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().
82 #define PKT_RX_VLAN_PKT (1ULL << 0) /**< RX packet is a 802.1q VLAN packet. */
83 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
84 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
85 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
86 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
87 #define PKT_RX_EIP_CKSUM_BAD (1ULL << 5) /**< External IP header checksum error. */
88 #define PKT_RX_OVERSIZE (0ULL << 0) /**< Num of desc of an RX pkt oversize. */
89 #define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
90 #define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
91 #define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
92 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
93 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
94 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
95 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
96 #define PKT_RX_QINQ_PKT (1ULL << 15) /**< RX packet with double VLAN stripped. */
97 /* add new RX flags here */
99 /* add new TX flags here */
102 * Second VLAN insertion (QinQ) flag.
104 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
107 * TCP segmentation offload. To enable this offload feature for a
108 * packet to be transmitted on hardware supporting TSO:
109 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
111 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
112 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
114 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
115 * - calculate the pseudo header checksum without taking ip_len in account,
116 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
117 * rte_ipv6_phdr_cksum() that can be used as helpers.
119 #define PKT_TX_TCP_SEG (1ULL << 50)
121 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
124 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
125 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
126 * L4 checksum offload, the user needs to:
127 * - fill l2_len and l3_len in mbuf
128 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
129 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
130 * - calculate the pseudo header checksum and set it in the L4 header (only
131 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
132 * For SCTP, set the crc field to 0.
134 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
135 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
136 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
137 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
138 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
141 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
142 * also be set by the application, although a PMD will only check
144 * - set the IP checksum field in the packet to 0
145 * - fill the mbuf offload information: l2_len, l3_len
147 #define PKT_TX_IP_CKSUM (1ULL << 54)
150 * Packet is IPv4. This flag must be set when using any offload feature
151 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
152 * packet. If the packet is a tunneled packet, this flag is related to
155 #define PKT_TX_IPV4 (1ULL << 55)
158 * Packet is IPv6. This flag must be set when using an offload feature
159 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
160 * packet. If the packet is a tunneled packet, this flag is related to
163 #define PKT_TX_IPV6 (1ULL << 56)
165 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
168 * Offload the IP checksum of an external header in the hardware. The
169 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
170 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
171 * packet must be set to 0.
172 * - set the outer IP checksum field in the packet to 0
173 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
175 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
178 * Packet outer header is IPv4. This flag must be set when using any
179 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
180 * outer header of the tunneled packet is an IPv4 packet.
182 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
185 * Packet outer header is IPv6. This flag must be set when using any
186 * outer offload feature (L4 checksum) to tell the NIC that the outer
187 * header of the tunneled packet is an IPv6 packet.
189 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
191 #define __RESERVED (1ULL << 61) /**< reserved for future mbuf use */
193 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
195 /* Use final bit of flags to indicate a control mbuf */
196 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
199 * 32 bits are divided into several fields to mark packet types. Note that
200 * each field is indexical.
201 * - Bit 3:0 is for L2 types.
202 * - Bit 7:4 is for L3 or outer L3 (for tunneling case) types.
203 * - Bit 11:8 is for L4 or outer L4 (for tunneling case) types.
204 * - Bit 15:12 is for tunnel types.
205 * - Bit 19:16 is for inner L2 types.
206 * - Bit 23:20 is for inner L3 types.
207 * - Bit 27:24 is for inner L4 types.
208 * - Bit 31:28 is reserved.
210 * To be compatible with Vector PMD, RTE_PTYPE_L3_IPV4, RTE_PTYPE_L3_IPV4_EXT,
211 * RTE_PTYPE_L3_IPV6, RTE_PTYPE_L3_IPV6_EXT, RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP
212 * and RTE_PTYPE_L4_SCTP should be kept as below in a contiguous 7 bits.
214 * Note that L3 types values are selected for checking IPV4/IPV6 header from
215 * performance point of view. Reading annotations of RTE_ETH_IS_IPV4_HDR and
216 * RTE_ETH_IS_IPV6_HDR is needed for any future changes of L3 type values.
218 * Note that the packet types of the same packet recognized by different
219 * hardware may be different, as different hardware may have different
220 * capability of packet type recognition.
223 * <'ether type'=0x0800
224 * | 'version'=4, 'protocol'=0x29
225 * | 'version'=6, 'next header'=0x3A
227 * will be recognized on i40e hardware as packet type combination of,
228 * RTE_PTYPE_L2_ETHER |
229 * RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
230 * RTE_PTYPE_TUNNEL_IP |
231 * RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
232 * RTE_PTYPE_INNER_L4_ICMP.
234 * <'ether type'=0x86DD
235 * | 'version'=6, 'next header'=0x2F
237 * | 'version'=6, 'next header'=0x11
239 * will be recognized on i40e hardware as packet type combination of,
240 * RTE_PTYPE_L2_ETHER |
241 * RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
242 * RTE_PTYPE_TUNNEL_GRENAT |
243 * RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
244 * RTE_PTYPE_INNER_L4_UDP.
246 #define RTE_PTYPE_UNKNOWN 0x00000000
248 * Ethernet packet type.
249 * It is used for outer packet for tunneling cases.
252 * <'ether type'=[0x0800|0x86DD]>
254 #define RTE_PTYPE_L2_ETHER 0x00000001
256 * Ethernet packet type for time sync.
259 * <'ether type'=0x88F7>
261 #define RTE_PTYPE_L2_ETHER_TIMESYNC 0x00000002
263 * ARP (Address Resolution Protocol) packet type.
266 * <'ether type'=0x0806>
268 #define RTE_PTYPE_L2_ETHER_ARP 0x00000003
270 * LLDP (Link Layer Discovery Protocol) packet type.
