packet_mmap.txt 36.3 KB
Newer Older
Linus Torvalds's avatar
Linus Torvalds committed
1 2 3 4
--------------------------------------------------------------------------------
+ ABSTRACT
--------------------------------------------------------------------------------

5
This file documents the mmap() facility available with the PACKET
6 7 8
socket interface on 2.4/2.6/3.x kernels. This type of sockets is used for
i) capture network traffic with utilities like tcpdump, ii) transmit network
traffic, or any other that needs raw access to network interface.
Linus Torvalds's avatar
Linus Torvalds committed
9

10
You can find the latest version of this document at:
11
    http://wiki.ipxwarzone.com/index.php5?title=Linux_packet_mmap
Linus Torvalds's avatar
Linus Torvalds committed
12

13 14
Howto can be found at:
    http://wiki.gnu-log.net (packet_mmap)
Linus Torvalds's avatar
Linus Torvalds committed
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>
Linus Torvalds's avatar
Linus Torvalds committed
19 20 21 22 23

-------------------------------------------------------------------------------
+ Why use PACKET_MMAP
--------------------------------------------------------------------------------

24 25 26 27
In Linux 2.4/2.6/3.x if PACKET_MMAP is not enabled, the capture process is very
inefficient. It uses very limited buffers and requires one system call to
capture each packet, it requires two if you want to get packet's timestamp
(like libpcap always does).
Linus Torvalds's avatar
Linus Torvalds committed
28 29

In the other hand PACKET_MMAP is very efficient. PACKET_MMAP provides a size 
30 31 32 33
configurable circular buffer mapped in user space that can be used to either
send or receive packets. This way reading packets just needs to wait for them,
most of the time there is no need to issue a single system call. Concerning
transmission, multiple packets can be sent through one system call to get the
34 35
highest bandwidth. By using a shared buffer between the kernel and the user
also has the benefit of minimizing packet copies.
36 37 38 39 40 41 42

It's fine to use PACKET_MMAP to improve the performance of the capture and
transmission process, but it isn't everything. At least, if you are capturing
at high speeds (this is relative to the cpu speed), you should check if the
device driver of your network interface card supports some sort of interrupt
load mitigation or (even better) if it supports NAPI, also make sure it is
enabled. For transmission, check the MTU (Maximum Transmission Unit) used and
43 44
supported by devices of your network. CPU IRQ pinning of your network interface
card can also be an advantage.
Linus Torvalds's avatar
Linus Torvalds committed
45 46

--------------------------------------------------------------------------------
47
+ How to use mmap() to improve capture process
Linus Torvalds's avatar
Linus Torvalds committed
48 49
--------------------------------------------------------------------------------

50
From the user standpoint, you should use the higher level libpcap library, which
Linus Torvalds's avatar
Linus Torvalds committed
51 52 53 54 55 56 57 58
is a de facto standard, portable across nearly all operating systems
including Win32. 

Said that, at time of this writing, official libpcap 0.8.1 is out and doesn't include
support for PACKET_MMAP, and also probably the libpcap included in your distribution. 

I'm aware of two implementations of PACKET_MMAP in libpcap:

59
    http://wiki.ipxwarzone.com/		     (by Simon Patarin, based on libpcap 0.6.2)
Linus Torvalds's avatar
Linus Torvalds committed
60 61 62 63 64 65 66
    http://public.lanl.gov/cpw/              (by Phil Wood, based on lastest libpcap)

The rest of this document is intended for people who want to understand
the low level details or want to improve libpcap by including PACKET_MMAP
support.

--------------------------------------------------------------------------------
67
+ How to use mmap() directly to improve capture process
Linus Torvalds's avatar
Linus Torvalds committed
68 69 70 71 72 73 74 75
--------------------------------------------------------------------------------

From the system calls stand point, the use of PACKET_MMAP involves
the following process:


[setup]     socket() -------> creation of the capture socket
            setsockopt() ---> allocation of the circular buffer (ring)
76
                              option: PACKET_RX_RING
77
            mmap() ---------> mapping of the allocated buffer to the
Linus Torvalds's avatar
Linus Torvalds committed
78 79 80 81 82 83 84 85 86 87 88 89
                              user process

[capture]   poll() ---------> to wait for incoming packets

[shutdown]  close() --------> destruction of the capture socket and
                              deallocation of all associated 
                              resources.


socket creation and destruction is straight forward, and is done 
the same way with or without PACKET_MMAP:

90
 int fd = socket(PF_PACKET, mode, htons(ETH_P_ALL));
Linus Torvalds's avatar
Linus Torvalds committed
91 92 93 94 95 96 97 98 99 100

where mode is SOCK_RAW for the raw interface were link level
information can be captured or SOCK_DGRAM for the cooked
interface where link level information capture is not 
supported and a link level pseudo-header is provided 
by the kernel.

The destruction of the socket and all associated resources
is done by a simple call to close(fd).

101 102 103 104 105
Similarly as without PACKET_MMAP, it is possible to use one socket
for capture and transmission. This can be done by mapping the
allocated RX and TX buffer ring with a single mmap() call.
See "Mapping and use of the circular buffer (ring)".

106
Next I will describe PACKET_MMAP settings and its constraints,
107
also the mapping of the circular buffer in the user process and 
Linus Torvalds's avatar
Linus Torvalds committed
108 109
the use of this buffer.

