4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
5 * Copyright 2014 6WIND S.A.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * * Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * * Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * * Neither the name of Intel Corporation nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
25 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
26 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
42 * The mbuf library provides the ability to create and destroy buffers
43 * that may be used by the RTE application to store message
44 * buffers. The message buffers are stored in a mempool, using the
45 * RTE mempool library.
47 * This library provide an API to allocate/free packet mbufs, which are
48 * used to carry network packets.
50 * To understand the concepts of packet buffers or mbufs, you
51 * should read "TCP/IP Illustrated, Volume 2: The Implementation,
52 * Addison-Wesley, 1995, ISBN 0-201-63354-X from Richard Stevens"
53 * http://www.kohala.com/start/tcpipiv2.html
57 #include <rte_common.h>
58 #include <rte_mempool.h>
59 #include <rte_memory.h>
60 #include <rte_atomic.h>
61 #include <rte_prefetch.h>
62 #include <rte_branch_prediction.h>
68 /* deprecated options */
69 #pragma GCC poison RTE_MBUF_SCATTER_GATHER
70 #pragma GCC poison RTE_MBUF_REFCNT
73 * Packet Offload Features Flags. It also carry packet type information.
74 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
76 * - RX flags start at bit position zero, and get added to the left of previous
78 * - The most-significant 3 bits are reserved for generic mbuf flags
79 * - TX flags therefore start at bit position 60 (i.e. 63-3), and new flags get
80 * added to the right of the previously defined flags i.e. they should count
81 * downwards, not upwards.
83 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
84 * rte_get_tx_ol_flag_name().
86 #define PKT_RX_VLAN_PKT (1ULL << 0) /**< RX packet is a 802.1q VLAN packet. */
87 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
88 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
89 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
90 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
91 #define PKT_RX_EIP_CKSUM_BAD (0ULL << 0) /**< External IP header checksum error. */
92 #define PKT_RX_OVERSIZE (0ULL << 0) /**< Num of desc of an RX pkt oversize. */
93 #define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
94 #define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
95 #define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
96 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
97 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
98 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
99 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
100 #define PKT_RX_QINQ_PKT (1ULL << 15) /**< RX packet with double VLAN stripped. */
101 /* add new RX flags here */
103 /* add new TX flags here */
106 * Second VLAN insertion (QinQ) flag.
108 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
111 * TCP segmentation offload. To enable this offload feature for a
112 * packet to be transmitted on hardware supporting TSO:
113 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
115 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
116 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
118 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
119 * - calculate the pseudo header checksum without taking ip_len in account,
120 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
121 * rte_ipv6_phdr_cksum() that can be used as helpers.
123 #define PKT_TX_TCP_SEG (1ULL << 50)
125 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
128 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
129 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
130 * L4 checksum offload, the user needs to:
131 * - fill l2_len and l3_len in mbuf
132 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
133 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
134 * - calculate the pseudo header checksum and set it in the L4 header (only
135 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
136 * For SCTP, set the crc field to 0.
138 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
139 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
140 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
141 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
142 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
145 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
146 * also be set by the application, although a PMD will only check
148 * - set the IP checksum field in the packet to 0
149 * - fill the mbuf offload information: l2_len, l3_len
151 #define PKT_TX_IP_CKSUM (1ULL << 54)
154 * Packet is IPv4. This flag must be set when using any offload feature
155 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
156 * packet. If the packet is a tunneled packet, this flag is related to
159 #define PKT_TX_IPV4 (1ULL << 55)
162 * Packet is IPv6. This flag must be set when using an offload feature
163 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
164 * packet. If the packet is a tunneled packet, this flag is related to
167 #define PKT_TX_IPV6 (1ULL << 56)
169 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
172 * Offload the IP checksum of an external header in the hardware. The
173 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
174 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
175 * packet must be set to 0.
176 * - set the outer IP checksum field in the packet to 0
177 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
179 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
182 * Packet outer header is IPv4. This flag must be set when using any
183 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
184 * outer header of the tunneled packet is an IPv4 packet.
186 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
189 * Packet outer header is IPv6. This flag must be set when using any
190 * outer offload feature (L4 checksum) to tell the NIC that the outer
191 * header of the tunneled packet is an IPv6 packet.
193 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
195 #define __RESERVED (1ULL << 61) /**< reserved for future mbuf use */
197 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
199 /* Use final bit of flags to indicate a control mbuf */
200 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
203 * 32 bits are divided into several fields to mark packet types. Note that
204 * each field is indexical.
205 * - Bit 3:0 is for L2 types.
206 * - Bit 7:4 is for L3 or outer L3 (for tunneling case) types.
207 * - Bit 11:8 is for L4 or outer L4 (for tunneling case) types.
208 * - Bit 15:12 is for tunnel types.
209 * - Bit 19:16 is for inner L2 types.
210 * - Bit 23:20 is for inner L3 types.
211 * - Bit 27:24 is for inner L4 types.
212 * - Bit 31:28 is reserved.
214 * To be compatible with Vector PMD, RTE_PTYPE_L3_IPV4, RTE_PTYPE_L3_IPV4_EXT,
215 * RTE_PTYPE_L3_IPV6, RTE_PTYPE_L3_IPV6_EXT, RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP
216 * and RTE_PTYPE_L4_SCTP should be kept as below in a contiguous 7 bits.
218 * Note that L3 types values are selected for checking IPV4/IPV6 header from
219 * performance point of view. Reading annotations of RTE_ETH_IS_IPV4_HDR and
220 * RTE_ETH_IS_IPV6_HDR is needed for any future changes of L3 type values.
222 * Note that the packet types of the same packet recognized by different
223 * hardware may be different, as different hardware may have different
224 * capability of packet type recognition.
