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>
69 * Packet Offload Features Flags. It also carry packet type information.
70 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
72 * - RX flags start at bit position zero, and get added to the left of previous
74 * - The most-significant 3 bits are reserved for generic mbuf flags
75 * - TX flags therefore start at bit position 60 (i.e. 63-3), and new flags get
76 * added to the right of the previously defined flags i.e. they should count
77 * downwards, not upwards.
79 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
80 * rte_get_tx_ol_flag_name().
84 * RX packet is a 802.1q VLAN packet. This flag was set by PMDs when
85 * the packet is recognized as a VLAN, but the behavior between PMDs
86 * was not the same. This flag is kept for some time to avoid breaking
87 * applications and should be replaced by PKT_RX_VLAN_STRIPPED.
89 #define PKT_RX_VLAN_PKT (1ULL << 0)
91 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
92 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
93 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
94 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
95 #define PKT_RX_EIP_CKSUM_BAD (1ULL << 5) /**< External IP header checksum error. */
98 * A vlan has been stripped by the hardware and its tci is saved in
99 * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
100 * in the RX configuration of the PMD.
102 #define PKT_RX_VLAN_STRIPPED (1ULL << 6)
104 /* hole, some bits can be reused here */
106 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
107 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
108 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
109 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
112 * The 2 vlans have been stripped by the hardware and their tci are
113 * saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
114 * This can only happen if vlan stripping is enabled in the RX
115 * configuration of the PMD. If this flag is set, PKT_RX_VLAN_STRIPPED
118 #define PKT_RX_QINQ_STRIPPED (1ULL << 15)
122 * RX packet with double VLAN stripped.
123 * This flag is replaced by PKT_RX_QINQ_STRIPPED.
125 #define PKT_RX_QINQ_PKT PKT_RX_QINQ_STRIPPED
127 /* add new RX flags here */
129 /* add new TX flags here */
132 * Bits 45:48 used for the tunnel type.
133 * When doing Tx offload like TSO or checksum, the HW needs to configure the
134 * tunnel type into the HW descriptors.
136 #define PKT_TX_TUNNEL_VXLAN (0x1ULL << 45)
137 #define PKT_TX_TUNNEL_GRE (0x2ULL << 45)
138 #define PKT_TX_TUNNEL_IPIP (0x3ULL << 45)
139 #define PKT_TX_TUNNEL_GENEVE (0x4ULL << 45)
140 /* add new TX TUNNEL type here */
141 #define PKT_TX_TUNNEL_MASK (0xFULL << 45)
144 * Second VLAN insertion (QinQ) flag.
146 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
149 * TCP segmentation offload. To enable this offload feature for a
150 * packet to be transmitted on hardware supporting TSO:
151 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
153 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
154 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
156 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
157 * - calculate the pseudo header checksum without taking ip_len in account,
158 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
159 * rte_ipv6_phdr_cksum() that can be used as helpers.
161 #define PKT_TX_TCP_SEG (1ULL << 50)
163 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
166 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
167 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
168 * L4 checksum offload, the user needs to:
169 * - fill l2_len and l3_len in mbuf
170 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
171 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
172 * - calculate the pseudo header checksum and set it in the L4 header (only
173 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
174 * For SCTP, set the crc field to 0.
176 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
177 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
178 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
179 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
180 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
183 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
184 * also be set by the application, although a PMD will only check
186 * - set the IP checksum field in the packet to 0
187 * - fill the mbuf offload information: l2_len, l3_len
189 #define PKT_TX_IP_CKSUM (1ULL << 54)
192 * Packet is IPv4. This flag must be set when using any offload feature
193 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
194 * packet. If the packet is a tunneled packet, this flag is related to
197 #define PKT_TX_IPV4 (1ULL << 55)
200 * Packet is IPv6. This flag must be set when using an offload feature
201 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
202 * packet. If the packet is a tunneled packet, this flag is related to
205 #define PKT_TX_IPV6 (1ULL << 56)
207 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
210 * Offload the IP checksum of an external header in the hardware. The
211 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
212 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
213 * packet must be set to 0.
214 * - set the outer IP checksum field in the packet to 0
215 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
217 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
220 * Packet outer header is IPv4. This flag must be set when using any
221 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
222 * outer header of the tunneled packet is an IPv4 packet.
224 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
227 * Packet outer header is IPv6. This flag must be set when using any
228 * outer offload feature (L4 checksum) to tell the NIC that the outer
229 * header of the tunneled packet is an IPv6 packet.
231 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
233 #define __RESERVED (1ULL << 61) /**< reserved for future mbuf use */
235 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
237 /* Use final bit of flags to indicate a control mbuf */
238 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
241 * 32 bits are divided into several fields to mark packet types. Note that
242 * each field is indexical.
243 * - Bit 3:0 is for L2 types.
244 * - Bit 7:4 is for L3 or outer L3 (for tunneling case) types.
245 * - Bit 11:8 is for L4 or outer L4 (for tunneling case) types.
246 * - Bit 15:12 is for tunnel types.
247 * - Bit 19:16 is for inner L2 types.
248 * - Bit 23:20 is for inner L3 types.
249 * - Bit 27:24 is for inner L4 types.
250 * - Bit 31:28 is reserved.
252 * To be compatible with Vector PMD, RTE_PTYPE_L3_IPV4, RTE_PTYPE_L3_IPV4_EXT,
253 * RTE_PTYPE_L3_IPV6, RTE_PTYPE_L3_IPV6_EXT, RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP
254 * and RTE_PTYPE_L4_SCTP should be kept as below in a contiguous 7 bits.
256 * Note that L3 types values are selected for checking IPV4/IPV6 header from
257 * performance point of view. Reading annotations of RTE_ETH_IS_IPV4_HDR and
258 * RTE_ETH_IS_IPV6_HDR is needed for any future changes of L3 type values.
260 * Note that the packet types of the same packet recognized by different
261 * hardware may be different, as different hardware may have different
262 * capability of packet type recognition.
