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;
588 * Prefetch the first part of the mbuf
590 * The first 64 bytes of the mbuf corresponds to fields that are used early
591 * in the receive path. If the cache line of the architecture is higher than
592 * 64B, the second part will also be prefetched.
595 * The pointer to the mbuf.
598 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
600 rte_prefetch0(&m->cacheline0);
604 * Prefetch the second part of the mbuf
606 * The next 64 bytes of the mbuf corresponds to fields that are used in the
607 * transmit path. If the cache line of the architecture is higher than 64B,
608 * this function does nothing as it is expected that the full mbuf is
612 * The pointer to the mbuf.
615 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
617 #if RTE_CACHE_LINE_SIZE == 64
618 rte_prefetch0(&m->cacheline1);
625 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
628 * Return the IO address of the beginning of the mbuf data
631 * The pointer to the mbuf.
633 * The IO address of the beginning of the mbuf data
635 static inline rte_iova_t
636 rte_mbuf_data_iova(const struct rte_mbuf *mb)
638 return mb->buf_iova + mb->data_off;
642 static inline phys_addr_t
643 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
645 return rte_mbuf_data_iova(mb);
649 * Return the default IO address of the beginning of the mbuf data
651 * This function is used by drivers in their receive function, as it
652 * returns the location where data should be written by the NIC, taking
653 * the default headroom in account.
656 * The pointer to the mbuf.
658 * The IO address of the beginning of the mbuf data
660 static inline rte_iova_t
661 rte_mbuf_data_iova_default(const struct rte_mbuf *mb)
663 return mb->buf_iova + RTE_PKTMBUF_HEADROOM;
667 static inline phys_addr_t
668 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
670 return rte_mbuf_data_iova_default(mb);
674 * Return the mbuf owning the data buffer address of an indirect mbuf.
677 * The pointer to the indirect mbuf.
679 * The address of the direct mbuf corresponding to buffer_addr.
681 static inline struct rte_mbuf *
682 rte_mbuf_from_indirect(struct rte_mbuf *mi)
684 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
688 * Return the buffer address embedded in the given mbuf.
691 * The pointer to the mbuf.
693 * The address of the data buffer owned by the mbuf.
696 rte_mbuf_to_baddr(struct rte_mbuf *md)
699 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
704 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
706 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
709 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
711 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
714 * Private data in case of pktmbuf pool.
716 * A structure that contains some pktmbuf_pool-specific data that are
717 * appended after the mempool structure (in private data).
719 struct rte_pktmbuf_pool_private {
720 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
721 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
724 #ifdef RTE_LIBRTE_MBUF_DEBUG
726 /** check mbuf type in debug mode */
727 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
729 #else /* RTE_LIBRTE_MBUF_DEBUG */
731 /** check mbuf type in debug mode */
732 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
734 #endif /* RTE_LIBRTE_MBUF_DEBUG */
736 #ifdef RTE_MBUF_REFCNT_ATOMIC
739 * Reads the value of an mbuf's refcnt.
743 * Reference count number.
745 static inline uint16_t
746 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
748 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
752 * Sets an mbuf's refcnt to a defined value.
759 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
761 rte_atomic16_set(&m->refcnt_atomic, new_value);
765 * Adds given value to an mbuf's refcnt and returns its new value.
769 * Value to add/subtract
773 static inline uint16_t
774 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
777 * The atomic_add is an expensive operation, so we don't want to
778 * call it in the case where we know we are the uniq holder of
779 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
780 * operation has to be used because concurrent accesses on the
781 * reference counter can occur.
783 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
784 rte_mbuf_refcnt_set(m, 1 + value);
788 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
791 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
794 * Adds given value to an mbuf's refcnt and returns its new value.
796 static inline uint16_t
797 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
799 m->refcnt = (uint16_t)(m->refcnt + value);
804 * Reads the value of an mbuf's refcnt.
806 static inline uint16_t
807 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
813 * Sets an mbuf's refcnt to the defined value.
816 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
818 m->refcnt = new_value;
821 #endif /* RTE_MBUF_REFCNT_ATOMIC */
824 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
831 * Sanity checks on an mbuf.
833 * Check the consistency of the given mbuf. The function will cause a
834 * panic if corruption is detected.
837 * The mbuf to be checked.
