1.. _hugetlbpage:
2
3=============
4HugeTLB Pages
5=============
6
7Overview
8========
9
10The intent of this file is to give a brief summary of hugetlbpage support in
11the Linux kernel.  This support is built on top of multiple page size support
12that is provided by most modern architectures.  For example, x86 CPUs normally
13support 4K and 2M (1G if architecturally supported) page sizes, ia64
14architecture supports multiple page sizes 4K, 8K, 64K, 256K, 1M, 4M, 16M,
15256M and ppc64 supports 4K and 16M.  A TLB is a cache of virtual-to-physical
16translations.  Typically this is a very scarce resource on processor.
17Operating systems try to make best use of limited number of TLB resources.
18This optimization is more critical now as bigger and bigger physical memories
19(several GBs) are more readily available.
20
21Users can use the huge page support in Linux kernel by either using the mmap
22system call or standard SYSV shared memory system calls (shmget, shmat).
23
24First the Linux kernel needs to be built with the CONFIG_HUGETLBFS
25(present under "File systems") and CONFIG_HUGETLB_PAGE (selected
26automatically when CONFIG_HUGETLBFS is selected) configuration
27options.
28
29The ``/proc/meminfo`` file provides information about the total number of
30persistent hugetlb pages in the kernel's huge page pool.  It also displays
31default huge page size and information about the number of free, reserved
32and surplus huge pages in the pool of huge pages of default size.
33The huge page size is needed for generating the proper alignment and
34size of the arguments to system calls that map huge page regions.
35
36The output of ``cat /proc/meminfo`` will include lines like::
37
38	HugePages_Total: uuu
39	HugePages_Free:  vvv
40	HugePages_Rsvd:  www
41	HugePages_Surp:  xxx
42	Hugepagesize:    yyy kB
43	Hugetlb:         zzz kB
44
45where:
46
47HugePages_Total
48	is the size of the pool of huge pages.
49HugePages_Free
50	is the number of huge pages in the pool that are not yet
51        allocated.
52HugePages_Rsvd
53	is short for "reserved," and is the number of huge pages for
54        which a commitment to allocate from the pool has been made,
55        but no allocation has yet been made.  Reserved huge pages
56        guarantee that an application will be able to allocate a
57        huge page from the pool of huge pages at fault time.
58HugePages_Surp
59	is short for "surplus," and is the number of huge pages in
60        the pool above the value in ``/proc/sys/vm/nr_hugepages``. The
61        maximum number of surplus huge pages is controlled by
62        ``/proc/sys/vm/nr_overcommit_hugepages``.
63	Note: When the feature of freeing unused vmemmap pages associated
64	with each hugetlb page is enabled, the number of surplus huge pages
65	may be temporarily larger than the maximum number of surplus huge
66	pages when the system is under memory pressure.
67Hugepagesize
68	is the default hugepage size (in Kb).
69Hugetlb
70        is the total amount of memory (in kB), consumed by huge
71        pages of all sizes.
72        If huge pages of different sizes are in use, this number
73        will exceed HugePages_Total \* Hugepagesize. To get more
74        detailed information, please, refer to
75        ``/sys/kernel/mm/hugepages`` (described below).
76
77
78``/proc/filesystems`` should also show a filesystem of type "hugetlbfs"
79configured in the kernel.
80
81``/proc/sys/vm/nr_hugepages`` indicates the current number of "persistent" huge
82pages in the kernel's huge page pool.  "Persistent" huge pages will be
83returned to the huge page pool when freed by a task.  A user with root
84privileges can dynamically allocate more or free some persistent huge pages
85by increasing or decreasing the value of ``nr_hugepages``.
86
87Note: When the feature of freeing unused vmemmap pages associated with each
88hugetlb page is enabled, we can fail to free the huge pages triggered by
89the user when ths system is under memory pressure.  Please try again later.
90
91Pages that are used as huge pages are reserved inside the kernel and cannot
92be used for other purposes.  Huge pages cannot be swapped out under
93memory pressure.
94
95Once a number of huge pages have been pre-allocated to the kernel huge page
96pool, a user with appropriate privilege can use either the mmap system call
97or shared memory system calls to use the huge pages.  See the discussion of
98:ref:`Using Huge Pages <using_huge_pages>`, below.