273 * <'ether type'=0x88CC>
275 #define RTE_PTYPE_L2_ETHER_LLDP 0x00000004
277 * Mask of layer 2 packet types.
278 * It is used for outer packet for tunneling cases.
280 #define RTE_PTYPE_L2_MASK 0x0000000f
282 * IP (Internet Protocol) version 4 packet type.
283 * It is used for outer packet for tunneling cases, and does not contain any
287 * <'ether type'=0x0800
288 * | 'version'=4, 'ihl'=5>
290 #define RTE_PTYPE_L3_IPV4 0x00000010
292 * IP (Internet Protocol) version 4 packet type.
293 * It is used for outer packet for tunneling cases, and contains header
297 * <'ether type'=0x0800
298 * | 'version'=4, 'ihl'=[6-15], 'options'>
300 #define RTE_PTYPE_L3_IPV4_EXT 0x00000030
302 * IP (Internet Protocol) version 6 packet type.
303 * It is used for outer packet for tunneling cases, and does not contain any
307 * <'ether type'=0x86DD
308 * | 'version'=6, 'next header'=0x3B>
310 #define RTE_PTYPE_L3_IPV6 0x00000040
312 * IP (Internet Protocol) version 4 packet type.
313 * It is used for outer packet for tunneling cases, and may or maynot contain
317 * <'ether type'=0x0800
318 * | 'version'=4, 'ihl'=[5-15], <'options'>>
320 #define RTE_PTYPE_L3_IPV4_EXT_UNKNOWN 0x00000090
322 * IP (Internet Protocol) version 6 packet type.
323 * It is used for outer packet for tunneling cases, and contains extension
327 * <'ether type'=0x86DD
328 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
329 * 'extension headers'>
331 #define RTE_PTYPE_L3_IPV6_EXT 0x000000c0
333 * IP (Internet Protocol) version 6 packet type.
334 * It is used for outer packet for tunneling cases, and may or maynot contain
338 * <'ether type'=0x86DD
339 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
340 * <'extension headers'>>
342 #define RTE_PTYPE_L3_IPV6_EXT_UNKNOWN 0x000000e0
344 * Mask of layer 3 packet types.
345 * It is used for outer packet for tunneling cases.
347 #define RTE_PTYPE_L3_MASK 0x000000f0
349 * TCP (Transmission Control Protocol) packet type.
350 * It is used for outer packet for tunneling cases.
353 * <'ether type'=0x0800
354 * | 'version'=4, 'protocol'=6, 'MF'=0>
356 * <'ether type'=0x86DD
357 * | 'version'=6, 'next header'=6>
359 #define RTE_PTYPE_L4_TCP 0x00000100
361 * UDP (User Datagram Protocol) packet type.
362 * It is used for outer packet for tunneling cases.
365 * <'ether type'=0x0800
366 * | 'version'=4, 'protocol'=17, 'MF'=0>
368 * <'ether type'=0x86DD
369 * | 'version'=6, 'next header'=17>
371 #define RTE_PTYPE_L4_UDP 0x00000200
373 * Fragmented IP (Internet Protocol) packet type.
374 * It is used for outer packet for tunneling cases.
376 * It refers to those packets of any IP types, which can be recognized as
377 * fragmented. A fragmented packet cannot be recognized as any other L4 types
378 * (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP,
379 * RTE_PTYPE_L4_NONFRAG).
382 * <'ether type'=0x0800
383 * | 'version'=4, 'MF'=1>
385 * <'ether type'=0x86DD
386 * | 'version'=6, 'next header'=44>
388 #define RTE_PTYPE_L4_FRAG 0x00000300
390 * SCTP (Stream Control Transmission Protocol) packet type.
391 * It is used for outer packet for tunneling cases.
394 * <'ether type'=0x0800
395 * | 'version'=4, 'protocol'=132, 'MF'=0>
397 * <'ether type'=0x86DD
398 * | 'version'=6, 'next header'=132>
400 #define RTE_PTYPE_L4_SCTP 0x00000400
402 * ICMP (Internet Control Message Protocol) packet type.
403 * It is used for outer packet for tunneling cases.
406 * <'ether type'=0x0800
407 * | 'version'=4, 'protocol'=1, 'MF'=0>
409 * <'ether type'=0x86DD
410 * | 'version'=6, 'next header'=1>
412 #define RTE_PTYPE_L4_ICMP 0x00000500
414 * Non-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, while cannot be recognized as
418 * any of above L4 types (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP,
419 * RTE_PTYPE_L4_FRAG, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP).
422 * <'ether type'=0x0800
423 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
425 * <'ether type'=0x86DD
426 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
428 #define RTE_PTYPE_L4_NONFRAG 0x00000600
430 * Mask of layer 4 packet types.
431 * It is used for outer packet for tunneling cases.
433 #define RTE_PTYPE_L4_MASK 0x00000f00
435 * IP (Internet Protocol) in IP (Internet Protocol) tunneling packet type.
438 * <'ether type'=0x0800
439 * | 'version'=4, 'protocol'=[4|41]>
441 * <'ether type'=0x86DD
442 * | 'version'=6, 'next header'=[4|41]>
444 #define RTE_PTYPE_TUNNEL_IP 0x00001000
446 * GRE (Generic Routing Encapsulation) tunneling packet type.
449 * <'ether type'=0x0800
450 * | 'version'=4, 'protocol'=47>
452 * <'ether type'=0x86DD
453 * | 'version'=6, 'next header'=47>
455 #define RTE_PTYPE_TUNNEL_GRE 0x00002000
457 * VXLAN (Virtual eXtensible Local Area Network) tunneling packet type.
