TEXT   50

packet_mmap.txt

Guest on 5th August 2021 01:48:17 AM

  1. --------------------------------------------------------------------------------
  2. + ABSTRACT
  3. --------------------------------------------------------------------------------
  4.  
  5. This file documents the CONFIG_PACKET_MMAP option available with the PACKET
  6. socket interface on 2.4 and 2.6 kernels. This type of sockets is used for
  7. capture network traffic with utilities like tcpdump or any other that needs
  8. raw access to network interface.
  9.  
  10. You can find the latest version of this document at:
  11.     http://pusa.uv.es/~ulisses/packet_mmap/
  12.  
  13. Howto can be found at:
  14.     http://wiki.gnu-log.net (packet_mmap)
  15.  
  16. Please send your comments to
  17.     Ulisses Alonso Camaró <uaca@i.hate.spam.alumni.uv.es>
  18.     Johann Baudy <johann.baudy@gnu-log.net>
  19.  
  20. -------------------------------------------------------------------------------
  21. + Why use PACKET_MMAP
  22. --------------------------------------------------------------------------------
  23.  
  24. In Linux 2.4/2.6 if PACKET_MMAP is not enabled, the capture process is very
  25. inefficient. It uses very limited buffers and requires one system call
  26. to capture each packet, it requires two if you want to get packet's
  27. timestamp (like libpcap always does).
  28.  
  29. In the other hand PACKET_MMAP is very efficient. PACKET_MMAP provides a size
  30. configurable circular buffer mapped in user space that can be used to either
  31. send or receive packets. This way reading packets just needs to wait for them,
  32. most of the time there is no need to issue a single system call. Concerning
  33. transmission, multiple packets can be sent through one system call to get the
  34. highest bandwidth.
  35. By using a shared buffer between the kernel and the user also has the benefit
  36. of minimizing packet copies.
  37.  
  38. It's fine to use PACKET_MMAP to improve the performance of the capture and
  39. transmission process, but it isn't everything. At least, if you are capturing
  40. at high speeds (this is relative to the cpu speed), you should check if the
  41. device driver of your network interface card supports some sort of interrupt
  42. load mitigation or (even better) if it supports NAPI, also make sure it is
  43. enabled. For transmission, check the MTU (Maximum Transmission Unit) used and
  44. supported by devices of your network.
  45.  
  46. --------------------------------------------------------------------------------
  47. + How to use CONFIG_PACKET_MMAP to improve capture process
  48. --------------------------------------------------------------------------------
  49.  
  50. From the user standpoint, you should use the higher level libpcap library, which
  51. is a de facto standard, portable across nearly all operating systems
  52. including Win32.
  53.  
  54. Said that, at time of this writing, official libpcap 0.8.1 is out and doesn't include
  55. support for PACKET_MMAP, and also probably the libpcap included in your distribution.
  56.  
  57. I'm aware of two implementations of PACKET_MMAP in libpcap:
  58.  
  59.     http://pusa.uv.es/~ulisses/packet_mmap/  (by Simon Patarin, based on libpcap 0.6.2)
  60.     http://public.lanl.gov/cpw/              (by Phil Wood, based on lastest libpcap)
  61.  
  62. The rest of this document is intended for people who want to understand
  63. the low level details or want to improve libpcap by including PACKET_MMAP
  64. support.
  65.  
  66. --------------------------------------------------------------------------------
  67. + How to use CONFIG_PACKET_MMAP directly to improve capture process
  68. --------------------------------------------------------------------------------
  69.  
  70. From the system calls stand point, the use of PACKET_MMAP involves
  71. the following process:
  72.  
  73.  
  74. [setup]     socket() -------> creation of the capture socket
  75.             setsockopt() ---> allocation of the circular buffer (ring)
  76.                               option: PACKET_RX_RING
  77.             mmap() ---------> mapping of the allocated buffer to the
  78.                               user process
  79.  
  80. [capture]   poll() ---------> to wait for incoming packets
  81.  
  82. [shutdown]  close() --------> destruction of the capture socket and
  83.                               deallocation of all associated
  84.                               resources.
  85.  
  86.  
  87. socket creation and destruction is straight forward, and is done
  88. the same way with or without PACKET_MMAP:
  89.  
  90. int fd;
  91.  
  92. fd= socket(PF_PACKET, mode, htons(ETH_P_ALL))
  93.  
  94. where mode is SOCK_RAW for the raw interface were link level
  95. information can be captured or SOCK_DGRAM for the cooked
  96. interface where link level information capture is not
  97. supported and a link level pseudo-header is provided
  98. by the kernel.
