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 * RX packet is a 802.1q VLAN packet. This flag was set by PMDs when
93 * the packet is recognized as a VLAN, but the behavior between PMDs
94 * was not the same. This flag is kept for some time to avoid breaking
95 * applications and should be replaced by PKT_RX_VLAN_STRIPPED.
97 #define PKT_RX_VLAN_PKT (1ULL << 0)
99 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
100 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
104 * Checking this flag alone is deprecated: check the 2 bits of
105 * PKT_RX_L4_CKSUM_MASK.
106 * This flag was set when the L4 checksum of a packet was detected as
107 * wrong by the hardware.
109 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
113 * Checking this flag alone is deprecated: check the 2 bits of
114 * PKT_RX_IP_CKSUM_MASK.
115 * This flag was set when the IP checksum of a packet was detected as
116 * wrong by the hardware.
118 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
120 #define PKT_RX_EIP_CKSUM_BAD (1ULL << 5) /**< External IP header checksum error. */
123 * A vlan has been stripped by the hardware and its tci is saved in
124 * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
125 * in the RX configuration of the PMD.
127 #define PKT_RX_VLAN_STRIPPED (1ULL << 6)
130 * Mask of bits used to determine the status of RX IP checksum.
131 * - PKT_RX_IP_CKSUM_UNKNOWN: no information about the RX IP checksum
132 * - PKT_RX_IP_CKSUM_BAD: the IP checksum in the packet is wrong
133 * - PKT_RX_IP_CKSUM_GOOD: the IP checksum in the packet is valid
134 * - PKT_RX_IP_CKSUM_NONE: the IP checksum is not correct in the packet
135 * data, but the integrity of the IP header is verified.
137 #define PKT_RX_IP_CKSUM_MASK ((1ULL << 4) | (1ULL << 7))
139 #define PKT_RX_IP_CKSUM_UNKNOWN 0
140 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
141 #define PKT_RX_IP_CKSUM_GOOD (1ULL << 7)
142 #define PKT_RX_IP_CKSUM_NONE ((1ULL << 4) | (1ULL << 7))
145 * Mask of bits used to determine the status of RX L4 checksum.
146 * - PKT_RX_L4_CKSUM_UNKNOWN: no information about the RX L4 checksum
147 * - PKT_RX_L4_CKSUM_BAD: the L4 checksum in the packet is wrong
148 * - PKT_RX_L4_CKSUM_GOOD: the L4 checksum in the packet is valid
149 * - PKT_RX_L4_CKSUM_NONE: the L4 checksum is not correct in the packet
150 * data, but the integrity of the L4 data is verified.
152 #define PKT_RX_L4_CKSUM_MASK ((1ULL << 3) | (1ULL << 8))
154 #define PKT_RX_L4_CKSUM_UNKNOWN 0
155 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
156 #define PKT_RX_L4_CKSUM_GOOD (1ULL << 8)
157 #define PKT_RX_L4_CKSUM_NONE ((1ULL << 3) | (1ULL << 8))
159 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
160 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
161 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
162 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
165 * The 2 vlans have been stripped by the hardware and their tci are
166 * saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
167 * This can only happen if vlan stripping is enabled in the RX
168 * configuration of the PMD. If this flag is set, PKT_RX_VLAN_STRIPPED
171 #define PKT_RX_QINQ_STRIPPED (1ULL << 15)
175 * RX packet with double VLAN stripped.
176 * This flag is replaced by PKT_RX_QINQ_STRIPPED.
178 #define PKT_RX_QINQ_PKT PKT_RX_QINQ_STRIPPED
181 * When packets are coalesced by a hardware or virtual driver, this flag
182 * can be set in the RX mbuf, meaning that the m->tso_segsz field is
183 * valid and is set to the segment size of original packets.
185 #define PKT_RX_LRO (1ULL << 16)
187 /* add new RX flags here */
189 /* add new TX flags here */
192 * Offload the MACsec. This flag must be set by the application to enable
193 * this offload feature for a packet to be transmitted.
195 #define PKT_TX_MACSEC (1ULL << 44)
198 * Bits 45:48 used for the tunnel type.
199 * When doing Tx offload like TSO or checksum, the HW needs to configure the
200 * tunnel type into the HW descriptors.
202 #define PKT_TX_TUNNEL_VXLAN (0x1ULL << 45)
203 #define PKT_TX_TUNNEL_GRE (0x2ULL << 45)
204 #define PKT_TX_TUNNEL_IPIP (0x3ULL << 45)
205 #define PKT_TX_TUNNEL_GENEVE (0x4ULL << 45)
206 /* add new TX TUNNEL type here */
207 #define PKT_TX_TUNNEL_MASK (0xFULL << 45)
210 * Second VLAN insertion (QinQ) flag.
212 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
215 * TCP segmentation offload. To enable this offload feature for a
216 * packet to be transmitted on hardware supporting TSO:
217 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
219 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
220 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
222 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
223 * - calculate the pseudo header checksum without taking ip_len in account,
224 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
225 * rte_ipv6_phdr_cksum() that can be used as helpers.
