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 * The preferred way to create a mbuf pool is to use
48 * rte_pktmbuf_pool_create(). However, in some situations, an
49 * application may want to have more control (ex: populate the pool with
50 * specific memory), in this case it is possible to use functions from
51 * rte_mempool. See how rte_pktmbuf_pool_create() is implemented for
54 * This library provides an API to allocate/free packet mbufs, which are
55 * used to carry network packets.
57 * To understand the concepts of packet buffers or mbufs, you
58 * should read "TCP/IP Illustrated, Volume 2: The Implementation,
59 * Addison-Wesley, 1995, ISBN 0-201-63354-X from Richard Stevens"
60 * http://www.kohala.com/start/tcpipiv2.html
64 #include <rte_common.h>
65 #include <rte_mempool.h>
66 #include <rte_memory.h>
67 #include <rte_atomic.h>
68 #include <rte_prefetch.h>
69 #include <rte_branch_prediction.h>
70 #include <rte_mbuf_ptype.h>
77 * Packet Offload Features Flags. It also carry packet type information.
78 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
80 * - RX flags start at bit position zero, and get added to the left of previous
82 * - The most-significant 3 bits are reserved for generic mbuf flags
83 * - TX flags therefore start at bit position 60 (i.e. 63-3), and new flags get
84 * added to the right of the previously defined flags i.e. they should count
85 * downwards, not upwards.
87 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
88 * rte_get_tx_ol_flag_name().
92 * The RX packet is a 802.1q VLAN packet, and the tci has been
93 * saved in in mbuf->vlan_tci.
94 * If the flag PKT_RX_VLAN_STRIPPED is also present, the VLAN
95 * header has been stripped from mbuf data, else it is still
98 #define PKT_RX_VLAN (1ULL << 0)
100 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
101 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
105 * Checking this flag alone is deprecated: check the 2 bits of
106 * PKT_RX_L4_CKSUM_MASK.
107 * This flag was set when the L4 checksum of a packet was detected as
108 * wrong by the hardware.
110 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
114 * Checking this flag alone is deprecated: check the 2 bits of
115 * PKT_RX_IP_CKSUM_MASK.
116 * This flag was set when the IP checksum of a packet was detected as
117 * wrong by the hardware.
119 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
121 #define PKT_RX_EIP_CKSUM_BAD (1ULL << 5) /**< External IP header checksum error. */
124 * A vlan has been stripped by the hardware and its tci is saved in
125 * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
126 * in the RX configuration of the PMD.
127 * When PKT_RX_VLAN_STRIPPED is set, PKT_RX_VLAN must also be set.
129 #define PKT_RX_VLAN_STRIPPED (1ULL << 6)
132 * Mask of bits used to determine the status of RX IP checksum.
133 * - PKT_RX_IP_CKSUM_UNKNOWN: no information about the RX IP checksum
134 * - PKT_RX_IP_CKSUM_BAD: the IP checksum in the packet is wrong
135 * - PKT_RX_IP_CKSUM_GOOD: the IP checksum in the packet is valid
136 * - PKT_RX_IP_CKSUM_NONE: the IP checksum is not correct in the packet
137 * data, but the integrity of the IP header is verified.
139 #define PKT_RX_IP_CKSUM_MASK ((1ULL << 4) | (1ULL << 7))
141 #define PKT_RX_IP_CKSUM_UNKNOWN 0
142 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
143 #define PKT_RX_IP_CKSUM_GOOD (1ULL << 7)
144 #define PKT_RX_IP_CKSUM_NONE ((1ULL << 4) | (1ULL << 7))
147 * Mask of bits used to determine the status of RX L4 checksum.
148 * - PKT_RX_L4_CKSUM_UNKNOWN: no information about the RX L4 checksum
149 * - PKT_RX_L4_CKSUM_BAD: the L4 checksum in the packet is wrong
150 * - PKT_RX_L4_CKSUM_GOOD: the L4 checksum in the packet is valid
151 * - PKT_RX_L4_CKSUM_NONE: the L4 checksum is not correct in the packet
152 * data, but the integrity of the L4 data is verified.
154 #define PKT_RX_L4_CKSUM_MASK ((1ULL << 3) | (1ULL << 8))
156 #define PKT_RX_L4_CKSUM_UNKNOWN 0
157 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
158 #define PKT_RX_L4_CKSUM_GOOD (1ULL << 8)
159 #define PKT_RX_L4_CKSUM_NONE ((1ULL << 3) | (1ULL << 8))
161 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
162 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
163 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
164 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
167 * The 2 vlans have been stripped by the hardware and their tci are
168 * saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
169 * This can only happen if vlan stripping is enabled in the RX
170 * configuration of the PMD. If this flag is set,
171 * When PKT_RX_QINQ_STRIPPED is set, the flags (PKT_RX_VLAN |
172 * PKT_RX_VLAN_STRIPPED | PKT_RX_QINQ) must also be set.
174 #define PKT_RX_QINQ_STRIPPED (1ULL << 15)
177 * When packets are coalesced by a hardware or virtual driver, this flag
178 * can be set in the RX mbuf, meaning that the m->tso_segsz field is
179 * valid and is set to the segment size of original packets.
181 #define PKT_RX_LRO (1ULL << 16)
184 * Indicate that the timestamp field in the mbuf is valid.
186 #define PKT_RX_TIMESTAMP (1ULL << 17)
189 * Indicate that security offload processing was applied on the RX packet.
191 #define PKT_RX_SEC_OFFLOAD (1ULL << 18)
194 * Indicate that security offload processing failed on the RX packet.