99
100The administrator can allocate persistent huge pages on the kernel boot
101command line by specifying the "hugepages=N" parameter, where 'N' = the
102number of huge pages requested.  This is the most reliable method of
103allocating huge pages as memory has not yet become fragmented.
104
105Some platforms support multiple huge page sizes.  To allocate huge pages
106of a specific size, one must precede the huge pages boot command parameters
107with a huge page size selection parameter "hugepagesz=<size>".  <size> must
108be specified in bytes with optional scale suffix [kKmMgG].  The default huge
109page size may be selected with the "default_hugepagesz=<size>" boot parameter.
110
111Hugetlb boot command line parameter semantics
112
113hugepagesz
114	Specify a huge page size.  Used in conjunction with hugepages
115	parameter to preallocate a number of huge pages of the specified
116	size.  Hence, hugepagesz and hugepages are typically specified in
117	pairs such as::
118
119		hugepagesz=2M hugepages=512
120
121	hugepagesz can only be specified once on the command line for a
122	specific huge page size.  Valid huge page sizes are architecture
123	dependent.
124hugepages
125	Specify the number of huge pages to preallocate.  This typically
126	follows a valid hugepagesz or default_hugepagesz parameter.  However,
127	if hugepages is the first or only hugetlb command line parameter it
128	implicitly specifies the number of huge pages of default size to
129	allocate.  If the number of huge pages of default size is implicitly
130	specified, it can not be overwritten by a hugepagesz,hugepages
131	parameter pair for the default size.  This parameter also has a
132	node format.  The node format specifies the number of huge pages
133	to allocate on specific nodes.
134
135	For example, on an architecture with 2M default huge page size::
136
137		hugepages=256 hugepagesz=2M hugepages=512
138
139	will result in 256 2M huge pages being allocated and a warning message
140	indicating that the hugepages=512 parameter is ignored.  If a hugepages
141	parameter is preceded by an invalid hugepagesz parameter, it will
142	be ignored.
143
144	Node format example::
145
146		hugepagesz=2M hugepages=0:1,1:2
147
148	It will allocate 1 2M hugepage on node0 and 2 2M hugepages on node1.
149	If the node number is invalid,  the parameter will be ignored.
150
151default_hugepagesz
152	Specify the default huge page size.  This parameter can
153	only be specified once on the command line.  default_hugepagesz can
154	optionally be followed by the hugepages parameter to preallocate a
155	specific number of huge pages of default size.  The number of default
156	sized huge pages to preallocate can also be implicitly specified as
157	mentioned in the hugepages section above.  Therefore, on an
158	architecture with 2M default huge page size::
159
160		hugepages=256
161		default_hugepagesz=2M hugepages=256
162		hugepages=256 default_hugepagesz=2M
163
164	will all result in 256 2M huge pages being allocated.  Valid default
165	huge page size is architecture dependent.
166hugetlb_free_vmemmap
167	When CONFIG_HUGETLB_PAGE_OPTIMIZE_VMEMMAP is set, this enables HugeTLB
168	Vmemmap Optimization (HVO).
169
170When multiple huge page sizes are supported, ``/proc/sys/vm/nr_hugepages``
171indicates the current number of pre-allocated huge pages of the default size.
172Thus, one can use the following command to dynamically allocate/deallocate
173default sized persistent huge pages::
174
175	echo 20 > /proc/sys/vm/nr_hugepages
176
177This command will try to adjust the number of default sized huge pages in the
178huge page pool to 20, allocating or freeing huge pages, as required.
179
180On a NUMA platform, the kernel will attempt to distribute the huge page pool
181over all the set of allowed nodes specified by the NUMA memory policy of the
182task that modifies ``nr_hugepages``. The default for the allowed nodes--when the
183task has default memory policy--is all on-line nodes with memory.  Allowed
184nodes with insufficient available, contiguous memory for a huge page will be
185silently skipped when allocating persistent huge pages.  See the
186:ref:`discussion below <mem_policy_and_hp_alloc>`
187of the interaction of task memory policy, cpusets and per node attributes
188with the allocation and freeing of persistent huge pages.
189
190The success or failure of huge page allocation depends on the amount of
191physically contiguous memory that is present in system at the time of the
192allocation attempt.  If the kernel is unable to allocate huge pages from
193some nodes in a NUMA system, it will attempt to make up the difference by
194allocating extra pages on other nodes with sufficient available contiguous
195memory, if any.