110
--------------------------------------------------------------------------------
111
+ How to use mmap() directly to improve transmission process
112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
--------------------------------------------------------------------------------
Transmission process is similar to capture as shown below.

[setup]          socket() -------> creation of the transmission socket
                 setsockopt() ---> allocation of the circular buffer (ring)
                                   option: PACKET_TX_RING
                 bind() ---------> bind transmission socket with a network interface
                 mmap() ---------> mapping of the allocated buffer to the
                                   user process

[transmission]   poll() ---------> wait for free packets (optional)
                 send() ---------> send all packets that are set as ready in
                                   the ring
                                   The flag MSG_DONTWAIT can be used to return
                                   before end of transfer.

[shutdown]  close() --------> destruction of the transmission socket and
                              deallocation of all associated resources.

131 132 133 134 135 136 137 138 139 140
Socket creation and destruction is also straight forward, and is done
the same way as in capturing described in the previous paragraph:

 int fd = socket(PF_PACKET, mode, 0);

The protocol can optionally be 0 in case we only want to transmit
via this socket, which avoids an expensive call to packet_rcv().
In this case, you also need to bind(2) the TX_RING with sll_protocol = 0
set. Otherwise, htons(ETH_P_ALL) or any other protocol, for example.

141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170
Binding the socket to your network interface is mandatory (with zero copy) to
know the header size of frames used in the circular buffer.

As capture, each frame contains two parts:

 --------------------
| struct tpacket_hdr | Header. It contains the status of
|                    | of this frame
|--------------------|
| data buffer        |
.                    .  Data that will be sent over the network interface.
.                    .
 --------------------

 bind() associates the socket to your network interface thanks to
 sll_ifindex parameter of struct sockaddr_ll.

 Initialization example:

 struct sockaddr_ll my_addr;
 struct ifreq s_ifr;
 ...

 strncpy (s_ifr.ifr_name, "eth0", sizeof(s_ifr.ifr_name));

 /* get interface index of eth0 */
 ioctl(this->socket, SIOCGIFINDEX, &s_ifr);

 /* fill sockaddr_ll struct to prepare binding */
 my_addr.sll_family = AF_PACKET;
171
 my_addr.sll_protocol = htons(ETH_P_ALL);
172 173 174 175 176 177 178
 my_addr.sll_ifindex =  s_ifr.ifr_ifindex;

 /* bind socket to eth0 */
 bind(this->socket, (struct sockaddr *)&my_addr, sizeof(struct sockaddr_ll));

 A complete tutorial is available at: http://wiki.gnu-log.net/

179 180 181 182 183 184 185 186 187 188 189 190 191
By default, the user should put data at :
 frame base + TPACKET_HDRLEN - sizeof(struct sockaddr_ll)

So, whatever you choose for the socket mode (SOCK_DGRAM or SOCK_RAW),
the beginning of the user data will be at :
 frame base + TPACKET_ALIGN(sizeof(struct tpacket_hdr))

If you wish to put user data at a custom offset from the beginning of
the frame (for payload alignment with SOCK_RAW mode for instance) you
can set tp_net (with SOCK_DGRAM) or tp_mac (with SOCK_RAW). In order
to make this work it must be enabled previously with setsockopt()
and the PACKET_TX_HAS_OFF option.

Linus Torvalds's avatar
Linus Torvalds committed
192 193 194 195 196 197
--------------------------------------------------------------------------------
+ PACKET_MMAP settings
--------------------------------------------------------------------------------

To setup PACKET_MMAP from user level code is done with a call like

198
 - Capture process
Linus Torvalds's avatar
Linus Torvalds committed
199
     setsockopt(fd, SOL_PACKET, PACKET_RX_RING, (void *) &req, sizeof(req))
200 201
 - Transmission process
     setsockopt(fd, SOL_PACKET, PACKET_TX_RING, (void *) &req, sizeof(req))
Linus Torvalds's avatar
Linus Torvalds committed
202 203 204 205 206 207 208 209 210 211 212 213 214

The most significant argument in the previous call is the req parameter, 
this parameter must to have the following structure:

    struct tpacket_req
    {
        unsigned int    tp_block_size;  /* Minimal size of contiguous block */
        unsigned int    tp_block_nr;    /* Number of blocks */
        unsigned int    tp_frame_size;  /* Size of frame */
        unsigned int    tp_frame_nr;    /* Total number of frames */
    };

This structure is defined in /usr/include/linux/if_packet.h and establishes a 
215
circular buffer (ring) of unswappable memory.
Linus Torvalds's avatar
Linus Torvalds committed
216 217 218
Being mapped in the capture process allows reading the captured frames and 
related meta-information like timestamps without requiring a system call.

219
Frames are grouped in blocks. Each block is a physically contiguous
Linus Torvalds's avatar
Linus Torvalds committed
220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249
region of memory and holds tp_block_size/tp_frame_size frames. The total number 
of blocks is tp_block_nr. Note that tp_frame_nr is a redundant parameter because

    frames_per_block = tp_block_size/tp_frame_size

indeed, packet_set_ring checks that the following condition is true

    frames_per_block * tp_block_nr == tp_frame_nr

Lets see an example, with the following values:

     tp_block_size= 4096
     tp_frame_size= 2048
     tp_block_nr  = 4
     tp_frame_nr  = 8

we will get the following buffer structure:

        block #1                 block #2         
+---------+---------+    +---------+---------+    
| frame 1 | frame 2 |    | frame 3 | frame 4 |    
+---------+---------+    +---------+---------+    

        block #3                 block #4
+---------+---------+    +---------+---------+
| frame 5 | frame 6 |    | frame 7 | frame 8 |
+---------+---------+    +---------+---------+

A frame can be of any size with the only condition it can fit in a block. A block
can only hold an integer number of frames, or in other words, a frame cannot 
250
be spawned across two blocks, so there are some details you have to take into 
251
account when choosing the frame_size. See "Mapping and use of the circular 
Linus Torvalds's avatar
Linus Torvalds committed
252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308
buffer (ring)".

