4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
5 * Copyright 2014 6WIND S.A.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * * Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * * Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * * Neither the name of Intel Corporation nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
25 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
26 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
42 * The mbuf library provides the ability to create and destroy buffers
43 * that may be used by the RTE application to store message
44 * buffers. The message buffers are stored in a mempool, using the
45 * RTE mempool library.
47 * This library provide an API to allocate/free packet mbufs, which are
48 * used to carry network packets.
50 * To understand the concepts of packet buffers or mbufs, you
51 * should read "TCP/IP Illustrated, Volume 2: The Implementation,
52 * Addison-Wesley, 1995, ISBN 0-201-63354-X from Richard Stevens"
53 * http://www.kohala.com/start/tcpipiv2.html
57 #include <rte_common.h>
58 #include <rte_mempool.h>
59 #include <rte_memory.h>
60 #include <rte_atomic.h>
61 #include <rte_prefetch.h>
62 #include <rte_branch_prediction.h>
69 * Packet Offload Features Flags. It also carry packet type information.
70 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
72 * - RX flags start at bit position zero, and get added to the left of previous
74 * - The most-significant 3 bits are reserved for generic mbuf flags
75 * - TX flags therefore start at bit position 60 (i.e. 63-3), and new flags get
76 * added to the right of the previously defined flags i.e. they should count
77 * downwards, not upwards.
79 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
80 * rte_get_tx_ol_flag_name().
84 * RX packet is a 802.1q VLAN packet. This flag was set by PMDs when
85 * the packet is recognized as a VLAN, but the behavior between PMDs
86 * was not the same. This flag is kept for some time to avoid breaking
87 * applications and should be replaced by PKT_RX_VLAN_STRIPPED.
89 #define PKT_RX_VLAN_PKT (1ULL << 0)
91 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
92 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
93 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
94 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
95 #define PKT_RX_EIP_CKSUM_BAD (1ULL << 5) /**< External IP header checksum error. */
98 * A vlan has been stripped by the hardware and its tci is saved in
99 * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
100 * in the RX configuration of the PMD.
102 #define PKT_RX_VLAN_STRIPPED (1ULL << 6)
104 /* hole, some bits can be reused here */
106 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
107 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
108 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
109 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
112 * The 2 vlans have been stripped by the hardware and their tci are
113 * saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
114 * This can only happen if vlan stripping is enabled in the RX
115 * configuration of the PMD. If this flag is set, PKT_RX_VLAN_STRIPPED
118 #define PKT_RX_QINQ_STRIPPED (1ULL << 15)
122 * RX packet with double VLAN stripped.
123 * This flag is replaced by PKT_RX_QINQ_STRIPPED.
125 #define PKT_RX_QINQ_PKT PKT_RX_QINQ_STRIPPED
127 /* add new RX flags here */
129 /* add new TX flags here */
132 * Second VLAN insertion (QinQ) flag.
134 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
137 * TCP segmentation offload. To enable this offload feature for a
138 * packet to be transmitted on hardware supporting TSO:
139 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
141 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
142 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
144 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
145 * - calculate the pseudo header checksum without taking ip_len in account,
146 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
147 * rte_ipv6_phdr_cksum() that can be used as helpers.
149 #define PKT_TX_TCP_SEG (1ULL << 50)
151 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
154 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
155 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
156 * L4 checksum offload, the user needs to:
157 * - fill l2_len and l3_len in mbuf
158 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
159 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
160 * - calculate the pseudo header checksum and set it in the L4 header (only
161 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
162 * For SCTP, set the crc field to 0.
164 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
165 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
166 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
167 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
168 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
171 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
172 * also be set by the application, although a PMD will only check
174 * - set the IP checksum field in the packet to 0
175 * - fill the mbuf offload information: l2_len, l3_len
177 #define PKT_TX_IP_CKSUM (1ULL << 54)
180 * Packet is IPv4. This flag must be set when using any offload feature
181 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
182 * packet. If the packet is a tunneled packet, this flag is related to
185 #define PKT_TX_IPV4 (1ULL << 55)
188 * Packet is IPv6. This flag must be set when using an offload feature
189 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
190 * packet. If the packet is a tunneled packet, this flag is related to
193 #define PKT_TX_IPV6 (1ULL << 56)
195 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
198 * Offload the IP checksum of an external header in the hardware. The
199 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
200 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
201 * packet must be set to 0.
202 * - set the outer IP checksum field in the packet to 0
203 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
205 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
208 * Packet outer header is IPv4. This flag must be set when using any
209 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
210 * outer header of the tunneled packet is an IPv4 packet.
212 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
215 * Packet outer header is IPv6. This flag must be set when using any
216 * outer offload feature (L4 checksum) to tell the NIC that the outer
217 * header of the tunneled packet is an IPv6 packet.
219 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
221 #define __RESERVED (1ULL << 61) /**< reserved for future mbuf use */
223 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
225 /* Use final bit of flags to indicate a control mbuf */
226 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
229 * 32 bits are divided into several fields to mark packet types. Note that
230 * each field is indexical.
231 * - Bit 3:0 is for L2 types.
232 * - Bit 7:4 is for L3 or outer L3 (for tunneling case) types.
233 * - Bit 11:8 is for L4 or outer L4 (for tunneling case) types.
234 * - Bit 15:12 is for tunnel types.
235 * - Bit 19:16 is for inner L2 types.
236 * - Bit 23:20 is for inner L3 types.
237 * - Bit 27:24 is for inner L4 types.
238 * - Bit 31:28 is reserved.
240 * To be compatible with Vector PMD, RTE_PTYPE_L3_IPV4, RTE_PTYPE_L3_IPV4_EXT,
241 * RTE_PTYPE_L3_IPV6, RTE_PTYPE_L3_IPV6_EXT, RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP
242 * and RTE_PTYPE_L4_SCTP should be kept as below in a contiguous 7 bits.
244 * Note that L3 types values are selected for checking IPV4/IPV6 header from
245 * performance point of view. Reading annotations of RTE_ETH_IS_IPV4_HDR and
246 * RTE_ETH_IS_IPV6_HDR is needed for any future changes of L3 type values.
