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 options */
68 #pragma GCC poison RTE_MBUF_SCATTER_GATHER
69 #pragma GCC poison RTE_MBUF_REFCNT
72 * Packet Offload Features Flags. It also carry packet type information.
73 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
75 * - RX flags start at bit position zero, and get added to the left of previous
77 * - The most-significant 8 bits are reserved for generic mbuf flags
78 * - TX flags therefore start at bit position 55 (i.e. 63-8), and new flags get
79 * added to the right of the previously defined flags
81 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
82 * rte_get_tx_ol_flag_name().
84 #define PKT_RX_VLAN_PKT (1ULL << 0) /**< RX packet is a 802.1q VLAN packet. */
85 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
86 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
87 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
88 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
89 #define PKT_RX_EIP_CKSUM_BAD (0ULL << 0) /**< External IP header checksum error. */
90 #define PKT_RX_OVERSIZE (0ULL << 0) /**< Num of desc of an RX pkt oversize. */
91 #define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
92 #define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
93 #define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
94 #define PKT_RX_IPV4_HDR (1ULL << 5) /**< RX packet with IPv4 header. */
95 #define PKT_RX_IPV4_HDR_EXT (1ULL << 6) /**< RX packet with extended IPv4 header. */
96 #define PKT_RX_IPV6_HDR (1ULL << 7) /**< RX packet with IPv6 header. */
97 #define PKT_RX_IPV6_HDR_EXT (1ULL << 8) /**< RX packet with extended IPv6 header. */
98 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
99 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
100 #define PKT_RX_TUNNEL_IPV4_HDR (1ULL << 11) /**< RX tunnel packet with IPv4 header.*/
101 #define PKT_RX_TUNNEL_IPV6_HDR (1ULL << 12) /**< RX tunnel packet with IPv6 header. */
102 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
103 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
104 /* add new RX flags here */
106 /* add new TX flags here */
109 * TCP segmentation offload. To enable this offload feature for a
110 * packet to be transmitted on hardware supporting TSO:
111 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
113 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
114 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
116 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
117 * - calculate the pseudo header checksum without taking ip_len in account,
118 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
119 * rte_ipv6_phdr_cksum() that can be used as helpers.
121 #define PKT_TX_TCP_SEG (1ULL << 50)
123 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
126 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
127 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
128 * L4 checksum offload, the user needs to:
129 * - fill l2_len and l3_len in mbuf
130 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
131 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
132 * - calculate the pseudo header checksum and set it in the L4 header (only
133 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
134 * For SCTP, set the crc field to 0.
136 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
137 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
138 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
139 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
140 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
143 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
144 * also be set by the application, although a PMD will only check
146 * - set the IP checksum field in the packet to 0
147 * - fill the mbuf offload information: l2_len, l3_len
149 #define PKT_TX_IP_CKSUM (1ULL << 54)
152 * Packet is IPv4. This flag must be set when using any offload feature
153 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
154 * packet. If the packet is a tunneled packet, this flag is related to
157 #define PKT_TX_IPV4 (1ULL << 55)
160 * Packet is IPv6. This flag must be set when using an offload feature
161 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
162 * packet. If the packet is a tunneled packet, this flag is related to
165 #define PKT_TX_IPV6 (1ULL << 56)
167 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
170 * Offload the IP checksum of an external header in the hardware. The
171 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
172 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
173 * packet must be set to 0.
174 * - set the outer IP checksum field in the packet to 0
175 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
177 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
180 * Packet outer header is IPv4. This flag must be set when using any
181 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
182 * outer header of the tunneled packet is an IPv4 packet.
184 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
187 * Packet outer header is IPv6. This flag must be set when using any
188 * outer offload feature (L4 checksum) to tell the NIC that the outer
189 * header of the tunneled packet is an IPv6 packet.
191 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
193 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
195 /* Use final bit of flags to indicate a control mbuf */
196 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
199 * Get the name of a RX offload flag
202 * The mask describing the flag.
204 * The name of this flag, or NULL if it's not a valid RX flag.