460 * <'ether type'=0x0800
461 * | 'version'=4, 'protocol'=17
462 * | 'destination port'=4798>
464 * <'ether type'=0x86DD
465 * | 'version'=6, 'next header'=17
466 * | 'destination port'=4798>
468 #define RTE_PTYPE_TUNNEL_VXLAN 0x00003000
470 * NVGRE (Network Virtualization using Generic Routing Encapsulation) tunneling
474 * <'ether type'=0x0800
475 * | 'version'=4, 'protocol'=47
476 * | 'protocol type'=0x6558>
478 * <'ether type'=0x86DD
479 * | 'version'=6, 'next header'=47
480 * | 'protocol type'=0x6558'>
482 #define RTE_PTYPE_TUNNEL_NVGRE 0x00004000
484 * GENEVE (Generic Network Virtualization Encapsulation) tunneling packet type.
487 * <'ether type'=0x0800
488 * | 'version'=4, 'protocol'=17
489 * | 'destination port'=6081>
491 * <'ether type'=0x86DD
492 * | 'version'=6, 'next header'=17
493 * | 'destination port'=6081>
495 #define RTE_PTYPE_TUNNEL_GENEVE 0x00005000
497 * Tunneling packet type of Teredo, VXLAN (Virtual eXtensible Local Area
498 * Network) or GRE (Generic Routing Encapsulation) could be recognized as this
499 * packet type, if they can not be recognized independently as of hardware
502 #define RTE_PTYPE_TUNNEL_GRENAT 0x00006000
504 * Mask of tunneling packet types.
506 #define RTE_PTYPE_TUNNEL_MASK 0x0000f000
508 * Ethernet packet type.
509 * It is used for inner packet type only.
511 * Packet format (inner only):
512 * <'ether type'=[0x800|0x86DD]>
514 #define RTE_PTYPE_INNER_L2_ETHER 0x00010000
516 * Ethernet packet type with VLAN (Virtual Local Area Network) tag.
518 * Packet format (inner only):
519 * <'ether type'=[0x800|0x86DD], vlan=[1-4095]>
521 #define RTE_PTYPE_INNER_L2_ETHER_VLAN 0x00020000
523 * Mask of inner layer 2 packet types.
525 #define RTE_PTYPE_INNER_L2_MASK 0x000f0000
527 * IP (Internet Protocol) version 4 packet type.
528 * It is used for inner packet only, and does not contain any header option.
530 * Packet format (inner only):
531 * <'ether type'=0x0800
532 * | 'version'=4, 'ihl'=5>
534 #define RTE_PTYPE_INNER_L3_IPV4 0x00100000
536 * IP (Internet Protocol) version 4 packet type.
537 * It is used for inner packet only, and contains header options.
539 * Packet format (inner only):
540 * <'ether type'=0x0800
541 * | 'version'=4, 'ihl'=[6-15], 'options'>
543 #define RTE_PTYPE_INNER_L3_IPV4_EXT 0x00200000
545 * IP (Internet Protocol) version 6 packet type.
546 * It is used for inner packet only, and does not contain any extension header.
548 * Packet format (inner only):
549 * <'ether type'=0x86DD
550 * | 'version'=6, 'next header'=0x3B>
552 #define RTE_PTYPE_INNER_L3_IPV6 0x00300000
554 * IP (Internet Protocol) version 4 packet type.
555 * It is used for inner packet only, and may or maynot contain header options.
557 * Packet format (inner only):
558 * <'ether type'=0x0800
559 * | 'version'=4, 'ihl'=[5-15], <'options'>>
561 #define RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN 0x00400000
563 * IP (Internet Protocol) version 6 packet type.
564 * It is used for inner packet only, and contains extension headers.
566 * Packet format (inner only):
567 * <'ether type'=0x86DD
568 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
569 * 'extension headers'>
571 #define RTE_PTYPE_INNER_L3_IPV6_EXT 0x00500000
573 * IP (Internet Protocol) version 6 packet type.
574 * It is used for inner packet only, and may or maynot contain extension
577 * Packet format (inner only):
578 * <'ether type'=0x86DD
579 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
580 * <'extension headers'>>
582 #define RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN 0x00600000
584 * Mask of inner layer 3 packet types.
586 #define RTE_PTYPE_INNER_L3_MASK 0x00f00000
588 * TCP (Transmission Control Protocol) packet type.
589 * It is used for inner packet only.
591 * Packet format (inner only):
592 * <'ether type'=0x0800
593 * | 'version'=4, 'protocol'=6, 'MF'=0>
595 * <'ether type'=0x86DD
596 * | 'version'=6, 'next header'=6>
598 #define RTE_PTYPE_INNER_L4_TCP 0x01000000
600 * UDP (User Datagram Protocol) packet type.
601 * It is used for inner packet only.
603 * Packet format (inner only):
604 * <'ether type'=0x0800
605 * | 'version'=4, 'protocol'=17, 'MF'=0>
607 * <'ether type'=0x86DD
608 * | 'version'=6, 'next header'=17>
610 #define RTE_PTYPE_INNER_L4_UDP 0x02000000
612 * Fragmented IP (Internet Protocol) packet type.
613 * It is used for inner packet only, and may or maynot have layer 4 packet.
615 * Packet format (inner only):
616 * <'ether type'=0x0800
617 * | 'version'=4, 'MF'=1>
619 * <'ether type'=0x86DD
620 * | 'version'=6, 'next header'=44>
622 #define RTE_PTYPE_INNER_L4_FRAG 0x03000000
624 * SCTP (Stream Control Transmission Protocol) packet type.
625 * It is used for inner packet only.