248 * Note that the packet types of the same packet recognized by different
249 * hardware may be different, as different hardware may have different
250 * capability of packet type recognition.
253 * <'ether type'=0x0800
254 * | 'version'=4, 'protocol'=0x29
255 * | 'version'=6, 'next header'=0x3A
257 * will be recognized on i40e hardware as packet type combination of,
258 * RTE_PTYPE_L2_ETHER |
259 * RTE_PTYPE_L3_IPV4_EXT_UNKNOWN |
260 * RTE_PTYPE_TUNNEL_IP |
261 * RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
262 * RTE_PTYPE_INNER_L4_ICMP.
264 * <'ether type'=0x86DD
265 * | 'version'=6, 'next header'=0x2F
267 * | 'version'=6, 'next header'=0x11
269 * will be recognized on i40e hardware as packet type combination of,
270 * RTE_PTYPE_L2_ETHER |
271 * RTE_PTYPE_L3_IPV6_EXT_UNKNOWN |
272 * RTE_PTYPE_TUNNEL_GRENAT |
273 * RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN |
274 * RTE_PTYPE_INNER_L4_UDP.
276 #define RTE_PTYPE_UNKNOWN 0x00000000
278 * Ethernet packet type.
279 * It is used for outer packet for tunneling cases.
282 * <'ether type'=[0x0800|0x86DD]>
284 #define RTE_PTYPE_L2_ETHER 0x00000001
286 * Ethernet packet type for time sync.
289 * <'ether type'=0x88F7>
291 #define RTE_PTYPE_L2_ETHER_TIMESYNC 0x00000002
293 * ARP (Address Resolution Protocol) packet type.
296 * <'ether type'=0x0806>
298 #define RTE_PTYPE_L2_ETHER_ARP 0x00000003
300 * LLDP (Link Layer Discovery Protocol) packet type.
303 * <'ether type'=0x88CC>
305 #define RTE_PTYPE_L2_ETHER_LLDP 0x00000004
307 * NSH (Network Service Header) packet type.
310 * <'ether type'=0x894F>
312 #define RTE_PTYPE_L2_ETHER_NSH 0x00000005
314 * Mask of layer 2 packet types.
315 * It is used for outer packet for tunneling cases.
317 #define RTE_PTYPE_L2_MASK 0x0000000f
319 * IP (Internet Protocol) version 4 packet type.
320 * It is used for outer packet for tunneling cases, and does not contain any
324 * <'ether type'=0x0800
325 * | 'version'=4, 'ihl'=5>
327 #define RTE_PTYPE_L3_IPV4 0x00000010
329 * IP (Internet Protocol) version 4 packet type.
330 * It is used for outer packet for tunneling cases, and contains header
334 * <'ether type'=0x0800
335 * | 'version'=4, 'ihl'=[6-15], 'options'>
337 #define RTE_PTYPE_L3_IPV4_EXT 0x00000030
339 * IP (Internet Protocol) version 6 packet type.
340 * It is used for outer packet for tunneling cases, and does not contain any
344 * <'ether type'=0x86DD
345 * | 'version'=6, 'next header'=0x3B>
347 #define RTE_PTYPE_L3_IPV6 0x00000040
349 * IP (Internet Protocol) version 4 packet type.
350 * It is used for outer packet for tunneling cases, and may or maynot contain
354 * <'ether type'=0x0800
355 * | 'version'=4, 'ihl'=[5-15], <'options'>>
357 #define RTE_PTYPE_L3_IPV4_EXT_UNKNOWN 0x00000090
359 * IP (Internet Protocol) version 6 packet type.
360 * It is used for outer packet for tunneling cases, and contains extension
364 * <'ether type'=0x86DD
365 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
366 * 'extension headers'>
368 #define RTE_PTYPE_L3_IPV6_EXT 0x000000c0
370 * IP (Internet Protocol) version 6 packet type.
371 * It is used for outer packet for tunneling cases, and may or maynot contain
375 * <'ether type'=0x86DD
376 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
377 * <'extension headers'>>
379 #define RTE_PTYPE_L3_IPV6_EXT_UNKNOWN 0x000000e0
381 * Mask of layer 3 packet types.
382 * It is used for outer packet for tunneling cases.
384 #define RTE_PTYPE_L3_MASK 0x000000f0
386 * TCP (Transmission Control Protocol) packet type.
387 * It is used for outer packet for tunneling cases.
390 * <'ether type'=0x0800
391 * | 'version'=4, 'protocol'=6, 'MF'=0>
393 * <'ether type'=0x86DD
394 * | 'version'=6, 'next header'=6>
396 #define RTE_PTYPE_L4_TCP 0x00000100
398 * UDP (User Datagram Protocol) packet type.
399 * It is used for outer packet for tunneling cases.
402 * <'ether type'=0x0800
403 * | 'version'=4, 'protocol'=17, 'MF'=0>
405 * <'ether type'=0x86DD
406 * | 'version'=6, 'next header'=17>
408 #define RTE_PTYPE_L4_UDP 0x00000200
410 * Fragmented IP (Internet Protocol) packet type.
411 * It is used for outer packet for tunneling cases.
413 * It refers to those packets of any IP types, which can be recognized as
414 * fragmented. A fragmented packet cannot be recognized as any other L4 types
415 * (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP,
416 * RTE_PTYPE_L4_NONFRAG).
419 * <'ether type'=0x0800
420 * | 'version'=4, 'MF'=1>
422 * <'ether type'=0x86DD
423 * | 'version'=6, 'next header'=44>
425 #define RTE_PTYPE_L4_FRAG 0x00000300
427 * SCTP (Stream Control Transmission Protocol) packet type.
428 * It is used for outer packet for tunneling cases.
431 * <'ether type'=0x0800
432 * | 'version'=4, 'protocol'=132, 'MF'=0>
434 * <'ether type'=0x86DD
435 * | 'version'=6, 'next header'=132>
437 #define RTE_PTYPE_L4_SCTP 0x00000400
439 * ICMP (Internet Control Message Protocol) packet type.
440 * It is used for outer packet for tunneling cases.