227 * <'ether type'=0x0800
228 * | 'version'=4, 'protocol'=0x29
229 * | 'version'=6, 'next header'=0x3A
231 * will be recognized on i40e hardware as packet type combination of,
232 * RTE_PTYPE_L2_ETHER |
233 * RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
234 * RTE_PTYPE_TUNNEL_IP |
235 * RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
236 * RTE_PTYPE_INNER_L4_ICMP.
238 * <'ether type'=0x86DD
239 * | 'version'=6, 'next header'=0x2F
241 * | 'version'=6, 'next header'=0x11
243 * will be recognized on i40e hardware as packet type combination of,
244 * RTE_PTYPE_L2_ETHER |
245 * RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
246 * RTE_PTYPE_TUNNEL_GRENAT |
247 * RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
248 * RTE_PTYPE_INNER_L4_UDP.
250 #define RTE_PTYPE_UNKNOWN 0x00000000
252 * Ethernet packet type.
253 * It is used for outer packet for tunneling cases.
256 * <'ether type'=[0x0800|0x86DD]>
258 #define RTE_PTYPE_L2_ETHER 0x00000001
260 * Ethernet packet type for time sync.
263 * <'ether type'=0x88F7>
265 #define RTE_PTYPE_L2_ETHER_TIMESYNC 0x00000002
267 * ARP (Address Resolution Protocol) packet type.
270 * <'ether type'=0x0806>
272 #define RTE_PTYPE_L2_ETHER_ARP 0x00000003
274 * LLDP (Link Layer Discovery Protocol) packet type.
277 * <'ether type'=0x88CC>
279 #define RTE_PTYPE_L2_ETHER_LLDP 0x00000004
281 * Mask of layer 2 packet types.
282 * It is used for outer packet for tunneling cases.
284 #define RTE_PTYPE_L2_MASK 0x0000000f
286 * IP (Internet Protocol) version 4 packet type.
287 * It is used for outer packet for tunneling cases, and does not contain any
291 * <'ether type'=0x0800
292 * | 'version'=4, 'ihl'=5>
294 #define RTE_PTYPE_L3_IPV4 0x00000010
296 * IP (Internet Protocol) version 4 packet type.
297 * It is used for outer packet for tunneling cases, and contains header
301 * <'ether type'=0x0800
302 * | 'version'=4, 'ihl'=[6-15], 'options'>
304 #define RTE_PTYPE_L3_IPV4_EXT 0x00000030
306 * IP (Internet Protocol) version 6 packet type.
307 * It is used for outer packet for tunneling cases, and does not contain any
311 * <'ether type'=0x86DD
312 * | 'version'=6, 'next header'=0x3B>
314 #define RTE_PTYPE_L3_IPV6 0x00000040
316 * IP (Internet Protocol) version 4 packet type.
317 * It is used for outer packet for tunneling cases, and may or maynot contain
321 * <'ether type'=0x0800
322 * | 'version'=4, 'ihl'=[5-15], <'options'>>
324 #define RTE_PTYPE_L3_IPV4_EXT_UNKNOWN 0x00000090
326 * IP (Internet Protocol) version 6 packet type.
327 * It is used for outer packet for tunneling cases, and contains extension
331 * <'ether type'=0x86DD
332 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
333 * 'extension headers'>
335 #define RTE_PTYPE_L3_IPV6_EXT 0x000000c0
337 * IP (Internet Protocol) version 6 packet type.
338 * It is used for outer packet for tunneling cases, and may or maynot contain
342 * <'ether type'=0x86DD
343 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
344 * <'extension headers'>>
346 #define RTE_PTYPE_L3_IPV6_EXT_UNKNOWN 0x000000e0
348 * Mask of layer 3 packet types.
349 * It is used for outer packet for tunneling cases.
351 #define RTE_PTYPE_L3_MASK 0x000000f0
353 * TCP (Transmission Control Protocol) packet type.
354 * It is used for outer packet for tunneling cases.
357 * <'ether type'=0x0800
358 * | 'version'=4, 'protocol'=6, 'MF'=0>
360 * <'ether type'=0x86DD
361 * | 'version'=6, 'next header'=6>
363 #define RTE_PTYPE_L4_TCP 0x00000100
365 * UDP (User Datagram Protocol) packet type.
366 * It is used for outer packet for tunneling cases.
369 * <'ether type'=0x0800
370 * | 'version'=4, 'protocol'=17, 'MF'=0>
372 * <'ether type'=0x86DD
373 * | 'version'=6, 'next header'=17>
375 #define RTE_PTYPE_L4_UDP 0x00000200
377 * Fragmented IP (Internet Protocol) packet type.
378 * It is used for outer packet for tunneling cases.
380 * It refers to those packets of any IP types, which can be recognized as
381 * fragmented. A fragmented packet cannot be recognized as any other L4 types
382 * (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP,
383 * RTE_PTYPE_L4_NONFRAG).
386 * <'ether type'=0x0800
387 * | 'version'=4, 'MF'=1>
389 * <'ether type'=0x86DD
390 * | 'version'=6, 'next header'=44>
392 #define RTE_PTYPE_L4_FRAG 0x00000300
394 * SCTP (Stream Control Transmission Protocol) packet type.
395 * It is used for outer packet for tunneling cases.
398 * <'ether type'=0x0800
399 * | 'version'=4, 'protocol'=132, 'MF'=0>
401 * <'ether type'=0x86DD
402 * | 'version'=6, 'next header'=132>
404 #define RTE_PTYPE_L4_SCTP 0x00000400
406 * ICMP (Internet Control Message Protocol) packet type.
407 * It is used for outer packet for tunneling cases.
410 * <'ether type'=0x0800
411 * | 'version'=4, 'protocol'=1, 'MF'=0>
413 * <'ether type'=0x86DD
414 * | 'version'=6, 'next header'=1>
416 #define RTE_PTYPE_L4_ICMP 0x00000500
418 * Non-fragmented IP (Internet Protocol) packet type.
419 * It is used for outer packet for tunneling cases.
