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_atomic.h>
59 #include <rte_prefetch.h>
60 #include <rte_branch_prediction.h>
66 /* deprecated feature, renamed in RTE_MBUF_REFCNT */
67 #pragma GCC poison RTE_MBUF_SCATTER_GATHER
70 * Packet Offload Features Flags. It also carry packet type information.
71 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
73 * - RX flags start at bit position zero, and get added to the left of previous
75 * - The most-significant 8 bits are reserved for generic mbuf flags
76 * - TX flags therefore start at bit position 55 (i.e. 63-8), and new flags get
77 * added to the right of the previously defined flags
79 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
80 * rte_get_tx_ol_flag_name().
82 #define PKT_RX_VLAN_PKT (1ULL << 0) /**< RX packet is a 802.1q VLAN packet. */
83 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
84 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
85 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
86 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
87 #define PKT_RX_EIP_CKSUM_BAD (0ULL << 0) /**< External IP header checksum error. */
88 #define PKT_RX_OVERSIZE (0ULL << 0) /**< Num of desc of an RX pkt oversize. */
89 #define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
90 #define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
91 #define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
92 #define PKT_RX_IPV4_HDR (1ULL << 5) /**< RX packet with IPv4 header. */
93 #define PKT_RX_IPV4_HDR_EXT (1ULL << 6) /**< RX packet with extended IPv4 header. */
94 #define PKT_RX_IPV6_HDR (1ULL << 7) /**< RX packet with IPv6 header. */
95 #define PKT_RX_IPV6_HDR_EXT (1ULL << 8) /**< RX packet with extended IPv6 header. */
96 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
97 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
98 #define PKT_RX_TUNNEL_IPV4_HDR (1ULL << 11) /**< RX tunnel packet with IPv4 header.*/
99 #define PKT_RX_TUNNEL_IPV6_HDR (1ULL << 12) /**< RX tunnel packet with IPv6 header. */
100 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
101 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
102 /* add new RX flags here */
104 /* add new TX flags here */
107 * TCP segmentation offload. To enable this offload feature for a
108 * packet to be transmitted on hardware supporting TSO:
109 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
111 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
112 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
114 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
115 * - calculate the pseudo header checksum without taking ip_len in account,
116 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
117 * rte_ipv6_phdr_cksum() that can be used as helpers.
119 #define PKT_TX_TCP_SEG (1ULL << 49)
121 #define PKT_TX_VXLAN_CKSUM (1ULL << 50) /**< TX checksum of VXLAN computed by NIC */
122 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
125 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
126 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
127 * L4 checksum offload, the user needs to:
128 * - fill l2_len and l3_len in mbuf
129 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
130 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
131 * - calculate the pseudo header checksum and set it in the L4 header (only
132 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
133 * For SCTP, set the crc field to 0.
135 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
136 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
137 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
138 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
139 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
141 #define PKT_TX_IP_CKSUM (1ULL << 54) /**< IP cksum of TX pkt. computed by NIC. */
142 #define PKT_TX_IPV4_CSUM PKT_TX_IP_CKSUM /**< Alias of PKT_TX_IP_CKSUM. */
144 /** Tell the NIC it's an IPv4 packet. Required for L4 checksum offload or TSO. */
145 #define PKT_TX_IPV4 PKT_RX_IPV4_HDR
147 /** Tell the NIC it's an IPv6 packet. Required for L4 checksum offload or TSO. */
148 #define PKT_TX_IPV6 PKT_RX_IPV6_HDR
150 #define PKT_TX_VLAN_PKT (1ULL << 55) /**< TX packet is a 802.1q VLAN packet. */
152 /* Use final bit of flags to indicate a control mbuf */
153 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
156 * Get the name of a RX offload flag
159 * The mask describing the flag.
161 * The name of this flag, or NULL if it's not a valid RX flag.
163 const char *rte_get_rx_ol_flag_name(uint64_t mask);
166 * Get the name of a TX offload flag
169 * The mask describing the flag. Usually only one bit must be set.
170 * Several bits can be given if they belong to the same mask.
171 * Ex: PKT_TX_L4_MASK.
173 * The name of this flag, or NULL if it's not a valid TX flag.
