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. */
98 /* add new RX flags here */
100 /* add new TX flags here */
101 #define PKT_TX_VXLAN_CKSUM (1ULL << 50) /**< TX checksum of VXLAN computed by NIC */
102 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
105 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
106 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
107 * L4 checksum offload, the user needs to:
108 * - fill l2_len and l3_len in mbuf
109 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
110 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
111 * - calculate the pseudo header checksum and set it in the L4 header (only
112 * for TCP or UDP). For SCTP, set the crc field to 0.
114 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
115 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
116 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
117 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
118 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
120 #define PKT_TX_IP_CKSUM (1ULL << 54) /**< IP cksum of TX pkt. computed by NIC. */
121 #define PKT_TX_IPV4_CSUM PKT_TX_IP_CKSUM /**< Alias of PKT_TX_IP_CKSUM. */
123 /** Tell the NIC it's an IPv4 packet. Required for L4 checksum offload. */
124 #define PKT_TX_IPV4 PKT_RX_IPV4_HDR
126 /** Tell the NIC it's an IPv6 packet. Required for L4 checksum offload. */
127 #define PKT_TX_IPV6 PKT_RX_IPV6_HDR
129 #define PKT_TX_VLAN_PKT (1ULL << 55) /**< TX packet is a 802.1q VLAN packet. */
131 /* Use final bit of flags to indicate a control mbuf */
132 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
135 * Bit Mask to indicate what bits required for building TX context
137 #define PKT_TX_OFFLOAD_MASK (PKT_TX_VLAN_PKT | PKT_TX_IP_CKSUM | PKT_TX_L4_MASK)
139 /* define a set of marker types that can be used to refer to set points in the
141 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
142 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
143 * with a single assignment */
145 * The generic rte_mbuf, containing a packet mbuf.
150 void *buf_addr; /**< Virtual address of segment buffer. */
151 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
153 /* next 8 bytes are initialised on RX descriptor rearm */
155 uint16_t buf_len; /**< Length of segment buffer. */
159 * 16-bit Reference counter.
160 * It should only be accessed using the following functions:
161 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
162 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
163 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
167 #ifdef RTE_MBUF_REFCNT
168 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
169 uint16_t refcnt; /**< Non-atomically accessed refcnt */
171 uint16_t refcnt_reserved; /**< Do not use this field */
173 uint8_t nb_segs; /**< Number of segments. */
174 uint8_t port; /**< Input port. */
176 uint64_t ol_flags; /**< Offload features. */
178 /* remaining bytes are set on RX when pulling packet from descriptor */
179 MARKER rx_descriptor_fields1;
182 * The packet type, which is used to indicate ordinary packet and also
183 * tunneled packet format, i.e. each number is represented a type of
186 uint16_t packet_type;
188 uint16_t data_len; /**< Amount of data in segment buffer. */
189 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
190 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
193 uint32_t rss; /**< RSS hash result if RSS enabled */
201 /**< Second 4 flexible bytes */
204 /**< First 4 flexible bytes or FD ID, dependent on
205 PKT_RX_FDIR_* flag in ol_flags. */
206 } fdir; /**< Filter identifier if FDIR enabled */
207 uint32_t sched; /**< Hierarchical scheduler */
208 uint32_t usr; /**< User defined tags. See @rte_distributor_process */
209 } hash; /**< hash information */
211 /* second cache line - fields only used in slow path or on TX */
212 MARKER cacheline1 __rte_cache_aligned;
215 void *userdata; /**< Can be used for external metadata */
216 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
219 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
220 struct rte_mbuf *next; /**< Next segment of scattered packet. */
222 /* fields to support TX offloads */
224 uint16_t l2_l3_len; /**< combined l2/l3 lengths as single var */
226 uint16_t l3_len:9; /**< L3 (IP) Header Length. */
227 uint16_t l2_len:7; /**< L2 (MAC) Header Length. */
231 /* fields for TX offloading of tunnels */
233 uint16_t inner_l2_l3_len;
234 /**< combined inner l2/l3 lengths as single var */
236 uint16_t inner_l3_len:9;
237 /**< inner L3 (IP) Header Length. */
238 uint16_t inner_l2_len:7;
239 /**< inner L2 (MAC) Header Length. */
242 } __rte_cache_aligned;
245 * Given the buf_addr returns the pointer to corresponding mbuf.