421 * It refers to those packets of any IP types, while cannot be recognized as
422 * any of above L4 types (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP,
423 * RTE_PTYPE_L4_FRAG, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP).
426 * <'ether type'=0x0800
427 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
429 * <'ether type'=0x86DD
430 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
432 #define RTE_PTYPE_L4_NONFRAG 0x00000600
434 * Mask of layer 4 packet types.
435 * It is used for outer packet for tunneling cases.
437 #define RTE_PTYPE_L4_MASK 0x00000f00
439 * IP (Internet Protocol) in IP (Internet Protocol) tunneling packet type.
442 * <'ether type'=0x0800
443 * | 'version'=4, 'protocol'=[4|41]>
445 * <'ether type'=0x86DD
446 * | 'version'=6, 'next header'=[4|41]>
448 #define RTE_PTYPE_TUNNEL_IP 0x00001000
450 * GRE (Generic Routing Encapsulation) tunneling packet type.
453 * <'ether type'=0x0800
454 * | 'version'=4, 'protocol'=47>
456 * <'ether type'=0x86DD
457 * | 'version'=6, 'next header'=47>
459 #define RTE_PTYPE_TUNNEL_GRE 0x00002000
461 * VXLAN (Virtual eXtensible Local Area Network) tunneling packet type.
464 * <'ether type'=0x0800
465 * | 'version'=4, 'protocol'=17
466 * | 'destination port'=4798>
468 * <'ether type'=0x86DD
469 * | 'version'=6, 'next header'=17
470 * | 'destination port'=4798>
472 #define RTE_PTYPE_TUNNEL_VXLAN 0x00003000
474 * NVGRE (Network Virtualization using Generic Routing Encapsulation) tunneling
478 * <'ether type'=0x0800
479 * | 'version'=4, 'protocol'=47
480 * | 'protocol type'=0x6558>
482 * <'ether type'=0x86DD
483 * | 'version'=6, 'next header'=47
484 * | 'protocol type'=0x6558'>
486 #define RTE_PTYPE_TUNNEL_NVGRE 0x00004000
488 * GENEVE (Generic Network Virtualization Encapsulation) tunneling packet type.
491 * <'ether type'=0x0800
492 * | 'version'=4, 'protocol'=17
493 * | 'destination port'=6081>
495 * <'ether type'=0x86DD
496 * | 'version'=6, 'next header'=17
497 * | 'destination port'=6081>
499 #define RTE_PTYPE_TUNNEL_GENEVE 0x00005000
501 * Tunneling packet type of Teredo, VXLAN (Virtual eXtensible Local Area
502 * Network) or GRE (Generic Routing Encapsulation) could be recognized as this
503 * packet type, if they can not be recognized independently as of hardware
506 #define RTE_PTYPE_TUNNEL_GRENAT 0x00006000
508 * Mask of tunneling packet types.
510 #define RTE_PTYPE_TUNNEL_MASK 0x0000f000
512 * Ethernet packet type.
513 * It is used for inner packet type only.
515 * Packet format (inner only):
516 * <'ether type'=[0x800|0x86DD]>
518 #define RTE_PTYPE_INNER_L2_ETHER 0x00010000
520 * Ethernet packet type with VLAN (Virtual Local Area Network) tag.
522 * Packet format (inner only):
523 * <'ether type'=[0x800|0x86DD], vlan=[1-4095]>
525 #define RTE_PTYPE_INNER_L2_ETHER_VLAN 0x00020000
527 * Mask of inner layer 2 packet types.
529 #define RTE_PTYPE_INNER_L2_MASK 0x000f0000
531 * IP (Internet Protocol) version 4 packet type.
532 * It is used for inner packet only, and does not contain any header option.
534 * Packet format (inner only):
535 * <'ether type'=0x0800
536 * | 'version'=4, 'ihl'=5>
538 #define RTE_PTYPE_INNER_L3_IPV4 0x00100000
540 * IP (Internet Protocol) version 4 packet type.
541 * It is used for inner packet only, and contains header options.
543 * Packet format (inner only):
544 * <'ether type'=0x0800
545 * | 'version'=4, 'ihl'=[6-15], 'options'>
547 #define RTE_PTYPE_INNER_L3_IPV4_EXT 0x00200000
549 * IP (Internet Protocol) version 6 packet type.
550 * It is used for inner packet only, and does not contain any extension header.
552 * Packet format (inner only):
553 * <'ether type'=0x86DD
554 * | 'version'=6, 'next header'=0x3B>
556 #define RTE_PTYPE_INNER_L3_IPV6 0x00300000
558 * IP (Internet Protocol) version 4 packet type.
559 * It is used for inner packet only, and may or maynot contain header options.
561 * Packet format (inner only):
562 * <'ether type'=0x0800
563 * | 'version'=4, 'ihl'=[5-15], <'options'>>
565 #define RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN 0x00400000
567 * IP (Internet Protocol) version 6 packet type.
568 * It is used for inner packet only, and contains extension headers.
570 * Packet format (inner only):
571 * <'ether type'=0x86DD
572 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
573 * 'extension headers'>
575 #define RTE_PTYPE_INNER_L3_IPV6_EXT 0x00500000
577 * IP (Internet Protocol) version 6 packet type.
578 * It is used for inner packet only, and may or maynot contain extension
581 * Packet format (inner only):
582 * <'ether type'=0x86DD
583 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
584 * <'extension headers'>>
586 #define RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN 0x00600000
588 * Mask of inner layer 3 packet types.
590 #define RTE_PTYPE_INNER_L3_MASK 0x00f00000
592 * TCP (Transmission Control Protocol) packet type.
593 * It is used for inner packet only.
595 * Packet format (inner only):
596 * <'ether type'=0x0800
597 * | 'version'=4, 'protocol'=6, 'MF'=0>
599 * <'ether type'=0x86DD
600 * | 'version'=6, 'next header'=6>
602 #define RTE_PTYPE_INNER_L4_TCP 0x01000000
604 * UDP (User Datagram Protocol) packet type.