265 * <'ether type'=0x0800
266 * | 'version'=4, 'protocol'=0x29
267 * | 'version'=6, 'next header'=0x3A
269 * will be recognized on i40e hardware as packet type combination of,
270 * RTE_PTYPE_L2_ETHER |
271 * RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
272 * RTE_PTYPE_TUNNEL_IP |
273 * RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
274 * RTE_PTYPE_INNER_L4_ICMP.
276 * <'ether type'=0x86DD
277 * | 'version'=6, 'next header'=0x2F
279 * | 'version'=6, 'next header'=0x11
281 * will be recognized on i40e hardware as packet type combination of,
282 * RTE_PTYPE_L2_ETHER |
283 * RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
284 * RTE_PTYPE_TUNNEL_GRENAT |
285 * RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
286 * RTE_PTYPE_INNER_L4_UDP.
288 #define RTE_PTYPE_UNKNOWN 0x00000000
290 * Ethernet packet type.
291 * It is used for outer packet for tunneling cases.
294 * <'ether type'=[0x0800|0x86DD]>
296 #define RTE_PTYPE_L2_ETHER 0x00000001
298 * Ethernet packet type for time sync.
301 * <'ether type'=0x88F7>
303 #define RTE_PTYPE_L2_ETHER_TIMESYNC 0x00000002
305 * ARP (Address Resolution Protocol) packet type.
308 * <'ether type'=0x0806>
310 #define RTE_PTYPE_L2_ETHER_ARP 0x00000003
312 * LLDP (Link Layer Discovery Protocol) packet type.
315 * <'ether type'=0x88CC>
317 #define RTE_PTYPE_L2_ETHER_LLDP 0x00000004
319 * NSH (Network Service Header) packet type.
322 * <'ether type'=0x894F>
324 #define RTE_PTYPE_L2_ETHER_NSH 0x00000005
326 * Mask of layer 2 packet types.
327 * It is used for outer packet for tunneling cases.
329 #define RTE_PTYPE_L2_MASK 0x0000000f
331 * IP (Internet Protocol) version 4 packet type.
332 * It is used for outer packet for tunneling cases, and does not contain any
336 * <'ether type'=0x0800
337 * | 'version'=4, 'ihl'=5>
339 #define RTE_PTYPE_L3_IPV4 0x00000010
341 * IP (Internet Protocol) version 4 packet type.
342 * It is used for outer packet for tunneling cases, and contains header
346 * <'ether type'=0x0800
347 * | 'version'=4, 'ihl'=[6-15], 'options'>
349 #define RTE_PTYPE_L3_IPV4_EXT 0x00000030
351 * IP (Internet Protocol) version 6 packet type.
352 * It is used for outer packet for tunneling cases, and does not contain any
356 * <'ether type'=0x86DD
357 * | 'version'=6, 'next header'=0x3B>
359 #define RTE_PTYPE_L3_IPV6 0x00000040
361 * IP (Internet Protocol) version 4 packet type.
362 * It is used for outer packet for tunneling cases, and may or maynot contain
366 * <'ether type'=0x0800
367 * | 'version'=4, 'ihl'=[5-15], <'options'>>
369 #define RTE_PTYPE_L3_IPV4_EXT_UNKNOWN 0x00000090
371 * IP (Internet Protocol) version 6 packet type.
372 * It is used for outer packet for tunneling cases, and contains extension
376 * <'ether type'=0x86DD
377 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
378 * 'extension headers'>
380 #define RTE_PTYPE_L3_IPV6_EXT 0x000000c0
382 * IP (Internet Protocol) version 6 packet type.
383 * It is used for outer packet for tunneling cases, and may or maynot contain
387 * <'ether type'=0x86DD
388 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
389 * <'extension headers'>>
391 #define RTE_PTYPE_L3_IPV6_EXT_UNKNOWN 0x000000e0
393 * Mask of layer 3 packet types.
394 * It is used for outer packet for tunneling cases.
396 #define RTE_PTYPE_L3_MASK 0x000000f0
398 * TCP (Transmission Control Protocol) packet type.
399 * It is used for outer packet for tunneling cases.
402 * <'ether type'=0x0800
403 * | 'version'=4, 'protocol'=6, 'MF'=0>
405 * <'ether type'=0x86DD
406 * | 'version'=6, 'next header'=6>
408 #define RTE_PTYPE_L4_TCP 0x00000100
410 * UDP (User Datagram Protocol) packet type.
411 * It is used for outer packet for tunneling cases.
414 * <'ether type'=0x0800
415 * | 'version'=4, 'protocol'=17, 'MF'=0>
417 * <'ether type'=0x86DD
418 * | 'version'=6, 'next header'=17>
420 #define RTE_PTYPE_L4_UDP 0x00000200
422 * Fragmented IP (Internet Protocol) packet type.
423 * It is used for outer packet for tunneling cases.
425 * It refers to those packets of any IP types, which can be recognized as
426 * fragmented. A fragmented packet cannot be recognized as any other L4 types
427 * (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP,
428 * RTE_PTYPE_L4_NONFRAG).
431 * <'ether type'=0x0800
432 * | 'version'=4, 'MF'=1>
434 * <'ether type'=0x86DD
435 * | 'version'=6, 'next header'=44>
437 #define RTE_PTYPE_L4_FRAG 0x00000300
439 * SCTP (Stream Control Transmission Protocol) packet type.
440 * It is used for outer packet for tunneling cases.
443 * <'ether type'=0x0800
444 * | 'version'=4, 'protocol'=132, 'MF'=0>
446 * <'ether type'=0x86DD
447 * | 'version'=6, 'next header'=132>
449 #define RTE_PTYPE_L4_SCTP 0x00000400
451 * ICMP (Internet Control Message Protocol) packet type.
452 * It is used for outer packet for tunneling cases.
455 * <'ether type'=0x0800
456 * | 'version'=4, 'protocol'=1, 'MF'=0>
458 * <'ether type'=0x86DD
459 * | 'version'=6, 'next header'=1>
461 #define RTE_PTYPE_L4_ICMP 0x00000500
463 * Non-fragmented IP (Internet Protocol) packet type.
464 * It is used for outer packet for tunneling cases.
