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 four 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 (since 4.2)
21 * Bit 61 page is file-page or shared-anon (since 3.5)
25 Since Linux 4.0 only users with the CAP_SYS_ADMIN capability can get PFNs.
26 In 4.0 and 4.1 opens by unprivileged fail with -EPERM. Starting from
27 4.2 the PFN field is zeroed if the user does not have CAP_SYS_ADMIN.
28 Reason: information about PFNs helps in exploiting Rowhammer vulnerability.
30 If the page is not present but in swap, then the PFN contains an
31 encoding of the swap file number and the page's offset into the
32 swap. Unmapped pages return a null PFN. This allows determining
33 precisely which pages are mapped (or in swap) and comparing mapped
34 pages between processes.
36 Efficient users of this interface will use /proc/pid/maps to
37 determine which areas of memory are actually mapped and llseek to
38 skip over unmapped regions.
40 * /proc/kpagecount. This file contains a 64-bit count of the number of
41 times each page is mapped, indexed by PFN.
43 * /proc/kpageflags. This file contains a 64-bit set of flags for each
46 The flags are (from fs/proc/page.c, above kpageflags_read):
74 * /proc/kpagecgroup. This file contains a 64-bit inode number of the
75 memory cgroup each page is charged to, indexed by PFN. Only available when
78 Short descriptions to the page flags:
81 page is being locked for exclusive access, eg. by undergoing read/write IO
84 page is managed by the SLAB/SLOB/SLUB/SLQB kernel memory allocator
85 When compound page is used, SLUB/SLQB will only set this flag on the head
86 page; SLOB will not flag it at all.
89 a free memory block managed by the buddy system allocator
90 The buddy system organizes free memory in blocks of various orders.
91 An order N block has 2^N physically contiguous pages, with the BUDDY flag
92 set for and _only_ for the first page.
96 A compound page with order N consists of 2^N physically contiguous pages.
97 A compound page with order 2 takes the form of "HTTT", where H donates its
98 head page and T donates its tail page(s). The major consumers of compound
99 pages are hugeTLB pages (Documentation/vm/hugetlbpage.txt), the SLUB etc.
100 memory allocators and various device drivers. However in this interface,
101 only huge/giga pages are made visible to end users.
103 this is an integral part of a HugeTLB page
106 hardware detected memory corruption on this page: don't touch the data!
109 no page frame exists at the requested address
112 identical memory pages dynamically shared between one or more processes
115 contiguous pages which construct transparent hugepages
118 balloon compaction page
121 zero page for pfn_zero or huge_zero page
123 [IO related page flags]
124 1. ERROR IO error occurred
125 3. UPTODATE page has up-to-date data
126 ie. for file backed page: (in-memory data revision >= on-disk one)
127 4. DIRTY page has been written to, hence contains new data
128 ie. for file backed page: (in-memory data revision > on-disk one)
129 8. WRITEBACK page is being synced to disk
131 [LRU related page flags]
132 5. LRU page is in one of the LRU lists
133 6. ACTIVE page is in the active LRU list
134 18. UNEVICTABLE page is in the unevictable (non-)LRU list
135 It is somehow pinned and not a candidate for LRU page reclaims,
136 eg. ramfs pages, shmctl(SHM_LOCK) and mlock() memory segments
137 2. REFERENCED page has been referenced since last LRU list enqueue/requeue
138 9. RECLAIM page will be reclaimed soon after its pageout IO completed
139 11. MMAP a memory mapped page
140 12. ANON a memory mapped page that is not part of a file
141 13. SWAPCACHE page is mapped to swap space, ie. has an associated swap entry
142 14. SWAPBACKED page is backed by swap/RAM
144 The page-types tool in the tools/vm directory can be used to query the
147 Using pagemap to do something useful:
149 The general procedure for using pagemap to find out about a process' memory
150 usage goes like this:
152 1. Read /proc/pid/maps to determine which parts of the memory space are
154 2. Select the maps you are interested in -- all of them, or a particular
155 library, or the stack or the heap, etc.
156 3. Open /proc/pid/pagemap and seek to the pages you would like to examine.
157 4. Read a u64 for each page from pagemap.
158 5. Open /proc/kpagecount and/or /proc/kpageflags. For each PFN you just
159 read, seek to that entry in the file, and read the data you want.
161 For example, to find the "unique set size" (USS), which is the amount of
162 memory that a process is using that is not shared with any other process,
163 you can go through every map in the process, find the PFNs, look those up
164 in kpagecount, and tally up the number of pages that are only referenced
169 Reading from any of the files will return -EINVAL if you are not starting
170 the read on an 8-byte boundary (e.g., if you sought an odd number of bytes
171 into the file), or if the size of the read is not a multiple of 8 bytes.
173 Before Linux 3.11 pagemap bits 55-60 were used for "page-shift" (which is
174 always 12 at most architectures). Since Linux 3.11 their meaning changes
175 after first clear of soft-dirty bits. Since Linux 4.2 they are used for
176 flags unconditionally.