1 .. _admin_guide_transhuge:
3 ============================
4 Transparent Hugepage Support
5 ============================
10 Performance critical computing applications dealing with large memory
11 working sets are already running on top of libhugetlbfs and in turn
12 hugetlbfs. Transparent HugePage Support (THP) is an alternative mean of
13 using huge pages for the backing of virtual memory with huge pages
14 that supports the automatic promotion and demotion of page sizes and
15 without the shortcomings of hugetlbfs.
17 Currently THP only works for anonymous memory mappings and tmpfs/shmem.
18 But in the future it can expand to other filesystems.
21 in the examples below we presume that the basic page size is 4K and
22 the huge page size is 2M, although the actual numbers may vary
23 depending on the CPU architecture.
25 The reason applications are running faster is because of two
26 factors. The first factor is almost completely irrelevant and it's not
27 of significant interest because it'll also have the downside of
28 requiring larger clear-page copy-page in page faults which is a
29 potentially negative effect. The first factor consists in taking a
30 single page fault for each 2M virtual region touched by userland (so
31 reducing the enter/exit kernel frequency by a 512 times factor). This
32 only matters the first time the memory is accessed for the lifetime of
33 a memory mapping. The second long lasting and much more important
34 factor will affect all subsequent accesses to the memory for the whole
35 runtime of the application. The second factor consist of two
38 1) the TLB miss will run faster (especially with virtualization using
39 nested pagetables but almost always also on bare metal without
42 2) a single TLB entry will be mapping a much larger amount of virtual
43 memory in turn reducing the number of TLB misses. With
44 virtualization and nested pagetables the TLB can be mapped of
45 larger size only if both KVM and the Linux guest are using
46 hugepages but a significant speedup already happens if only one of
47 the two is using hugepages just because of the fact the TLB miss is
50 THP can be enabled system wide or restricted to certain tasks or even
51 memory ranges inside task's address space. Unless THP is completely
52 disabled, there is ``khugepaged`` daemon that scans memory and
53 collapses sequences of basic pages into huge pages.
55 The THP behaviour is controlled via :ref:`sysfs <thp_sysfs>`
56 interface and using madvise(2) and prctl(2) system calls.
58 Transparent Hugepage Support maximizes the usefulness of free memory
59 if compared to the reservation approach of hugetlbfs by allowing all
60 unused memory to be used as cache or other movable (or even unmovable
61 entities). It doesn't require reservation to prevent hugepage
62 allocation failures to be noticeable from userland. It allows paging
63 and all other advanced VM features to be available on the
64 hugepages. It requires no modifications for applications to take
67 Applications however can be further optimized to take advantage of
68 this feature, like for example they've been optimized before to avoid
69 a flood of mmap system calls for every malloc(4k). Optimizing userland
70 is by far not mandatory and khugepaged already can take care of long
71 lived page allocations even for hugepage unaware applications that
72 deals with large amounts of memory.
74 In certain cases when hugepages are enabled system wide, application
75 may end up allocating more memory resources. An application may mmap a
76 large region but only touch 1 byte of it, in that case a 2M page might
77 be allocated instead of a 4k page for no good. This is why it's
78 possible to disable hugepages system-wide and to only have them inside
79 MADV_HUGEPAGE madvise regions.
81 Embedded systems should enable hugepages only inside madvise regions
82 to eliminate any risk of wasting any precious byte of memory and to
85 Applications that gets a lot of benefit from hugepages and that don't
86 risk to lose memory by using hugepages, should use
87 madvise(MADV_HUGEPAGE) on their critical mmapped regions.
97 Transparent Hugepage Support for anonymous memory can be entirely disabled
98 (mostly for debugging purposes) or only enabled inside MADV_HUGEPAGE
99 regions (to avoid the risk of consuming more memory resources) or enabled
100 system wide. This can be achieved with one of::
102 echo always >/sys/kernel/mm/transparent_hugepage/enabled
103 echo madvise >/sys/kernel/mm/transparent_hugepage/enabled
104 echo never >/sys/kernel/mm/transparent_hugepage/enabled
106 It's also possible to limit defrag efforts in the VM to generate
107 anonymous hugepages in case they're not immediately free to madvise
108 regions or to never try to defrag memory and simply fallback to regular
109 pages unless hugepages are immediately available. Clearly if we spend CPU
110 time to defrag memory, we would expect to gain even more by the fact we
111 use hugepages later instead of regular pages. This isn't always
112 guaranteed, but it may be more likely in case the allocation is for a
113 MADV_HUGEPAGE region.
117 echo always >/sys/kernel/mm/transparent_hugepage/defrag
118 echo defer >/sys/kernel/mm/transparent_hugepage/defrag
119 echo defer+madvise >/sys/kernel/mm/transparent_hugepage/defrag
120 echo madvise >/sys/kernel/mm/transparent_hugepage/defrag
121 echo never >/sys/kernel/mm/transparent_hugepage/defrag
124 means that an application requesting THP will stall on
125 allocation failure and directly reclaim pages and compact
126 memory in an effort to allocate a THP immediately. This may be
127 desirable for virtual machines that benefit heavily from THP
128 use and are willing to delay the VM start to utilise them.
