1 ============================= 1 =============================
2 Examining Process Page Tables 2 Examining Process Page Tables
3 ============================= 3 =============================
4 4
5 pagemap is a new (as of 2.6.25) set of interfa 5 pagemap is a new (as of 2.6.25) set of interfaces in the kernel that allow
6 userspace programs to examine the page tables 6 userspace programs to examine the page tables and related information by
7 reading files in ``/proc``. 7 reading files in ``/proc``.
8 8
9 There are four components to pagemap: 9 There are four components to pagemap:
10 10
11 * ``/proc/pid/pagemap``. This file lets a us 11 * ``/proc/pid/pagemap``. This file lets a userspace process find out which
12 physical frame each virtual page is mapped 12 physical frame each virtual page is mapped to. It contains one 64-bit
13 value for each virtual page, containing the 13 value for each virtual page, containing the following data (from
14 ``fs/proc/task_mmu.c``, above pagemap_read) 14 ``fs/proc/task_mmu.c``, above pagemap_read):
15 15
16 * Bits 0-54 page frame number (PFN) if pr 16 * Bits 0-54 page frame number (PFN) if present
17 * Bits 0-4 swap type if swapped 17 * Bits 0-4 swap type if swapped
18 * Bits 5-54 swap offset if swapped 18 * Bits 5-54 swap offset if swapped
19 * Bit 55 pte is soft-dirty (see 19 * Bit 55 pte is soft-dirty (see
20 Documentation/admin-guide/mm/soft-dirty. 20 Documentation/admin-guide/mm/soft-dirty.rst)
21 * Bit 56 page exclusively mapped (sinc 21 * Bit 56 page exclusively mapped (since 4.2)
22 * Bit 57 pte is uffd-wp write-protecte 22 * Bit 57 pte is uffd-wp write-protected (since 5.13) (see
23 Documentation/admin-guide/mm/userfaultfd 23 Documentation/admin-guide/mm/userfaultfd.rst)
24 * Bits 58-60 zero 24 * Bits 58-60 zero
25 * Bit 61 page is file-page or shared-a 25 * Bit 61 page is file-page or shared-anon (since 3.5)
26 * Bit 62 page swapped 26 * Bit 62 page swapped
27 * Bit 63 page present 27 * Bit 63 page present
28 28
29 Since Linux 4.0 only users with the CAP_SYS 29 Since Linux 4.0 only users with the CAP_SYS_ADMIN capability can get PFNs.
30 In 4.0 and 4.1 opens by unprivileged fail w 30 In 4.0 and 4.1 opens by unprivileged fail with -EPERM. Starting from
31 4.2 the PFN field is zeroed if the user doe 31 4.2 the PFN field is zeroed if the user does not have CAP_SYS_ADMIN.
32 Reason: information about PFNs helps in exp 32 Reason: information about PFNs helps in exploiting Rowhammer vulnerability.
33 33
34 If the page is not present but in swap, the 34 If the page is not present but in swap, then the PFN contains an
35 encoding of the swap file number and the pa 35 encoding of the swap file number and the page's offset into the
36 swap. Unmapped pages return a null PFN. Thi 36 swap. Unmapped pages return a null PFN. This allows determining
37 precisely which pages are mapped (or in swa 37 precisely which pages are mapped (or in swap) and comparing mapped
38 pages between processes. 38 pages between processes.
39 39
40 Efficient users of this interface will use 40 Efficient users of this interface will use ``/proc/pid/maps`` to
41 determine which areas of memory are actuall 41 determine which areas of memory are actually mapped and llseek to
42 skip over unmapped regions. 42 skip over unmapped regions.
43 43
44 * ``/proc/kpagecount``. This file contains a 44 * ``/proc/kpagecount``. This file contains a 64-bit count of the number of
45 times each page is mapped, indexed by PFN. 45 times each page is mapped, indexed by PFN.
46 46
47 The page-types tool in the tools/mm directory 47 The page-types tool in the tools/mm directory can be used to query the
48 number of times a page is mapped. 48 number of times a page is mapped.
49 49
50 * ``/proc/kpageflags``. This file contains a 50 * ``/proc/kpageflags``. This file contains a 64-bit set of flags for each
51 page, indexed by PFN. 51 page, indexed by PFN.
