4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
5 * Copyright 2014 6WIND S.A.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * * Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * * Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * * Neither the name of Intel Corporation nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
25 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
26 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
42 * The mbuf library provides the ability to create and destroy buffers
43 * that may be used by the RTE application to store message
44 * buffers. The message buffers are stored in a mempool, using the
45 * RTE mempool library.
47 * This library provide an API to allocate/free packet mbufs, which are
48 * used to carry network packets.
50 * To understand the concepts of packet buffers or mbufs, you
51 * should read "TCP/IP Illustrated, Volume 2: The Implementation,
52 * Addison-Wesley, 1995, ISBN 0-201-63354-X from Richard Stevens"
53 * http://www.kohala.com/start/tcpipiv2.html
57 #include <rte_mempool.h>
58 #include <rte_memory.h>
59 #include <rte_atomic.h>
60 #include <rte_prefetch.h>
61 #include <rte_branch_prediction.h>
67 /* deprecated feature, renamed in RTE_MBUF_REFCNT */
68 #pragma GCC poison RTE_MBUF_SCATTER_GATHER
71 * Packet Offload Features Flags. It also carry packet type information.
72 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
74 * - RX flags start at bit position zero, and get added to the left of previous
76 * - The most-significant 8 bits are reserved for generic mbuf flags
77 * - TX flags therefore start at bit position 55 (i.e. 63-8), and new flags get
78 * added to the right of the previously defined flags
80 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
81 * rte_get_tx_ol_flag_name().
83 #define PKT_RX_VLAN_PKT (1ULL << 0) /**< RX packet is a 802.1q VLAN packet. */
84 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
85 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
86 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
87 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
88 #define PKT_RX_EIP_CKSUM_BAD (0ULL << 0) /**< External IP header checksum error. */
89 #define PKT_RX_OVERSIZE (0ULL << 0) /**< Num of desc of an RX pkt oversize. */
90 #define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
91 #define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
92 #define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
93 #define PKT_RX_IPV4_HDR (1ULL << 5) /**< RX packet with IPv4 header. */
94 #define PKT_RX_IPV4_HDR_EXT (1ULL << 6) /**< RX packet with extended IPv4 header. */
95 #define PKT_RX_IPV6_HDR (1ULL << 7) /**< RX packet with IPv6 header. */
96 #define PKT_RX_IPV6_HDR_EXT (1ULL << 8) /**< RX packet with extended IPv6 header. */
97 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
98 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
99 #define PKT_RX_TUNNEL_IPV4_HDR (1ULL << 11) /**< RX tunnel packet with IPv4 header.*/
100 #define PKT_RX_TUNNEL_IPV6_HDR (1ULL << 12) /**< RX tunnel packet with IPv6 header. */
101 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
102 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
103 /* add new RX flags here */
105 /* add new TX flags here */
108 * TCP segmentation offload. To enable this offload feature for a
109 * packet to be transmitted on hardware supporting TSO:
110 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
112 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
113 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
115 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
116 * - calculate the pseudo header checksum without taking ip_len in account,
117 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
118 * rte_ipv6_phdr_cksum() that can be used as helpers.
120 #define PKT_TX_TCP_SEG (1ULL << 49)
122 /** TX packet is an UDP tunneled packet. It must be specified when using
123 * outer checksum offload (PKT_TX_OUTER_IP_CKSUM) */
124 #define PKT_TX_UDP_TUNNEL_PKT (1ULL << 50) /**< TX packet is an UDP tunneled packet */
125 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
128 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
129 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
130 * L4 checksum offload, the user needs to:
131 * - fill l2_len and l3_len in mbuf
132 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
133 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
134 * - calculate the pseudo header checksum and set it in the L4 header (only
135 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
136 * For SCTP, set the crc field to 0.
138 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
139 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
140 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
141 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
142 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
144 #define PKT_TX_IP_CKSUM (1ULL << 54) /**< IP cksum of TX pkt. computed by NIC. */
145 #define PKT_TX_IPV4_CSUM PKT_TX_IP_CKSUM /**< Alias of PKT_TX_IP_CKSUM. */
147 /** Packet is IPv4 without requiring IP checksum offload. */
148 #define PKT_TX_IPV4 (1ULL << 55)
150 /** Tell the NIC it's an IPv6 packet.*/
151 #define PKT_TX_IPV6 (1ULL << 56)
153 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
155 /** Outer IP checksum of TX packet, computed by NIC for tunneling packet.