196
197System administrators may want to put this command in one of the local rc
198init files.  This will enable the kernel to allocate huge pages early in
199the boot process when the possibility of getting physical contiguous pages
200is still very high.  Administrators can verify the number of huge pages
201actually allocated by checking the sysctl or meminfo.  To check the per node
202distribution of huge pages in a NUMA system, use::
203
204	cat /sys/devices/system/node/node*/meminfo | fgrep Huge
205
206``/proc/sys/vm/nr_overcommit_hugepages`` specifies how large the pool of
207huge pages can grow, if more huge pages than ``/proc/sys/vm/nr_hugepages`` are
208requested by applications.  Writing any non-zero value into this file
209indicates that the hugetlb subsystem is allowed to try to obtain that
210number of "surplus" huge pages from the kernel's normal page pool, when the
211persistent huge page pool is exhausted. As these surplus huge pages become
212unused, they are freed back to the kernel's normal page pool.
213
214When increasing the huge page pool size via ``nr_hugepages``, any existing
215surplus pages will first be promoted to persistent huge pages.  Then, additional
216huge pages will be allocated, if necessary and if possible, to fulfill
217the new persistent huge page pool size.
218
219The administrator may shrink the pool of persistent huge pages for
220the default huge page size by setting the ``nr_hugepages`` sysctl to a
221smaller value.  The kernel will attempt to balance the freeing of huge pages
222across all nodes in the memory policy of the task modifying ``nr_hugepages``.
223Any free huge pages on the selected nodes will be freed back to the kernel's
224normal page pool.
225
226Caveat: Shrinking the persistent huge page pool via ``nr_hugepages`` such that
227it becomes less than the number of huge pages in use will convert the balance
228of the in-use huge pages to surplus huge pages.  This will occur even if
229the number of surplus pages would exceed the overcommit value.  As long as
230this condition holds--that is, until ``nr_hugepages+nr_overcommit_hugepages`` is
231increased sufficiently, or the surplus huge pages go out of use and are freed--
232no more surplus huge pages will be allowed to be allocated.
233
234With support for multiple huge page pools at run-time available, much of
235the huge page userspace interface in ``/proc/sys/vm`` has been duplicated in
236sysfs.
237The ``/proc`` interfaces discussed above have been retained for backwards
238compatibility. The root huge page control directory in sysfs is::
239
240	/sys/kernel/mm/hugepages
241
242For each huge page size supported by the running kernel, a subdirectory
243will exist, of the form::
244
245	hugepages-${size}kB
246
247Inside each of these directories, the set of files contained in ``/proc``
248will exist.  In addition, two additional interfaces for demoting huge
249pages may exist::
250
251        demote
252        demote_size
253	nr_hugepages
254	nr_hugepages_mempolicy
255	nr_overcommit_hugepages
256	free_hugepages
257	resv_hugepages
258	surplus_hugepages
259
260The demote interfaces provide the ability to split a huge page into
261smaller huge pages.  For example, the x86 architecture supports both
2621GB and 2MB huge pages sizes.  A 1GB huge page can be split into 512
2632MB huge pages.  Demote interfaces are not available for the smallest
264huge page size.  The demote interfaces are:
265
266demote_size
267        is the size of demoted pages.  When a page is demoted a corresponding
268        number of huge pages of demote_size will be created.  By default,
269        demote_size is set to the next smaller huge page size.  If there are
270        multiple smaller huge page sizes, demote_size can be set to any of
271        these smaller sizes.  Only huge page sizes less than the current huge
272        pages size are allowed.
273
274demote
275        is used to demote a number of huge pages.  A user with root privileges
276        can write to this file.  It may not be possible to demote the
277        requested number of huge pages.  To determine how many pages were
278        actually demoted, compare the value of nr_hugepages before and after
279        writing to the demote interface.  demote is a write only interface.
280
281The interfaces which are the same as in ``/proc`` (all except demote and
282demote_size) function as described above for the default huge page-sized case.