227 #define PKT_TX_TCP_SEG (1ULL << 50)
229 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
232 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
233 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
234 * L4 checksum offload, the user needs to:
235 * - fill l2_len and l3_len in mbuf
236 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
237 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
238 * - calculate the pseudo header checksum and set it in the L4 header (only
239 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
240 * For SCTP, set the crc field to 0.
242 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
243 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
244 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
245 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
246 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
249 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
250 * also be set by the application, although a PMD will only check
252 * - set the IP checksum field in the packet to 0
253 * - fill the mbuf offload information: l2_len, l3_len
255 #define PKT_TX_IP_CKSUM (1ULL << 54)
258 * Packet is IPv4. This flag must be set when using any offload feature
259 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
260 * packet. If the packet is a tunneled packet, this flag is related to
263 #define PKT_TX_IPV4 (1ULL << 55)
266 * Packet is IPv6. This flag must be set when using an offload feature
267 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
268 * packet. If the packet is a tunneled packet, this flag is related to
271 #define PKT_TX_IPV6 (1ULL << 56)
273 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
276 * Offload the IP checksum of an external header in the hardware. The
277 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
278 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
279 * packet must be set to 0.
280 * - set the outer IP checksum field in the packet to 0
281 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
283 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
286 * Packet outer header is IPv4. This flag must be set when using any
287 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
288 * outer header of the tunneled packet is an IPv4 packet.
290 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
293 * Packet outer header is IPv6. This flag must be set when using any
294 * outer offload feature (L4 checksum) to tell the NIC that the outer
295 * header of the tunneled packet is an IPv6 packet.
297 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
300 * Bitmask of all supported packet Tx offload features flags,
301 * which can be set for packet.
303 #define PKT_TX_OFFLOAD_MASK ( \
306 PKT_TX_OUTER_IP_CKSUM | \
308 PKT_TX_IEEE1588_TMST | \
311 PKT_TX_TUNNEL_MASK | \
314 #define __RESERVED (1ULL << 61) /**< reserved for future mbuf use */
316 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
318 /* Use final bit of flags to indicate a control mbuf */
319 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
321 /** Alignment constraint of mbuf private area. */
322 #define RTE_MBUF_PRIV_ALIGN 8
325 * Get the name of a RX offload flag
328 * The mask describing the flag.
330 * The name of this flag, or NULL if it's not a valid RX flag.
332 const char *rte_get_rx_ol_flag_name(uint64_t mask);
335 * Dump the list of RX offload flags in a buffer
338 * The mask describing the RX flags.
342 * The length of the buffer.
344 * 0 on success, (-1) on error.
346 int rte_get_rx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
349 * Get the name of a TX offload flag
352 * The mask describing the flag. Usually only one bit must be set.
353 * Several bits can be given if they belong to the same mask.
354 * Ex: PKT_TX_L4_MASK.
356 * The name of this flag, or NULL if it's not a valid TX flag.
358 const char *rte_get_tx_ol_flag_name(uint64_t mask);
361 * Dump the list of TX offload flags in a buffer
364 * The mask describing the TX flags.
368 * The length of the buffer.
370 * 0 on success, (-1) on error.
372 int rte_get_tx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
375 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
376 * splitting it into multiple segments.
377 * So, for mbufs that planned to be involved into RX/TX, the recommended
378 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
380 #define RTE_MBUF_DEFAULT_DATAROOM 2048
381 #define RTE_MBUF_DEFAULT_BUF_SIZE \
382 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
384 /* define a set of marker types that can be used to refer to set points in the
387 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
389 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
391 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
392 * with a single assignment */
395 * The generic rte_mbuf, containing a packet mbuf.
400 void *buf_addr; /**< Virtual address of segment buffer. */
401 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
403 /* next 8 bytes are initialised on RX descriptor rearm */
408 * Reference counter. Its size should at least equal to the size
409 * of port field (16 bits), to support zero-copy broadcast.
410 * It should only be accessed using the following functions:
411 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
412 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
413 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
418 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
419 uint16_t refcnt; /**< Non-atomically accessed refcnt */
421 uint16_t nb_segs; /**< Number of segments. */
423 /** Input port (16 bits to support more than 256 virtual ports). */
426 uint64_t ol_flags; /**< Offload features. */
428 /* remaining bytes are set on RX when pulling packet from descriptor */
429 MARKER rx_descriptor_fields1;
432 * The packet type, which is the combination of outer/inner L2, L3, L4
433 * and tunnel types. The packet_type is about data really present in the
434 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
435 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
436 * vlan is stripped from the data.