466 * It refers to those packets of any IP types, while cannot be recognized as
467 * any of above L4 types (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP,
468 * RTE_PTYPE_L4_FRAG, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP).
471 * <'ether type'=0x0800
472 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
474 * <'ether type'=0x86DD
475 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
477 #define RTE_PTYPE_L4_NONFRAG 0x00000600
479 * Mask of layer 4 packet types.
480 * It is used for outer packet for tunneling cases.
482 #define RTE_PTYPE_L4_MASK 0x00000f00
484 * IP (Internet Protocol) in IP (Internet Protocol) tunneling packet type.
487 * <'ether type'=0x0800
488 * | 'version'=4, 'protocol'=[4|41]>
490 * <'ether type'=0x86DD
491 * | 'version'=6, 'next header'=[4|41]>
493 #define RTE_PTYPE_TUNNEL_IP 0x00001000
495 * GRE (Generic Routing Encapsulation) tunneling packet type.
498 * <'ether type'=0x0800
499 * | 'version'=4, 'protocol'=47>
501 * <'ether type'=0x86DD
502 * | 'version'=6, 'next header'=47>
504 #define RTE_PTYPE_TUNNEL_GRE 0x00002000
506 * VXLAN (Virtual eXtensible Local Area Network) tunneling packet type.
509 * <'ether type'=0x0800
510 * | 'version'=4, 'protocol'=17
511 * | 'destination port'=4798>
513 * <'ether type'=0x86DD
514 * | 'version'=6, 'next header'=17
515 * | 'destination port'=4798>
517 #define RTE_PTYPE_TUNNEL_VXLAN 0x00003000
519 * NVGRE (Network Virtualization using Generic Routing Encapsulation) tunneling
523 * <'ether type'=0x0800
524 * | 'version'=4, 'protocol'=47
525 * | 'protocol type'=0x6558>
527 * <'ether type'=0x86DD
528 * | 'version'=6, 'next header'=47
529 * | 'protocol type'=0x6558'>
531 #define RTE_PTYPE_TUNNEL_NVGRE 0x00004000
533 * GENEVE (Generic Network Virtualization Encapsulation) tunneling packet type.
536 * <'ether type'=0x0800
537 * | 'version'=4, 'protocol'=17
538 * | 'destination port'=6081>
540 * <'ether type'=0x86DD
541 * | 'version'=6, 'next header'=17
542 * | 'destination port'=6081>
544 #define RTE_PTYPE_TUNNEL_GENEVE 0x00005000
546 * Tunneling packet type of Teredo, VXLAN (Virtual eXtensible Local Area
547 * Network) or GRE (Generic Routing Encapsulation) could be recognized as this
548 * packet type, if they can not be recognized independently as of hardware
551 #define RTE_PTYPE_TUNNEL_GRENAT 0x00006000
553 * Mask of tunneling packet types.
555 #define RTE_PTYPE_TUNNEL_MASK 0x0000f000
557 * Ethernet packet type.
558 * It is used for inner packet type only.
560 * Packet format (inner only):
561 * <'ether type'=[0x800|0x86DD]>
563 #define RTE_PTYPE_INNER_L2_ETHER 0x00010000
565 * Ethernet packet type with VLAN (Virtual Local Area Network) tag.
567 * Packet format (inner only):
568 * <'ether type'=[0x800|0x86DD], vlan=[1-4095]>
570 #define RTE_PTYPE_INNER_L2_ETHER_VLAN 0x00020000
572 * Mask of inner layer 2 packet types.
574 #define RTE_PTYPE_INNER_L2_MASK 0x000f0000
576 * IP (Internet Protocol) version 4 packet type.
577 * It is used for inner packet only, and does not contain any header option.
579 * Packet format (inner only):
580 * <'ether type'=0x0800
581 * | 'version'=4, 'ihl'=5>
583 #define RTE_PTYPE_INNER_L3_IPV4 0x00100000
585 * IP (Internet Protocol) version 4 packet type.
586 * It is used for inner packet only, and contains header options.
588 * Packet format (inner only):
589 * <'ether type'=0x0800
590 * | 'version'=4, 'ihl'=[6-15], 'options'>
592 #define RTE_PTYPE_INNER_L3_IPV4_EXT 0x00200000
594 * IP (Internet Protocol) version 6 packet type.
595 * It is used for inner packet only, and does not contain any extension header.
597 * Packet format (inner only):
598 * <'ether type'=0x86DD
599 * | 'version'=6, 'next header'=0x3B>
601 #define RTE_PTYPE_INNER_L3_IPV6 0x00300000
603 * IP (Internet Protocol) version 4 packet type.
604 * It is used for inner packet only, and may or maynot contain header options.
606 * Packet format (inner only):
607 * <'ether type'=0x0800
608 * | 'version'=4, 'ihl'=[5-15], <'options'>>
610 #define RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN 0x00400000
612 * IP (Internet Protocol) version 6 packet type.
613 * It is used for inner packet only, and contains extension headers.
615 * Packet format (inner only):
616 * <'ether type'=0x86DD
617 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
618 * 'extension headers'>
620 #define RTE_PTYPE_INNER_L3_IPV6_EXT 0x00500000
622 * IP (Internet Protocol) version 6 packet type.
623 * It is used for inner packet only, and may or maynot contain extension
626 * Packet format (inner only):
627 * <'ether type'=0x86DD
628 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
629 * <'extension headers'>>
631 #define RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN 0x00600000
633 * Mask of inner layer 3 packet types.
635 #define RTE_PTYPE_INNER_L3_MASK 0x00f00000
637 * TCP (Transmission Control Protocol) packet type.
638 * It is used for inner packet only.
640 * Packet format (inner only):
641 * <'ether type'=0x0800
642 * | 'version'=4, 'protocol'=6, 'MF'=0>
644 * <'ether type'=0x86DD
645 * | 'version'=6, 'next header'=6>
647 #define RTE_PTYPE_INNER_L4_TCP 0x01000000
649 * UDP (User Datagram Protocol) packet type.
650 * It is used for inner packet only.