443 * <'ether type'=0x0800
444 * | 'version'=4, 'protocol'=1, 'MF'=0>
446 * <'ether type'=0x86DD
447 * | 'version'=6, 'next header'=1>
449 #define RTE_PTYPE_L4_ICMP 0x00000500
451 * Non-fragmented IP (Internet Protocol) packet type.
452 * It is used for outer packet for tunneling cases.
454 * It refers to those packets of any IP types, while cannot be recognized as
455 * any of above L4 types (RTE_PTYPE_L4_TCP, RTE_PTYPE_L4_UDP,
456 * RTE_PTYPE_L4_FRAG, RTE_PTYPE_L4_SCTP, RTE_PTYPE_L4_ICMP).
459 * <'ether type'=0x0800
460 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
462 * <'ether type'=0x86DD
463 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
465 #define RTE_PTYPE_L4_NONFRAG 0x00000600
467 * Mask of layer 4 packet types.
468 * It is used for outer packet for tunneling cases.
470 #define RTE_PTYPE_L4_MASK 0x00000f00
472 * IP (Internet Protocol) in IP (Internet Protocol) tunneling packet type.
475 * <'ether type'=0x0800
476 * | 'version'=4, 'protocol'=[4|41]>
478 * <'ether type'=0x86DD
479 * | 'version'=6, 'next header'=[4|41]>
481 #define RTE_PTYPE_TUNNEL_IP 0x00001000
483 * GRE (Generic Routing Encapsulation) tunneling packet type.
486 * <'ether type'=0x0800
487 * | 'version'=4, 'protocol'=47>
489 * <'ether type'=0x86DD
490 * | 'version'=6, 'next header'=47>
492 #define RTE_PTYPE_TUNNEL_GRE 0x00002000
494 * VXLAN (Virtual eXtensible Local Area Network) tunneling packet type.
497 * <'ether type'=0x0800
498 * | 'version'=4, 'protocol'=17
499 * | 'destination port'=4798>
501 * <'ether type'=0x86DD
502 * | 'version'=6, 'next header'=17
503 * | 'destination port'=4798>
505 #define RTE_PTYPE_TUNNEL_VXLAN 0x00003000
507 * NVGRE (Network Virtualization using Generic Routing Encapsulation) tunneling
511 * <'ether type'=0x0800
512 * | 'version'=4, 'protocol'=47
513 * | 'protocol type'=0x6558>
515 * <'ether type'=0x86DD
516 * | 'version'=6, 'next header'=47
517 * | 'protocol type'=0x6558'>
519 #define RTE_PTYPE_TUNNEL_NVGRE 0x00004000
521 * GENEVE (Generic Network Virtualization Encapsulation) tunneling packet type.
524 * <'ether type'=0x0800
525 * | 'version'=4, 'protocol'=17
526 * | 'destination port'=6081>
528 * <'ether type'=0x86DD
529 * | 'version'=6, 'next header'=17
530 * | 'destination port'=6081>
532 #define RTE_PTYPE_TUNNEL_GENEVE 0x00005000
534 * Tunneling packet type of Teredo, VXLAN (Virtual eXtensible Local Area
535 * Network) or GRE (Generic Routing Encapsulation) could be recognized as this
536 * packet type, if they can not be recognized independently as of hardware
539 #define RTE_PTYPE_TUNNEL_GRENAT 0x00006000
541 * Mask of tunneling packet types.
543 #define RTE_PTYPE_TUNNEL_MASK 0x0000f000
545 * Ethernet packet type.
546 * It is used for inner packet type only.
548 * Packet format (inner only):
549 * <'ether type'=[0x800|0x86DD]>
551 #define RTE_PTYPE_INNER_L2_ETHER 0x00010000
553 * Ethernet packet type with VLAN (Virtual Local Area Network) tag.
555 * Packet format (inner only):
556 * <'ether type'=[0x800|0x86DD], vlan=[1-4095]>
558 #define RTE_PTYPE_INNER_L2_ETHER_VLAN 0x00020000
560 * Mask of inner layer 2 packet types.
562 #define RTE_PTYPE_INNER_L2_MASK 0x000f0000
564 * IP (Internet Protocol) version 4 packet type.
565 * It is used for inner packet only, and does not contain any header option.
567 * Packet format (inner only):
568 * <'ether type'=0x0800
569 * | 'version'=4, 'ihl'=5>
571 #define RTE_PTYPE_INNER_L3_IPV4 0x00100000
573 * IP (Internet Protocol) version 4 packet type.
574 * It is used for inner packet only, and contains header options.
576 * Packet format (inner only):
577 * <'ether type'=0x0800
578 * | 'version'=4, 'ihl'=[6-15], 'options'>
580 #define RTE_PTYPE_INNER_L3_IPV4_EXT 0x00200000
582 * IP (Internet Protocol) version 6 packet type.
583 * It is used for inner packet only, and does not contain any extension header.
585 * Packet format (inner only):
586 * <'ether type'=0x86DD
587 * | 'version'=6, 'next header'=0x3B>
589 #define RTE_PTYPE_INNER_L3_IPV6 0x00300000
591 * IP (Internet Protocol) version 4 packet type.
592 * It is used for inner packet only, and may or maynot contain header options.
594 * Packet format (inner only):
595 * <'ether type'=0x0800
596 * | 'version'=4, 'ihl'=[5-15], <'options'>>
598 #define RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN 0x00400000
600 * IP (Internet Protocol) version 6 packet type.
601 * It is used for inner packet only, and contains extension headers.
603 * Packet format (inner only):
604 * <'ether type'=0x86DD
605 * | 'version'=6, 'next header'=[0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
606 * 'extension headers'>
608 #define RTE_PTYPE_INNER_L3_IPV6_EXT 0x00500000
610 * IP (Internet Protocol) version 6 packet type.
611 * It is used for inner packet only, and may or maynot contain extension
614 * Packet format (inner only):
615 * <'ether type'=0x86DD
616 * | 'version'=6, 'next header'=[0x3B|0x0|0x2B|0x2C|0x32|0x33|0x3C|0x87],
617 * <'extension headers'>>
619 #define RTE_PTYPE_INNER_L3_IPV6_EXT_UNKNOWN 0x00600000
621 * Mask of inner layer 3 packet types.
623 #define RTE_PTYPE_INNER_L3_MASK 0x00f00000
625 * TCP (Transmission Control Protocol) packet type.
