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. */
96 #define PKT_RX_OVERSIZE (0ULL << 0) /**< Num of desc of an RX pkt oversize. */
97 #define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
98 #define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
99 #define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
102 * A vlan has been stripped by the hardware and its tci is saved in
103 * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
104 * in the RX configuration of the PMD.
106 #define PKT_RX_VLAN_STRIPPED (1ULL << 6)
108 /* hole, some bits can be reused here */
110 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
111 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
112 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
113 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
116 * The 2 vlans have been stripped by the hardware and their tci are
117 * saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
118 * This can only happen if vlan stripping is enabled in the RX
119 * configuration of the PMD. If this flag is set, PKT_RX_VLAN_STRIPPED
122 #define PKT_RX_QINQ_STRIPPED (1ULL << 15)
126 * RX packet with double VLAN stripped.
127 * This flag is replaced by PKT_RX_QINQ_STRIPPED.
129 #define PKT_RX_QINQ_PKT PKT_RX_QINQ_STRIPPED
131 /* add new RX flags here */
133 /* add new TX flags here */
136 * Second VLAN insertion (QinQ) flag.
138 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
141 * TCP segmentation offload. To enable this offload feature for a
142 * packet to be transmitted on hardware supporting TSO:
143 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
145 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
146 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
148 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
149 * - calculate the pseudo header checksum without taking ip_len in account,
150 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
151 * rte_ipv6_phdr_cksum() that can be used as helpers.
153 #define PKT_TX_TCP_SEG (1ULL << 50)
155 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
158 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
159 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
160 * L4 checksum offload, the user needs to:
161 * - fill l2_len and l3_len in mbuf
162 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
163 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
164 * - calculate the pseudo header checksum and set it in the L4 header (only
165 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
166 * For SCTP, set the crc field to 0.
168 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
169 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
170 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
171 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
172 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
175 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
176 * also be set by the application, although a PMD will only check
178 * - set the IP checksum field in the packet to 0
179 * - fill the mbuf offload information: l2_len, l3_len
181 #define PKT_TX_IP_CKSUM (1ULL << 54)
184 * Packet is IPv4. This flag must be set when using any offload feature
185 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
186 * packet. If the packet is a tunneled packet, this flag is related to
189 #define PKT_TX_IPV4 (1ULL << 55)
192 * Packet is IPv6. This flag must be set when using an offload feature
193 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
194 * packet. If the packet is a tunneled packet, this flag is related to
197 #define PKT_TX_IPV6 (1ULL << 56)
199 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
202 * Offload the IP checksum of an external header in the hardware. The
203 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
204 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
205 * packet must be set to 0.
206 * - set the outer IP checksum field in the packet to 0
207 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
209 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
212 * Packet outer header is IPv4. This flag must be set when using any
213 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
214 * outer header of the tunneled packet is an IPv4 packet.
216 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
219 * Packet outer header is IPv6. This flag must be set when using any
220 * outer offload feature (L4 checksum) to tell the NIC that the outer
221 * header of the tunneled packet is an IPv6 packet.
223 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
225 #define __RESERVED (1ULL << 61) /**< reserved for future mbuf use */
227 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
229 /* Use final bit of flags to indicate a control mbuf */
230 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
233 * 32 bits are divided into several fields to mark packet types. Note that
234 * each field is indexical.
235 * - Bit 3:0 is for L2 types.
236 * - Bit 7:4 is for L3 or outer L3 (for tunneling case) types.
237 * - Bit 11:8 is for L4 or outer L4 (for tunneling case) types.
238 * - Bit 15:12 is for tunnel types.
239 * - Bit 19:16 is for inner L2 types.
240 * - Bit 23:20 is for inner L3 types.
241 * - Bit 27:24 is for inner L4 types.
242 * - Bit 31:28 is reserved.
244 * To be compatible with Vector PMD, RTE_PTYPE_L3_IPV4, RTE_PTYPE_L3_IPV4_EXT,
245 * RTE_PTYPE_L3_IPV6, RTE_PTYPE_L3_IPV6_EXT, RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP
246 * and RTE_PTYPE_L4_SCTP should be kept as below in a contiguous 7 bits.
248 * Note that L3 types values are selected for checking IPV4/IPV6 header from
249 * performance point of view. Reading annotations of RTE_ETH_IS_IPV4_HDR and
250 * RTE_ETH_IS_IPV6_HDR is needed for any future changes of L3 type values.
252 * Note that the packet types of the same packet recognized by different
253 * hardware may be different, as different hardware may have different
254 * capability of packet type recognition.
257 * <'ether type'=0x0800
258 * | 'version'=4, 'protocol'=0x29
259 * | 'version'=6, 'next header'=0x3A
261 * will be recognized on i40e hardware as packet type combination of,
262 * RTE_PTYPE_L2_ETHER |
263 * RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
264 * RTE_PTYPE_TUNNEL_IP |
265 * RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
266 * RTE_PTYPE_INNER_L4_ICMP.
268 * <'ether type'=0x86DD
269 * | 'version'=6, 'next header'=0x2F
271 * | 'version'=6, 'next header'=0x11
273 * will be recognized on i40e hardware as packet type combination of,
274 * RTE_PTYPE_L2_ETHER |
275 * RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
276 * RTE_PTYPE_TUNNEL_GRENAT |
277 * RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
278 * RTE_PTYPE_INNER_L4_UDP.
280 #define RTE_PTYPE_UNKNOWN 0x00000000
282 * Ethernet packet type.
283 * It is used for outer packet for tunneling cases.
286 * <'ether type'=[0x0800|0x86DD]>
288 #define RTE_PTYPE_L2_ETHER 0x00000001
290 * Ethernet packet type for time sync.
293 * <'ether type'=0x88F7>
295 #define RTE_PTYPE_L2_ETHER_TIMESYNC 0x00000002
297 * ARP (Address Resolution Protocol) packet type.