52 52
53 The flags are (from ``fs/proc/page.c``, abo 53 The flags are (from ``fs/proc/page.c``, above kpageflags_read):
54 54
55 0. LOCKED 55 0. LOCKED
56 1. ERROR 56 1. ERROR
57 2. REFERENCED 57 2. REFERENCED
58 3. UPTODATE 58 3. UPTODATE
59 4. DIRTY 59 4. DIRTY
60 5. LRU 60 5. LRU
61 6. ACTIVE 61 6. ACTIVE
62 7. SLAB 62 7. SLAB
63 8. WRITEBACK 63 8. WRITEBACK
64 9. RECLAIM 64 9. RECLAIM
65 10. BUDDY 65 10. BUDDY
66 11. MMAP 66 11. MMAP
67 12. ANON 67 12. ANON
68 13. SWAPCACHE 68 13. SWAPCACHE
69 14. SWAPBACKED 69 14. SWAPBACKED
70 15. COMPOUND_HEAD 70 15. COMPOUND_HEAD
71 16. COMPOUND_TAIL 71 16. COMPOUND_TAIL
72 17. HUGE 72 17. HUGE
73 18. UNEVICTABLE 73 18. UNEVICTABLE
74 19. HWPOISON 74 19. HWPOISON
75 20. NOPAGE 75 20. NOPAGE
76 21. KSM 76 21. KSM
77 22. THP 77 22. THP
78 23. OFFLINE 78 23. OFFLINE
79 24. ZERO_PAGE 79 24. ZERO_PAGE
80 25. IDLE 80 25. IDLE
81 26. PGTABLE 81 26. PGTABLE
82 82
83 * ``/proc/kpagecgroup``. This file contains 83 * ``/proc/kpagecgroup``. This file contains a 64-bit inode number of the
84 memory cgroup each page is charged to, inde 84 memory cgroup each page is charged to, indexed by PFN. Only available when
85 CONFIG_MEMCG is set. 85 CONFIG_MEMCG is set.
86 86
87 Short descriptions to the page flags 87 Short descriptions to the page flags
88 ==================================== 88 ====================================
89 89
90 0 - LOCKED 90 0 - LOCKED
91 The page is being locked for exclusive acce 91 The page is being locked for exclusive access, e.g. by undergoing read/write
92 IO. 92 IO.
93 7 - SLAB 93 7 - SLAB
94 The page is managed by the SLAB/SLUB kernel 94 The page is managed by the SLAB/SLUB kernel memory allocator.
95 When compound page is used, either will onl 95 When compound page is used, either will only set this flag on the head
96 page. 96 page.
97 10 - BUDDY 97 10 - BUDDY
98 A free memory block managed by the buddy s 98 A free memory block managed by the buddy system allocator.
99 The buddy system organizes free memory in 99 The buddy system organizes free memory in blocks of various orders.
100 An order N block has 2^N physically contig 100 An order N block has 2^N physically contiguous pages, with the BUDDY flag
101 set for and _only_ for the first page. 101 set for and _only_ for the first page.
102 15 - COMPOUND_HEAD 102 15 - COMPOUND_HEAD
103 A compound page with order N consists of 2 103 A compound page with order N consists of 2^N physically contiguous pages.
104 A compound page with order 2 takes the for 104 A compound page with order 2 takes the form of "HTTT", where H donates its
105 head page and T donates its tail page(s). 105 head page and T donates its tail page(s). The major consumers of compound
106 pages are hugeTLB pages (Documentation/adm 106 pages are hugeTLB pages (Documentation/admin-guide/mm/hugetlbpage.rst),
107 the SLUB etc. memory allocators and vario 107 the SLUB etc. memory allocators and various device drivers.
108 However in this interface, only huge/giga 108 However in this interface, only huge/giga pages are made visible
109 to end users. 109 to end users.
110 16 - COMPOUND_TAIL 110 16 - COMPOUND_TAIL
111 A compound page tail (see description abov 111 A compound page tail (see description above).
112 17 - HUGE 112 17 - HUGE
113 This is an integral part of a HugeTLB page 113 This is an integral part of a HugeTLB page.
114 19 - HWPOISON 114 19 - HWPOISON
115 Hardware detected memory corruption on thi 115 Hardware detected memory corruption on this page: don't touch the data!