247 #define RTE_MBUF_FROM_BADDR(ba) (((struct rte_mbuf *)(ba)) - 1)
250 * Given the pointer to mbuf returns an address where it's buf_addr
253 #define RTE_MBUF_TO_BADDR(mb) (((struct rte_mbuf *)(mb)) + 1)
256 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
258 #define RTE_MBUF_INDIRECT(mb) (RTE_MBUF_FROM_BADDR((mb)->buf_addr) != (mb))
261 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
263 #define RTE_MBUF_DIRECT(mb) (RTE_MBUF_FROM_BADDR((mb)->buf_addr) == (mb))
267 * Private data in case of pktmbuf pool.
269 * A structure that contains some pktmbuf_pool-specific data that are
270 * appended after the mempool structure (in private data).
272 struct rte_pktmbuf_pool_private {
273 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf.*/
276 #ifdef RTE_LIBRTE_MBUF_DEBUG
278 /** check mbuf type in debug mode */
279 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
281 /** check mbuf type in debug mode if mbuf pointer is not null */
282 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
284 rte_mbuf_sanity_check(m, is_h); \
287 /** MBUF asserts in debug mode */
288 #define RTE_MBUF_ASSERT(exp) \
290 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
293 #else /* RTE_LIBRTE_MBUF_DEBUG */
295 /** check mbuf type in debug mode */
296 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
298 /** check mbuf type in debug mode if mbuf pointer is not null */
299 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
301 /** MBUF asserts in debug mode */
302 #define RTE_MBUF_ASSERT(exp) do { } while (0)
304 #endif /* RTE_LIBRTE_MBUF_DEBUG */
306 #ifdef RTE_MBUF_REFCNT
307 #ifdef RTE_MBUF_REFCNT_ATOMIC
310 * Adds given value to an mbuf's refcnt and returns its new value.
314 * Value to add/subtract
318 static inline uint16_t
319 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
321 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
325 * Reads the value of an mbuf's refcnt.
329 * Reference count number.
331 static inline uint16_t
332 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
334 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
338 * Sets an mbuf's refcnt to a defined value.
345 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
347 rte_atomic16_set(&m->refcnt_atomic, new_value);
350 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
353 * Adds given value to an mbuf's refcnt and returns its new value.
355 static inline uint16_t
356 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
358 m->refcnt = (uint16_t)(m->refcnt + value);
363 * Reads the value of an mbuf's refcnt.
365 static inline uint16_t
366 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
372 * Sets an mbuf's refcnt to the defined value.
375 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
377 m->refcnt = new_value;
380 #endif /* RTE_MBUF_REFCNT_ATOMIC */
383 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
388 #else /* ! RTE_MBUF_REFCNT */
391 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { } while(0)
393 #define rte_mbuf_refcnt_set(m,v) do { } while(0)
395 #endif /* RTE_MBUF_REFCNT */
399 * Sanity checks on an mbuf.
401 * Check the consistency of the given mbuf. The function will cause a
402 * panic if corruption is detected.
405 * The mbuf to be checked.
407 * True if the mbuf is a packet header, false if it is a sub-segment
408 * of a packet (in this case, some fields like nb_segs are not checked)
411 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
414 * @internal Allocate a new mbuf from mempool *mp*.
415 * The use of that function is reserved for RTE internal needs.
416 * Please use rte_pktmbuf_alloc().
419 * The mempool from which mbuf is allocated.
421 * - The pointer to the new mbuf on success.
422 * - NULL if allocation failed.
424 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
428 if (rte_mempool_get(mp, &mb) < 0)
430 m = (struct rte_mbuf *)mb;
431 #ifdef RTE_MBUF_REFCNT
432 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
433 rte_mbuf_refcnt_set(m, 1);
434 #endif /* RTE_MBUF_REFCNT */
439 * @internal Put mbuf back into its original mempool.
440 * The use of that function is reserved for RTE internal needs.
441 * Please use rte_pktmbuf_free().
444 * The mbuf to be freed.
446 static inline void __attribute__((always_inline))
447 __rte_mbuf_raw_free(struct rte_mbuf *m)
449 #ifdef RTE_MBUF_REFCNT
450 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
451 #endif /* RTE_MBUF_REFCNT */
452 rte_mempool_put(m->pool, m);
455 /* Operations on ctrl mbuf */
458 * The control mbuf constructor.