300 * <'ether type'=0x0806>
302 #define RTE_PTYPE_L2_ETHER_ARP 0x00000003
304 * LLDP (Link Layer Discovery Protocol) packet type.
307 * <'ether type'=0x88CC>
309 #define RTE_PTYPE_L2_ETHER_LLDP 0x00000004
311 * NSH (Network Service Header) packet type.
314 * <'ether type'=0x894F>
316 #define RTE_PTYPE_L2_ETHER_NSH 0x00000005
318 * Mask of layer 2 packet types.
319 * It is used for outer packet for tunneling cases.
321 #define RTE_PTYPE_L2_MASK 0x0000000f
323 * IP (Internet Protocol) version 4 packet type.
324 * It is used for outer packet for tunneling cases, and does not contain any
328 * <'ether type'=0x0800
329 * | 'version'=4, 'ihl'=5>
331 #define RTE_PTYPE_L3_IPV4 0x00000010
333 * IP (Internet Protocol) version 4 packet type.
334 * It is used for outer packet for tunneling cases, and contains header
338 * <'ether type'=0x0800
339 * | 'version'=4, 'ihl'=[6-15], 'options'>
341 #define RTE_PTYPE_L3_IPV4_EXT 0x00000030
343 * IP (Internet Protocol) version 6 packet type.
344 * It is used for outer packet for tunneling cases, and does not contain any
348 * <'ether type'=0x86DD
349 * | 'version'=6, 'next header'=0x3B>
351 #define RTE_PTYPE_L3_IPV6 0x00000040
353 * IP (Internet Protocol) version 4 packet type.
354 * It is used for outer packet for tunneling cases, and may or maynot contain
358 * <'ether type'=0x0800
359 * | 'version'=4, 'ihl'=[5-15], <'options'>>
361 #define RTE_PTYPE_L3_IPV4_EXT_UNKNOWN 0x00000090
363 * IP (Internet Protocol) version 6 packet type.
364 * It is used for outer packet for tunneling cases, and contains extension
368 * <'ether type'=0x86DD
369 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
370 * 'extension headers'>
372 #define RTE_PTYPE_L3_IPV6_EXT 0x000000c0
374 * IP (Internet Protocol) version 6 packet type.
375 * It is used for outer packet for tunneling cases, and may or maynot contain
379 * <'ether type'=0x86DD
380 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
381 * <'extension headers'>>
383 #define RTE_PTYPE_L3_IPV6_EXT_UNKNOWN 0x000000e0
385 * Mask of layer 3 packet types.
386 * It is used for outer packet for tunneling cases.
388 #define RTE_PTYPE_L3_MASK 0x000000f0
390 * TCP (Transmission Control Protocol) packet type.
391 * It is used for outer packet for tunneling cases.
394 * <'ether type'=0x0800
395 * | 'version'=4, 'protocol'=6, 'MF'=0>
397 * <'ether type'=0x86DD
398 * | 'version'=6, 'next header'=6>
400 #define RTE_PTYPE_L4_TCP 0x00000100
402 * UDP (User Datagram Protocol) packet type.
403 * It is used for outer packet for tunneling cases.
406 * <'ether type'=0x0800
407 * | 'version'=4, 'protocol'=17, 'MF'=0>
409 * <'ether type'=0x86DD
410 * | 'version'=6, 'next header'=17>
412 #define RTE_PTYPE_L4_UDP 0x00000200
414 * Fragmented IP (Internet Protocol) packet type.
415 * It is used for outer packet for tunneling cases.
417 * It refers to those packets of any IP types, which can be recognized as
418 * fragmented. A fragmented packet cannot be recognized as any other L4 types
419 * (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP,
420 * RTE_PTYPE_L4_NONFRAG).
423 * <'ether type'=0x0800
424 * | 'version'=4, 'MF'=1>
426 * <'ether type'=0x86DD
427 * | 'version'=6, 'next header'=44>
429 #define RTE_PTYPE_L4_FRAG 0x00000300
431 * SCTP (Stream Control Transmission Protocol) packet type.
432 * It is used for outer packet for tunneling cases.
435 * <'ether type'=0x0800
436 * | 'version'=4, 'protocol'=132, 'MF'=0>
438 * <'ether type'=0x86DD
439 * | 'version'=6, 'next header'=132>
441 #define RTE_PTYPE_L4_SCTP 0x00000400
443 * ICMP (Internet Control Message Protocol) packet type.
444 * It is used for outer packet for tunneling cases.
447 * <'ether type'=0x0800
448 * | 'version'=4, 'protocol'=1, 'MF'=0>
450 * <'ether type'=0x86DD
451 * | 'version'=6, 'next header'=1>
453 #define RTE_PTYPE_L4_ICMP 0x00000500
455 * Non-fragmented IP (Internet Protocol) packet type.
456 * It is used for outer packet for tunneling cases.
458 * It refers to those packets of any IP types, while cannot be recognized as
459 * any of above L4 types (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP,
460 * RTE_PTYPE_L4_FRAG, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP).
463 * <'ether type'=0x0800
464 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
466 * <'ether type'=0x86DD
467 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
469 #define RTE_PTYPE_L4_NONFRAG 0x00000600
471 * Mask of layer 4 packet types.
472 * It is used for outer packet for tunneling cases.
474 #define RTE_PTYPE_L4_MASK 0x00000f00
476 * IP (Internet Protocol) in IP (Internet Protocol) tunneling packet type.
479 * <'ether type'=0x0800
480 * | 'version'=4, 'protocol'=[4|41]>
482 * <'ether type'=0x86DD
483 * | 'version'=6, 'next header'=[4|41]>
485 #define RTE_PTYPE_TUNNEL_IP 0x00001000
487 * GRE (Generic Routing Encapsulation) tunneling packet type.