626 * It is used for inner packet only.
628 * Packet format (inner only):
629 * <'ether type'=0x0800
630 * | 'version'=4, 'protocol'=6, 'MF'=0>
632 * <'ether type'=0x86DD
633 * | 'version'=6, 'next header'=6>
635 #define RTE_PTYPE_INNER_L4_TCP 0x01000000
637 * UDP (User Datagram Protocol) packet type.
638 * It is used for inner packet only.
640 * Packet format (inner only):
641 * <'ether type'=0x0800
642 * | 'version'=4, 'protocol'=17, 'MF'=0>
644 * <'ether type'=0x86DD
645 * | 'version'=6, 'next header'=17>
647 #define RTE_PTYPE_INNER_L4_UDP 0x02000000
649 * Fragmented IP (Internet Protocol) packet type.
650 * It is used for inner packet only, and may or maynot have layer 4 packet.
652 * Packet format (inner only):
653 * <'ether type'=0x0800
654 * | 'version'=4, 'MF'=1>
656 * <'ether type'=0x86DD
657 * | 'version'=6, 'next header'=44>
659 #define RTE_PTYPE_INNER_L4_FRAG 0x03000000
661 * SCTP (Stream Control Transmission Protocol) packet type.
662 * It is used for inner packet only.
664 * Packet format (inner only):
665 * <'ether type'=0x0800
666 * | 'version'=4, 'protocol'=132, 'MF'=0>
668 * <'ether type'=0x86DD
669 * | 'version'=6, 'next header'=132>
671 #define RTE_PTYPE_INNER_L4_SCTP 0x04000000
673 * ICMP (Internet Control Message Protocol) packet type.
674 * It is used for inner packet only.
676 * Packet format (inner only):
677 * <'ether type'=0x0800
678 * | 'version'=4, 'protocol'=1, 'MF'=0>
680 * <'ether type'=0x86DD
681 * | 'version'=6, 'next header'=1>
683 #define RTE_PTYPE_INNER_L4_ICMP 0x05000000
685 * Non-fragmented IP (Internet Protocol) packet type.
686 * It is used for inner packet only, and may or maynot have other unknown layer
689 * Packet format (inner only):
690 * <'ether type'=0x0800
691 * | 'version'=4, 'protocol'!=[6|17|132|1], 'MF'=0>
693 * <'ether type'=0x86DD
694 * | 'version'=6, 'next header'!=[6|17|44|132|1]>
696 #define RTE_PTYPE_INNER_L4_NONFRAG 0x06000000
698 * Mask of inner layer 4 packet types.
700 #define RTE_PTYPE_INNER_L4_MASK 0x0f000000
703 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
704 * one, bit 4 is selected to be used for IPv4 only. Then checking bit 4 can
705 * determine if it is an IPV4 packet.
707 #define RTE_ETH_IS_IPV4_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV4)
710 * Check if the (outer) L3 header is IPv4. To avoid comparing IPv4 types one by
711 * one, bit 6 is selected to be used for IPv4 only. Then checking bit 6 can
712 * determine if it is an IPV4 packet.
714 #define RTE_ETH_IS_IPV6_HDR(ptype) ((ptype) & RTE_PTYPE_L3_IPV6)
716 /* Check if it is a tunneling packet */
717 #define RTE_ETH_IS_TUNNEL_PKT(ptype) ((ptype) & (RTE_PTYPE_TUNNEL_MASK | \
718 RTE_PTYPE_INNER_L2_MASK | \
719 RTE_PTYPE_INNER_L3_MASK | \
720 RTE_PTYPE_INNER_L4_MASK))
722 /** Alignment constraint of mbuf private area. */
723 #define RTE_MBUF_PRIV_ALIGN 8
726 * Get the name of a RX offload flag
729 * The mask describing the flag.
731 * The name of this flag, or NULL if it's not a valid RX flag.
733 const char *rte_get_rx_ol_flag_name(uint64_t mask);
736 * Get the name of a TX offload flag
739 * The mask describing the flag. Usually only one bit must be set.
740 * Several bits can be given if they belong to the same mask.
741 * Ex: PKT_TX_L4_MASK.
743 * The name of this flag, or NULL if it's not a valid TX flag.
745 const char *rte_get_tx_ol_flag_name(uint64_t mask);
748 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
749 * splitting it into multiple segments.
750 * So, for mbufs that planned to be involved into RX/TX, the recommended
751 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
753 #define RTE_MBUF_DEFAULT_DATAROOM 2048
754 #define RTE_MBUF_DEFAULT_BUF_SIZE \
755 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
757 /* define a set of marker types that can be used to refer to set points in the
759 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
760 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
761 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
762 * with a single assignment */
765 * The generic rte_mbuf, containing a packet mbuf.
770 void *buf_addr; /**< Virtual address of segment buffer. */
771 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
773 uint16_t buf_len; /**< Length of segment buffer. */
775 /* next 6 bytes are initialised on RX descriptor rearm */
780 * 16-bit Reference counter.
781 * It should only be accessed using the following functions:
782 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
783 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
784 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
788 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
789 uint16_t refcnt; /**< Non-atomically accessed refcnt */
791 uint8_t nb_segs; /**< Number of segments. */
792 uint8_t port; /**< Input port. */
794 uint64_t ol_flags; /**< Offload features. */
796 /* remaining bytes are set on RX when pulling packet from descriptor */
797 MARKER rx_descriptor_fields1;
800 * The packet type, which is the combination of outer/inner L2, L3, L4
801 * and tunnel types. The packet_type is about data really present in the
802 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
803 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
804 * vlan is stripped from the data.