460 * This function initializes some fields in an mbuf structure that are
461 * not modified by the user once created (mbuf type, origin pool, buffer
462 * start address, and so on). This function is given as a callback function
463 * to rte_mempool_create() at pool creation time.
466 * The mempool from which the mbuf is allocated.
468 * A pointer that can be used by the user to retrieve useful information
469 * for mbuf initialization. This pointer comes from the ``init_arg``
470 * parameter of rte_mempool_create().
472 * The mbuf to initialize.
474 * The index of the mbuf in the pool table.
476 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
477 void *m, unsigned i);
480 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
482 * This new mbuf is initialized with data pointing to the beginning of
483 * buffer, and with a length of zero.
486 * The mempool from which the mbuf is allocated.
488 * - The pointer to the new mbuf on success.
489 * - NULL if allocation failed.
491 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
494 * Free a control mbuf back into its original mempool.
497 * The control mbuf to be freed.
499 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
502 * A macro that returns the pointer to the carried data.
504 * The value that can be read or assigned.
509 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
512 * A macro that returns the length of the carried data.
514 * The value that can be read or assigned.
519 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
522 * Tests if an mbuf is a control mbuf
525 * The mbuf to be tested
527 * - True (1) if the mbuf is a control mbuf
528 * - False(0) otherwise
531 rte_is_ctrlmbuf(struct rte_mbuf *m)
533 return (!!(m->ol_flags & CTRL_MBUF_FLAG));
536 /* Operations on pkt mbuf */
539 * The packet mbuf constructor.
541 * This function initializes some fields in the mbuf structure that are
542 * not modified by the user once created (origin pool, buffer start
543 * address, and so on). This function is given as a callback function to
544 * rte_mempool_create() at pool creation time.
547 * The mempool from which mbufs originate.
549 * A pointer that can be used by the user to retrieve useful information
550 * for mbuf initialization. This pointer comes from the ``init_arg``
551 * parameter of rte_mempool_create().
553 * The mbuf to initialize.
555 * The index of the mbuf in the pool table.
557 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
558 void *m, unsigned i);
562 * A packet mbuf pool constructor.
564 * This function initializes the mempool private data in the case of a
565 * pktmbuf pool. This private data is needed by the driver. The
566 * function is given as a callback function to rte_mempool_create() at
567 * pool creation. It can be extended by the user, for example, to
568 * provide another packet size.
571 * The mempool from which mbufs originate.
573 * A pointer that can be used by the user to retrieve useful information
574 * for mbuf initialization. This pointer comes from the ``init_arg``
575 * parameter of rte_mempool_create().
577 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
580 * Reset the fields of a packet mbuf to their default values.
582 * The given mbuf must have only one segment.
585 * The packet mbuf to be resetted.
587 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
592 m->inner_l2_l3_len = 0;
599 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
600 RTE_PKTMBUF_HEADROOM : m->buf_len;
603 __rte_mbuf_sanity_check(m, 1);
607 * Allocate a new mbuf from a mempool.
609 * This new mbuf contains one segment, which has a length of 0. The pointer
610 * to data is initialized to have some bytes of headroom in the buffer
611 * (if buffer size allows).
614 * The mempool from which the mbuf is allocated.
616 * - The pointer to the new mbuf on success.
617 * - NULL if allocation failed.
619 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
622 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
623 rte_pktmbuf_reset(m);
627 #ifdef RTE_MBUF_REFCNT
630 * Attach packet mbuf to another packet mbuf.
631 * After attachment we refer the mbuf we attached as 'indirect',
632 * while mbuf we attached to as 'direct'.
633 * Right now, not supported:
634 * - attachment to indirect mbuf (e.g. - md has to be direct).
635 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
636 * - mbuf we trying to attach (mi) is used by someone else
637 * e.g. it's reference counter is greater then 1.
640 * The indirect packet mbuf.
642 * The direct packet mbuf.