627 * Packet format (inner only):
628 * <'ether type'=0x0800
629 * | 'version'=4, 'protocol'=132, 'MF'=0>
631 * <'ether type'=0x86DD
632 * | 'version'=6, 'next header'=132>
634 #define RTE_PTYPE_INNER_L4_SCTP 0x04000000
636 * ICMP (Internet Control Message Protocol) packet type.
637 * It is used for inner packet only.
639 * Packet format (inner only):
640 * <'ether type'=0x0800
641 * | 'version'=4, 'protocol'=1, 'MF'=0>
643 * <'ether type'=0x86DD
644 * | 'version'=6, 'next header'=1>
646 #define RTE_PTYPE_INNER_L4_ICMP 0x05000000
648 * Non-fragmented IP (Internet Protocol) packet type.
649 * It is used for inner packet only, and may or maynot have other unknown layer
652 * Packet format (inner only):
653 * <'ether type'=0x0800
654 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
656 * <'ether type'=0x86DD
657 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
659 #define RTE_PTYPE_INNER_L4_NONFRAG 0x06000000
661 * Mask of inner layer 4 packet types.
663 #define RTE_PTYPE_INNER_L4_MASK 0x0f000000
666 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
667 * one, bit 4 is selected to be used for IPv4 only. Then checking bit 4 can
668 * determine if it is an IPV4 packet.
670 #define RTE_ETH_IS_IPV4_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV4)
673 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
674 * one, bit 6 is selected to be used for IPv4 only. Then checking bit 6 can
675 * determine if it is an IPV4 packet.
677 #define RTE_ETH_IS_IPV6_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV6)
679 /* Check if it is a tunneling packet */
680 #define RTE_ETH_IS_TUNNEL_PKT(ptype) ((ptype) & (RTE_PTYPE_TUNNEL_MASK | \
681 RTE_PTYPE_INNER_L2_MASK | \
682 RTE_PTYPE_INNER_L3_MASK | \
683 RTE_PTYPE_INNER_L4_MASK))
685 /** Alignment constraint of mbuf private area. */
686 #define RTE_MBUF_PRIV_ALIGN 8
689 * Get the name of a RX offload flag
692 * The mask describing the flag.
694 * The name of this flag, or NULL if it's not a valid RX flag.
696 const char *rte_get_rx_ol_flag_name(uint64_t mask);
699 * Get the name of a TX offload flag
702 * The mask describing the flag. Usually only one bit must be set.
703 * Several bits can be given if they belong to the same mask.
704 * Ex: PKT_TX_L4_MASK.
706 * The name of this flag, or NULL if it's not a valid TX flag.
708 const char *rte_get_tx_ol_flag_name(uint64_t mask);
711 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
712 * splitting it into multiple segments.
713 * So, for mbufs that planned to be involved into RX/TX, the recommended
714 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
716 #define RTE_MBUF_DEFAULT_DATAROOM 2048
717 #define RTE_MBUF_DEFAULT_BUF_SIZE \
718 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
720 /* define a set of marker types that can be used to refer to set points in the
722 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
723 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
724 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
725 * with a single assignment */
728 * The generic rte_mbuf, containing a packet mbuf.
733 void *buf_addr; /**< Virtual address of segment buffer. */
734 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
736 uint16_t buf_len; /**< Length of segment buffer. */
738 /* next 6 bytes are initialised on RX descriptor rearm */
743 * 16-bit Reference counter.
744 * It should only be accessed using the following functions:
745 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
746 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
747 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
751 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
752 uint16_t refcnt; /**< Non-atomically accessed refcnt */
754 uint8_t nb_segs; /**< Number of segments. */
755 uint8_t port; /**< Input port. */
757 uint64_t ol_flags; /**< Offload features. */
759 /* remaining bytes are set on RX when pulling packet from descriptor */
760 MARKER rx_descriptor_fields1;
763 * The packet type, which is the combination of outer/inner L2, L3, L4
767 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
769 uint32_t l2_type:4; /**< (Outer) L2 type. */
770 uint32_t l3_type:4; /**< (Outer) L3 type. */
771 uint32_t l4_type:4; /**< (Outer) L4 type. */
772 uint32_t tun_type:4; /**< Tunnel type. */
773 uint32_t inner_l2_type:4; /**< Inner L2 type. */
774 uint32_t inner_l3_type:4; /**< Inner L3 type. */
775 uint32_t inner_l4_type:4; /**< Inner L4 type. */
779 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
780 uint16_t data_len; /**< Amount of data in segment buffer. */
781 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
784 uint32_t rss; /**< RSS hash result if RSS enabled */
792 /**< Second 4 flexible bytes */
795 /**< First 4 flexible bytes or FD ID, dependent on
796 PKT_RX_FDIR_* flag in ol_flags. */
797 } fdir; /**< Filter identifier if FDIR enabled */
801 } sched; /**< Hierarchical scheduler */
802 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
803 } hash; /**< hash information */
805 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
807 uint16_t vlan_tci_outer; /**< Outer VLAN Tag Control Identifier (CPU order) */
809 /* second cache line - fields only used in slow path or on TX */
810 MARKER cacheline1 __rte_cache_min_aligned;
813 void *userdata; /**< Can be used for external metadata */
814 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
817 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
818 struct rte_mbuf *next; /**< Next segment of scattered packet. */
820 /* fields to support TX offloads */
822 uint64_t tx_offload; /**< combined for easy fetch */
824 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
825 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
826 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
827 uint64_t tso_segsz:16; /**< TCP TSO segment size */
829 /* fields for TX offloading of tunnels */
830 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
831 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
833 /* uint64_t unused:8; */
837 /** Size of the application private data. In case of an indirect
838 * mbuf, it stores the direct mbuf private data size. */
841 /** Timesync flags for use with IEEE1588. */
843 } __rte_cache_aligned;
845 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
848 * Return the DMA address of the beginning of the mbuf data
851 * The pointer to the mbuf.