440 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
442 uint32_t l2_type:4; /**< (Outer) L2 type. */
443 uint32_t l3_type:4; /**< (Outer) L3 type. */
444 uint32_t l4_type:4; /**< (Outer) L4 type. */
445 uint32_t tun_type:4; /**< Tunnel type. */
446 uint32_t inner_l2_type:4; /**< Inner L2 type. */
447 uint32_t inner_l3_type:4; /**< Inner L3 type. */
448 uint32_t inner_l4_type:4; /**< Inner L4 type. */
452 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
453 uint16_t data_len; /**< Amount of data in segment buffer. */
454 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN_STRIPPED is set. */
458 uint32_t rss; /**< RSS hash result if RSS enabled */
467 /**< Second 4 flexible bytes */
470 /**< First 4 flexible bytes or FD ID, dependent on
471 PKT_RX_FDIR_* flag in ol_flags. */
472 } fdir; /**< Filter identifier if FDIR enabled */
476 } sched; /**< Hierarchical scheduler */
477 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
478 } hash; /**< hash information */
480 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
482 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ_STRIPPED is set. */
483 uint16_t vlan_tci_outer;
485 uint16_t buf_len; /**< Length of segment buffer. */
486 /* second cache line - fields only used in slow path or on TX */
487 MARKER cacheline1 __rte_cache_min_aligned;
491 void *userdata; /**< Can be used for external metadata */
492 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
495 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
496 struct rte_mbuf *next; /**< Next segment of scattered packet. */
498 /* fields to support TX offloads */
501 uint64_t tx_offload; /**< combined for easy fetch */
505 /**< L2 (MAC) Header Length for non-tunneling pkt.
506 * Outer_L4_len + ... + Inner_L2_len for tunneling pkt.
508 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
509 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
510 uint64_t tso_segsz:16; /**< TCP TSO segment size */
512 /* fields for TX offloading of tunnels */
513 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
514 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
516 /* uint64_t unused:8; */
520 /** Size of the application private data. In case of an indirect
521 * mbuf, it stores the direct mbuf private data size. */
524 /** Timesync flags for use with IEEE1588. */
526 } __rte_cache_aligned;
529 * Prefetch the first part of the mbuf
531 * The first 64 bytes of the mbuf corresponds to fields that are used early
532 * in the receive path. If the cache line of the architecture is higher than
533 * 64B, the second part will also be prefetched.
536 * The pointer to the mbuf.
539 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
541 rte_prefetch0(&m->cacheline0);
545 * Prefetch the second part of the mbuf
547 * The next 64 bytes of the mbuf corresponds to fields that are used in the
548 * transmit path. If the cache line of the architecture is higher than 64B,
549 * this function does nothing as it is expected that the full mbuf is
553 * The pointer to the mbuf.
556 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
558 #if RTE_CACHE_LINE_SIZE == 64
559 rte_prefetch0(&m->cacheline1);
566 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
569 * Return the DMA address of the beginning of the mbuf data
572 * The pointer to the mbuf.
574 * The physical address of the beginning of the mbuf data
576 static inline phys_addr_t
577 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
579 return mb->buf_physaddr + mb->data_off;
583 * Return the default DMA address of the beginning of the mbuf data
585 * This function is used by drivers in their receive function, as it
586 * returns the location where data should be written by the NIC, taking
587 * the default headroom in account.
590 * The pointer to the mbuf.
592 * The physical address of the beginning of the mbuf data
594 static inline phys_addr_t
595 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
597 return mb->buf_physaddr + RTE_PKTMBUF_HEADROOM;
601 * Return the mbuf owning the data buffer address of an indirect mbuf.
604 * The pointer to the indirect mbuf.
606 * The address of the direct mbuf corresponding to buffer_addr.
608 static inline struct rte_mbuf *
609 rte_mbuf_from_indirect(struct rte_mbuf *mi)
611 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
615 * Return the buffer address embedded in the given mbuf.
618 * The pointer to the mbuf.
620 * The address of the data buffer owned by the mbuf.
623 rte_mbuf_to_baddr(struct rte_mbuf *md)
626 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
631 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
633 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
636 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
638 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
641 * Private data in case of pktmbuf pool.
643 * A structure that contains some pktmbuf_pool-specific data that are
644 * appended after the mempool structure (in private data).
646 struct rte_pktmbuf_pool_private {
647 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
648 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
651 #ifdef RTE_LIBRTE_MBUF_DEBUG
653 /** check mbuf type in debug mode */
654 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
656 #else /* RTE_LIBRTE_MBUF_DEBUG */
658 /** check mbuf type in debug mode */
659 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
661 #endif /* RTE_LIBRTE_MBUF_DEBUG */
663 #ifdef RTE_MBUF_REFCNT_ATOMIC
666 * Reads the value of an mbuf's refcnt.
670 * Reference count number.
672 static inline uint16_t
673 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
675 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
679 * Sets an mbuf's refcnt to a defined value.
686 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
688 rte_atomic16_set(&m->refcnt_atomic, new_value);
692 * Adds given value to an mbuf's refcnt and returns its new value.
696 * Value to add/subtract
700 static inline uint16_t
701 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
704 * The atomic_add is an expensive operation, so we don't want to
705 * call it in the case where we know we are the uniq holder of
706 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
707 * operation has to be used because concurrent accesses on the
708 * reference counter can occur.
710 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
711 rte_mbuf_refcnt_set(m, 1 + value);
715 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
718 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
721 * Adds given value to an mbuf's refcnt and returns its new value.
723 static inline uint16_t
724 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
726 m->refcnt = (uint16_t)(m->refcnt + value);
731 * Reads the value of an mbuf's refcnt.