645 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *md)
647 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(md) &&
648 RTE_MBUF_DIRECT(mi) &&
649 rte_mbuf_refcnt_read(mi) == 1);
651 rte_mbuf_refcnt_update(md, 1);
652 mi->buf_physaddr = md->buf_physaddr;
653 mi->buf_addr = md->buf_addr;
654 mi->buf_len = md->buf_len;
657 mi->data_off = md->data_off;
658 mi->data_len = md->data_len;
660 mi->vlan_tci = md->vlan_tci;
661 mi->l2_l3_len = md->l2_l3_len;
662 mi->inner_l2_l3_len = md->inner_l2_l3_len;
666 mi->pkt_len = mi->data_len;
668 mi->ol_flags = md->ol_flags;
669 mi->packet_type = md->packet_type;
671 __rte_mbuf_sanity_check(mi, 1);
672 __rte_mbuf_sanity_check(md, 0);
676 * Detach an indirect packet mbuf -
677 * - restore original mbuf address and length values.
678 * - reset pktmbuf data and data_len to their default values.
679 * All other fields of the given packet mbuf will be left intact.
682 * The indirect attached packet mbuf.
685 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
687 const struct rte_mempool *mp = m->pool;
688 void *buf = RTE_MBUF_TO_BADDR(m);
689 uint32_t buf_len = mp->elt_size - sizeof(*m);
690 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + sizeof (*m);
693 m->buf_len = (uint16_t)buf_len;
695 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
696 RTE_PKTMBUF_HEADROOM : m->buf_len;
701 #endif /* RTE_MBUF_REFCNT */
704 static inline struct rte_mbuf* __attribute__((always_inline))
705 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
707 __rte_mbuf_sanity_check(m, 0);
709 #ifdef RTE_MBUF_REFCNT
710 if (likely (rte_mbuf_refcnt_read(m) == 1) ||
711 likely (rte_mbuf_refcnt_update(m, -1) == 0)) {
712 struct rte_mbuf *md = RTE_MBUF_FROM_BADDR(m->buf_addr);
714 rte_mbuf_refcnt_set(m, 0);
716 /* if this is an indirect mbuf, then
718 * - free attached mbuf segment
720 if (unlikely (md != m)) {
721 rte_pktmbuf_detach(m);
722 if (rte_mbuf_refcnt_update(md, -1) == 0)
723 __rte_mbuf_raw_free(md);
727 #ifdef RTE_MBUF_REFCNT
734 * Free a segment of a packet mbuf into its original mempool.
736 * Free an mbuf, without parsing other segments in case of chained
740 * The packet mbuf segment to be freed.
742 static inline void __attribute__((always_inline))
743 rte_pktmbuf_free_seg(struct rte_mbuf *m)
745 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
747 __rte_mbuf_raw_free(m);
752 * Free a packet mbuf back into its original mempool.
754 * Free an mbuf, and all its segments in case of chained buffers. Each
755 * segment is added back into its original mempool.
758 * The packet mbuf to be freed.
760 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
762 struct rte_mbuf *m_next;
764 __rte_mbuf_sanity_check(m, 1);
768 rte_pktmbuf_free_seg(m);
773 #ifdef RTE_MBUF_REFCNT
776 * Creates a "clone" of the given packet mbuf.
778 * Walks through all segments of the given packet mbuf, and for each of them:
779 * - Creates a new packet mbuf from the given pool.
780 * - Attaches newly created mbuf to the segment.
781 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
782 * from the original packet mbuf.
785 * The packet mbuf to be cloned.
787 * The mempool from which the "clone" mbufs are allocated.
789 * - The pointer to the new "clone" mbuf on success.
790 * - NULL if allocation fails.
792 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
793 struct rte_mempool *mp)
795 struct rte_mbuf *mc, *mi, **prev;
799 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
804 pktlen = md->pkt_len;
809 rte_pktmbuf_attach(mi, md);
812 } while ((md = md->next) != NULL &&
813 (mi = rte_pktmbuf_alloc(mp)) != NULL);
817 mc->pkt_len = pktlen;
819 /* Allocation of new indirect segment failed */
820 if (unlikely (mi == NULL)) {
821 rte_pktmbuf_free(mc);
825 __rte_mbuf_sanity_check(mc, 1);
830 * Adds given value to the refcnt of all packet mbuf segments.
832 * Walks through all segments of given packet mbuf and for each of them
833 * invokes rte_mbuf_refcnt_update().
836 * The packet mbuf whose refcnt to be updated.
838 * The value to add to the mbuf's segments refcnt.
840 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
842 __rte_mbuf_sanity_check(m, 1);
845 rte_mbuf_refcnt_update(m, v);
846 } while ((m = m->next) != NULL);
849 #endif /* RTE_MBUF_REFCNT */
852 * Get the headroom in a packet mbuf.