196 #define PKT_RX_SEC_OFFLOAD_FAILED (1ULL << 19)
199 * The RX packet is a double VLAN, and the outer tci has been
200 * saved in in mbuf->vlan_tci_outer.
201 * If the flag PKT_RX_QINQ_STRIPPED is also present, both VLANs
202 * headers have been stripped from mbuf data, else they are still
205 #define PKT_RX_QINQ (1ULL << 20)
207 /* add new RX flags here */
209 /* add new TX flags here */
212 * Request security offload processing on the TX packet.
214 #define PKT_TX_SEC_OFFLOAD (1ULL << 43)
217 * Offload the MACsec. This flag must be set by the application to enable
218 * this offload feature for a packet to be transmitted.
220 #define PKT_TX_MACSEC (1ULL << 44)
223 * Bits 45:48 used for the tunnel type.
224 * When doing Tx offload like TSO or checksum, the HW needs to configure the
225 * tunnel type into the HW descriptors.
227 #define PKT_TX_TUNNEL_VXLAN (0x1ULL << 45)
228 #define PKT_TX_TUNNEL_GRE (0x2ULL << 45)
229 #define PKT_TX_TUNNEL_IPIP (0x3ULL << 45)
230 #define PKT_TX_TUNNEL_GENEVE (0x4ULL << 45)
231 /**< TX packet with MPLS-in-UDP RFC 7510 header. */
232 #define PKT_TX_TUNNEL_MPLSINUDP (0x5ULL << 45)
233 /* add new TX TUNNEL type here */
234 #define PKT_TX_TUNNEL_MASK (0xFULL << 45)
237 * Second VLAN insertion (QinQ) flag.
239 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
242 * TCP segmentation offload. To enable this offload feature for a
243 * packet to be transmitted on hardware supporting TSO:
244 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
246 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
247 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
249 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
250 * - calculate the pseudo header checksum without taking ip_len in account,
251 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
252 * rte_ipv6_phdr_cksum() that can be used as helpers.
254 #define PKT_TX_TCP_SEG (1ULL << 50)
256 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
259 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
260 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
261 * L4 checksum offload, the user needs to:
262 * - fill l2_len and l3_len in mbuf
263 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
264 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
265 * - calculate the pseudo header checksum and set it in the L4 header (only
266 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
267 * For SCTP, set the crc field to 0.
269 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
270 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
271 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
272 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
273 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
276 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
277 * also be set by the application, although a PMD will only check
279 * - set the IP checksum field in the packet to 0
280 * - fill the mbuf offload information: l2_len, l3_len
282 #define PKT_TX_IP_CKSUM (1ULL << 54)
285 * Packet is IPv4. This flag must be set when using any offload feature
286 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
287 * packet. If the packet is a tunneled packet, this flag is related to
290 #define PKT_TX_IPV4 (1ULL << 55)
293 * Packet is IPv6. This flag must be set when using an offload feature
294 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
295 * packet. If the packet is a tunneled packet, this flag is related to
298 #define PKT_TX_IPV6 (1ULL << 56)
300 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
303 * Offload the IP checksum of an external header in the hardware. The
304 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
305 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
306 * packet must be set to 0.
307 * - set the outer IP checksum field in the packet to 0
308 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
310 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
313 * Packet outer header is IPv4. This flag must be set when using any
314 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
315 * outer header of the tunneled packet is an IPv4 packet.
317 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
320 * Packet outer header is IPv6. This flag must be set when using any
321 * outer offload feature (L4 checksum) to tell the NIC that the outer
322 * header of the tunneled packet is an IPv6 packet.
324 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
327 * Bitmask of all supported packet Tx offload features flags,
328 * which can be set for packet.
330 #define PKT_TX_OFFLOAD_MASK ( \
333 PKT_TX_OUTER_IP_CKSUM | \
335 PKT_TX_IEEE1588_TMST | \
338 PKT_TX_TUNNEL_MASK | \
342 #define __RESERVED (1ULL << 61) /**< reserved for future mbuf use */
344 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
346 /* Use final bit of flags to indicate a control mbuf */
347 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
349 /** Alignment constraint of mbuf private area. */
350 #define RTE_MBUF_PRIV_ALIGN 8
353 * Get the name of a RX offload flag
356 * The mask describing the flag.
358 * The name of this flag, or NULL if it's not a valid RX flag.
360 const char *rte_get_rx_ol_flag_name(uint64_t mask);
363 * Dump the list of RX offload flags in a buffer
366 * The mask describing the RX flags.
370 * The length of the buffer.
372 * 0 on success, (-1) on error.
374 int rte_get_rx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
377 * Get the name of a TX offload flag
380 * The mask describing the flag. Usually only one bit must be set.
381 * Several bits can be given if they belong to the same mask.
382 * Ex: PKT_TX_L4_MASK.
384 * The name of this flag, or NULL if it's not a valid TX flag.
386 const char *rte_get_tx_ol_flag_name(uint64_t mask);
389 * Dump the list of TX offload flags in a buffer
392 * The mask describing the TX flags.
396 * The length of the buffer.
398 * 0 on success, (-1) on error.
400 int rte_get_tx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
403 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
404 * splitting it into multiple segments.
405 * So, for mbufs that planned to be involved into RX/TX, the recommended
406 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
408 #define RTE_MBUF_DEFAULT_DATAROOM 2048
409 #define RTE_MBUF_DEFAULT_BUF_SIZE \
410 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
412 /* define a set of marker types that can be used to refer to set points in the
415 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
417 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
419 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
420 * with a single assignment */
423 * The generic rte_mbuf, containing a packet mbuf.
