4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
11 * * Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * * Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in
15 * the documentation and/or other materials provided with the
17 * * Neither the name of Intel Corporation nor the names of its
18 * contributors may be used to endorse or promote products derived
19 * from this software without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
41 * The mbuf library provides the ability to create and destroy buffers
42 * that may be used by the RTE application to store message
43 * buffers. The message buffers are stored in a mempool, using the
44 * RTE mempool library.
46 * This library provide an API to allocate/free packet mbufs, which are
47 * used to carry network packets.
49 * To understand the concepts of packet buffers or mbufs, you
50 * should read "TCP/IP Illustrated, Volume 2: The Implementation,
51 * Addison-Wesley, 1995, ISBN 0-201-63354-X from Richard Stevens"
52 * http://www.kohala.com/start/tcpipiv2.html
56 #include <rte_mempool.h>
57 #include <rte_atomic.h>
58 #include <rte_prefetch.h>
59 #include <rte_branch_prediction.h>
65 /* deprecated feature, renamed in RTE_MBUF_REFCNT */
66 #pragma GCC poison RTE_MBUF_SCATTER_GATHER
69 * Packet Offload Features Flags. It also carry packet type information.
70 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
72 * - RX flags start at bit position zero, and get added to the left of previous
74 * - The most-significant 8 bits are reserved for generic mbuf flags
75 * - TX flags therefore start at bit position 55 (i.e. 63-8), and new flags get
76 * added to the right of the previously defined flags
78 #define PKT_RX_VLAN_PKT (1ULL << 0) /**< RX packet is a 802.1q VLAN packet. */
79 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
80 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
81 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
82 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
83 #define PKT_RX_EIP_CKSUM_BAD (0ULL << 0) /**< External IP header checksum error. */
84 #define PKT_RX_OVERSIZE (0ULL << 0) /**< Num of desc of an RX pkt oversize. */
85 #define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
86 #define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
87 #define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
88 #define PKT_RX_IPV4_HDR (1ULL << 5) /**< RX packet with IPv4 header. */
89 #define PKT_RX_IPV4_HDR_EXT (1ULL << 6) /**< RX packet with extended IPv4 header. */
90 #define PKT_RX_IPV6_HDR (1ULL << 7) /**< RX packet with IPv6 header. */
91 #define PKT_RX_IPV6_HDR_EXT (1ULL << 8) /**< RX packet with extended IPv6 header. */
92 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
93 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
94 #define PKT_RX_TUNNEL_IPV4_HDR (1ULL << 11) /**< RX tunnel packet with IPv4 header.*/
95 #define PKT_RX_TUNNEL_IPV6_HDR (1ULL << 12) /**< RX tunnel packet with IPv6 header. */
96 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
97 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
99 #define PKT_TX_VLAN_PKT (1ULL << 55) /**< TX packet is a 802.1q VLAN packet. */
100 #define PKT_TX_IP_CKSUM (1ULL << 54) /**< IP cksum of TX pkt. computed by NIC. */
101 #define PKT_TX_VXLAN_CKSUM (1ULL << 50) /**< TX checksum of VXLAN computed by NIC */
102 #define PKT_TX_IPV4_CSUM PKT_TX_IP_CKSUM /**< Alias of PKT_TX_IP_CKSUM. */
103 #define PKT_TX_IPV4 PKT_RX_IPV4_HDR /**< IPv4 with no IP checksum offload. */
104 #define PKT_TX_IPV6 PKT_RX_IPV6_HDR /**< IPv6 packet */
107 * Bits 52+53 used for L4 packet type with checksum enabled.
113 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
114 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
115 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
116 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
117 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
119 /* Bit 51 - IEEE1588*/
120 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
122 /* Use final bit of flags to indicate a control mbuf */
123 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
126 * Bit Mask to indicate what bits required for building TX context
128 #define PKT_TX_OFFLOAD_MASK (PKT_TX_VLAN_PKT | PKT_TX_IP_CKSUM | PKT_TX_L4_MASK)
130 /* define a set of marker types that can be used to refer to set points in the
132 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
133 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
134 * with a single assignment */
136 * The generic rte_mbuf, containing a packet mbuf.
