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 * Mask of layer 2 packet types.
312 * It is used for outer packet for tunneling cases.
314 #define RTE_PTYPE_L2_MASK 0x0000000f
316 * IP (Internet Protocol) version 4 packet type.
317 * It is used for outer packet for tunneling cases, and does not contain any
321 * <'ether type'=0x0800
322 * | 'version'=4, 'ihl'=5>
324 #define RTE_PTYPE_L3_IPV4 0x00000010
326 * IP (Internet Protocol) version 4 packet type.
327 * It is used for outer packet for tunneling cases, and contains header
331 * <'ether type'=0x0800
332 * | 'version'=4, 'ihl'=[6-15], 'options'>
334 #define RTE_PTYPE_L3_IPV4_EXT 0x00000030
336 * IP (Internet Protocol) version 6 packet type.
337 * It is used for outer packet for tunneling cases, and does not contain any
341 * <'ether type'=0x86DD
342 * | 'version'=6, 'next header'=0x3B>
344 #define RTE_PTYPE_L3_IPV6 0x00000040
346 * IP (Internet Protocol) version 4 packet type.
347 * It is used for outer packet for tunneling cases, and may or maynot contain
351 * <'ether type'=0x0800
352 * | 'version'=4, 'ihl'=[5-15], <'options'>>
354 #define RTE_PTYPE_L3_IPV4_EXT_UNKNOWN 0x00000090
356 * IP (Internet Protocol) version 6 packet type.
357 * It is used for outer packet for tunneling cases, and contains extension
361 * <'ether type'=0x86DD
362 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
363 * 'extension headers'>
365 #define RTE_PTYPE_L3_IPV6_EXT 0x000000c0
367 * IP (Internet Protocol) version 6 packet type.
368 * It is used for outer packet for tunneling cases, and may or maynot contain
372 * <'ether type'=0x86DD
373 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
374 * <'extension headers'>>
376 #define RTE_PTYPE_L3_IPV6_EXT_UNKNOWN 0x000000e0
378 * Mask of layer 3 packet types.
379 * It is used for outer packet for tunneling cases.
381 #define RTE_PTYPE_L3_MASK 0x000000f0
383 * TCP (Transmission Control Protocol) packet type.
384 * It is used for outer packet for tunneling cases.
387 * <'ether type'=0x0800
388 * | 'version'=4, 'protocol'=6, 'MF'=0>
390 * <'ether type'=0x86DD
391 * | 'version'=6, 'next header'=6>
393 #define RTE_PTYPE_L4_TCP 0x00000100
395 * UDP (User Datagram Protocol) packet type.
396 * It is used for outer packet for tunneling cases.
399 * <'ether type'=0x0800
400 * | 'version'=4, 'protocol'=17, 'MF'=0>
402 * <'ether type'=0x86DD
403 * | 'version'=6, 'next header'=17>
405 #define RTE_PTYPE_L4_UDP 0x00000200
407 * Fragmented IP (Internet Protocol) packet type.
408 * It is used for outer packet for tunneling cases.
410 * It refers to those packets of any IP types, which can be recognized as
411 * fragmented. A fragmented packet cannot be recognized as any other L4 types
412 * (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP,
413 * RTE_PTYPE_L4_NONFRAG).
416 * <'ether type'=0x0800
417 * | 'version'=4, 'MF'=1>
419 * <'ether type'=0x86DD
420 * | 'version'=6, 'next header'=44>
422 #define RTE_PTYPE_L4_FRAG 0x00000300
424 * SCTP (Stream Control Transmission Protocol) packet type.
425 * It is used for outer packet for tunneling cases.
428 * <'ether type'=0x0800
429 * | 'version'=4, 'protocol'=132, 'MF'=0>
431 * <'ether type'=0x86DD
432 * | 'version'=6, 'next header'=132>
434 #define RTE_PTYPE_L4_SCTP 0x00000400
436 * ICMP (Internet Control Message Protocol) packet type.
437 * It is used for outer packet for tunneling cases.
440 * <'ether type'=0x0800
441 * | 'version'=4, 'protocol'=1, 'MF'=0>
443 * <'ether type'=0x86DD
444 * | 'version'=6, 'next header'=1>
446 #define RTE_PTYPE_L4_ICMP 0x00000500
448 * Non-fragmented IP (Internet Protocol) packet type.
449 * It is used for outer packet for tunneling cases.
451 * It refers to those packets of any IP types, while cannot be recognized as
452 * any of above L4 types (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP,
453 * RTE_PTYPE_L4_FRAG, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP).
456 * <'ether type'=0x0800
457 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
459 * <'ether type'=0x86DD
460 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
462 #define RTE_PTYPE_L4_NONFRAG 0x00000600
464 * Mask of layer 4 packet types.
465 * It is used for outer packet for tunneling cases.
467 #define RTE_PTYPE_L4_MASK 0x00000f00
469 * IP (Internet Protocol) in IP (Internet Protocol) tunneling packet type.
