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_config.h>
66 #include <rte_mempool.h>
67 #include <rte_memory.h>
68 #include <rte_atomic.h>
69 #include <rte_prefetch.h>
70 #include <rte_branch_prediction.h>
71 #include <rte_mbuf_ptype.h>
78 * Packet Offload Features Flags. It also carry packet type information.
79 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
81 * - RX flags start at bit position zero, and get added to the left of previous
83 * - The most-significant 3 bits are reserved for generic mbuf flags
84 * - TX flags therefore start at bit position 60 (i.e. 63-3), and new flags get
85 * added to the right of the previously defined flags i.e. they should count
86 * downwards, not upwards.
88 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
89 * rte_get_tx_ol_flag_name().
93 * The RX packet is a 802.1q VLAN packet, and the tci has been
94 * saved in in mbuf->vlan_tci.
95 * If the flag PKT_RX_VLAN_STRIPPED is also present, the VLAN
96 * header has been stripped from mbuf data, else it is still
99 #define PKT_RX_VLAN (1ULL << 0)
101 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
102 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
106 * Checking this flag alone is deprecated: check the 2 bits of
107 * PKT_RX_L4_CKSUM_MASK.
108 * This flag was set when the L4 checksum of a packet was detected as
109 * wrong by the hardware.
111 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
115 * Checking this flag alone is deprecated: check the 2 bits of
116 * PKT_RX_IP_CKSUM_MASK.
117 * This flag was set when the IP checksum of a packet was detected as
118 * wrong by the hardware.
120 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
122 #define PKT_RX_EIP_CKSUM_BAD (1ULL << 5) /**< External IP header checksum error. */
125 * A vlan has been stripped by the hardware and its tci is saved in
126 * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
127 * in the RX configuration of the PMD.
128 * When PKT_RX_VLAN_STRIPPED is set, PKT_RX_VLAN must also be set.
130 #define PKT_RX_VLAN_STRIPPED (1ULL << 6)
133 * Mask of bits used to determine the status of RX IP checksum.
134 * - PKT_RX_IP_CKSUM_UNKNOWN: no information about the RX IP checksum
135 * - PKT_RX_IP_CKSUM_BAD: the IP checksum in the packet is wrong
136 * - PKT_RX_IP_CKSUM_GOOD: the IP checksum in the packet is valid
137 * - PKT_RX_IP_CKSUM_NONE: the IP checksum is not correct in the packet
138 * data, but the integrity of the IP header is verified.
140 #define PKT_RX_IP_CKSUM_MASK ((1ULL << 4) | (1ULL << 7))
142 #define PKT_RX_IP_CKSUM_UNKNOWN 0
143 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
144 #define PKT_RX_IP_CKSUM_GOOD (1ULL << 7)
145 #define PKT_RX_IP_CKSUM_NONE ((1ULL << 4) | (1ULL << 7))
148 * Mask of bits used to determine the status of RX L4 checksum.
149 * - PKT_RX_L4_CKSUM_UNKNOWN: no information about the RX L4 checksum
150 * - PKT_RX_L4_CKSUM_BAD: the L4 checksum in the packet is wrong
151 * - PKT_RX_L4_CKSUM_GOOD: the L4 checksum in the packet is valid
152 * - PKT_RX_L4_CKSUM_NONE: the L4 checksum is not correct in the packet
153 * data, but the integrity of the L4 data is verified.
155 #define PKT_RX_L4_CKSUM_MASK ((1ULL << 3) | (1ULL << 8))
157 #define PKT_RX_L4_CKSUM_UNKNOWN 0
158 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
159 #define PKT_RX_L4_CKSUM_GOOD (1ULL << 8)
160 #define PKT_RX_L4_CKSUM_NONE ((1ULL << 3) | (1ULL << 8))
162 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
163 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
164 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
165 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
168 * The 2 vlans have been stripped by the hardware and their tci are
169 * saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
170 * This can only happen if vlan stripping is enabled in the RX
171 * configuration of the PMD. If this flag is set,
172 * When PKT_RX_QINQ_STRIPPED is set, the flags (PKT_RX_VLAN |
173 * PKT_RX_VLAN_STRIPPED | PKT_RX_QINQ) must also be set.
175 #define PKT_RX_QINQ_STRIPPED (1ULL << 15)
178 * When packets are coalesced by a hardware or virtual driver, this flag
179 * can be set in the RX mbuf, meaning that the m->tso_segsz field is
180 * valid and is set to the segment size of original packets.
182 #define PKT_RX_LRO (1ULL << 16)
185 * Indicate that the timestamp field in the mbuf is valid.
187 #define PKT_RX_TIMESTAMP (1ULL << 17)
190 * Indicate that security offload processing was applied on the RX packet.
192 #define PKT_RX_SEC_OFFLOAD (1ULL << 18)
195 * Indicate that security offload processing failed on the RX packet.
197 #define PKT_RX_SEC_OFFLOAD_FAILED (1ULL << 19)
200 * The RX packet is a double VLAN, and the outer tci has been
201 * saved in in mbuf->vlan_tci_outer.
202 * If the flag PKT_RX_QINQ_STRIPPED is also present, both VLANs
203 * headers have been stripped from mbuf data, else they are still
206 #define PKT_RX_QINQ (1ULL << 20)
208 /* add new RX flags here */
210 /* add new TX flags here */
213 * UDP Fragmentation Offload flag. This flag is used for enabling UDP
214 * fragmentation in SW or in HW. When use UFO, mbuf->tso_segsz is used
215 * to store the MSS of UDP fragments.