141 void *buf_addr; /**< Virtual address of segment buffer. */
142 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
144 /* next 8 bytes are initialised on RX descriptor rearm */
146 uint16_t buf_len; /**< Length of segment buffer. */
150 * 16-bit Reference counter.
151 * It should only be accessed using the following functions:
152 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
153 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
154 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
158 #ifdef RTE_MBUF_REFCNT
159 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
160 uint16_t refcnt; /**< Non-atomically accessed refcnt */
162 uint16_t refcnt_reserved; /**< Do not use this field */
164 uint8_t nb_segs; /**< Number of segments. */
165 uint8_t port; /**< Input port. */
167 uint64_t ol_flags; /**< Offload features. */
169 /* remaining bytes are set on RX when pulling packet from descriptor */
170 MARKER rx_descriptor_fields1;
173 * The packet type, which is used to indicate ordinary packet and also
174 * tunneled packet format, i.e. each number is represented a type of
177 uint16_t packet_type;
179 uint16_t data_len; /**< Amount of data in segment buffer. */
180 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
181 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
184 uint32_t rss; /**< RSS hash result if RSS enabled */
192 /**< Second 4 flexible bytes */
195 /**< First 4 flexible bytes or FD ID, dependent on
196 PKT_RX_FDIR_* flag in ol_flags. */
197 } fdir; /**< Filter identifier if FDIR enabled */
198 uint32_t sched; /**< Hierarchical scheduler */
199 uint32_t usr; /**< User defined tags. See @rte_distributor_process */
200 } hash; /**< hash information */
202 /* second cache line - fields only used in slow path or on TX */
203 MARKER cacheline1 __rte_cache_aligned;
206 void *userdata; /**< Can be used for external metadata */
207 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
210 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
211 struct rte_mbuf *next; /**< Next segment of scattered packet. */
213 /* fields to support TX offloads */
215 uint16_t l2_l3_len; /**< combined l2/l3 lengths as single var */
217 uint16_t l3_len:9; /**< L3 (IP) Header Length. */
218 uint16_t l2_len:7; /**< L2 (MAC) Header Length. */
222 /* fields for TX offloading of tunnels */
224 uint16_t inner_l2_l3_len;
225 /**< combined inner l2/l3 lengths as single var */
227 uint16_t inner_l3_len:9;
228 /**< inner L3 (IP) Header Length. */
229 uint16_t inner_l2_len:7;
230 /**< inner L2 (MAC) Header Length. */
233 } __rte_cache_aligned;
236 * Given the buf_addr returns the pointer to corresponding mbuf.
238 #define RTE_MBUF_FROM_BADDR(ba) (((struct rte_mbuf *)(ba)) - 1)
241 * Given the pointer to mbuf returns an address where it's buf_addr
244 #define RTE_MBUF_TO_BADDR(mb) (((struct rte_mbuf *)(mb)) + 1)
247 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
249 #define RTE_MBUF_INDIRECT(mb) (RTE_MBUF_FROM_BADDR((mb)->buf_addr) != (mb))
252 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
254 #define RTE_MBUF_DIRECT(mb) (RTE_MBUF_FROM_BADDR((mb)->buf_addr) == (mb))
258 * Private data in case of pktmbuf pool.
260 * A structure that contains some pktmbuf_pool-specific data that are
261 * appended after the mempool structure (in private data).