807 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
809 uint32_t l2_type:4; /**< (Outer) L2 type. */
810 uint32_t l3_type:4; /**< (Outer) L3 type. */
811 uint32_t l4_type:4; /**< (Outer) L4 type. */
812 uint32_t tun_type:4; /**< Tunnel type. */
813 uint32_t inner_l2_type:4; /**< Inner L2 type. */
814 uint32_t inner_l3_type:4; /**< Inner L3 type. */
815 uint32_t inner_l4_type:4; /**< Inner L4 type. */
819 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
820 uint16_t data_len; /**< Amount of data in segment buffer. */
821 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN_STRIPPED is set. */
825 uint32_t rss; /**< RSS hash result if RSS enabled */
833 /**< Second 4 flexible bytes */
836 /**< First 4 flexible bytes or FD ID, dependent on
837 PKT_RX_FDIR_* flag in ol_flags. */
838 } fdir; /**< Filter identifier if FDIR enabled */
842 } sched; /**< Hierarchical scheduler */
843 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
844 } hash; /**< hash information */
846 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
848 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ_STRIPPED is set. */
849 uint16_t vlan_tci_outer;
851 /* second cache line - fields only used in slow path or on TX */
852 MARKER cacheline1 __rte_cache_min_aligned;
855 void *userdata; /**< Can be used for external metadata */
856 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
859 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
860 struct rte_mbuf *next; /**< Next segment of scattered packet. */
862 /* fields to support TX offloads */
864 uint64_t tx_offload; /**< combined for easy fetch */
866 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
867 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
868 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
869 uint64_t tso_segsz:16; /**< TCP TSO segment size */
871 /* fields for TX offloading of tunnels */
872 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
873 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
875 /* uint64_t unused:8; */
879 /** Size of the application private data. In case of an indirect
880 * mbuf, it stores the direct mbuf private data size. */
883 /** Timesync flags for use with IEEE1588. */
885 } __rte_cache_aligned;
888 * Prefetch the first part of the mbuf
890 * The first 64 bytes of the mbuf corresponds to fields that are used early
891 * in the receive path. If the cache line of the architecture is higher than
892 * 64B, the second part will also be prefetched.
895 * The pointer to the mbuf.
898 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
900 rte_prefetch0(&m->cacheline0);
904 * Prefetch the second part of the mbuf
906 * The next 64 bytes of the mbuf corresponds to fields that are used in the
907 * transmit path. If the cache line of the architecture is higher than 64B,
908 * this function does nothing as it is expected that the full mbuf is
912 * The pointer to the mbuf.
915 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
917 #if RTE_CACHE_LINE_SIZE == 64
918 rte_prefetch0(&m->cacheline1);
925 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
928 * Return the DMA address of the beginning of the mbuf data
931 * The pointer to the mbuf.
933 * The physical address of the beginning of the mbuf data
935 static inline phys_addr_t
936 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
938 return mb->buf_physaddr + mb->data_off;
942 * Return the default DMA address of the beginning of the mbuf data
944 * This function is used by drivers in their receive function, as it
945 * returns the location where data should be written by the NIC, taking
946 * the default headroom in account.
949 * The pointer to the mbuf.
951 * The physical address of the beginning of the mbuf data
953 static inline phys_addr_t
954 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
956 return mb->buf_physaddr + RTE_PKTMBUF_HEADROOM;
960 * Return the mbuf owning the data buffer address of an indirect mbuf.
963 * The pointer to the indirect mbuf.
965 * The address of the direct mbuf corresponding to buffer_addr.
967 static inline struct rte_mbuf *
968 rte_mbuf_from_indirect(struct rte_mbuf *mi)
970 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
974 * Return the buffer address embedded in the given mbuf.
977 * The pointer to the mbuf.
979 * The address of the data buffer owned by the mbuf.
982 rte_mbuf_to_baddr(struct rte_mbuf *md)
985 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
990 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
992 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
995 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
997 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
1000 * Private data in case of pktmbuf pool.
1002 * A structure that contains some pktmbuf_pool-specific data that are
1003 * appended after the mempool structure (in private data).
1005 struct rte_pktmbuf_pool_private {
1006 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
1007 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
1010 #ifdef RTE_LIBRTE_MBUF_DEBUG
1012 /** check mbuf type in debug mode */
1013 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
1015 #else /* RTE_LIBRTE_MBUF_DEBUG */
1017 /** check mbuf type in debug mode */
1018 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
1020 #endif /* RTE_LIBRTE_MBUF_DEBUG */
1022 #ifdef RTE_MBUF_REFCNT_ATOMIC
1025 * Reads the value of an mbuf's refcnt.
1029 * Reference count number.
1031 static inline uint16_t
1032 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1034 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
1038 * Sets an mbuf's refcnt to a defined value.
1045 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1047 rte_atomic16_set(&m->refcnt_atomic, new_value);
1051 * Adds given value to an mbuf's refcnt and returns its new value.
1055 * Value to add/subtract
1059 static inline uint16_t
1060 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1063 * The atomic_add is an expensive operation, so we don't want to
1064 * call it in the case where we know we are the uniq holder of
1065 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
1066 * operation has to be used because concurrent accesses on the
1067 * reference counter can occur.
1069 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1070 rte_mbuf_refcnt_set(m, 1 + value);
1074 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
1077 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
1080 * Adds given value to an mbuf's refcnt and returns its new value.
1082 static inline uint16_t
1083 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
1085 m->refcnt = (uint16_t)(m->refcnt + value);
1090 * Reads the value of an mbuf's refcnt.
1092 static inline uint16_t
1093 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
1099 * Sets an mbuf's refcnt to the defined value.
1102 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
1104 m->refcnt = new_value;
1107 #endif /* RTE_MBUF_REFCNT_ATOMIC */
1109 /** Mbuf prefetch */
1110 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
1117 * Sanity checks on an mbuf.
1119 * Check the consistency of the given mbuf. The function will cause a
1120 * panic if corruption is detected.
1123 * The mbuf to be checked.
1125 * True if the mbuf is a packet header, false if it is a sub-segment
1126 * of a packet (in this case, some fields like nb_segs are not checked)
1129 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
1132 * Allocate an unitialized mbuf from mempool *mp*.
1134 * This function can be used by PMDs (especially in RX functions) to
1135 * allocate an unitialized mbuf. The driver is responsible of
1136 * initializing all the required fields. See rte_pktmbuf_reset().
1137 * For standard needs, prefer rte_pktmbuf_alloc().
1140 * The mempool from which mbuf is allocated.
1142 * - The pointer to the new mbuf on success.
1143 * - NULL if allocation failed.