217 #define PKT_TX_UDP_SEG (1ULL << 42)
220 * Request security offload processing on the TX packet.
222 #define PKT_TX_SEC_OFFLOAD (1ULL << 43)
225 * Offload the MACsec. This flag must be set by the application to enable
226 * this offload feature for a packet to be transmitted.
228 #define PKT_TX_MACSEC (1ULL << 44)
231 * Bits 45:48 used for the tunnel type.
232 * When doing Tx offload like TSO or checksum, the HW needs to configure the
233 * tunnel type into the HW descriptors.
235 #define PKT_TX_TUNNEL_VXLAN (0x1ULL << 45)
236 #define PKT_TX_TUNNEL_GRE (0x2ULL << 45)
237 #define PKT_TX_TUNNEL_IPIP (0x3ULL << 45)
238 #define PKT_TX_TUNNEL_GENEVE (0x4ULL << 45)
239 /**< TX packet with MPLS-in-UDP RFC 7510 header. */
240 #define PKT_TX_TUNNEL_MPLSINUDP (0x5ULL << 45)
241 /* add new TX TUNNEL type here */
242 #define PKT_TX_TUNNEL_MASK (0xFULL << 45)
245 * Second VLAN insertion (QinQ) flag.
247 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
250 * TCP segmentation offload. To enable this offload feature for a
251 * packet to be transmitted on hardware supporting TSO:
252 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
254 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
255 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
257 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
258 * - calculate the pseudo header checksum without taking ip_len in account,
259 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
260 * rte_ipv6_phdr_cksum() that can be used as helpers.
262 #define PKT_TX_TCP_SEG (1ULL << 50)
264 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
267 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
268 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
269 * L4 checksum offload, the user needs to:
270 * - fill l2_len and l3_len in mbuf
271 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
272 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
273 * - calculate the pseudo header checksum and set it in the L4 header (only
274 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
275 * For SCTP, set the crc field to 0.
277 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
278 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
279 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
280 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
281 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
284 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
285 * also be set by the application, although a PMD will only check
287 * - set the IP checksum field in the packet to 0
288 * - fill the mbuf offload information: l2_len, l3_len
290 #define PKT_TX_IP_CKSUM (1ULL << 54)
293 * Packet is IPv4. This flag must be set when using any offload feature
294 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
295 * packet. If the packet is a tunneled packet, this flag is related to
298 #define PKT_TX_IPV4 (1ULL << 55)
301 * Packet is IPv6. This flag must be set when using an offload feature
302 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
303 * packet. If the packet is a tunneled packet, this flag is related to
306 #define PKT_TX_IPV6 (1ULL << 56)
308 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
311 * Offload the IP checksum of an external header in the hardware. The
312 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
313 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
314 * packet must be set to 0.
315 * - set the outer IP checksum field in the packet to 0
316 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
318 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
321 * Packet outer header is IPv4. This flag must be set when using any
322 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
323 * outer header of the tunneled packet is an IPv4 packet.
325 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
328 * Packet outer header is IPv6. This flag must be set when using any
329 * outer offload feature (L4 checksum) to tell the NIC that the outer
330 * header of the tunneled packet is an IPv6 packet.
332 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
335 * Bitmask of all supported packet Tx offload features flags,
336 * which can be set for packet.
338 #define PKT_TX_OFFLOAD_MASK ( \
341 PKT_TX_OUTER_IP_CKSUM | \
343 PKT_TX_IEEE1588_TMST | \
346 PKT_TX_TUNNEL_MASK | \
350 #define __RESERVED (1ULL << 61) /**< reserved for future mbuf use */
352 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
354 /* Use final bit of flags to indicate a control mbuf */
355 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
357 /** Alignment constraint of mbuf private area. */
358 #define RTE_MBUF_PRIV_ALIGN 8
361 * Get the name of a RX offload flag
364 * The mask describing the flag.
366 * The name of this flag, or NULL if it's not a valid RX flag.
368 const char *rte_get_rx_ol_flag_name(uint64_t mask);
371 * Dump the list of RX offload flags in a buffer
374 * The mask describing the RX flags.
378 * The length of the buffer.
380 * 0 on success, (-1) on error.
382 int rte_get_rx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
385 * Get the name of a TX offload flag
388 * The mask describing the flag. Usually only one bit must be set.
389 * Several bits can be given if they belong to the same mask.
390 * Ex: PKT_TX_L4_MASK.
392 * The name of this flag, or NULL if it's not a valid TX flag.
394 const char *rte_get_tx_ol_flag_name(uint64_t mask);
397 * Dump the list of TX offload flags in a buffer
400 * The mask describing the TX flags.
404 * The length of the buffer.
406 * 0 on success, (-1) on error.
408 int rte_get_tx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
411 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
412 * splitting it into multiple segments.
413 * So, for mbufs that planned to be involved into RX/TX, the recommended
414 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
416 #define RTE_MBUF_DEFAULT_DATAROOM 2048
417 #define RTE_MBUF_DEFAULT_BUF_SIZE \
418 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
420 /* define a set of marker types that can be used to refer to set points in the
423 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
425 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
427 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
428 * with a single assignment */
431 * The generic rte_mbuf, containing a packet mbuf.
436 void *buf_addr; /**< Virtual address of segment buffer. */
438 * Physical address of segment buffer.
439 * Force alignment to 8-bytes, so as to ensure we have the exact
440 * same mbuf cacheline0 layout for 32-bit and 64-bit. This makes
441 * working on vector drivers easier.