853 * The physical address of the beginning of the mbuf data
855 static inline phys_addr_t
856 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
858 return mb->buf_physaddr + mb->data_off;
862 * Return the default DMA address of the beginning of the mbuf data
864 * This function is used by drivers in their receive function, as it
865 * returns the location where data should be written by the NIC, taking
866 * the default headroom in account.
869 * The pointer to the mbuf.
871 * The physical address of the beginning of the mbuf data
873 static inline phys_addr_t
874 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
876 return mb->buf_physaddr + RTE_PKTMBUF_HEADROOM;
880 * Return the mbuf owning the data buffer address of an indirect mbuf.
883 * The pointer to the indirect mbuf.
885 * The address of the direct mbuf corresponding to buffer_addr.
887 static inline struct rte_mbuf *
888 rte_mbuf_from_indirect(struct rte_mbuf *mi)
890 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
894 * Return the buffer address embedded in the given mbuf.
897 * The pointer to the mbuf.
899 * The address of the data buffer owned by the mbuf.
902 rte_mbuf_to_baddr(struct rte_mbuf *md)
905 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
910 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
912 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
915 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
917 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
920 * Private data in case of pktmbuf pool.
922 * A structure that contains some pktmbuf_pool-specific data that are
923 * appended after the mempool structure (in private data).
925 struct rte_pktmbuf_pool_private {
926 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
927 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
930 #ifdef RTE_LIBRTE_MBUF_DEBUG
932 /** check mbuf type in debug mode */
933 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
935 /** check mbuf type in debug mode if mbuf pointer is not null */
936 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
938 rte_mbuf_sanity_check(m, is_h); \
941 /** MBUF asserts in debug mode */
942 #define RTE_MBUF_ASSERT(exp) \
944 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
947 #else /* RTE_LIBRTE_MBUF_DEBUG */
949 /** check mbuf type in debug mode */
950 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
952 /** check mbuf type in debug mode if mbuf pointer is not null */
953 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
955 /** MBUF asserts in debug mode */
956 #define RTE_MBUF_ASSERT(exp) do { } while (0)
958 #endif /* RTE_LIBRTE_MBUF_DEBUG */
960 #ifdef RTE_MBUF_REFCNT_ATOMIC
963 * Reads the value of an mbuf's refcnt.
967 * Reference count number.
969 static inline uint16_t
970 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
972 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
976 * Sets an mbuf's refcnt to a defined value.
983 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
985 rte_atomic16_set(&m->refcnt_atomic, new_value);
989 * Adds given value to an mbuf's refcnt and returns its new value.
993 * Value to add/subtract
997 static inline uint16_t
998 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1001 * The atomic_add is an expensive operation, so we don't want to
1002 * call it in the case where we know we are the uniq holder of
1003 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
1004 * operation has to be used because concurrent accesses on the
1005 * reference counter can occur.
1007 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1008 rte_mbuf_refcnt_set(m, 1 + value);
1012 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
1015 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
1018 * Adds given value to an mbuf's refcnt and returns its new value.
1020 static inline uint16_t
1021 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1023 m->refcnt = (uint16_t)(m->refcnt + value);
1028 * Reads the value of an mbuf's refcnt.
1030 static inline uint16_t
1031 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1037 * Sets an mbuf's refcnt to the defined value.
1040 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1042 m->refcnt = new_value;
1045 #endif /* RTE_MBUF_REFCNT_ATOMIC */
1047 /** Mbuf prefetch */
1048 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
1055 * Sanity checks on an mbuf.
1057 * Check the consistency of the given mbuf. The function will cause a
1058 * panic if corruption is detected.
1061 * The mbuf to be checked.
1063 * True if the mbuf is a packet header, false if it is a sub-segment
1064 * of a packet (in this case, some fields like nb_segs are not checked)
1067 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
1070 * @internal Allocate a new mbuf from mempool *mp*.
1071 * The use of that function is reserved for RTE internal needs.
1072 * Please use rte_pktmbuf_alloc().
1075 * The mempool from which mbuf is allocated.
1077 * - The pointer to the new mbuf on success.
1078 * - NULL if allocation failed.
1080 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
1084 if (rte_mempool_get(mp, &mb) < 0)
1086 m = (struct rte_mbuf *)mb;
1087 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1088 rte_mbuf_refcnt_set(m, 1);
1093 * @internal Put mbuf back into its original mempool.
1094 * The use of that function is reserved for RTE internal needs.
1095 * Please use rte_pktmbuf_free().
1098 * The mbuf to be freed.
1100 static inline void __attribute__((always_inline))
1101 __rte_mbuf_raw_free(struct rte_mbuf *m)
1103 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1104 rte_mempool_put(m->pool, m);
1107 /* Operations on ctrl mbuf */
1110 * The control mbuf constructor.
1112 * This function initializes some fields in an mbuf structure that are
1113 * not modified by the user once created (mbuf type, origin pool, buffer
1114 * start address, and so on). This function is given as a callback function
1115 * to rte_mempool_create() at pool creation time.
1118 * The mempool from which the mbuf is allocated.
1120 * A pointer that can be used by the user to retrieve useful information
1121 * for mbuf initialization. This pointer comes from the ``init_arg``
1122 * parameter of rte_mempool_create().
1124 * The mbuf to initialize.
1126 * The index of the mbuf in the pool table.
1128 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1129 void *m, unsigned i);
1132 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
1134 * This new mbuf is initialized with data pointing to the beginning of
1135 * buffer, and with a length of zero.