605 * It is used for inner packet only.
607 * Packet format (inner only):
608 * <'ether type'=0x0800
609 * | 'version'=4, 'protocol'=17, 'MF'=0>
611 * <'ether type'=0x86DD
612 * | 'version'=6, 'next header'=17>
614 #define RTE_PTYPE_INNER_L4_UDP 0x02000000
616 * Fragmented IP (Internet Protocol) packet type.
617 * It is used for inner packet only, and may or maynot have layer 4 packet.
619 * Packet format (inner only):
620 * <'ether type'=0x0800
621 * | 'version'=4, 'MF'=1>
623 * <'ether type'=0x86DD
624 * | 'version'=6, 'next header'=44>
626 #define RTE_PTYPE_INNER_L4_FRAG 0x03000000
628 * SCTP (Stream Control Transmission Protocol) packet type.
629 * It is used for inner packet only.
631 * Packet format (inner only):
632 * <'ether type'=0x0800
633 * | 'version'=4, 'protocol'=132, 'MF'=0>
635 * <'ether type'=0x86DD
636 * | 'version'=6, 'next header'=132>
638 #define RTE_PTYPE_INNER_L4_SCTP 0x04000000
640 * ICMP (Internet Control Message Protocol) packet type.
641 * It is used for inner packet only.
643 * Packet format (inner only):
644 * <'ether type'=0x0800
645 * | 'version'=4, 'protocol'=1, 'MF'=0>
647 * <'ether type'=0x86DD
648 * | 'version'=6, 'next header'=1>
650 #define RTE_PTYPE_INNER_L4_ICMP 0x05000000
652 * Non-fragmented IP (Internet Protocol) packet type.
653 * It is used for inner packet only, and may or maynot have other unknown layer
656 * Packet format (inner only):
657 * <'ether type'=0x0800
658 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
660 * <'ether type'=0x86DD
661 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
663 #define RTE_PTYPE_INNER_L4_NONFRAG 0x06000000
665 * Mask of inner layer 4 packet types.
667 #define RTE_PTYPE_INNER_L4_MASK 0x0f000000
670 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
671 * one, bit 4 is selected to be used for IPv4 only. Then checking bit 4 can
672 * determine if it is an IPV4 packet.
674 #define RTE_ETH_IS_IPV4_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV4)
677 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
678 * one, bit 6 is selected to be used for IPv4 only. Then checking bit 6 can
679 * determine if it is an IPV4 packet.
681 #define RTE_ETH_IS_IPV6_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV6)
683 /* Check if it is a tunneling packet */
684 #define RTE_ETH_IS_TUNNEL_PKT(ptype) ((ptype) & (RTE_PTYPE_TUNNEL_MASK | \
685 RTE_PTYPE_INNER_L2_MASK | \
686 RTE_PTYPE_INNER_L3_MASK | \
687 RTE_PTYPE_INNER_L4_MASK))
689 /** Alignment constraint of mbuf private area. */
690 #define RTE_MBUF_PRIV_ALIGN 8
693 * Get the name of a RX offload flag
696 * The mask describing the flag.
698 * The name of this flag, or NULL if it's not a valid RX flag.
700 const char *rte_get_rx_ol_flag_name(uint64_t mask);
703 * Get the name of a TX offload flag
706 * The mask describing the flag. Usually only one bit must be set.
707 * Several bits can be given if they belong to the same mask.
708 * Ex: PKT_TX_L4_MASK.
710 * The name of this flag, or NULL if it's not a valid TX flag.
712 const char *rte_get_tx_ol_flag_name(uint64_t mask);
715 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
716 * splitting it into multiple segments.
717 * So, for mbufs that planned to be involved into RX/TX, the recommended
718 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
720 #define RTE_MBUF_DEFAULT_DATAROOM 2048
721 #define RTE_MBUF_DEFAULT_BUF_SIZE \
722 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
724 /* define a set of marker types that can be used to refer to set points in the
726 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
727 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
728 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
729 * with a single assignment */
731 /** Opaque rte_mbuf_offload structure declarations */
732 struct rte_mbuf_offload;
735 * The generic rte_mbuf, containing a packet mbuf.
740 void *buf_addr; /**< Virtual address of segment buffer. */
741 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
743 uint16_t buf_len; /**< Length of segment buffer. */
745 /* next 6 bytes are initialised on RX descriptor rearm */
750 * 16-bit Reference counter.
751 * It should only be accessed using the following functions:
752 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
753 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
754 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
758 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
759 uint16_t refcnt; /**< Non-atomically accessed refcnt */
761 uint8_t nb_segs; /**< Number of segments. */
762 uint8_t port; /**< Input port. */
764 uint64_t ol_flags; /**< Offload features. */
766 /* remaining bytes are set on RX when pulling packet from descriptor */
767 MARKER rx_descriptor_fields1;
770 * The packet type, which is the combination of outer/inner L2, L3, L4
774 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
776 uint32_t l2_type:4; /**< (Outer) L2 type. */
777 uint32_t l3_type:4; /**< (Outer) L3 type. */
778 uint32_t l4_type:4; /**< (Outer) L4 type. */
779 uint32_t tun_type:4; /**< Tunnel type. */
780 uint32_t inner_l2_type:4; /**< Inner L2 type. */
781 uint32_t inner_l3_type:4; /**< Inner L3 type. */
782 uint32_t inner_l4_type:4; /**< Inner L4 type. */
786 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
787 uint16_t data_len; /**< Amount of data in segment buffer. */
788 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
791 uint32_t rss; /**< RSS hash result if RSS enabled */
799 /**< Second 4 flexible bytes */
802 /**< First 4 flexible bytes or FD ID, dependent on
803 PKT_RX_FDIR_* flag in ol_flags. */
804 } fdir; /**< Filter identifier if FDIR enabled */
808 } sched; /**< Hierarchical scheduler */
809 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
810 } hash; /**< hash information */
812 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
814 uint16_t vlan_tci_outer; /**< Outer VLAN Tag Control Identifier (CPU order) */
816 /* second cache line - fields only used in slow path or on TX */
817 MARKER cacheline1 __rte_cache_min_aligned;
820 void *userdata; /**< Can be used for external metadata */
821 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
824 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
825 struct rte_mbuf *next; /**< Next segment of scattered packet. */
827 /* fields to support TX offloads */
829 uint64_t tx_offload; /**< combined for easy fetch */
831 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
832 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
833 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
834 uint64_t tso_segsz:16; /**< TCP TSO segment size */
836 /* fields for TX offloading of tunnels */
837 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
838 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
840 /* uint64_t unused:8; */
844 /** Size of the application private data. In case of an indirect
845 * mbuf, it stores the direct mbuf private data size. */
848 /** Timesync flags for use with IEEE1588. */
851 /* Chain of off-load operations to perform on mbuf */
852 struct rte_mbuf_offload *offload_ops;
853 } __rte_cache_aligned;
855 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
858 * Return the DMA address of the beginning of the mbuf data
861 * The pointer to the mbuf.