733 static inline uint16_t
734 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
740 * Sets an mbuf's refcnt to the defined value.
743 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
745 m->refcnt = new_value;
748 #endif /* RTE_MBUF_REFCNT_ATOMIC */
751 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
758 * Sanity checks on an mbuf.
760 * Check the consistency of the given mbuf. The function will cause a
761 * panic if corruption is detected.
764 * The mbuf to be checked.
766 * True if the mbuf is a packet header, false if it is a sub-segment
767 * of a packet (in this case, some fields like nb_segs are not checked)
770 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
773 * Allocate an unitialized mbuf from mempool *mp*.
775 * This function can be used by PMDs (especially in RX functions) to
776 * allocate an unitialized mbuf. The driver is responsible of
777 * initializing all the required fields. See rte_pktmbuf_reset().
778 * For standard needs, prefer rte_pktmbuf_alloc().
780 * The caller can expect that the following fields of the mbuf structure
781 * are initialized: buf_addr, buf_physaddr, buf_len, refcnt=1, nb_segs=1,
782 * next=NULL, pool, priv_size. The other fields must be initialized
786 * The mempool from which mbuf is allocated.
788 * - The pointer to the new mbuf on success.
789 * - NULL if allocation failed.
791 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
796 if (rte_mempool_get(mp, &mb) < 0)
798 m = (struct rte_mbuf *)mb;
799 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
800 RTE_ASSERT(m->next == NULL);
801 RTE_ASSERT(m->nb_segs == 1);
802 __rte_mbuf_sanity_check(m, 0);
808 * Put mbuf back into its original mempool.
810 * The caller must ensure that the mbuf is direct and properly
811 * reinitialized (refcnt=1, next=NULL, nb_segs=1), as done by
812 * rte_pktmbuf_prefree_seg().
814 * This function should be used with care, when optimization is
815 * required. For standard needs, prefer rte_pktmbuf_free() or
816 * rte_pktmbuf_free_seg().
819 * The mbuf to be freed.
821 static inline void __attribute__((always_inline))
822 rte_mbuf_raw_free(struct rte_mbuf *m)
824 RTE_ASSERT(RTE_MBUF_DIRECT(m));
825 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
826 RTE_ASSERT(m->next == NULL);
827 RTE_ASSERT(m->nb_segs == 1);
828 __rte_mbuf_sanity_check(m, 0);
829 rte_mempool_put(m->pool, m);
832 /* compat with older versions */
835 __rte_mbuf_raw_free(struct rte_mbuf *m)
837 rte_mbuf_raw_free(m);
840 /* Operations on ctrl mbuf */
843 * The control mbuf constructor.
845 * This function initializes some fields in an mbuf structure that are
846 * not modified by the user once created (mbuf type, origin pool, buffer
847 * start address, and so on). This function is given as a callback function
848 * to rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
851 * The mempool from which the mbuf is allocated.
853 * A pointer that can be used by the user to retrieve useful information
854 * for mbuf initialization. This pointer is the opaque argument passed to
855 * rte_mempool_obj_iter() or rte_mempool_create().
857 * The mbuf to initialize.
859 * The index of the mbuf in the pool table.
861 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
862 void *m, unsigned i);
865 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
867 * This new mbuf is initialized with data pointing to the beginning of
868 * buffer, and with a length of zero.
871 * The mempool from which the mbuf is allocated.
873 * - The pointer to the new mbuf on success.
874 * - NULL if allocation failed.
876 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
879 * Free a control mbuf back into its original mempool.
882 * The control mbuf to be freed.
884 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
887 * A macro that returns the pointer to the carried data.
889 * The value that can be read or assigned.
894 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
897 * A macro that returns the length of the carried data.
899 * The value that can be read or assigned.
904 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
907 * Tests if an mbuf is a control mbuf
910 * The mbuf to be tested
912 * - True (1) if the mbuf is a control mbuf
913 * - False(0) otherwise
916 rte_is_ctrlmbuf(struct rte_mbuf *m)
918 return !!(m->ol_flags & CTRL_MBUF_FLAG);
921 /* Operations on pkt mbuf */
924 * The packet mbuf constructor.
926 * This function initializes some fields in the mbuf structure that are
927 * not modified by the user once created (origin pool, buffer start
928 * address, and so on). This function is given as a callback function to
929 * rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
932 * The mempool from which mbufs originate.
934 * A pointer that can be used by the user to retrieve useful information
935 * for mbuf initialization. This pointer is the opaque argument passed to
936 * rte_mempool_obj_iter() or rte_mempool_create().
938 * The mbuf to initialize.
940 * The index of the mbuf in the pool table.
942 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
943 void *m, unsigned i);
947 * A packet mbuf pool constructor.
949 * This function initializes the mempool private data in the case of a
950 * pktmbuf pool. This private data is needed by the driver. The
951 * function must be called on the mempool before it is used, or it
952 * can be given as a callback function to rte_mempool_create() at
953 * pool creation. It can be extended by the user, for example, to
954 * provide another packet size.
957 * The mempool from which mbufs originate.
959 * A pointer that can be used by the user to retrieve useful information
960 * for mbuf initialization. This pointer is the opaque argument passed to
961 * rte_mempool_create().