839 * True if the mbuf is a packet header, false if it is a sub-segment
840 * of a packet (in this case, some fields like nb_segs are not checked)
843 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
845 #define MBUF_RAW_ALLOC_CHECK(m) do { \
846 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1); \
847 RTE_ASSERT((m)->next == NULL); \
848 RTE_ASSERT((m)->nb_segs == 1); \
849 __rte_mbuf_sanity_check(m, 0); \
853 * Allocate an unitialized mbuf from mempool *mp*.
855 * This function can be used by PMDs (especially in RX functions) to
856 * allocate an unitialized mbuf. The driver is responsible of
857 * initializing all the required fields. See rte_pktmbuf_reset().
858 * For standard needs, prefer rte_pktmbuf_alloc().
860 * The caller can expect that the following fields of the mbuf structure
861 * are initialized: buf_addr, buf_iova, buf_len, refcnt=1, nb_segs=1,
862 * next=NULL, pool, priv_size. The other fields must be initialized
866 * The mempool from which mbuf is allocated.
868 * - The pointer to the new mbuf on success.
869 * - NULL if allocation failed.
871 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
876 if (rte_mempool_get(mp, &mb) < 0)
878 m = (struct rte_mbuf *)mb;
879 MBUF_RAW_ALLOC_CHECK(m);
884 * Put mbuf back into its original mempool.
886 * The caller must ensure that the mbuf is direct and properly
887 * reinitialized (refcnt=1, next=NULL, nb_segs=1), as done by
888 * rte_pktmbuf_prefree_seg().
890 * This function should be used with care, when optimization is
891 * required. For standard needs, prefer rte_pktmbuf_free() or
892 * rte_pktmbuf_free_seg().
895 * The mbuf to be freed.
897 static __rte_always_inline void
898 rte_mbuf_raw_free(struct rte_mbuf *m)
900 RTE_ASSERT(RTE_MBUF_DIRECT(m));
901 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
902 RTE_ASSERT(m->next == NULL);
903 RTE_ASSERT(m->nb_segs == 1);
904 __rte_mbuf_sanity_check(m, 0);
905 rte_mempool_put(m->pool, m);
908 /* compat with older versions */
911 __rte_mbuf_raw_free(struct rte_mbuf *m)
913 rte_mbuf_raw_free(m);
916 /* Operations on ctrl mbuf */
919 * The control mbuf constructor.
921 * This function initializes some fields in an mbuf structure that are
922 * not modified by the user once created (mbuf type, origin pool, buffer
923 * start address, and so on). This function is given as a callback function
924 * to rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
927 * The mempool from which the mbuf is allocated.
929 * A pointer that can be used by the user to retrieve useful information
930 * for mbuf initialization. This pointer is the opaque argument passed to
931 * rte_mempool_obj_iter() or rte_mempool_create().
933 * The mbuf to initialize.
935 * The index of the mbuf in the pool table.
937 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
938 void *m, unsigned i);
941 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
943 * This new mbuf is initialized with data pointing to the beginning of
944 * buffer, and with a length of zero.
947 * The mempool from which the mbuf is allocated.
949 * - The pointer to the new mbuf on success.
950 * - NULL if allocation failed.
952 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
955 * Free a control mbuf back into its original mempool.
958 * The control mbuf to be freed.
960 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
963 * A macro that returns the pointer to the carried data.
965 * The value that can be read or assigned.
970 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
973 * A macro that returns the length of the carried data.
975 * The value that can be read or assigned.
980 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
983 * Tests if an mbuf is a control mbuf
986 * The mbuf to be tested
988 * - True (1) if the mbuf is a control mbuf
989 * - False(0) otherwise
992 rte_is_ctrlmbuf(struct rte_mbuf *m)
994 return !!(m->ol_flags & CTRL_MBUF_FLAG);
997 /* Operations on pkt mbuf */
1000 * The packet mbuf constructor.
1002 * This function initializes some fields in the mbuf structure that are
1003 * not modified by the user once created (origin pool, buffer start
1004 * address, and so on). This function is given as a callback function to
1005 * rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
1008 * The mempool from which mbufs originate.
1010 * A pointer that can be used by the user to retrieve useful information
1011 * for mbuf initialization. This pointer is the opaque argument passed to
1012 * rte_mempool_obj_iter() or rte_mempool_create().