  99.  
  100. The destruction of the socket and all associated resources
  101. is done by a simple call to close(fd).
  102.  
  103. Next I will describe PACKET_MMAP settings and it's constraints,
  104. also the mapping of the circular buffer in the user process and
  105. the use of this buffer.
  106.  
  107. --------------------------------------------------------------------------------
  108. + How to use CONFIG_PACKET_MMAP directly to improve transmission process
  109. --------------------------------------------------------------------------------
  110. Transmission process is similar to capture as shown below.
  111.  
  112. [setup]          socket() -------> creation of the transmission socket
  113.                  setsockopt() ---> allocation of the circular buffer (ring)
  114.                                    option: PACKET_TX_RING
  115.                  bind() ---------> bind transmission socket with a network interface
  116.                  mmap() ---------> mapping of the allocated buffer to the
  117.                                    user process
  118.  
  119. [transmission]   poll() ---------> wait for free packets (optional)
  120.                  send() ---------> send all packets that are set as ready in
  121.                                    the ring
  122.                                    The flag MSG_DONTWAIT can be used to return
  123.                                    before end of transfer.
  124.  
  125. [shutdown]  close() --------> destruction of the transmission socket and
  126.                               deallocation of all associated resources.
  127.  
  128. Binding the socket to your network interface is mandatory (with zero copy) to
  129. know the header size of frames used in the circular buffer.
  130.  
  131. As capture, each frame contains two parts:
  132.  
  133.  --------------------
  134. | struct tpacket_hdr | Header. It contains the status of
  135. |                    | of this frame
  136. |--------------------|
  137. | data buffer        |
  138. .                    .  Data that will be sent over the network interface.
  139. .                    .
  140.  --------------------
  141.  
  142.  bind() associates the socket to your network interface thanks to
  143.  sll_ifindex parameter of struct sockaddr_ll.
  144.  
  145.  Initialization example:
  146.  
  147.  struct sockaddr_ll my_addr;
  148.  struct ifreq s_ifr;
  149.  ...
  150.  
  151.  strncpy (s_ifr.ifr_name, "eth0", sizeof(s_ifr.ifr_name));
  152.  
  153.  /* get interface index of eth0 */
  154.  ioctl(this->socket, SIOCGIFINDEX, &s_ifr);
  155.  
  156.  /* fill sockaddr_ll struct to prepare binding */
  157.  my_addr.sll_family = AF_PACKET;
  158.  my_addr.sll_protocol = ETH_P_ALL;
  159.  my_addr.sll_ifindex =  s_ifr.ifr_ifindex;
  160.  
  161.  /* bind socket to eth0 */
  162.  bind(this->socket, (struct sockaddr *)&my_addr, sizeof(struct sockaddr_ll));
  163.  
  164.  A complete tutorial is available at: http://wiki.gnu-log.net/
  165.  
  166. --------------------------------------------------------------------------------
  167. + PACKET_MMAP settings
  168. --------------------------------------------------------------------------------
  169.  
  170.  
  171. To setup PACKET_MMAP from user level code is done with a call like
  172.  
  173.  - Capture process
  174.      setsockopt(fd, SOL_PACKET, PACKET_RX_RING, (void *) &req, sizeof(req))
  175.  - Transmission process
  176.      setsockopt(fd, SOL_PACKET, PACKET_TX_RING, (void *) &req, sizeof(req))
  177.  
  178. The most significant argument in the previous call is the req parameter,
  179. this parameter must to have the following structure:
  180.  
  181.     struct tpacket_req
  182.     {
  183.         unsigned int    tp_block_size;  /* Minimal size of contiguous block */
  184.         unsigned int    tp_block_nr;    /* Number of blocks */
  185.         unsigned int    tp_frame_size;  /* Size of frame */
  186.         unsigned int    tp_frame_nr;    /* Total number of frames */
  187.     };
  188.  
  189. This structure is defined in /usr/include/linux/if_packet.h and establishes a
  190. circular buffer (ring) of unswappable memory.
  191. Being mapped in the capture process allows reading the captured frames and
  192. related meta-information like timestamps without requiring a system call.
  193.  
  194. Frames are grouped in blocks. Each block is a physically contiguous
  195. region of memory and holds tp_block_size/tp_frame_size frames. The total number
  196. of blocks is tp_block_nr. Note that tp_frame_nr is a redundant parameter because
  197.  
  198.     frames_per_block = tp_block_size/tp_frame_size
  199.  