472 * <'ether type'=0x0800
473 * | 'version'=4, 'protocol'=[4|41]>
475 * <'ether type'=0x86DD
476 * | 'version'=6, 'next header'=[4|41]>
478 #define RTE_PTYPE_TUNNEL_IP 0x00001000
480 * GRE (Generic Routing Encapsulation) tunneling packet type.
483 * <'ether type'=0x0800
484 * | 'version'=4, 'protocol'=47>
486 * <'ether type'=0x86DD
487 * | 'version'=6, 'next header'=47>
489 #define RTE_PTYPE_TUNNEL_GRE 0x00002000
491 * VXLAN (Virtual eXtensible Local Area Network) tunneling packet type.
494 * <'ether type'=0x0800
495 * | 'version'=4, 'protocol'=17
496 * | 'destination port'=4798>
498 * <'ether type'=0x86DD
499 * | 'version'=6, 'next header'=17
500 * | 'destination port'=4798>
502 #define RTE_PTYPE_TUNNEL_VXLAN 0x00003000
504 * NVGRE (Network Virtualization using Generic Routing Encapsulation) tunneling
508 * <'ether type'=0x0800
509 * | 'version'=4, 'protocol'=47
510 * | 'protocol type'=0x6558>
512 * <'ether type'=0x86DD
513 * | 'version'=6, 'next header'=47
514 * | 'protocol type'=0x6558'>
516 #define RTE_PTYPE_TUNNEL_NVGRE 0x00004000
518 * GENEVE (Generic Network Virtualization Encapsulation) tunneling packet type.
521 * <'ether type'=0x0800
522 * | 'version'=4, 'protocol'=17
523 * | 'destination port'=6081>
525 * <'ether type'=0x86DD
526 * | 'version'=6, 'next header'=17
527 * | 'destination port'=6081>
529 #define RTE_PTYPE_TUNNEL_GENEVE 0x00005000
531 * Tunneling packet type of Teredo, VXLAN (Virtual eXtensible Local Area
532 * Network) or GRE (Generic Routing Encapsulation) could be recognized as this
533 * packet type, if they can not be recognized independently as of hardware
536 #define RTE_PTYPE_TUNNEL_GRENAT 0x00006000
538 * Mask of tunneling packet types.
540 #define RTE_PTYPE_TUNNEL_MASK 0x0000f000
542 * Ethernet packet type.
543 * It is used for inner packet type only.
545 * Packet format (inner only):
546 * <'ether type'=[0x800|0x86DD]>
548 #define RTE_PTYPE_INNER_L2_ETHER 0x00010000
550 * Ethernet packet type with VLAN (Virtual Local Area Network) tag.
552 * Packet format (inner only):
553 * <'ether type'=[0x800|0x86DD], vlan=[1-4095]>
555 #define RTE_PTYPE_INNER_L2_ETHER_VLAN 0x00020000
557 * Mask of inner layer 2 packet types.
559 #define RTE_PTYPE_INNER_L2_MASK 0x000f0000
561 * IP (Internet Protocol) version 4 packet type.
562 * It is used for inner packet only, and does not contain any header option.
564 * Packet format (inner only):
565 * <'ether type'=0x0800
566 * | 'version'=4, 'ihl'=5>
568 #define RTE_PTYPE_INNER_L3_IPV4 0x00100000
570 * IP (Internet Protocol) version 4 packet type.
571 * It is used for inner packet only, and contains header options.
573 * Packet format (inner only):
574 * <'ether type'=0x0800
575 * | 'version'=4, 'ihl'=[6-15], 'options'>
577 #define RTE_PTYPE_INNER_L3_IPV4_EXT 0x00200000
579 * IP (Internet Protocol) version 6 packet type.
580 * It is used for inner packet only, and does not contain any extension header.
582 * Packet format (inner only):
583 * <'ether type'=0x86DD
584 * | 'version'=6, 'next header'=0x3B>
586 #define RTE_PTYPE_INNER_L3_IPV6 0x00300000
588 * IP (Internet Protocol) version 4 packet type.
589 * It is used for inner packet only, and may or maynot contain header options.
591 * Packet format (inner only):
592 * <'ether type'=0x0800
593 * | 'version'=4, 'ihl'=[5-15], <'options'>>
595 #define RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN 0x00400000
597 * IP (Internet Protocol) version 6 packet type.
598 * It is used for inner packet only, and contains extension headers.
600 * Packet format (inner only):
601 * <'ether type'=0x86DD
602 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
603 * 'extension headers'>
605 #define RTE_PTYPE_INNER_L3_IPV6_EXT 0x00500000
607 * IP (Internet Protocol) version 6 packet type.
608 * It is used for inner packet only, and may or maynot contain extension
611 * Packet format (inner only):
612 * <'ether type'=0x86DD
613 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
614 * <'extension headers'>>
616 #define RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN 0x00600000
618 * Mask of inner layer 3 packet types.
620 #define RTE_PTYPE_INNER_L3_MASK 0x00f00000
622 * TCP (Transmission Control Protocol) packet type.
623 * It is used for inner packet only.