263 struct rte_pktmbuf_pool_private {
264 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf.*/
267 #ifdef RTE_LIBRTE_MBUF_DEBUG
269 /** check mbuf type in debug mode */
270 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
272 /** check mbuf type in debug mode if mbuf pointer is not null */
273 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
275 rte_mbuf_sanity_check(m, is_h); \
278 /** MBUF asserts in debug mode */
279 #define RTE_MBUF_ASSERT(exp) \
281 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
284 #else /* RTE_LIBRTE_MBUF_DEBUG */
286 /** check mbuf type in debug mode */
287 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
289 /** check mbuf type in debug mode if mbuf pointer is not null */
290 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
292 /** MBUF asserts in debug mode */
293 #define RTE_MBUF_ASSERT(exp) do { } while (0)
295 #endif /* RTE_LIBRTE_MBUF_DEBUG */
297 #ifdef RTE_MBUF_REFCNT
298 #ifdef RTE_MBUF_REFCNT_ATOMIC
301 * Adds given value to an mbuf's refcnt and returns its new value.
305 * Value to add/subtract
309 static inline uint16_t
310 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
312 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
316 * Reads the value of an mbuf's refcnt.
320 * Reference count number.
322 static inline uint16_t
323 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
325 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
329 * Sets an mbuf's refcnt to a defined value.
336 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
338 rte_atomic16_set(&m->refcnt_atomic, new_value);
341 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
344 * Adds given value to an mbuf's refcnt and returns its new value.
346 static inline uint16_t
347 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
349 m->refcnt = (uint16_t)(m->refcnt + value);
354 * Reads the value of an mbuf's refcnt.
356 static inline uint16_t
357 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
363 * Sets an mbuf's refcnt to the defined value.
366 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
368 m->refcnt = new_value;
371 #endif /* RTE_MBUF_REFCNT_ATOMIC */
374 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
379 #else /* ! RTE_MBUF_REFCNT */
382 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { } while(0)
384 #define rte_mbuf_refcnt_set(m,v) do { } while(0)
386 #endif /* RTE_MBUF_REFCNT */
390 * Sanity checks on an mbuf.
392 * Check the consistency of the given mbuf. The function will cause a
393 * panic if corruption is detected.
396 * The mbuf to be checked.
398 * True if the mbuf is a packet header, false if it is a sub-segment
399 * of a packet (in this case, some fields like nb_segs are not checked)
402 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
405 * @internal Allocate a new mbuf from mempool *mp*.
406 * The use of that function is reserved for RTE internal needs.
407 * Please use rte_pktmbuf_alloc().
410 * The mempool from which mbuf is allocated.
412 * - The pointer to the new mbuf on success.
413 * - NULL if allocation failed.
415 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
419 if (rte_mempool_get(mp, &mb) < 0)
421 m = (struct rte_mbuf *)mb;
422 #ifdef RTE_MBUF_REFCNT
423 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
424 rte_mbuf_refcnt_set(m, 1);
425 #endif /* RTE_MBUF_REFCNT */
430 * @internal Put mbuf back into its original mempool.
431 * The use of that function is reserved for RTE internal needs.
432 * Please use rte_pktmbuf_free().
435 * The mbuf to be freed.
437 static inline void __attribute__((always_inline))
438 __rte_mbuf_raw_free(struct rte_mbuf *m)
440 #ifdef RTE_MBUF_REFCNT
441 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
442 #endif /* RTE_MBUF_REFCNT */
443 rte_mempool_put(m->pool, m);
446 /* Operations on ctrl mbuf */
449 * The control mbuf constructor.
451 * This function initializes some fields in an mbuf structure that are
452 * not modified by the user once created (mbuf type, origin pool, buffer
453 * start address, and so on). This function is given as a callback function
454 * to rte_mempool_create() at pool creation time.
457 * The mempool from which the mbuf is allocated.
459 * A pointer that can be used by the user to retrieve useful information
460 * for mbuf initialization. This pointer comes from the ``init_arg``
461 * parameter of rte_mempool_create().
463 * The mbuf to initialize.
465 * The index of the mbuf in the pool table.
467 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
468 void *m, unsigned i);
471 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
473 * This new mbuf is initialized with data pointing to the beginning of
474 * buffer, and with a length of zero.
477 * The mempool from which the mbuf is allocated.
479 * - The pointer to the new mbuf on success.
480 * - NULL if allocation failed.