206 const char *rte_get_rx_ol_flag_name(uint64_t mask);
209 * Get the name of a TX offload flag
212 * The mask describing the flag. Usually only one bit must be set.
213 * Several bits can be given if they belong to the same mask.
214 * Ex: PKT_TX_L4_MASK.
216 * The name of this flag, or NULL if it's not a valid TX flag.
218 const char *rte_get_tx_ol_flag_name(uint64_t mask);
220 /* define a set of marker types that can be used to refer to set points in the
222 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
223 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
224 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
225 * with a single assignment */
228 * The generic rte_mbuf, containing a packet mbuf.
233 void *buf_addr; /**< Virtual address of segment buffer. */
234 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
236 uint16_t buf_len; /**< Length of segment buffer. */
238 /* next 6 bytes are initialised on RX descriptor rearm */
243 * 16-bit Reference counter.
244 * It should only be accessed using the following functions:
245 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
246 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
247 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
251 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
252 uint16_t refcnt; /**< Non-atomically accessed refcnt */
254 uint8_t nb_segs; /**< Number of segments. */
255 uint8_t port; /**< Input port. */
257 uint64_t ol_flags; /**< Offload features. */
259 /* remaining bytes are set on RX when pulling packet from descriptor */
260 MARKER rx_descriptor_fields1;
263 * The packet type, which is used to indicate ordinary packet and also
264 * tunneled packet format, i.e. each number is represented a type of
267 uint16_t packet_type;
269 uint16_t data_len; /**< Amount of data in segment buffer. */
270 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
271 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
274 uint32_t rss; /**< RSS hash result if RSS enabled */
282 /**< Second 4 flexible bytes */
285 /**< First 4 flexible bytes or FD ID, dependent on
286 PKT_RX_FDIR_* flag in ol_flags. */
287 } fdir; /**< Filter identifier if FDIR enabled */
288 uint32_t sched; /**< Hierarchical scheduler */
289 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
290 } hash; /**< hash information */
292 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
294 /* second cache line - fields only used in slow path or on TX */
295 MARKER cacheline1 __rte_cache_aligned;
298 void *userdata; /**< Can be used for external metadata */
299 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
302 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
303 struct rte_mbuf *next; /**< Next segment of scattered packet. */
305 /* fields to support TX offloads */
307 uint64_t tx_offload; /**< combined for easy fetch */
309 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
310 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
311 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
312 uint64_t tso_segsz:16; /**< TCP TSO segment size */
314 /* fields for TX offloading of tunnels */
315 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
316 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
318 /* uint64_t unused:8; */
321 } __rte_cache_aligned;
324 * Given the buf_addr returns the pointer to corresponding mbuf.
326 #define RTE_MBUF_FROM_BADDR(ba) (((struct rte_mbuf *)(ba)) - 1)
329 * Given the pointer to mbuf returns an address where it's buf_addr
332 #define RTE_MBUF_TO_BADDR(mb) (((struct rte_mbuf *)(mb)) + 1)
335 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
337 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
340 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
342 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
345 * Private data in case of pktmbuf pool.
347 * A structure that contains some pktmbuf_pool-specific data that are
348 * appended after the mempool structure (in private data).
350 struct rte_pktmbuf_pool_private {
351 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
352 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
355 #ifdef RTE_LIBRTE_MBUF_DEBUG
357 /** check mbuf type in debug mode */
358 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
360 /** check mbuf type in debug mode if mbuf pointer is not null */
361 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
363 rte_mbuf_sanity_check(m, is_h); \
366 /** MBUF asserts in debug mode */
367 #define RTE_MBUF_ASSERT(exp) \
369 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
372 #else /* RTE_LIBRTE_MBUF_DEBUG */
374 /** check mbuf type in debug mode */
375 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
377 /** check mbuf type in debug mode if mbuf pointer is not null */
378 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
380 /** MBUF asserts in debug mode */
381 #define RTE_MBUF_ASSERT(exp) do { } while (0)
383 #endif /* RTE_LIBRTE_MBUF_DEBUG */
385 #ifdef RTE_MBUF_REFCNT_ATOMIC
388 * Adds given value to an mbuf's refcnt and returns its new value.