175 const char *rte_get_tx_ol_flag_name(uint64_t mask);
177 /* define a set of marker types that can be used to refer to set points in the
179 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
180 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
181 * with a single assignment */
183 * The generic rte_mbuf, containing a packet mbuf.
188 void *buf_addr; /**< Virtual address of segment buffer. */
189 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
191 /* next 8 bytes are initialised on RX descriptor rearm */
193 uint16_t buf_len; /**< Length of segment buffer. */
197 * 16-bit Reference counter.
198 * It should only be accessed using the following functions:
199 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
200 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
201 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
205 #ifdef RTE_MBUF_REFCNT
206 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
207 uint16_t refcnt; /**< Non-atomically accessed refcnt */
209 uint16_t refcnt_reserved; /**< Do not use this field */
211 uint8_t nb_segs; /**< Number of segments. */
212 uint8_t port; /**< Input port. */
214 uint64_t ol_flags; /**< Offload features. */
216 /* remaining bytes are set on RX when pulling packet from descriptor */
217 MARKER rx_descriptor_fields1;
220 * The packet type, which is used to indicate ordinary packet and also
221 * tunneled packet format, i.e. each number is represented a type of
224 uint16_t packet_type;
226 uint16_t data_len; /**< Amount of data in segment buffer. */
227 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
228 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
231 uint32_t rss; /**< RSS hash result if RSS enabled */
239 /**< Second 4 flexible bytes */
242 /**< First 4 flexible bytes or FD ID, dependent on
243 PKT_RX_FDIR_* flag in ol_flags. */
244 } fdir; /**< Filter identifier if FDIR enabled */
245 uint32_t sched; /**< Hierarchical scheduler */
246 uint32_t usr; /**< User defined tags. See @rte_distributor_process */
247 } hash; /**< hash information */
249 /* second cache line - fields only used in slow path or on TX */
250 MARKER cacheline1 __rte_cache_aligned;
253 void *userdata; /**< Can be used for external metadata */
254 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
257 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
258 struct rte_mbuf *next; /**< Next segment of scattered packet. */
260 /* fields to support TX offloads */
262 uint64_t tx_offload; /**< combined for easy fetch */
264 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
265 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
266 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
267 uint64_t tso_segsz:16; /**< TCP TSO segment size */
269 /* fields for TX offloading of tunnels */
270 uint64_t inner_l3_len:9; /**< inner L3 (IP) Hdr Length. */
271 uint64_t inner_l2_len:7; /**< inner L2 (MAC) Hdr Length. */
273 /* uint64_t unused:8; */
276 } __rte_cache_aligned;
279 * Given the buf_addr returns the pointer to corresponding mbuf.
281 #define RTE_MBUF_FROM_BADDR(ba) (((struct rte_mbuf *)(ba)) - 1)
284 * Given the pointer to mbuf returns an address where it's buf_addr
287 #define RTE_MBUF_TO_BADDR(mb) (((struct rte_mbuf *)(mb)) + 1)
290 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
292 #define RTE_MBUF_INDIRECT(mb) (RTE_MBUF_FROM_BADDR((mb)->buf_addr) != (mb))
295 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
297 #define RTE_MBUF_DIRECT(mb) (RTE_MBUF_FROM_BADDR((mb)->buf_addr) == (mb))
301 * Private data in case of pktmbuf pool.
303 * A structure that contains some pktmbuf_pool-specific data that are
304 * appended after the mempool structure (in private data).
306 struct rte_pktmbuf_pool_private {
307 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf.*/
310 #ifdef RTE_LIBRTE_MBUF_DEBUG
312 /** check mbuf type in debug mode */
313 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
315 /** check mbuf type in debug mode if mbuf pointer is not null */
316 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
318 rte_mbuf_sanity_check(m, is_h); \
321 /** MBUF asserts in debug mode */
322 #define RTE_MBUF_ASSERT(exp) \
324 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
327 #else /* RTE_LIBRTE_MBUF_DEBUG */
329 /** check mbuf type in debug mode */
330 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
332 /** check mbuf type in debug mode if mbuf pointer is not null */
333 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
335 /** MBUF asserts in debug mode */
336 #define RTE_MBUF_ASSERT(exp) do { } while (0)
338 #endif /* RTE_LIBRTE_MBUF_DEBUG */
340 #ifdef RTE_MBUF_REFCNT
341 #ifdef RTE_MBUF_REFCNT_ATOMIC
344 * Adds given value to an mbuf's refcnt and returns its new value.