963 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
966 * Create a mbuf pool.
968 * This function creates and initializes a packet mbuf pool. It is
969 * a wrapper to rte_mempool functions.
972 * The name of the mbuf pool.
974 * The number of elements in the mbuf pool. The optimum size (in terms
975 * of memory usage) for a mempool is when n is a power of two minus one:
978 * Size of the per-core object cache. See rte_mempool_create() for
981 * Size of application private are between the rte_mbuf structure
982 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
983 * @param data_room_size
984 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
986 * The socket identifier where the memory should be allocated. The
987 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
990 * The pointer to the new allocated mempool, on success. NULL on error
991 * with rte_errno set appropriately. Possible rte_errno values include:
992 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
993 * - E_RTE_SECONDARY - function was called from a secondary process instance
994 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
995 * - ENOSPC - the maximum number of memzones has already been allocated
996 * - EEXIST - a memzone with the same name already exists
997 * - ENOMEM - no appropriate memory area found in which to create memzone
1000 rte_pktmbuf_pool_create(const char *name, unsigned n,
1001 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1005 * Get the data room size of mbufs stored in a pktmbuf_pool
1007 * The data room size is the amount of data that can be stored in a
1008 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1011 * The packet mbuf pool.
1013 * The data room size of mbufs stored in this mempool.
1015 static inline uint16_t
1016 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1018 struct rte_pktmbuf_pool_private *mbp_priv;
1020 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1021 return mbp_priv->mbuf_data_room_size;
1025 * Get the application private size of mbufs stored in a pktmbuf_pool
1027 * The private size of mbuf is a zone located between the rte_mbuf
1028 * structure and the data buffer where an application can store data
1029 * associated to a packet.
1032 * The packet mbuf pool.
1034 * The private size of mbufs stored in this mempool.
1036 static inline uint16_t
1037 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1039 struct rte_pktmbuf_pool_private *mbp_priv;
1041 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1042 return mbp_priv->mbuf_priv_size;
1046 * Reset the data_off field of a packet mbuf to its default value.
1048 * The given mbuf must have only one segment, which should be empty.
1051 * The packet mbuf's data_off field has to be reset.
1053 static inline void rte_pktmbuf_reset_headroom(struct rte_mbuf *m)
1055 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1059 * Reset the fields of a packet mbuf to their default values.
1061 * The given mbuf must have only one segment.
1064 * The packet mbuf to be resetted.
1066 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1072 m->vlan_tci_outer = 0;
1078 rte_pktmbuf_reset_headroom(m);
1081 __rte_mbuf_sanity_check(m, 1);
1085 * Allocate a new mbuf from a mempool.
1087 * This new mbuf contains one segment, which has a length of 0. The pointer
1088 * to data is initialized to have some bytes of headroom in the buffer
1089 * (if buffer size allows).
1092 * The mempool from which the mbuf is allocated.
1094 * - The pointer to the new mbuf on success.
1095 * - NULL if allocation failed.
1097 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1100 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1101 rte_pktmbuf_reset(m);
1106 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1110 * The mempool from which mbufs are allocated.
1112 * Array of pointers to mbufs
1118 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1119 struct rte_mbuf **mbufs, unsigned count)
1124 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1128 /* To understand duff's device on loop unwinding optimization, see
1129 * https://en.wikipedia.org/wiki/Duff's_device.
1130 * Here while() loop is used rather than do() while{} to avoid extra
1131 * check if count is zero.
1133 switch (count % 4) {
1135 while (idx != count) {
1136 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1137 rte_mbuf_refcnt_set(mbufs[idx], 1);
1138 rte_pktmbuf_reset(mbufs[idx]);
1141 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1142 rte_mbuf_refcnt_set(mbufs[idx], 1);
1143 rte_pktmbuf_reset(mbufs[idx]);
1146 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1147 rte_mbuf_refcnt_set(mbufs[idx], 1);
1148 rte_pktmbuf_reset(mbufs[idx]);
1151 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1152 rte_mbuf_refcnt_set(mbufs[idx], 1);
1153 rte_pktmbuf_reset(mbufs[idx]);
1161 * Attach packet mbuf to another packet mbuf.
1163 * After attachment we refer the mbuf we attached as 'indirect',
1164 * while mbuf we attached to as 'direct'.
1165 * The direct mbuf's reference counter is incremented.
1167 * Right now, not supported:
1168 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1169 * - mbuf we trying to attach (mi) is used by someone else
1170 * e.g. it's reference counter is greater then 1.
1173 * The indirect packet mbuf.
1175 * The packet mbuf we're attaching to.