283
284.. _mem_policy_and_hp_alloc:
285
286Interaction of Task Memory Policy with Huge Page Allocation/Freeing
287===================================================================
288
289Whether huge pages are allocated and freed via the ``/proc`` interface or
290the ``/sysfs`` interface using the ``nr_hugepages_mempolicy`` attribute, the
291NUMA nodes from which huge pages are allocated or freed are controlled by the
292NUMA memory policy of the task that modifies the ``nr_hugepages_mempolicy``
293sysctl or attribute.  When the ``nr_hugepages`` attribute is used, mempolicy
294is ignored.
295
296The recommended method to allocate or free huge pages to/from the kernel
297huge page pool, using the ``nr_hugepages`` example above, is::
298
299    numactl --interleave <node-list> echo 20 \
300				>/proc/sys/vm/nr_hugepages_mempolicy
301
302or, more succinctly::
303
304    numactl -m <node-list> echo 20 >/proc/sys/vm/nr_hugepages_mempolicy
305
306This will allocate or free ``abs(20 - nr_hugepages)`` to or from the nodes
307specified in <node-list>, depending on whether number of persistent huge pages
308is initially less than or greater than 20, respectively.  No huge pages will be
309allocated nor freed on any node not included in the specified <node-list>.
310
311When adjusting the persistent hugepage count via ``nr_hugepages_mempolicy``, any
312memory policy mode--bind, preferred, local or interleave--may be used.  The
313resulting effect on persistent huge page allocation is as follows:
314
315#. Regardless of mempolicy mode [see
316   :ref:`Documentation/admin-guide/mm/numa_memory_policy.rst <numa_memory_policy>`],
317   persistent huge pages will be distributed across the node or nodes
318   specified in the mempolicy as if "interleave" had been specified.
319   However, if a node in the policy does not contain sufficient contiguous
320   memory for a huge page, the allocation will not "fallback" to the nearest
321   neighbor node with sufficient contiguous memory.  To do this would cause
322   undesirable imbalance in the distribution of the huge page pool, or
323   possibly, allocation of persistent huge pages on nodes not allowed by
324   the task's memory policy.
325
326#. One or more nodes may be specified with the bind or interleave policy.
327   If more than one node is specified with the preferred policy, only the
328   lowest numeric id will be used.  Local policy will select the node where
329   the task is running at the time the nodes_allowed mask is constructed.
330   For local policy to be deterministic, the task must be bound to a cpu or
331   cpus in a single node.  Otherwise, the task could be migrated to some
332   other node at any time after launch and the resulting node will be
333   indeterminate.  Thus, local policy is not very useful for this purpose.
334   Any of the other mempolicy modes may be used to specify a single node.
335
336#. The nodes allowed mask will be derived from any non-default task mempolicy,
337   whether this policy was set explicitly by the task itself or one of its
338   ancestors, such as numactl.  This means that if the task is invoked from a
339   shell with non-default policy, that policy will be used.  One can specify a
340   node list of "all" with numactl --interleave or --membind [-m] to achieve
341   interleaving over all nodes in the system or cpuset.
342
343#. Any task mempolicy specified--e.g., using numactl--will be constrained by
344   the resource limits of any cpuset in which the task runs.  Thus, there will
345   be no way for a task with non-default policy running in a cpuset with a
346   subset of the system nodes to allocate huge pages outside the cpuset
347   without first moving to a cpuset that contains all of the desired nodes.
348
349#. Boot-time huge page allocation attempts to distribute the requested number
350   of huge pages over all on-lines nodes with memory.
351
352Per Node Hugepages Attributes
353=============================
354
355A subset of the contents of the root huge page control directory in sysfs,
356described above, will be replicated under each the system device of each
357NUMA node with memory in::
358
359	/sys/devices/system/node/node[0-9]*/hugepages/
360
361Under this directory, the subdirectory for each supported huge page size
362contains the following attribute files::
363
364	nr_hugepages
365	free_hugepages
366	surplus_hugepages
367
368The free\_' and surplus\_' attribute files are read-only.  They return the number
369of free and surplus [overcommitted] huge pages, respectively, on the parent
370node.
371
372The ``nr_hugepages`` attribute returns the total number of huge pages on the
373specified node.  When this attribute is written, the number of persistent huge
374pages on the parent node will be adjusted to the specified value, if sufficient
375resources exist, regardless of the task's mempolicy or cpuset constraints.