348 * Value to add/subtract
352 static inline uint16_t
353 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
355 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
359 * Reads the value of an mbuf's refcnt.
363 * Reference count number.
365 static inline uint16_t
366 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
368 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
372 * Sets an mbuf's refcnt to a defined value.
379 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
381 rte_atomic16_set(&m->refcnt_atomic, new_value);
384 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
387 * Adds given value to an mbuf's refcnt and returns its new value.
389 static inline uint16_t
390 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
392 m->refcnt = (uint16_t)(m->refcnt + value);
397 * Reads the value of an mbuf's refcnt.
399 static inline uint16_t
400 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
406 * Sets an mbuf's refcnt to the defined value.
409 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
411 m->refcnt = new_value;
414 #endif /* RTE_MBUF_REFCNT_ATOMIC */
417 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
422 #else /* ! RTE_MBUF_REFCNT */
425 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { } while(0)
427 #define rte_mbuf_refcnt_set(m,v) do { } while(0)
429 #endif /* RTE_MBUF_REFCNT */
433 * Sanity checks on an mbuf.
435 * Check the consistency of the given mbuf. The function will cause a
436 * panic if corruption is detected.
439 * The mbuf to be checked.
441 * True if the mbuf is a packet header, false if it is a sub-segment
442 * of a packet (in this case, some fields like nb_segs are not checked)
445 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
448 * @internal Allocate a new mbuf from mempool *mp*.
449 * The use of that function is reserved for RTE internal needs.
450 * Please use rte_pktmbuf_alloc().
453 * The mempool from which mbuf is allocated.
455 * - The pointer to the new mbuf on success.
456 * - NULL if allocation failed.
458 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
462 if (rte_mempool_get(mp, &mb) < 0)
464 m = (struct rte_mbuf *)mb;
465 #ifdef RTE_MBUF_REFCNT
466 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
467 rte_mbuf_refcnt_set(m, 1);
468 #endif /* RTE_MBUF_REFCNT */
473 * @internal Put mbuf back into its original mempool.
474 * The use of that function is reserved for RTE internal needs.
475 * Please use rte_pktmbuf_free().
478 * The mbuf to be freed.
480 static inline void __attribute__((always_inline))
481 __rte_mbuf_raw_free(struct rte_mbuf *m)
483 #ifdef RTE_MBUF_REFCNT
484 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
485 #endif /* RTE_MBUF_REFCNT */
486 rte_mempool_put(m->pool, m);
489 /* Operations on ctrl mbuf */
492 * The control mbuf constructor.
494 * This function initializes some fields in an mbuf structure that are
495 * not modified by the user once created (mbuf type, origin pool, buffer
496 * start address, and so on). This function is given as a callback function
497 * to rte_mempool_create() at pool creation time.
500 * The mempool from which the mbuf is allocated.
502 * A pointer that can be used by the user to retrieve useful information
503 * for mbuf initialization. This pointer comes from the ``init_arg``
504 * parameter of rte_mempool_create().
506 * The mbuf to initialize.
508 * The index of the mbuf in the pool table.
510 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
511 void *m, unsigned i);
514 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
516 * This new mbuf is initialized with data pointing to the beginning of
517 * buffer, and with a length of zero.
520 * The mempool from which the mbuf is allocated.
522 * - The pointer to the new mbuf on success.
523 * - NULL if allocation failed.
525 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
528 * Free a control mbuf back into its original mempool.
531 * The control mbuf to be freed.
533 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
536 * A macro that returns the pointer to the carried data.
538 * The value that can be read or assigned.
543 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
546 * A macro that returns the length of the carried data.
548 * The value that can be read or assigned.