482 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
485 * Free a control mbuf back into its original mempool.
488 * The control mbuf to be freed.
490 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
493 * A macro that returns the pointer to the carried data.
495 * The value that can be read or assigned.
500 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
503 * A macro that returns the length of the carried data.
505 * The value that can be read or assigned.
510 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
513 * Tests if an mbuf is a control mbuf
516 * The mbuf to be tested
518 * - True (1) if the mbuf is a control mbuf
519 * - False(0) otherwise
522 rte_is_ctrlmbuf(struct rte_mbuf *m)
524 return (!!(m->ol_flags & CTRL_MBUF_FLAG));
527 /* Operations on pkt mbuf */
530 * The packet mbuf constructor.
532 * This function initializes some fields in the mbuf structure that are
533 * not modified by the user once created (origin pool, buffer start
534 * address, and so on). This function is given as a callback function to
535 * rte_mempool_create() at pool creation time.
538 * The mempool from which mbufs originate.
540 * A pointer that can be used by the user to retrieve useful information
541 * for mbuf initialization. This pointer comes from the ``init_arg``
542 * parameter of rte_mempool_create().
544 * The mbuf to initialize.
546 * The index of the mbuf in the pool table.
548 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
549 void *m, unsigned i);
553 * A packet mbuf pool constructor.
555 * This function initializes the mempool private data in the case of a
556 * pktmbuf pool. This private data is needed by the driver. The
557 * function is given as a callback function to rte_mempool_create() at
558 * pool creation. It can be extended by the user, for example, to
559 * provide another packet size.
562 * The mempool from which mbufs originate.
564 * A pointer that can be used by the user to retrieve useful information
565 * for mbuf initialization. This pointer comes from the ``init_arg``
566 * parameter of rte_mempool_create().
568 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
571 * Reset the fields of a packet mbuf to their default values.
573 * The given mbuf must have only one segment.
576 * The packet mbuf to be resetted.
578 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
583 m->inner_l2_l3_len = 0;
590 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
591 RTE_PKTMBUF_HEADROOM : m->buf_len;
594 __rte_mbuf_sanity_check(m, 1);
598 * Allocate a new mbuf from a mempool.
600 * This new mbuf contains one segment, which has a length of 0. The pointer
601 * to data is initialized to have some bytes of headroom in the buffer
602 * (if buffer size allows).
605 * The mempool from which the mbuf is allocated.
607 * - The pointer to the new mbuf on success.
608 * - NULL if allocation failed.
610 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
613 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
614 rte_pktmbuf_reset(m);
618 #ifdef RTE_MBUF_REFCNT
621 * Attach packet mbuf to another packet mbuf.
622 * After attachment we refer the mbuf we attached as 'indirect',
623 * while mbuf we attached to as 'direct'.
624 * Right now, not supported:
625 * - attachment to indirect mbuf (e.g. - md has to be direct).
626 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
627 * - mbuf we trying to attach (mi) is used by someone else
628 * e.g. it's reference counter is greater then 1.
631 * The indirect packet mbuf.
633 * The direct packet mbuf.
636 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *md)
638 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(md) &&
639 RTE_MBUF_DIRECT(mi) &&
640 rte_mbuf_refcnt_read(mi) == 1);
642 rte_mbuf_refcnt_update(md, 1);
643 mi->buf_physaddr = md->buf_physaddr;
644 mi->buf_addr = md->buf_addr;
645 mi->buf_len = md->buf_len;
648 mi->data_off = md->data_off;
649 mi->data_len = md->data_len;
651 mi->vlan_tci = md->vlan_tci;
652 mi->l2_l3_len = md->l2_l3_len;
653 mi->inner_l2_l3_len = md->inner_l2_l3_len;
657 mi->pkt_len = mi->data_len;
659 mi->ol_flags = md->ol_flags;
660 mi->packet_type = md->packet_type;
662 __rte_mbuf_sanity_check(mi, 1);
663 __rte_mbuf_sanity_check(md, 0);
667 * Detach an indirect packet mbuf -
668 * - restore original mbuf address and length values.
