Documentation/networking/netlink_mmap.txt

   1 This file documents how to use memory mapped I/O with netlink.
   2
   3 Author: Patrick McHardy <kaber@trash.net>
   4
   5 Overview
   6 --------
   7
   8 Memory mapped netlink I/O can be used to increase throughput and decrease
   9 overhead of unicast receive and transmit operations. Some netlink subsystems
  10 require high throughput, these are mainly the netfilter subsystems
  11 nfnetlink_queue and nfnetlink_log, but it can also help speed up large
  12 dump operations of f.i. the routing database.
  13
  14 Memory mapped netlink I/O used two circular ring buffers for RX and TX which
  15 are mapped into the processes address space.
  16
  17 The RX ring is used by the kernel to directly construct netlink messages into
  18 user-space memory without copying them as done with regular socket I/O,
  19 additionally as long as the ring contains messages no recvmsg() or poll()
  20 syscalls have to be issued by user-space to get more message.
  21
  22 The TX ring is used to process messages directly from user-space memory, the
  23 kernel processes all messages contained in the ring using a single sendmsg()
  24 call.
  25
  26 Usage overview
  27 --------------
  28
  29 In order to use memory mapped netlink I/O, user-space needs three main changes:
  30
  31 - ring setup
  32 - conversion of the RX path to get messages from the ring instead of recvmsg()
  33 - conversion of the TX path to construct messages into the ring
  34
  35 Ring setup is done using setsockopt() to provide the ring parameters to the
  36 kernel, then a call to mmap() to map the ring into the processes address space:
  37
  38 - setsockopt(fd, SOL_NETLINK, NETLINK_RX_RING, &params, sizeof(params));
  39 - setsockopt(fd, SOL_NETLINK, NETLINK_TX_RING, &params, sizeof(params));
  40 - ring = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED, fd, 0)
  41
  42 Usage of either ring is optional, but even if only the RX ring is used the
  43 mapping still needs to be writable in order to update the frame status after
  44 processing.
  45
  46 Conversion of the reception path involves calling poll() on the file
  47 descriptor, once the socket is readable the frames from the ring are
  48 processsed in order until no more messages are available, as indicated by
  49 a status word in the frame header.
  50
  51 On kernel side, in order to make use of memory mapped I/O on receive, the
  52 originating netlink subsystem needs to support memory mapped I/O, otherwise
  53 it will use an allocated socket buffer as usual and the contents will be
  54  copied to the ring on transmission, nullifying most of the performance gains.
  55 Dumps of kernel databases automatically support memory mapped I/O.
  56
  57 Conversion of the transmit path involves changing message contruction to
  58 use memory from the TX ring instead of (usually) a buffer declared on the
  59 stack and setting up the frame header approriately. Optionally poll() can
  60 be used to wait for free frames in the TX ring.
  61
  62 Structured and definitions for using memory mapped I/O are contained in
  63 <linux/netlink.h>.
  64
  65 RX and TX rings
  66 ----------------
  67
  68 Each ring contains a number of continous memory blocks, containing frames of
  69 fixed size dependant on the parameters used for ring setup.
  70
  71 Ring:   [ block 0 ]
  72                 [ frame 0 ]
  73                 [ frame 1 ]
  74         [ block 1 ]
  75                 [ frame 2 ]
  76                 [ frame 3 ]
  77         ...
  78         [ block n ]
  79                 [ frame 2 * n ]
  80                 [ frame 2 * n + 1 ]
  81
  82 The blocks are only visible to the kernel, from the point of view of user-space
  83 the ring just contains the frames in a continous memory zone.
  84
  85 The ring parameters used for setting up the ring are defined as follows:
  86
  87 struct nl_mmap_req {
  88         unsigned int    nm_block_size;
  89         unsigned int    nm_block_nr;
  90         unsigned int    nm_frame_size;
  91         unsigned int    nm_frame_nr;
  92 };
  93
  94 Frames are grouped into blocks, where each block is a continous region of memory
  95 and holds nm_block_size / nm_frame_size frames. The total number of frames in
  96 the ring is nm_frame_nr. The following invariants hold:
  97
  98 - frames_per_block = nm_block_size / nm_frame_size
  99
 100 - nm_frame_nr = frames_per_block * nm_block_nr
 101
 102 Some parameters are constrained, specifically:
 103
 104 - nm_block_size must be a multiple of the architectures memory page size.