625 * Packet format (inner only):
626 * <'ether type'=0x0800
627 * | 'version'=4, 'protocol'=6, 'MF'=0>
629 * <'ether type'=0x86DD
630 * | 'version'=6, 'next header'=6>
632 #define RTE_PTYPE_INNER_L4_TCP 0x01000000
634 * UDP (User Datagram Protocol) packet type.
635 * It is used for inner packet only.
637 * Packet format (inner only):
638 * <'ether type'=0x0800
639 * | 'version'=4, 'protocol'=17, 'MF'=0>
641 * <'ether type'=0x86DD
642 * | 'version'=6, 'next header'=17>
644 #define RTE_PTYPE_INNER_L4_UDP 0x02000000
646 * Fragmented IP (Internet Protocol) packet type.
647 * It is used for inner packet only, and may or maynot have layer 4 packet.
649 * Packet format (inner only):
650 * <'ether type'=0x0800
651 * | 'version'=4, 'MF'=1>
653 * <'ether type'=0x86DD
654 * | 'version'=6, 'next header'=44>
656 #define RTE_PTYPE_INNER_L4_FRAG 0x03000000
658 * SCTP (Stream Control Transmission Protocol) packet type.
659 * It is used for inner packet only.
661 * Packet format (inner only):
662 * <'ether type'=0x0800
663 * | 'version'=4, 'protocol'=132, 'MF'=0>
665 * <'ether type'=0x86DD
666 * | 'version'=6, 'next header'=132>
668 #define RTE_PTYPE_INNER_L4_SCTP 0x04000000
670 * ICMP (Internet Control Message Protocol) packet type.
671 * It is used for inner packet only.
673 * Packet format (inner only):
674 * <'ether type'=0x0800
675 * | 'version'=4, 'protocol'=1, 'MF'=0>
677 * <'ether type'=0x86DD
678 * | 'version'=6, 'next header'=1>
680 #define RTE_PTYPE_INNER_L4_ICMP 0x05000000
682 * Non-fragmented IP (Internet Protocol) packet type.
683 * It is used for inner packet only, and may or maynot have other unknown layer
686 * Packet format (inner only):
687 * <'ether type'=0x0800
688 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
690 * <'ether type'=0x86DD
691 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
693 #define RTE_PTYPE_INNER_L4_NONFRAG 0x06000000
695 * Mask of inner layer 4 packet types.
697 #define RTE_PTYPE_INNER_L4_MASK 0x0f000000
700 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
701 * one, bit 4 is selected to be used for IPv4 only. Then checking bit 4 can
702 * determine if it is an IPV4 packet.
704 #define RTE_ETH_IS_IPV4_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV4)
707 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
708 * one, bit 6 is selected to be used for IPv4 only. Then checking bit 6 can
709 * determine if it is an IPV4 packet.
711 #define RTE_ETH_IS_IPV6_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV6)
713 /* Check if it is a tunneling packet */
714 #define RTE_ETH_IS_TUNNEL_PKT(ptype) ((ptype) & (RTE_PTYPE_TUNNEL_MASK | \
715 RTE_PTYPE_INNER_L2_MASK | \
716 RTE_PTYPE_INNER_L3_MASK | \
717 RTE_PTYPE_INNER_L4_MASK))
719 /** Alignment constraint of mbuf private area. */
720 #define RTE_MBUF_PRIV_ALIGN 8
723 * Get the name of a RX offload flag
726 * The mask describing the flag.
728 * The name of this flag, or NULL if it's not a valid RX flag.
730 const char *rte_get_rx_ol_flag_name(uint64_t mask);
733 * Get the name of a TX offload flag
736 * The mask describing the flag. Usually only one bit must be set.
737 * Several bits can be given if they belong to the same mask.
738 * Ex: PKT_TX_L4_MASK.
740 * The name of this flag, or NULL if it's not a valid TX flag.
742 const char *rte_get_tx_ol_flag_name(uint64_t mask);
745 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
746 * splitting it into multiple segments.
747 * So, for mbufs that planned to be involved into RX/TX, the recommended
748 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
750 #define RTE_MBUF_DEFAULT_DATAROOM 2048
751 #define RTE_MBUF_DEFAULT_BUF_SIZE \
752 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
754 /* define a set of marker types that can be used to refer to set points in the
756 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
757 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
758 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
759 * with a single assignment */
762 * The generic rte_mbuf, containing a packet mbuf.
767 void *buf_addr; /**< Virtual address of segment buffer. */
768 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
770 uint16_t buf_len; /**< Length of segment buffer. */
772 /* next 6 bytes are initialised on RX descriptor rearm */
777 * 16-bit Reference counter.
778 * It should only be accessed using the following functions:
779 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
780 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
781 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
785 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
786 uint16_t refcnt; /**< Non-atomically accessed refcnt */
788 uint8_t nb_segs; /**< Number of segments. */
789 uint8_t port; /**< Input port. */
791 uint64_t ol_flags; /**< Offload features. */
793 /* remaining bytes are set on RX when pulling packet from descriptor */
794 MARKER rx_descriptor_fields1;
797 * The packet type, which is the combination of outer/inner L2, L3, L4
798 * and tunnel types. The packet_type is about data really present in the
799 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
800 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
801 * vlan is stripped from the data.