428 void *buf_addr; /**< Virtual address of segment buffer. */
430 * Physical address of segment buffer.
431 * Force alignment to 8-bytes, so as to ensure we have the exact
432 * same mbuf cacheline0 layout for 32-bit and 64-bit. This makes
433 * working on vector drivers easier.
438 rte_iova_t buf_physaddr; /**< deprecated */
439 } __rte_aligned(sizeof(rte_iova_t));
441 /* next 8 bytes are initialised on RX descriptor rearm */
446 * Reference counter. Its size should at least equal to the size
447 * of port field (16 bits), to support zero-copy broadcast.
448 * It should only be accessed using the following functions:
449 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
450 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
451 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
456 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
457 uint16_t refcnt; /**< Non-atomically accessed refcnt */
459 uint16_t nb_segs; /**< Number of segments. */
461 /** Input port (16 bits to support more than 256 virtual ports). */
464 uint64_t ol_flags; /**< Offload features. */
466 /* remaining bytes are set on RX when pulling packet from descriptor */
467 MARKER rx_descriptor_fields1;
470 * The packet type, which is the combination of outer/inner L2, L3, L4
471 * and tunnel types. The packet_type is about data really present in the
472 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
473 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
474 * vlan is stripped from the data.
478 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
480 uint32_t l2_type:4; /**< (Outer) L2 type. */
481 uint32_t l3_type:4; /**< (Outer) L3 type. */
482 uint32_t l4_type:4; /**< (Outer) L4 type. */
483 uint32_t tun_type:4; /**< Tunnel type. */
486 uint8_t inner_esp_next_proto;
487 /**< ESP next protocol type, valid if
488 * RTE_PTYPE_TUNNEL_ESP tunnel type is set
493 uint8_t inner_l2_type:4;
494 /**< Inner L2 type. */
495 uint8_t inner_l3_type:4;
496 /**< Inner L3 type. */
499 uint32_t inner_l4_type:4; /**< Inner L4 type. */
503 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
504 uint16_t data_len; /**< Amount of data in segment buffer. */
505 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN_STRIPPED is set. */
509 uint32_t rss; /**< RSS hash result if RSS enabled */
518 /**< Second 4 flexible bytes */
521 /**< First 4 flexible bytes or FD ID, dependent on
522 PKT_RX_FDIR_* flag in ol_flags. */
523 } fdir; /**< Filter identifier if FDIR enabled */
527 } sched; /**< Hierarchical scheduler */
528 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
529 } hash; /**< hash information */
531 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ_STRIPPED is set. */
532 uint16_t vlan_tci_outer;
534 uint16_t buf_len; /**< Length of segment buffer. */
536 /** Valid if PKT_RX_TIMESTAMP is set. The unit and time reference
537 * are not normalized but are always the same for a given port.
541 /* second cache line - fields only used in slow path or on TX */
542 MARKER cacheline1 __rte_cache_min_aligned;
546 void *userdata; /**< Can be used for external metadata */
547 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
550 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
551 struct rte_mbuf *next; /**< Next segment of scattered packet. */
553 /* fields to support TX offloads */
556 uint64_t tx_offload; /**< combined for easy fetch */
560 /**< L2 (MAC) Header Length for non-tunneling pkt.
561 * Outer_L4_len + ... + Inner_L2_len for tunneling pkt.
563 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
564 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
565 uint64_t tso_segsz:16; /**< TCP TSO segment size */
567 /* fields for TX offloading of tunnels */
568 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
569 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
571 /* uint64_t unused:8; */
575 /** Size of the application private data. In case of an indirect
576 * mbuf, it stores the direct mbuf private data size. */
579 /** Timesync flags for use with IEEE1588. */
582 /** Sequence number. See also rte_reorder_insert(). */
585 } __rte_cache_aligned;
587 /**< Maximum number of nb_segs allowed. */
588 #define RTE_MBUF_MAX_NB_SEGS UINT16_MAX
591 * Prefetch the first part of the mbuf
593 * The first 64 bytes of the mbuf corresponds to fields that are used early
594 * in the receive path. If the cache line of the architecture is higher than
595 * 64B, the second part will also be prefetched.
598 * The pointer to the mbuf.
601 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
603 rte_prefetch0(&m->cacheline0);
607 * Prefetch the second part of the mbuf
609 * The next 64 bytes of the mbuf corresponds to fields that are used in the
610 * transmit path. If the cache line of the architecture is higher than 64B,
611 * this function does nothing as it is expected that the full mbuf is
615 * The pointer to the mbuf.
618 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
620 #if RTE_CACHE_LINE_SIZE == 64
621 rte_prefetch0(&m->cacheline1);
628 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
631 * Return the IO address of the beginning of the mbuf data
634 * The pointer to the mbuf.
636 * The IO address of the beginning of the mbuf data
638 static inline rte_iova_t
639 rte_mbuf_data_iova(const struct rte_mbuf *mb)
641 return mb->buf_iova + mb->data_off;
645 static inline phys_addr_t
646 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
648 return rte_mbuf_data_iova(mb);
652 * Return the default IO address of the beginning of the mbuf data
654 * This function is used by drivers in their receive function, as it
655 * returns the location where data should be written by the NIC, taking
656 * the default headroom in account.
659 * The pointer to the mbuf.
661 * The IO address of the beginning of the mbuf data
663 static inline rte_iova_t
664 rte_mbuf_data_iova_default(const struct rte_mbuf *mb)
666 return mb->buf_iova + RTE_PKTMBUF_HEADROOM;
670 static inline phys_addr_t
671 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
673 return rte_mbuf_data_iova_default(mb);
677 * Return the mbuf owning the data buffer address of an indirect mbuf.