 105   The getpagesize() function can be used to get the page size.
 106
 107 - nm_frame_size must be equal or larger to NL_MMAP_HDRLEN, IOW a frame must be
 108   able to hold at least the frame header
 109
 110 - nm_frame_size must be smaller or equal to nm_block_size
 111
 112 - nm_frame_size must be a multiple of NL_MMAP_MSG_ALIGNMENT
 113
 114 - nm_frame_nr must equal the actual number of frames as specified above.
 115
 116 When the kernel can't allocate phsyically continous memory for a ring block,
 117 it will fall back to use physically discontinous memory. This might affect
 118 performance negatively, in order to avoid this the nm_frame_size parameter
 119 should be chosen to be as small as possible for the required frame size and
 120 the number of blocks should be increased instead.
 121
 122 Ring frames
 123 ------------
 124
 125 Each frames contain a frame header, consisting of a synchronization word and some
 126 meta-data, and the message itself.
 127
 128 Frame:  [ header message ]
 129
 130 The frame header is defined as follows:
 131
 132 struct nl_mmap_hdr {
 133         unsigned int    nm_status;
 134         unsigned int    nm_len;
 135         __u32           nm_group;
 136         /* credentials */
 137         __u32           nm_pid;
 138         __u32           nm_uid;
 139         __u32           nm_gid;
 140 };
 141
 142 - nm_status is used for synchronizing processing between the kernel and user-
 143   space and specifies ownership of the frame as well as the operation to perform
 144
 145 - nm_len contains the length of the message contained in the data area
 146
 147 - nm_group specified the destination multicast group of message
 148
 149 - nm_pid, nm_uid and nm_gid contain the netlink pid, UID and GID of the sending
 150   process. These values correspond to the data available using SOCK_PASSCRED in
 151   the SCM_CREDENTIALS cmsg.
 152
 153 The possible values in the status word are:
 154
 155 - NL_MMAP_STATUS_UNUSED:
 156         RX ring:        frame belongs to the kernel and contains no message
 157                         for user-space. Approriate action is to invoke poll()
 158                         to wait for new messages.
 159
 160         TX ring:        frame belongs to user-space and can be used for
 161                         message construction.
 162
 163 - NL_MMAP_STATUS_RESERVED:
 164         RX ring only:   frame is currently used by the kernel for message
 165                         construction and contains no valid message yet.
 166                         Appropriate action is to invoke poll() to wait for
 167                         new messages.
 168
 169 - NL_MMAP_STATUS_VALID:
 170         RX ring:        frame contains a valid message. Approriate action is
 171                         to process the message and release the frame back to
 172                         the kernel by setting the status to
 173                         NL_MMAP_STATUS_UNUSED or queue the frame by setting the
 174                         status to NL_MMAP_STATUS_SKIP.
 175
 176         TX ring:        the frame contains a valid message from user-space to
 177                         be processed by the kernel. After completing processing
 178                         the kernel will release the frame back to user-space by
 179                         setting the status to NL_MMAP_STATUS_UNUSED.
 180
 181 - NL_MMAP_STATUS_COPY:
 182         RX ring only:   a message is ready to be processed but could not be
 183                         stored in the ring, either because it exceeded the
 184                         frame size or because the originating subsystem does
 185                         not support memory mapped I/O. Appropriate action is
 186                         to invoke recvmsg() to receive the message and release
 187                         the frame back to the kernel by setting the status to
 188                         NL_MMAP_STATUS_UNUSED.
 189
 190 - NL_MMAP_STATUS_SKIP:
 191         RX ring only:   user-space queued the message for later processing, but
 192                         processed some messages following it in the ring. The
 193                         kernel should skip this frame when looking for unused
 194                         frames.
 195
 196 The data area of a frame begins at a offset of NL_MMAP_HDRLEN relative to the
 197 frame header.
 198
 199 TX limitations
 200 --------------
 201
 202 Kernel processing usually involves validation of the message received by
 203 user-space, then processing its contents. The kernel must assure that
 204 userspace is not able to modify the message contents after they have been
 205 validated. In order to do so, the message is copied from the ring frame
 206 to an allocated buffer if either of these conditions is false:
