned by kmalloc(). So handle arbitrary pointers here. | |
121 | return (PAGE_SIZE << page->index); | 127 | */ |
128 | return PAGE_SIZE << compound_order(page); | ||
122 | } | 129 | } |
123 | 130 | ||
124 | /* | 131 | /* |
@@ -1410,7 +1417,7 @@ int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin) | |||
1410 | * cast `allowed' as a signed long because vm_committed_space | 1417 | * cast `allowed' as a signed long because vm_committed_space |
1411 | * sometimes has a negative value | 1418 | * sometimes has a negative value |
1412 | */ | 1419 | */ |
1413 | if (atomic_read(&vm_committed_space) < (long)allowed) | 1420 | if (atomic_long_read(&vm_committed_space) < (long)allowed) |
1414 | return 0; | 1421 | return 0; |
1415 | error: | 1422 | error: |
1416 | vm_unacct_memory(pages); | 1423 | vm_unacct_memory(pages); |
diff --git a/mm/page_alloc.c b/mm/page_alloc.c index 63835579323a..2f552955a02f 100644 --- a/mm/page_alloc.c +++ b/mm/page_alloc.c | |||
@@ -237,16 +237,7 @@ static void bad_page(struct page *page) | |||
237 | printk(KERN_EMERG "Trying to fix it up, but a reboot is needed\n" | 237 | printk(KERN_EMERG "Trying to fix it up, but a reboot is needed\n" |
238 | KERN_EMERG "Backtrace:\n"); | 238 | KERN_EMERG "Backtrace:\n"); |
239 | dump_stack(); | 239 | dump_stack(); |
240 | page->flags &= ~(1 << PG_lru | | 240 | page->flags &= ~PAGE_FLAGS_CLEAR_WHEN_BAD; |
241 | 1 << PG_private | | ||
242 | 1 << PG_locked | | ||
243 | 1 << PG_active | | ||
244 | 1 << PG_dirty | | ||
245 | 1 << PG_reclaim | | ||
246 | 1 << PG_slab | | ||
247 | 1 << PG_swapcache | | ||
248 | 1 << PG_writeback | | ||
249 | 1 << PG_buddy ); | ||
250 | set_page_count(page, 0); | 241 | set_page_count(page, 0); |
251 | reset_page_mapcount(page); | 242 | reset_page_mapcount(page); |
252 | page->mapping = NULL; | 243 | page->mapping = NULL; |
@@ -463,16 +454,7 @@ static inline int free_pages_check(struct page *page) | |||
463 | (page->mapping != NULL) | | 454 | (page->mapping != NULL) | |
464 | (page_get_page_cgroup(page) != NULL) | | 455 | (page_get_page_cgroup(page) != NULL) | |
465 | (page_count(page) != 0) | | 456 | (page_count(page) != 0) | |
466 | (page->flags & ( | 457 | (page->flags & PAGE_FLAGS_CHECK_AT_FREE))) |
467 | 1 << PG_lru | | ||
468 | 1 << PG_private | | ||
469 | 1 << PG_locked | | ||
470 | 1 << PG_active | | ||
471 | 1 << PG_slab | | ||
472 | 1 << PG_swapcache | | ||
473 | 1 << PG_writeback | | ||
474 | 1 << PG_reserved | | ||
475 | 1 << PG_buddy )))) | ||
476 | bad_page(page); | 458 | bad_page(page); |
477 | if (PageDirty(page)) | 459 | if (PageDirty(page)) |
478 | __ClearPageDirty(page); | 460 | __ClearPageDirty(page); |
@@ -616,17 +598,7 @@ static int prep_new_page(struct page *page, int order, gfp_t gfp_flags) | |||
616 | (page->mapping != NULL) | | 598 | (page->mapping != NULL) | |
617 | (page_get_page_cgroup(page) != NULL) | | 599 | (page_get_page_cgroup(page) != NULL) | |
618 | (page_count(page) != 0) | | 600 | (page_count(page) != 0) | |
619 | (page->flags & ( | 601 | (page->flags & PAGE_FLAGS_CHECK_AT_PREP))) |
620 | 1 << PG_lru | | ||
621 | 1 << PG_private | | ||
622 | 1 << PG_locked | | ||
623 | 1 << PG_active | | ||
624 | 1 << PG_dirty | | ||
625 | 1 << PG_slab | | ||
626 | 1 << PG_swapcache | | ||
627 | 1 << PG_writeback | | ||
628 | 1 << PG_reserved | | ||
629 | 1 << PG_buddy )))) | ||