--------------------------------------------------------------------------------
+ PACKET_MMAP setting constraints
--------------------------------------------------------------------------------

In kernel versions prior to 2.4.26 (for the 2.4 branch) and 2.6.5 (2.6 branch),
the PACKET_MMAP buffer could hold only 32768 frames in a 32 bit architecture or
16384 in a 64 bit architecture. For information on these kernel versions
see http://pusa.uv.es/~ulisses/packet_mmap/packet_mmap.pre-2.4.26_2.6.5.txt

 Block size limit
------------------

As stated earlier, each block is a contiguous physical region of memory. These 
memory regions are allocated with calls to the __get_free_pages() function. As 
the name indicates, this function allocates pages of memory, and the second
argument is "order" or a power of two number of pages, that is 
(for PAGE_SIZE == 4096) order=0 ==> 4096 bytes, order=1 ==> 8192 bytes, 
order=2 ==> 16384 bytes, etc. The maximum size of a 
region allocated by __get_free_pages is determined by the MAX_ORDER macro. More 
precisely the limit can be calculated as:

   PAGE_SIZE << MAX_ORDER

   In a i386 architecture PAGE_SIZE is 4096 bytes 
   In a 2.4/i386 kernel MAX_ORDER is 10
   In a 2.6/i386 kernel MAX_ORDER is 11

So get_free_pages can allocate as much as 4MB or 8MB in a 2.4/2.6 kernel 
respectively, with an i386 architecture.

User space programs can include /usr/include/sys/user.h and 
/usr/include/linux/mmzone.h to get PAGE_SIZE MAX_ORDER declarations.

The pagesize can also be determined dynamically with the getpagesize (2) 
system call. 

 Block number limit
--------------------

To understand the constraints of PACKET_MMAP, we have to see the structure 
used to hold the pointers to each block.

Currently, this structure is a dynamically allocated vector with kmalloc 
called pg_vec, its size limits the number of blocks that can be allocated.

    +---+---+---+---+
    | x | x | x | x |
    +---+---+---+---+
      |   |   |   |
      |   |   |   v
      |   |   v  block #4
      |   v  block #3
      v  block #2
     block #1

309 310
kmalloc allocates any number of bytes of physically contiguous memory from 
a pool of pre-determined sizes. This pool of memory is maintained by the slab 
311 312
allocator which is at the end the responsible for doing the allocation and 
hence which imposes the maximum memory that kmalloc can allocate. 
Linus Torvalds's avatar
Linus Torvalds committed
313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346

In a 2.4/2.6 kernel and the i386 architecture, the limit is 131072 bytes. The 
predetermined sizes that kmalloc uses can be checked in the "size-<bytes>" 
entries of /proc/slabinfo

In a 32 bit architecture, pointers are 4 bytes long, so the total number of 
pointers to blocks is

     131072/4 = 32768 blocks

 PACKET_MMAP buffer size calculator
------------------------------------

Definitions:

<size-max>    : is the maximum size of allocable with kmalloc (see /proc/slabinfo)
<pointer size>: depends on the architecture -- sizeof(void *)
<page size>   : depends on the architecture -- PAGE_SIZE or getpagesize (2)
<max-order>   : is the value defined with MAX_ORDER
<frame size>  : it's an upper bound of frame's capture size (more on this later)

from these definitions we will derive 

	<block number> = <size-max>/<pointer size>
	<block size> = <pagesize> << <max-order>

so, the max buffer size is

	<block number> * <block size>

and, the number of frames be

	<block number> * <block size> / <frame size>

347
Suppose the following parameters, which apply for 2.6 kernel and an
Linus Torvalds's avatar
Linus Torvalds committed
348 349 350 351 352 353 354
i386 architecture:

	<size-max> = 131072 bytes
	<pointer size> = 4 bytes
	<pagesize> = 4096 bytes
	<max-order> = 11

355
and a value for <frame size> of 2048 bytes. These parameters will yield
Linus Torvalds's avatar
Linus Torvalds committed
356 357 358 359 360 361 362 363 364 365 366 367 368 369

	<block number> = 131072/4 = 32768 blocks
	<block size> = 4096 << 11 = 8 MiB.

and hence the buffer will have a 262144 MiB size. So it can hold 
262144 MiB / 2048 bytes = 134217728 frames

Actually, this buffer size is not possible with an i386 architecture. 
Remember that the memory is allocated in kernel space, in the case of 
an i386 kernel's memory size is limited to 1GiB.