490 * <'ether type'=0x0800
491 * | 'version'=4, 'protocol'=47>
493 * <'ether type'=0x86DD
494 * | 'version'=6, 'next header'=47>
496 #define RTE_PTYPE_TUNNEL_GRE 0x00002000
498 * VXLAN (Virtual eXtensible Local Area Network) tunneling packet type.
501 * <'ether type'=0x0800
502 * | 'version'=4, 'protocol'=17
503 * | 'destination port'=4798>
505 * <'ether type'=0x86DD
506 * | 'version'=6, 'next header'=17
507 * | 'destination port'=4798>
509 #define RTE_PTYPE_TUNNEL_VXLAN 0x00003000
511 * NVGRE (Network Virtualization using Generic Routing Encapsulation) tunneling
515 * <'ether type'=0x0800
516 * | 'version'=4, 'protocol'=47
517 * | 'protocol type'=0x6558>
519 * <'ether type'=0x86DD
520 * | 'version'=6, 'next header'=47
521 * | 'protocol type'=0x6558'>
523 #define RTE_PTYPE_TUNNEL_NVGRE 0x00004000
525 * GENEVE (Generic Network Virtualization Encapsulation) tunneling packet type.
528 * <'ether type'=0x0800
529 * | 'version'=4, 'protocol'=17
530 * | 'destination port'=6081>
532 * <'ether type'=0x86DD
533 * | 'version'=6, 'next header'=17
534 * | 'destination port'=6081>
536 #define RTE_PTYPE_TUNNEL_GENEVE 0x00005000
538 * Tunneling packet type of Teredo, VXLAN (Virtual eXtensible Local Area
539 * Network) or GRE (Generic Routing Encapsulation) could be recognized as this
540 * packet type, if they can not be recognized independently as of hardware
543 #define RTE_PTYPE_TUNNEL_GRENAT 0x00006000
545 * Mask of tunneling packet types.
547 #define RTE_PTYPE_TUNNEL_MASK 0x0000f000
549 * Ethernet packet type.
550 * It is used for inner packet type only.
552 * Packet format (inner only):
553 * <'ether type'=[0x800|0x86DD]>
555 #define RTE_PTYPE_INNER_L2_ETHER 0x00010000
557 * Ethernet packet type with VLAN (Virtual Local Area Network) tag.
559 * Packet format (inner only):
560 * <'ether type'=[0x800|0x86DD], vlan=[1-4095]>
562 #define RTE_PTYPE_INNER_L2_ETHER_VLAN 0x00020000
564 * Mask of inner layer 2 packet types.
566 #define RTE_PTYPE_INNER_L2_MASK 0x000f0000
568 * IP (Internet Protocol) version 4 packet type.
569 * It is used for inner packet only, and does not contain any header option.
571 * Packet format (inner only):
572 * <'ether type'=0x0800
573 * | 'version'=4, 'ihl'=5>
575 #define RTE_PTYPE_INNER_L3_IPV4 0x00100000
577 * IP (Internet Protocol) version 4 packet type.
578 * It is used for inner packet only, and contains header options.
580 * Packet format (inner only):
581 * <'ether type'=0x0800
582 * | 'version'=4, 'ihl'=[6-15], 'options'>
584 #define RTE_PTYPE_INNER_L3_IPV4_EXT 0x00200000
586 * IP (Internet Protocol) version 6 packet type.
587 * It is used for inner packet only, and does not contain any extension header.
589 * Packet format (inner only):
590 * <'ether type'=0x86DD
591 * | 'version'=6, 'next header'=0x3B>
593 #define RTE_PTYPE_INNER_L3_IPV6 0x00300000
595 * IP (Internet Protocol) version 4 packet type.
596 * It is used for inner packet only, and may or maynot contain header options.
598 * Packet format (inner only):
599 * <'ether type'=0x0800
600 * | 'version'=4, 'ihl'=[5-15], <'options'>>
602 #define RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN 0x00400000
604 * IP (Internet Protocol) version 6 packet type.
605 * It is used for inner packet only, and contains extension headers.
607 * Packet format (inner only):
608 * <'ether type'=0x86DD
609 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
610 * 'extension headers'>
612 #define RTE_PTYPE_INNER_L3_IPV6_EXT 0x00500000
614 * IP (Internet Protocol) version 6 packet type.
615 * It is used for inner packet only, and may or maynot contain extension
618 * Packet format (inner only):
619 * <'ether type'=0x86DD
620 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
621 * <'extension headers'>>
623 #define RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN 0x00600000
625 * Mask of inner layer 3 packet types.
627 #define RTE_PTYPE_INNER_L3_MASK 0x00f00000
629 * TCP (Transmission Control Protocol) packet type.
630 * It is used for inner packet only.
632 * Packet format (inner only):
633 * <'ether type'=0x0800
634 * | 'version'=4, 'protocol'=6, 'MF'=0>
636 * <'ether type'=0x86DD
637 * | 'version'=6, 'next header'=6>
639 #define RTE_PTYPE_INNER_L4_TCP 0x01000000
641 * UDP (User Datagram Protocol) packet type.
642 * It is used for inner packet only.
644 * Packet format (inner only):
645 * <'ether type'=0x0800
646 * | 'version'=4, 'protocol'=17, 'MF'=0>
648 * <'ether type'=0x86DD
649 * | 'version'=6, 'next header'=17>
651 #define RTE_PTYPE_INNER_L4_UDP 0x02000000
653 * Fragmented IP (Internet Protocol) packet type.
654 * It is used for inner packet only, and may or maynot have layer 4 packet.
656 * Packet format (inner only):
657 * <'ether type'=0x0800
658 * | 'version'=4, 'MF'=1>
660 * <'ether type'=0x86DD
661 * | 'version'=6, 'next header'=44>
663 #define RTE_PTYPE_INNER_L4_FRAG 0x03000000
665 * SCTP (Stream Control Transmission Protocol) packet type.
666 * It is used for inner packet only.