116 20 - NOPAGE 116 20 - NOPAGE
117 No page frame exists at the requested addr 117 No page frame exists at the requested address.
118 21 - KSM 118 21 - KSM
119 Identical memory pages dynamically shared 119 Identical memory pages dynamically shared between one or more processes.
120 22 - THP 120 22 - THP
121 Contiguous pages which construct THP of an !! 121 Contiguous pages which construct transparent hugepages.
122 23 - OFFLINE 122 23 - OFFLINE
123 The page is logically offline. 123 The page is logically offline.
124 24 - ZERO_PAGE 124 24 - ZERO_PAGE
125 Zero page for pfn_zero or huge_zero page. 125 Zero page for pfn_zero or huge_zero page.
126 25 - IDLE 126 25 - IDLE
127 The page has not been accessed since it wa 127 The page has not been accessed since it was marked idle (see
128 Documentation/admin-guide/mm/idle_page_tra 128 Documentation/admin-guide/mm/idle_page_tracking.rst).
129 Note that this flag may be stale in case t 129 Note that this flag may be stale in case the page was accessed via
130 a PTE. To make sure the flag is up-to-date 130 a PTE. To make sure the flag is up-to-date one has to read
131 ``/sys/kernel/mm/page_idle/bitmap`` first. 131 ``/sys/kernel/mm/page_idle/bitmap`` first.
132 26 - PGTABLE 132 26 - PGTABLE
133 The page is in use as a page table. 133 The page is in use as a page table.
134 134
135 IO related page flags 135 IO related page flags
136 --------------------- 136 ---------------------
137 137
138 1 - ERROR 138 1 - ERROR
139 IO error occurred. 139 IO error occurred.
140 3 - UPTODATE 140 3 - UPTODATE
141 The page has up-to-date data. 141 The page has up-to-date data.
142 ie. for file backed page: (in-memory data r 142 ie. for file backed page: (in-memory data revision >= on-disk one)
143 4 - DIRTY 143 4 - DIRTY
144 The page has been written to, hence contain 144 The page has been written to, hence contains new data.
145 i.e. for file backed page: (in-memory data 145 i.e. for file backed page: (in-memory data revision > on-disk one)
146 8 - WRITEBACK 146 8 - WRITEBACK
147 The page is being synced to disk. 147 The page is being synced to disk.
148 148
149 LRU related page flags 149 LRU related page flags
150 ---------------------- 150 ----------------------
151 151
152 5 - LRU 152 5 - LRU
153 The page is in one of the LRU lists. 153 The page is in one of the LRU lists.
154 6 - ACTIVE 154 6 - ACTIVE
155 The page is in the active LRU list. 155 The page is in the active LRU list.
156 18 - UNEVICTABLE 156 18 - UNEVICTABLE
157 The page is in the unevictable (non-)LRU li 157 The page is in the unevictable (non-)LRU list It is somehow pinned and
158 not a candidate for LRU page reclaims, e.g. 158 not a candidate for LRU page reclaims, e.g. ramfs pages,
159 shmctl(SHM_LOCK) and mlock() memory segment 159 shmctl(SHM_LOCK) and mlock() memory segments.
160 2 - REFERENCED 160 2 - REFERENCED
161 The page has been referenced since last LRU 161 The page has been referenced since last LRU list enqueue/requeue.
162 9 - RECLAIM 162 9 - RECLAIM
163 The page will be reclaimed soon after its p 163 The page will be reclaimed soon after its pageout IO completed.
164 11 - MMAP 164 11 - MMAP
165 A memory mapped page. 165 A memory mapped page.
166 12 - ANON 166 12 - ANON
167 A memory mapped page that is not part of a 167 A memory mapped page that is not part of a file.
168 13 - SWAPCACHE 168 13 - SWAPCACHE
169 The page is mapped to swap space, i.e. has 169 The page is mapped to swap space, i.e. has an associated swap entry.
170 14 - SWAPBACKED 170 14 - SWAPBACKED
171 The page is backed by swap/RAM. 171 The page is backed by swap/RAM.
172 172
173 The page-types tool in the tools/mm directory 173 The page-types tool in the tools/mm directory can be used to query the
174 above flags. 174 above flags.