630 | bad_page(page); | 602 | bad_page(page); |
631 | 603 | ||
632 | /* | 604 | /* |
@@ -1396,6 +1368,9 @@ get_page_from_freelist(gfp_t gfp_mask, nodemask_t *nodemask, unsigned int order, | |||
1396 | 1368 | ||
1397 | (void)first_zones_zonelist(zonelist, high_zoneidx, nodemask, | 1369 | (void)first_zones_zonelist(zonelist, high_zoneidx, nodemask, |
1398 | &preferred_zone); | 1370 | &preferred_zone); |
1371 | if (!preferred_zone) | ||
1372 | return NULL; | ||
1373 | |||
1399 | classzone_idx = zone_idx(preferred_zone); | 1374 | classzone_idx = zone_idx(preferred_zone); |
1400 | 1375 | ||
1401 | zonelist_scan: | 1376 | zonelist_scan: |
@@ -2804,7 +2779,7 @@ int zone_wait_table_init(struct zone *zone, unsigned long zone_size_pages) | |||
2804 | alloc_size = zone->wait_table_hash_nr_entries | 2779 | alloc_size = zone->wait_table_hash_nr_entries |
2805 | * sizeof(wait_queue_head_t); | 2780 | * sizeof(wait_queue_head_t); |
2806 | 2781 | ||
2807 | if (system_state == SYSTEM_BOOTING) { | 2782 | if (!slab_is_available()) { |
2808 | zone->wait_table = (wait_queue_head_t *) | 2783 | zone->wait_table = (wait_queue_head_t *) |
2809 | alloc_bootmem_node(pgdat, alloc_size); | 2784 | alloc_bootmem_node(pgdat, alloc_size); |
2810 | } else { | 2785 | } else { |
@@ -3378,7 +3353,8 @@ static void __paginginit free_area_init_core(struct pglist_data *pgdat, | |||
3378 | * is used by this zone for memmap. This affects the watermark | 3353 | * is used by this zone for memmap. This affects the watermark |
3379 | * and per-cpu initialisations | 3354 | * and per-cpu initialisations |
3380 | */ | 3355 | */ |
3381 | memmap_pages = (size * sizeof(struct page)) >> PAGE_SHIFT; | 3356 | memmap_pages = |
3357 | PAGE_ALIGN(size * sizeof(struct page)) >> PAGE_SHIFT; | ||
3382 | if (realsize >= memmap_pages) { | 3358 | if (realsize >= memmap_pages) { |
3383 | realsize -= memmap_pages; | 3359 | realsize -= memmap_pages; |
3384 | printk(KERN_DEBUG | 3360 | printk(KERN_DEBUG |
diff --git a/mm/pagewalk.c b/mm/pagewalk.c index 0afd2387e507..d5878bed7841 100644 --- a/mm/pagewalk.c +++ b/mm/pagewalk.c | |||
@@ -3,14 +3,14 @@ | |||
3 | #include <linux/sched.h> | 3 | #include <linux/sched.h> |
4 | 4 | ||
5 | static int walk_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, | 5 | static int walk_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, |
6 | const struct mm_walk *walk, void *private) | 6 | struct mm_walk *walk) |
7 | { | 7 | { |
8 | pte_t *pte; | 8 | pte_t *pte; |
9 | int err = 0; | 9 | int err = 0; |
10 | 10 | ||
11 | pte = pte_offset_map(pmd, addr); | 11 | pte = pte_offset_map(pmd, addr); |
12 | for (;;) { | 12 | for (;;) { |
13 | err = walk->pte_entry(pte, addr, addr + PAGE_SIZE, private); | 13 | err = walk->pte_entry(pte, addr, addr + PAGE_SIZE, walk); |
14 | if (err) | 14 | if (err) |
15 | break; | 15 | break; |
16 | addr += PAGE_SIZE; | 16 | addr += PAGE_SIZE; |
@@ -24,7 +24,7 @@ static int walk_pte_range(pmd_t *pmd, unsigned long addr, unsigned long end, | |||
24 | } | 24 | } |
25 | 25 | ||
26 | static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end, | 26 | static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end, |
27 | const struct mm_walk *walk, void *private) | 27 | struct mm_walk *walk) |
28 | { | 28 | { |
29 | pmd_t *pmd; | 29 | pmd_t *pmd; |
30 | unsigned long next; | 30 | unsigned long next; |