376
377Note that the number of overcommit and reserve pages remain global quantities,
378as we don't know until fault time, when the faulting task's mempolicy is
379applied, from which node the huge page allocation will be attempted.
380
381.. _using_huge_pages:
382
383Using Huge Pages
384================
385
386If the user applications are going to request huge pages using mmap system
387call, then it is required that system administrator mount a file system of
388type hugetlbfs::
389
390  mount -t hugetlbfs \
391	-o uid=<value>,gid=<value>,mode=<value>,pagesize=<value>,size=<value>,\
392	min_size=<value>,nr_inodes=<value> none /mnt/huge
393
394This command mounts a (pseudo) filesystem of type hugetlbfs on the directory
395``/mnt/huge``.  Any file created on ``/mnt/huge`` uses huge pages.
396
397The ``uid`` and ``gid`` options sets the owner and group of the root of the
398file system.  By default the ``uid`` and ``gid`` of the current process
399are taken.
400
401The ``mode`` option sets the mode of root of file system to value & 01777.
402This value is given in octal. By default the value 0755 is picked.
403
404If the platform supports multiple huge page sizes, the ``pagesize`` option can
405be used to specify the huge page size and associated pool. ``pagesize``
406is specified in bytes. If ``pagesize`` is not specified the platform's
407default huge page size and associated pool will be used.
408
409The ``size`` option sets the maximum value of memory (huge pages) allowed
410for that filesystem (``/mnt/huge``). The ``size`` option can be specified
411in bytes, or as a percentage of the specified huge page pool (``nr_hugepages``).
412The size is rounded down to HPAGE_SIZE boundary.
413
414The ``min_size`` option sets the minimum value of memory (huge pages) allowed
415for the filesystem. ``min_size`` can be specified in the same way as ``size``,
416either bytes or a percentage of the huge page pool.
417At mount time, the number of huge pages specified by ``min_size`` are reserved
418for use by the filesystem.
419If there are not enough free huge pages available, the mount will fail.
420As huge pages are allocated to the filesystem and freed, the reserve count
421is adjusted so that the sum of allocated and reserved huge pages is always
422at least ``min_size``.
423
424The option ``nr_inodes`` sets the maximum number of inodes that ``/mnt/huge``
425can use.
426
427If the ``size``, ``min_size`` or ``nr_inodes`` option is not provided on
428command line then no limits are set.
429
430For ``pagesize``, ``size``, ``min_size`` and ``nr_inodes`` options, you can
431use [G|g]/[M|m]/[K|k] to represent giga/mega/kilo.
432For example, size=2K has the same meaning as size=2048.
433
434While read system calls are supported on files that reside on hugetlb
435file systems, write system calls are not.
436
437Regular chown, chgrp, and chmod commands (with right permissions) could be
438used to change the file attributes on hugetlbfs.
439
440Also, it is important to note that no such mount command is required if
441applications are going to use only shmat/shmget system calls or mmap with
442MAP_HUGETLB.  For an example of how to use mmap with MAP_HUGETLB see
443:ref:`map_hugetlb <map_hugetlb>` below.
444
445Users who wish to use hugetlb memory via shared memory segment should be
446members of a supplementary group and system admin needs to configure that gid
447into ``/proc/sys/vm/hugetlb_shm_group``.  It is possible for same or different
448applications to use any combination of mmaps and shm* calls, though the mount of
449filesystem will be required for using mmap calls without MAP_HUGETLB.
450
451Syscalls that operate on memory backed by hugetlb pages only have their lengths
452aligned to the native page size of the processor; they will normally fail with
453errno set to EINVAL or exclude hugetlb pages that extend beyond the length if
454not hugepage aligned.  For example, munmap(2) will fail if memory is backed by
455a hugetlb page and the length is smaller than the hugepage size.
456
457
458Examples
459========
460
461.. _map_hugetlb:
462
463``map_hugetlb``
464	see tools/testing/selftests/vm/map_hugetlb.c
465
466``hugepage-shm``
467	see tools/testing/selftests/vm/hugepage-shm.c
468
469``hugepage-mmap``
470	see tools/testing/selftests/vm/hugepage-mmap.c
471
472The `libhugetlbfs`_  library provides a wide range of userspace tools
473to help with huge page usability, environment setup, and control.
474
475.. _libhugetlbfs: https://github.com/libhugetlbfs/libhugetlbfs
476