804 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
806 uint32_t l2_type:4; /**< (Outer) L2 type. */
807 uint32_t l3_type:4; /**< (Outer) L3 type. */
808 uint32_t l4_type:4; /**< (Outer) L4 type. */
809 uint32_t tun_type:4; /**< Tunnel type. */
810 uint32_t inner_l2_type:4; /**< Inner L2 type. */
811 uint32_t inner_l3_type:4; /**< Inner L3 type. */
812 uint32_t inner_l4_type:4; /**< Inner L4 type. */
816 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
817 uint16_t data_len; /**< Amount of data in segment buffer. */
818 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN_STRIPPED is set. */
822 uint32_t rss; /**< RSS hash result if RSS enabled */
830 /**< Second 4 flexible bytes */
833 /**< First 4 flexible bytes or FD ID, dependent on
834 PKT_RX_FDIR_* flag in ol_flags. */
835 } fdir; /**< Filter identifier if FDIR enabled */
839 } sched; /**< Hierarchical scheduler */
840 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
841 } hash; /**< hash information */
843 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
845 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ_STRIPPED is set. */
846 uint16_t vlan_tci_outer;
848 /* second cache line - fields only used in slow path or on TX */
849 MARKER cacheline1 __rte_cache_min_aligned;
852 void *userdata; /**< Can be used for external metadata */
853 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
856 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
857 struct rte_mbuf *next; /**< Next segment of scattered packet. */
859 /* fields to support TX offloads */
861 uint64_t tx_offload; /**< combined for easy fetch */
863 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
864 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
865 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
866 uint64_t tso_segsz:16; /**< TCP TSO segment size */
868 /* fields for TX offloading of tunnels */
869 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
870 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
872 /* uint64_t unused:8; */
876 /** Size of the application private data. In case of an indirect
877 * mbuf, it stores the direct mbuf private data size. */
880 /** Timesync flags for use with IEEE1588. */
882 } __rte_cache_aligned;
885 * Prefetch the first part of the mbuf
887 * The first 64 bytes of the mbuf corresponds to fields that are used early
888 * in the receive path. If the cache line of the architecture is higher than
889 * 64B, the second part will also be prefetched.
892 * The pointer to the mbuf.
895 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
897 rte_prefetch0(&m->cacheline0);
901 * Prefetch the second part of the mbuf
903 * The next 64 bytes of the mbuf corresponds to fields that are used in the
904 * transmit path. If the cache line of the architecture is higher than 64B,
905 * this function does nothing as it is expected that the full mbuf is
909 * The pointer to the mbuf.
912 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
914 #if RTE_CACHE_LINE_SIZE == 64
915 rte_prefetch0(&m->cacheline1);
922 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
925 * Return the DMA address of the beginning of the mbuf data
928 * The pointer to the mbuf.
930 * The physical address of the beginning of the mbuf data
932 static inline phys_addr_t
933 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
935 return mb->buf_physaddr + mb->data_off;
939 * Return the default DMA address of the beginning of the mbuf data
941 * This function is used by drivers in their receive function, as it
942 * returns the location where data should be written by the NIC, taking
943 * the default headroom in account.
946 * The pointer to the mbuf.
948 * The physical address of the beginning of the mbuf data
950 static inline phys_addr_t
951 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
953 return mb->buf_physaddr + RTE_PKTMBUF_HEADROOM;
957 * Return the mbuf owning the data buffer address of an indirect mbuf.
960 * The pointer to the indirect mbuf.
962 * The address of the direct mbuf corresponding to buffer_addr.
964 static inline struct rte_mbuf *
965 rte_mbuf_from_indirect(struct rte_mbuf *mi)
967 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
971 * Return the buffer address embedded in the given mbuf.
974 * The pointer to the mbuf.
976 * The address of the data buffer owned by the mbuf.
979 rte_mbuf_to_baddr(struct rte_mbuf *md)
982 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
987 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
989 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
992 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
994 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
997 * Private data in case of pktmbuf pool.
999 * A structure that contains some pktmbuf_pool-specific data that are
1000 * appended after the mempool structure (in private data).
1002 struct rte_pktmbuf_pool_private {
1003 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
1004 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
1007 #ifdef RTE_LIBRTE_MBUF_DEBUG
1009 /** check mbuf type in debug mode */
1010 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
1012 #else /* RTE_LIBRTE_MBUF_DEBUG */
1014 /** check mbuf type in debug mode */
1015 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
1017 #endif /* RTE_LIBRTE_MBUF_DEBUG */
1019 #ifdef RTE_MBUF_REFCNT_ATOMIC
1022 * Reads the value of an mbuf's refcnt.
1026 * Reference count number.
1028 static inline uint16_t
1029 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1031 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
1035 * Sets an mbuf's refcnt to a defined value.
1042 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1044 rte_atomic16_set(&m->refcnt_atomic, new_value);
1048 * Adds given value to an mbuf's refcnt and returns its new value.