175 175
>> 176 Using pagemap to do something useful
>> 177 ====================================
>> 178
>> 179 The general procedure for using pagemap to find out about a process' memory
>> 180 usage goes like this:
>> 181
>> 182 1. Read ``/proc/pid/maps`` to determine which parts of the memory space are
>> 183 mapped to what.
>> 184 2. Select the maps you are interested in -- all of them, or a particular
>> 185 library, or the stack or the heap, etc.
>> 186 3. Open ``/proc/pid/pagemap`` and seek to the pages you would like to examine.
>> 187 4. Read a u64 for each page from pagemap.
>> 188 5. Open ``/proc/kpagecount`` and/or ``/proc/kpageflags``. For each PFN you
>> 189 just read, seek to that entry in the file, and read the data you want.
>> 190
>> 191 For example, to find the "unique set size" (USS), which is the amount of
>> 192 memory that a process is using that is not shared with any other process,
>> 193 you can go through every map in the process, find the PFNs, look those up
>> 194 in kpagecount, and tally up the number of pages that are only referenced
>> 195 once.
>> 196
176 Exceptions for Shared Memory 197 Exceptions for Shared Memory
177 ============================ 198 ============================
178 199
179 Page table entries for shared pages are cleare 200 Page table entries for shared pages are cleared when the pages are zapped or
180 swapped out. This makes swapped out pages indi 201 swapped out. This makes swapped out pages indistinguishable from never-allocated
181 ones. 202 ones.
182 203
183 In kernel space, the swap location can still b 204 In kernel space, the swap location can still be retrieved from the page cache.
184 However, values stored only on the normal PTE 205 However, values stored only on the normal PTE get lost irretrievably when the
185 page is swapped out (i.e. SOFT_DIRTY). 206 page is swapped out (i.e. SOFT_DIRTY).
186 207
187 In user space, whether the page is present, sw 208 In user space, whether the page is present, swapped or none can be deduced with
188 the help of lseek and/or mincore system calls. 209 the help of lseek and/or mincore system calls.
189 210
190 lseek() can differentiate between accessed pag 211 lseek() can differentiate between accessed pages (present or swapped out) and
191 holes (none/non-allocated) by specifying the S 212 holes (none/non-allocated) by specifying the SEEK_DATA flag on the file where
192 the pages are backed. For anonymous shared pag 213 the pages are backed. For anonymous shared pages, the file can be found in
193 ``/proc/pid/map_files/``. 214 ``/proc/pid/map_files/``.
194 215
195 mincore() can differentiate between pages in m 216 mincore() can differentiate between pages in memory (present, including swap
196 cache) and out of memory (swapped out or none/ 217 cache) and out of memory (swapped out or none/non-allocated).
197 218
198 Other notes 219 Other notes
199 =========== 220 ===========
200 221
201 Reading from any of the files will return -EIN 222 Reading from any of the files will return -EINVAL if you are not starting
202 the read on an 8-byte boundary (e.g., if you s 223 the read on an 8-byte boundary (e.g., if you sought an odd number of bytes
203 into the file), or if the size of the read is 224 into the file), or if the size of the read is not a multiple of 8 bytes.
204 225
205 Before Linux 3.11 pagemap bits 55-60 were used 226 Before Linux 3.11 pagemap bits 55-60 were used for "page-shift" (which is
206 always 12 at most architectures). Since Linux 227 always 12 at most architectures). Since Linux 3.11 their meaning changes
207 after first clear of soft-dirty bits. Since Li 228 after first clear of soft-dirty bits. Since Linux 4.2 they are used for
208 flags unconditionally. 229 flags unconditionally.
209 230
210 Pagemap Scan IOCTL 231 Pagemap Scan IOCTL
211 ================== 232 ==================
212 233
213 The ``PAGEMAP_SCAN`` IOCTL on the pagemap file 234 The ``PAGEMAP_SCAN`` IOCTL on the pagemap file can be used to get or optionally
214 clear the info about page table entries. The f 235 clear the info about page table entries. The following operations are supported
215 in this IOCTL: 236 in this IOCTL:
216 237
217 - Scan the address range and get the memory ra 238 - Scan the address range and get the memory ranges matching the provided criteria.