131 means that an application will wake kswapd in the background
132 to reclaim pages and wake kcompactd to compact memory so that
133 THP is available in the near future. It's the responsibility
134 of khugepaged to then install the THP pages later.
137 will enter direct reclaim and compaction like ``always``, but
138 only for regions that have used madvise(MADV_HUGEPAGE); all
139 other regions will wake kswapd in the background to reclaim
140 pages and wake kcompactd to compact memory so that THP is
141 available in the near future.
144 will enter direct reclaim like ``always`` but only for regions
145 that are have used madvise(MADV_HUGEPAGE). This is the default
149 should be self-explanatory.
151 By default kernel tries to use huge zero page on read page fault to
152 anonymous mapping. It's possible to disable huge zero page by writing 0
153 or enable it back by writing 1::
155 echo 0 >/sys/kernel/mm/transparent_hugepage/use_zero_page
156 echo 1 >/sys/kernel/mm/transparent_hugepage/use_zero_page
158 Some userspace (such as a test program, or an optimized memory allocation
159 library) may want to know the size (in bytes) of a transparent hugepage::
161 cat /sys/kernel/mm/transparent_hugepage/hpage_pmd_size
163 khugepaged will be automatically started when
164 transparent_hugepage/enabled is set to "always" or "madvise, and it'll
165 be automatically shutdown if it's set to "never".
170 khugepaged runs usually at low frequency so while one may not want to
171 invoke defrag algorithms synchronously during the page faults, it
172 should be worth invoking defrag at least in khugepaged. However it's
173 also possible to disable defrag in khugepaged by writing 0 or enable
174 defrag in khugepaged by writing 1::
176 echo 0 >/sys/kernel/mm/transparent_hugepage/khugepaged/defrag
177 echo 1 >/sys/kernel/mm/transparent_hugepage/khugepaged/defrag
179 You can also control how many pages khugepaged should scan at each
182 /sys/kernel/mm/transparent_hugepage/khugepaged/pages_to_scan
184 and how many milliseconds to wait in khugepaged between each pass (you
185 can set this to 0 to run khugepaged at 100% utilization of one core)::
187 /sys/kernel/mm/transparent_hugepage/khugepaged/scan_sleep_millisecs
189 and how many milliseconds to wait in khugepaged if there's an hugepage
190 allocation failure to throttle the next allocation attempt::
192 /sys/kernel/mm/transparent_hugepage/khugepaged/alloc_sleep_millisecs
194 The khugepaged progress can be seen in the number of pages collapsed::
196 /sys/kernel/mm/transparent_hugepage/khugepaged/pages_collapsed
200 /sys/kernel/mm/transparent_hugepage/khugepaged/full_scans
202 ``max_ptes_none`` specifies how many extra small pages (that are
203 not already mapped) can be allocated when collapsing a group
204 of small pages into one large page::
206 /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_none
208 A higher value leads to use additional memory for programs.
209 A lower value leads to gain less thp performance. Value of
210 max_ptes_none can waste cpu time very little, you can
213 ``max_ptes_swap`` specifies how many pages can be brought in from
214 swap when collapsing a group of pages into a transparent huge page::
216 /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_swap
218 A higher value can cause excessive swap IO and waste
219 memory. A lower value can prevent THPs from being
220 collapsed, resulting fewer pages being collapsed into
221 THPs, and lower memory access performance.
223 ``max_ptes_shared`` specifies how many pages can be shared across multiple
224 processes. Exceeding the number would block the collapse::
226 /sys/kernel/mm/transparent_hugepage/khugepaged/max_ptes_shared
228 A higher value may increase memory footprint for some workloads.
233 You can change the sysfs boot time defaults of Transparent Hugepage
234 Support by passing the parameter ``transparent_hugepage=always`` or
235 ``transparent_hugepage=madvise`` or ``transparent_hugepage=never``
236 to the kernel command line.
238 Hugepages in tmpfs/shmem
239 ========================
241 You can control hugepage allocation policy in tmpfs with mount option
242 ``huge=``. It can have following values:
245 Attempt to allocate huge pages every time we need a new page;
248 Do not allocate huge pages;
251 Only allocate huge page if it will be fully within i_size.
252 Also respect fadvise()/madvise() hints;
255 Only allocate huge pages if requested with fadvise()/madvise();
257 The default policy is ``never``.
259 ``mount -o remount,huge= /mountpoint`` works fine after mount: remounting
260 ``huge=never`` will not attempt to break up huge pages at all, just stop more
261 from being allocated.