1145 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
1150 if (rte_mempool_get(mp, &mb) < 0)
1152 m = (struct rte_mbuf *)mb;
1153 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1154 rte_mbuf_refcnt_set(m, 1);
1155 __rte_mbuf_sanity_check(m, 0);
1160 /* compat with older versions */
1161 __rte_deprecated static inline struct rte_mbuf *
1162 __rte_mbuf_raw_alloc(struct rte_mempool *mp)
1164 return rte_mbuf_raw_alloc(mp);
1168 * @internal Put mbuf back into its original mempool.
1169 * The use of that function is reserved for RTE internal needs.
1170 * Please use rte_pktmbuf_free().
1173 * The mbuf to be freed.
1175 static inline void __attribute__((always_inline))
1176 __rte_mbuf_raw_free(struct rte_mbuf *m)
1178 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 0);
1179 rte_mempool_put(m->pool, m);
1182 /* Operations on ctrl mbuf */
1185 * The control mbuf constructor.
1187 * This function initializes some fields in an mbuf structure that are
1188 * not modified by the user once created (mbuf type, origin pool, buffer
1189 * start address, and so on). This function is given as a callback function
1190 * to rte_mempool_create() at pool creation time.
1193 * The mempool from which the mbuf is allocated.
1195 * A pointer that can be used by the user to retrieve useful information
1196 * for mbuf initialization. This pointer comes from the ``init_arg``
1197 * parameter of rte_mempool_create().
1199 * The mbuf to initialize.
1201 * The index of the mbuf in the pool table.
1203 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1204 void *m, unsigned i);
1207 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
1209 * This new mbuf is initialized with data pointing to the beginning of
1210 * buffer, and with a length of zero.
1213 * The mempool from which the mbuf is allocated.
1215 * - The pointer to the new mbuf on success.
1216 * - NULL if allocation failed.
1218 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
1221 * Free a control mbuf back into its original mempool.
1224 * The control mbuf to be freed.
1226 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
1229 * A macro that returns the pointer to the carried data.
1231 * The value that can be read or assigned.
1236 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
1239 * A macro that returns the length of the carried data.
1241 * The value that can be read or assigned.
1246 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
1249 * Tests if an mbuf is a control mbuf
1252 * The mbuf to be tested
1254 * - True (1) if the mbuf is a control mbuf
1255 * - False(0) otherwise
1258 rte_is_ctrlmbuf(struct rte_mbuf *m)
1260 return !!(m->ol_flags & CTRL_MBUF_FLAG);
1263 /* Operations on pkt mbuf */
1266 * The packet mbuf constructor.
1268 * This function initializes some fields in the mbuf structure that are
1269 * not modified by the user once created (origin pool, buffer start
1270 * address, and so on). This function is given as a callback function to
1271 * rte_mempool_create() at pool creation time.
1274 * The mempool from which mbufs originate.
1276 * A pointer that can be used by the user to retrieve useful information
1277 * for mbuf initialization. This pointer comes from the ``init_arg``
1278 * parameter of rte_mempool_create().
1280 * The mbuf to initialize.
1282 * The index of the mbuf in the pool table.
1284 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1285 void *m, unsigned i);
1289 * A packet mbuf pool constructor.
1291 * This function initializes the mempool private data in the case of a
1292 * pktmbuf pool. This private data is needed by the driver. The
1293 * function is given as a callback function to rte_mempool_create() at
1294 * pool creation. It can be extended by the user, for example, to
1295 * provide another packet size.
1298 * The mempool from which mbufs originate.
1300 * A pointer that can be used by the user to retrieve useful information
1301 * for mbuf initialization. This pointer comes from the ``init_arg``
1302 * parameter of rte_mempool_create().
1304 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1307 * Create a mbuf pool.
1309 * This function creates and initializes a packet mbuf pool. It is
1310 * a wrapper to rte_mempool_create() with the proper packet constructor
1311 * and mempool constructor.
1314 * The name of the mbuf pool.
1316 * The number of elements in the mbuf pool. The optimum size (in terms
1317 * of memory usage) for a mempool is when n is a power of two minus one:
1320 * Size of the per-core object cache. See rte_mempool_create() for
1323 * Size of application private are between the rte_mbuf structure
1324 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1325 * @param data_room_size
1326 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1328 * The socket identifier where the memory should be allocated. The
1329 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1332 * The pointer to the new allocated mempool, on success. NULL on error
1333 * with rte_errno set appropriately. Possible rte_errno values include:
1334 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1335 * - E_RTE_SECONDARY - function was called from a secondary process instance
1336 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1337 * - ENOSPC - the maximum number of memzones has already been allocated
1338 * - EEXIST - a memzone with the same name already exists
1339 * - ENOMEM - no appropriate memory area found in which to create memzone
1341 struct rte_mempool *
1342 rte_pktmbuf_pool_create(const char *name, unsigned n,
1343 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1347 * Get the data room size of mbufs stored in a pktmbuf_pool
1349 * The data room size is the amount of data that can be stored in a
1350 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1353 * The packet mbuf pool.
1355 * The data room size of mbufs stored in this mempool.
1357 static inline uint16_t
1358 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1360 struct rte_pktmbuf_pool_private *mbp_priv;
1362 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1363 return mbp_priv->mbuf_data_room_size;
1367 * Get the application private size of mbufs stored in a pktmbuf_pool
1369 * The private size of mbuf is a zone located between the rte_mbuf
1370 * structure and the data buffer where an application can store data
1371 * associated to a packet.
1374 * The packet mbuf pool.
1376 * The private size of mbufs stored in this mempool.
1378 static inline uint16_t
1379 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1381 struct rte_pktmbuf_pool_private *mbp_priv;
1383 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1384 return mbp_priv->mbuf_priv_size;
1388 * Reset the fields of a packet mbuf to their default values.
1390 * The given mbuf must have only one segment.
1393 * The packet mbuf to be resetted.
1395 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1401 m->vlan_tci_outer = 0;
1407 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
1408 RTE_PKTMBUF_HEADROOM : m->buf_len;
1411 __rte_mbuf_sanity_check(m, 1);
1415 * Allocate a new mbuf from a mempool.
1417 * This new mbuf contains one segment, which has a length of 0. The pointer
1418 * to data is initialized to have some bytes of headroom in the buffer
1419 * (if buffer size allows).