857 * The length of the headroom.
859 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
861 __rte_mbuf_sanity_check(m, 1);
866 * Get the tailroom of a packet mbuf.
871 * The length of the tailroom.
873 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
875 __rte_mbuf_sanity_check(m, 1);
876 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
881 * Get the last segment of the packet.
886 * The last segment of the given mbuf.
888 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
890 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
892 __rte_mbuf_sanity_check(m, 1);
893 while (m2->next != NULL)
899 * A macro that points to the start of the data in the mbuf.
901 * The returned pointer is cast to type t. Before using this
902 * function, the user must ensure that m_headlen(m) is large enough to
908 * The type to cast the result into.
910 #define rte_pktmbuf_mtod(m, t) ((t)((char *)(m)->buf_addr + (m)->data_off))
913 * A macro that returns the length of the packet.
915 * The value can be read or assigned.
920 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
923 * A macro that returns the length of the segment.
925 * The value can be read or assigned.
930 #define rte_pktmbuf_data_len(m) ((m)->data_len)
933 * Prepend len bytes to an mbuf data area.
935 * Returns a pointer to the new
936 * data start address. If there is not enough headroom in the first
937 * segment, the function will return NULL, without modifying the mbuf.
942 * The amount of data to prepend (in bytes).
944 * A pointer to the start of the newly prepended data, or
945 * NULL if there is not enough headroom space in the first segment
947 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
950 __rte_mbuf_sanity_check(m, 1);
952 if (unlikely(len > rte_pktmbuf_headroom(m)))
956 m->data_len = (uint16_t)(m->data_len + len);
957 m->pkt_len = (m->pkt_len + len);
959 return (char *)m->buf_addr + m->data_off;
963 * Append len bytes to an mbuf.
965 * Append len bytes to an mbuf and return a pointer to the start address
966 * of the added data. If there is not enough tailroom in the last
967 * segment, the function will return NULL, without modifying the mbuf.
972 * The amount of data to append (in bytes).
974 * A pointer to the start of the newly appended data, or
975 * NULL if there is not enough tailroom space in the last segment
977 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
980 struct rte_mbuf *m_last;
982 __rte_mbuf_sanity_check(m, 1);
984 m_last = rte_pktmbuf_lastseg(m);
985 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
988 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
989 m_last->data_len = (uint16_t)(m_last->data_len + len);
990 m->pkt_len = (m->pkt_len + len);
995 * Remove len bytes at the beginning of an mbuf.
997 * Returns a pointer to the start address of the new data area. If the
998 * length is greater than the length of the first segment, then the
999 * function will fail and return NULL, without modifying the mbuf.
1004 * The amount of data to remove (in bytes).
1006 * A pointer to the new start of the data.
1008 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1010 __rte_mbuf_sanity_check(m, 1);
1012 if (unlikely(len > m->data_len))
1015 m->data_len = (uint16_t)(m->data_len - len);
1017 m->pkt_len = (m->pkt_len - len);
1018 return (char *)m->buf_addr + m->data_off;
1022 * Remove len bytes of data at the end of the mbuf.
1024 * If the length is greater than the length of the last segment, the
1025 * function will fail and return -1 without modifying the mbuf.
1030 * The amount of data to remove (in bytes).
1035 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1037 struct rte_mbuf *m_last;
1039 __rte_mbuf_sanity_check(m, 1);
1041 m_last = rte_pktmbuf_lastseg(m);
1042 if (unlikely(len > m_last->data_len))
1045 m_last->data_len = (uint16_t)(m_last->data_len - len);
1046 m->pkt_len = (m->pkt_len - len);
1051 * Test if mbuf data is contiguous.
1056 * - 1, if all data is contiguous (one segment).
1057 * - 0, if there is several segments.
1059 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1061 __rte_mbuf_sanity_check(m, 1);
1062 return !!(m->nb_segs == 1);
1066 * Dump an mbuf structure to the console.
1068 * Dump all fields for the given packet mbuf and all its associated
1069 * segments (in the case of a chained buffer).
1072 * A pointer to a file for output
1076 * If dump_len != 0, also dump the "dump_len" first data bytes of
1079 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1085 #endif /* _RTE_MBUF_H_ */