All memory allocations are not freed until the socket is closed. The memory 
allocations are done with GFP_KERNEL priority, this basically means that 
the allocation can wait and swap other process' memory in order to allocate 
370
the necessary memory, so normally limits can be reached.
Linus Torvalds's avatar
Linus Torvalds committed
371 372 373 374 375

 Other constraints
-------------------

If you check the source code you will see that what I draw here as a frame
376
is not only the link level frame. At the beginning of each frame there is a 
Linus Torvalds's avatar
Linus Torvalds committed
377 378 379 380 381 382 383 384 385 386 387
header called struct tpacket_hdr used in PACKET_MMAP to hold link level's frame
meta information like timestamp. So what we draw here a frame it's really 
the following (from include/linux/if_packet.h):

/*
   Frame structure:

   - Start. Frame must be aligned to TPACKET_ALIGNMENT=16
   - struct tpacket_hdr
   - pad to TPACKET_ALIGNMENT=16
   - struct sockaddr_ll
388
   - Gap, chosen so that packet data (Start+tp_net) aligns to 
Linus Torvalds's avatar
Linus Torvalds committed
389 390 391 392 393 394 395 396 397 398 399 400 401
     TPACKET_ALIGNMENT=16
   - Start+tp_mac: [ Optional MAC header ]
   - Start+tp_net: Packet data, aligned to TPACKET_ALIGNMENT=16.
   - Pad to align to TPACKET_ALIGNMENT=16
 */
 
 The following are conditions that are checked in packet_set_ring

   tp_block_size must be a multiple of PAGE_SIZE (1)
   tp_frame_size must be greater than TPACKET_HDRLEN (obvious)
   tp_frame_size must be a multiple of TPACKET_ALIGNMENT
   tp_frame_nr   must be exactly frames_per_block*tp_block_nr

402
Note that tp_block_size should be chosen to be a power of two or there will
Linus Torvalds's avatar
Linus Torvalds committed
403 404 405
be a waste of memory.

--------------------------------------------------------------------------------
406
+ Mapping and use of the circular buffer (ring)
Linus Torvalds's avatar
Linus Torvalds committed
407 408
--------------------------------------------------------------------------------

409
The mapping of the buffer in the user process is done with the conventional 
Linus Torvalds's avatar
Linus Torvalds committed
410 411 412 413 414 415 416
mmap function. Even the circular buffer is compound of several physically
discontiguous blocks of memory, they are contiguous to the user space, hence
just one call to mmap is needed:

    mmap(0, size, PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0);

If tp_frame_size is a divisor of tp_block_size frames will be 
417
contiguously spaced by tp_frame_size bytes. If not, each
Linus Torvalds's avatar
Linus Torvalds committed
418 419 420 421
tp_block_size/tp_frame_size frames there will be a gap between 
the frames. This is because a frame cannot be spawn across two
blocks. 

422 423 424 425 426 427 428 429 430 431 432 433 434
To use one socket for capture and transmission, the mapping of both the
RX and TX buffer ring has to be done with one call to mmap:

    ...
    setsockopt(fd, SOL_PACKET, PACKET_RX_RING, &foo, sizeof(foo));
    setsockopt(fd, SOL_PACKET, PACKET_TX_RING, &bar, sizeof(bar));
    ...
    rx_ring = mmap(0, size * 2, PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0);
    tx_ring = rx_ring + size;

RX must be the first as the kernel maps the TX ring memory right
after the RX one.

Linus Torvalds's avatar
Linus Torvalds committed
435 436 437 438 439
At the beginning of each frame there is an status field (see 
struct tpacket_hdr). If this field is 0 means that the frame is ready
to be used for the kernel, If not, there is a frame the user can read 
and the following flags apply:

440
+++ Capture process:
Linus Torvalds's avatar
Linus Torvalds committed
441 442
     from include/linux/if_packet.h

443 444 445 446
     #define TP_STATUS_COPY          (1 << 1)
     #define TP_STATUS_LOSING        (1 << 2)
     #define TP_STATUS_CSUMNOTREADY  (1 << 3)
     #define TP_STATUS_CSUM_VALID    (1 << 7)
Linus Torvalds's avatar
Linus Torvalds committed
447 448 449 450 451 452 453 454 455 456

TP_STATUS_COPY        : This flag indicates that the frame (and associated
                        meta information) has been truncated because it's 
                        larger than tp_frame_size. This packet can be 
                        read entirely with recvfrom().
                        
                        In order to make this work it must to be
                        enabled previously with setsockopt() and 
                        the PACKET_COPY_THRESH option. 

457
                        The number of frames that can be buffered to
Linus Torvalds's avatar
Linus Torvalds committed
458 459 460 461 462 463 464
                        be read with recvfrom is limited like a normal socket.
                        See the SO_RCVBUF option in the socket (7) man page.

TP_STATUS_LOSING      : indicates there were packet drops from last time 
                        statistics where checked with getsockopt() and
                        the PACKET_STATISTICS option.

465
TP_STATUS_CSUMNOTREADY: currently it's used for outgoing IP packets which 
466
                        its checksum will be done in hardware. So while
Linus Torvalds's avatar
Linus Torvalds committed
467 468 469
                        reading the packet we should not try to check the 
                        checksum. 

470 471 472 473 474 475
TP_STATUS_CSUM_VALID  : This flag indicates that at least the transport
                        header checksum of the packet has been already
                        validated on the kernel side. If the flag is not set
                        then we are free to check the checksum by ourselves
                        provided that TP_STATUS_CSUMNOTREADY is also not set.

Linus Torvalds's avatar
Linus Torvalds committed
476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501
for convenience there are also the following defines:

     #define TP_STATUS_KERNEL        0
     #define TP_STATUS_USER          1

The kernel initializes all frames to TP_STATUS_KERNEL, when the kernel
receives a packet it puts in the buffer and updates the status with
at least the TP_STATUS_USER flag. Then the user can read the packet,
once the packet is read the user must zero the status field, so the kernel 
can use again that frame buffer.