156 * The tunnel type must also be specified, ex: PKT_TX_UDP_TUNNEL_PKT. */
157 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
159 /** Packet is outer IPv4 without requiring IP checksum offload for tunneling packet. */
160 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
162 /** Tell the NIC it's an outer IPv6 packet for tunneling packet */
163 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
165 /* Use final bit of flags to indicate a control mbuf */
166 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
169 * Get the name of a RX offload flag
172 * The mask describing the flag.
174 * The name of this flag, or NULL if it's not a valid RX flag.
176 const char *rte_get_rx_ol_flag_name(uint64_t mask);
179 * Get the name of a TX offload flag
182 * The mask describing the flag. Usually only one bit must be set.
183 * Several bits can be given if they belong to the same mask.
184 * Ex: PKT_TX_L4_MASK.
186 * The name of this flag, or NULL if it's not a valid TX flag.
188 const char *rte_get_tx_ol_flag_name(uint64_t mask);
190 /* define a set of marker types that can be used to refer to set points in the
192 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
193 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
194 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
195 * with a single assignment */
198 * The generic rte_mbuf, containing a packet mbuf.
203 void *buf_addr; /**< Virtual address of segment buffer. */
204 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
206 uint16_t buf_len; /**< Length of segment buffer. */
208 /* next 6 bytes are initialised on RX descriptor rearm */
213 * 16-bit Reference counter.
214 * It should only be accessed using the following functions:
215 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
216 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
217 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
221 #ifdef RTE_MBUF_REFCNT
222 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
223 uint16_t refcnt; /**< Non-atomically accessed refcnt */
225 uint16_t refcnt_reserved; /**< Do not use this field */
227 uint8_t nb_segs; /**< Number of segments. */
228 uint8_t port; /**< Input port. */
230 uint64_t ol_flags; /**< Offload features. */
232 /* remaining bytes are set on RX when pulling packet from descriptor */
233 MARKER rx_descriptor_fields1;
236 * The packet type, which is used to indicate ordinary packet and also
237 * tunneled packet format, i.e. each number is represented a type of
240 uint16_t packet_type;
242 uint16_t data_len; /**< Amount of data in segment buffer. */
243 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
244 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
247 uint32_t rss; /**< RSS hash result if RSS enabled */
255 /**< Second 4 flexible bytes */
258 /**< First 4 flexible bytes or FD ID, dependent on
259 PKT_RX_FDIR_* flag in ol_flags. */
260 } fdir; /**< Filter identifier if FDIR enabled */
261 uint32_t sched; /**< Hierarchical scheduler */
262 uint32_t usr; /**< User defined tags. See @rte_distributor_process */
263 } hash; /**< hash information */
265 /* second cache line - fields only used in slow path or on TX */
266 MARKER cacheline1 __rte_cache_aligned;
269 void *userdata; /**< Can be used for external metadata */
270 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
273 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
274 struct rte_mbuf *next; /**< Next segment of scattered packet. */
276 /* fields to support TX offloads */
278 uint64_t tx_offload; /**< combined for easy fetch */
280 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
281 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
282 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
283 uint64_t tso_segsz:16; /**< TCP TSO segment size */
285 /* fields for TX offloading of tunnels */
286 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
287 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
289 /* uint64_t unused:8; */
292 } __rte_cache_aligned;
295 * Given the buf_addr returns the pointer to corresponding mbuf.
297 #define RTE_MBUF_FROM_BADDR(ba) (((struct rte_mbuf *)(ba)) - 1)
300 * Given the pointer to mbuf returns an address where it's buf_addr
303 #define RTE_MBUF_TO_BADDR(mb) (((struct rte_mbuf *)(mb)) + 1)
306 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
308 #define RTE_MBUF_INDIRECT(mb) (RTE_MBUF_FROM_BADDR((mb)->buf_addr) != (mb))
311 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
313 #define RTE_MBUF_DIRECT(mb) (RTE_MBUF_FROM_BADDR((mb)->buf_addr) == (mb))
317 * Private data in case of pktmbuf pool.