@@ -35,15 +35,15 @@ static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end, | |||
35 | next = pmd_addr_end(addr, end); | 35 | next = pmd_addr_end(addr, end); |
36 | if (pmd_none_or_clear_bad(pmd)) { | 36 | if (pmd_none_or_clear_bad(pmd)) { |
37 | if (walk->pte_hole) | 37 | if (walk->pte_hole) |
38 | err = walk->pte_hole(addr, next, private); | 38 | err = walk->pte_hole(addr, next, walk); |
39 | if (err) | 39 | if (err) |
40 | break; | 40 | break; |
41 | continue; | 41 | continue; |
42 | } | 42 | } |
43 | if (walk->pmd_entry) | 43 | if (walk->pmd_entry) |
44 | err = walk->pmd_entry(pmd, addr, next, private); | 44 | err = walk->pmd_entry(pmd, addr, next, walk); |
45 | if (!err && walk->pte_entry) | 45 | if (!err && walk->pte_entry) |
46 | err = walk_pte_range(pmd, addr, next, walk, private); | 46 | err = walk_pte_range(pmd, addr, next, walk); |
47 | if (err) | 47 | if (err) |
48 | break; | 48 | break; |
49 | } while (pmd++, addr = next, addr != end); | 49 | } while (pmd++, addr = next, addr != end); |
@@ -52,7 +52,7 @@ static int walk_pmd_range(pud_t *pud, unsigned long addr, unsigned long end, | |||
52 | } | 52 | } |
53 | 53 | ||
54 | static int walk_pud_range(pgd_t *pgd, unsigned long addr, unsigned long end, | 54 | static int walk_pud_range(pgd_t *pgd, unsigned long addr, unsigned long end, |
55 | const struct mm_walk *walk, void *private) | 55 | struct mm_walk *walk) |
56 | { | 56 | { |
57 | pud_t *pud; | 57 | pud_t *pud; |
58 | unsigned long next; | 58 | unsigned long next; |
@@ -63,15 +63,15 @@ static int walk_pud_range(pgd_t *pgd, unsigned long addr, unsigned long end, | |||
63 | next = pud_addr_end(addr, end); | 63 | next = pud_addr_end(addr, end); |
64 | if (pud_none_or_clear_bad(pud)) { | 64 | if (pud_none_or_clear_bad(pud)) { |
65 | if (walk->pte_hole) | 65 | if (walk->pte_hole) |
66 | err = walk->pte_hole(addr, next, private); | 66 | err = walk->pte_hole(addr, next, walk); |
67 | if (err) | 67 | if (err) |
68 | break; | 68 | break; |
69 | continue; | 69 | continue; |
70 | } | 70 | } |
71 | if (walk->pud_entry) | 71 | if (walk->pud_entry) |
72 | err = walk->pud_entry(pud, addr, next, private); | 72 | err = walk->pud_entry(pud, addr, next, walk); |
73 | if (!err && (walk->pmd_entry || walk->pte_entry)) | 73 | if (!err && (walk->pmd_entry || walk->pte_entry)) |
74 | err = walk_pmd_range(pud, addr, next, walk, private); | 74 | err = walk_pmd_range(pud, addr, next, walk); |
75 | if (err) | 75 | if (err) |
76 | break; | 76 | break; |
77 | } while (pud++, addr = next, addr != end); | 77 | } while (pud++, addr = next, addr != end); |
@@ -85,15 +85,15 @@ static int walk_pud_range(pgd_t *pgd, unsigned long addr, unsigned long end, | |||
85 | * @addr: starting address | 85 | * @addr: starting address |
86 | * @end: ending address | 86 | * @end: ending address |
87 | * @walk: set of callbacks to invoke for each level of the tree | 87 | * @walk: set of callbacks to invoke for each level of the tree |
88 | * @private: private data passed to the callback function | ||
89 | * | 88 | * |
90 | * Recursively walk the page table for the memory area in a VMA, | 89 | * Recursively walk the page table for the memory area in a VMA, |
91 | * calling supplied callbacks. Callbacks are called in-order (first | 90 | * calling supplied callbacks. Callbacks are called in-order (first |