1052 * Value to add/subtract
1056 static inline uint16_t
1057 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1060 * The atomic_add is an expensive operation, so we don't want to
1061 * call it in the case where we know we are the uniq holder of
1062 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
1063 * operation has to be used because concurrent accesses on the
1064 * reference counter can occur.
1066 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1067 rte_mbuf_refcnt_set(m, 1 + value);
1071 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
1074 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
1077 * Adds given value to an mbuf's refcnt and returns its new value.
1079 static inline uint16_t
1080 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1082 m->refcnt = (uint16_t)(m->refcnt + value);
1087 * Reads the value of an mbuf's refcnt.
1089 static inline uint16_t
1090 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1096 * Sets an mbuf's refcnt to the defined value.
1099 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1101 m->refcnt = new_value;
1104 #endif /* RTE_MBUF_REFCNT_ATOMIC */
1106 /** Mbuf prefetch */
1107 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
1114 * Sanity checks on an mbuf.
1116 * Check the consistency of the given mbuf. The function will cause a
1117 * panic if corruption is detected.
1120 * The mbuf to be checked.
1122 * True if the mbuf is a packet header, false if it is a sub-segment
1123 * of a packet (in this case, some fields like nb_segs are not checked)
1126 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
1129 * Allocate an unitialized mbuf from mempool *mp*.
1131 * This function can be used by PMDs (especially in RX functions) to
1132 * allocate an unitialized mbuf. The driver is responsible of
1133 * initializing all the required fields. See rte_pktmbuf_reset().
1134 * For standard needs, prefer rte_pktmbuf_alloc().
1137 * The mempool from which mbuf is allocated.
1139 * - The pointer to the new mbuf on success.
1140 * - NULL if allocation failed.
1142 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
1147 if (rte_mempool_get(mp, &mb) < 0)
1149 m = (struct rte_mbuf *)mb;
1150 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1151 rte_mbuf_refcnt_set(m, 1);
1152 __rte_mbuf_sanity_check(m, 0);
1157 /* compat with older versions */
1158 __rte_deprecated static inline struct rte_mbuf *
1159 __rte_mbuf_raw_alloc(struct rte_mempool *mp)
1161 return rte_mbuf_raw_alloc(mp);
1165 * @internal Put mbuf back into its original mempool.
1166 * The use of that function is reserved for RTE internal needs.
1167 * Please use rte_pktmbuf_free().
1170 * The mbuf to be freed.
1172 static inline void __attribute__((always_inline))
1173 __rte_mbuf_raw_free(struct rte_mbuf *m)
1175 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1176 rte_mempool_put(m->pool, m);
1179 /* Operations on ctrl mbuf */
1182 * The control mbuf constructor.
1184 * This function initializes some fields in an mbuf structure that are
1185 * not modified by the user once created (mbuf type, origin pool, buffer
1186 * start address, and so on). This function is given as a callback function
1187 * to rte_mempool_create() at pool creation time.
1190 * The mempool from which the mbuf is allocated.
1192 * A pointer that can be used by the user to retrieve useful information
1193 * for mbuf initialization. This pointer comes from the ``init_arg``
1194 * parameter of rte_mempool_create().
1196 * The mbuf to initialize.
1198 * The index of the mbuf in the pool table.
1200 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1201 void *m, unsigned i);
1204 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
1206 * This new mbuf is initialized with data pointing to the beginning of
1207 * buffer, and with a length of zero.
1210 * The mempool from which the mbuf is allocated.
1212 * - The pointer to the new mbuf on success.
1213 * - NULL if allocation failed.
1215 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
1218 * Free a control mbuf back into its original mempool.
1221 * The control mbuf to be freed.
1223 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
1226 * A macro that returns the pointer to the carried data.
1228 * The value that can be read or assigned.
1233 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
1236 * A macro that returns the length of the carried data.
1238 * The value that can be read or assigned.
1243 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
1246 * Tests if an mbuf is a control mbuf
1249 * The mbuf to be tested
1251 * - True (1) if the mbuf is a control mbuf
1252 * - False(0) otherwise
1255 rte_is_ctrlmbuf(struct rte_mbuf *m)
1257 return !!(m->ol_flags & CTRL_MBUF_FLAG);
1260 /* Operations on pkt mbuf */
1263 * The packet mbuf constructor.
1265 * This function initializes some fields in the mbuf structure that are
1266 * not modified by the user once created (origin pool, buffer start
1267 * address, and so on). This function is given as a callback function to
1268 * rte_mempool_create() at pool creation time.
1271 * The mempool from which mbufs originate.
1273 * A pointer that can be used by the user to retrieve useful information
1274 * for mbuf initialization. This pointer comes from the ``init_arg``
1275 * parameter of rte_mempool_create().
1277 * The mbuf to initialize.
1279 * The index of the mbuf in the pool table.
1281 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1282 void *m, unsigned i);
1286 * A packet mbuf pool constructor.
1288 * This function initializes the mempool private data in the case of a
1289 * pktmbuf pool. This private data is needed by the driver. The
1290 * function is given as a callback function to rte_mempool_create() at
1291 * pool creation. It can be extended by the user, for example, to
1292 * provide another packet size.
1295 * The mempool from which mbufs originate.
