Linux内核---62.用户空间获得变量的物理地址

1.  linux内核文档Documentation/vm/pagemap.txt

1.2  用户空间获取物理地址的代码

1. cong@msi:/work/test/ctest/vaddr$ cat test.c 
   
2. #include <stdio.h>
   
3. #include <unistd.h>
   
4. #include <inttypes.h>
   
5. #include <stdint.h>
   
6. #include <sys/types.h>
   
7. #include <sys/stat.h>
   
8. #include <fcntl.h>
   
9. 
   
10. #define page_map_file "/proc/self/pagemap"
    
11. 
    
12. #define PFN_MASK ((((uint64_t)1)<<55)-1)
    
13. 
    
14. #define PFN_PRESENT_FLAG (((uint64_t)1)<<63)
    
15. 
    
16. int mem_addr_vir2phy(unsigned long vir, unsigned long *phy)
    
17. {
    
18.     int fd;
    
19.     int page_size=getpagesize();
    
20.     unsigned long vir_page_idx = vir/page_size;
    
21.     unsigned long pfn_item_offset = vir_page_idx*sizeof(uint64_t);
    
22.     uint64_t pfn_item;
    
23. 
    
24.     fd = open(page_map_file, O_RDONLY);
    
25.     if (fd<0)
    
26.     {
    
27.         printf("open %s failed", page_map_file);
    
28.         return -1;
    
29.     }
    
30.     if ((off_t)-1 == lseek(fd, pfn_item_offset, SEEK_SET))
    
31.     {
    
32.         printf("lseek %s failed", page_map_file);
    
33.         return -1;
    
34.     }
    
35.     if (sizeof(uint64_t) != read(fd, &pfn_item, sizeof(uint64_t)))
    
36.     {
    
37.         printf("read %s failed", page_map_file);
    
38.         return -1;
    
39.     }
    
40.     if (0==(pfn_item & PFN_PRESENT_FLAG))
    
41.     {
    
42.         printf("page is not present");
    
43.         return -1;
    
44.     }
    
45.     *phy = (pfn_item & PFN_MASK)*page_size + vir % page_size;
    
46.     return 0;
    
47. }
    
48. 
    
49. void main()
    
50. {
    
51.     int a=0x12345678;
    
52.     unsigned long phy;
    
53.     mem_addr_vir2phy((unsigned long)&a, &phy);
    
54.     printf("vaddr=%p,phy=0x%lx\n", &a, phy);
    
55.     while(1)
    
56.     {
    
57.         sleep(100);
    
58.     }
    
59. }

注：函数mem_addr_vir2phy代码出自 
《Linux下获取虚拟地址对应的物理地址的方法》
http://www.lai18.com/content/7598292.html
这儿我只是打了一个main函数进行测试

---

二.验证结果的正确性
2.1 使用方法
cong@msi:/work/test/ctest/vaddr/Access_Physical_Memory$ sudo mknod /dev/dram c 85 0

cong@msi:/work/test/ctest/vaddr/Access_Physical_Memory$ ./fileview /dev/dram

注:上述驱动与fileview出自
《Linux用户程序如何访问物理内存》
http://ilinuxkernel.com/?p=1248
2.2.2 我的ubuntun 的内核版本是 3.13.0-49-generic
作了少许改动

1. 27d26
   
2. < #include <linux/mm.h>
   
3. 47,49c46,47
   
4. <     unsigned long pages = get_num_physpages();    
   
5. <     //dram_size = (loff_t)num_physpages << PAGE_SHIFT;
   
6. <     dram_size = pages << PAGE_SHIFT;
   
7. ---
   
8. >     
   
9. >     dram_size = (loff_t)num_physpages << PAGE_SHIFT;
   
10. 77c75
    
11. < #if 1
    
12. ---
    
13. > #if 0

2.3 验证一下结果
2.3.1 这个是用户空间打印出来的物理地址是0x41F47284


2.3.2 这是物理内存0x41F47284处的数据

完全吻合，说明成功
2.4 代码下载
包括fileview和驱动
Access_Physical_Memory.rar(下载后改名为Access_Physical_Memory.tar.bz2)
这个代码出自 http://cs.usfca.edu/~cruse/cs635/

---

附录: 内核文档 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 reserved for future use
    
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. 
    
64. Short descriptions to the page flags:
    
65. 
    