446 rte_iova_t buf_physaddr; /**< deprecated */
447 } __rte_aligned(sizeof(rte_iova_t));
449 /* next 8 bytes are initialised on RX descriptor rearm */
454 * Reference counter. Its size should at least equal to the size
455 * of port field (16 bits), to support zero-copy broadcast.
456 * It should only be accessed using the following functions:
457 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
458 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
459 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
464 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
465 uint16_t refcnt; /**< Non-atomically accessed refcnt */
467 uint16_t nb_segs; /**< Number of segments. */
469 /** Input port (16 bits to support more than 256 virtual ports). */
472 uint64_t ol_flags; /**< Offload features. */
474 /* remaining bytes are set on RX when pulling packet from descriptor */
475 MARKER rx_descriptor_fields1;
478 * The packet type, which is the combination of outer/inner L2, L3, L4
479 * and tunnel types. The packet_type is about data really present in the
480 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
481 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
482 * vlan is stripped from the data.
486 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
488 uint32_t l2_type:4; /**< (Outer) L2 type. */
489 uint32_t l3_type:4; /**< (Outer) L3 type. */
490 uint32_t l4_type:4; /**< (Outer) L4 type. */
491 uint32_t tun_type:4; /**< Tunnel type. */
494 uint8_t inner_esp_next_proto;
495 /**< ESP next protocol type, valid if
496 * RTE_PTYPE_TUNNEL_ESP tunnel type is set
501 uint8_t inner_l2_type:4;
502 /**< Inner L2 type. */
503 uint8_t inner_l3_type:4;
504 /**< Inner L3 type. */
507 uint32_t inner_l4_type:4; /**< Inner L4 type. */
511 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
512 uint16_t data_len; /**< Amount of data in segment buffer. */
513 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN_STRIPPED is set. */
517 uint32_t rss; /**< RSS hash result if RSS enabled */
526 /**< Second 4 flexible bytes */
529 /**< First 4 flexible bytes or FD ID, dependent on
530 PKT_RX_FDIR_* flag in ol_flags. */
531 } fdir; /**< Filter identifier if FDIR enabled */
535 } sched; /**< Hierarchical scheduler */
536 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
537 } hash; /**< hash information */
539 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ_STRIPPED is set. */
540 uint16_t vlan_tci_outer;
542 uint16_t buf_len; /**< Length of segment buffer. */
544 /** Valid if PKT_RX_TIMESTAMP is set. The unit and time reference
545 * are not normalized but are always the same for a given port.
549 /* second cache line - fields only used in slow path or on TX */
550 MARKER cacheline1 __rte_cache_min_aligned;
554 void *userdata; /**< Can be used for external metadata */
555 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
558 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
559 struct rte_mbuf *next; /**< Next segment of scattered packet. */
561 /* fields to support TX offloads */
564 uint64_t tx_offload; /**< combined for easy fetch */
568 /**< L2 (MAC) Header Length for non-tunneling pkt.
569 * Outer_L4_len + ... + Inner_L2_len for tunneling pkt.
571 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
572 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
573 uint64_t tso_segsz:16; /**< TCP TSO segment size */
575 /* fields for TX offloading of tunnels */
576 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
577 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
579 /* uint64_t unused:8; */
583 /** Size of the application private data. In case of an indirect
584 * mbuf, it stores the direct mbuf private data size. */
587 /** Timesync flags for use with IEEE1588. */
590 /** Sequence number. See also rte_reorder_insert(). */
593 } __rte_cache_aligned;
595 /**< Maximum number of nb_segs allowed. */
596 #define RTE_MBUF_MAX_NB_SEGS UINT16_MAX
599 * Prefetch the first part of the mbuf
601 * The first 64 bytes of the mbuf corresponds to fields that are used early
602 * in the receive path. If the cache line of the architecture is higher than
603 * 64B, the second part will also be prefetched.
606 * The pointer to the mbuf.
609 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
611 rte_prefetch0(&m->cacheline0);
615 * Prefetch the second part of the mbuf
617 * The next 64 bytes of the mbuf corresponds to fields that are used in the
618 * transmit path. If the cache line of the architecture is higher than 64B,
619 * this function does nothing as it is expected that the full mbuf is
623 * The pointer to the mbuf.
626 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
628 #if RTE_CACHE_LINE_SIZE == 64
629 rte_prefetch0(&m->cacheline1);
636 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
639 * Return the IO address of the beginning of the mbuf data
642 * The pointer to the mbuf.
644 * The IO address of the beginning of the mbuf data
646 static inline rte_iova_t
647 rte_mbuf_data_iova(const struct rte_mbuf *mb)
649 return mb->buf_iova + mb->data_off;
653 static inline phys_addr_t
654 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
656 return rte_mbuf_data_iova(mb);
660 * Return the default IO address of the beginning of the mbuf data
662 * This function is used by drivers in their receive function, as it
663 * returns the location where data should be written by the NIC, taking
664 * the default headroom in account.
667 * The pointer to the mbuf.
669 * The IO address of the beginning of the mbuf data
671 static inline rte_iova_t
672 rte_mbuf_data_iova_default(const struct rte_mbuf *mb)
674 return mb->buf_iova + RTE_PKTMBUF_HEADROOM;
678 static inline phys_addr_t
679 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
681 return rte_mbuf_data_iova_default(mb);
685 * Return the mbuf owning the data buffer address of an indirect mbuf.
688 * The pointer to the indirect mbuf.
690 * The address of the direct mbuf corresponding to buffer_addr.
692 static inline struct rte_mbuf *
693 rte_mbuf_from_indirect(struct rte_mbuf *mi)
695 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
699 * Return the buffer address embedded in the given mbuf.