668 * Packet format (inner only):
669 * <'ether type'=0x0800
670 * | 'version'=4, 'protocol'=132, 'MF'=0>
672 * <'ether type'=0x86DD
673 * | 'version'=6, 'next header'=132>
675 #define RTE_PTYPE_INNER_L4_SCTP 0x04000000
677 * ICMP (Internet Control Message Protocol) packet type.
678 * It is used for inner packet only.
680 * Packet format (inner only):
681 * <'ether type'=0x0800
682 * | 'version'=4, 'protocol'=1, 'MF'=0>
684 * <'ether type'=0x86DD
685 * | 'version'=6, 'next header'=1>
687 #define RTE_PTYPE_INNER_L4_ICMP 0x05000000
689 * Non-fragmented IP (Internet Protocol) packet type.
690 * It is used for inner packet only, and may or maynot have other unknown layer
693 * Packet format (inner only):
694 * <'ether type'=0x0800
695 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
697 * <'ether type'=0x86DD
698 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
700 #define RTE_PTYPE_INNER_L4_NONFRAG 0x06000000
702 * Mask of inner layer 4 packet types.
704 #define RTE_PTYPE_INNER_L4_MASK 0x0f000000
707 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
708 * one, bit 4 is selected to be used for IPv4 only. Then checking bit 4 can
709 * determine if it is an IPV4 packet.
711 #define RTE_ETH_IS_IPV4_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV4)
714 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
715 * one, bit 6 is selected to be used for IPv4 only. Then checking bit 6 can
716 * determine if it is an IPV4 packet.
718 #define RTE_ETH_IS_IPV6_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV6)
720 /* Check if it is a tunneling packet */
721 #define RTE_ETH_IS_TUNNEL_PKT(ptype) ((ptype) & (RTE_PTYPE_TUNNEL_MASK | \
722 RTE_PTYPE_INNER_L2_MASK | \
723 RTE_PTYPE_INNER_L3_MASK | \
724 RTE_PTYPE_INNER_L4_MASK))
726 /** Alignment constraint of mbuf private area. */
727 #define RTE_MBUF_PRIV_ALIGN 8
730 * Get the name of a RX offload flag
733 * The mask describing the flag.
735 * The name of this flag, or NULL if it's not a valid RX flag.
737 const char *rte_get_rx_ol_flag_name(uint64_t mask);
740 * Get the name of a TX offload flag
743 * The mask describing the flag. Usually only one bit must be set.
744 * Several bits can be given if they belong to the same mask.
745 * Ex: PKT_TX_L4_MASK.
747 * The name of this flag, or NULL if it's not a valid TX flag.
749 const char *rte_get_tx_ol_flag_name(uint64_t mask);
752 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
753 * splitting it into multiple segments.
754 * So, for mbufs that planned to be involved into RX/TX, the recommended
755 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
757 #define RTE_MBUF_DEFAULT_DATAROOM 2048
758 #define RTE_MBUF_DEFAULT_BUF_SIZE \
759 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
761 /* define a set of marker types that can be used to refer to set points in the
763 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
764 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
765 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
766 * with a single assignment */
769 * The generic rte_mbuf, containing a packet mbuf.
774 void *buf_addr; /**< Virtual address of segment buffer. */
775 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
777 uint16_t buf_len; /**< Length of segment buffer. */
779 /* next 6 bytes are initialised on RX descriptor rearm */
784 * 16-bit Reference counter.
785 * It should only be accessed using the following functions:
786 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
787 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
788 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
792 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
793 uint16_t refcnt; /**< Non-atomically accessed refcnt */
795 uint8_t nb_segs; /**< Number of segments. */
796 uint8_t port; /**< Input port. */
798 uint64_t ol_flags; /**< Offload features. */
800 /* remaining bytes are set on RX when pulling packet from descriptor */
801 MARKER rx_descriptor_fields1;
804 * The packet type, which is the combination of outer/inner L2, L3, L4
805 * and tunnel types. The packet_type is about data really present in the
806 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
807 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
808 * vlan is stripped from the data.
811 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
813 uint32_t l2_type:4; /**< (Outer) L2 type. */
814 uint32_t l3_type:4; /**< (Outer) L3 type. */
815 uint32_t l4_type:4; /**< (Outer) L4 type. */
816 uint32_t tun_type:4; /**< Tunnel type. */
817 uint32_t inner_l2_type:4; /**< Inner L2 type. */
818 uint32_t inner_l3_type:4; /**< Inner L3 type. */
819 uint32_t inner_l4_type:4; /**< Inner L4 type. */
823 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
824 uint16_t data_len; /**< Amount of data in segment buffer. */
825 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN_STRIPPED is set. */
829 uint32_t rss; /**< RSS hash result if RSS enabled */
837 /**< Second 4 flexible bytes */
840 /**< First 4 flexible bytes or FD ID, dependent on
841 PKT_RX_FDIR_* flag in ol_flags. */
842 } fdir; /**< Filter identifier if FDIR enabled */
846 } sched; /**< Hierarchical scheduler */
847 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
848 } hash; /**< hash information */
850 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
852 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ_STRIPPED is set. */
853 uint16_t vlan_tci_outer;
855 /* second cache line - fields only used in slow path or on TX */
856 MARKER cacheline1 __rte_cache_min_aligned;
859 void *userdata; /**< Can be used for external metadata */
860 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
863 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
864 struct rte_mbuf *next; /**< Next segment of scattered packet. */
866 /* fields to support TX offloads */
868 uint64_t tx_offload; /**< combined for easy fetch */
870 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
871 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
872 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
873 uint64_t tso_segsz:16; /**< TCP TSO segment size */
875 /* fields for TX offloading of tunnels */
876 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
877 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
879 /* uint64_t unused:8; */
883 /** Size of the application private data. In case of an indirect
884 * mbuf, it stores the direct mbuf private data size. */
887 /** Timesync flags for use with IEEE1588. */
889 } __rte_cache_aligned;
892 * Prefetch the first part of the mbuf
894 * The first 64 bytes of the mbuf corresponds to fields that are used early
895 * in the receive path. If the cache line of the architecture is higher than
896 * 64B, the second part will also be prefetched.