863 * The physical address of the beginning of the mbuf data
865 static inline phys_addr_t
866 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
868 return mb->buf_physaddr + mb->data_off;
872 * Return the default DMA address of the beginning of the mbuf data
874 * This function is used by drivers in their receive function, as it
875 * returns the location where data should be written by the NIC, taking
876 * the default headroom in account.
879 * The pointer to the mbuf.
881 * The physical address of the beginning of the mbuf data
883 static inline phys_addr_t
884 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
886 return mb->buf_physaddr + RTE_PKTMBUF_HEADROOM;
890 * Return the mbuf owning the data buffer address of an indirect mbuf.
893 * The pointer to the indirect mbuf.
895 * The address of the direct mbuf corresponding to buffer_addr.
897 static inline struct rte_mbuf *
898 rte_mbuf_from_indirect(struct rte_mbuf *mi)
900 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
904 * Return the buffer address embedded in the given mbuf.
907 * The pointer to the mbuf.
909 * The address of the data buffer owned by the mbuf.
912 rte_mbuf_to_baddr(struct rte_mbuf *md)
915 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
920 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
922 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
925 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
927 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
930 * Private data in case of pktmbuf pool.
932 * A structure that contains some pktmbuf_pool-specific data that are
933 * appended after the mempool structure (in private data).
935 struct rte_pktmbuf_pool_private {
936 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
937 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
940 #ifdef RTE_LIBRTE_MBUF_DEBUG
942 /** check mbuf type in debug mode */
943 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
945 /** check mbuf type in debug mode if mbuf pointer is not null */
946 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
948 rte_mbuf_sanity_check(m, is_h); \
951 /** MBUF asserts in debug mode */
952 #define RTE_MBUF_ASSERT(exp) \
954 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
957 #else /* RTE_LIBRTE_MBUF_DEBUG */
959 /** check mbuf type in debug mode */
960 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
962 /** check mbuf type in debug mode if mbuf pointer is not null */
963 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
965 /** MBUF asserts in debug mode */
966 #define RTE_MBUF_ASSERT(exp) do { } while (0)
968 #endif /* RTE_LIBRTE_MBUF_DEBUG */
970 #ifdef RTE_MBUF_REFCNT_ATOMIC
973 * Reads the value of an mbuf's refcnt.
977 * Reference count number.
979 static inline uint16_t
980 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
982 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
986 * Sets an mbuf's refcnt to a defined value.
993 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
995 rte_atomic16_set(&m->refcnt_atomic, new_value);
999 * Adds given value to an mbuf's refcnt and returns its new value.
1003 * Value to add/subtract
1007 static inline uint16_t
1008 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1011 * The atomic_add is an expensive operation, so we don't want to
1012 * call it in the case where we know we are the uniq holder of
1013 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
1014 * operation has to be used because concurrent accesses on the
1015 * reference counter can occur.
1017 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1018 rte_mbuf_refcnt_set(m, 1 + value);
1022 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
1025 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
1028 * Adds given value to an mbuf's refcnt and returns its new value.
1030 static inline uint16_t
1031 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1033 m->refcnt = (uint16_t)(m->refcnt + value);
1038 * Reads the value of an mbuf's refcnt.
1040 static inline uint16_t
1041 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1047 * Sets an mbuf's refcnt to the defined value.
1050 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1052 m->refcnt = new_value;
1055 #endif /* RTE_MBUF_REFCNT_ATOMIC */
1057 /** Mbuf prefetch */
1058 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
1065 * Sanity checks on an mbuf.
1067 * Check the consistency of the given mbuf. The function will cause a
1068 * panic if corruption is detected.
1071 * The mbuf to be checked.
1073 * True if the mbuf is a packet header, false if it is a sub-segment
1074 * of a packet (in this case, some fields like nb_segs are not checked)
1077 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
1080 * @internal Allocate a new mbuf from mempool *mp*.
1081 * The use of that function is reserved for RTE internal needs.
1082 * Please use rte_pktmbuf_alloc().
1085 * The mempool from which mbuf is allocated.
1087 * - The pointer to the new mbuf on success.
1088 * - NULL if allocation failed.
1090 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
1094 if (rte_mempool_get(mp, &mb) < 0)
1096 m = (struct rte_mbuf *)mb;
1097 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1098 rte_mbuf_refcnt_set(m, 1);
1103 * @internal Put mbuf back into its original mempool.
1104 * The use of that function is reserved for RTE internal needs.
1105 * Please use rte_pktmbuf_free().
1108 * The mbuf to be freed.
1110 static inline void __attribute__((always_inline))
1111 __rte_mbuf_raw_free(struct rte_mbuf *m)
1113 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1114 rte_mempool_put(m->pool, m);
1117 /* Operations on ctrl mbuf */
1120 * The control mbuf constructor.