553 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
556 * Tests if an mbuf is a control mbuf
559 * The mbuf to be tested
561 * - True (1) if the mbuf is a control mbuf
562 * - False(0) otherwise
565 rte_is_ctrlmbuf(struct rte_mbuf *m)
567 return (!!(m->ol_flags & CTRL_MBUF_FLAG));
570 /* Operations on pkt mbuf */
573 * The packet mbuf constructor.
575 * This function initializes some fields in the mbuf structure that are
576 * not modified by the user once created (origin pool, buffer start
577 * address, and so on). This function is given as a callback function to
578 * rte_mempool_create() at pool creation time.
581 * The mempool from which mbufs originate.
583 * A pointer that can be used by the user to retrieve useful information
584 * for mbuf initialization. This pointer comes from the ``init_arg``
585 * parameter of rte_mempool_create().
587 * The mbuf to initialize.
589 * The index of the mbuf in the pool table.
591 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
592 void *m, unsigned i);
596 * A packet mbuf pool constructor.
598 * This function initializes the mempool private data in the case of a
599 * pktmbuf pool. This private data is needed by the driver. The
600 * function is given as a callback function to rte_mempool_create() at
601 * pool creation. It can be extended by the user, for example, to
602 * provide another packet size.
605 * The mempool from which mbufs originate.
607 * A pointer that can be used by the user to retrieve useful information
608 * for mbuf initialization. This pointer comes from the ``init_arg``
609 * parameter of rte_mempool_create().
611 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
614 * Reset the fields of a packet mbuf to their default values.
616 * The given mbuf must have only one segment.
619 * The packet mbuf to be resetted.
621 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
632 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
633 RTE_PKTMBUF_HEADROOM : m->buf_len;
636 __rte_mbuf_sanity_check(m, 1);
640 * Allocate a new mbuf from a mempool.
642 * This new mbuf contains one segment, which has a length of 0. The pointer
643 * to data is initialized to have some bytes of headroom in the buffer
644 * (if buffer size allows).
647 * The mempool from which the mbuf is allocated.
649 * - The pointer to the new mbuf on success.
650 * - NULL if allocation failed.
652 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
655 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
656 rte_pktmbuf_reset(m);
660 #ifdef RTE_MBUF_REFCNT
663 * Attach packet mbuf to another packet mbuf.
664 * After attachment we refer the mbuf we attached as 'indirect',
665 * while mbuf we attached to as 'direct'.
666 * Right now, not supported:
667 * - attachment to indirect mbuf (e.g. - md has to be direct).
668 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
669 * - mbuf we trying to attach (mi) is used by someone else
670 * e.g. it's reference counter is greater then 1.
673 * The indirect packet mbuf.
675 * The direct packet mbuf.
678 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *md)
680 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(md) &&
681 RTE_MBUF_DIRECT(mi) &&
682 rte_mbuf_refcnt_read(mi) == 1);
684 rte_mbuf_refcnt_update(md, 1);
685 mi->buf_physaddr = md->buf_physaddr;
686 mi->buf_addr = md->buf_addr;
687 mi->buf_len = md->buf_len;
690 mi->data_off = md->data_off;
691 mi->data_len = md->data_len;
693 mi->vlan_tci = md->vlan_tci;
694 mi->tx_offload = md->tx_offload;
698 mi->pkt_len = mi->data_len;
700 mi->ol_flags = md->ol_flags;
701 mi->packet_type = md->packet_type;
703 __rte_mbuf_sanity_check(mi, 1);
704 __rte_mbuf_sanity_check(md, 0);
708 * Detach an indirect packet mbuf -
709 * - restore original mbuf address and length values.
710 * - reset pktmbuf data and data_len to their default values.
711 * All other fields of the given packet mbuf will be left intact.
714 * The indirect attached packet mbuf.