319 * A structure that contains some pktmbuf_pool-specific data that are
320 * appended after the mempool structure (in private data).
322 struct rte_pktmbuf_pool_private {
323 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf.*/
326 #ifdef RTE_LIBRTE_MBUF_DEBUG
328 /** check mbuf type in debug mode */
329 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
331 /** check mbuf type in debug mode if mbuf pointer is not null */
332 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
334 rte_mbuf_sanity_check(m, is_h); \
337 /** MBUF asserts in debug mode */
338 #define RTE_MBUF_ASSERT(exp) \
340 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
343 #else /* RTE_LIBRTE_MBUF_DEBUG */
345 /** check mbuf type in debug mode */
346 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
348 /** check mbuf type in debug mode if mbuf pointer is not null */
349 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
351 /** MBUF asserts in debug mode */
352 #define RTE_MBUF_ASSERT(exp) do { } while (0)
354 #endif /* RTE_LIBRTE_MBUF_DEBUG */
356 #ifdef RTE_MBUF_REFCNT
357 #ifdef RTE_MBUF_REFCNT_ATOMIC
360 * Adds given value to an mbuf's refcnt and returns its new value.
364 * Value to add/subtract
368 static inline uint16_t
369 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
371 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
375 * Reads the value of an mbuf's refcnt.
379 * Reference count number.
381 static inline uint16_t
382 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
384 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
388 * Sets an mbuf's refcnt to a defined value.
395 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
397 rte_atomic16_set(&m->refcnt_atomic, new_value);
400 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
403 * Adds given value to an mbuf's refcnt and returns its new value.
405 static inline uint16_t
406 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
408 m->refcnt = (uint16_t)(m->refcnt + value);
413 * Reads the value of an mbuf's refcnt.
415 static inline uint16_t
416 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
422 * Sets an mbuf's refcnt to the defined value.
425 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
427 m->refcnt = new_value;
430 #endif /* RTE_MBUF_REFCNT_ATOMIC */
433 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
438 #else /* ! RTE_MBUF_REFCNT */
441 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { } while(0)
443 #define rte_mbuf_refcnt_set(m,v) do { } while(0)
445 #endif /* RTE_MBUF_REFCNT */
449 * Sanity checks on an mbuf.
451 * Check the consistency of the given mbuf. The function will cause a
452 * panic if corruption is detected.
455 * The mbuf to be checked.
457 * True if the mbuf is a packet header, false if it is a sub-segment
458 * of a packet (in this case, some fields like nb_segs are not checked)
461 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
464 * @internal Allocate a new mbuf from mempool *mp*.
465 * The use of that function is reserved for RTE internal needs.
466 * Please use rte_pktmbuf_alloc().
469 * The mempool from which mbuf is allocated.
471 * - The pointer to the new mbuf on success.
472 * - NULL if allocation failed.
474 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
478 if (rte_mempool_get(mp, &mb) < 0)
480 m = (struct rte_mbuf *)mb;
481 #ifdef RTE_MBUF_REFCNT
482 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
483 rte_mbuf_refcnt_set(m, 1);
484 #endif /* RTE_MBUF_REFCNT */
489 * @internal Put mbuf back into its original mempool.
490 * The use of that function is reserved for RTE internal needs.
491 * Please use rte_pktmbuf_free().
494 * The mbuf to be freed.
496 static inline void __attribute__((always_inline))
497 __rte_mbuf_raw_free(struct rte_mbuf *m)
499 #ifdef RTE_MBUF_REFCNT
500 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
501 #endif /* RTE_MBUF_REFCNT */
502 rte_mempool_put(m->pool, m);
505 /* Operations on ctrl mbuf */
508 * The control mbuf constructor.
510 * This function initializes some fields in an mbuf structure that are
511 * not modified by the user once created (mbuf type, origin pool, buffer
512 * start address, and so on). This function is given as a callback function
513 * to rte_mempool_create() at pool creation time.
516 * The mempool from which the mbuf is allocated.