680 * The pointer to the indirect mbuf.
682 * The address of the direct mbuf corresponding to buffer_addr.
684 static inline struct rte_mbuf *
685 rte_mbuf_from_indirect(struct rte_mbuf *mi)
687 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
691 * Return the buffer address embedded in the given mbuf.
694 * The pointer to the mbuf.
696 * The address of the data buffer owned by the mbuf.
699 rte_mbuf_to_baddr(struct rte_mbuf *md)
702 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
707 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
709 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
712 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
714 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
717 * Private data in case of pktmbuf pool.
719 * A structure that contains some pktmbuf_pool-specific data that are
720 * appended after the mempool structure (in private data).
722 struct rte_pktmbuf_pool_private {
723 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
724 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
727 #ifdef RTE_LIBRTE_MBUF_DEBUG
729 /** check mbuf type in debug mode */
730 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
732 #else /* RTE_LIBRTE_MBUF_DEBUG */
734 /** check mbuf type in debug mode */
735 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
737 #endif /* RTE_LIBRTE_MBUF_DEBUG */
739 #ifdef RTE_MBUF_REFCNT_ATOMIC
742 * Reads the value of an mbuf's refcnt.
746 * Reference count number.
748 static inline uint16_t
749 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
751 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
755 * Sets an mbuf's refcnt to a defined value.
762 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
764 rte_atomic16_set(&m->refcnt_atomic, new_value);
768 * Adds given value to an mbuf's refcnt and returns its new value.
772 * Value to add/subtract
776 static inline uint16_t
777 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
780 * The atomic_add is an expensive operation, so we don't want to
781 * call it in the case where we know we are the uniq holder of
782 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
783 * operation has to be used because concurrent accesses on the
784 * reference counter can occur.
786 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
787 rte_mbuf_refcnt_set(m, 1 + value);
791 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
794 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
797 * Adds given value to an mbuf's refcnt and returns its new value.
799 static inline uint16_t
800 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
802 m->refcnt = (uint16_t)(m->refcnt + value);
807 * Reads the value of an mbuf's refcnt.
809 static inline uint16_t
810 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
816 * Sets an mbuf's refcnt to the defined value.
819 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
821 m->refcnt = new_value;
824 #endif /* RTE_MBUF_REFCNT_ATOMIC */
827 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
834 * Sanity checks on an mbuf.
836 * Check the consistency of the given mbuf. The function will cause a
837 * panic if corruption is detected.
840 * The mbuf to be checked.
842 * True if the mbuf is a packet header, false if it is a sub-segment
843 * of a packet (in this case, some fields like nb_segs are not checked)
846 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
848 #define MBUF_RAW_ALLOC_CHECK(m) do { \
849 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1); \
850 RTE_ASSERT((m)->next == NULL); \
851 RTE_ASSERT((m)->nb_segs == 1); \
852 __rte_mbuf_sanity_check(m, 0); \
856 * Allocate an uninitialized mbuf from mempool *mp*.
858 * This function can be used by PMDs (especially in RX functions) to
859 * allocate an uninitialized mbuf. The driver is responsible of
860 * initializing all the required fields. See rte_pktmbuf_reset().
861 * For standard needs, prefer rte_pktmbuf_alloc().
863 * The caller can expect that the following fields of the mbuf structure
864 * are initialized: buf_addr, buf_iova, buf_len, refcnt=1, nb_segs=1,
865 * next=NULL, pool, priv_size. The other fields must be initialized
869 * The mempool from which mbuf is allocated.
871 * - The pointer to the new mbuf on success.
872 * - NULL if allocation failed.
874 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
879 if (rte_mempool_get(mp, &mb) < 0)
881 m = (struct rte_mbuf *)mb;
882 MBUF_RAW_ALLOC_CHECK(m);
887 * Put mbuf back into its original mempool.
889 * The caller must ensure that the mbuf is direct and properly
890 * reinitialized (refcnt=1, next=NULL, nb_segs=1), as done by
891 * rte_pktmbuf_prefree_seg().
893 * This function should be used with care, when optimization is
894 * required. For standard needs, prefer rte_pktmbuf_free() or
895 * rte_pktmbuf_free_seg().
898 * The mbuf to be freed.
900 static __rte_always_inline void
901 rte_mbuf_raw_free(struct rte_mbuf *m)
903 RTE_ASSERT(RTE_MBUF_DIRECT(m));
904 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
905 RTE_ASSERT(m->next == NULL);
906 RTE_ASSERT(m->nb_segs == 1);
907 __rte_mbuf_sanity_check(m, 0);
908 rte_mempool_put(m->pool, m);
911 /* compat with older versions */
914 __rte_mbuf_raw_free(struct rte_mbuf *m)
916 rte_mbuf_raw_free(m);
919 /* Operations on ctrl mbuf */
922 * The control mbuf constructor.
924 * This function initializes some fields in an mbuf structure that are
925 * not modified by the user once created (mbuf type, origin pool, buffer
926 * start address, and so on). This function is given as a callback function
927 * to rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
930 * The mempool from which the mbuf is allocated.
932 * A pointer that can be used by the user to retrieve useful information
933 * for mbuf initialization. This pointer is the opaque argument passed to
934 * rte_mempool_obj_iter() or rte_mempool_create().
936 * The mbuf to initialize.
938 * The index of the mbuf in the pool table.
940 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
941 void *m, unsigned i);
944 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
946 * This new mbuf is initialized with data pointing to the beginning of
947 * buffer, and with a length of zero.
950 * The mempool from which the mbuf is allocated.
952 * - The pointer to the new mbuf on success.