899 * The pointer to the mbuf.
902 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
904 rte_prefetch0(&m->cacheline0);
908 * Prefetch the second part of the mbuf
910 * The next 64 bytes of the mbuf corresponds to fields that are used in the
911 * transmit path. If the cache line of the architecture is higher than 64B,
912 * this function does nothing as it is expected that the full mbuf is
916 * The pointer to the mbuf.
919 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
921 #if RTE_CACHE_LINE_SIZE == 64
922 rte_prefetch0(&m->cacheline1);
929 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
932 * Return the DMA address of the beginning of the mbuf data
935 * The pointer to the mbuf.
937 * The physical address of the beginning of the mbuf data
939 static inline phys_addr_t
940 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
942 return mb->buf_physaddr + mb->data_off;
946 * Return the default DMA address of the beginning of the mbuf data
948 * This function is used by drivers in their receive function, as it
949 * returns the location where data should be written by the NIC, taking
950 * the default headroom in account.
953 * The pointer to the mbuf.
955 * The physical address of the beginning of the mbuf data
957 static inline phys_addr_t
958 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
960 return mb->buf_physaddr + RTE_PKTMBUF_HEADROOM;
964 * Return the mbuf owning the data buffer address of an indirect mbuf.
967 * The pointer to the indirect mbuf.
969 * The address of the direct mbuf corresponding to buffer_addr.
971 static inline struct rte_mbuf *
972 rte_mbuf_from_indirect(struct rte_mbuf *mi)
974 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
978 * Return the buffer address embedded in the given mbuf.
981 * The pointer to the mbuf.
983 * The address of the data buffer owned by the mbuf.
986 rte_mbuf_to_baddr(struct rte_mbuf *md)
989 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
994 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
996 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
999 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
1001 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
1004 * Private data in case of pktmbuf pool.
1006 * A structure that contains some pktmbuf_pool-specific data that are
1007 * appended after the mempool structure (in private data).
1009 struct rte_pktmbuf_pool_private {
1010 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
1011 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
1014 #ifdef RTE_LIBRTE_MBUF_DEBUG
1016 /** check mbuf type in debug mode */
1017 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
1019 #else /* RTE_LIBRTE_MBUF_DEBUG */
1021 /** check mbuf type in debug mode */
1022 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
1024 #endif /* RTE_LIBRTE_MBUF_DEBUG */
1026 #ifdef RTE_MBUF_REFCNT_ATOMIC
1029 * Reads the value of an mbuf's refcnt.
1033 * Reference count number.
1035 static inline uint16_t
1036 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1038 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
1042 * Sets an mbuf's refcnt to a defined value.
1049 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1051 rte_atomic16_set(&m->refcnt_atomic, new_value);
1055 * Adds given value to an mbuf's refcnt and returns its new value.
1059 * Value to add/subtract
1063 static inline uint16_t
1064 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1067 * The atomic_add is an expensive operation, so we don't want to
1068 * call it in the case where we know we are the uniq holder of
1069 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
1070 * operation has to be used because concurrent accesses on the
1071 * reference counter can occur.
1073 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1074 rte_mbuf_refcnt_set(m, 1 + value);
1078 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
1081 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
1084 * Adds given value to an mbuf's refcnt and returns its new value.
1086 static inline uint16_t
1087 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1089 m->refcnt = (uint16_t)(m->refcnt + value);
1094 * Reads the value of an mbuf's refcnt.
1096 static inline uint16_t
1097 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1103 * Sets an mbuf's refcnt to the defined value.
1106 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1108 m->refcnt = new_value;
1111 #endif /* RTE_MBUF_REFCNT_ATOMIC */
1113 /** Mbuf prefetch */
1114 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
1121 * Sanity checks on an mbuf.
1123 * Check the consistency of the given mbuf. The function will cause a
1124 * panic if corruption is detected.
1127 * The mbuf to be checked.
1129 * True if the mbuf is a packet header, false if it is a sub-segment
1130 * of a packet (in this case, some fields like nb_segs are not checked)
1133 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
1136 * Allocate an unitialized mbuf from mempool *mp*.
1138 * This function can be used by PMDs (especially in RX functions) to
1139 * allocate an unitialized mbuf. The driver is responsible of
1140 * initializing all the required fields. See rte_pktmbuf_reset().
1141 * For standard needs, prefer rte_pktmbuf_alloc().
1144 * The mempool from which mbuf is allocated.
1146 * - The pointer to the new mbuf on success.
1147 * - NULL if allocation failed.
1149 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
1154 if (rte_mempool_get(mp, &mb) < 0)
1156 m = (struct rte_mbuf *)mb;
1157 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1158 rte_mbuf_refcnt_set(m, 1);
1159 __rte_mbuf_sanity_check(m, 0);
1164 /* compat with older versions */
1165 __rte_deprecated static inline struct rte_mbuf *
1166 __rte_mbuf_raw_alloc(struct rte_mempool *mp)
1168 return rte_mbuf_raw_alloc(mp);
1172 * @internal Put mbuf back into its original mempool.
1173 * The use of that function is reserved for RTE internal needs.
1174 * Please use rte_pktmbuf_free().
1177 * The mbuf to be freed.
1179 static inline void __attribute__((always_inline))
1180 __rte_mbuf_raw_free(struct rte_mbuf *m)
1182 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1183 rte_mempool_put(m->pool, m);
1186 /* Operations on ctrl mbuf */
1189 * The control mbuf constructor.
1191 * This function initializes some fields in an mbuf structure that are
1192 * not modified by the user once created (mbuf type, origin pool, buffer
1193 * start address, and so on). This function is given as a callback function
1194 * to rte_mempool_create() at pool creation time.
