Documentation/vm/pagemap.txt

   1 pagemap, from the userspace perspective
   2 ---------------------------------------
   3
   4 pagemap is a new (as of 2.6.25) set of interfaces in the kernel that allow
   5 userspace programs to examine the page tables and related information by
   6 reading files in /proc.
   7
   8 There 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     * Bits 55-60 page shift (page size = 1<<page shift)
  19     * Bit  61    page is file-page or shared-anon
  20     * Bit  62    page swapped
  21     * Bit  63    page present
  22
  23    If the page is not present but in swap, then the PFN contains an
  24    encoding of the swap file number and the page's offset into the
  25    swap. Unmapped pages return a null PFN. This allows determining
  26    precisely which pages are mapped (or in swap) and comparing mapped
  27    pages between processes.
  28
  29    Efficient users of this interface will use /proc/pid/maps to
  30    determine which areas of memory are actually mapped and llseek to
  31    skip over unmapped regions.
  32
  33  * /proc/kpagecount.  This file contains a 64-bit count of the number of
  34    times each page is mapped, indexed by PFN.
  35
  36  * /proc/kpageflags.  This file contains a 64-bit set of flags for each
  37    page, indexed by PFN.
  38
  39    The flags are (from fs/proc/page.c, above kpageflags_read):
  40
  41      0. LOCKED
  42      1. ERROR
  43      2. REFERENCED
  44      3. UPTODATE
  45      4. DIRTY
  46      5. LRU
  47      6. ACTIVE
  48      7. SLAB
  49      8. WRITEBACK
  50      9. RECLAIM
  51     10. BUDDY
  52     11. MMAP
  53     12. ANON
  54     13. SWAPCACHE
  55     14. SWAPBACKED
  56     15. COMPOUND_HEAD
  57     16. COMPOUND_TAIL
  58     16. HUGE
  59     18. UNEVICTABLE
  60     19. HWPOISON
  61     20. NOPAGE
  62     21. KSM
  63     22. THP
  64
  65 Short descriptions to the page flags:
  66
  67  0. LOCKED
  68     page is being locked for exclusive access, eg. by undergoing read/write IO
  69
  70  7. SLAB
  71     page is managed by the SLAB/SLOB/SLUB/SLQB kernel memory allocator
  72     When compound page is used, SLUB/SLQB will only set this flag on the head
  73     page; SLOB will not flag it at all.
  74
  75 10. BUDDY
  76     a free memory block managed by the buddy system allocator
  77     The buddy system organizes free memory in blocks of various orders.
  78     An order N block has 2^N physically contiguous pages, with the BUDDY flag
  79     set for and _only_ for the first page.
  80
  81 15. COMPOUND_HEAD
  82 16. COMPOUND_TAIL
  83     A compound page with order N consists of 2^N physically contiguous pages.
  84     A compound page with order 2 takes the form of "HTTT", where H donates its
  85     head page and T donates its tail page(s).  The major consumers of compound
  86     pages are hugeTLB pages (Documentation/vm/hugetlbpage.txt), the SLUB etc.
  87     memory allocators and various device drivers. However in this interface,
  88     only huge/giga pages are made visible to end users.
  89 17. HUGE
  90     this is an integral part of a HugeTLB page
  91
  92 19. HWPOISON
  93     hardware detected memory corruption on this page: don't touch the data!
  94
  95 20. NOPAGE
  96     no page frame exists at the requested address
  97
  98 21. KSM
  99     identical memory pages dynamically shared between one or more processes
 100
 101 22. THP
 102     contiguous pages which construct transparent hugepages
 103
 104     [IO related page flags]
 105  1. ERROR     IO error occurred
 106  3. UPTODATE  page has up-to-date data
 107               ie. for file backed page: (in-memory data revision >= on-disk one)
 108  4. DIRTY     page has been written to, hence contains new data
 109               ie. for file backed page: (in-memory data revision >  on-disk one)
 110  8. WRITEBACK page is being synced to disk
 111
 112     [LRU related page flags]
 113  5. LRU         page is in one of the LRU lists
 114  6. ACTIVE      page is in the active LRU list
 115 18. UNEVICTABLE page is in the unevictable (non-)LRU list
 116                 It is somehow pinned and not a candidate for LRU page reclaims,
 117                 eg. ramfs pages, shmctl(SHM_LOCK) and mlock() memory segments
 118  2. REFERENCED  page has been referenced since last LRU list enqueue/requeue
 119  9. RECLAIM     page will be reclaimed soon after its pageout IO completed
 120 11. MMAP        a memory mapped page
 121 12. ANON        a memory mapped page that is not part of a file
 122 13. SWAPCACHE   page is mapped to swap space, ie. has an associated swap entry
 123 14. SWAPBACKED  page is backed by swap/RAM
 124
 125 The page-types tool in this directory can be used to query the above flags.
 126
 127 Using pagemap to do something useful:
 128
 129 The general procedure for using pagemap to find out about a process' memory
 130 usage goes like this:
 131
 132  1. Read /proc/pid/maps to determine which parts of the memory space are
 133     mapped to what.
 134  2. Select the maps you are interested in -- all of them, or a particular
 135     library, or the stack or the heap, etc.
 136  3. Open /proc/pid/pagemap and seek to the pages you would like to examine.
 137  4. Read a u64 for each page from pagemap.
 138  5. Open /proc/kpagecount and/or /proc/kpageflags.  For each PFN you just
 139     read, seek to that entry in the file, and read the data you want.
 140
 141 For example, to find the "unique set size" (USS), which is the amount of
 142 memory that a process is using that is not shared with any other process,
 143 you can go through every map in the process, find the PFNs, look those up
 144 in kpagecount, and tally up the number of pages that are only referenced
 145 once.
 146
 147 Other notes:
 148
 149 Reading from any of the files will return -EINVAL if you are not starting
 150 the read on an 8-byte boundary (e.g., if you seeked an odd number of bytes
 151 into the file), or if the size of the read is not a multiple of 8 bytes.