392 * Value to add/subtract
396 static inline uint16_t
397 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
399 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
403 * Reads the value of an mbuf's refcnt.
407 * Reference count number.
409 static inline uint16_t
410 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
412 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
416 * Sets an mbuf's refcnt to a defined value.
423 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
425 rte_atomic16_set(&m->refcnt_atomic, new_value);
428 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
431 * Adds given value to an mbuf's refcnt and returns its new value.
433 static inline uint16_t
434 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
436 m->refcnt = (uint16_t)(m->refcnt + value);
441 * Reads the value of an mbuf's refcnt.
443 static inline uint16_t
444 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
450 * Sets an mbuf's refcnt to the defined value.
453 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
455 m->refcnt = new_value;
458 #endif /* RTE_MBUF_REFCNT_ATOMIC */
461 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
468 * Sanity checks on an mbuf.
470 * Check the consistency of the given mbuf. The function will cause a
471 * panic if corruption is detected.
474 * The mbuf to be checked.
476 * True if the mbuf is a packet header, false if it is a sub-segment
477 * of a packet (in this case, some fields like nb_segs are not checked)
480 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
483 * @internal Allocate a new mbuf from mempool *mp*.
484 * The use of that function is reserved for RTE internal needs.
485 * Please use rte_pktmbuf_alloc().
488 * The mempool from which mbuf is allocated.
490 * - The pointer to the new mbuf on success.
491 * - NULL if allocation failed.
493 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
497 if (rte_mempool_get(mp, &mb) < 0)
499 m = (struct rte_mbuf *)mb;
500 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
501 rte_mbuf_refcnt_set(m, 1);
506 * @internal Put mbuf back into its original mempool.
507 * The use of that function is reserved for RTE internal needs.
508 * Please use rte_pktmbuf_free().
511 * The mbuf to be freed.
513 static inline void __attribute__((always_inline))
514 __rte_mbuf_raw_free(struct rte_mbuf *m)
516 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
517 rte_mempool_put(m->pool, m);
520 /* Operations on ctrl mbuf */
523 * The control mbuf constructor.
525 * This function initializes some fields in an mbuf structure that are
526 * not modified by the user once created (mbuf type, origin pool, buffer
527 * start address, and so on). This function is given as a callback function
528 * to rte_mempool_create() at pool creation time.
531 * The mempool from which the mbuf is allocated.
533 * A pointer that can be used by the user to retrieve useful information
534 * for mbuf initialization. This pointer comes from the ``init_arg``
535 * parameter of rte_mempool_create().
537 * The mbuf to initialize.
539 * The index of the mbuf in the pool table.
541 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
542 void *m, unsigned i);
545 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
547 * This new mbuf is initialized with data pointing to the beginning of
548 * buffer, and with a length of zero.
551 * The mempool from which the mbuf is allocated.
553 * - The pointer to the new mbuf on success.
554 * - NULL if allocation failed.
556 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
559 * Free a control mbuf back into its original mempool.
562 * The control mbuf to be freed.
564 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
567 * A macro that returns the pointer to the carried data.
569 * The value that can be read or assigned.
574 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
577 * A macro that returns the length of the carried data.
579 * The value that can be read or assigned.
584 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
587 * Tests if an mbuf is a control mbuf
590 * The mbuf to be tested
592 * - True (1) if the mbuf is a control mbuf
593 * - False(0) otherwise
596 rte_is_ctrlmbuf(struct rte_mbuf *m)
598 return (!!(m->ol_flags & CTRL_MBUF_FLAG));
601 /* Operations on pkt mbuf */
604 * The packet mbuf constructor.
606 * This function initializes some fields in the mbuf structure that are
607 * not modified by the user once created (origin pool, buffer start
608 * address, and so on). This function is given as a callback function to
609 * rte_mempool_create() at pool creation time.
612 * The mempool from which mbufs originate.