652 * Packet format (inner only):
653 * <'ether type'=0x0800
654 * | 'version'=4, 'protocol'=17, 'MF'=0>
656 * <'ether type'=0x86DD
657 * | 'version'=6, 'next header'=17>
659 #define RTE_PTYPE_INNER_L4_UDP 0x02000000
661 * Fragmented IP (Internet Protocol) packet type.
662 * It is used for inner packet only, and may or maynot have layer 4 packet.
664 * Packet format (inner only):
665 * <'ether type'=0x0800
666 * | 'version'=4, 'MF'=1>
668 * <'ether type'=0x86DD
669 * | 'version'=6, 'next header'=44>
671 #define RTE_PTYPE_INNER_L4_FRAG 0x03000000
673 * SCTP (Stream Control Transmission Protocol) packet type.
674 * It is used for inner packet only.
676 * Packet format (inner only):
677 * <'ether type'=0x0800
678 * | 'version'=4, 'protocol'=132, 'MF'=0>
680 * <'ether type'=0x86DD
681 * | 'version'=6, 'next header'=132>
683 #define RTE_PTYPE_INNER_L4_SCTP 0x04000000
685 * ICMP (Internet Control Message Protocol) packet type.
686 * It is used for inner packet only.
688 * Packet format (inner only):
689 * <'ether type'=0x0800
690 * | 'version'=4, 'protocol'=1, 'MF'=0>
692 * <'ether type'=0x86DD
693 * | 'version'=6, 'next header'=1>
695 #define RTE_PTYPE_INNER_L4_ICMP 0x05000000
697 * Non-fragmented IP (Internet Protocol) packet type.
698 * It is used for inner packet only, and may or maynot have other unknown layer
701 * Packet format (inner only):
702 * <'ether type'=0x0800
703 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
705 * <'ether type'=0x86DD
706 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
708 #define RTE_PTYPE_INNER_L4_NONFRAG 0x06000000
710 * Mask of inner layer 4 packet types.
712 #define RTE_PTYPE_INNER_L4_MASK 0x0f000000
715 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
716 * one, bit 4 is selected to be used for IPv4 only. Then checking bit 4 can
717 * determine if it is an IPV4 packet.
719 #define RTE_ETH_IS_IPV4_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV4)
722 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
723 * one, bit 6 is selected to be used for IPv4 only. Then checking bit 6 can
724 * determine if it is an IPV4 packet.
726 #define RTE_ETH_IS_IPV6_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV6)
728 /* Check if it is a tunneling packet */
729 #define RTE_ETH_IS_TUNNEL_PKT(ptype) ((ptype) & (RTE_PTYPE_TUNNEL_MASK | \
730 RTE_PTYPE_INNER_L2_MASK | \
731 RTE_PTYPE_INNER_L3_MASK | \
732 RTE_PTYPE_INNER_L4_MASK))
734 /** Alignment constraint of mbuf private area. */
735 #define RTE_MBUF_PRIV_ALIGN 8
738 * Get the name of a RX offload flag
741 * The mask describing the flag.
743 * The name of this flag, or NULL if it's not a valid RX flag.
745 const char *rte_get_rx_ol_flag_name(uint64_t mask);
748 * Get the name of a TX offload flag
751 * The mask describing the flag. Usually only one bit must be set.
752 * Several bits can be given if they belong to the same mask.
753 * Ex: PKT_TX_L4_MASK.
755 * The name of this flag, or NULL if it's not a valid TX flag.
757 const char *rte_get_tx_ol_flag_name(uint64_t mask);
760 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
761 * splitting it into multiple segments.
762 * So, for mbufs that planned to be involved into RX/TX, the recommended
763 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
765 #define RTE_MBUF_DEFAULT_DATAROOM 2048
766 #define RTE_MBUF_DEFAULT_BUF_SIZE \
767 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
769 /* define a set of marker types that can be used to refer to set points in the
772 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
774 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
776 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
777 * with a single assignment */
780 * The generic rte_mbuf, containing a packet mbuf.
785 void *buf_addr; /**< Virtual address of segment buffer. */
786 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
788 uint16_t buf_len; /**< Length of segment buffer. */
790 /* next 6 bytes are initialised on RX descriptor rearm */
795 * 16-bit Reference counter.
796 * It should only be accessed using the following functions:
797 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
798 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
799 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
804 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
805 uint16_t refcnt; /**< Non-atomically accessed refcnt */
807 uint8_t nb_segs; /**< Number of segments. */
808 uint8_t port; /**< Input port. */
810 uint64_t ol_flags; /**< Offload features. */
812 /* remaining bytes are set on RX when pulling packet from descriptor */
813 MARKER rx_descriptor_fields1;
816 * The packet type, which is the combination of outer/inner L2, L3, L4
817 * and tunnel types. The packet_type is about data really present in the
818 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
819 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
820 * vlan is stripped from the data.
824 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
826 uint32_t l2_type:4; /**< (Outer) L2 type. */
827 uint32_t l3_type:4; /**< (Outer) L3 type. */
828 uint32_t l4_type:4; /**< (Outer) L4 type. */
829 uint32_t tun_type:4; /**< Tunnel type. */
830 uint32_t inner_l2_type:4; /**< Inner L2 type. */
831 uint32_t inner_l3_type:4; /**< Inner L3 type. */
832 uint32_t inner_l4_type:4; /**< Inner L4 type. */
836 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
837 uint16_t data_len; /**< Amount of data in segment buffer. */
838 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN_STRIPPED is set. */
842 uint32_t rss; /**< RSS hash result if RSS enabled */
851 /**< Second 4 flexible bytes */
854 /**< First 4 flexible bytes or FD ID, dependent on
855 PKT_RX_FDIR_* flag in ol_flags. */
856 } fdir; /**< Filter identifier if FDIR enabled */
860 } sched; /**< Hierarchical scheduler */
861 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
862 } hash; /**< hash information */
864 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
866 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ_STRIPPED is set. */
867 uint16_t vlan_tci_outer;
869 /* second cache line - fields only used in slow path or on TX */
870 MARKER cacheline1 __rte_cache_min_aligned;
874 void *userdata; /**< Can be used for external metadata */
875 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
878 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
879 struct rte_mbuf *next; /**< Next segment of scattered packet. */
881 /* fields to support TX offloads */
884 uint64_t tx_offload; /**< combined for easy fetch */
888 /**< L2 (MAC) Header Length for non-tunneling pkt.