1014 * The mbuf to initialize.
1016 * The index of the mbuf in the pool table.
1018 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1019 void *m, unsigned i);
1023 * A packet mbuf pool constructor.
1025 * This function initializes the mempool private data in the case of a
1026 * pktmbuf pool. This private data is needed by the driver. The
1027 * function must be called on the mempool before it is used, or it
1028 * can be given as a callback function to rte_mempool_create() at
1029 * pool creation. It can be extended by the user, for example, to
1030 * provide another packet size.
1033 * The mempool from which mbufs originate.
1035 * A pointer that can be used by the user to retrieve useful information
1036 * for mbuf initialization. This pointer is the opaque argument passed to
1037 * rte_mempool_create().
1039 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1042 * Create a mbuf pool.
1044 * This function creates and initializes a packet mbuf pool. It is
1045 * a wrapper to rte_mempool functions.
1048 * The name of the mbuf pool.
1050 * The number of elements in the mbuf pool. The optimum size (in terms
1051 * of memory usage) for a mempool is when n is a power of two minus one:
1054 * Size of the per-core object cache. See rte_mempool_create() for
1057 * Size of application private are between the rte_mbuf structure
1058 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1059 * @param data_room_size
1060 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1062 * The socket identifier where the memory should be allocated. The
1063 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1066 * The pointer to the new allocated mempool, on success. NULL on error
1067 * with rte_errno set appropriately. Possible rte_errno values include:
1068 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1069 * - E_RTE_SECONDARY - function was called from a secondary process instance
1070 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1071 * - ENOSPC - the maximum number of memzones has already been allocated
1072 * - EEXIST - a memzone with the same name already exists
1073 * - ENOMEM - no appropriate memory area found in which to create memzone
1075 struct rte_mempool *
1076 rte_pktmbuf_pool_create(const char *name, unsigned n,
1077 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1081 * Get the data room size of mbufs stored in a pktmbuf_pool
1083 * The data room size is the amount of data that can be stored in a
1084 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1087 * The packet mbuf pool.
1089 * The data room size of mbufs stored in this mempool.
1091 static inline uint16_t
1092 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1094 struct rte_pktmbuf_pool_private *mbp_priv;
1096 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1097 return mbp_priv->mbuf_data_room_size;
1101 * Get the application private size of mbufs stored in a pktmbuf_pool
1103 * The private size of mbuf is a zone located between the rte_mbuf
1104 * structure and the data buffer where an application can store data
1105 * associated to a packet.
1108 * The packet mbuf pool.
1110 * The private size of mbufs stored in this mempool.
1112 static inline uint16_t
1113 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1115 struct rte_pktmbuf_pool_private *mbp_priv;
1117 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1118 return mbp_priv->mbuf_priv_size;
1122 * Reset the data_off field of a packet mbuf to its default value.
1124 * The given mbuf must have only one segment, which should be empty.
1127 * The packet mbuf's data_off field has to be reset.
1129 static inline void rte_pktmbuf_reset_headroom(struct rte_mbuf *m)
1131 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1135 * Reset the fields of a packet mbuf to their default values.
1137 * The given mbuf must have only one segment.
1140 * The packet mbuf to be resetted.
1142 #define MBUF_INVALID_PORT UINT16_MAX
1144 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1150 m->vlan_tci_outer = 0;
1152 m->port = MBUF_INVALID_PORT;
1156 rte_pktmbuf_reset_headroom(m);
1159 __rte_mbuf_sanity_check(m, 1);
1163 * Allocate a new mbuf from a mempool.
1165 * This new mbuf contains one segment, which has a length of 0. The pointer
1166 * to data is initialized to have some bytes of headroom in the buffer
1167 * (if buffer size allows).
1170 * The mempool from which the mbuf is allocated.
1172 * - The pointer to the new mbuf on success.
1173 * - NULL if allocation failed.
1175 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1178 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1179 rte_pktmbuf_reset(m);
1184 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1188 * The mempool from which mbufs are allocated.
1190 * Array of pointers to mbufs
1195 * - -ENOENT: Not enough entries in the mempool; no mbufs are retrieved.
1197 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1198 struct rte_mbuf **mbufs, unsigned count)
1203 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1207 /* To understand duff's device on loop unwinding optimization, see
1208 * https://en.wikipedia.org/wiki/Duff's_device.