702 * The pointer to the mbuf.
704 * The address of the data buffer owned by the mbuf.
707 rte_mbuf_to_baddr(struct rte_mbuf *md)
710 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
715 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
717 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
720 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
722 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
725 * Private data in case of pktmbuf pool.
727 * A structure that contains some pktmbuf_pool-specific data that are
728 * appended after the mempool structure (in private data).
730 struct rte_pktmbuf_pool_private {
731 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
732 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
735 #ifdef RTE_LIBRTE_MBUF_DEBUG
737 /** check mbuf type in debug mode */
738 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
740 #else /* RTE_LIBRTE_MBUF_DEBUG */
742 /** check mbuf type in debug mode */
743 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
745 #endif /* RTE_LIBRTE_MBUF_DEBUG */
747 #ifdef RTE_MBUF_REFCNT_ATOMIC
750 * Reads the value of an mbuf's refcnt.
754 * Reference count number.
756 static inline uint16_t
757 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
759 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
763 * Sets an mbuf's refcnt to a defined value.
770 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
772 rte_atomic16_set(&m->refcnt_atomic, new_value);
776 static inline uint16_t
777 __rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
779 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
783 * Adds given value to an mbuf's refcnt and returns its new value.
787 * Value to add/subtract
791 static inline uint16_t
792 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
795 * The atomic_add is an expensive operation, so we don't want to
796 * call it in the case where we know we are the uniq holder of
797 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
798 * operation has to be used because concurrent accesses on the
799 * reference counter can occur.
801 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
802 rte_mbuf_refcnt_set(m, 1 + value);
806 return __rte_mbuf_refcnt_update(m, value);
809 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
812 static inline uint16_t
813 __rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
815 m->refcnt = (uint16_t)(m->refcnt + value);
820 * Adds given value to an mbuf's refcnt and returns its new value.
822 static inline uint16_t
823 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
825 return __rte_mbuf_refcnt_update(m, value);
829 * Reads the value of an mbuf's refcnt.
831 static inline uint16_t
832 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
838 * Sets an mbuf's refcnt to the defined value.
841 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
843 m->refcnt = new_value;
846 #endif /* RTE_MBUF_REFCNT_ATOMIC */
849 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
856 * Sanity checks on an mbuf.
858 * Check the consistency of the given mbuf. The function will cause a
859 * panic if corruption is detected.
862 * The mbuf to be checked.
864 * True if the mbuf is a packet header, false if it is a sub-segment
865 * of a packet (in this case, some fields like nb_segs are not checked)
868 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
870 #define MBUF_RAW_ALLOC_CHECK(m) do { \
871 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1); \
872 RTE_ASSERT((m)->next == NULL); \
873 RTE_ASSERT((m)->nb_segs == 1); \
874 __rte_mbuf_sanity_check(m, 0); \
878 * Allocate an uninitialized mbuf from mempool *mp*.
880 * This function can be used by PMDs (especially in RX functions) to
881 * allocate an uninitialized mbuf. The driver is responsible of
882 * initializing all the required fields. See rte_pktmbuf_reset().
883 * For standard needs, prefer rte_pktmbuf_alloc().
885 * The caller can expect that the following fields of the mbuf structure
886 * are initialized: buf_addr, buf_iova, buf_len, refcnt=1, nb_segs=1,
887 * next=NULL, pool, priv_size. The other fields must be initialized
891 * The mempool from which mbuf is allocated.
893 * - The pointer to the new mbuf on success.
894 * - NULL if allocation failed.
896 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
901 if (rte_mempool_get(mp, &mb) < 0)
903 m = (struct rte_mbuf *)mb;
904 MBUF_RAW_ALLOC_CHECK(m);
909 * Put mbuf back into its original mempool.
911 * The caller must ensure that the mbuf is direct and properly
912 * reinitialized (refcnt=1, next=NULL, nb_segs=1), as done by
913 * rte_pktmbuf_prefree_seg().
915 * This function should be used with care, when optimization is
916 * required. For standard needs, prefer rte_pktmbuf_free() or
917 * rte_pktmbuf_free_seg().
920 * The mbuf to be freed.
922 static __rte_always_inline void
923 rte_mbuf_raw_free(struct rte_mbuf *m)
925 RTE_ASSERT(RTE_MBUF_DIRECT(m));
926 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
927 RTE_ASSERT(m->next == NULL);
928 RTE_ASSERT(m->nb_segs == 1);
929 __rte_mbuf_sanity_check(m, 0);
930 rte_mempool_put(m->pool, m);
933 /* compat with older versions */
936 __rte_mbuf_raw_free(struct rte_mbuf *m)
938 rte_mbuf_raw_free(m);
941 /* Operations on ctrl mbuf */
944 * The control mbuf constructor.
946 * This function initializes some fields in an mbuf structure that are
947 * not modified by the user once created (mbuf type, origin pool, buffer
948 * start address, and so on). This function is given as a callback function
949 * to rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
952 * The mempool from which the mbuf is allocated.
954 * A pointer that can be used by the user to retrieve useful information
955 * for mbuf initialization. This pointer is the opaque argument passed to
956 * rte_mempool_obj_iter() or rte_mempool_create().
958 * The mbuf to initialize.
960 * The index of the mbuf in the pool table.
962 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
963 void *m, unsigned i);
966 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
968 * This new mbuf is initialized with data pointing to the beginning of
969 * buffer, and with a length of zero.
972 * The mempool from which the mbuf is allocated.
974 * - The pointer to the new mbuf on success.
975 * - NULL if allocation failed.