1177 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1179 struct rte_mbuf *md;
1181 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1182 rte_mbuf_refcnt_read(mi) == 1);
1184 /* if m is not direct, get the mbuf that embeds the data */
1185 if (RTE_MBUF_DIRECT(m))
1188 md = rte_mbuf_from_indirect(m);
1190 rte_mbuf_refcnt_update(md, 1);
1191 mi->priv_size = m->priv_size;
1192 mi->buf_physaddr = m->buf_physaddr;
1193 mi->buf_addr = m->buf_addr;
1194 mi->buf_len = m->buf_len;
1196 mi->data_off = m->data_off;
1197 mi->data_len = m->data_len;
1199 mi->vlan_tci = m->vlan_tci;
1200 mi->vlan_tci_outer = m->vlan_tci_outer;
1201 mi->tx_offload = m->tx_offload;
1205 mi->pkt_len = mi->data_len;
1207 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1208 mi->packet_type = m->packet_type;
1210 __rte_mbuf_sanity_check(mi, 1);
1211 __rte_mbuf_sanity_check(m, 0);
1215 * Detach an indirect packet mbuf.
1217 * - restore original mbuf address and length values.
1218 * - reset pktmbuf data and data_len to their default values.
1219 * - decrement the direct mbuf's reference counter. When the
1220 * reference counter becomes 0, the direct mbuf is freed.
1222 * All other fields of the given packet mbuf will be left intact.
1225 * The indirect attached packet mbuf.
1227 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1229 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1230 struct rte_mempool *mp = m->pool;
1231 uint32_t mbuf_size, buf_len, priv_size;
1233 priv_size = rte_pktmbuf_priv_size(mp);
1234 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1235 buf_len = rte_pktmbuf_data_room_size(mp);
1237 m->priv_size = priv_size;
1238 m->buf_addr = (char *)m + mbuf_size;
1239 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
1240 m->buf_len = (uint16_t)buf_len;
1241 rte_pktmbuf_reset_headroom(m);
1245 if (rte_mbuf_refcnt_update(md, -1) == 0) {
1248 rte_mbuf_refcnt_set(md, 1);
1249 rte_mbuf_raw_free(md);
1254 * Decrease reference counter and unlink a mbuf segment
1256 * This function does the same than a free, except that it does not
1257 * return the segment to its pool.
1258 * It decreases the reference counter, and if it reaches 0, it is
1259 * detached from its parent for an indirect mbuf.
1262 * The mbuf to be unlinked
1264 * - (m) if it is the last reference. It can be recycled or freed.
1265 * - (NULL) if the mbuf still has remaining references on it.
1267 __attribute__((always_inline))
1268 static inline struct rte_mbuf *
1269 rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1271 __rte_mbuf_sanity_check(m, 0);
1273 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1275 if (RTE_MBUF_INDIRECT(m))
1276 rte_pktmbuf_detach(m);
1278 if (m->next != NULL) {
1285 } else if (rte_atomic16_add_return(&m->refcnt_atomic, -1) == 0) {
1288 if (RTE_MBUF_INDIRECT(m))
1289 rte_pktmbuf_detach(m);
1291 if (m->next != NULL) {
1295 rte_mbuf_refcnt_set(m, 1);
1302 /* deprecated, replaced by rte_pktmbuf_prefree_seg() */
1304 static inline struct rte_mbuf *
1305 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1307 return rte_pktmbuf_prefree_seg(m);
1311 * Free a segment of a packet mbuf into its original mempool.
1313 * Free an mbuf, without parsing other segments in case of chained
1317 * The packet mbuf segment to be freed.
1319 static inline void __attribute__((always_inline))
1320 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1322 m = rte_pktmbuf_prefree_seg(m);
1323 if (likely(m != NULL))
1324 rte_mbuf_raw_free(m);
1328 * Free a packet mbuf back into its original mempool.
1330 * Free an mbuf, and all its segments in case of chained buffers. Each
1331 * segment is added back into its original mempool.
1334 * The packet mbuf to be freed.
1336 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1338 struct rte_mbuf *m_next;
1340 __rte_mbuf_sanity_check(m, 1);
1344 rte_pktmbuf_free_seg(m);
1350 * Creates a "clone" of the given packet mbuf.
1352 * Walks through all segments of the given packet mbuf, and for each of them:
1353 * - Creates a new packet mbuf from the given pool.
1354 * - Attaches newly created mbuf to the segment.
1355 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1356 * from the original packet mbuf.
1359 * The packet mbuf to be cloned.
1361 * The mempool from which the "clone" mbufs are allocated.
1363 * - The pointer to the new "clone" mbuf on success.
1364 * - NULL if allocation fails.
1366 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1367 struct rte_mempool *mp)
1369 struct rte_mbuf *mc, *mi, **prev;
1373 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1378 pktlen = md->pkt_len;
1383 rte_pktmbuf_attach(mi, md);
1386 } while ((md = md->next) != NULL &&
1387 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1391 mc->pkt_len = pktlen;
1393 /* Allocation of new indirect segment failed */
1394 if (unlikely (mi == NULL)) {
1395 rte_pktmbuf_free(mc);
1399 __rte_mbuf_sanity_check(mc, 1);
1404 * Adds given value to the refcnt of all packet mbuf segments.
1406 * Walks through all segments of given packet mbuf and for each of them
1407 * invokes rte_mbuf_refcnt_update().
1410 * The packet mbuf whose refcnt to be updated.
1412 * The value to add to the mbuf's segments refcnt.