1122 * This function initializes some fields in an mbuf structure that are
1123 * not modified by the user once created (mbuf type, origin pool, buffer
1124 * start address, and so on). This function is given as a callback function
1125 * to rte_mempool_create() at pool creation time.
1128 * The mempool from which the mbuf is allocated.
1130 * A pointer that can be used by the user to retrieve useful information
1131 * for mbuf initialization. This pointer comes from the ``init_arg``
1132 * parameter of rte_mempool_create().
1134 * The mbuf to initialize.
1136 * The index of the mbuf in the pool table.
1138 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1139 void *m, unsigned i);
1142 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
1144 * This new mbuf is initialized with data pointing to the beginning of
1145 * buffer, and with a length of zero.
1148 * The mempool from which the mbuf is allocated.
1150 * - The pointer to the new mbuf on success.
1151 * - NULL if allocation failed.
1153 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
1156 * Free a control mbuf back into its original mempool.
1159 * The control mbuf to be freed.
1161 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
1164 * A macro that returns the pointer to the carried data.
1166 * The value that can be read or assigned.
1171 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
1174 * A macro that returns the length of the carried data.
1176 * The value that can be read or assigned.
1181 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
1184 * Tests if an mbuf is a control mbuf
1187 * The mbuf to be tested
1189 * - True (1) if the mbuf is a control mbuf
1190 * - False(0) otherwise
1193 rte_is_ctrlmbuf(struct rte_mbuf *m)
1195 return !!(m->ol_flags & CTRL_MBUF_FLAG);
1198 /* Operations on pkt mbuf */
1201 * The packet mbuf constructor.
1203 * This function initializes some fields in the mbuf structure that are
1204 * not modified by the user once created (origin pool, buffer start
1205 * address, and so on). This function is given as a callback function to
1206 * rte_mempool_create() at pool creation time.
1209 * The mempool from which mbufs originate.
1211 * A pointer that can be used by the user to retrieve useful information
1212 * for mbuf initialization. This pointer comes from the ``init_arg``
1213 * parameter of rte_mempool_create().
1215 * The mbuf to initialize.
1217 * The index of the mbuf in the pool table.
1219 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1220 void *m, unsigned i);
1224 * A packet mbuf pool constructor.
1226 * This function initializes the mempool private data in the case of a
1227 * pktmbuf pool. This private data is needed by the driver. The
1228 * function is given as a callback function to rte_mempool_create() at
1229 * pool creation. It can be extended by the user, for example, to
1230 * provide another packet size.
1233 * The mempool from which mbufs originate.
1235 * A pointer that can be used by the user to retrieve useful information
1236 * for mbuf initialization. This pointer comes from the ``init_arg``
1237 * parameter of rte_mempool_create().
1239 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1242 * Create a mbuf pool.
1244 * This function creates and initializes a packet mbuf pool. It is
1245 * a wrapper to rte_mempool_create() with the proper packet constructor
1246 * and mempool constructor.
1249 * The name of the mbuf pool.
1251 * The number of elements in the mbuf pool. The optimum size (in terms
1252 * of memory usage) for a mempool is when n is a power of two minus one:
1255 * Size of the per-core object cache. See rte_mempool_create() for
1258 * Size of application private are between the rte_mbuf structure
1259 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1260 * @param data_room_size
1261 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1263 * The socket identifier where the memory should be allocated. The
1264 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1267 * The pointer to the new allocated mempool, on success. NULL on error
1268 * with rte_errno set appropriately. Possible rte_errno values include:
1269 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1270 * - E_RTE_SECONDARY - function was called from a secondary process instance
1271 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1272 * - ENOSPC - the maximum number of memzones has already been allocated
1273 * - EEXIST - a memzone with the same name already exists
1274 * - ENOMEM - no appropriate memory area found in which to create memzone
1276 struct rte_mempool *
1277 rte_pktmbuf_pool_create(const char *name, unsigned n,
1278 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1282 * Get the data room size of mbufs stored in a pktmbuf_pool
1284 * The data room size is the amount of data that can be stored in a
1285 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1288 * The packet mbuf pool.
1290 * The data room size of mbufs stored in this mempool.
1292 static inline uint16_t
1293 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1295 struct rte_pktmbuf_pool_private *mbp_priv;
1297 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1298 return mbp_priv->mbuf_data_room_size;
1302 * Get the application private size of mbufs stored in a pktmbuf_pool
1304 * The private size of mbuf is a zone located between the rte_mbuf
1305 * structure and the data buffer where an application can store data
1306 * associated to a packet.
1309 * The packet mbuf pool.
1311 * The private size of mbufs stored in this mempool.
1313 static inline uint16_t
1314 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1316 struct rte_pktmbuf_pool_private *mbp_priv;
1318 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1319 return mbp_priv->mbuf_priv_size;
1323 * Reset the fields of a packet mbuf to their default values.
1325 * The given mbuf must have only one segment.
1328 * The packet mbuf to be resetted.
1330 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1336 m->vlan_tci_outer = 0;
1342 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
1343 RTE_PKTMBUF_HEADROOM : m->buf_len;
1346 __rte_mbuf_sanity_check(m, 1);
1350 * Allocate a new mbuf from a mempool.
1352 * This new mbuf contains one segment, which has a length of 0. The pointer
1353 * to data is initialized to have some bytes of headroom in the buffer
1354 * (if buffer size allows).
1357 * The mempool from which the mbuf is allocated.
1359 * - The pointer to the new mbuf on success.
1360 * - NULL if allocation failed.
1362 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1365 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
1366 rte_pktmbuf_reset(m);
1371 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1375 * The mempool from which mbufs are allocated.
1377 * Array of pointers to mbufs
1383 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1384 struct rte_mbuf **mbufs, unsigned count)
1389 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1393 /* To understand duff's device on loop unwinding optimization, see
1394 * https://en.wikipedia.org/wiki/Duff's_device.