1138 * The mempool from which the mbuf is allocated.
1140 * - The pointer to the new mbuf on success.
1141 * - NULL if allocation failed.
1143 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
1146 * Free a control mbuf back into its original mempool.
1149 * The control mbuf to be freed.
1151 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
1154 * A macro that returns the pointer to the carried data.
1156 * The value that can be read or assigned.
1161 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
1164 * A macro that returns the length of the carried data.
1166 * The value that can be read or assigned.
1171 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
1174 * Tests if an mbuf is a control mbuf
1177 * The mbuf to be tested
1179 * - True (1) if the mbuf is a control mbuf
1180 * - False(0) otherwise
1183 rte_is_ctrlmbuf(struct rte_mbuf *m)
1185 return !!(m->ol_flags & CTRL_MBUF_FLAG);
1188 /* Operations on pkt mbuf */
1191 * The packet mbuf constructor.
1193 * This function initializes some fields in the mbuf structure that are
1194 * not modified by the user once created (origin pool, buffer start
1195 * address, and so on). This function is given as a callback function to
1196 * rte_mempool_create() at pool creation time.
1199 * The mempool from which mbufs originate.
1201 * A pointer that can be used by the user to retrieve useful information
1202 * for mbuf initialization. This pointer comes from the ``init_arg``
1203 * parameter of rte_mempool_create().
1205 * The mbuf to initialize.
1207 * The index of the mbuf in the pool table.
1209 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1210 void *m, unsigned i);
1214 * A packet mbuf pool constructor.
1216 * This function initializes the mempool private data in the case of a
1217 * pktmbuf pool. This private data is needed by the driver. The
1218 * function is given as a callback function to rte_mempool_create() at
1219 * pool creation. It can be extended by the user, for example, to
1220 * provide another packet size.
1223 * The mempool from which mbufs originate.
1225 * A pointer that can be used by the user to retrieve useful information
1226 * for mbuf initialization. This pointer comes from the ``init_arg``
1227 * parameter of rte_mempool_create().
1229 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1232 * Create a mbuf pool.
1234 * This function creates and initializes a packet mbuf pool. It is
1235 * a wrapper to rte_mempool_create() with the proper packet constructor
1236 * and mempool constructor.
1239 * The name of the mbuf pool.
1241 * The number of elements in the mbuf pool. The optimum size (in terms
1242 * of memory usage) for a mempool is when n is a power of two minus one:
1245 * Size of the per-core object cache. See rte_mempool_create() for
1248 * Size of application private are between the rte_mbuf structure
1249 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1250 * @param data_room_size
1251 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1253 * The socket identifier where the memory should be allocated. The
1254 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1257 * The pointer to the new allocated mempool, on success. NULL on error
1258 * with rte_errno set appropriately. Possible rte_errno values include:
1259 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1260 * - E_RTE_SECONDARY - function was called from a secondary process instance
1261 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1262 * - ENOSPC - the maximum number of memzones has already been allocated
1263 * - EEXIST - a memzone with the same name already exists
1264 * - ENOMEM - no appropriate memory area found in which to create memzone
1266 struct rte_mempool *
1267 rte_pktmbuf_pool_create(const char *name, unsigned n,
1268 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1272 * Get the data room size of mbufs stored in a pktmbuf_pool
1274 * The data room size is the amount of data that can be stored in a
1275 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1278 * The packet mbuf pool.
1280 * The data room size of mbufs stored in this mempool.
1282 static inline uint16_t
1283 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1285 struct rte_pktmbuf_pool_private *mbp_priv;
1287 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1288 return mbp_priv->mbuf_data_room_size;
1292 * Get the application private size of mbufs stored in a pktmbuf_pool
1294 * The private size of mbuf is a zone located between the rte_mbuf
1295 * structure and the data buffer where an application can store data
1296 * associated to a packet.
1299 * The packet mbuf pool.
1301 * The private size of mbufs stored in this mempool.
1303 static inline uint16_t
1304 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1306 struct rte_pktmbuf_pool_private *mbp_priv;
1308 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1309 return mbp_priv->mbuf_priv_size;
1313 * Reset the fields of a packet mbuf to their default values.
1315 * The given mbuf must have only one segment.
1318 * The packet mbuf to be resetted.
1320 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1326 m->vlan_tci_outer = 0;
1332 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
1333 RTE_PKTMBUF_HEADROOM : m->buf_len;
1336 __rte_mbuf_sanity_check(m, 1);
1340 * Allocate a new mbuf from a mempool.
1342 * This new mbuf contains one segment, which has a length of 0. The pointer
1343 * to data is initialized to have some bytes of headroom in the buffer
1344 * (if buffer size allows).
1347 * The mempool from which the mbuf is allocated.
1349 * - The pointer to the new mbuf on success.
1350 * - NULL if allocation failed.
1352 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1355 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
1356 rte_pktmbuf_reset(m);
1361 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1365 * The mempool from which mbufs are allocated.
1367 * Array of pointers to mbufs
1373 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1374 struct rte_mbuf **mbufs, unsigned count)
1379 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1383 /* To understand duff's device on loop unwinding optimization, see
1384 * https://en.wikipedia.org/wiki/Duff's_device.
1385 * Here while() loop is used rather than do() while{} to avoid extra
1386 * check if count is zero.