1414 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1416 __rte_mbuf_sanity_check(m, 1);
1419 rte_mbuf_refcnt_update(m, v);
1420 } while ((m = m->next) != NULL);
1424 * Get the headroom in a packet mbuf.
1429 * The length of the headroom.
1431 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1433 __rte_mbuf_sanity_check(m, 1);
1438 * Get the tailroom of a packet mbuf.
1443 * The length of the tailroom.
1445 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1447 __rte_mbuf_sanity_check(m, 1);
1448 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1453 * Get the last segment of the packet.
1458 * The last segment of the given mbuf.
1460 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1462 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1464 __rte_mbuf_sanity_check(m, 1);
1465 while (m2->next != NULL)
1471 * A macro that points to an offset into the data in the mbuf.
1473 * The returned pointer is cast to type t. Before using this
1474 * function, the user must ensure that the first segment is large
1475 * enough to accommodate its data.
1480 * The offset into the mbuf data.
1482 * The type to cast the result into.
1484 #define rte_pktmbuf_mtod_offset(m, t, o) \
1485 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1488 * A macro that points to the start of the data in the mbuf.
1490 * The returned pointer is cast to type t. Before using this
1491 * function, the user must ensure that the first segment is large
1492 * enough to accommodate its data.
1497 * The type to cast the result into.
1499 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1502 * A macro that returns the physical address that points to an offset of the
1503 * start of the data in the mbuf
1508 * The offset into the data to calculate address from.
1510 #define rte_pktmbuf_mtophys_offset(m, o) \
1511 (phys_addr_t)((m)->buf_physaddr + (m)->data_off + (o))
1514 * A macro that returns the physical address that points to the start of the
1520 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_mtophys_offset(m, 0)
1523 * A macro that returns the length of the packet.
1525 * The value can be read or assigned.
1530 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1533 * A macro that returns the length of the segment.
1535 * The value can be read or assigned.
1540 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1543 * Prepend len bytes to an mbuf data area.
1545 * Returns a pointer to the new
1546 * data start address. If there is not enough headroom in the first
1547 * segment, the function will return NULL, without modifying the mbuf.
1552 * The amount of data to prepend (in bytes).
1554 * A pointer to the start of the newly prepended data, or
1555 * NULL if there is not enough headroom space in the first segment
1557 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1560 __rte_mbuf_sanity_check(m, 1);
1562 if (unlikely(len > rte_pktmbuf_headroom(m)))
1566 m->data_len = (uint16_t)(m->data_len + len);
1567 m->pkt_len = (m->pkt_len + len);
1569 return (char *)m->buf_addr + m->data_off;
1573 * Append len bytes to an mbuf.
1575 * Append len bytes to an mbuf and return a pointer to the start address
1576 * of the added data. If there is not enough tailroom in the last
1577 * segment, the function will return NULL, without modifying the mbuf.
1582 * The amount of data to append (in bytes).
1584 * A pointer to the start of the newly appended data, or
1585 * NULL if there is not enough tailroom space in the last segment
1587 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1590 struct rte_mbuf *m_last;
1592 __rte_mbuf_sanity_check(m, 1);
1594 m_last = rte_pktmbuf_lastseg(m);
1595 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1598 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1599 m_last->data_len = (uint16_t)(m_last->data_len + len);
1600 m->pkt_len = (m->pkt_len + len);
1601 return (char*) tail;
1605 * Remove len bytes at the beginning of an mbuf.
1607 * Returns a pointer to the start address of the new data area. If the
1608 * length is greater than the length of the first segment, then the
1609 * function will fail and return NULL, without modifying the mbuf.
1614 * The amount of data to remove (in bytes).
1616 * A pointer to the new start of the data.
1618 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1620 __rte_mbuf_sanity_check(m, 1);
1622 if (unlikely(len > m->data_len))
1625 m->data_len = (uint16_t)(m->data_len - len);
1627 m->pkt_len = (m->pkt_len - len);
1628 return (char *)m->buf_addr + m->data_off;
1632 * Remove len bytes of data at the end of the mbuf.
1634 * If the length is greater than the length of the last segment, the
1635 * function will fail and return -1 without modifying the mbuf.
1640 * The amount of data to remove (in bytes).
1645 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1647 struct rte_mbuf *m_last;
1649 __rte_mbuf_sanity_check(m, 1);
1651 m_last = rte_pktmbuf_lastseg(m);
1652 if (unlikely(len > m_last->data_len))
1655 m_last->data_len = (uint16_t)(m_last->data_len - len);
1656 m->pkt_len = (m->pkt_len - len);
1661 * Test if mbuf data is contiguous.
1666 * - 1, if all data is contiguous (one segment).
1667 * - 0, if there is several segments.
1669 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1671 __rte_mbuf_sanity_check(m, 1);
1672 return !!(m->nb_segs == 1);
1676 * @internal used by rte_pktmbuf_read().
1678 const void *__rte_pktmbuf_read(const struct rte_mbuf *m, uint32_t off,
1679 uint32_t len, void *buf);
1682 * Read len data bytes in a mbuf at specified offset.
1684 * If the data is contiguous, return the pointer in the mbuf data, else
1685 * copy the data in the buffer provided by the user and return its
1689 * The pointer to the mbuf.