614 * A pointer that can be used by the user to retrieve useful information
615 * for mbuf initialization. This pointer comes from the ``init_arg``
616 * parameter of rte_mempool_create().
618 * The mbuf to initialize.
620 * The index of the mbuf in the pool table.
622 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
623 void *m, unsigned i);
627 * A packet mbuf pool constructor.
629 * This function initializes the mempool private data in the case of a
630 * pktmbuf pool. This private data is needed by the driver. The
631 * function is given as a callback function to rte_mempool_create() at
632 * pool creation. It can be extended by the user, for example, to
633 * provide another packet size.
636 * The mempool from which mbufs originate.
638 * A pointer that can be used by the user to retrieve useful information
639 * for mbuf initialization. This pointer comes from the ``init_arg``
640 * parameter of rte_mempool_create().
642 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
645 * Get the data room size of mbufs stored in a pktmbuf_pool
647 * The data room size is the amount of data that can be stored in a
648 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
651 * The packet mbuf pool.
653 * The data room size of mbufs stored in this mempool.
655 static inline uint16_t
656 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
658 struct rte_pktmbuf_pool_private *mbp_priv;
660 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
661 return mbp_priv->mbuf_data_room_size;
665 * Get the application private size of mbufs stored in a pktmbuf_pool
667 * The private size of mbuf is a zone located between the rte_mbuf
668 * structure and the data buffer where an application can store data
669 * associated to a packet.
672 * The packet mbuf pool.
674 * The private size of mbufs stored in this mempool.
676 static inline uint16_t
677 rte_pktmbuf_priv_size(struct rte_mempool *mp)
679 struct rte_pktmbuf_pool_private *mbp_priv;
681 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
682 return mbp_priv->mbuf_priv_size;
686 * Reset the fields of a packet mbuf to their default values.
688 * The given mbuf must have only one segment.
691 * The packet mbuf to be resetted.
693 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
704 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
705 RTE_PKTMBUF_HEADROOM : m->buf_len;
708 __rte_mbuf_sanity_check(m, 1);
712 * Allocate a new mbuf from a mempool.
714 * This new mbuf contains one segment, which has a length of 0. The pointer
715 * to data is initialized to have some bytes of headroom in the buffer
716 * (if buffer size allows).
719 * The mempool from which the mbuf is allocated.
721 * - The pointer to the new mbuf on success.
722 * - NULL if allocation failed.
724 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
727 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
728 rte_pktmbuf_reset(m);
733 * Attach packet mbuf to another packet mbuf.
734 * After attachment we refer the mbuf we attached as 'indirect',
735 * while mbuf we attached to as 'direct'.
736 * Right now, not supported:
737 * - attachment to indirect mbuf (e.g. - md has to be direct).
738 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
739 * - mbuf we trying to attach (mi) is used by someone else
740 * e.g. it's reference counter is greater then 1.
743 * The indirect packet mbuf.
745 * The direct packet mbuf.
748 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *md)
750 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(md) &&
751 RTE_MBUF_DIRECT(mi) &&
752 rte_mbuf_refcnt_read(mi) == 1);
754 rte_mbuf_refcnt_update(md, 1);
755 mi->buf_physaddr = md->buf_physaddr;
756 mi->buf_addr = md->buf_addr;
757 mi->buf_len = md->buf_len;
760 mi->data_off = md->data_off;
761 mi->data_len = md->data_len;
763 mi->vlan_tci = md->vlan_tci;
764 mi->tx_offload = md->tx_offload;
768 mi->pkt_len = mi->data_len;
770 mi->ol_flags = md->ol_flags | IND_ATTACHED_MBUF;
771 mi->packet_type = md->packet_type;
773 __rte_mbuf_sanity_check(mi, 1);
774 __rte_mbuf_sanity_check(md, 0);
778 * Detach an indirect packet mbuf -
779 * - restore original mbuf address and length values.
780 * - reset pktmbuf data and data_len to their default values.
781 * All other fields of the given packet mbuf will be left intact.
784 * The indirect attached packet mbuf.