889 * Outer_L4_len + ... + Inner_L2_len for tunneling pkt.
891 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
892 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
893 uint64_t tso_segsz:16; /**< TCP TSO segment size */
895 /* fields for TX offloading of tunnels */
896 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
897 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
899 /* uint64_t unused:8; */
903 /** Size of the application private data. In case of an indirect
904 * mbuf, it stores the direct mbuf private data size. */
907 /** Timesync flags for use with IEEE1588. */
909 } __rte_cache_aligned;
912 * Prefetch the first part of the mbuf
914 * The first 64 bytes of the mbuf corresponds to fields that are used early
915 * in the receive path. If the cache line of the architecture is higher than
916 * 64B, the second part will also be prefetched.
919 * The pointer to the mbuf.
922 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
924 rte_prefetch0(&m->cacheline0);
928 * Prefetch the second part of the mbuf
930 * The next 64 bytes of the mbuf corresponds to fields that are used in the
931 * transmit path. If the cache line of the architecture is higher than 64B,
932 * this function does nothing as it is expected that the full mbuf is
936 * The pointer to the mbuf.
939 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
941 #if RTE_CACHE_LINE_SIZE == 64
942 rte_prefetch0(&m->cacheline1);
949 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
952 * Return the DMA address of the beginning of the mbuf data
955 * The pointer to the mbuf.
957 * The physical address of the beginning of the mbuf data
959 static inline phys_addr_t
960 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
962 return mb->buf_physaddr + mb->data_off;
966 * Return the default DMA address of the beginning of the mbuf data
968 * This function is used by drivers in their receive function, as it
969 * returns the location where data should be written by the NIC, taking
970 * the default headroom in account.
973 * The pointer to the mbuf.
975 * The physical address of the beginning of the mbuf data
977 static inline phys_addr_t
978 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
980 return mb->buf_physaddr + RTE_PKTMBUF_HEADROOM;
984 * Return the mbuf owning the data buffer address of an indirect mbuf.
987 * The pointer to the indirect mbuf.
989 * The address of the direct mbuf corresponding to buffer_addr.
991 static inline struct rte_mbuf *
992 rte_mbuf_from_indirect(struct rte_mbuf *mi)
994 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
998 * Return the buffer address embedded in the given mbuf.
1001 * The pointer to the mbuf.
1003 * The address of the data buffer owned by the mbuf.
1005 static inline char *
1006 rte_mbuf_to_baddr(struct rte_mbuf *md)
1009 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
1014 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
1016 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
1019 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
1021 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
1024 * Private data in case of pktmbuf pool.
1026 * A structure that contains some pktmbuf_pool-specific data that are
1027 * appended after the mempool structure (in private data).
1029 struct rte_pktmbuf_pool_private {
1030 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
1031 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
1034 #ifdef RTE_LIBRTE_MBUF_DEBUG
1036 /** check mbuf type in debug mode */
1037 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
1039 #else /* RTE_LIBRTE_MBUF_DEBUG */
1041 /** check mbuf type in debug mode */
1042 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
1044 #endif /* RTE_LIBRTE_MBUF_DEBUG */
1046 #ifdef RTE_MBUF_REFCNT_ATOMIC
1049 * Reads the value of an mbuf's refcnt.
1053 * Reference count number.
1055 static inline uint16_t
1056 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1058 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
1062 * Sets an mbuf's refcnt to a defined value.
1069 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1071 rte_atomic16_set(&m->refcnt_atomic, new_value);
1075 * Adds given value to an mbuf's refcnt and returns its new value.
1079 * Value to add/subtract
1083 static inline uint16_t
1084 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1087 * The atomic_add is an expensive operation, so we don't want to
1088 * call it in the case where we know we are the uniq holder of
1089 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
1090 * operation has to be used because concurrent accesses on the
1091 * reference counter can occur.
1093 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1094 rte_mbuf_refcnt_set(m, 1 + value);
1098 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
1101 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
1104 * Adds given value to an mbuf's refcnt and returns its new value.
1106 static inline uint16_t
1107 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1109 m->refcnt = (uint16_t)(m->refcnt + value);
1114 * Reads the value of an mbuf's refcnt.
1116 static inline uint16_t
1117 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1123 * Sets an mbuf's refcnt to the defined value.
1126 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1128 m->refcnt = new_value;
1131 #endif /* RTE_MBUF_REFCNT_ATOMIC */
1133 /** Mbuf prefetch */
1134 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
1141 * Sanity checks on an mbuf.
1143 * Check the consistency of the given mbuf. The function will cause a
1144 * panic if corruption is detected.
1147 * The mbuf to be checked.
1149 * True if the mbuf is a packet header, false if it is a sub-segment
1150 * of a packet (in this case, some fields like nb_segs are not checked)
1153 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
1156 * Allocate an unitialized mbuf from mempool *mp*.
1158 * This function can be used by PMDs (especially in RX functions) to
1159 * allocate an unitialized mbuf. The driver is responsible of
1160 * initializing all the required fields. See rte_pktmbuf_reset().
1161 * For standard needs, prefer rte_pktmbuf_alloc().
1164 * The mempool from which mbuf is allocated.
1166 * - The pointer to the new mbuf on success.
1167 * - NULL if allocation failed.
1169 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
1174 if (rte_mempool_get(mp, &mb) < 0)
1176 m = (struct rte_mbuf *)mb;
1177 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1178 rte_mbuf_refcnt_set(m, 1);
1179 __rte_mbuf_sanity_check(m, 0);
1185 * @internal Put mbuf back into its original mempool.
1186 * The use of that function is reserved for RTE internal needs.
1187 * Please use rte_pktmbuf_free().
1190 * The mbuf to be freed.