977 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
980 * Free a control mbuf back into its original mempool.
983 * The control mbuf to be freed.
985 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
988 * A macro that returns the pointer to the carried data.
990 * The value that can be read or assigned.
995 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
998 * A macro that returns the length of the carried data.
1000 * The value that can be read or assigned.
1005 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
1008 * Tests if an mbuf is a control mbuf
1011 * The mbuf to be tested
1013 * - True (1) if the mbuf is a control mbuf
1014 * - False(0) otherwise
1017 rte_is_ctrlmbuf(struct rte_mbuf *m)
1019 return !!(m->ol_flags & CTRL_MBUF_FLAG);
1022 /* Operations on pkt mbuf */
1025 * The packet mbuf constructor.
1027 * This function initializes some fields in the mbuf structure that are
1028 * not modified by the user once created (origin pool, buffer start
1029 * address, and so on). This function is given as a callback function to
1030 * rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
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_obj_iter() or rte_mempool_create().
1039 * The mbuf to initialize.
1041 * The index of the mbuf in the pool table.
1043 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
1044 void *m, unsigned i);
1048 * A packet mbuf pool constructor.
1050 * This function initializes the mempool private data in the case of a
1051 * pktmbuf pool. This private data is needed by the driver. The
1052 * function must be called on the mempool before it is used, or it
1053 * can be given as a callback function to rte_mempool_create() at
1054 * pool creation. It can be extended by the user, for example, to
1055 * provide another packet size.
1058 * The mempool from which mbufs originate.
1060 * A pointer that can be used by the user to retrieve useful information
1061 * for mbuf initialization. This pointer is the opaque argument passed to
1062 * rte_mempool_create().
1064 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
1067 * Create a mbuf pool.
1069 * This function creates and initializes a packet mbuf pool. It is
1070 * a wrapper to rte_mempool functions.
1073 * The name of the mbuf pool.
1075 * The number of elements in the mbuf pool. The optimum size (in terms
1076 * of memory usage) for a mempool is when n is a power of two minus one:
1079 * Size of the per-core object cache. See rte_mempool_create() for
1082 * Size of application private are between the rte_mbuf structure
1083 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
1084 * @param data_room_size
1085 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
1087 * The socket identifier where the memory should be allocated. The
1088 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1091 * The pointer to the new allocated mempool, on success. NULL on error
1092 * with rte_errno set appropriately. Possible rte_errno values include:
1093 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1094 * - E_RTE_SECONDARY - function was called from a secondary process instance
1095 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1096 * - ENOSPC - the maximum number of memzones has already been allocated
1097 * - EEXIST - a memzone with the same name already exists
1098 * - ENOMEM - no appropriate memory area found in which to create memzone
1100 struct rte_mempool *
1101 rte_pktmbuf_pool_create(const char *name, unsigned n,
1102 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1106 * Get the data room size of mbufs stored in a pktmbuf_pool
1108 * The data room size is the amount of data that can be stored in a
1109 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1112 * The packet mbuf pool.
1114 * The data room size of mbufs stored in this mempool.
1116 static inline uint16_t
1117 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1119 struct rte_pktmbuf_pool_private *mbp_priv;
1121 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1122 return mbp_priv->mbuf_data_room_size;
1126 * Get the application private size of mbufs stored in a pktmbuf_pool
1128 * The private size of mbuf is a zone located between the rte_mbuf
1129 * structure and the data buffer where an application can store data
1130 * associated to a packet.
1133 * The packet mbuf pool.
1135 * The private size of mbufs stored in this mempool.
1137 static inline uint16_t
1138 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1140 struct rte_pktmbuf_pool_private *mbp_priv;
1142 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1143 return mbp_priv->mbuf_priv_size;
1147 * Reset the data_off field of a packet mbuf to its default value.
1149 * The given mbuf must have only one segment, which should be empty.
1152 * The packet mbuf's data_off field has to be reset.
1154 static inline void rte_pktmbuf_reset_headroom(struct rte_mbuf *m)
1156 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1160 * Reset the fields of a packet mbuf to their default values.
1162 * The given mbuf must have only one segment.
1165 * The packet mbuf to be resetted.
1167 #define MBUF_INVALID_PORT UINT16_MAX
1169 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1175 m->vlan_tci_outer = 0;
1177 m->port = MBUF_INVALID_PORT;
1181 rte_pktmbuf_reset_headroom(m);
1184 __rte_mbuf_sanity_check(m, 1);
1188 * Allocate a new mbuf from a mempool.
1190 * This new mbuf contains one segment, which has a length of 0. The pointer
1191 * to data is initialized to have some bytes of headroom in the buffer
1192 * (if buffer size allows).
1195 * The mempool from which the mbuf is allocated.
1197 * - The pointer to the new mbuf on success.
1198 * - NULL if allocation failed.
1200 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1203 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1204 rte_pktmbuf_reset(m);
1209 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1213 * The mempool from which mbufs are allocated.
1215 * Array of pointers to mbufs
1220 * - -ENOENT: Not enough entries in the mempool; no mbufs are retrieved.
1222 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1223 struct rte_mbuf **mbufs, unsigned count)
1228 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1232 /* To understand duff's device on loop unwinding optimization, see
1233 * https://en.wikipedia.org/wiki/Duff's_device.
1234 * Here while() loop is used rather than do() while{} to avoid extra
1235 * check if count is zero.