  200. indeed, packet_set_ring checks that the following condition is true
  201.  
  202.     frames_per_block * tp_block_nr == tp_frame_nr
  203.  
  204.  
  205. Lets see an example, with the following values:
  206.  
  207.      tp_block_size= 4096
  208.      tp_frame_size= 2048
  209.      tp_block_nr  = 4
  210.      tp_frame_nr  = 8
  211.  
  212. we will get the following buffer structure:
  213.  
  214.         block #1                 block #2        
  215. +---------+---------+    +---------+---------+    
  216. | frame 1 | frame 2 |    | frame 3 | frame 4 |    
  217. +---------+---------+    +---------+---------+    
  218.  
  219.         block #3                 block #4
  220. +---------+---------+    +---------+---------+
  221. | frame 5 | frame 6 |    | frame 7 | frame 8 |
  222. +---------+---------+    +---------+---------+
  223.  
  224. A frame can be of any size with the only condition it can fit in a block. A block
  225. can only hold an integer number of frames, or in other words, a frame cannot
  226. be spawned accross two blocks, so there are some details you have to take into
  227. account when choosing the frame_size. See "Mapping and use of the circular
  228. buffer (ring)".
  229.  
  230.  
  231. --------------------------------------------------------------------------------
  232. + PACKET_MMAP setting constraints
  233. --------------------------------------------------------------------------------
  234.  
  235. In kernel versions prior to 2.4.26 (for the 2.4 branch) and 2.6.5 (2.6 branch),
  236. the PACKET_MMAP buffer could hold only 32768 frames in a 32 bit architecture or
  237. 16384 in a 64 bit architecture. For information on these kernel versions
  238. see http://pusa.uv.es/~ulisses/packet_mmap/packet_mmap.pre-2.4.26_2.6.5.txt
  239.  
  240.  Block size limit
  241. ------------------
  242.  
  243. As stated earlier, each block is a contiguous physical region of memory. These
  244. memory regions are allocated with calls to the __get_free_pages() function. As
  245. the name indicates, this function allocates pages of memory, and the second
  246. argument is "order" or a power of two number of pages, that is
  247. (for PAGE_SIZE == 4096) order=0 ==> 4096 bytes, order=1 ==> 8192 bytes,
  248. order=2 ==> 16384 bytes, etc. The maximum size of a
  249. region allocated by __get_free_pages is determined by the MAX_ORDER macro. More
  250. precisely the limit can be calculated as:
  251.  
  252.    PAGE_SIZE << MAX_ORDER
  253.  
  254.    In a i386 architecture PAGE_SIZE is 4096 bytes
  255.    In a 2.4/i386 kernel MAX_ORDER is 10
  256.    In a 2.6/i386 kernel MAX_ORDER is 11
  257.  
  258. So get_free_pages can allocate as much as 4MB or 8MB in a 2.4/2.6 kernel
  259. respectively, with an i386 architecture.
  260.  
  261. User space programs can include /usr/include/sys/user.h and
  262. /usr/include/linux/mmzone.h to get PAGE_SIZE MAX_ORDER declarations.
  263.  
  264. The pagesize can also be determined dynamically with the getpagesize (2)
  265. system call.
  266.  
  267.  
  268.  Block number limit
  269. --------------------
  270.  
  271. To understand the constraints of PACKET_MMAP, we have to see the structure
  272. used to hold the pointers to each block.
  273.  
  274. Currently, this structure is a dynamically allocated vector with kmalloc
  275. called pg_vec, its size limits the number of blocks that can be allocated.
  276.  
  277.     +---+---+---+---+
  278.     | x | x | x | x |
  279.     +---+---+---+---+
  280.       |   |   |   |
  281.       |   |   |   v
  282.       |   |   v  block #4
  283.       |   v  block #3
  284.       v  block #2
  285.      block #1
  286.  
  287.  
  288. kmalloc allocates any number of bytes of physically contiguous memory from
  289. a pool of pre-determined sizes. This pool of memory is maintained by the slab
  290. allocator which is at the end the responsible for doing the allocation and
  291. hence which imposes the maximum memory that kmalloc can allocate.
  292.  
  293. In a 2.4/2.6 kernel and the i386 architecture, the limit is 131072 bytes. The
  294. predetermined sizes that kmalloc uses can be checked in the "size-<bytes>"
  295. entries of /proc/slabinfo
  296.  
  297. In a 32 bit architecture, pointers are 4 bytes long, so the total number of
  298. pointers to blocks is
  299.  
