1pagemap, from the userspace perspective
2---------------------------------------
3
4pagemap is a new (as of 2.6.25) set of interfaces in the kernel that allow
5userspace programs to examine the page tables and related information by
6reading files in /proc.
7
8There are three components to pagemap:
9
10 * /proc/pid/pagemap.  This file lets a userspace process find out which
11   physical frame each virtual page is mapped to.  It contains one 64-bit
12   value for each virtual page, containing the following data (from
13   fs/proc/task_mmu.c, above pagemap_read):
14
15    * Bits 0-54  page frame number (PFN) if present
16    * Bits 0-4   swap type if swapped
17    * Bits 5-54  swap offset if swapped
18    * Bit  55    pte is soft-dirty (see Documentation/vm/soft-dirty.txt)
19    * Bits 56-60 zero
20    * Bit  61    page is file-page or shared-anon
21    * Bit  62    page swapped
22    * Bit  63    page present
23
24   If the page is not present but in swap, then the PFN contains an
25   encoding of the swap file number and the page's offset into the
26   swap. Unmapped pages return a null PFN. This allows determining
27   precisely which pages are mapped (or in swap) and comparing mapped
28   pages between processes.
29
30   Efficient users of this interface will use /proc/pid/maps to
31   determine which areas of memory are actually mapped and llseek to
32   skip over unmapped regions.
33
34 * /proc/kpagecount.  This file contains a 64-bit count of the number of
35   times each page is mapped, indexed by PFN.
36
37 * /proc/kpageflags.  This file contains a 64-bit set of flags for each
38   page, indexed by PFN.
39
40   The flags are (from fs/proc/page.c, above kpageflags_read):
41
42     0. LOCKED
43     1. ERROR
44     2. REFERENCED
45     3. UPTODATE
46     4. DIRTY
47     5. LRU
48     6. ACTIVE
49     7. SLAB
50     8. WRITEBACK
51     9. RECLAIM
52    10. BUDDY
53    11. MMAP
54    12. ANON
55    13. SWAPCACHE
56    14. SWAPBACKED
57    15. COMPOUND_HEAD
58    16. COMPOUND_TAIL
59    16. HUGE
60    18. UNEVICTABLE
61    19. HWPOISON
62    20. NOPAGE
63    21. KSM
64    22. THP
65    23. BALLOON
66    24. ZERO_PAGE
67
68Short descriptions to the page flags:
69
70 0. LOCKED
71    page is being locked for exclusive access, eg. by undergoing read/write IO
72
73 7. SLAB
74    page is managed by the SLAB/SLOB/SLUB/SLQB kernel memory allocator
75    When compound page is used, SLUB/SLQB will only set this flag on the head
76    page; SLOB will not flag it at all.
77
7810. BUDDY
79    a free memory block managed by the buddy system allocator
80    The buddy system organizes free memory in blocks of various orders.
81    An order N block has 2^N physically contiguous pages, with the BUDDY flag
82    set for and _only_ for the first page.
83
8415. COMPOUND_HEAD
8516. COMPOUND_TAIL
86    A compound page with order N consists of 2^N physically contiguous pages.
87    A compound page with order 2 takes the form of "HTTT", where H donates its
88    head page and T donates its tail page(s).  The major consumers of compound
89    pages are hugeTLB pages (Documentation/vm/hugetlbpage.txt), the SLUB etc.
90    memory allocators and various device drivers. However in this interface,
91    only huge/giga pages are made visible to end users.
9217. HUGE
93    this is an integral part of a HugeTLB page
94
9519. HWPOISON
96    hardware detected memory corruption on this page: don't touch the data!
97
9820. NOPAGE
99    no page frame exists at the requested address
100
10121. KSM
102    identical memory pages dynamically shared between one or more processes
103
10422. THP
105    contiguous pages which construct transparent hugepages
106
10723. BALLOON
108    balloon compaction page
109
11024. ZERO_PAGE
111    zero page for pfn_zero or huge_zero page
112
113    [IO related page flags]
114 1. ERROR     IO error occurred
115 3. UPTODATE  page has up-to-date data
116              ie. for file backed page: (in-memory data revision >= on-disk one)
117 4. DIRTY     page has been written to, hence contains new data
118              ie. for file backed page: (in-memory data revision >  on-disk one)
119 8. WRITEBACK page is being synced to disk
120
121    [LRU related page flags]
122 5. LRU         page is in one of the LRU lists
123 6. ACTIVE      page is in the active LRU list
12418. UNEVICTABLE page is in the unevictable (non-)LRU list
125                It is somehow pinned and not a candidate for LRU page reclaims,
126		eg. ramfs pages, shmctl(SHM_LOCK) and mlock() memory segments
127 2. REFERENCED  page has been referenced since last LRU list enqueue/requeue
128 9. RECLAIM     page will be reclaimed soon after its pageout IO completed
12911. MMAP        a memory mapped page
13012. ANON        a memory mapped page that is not part of a file
13113. SWAPCACHE   page is mapped to swap space, ie. has an associated swap entry
13214. SWAPBACKED  page is backed by swap/RAM
133
134The page-types tool in the tools/vm directory can be used to query the
135above flags.
136
137Using pagemap to do something useful:
138
139The general procedure for using pagemap to find out about a process' memory
140usage goes like this:
141
142 1. Read /proc/pid/maps to determine which parts of the memory space are
143    mapped to what.
144 2. Select the maps you are interested in -- all of them, or a particular
145    library, or the stack or the heap, etc.
146 3. Open /proc/pid/pagemap and seek to the pages you would like to examine.
147 4. Read a u64 for each page from pagemap.
148 5. Open /proc/kpagecount and/or /proc/kpageflags.  For each PFN you just
149    read, seek to that entry in the file, and read the data you want.
150
151For example, to find the "unique set size" (USS), which is the amount of
152memory that a process is using that is not shared with any other process,
153you can go through every map in the process, find the PFNs, look those up
154in kpagecount, and tally up the number of pages that are only referenced
155once.
156
157Other notes:
158
159Reading from any of the files will return -EINVAL if you are not starting
160the read on an 8-byte boundary (e.g., if you sought an odd number of bytes
161into the file), or if the size of the read is not a multiple of 8 bytes.
162