953 * - NULL if allocation failed.
955 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
958 * Free a control mbuf back into its original mempool.
961 * The control mbuf to be freed.
963 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
966 * A macro that returns the pointer to the carried data.
968 * The value that can be read or assigned.
973 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
976 * A macro that returns the length of the carried data.
978 * The value that can be read or assigned.
983 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
986 * Tests if an mbuf is a control mbuf
989 * The mbuf to be tested
991 * - True (1) if the mbuf is a control mbuf
992 * - False(0) otherwise
995 rte_is_ctrlmbuf(struct rte_mbuf *m)
997 return !!(m->ol_flags & CTRL_MBUF_FLAG);
1000 /* Operations on pkt mbuf */
1003 * The packet mbuf constructor.
1005 * This function initializes some fields in the mbuf structure that are
1006 * not modified by the user once created (origin pool, buffer start
1007 * address, and so on). This function is given as a callback function to
1008 * rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
1011 * The mempool from which mbufs originate.
1013 * A pointer that can be used by the user to retrieve useful information
1014 * for mbuf initialization. This pointer is the opaque argument passed to
1015 * rte_mempool_obj_iter() or rte_mempool_create().
1017 * The mbuf to initialize.
1019 * The index of the mbuf in the pool table.
1021 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1022 void *m, unsigned i);
1026 * A packet mbuf pool constructor.
1028 * This function initializes the mempool private data in the case of a
1029 * pktmbuf pool. This private data is needed by the driver. The
1030 * function must be called on the mempool before it is used, or it
1031 * can be given as a callback function to rte_mempool_create() at
1032 * pool creation. It can be extended by the user, for example, to
1033 * provide another packet size.
1036 * The mempool from which mbufs originate.
1038 * A pointer that can be used by the user to retrieve useful information
1039 * for mbuf initialization. This pointer is the opaque argument passed to
1040 * rte_mempool_create().
1042 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1045 * Create a mbuf pool.
1047 * This function creates and initializes a packet mbuf pool. It is
1048 * a wrapper to rte_mempool functions.
1051 * The name of the mbuf pool.
1053 * The number of elements in the mbuf pool. The optimum size (in terms
1054 * of memory usage) for a mempool is when n is a power of two minus one:
1057 * Size of the per-core object cache. See rte_mempool_create() for
1060 * Size of application private are between the rte_mbuf structure
1061 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1062 * @param data_room_size
1063 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1065 * The socket identifier where the memory should be allocated. The
1066 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1069 * The pointer to the new allocated mempool, on success. NULL on error
1070 * with rte_errno set appropriately. Possible rte_errno values include:
1071 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1072 * - E_RTE_SECONDARY - function was called from a secondary process instance
1073 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1074 * - ENOSPC - the maximum number of memzones has already been allocated
1075 * - EEXIST - a memzone with the same name already exists
1076 * - ENOMEM - no appropriate memory area found in which to create memzone
1078 struct rte_mempool *
1079 rte_pktmbuf_pool_create(const char *name, unsigned n,
1080 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1084 * Get the data room size of mbufs stored in a pktmbuf_pool
1086 * The data room size is the amount of data that can be stored in a
1087 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1090 * The packet mbuf pool.
1092 * The data room size of mbufs stored in this mempool.
1094 static inline uint16_t
1095 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1097 struct rte_pktmbuf_pool_private *mbp_priv;
1099 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1100 return mbp_priv->mbuf_data_room_size;
1104 * Get the application private size of mbufs stored in a pktmbuf_pool
1106 * The private size of mbuf is a zone located between the rte_mbuf
1107 * structure and the data buffer where an application can store data
1108 * associated to a packet.
1111 * The packet mbuf pool.
1113 * The private size of mbufs stored in this mempool.
1115 static inline uint16_t
1116 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1118 struct rte_pktmbuf_pool_private *mbp_priv;
1120 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1121 return mbp_priv->mbuf_priv_size;
1125 * Reset the data_off field of a packet mbuf to its default value.
1127 * The given mbuf must have only one segment, which should be empty.
1130 * The packet mbuf's data_off field has to be reset.
1132 static inline void rte_pktmbuf_reset_headroom(struct rte_mbuf *m)
1134 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1138 * Reset the fields of a packet mbuf to their default values.
1140 * The given mbuf must have only one segment.
1143 * The packet mbuf to be resetted.
1145 #define MBUF_INVALID_PORT UINT16_MAX
1147 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1153 m->vlan_tci_outer = 0;
1155 m->port = MBUF_INVALID_PORT;
1159 rte_pktmbuf_reset_headroom(m);
1162 __rte_mbuf_sanity_check(m, 1);
1166 * Allocate a new mbuf from a mempool.
1168 * This new mbuf contains one segment, which has a length of 0. The pointer
1169 * to data is initialized to have some bytes of headroom in the buffer
1170 * (if buffer size allows).
1173 * The mempool from which the mbuf is allocated.
1175 * - The pointer to the new mbuf on success.
1176 * - NULL if allocation failed.
1178 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1181 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1182 rte_pktmbuf_reset(m);
1187 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1191 * The mempool from which mbufs are allocated.
1193 * Array of pointers to mbufs
1198 * - -ENOENT: Not enough entries in the mempool; no mbufs are retrieved.
1200 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1201 struct rte_mbuf **mbufs, unsigned count)
1206 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1210 /* To understand duff's device on loop unwinding optimization, see
1211 * https://en.wikipedia.org/wiki/Duff's_device.
1212 * Here while() loop is used rather than do() while{} to avoid extra
1213 * check if count is zero.