1192 static inline void __attribute__((always_inline))
1193 __rte_mbuf_raw_free(struct rte_mbuf *m)
1195 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1196 rte_mempool_put(m->pool, m);
1199 /* Operations on ctrl mbuf */
1202 * The control mbuf constructor.
1204 * This function initializes some fields in an mbuf structure that are
1205 * not modified by the user once created (mbuf type, origin pool, buffer
1206 * start address, and so on). This function is given as a callback function
1207 * to rte_mempool_create() at pool creation time.
1210 * The mempool from which the mbuf is allocated.
1212 * A pointer that can be used by the user to retrieve useful information
1213 * for mbuf initialization. This pointer comes from the ``init_arg``
1214 * parameter of rte_mempool_create().
1216 * The mbuf to initialize.
1218 * The index of the mbuf in the pool table.
1220 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1221 void *m, unsigned i);
1224 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
1226 * This new mbuf is initialized with data pointing to the beginning of
1227 * buffer, and with a length of zero.
1230 * The mempool from which the mbuf is allocated.
1232 * - The pointer to the new mbuf on success.
1233 * - NULL if allocation failed.
1235 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
1238 * Free a control mbuf back into its original mempool.
1241 * The control mbuf to be freed.
1243 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
1246 * A macro that returns the pointer to the carried data.
1248 * The value that can be read or assigned.
1253 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
1256 * A macro that returns the length of the carried data.
1258 * The value that can be read or assigned.
1263 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
1266 * Tests if an mbuf is a control mbuf
1269 * The mbuf to be tested
1271 * - True (1) if the mbuf is a control mbuf
1272 * - False(0) otherwise
1275 rte_is_ctrlmbuf(struct rte_mbuf *m)
1277 return !!(m->ol_flags & CTRL_MBUF_FLAG);
1280 /* Operations on pkt mbuf */
1283 * The packet mbuf constructor.
1285 * This function initializes some fields in the mbuf structure that are
1286 * not modified by the user once created (origin pool, buffer start
1287 * address, and so on). This function is given as a callback function to
1288 * rte_mempool_create() at pool creation time.
1291 * The mempool from which mbufs originate.
1293 * A pointer that can be used by the user to retrieve useful information
1294 * for mbuf initialization. This pointer comes from the ``init_arg``
1295 * parameter of rte_mempool_create().
1297 * The mbuf to initialize.
1299 * The index of the mbuf in the pool table.
1301 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1302 void *m, unsigned i);
1306 * A packet mbuf pool constructor.
1308 * This function initializes the mempool private data in the case of a
1309 * pktmbuf pool. This private data is needed by the driver. The
1310 * function is given as a callback function to rte_mempool_create() at
1311 * pool creation. It can be extended by the user, for example, to
1312 * provide another packet size.
1315 * The mempool from which mbufs originate.
1317 * A pointer that can be used by the user to retrieve useful information
1318 * for mbuf initialization. This pointer comes from the ``init_arg``
1319 * parameter of rte_mempool_create().
1321 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1324 * Create a mbuf pool.
1326 * This function creates and initializes a packet mbuf pool. It is
1327 * a wrapper to rte_mempool_create() with the proper packet constructor
1328 * and mempool constructor.
1331 * The name of the mbuf pool.
1333 * The number of elements in the mbuf pool. The optimum size (in terms
1334 * of memory usage) for a mempool is when n is a power of two minus one:
1337 * Size of the per-core object cache. See rte_mempool_create() for
1340 * Size of application private are between the rte_mbuf structure
1341 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1342 * @param data_room_size
1343 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1345 * The socket identifier where the memory should be allocated. The
1346 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1349 * The pointer to the new allocated mempool, on success. NULL on error
1350 * with rte_errno set appropriately. Possible rte_errno values include:
1351 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1352 * - E_RTE_SECONDARY - function was called from a secondary process instance
1353 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1354 * - ENOSPC - the maximum number of memzones has already been allocated
1355 * - EEXIST - a memzone with the same name already exists
1356 * - ENOMEM - no appropriate memory area found in which to create memzone
1358 struct rte_mempool *
1359 rte_pktmbuf_pool_create(const char *name, unsigned n,
1360 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1364 * Get the data room size of mbufs stored in a pktmbuf_pool
1366 * The data room size is the amount of data that can be stored in a
1367 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1370 * The packet mbuf pool.
1372 * The data room size of mbufs stored in this mempool.
1374 static inline uint16_t
1375 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1377 struct rte_pktmbuf_pool_private *mbp_priv;
1379 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1380 return mbp_priv->mbuf_data_room_size;
1384 * Get the application private size of mbufs stored in a pktmbuf_pool
1386 * The private size of mbuf is a zone located between the rte_mbuf
1387 * structure and the data buffer where an application can store data
1388 * associated to a packet.
1391 * The packet mbuf pool.
1393 * The private size of mbufs stored in this mempool.
1395 static inline uint16_t
1396 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1398 struct rte_pktmbuf_pool_private *mbp_priv;
1400 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1401 return mbp_priv->mbuf_priv_size;
1405 * Reset the data_off field of a packet mbuf to its default value.
1407 * The given mbuf must have only one segment, which should be empty.
1410 * The packet mbuf's data_off field has to be reset.
1412 static inline void rte_pktmbuf_reset_headroom(struct rte_mbuf *m)
1414 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1418 * Reset the fields of a packet mbuf to their default values.
1420 * The given mbuf must have only one segment.
1423 * The packet mbuf to be resetted.
1425 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1431 m->vlan_tci_outer = 0;
1437 rte_pktmbuf_reset_headroom(m);
1440 __rte_mbuf_sanity_check(m, 1);
1444 * Allocate a new mbuf from a mempool.
1446 * This new mbuf contains one segment, which has a length of 0. The pointer
1447 * to data is initialized to have some bytes of headroom in the buffer
1448 * (if buffer size allows).
1451 * The mempool from which the mbuf is allocated.
1453 * - The pointer to the new mbuf on success.
1454 * - NULL if allocation failed.
1456 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1459 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1460 rte_pktmbuf_reset(m);
1465 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1469 * The mempool from which mbufs are allocated.