1197 * The mempool from which the mbuf is allocated.
1199 * A pointer that can be used by the user to retrieve useful information
1200 * for mbuf initialization. This pointer comes from the ``init_arg``
1201 * parameter of rte_mempool_create().
1203 * The mbuf to initialize.
1205 * The index of the mbuf in the pool table.
1207 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1208 void *m, unsigned i);
1211 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
1213 * This new mbuf is initialized with data pointing to the beginning of
1214 * buffer, and with a length of zero.
1217 * The mempool from which the mbuf is allocated.
1219 * - The pointer to the new mbuf on success.
1220 * - NULL if allocation failed.
1222 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
1225 * Free a control mbuf back into its original mempool.
1228 * The control mbuf to be freed.
1230 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
1233 * A macro that returns the pointer to the carried data.
1235 * The value that can be read or assigned.
1240 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
1243 * A macro that returns the length of the carried data.
1245 * The value that can be read or assigned.
1250 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
1253 * Tests if an mbuf is a control mbuf
1256 * The mbuf to be tested
1258 * - True (1) if the mbuf is a control mbuf
1259 * - False(0) otherwise
1262 rte_is_ctrlmbuf(struct rte_mbuf *m)
1264 return !!(m->ol_flags & CTRL_MBUF_FLAG);
1267 /* Operations on pkt mbuf */
1270 * The packet mbuf constructor.
1272 * This function initializes some fields in the mbuf structure that are
1273 * not modified by the user once created (origin pool, buffer start
1274 * address, and so on). This function is given as a callback function to
1275 * rte_mempool_create() at pool creation time.
1278 * The mempool from which mbufs originate.
1280 * A pointer that can be used by the user to retrieve useful information
1281 * for mbuf initialization. This pointer comes from the ``init_arg``
1282 * parameter of rte_mempool_create().
1284 * The mbuf to initialize.
1286 * The index of the mbuf in the pool table.
1288 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1289 void *m, unsigned i);
1293 * A packet mbuf pool constructor.
1295 * This function initializes the mempool private data in the case of a
1296 * pktmbuf pool. This private data is needed by the driver. The
1297 * function is given as a callback function to rte_mempool_create() at
1298 * pool creation. It can be extended by the user, for example, to
1299 * provide another packet size.
1302 * The mempool from which mbufs originate.
1304 * A pointer that can be used by the user to retrieve useful information
1305 * for mbuf initialization. This pointer comes from the ``init_arg``
1306 * parameter of rte_mempool_create().
1308 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1311 * Create a mbuf pool.
1313 * This function creates and initializes a packet mbuf pool. It is
1314 * a wrapper to rte_mempool_create() with the proper packet constructor
1315 * and mempool constructor.
1318 * The name of the mbuf pool.
1320 * The number of elements in the mbuf pool. The optimum size (in terms
1321 * of memory usage) for a mempool is when n is a power of two minus one:
1324 * Size of the per-core object cache. See rte_mempool_create() for
1327 * Size of application private are between the rte_mbuf structure
1328 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1329 * @param data_room_size
1330 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1332 * The socket identifier where the memory should be allocated. The
1333 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1336 * The pointer to the new allocated mempool, on success. NULL on error
1337 * with rte_errno set appropriately. Possible rte_errno values include:
1338 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1339 * - E_RTE_SECONDARY - function was called from a secondary process instance
1340 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1341 * - ENOSPC - the maximum number of memzones has already been allocated
1342 * - EEXIST - a memzone with the same name already exists
1343 * - ENOMEM - no appropriate memory area found in which to create memzone
1345 struct rte_mempool *
1346 rte_pktmbuf_pool_create(const char *name, unsigned n,
1347 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1351 * Get the data room size of mbufs stored in a pktmbuf_pool
1353 * The data room size is the amount of data that can be stored in a
1354 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1357 * The packet mbuf pool.
1359 * The data room size of mbufs stored in this mempool.
1361 static inline uint16_t
1362 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1364 struct rte_pktmbuf_pool_private *mbp_priv;
1366 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1367 return mbp_priv->mbuf_data_room_size;
1371 * Get the application private size of mbufs stored in a pktmbuf_pool
1373 * The private size of mbuf is a zone located between the rte_mbuf
1374 * structure and the data buffer where an application can store data
1375 * associated to a packet.
1378 * The packet mbuf pool.
1380 * The private size of mbufs stored in this mempool.
1382 static inline uint16_t
1383 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1385 struct rte_pktmbuf_pool_private *mbp_priv;
1387 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1388 return mbp_priv->mbuf_priv_size;
1392 * Reset the fields of a packet mbuf to their default values.
1394 * The given mbuf must have only one segment.
1397 * The packet mbuf to be resetted.
1399 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1405 m->vlan_tci_outer = 0;
1411 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
1412 RTE_PKTMBUF_HEADROOM : m->buf_len;
1415 __rte_mbuf_sanity_check(m, 1);
1419 * Allocate a new mbuf from a mempool.
1421 * This new mbuf contains one segment, which has a length of 0. The pointer
1422 * to data is initialized to have some bytes of headroom in the buffer
1423 * (if buffer size allows).
1426 * The mempool from which the mbuf is allocated.
1428 * - The pointer to the new mbuf on success.
1429 * - NULL if allocation failed.
1431 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1434 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1435 rte_pktmbuf_reset(m);
1440 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1444 * The mempool from which mbufs are allocated.
1446 * Array of pointers to mbufs
1452 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1453 struct rte_mbuf **mbufs, unsigned count)
1458 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1462 /* To understand duff's device on loop unwinding optimization, see
1463 * https://en.wikipedia.org/wiki/Duff's_device.
1464 * Here while() loop is used rather than do() while{} to avoid extra
1465 * check if count is zero.