1237 switch (count % 4) {
1239 while (idx != count) {
1240 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1241 rte_pktmbuf_reset(mbufs[idx]);
1245 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1246 rte_pktmbuf_reset(mbufs[idx]);
1250 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1251 rte_pktmbuf_reset(mbufs[idx]);
1255 MBUF_RAW_ALLOC_CHECK(mbufs[idx]);
1256 rte_pktmbuf_reset(mbufs[idx]);
1265 * Attach packet mbuf to another packet mbuf.
1267 * After attachment we refer the mbuf we attached as 'indirect',
1268 * while mbuf we attached to as 'direct'.
1269 * The direct mbuf's reference counter is incremented.
1271 * Right now, not supported:
1272 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1273 * - mbuf we trying to attach (mi) is used by someone else
1274 * e.g. it's reference counter is greater then 1.
1277 * The indirect packet mbuf.
1279 * The packet mbuf we're attaching to.
1281 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1283 struct rte_mbuf *md;
1285 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1286 rte_mbuf_refcnt_read(mi) == 1);
1288 /* if m is not direct, get the mbuf that embeds the data */
1289 if (RTE_MBUF_DIRECT(m))
1292 md = rte_mbuf_from_indirect(m);
1294 rte_mbuf_refcnt_update(md, 1);
1295 mi->priv_size = m->priv_size;
1296 mi->buf_iova = m->buf_iova;
1297 mi->buf_addr = m->buf_addr;
1298 mi->buf_len = m->buf_len;
1300 mi->data_off = m->data_off;
1301 mi->data_len = m->data_len;
1303 mi->vlan_tci = m->vlan_tci;
1304 mi->vlan_tci_outer = m->vlan_tci_outer;
1305 mi->tx_offload = m->tx_offload;
1309 mi->pkt_len = mi->data_len;
1311 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1312 mi->packet_type = m->packet_type;
1313 mi->timestamp = m->timestamp;
1315 __rte_mbuf_sanity_check(mi, 1);
1316 __rte_mbuf_sanity_check(m, 0);
1320 * Detach an indirect packet mbuf.
1322 * - restore original mbuf address and length values.
1323 * - reset pktmbuf data and data_len to their default values.
1324 * - decrement the direct mbuf's reference counter. When the
1325 * reference counter becomes 0, the direct mbuf is freed.
1327 * All other fields of the given packet mbuf will be left intact.
1330 * The indirect attached packet mbuf.
1332 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1334 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1335 struct rte_mempool *mp = m->pool;
1336 uint32_t mbuf_size, buf_len, priv_size;
1338 priv_size = rte_pktmbuf_priv_size(mp);
1339 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1340 buf_len = rte_pktmbuf_data_room_size(mp);
1342 m->priv_size = priv_size;
1343 m->buf_addr = (char *)m + mbuf_size;
1344 m->buf_iova = rte_mempool_virt2iova(m) + mbuf_size;
1345 m->buf_len = (uint16_t)buf_len;
1346 rte_pktmbuf_reset_headroom(m);
1350 if (rte_mbuf_refcnt_update(md, -1) == 0) {
1353 rte_mbuf_refcnt_set(md, 1);
1354 rte_mbuf_raw_free(md);
1359 * Decrease reference counter and unlink a mbuf segment
1361 * This function does the same than a free, except that it does not
1362 * return the segment to its pool.
1363 * It decreases the reference counter, and if it reaches 0, it is
1364 * detached from its parent for an indirect mbuf.
1367 * The mbuf to be unlinked
1369 * - (m) if it is the last reference. It can be recycled or freed.
1370 * - (NULL) if the mbuf still has remaining references on it.
1372 static __rte_always_inline struct rte_mbuf *
1373 rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1375 __rte_mbuf_sanity_check(m, 0);
1377 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1379 if (RTE_MBUF_INDIRECT(m))
1380 rte_pktmbuf_detach(m);
1382 if (m->next != NULL) {
1389 } else if (__rte_mbuf_refcnt_update(m, -1) == 0) {
1391 if (RTE_MBUF_INDIRECT(m))
1392 rte_pktmbuf_detach(m);
1394 if (m->next != NULL) {
1398 rte_mbuf_refcnt_set(m, 1);
1405 /* deprecated, replaced by rte_pktmbuf_prefree_seg() */
1407 static inline struct rte_mbuf *
1408 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1410 return rte_pktmbuf_prefree_seg(m);
1414 * Free a segment of a packet mbuf into its original mempool.
1416 * Free an mbuf, without parsing other segments in case of chained
1420 * The packet mbuf segment to be freed.
1422 static __rte_always_inline void
1423 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1425 m = rte_pktmbuf_prefree_seg(m);
1426 if (likely(m != NULL))
1427 rte_mbuf_raw_free(m);
1431 * Free a packet mbuf back into its original mempool.
1433 * Free an mbuf, and all its segments in case of chained buffers. Each
1434 * segment is added back into its original mempool.
1437 * The packet mbuf to be freed.
1439 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1441 struct rte_mbuf *m_next;
1443 __rte_mbuf_sanity_check(m, 1);
1447 rte_pktmbuf_free_seg(m);
1453 * Creates a "clone" of the given packet mbuf.
1455 * Walks through all segments of the given packet mbuf, and for each of them:
1456 * - Creates a new packet mbuf from the given pool.
1457 * - Attaches newly created mbuf to the segment.
1458 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1459 * from the original packet mbuf.
1462 * The packet mbuf to be cloned.
1464 * The mempool from which the "clone" mbufs are allocated.
1466 * - The pointer to the new "clone" mbuf on success.
1467 * - NULL if allocation fails.