263 There's also sysfs knob to control hugepage allocation policy for internal
264 shmem mount: /sys/kernel/mm/transparent_hugepage/shmem_enabled. The mount
265 is used for SysV SHM, memfds, shared anonymous mmaps (of /dev/zero or
266 MAP_ANONYMOUS), GPU drivers' DRM objects, Ashmem.
268 In addition to policies listed above, shmem_enabled allows two further
272 For use in emergencies, to force the huge option off from
275 Force the huge option on for all - very useful for testing;
277 Need of application restart
278 ===========================
280 The transparent_hugepage/enabled values and tmpfs mount option only affect
281 future behavior. So to make them effective you need to restart any
282 application that could have been using hugepages. This also applies to the
283 regions registered in khugepaged.
288 The number of anonymous transparent huge pages currently used by the
289 system is available by reading the AnonHugePages field in ``/proc/meminfo``.
290 To identify what applications are using anonymous transparent huge pages,
291 it is necessary to read ``/proc/PID/smaps`` and count the AnonHugePages fields
294 The number of file transparent huge pages mapped to userspace is available
295 by reading ShmemPmdMapped and ShmemHugePages fields in ``/proc/meminfo``.
296 To identify what applications are mapping file transparent huge pages, it
297 is necessary to read ``/proc/PID/smaps`` and count the FileHugeMapped fields
300 Note that reading the smaps file is expensive and reading it
301 frequently will incur overhead.
303 There are a number of counters in ``/proc/vmstat`` that may be used to
304 monitor how successfully the system is providing huge pages for use.
307 is incremented every time a huge page is successfully
308 allocated to handle a page fault.
311 is incremented by khugepaged when it has found
312 a range of pages to collapse into one huge page and has
313 successfully allocated a new huge page to store the data.
316 is incremented if a page fault fails to allocate
317 a huge page and instead falls back to using small pages.
319 thp_fault_fallback_charge
320 is incremented if a page fault fails to charge a huge page and
321 instead falls back to using small pages even though the
322 allocation was successful.
324 thp_collapse_alloc_failed
325 is incremented if khugepaged found a range
326 of pages that should be collapsed into one huge page but failed
330 is incremented every time a file huge page is successfully
334 is incremented if a file huge page is attempted to be allocated
335 but fails and instead falls back to using small pages.
337 thp_file_fallback_charge
338 is incremented if a file huge page cannot be charged and instead
339 falls back to using small pages even though the allocation was
343 is incremented every time a file huge page is mapped into
347 is incremented every time a huge page is split into base
348 pages. This can happen for a variety of reasons but a common
349 reason is that a huge page is old and is being reclaimed.
350 This action implies splitting all PMD the page mapped with.
352 thp_split_page_failed
353 is incremented if kernel fails to split huge
354 page. This can happen if the page was pinned by somebody.
356 thp_deferred_split_page
357 is incremented when a huge page is put onto split
358 queue. This happens when a huge page is partially unmapped and
359 splitting it would free up some memory. Pages on split queue are
360 going to be split under memory pressure.
363 is incremented every time a PMD split into table of PTEs.
364 This can happen, for instance, when application calls mprotect() or
365 munmap() on part of huge page. It doesn't split huge page, only
369 is incremented every time a huge zero page is
370 successfully allocated. It includes allocations which where
371 dropped due race with other allocation. Note, it doesn't count
372 every map of the huge zero page, only its allocation.
374 thp_zero_page_alloc_failed
375 is incremented if kernel fails to allocate
376 huge zero page and falls back to using small pages.
379 is incremented every time a huge page is swapout in one
380 piece without splitting.
383 is incremented if a huge page has to be split before swapout.
384 Usually because failed to allocate some continuous swap space
387 As the system ages, allocating huge pages may be expensive as the
388 system uses memory compaction to copy data around memory to free a
389 huge page for use. There are some counters in ``/proc/vmstat`` to help
390 monitor this overhead.
393 is incremented every time a process stalls to run
394 memory compaction so that a huge page is free for use.
397 is incremented if the system compacted memory and
398 freed a huge page for use.
401 is incremented if the system tries to compact memory
404 It is possible to establish how long the stalls were using the function
405 tracer to record how long was spent in __alloc_pages_nodemask and
406 using the mm_page_alloc tracepoint to identify which allocations were
409 Optimizing the applications
410 ===========================
412 To be guaranteed that the kernel will map a 2M page immediately in any
413 memory region, the mmap region has to be hugepage naturally
414 aligned. posix_memalign() can provide that guarantee.
419 You can use hugetlbfs on a kernel that has transparent hugepage
420 support enabled just fine as always. No difference can be noted in
421 hugetlbfs other than there will be less overall fragmentation. All
422 usual features belonging to hugetlbfs are preserved and
423 unaffected. libhugetlbfs will also work fine as usual.