1467 switch (count % 4) {
1469 while (idx != count) {
1470 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1471 rte_mbuf_refcnt_set(mbufs[idx], 1);
1472 rte_pktmbuf_reset(mbufs[idx]);
1475 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1476 rte_mbuf_refcnt_set(mbufs[idx], 1);
1477 rte_pktmbuf_reset(mbufs[idx]);
1480 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1481 rte_mbuf_refcnt_set(mbufs[idx], 1);
1482 rte_pktmbuf_reset(mbufs[idx]);
1485 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1486 rte_mbuf_refcnt_set(mbufs[idx], 1);
1487 rte_pktmbuf_reset(mbufs[idx]);
1495 * Attach packet mbuf to another packet mbuf.
1497 * After attachment we refer the mbuf we attached as 'indirect',
1498 * while mbuf we attached to as 'direct'.
1499 * The direct mbuf's reference counter is incremented.
1501 * Right now, not supported:
1502 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1503 * - mbuf we trying to attach (mi) is used by someone else
1504 * e.g. it's reference counter is greater then 1.
1507 * The indirect packet mbuf.
1509 * The packet mbuf we're attaching to.
1511 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1513 struct rte_mbuf *md;
1515 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1516 rte_mbuf_refcnt_read(mi) == 1);
1518 /* if m is not direct, get the mbuf that embeds the data */
1519 if (RTE_MBUF_DIRECT(m))
1522 md = rte_mbuf_from_indirect(m);
1524 rte_mbuf_refcnt_update(md, 1);
1525 mi->priv_size = m->priv_size;
1526 mi->buf_physaddr = m->buf_physaddr;
1527 mi->buf_addr = m->buf_addr;
1528 mi->buf_len = m->buf_len;
1531 mi->data_off = m->data_off;
1532 mi->data_len = m->data_len;
1534 mi->vlan_tci = m->vlan_tci;
1535 mi->vlan_tci_outer = m->vlan_tci_outer;
1536 mi->tx_offload = m->tx_offload;
1540 mi->pkt_len = mi->data_len;
1542 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1543 mi->packet_type = m->packet_type;
1545 __rte_mbuf_sanity_check(mi, 1);
1546 __rte_mbuf_sanity_check(m, 0);
1550 * Detach an indirect packet mbuf.
1552 * - restore original mbuf address and length values.
1553 * - reset pktmbuf data and data_len to their default values.
1554 * - decrement the direct mbuf's reference counter. When the
1555 * reference counter becomes 0, the direct mbuf is freed.
1557 * All other fields of the given packet mbuf will be left intact.
1560 * The indirect attached packet mbuf.
1562 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1564 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1565 struct rte_mempool *mp = m->pool;
1566 uint32_t mbuf_size, buf_len, priv_size;
1568 priv_size = rte_pktmbuf_priv_size(mp);
1569 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1570 buf_len = rte_pktmbuf_data_room_size(mp);
1572 m->priv_size = priv_size;
1573 m->buf_addr = (char *)m + mbuf_size;
1574 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
1575 m->buf_len = (uint16_t)buf_len;
1576 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1580 if (rte_mbuf_refcnt_update(md, -1) == 0)
1581 __rte_mbuf_raw_free(md);
1584 static inline struct rte_mbuf* __attribute__((always_inline))
1585 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1587 __rte_mbuf_sanity_check(m, 0);
1589 if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
1590 /* if this is an indirect mbuf, it is detached. */
1591 if (RTE_MBUF_INDIRECT(m))
1592 rte_pktmbuf_detach(m);
1599 * Free a segment of a packet mbuf into its original mempool.
1601 * Free an mbuf, without parsing other segments in case of chained
1605 * The packet mbuf segment to be freed.
1607 static inline void __attribute__((always_inline))
1608 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1610 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
1612 __rte_mbuf_raw_free(m);
1617 * Free a packet mbuf back into its original mempool.
1619 * Free an mbuf, and all its segments in case of chained buffers. Each
1620 * segment is added back into its original mempool.
1623 * The packet mbuf to be freed.
1625 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1627 struct rte_mbuf *m_next;
1629 __rte_mbuf_sanity_check(m, 1);
1633 rte_pktmbuf_free_seg(m);
1639 * Creates a "clone" of the given packet mbuf.
1641 * Walks through all segments of the given packet mbuf, and for each of them:
1642 * - Creates a new packet mbuf from the given pool.
1643 * - Attaches newly created mbuf to the segment.
1644 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1645 * from the original packet mbuf.
1648 * The packet mbuf to be cloned.
1650 * The mempool from which the "clone" mbufs are allocated.
1652 * - The pointer to the new "clone" mbuf on success.
1653 * - NULL if allocation fails.
1655 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1656 struct rte_mempool *mp)
1658 struct rte_mbuf *mc, *mi, **prev;
1662 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1667 pktlen = md->pkt_len;
1672 rte_pktmbuf_attach(mi, md);
1675 } while ((md = md->next) != NULL &&
1676 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1680 mc->pkt_len = pktlen;
1682 /* Allocation of new indirect segment failed */
1683 if (unlikely (mi == NULL)) {
1684 rte_pktmbuf_free(mc);
1688 __rte_mbuf_sanity_check(mc, 1);
1693 * Adds given value to the refcnt of all packet mbuf segments.
1695 * Walks through all segments of given packet mbuf and for each of them
1696 * invokes rte_mbuf_refcnt_update().
1699 * The packet mbuf whose refcnt to be updated.
1701 * The value to add to the mbuf's segments refcnt.
1703 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1705 __rte_mbuf_sanity_check(m, 1);
1708 rte_mbuf_refcnt_update(m, v);
1709 } while ((m = m->next) != NULL);
1713 * Get the headroom in a packet mbuf.
1718 * The length of the headroom.
1720 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1722 __rte_mbuf_sanity_check(m, 1);
1727 * Get the tailroom of a packet mbuf.