518 * A pointer that can be used by the user to retrieve useful information
519 * for mbuf initialization. This pointer comes from the ``init_arg``
520 * parameter of rte_mempool_create().
522 * The mbuf to initialize.
524 * The index of the mbuf in the pool table.
526 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
527 void *m, unsigned i);
530 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
532 * This new mbuf is initialized with data pointing to the beginning of
533 * buffer, and with a length of zero.
536 * The mempool from which the mbuf is allocated.
538 * - The pointer to the new mbuf on success.
539 * - NULL if allocation failed.
541 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
544 * Free a control mbuf back into its original mempool.
547 * The control mbuf to be freed.
549 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
552 * A macro that returns the pointer to the carried data.
554 * The value that can be read or assigned.
559 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
562 * A macro that returns the length of the carried data.
564 * The value that can be read or assigned.
569 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
572 * Tests if an mbuf is a control mbuf
575 * The mbuf to be tested
577 * - True (1) if the mbuf is a control mbuf
578 * - False(0) otherwise
581 rte_is_ctrlmbuf(struct rte_mbuf *m)
583 return (!!(m->ol_flags & CTRL_MBUF_FLAG));
586 /* Operations on pkt mbuf */
589 * The packet mbuf constructor.
591 * This function initializes some fields in the mbuf structure that are
592 * not modified by the user once created (origin pool, buffer start
593 * address, and so on). This function is given as a callback function to
594 * rte_mempool_create() at pool creation time.
597 * The mempool from which mbufs originate.
599 * A pointer that can be used by the user to retrieve useful information
600 * for mbuf initialization. This pointer comes from the ``init_arg``
601 * parameter of rte_mempool_create().
603 * The mbuf to initialize.
605 * The index of the mbuf in the pool table.
607 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
608 void *m, unsigned i);
612 * A packet mbuf pool constructor.
614 * This function initializes the mempool private data in the case of a
615 * pktmbuf pool. This private data is needed by the driver. The
616 * function is given as a callback function to rte_mempool_create() at
617 * pool creation. It can be extended by the user, for example, to
618 * provide another packet size.
621 * The mempool from which mbufs originate.
623 * A pointer that can be used by the user to retrieve useful information
624 * for mbuf initialization. This pointer comes from the ``init_arg``
625 * parameter of rte_mempool_create().
627 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
630 * Reset the fields of a packet mbuf to their default values.
632 * The given mbuf must have only one segment.
635 * The packet mbuf to be resetted.
637 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
648 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
649 RTE_PKTMBUF_HEADROOM : m->buf_len;
652 __rte_mbuf_sanity_check(m, 1);
656 * Allocate a new mbuf from a mempool.
658 * This new mbuf contains one segment, which has a length of 0. The pointer
659 * to data is initialized to have some bytes of headroom in the buffer
660 * (if buffer size allows).
663 * The mempool from which the mbuf is allocated.
665 * - The pointer to the new mbuf on success.
666 * - NULL if allocation failed.
668 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
671 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
672 rte_pktmbuf_reset(m);
676 #ifdef RTE_MBUF_REFCNT
679 * Attach packet mbuf to another packet mbuf.
680 * After attachment we refer the mbuf we attached as 'indirect',
681 * while mbuf we attached to as 'direct'.
682 * Right now, not supported:
683 * - attachment to indirect mbuf (e.g. - md has to be direct).
684 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
685 * - mbuf we trying to attach (mi) is used by someone else
686 * e.g. it's reference counter is greater then 1.
689 * The indirect packet mbuf.
691 * The direct packet mbuf.
694 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *md)
696 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(md) &&
697 RTE_MBUF_DIRECT(mi) &&
698 rte_mbuf_refcnt_read(mi) == 1);
700 rte_mbuf_refcnt_update(md, 1);
701 mi->buf_physaddr = md->buf_physaddr;
702 mi->buf_addr = md->buf_addr;
703 mi->buf_len = md->buf_len;
706 mi->data_off = md->data_off;
707 mi->data_len = md->data_len;
709 mi->vlan_tci = md->vlan_tci;
710 mi->tx_offload = md->tx_offload;
714 mi->pkt_len = mi->data_len;
716 mi->ol_flags = md->ol_flags;
717 mi->packet_type = md->packet_type;
719 __rte_mbuf_sanity_check(mi, 1);
720 __rte_mbuf_sanity_check(md, 0);
724 * Detach an indirect packet mbuf -
725 * - restore original mbuf address and length values.