The user can use poll (any other variant should apply too) to check if new
packets are in the ring:

    struct pollfd pfd;

    pfd.fd = fd;
    pfd.revents = 0;
    pfd.events = POLLIN|POLLRDNORM|POLLERR;

    if (status == TP_STATUS_KERNEL)
        retval = poll(&pfd, 1, timeout);

It doesn't incur in a race condition to first check the status value and 
then poll for frames.

502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531
++ Transmission process
Those defines are also used for transmission:

     #define TP_STATUS_AVAILABLE        0 // Frame is available
     #define TP_STATUS_SEND_REQUEST     1 // Frame will be sent on next send()
     #define TP_STATUS_SENDING          2 // Frame is currently in transmission
     #define TP_STATUS_WRONG_FORMAT     4 // Frame format is not correct

First, the kernel initializes all frames to TP_STATUS_AVAILABLE. To send a
packet, the user fills a data buffer of an available frame, sets tp_len to
current data buffer size and sets its status field to TP_STATUS_SEND_REQUEST.
This can be done on multiple frames. Once the user is ready to transmit, it
calls send(). Then all buffers with status equal to TP_STATUS_SEND_REQUEST are
forwarded to the network device. The kernel updates each status of sent
frames with TP_STATUS_SENDING until the end of transfer.
At the end of each transfer, buffer status returns to TP_STATUS_AVAILABLE.

    header->tp_len = in_i_size;
    header->tp_status = TP_STATUS_SEND_REQUEST;
    retval = send(this->socket, NULL, 0, 0);

The user can also use poll() to check if a buffer is available:
(status == TP_STATUS_SENDING)

    struct pollfd pfd;
    pfd.fd = fd;
    pfd.revents = 0;
    pfd.events = POLLOUT;
    retval = poll(&pfd, 1, timeout);

532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551
-------------------------------------------------------------------------------
+ What TPACKET versions are available and when to use them?
-------------------------------------------------------------------------------

 int val = tpacket_version;
 setsockopt(fd, SOL_PACKET, PACKET_VERSION, &val, sizeof(val));
 getsockopt(fd, SOL_PACKET, PACKET_VERSION, &val, sizeof(val));

where 'tpacket_version' can be TPACKET_V1 (default), TPACKET_V2, TPACKET_V3.

TPACKET_V1:
	- Default if not otherwise specified by setsockopt(2)
	- RX_RING, TX_RING available

TPACKET_V1 --> TPACKET_V2:
	- Made 64 bit clean due to unsigned long usage in TPACKET_V1
	  structures, thus this also works on 64 bit kernel with 32 bit
	  userspace and the like
	- Timestamp resolution in nanoseconds instead of microseconds
	- RX_RING, TX_RING available
552 553 554 555 556 557 558
	- VLAN metadata information available for packets
	  (TP_STATUS_VLAN_VALID, TP_STATUS_VLAN_TPID_VALID),
	  in the tpacket2_hdr structure:
		- TP_STATUS_VLAN_VALID bit being set into the tp_status field indicates
		  that the tp_vlan_tci field has valid VLAN TCI value
		- TP_STATUS_VLAN_TPID_VALID bit being set into the tp_status field
		  indicates that the tp_vlan_tpid field has valid VLAN TPID value
559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585
	- How to switch to TPACKET_V2:
		1. Replace struct tpacket_hdr by struct tpacket2_hdr
		2. Query header len and save
		3. Set protocol version to 2, set up ring as usual
		4. For getting the sockaddr_ll,
		   use (void *)hdr + TPACKET_ALIGN(hdrlen) instead of
		   (void *)hdr + TPACKET_ALIGN(sizeof(struct tpacket_hdr))

TPACKET_V2 --> TPACKET_V3:
	- Flexible buffer implementation:
		1. Blocks can be configured with non-static frame-size
		2. Read/poll is at a block-level (as opposed to packet-level)
		3. Added poll timeout to avoid indefinite user-space wait
		   on idle links
		4. Added user-configurable knobs:
			4.1 block::timeout
			4.2 tpkt_hdr::sk_rxhash
	- RX Hash data available in user space
	- Currently only RX_RING available

-------------------------------------------------------------------------------
+ AF_PACKET fanout mode
-------------------------------------------------------------------------------

In the AF_PACKET fanout mode, packet reception can be load balanced among
processes. This also works in combination with mmap(2) on packet sockets.

586 587
Currently implemented fanout policies are:

588
  - PACKET_FANOUT_HASH: schedule to socket by skb's packet hash
589 590 591 592
  - PACKET_FANOUT_LB: schedule to socket by round-robin
  - PACKET_FANOUT_CPU: schedule to socket by CPU packet arrives on
  - PACKET_FANOUT_RND: schedule to socket by random selection
  - PACKET_FANOUT_ROLLOVER: if one socket is full, rollover to another
593
  - PACKET_FANOUT_QM: schedule to socket by skbs recorded queue_mapping
594

595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735
Minimal example code by David S. Miller (try things like "./test eth0 hash",
"./test eth0 lb", etc.):

#include <stddef.h>
#include <stdlib.h>
#include <stdio.h>
#include <string.h>

#include <sys/types.h>
#include <sys/wait.h>
#include <sys/socket.h>
#include <sys/ioctl.h>