1388 switch (count % 4) {
1390 while (idx != count) {
1391 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1392 rte_mbuf_refcnt_set(mbufs[idx], 1);
1393 rte_pktmbuf_reset(mbufs[idx]);
1396 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1397 rte_mbuf_refcnt_set(mbufs[idx], 1);
1398 rte_pktmbuf_reset(mbufs[idx]);
1401 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1402 rte_mbuf_refcnt_set(mbufs[idx], 1);
1403 rte_pktmbuf_reset(mbufs[idx]);
1406 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1407 rte_mbuf_refcnt_set(mbufs[idx], 1);
1408 rte_pktmbuf_reset(mbufs[idx]);
1416 * Attach packet mbuf to another packet mbuf.
1418 * After attachment we refer the mbuf we attached as 'indirect',
1419 * while mbuf we attached to as 'direct'.
1420 * Right now, not supported:
1421 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1422 * - mbuf we trying to attach (mi) is used by someone else
1423 * e.g. it's reference counter is greater then 1.
1426 * The indirect packet mbuf.
1428 * The packet mbuf we're attaching to.
1430 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1432 struct rte_mbuf *md;
1434 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(mi) &&
1435 rte_mbuf_refcnt_read(mi) == 1);
1437 /* if m is not direct, get the mbuf that embeds the data */
1438 if (RTE_MBUF_DIRECT(m))
1441 md = rte_mbuf_from_indirect(m);
1443 rte_mbuf_refcnt_update(md, 1);
1444 mi->priv_size = m->priv_size;
1445 mi->buf_physaddr = m->buf_physaddr;
1446 mi->buf_addr = m->buf_addr;
1447 mi->buf_len = m->buf_len;
1450 mi->data_off = m->data_off;
1451 mi->data_len = m->data_len;
1453 mi->vlan_tci = m->vlan_tci;
1454 mi->vlan_tci_outer = m->vlan_tci_outer;
1455 mi->tx_offload = m->tx_offload;
1459 mi->pkt_len = mi->data_len;
1461 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1462 mi->packet_type = m->packet_type;
1464 __rte_mbuf_sanity_check(mi, 1);
1465 __rte_mbuf_sanity_check(m, 0);
1469 * Detach an indirect packet mbuf.
1471 * - restore original mbuf address and length values.
1472 * - reset pktmbuf data and data_len to their default values.
1473 * All other fields of the given packet mbuf will be left intact.
1476 * The indirect attached packet mbuf.
1478 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1480 struct rte_mempool *mp = m->pool;
1481 uint32_t mbuf_size, buf_len, priv_size;
1483 priv_size = rte_pktmbuf_priv_size(mp);
1484 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1485 buf_len = rte_pktmbuf_data_room_size(mp);
1487 m->priv_size = priv_size;
1488 m->buf_addr = (char *)m + mbuf_size;
1489 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
1490 m->buf_len = (uint16_t)buf_len;
1491 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1496 static inline struct rte_mbuf* __attribute__((always_inline))
1497 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1499 __rte_mbuf_sanity_check(m, 0);
1501 if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
1503 /* if this is an indirect mbuf, then
1505 * - free attached mbuf segment
1507 if (RTE_MBUF_INDIRECT(m)) {
1508 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1509 rte_pktmbuf_detach(m);
1510 if (rte_mbuf_refcnt_update(md, -1) == 0)
1511 __rte_mbuf_raw_free(md);
1519 * Free a segment of a packet mbuf into its original mempool.
1521 * Free an mbuf, without parsing other segments in case of chained
1525 * The packet mbuf segment to be freed.
1527 static inline void __attribute__((always_inline))
1528 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1530 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
1532 __rte_mbuf_raw_free(m);
1537 * Free a packet mbuf back into its original mempool.
1539 * Free an mbuf, and all its segments in case of chained buffers. Each
1540 * segment is added back into its original mempool.
1543 * The packet mbuf to be freed.
1545 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1547 struct rte_mbuf *m_next;
1549 __rte_mbuf_sanity_check(m, 1);
1553 rte_pktmbuf_free_seg(m);
1559 * Creates a "clone" of the given packet mbuf.
1561 * Walks through all segments of the given packet mbuf, and for each of them:
1562 * - Creates a new packet mbuf from the given pool.
1563 * - Attaches newly created mbuf to the segment.
1564 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1565 * from the original packet mbuf.
1568 * The packet mbuf to be cloned.
1570 * The mempool from which the "clone" mbufs are allocated.
1572 * - The pointer to the new "clone" mbuf on success.
1573 * - NULL if allocation fails.
1575 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1576 struct rte_mempool *mp)
1578 struct rte_mbuf *mc, *mi, **prev;
1582 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1587 pktlen = md->pkt_len;
1592 rte_pktmbuf_attach(mi, md);
1595 } while ((md = md->next) != NULL &&
1596 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1600 mc->pkt_len = pktlen;
1602 /* Allocation of new indirect segment failed */
1603 if (unlikely (mi == NULL)) {
1604 rte_pktmbuf_free(mc);
1608 __rte_mbuf_sanity_check(mc, 1);
1613 * Adds given value to the refcnt of all packet mbuf segments.
1615 * Walks through all segments of given packet mbuf and for each of them
1616 * invokes rte_mbuf_refcnt_update().
1619 * The packet mbuf whose refcnt to be updated.
1621 * The value to add to the mbuf's segments refcnt.
1623 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1625 __rte_mbuf_sanity_check(m, 1);
1628 rte_mbuf_refcnt_update(m, v);
1629 } while ((m = m->next) != NULL);
1633 * Get the headroom in a packet mbuf.
1638 * The length of the headroom.
1640 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1642 __rte_mbuf_sanity_check(m, 1);
1647 * Get the tailroom of a packet mbuf.
1652 * The length of the tailroom.
1654 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1656 __rte_mbuf_sanity_check(m, 1);
1657 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1662 * Get the last segment of the packet.