218 This is performed when the output buffer is 239 This is performed when the output buffer is specified.
219 - Write-protect the pages. The ``PM_SCAN_WP_MA 240 - Write-protect the pages. The ``PM_SCAN_WP_MATCHING`` is used to write-protect
220 the pages of interest. The ``PM_SCAN_CHECK_W 241 the pages of interest. The ``PM_SCAN_CHECK_WPASYNC`` aborts the operation if
221 non-Async Write Protected pages are found. T 242 non-Async Write Protected pages are found. The ``PM_SCAN_WP_MATCHING`` can be
222 used with or without ``PM_SCAN_CHECK_WPASYNC 243 used with or without ``PM_SCAN_CHECK_WPASYNC``.
223 - Both of those operations can be combined int 244 - Both of those operations can be combined into one atomic operation where we can
224 get and write protect the pages as well. 245 get and write protect the pages as well.
225 246
226 Following flags about pages are currently supp 247 Following flags about pages are currently supported:
227 248
228 - ``PAGE_IS_WPALLOWED`` - Page has async-write 249 - ``PAGE_IS_WPALLOWED`` - Page has async-write-protection enabled
229 - ``PAGE_IS_WRITTEN`` - Page has been written 250 - ``PAGE_IS_WRITTEN`` - Page has been written to from the time it was write protected
230 - ``PAGE_IS_FILE`` - Page is file backed 251 - ``PAGE_IS_FILE`` - Page is file backed
231 - ``PAGE_IS_PRESENT`` - Page is present in the 252 - ``PAGE_IS_PRESENT`` - Page is present in the memory
232 - ``PAGE_IS_SWAPPED`` - Page is in swapped 253 - ``PAGE_IS_SWAPPED`` - Page is in swapped
233 - ``PAGE_IS_PFNZERO`` - Page has zero PFN 254 - ``PAGE_IS_PFNZERO`` - Page has zero PFN
234 - ``PAGE_IS_HUGE`` - Page is PMD-mapped THP or !! 255 - ``PAGE_IS_HUGE`` - Page is THP or Hugetlb backed
235 - ``PAGE_IS_SOFT_DIRTY`` - Page is soft-dirty 256 - ``PAGE_IS_SOFT_DIRTY`` - Page is soft-dirty
236 257
237 The ``struct pm_scan_arg`` is used as the argu 258 The ``struct pm_scan_arg`` is used as the argument of the IOCTL.
238 259
239 1. The size of the ``struct pm_scan_arg`` mus 260 1. The size of the ``struct pm_scan_arg`` must be specified in the ``size``
240 field. This field will be helpful in recog 261 field. This field will be helpful in recognizing the structure if extensions
241 are done later. 262 are done later.
242 2. The flags can be specified in the ``flags` 263 2. The flags can be specified in the ``flags`` field. The ``PM_SCAN_WP_MATCHING``
243 and ``PM_SCAN_CHECK_WPASYNC`` are the only 264 and ``PM_SCAN_CHECK_WPASYNC`` are the only added flags at this time. The get
244 operation is optionally performed dependin 265 operation is optionally performed depending upon if the output buffer is
245 provided or not. 266 provided or not.
246 3. The range is specified through ``start`` a 267 3. The range is specified through ``start`` and ``end``.
247 4. The walk can abort before visiting the com 268 4. The walk can abort before visiting the complete range such as the user buffer
248 can get full etc. The walk ending address 269 can get full etc. The walk ending address is specified in``end_walk``.
249 5. The output buffer of ``struct page_region` 270 5. The output buffer of ``struct page_region`` array and size is specified in
250 ``vec`` and ``vec_len``. 271 ``vec`` and ``vec_len``.
251 6. The optional maximum requested pages are s 272 6. The optional maximum requested pages are specified in the ``max_pages``.
252 7. The masks are specified in ``category_mask 273 7. The masks are specified in ``category_mask``, ``category_anyof_mask``,
253 ``category_inverted`` and ``return_mask``. 274 ``category_inverted`` and ``return_mask``.