#include <unistd.h>

#include <linux/if_ether.h>
#include <linux/if_packet.h>

#include <net/if.h>

static const char *device_name;
static int fanout_type;
static int fanout_id;

#ifndef PACKET_FANOUT
# define PACKET_FANOUT			18
# define PACKET_FANOUT_HASH		0
# define PACKET_FANOUT_LB		1
#endif

static int setup_socket(void)
{
	int err, fd = socket(AF_PACKET, SOCK_RAW, htons(ETH_P_IP));
	struct sockaddr_ll ll;
	struct ifreq ifr;
	int fanout_arg;

	if (fd < 0) {
		perror("socket");
		return EXIT_FAILURE;
	}

	memset(&ifr, 0, sizeof(ifr));
	strcpy(ifr.ifr_name, device_name);
	err = ioctl(fd, SIOCGIFINDEX, &ifr);
	if (err < 0) {
		perror("SIOCGIFINDEX");
		return EXIT_FAILURE;
	}

	memset(&ll, 0, sizeof(ll));
	ll.sll_family = AF_PACKET;
	ll.sll_ifindex = ifr.ifr_ifindex;
	err = bind(fd, (struct sockaddr *) &ll, sizeof(ll));
	if (err < 0) {
		perror("bind");
		return EXIT_FAILURE;
	}

	fanout_arg = (fanout_id | (fanout_type << 16));
	err = setsockopt(fd, SOL_PACKET, PACKET_FANOUT,
			 &fanout_arg, sizeof(fanout_arg));
	if (err) {
		perror("setsockopt");
		return EXIT_FAILURE;
	}

	return fd;
}

static void fanout_thread(void)
{
	int fd = setup_socket();
	int limit = 10000;

	if (fd < 0)
		exit(fd);

	while (limit-- > 0) {
		char buf[1600];
		int err;

		err = read(fd, buf, sizeof(buf));
		if (err < 0) {
			perror("read");
			exit(EXIT_FAILURE);
		}
		if ((limit % 10) == 0)
			fprintf(stdout, "(%d) \n", getpid());
	}

	fprintf(stdout, "%d: Received 10000 packets\n", getpid());

	close(fd);
	exit(0);
}

int main(int argc, char **argp)
{
	int fd, err;
	int i;

	if (argc != 3) {
		fprintf(stderr, "Usage: %s INTERFACE {hash|lb}\n", argp[0]);
		return EXIT_FAILURE;
	}

	if (!strcmp(argp[2], "hash"))
		fanout_type = PACKET_FANOUT_HASH;
	else if (!strcmp(argp[2], "lb"))
		fanout_type = PACKET_FANOUT_LB;
	else {
		fprintf(stderr, "Unknown fanout type [%s]\n", argp[2]);
		exit(EXIT_FAILURE);
	}

	device_name = argp[1];
	fanout_id = getpid() & 0xffff;

	for (i = 0; i < 4; i++) {
		pid_t pid = fork();

		switch (pid) {
		case 0:
			fanout_thread();

		case -1:
			perror("fork");
			exit(EXIT_FAILURE);
		}
	}

	for (i = 0; i < 4; i++) {
		int status;

		wait(&status);
	}

	return 0;
}

736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755
-------------------------------------------------------------------------------
+ AF_PACKET TPACKET_V3 example
-------------------------------------------------------------------------------

AF_PACKET's TPACKET_V3 ring buffer can be configured to use non-static frame
sizes by doing it's own memory management. It is based on blocks where polling
works on a per block basis instead of per ring as in TPACKET_V2 and predecessor.

It is said that TPACKET_V3 brings the following benefits:
 *) ~15 - 20% reduction in CPU-usage
 *) ~20% increase in packet capture rate
 *) ~2x increase in packet density
 *) Port aggregation analysis
 *) Non static frame size to capture entire packet payload

So it seems to be a good candidate to be used with packet fanout.

Minimal example code by Daniel Borkmann based on Chetan Loke's lolpcap (compile
it with gcc -Wall -O2 blob.c, and try things like "./a.out eth0", etc.):

756 757 758 759 760 761
/* Written from scratch, but kernel-to-user space API usage
 * dissected from lolpcap:
 *  Copyright 2011, Chetan Loke <loke.chetan@gmail.com>
 *  License: GPL, version 2.0
 */

762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801
#include <stdio.h>
#include <stdlib.h>
#include <stdint.h>
#include <string.h>
#include <assert.h>
#include <net/if.h>
#include <arpa/inet.h>
#include <netdb.h>
#include <poll.h>
#include <unistd.h>
#include <signal.h>
#include <inttypes.h>
#include <sys/socket.h>
#include <sys/mman.h>
#include <linux/if_packet.h>
#include <linux/if_ether.h>
#include <linux/ip.h>

#ifndef likely
# define likely(x)		__builtin_expect(!!(x), 1)
#endif
#ifndef unlikely
# define unlikely(x)		__builtin_expect(!!(x), 0)
#endif

struct block_desc {
	uint32_t version;
	uint32_t offset_to_priv;
	struct tpacket_hdr_v1 h1;
};

struct ring {
	struct iovec *rd;
	uint8_t *map;
	struct tpacket_req3 req;
};

static unsigned long packets_total = 0, bytes_total = 0;
static sig_atomic_t sigint = 0;