1297 * A pointer that can be used by the user to retrieve useful information
1298 * for mbuf initialization. This pointer comes from the ``init_arg``
1299 * parameter of rte_mempool_create().
1301 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1304 * Create a mbuf pool.
1306 * This function creates and initializes a packet mbuf pool. It is
1307 * a wrapper to rte_mempool_create() with the proper packet constructor
1308 * and mempool constructor.
1311 * The name of the mbuf pool.
1313 * The number of elements in the mbuf pool. The optimum size (in terms
1314 * of memory usage) for a mempool is when n is a power of two minus one:
1317 * Size of the per-core object cache. See rte_mempool_create() for
1320 * Size of application private are between the rte_mbuf structure
1321 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1322 * @param data_room_size
1323 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1325 * The socket identifier where the memory should be allocated. The
1326 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1329 * The pointer to the new allocated mempool, on success. NULL on error
1330 * with rte_errno set appropriately. Possible rte_errno values include:
1331 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1332 * - E_RTE_SECONDARY - function was called from a secondary process instance
1333 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1334 * - ENOSPC - the maximum number of memzones has already been allocated
1335 * - EEXIST - a memzone with the same name already exists
1336 * - ENOMEM - no appropriate memory area found in which to create memzone
1338 struct rte_mempool *
1339 rte_pktmbuf_pool_create(const char *name, unsigned n,
1340 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1344 * Get the data room size of mbufs stored in a pktmbuf_pool
1346 * The data room size is the amount of data that can be stored in a
1347 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1350 * The packet mbuf pool.
1352 * The data room size of mbufs stored in this mempool.
1354 static inline uint16_t
1355 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1357 struct rte_pktmbuf_pool_private *mbp_priv;
1359 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1360 return mbp_priv->mbuf_data_room_size;
1364 * Get the application private size of mbufs stored in a pktmbuf_pool
1366 * The private size of mbuf is a zone located between the rte_mbuf
1367 * structure and the data buffer where an application can store data
1368 * associated to a packet.
1371 * The packet mbuf pool.
1373 * The private size of mbufs stored in this mempool.
1375 static inline uint16_t
1376 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1378 struct rte_pktmbuf_pool_private *mbp_priv;
1380 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1381 return mbp_priv->mbuf_priv_size;
1385 * Reset the fields of a packet mbuf to their default values.
1387 * The given mbuf must have only one segment.
1390 * The packet mbuf to be resetted.
1392 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1398 m->vlan_tci_outer = 0;
1404 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
1405 RTE_PKTMBUF_HEADROOM : m->buf_len;
1408 __rte_mbuf_sanity_check(m, 1);
1412 * Allocate a new mbuf from a mempool.
1414 * This new mbuf contains one segment, which has a length of 0. The pointer
1415 * to data is initialized to have some bytes of headroom in the buffer
1416 * (if buffer size allows).
1419 * The mempool from which the mbuf is allocated.
1421 * - The pointer to the new mbuf on success.
1422 * - NULL if allocation failed.
1424 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1427 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1428 rte_pktmbuf_reset(m);
1433 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1437 * The mempool from which mbufs are allocated.
1439 * Array of pointers to mbufs
1445 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1446 struct rte_mbuf **mbufs, unsigned count)
1451 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1455 /* To understand duff's device on loop unwinding optimization, see
1456 * https://en.wikipedia.org/wiki/Duff's_device.
1457 * Here while() loop is used rather than do() while{} to avoid extra
1458 * check if count is zero.
1460 switch (count % 4) {
1462 while (idx != count) {
1463 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1464 rte_mbuf_refcnt_set(mbufs[idx], 1);
1465 rte_pktmbuf_reset(mbufs[idx]);
1468 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1469 rte_mbuf_refcnt_set(mbufs[idx], 1);
1470 rte_pktmbuf_reset(mbufs[idx]);
1473 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1474 rte_mbuf_refcnt_set(mbufs[idx], 1);
1475 rte_pktmbuf_reset(mbufs[idx]);
1478 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1479 rte_mbuf_refcnt_set(mbufs[idx], 1);
1480 rte_pktmbuf_reset(mbufs[idx]);
1488 * Attach packet mbuf to another packet mbuf.
1490 * After attachment we refer the mbuf we attached as 'indirect',
1491 * while mbuf we attached to as 'direct'.
1492 * The direct mbuf's reference counter is incremented.
1494 * Right now, not supported:
1495 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1496 * - mbuf we trying to attach (mi) is used by someone else
1497 * e.g. it's reference counter is greater then 1.
1500 * The indirect packet mbuf.
1502 * The packet mbuf we're attaching to.