1209 * Here while() loop is used rather than do() while{} to avoid extra
1210 * check if count is zero.
1212 switch (count % 4) {
1214 while (idx != count) {
1215 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1216 rte_pktmbuf_reset(mbufs[idx]);
1220 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1221 rte_pktmbuf_reset(mbufs[idx]);
1225 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1226 rte_pktmbuf_reset(mbufs[idx]);
1230 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1231 rte_pktmbuf_reset(mbufs[idx]);
1240 * Attach packet mbuf to another packet mbuf.
1242 * After attachment we refer the mbuf we attached as 'indirect',
1243 * while mbuf we attached to as 'direct'.
1244 * The direct mbuf's reference counter is incremented.
1246 * Right now, not supported:
1247 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1248 * - mbuf we trying to attach (mi) is used by someone else
1249 * e.g. it's reference counter is greater then 1.
1252 * The indirect packet mbuf.
1254 * The packet mbuf we're attaching to.
1256 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1258 struct rte_mbuf *md;
1260 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1261 rte_mbuf_refcnt_read(mi) == 1);
1263 /* if m is not direct, get the mbuf that embeds the data */
1264 if (RTE_MBUF_DIRECT(m))
1267 md = rte_mbuf_from_indirect(m);
1269 rte_mbuf_refcnt_update(md, 1);
1270 mi->priv_size = m->priv_size;
1271 mi->buf_iova = m->buf_iova;
1272 mi->buf_addr = m->buf_addr;
1273 mi->buf_len = m->buf_len;
1275 mi->data_off = m->data_off;
1276 mi->data_len = m->data_len;
1278 mi->vlan_tci = m->vlan_tci;
1279 mi->vlan_tci_outer = m->vlan_tci_outer;
1280 mi->tx_offload = m->tx_offload;
1284 mi->pkt_len = mi->data_len;
1286 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1287 mi->packet_type = m->packet_type;
1288 mi->timestamp = m->timestamp;
1290 __rte_mbuf_sanity_check(mi, 1);
1291 __rte_mbuf_sanity_check(m, 0);
1295 * Detach an indirect packet mbuf.
1297 * - restore original mbuf address and length values.
1298 * - reset pktmbuf data and data_len to their default values.
1299 * - decrement the direct mbuf's reference counter. When the
1300 * reference counter becomes 0, the direct mbuf is freed.
1302 * All other fields of the given packet mbuf will be left intact.
1305 * The indirect attached packet mbuf.
1307 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1309 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1310 struct rte_mempool *mp = m->pool;
1311 uint32_t mbuf_size, buf_len, priv_size;
1313 priv_size = rte_pktmbuf_priv_size(mp);
1314 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1315 buf_len = rte_pktmbuf_data_room_size(mp);
1317 m->priv_size = priv_size;
1318 m->buf_addr = (char *)m + mbuf_size;
1319 m->buf_iova = rte_mempool_virt2iova(m) + mbuf_size;
1320 m->buf_len = (uint16_t)buf_len;
1321 rte_pktmbuf_reset_headroom(m);
1325 if (rte_mbuf_refcnt_update(md, -1) == 0) {
1328 rte_mbuf_refcnt_set(md, 1);
1329 rte_mbuf_raw_free(md);
1334 * Decrease reference counter and unlink a mbuf segment
1336 * This function does the same than a free, except that it does not
1337 * return the segment to its pool.
1338 * It decreases the reference counter, and if it reaches 0, it is
1339 * detached from its parent for an indirect mbuf.
1342 * The mbuf to be unlinked
1344 * - (m) if it is the last reference. It can be recycled or freed.
1345 * - (NULL) if the mbuf still has remaining references on it.
1347 static __rte_always_inline struct rte_mbuf *
1348 rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1350 __rte_mbuf_sanity_check(m, 0);
1352 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1354 if (RTE_MBUF_INDIRECT(m))
1355 rte_pktmbuf_detach(m);
1357 if (m->next != NULL) {
1364 } else if (rte_atomic16_add_return(&m->refcnt_atomic, -1) == 0) {
1367 if (RTE_MBUF_INDIRECT(m))
1368 rte_pktmbuf_detach(m);
1370 if (m->next != NULL) {
1374 rte_mbuf_refcnt_set(m, 1);
1381 /* deprecated, replaced by rte_pktmbuf_prefree_seg() */
1383 static inline struct rte_mbuf *
1384 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1386 return rte_pktmbuf_prefree_seg(m);
1390 * Free a segment of a packet mbuf into its original mempool.