1471 * Array of pointers to mbufs
1477 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1478 struct rte_mbuf **mbufs, unsigned count)
1483 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1487 /* To understand duff's device on loop unwinding optimization, see
1488 * https://en.wikipedia.org/wiki/Duff's_device.
1489 * Here while() loop is used rather than do() while{} to avoid extra
1490 * check if count is zero.
1492 switch (count % 4) {
1494 while (idx != count) {
1495 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1496 rte_mbuf_refcnt_set(mbufs[idx], 1);
1497 rte_pktmbuf_reset(mbufs[idx]);
1500 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1501 rte_mbuf_refcnt_set(mbufs[idx], 1);
1502 rte_pktmbuf_reset(mbufs[idx]);
1505 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1506 rte_mbuf_refcnt_set(mbufs[idx], 1);
1507 rte_pktmbuf_reset(mbufs[idx]);
1510 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1511 rte_mbuf_refcnt_set(mbufs[idx], 1);
1512 rte_pktmbuf_reset(mbufs[idx]);
1520 * Attach packet mbuf to another packet mbuf.
1522 * After attachment we refer the mbuf we attached as 'indirect',
1523 * while mbuf we attached to as 'direct'.
1524 * The direct mbuf's reference counter is incremented.
1526 * Right now, not supported:
1527 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1528 * - mbuf we trying to attach (mi) is used by someone else
1529 * e.g. it's reference counter is greater then 1.
1532 * The indirect packet mbuf.
1534 * The packet mbuf we're attaching to.
1536 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1538 struct rte_mbuf *md;
1540 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1541 rte_mbuf_refcnt_read(mi) == 1);
1543 /* if m is not direct, get the mbuf that embeds the data */
1544 if (RTE_MBUF_DIRECT(m))
1547 md = rte_mbuf_from_indirect(m);
1549 rte_mbuf_refcnt_update(md, 1);
1550 mi->priv_size = m->priv_size;
1551 mi->buf_physaddr = m->buf_physaddr;
1552 mi->buf_addr = m->buf_addr;
1553 mi->buf_len = m->buf_len;
1556 mi->data_off = m->data_off;
1557 mi->data_len = m->data_len;
1559 mi->vlan_tci = m->vlan_tci;
1560 mi->vlan_tci_outer = m->vlan_tci_outer;
1561 mi->tx_offload = m->tx_offload;
1565 mi->pkt_len = mi->data_len;
1567 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1568 mi->packet_type = m->packet_type;
1570 __rte_mbuf_sanity_check(mi, 1);
1571 __rte_mbuf_sanity_check(m, 0);
1575 * Detach an indirect packet mbuf.
1577 * - restore original mbuf address and length values.
1578 * - reset pktmbuf data and data_len to their default values.
1579 * - decrement the direct mbuf's reference counter. When the
1580 * reference counter becomes 0, the direct mbuf is freed.
1582 * All other fields of the given packet mbuf will be left intact.
1585 * The indirect attached packet mbuf.
1587 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1589 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1590 struct rte_mempool *mp = m->pool;
1591 uint32_t mbuf_size, buf_len, priv_size;
1593 priv_size = rte_pktmbuf_priv_size(mp);
1594 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1595 buf_len = rte_pktmbuf_data_room_size(mp);
1597 m->priv_size = priv_size;
1598 m->buf_addr = (char *)m + mbuf_size;
1599 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
1600 m->buf_len = (uint16_t)buf_len;
1601 rte_pktmbuf_reset_headroom(m);
1605 if (rte_mbuf_refcnt_update(md, -1) == 0)
1606 __rte_mbuf_raw_free(md);
1609 static inline struct rte_mbuf* __attribute__((always_inline))
1610 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1612 __rte_mbuf_sanity_check(m, 0);
1614 if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
1615 /* if this is an indirect mbuf, it is detached. */
1616 if (RTE_MBUF_INDIRECT(m))
1617 rte_pktmbuf_detach(m);
1624 * Free a segment of a packet mbuf into its original mempool.
1626 * Free an mbuf, without parsing other segments in case of chained
1630 * The packet mbuf segment to be freed.
1632 static inline void __attribute__((always_inline))
1633 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1635 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
1637 __rte_mbuf_raw_free(m);
1642 * Free a packet mbuf back into its original mempool.
1644 * Free an mbuf, and all its segments in case of chained buffers. Each
1645 * segment is added back into its original mempool.
1648 * The packet mbuf to be freed.
1650 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1652 struct rte_mbuf *m_next;
1654 __rte_mbuf_sanity_check(m, 1);
1658 rte_pktmbuf_free_seg(m);
1664 * Creates a "clone" of the given packet mbuf.
1666 * Walks through all segments of the given packet mbuf, and for each of them:
1667 * - Creates a new packet mbuf from the given pool.
1668 * - Attaches newly created mbuf to the segment.
1669 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1670 * from the original packet mbuf.
1673 * The packet mbuf to be cloned.
1675 * The mempool from which the "clone" mbufs are allocated.
1677 * - The pointer to the new "clone" mbuf on success.
1678 * - NULL if allocation fails.
1680 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1681 struct rte_mempool *mp)
1683 struct rte_mbuf *mc, *mi, **prev;
1687 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1692 pktlen = md->pkt_len;
1697 rte_pktmbuf_attach(mi, md);
1700 } while ((md = md->next) != NULL &&
1701 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1705 mc->pkt_len = pktlen;
1707 /* Allocation of new indirect segment failed */
1708 if (unlikely (mi == NULL)) {
1709 rte_pktmbuf_free(mc);
1713 __rte_mbuf_sanity_check(mc, 1);
1718 * Adds given value to the refcnt of all packet mbuf segments.
1720 * Walks through all segments of given packet mbuf and for each of them
1721 * invokes rte_mbuf_refcnt_update().
1724 * The packet mbuf whose refcnt to be updated.
1726 * The value to add to the mbuf's segments refcnt.
1728 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1730 __rte_mbuf_sanity_check(m, 1);
1733 rte_mbuf_refcnt_update(m, v);
1734 } while ((m = m->next) != NULL);
1738 * Get the headroom in a packet mbuf.