1215 switch (count % 4) {
1217 while (idx != count) {
1218 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1219 rte_pktmbuf_reset(mbufs[idx]);
1223 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1224 rte_pktmbuf_reset(mbufs[idx]);
1228 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1229 rte_pktmbuf_reset(mbufs[idx]);
1233 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1234 rte_pktmbuf_reset(mbufs[idx]);
1243 * Attach packet mbuf to another packet mbuf.
1245 * After attachment we refer the mbuf we attached as 'indirect',
1246 * while mbuf we attached to as 'direct'.
1247 * The direct mbuf's reference counter is incremented.
1249 * Right now, not supported:
1250 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1251 * - mbuf we trying to attach (mi) is used by someone else
1252 * e.g. it's reference counter is greater then 1.
1255 * The indirect packet mbuf.
1257 * The packet mbuf we're attaching to.
1259 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1261 struct rte_mbuf *md;
1263 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1264 rte_mbuf_refcnt_read(mi) == 1);
1266 /* if m is not direct, get the mbuf that embeds the data */
1267 if (RTE_MBUF_DIRECT(m))
1270 md = rte_mbuf_from_indirect(m);
1272 rte_mbuf_refcnt_update(md, 1);
1273 mi->priv_size = m->priv_size;
1274 mi->buf_iova = m->buf_iova;
1275 mi->buf_addr = m->buf_addr;
1276 mi->buf_len = m->buf_len;
1278 mi->data_off = m->data_off;
1279 mi->data_len = m->data_len;
1281 mi->vlan_tci = m->vlan_tci;
1282 mi->vlan_tci_outer = m->vlan_tci_outer;
1283 mi->tx_offload = m->tx_offload;
1287 mi->pkt_len = mi->data_len;
1289 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1290 mi->packet_type = m->packet_type;
1291 mi->timestamp = m->timestamp;
1293 __rte_mbuf_sanity_check(mi, 1);
1294 __rte_mbuf_sanity_check(m, 0);
1298 * Detach an indirect packet mbuf.
1300 * - restore original mbuf address and length values.
1301 * - reset pktmbuf data and data_len to their default values.
1302 * - decrement the direct mbuf's reference counter. When the
1303 * reference counter becomes 0, the direct mbuf is freed.
1305 * All other fields of the given packet mbuf will be left intact.
1308 * The indirect attached packet mbuf.
1310 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1312 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1313 struct rte_mempool *mp = m->pool;
1314 uint32_t mbuf_size, buf_len, priv_size;
1316 priv_size = rte_pktmbuf_priv_size(mp);
1317 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1318 buf_len = rte_pktmbuf_data_room_size(mp);
1320 m->priv_size = priv_size;
1321 m->buf_addr = (char *)m + mbuf_size;
1322 m->buf_iova = rte_mempool_virt2iova(m) + mbuf_size;
1323 m->buf_len = (uint16_t)buf_len;
1324 rte_pktmbuf_reset_headroom(m);
1328 if (rte_mbuf_refcnt_update(md, -1) == 0) {
1331 rte_mbuf_refcnt_set(md, 1);
1332 rte_mbuf_raw_free(md);
1337 * Decrease reference counter and unlink a mbuf segment
1339 * This function does the same than a free, except that it does not
1340 * return the segment to its pool.
1341 * It decreases the reference counter, and if it reaches 0, it is
1342 * detached from its parent for an indirect mbuf.
1345 * The mbuf to be unlinked
1347 * - (m) if it is the last reference. It can be recycled or freed.
1348 * - (NULL) if the mbuf still has remaining references on it.
1350 static __rte_always_inline struct rte_mbuf *
1351 rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1353 __rte_mbuf_sanity_check(m, 0);
1355 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1357 if (RTE_MBUF_INDIRECT(m))
1358 rte_pktmbuf_detach(m);
1360 if (m->next != NULL) {
1367 } else if (rte_atomic16_add_return(&m->refcnt_atomic, -1) == 0) {
1370 if (RTE_MBUF_INDIRECT(m))
1371 rte_pktmbuf_detach(m);
1373 if (m->next != NULL) {
1377 rte_mbuf_refcnt_set(m, 1);
1384 /* deprecated, replaced by rte_pktmbuf_prefree_seg() */
1386 static inline struct rte_mbuf *
1387 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1389 return rte_pktmbuf_prefree_seg(m);
1393 * Free a segment of a packet mbuf into its original mempool.
1395 * Free an mbuf, without parsing other segments in case of chained
1399 * The packet mbuf segment to be freed.
1401 static __rte_always_inline void
1402 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1404 m = rte_pktmbuf_prefree_seg(m);
1405 if (likely(m != NULL))
1406 rte_mbuf_raw_free(m);
1410 * Free a packet mbuf back into its original mempool.
1412 * Free an mbuf, and all its segments in case of chained buffers. Each
1413 * segment is added back into its original mempool.
1416 * The packet mbuf to be freed.
1418 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1420 struct rte_mbuf *m_next;
1422 __rte_mbuf_sanity_check(m, 1);
1426 rte_pktmbuf_free_seg(m);
1432 * Creates a "clone" of the given packet mbuf.
1434 * Walks through all segments of the given packet mbuf, and for each of them:
1435 * - Creates a new packet mbuf from the given pool.
1436 * - Attaches newly created mbuf to the segment.
1437 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1438 * from the original packet mbuf.
1441 * The packet mbuf to be cloned.
1443 * The mempool from which the "clone" mbufs are allocated.
1445 * - The pointer to the new "clone" mbuf on success.
1446 * - NULL if allocation fails.