1392 * Free an mbuf, without parsing other segments in case of chained
1396 * The packet mbuf segment to be freed.
1398 static __rte_always_inline void
1399 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1401 m = rte_pktmbuf_prefree_seg(m);
1402 if (likely(m != NULL))
1403 rte_mbuf_raw_free(m);
1407 * Free a packet mbuf back into its original mempool.
1409 * Free an mbuf, and all its segments in case of chained buffers. Each
1410 * segment is added back into its original mempool.
1413 * The packet mbuf to be freed.
1415 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1417 struct rte_mbuf *m_next;
1419 __rte_mbuf_sanity_check(m, 1);
1423 rte_pktmbuf_free_seg(m);
1429 * Creates a "clone" of the given packet mbuf.
1431 * Walks through all segments of the given packet mbuf, and for each of them:
1432 * - Creates a new packet mbuf from the given pool.
1433 * - Attaches newly created mbuf to the segment.
1434 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1435 * from the original packet mbuf.
1438 * The packet mbuf to be cloned.
1440 * The mempool from which the "clone" mbufs are allocated.
1442 * - The pointer to the new "clone" mbuf on success.
1443 * - NULL if allocation fails.
1445 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1446 struct rte_mempool *mp)
1448 struct rte_mbuf *mc, *mi, **prev;
1452 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1457 pktlen = md->pkt_len;
1462 rte_pktmbuf_attach(mi, md);
1465 } while ((md = md->next) != NULL &&
1466 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1470 mc->pkt_len = pktlen;
1472 /* Allocation of new indirect segment failed */
1473 if (unlikely (mi == NULL)) {
1474 rte_pktmbuf_free(mc);
1478 __rte_mbuf_sanity_check(mc, 1);
1483 * Adds given value to the refcnt of all packet mbuf segments.
1485 * Walks through all segments of given packet mbuf and for each of them
1486 * invokes rte_mbuf_refcnt_update().
1489 * The packet mbuf whose refcnt to be updated.
1491 * The value to add to the mbuf's segments refcnt.
1493 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1495 __rte_mbuf_sanity_check(m, 1);
1498 rte_mbuf_refcnt_update(m, v);
1499 } while ((m = m->next) != NULL);
1503 * Get the headroom in a packet mbuf.
1508 * The length of the headroom.
1510 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1512 __rte_mbuf_sanity_check(m, 0);
1517 * Get the tailroom of a packet mbuf.
1522 * The length of the tailroom.
1524 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1526 __rte_mbuf_sanity_check(m, 0);
1527 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1532 * Get the last segment of the packet.
1537 * The last segment of the given mbuf.
1539 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1541 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1543 __rte_mbuf_sanity_check(m, 1);
1544 while (m2->next != NULL)
1550 * A macro that points to an offset into the data in the mbuf.
1552 * The returned pointer is cast to type t. Before using this
1553 * function, the user must ensure that the first segment is large
1554 * enough to accommodate its data.
1559 * The offset into the mbuf data.
1561 * The type to cast the result into.
1563 #define rte_pktmbuf_mtod_offset(m, t, o) \
1564 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1567 * A macro that points to the start of the data in the mbuf.
1569 * The returned pointer is cast to type t. Before using this
1570 * function, the user must ensure that the first segment is large
1571 * enough to accommodate its data.
1576 * The type to cast the result into.
1578 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1581 * A macro that returns the IO address that points to an offset of the
1582 * start of the data in the mbuf
1587 * The offset into the data to calculate address from.
1589 #define rte_pktmbuf_iova_offset(m, o) \
1590 (rte_iova_t)((m)->buf_iova + (m)->data_off + (o))
1593 #define rte_pktmbuf_mtophys_offset(m, o) \
1594 rte_pktmbuf_iova_offset(m, o)
1597 * A macro that returns the IO address that points to the start of the
1603 #define rte_pktmbuf_iova(m) rte_pktmbuf_iova_offset(m, 0)
1606 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_iova(m)
1609 * A macro that returns the length of the packet.