1469 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1470 struct rte_mempool *mp)
1472 struct rte_mbuf *mc, *mi, **prev;
1476 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1481 pktlen = md->pkt_len;
1486 rte_pktmbuf_attach(mi, md);
1489 } while ((md = md->next) != NULL &&
1490 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1494 mc->pkt_len = pktlen;
1496 /* Allocation of new indirect segment failed */
1497 if (unlikely (mi == NULL)) {
1498 rte_pktmbuf_free(mc);
1502 __rte_mbuf_sanity_check(mc, 1);
1507 * Adds given value to the refcnt of all packet mbuf segments.
1509 * Walks through all segments of given packet mbuf and for each of them
1510 * invokes rte_mbuf_refcnt_update().
1513 * The packet mbuf whose refcnt to be updated.
1515 * The value to add to the mbuf's segments refcnt.
1517 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1519 __rte_mbuf_sanity_check(m, 1);
1522 rte_mbuf_refcnt_update(m, v);
1523 } while ((m = m->next) != NULL);
1527 * Get the headroom in a packet mbuf.
1532 * The length of the headroom.
1534 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1536 __rte_mbuf_sanity_check(m, 0);
1541 * Get the tailroom of a packet mbuf.
1546 * The length of the tailroom.
1548 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1550 __rte_mbuf_sanity_check(m, 0);
1551 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1556 * Get the last segment of the packet.
1561 * The last segment of the given mbuf.
1563 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1565 __rte_mbuf_sanity_check(m, 1);
1566 while (m->next != NULL)
1572 * A macro that points to an offset into the data in the mbuf.
1574 * The returned pointer is cast to type t. Before using this
1575 * function, the user must ensure that the first segment is large
1576 * enough to accommodate its data.
1581 * The offset into the mbuf data.
1583 * The type to cast the result into.
1585 #define rte_pktmbuf_mtod_offset(m, t, o) \
1586 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1589 * A macro that points to the start of the data in the mbuf.
1591 * The returned pointer is cast to type t. Before using this
1592 * function, the user must ensure that the first segment is large
1593 * enough to accommodate its data.
1598 * The type to cast the result into.
1600 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1603 * A macro that returns the IO address that points to an offset of the
1604 * start of the data in the mbuf
1609 * The offset into the data to calculate address from.
1611 #define rte_pktmbuf_iova_offset(m, o) \
1612 (rte_iova_t)((m)->buf_iova + (m)->data_off + (o))
1615 #define rte_pktmbuf_mtophys_offset(m, o) \
1616 rte_pktmbuf_iova_offset(m, o)
1619 * A macro that returns the IO address that points to the start of the
1625 #define rte_pktmbuf_iova(m) rte_pktmbuf_iova_offset(m, 0)
1628 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_iova(m)
1631 * A macro that returns the length of the packet.
1633 * The value can be read or assigned.
1638 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1641 * A macro that returns the length of the segment.
1643 * The value can be read or assigned.
1648 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1651 * Prepend len bytes to an mbuf data area.
1653 * Returns a pointer to the new
1654 * data start address. If there is not enough headroom in the first
1655 * segment, the function will return NULL, without modifying the mbuf.
1660 * The amount of data to prepend (in bytes).
1662 * A pointer to the start of the newly prepended data, or
1663 * NULL if there is not enough headroom space in the first segment
1665 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1668 __rte_mbuf_sanity_check(m, 1);
1670 if (unlikely(len > rte_pktmbuf_headroom(m)))
1674 m->data_len = (uint16_t)(m->data_len + len);
1675 m->pkt_len = (m->pkt_len + len);
1677 return (char *)m->buf_addr + m->data_off;
1681 * Append len bytes to an mbuf.
1683 * Append len bytes to an mbuf and return a pointer to the start address
1684 * of the added data. If there is not enough tailroom in the last
1685 * segment, the function will return NULL, without modifying the mbuf.
1690 * The amount of data to append (in bytes).
1692 * A pointer to the start of the newly appended data, or
1693 * NULL if there is not enough tailroom space in the last segment
1695 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1698 struct rte_mbuf *m_last;
1700 __rte_mbuf_sanity_check(m, 1);
1702 m_last = rte_pktmbuf_lastseg(m);
1703 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1706 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1707 m_last->data_len = (uint16_t)(m_last->data_len + len);
1708 m->pkt_len = (m->pkt_len + len);
1709 return (char*) tail;
1713 * Remove len bytes at the beginning of an mbuf.
1715 * Returns a pointer to the start address of the new data area. If the
1716 * length is greater than the length of the first segment, then the
1717 * function will fail and return NULL, without modifying the mbuf.
1722 * The amount of data to remove (in bytes).
1724 * A pointer to the new start of the data.
1726 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1728 __rte_mbuf_sanity_check(m, 1);
1730 if (unlikely(len > m->data_len))
1733 m->data_len = (uint16_t)(m->data_len - len);
1735 m->pkt_len = (m->pkt_len - len);
1736 return (char *)m->buf_addr + m->data_off;
1740 * Remove len bytes of data at the end of the mbuf.
1742 * If the length is greater than the length of the last segment, the
1743 * function will fail and return -1 without modifying the mbuf.
1748 * The amount of data to remove (in bytes).
1753 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1755 struct rte_mbuf *m_last;
1757 __rte_mbuf_sanity_check(m, 1);
1759 m_last = rte_pktmbuf_lastseg(m);
1760 if (unlikely(len > m_last->data_len))
1763 m_last->data_len = (uint16_t)(m_last->data_len - len);
1764 m->pkt_len = (m->pkt_len - len);
1769 * Test if mbuf data is contiguous.
1774 * - 1, if all data is contiguous (one segment).
1775 * - 0, if there is several segments.