 207
 208 - only a single mapping of the ring exists
 209 - the file descriptor is not shared between processes
 210
 211 This means that for threaded programs, the kernel will fall back to copying.
 212
 213 Example
 214 -------
 215
 216 Ring setup:
 217
 218         unsigned int block_size = 16 * getpagesize();
 219         struct nl_mmap_req req = {
 220                 .nm_block_size          = block_size,
 221                 .nm_block_nr            = 64,
 222                 .nm_frame_size          = 16384,
 223                 .nm_frame_nr            = 64 * block_size / 16384,
 224         };
 225         unsigned int ring_size;
 226         void *rx_ring, *tx_ring;
 227
 228         /* Configure ring parameters */
 229         if (setsockopt(fd, NETLINK_RX_RING, &req, sizeof(req)) < 0)
 230                 exit(1);
 231         if (setsockopt(fd, NETLINK_TX_RING, &req, sizeof(req)) < 0)
 232                 exit(1)
 233
 234         /* Calculate size of each invididual ring */
 235         ring_size = req.nm_block_nr * req.nm_block_size;
 236
 237         /* Map RX/TX rings. The TX ring is located after the RX ring */
 238         rx_ring = mmap(NULL, 2 * ring_size, PROT_READ | PROT_WRITE,
 239                        MAP_SHARED, fd, 0);
 240         if ((long)rx_ring == -1L)
 241                 exit(1);
 242         tx_ring = rx_ring + ring_size:
 243
 244 Message reception:
 245
 246 This example assumes some ring parameters of the ring setup are available.
 247
 248         unsigned int frame_offset = 0;
 249         struct nl_mmap_hdr *hdr;
 250         struct nlmsghdr *nlh;
 251         unsigned char buf[16384];
 252         ssize_t len;
 253
 254         while (1) {
 255                 struct pollfd pfds[1];
 256
 257                 pfds[0].fd      = fd;
 258                 pfds[0].events  = POLLIN | POLLERR;
 259                 pfds[0].revents = 0;
 260
 261                 if (poll(pfds, 1, -1) < 0 && errno != -EINTR)
 262                         exit(1);
 263
 264                 /* Check for errors. Error handling omitted */
 265                 if (pfds[0].revents & POLLERR)
 266                         <handle error>
 267
 268                 /* If no new messages, poll again */
 269                 if (!(pfds[0].revents & POLLIN))
 270                         continue;
 271
 272                 /* Process all frames */
 273                 while (1) {
 274                         /* Get next frame header */
 275                         hdr = rx_ring + frame_offset;
 276
 277                         if (hdr->nm_status == NL_MMAP_STATUS_VALID)
 278                                 /* Regular memory mapped frame */
 279                                 nlh = (void *hdr) + NL_MMAP_HDRLEN;
 280                                 len = hdr->nm_len;
 281
 282                                 /* Release empty message immediately. May happen
 283                                  * on error during message construction.
 284                                  */
 285                                 if (len == 0)
 286                                         goto release;
 287                         } else if (hdr->nm_status == NL_MMAP_STATUS_COPY) {
 288                                 /* Frame queued to socket receive queue */
 289                                 len = recv(fd, buf, sizeof(buf), MSG_DONTWAIT);
 290                                 if (len <= 0)
 291                                         break;
 292                                 nlh = buf;
 293                         } else
 294                                 /* No more messages to process, continue polling */
 295                                 break;
 296
 297                         process_msg(nlh);
 298 release:
 299                         /* Release frame back to the kernel */
 300                         hdr->nm_status = NL_MMAP_STATUS_UNUSED;
 301
 302                         /* Advance frame offset to next frame */
 303                         frame_offset = (frame_offset + frame_size) % ring_size;
 304                 }
 305         }
 306
 307 Message transmission:
 308
 309 This example assumes some ring parameters of the ring setup are available.
 310 A single message is constructed and transmitted, to send multiple messages
 311 at once they would be constructed in consecutive frames before a final call
 312 to sendto().
 313
 314         unsigned int frame_offset = 0;
 315         struct nl_mmap_hdr *hdr;
 316         struct nlmsghdr *nlh;
 317         struct sockaddr_nl addr = {
 318                 .nl_family      = AF_NETLINK,
 319         };
 320
 321         hdr = tx_ring + frame_offset;
 322         if (hdr->nm_status != NL_MMAP_STATUS_UNUSED)
 323                 /* No frame available. Use poll() to avoid. */
 324                 exit(1);
 325
 326         nlh = (void *)hdr + NL_MMAP_HDRLEN;
 327
 328         /* Build message */
 329         build_message(nlh);
 330
 331         /* Fill frame header: length and status need to be set */
 332         hdr->nm_len     = nlh->nlmsg_len;
 333         hdr->nm_status  = NL_MMAP_STATUS_VALID;
 334
 335         if (sendto(fd, NULL, 0, 0, &addr, sizeof(addr)) < 0)
 336                 exit(1);
 337
 338         /* Advance frame offset to next frame */
 339         frame_offset = (frame_offset + frame_size) % ring_size;