1504 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1506 struct rte_mbuf *md;
1508 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1509 rte_mbuf_refcnt_read(mi) == 1);
1511 /* if m is not direct, get the mbuf that embeds the data */
1512 if (RTE_MBUF_DIRECT(m))
1515 md = rte_mbuf_from_indirect(m);
1517 rte_mbuf_refcnt_update(md, 1);
1518 mi->priv_size = m->priv_size;
1519 mi->buf_physaddr = m->buf_physaddr;
1520 mi->buf_addr = m->buf_addr;
1521 mi->buf_len = m->buf_len;
1524 mi->data_off = m->data_off;
1525 mi->data_len = m->data_len;
1527 mi->vlan_tci = m->vlan_tci;
1528 mi->vlan_tci_outer = m->vlan_tci_outer;
1529 mi->tx_offload = m->tx_offload;
1533 mi->pkt_len = mi->data_len;
1535 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1536 mi->packet_type = m->packet_type;
1538 __rte_mbuf_sanity_check(mi, 1);
1539 __rte_mbuf_sanity_check(m, 0);
1543 * Detach an indirect packet mbuf.
1545 * - restore original mbuf address and length values.
1546 * - reset pktmbuf data and data_len to their default values.
1547 * - decrement the direct mbuf's reference counter. When the
1548 * reference counter becomes 0, the direct mbuf is freed.
1550 * All other fields of the given packet mbuf will be left intact.
1553 * The indirect attached packet mbuf.
1555 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1557 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1558 struct rte_mempool *mp = m->pool;
1559 uint32_t mbuf_size, buf_len, priv_size;
1561 priv_size = rte_pktmbuf_priv_size(mp);
1562 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1563 buf_len = rte_pktmbuf_data_room_size(mp);
1565 m->priv_size = priv_size;
1566 m->buf_addr = (char *)m + mbuf_size;
1567 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
1568 m->buf_len = (uint16_t)buf_len;
1569 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1573 if (rte_mbuf_refcnt_update(md, -1) == 0)
1574 __rte_mbuf_raw_free(md);
1577 static inline struct rte_mbuf* __attribute__((always_inline))
1578 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1580 __rte_mbuf_sanity_check(m, 0);
1582 if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
1583 /* if this is an indirect mbuf, it is detached. */
1584 if (RTE_MBUF_INDIRECT(m))
1585 rte_pktmbuf_detach(m);
1592 * Free a segment of a packet mbuf into its original mempool.
1594 * Free an mbuf, without parsing other segments in case of chained
1598 * The packet mbuf segment to be freed.
1600 static inline void __attribute__((always_inline))
1601 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1603 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
1605 __rte_mbuf_raw_free(m);
1610 * Free a packet mbuf back into its original mempool.
1612 * Free an mbuf, and all its segments in case of chained buffers. Each
1613 * segment is added back into its original mempool.
1616 * The packet mbuf to be freed.
1618 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1620 struct rte_mbuf *m_next;
1622 __rte_mbuf_sanity_check(m, 1);
1626 rte_pktmbuf_free_seg(m);
1632 * Creates a "clone" of the given packet mbuf.
1634 * Walks through all segments of the given packet mbuf, and for each of them:
1635 * - Creates a new packet mbuf from the given pool.
1636 * - Attaches newly created mbuf to the segment.
1637 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1638 * from the original packet mbuf.
1641 * The packet mbuf to be cloned.
1643 * The mempool from which the "clone" mbufs are allocated.
1645 * - The pointer to the new "clone" mbuf on success.
1646 * - NULL if allocation fails.
1648 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1649 struct rte_mempool *mp)
1651 struct rte_mbuf *mc, *mi, **prev;
1655 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1660 pktlen = md->pkt_len;
1665 rte_pktmbuf_attach(mi, md);
1668 } while ((md = md->next) != NULL &&
1669 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1673 mc->pkt_len = pktlen;
1675 /* Allocation of new indirect segment failed */
1676 if (unlikely (mi == NULL)) {
1677 rte_pktmbuf_free(mc);
1681 __rte_mbuf_sanity_check(mc, 1);
1686 * Adds given value to the refcnt of all packet mbuf segments.
1688 * Walks through all segments of given packet mbuf and for each of them
1689 * invokes rte_mbuf_refcnt_update().
1692 * The packet mbuf whose refcnt to be updated.
1694 * The value to add to the mbuf's segments refcnt.
1696 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1698 __rte_mbuf_sanity_check(m, 1);
1701 rte_mbuf_refcnt_update(m, v);
1702 } while ((m = m->next) != NULL);
1706 * Get the headroom in a packet mbuf.
1711 * The length of the headroom.
1713 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1715 __rte_mbuf_sanity_check(m, 1);
1720 * Get the tailroom of a packet mbuf.
1725 * The length of the tailroom.
1727 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1729 __rte_mbuf_sanity_check(m, 1);
1730 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1735 * Get the last segment of the packet.
1740 * The last segment of the given mbuf.
1742 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1744 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1746 __rte_mbuf_sanity_check(m, 1);
1747 while (m2->next != NULL)
1753 * A macro that points to an offset into the data in the mbuf.
1755 * The returned pointer is cast to type t. Before using this
1756 * function, the user must ensure that the first segment is large
1757 * enough to accommodate its data.
1762 * The offset into the mbuf data.
1764 * The type to cast the result into.
1766 #define rte_pktmbuf_mtod_offset(m, t, o) \
1767 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1770 * A macro that points to the start of the data in the mbuf.