1743 * The length of the headroom.
1745 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1747 __rte_mbuf_sanity_check(m, 1);
1752 * Get the tailroom of a packet mbuf.
1757 * The length of the tailroom.
1759 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1761 __rte_mbuf_sanity_check(m, 1);
1762 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1767 * Get the last segment of the packet.
1772 * The last segment of the given mbuf.
1774 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1776 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1778 __rte_mbuf_sanity_check(m, 1);
1779 while (m2->next != NULL)
1785 * A macro that points to an offset into the data in the mbuf.
1787 * The returned pointer is cast to type t. Before using this
1788 * function, the user must ensure that the first segment is large
1789 * enough to accommodate its data.
1794 * The offset into the mbuf data.
1796 * The type to cast the result into.
1798 #define rte_pktmbuf_mtod_offset(m, t, o) \
1799 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1802 * A macro that points to the start of the data in the mbuf.
1804 * The returned pointer is cast to type t. Before using this
1805 * function, the user must ensure that the first segment is large
1806 * enough to accommodate its data.
1811 * The type to cast the result into.
1813 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1816 * A macro that returns the physical address that points to an offset of the
1817 * start of the data in the mbuf
1822 * The offset into the data to calculate address from.
1824 #define rte_pktmbuf_mtophys_offset(m, o) \
1825 (phys_addr_t)((m)->buf_physaddr + (m)->data_off + (o))
1828 * A macro that returns the physical address that points to the start of the
1834 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_mtophys_offset(m, 0)
1837 * A macro that returns the length of the packet.
1839 * The value can be read or assigned.
1844 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1847 * A macro that returns the length of the segment.
1849 * The value can be read or assigned.
1854 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1857 * Prepend len bytes to an mbuf data area.
1859 * Returns a pointer to the new
1860 * data start address. If there is not enough headroom in the first
1861 * segment, the function will return NULL, without modifying the mbuf.
1866 * The amount of data to prepend (in bytes).
1868 * A pointer to the start of the newly prepended data, or
1869 * NULL if there is not enough headroom space in the first segment
1871 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1874 __rte_mbuf_sanity_check(m, 1);
1876 if (unlikely(len > rte_pktmbuf_headroom(m)))
1880 m->data_len = (uint16_t)(m->data_len + len);
1881 m->pkt_len = (m->pkt_len + len);
1883 return (char *)m->buf_addr + m->data_off;
1887 * Append len bytes to an mbuf.
1889 * Append len bytes to an mbuf and return a pointer to the start address
1890 * of the added data. If there is not enough tailroom in the last
1891 * segment, the function will return NULL, without modifying the mbuf.
1896 * The amount of data to append (in bytes).
1898 * A pointer to the start of the newly appended data, or
1899 * NULL if there is not enough tailroom space in the last segment
1901 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1904 struct rte_mbuf *m_last;
1906 __rte_mbuf_sanity_check(m, 1);
1908 m_last = rte_pktmbuf_lastseg(m);
1909 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1912 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1913 m_last->data_len = (uint16_t)(m_last->data_len + len);
1914 m->pkt_len = (m->pkt_len + len);
1915 return (char*) tail;
1919 * Remove len bytes at the beginning of an mbuf.
1921 * Returns a pointer to the start address of the new data area. If the
1922 * length is greater than the length of the first segment, then the
1923 * function will fail and return NULL, without modifying the mbuf.
1928 * The amount of data to remove (in bytes).
1930 * A pointer to the new start of the data.
1932 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1934 __rte_mbuf_sanity_check(m, 1);
1936 if (unlikely(len > m->data_len))
1939 m->data_len = (uint16_t)(m->data_len - len);
1941 m->pkt_len = (m->pkt_len - len);
1942 return (char *)m->buf_addr + m->data_off;
1946 * Remove len bytes of data at the end of the mbuf.
1948 * If the length is greater than the length of the last segment, the
1949 * function will fail and return -1 without modifying the mbuf.
1954 * The amount of data to remove (in bytes).
1959 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1961 struct rte_mbuf *m_last;
1963 __rte_mbuf_sanity_check(m, 1);
1965 m_last = rte_pktmbuf_lastseg(m);
1966 if (unlikely(len > m_last->data_len))
1969 m_last->data_len = (uint16_t)(m_last->data_len - len);
1970 m->pkt_len = (m->pkt_len - len);
1975 * Test if mbuf data is contiguous.
1980 * - 1, if all data is contiguous (one segment).
1981 * - 0, if there is several segments.
1983 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1985 __rte_mbuf_sanity_check(m, 1);
1986 return !!(m->nb_segs == 1);
1990 * Chain an mbuf to another, thereby creating a segmented packet.
1992 * Note: The implementation will do a linear walk over the segments to find
1993 * the tail entry. For cases when there are many segments, it's better to
1994 * chain the entries manually.
1997 * The head of the mbuf chain (the first packet)
1999 * The mbuf to put last in the chain
2003 * - -EOVERFLOW, if the chain is full (256 entries)
2005 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
2007 struct rte_mbuf *cur_tail;
2009 /* Check for number-of-segments-overflow */
2010 if (head->nb_segs + tail->nb_segs >= 1 << (sizeof(head->nb_segs) * 8))
2013 /* Chain 'tail' onto the old tail */
2014 cur_tail = rte_pktmbuf_lastseg(head);
2015 cur_tail->next = tail;
2017 /* accumulate number of segments and total length. */
2018 head->nb_segs = (uint8_t)(head->nb_segs + tail->nb_segs);
2019 head->pkt_len += tail->pkt_len;
2021 /* pkt_len is only set in the head */
2022 tail->pkt_len = tail->data_len;
2028 * Dump an mbuf structure to the console.
2030 * Dump all fields for the given packet mbuf and all its associated
2031 * segments (in the case of a chained buffer).
2034 * A pointer to a file for output
2038 * If dump_len != 0, also dump the "dump_len" first data bytes of
2041 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
2047 #endif /* _RTE_MBUF_H_ */