1395 * Here while() loop is used rather than do() while{} to avoid extra
1396 * check if count is zero.
1398 switch (count % 4) {
1400 while (idx != count) {
1401 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1402 rte_mbuf_refcnt_set(mbufs[idx], 1);
1403 rte_pktmbuf_reset(mbufs[idx]);
1406 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1407 rte_mbuf_refcnt_set(mbufs[idx], 1);
1408 rte_pktmbuf_reset(mbufs[idx]);
1411 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1412 rte_mbuf_refcnt_set(mbufs[idx], 1);
1413 rte_pktmbuf_reset(mbufs[idx]);
1416 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1417 rte_mbuf_refcnt_set(mbufs[idx], 1);
1418 rte_pktmbuf_reset(mbufs[idx]);
1426 * Attach packet mbuf to another packet mbuf.
1428 * After attachment we refer the mbuf we attached as 'indirect',
1429 * while mbuf we attached to as 'direct'.
1430 * Right now, not supported:
1431 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1432 * - mbuf we trying to attach (mi) is used by someone else
1433 * e.g. it's reference counter is greater then 1.
1436 * The indirect packet mbuf.
1438 * The packet mbuf we're attaching to.
1440 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1442 struct rte_mbuf *md;
1444 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(mi) &&
1445 rte_mbuf_refcnt_read(mi) == 1);
1447 /* if m is not direct, get the mbuf that embeds the data */
1448 if (RTE_MBUF_DIRECT(m))
1451 md = rte_mbuf_from_indirect(m);
1453 rte_mbuf_refcnt_update(md, 1);
1454 mi->priv_size = m->priv_size;
1455 mi->buf_physaddr = m->buf_physaddr;
1456 mi->buf_addr = m->buf_addr;
1457 mi->buf_len = m->buf_len;
1460 mi->data_off = m->data_off;
1461 mi->data_len = m->data_len;
1463 mi->vlan_tci = m->vlan_tci;
1464 mi->vlan_tci_outer = m->vlan_tci_outer;
1465 mi->tx_offload = m->tx_offload;
1469 mi->pkt_len = mi->data_len;
1471 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1472 mi->packet_type = m->packet_type;
1474 __rte_mbuf_sanity_check(mi, 1);
1475 __rte_mbuf_sanity_check(m, 0);
1479 * Detach an indirect packet mbuf.
1481 * - restore original mbuf address and length values.
1482 * - reset pktmbuf data and data_len to their default values.
1483 * All other fields of the given packet mbuf will be left intact.
1486 * The indirect attached packet mbuf.
1488 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1490 struct rte_mempool *mp = m->pool;
1491 uint32_t mbuf_size, buf_len, priv_size;
1493 priv_size = rte_pktmbuf_priv_size(mp);
1494 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1495 buf_len = rte_pktmbuf_data_room_size(mp);
1497 m->priv_size = priv_size;
1498 m->buf_addr = (char *)m + mbuf_size;
1499 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
1500 m->buf_len = (uint16_t)buf_len;
1501 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1506 static inline struct rte_mbuf* __attribute__((always_inline))
1507 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1509 __rte_mbuf_sanity_check(m, 0);
1511 if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
1513 /* if this is an indirect mbuf, then
1515 * - free attached mbuf segment
1517 if (RTE_MBUF_INDIRECT(m)) {
1518 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1519 rte_pktmbuf_detach(m);
1520 if (rte_mbuf_refcnt_update(md, -1) == 0)
1521 __rte_mbuf_raw_free(md);
1529 * Free a segment of a packet mbuf into its original mempool.
1531 * Free an mbuf, without parsing other segments in case of chained
1535 * The packet mbuf segment to be freed.
1537 static inline void __attribute__((always_inline))
1538 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1540 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
1542 __rte_mbuf_raw_free(m);
1547 * Free a packet mbuf back into its original mempool.
1549 * Free an mbuf, and all its segments in case of chained buffers. Each
1550 * segment is added back into its original mempool.
1553 * The packet mbuf to be freed.
1555 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1557 struct rte_mbuf *m_next;
1559 __rte_mbuf_sanity_check(m, 1);
1563 rte_pktmbuf_free_seg(m);
1569 * Creates a "clone" of the given packet mbuf.
1571 * Walks through all segments of the given packet mbuf, and for each of them:
1572 * - Creates a new packet mbuf from the given pool.
1573 * - Attaches newly created mbuf to the segment.
1574 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1575 * from the original packet mbuf.
1578 * The packet mbuf to be cloned.
1580 * The mempool from which the "clone" mbufs are allocated.
1582 * - The pointer to the new "clone" mbuf on success.
1583 * - NULL if allocation fails.
1585 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1586 struct rte_mempool *mp)
1588 struct rte_mbuf *mc, *mi, **prev;
1592 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1597 pktlen = md->pkt_len;
1602 rte_pktmbuf_attach(mi, md);
1605 } while ((md = md->next) != NULL &&
1606 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1610 mc->pkt_len = pktlen;
1612 /* Allocation of new indirect segment failed */
1613 if (unlikely (mi == NULL)) {
1614 rte_pktmbuf_free(mc);
1618 __rte_mbuf_sanity_check(mc, 1);
1623 * Adds given value to the refcnt of all packet mbuf segments.
1625 * Walks through all segments of given packet mbuf and for each of them
1626 * invokes rte_mbuf_refcnt_update().
1629 * The packet mbuf whose refcnt to be updated.
1631 * The value to add to the mbuf's segments refcnt.
1633 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1635 __rte_mbuf_sanity_check(m, 1);
1638 rte_mbuf_refcnt_update(m, v);
1639 } while ((m = m->next) != NULL);
1643 * Get the headroom in a packet mbuf.
1648 * The length of the headroom.