1611 * The value can be read or assigned.
1616 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1619 * A macro that returns the length of the segment.
1621 * The value can be read or assigned.
1626 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1629 * Prepend len bytes to an mbuf data area.
1631 * Returns a pointer to the new
1632 * data start address. If there is not enough headroom in the first
1633 * segment, the function will return NULL, without modifying the mbuf.
1638 * The amount of data to prepend (in bytes).
1640 * A pointer to the start of the newly prepended data, or
1641 * NULL if there is not enough headroom space in the first segment
1643 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1646 __rte_mbuf_sanity_check(m, 1);
1648 if (unlikely(len > rte_pktmbuf_headroom(m)))
1652 m->data_len = (uint16_t)(m->data_len + len);
1653 m->pkt_len = (m->pkt_len + len);
1655 return (char *)m->buf_addr + m->data_off;
1659 * Append len bytes to an mbuf.
1661 * Append len bytes to an mbuf and return a pointer to the start address
1662 * of the added data. If there is not enough tailroom in the last
1663 * segment, the function will return NULL, without modifying the mbuf.
1668 * The amount of data to append (in bytes).
1670 * A pointer to the start of the newly appended data, or
1671 * NULL if there is not enough tailroom space in the last segment
1673 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1676 struct rte_mbuf *m_last;
1678 __rte_mbuf_sanity_check(m, 1);
1680 m_last = rte_pktmbuf_lastseg(m);
1681 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1684 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1685 m_last->data_len = (uint16_t)(m_last->data_len + len);
1686 m->pkt_len = (m->pkt_len + len);
1687 return (char*) tail;
1691 * Remove len bytes at the beginning of an mbuf.
1693 * Returns a pointer to the start address of the new data area. If the
1694 * length is greater than the length of the first segment, then the
1695 * function will fail and return NULL, without modifying the mbuf.
1700 * The amount of data to remove (in bytes).
1702 * A pointer to the new start of the data.
1704 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1706 __rte_mbuf_sanity_check(m, 1);
1708 if (unlikely(len > m->data_len))
1711 m->data_len = (uint16_t)(m->data_len - len);
1713 m->pkt_len = (m->pkt_len - len);
1714 return (char *)m->buf_addr + m->data_off;
1718 * Remove len bytes of data at the end of the mbuf.
1720 * If the length is greater than the length of the last segment, the
1721 * function will fail and return -1 without modifying the mbuf.
1726 * The amount of data to remove (in bytes).
1731 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1733 struct rte_mbuf *m_last;
1735 __rte_mbuf_sanity_check(m, 1);
1737 m_last = rte_pktmbuf_lastseg(m);
1738 if (unlikely(len > m_last->data_len))
1741 m_last->data_len = (uint16_t)(m_last->data_len - len);
1742 m->pkt_len = (m->pkt_len - len);
1747 * Test if mbuf data is contiguous.
1752 * - 1, if all data is contiguous (one segment).
1753 * - 0, if there is several segments.
1755 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1757 __rte_mbuf_sanity_check(m, 1);
1758 return !!(m->nb_segs == 1);
1762 * @internal used by rte_pktmbuf_read().
1764 const void *__rte_pktmbuf_read(const struct rte_mbuf *m, uint32_t off,
1765 uint32_t len, void *buf);
1768 * Read len data bytes in a mbuf at specified offset.
1770 * If the data is contiguous, return the pointer in the mbuf data, else
1771 * copy the data in the buffer provided by the user and return its
1775 * The pointer to the mbuf.
1777 * The offset of the data in the mbuf.
1779 * The amount of bytes to read.
1781 * The buffer where data is copied if it is not contigous in mbuf
1782 * data. Its length should be at least equal to the len parameter.
1784 * The pointer to the data, either in the mbuf if it is contiguous,
1785 * or in the user buffer. If mbuf is too small, NULL is returned.
1787 static inline const void *rte_pktmbuf_read(const struct rte_mbuf *m,
1788 uint32_t off, uint32_t len, void *buf)
1790 if (likely(off + len <= rte_pktmbuf_data_len(m)))
1791 return rte_pktmbuf_mtod_offset(m, char *, off);
1793 return __rte_pktmbuf_read(m, off, len, buf);
1797 * Chain an mbuf to another, thereby creating a segmented packet.