669 * - reset pktmbuf data and data_len to their default values.
670 * All other fields of the given packet mbuf will be left intact.
673 * The indirect attached packet mbuf.
676 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
678 const struct rte_mempool *mp = m->pool;
679 void *buf = RTE_MBUF_TO_BADDR(m);
680 uint32_t buf_len = mp->elt_size - sizeof(*m);
681 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + sizeof (*m);
684 m->buf_len = (uint16_t)buf_len;
686 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
687 RTE_PKTMBUF_HEADROOM : m->buf_len;
692 #endif /* RTE_MBUF_REFCNT */
695 static inline struct rte_mbuf* __attribute__((always_inline))
696 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
698 __rte_mbuf_sanity_check(m, 0);
700 #ifdef RTE_MBUF_REFCNT
701 if (likely (rte_mbuf_refcnt_read(m) == 1) ||
702 likely (rte_mbuf_refcnt_update(m, -1) == 0)) {
703 struct rte_mbuf *md = RTE_MBUF_FROM_BADDR(m->buf_addr);
705 rte_mbuf_refcnt_set(m, 0);
707 /* if this is an indirect mbuf, then
709 * - free attached mbuf segment
711 if (unlikely (md != m)) {
712 rte_pktmbuf_detach(m);
713 if (rte_mbuf_refcnt_update(md, -1) == 0)
714 __rte_mbuf_raw_free(md);
718 #ifdef RTE_MBUF_REFCNT
725 * Free a segment of a packet mbuf into its original mempool.
727 * Free an mbuf, without parsing other segments in case of chained
731 * The packet mbuf segment to be freed.
733 static inline void __attribute__((always_inline))
734 rte_pktmbuf_free_seg(struct rte_mbuf *m)
736 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
738 __rte_mbuf_raw_free(m);
743 * Free a packet mbuf back into its original mempool.
745 * Free an mbuf, and all its segments in case of chained buffers. Each
746 * segment is added back into its original mempool.
749 * The packet mbuf to be freed.
751 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
753 struct rte_mbuf *m_next;
755 __rte_mbuf_sanity_check(m, 1);
759 rte_pktmbuf_free_seg(m);
764 #ifdef RTE_MBUF_REFCNT
767 * Creates a "clone" of the given packet mbuf.
769 * Walks through all segments of the given packet mbuf, and for each of them:
770 * - Creates a new packet mbuf from the given pool.
771 * - Attaches newly created mbuf to the segment.
772 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
773 * from the original packet mbuf.
776 * The packet mbuf to be cloned.
778 * The mempool from which the "clone" mbufs are allocated.
780 * - The pointer to the new "clone" mbuf on success.
781 * - NULL if allocation fails.
783 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
784 struct rte_mempool *mp)
786 struct rte_mbuf *mc, *mi, **prev;
790 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
795 pktlen = md->pkt_len;
800 rte_pktmbuf_attach(mi, md);
803 } while ((md = md->next) != NULL &&
804 (mi = rte_pktmbuf_alloc(mp)) != NULL);
808 mc->pkt_len = pktlen;
810 /* Allocation of new indirect segment failed */
811 if (unlikely (mi == NULL)) {
812 rte_pktmbuf_free(mc);
816 __rte_mbuf_sanity_check(mc, 1);
821 * Adds given value to the refcnt of all packet mbuf segments.
823 * Walks through all segments of given packet mbuf and for each of them
824 * invokes rte_mbuf_refcnt_update().
827 * The packet mbuf whose refcnt to be updated.
829 * The value to add to the mbuf's segments refcnt.
831 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
833 __rte_mbuf_sanity_check(m, 1);
836 rte_mbuf_refcnt_update(m, v);
837 } while ((m = m->next) != NULL);
840 #endif /* RTE_MBUF_REFCNT */
843 * Get the headroom in a packet mbuf.
848 * The length of the headroom.