787 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
789 const struct rte_mempool *mp = m->pool;
790 void *buf = RTE_MBUF_TO_BADDR(m);
791 uint32_t buf_len = mp->elt_size - sizeof(*m);
792 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + sizeof (*m);
795 m->buf_len = (uint16_t)buf_len;
797 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
798 RTE_PKTMBUF_HEADROOM : m->buf_len;
805 static inline struct rte_mbuf* __attribute__((always_inline))
806 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
808 __rte_mbuf_sanity_check(m, 0);
811 * Check to see if this is the last reference to the mbuf.
812 * Note: the double check here is deliberate. If the ref_cnt is "atomic"
813 * the call to "refcnt_update" is a very expensive operation, so we
814 * don't want to call it in the case where we know we are the holder
815 * of the last reference to this mbuf i.e. ref_cnt == 1.
816 * If however, ref_cnt != 1, it's still possible that we may still be
817 * the final decrementer of the count, so we need to check that
818 * result also, to make sure the mbuf is freed properly.
820 if (likely (rte_mbuf_refcnt_read(m) == 1) ||
821 likely (rte_mbuf_refcnt_update(m, -1) == 0)) {
823 rte_mbuf_refcnt_set(m, 0);
825 /* if this is an indirect mbuf, then
827 * - free attached mbuf segment
829 if (RTE_MBUF_INDIRECT(m)) {
830 struct rte_mbuf *md = RTE_MBUF_FROM_BADDR(m->buf_addr);
831 rte_pktmbuf_detach(m);
832 if (rte_mbuf_refcnt_update(md, -1) == 0)
833 __rte_mbuf_raw_free(md);
841 * Free a segment of a packet mbuf into its original mempool.
843 * Free an mbuf, without parsing other segments in case of chained
847 * The packet mbuf segment to be freed.
849 static inline void __attribute__((always_inline))
850 rte_pktmbuf_free_seg(struct rte_mbuf *m)
852 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
854 __rte_mbuf_raw_free(m);
859 * Free a packet mbuf back into its original mempool.
861 * Free an mbuf, and all its segments in case of chained buffers. Each
862 * segment is added back into its original mempool.
865 * The packet mbuf to be freed.
867 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
869 struct rte_mbuf *m_next;
871 __rte_mbuf_sanity_check(m, 1);
875 rte_pktmbuf_free_seg(m);
881 * Creates a "clone" of the given packet mbuf.
883 * Walks through all segments of the given packet mbuf, and for each of them:
884 * - Creates a new packet mbuf from the given pool.
885 * - Attaches newly created mbuf to the segment.
886 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
887 * from the original packet mbuf.
890 * The packet mbuf to be cloned.
892 * The mempool from which the "clone" mbufs are allocated.
894 * - The pointer to the new "clone" mbuf on success.
895 * - NULL if allocation fails.
897 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
898 struct rte_mempool *mp)
900 struct rte_mbuf *mc, *mi, **prev;
904 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
909 pktlen = md->pkt_len;
914 rte_pktmbuf_attach(mi, md);
917 } while ((md = md->next) != NULL &&
918 (mi = rte_pktmbuf_alloc(mp)) != NULL);
922 mc->pkt_len = pktlen;
924 /* Allocation of new indirect segment failed */
925 if (unlikely (mi == NULL)) {
926 rte_pktmbuf_free(mc);
930 __rte_mbuf_sanity_check(mc, 1);
935 * Adds given value to the refcnt of all packet mbuf segments.
937 * Walks through all segments of given packet mbuf and for each of them
938 * invokes rte_mbuf_refcnt_update().
941 * The packet mbuf whose refcnt to be updated.
943 * The value to add to the mbuf's segments refcnt.
945 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
947 __rte_mbuf_sanity_check(m, 1);
950 rte_mbuf_refcnt_update(m, v);
951 } while ((m = m->next) != NULL);
955 * Get the headroom in a packet mbuf.
960 * The length of the headroom.
962 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
964 __rte_mbuf_sanity_check(m, 1);
969 * Get the tailroom of a packet mbuf.
974 * The length of the tailroom.
