1 pagemap, from the userspace perspective
2 ---------------------------------------
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.
8 There are three components to pagemap:
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):
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 * Bit 56 page exclusively mapped
21 * Bit 61 page is file-page or shared-anon
25 If the page is not present but in swap, then the PFN contains an
26 encoding of the swap file number and the page's offset into the
27 swap. Unmapped pages return a null PFN. This allows determining
28 precisely which pages are mapped (or in swap) and comparing mapped
29 pages between processes.
31 Efficient users of this interface will use /proc/pid/maps to
32 determine which areas of memory are actually mapped and llseek to
33 skip over unmapped regions.
35 * /proc/kpagecount. This file contains a 64-bit count of the number of
36 times each page is mapped, indexed by PFN.
38 * /proc/kpageflags. This file contains a 64-bit set of flags for each
41 The flags are (from fs/proc/page.c, above kpageflags_read):
69 Short descriptions to the page flags:
72 page is being locked for exclusive access, eg. by undergoing read/write IO
75 page is managed by the SLAB/SLOB/SLUB/SLQB kernel memory allocator
76 When compound page is used, SLUB/SLQB will only set this flag on the head
77 page; SLOB will not flag it at all.
80 a free memory block managed by the buddy system allocator
81 The buddy system organizes free memory in blocks of various orders.
82 An order N block has 2^N physically contiguous pages, with the BUDDY flag
83 set for and _only_ for the first page.
87 A compound page with order N consists of 2^N physically contiguous pages.
88 A compound page with order 2 takes the form of "HTTT", where H donates its
89 head page and T donates its tail page(s). The major consumers of compound
90 pages are hugeTLB pages (Documentation/vm/hugetlbpage.txt), the SLUB etc.
91 memory allocators and various device drivers. However in this interface,
92 only huge/giga pages are made visible to end users.
94 this is an integral part of a HugeTLB page
97 hardware detected memory corruption on this page: don't touch the data!
100 no page frame exists at the requested address
103 identical memory pages dynamically shared between one or more processes
106 contiguous pages which construct transparent hugepages
109 balloon compaction page
112 zero page for pfn_zero or huge_zero page
114 [IO related page flags]
115 1. ERROR IO error occurred
116 3. UPTODATE page has up-to-date data
117 ie. for file backed page: (in-memory data revision >= on-disk one)
118 4. DIRTY page has been written to, hence contains new data
119 ie. for file backed page: (in-memory data revision > on-disk one)
120 8. WRITEBACK page is being synced to disk
122 [LRU related page flags]
123 5. LRU page is in one of the LRU lists
124 6. ACTIVE page is in the active LRU list
125 18. UNEVICTABLE page is in the unevictable (non-)LRU list
126 It is somehow pinned and not a candidate for LRU page reclaims,
127 eg. ramfs pages, shmctl(SHM_LOCK) and mlock() memory segments
128 2. REFERENCED page has been referenced since last LRU list enqueue/requeue
129 9. RECLAIM page will be reclaimed soon after its pageout IO completed
130 11. MMAP a memory mapped page
131 12. ANON a memory mapped page that is not part of a file
132 13. SWAPCACHE page is mapped to swap space, ie. has an associated swap entry
133 14. SWAPBACKED page is backed by swap/RAM
135 The page-types tool in the tools/vm directory can be used to query the
138 Using pagemap to do something useful:
140 The general procedure for using pagemap to find out about a process' memory
141 usage goes like this:
143 1. Read /proc/pid/maps to determine which parts of the memory space are
145 2. Select the maps you are interested in -- all of them, or a particular
146 library, or the stack or the heap, etc.
147 3. Open /proc/pid/pagemap and seek to the pages you would like to examine.
148 4. Read a u64 for each page from pagemap.
149 5. Open /proc/kpagecount and/or /proc/kpageflags. For each PFN you just
150 read, seek to that entry in the file, and read the data you want.
152 For example, to find the "unique set size" (USS), which is the amount of
153 memory that a process is using that is not shared with any other process,
154 you can go through every map in the process, find the PFNs, look those up
155 in kpagecount, and tally up the number of pages that are only referenced
160 Reading from any of the files will return -EINVAL if you are not starting
161 the read on an 8-byte boundary (e.g., if you sought an odd number of bytes
162 into the file), or if the size of the read is not a multiple of 8 bytes.