850 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
852 __rte_mbuf_sanity_check(m, 1);
857 * Get the tailroom of a packet mbuf.
862 * The length of the tailroom.
864 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
866 __rte_mbuf_sanity_check(m, 1);
867 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
872 * Get the last segment of the packet.
877 * The last segment of the given mbuf.
879 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
881 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
883 __rte_mbuf_sanity_check(m, 1);
884 while (m2->next != NULL)
890 * A macro that points to the start of the data in the mbuf.
892 * The returned pointer is cast to type t. Before using this
893 * function, the user must ensure that m_headlen(m) is large enough to
899 * The type to cast the result into.
901 #define rte_pktmbuf_mtod(m, t) ((t)((char *)(m)->buf_addr + (m)->data_off))
904 * A macro that returns the length of the packet.
906 * The value can be read or assigned.
911 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
914 * A macro that returns the length of the segment.
916 * The value can be read or assigned.
921 #define rte_pktmbuf_data_len(m) ((m)->data_len)
924 * Prepend len bytes to an mbuf data area.
926 * Returns a pointer to the new
927 * data start address. If there is not enough headroom in the first
928 * segment, the function will return NULL, without modifying the mbuf.
933 * The amount of data to prepend (in bytes).
935 * A pointer to the start of the newly prepended data, or
936 * NULL if there is not enough headroom space in the first segment
938 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
941 __rte_mbuf_sanity_check(m, 1);
943 if (unlikely(len > rte_pktmbuf_headroom(m)))
947 m->data_len = (uint16_t)(m->data_len + len);
948 m->pkt_len = (m->pkt_len + len);
950 return (char *)m->buf_addr + m->data_off;
954 * Append len bytes to an mbuf.
956 * Append len bytes to an mbuf and return a pointer to the start address
957 * of the added data. If there is not enough tailroom in the last
958 * segment, the function will return NULL, without modifying the mbuf.
963 * The amount of data to append (in bytes).
965 * A pointer to the start of the newly appended data, or
966 * NULL if there is not enough tailroom space in the last segment
968 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
971 struct rte_mbuf *m_last;
973 __rte_mbuf_sanity_check(m, 1);
975 m_last = rte_pktmbuf_lastseg(m);
976 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
979 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
980 m_last->data_len = (uint16_t)(m_last->data_len + len);
981 m->pkt_len = (m->pkt_len + len);
986 * Remove len bytes at the beginning of an mbuf.
988 * Returns a pointer to the start address of the new data area. If the
989 * length is greater than the length of the first segment, then the
990 * function will fail and return NULL, without modifying the mbuf.
995 * The amount of data to remove (in bytes).
997 * A pointer to the new start of the data.
999 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1001 __rte_mbuf_sanity_check(m, 1);
1003 if (unlikely(len > m->data_len))
1006 m->data_len = (uint16_t)(m->data_len - len);
1008 m->pkt_len = (m->pkt_len - len);
1009 return (char *)m->buf_addr + m->data_off;
1013 * Remove len bytes of data at the end of the mbuf.
1015 * If the length is greater than the length of the last segment, the
1016 * function will fail and return -1 without modifying the mbuf.
1021 * The amount of data to remove (in bytes).
1026 static inline int rte_pktmbuf_trim(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 > m_last->data_len))
1036 m_last->data_len = (uint16_t)(m_last->data_len - len);
1037 m->pkt_len = (m->pkt_len - len);
1042 * Test if mbuf data is contiguous.
1047 * - 1, if all data is contiguous (one segment).
1048 * - 0, if there is several segments.
1050 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1052 __rte_mbuf_sanity_check(m, 1);
1053 return !!(m->nb_segs == 1);
1057 * Dump an mbuf structure to the console.
1059 * Dump all fields for the given packet mbuf and all its associated
1060 * segments (in the case of a chained buffer).
1063 * A pointer to a file for output
1067 * If dump_len != 0, also dump the "dump_len" first data bytes of
1070 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1076 #endif /* _RTE_MBUF_H_ */