976 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
978 __rte_mbuf_sanity_check(m, 1);
979 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
984 * Get the last segment of the packet.
989 * The last segment of the given mbuf.
991 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
993 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
995 __rte_mbuf_sanity_check(m, 1);
996 while (m2->next != NULL)
1002 * A macro that points to the start of the data in the mbuf.
1004 * The returned pointer is cast to type t. Before using this
1005 * function, the user must ensure that m_headlen(m) is large enough to
1011 * The type to cast the result into.
1013 #define rte_pktmbuf_mtod(m, t) ((t)((char *)(m)->buf_addr + (m)->data_off))
1016 * A macro that returns the length of the packet.
1018 * The value can be read or assigned.
1023 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1026 * A macro that returns the length of the segment.
1028 * The value can be read or assigned.
1033 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1036 * Prepend len bytes to an mbuf data area.
1038 * Returns a pointer to the new
1039 * data start address. If there is not enough headroom in the first
1040 * segment, the function will return NULL, without modifying the mbuf.
1045 * The amount of data to prepend (in bytes).
1047 * A pointer to the start of the newly prepended data, or
1048 * NULL if there is not enough headroom space in the first segment
1050 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1053 __rte_mbuf_sanity_check(m, 1);
1055 if (unlikely(len > rte_pktmbuf_headroom(m)))
1059 m->data_len = (uint16_t)(m->data_len + len);
1060 m->pkt_len = (m->pkt_len + len);
1062 return (char *)m->buf_addr + m->data_off;
1066 * Append len bytes to an mbuf.
1068 * Append len bytes to an mbuf and return a pointer to the start address
1069 * of the added data. If there is not enough tailroom in the last
1070 * segment, the function will return NULL, without modifying the mbuf.
1075 * The amount of data to append (in bytes).
1077 * A pointer to the start of the newly appended data, or
1078 * NULL if there is not enough tailroom space in the last segment
1080 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1083 struct rte_mbuf *m_last;
1085 __rte_mbuf_sanity_check(m, 1);
1087 m_last = rte_pktmbuf_lastseg(m);
1088 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1091 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1092 m_last->data_len = (uint16_t)(m_last->data_len + len);
1093 m->pkt_len = (m->pkt_len + len);
1094 return (char*) tail;
1098 * Remove len bytes at the beginning of an mbuf.
1100 * Returns a pointer to the start address of the new data area. If the
1101 * length is greater than the length of the first segment, then the
1102 * function will fail and return NULL, without modifying the mbuf.
1107 * The amount of data to remove (in bytes).
1109 * A pointer to the new start of the data.
1111 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1113 __rte_mbuf_sanity_check(m, 1);
1115 if (unlikely(len > m->data_len))
1118 m->data_len = (uint16_t)(m->data_len - len);
1120 m->pkt_len = (m->pkt_len - len);
1121 return (char *)m->buf_addr + m->data_off;
1125 * Remove len bytes of data at the end of the mbuf.
1127 * If the length is greater than the length of the last segment, the
1128 * function will fail and return -1 without modifying the mbuf.
1133 * The amount of data to remove (in bytes).
1138 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1140 struct rte_mbuf *m_last;
1142 __rte_mbuf_sanity_check(m, 1);
1144 m_last = rte_pktmbuf_lastseg(m);
1145 if (unlikely(len > m_last->data_len))
1148 m_last->data_len = (uint16_t)(m_last->data_len - len);
1149 m->pkt_len = (m->pkt_len - len);
1154 * Test if mbuf data is contiguous.
1159 * - 1, if all data is contiguous (one segment).
1160 * - 0, if there is several segments.
1162 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1164 __rte_mbuf_sanity_check(m, 1);
1165 return !!(m->nb_segs == 1);
1169 * Dump an mbuf structure to the console.
1171 * Dump all fields for the given packet mbuf and all its associated
1172 * segments (in the case of a chained buffer).
1175 * A pointer to a file for output
1179 * If dump_len != 0, also dump the "dump_len" first data bytes of
1182 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1188 #endif /* _RTE_MBUF_H_ */