726 * - reset pktmbuf data and data_len to their default values.
727 * All other fields of the given packet mbuf will be left intact.
730 * The indirect attached packet mbuf.
733 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
735 const struct rte_mempool *mp = m->pool;
736 void *buf = RTE_MBUF_TO_BADDR(m);
737 uint32_t buf_len = mp->elt_size - sizeof(*m);
738 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + sizeof (*m);
741 m->buf_len = (uint16_t)buf_len;
743 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
744 RTE_PKTMBUF_HEADROOM : m->buf_len;
749 #endif /* RTE_MBUF_REFCNT */
752 static inline struct rte_mbuf* __attribute__((always_inline))
753 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
755 __rte_mbuf_sanity_check(m, 0);
757 #ifdef RTE_MBUF_REFCNT
758 if (likely (rte_mbuf_refcnt_read(m) == 1) ||
759 likely (rte_mbuf_refcnt_update(m, -1) == 0)) {
760 struct rte_mbuf *md = RTE_MBUF_FROM_BADDR(m->buf_addr);
762 rte_mbuf_refcnt_set(m, 0);
764 /* if this is an indirect mbuf, then
766 * - free attached mbuf segment
768 if (unlikely (md != m)) {
769 rte_pktmbuf_detach(m);
770 if (rte_mbuf_refcnt_update(md, -1) == 0)
771 __rte_mbuf_raw_free(md);
775 #ifdef RTE_MBUF_REFCNT
782 * Free a segment of a packet mbuf into its original mempool.
784 * Free an mbuf, without parsing other segments in case of chained
788 * The packet mbuf segment to be freed.
790 static inline void __attribute__((always_inline))
791 rte_pktmbuf_free_seg(struct rte_mbuf *m)
793 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
795 __rte_mbuf_raw_free(m);
800 * Free a packet mbuf back into its original mempool.
802 * Free an mbuf, and all its segments in case of chained buffers. Each
803 * segment is added back into its original mempool.
806 * The packet mbuf to be freed.
808 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
810 struct rte_mbuf *m_next;
812 __rte_mbuf_sanity_check(m, 1);
816 rte_pktmbuf_free_seg(m);
821 #ifdef RTE_MBUF_REFCNT
824 * Creates a "clone" of the given packet mbuf.
826 * Walks through all segments of the given packet mbuf, and for each of them:
827 * - Creates a new packet mbuf from the given pool.
828 * - Attaches newly created mbuf to the segment.
829 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
830 * from the original packet mbuf.
833 * The packet mbuf to be cloned.
835 * The mempool from which the "clone" mbufs are allocated.
837 * - The pointer to the new "clone" mbuf on success.
838 * - NULL if allocation fails.
840 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
841 struct rte_mempool *mp)
843 struct rte_mbuf *mc, *mi, **prev;
847 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
852 pktlen = md->pkt_len;
857 rte_pktmbuf_attach(mi, md);
860 } while ((md = md->next) != NULL &&
861 (mi = rte_pktmbuf_alloc(mp)) != NULL);
865 mc->pkt_len = pktlen;
867 /* Allocation of new indirect segment failed */
868 if (unlikely (mi == NULL)) {
869 rte_pktmbuf_free(mc);
873 __rte_mbuf_sanity_check(mc, 1);
878 * Adds given value to the refcnt of all packet mbuf segments.
880 * Walks through all segments of given packet mbuf and for each of them
881 * invokes rte_mbuf_refcnt_update().
884 * The packet mbuf whose refcnt to be updated.
886 * The value to add to the mbuf's segments refcnt.
888 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
890 __rte_mbuf_sanity_check(m, 1);
893 rte_mbuf_refcnt_update(m, v);
894 } while ((m = m->next) != NULL);
897 #endif /* RTE_MBUF_REFCNT */
900 * Get the headroom in a packet mbuf.
905 * The length of the headroom.
907 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
909 __rte_mbuf_sanity_check(m, 1);
914 * Get the tailroom of a packet mbuf.