802
static void sighandler(int num)
803 804 805 806 807 808 809 810
{
	sigint = 1;
}

static int setup_socket(struct ring *ring, char *netdev)
{
	int err, i, fd, v = TPACKET_V3;
	struct sockaddr_ll ll;
811 812
	unsigned int blocksiz = 1 << 22, framesiz = 1 << 11;
	unsigned int blocknum = 64;
813 814 815 816 817 818 819 820 821 822 823 824 825 826

	fd = socket(AF_PACKET, SOCK_RAW, htons(ETH_P_ALL));
	if (fd < 0) {
		perror("socket");
		exit(1);
	}

	err = setsockopt(fd, SOL_PACKET, PACKET_VERSION, &v, sizeof(v));
	if (err < 0) {
		perror("setsockopt");
		exit(1);
	}

	memset(&ring->req, 0, sizeof(ring->req));
827 828 829 830 831 832
	ring->req.tp_block_size = blocksiz;
	ring->req.tp_frame_size = framesiz;
	ring->req.tp_block_nr = blocknum;
	ring->req.tp_frame_nr = (blocksiz * blocknum) / framesiz;
	ring->req.tp_retire_blk_tov = 60;
	ring->req.tp_feature_req_word = TP_FT_REQ_FILL_RXHASH;
833 834 835 836 837 838 839 840 841

	err = setsockopt(fd, SOL_PACKET, PACKET_RX_RING, &ring->req,
			 sizeof(ring->req));
	if (err < 0) {
		perror("setsockopt");
		exit(1);
	}

	ring->map = mmap(NULL, ring->req.tp_block_size * ring->req.tp_block_nr,
842
			 PROT_READ | PROT_WRITE, MAP_SHARED | MAP_LOCKED, fd, 0);
843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900
	if (ring->map == MAP_FAILED) {
		perror("mmap");
		exit(1);
	}

	ring->rd = malloc(ring->req.tp_block_nr * sizeof(*ring->rd));
	assert(ring->rd);
	for (i = 0; i < ring->req.tp_block_nr; ++i) {
		ring->rd[i].iov_base = ring->map + (i * ring->req.tp_block_size);
		ring->rd[i].iov_len = ring->req.tp_block_size;
	}

	memset(&ll, 0, sizeof(ll));
	ll.sll_family = PF_PACKET;
	ll.sll_protocol = htons(ETH_P_ALL);
	ll.sll_ifindex = if_nametoindex(netdev);
	ll.sll_hatype = 0;
	ll.sll_pkttype = 0;
	ll.sll_halen = 0;

	err = bind(fd, (struct sockaddr *) &ll, sizeof(ll));
	if (err < 0) {
		perror("bind");
		exit(1);
	}

	return fd;
}

static void display(struct tpacket3_hdr *ppd)
{
	struct ethhdr *eth = (struct ethhdr *) ((uint8_t *) ppd + ppd->tp_mac);
	struct iphdr *ip = (struct iphdr *) ((uint8_t *) eth + ETH_HLEN);

	if (eth->h_proto == htons(ETH_P_IP)) {
		struct sockaddr_in ss, sd;
		char sbuff[NI_MAXHOST], dbuff[NI_MAXHOST];

		memset(&ss, 0, sizeof(ss));
		ss.sin_family = PF_INET;
		ss.sin_addr.s_addr = ip->saddr;
		getnameinfo((struct sockaddr *) &ss, sizeof(ss),
			    sbuff, sizeof(sbuff), NULL, 0, NI_NUMERICHOST);

		memset(&sd, 0, sizeof(sd));
		sd.sin_family = PF_INET;
		sd.sin_addr.s_addr = ip->daddr;
		getnameinfo((struct sockaddr *) &sd, sizeof(sd),
			    dbuff, sizeof(dbuff), NULL, 0, NI_NUMERICHOST);

		printf("%s -> %s, ", sbuff, dbuff);
	}

	printf("rxhash: 0x%x\n", ppd->hv1.tp_rxhash);
}

static void walk_block(struct block_desc *pbd, const int block_num)
{
901
	int num_pkts = pbd->h1.num_pkts, i;
902 903 904
	unsigned long bytes = 0;
	struct tpacket3_hdr *ppd;

905 906
	ppd = (struct tpacket3_hdr *) ((uint8_t *) pbd +
				       pbd->h1.offset_to_first_pkt);
907 908 909 910
	for (i = 0; i < num_pkts; ++i) {
		bytes += ppd->tp_snaplen;
		display(ppd);

911 912
		ppd = (struct tpacket3_hdr *) ((uint8_t *) ppd +
					       ppd->tp_next_offset);
913 914 915 916 917 918
	}

	packets_total += num_pkts;
	bytes_total += bytes;
}

919
static void flush_block(struct block_desc *pbd)
920
{
921
	pbd->h1.block_status = TP_STATUS_KERNEL;
922 923 924 925 926 927 928 929 930 931 932 933 934 935 936
}

static void teardown_socket(struct ring *ring, int fd)
{
	munmap(ring->map, ring->req.tp_block_size * ring->req.tp_block_nr);
	free(ring->rd);
	close(fd);
}

int main(int argc, char **argp)
{
	int fd, err;
	socklen_t len;
	struct ring ring;
	struct pollfd pfd;
937
	unsigned int block_num = 0, blocks = 64;
938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958
	struct block_desc *pbd;
	struct tpacket_stats_v3 stats;

	if (argc != 2) {
		fprintf(stderr, "Usage: %s INTERFACE\n", argp[0]);
		return EXIT_FAILURE;
	}

	signal(SIGINT, sighandler);