1777 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1779 __rte_mbuf_sanity_check(m, 1);
1780 return !!(m->nb_segs == 1);
1784 * @internal used by rte_pktmbuf_read().
1786 const void *__rte_pktmbuf_read(const struct rte_mbuf *m, uint32_t off,
1787 uint32_t len, void *buf);
1790 * Read len data bytes in a mbuf at specified offset.
1792 * If the data is contiguous, return the pointer in the mbuf data, else
1793 * copy the data in the buffer provided by the user and return its
1797 * The pointer to the mbuf.
1799 * The offset of the data in the mbuf.
1801 * The amount of bytes to read.
1803 * The buffer where data is copied if it is not contiguous in mbuf
1804 * data. Its length should be at least equal to the len parameter.
1806 * The pointer to the data, either in the mbuf if it is contiguous,
1807 * or in the user buffer. If mbuf is too small, NULL is returned.
1809 static inline const void *rte_pktmbuf_read(const struct rte_mbuf *m,
1810 uint32_t off, uint32_t len, void *buf)
1812 if (likely(off + len <= rte_pktmbuf_data_len(m)))
1813 return rte_pktmbuf_mtod_offset(m, char *, off);
1815 return __rte_pktmbuf_read(m, off, len, buf);
1819 * Chain an mbuf to another, thereby creating a segmented packet.
1821 * Note: The implementation will do a linear walk over the segments to find
1822 * the tail entry. For cases when there are many segments, it's better to
1823 * chain the entries manually.
1826 * The head of the mbuf chain (the first packet)
1828 * The mbuf to put last in the chain
1832 * - -EOVERFLOW, if the chain segment limit exceeded
1834 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1836 struct rte_mbuf *cur_tail;
1838 /* Check for number-of-segments-overflow */
1839 if (head->nb_segs + tail->nb_segs > RTE_MBUF_MAX_NB_SEGS)
1842 /* Chain 'tail' onto the old tail */
1843 cur_tail = rte_pktmbuf_lastseg(head);
1844 cur_tail->next = tail;
1846 /* accumulate number of segments and total length. */
1847 head->nb_segs += tail->nb_segs;
1848 head->pkt_len += tail->pkt_len;
1850 /* pkt_len is only set in the head */
1851 tail->pkt_len = tail->data_len;
1857 * Validate general requirements for Tx offload in mbuf.
1859 * This function checks correctness and completeness of Tx offload settings.
1862 * The packet mbuf to be validated.
1864 * 0 if packet is valid
1867 rte_validate_tx_offload(const struct rte_mbuf *m)
1869 uint64_t ol_flags = m->ol_flags;
1870 uint64_t inner_l3_offset = m->l2_len;
1872 /* Does packet set any of available offloads? */
1873 if (!(ol_flags & PKT_TX_OFFLOAD_MASK))
1876 if (ol_flags & PKT_TX_OUTER_IP_CKSUM)
1877 inner_l3_offset += m->outer_l2_len + m->outer_l3_len;
1879 /* Headers are fragmented */
1880 if (rte_pktmbuf_data_len(m) < inner_l3_offset + m->l3_len + m->l4_len)
1883 /* IP checksum can be counted only for IPv4 packet */
1884 if ((ol_flags & PKT_TX_IP_CKSUM) && (ol_flags & PKT_TX_IPV6))
1887 /* IP type not set when required */
1888 if (ol_flags & (PKT_TX_L4_MASK | PKT_TX_TCP_SEG))
1889 if (!(ol_flags & (PKT_TX_IPV4 | PKT_TX_IPV6)))
1892 /* Check requirements for TSO packet */
1893 if (ol_flags & PKT_TX_TCP_SEG)
1894 if ((m->tso_segsz == 0) ||
1895 ((ol_flags & PKT_TX_IPV4) &&
1896 !(ol_flags & PKT_TX_IP_CKSUM)))
1899 /* PKT_TX_OUTER_IP_CKSUM set for non outer IPv4 packet. */
1900 if ((ol_flags & PKT_TX_OUTER_IP_CKSUM) &&
1901 !(ol_flags & PKT_TX_OUTER_IPV4))
1908 * Linearize data in mbuf.
1910 * This function moves the mbuf data in the first segment if there is enough
1911 * tailroom. The subsequent segments are unchained and freed.
1920 rte_pktmbuf_linearize(struct rte_mbuf *mbuf)
1922 int seg_len, copy_len;
1924 struct rte_mbuf *m_next;
1927 if (rte_pktmbuf_is_contiguous(mbuf))
1930 /* Extend first segment to the total packet length */
1931 copy_len = rte_pktmbuf_pkt_len(mbuf) - rte_pktmbuf_data_len(mbuf);
1933 if (unlikely(copy_len > rte_pktmbuf_tailroom(mbuf)))
1936 buffer = rte_pktmbuf_mtod_offset(mbuf, char *, mbuf->data_len);
1937 mbuf->data_len = (uint16_t)(mbuf->pkt_len);
1939 /* Append data from next segments to the first one */
1944 seg_len = rte_pktmbuf_data_len(m);
1945 rte_memcpy(buffer, rte_pktmbuf_mtod(m, char *), seg_len);
1948 rte_pktmbuf_free_seg(m);
1959 * Dump an mbuf structure to a file.
1961 * Dump all fields for the given packet mbuf and all its associated
1962 * segments (in the case of a chained buffer).
1965 * A pointer to a file for output
1969 * If dump_len != 0, also dump the "dump_len" first data bytes of
1972 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1978 #endif /* _RTE_MBUF_H_ */