1422 * The mempool from which the mbuf is allocated.
1424 * - The pointer to the new mbuf on success.
1425 * - NULL if allocation failed.
1427 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1430 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1431 rte_pktmbuf_reset(m);
1436 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1440 * The mempool from which mbufs are allocated.
1442 * Array of pointers to mbufs
1448 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1449 struct rte_mbuf **mbufs, unsigned count)
1454 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1458 /* To understand duff's device on loop unwinding optimization, see
1459 * https://en.wikipedia.org/wiki/Duff's_device.
1460 * Here while() loop is used rather than do() while{} to avoid extra
1461 * check if count is zero.
1463 switch (count % 4) {
1465 while (idx != count) {
1466 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1467 rte_mbuf_refcnt_set(mbufs[idx], 1);
1468 rte_pktmbuf_reset(mbufs[idx]);
1471 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1472 rte_mbuf_refcnt_set(mbufs[idx], 1);
1473 rte_pktmbuf_reset(mbufs[idx]);
1476 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1477 rte_mbuf_refcnt_set(mbufs[idx], 1);
1478 rte_pktmbuf_reset(mbufs[idx]);
1481 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1482 rte_mbuf_refcnt_set(mbufs[idx], 1);
1483 rte_pktmbuf_reset(mbufs[idx]);
1491 * Attach packet mbuf to another packet mbuf.
1493 * After attachment we refer the mbuf we attached as 'indirect',
1494 * while mbuf we attached to as 'direct'.
1495 * The direct mbuf's reference counter is incremented.
1497 * Right now, not supported:
1498 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1499 * - mbuf we trying to attach (mi) is used by someone else
1500 * e.g. it's reference counter is greater then 1.
1503 * The indirect packet mbuf.
1505 * The packet mbuf we're attaching to.
1507 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1509 struct rte_mbuf *md;
1511 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1512 rte_mbuf_refcnt_read(mi) == 1);
1514 /* if m is not direct, get the mbuf that embeds the data */
1515 if (RTE_MBUF_DIRECT(m))
1518 md = rte_mbuf_from_indirect(m);
1520 rte_mbuf_refcnt_update(md, 1);
1521 mi->priv_size = m->priv_size;
1522 mi->buf_physaddr = m->buf_physaddr;
1523 mi->buf_addr = m->buf_addr;
1524 mi->buf_len = m->buf_len;
1527 mi->data_off = m->data_off;
1528 mi->data_len = m->data_len;
1530 mi->vlan_tci = m->vlan_tci;
1531 mi->vlan_tci_outer = m->vlan_tci_outer;
1532 mi->tx_offload = m->tx_offload;
1536 mi->pkt_len = mi->data_len;
1538 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1539 mi->packet_type = m->packet_type;
1541 __rte_mbuf_sanity_check(mi, 1);
1542 __rte_mbuf_sanity_check(m, 0);
1546 * Detach an indirect packet mbuf.
1548 * - restore original mbuf address and length values.
1549 * - reset pktmbuf data and data_len to their default values.
1550 * - decrement the direct mbuf's reference counter. When the
1551 * reference counter becomes 0, the direct mbuf is freed.
1553 * All other fields of the given packet mbuf will be left intact.
1556 * The indirect attached packet mbuf.
1558 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1560 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1561 struct rte_mempool *mp = m->pool;
1562 uint32_t mbuf_size, buf_len, priv_size;
1564 priv_size = rte_pktmbuf_priv_size(mp);
1565 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1566 buf_len = rte_pktmbuf_data_room_size(mp);
1568 m->priv_size = priv_size;
1569 m->buf_addr = (char *)m + mbuf_size;
1570 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
1571 m->buf_len = (uint16_t)buf_len;
1572 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1576 if (rte_mbuf_refcnt_update(md, -1) == 0)
1577 __rte_mbuf_raw_free(md);
1580 static inline struct rte_mbuf* __attribute__((always_inline))
1581 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1583 __rte_mbuf_sanity_check(m, 0);
1585 if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
1586 /* if this is an indirect mbuf, it is detached. */
1587 if (RTE_MBUF_INDIRECT(m))
1588 rte_pktmbuf_detach(m);
1595 * Free a segment of a packet mbuf into its original mempool.
1597 * Free an mbuf, without parsing other segments in case of chained
1601 * The packet mbuf segment to be freed.
1603 static inline void __attribute__((always_inline))
1604 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1606 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
1608 __rte_mbuf_raw_free(m);
1613 * Free a packet mbuf back into its original mempool.
1615 * Free an mbuf, and all its segments in case of chained buffers. Each
1616 * segment is added back into its original mempool.
1619 * The packet mbuf to be freed.
1621 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1623 struct rte_mbuf *m_next;
1625 __rte_mbuf_sanity_check(m, 1);
1629 rte_pktmbuf_free_seg(m);
1635 * Creates a "clone" of the given packet mbuf.
1637 * Walks through all segments of the given packet mbuf, and for each of them:
1638 * - Creates a new packet mbuf from the given pool.
1639 * - Attaches newly created mbuf to the segment.
1640 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1641 * from the original packet mbuf.
1644 * The packet mbuf to be cloned.
1646 * The mempool from which the "clone" mbufs are allocated.
1648 * - The pointer to the new "clone" mbuf on success.
1649 * - NULL if allocation fails.
1651 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1652 struct rte_mempool *mp)
1654 struct rte_mbuf *mc, *mi, **prev;
1658 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1663 pktlen = md->pkt_len;
1668 rte_pktmbuf_attach(mi, md);
1671 } while ((md = md->next) != NULL &&
1672 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1676 mc->pkt_len = pktlen;
1678 /* Allocation of new indirect segment failed */
1679 if (unlikely (mi == NULL)) {
1680 rte_pktmbuf_free(mc);
1684 __rte_mbuf_sanity_check(mc, 1);
1689 * Adds given value to the refcnt of all packet mbuf segments.
1691 * Walks through all segments of given packet mbuf and for each of them
1692 * invokes rte_mbuf_refcnt_update().
1695 * The packet mbuf whose refcnt to be updated.
1697 * The value to add to the mbuf's segments refcnt.