1732 * The length of the tailroom.
1734 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1736 __rte_mbuf_sanity_check(m, 1);
1737 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1742 * Get the last segment of the packet.
1747 * The last segment of the given mbuf.
1749 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1751 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1753 __rte_mbuf_sanity_check(m, 1);
1754 while (m2->next != NULL)
1760 * A macro that points to an offset into the data in the mbuf.
1762 * The returned pointer is cast to type t. Before using this
1763 * function, the user must ensure that the first segment is large
1764 * enough to accommodate its data.
1769 * The offset into the mbuf data.
1771 * The type to cast the result into.
1773 #define rte_pktmbuf_mtod_offset(m, t, o) \
1774 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1777 * A macro that points to the start of the data in the mbuf.
1779 * The returned pointer is cast to type t. Before using this
1780 * function, the user must ensure that the first segment is large
1781 * enough to accommodate its data.
1786 * The type to cast the result into.
1788 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1791 * A macro that returns the physical address that points to an offset of the
1792 * start of the data in the mbuf
1797 * The offset into the data to calculate address from.
1799 #define rte_pktmbuf_mtophys_offset(m, o) \
1800 (phys_addr_t)((m)->buf_physaddr + (m)->data_off + (o))
1803 * A macro that returns the physical address that points to the start of the
1809 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_mtophys_offset(m, 0)
1812 * A macro that returns the length of the packet.
1814 * The value can be read or assigned.
1819 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1822 * A macro that returns the length of the segment.
1824 * The value can be read or assigned.
1829 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1832 * Prepend len bytes to an mbuf data area.
1834 * Returns a pointer to the new
1835 * data start address. If there is not enough headroom in the first
1836 * segment, the function will return NULL, without modifying the mbuf.
1841 * The amount of data to prepend (in bytes).
1843 * A pointer to the start of the newly prepended data, or
1844 * NULL if there is not enough headroom space in the first segment
1846 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1849 __rte_mbuf_sanity_check(m, 1);
1851 if (unlikely(len > rte_pktmbuf_headroom(m)))
1855 m->data_len = (uint16_t)(m->data_len + len);
1856 m->pkt_len = (m->pkt_len + len);
1858 return (char *)m->buf_addr + m->data_off;
1862 * Append len bytes to an mbuf.
1864 * Append len bytes to an mbuf and return a pointer to the start address
1865 * of the added data. If there is not enough tailroom in the last
1866 * segment, the function will return NULL, without modifying the mbuf.
1871 * The amount of data to append (in bytes).
1873 * A pointer to the start of the newly appended data, or
1874 * NULL if there is not enough tailroom space in the last segment
1876 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1879 struct rte_mbuf *m_last;
1881 __rte_mbuf_sanity_check(m, 1);
1883 m_last = rte_pktmbuf_lastseg(m);
1884 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1887 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1888 m_last->data_len = (uint16_t)(m_last->data_len + len);
1889 m->pkt_len = (m->pkt_len + len);
1890 return (char*) tail;
1894 * Remove len bytes at the beginning of an mbuf.
1896 * Returns a pointer to the start address of the new data area. If the
1897 * length is greater than the length of the first segment, then the
1898 * function will fail and return NULL, without modifying the mbuf.
1903 * The amount of data to remove (in bytes).
1905 * A pointer to the new start of the data.
1907 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1909 __rte_mbuf_sanity_check(m, 1);
1911 if (unlikely(len > m->data_len))
1914 m->data_len = (uint16_t)(m->data_len - len);
1916 m->pkt_len = (m->pkt_len - len);
1917 return (char *)m->buf_addr + m->data_off;
1921 * Remove len bytes of data at the end of the mbuf.
1923 * If the length is greater than the length of the last segment, the
1924 * function will fail and return -1 without modifying the mbuf.
1929 * The amount of data to remove (in bytes).
1934 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1936 struct rte_mbuf *m_last;
1938 __rte_mbuf_sanity_check(m, 1);
1940 m_last = rte_pktmbuf_lastseg(m);
1941 if (unlikely(len > m_last->data_len))
1944 m_last->data_len = (uint16_t)(m_last->data_len - len);
1945 m->pkt_len = (m->pkt_len - len);
1950 * Test if mbuf data is contiguous.
1955 * - 1, if all data is contiguous (one segment).
1956 * - 0, if there is several segments.
1958 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1960 __rte_mbuf_sanity_check(m, 1);
1961 return !!(m->nb_segs == 1);
1965 * Chain an mbuf to another, thereby creating a segmented packet.
1967 * Note: The implementation will do a linear walk over the segments to find
1968 * the tail entry. For cases when there are many segments, it's better to
1969 * chain the entries manually.
1972 * The head of the mbuf chain (the first packet)
1974 * The mbuf to put last in the chain
1978 * - -EOVERFLOW, if the chain is full (256 entries)
1980 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1982 struct rte_mbuf *cur_tail;
1984 /* Check for number-of-segments-overflow */
1985 if (head->nb_segs + tail->nb_segs >= 1 << (sizeof(head->nb_segs) * 8))
1988 /* Chain 'tail' onto the old tail */
1989 cur_tail = rte_pktmbuf_lastseg(head);
1990 cur_tail->next = tail;
1992 /* accumulate number of segments and total length. */
1993 head->nb_segs = (uint8_t)(head->nb_segs + tail->nb_segs);
1994 head->pkt_len += tail->pkt_len;
1996 /* pkt_len is only set in the head */
1997 tail->pkt_len = tail->data_len;
2003 * Dump an mbuf structure to the console.
2005 * Dump all fields for the given packet mbuf and all its associated
2006 * segments (in the case of a chained buffer).
2009 * A pointer to a file for output
2013 * If dump_len != 0, also dump the "dump_len" first data bytes of
2016 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
2022 #endif /* _RTE_MBUF_H_ */