1667 * The last segment of the given mbuf.
1669 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1671 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1673 __rte_mbuf_sanity_check(m, 1);
1674 while (m2->next != NULL)
1680 * A macro that points to an offset into the data in the mbuf.
1682 * The returned pointer is cast to type t. Before using this
1683 * function, the user must ensure that the first segment is large
1684 * enough to accommodate its data.
1689 * The offset into the mbuf data.
1691 * The type to cast the result into.
1693 #define rte_pktmbuf_mtod_offset(m, t, o) \
1694 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1697 * A macro that points to the start of the data in the mbuf.
1699 * The returned pointer is cast to type t. Before using this
1700 * function, the user must ensure that the first segment is large
1701 * enough to accommodate its data.
1706 * The type to cast the result into.
1708 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1711 * A macro that returns the physical address that points to an offset of the
1712 * start of the data in the mbuf
1717 * The offset into the data to calculate address from.
1719 #define rte_pktmbuf_mtophys_offset(m, o) \
1720 (phys_addr_t)((m)->buf_physaddr + (m)->data_off + (o))
1723 * A macro that returns the physical address that points to the start of the
1729 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_mtophys_offset(m, 0)
1732 * A macro that returns the length of the packet.
1734 * The value can be read or assigned.
1739 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1742 * A macro that returns the length of the segment.
1744 * The value can be read or assigned.
1749 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1752 * Prepend len bytes to an mbuf data area.
1754 * Returns a pointer to the new
1755 * data start address. If there is not enough headroom in the first
1756 * segment, the function will return NULL, without modifying the mbuf.
1761 * The amount of data to prepend (in bytes).
1763 * A pointer to the start of the newly prepended data, or
1764 * NULL if there is not enough headroom space in the first segment
1766 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1769 __rte_mbuf_sanity_check(m, 1);
1771 if (unlikely(len > rte_pktmbuf_headroom(m)))
1775 m->data_len = (uint16_t)(m->data_len + len);
1776 m->pkt_len = (m->pkt_len + len);
1778 return (char *)m->buf_addr + m->data_off;
1782 * Append len bytes to an mbuf.
1784 * Append len bytes to an mbuf and return a pointer to the start address
1785 * of the added data. If there is not enough tailroom in the last
1786 * segment, the function will return NULL, without modifying the mbuf.
1791 * The amount of data to append (in bytes).
1793 * A pointer to the start of the newly appended data, or
1794 * NULL if there is not enough tailroom space in the last segment
1796 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1799 struct rte_mbuf *m_last;
1801 __rte_mbuf_sanity_check(m, 1);
1803 m_last = rte_pktmbuf_lastseg(m);
1804 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1807 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1808 m_last->data_len = (uint16_t)(m_last->data_len + len);
1809 m->pkt_len = (m->pkt_len + len);
1810 return (char*) tail;
1814 * Remove len bytes at the beginning of an mbuf.
1816 * Returns a pointer to the start address of the new data area. If the
1817 * length is greater than the length of the first segment, then the
1818 * function will fail and return NULL, without modifying the mbuf.
1823 * The amount of data to remove (in bytes).
1825 * A pointer to the new start of the data.
1827 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1829 __rte_mbuf_sanity_check(m, 1);
1831 if (unlikely(len > m->data_len))
1834 m->data_len = (uint16_t)(m->data_len - len);
1836 m->pkt_len = (m->pkt_len - len);
1837 return (char *)m->buf_addr + m->data_off;
1841 * Remove len bytes of data at the end of the mbuf.
1843 * If the length is greater than the length of the last segment, the
1844 * function will fail and return -1 without modifying the mbuf.
1849 * The amount of data to remove (in bytes).
1854 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1856 struct rte_mbuf *m_last;
1858 __rte_mbuf_sanity_check(m, 1);
1860 m_last = rte_pktmbuf_lastseg(m);
1861 if (unlikely(len > m_last->data_len))
1864 m_last->data_len = (uint16_t)(m_last->data_len - len);
1865 m->pkt_len = (m->pkt_len - len);
1870 * Test if mbuf data is contiguous.
1875 * - 1, if all data is contiguous (one segment).
1876 * - 0, if there is several segments.
1878 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1880 __rte_mbuf_sanity_check(m, 1);
1881 return !!(m->nb_segs == 1);
1885 * Chain an mbuf to another, thereby creating a segmented packet.
1887 * Note: The implementation will do a linear walk over the segments to find
1888 * the tail entry. For cases when there are many segments, it's better to
1889 * chain the entries manually.
1892 * The head of the mbuf chain (the first packet)
1894 * The mbuf to put last in the chain
1898 * - -EOVERFLOW, if the chain is full (256 entries)
1900 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1902 struct rte_mbuf *cur_tail;
1904 /* Check for number-of-segments-overflow */
1905 if (head->nb_segs + tail->nb_segs >= 1 << (sizeof(head->nb_segs) * 8))
1908 /* Chain 'tail' onto the old tail */
1909 cur_tail = rte_pktmbuf_lastseg(head);
1910 cur_tail->next = tail;
1912 /* accumulate number of segments and total length. */
1913 head->nb_segs = (uint8_t)(head->nb_segs + tail->nb_segs);
1914 head->pkt_len += tail->pkt_len;
1916 /* pkt_len is only set in the head */
1917 tail->pkt_len = tail->data_len;
1923 * Dump an mbuf structure to the console.
1925 * Dump all fields for the given packet mbuf and all its associated
1926 * segments (in the case of a chained buffer).
1929 * A pointer to a file for output
1933 * If dump_len != 0, also dump the "dump_len" first data bytes of
1936 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1942 #endif /* _RTE_MBUF_H_ */