1772 * The returned pointer is cast to type t. Before using this
1773 * function, the user must ensure that the first segment is large
1774 * enough to accommodate its data.
1779 * The type to cast the result into.
1781 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1784 * A macro that returns the physical address that points to an offset of the
1785 * start of the data in the mbuf
1790 * The offset into the data to calculate address from.
1792 #define rte_pktmbuf_mtophys_offset(m, o) \
1793 (phys_addr_t)((m)->buf_physaddr + (m)->data_off + (o))
1796 * A macro that returns the physical address that points to the start of the
1802 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_mtophys_offset(m, 0)
1805 * A macro that returns the length of the packet.
1807 * The value can be read or assigned.
1812 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1815 * A macro that returns the length of the segment.
1817 * The value can be read or assigned.
1822 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1825 * Prepend len bytes to an mbuf data area.
1827 * Returns a pointer to the new
1828 * data start address. If there is not enough headroom in the first
1829 * segment, the function will return NULL, without modifying the mbuf.
1834 * The amount of data to prepend (in bytes).
1836 * A pointer to the start of the newly prepended data, or
1837 * NULL if there is not enough headroom space in the first segment
1839 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1842 __rte_mbuf_sanity_check(m, 1);
1844 if (unlikely(len > rte_pktmbuf_headroom(m)))
1848 m->data_len = (uint16_t)(m->data_len + len);
1849 m->pkt_len = (m->pkt_len + len);
1851 return (char *)m->buf_addr + m->data_off;
1855 * Append len bytes to an mbuf.
1857 * Append len bytes to an mbuf and return a pointer to the start address
1858 * of the added data. If there is not enough tailroom in the last
1859 * segment, the function will return NULL, without modifying the mbuf.
1864 * The amount of data to append (in bytes).
1866 * A pointer to the start of the newly appended data, or
1867 * NULL if there is not enough tailroom space in the last segment
1869 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1872 struct rte_mbuf *m_last;
1874 __rte_mbuf_sanity_check(m, 1);
1876 m_last = rte_pktmbuf_lastseg(m);
1877 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1880 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1881 m_last->data_len = (uint16_t)(m_last->data_len + len);
1882 m->pkt_len = (m->pkt_len + len);
1883 return (char*) tail;
1887 * Remove len bytes at the beginning of an mbuf.
1889 * Returns a pointer to the start address of the new data area. If the
1890 * length is greater than the length of the first segment, then the
1891 * function will fail and return NULL, without modifying the mbuf.
1896 * The amount of data to remove (in bytes).
1898 * A pointer to the new start of the data.
1900 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1902 __rte_mbuf_sanity_check(m, 1);
1904 if (unlikely(len > m->data_len))
1907 m->data_len = (uint16_t)(m->data_len - len);
1909 m->pkt_len = (m->pkt_len - len);
1910 return (char *)m->buf_addr + m->data_off;
1914 * Remove len bytes of data at the end of the mbuf.
1916 * If the length is greater than the length of the last segment, the
1917 * function will fail and return -1 without modifying the mbuf.
1922 * The amount of data to remove (in bytes).
1927 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1929 struct rte_mbuf *m_last;
1931 __rte_mbuf_sanity_check(m, 1);
1933 m_last = rte_pktmbuf_lastseg(m);
1934 if (unlikely(len > m_last->data_len))
1937 m_last->data_len = (uint16_t)(m_last->data_len - len);
1938 m->pkt_len = (m->pkt_len - len);
1943 * Test if mbuf data is contiguous.
1948 * - 1, if all data is contiguous (one segment).
1949 * - 0, if there is several segments.
1951 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1953 __rte_mbuf_sanity_check(m, 1);
1954 return !!(m->nb_segs == 1);
1958 * Chain an mbuf to another, thereby creating a segmented packet.
1960 * Note: The implementation will do a linear walk over the segments to find
1961 * the tail entry. For cases when there are many segments, it's better to
1962 * chain the entries manually.
1965 * The head of the mbuf chain (the first packet)
1967 * The mbuf to put last in the chain
1971 * - -EOVERFLOW, if the chain is full (256 entries)
1973 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1975 struct rte_mbuf *cur_tail;
1977 /* Check for number-of-segments-overflow */
1978 if (head->nb_segs + tail->nb_segs >= 1 << (sizeof(head->nb_segs) * 8))
1981 /* Chain 'tail' onto the old tail */
1982 cur_tail = rte_pktmbuf_lastseg(head);
1983 cur_tail->next = tail;
1985 /* accumulate number of segments and total length. */
1986 head->nb_segs = (uint8_t)(head->nb_segs + tail->nb_segs);
1987 head->pkt_len += tail->pkt_len;
1989 /* pkt_len is only set in the head */
1990 tail->pkt_len = tail->data_len;
1996 * Dump an mbuf structure to the console.
1998 * Dump all fields for the given packet mbuf and all its associated
1999 * segments (in the case of a chained buffer).
2002 * A pointer to a file for output
2006 * If dump_len != 0, also dump the "dump_len" first data bytes of
2009 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
2015 #endif /* _RTE_MBUF_H_ */