1650 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1652 __rte_mbuf_sanity_check(m, 1);
1657 * Get the tailroom of a packet mbuf.
1662 * The length of the tailroom.
1664 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1666 __rte_mbuf_sanity_check(m, 1);
1667 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1672 * Get the last segment of the packet.
1677 * The last segment of the given mbuf.
1679 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1681 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1683 __rte_mbuf_sanity_check(m, 1);
1684 while (m2->next != NULL)
1690 * A macro that points to an offset into the data in the mbuf.
1692 * The returned pointer is cast to type t. Before using this
1693 * function, the user must ensure that the first segment is large
1694 * enough to accommodate its data.
1699 * The offset into the mbuf data.
1701 * The type to cast the result into.
1703 #define rte_pktmbuf_mtod_offset(m, t, o) \
1704 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1707 * A macro that points to the start of the data in the mbuf.
1709 * The returned pointer is cast to type t. Before using this
1710 * function, the user must ensure that the first segment is large
1711 * enough to accommodate its data.
1716 * The type to cast the result into.
1718 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1721 * A macro that returns the physical address that points to an offset of the
1722 * start of the data in the mbuf
1727 * The offset into the data to calculate address from.
1729 #define rte_pktmbuf_mtophys_offset(m, o) \
1730 (phys_addr_t)((m)->buf_physaddr + (m)->data_off + (o))
1733 * A macro that returns the physical address that points to the start of the
1739 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_mtophys_offset(m, 0)
1742 * A macro that returns the length of the packet.
1744 * The value can be read or assigned.
1749 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1752 * A macro that returns the length of the segment.
1754 * The value can be read or assigned.
1759 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1762 * Prepend len bytes to an mbuf data area.
1764 * Returns a pointer to the new
1765 * data start address. If there is not enough headroom in the first
1766 * segment, the function will return NULL, without modifying the mbuf.
1771 * The amount of data to prepend (in bytes).
1773 * A pointer to the start of the newly prepended data, or
1774 * NULL if there is not enough headroom space in the first segment
1776 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1779 __rte_mbuf_sanity_check(m, 1);
1781 if (unlikely(len > rte_pktmbuf_headroom(m)))
1785 m->data_len = (uint16_t)(m->data_len + len);
1786 m->pkt_len = (m->pkt_len + len);
1788 return (char *)m->buf_addr + m->data_off;
1792 * Append len bytes to an mbuf.
1794 * Append len bytes to an mbuf and return a pointer to the start address
1795 * of the added data. If there is not enough tailroom in the last
1796 * segment, the function will return NULL, without modifying the mbuf.
1801 * The amount of data to append (in bytes).
1803 * A pointer to the start of the newly appended data, or
1804 * NULL if there is not enough tailroom space in the last segment
1806 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1809 struct rte_mbuf *m_last;
1811 __rte_mbuf_sanity_check(m, 1);
1813 m_last = rte_pktmbuf_lastseg(m);
1814 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1817 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1818 m_last->data_len = (uint16_t)(m_last->data_len + len);
1819 m->pkt_len = (m->pkt_len + len);
1820 return (char*) tail;
1824 * Remove len bytes at the beginning of an mbuf.
1826 * Returns a pointer to the start address of the new data area. If the
1827 * length is greater than the length of the first segment, then the
1828 * function will fail and return NULL, without modifying the mbuf.
1833 * The amount of data to remove (in bytes).
1835 * A pointer to the new start of the data.
1837 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1839 __rte_mbuf_sanity_check(m, 1);
1841 if (unlikely(len > m->data_len))
1844 m->data_len = (uint16_t)(m->data_len - len);
1846 m->pkt_len = (m->pkt_len - len);
1847 return (char *)m->buf_addr + m->data_off;
1851 * Remove len bytes of data at the end of the mbuf.
1853 * If the length is greater than the length of the last segment, the
1854 * function will fail and return -1 without modifying the mbuf.
1859 * The amount of data to remove (in bytes).
1864 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1866 struct rte_mbuf *m_last;
1868 __rte_mbuf_sanity_check(m, 1);
1870 m_last = rte_pktmbuf_lastseg(m);
1871 if (unlikely(len > m_last->data_len))
1874 m_last->data_len = (uint16_t)(m_last->data_len - len);
1875 m->pkt_len = (m->pkt_len - len);
1880 * Test if mbuf data is contiguous.
1885 * - 1, if all data is contiguous (one segment).
1886 * - 0, if there is several segments.
1888 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1890 __rte_mbuf_sanity_check(m, 1);
1891 return !!(m->nb_segs == 1);
1895 * Chain an mbuf to another, thereby creating a segmented packet.
1897 * Note: The implementation will do a linear walk over the segments to find
1898 * the tail entry. For cases when there are many segments, it's better to
1899 * chain the entries manually.
1902 * The head of the mbuf chain (the first packet)
1904 * The mbuf to put last in the chain
1908 * - -EOVERFLOW, if the chain is full (256 entries)
1910 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1912 struct rte_mbuf *cur_tail;
1914 /* Check for number-of-segments-overflow */
1915 if (head->nb_segs + tail->nb_segs >= 1 << (sizeof(head->nb_segs) * 8))
1918 /* Chain 'tail' onto the old tail */
1919 cur_tail = rte_pktmbuf_lastseg(head);
1920 cur_tail->next = tail;
1922 /* accumulate number of segments and total length. */
1923 head->nb_segs = (uint8_t)(head->nb_segs + tail->nb_segs);
1924 head->pkt_len += tail->pkt_len;
1926 /* pkt_len is only set in the head */
1927 tail->pkt_len = tail->data_len;
1933 * Dump an mbuf structure to the console.
1935 * Dump all fields for the given packet mbuf and all its associated
1936 * segments (in the case of a chained buffer).
1939 * A pointer to a file for output
1943 * If dump_len != 0, also dump the "dump_len" first data bytes of
1946 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1952 #endif /* _RTE_MBUF_H_ */