1448 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1449 struct rte_mempool *mp)
1451 struct rte_mbuf *mc, *mi, **prev;
1455 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1460 pktlen = md->pkt_len;
1465 rte_pktmbuf_attach(mi, md);
1468 } while ((md = md->next) != NULL &&
1469 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1473 mc->pkt_len = pktlen;
1475 /* Allocation of new indirect segment failed */
1476 if (unlikely (mi == NULL)) {
1477 rte_pktmbuf_free(mc);
1481 __rte_mbuf_sanity_check(mc, 1);
1486 * Adds given value to the refcnt of all packet mbuf segments.
1488 * Walks through all segments of given packet mbuf and for each of them
1489 * invokes rte_mbuf_refcnt_update().
1492 * The packet mbuf whose refcnt to be updated.
1494 * The value to add to the mbuf's segments refcnt.
1496 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1498 __rte_mbuf_sanity_check(m, 1);
1501 rte_mbuf_refcnt_update(m, v);
1502 } while ((m = m->next) != NULL);
1506 * Get the headroom in a packet mbuf.
1511 * The length of the headroom.
1513 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1515 __rte_mbuf_sanity_check(m, 0);
1520 * Get the tailroom of a packet mbuf.
1525 * The length of the tailroom.
1527 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1529 __rte_mbuf_sanity_check(m, 0);
1530 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1535 * Get the last segment of the packet.
1540 * The last segment of the given mbuf.
1542 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1544 __rte_mbuf_sanity_check(m, 1);
1545 while (m->next != NULL)
1551 * A macro that points to an offset into the data in the mbuf.
1553 * The returned pointer is cast to type t. Before using this
1554 * function, the user must ensure that the first segment is large
1555 * enough to accommodate its data.
1560 * The offset into the mbuf data.
1562 * The type to cast the result into.
1564 #define rte_pktmbuf_mtod_offset(m, t, o) \
1565 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1568 * A macro that points to the start of the data in the mbuf.
1570 * The returned pointer is cast to type t. Before using this
1571 * function, the user must ensure that the first segment is large
1572 * enough to accommodate its data.
1577 * The type to cast the result into.
1579 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1582 * A macro that returns the IO address that points to an offset of the
1583 * start of the data in the mbuf
1588 * The offset into the data to calculate address from.
1590 #define rte_pktmbuf_iova_offset(m, o) \
1591 (rte_iova_t)((m)->buf_iova + (m)->data_off + (o))
1594 #define rte_pktmbuf_mtophys_offset(m, o) \
1595 rte_pktmbuf_iova_offset(m, o)
1598 * A macro that returns the IO address that points to the start of the
1604 #define rte_pktmbuf_iova(m) rte_pktmbuf_iova_offset(m, 0)
1607 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_iova(m)
1610 * A macro that returns the length of the packet.
1612 * The value can be read or assigned.
1617 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1620 * A macro that returns the length of the segment.
1622 * The value can be read or assigned.
1627 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1630 * Prepend len bytes to an mbuf data area.
1632 * Returns a pointer to the new
1633 * data start address. If there is not enough headroom in the first
1634 * segment, the function will return NULL, without modifying the mbuf.
1639 * The amount of data to prepend (in bytes).
1641 * A pointer to the start of the newly prepended data, or
1642 * NULL if there is not enough headroom space in the first segment
1644 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1647 __rte_mbuf_sanity_check(m, 1);
1649 if (unlikely(len > rte_pktmbuf_headroom(m)))
1653 m->data_len = (uint16_t)(m->data_len + len);
1654 m->pkt_len = (m->pkt_len + len);
1656 return (char *)m->buf_addr + m->data_off;
1660 * Append len bytes to an mbuf.
1662 * Append len bytes to an mbuf and return a pointer to the start address
1663 * of the added data. If there is not enough tailroom in the last
1664 * segment, the function will return NULL, without modifying the mbuf.
1669 * The amount of data to append (in bytes).
1671 * A pointer to the start of the newly appended data, or
1672 * NULL if there is not enough tailroom space in the last segment
1674 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1677 struct rte_mbuf *m_last;
1679 __rte_mbuf_sanity_check(m, 1);
1681 m_last = rte_pktmbuf_lastseg(m);
1682 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1685 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1686 m_last->data_len = (uint16_t)(m_last->data_len + len);
1687 m->pkt_len = (m->pkt_len + len);
1688 return (char*) tail;
1692 * Remove len bytes at the beginning of an mbuf.
1694 * Returns a pointer to the start address of the new data area. If the
1695 * length is greater than the length of the first segment, then the
1696 * function will fail and return NULL, without modifying the mbuf.
1701 * The amount of data to remove (in bytes).
1703 * A pointer to the new start of the data.
1705 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1707 __rte_mbuf_sanity_check(m, 1);
1709 if (unlikely(len > m->data_len))
1712 m->data_len = (uint16_t)(m->data_len - len);
1714 m->pkt_len = (m->pkt_len - len);
1715 return (char *)m->buf_addr + m->data_off;
1719 * Remove len bytes of data at the end of the mbuf.
1721 * If the length is greater than the length of the last segment, the
1722 * function will fail and return -1 without modifying the mbuf.
1727 * The amount of data to remove (in bytes).
1732 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1734 struct rte_mbuf *m_last;
1736 __rte_mbuf_sanity_check(m, 1);
1738 m_last = rte_pktmbuf_lastseg(m);
1739 if (unlikely(len > m_last->data_len))
1742 m_last->data_len = (uint16_t)(m_last->data_len - len);
1743 m->pkt_len = (m->pkt_len - len);
1748 * Test if mbuf data is contiguous.
1753 * - 1, if all data is contiguous (one segment).