  300.      131072/4 = 32768 blocks
  301.  
  302.  
  303.  PACKET_MMAP buffer size calculator
  304. ------------------------------------
  305.  
  306. Definitions:
  307.  
  308. <size-max>    : is the maximum size of allocable with kmalloc (see /proc/slabinfo)
  309. <pointer size>: depends on the architecture -- sizeof(void *)
  310. <page size>   : depends on the architecture -- PAGE_SIZE or getpagesize (2)
  311. <max-order>   : is the value defined with MAX_ORDER
  312. <frame size>  : it's an upper bound of frame's capture size (more on this later)
  313.  
  314. from these definitions we will derive
  315.  
  316.         <block number> = <size-max>/<pointer size>
  317.         <block size> = <pagesize> << <max-order>
  318.  
  319. so, the max buffer size is
  320.  
  321.         <block number> * <block size>
  322.  
  323. and, the number of frames be
  324.  
  325.         <block number> * <block size> / <frame size>
  326.  
  327. Suppose the following parameters, which apply for 2.6 kernel and an
  328. i386 architecture:
  329.  
  330.         <size-max> = 131072 bytes
  331.         <pointer size> = 4 bytes
  332.         <pagesize> = 4096 bytes
  333.         <max-order> = 11
  334.  
  335. and a value for <frame size> of 2048 bytes. These parameters will yield
  336.  
  337.         <block number> = 131072/4 = 32768 blocks
  338.         <block size> = 4096 << 11 = 8 MiB.
  339.  
  340. and hence the buffer will have a 262144 MiB size. So it can hold
  341. 262144 MiB / 2048 bytes = 134217728 frames
  342.  
  343.  
  344. Actually, this buffer size is not possible with an i386 architecture.
  345. Remember that the memory is allocated in kernel space, in the case of
  346. an i386 kernel's memory size is limited to 1GiB.
  347.  
  348. All memory allocations are not freed until the socket is closed. The memory
  349. allocations are done with GFP_KERNEL priority, this basically means that
  350. the allocation can wait and swap other process' memory in order to allocate
  351. the necessary memory, so normally limits can be reached.
  352.  
  353.  Other constraints
  354. -------------------
  355.  
  356. If you check the source code you will see that what I draw here as a frame
  357. is not only the link level frame. At the beginning of each frame there is a
  358. header called struct tpacket_hdr used in PACKET_MMAP to hold link level's frame
  359. meta information like timestamp. So what we draw here a frame it's really
  360. the following (from include/linux/if_packet.h):
  361.  
  362. /*
  363.    Frame structure:
  364.  
  365.    - Start. Frame must be aligned to TPACKET_ALIGNMENT=16
  366.    - struct tpacket_hdr
  367.    - pad to TPACKET_ALIGNMENT=16
  368.    - struct sockaddr_ll
  369.    - Gap, chosen so that packet data (Start+tp_net) aligns to
  370.      TPACKET_ALIGNMENT=16
  371.    - Start+tp_mac: [ Optional MAC header ]
  372.    - Start+tp_net: Packet data, aligned to TPACKET_ALIGNMENT=16.
  373.    - Pad to align to TPACKET_ALIGNMENT=16
  374.  */
  375.            
  376.  
  377.  The following are conditions that are checked in packet_set_ring
  378.  
  379.    tp_block_size must be a multiple of PAGE_SIZE (1)
  380.    tp_frame_size must be greater than TPACKET_HDRLEN (obvious)
  381.    tp_frame_size must be a multiple of TPACKET_ALIGNMENT
  382.    tp_frame_nr   must be exactly frames_per_block*tp_block_nr
  383.  
  384. Note that tp_block_size should be chosen to be a power of two or there will
  385. be a waste of memory.
  386.  
  387. --------------------------------------------------------------------------------
  388. + Mapping and use of the circular buffer (ring)
  389. --------------------------------------------------------------------------------
  390.  
  391. The mapping of the buffer in the user process is done with the conventional
  392. mmap function. Even the circular buffer is compound of several physically
  393. discontiguous blocks of memory, they are contiguous to the user space, hence
  394. just one call to mmap is needed:
  395.  
  396.     mmap(0, size, PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0);
  397.  
  398. If tp_frame_size is a divisor of tp_block_size frames will be
  399. contiguously spaced by tp_frame_size bytes. If not, each
  400. tp_block_size/tp_frame_size frames there will be a gap between
  401. the frames. This is because a frame cannot be spawn across two
  402. blocks.