66.  0. LOCKED
    
67.     page is being locked for exclusive access, eg. by undergoing read/write IO
    
68. 
    
69.  7. SLAB
    
70.     page is managed by the SLAB/SLOB/SLUB/SLQB kernel memory allocator
    
71.     When compound page is used, SLUB/SLQB will only set this flag on the head
    
72.     page; SLOB will not flag it at all.
    
73. 
    
74. 10. BUDDY
    
75.     a free memory block managed by the buddy system allocator
    
76.     The buddy system organizes free memory in blocks of various orders.
    
77.     An order N block has 2^N physically contiguous pages, with the BUDDY flag
    
78.     set for and _only_ for the first page.
    
79. 
    
80. 15. COMPOUND_HEAD
    
81. 16. COMPOUND_TAIL
    
82.     A compound page with order N consists of 2^N physically contiguous pages.
    
83.     A compound page with order 2 takes the form of "HTTT", where H donates its
    
84.     head page and T donates its tail page(s). The major consumers of compound
    
85.     pages are hugeTLB pages (Documentation/vm/hugetlbpage.txt), the SLUB etc.
    
86.     memory allocators and various device drivers. However in this interface,
    
87.     only huge/giga pages are made visible to end users.
    
88. 17. HUGE
    
89.     this is an integral part of a HugeTLB page
    
90. 
    
91. 19. HWPOISON
    
92.     hardware detected memory corruption on this page: don't touch the 
    
93. 
    
94. 20. NOPAGE
    
95.     no page frame exists at the requested address
    
96. 
    
97. 21. KSM
    
98.     identical memory pages dynamically shared between one or more processes
    
99. 
    
100.     [IO related page flags]
     
101.  1. ERROR IO error occurred
     
102.  3. UPTODATE page has up-to-date data
     
103.               ie. for file backed page: (in-memory data revision >= on-disk one)
     
104.  4. DIRTY page has been written to, hence contains new data
     
105.               ie. for file backed page: (in-memory data revision > on-disk one)
     
106.  8. WRITEBACK page is being synced to disk
     
107. 
     
108.     [LRU related page flags]
     
109.  5. LRU page is in one of the LRU lists
     
110.  6. ACTIVE page is in the active LRU list
     
111. 18. UNEVICTABLE page is in the unevictable (non-)LRU list
     
112.                 It is somehow pinned and not a candidate for LRU page reclaims,
     
113.         eg. ramfs pages, shmctl(SHM_LOCK) and mlock() memory segments
     
114.  2. REFERENCED page has been referenced since last LRU list enqueue/requeue
     
115.  9. RECLAIM page will be reclaimed soon after its pageout IO completed
     
116. 11. MMAP a memory mapped page
     
117. 12. ANON a memory mapped page that is not part of a file
     
118. 13. SWAPCACHE page is mapped to swap space, ie. has an associated swap entry
     
119. 14. SWAPBACKED page is backed by swap/RAM
     
120. 
     
121. The page-types tool in this directory can be used to query the above flags.
     
122. 
     
123. Using pagemap to do something useful:
     
124. 
     
125. The general procedure for using pagemap to find out about a process' memory
     
126. usage goes like this:
     
127. 
     
128.  1. Read /proc/pid/maps to determine which parts of the memory space are
     
129.     mapped to what.
     
130.  2. Select the maps you are interested in -- all of them, or a particular
     
131.     library, or the stack or the heap, etc.
     
132.  3. Open /proc/pid/pagemap and seek to the pages you would like to examine.
     
133.  4. Read a u64 for each page from pagemap.
     
134.  5. Open /proc/kpagecount and/or /proc/kpageflags. For each PFN you just
     
135.     read, seek to that entry in the file, and read the data you want.
     
136. 
     
137. For example, to find the "unique set size" (USS), which is the amount of
     
138. memory that a process is using that is not shared with any other process,
     
139. you can go through every map in the process, find the PFNs, look those up
     
140. in kpagecount, and tally up the number of pages that are only referenced
     
141. once.
     
142. 
     
143. Other notes:
     
144. 
     
145. Reading from any of the files will return -EINVAL if you are not starting
     
146. the read on an 8-byte boundary (e.g., if you seeked an odd number of bytes
     
147. into the file), or if the size of the read is not a multiple of 8 bytes.