919 * The length of the tailroom.
921 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
923 __rte_mbuf_sanity_check(m, 1);
924 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
929 * Get the last segment of the packet.
934 * The last segment of the given mbuf.
936 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
938 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
940 __rte_mbuf_sanity_check(m, 1);
941 while (m2->next != NULL)
947 * A macro that points to the start of the data in the mbuf.
949 * The returned pointer is cast to type t. Before using this
950 * function, the user must ensure that m_headlen(m) is large enough to
956 * The type to cast the result into.
958 #define rte_pktmbuf_mtod(m, t) ((t)((char *)(m)->buf_addr + (m)->data_off))
961 * A macro that returns the length of the packet.
963 * The value can be read or assigned.
968 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
971 * A macro that returns the length of the segment.
973 * The value can be read or assigned.
978 #define rte_pktmbuf_data_len(m) ((m)->data_len)
981 * Prepend len bytes to an mbuf data area.
983 * Returns a pointer to the new
984 * data start address. If there is not enough headroom in the first
985 * segment, the function will return NULL, without modifying the mbuf.
990 * The amount of data to prepend (in bytes).
992 * A pointer to the start of the newly prepended data, or
993 * NULL if there is not enough headroom space in the first segment
995 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
998 __rte_mbuf_sanity_check(m, 1);
1000 if (unlikely(len > rte_pktmbuf_headroom(m)))
1004 m->data_len = (uint16_t)(m->data_len + len);
1005 m->pkt_len = (m->pkt_len + len);
1007 return (char *)m->buf_addr + m->data_off;
1011 * Append len bytes to an mbuf.
1013 * Append len bytes to an mbuf and return a pointer to the start address
1014 * of the added data. If there is not enough tailroom in the last
1015 * segment, the function will return NULL, without modifying the mbuf.
1020 * The amount of data to append (in bytes).
1022 * A pointer to the start of the newly appended data, or
1023 * NULL if there is not enough tailroom space in the last segment
1025 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1028 struct rte_mbuf *m_last;
1030 __rte_mbuf_sanity_check(m, 1);
1032 m_last = rte_pktmbuf_lastseg(m);
1033 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1036 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1037 m_last->data_len = (uint16_t)(m_last->data_len + len);
1038 m->pkt_len = (m->pkt_len + len);
1039 return (char*) tail;
1043 * Remove len bytes at the beginning of an mbuf.
1045 * Returns a pointer to the start address of the new data area. If the
1046 * length is greater than the length of the first segment, then the
1047 * function will fail and return NULL, without modifying the mbuf.
1052 * The amount of data to remove (in bytes).
1054 * A pointer to the new start of the data.
1056 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1058 __rte_mbuf_sanity_check(m, 1);
1060 if (unlikely(len > m->data_len))
1063 m->data_len = (uint16_t)(m->data_len - len);
1065 m->pkt_len = (m->pkt_len - len);
1066 return (char *)m->buf_addr + m->data_off;
1070 * Remove len bytes of data at the end of the mbuf.
1072 * If the length is greater than the length of the last segment, the
1073 * function will fail and return -1 without modifying the mbuf.
1078 * The amount of data to remove (in bytes).
1083 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1085 struct rte_mbuf *m_last;
1087 __rte_mbuf_sanity_check(m, 1);
1089 m_last = rte_pktmbuf_lastseg(m);
1090 if (unlikely(len > m_last->data_len))
1093 m_last->data_len = (uint16_t)(m_last->data_len - len);
1094 m->pkt_len = (m->pkt_len - len);
1099 * Test if mbuf data is contiguous.
1104 * - 1, if all data is contiguous (one segment).
1105 * - 0, if there is several segments.
1107 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1109 __rte_mbuf_sanity_check(m, 1);
1110 return !!(m->nb_segs == 1);
1114 * Dump an mbuf structure to the console.
1116 * Dump all fields for the given packet mbuf and all its associated
1117 * segments (in the case of a chained buffer).
1120 * A pointer to a file for output
1124 * If dump_len != 0, also dump the "dump_len" first data bytes of
1127 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1133 #endif /* _RTE_MBUF_H_ */