1754 * - 0, if there is several segments.
1756 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1758 __rte_mbuf_sanity_check(m, 1);
1759 return !!(m->nb_segs == 1);
1763 * @internal used by rte_pktmbuf_read().
1765 const void *__rte_pktmbuf_read(const struct rte_mbuf *m, uint32_t off,
1766 uint32_t len, void *buf);
1769 * Read len data bytes in a mbuf at specified offset.
1771 * If the data is contiguous, return the pointer in the mbuf data, else
1772 * copy the data in the buffer provided by the user and return its
1776 * The pointer to the mbuf.
1778 * The offset of the data in the mbuf.
1780 * The amount of bytes to read.
1782 * The buffer where data is copied if it is not contiguous in mbuf
1783 * data. Its length should be at least equal to the len parameter.
1785 * The pointer to the data, either in the mbuf if it is contiguous,
1786 * or in the user buffer. If mbuf is too small, NULL is returned.
1788 static inline const void *rte_pktmbuf_read(const struct rte_mbuf *m,
1789 uint32_t off, uint32_t len, void *buf)
1791 if (likely(off + len <= rte_pktmbuf_data_len(m)))
1792 return rte_pktmbuf_mtod_offset(m, char *, off);
1794 return __rte_pktmbuf_read(m, off, len, buf);
1798 * Chain an mbuf to another, thereby creating a segmented packet.
1800 * Note: The implementation will do a linear walk over the segments to find
1801 * the tail entry. For cases when there are many segments, it's better to
1802 * chain the entries manually.
1805 * The head of the mbuf chain (the first packet)
1807 * The mbuf to put last in the chain
1811 * - -EOVERFLOW, if the chain segment limit exceeded
1813 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1815 struct rte_mbuf *cur_tail;
1817 /* Check for number-of-segments-overflow */
1818 if (head->nb_segs + tail->nb_segs > RTE_MBUF_MAX_NB_SEGS)
1821 /* Chain 'tail' onto the old tail */
1822 cur_tail = rte_pktmbuf_lastseg(head);
1823 cur_tail->next = tail;
1825 /* accumulate number of segments and total length. */
1826 head->nb_segs += tail->nb_segs;
1827 head->pkt_len += tail->pkt_len;
1829 /* pkt_len is only set in the head */
1830 tail->pkt_len = tail->data_len;
1836 * Validate general requirements for Tx offload in mbuf.
1838 * This function checks correctness and completeness of Tx offload settings.
1841 * The packet mbuf to be validated.
1843 * 0 if packet is valid
1846 rte_validate_tx_offload(const struct rte_mbuf *m)
1848 uint64_t ol_flags = m->ol_flags;
1849 uint64_t inner_l3_offset = m->l2_len;
1851 /* Does packet set any of available offloads? */
1852 if (!(ol_flags & PKT_TX_OFFLOAD_MASK))
1855 if (ol_flags & PKT_TX_OUTER_IP_CKSUM)
1856 inner_l3_offset += m->outer_l2_len + m->outer_l3_len;
1858 /* Headers are fragmented */
1859 if (rte_pktmbuf_data_len(m) < inner_l3_offset + m->l3_len + m->l4_len)
1862 /* IP checksum can be counted only for IPv4 packet */
1863 if ((ol_flags & PKT_TX_IP_CKSUM) && (ol_flags & PKT_TX_IPV6))
1866 /* IP type not set when required */
1867 if (ol_flags & (PKT_TX_L4_MASK | PKT_TX_TCP_SEG))
1868 if (!(ol_flags & (PKT_TX_IPV4 | PKT_TX_IPV6)))
1871 /* Check requirements for TSO packet */
1872 if (ol_flags & PKT_TX_TCP_SEG)
1873 if ((m->tso_segsz == 0) ||
1874 ((ol_flags & PKT_TX_IPV4) &&
1875 !(ol_flags & PKT_TX_IP_CKSUM)))
1878 /* PKT_TX_OUTER_IP_CKSUM set for non outer IPv4 packet. */
1879 if ((ol_flags & PKT_TX_OUTER_IP_CKSUM) &&
1880 !(ol_flags & PKT_TX_OUTER_IPV4))
1887 * Linearize data in mbuf.
1889 * This function moves the mbuf data in the first segment if there is enough
1890 * tailroom. The subsequent segments are unchained and freed.
1899 rte_pktmbuf_linearize(struct rte_mbuf *mbuf)
1901 int seg_len, copy_len;
1903 struct rte_mbuf *m_next;
1906 if (rte_pktmbuf_is_contiguous(mbuf))
1909 /* Extend first segment to the total packet length */
1910 copy_len = rte_pktmbuf_pkt_len(mbuf) - rte_pktmbuf_data_len(mbuf);
1912 if (unlikely(copy_len > rte_pktmbuf_tailroom(mbuf)))
1915 buffer = rte_pktmbuf_mtod_offset(mbuf, char *, mbuf->data_len);
1916 mbuf->data_len = (uint16_t)(mbuf->pkt_len);
1918 /* Append data from next segments to the first one */
1923 seg_len = rte_pktmbuf_data_len(m);
1924 rte_memcpy(buffer, rte_pktmbuf_mtod(m, char *), seg_len);
1927 rte_pktmbuf_free_seg(m);
1938 * Dump an mbuf structure to a file.
1940 * Dump all fields for the given packet mbuf and all its associated
1941 * segments (in the case of a chained buffer).
1944 * A pointer to a file for output
1948 * If dump_len != 0, also dump the "dump_len" first data bytes of
1951 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1957 #endif /* _RTE_MBUF_H_ */