	memset(&ring, 0, sizeof(ring));
	fd = setup_socket(&ring, argp[argc - 1]);
	assert(fd > 0);

	memset(&pfd, 0, sizeof(pfd));
	pfd.fd = fd;
	pfd.events = POLLIN | POLLERR;
	pfd.revents = 0;

	while (likely(!sigint)) {
		pbd = (struct block_desc *) ring.rd[block_num].iov_base;
959 960

		if ((pbd->h1.block_status & TP_STATUS_USER) == 0) {
961
			poll(&pfd, 1, -1);
962
			continue;
963 964 965 966
		}

		walk_block(pbd, block_num);
		flush_block(pbd);
967
		block_num = (block_num + 1) % blocks;
968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985
	}

	len = sizeof(stats);
	err = getsockopt(fd, SOL_PACKET, PACKET_STATISTICS, &stats, &len);
	if (err < 0) {
		perror("getsockopt");
		exit(1);
	}

	fflush(stdout);
	printf("\nReceived %u packets, %lu bytes, %u dropped, freeze_q_cnt: %u\n",
	       stats.tp_packets, bytes_total, stats.tp_drops,
	       stats.tp_freeze_q_cnt);

	teardown_socket(&ring, fd);
	return 0;
}

986 987 988 989 990 991 992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006
-------------------------------------------------------------------------------
+ PACKET_QDISC_BYPASS
-------------------------------------------------------------------------------

If there is a requirement to load the network with many packets in a similar
fashion as pktgen does, you might set the following option after socket
creation:

    int one = 1;
    setsockopt(fd, SOL_PACKET, PACKET_QDISC_BYPASS, &one, sizeof(one));

This has the side-effect, that packets sent through PF_PACKET will bypass the
kernel's qdisc layer and are forcedly pushed to the driver directly. Meaning,
packet are not buffered, tc disciplines are ignored, increased loss can occur
and such packets are also not visible to other PF_PACKET sockets anymore. So,
you have been warned; generally, this can be useful for stress testing various
components of a system.

On default, PACKET_QDISC_BYPASS is disabled and needs to be explicitly enabled
on PF_PACKET sockets.

1007 1008 1009 1010 1011
-------------------------------------------------------------------------------
+ PACKET_TIMESTAMP
-------------------------------------------------------------------------------

The PACKET_TIMESTAMP setting determines the source of the timestamp in
1012 1013 1014 1015 1016
the packet meta information for mmap(2)ed RX_RING and TX_RINGs.  If your
NIC is capable of timestamping packets in hardware, you can request those
hardware timestamps to be used. Note: you may need to enable the generation
of hardware timestamps with SIOCSHWTSTAMP (see related information from
Documentation/networking/timestamping.txt).
1017

1018 1019 1020
PACKET_TIMESTAMP accepts the same integer bit field as SO_TIMESTAMPING:

    int req = SOF_TIMESTAMPING_RAW_HARDWARE;
1021 1022
    setsockopt(fd, SOL_PACKET, PACKET_TIMESTAMP, (void *) &req, sizeof(req))

1023 1024 1025 1026 1027 1028 1029 1030 1031
For the mmap(2)ed ring buffers, such timestamps are stored in the
tpacket{,2,3}_hdr structure's tp_sec and tp_{n,u}sec members. To determine
what kind of timestamp has been reported, the tp_status field is binary |'ed
with the following possible bits ...

    TP_STATUS_TS_RAW_HARDWARE
    TP_STATUS_TS_SOFTWARE

... that are equivalent to its SOF_TIMESTAMPING_* counterparts. For the
1032 1033 1034
RX_RING, if neither is set (i.e. PACKET_TIMESTAMP is not set), then a
software fallback was invoked *within* PF_PACKET's processing code (less
precise).
1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051

Getting timestamps for the TX_RING works as follows: i) fill the ring frames,
ii) call sendto() e.g. in blocking mode, iii) wait for status of relevant
frames to be updated resp. the frame handed over to the application, iv) walk
through the frames to pick up the individual hw/sw timestamps.

Only (!) if transmit timestamping is enabled, then these bits are combined
with binary | with TP_STATUS_AVAILABLE, so you must check for that in your
application (e.g. !(tp_status & (TP_STATUS_SEND_REQUEST | TP_STATUS_SENDING))
in a first step to see if the frame belongs to the application, and then
one can extract the type of timestamp in a second step from tp_status)!

If you don't care about them, thus having it disabled, checking for
TP_STATUS_AVAILABLE resp. TP_STATUS_WRONG_FORMAT is sufficient. If in the
TX_RING part only TP_STATUS_AVAILABLE is set, then the tp_sec and tp_{n,u}sec
members do not contain a valid value. For TX_RINGs, by default no timestamp
is generated!
1052 1053 1054 1055

See include/linux/net_tstamp.h and Documentation/networking/timestamping
for more information on hardware timestamps.

1056 1057 1058 1059 1060 1061 1062
-------------------------------------------------------------------------------
+ Miscellaneous bits
-------------------------------------------------------------------------------

- Packet sockets work well together with Linux socket filters, thus you also
  might want to have a look at Documentation/networking/filter.txt

Linus Torvalds's avatar
Linus Torvalds committed
1063 1064 1065 1066 1067 1068
--------------------------------------------------------------------------------
+ THANKS
--------------------------------------------------------------------------------
   
   Jesse Brandeburg, for fixing my grammathical/spelling errors