  403.  
  404. At the beginning of each frame there is an status field (see
  405. struct tpacket_hdr). If this field is 0 means that the frame is ready
  406. to be used for the kernel, If not, there is a frame the user can read
  407. and the following flags apply:
  408.  
  409. +++ Capture process:
  410.      from include/linux/if_packet.h
  411.  
  412.      #define TP_STATUS_COPY          2
  413.      #define TP_STATUS_LOSING        4
  414.      #define TP_STATUS_CSUMNOTREADY  8
  415.  
  416.  
  417. TP_STATUS_COPY        : This flag indicates that the frame (and associated
  418.                         meta information) has been truncated because it's
  419.                         larger than tp_frame_size. This packet can be
  420.                         read entirely with recvfrom().
  421.                        
  422.                         In order to make this work it must to be
  423.                         enabled previously with setsockopt() and
  424.                         the PACKET_COPY_THRESH option.
  425.  
  426.                         The number of frames than can be buffered to
  427.                         be read with recvfrom is limited like a normal socket.
  428.                         See the SO_RCVBUF option in the socket (7) man page.
  429.  
  430. TP_STATUS_LOSING      : indicates there were packet drops from last time
  431.                         statistics where checked with getsockopt() and
  432.                         the PACKET_STATISTICS option.
  433.  
  434. TP_STATUS_CSUMNOTREADY: currently it's used for outgoing IP packets which
  435.                         it's checksum will be done in hardware. So while
  436.                         reading the packet we should not try to check the
  437.                         checksum.
  438.  
  439. for convenience there are also the following defines:
  440.  
  441.      #define TP_STATUS_KERNEL        0
  442.      #define TP_STATUS_USER          1
  443.  
  444. The kernel initializes all frames to TP_STATUS_KERNEL, when the kernel
  445. receives a packet it puts in the buffer and updates the status with
  446. at least the TP_STATUS_USER flag. Then the user can read the packet,
  447. once the packet is read the user must zero the status field, so the kernel
  448. can use again that frame buffer.
  449.  
  450. The user can use poll (any other variant should apply too) to check if new
  451. packets are in the ring:
  452.  
  453.     struct pollfd pfd;
  454.  
  455.     pfd.fd = fd;
  456.     pfd.revents = 0;
  457.     pfd.events = POLLIN|POLLRDNORM|POLLERR;
  458.  
  459.     if (status == TP_STATUS_KERNEL)
  460.         retval = poll(&pfd, 1, timeout);
  461.  
  462. It doesn't incur in a race condition to first check the status value and
  463. then poll for frames.
  464.  
  465.  
  466. ++ Transmission process
  467. Those defines are also used for transmission:
  468.  
  469.      #define TP_STATUS_AVAILABLE        0 // Frame is available
  470.      #define TP_STATUS_SEND_REQUEST     1 // Frame will be sent on next send()
  471.      #define TP_STATUS_SENDING          2 // Frame is currently in transmission
  472.      #define TP_STATUS_WRONG_FORMAT     4 // Frame format is not correct
  473.  
  474. First, the kernel initializes all frames to TP_STATUS_AVAILABLE. To send a
  475. packet, the user fills a data buffer of an available frame, sets tp_len to
  476. current data buffer size and sets its status field to TP_STATUS_SEND_REQUEST.
  477. This can be done on multiple frames. Once the user is ready to transmit, it
  478. calls send(). Then all buffers with status equal to TP_STATUS_SEND_REQUEST are
  479. forwarded to the network device. The kernel updates each status of sent
  480. frames with TP_STATUS_SENDING until the end of transfer.
  481. At the end of each transfer, buffer status returns to TP_STATUS_AVAILABLE.
  482.  
  483.     header->tp_len = in_i_size;
  484.     header->tp_status = TP_STATUS_SEND_REQUEST;
  485.     retval = send(this->socket, NULL, 0, 0);
  486.  
  487. The user can also use poll() to check if a buffer is available:
  488. (status == TP_STATUS_SENDING)
  489.  
  490.     struct pollfd pfd;
  491.     pfd.fd = fd;
  492.     pfd.revents = 0;
  493.     pfd.events = POLLOUT;
  494.     retval = poll(&pfd, 1, timeout);
  495.  
  496. --------------------------------------------------------------------------------
  497. + THANKS
  498. --------------------------------------------------------------------------------
  499.    
  500.    Jesse Brandeburg, for fixing my grammathical/spelling erro

Raw Paste


Login or Register to edit or fork this paste. It's free.