92 | * PGD, first PUD, first PMD, first PTE, second PTE... second PMD, | 91 | * PGD, first PUD, first PMD, first PTE, second PTE... second PMD, |
93 | * etc.). If lower-level callbacks are omitted, walking depth is reduced. | 92 | * etc.). If lower-level callbacks are omitted, walking depth is reduced. |
94 | * | 93 | * |
95 | * Each callback receives an entry pointer, the start and end of the | 94 | * Each callback receives an entry pointer and the start and end of the |
96 | * associated range, and a caller-supplied private data pointer. | 95 | * associated range, and a copy of the original mm_walk for access to |
96 | * the ->private or ->mm fields. | ||
97 | * | 97 | * |
98 | * No locks are taken, but the bottom level iterator will map PTE | 98 | * No locks are taken, but the bottom level iterator will map PTE |
99 | * directories from highmem if necessary. | 99 | * directories from highmem if necessary. |
@@ -101,9 +101,8 @@ static int walk_pud_range(pgd_t *pgd, unsigned long addr, unsigned long end, | |||
101 | * If any callback returns a non-zero value, the walk is aborted and | 101 | * If any callback returns a non-zero value, the walk is aborted and |
102 | * the return value is propagated back to the caller. Otherwise 0 is returned. | 102 | * the return value is propagated back to the caller. Otherwise 0 is returned. |
103 | */ | 103 | */ |
104 | int walk_page_range(const struct mm_struct *mm, | 104 | int walk_page_range(unsigned long addr, unsigned long end, |
105 | unsigned long addr, unsigned long end, | 105 | struct mm_walk *walk) |
106 | const struct mm_walk *walk, void *private) | ||
107 | { | 106 | { |
108 | pgd_t *pgd; | 107 | pgd_t *pgd; |
109 | unsigned long next; | 108 | unsigned long next; |
@@ -112,21 +111,24 @@ int walk_page_range(const struct mm_struct *mm, | |||
112 | if (addr >= end) | 111 | if (addr >= end) |
113 | return err; | 112 | return err; |
114 | 113 | ||
115 | pgd = pgd_offset(mm, addr); | 114 | if (!walk->mm) |
115 | return -EINVAL; | ||
116 | |||
117 | pgd = pgd_offset(walk->mm, addr); | ||
116 | do { | 118 | do { |
117 | next = pgd_addr_end(addr, end); | 119 | next = pgd_addr_end(addr, end); |
118 | if (pgd_none_or_clear_bad(pgd)) { | 120 | if (pgd_none_or_clear_bad(pgd)) { |
119 | if (walk->pte_hole) | 121 | if (walk->pte_hole) |
120 | err = walk->pte_hole(addr, next, private); | 122 | err = walk->pte_hole(addr, next, walk); |
121 | if (err) | 123 | if (err) |
122 | break; | 124 | break; |
123 | continue; | 125 | continue; |
124 | } | 126 | } |
125 | if (walk->pgd_entry) | 127 | if (walk->pgd_entry) |
126 | err = walk->pgd_entry(pgd, addr, next, private); | 128 | err = walk->pgd_entry(pgd, addr, next, walk); |
127 | if (!err && | 129 | if (!err && |
128 | (walk->pud_entry || walk->pmd_entry || walk->pte_entry)) | 130 | (walk->pud_entry || walk->pmd_entry || walk->pte_entry)) |
129 | err = walk_pud_range(pgd, addr, next, walk, private); | 131 | err = walk_pud_range(pgd, addr, next, walk); |
130 | if (err) | 132 | if (err) |
131 | break; | 133 | break; |
132 | } while (pgd++, addr = next, addr != end); | 134 | } while (pgd++, addr = next, addr != end); |
@@ -3263,9 +3263,12 @@ retry: | |||
3263 | 3263 | ||
3264 | if (cpuset_zone_allowed_hardwall(zone, flags) && | 3264 | if (cpuset_zone_allowed_hardwall(zone, flags) && |
3265 | cache->nodelists[nid] && | 3265 | cache->nodelists[nid] && |