1799 * Note: The implementation will do a linear walk over the segments to find
1800 * the tail entry. For cases when there are many segments, it's better to
1801 * chain the entries manually.
1804 * The head of the mbuf chain (the first packet)
1806 * The mbuf to put last in the chain
1810 * - -EOVERFLOW, if the chain is full (256 entries)
1812 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1814 struct rte_mbuf *cur_tail;
1816 /* Check for number-of-segments-overflow */
1817 if (head->nb_segs + tail->nb_segs >= 1 << (sizeof(head->nb_segs) * 8))
1820 /* Chain 'tail' onto the old tail */
1821 cur_tail = rte_pktmbuf_lastseg(head);
1822 cur_tail->next = tail;
1824 /* accumulate number of segments and total length. */
1825 head->nb_segs = (uint8_t)(head->nb_segs + tail->nb_segs);
1826 head->pkt_len += tail->pkt_len;
1828 /* pkt_len is only set in the head */
1829 tail->pkt_len = tail->data_len;
1835 * Validate general requirements for Tx offload in mbuf.
1837 * This function checks correctness and completeness of Tx offload settings.
1840 * The packet mbuf to be validated.
1842 * 0 if packet is valid
1845 rte_validate_tx_offload(const struct rte_mbuf *m)
1847 uint64_t ol_flags = m->ol_flags;
1848 uint64_t inner_l3_offset = m->l2_len;
1850 /* Does packet set any of available offloads? */
1851 if (!(ol_flags & PKT_TX_OFFLOAD_MASK))
1854 if (ol_flags & PKT_TX_OUTER_IP_CKSUM)
1855 inner_l3_offset += m->outer_l2_len + m->outer_l3_len;
1857 /* Headers are fragmented */
1858 if (rte_pktmbuf_data_len(m) < inner_l3_offset + m->l3_len + m->l4_len)
1861 /* IP checksum can be counted only for IPv4 packet */
1862 if ((ol_flags & PKT_TX_IP_CKSUM) && (ol_flags & PKT_TX_IPV6))
1865 /* IP type not set when required */
1866 if (ol_flags & (PKT_TX_L4_MASK | PKT_TX_TCP_SEG))
1867 if (!(ol_flags & (PKT_TX_IPV4 | PKT_TX_IPV6)))
1870 /* Check requirements for TSO packet */
1871 if (ol_flags & PKT_TX_TCP_SEG)
1872 if ((m->tso_segsz == 0) ||
1873 ((ol_flags & PKT_TX_IPV4) &&
1874 !(ol_flags & PKT_TX_IP_CKSUM)))
1877 /* PKT_TX_OUTER_IP_CKSUM set for non outer IPv4 packet. */
1878 if ((ol_flags & PKT_TX_OUTER_IP_CKSUM) &&
1879 !(ol_flags & PKT_TX_OUTER_IPV4))
1886 * Linearize data in mbuf.
1888 * This function moves the mbuf data in the first segment if there is enough
1889 * tailroom. The subsequent segments are unchained and freed.
1898 rte_pktmbuf_linearize(struct rte_mbuf *mbuf)
1900 int seg_len, copy_len;
1902 struct rte_mbuf *m_next;
1905 if (rte_pktmbuf_is_contiguous(mbuf))
1908 /* Extend first segment to the total packet length */
1909 copy_len = rte_pktmbuf_pkt_len(mbuf) - rte_pktmbuf_data_len(mbuf);
1911 if (unlikely(copy_len > rte_pktmbuf_tailroom(mbuf)))
1914 buffer = rte_pktmbuf_mtod_offset(mbuf, char *, mbuf->data_len);
1915 mbuf->data_len = (uint16_t)(mbuf->pkt_len);
1917 /* Append data from next segments to the first one */
1922 seg_len = rte_pktmbuf_data_len(m);
1923 rte_memcpy(buffer, rte_pktmbuf_mtod(m, char *), seg_len);
1926 rte_pktmbuf_free_seg(m);
1937 * Dump an mbuf structure to a file.
1939 * Dump all fields for the given packet mbuf and all its associated
1940 * segments (in the case of a chained buffer).
1943 * A pointer to a file for output
1947 * If dump_len != 0, also dump the "dump_len" first data bytes of
1950 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1956 #endif /* _RTE_MBUF_H_ */