254 275
255 Find pages which have been written and WP them 276 Find pages which have been written and WP them as well::
256 277
257 struct pm_scan_arg arg = { 278 struct pm_scan_arg arg = {
258 .size = sizeof(arg), 279 .size = sizeof(arg),
259 .flags = PM_SCAN_CHECK_WPASYNC | PM_SCAN_CH 280 .flags = PM_SCAN_CHECK_WPASYNC | PM_SCAN_CHECK_WPASYNC,
260 .. 281 ..
261 .category_mask = PAGE_IS_WRITTEN, 282 .category_mask = PAGE_IS_WRITTEN,
262 .return_mask = PAGE_IS_WRITTEN, 283 .return_mask = PAGE_IS_WRITTEN,
263 }; 284 };
264 285
265 Find pages which have been written, are file b 286 Find pages which have been written, are file backed, not swapped and either
266 present or huge:: 287 present or huge::
267 288
268 struct pm_scan_arg arg = { 289 struct pm_scan_arg arg = {
269 .size = sizeof(arg), 290 .size = sizeof(arg),
270 .flags = 0, 291 .flags = 0,
271 .. 292 ..
272 .category_mask = PAGE_IS_WRITTEN | PAGE_IS_ 293 .category_mask = PAGE_IS_WRITTEN | PAGE_IS_SWAPPED,
273 .category_inverted = PAGE_IS_SWAPPED, 294 .category_inverted = PAGE_IS_SWAPPED,
274 .category_anyof_mask = PAGE_IS_PRESENT | PA 295 .category_anyof_mask = PAGE_IS_PRESENT | PAGE_IS_HUGE,
275 .return_mask = PAGE_IS_WRITTEN | PAGE_IS_SW 296 .return_mask = PAGE_IS_WRITTEN | PAGE_IS_SWAPPED |
276 PAGE_IS_PRESENT | PAGE_IS_HU 297 PAGE_IS_PRESENT | PAGE_IS_HUGE,
277 }; 298 };
278 299
279 The ``PAGE_IS_WRITTEN`` flag can be considered 300 The ``PAGE_IS_WRITTEN`` flag can be considered as a better-performing alternative
280 of soft-dirty flag. It doesn't get affected by 301 of soft-dirty flag. It doesn't get affected by VMA merging of the kernel and hence
281 the user can find the true soft-dirty pages in 302 the user can find the true soft-dirty pages in case of normal pages. (There may
282 still be extra dirty pages reported for THP or 303 still be extra dirty pages reported for THP or Hugetlb pages.)
283 304
284 "PAGE_IS_WRITTEN" category is used with uffd w 305 "PAGE_IS_WRITTEN" category is used with uffd write protect-enabled ranges to
285 implement memory dirty tracking in userspace: 306 implement memory dirty tracking in userspace:
286 307
287 1. The userfaultfd file descriptor is created 308 1. The userfaultfd file descriptor is created with ``userfaultfd`` syscall.
288 2. The ``UFFD_FEATURE_WP_UNPOPULATED`` and `` 309 2. The ``UFFD_FEATURE_WP_UNPOPULATED`` and ``UFFD_FEATURE_WP_ASYNC`` features
289 are set by ``UFFDIO_API`` IOCTL. 310 are set by ``UFFDIO_API`` IOCTL.
290 3. The memory range is registered with ``UFFD 311 3. The memory range is registered with ``UFFDIO_REGISTER_MODE_WP`` mode
291 through ``UFFDIO_REGISTER`` IOCTL. 312 through ``UFFDIO_REGISTER`` IOCTL.
292 4. Then any part of the registered memory or 313 4. Then any part of the registered memory or the whole memory region must
293 be write protected using ``PAGEMAP_SCAN`` 314 be write protected using ``PAGEMAP_SCAN`` IOCTL with flag ``PM_SCAN_WP_MATCHING``
294 or the ``UFFDIO_WRITEPROTECT`` IOCTL can b 315 or the ``UFFDIO_WRITEPROTECT`` IOCTL can be used. Both of these perform the
295 same operation. The former is better in te 316 same operation. The former is better in terms of performance.
296 5. Now the ``PAGEMAP_SCAN`` IOCTL can be used 317 5. Now the ``PAGEMAP_SCAN`` IOCTL can be used to either just find pages which
297 have been written to since they were last 318 have been written to since they were last marked and/or optionally write protect
298 the pages as well. 319 the pages as well.
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.