3266 | cache->nodelists[nid]->free_objects) | 3266 | cache->nodelists[nid]->free_objects) { |
3267 | obj = ____cache_alloc_node(cache, | 3267 | obj = ____cache_alloc_node(cache, |
3268 | flags | GFP_THISNODE, nid); | 3268 | flags | GFP_THISNODE, nid); |
3269 | if (obj) | ||
3270 | break; | ||
3271 | } | ||
3269 | } | 3272 | } |
3270 | 3273 | ||
3271 | if (!obj) { | 3274 | if (!obj) { |
@@ -469,8 +469,9 @@ void *__kmalloc_node(size_t size, gfp_t gfp, int node) | |||
469 | return ZERO_SIZE_PTR; | 469 | return ZERO_SIZE_PTR; |
470 | 470 | ||
471 | m = slob_alloc(size + align, gfp, align, node); | 471 | m = slob_alloc(size + align, gfp, align, node); |
472 | if (m) | 472 | if (!m) |
473 | *m = size; | 473 | return NULL; |
474 | *m = size; | ||
474 | return (void *)m + align; | 475 | return (void *)m + align; |
475 | } else { | 476 | } else { |
476 | void *ret; | 477 | void *ret; |
@@ -2726,9 +2726,10 @@ size_t ksize(const void *object) | |||
2726 | 2726 | ||
2727 | page = virt_to_head_page(object); | 2727 | page = virt_to_head_page(object); |
2728 | 2728 | ||
2729 | if (unlikely(!PageSlab(page))) | 2729 | if (unlikely(!PageSlab(page))) { |
2730 | WARN_ON(!PageCompound(page)); | ||
2730 | return PAGE_SIZE << compound_order(page); | 2731 | return PAGE_SIZE << compound_order(page); |
2731 | 2732 | } | |
2732 | s = page->slab; | 2733 | s = page->slab; |
2733 | 2734 | ||
2734 | #ifdef CONFIG_SLUB_DEBUG | 2735 | #ifdef CONFIG_SLUB_DEBUG |
@@ -503,7 +503,7 @@ void vm_acct_memory(long pages) | |||
503 | local = &__get_cpu_var(committed_space); | 503 | local = &__get_cpu_var(committed_space); |
504 | *local += pages; | 504 | *local += pages; |
505 | if (*local > ACCT_THRESHOLD || *local < -ACCT_THRESHOLD) { | 505 | if (*local > ACCT_THRESHOLD || *local < -ACCT_THRESHOLD) { |
506 | atomic_add(*local, &vm_committed_space); | 506 | atomic_long_add(*local, &vm_committed_space); |
507 | *local = 0; | 507 | *local = 0; |
508 | } | 508 | } |
509 | preempt_enable(); | 509 | preempt_enable(); |
@@ -520,7 +520,7 @@ static int cpu_swap_callback(struct notifier_block *nfb, | |||
520 | 520 | ||
521 | committed = &per_cpu(committed_space, (long)hcpu); | 521 | committed = &per_cpu(committed_space, (long)hcpu); |
522 | if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) { | 522 | if (action == CPU_DEAD || action == CPU_DEAD_FROZEN) { |
523 | atomic_add(*committed, &vm_committed_space); | 523 | atomic_long_add(*committed, &vm_committed_space); |
524 | *committed = 0; | 524 | *committed = 0; |
525 | drain_cpu_pagevecs((long)hcpu); | 525 | drain_cpu_pagevecs((long)hcpu); |
526 | } | 526 | } |
diff --git a/mm/vmscan.c b/mm/vmscan.c index 9a29901ad3b3..967d30ccd92b 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c | |||
@@ -1307,7 +1307,7 @@ static unsigned long do_try_to_free_pages(struct zonelist *zonelist, | |||
1307 | struct scan_control *sc) | 1307 | struct scan_control *sc) |
1308 | { | 1308 | { |
1309 | int priority; | 1309 | int priority; |
1310 | int ret = 0; | 1310 | unsigned long ret = 0; |
1311 | unsigned long total_scanned = 0; | 1311 | unsigned long total_scanned = 0; |
1312 | unsigned long nr_reclaimed = 0; | 1312 | unsigned long nr_reclaimed = 0; |
1313 | struct reclaim_state *reclaim_state = current->reclaim_state; | 1313 | struct reclaim_state *reclaim_state = current->reclaim_state; |