1691 * The offset of the data in the mbuf.
1693 * The amount of bytes to read.
1695 * The buffer where data is copied if it is not contigous in mbuf
1696 * data. Its length should be at least equal to the len parameter.
1698 * The pointer to the data, either in the mbuf if it is contiguous,
1699 * or in the user buffer. If mbuf is too small, NULL is returned.
1701 static inline const void *rte_pktmbuf_read(const struct rte_mbuf *m,
1702 uint32_t off, uint32_t len, void *buf)
1704 if (likely(off + len <= rte_pktmbuf_data_len(m)))
1705 return rte_pktmbuf_mtod_offset(m, char *, off);
1707 return __rte_pktmbuf_read(m, off, len, buf);
1711 * Chain an mbuf to another, thereby creating a segmented packet.
1713 * Note: The implementation will do a linear walk over the segments to find
1714 * the tail entry. For cases when there are many segments, it's better to
1715 * chain the entries manually.
1718 * The head of the mbuf chain (the first packet)
1720 * The mbuf to put last in the chain
1724 * - -EOVERFLOW, if the chain is full (256 entries)
1726 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1728 struct rte_mbuf *cur_tail;
1730 /* Check for number-of-segments-overflow */
1731 if (head->nb_segs + tail->nb_segs >= 1 << (sizeof(head->nb_segs) * 8))
1734 /* Chain 'tail' onto the old tail */
1735 cur_tail = rte_pktmbuf_lastseg(head);
1736 cur_tail->next = tail;
1738 /* accumulate number of segments and total length. */
1739 head->nb_segs = (uint8_t)(head->nb_segs + tail->nb_segs);
1740 head->pkt_len += tail->pkt_len;
1742 /* pkt_len is only set in the head */
1743 tail->pkt_len = tail->data_len;
1749 * Validate general requirements for Tx offload in mbuf.
1751 * This function checks correctness and completeness of Tx offload settings.
1754 * The packet mbuf to be validated.
1756 * 0 if packet is valid
1759 rte_validate_tx_offload(const struct rte_mbuf *m)
1761 uint64_t ol_flags = m->ol_flags;
1762 uint64_t inner_l3_offset = m->l2_len;
1764 /* Does packet set any of available offloads? */
1765 if (!(ol_flags & PKT_TX_OFFLOAD_MASK))
1768 if (ol_flags & PKT_TX_OUTER_IP_CKSUM)
1769 inner_l3_offset += m->outer_l2_len + m->outer_l3_len;
1771 /* Headers are fragmented */
1772 if (rte_pktmbuf_data_len(m) < inner_l3_offset + m->l3_len + m->l4_len)
1775 /* IP checksum can be counted only for IPv4 packet */
1776 if ((ol_flags & PKT_TX_IP_CKSUM) && (ol_flags & PKT_TX_IPV6))
1779 /* IP type not set when required */
1780 if (ol_flags & (PKT_TX_L4_MASK | PKT_TX_TCP_SEG))
1781 if (!(ol_flags & (PKT_TX_IPV4 | PKT_TX_IPV6)))
1784 /* Check requirements for TSO packet */
1785 if (ol_flags & PKT_TX_TCP_SEG)
1786 if ((m->tso_segsz == 0) ||
1787 ((ol_flags & PKT_TX_IPV4) &&
1788 !(ol_flags & PKT_TX_IP_CKSUM)))
1791 /* PKT_TX_OUTER_IP_CKSUM set for non outer IPv4 packet. */
1792 if ((ol_flags & PKT_TX_OUTER_IP_CKSUM) &&
1793 !(ol_flags & PKT_TX_OUTER_IPV4))
1800 * Linearize data in mbuf.
1802 * This function moves the mbuf data in the first segment if there is enough
1803 * tailroom. The subsequent segments are unchained and freed.
1812 rte_pktmbuf_linearize(struct rte_mbuf *mbuf)
1814 int seg_len, copy_len;
1816 struct rte_mbuf *m_next;
1819 if (rte_pktmbuf_is_contiguous(mbuf))
1822 /* Extend first segment to the total packet length */
1823 copy_len = rte_pktmbuf_pkt_len(mbuf) - rte_pktmbuf_data_len(mbuf);
1825 if (unlikely(copy_len > rte_pktmbuf_tailroom(mbuf)))
1828 buffer = rte_pktmbuf_mtod_offset(mbuf, char *, mbuf->data_len);
1829 mbuf->data_len = (uint16_t)(mbuf->pkt_len);
1831 /* Append data from next segments to the first one */
1836 seg_len = rte_pktmbuf_data_len(m);
1837 rte_memcpy(buffer, rte_pktmbuf_mtod(m, char *), seg_len);
1840 rte_pktmbuf_free_seg(m);
1851 * Dump an mbuf structure to a file.
1853 * Dump all fields for the given packet mbuf and all its associated
1854 * segments (in the case of a chained buffer).
1857 * A pointer to a file for output
1861 * If dump_len != 0, also dump the "dump_len" first data bytes of
1864 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1870 #endif /* _RTE_MBUF_H_ */