1699 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1701 __rte_mbuf_sanity_check(m, 1);
1704 rte_mbuf_refcnt_update(m, v);
1705 } while ((m = m->next) != NULL);
1709 * Get the headroom in a packet mbuf.
1714 * The length of the headroom.
1716 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1718 __rte_mbuf_sanity_check(m, 1);
1723 * Get the tailroom of a packet mbuf.
1728 * The length of the tailroom.
1730 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1732 __rte_mbuf_sanity_check(m, 1);
1733 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1738 * Get the last segment of the packet.
1743 * The last segment of the given mbuf.
1745 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1747 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1749 __rte_mbuf_sanity_check(m, 1);
1750 while (m2->next != NULL)
1756 * A macro that points to an offset into the data in the mbuf.
1758 * The returned pointer is cast to type t. Before using this
1759 * function, the user must ensure that the first segment is large
1760 * enough to accommodate its data.
1765 * The offset into the mbuf data.
1767 * The type to cast the result into.
1769 #define rte_pktmbuf_mtod_offset(m, t, o) \
1770 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1773 * A macro that points to the start of the data in the mbuf.
1775 * The returned pointer is cast to type t. Before using this
1776 * function, the user must ensure that the first segment is large
1777 * enough to accommodate its data.
1782 * The type to cast the result into.
1784 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1787 * A macro that returns the physical address that points to an offset of the
1788 * start of the data in the mbuf
1793 * The offset into the data to calculate address from.
1795 #define rte_pktmbuf_mtophys_offset(m, o) \
1796 (phys_addr_t)((m)->buf_physaddr + (m)->data_off + (o))
1799 * A macro that returns the physical address that points to the start of the
1805 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_mtophys_offset(m, 0)
1808 * A macro that returns the length of the packet.
1810 * The value can be read or assigned.
1815 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1818 * A macro that returns the length of the segment.
1820 * The value can be read or assigned.
1825 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1828 * Prepend len bytes to an mbuf data area.
1830 * Returns a pointer to the new
1831 * data start address. If there is not enough headroom in the first
1832 * segment, the function will return NULL, without modifying the mbuf.
1837 * The amount of data to prepend (in bytes).
1839 * A pointer to the start of the newly prepended data, or
1840 * NULL if there is not enough headroom space in the first segment
1842 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1845 __rte_mbuf_sanity_check(m, 1);
1847 if (unlikely(len > rte_pktmbuf_headroom(m)))
1851 m->data_len = (uint16_t)(m->data_len + len);
1852 m->pkt_len = (m->pkt_len + len);
1854 return (char *)m->buf_addr + m->data_off;
1858 * Append len bytes to an mbuf.
1860 * Append len bytes to an mbuf and return a pointer to the start address
1861 * of the added data. If there is not enough tailroom in the last
1862 * segment, the function will return NULL, without modifying the mbuf.
1867 * The amount of data to append (in bytes).
1869 * A pointer to the start of the newly appended data, or
1870 * NULL if there is not enough tailroom space in the last segment
1872 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1875 struct rte_mbuf *m_last;
1877 __rte_mbuf_sanity_check(m, 1);
1879 m_last = rte_pktmbuf_lastseg(m);
1880 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1883 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1884 m_last->data_len = (uint16_t)(m_last->data_len + len);
1885 m->pkt_len = (m->pkt_len + len);
1886 return (char*) tail;
1890 * Remove len bytes at the beginning of an mbuf.
1892 * Returns a pointer to the start address of the new data area. If the
1893 * length is greater than the length of the first segment, then the
1894 * function will fail and return NULL, without modifying the mbuf.
1899 * The amount of data to remove (in bytes).
1901 * A pointer to the new start of the data.
1903 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1905 __rte_mbuf_sanity_check(m, 1);
1907 if (unlikely(len > m->data_len))
1910 m->data_len = (uint16_t)(m->data_len - len);
1912 m->pkt_len = (m->pkt_len - len);
1913 return (char *)m->buf_addr + m->data_off;
1917 * Remove len bytes of data at the end of the mbuf.
1919 * If the length is greater than the length of the last segment, the
1920 * function will fail and return -1 without modifying the mbuf.
1925 * The amount of data to remove (in bytes).
1930 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1932 struct rte_mbuf *m_last;
1934 __rte_mbuf_sanity_check(m, 1);
1936 m_last = rte_pktmbuf_lastseg(m);
1937 if (unlikely(len > m_last->data_len))
1940 m_last->data_len = (uint16_t)(m_last->data_len - len);
1941 m->pkt_len = (m->pkt_len - len);
1946 * Test if mbuf data is contiguous.
1951 * - 1, if all data is contiguous (one segment).
1952 * - 0, if there is several segments.
1954 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1956 __rte_mbuf_sanity_check(m, 1);
1957 return !!(m->nb_segs == 1);
1961 * Chain an mbuf to another, thereby creating a segmented packet.
1963 * Note: The implementation will do a linear walk over the segments to find
1964 * the tail entry. For cases when there are many segments, it's better to
1965 * chain the entries manually.
1968 * The head of the mbuf chain (the first packet)
1970 * The mbuf to put last in the chain
1974 * - -EOVERFLOW, if the chain is full (256 entries)
1976 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1978 struct rte_mbuf *cur_tail;
1980 /* Check for number-of-segments-overflow */
1981 if (head->nb_segs + tail->nb_segs >= 1 << (sizeof(head->nb_segs) * 8))
1984 /* Chain 'tail' onto the old tail */
1985 cur_tail = rte_pktmbuf_lastseg(head);
1986 cur_tail->next = tail;
1988 /* accumulate number of segments and total length. */
1989 head->nb_segs = (uint8_t)(head->nb_segs + tail->nb_segs);
1990 head->pkt_len += tail->pkt_len;
1992 /* pkt_len is only set in the head */
1993 tail->pkt_len = tail->data_len;
1999 * Dump an mbuf structure to the console.
2001 * Dump all fields for the given packet mbuf and all its associated
2002 * segments (in the case of a chained buffer).
2005 * A pointer to a file for output
2009 * If dump_len != 0, also dump the "dump_len" first data bytes of
2012 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
2018 #endif /* _RTE_MBUF_H_ */