717 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
719 const struct rte_mempool *mp = m->pool;
720 void *buf = RTE_MBUF_TO_BADDR(m);
721 uint32_t buf_len = mp->elt_size - sizeof(*m);
722 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + sizeof (*m);
725 m->buf_len = (uint16_t)buf_len;
727 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
728 RTE_PKTMBUF_HEADROOM : m->buf_len;
733 #endif /* RTE_MBUF_REFCNT */
736 static inline struct rte_mbuf* __attribute__((always_inline))
737 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
739 __rte_mbuf_sanity_check(m, 0);
741 #ifdef RTE_MBUF_REFCNT
742 if (likely (rte_mbuf_refcnt_read(m) == 1) ||
743 likely (rte_mbuf_refcnt_update(m, -1) == 0)) {
744 struct rte_mbuf *md = RTE_MBUF_FROM_BADDR(m->buf_addr);
746 rte_mbuf_refcnt_set(m, 0);
748 /* if this is an indirect mbuf, then
750 * - free attached mbuf segment
752 if (unlikely (md != m)) {
753 rte_pktmbuf_detach(m);
754 if (rte_mbuf_refcnt_update(md, -1) == 0)
755 __rte_mbuf_raw_free(md);
759 #ifdef RTE_MBUF_REFCNT
766 * Free a segment of a packet mbuf into its original mempool.
768 * Free an mbuf, without parsing other segments in case of chained
772 * The packet mbuf segment to be freed.
774 static inline void __attribute__((always_inline))
775 rte_pktmbuf_free_seg(struct rte_mbuf *m)
777 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
779 __rte_mbuf_raw_free(m);
784 * Free a packet mbuf back into its original mempool.
786 * Free an mbuf, and all its segments in case of chained buffers. Each
787 * segment is added back into its original mempool.
790 * The packet mbuf to be freed.
792 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
794 struct rte_mbuf *m_next;
796 __rte_mbuf_sanity_check(m, 1);
800 rte_pktmbuf_free_seg(m);
805 #ifdef RTE_MBUF_REFCNT
808 * Creates a "clone" of the given packet mbuf.
810 * Walks through all segments of the given packet mbuf, and for each of them:
811 * - Creates a new packet mbuf from the given pool.
812 * - Attaches newly created mbuf to the segment.
813 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
814 * from the original packet mbuf.
817 * The packet mbuf to be cloned.
819 * The mempool from which the "clone" mbufs are allocated.
821 * - The pointer to the new "clone" mbuf on success.
822 * - NULL if allocation fails.
824 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
825 struct rte_mempool *mp)
827 struct rte_mbuf *mc, *mi, **prev;
831 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
836 pktlen = md->pkt_len;
841 rte_pktmbuf_attach(mi, md);
844 } while ((md = md->next) != NULL &&
845 (mi = rte_pktmbuf_alloc(mp)) != NULL);
849 mc->pkt_len = pktlen;
851 /* Allocation of new indirect segment failed */
852 if (unlikely (mi == NULL)) {
853 rte_pktmbuf_free(mc);
857 __rte_mbuf_sanity_check(mc, 1);
862 * Adds given value to the refcnt of all packet mbuf segments.
864 * Walks through all segments of given packet mbuf and for each of them
865 * invokes rte_mbuf_refcnt_update().
868 * The packet mbuf whose refcnt to be updated.
870 * The value to add to the mbuf's segments refcnt.
872 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
874 __rte_mbuf_sanity_check(m, 1);
877 rte_mbuf_refcnt_update(m, v);
878 } while ((m = m->next) != NULL);
881 #endif /* RTE_MBUF_REFCNT */
884 * Get the headroom in a packet mbuf.
889 * The length of the headroom.
891 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
893 __rte_mbuf_sanity_check(m, 1);
898 * Get the tailroom of a packet mbuf.
903 * The length of the tailroom.
905 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
907 __rte_mbuf_sanity_check(m, 1);
908 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
913 * Get the last segment of the packet.
918 * The last segment of the given mbuf.
920 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
922 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
924 __rte_mbuf_sanity_check(m, 1);
925 while (m2->next != NULL)
931 * A macro that points to the start of the data in the mbuf.
933 * The returned pointer is cast to type t. Before using this
934 * function, the user must ensure that m_headlen(m) is large enough to
940 * The type to cast the result into.
942 #define rte_pktmbuf_mtod(m, t) ((t)((char *)(m)->buf_addr + (m)->data_off))
945 * A macro that returns the length of the packet.
947 * The value can be read or assigned.
952 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
955 * A macro that returns the length of the segment.
957 * The value can be read or assigned.
962 #define rte_pktmbuf_data_len(m) ((m)->data_len)
965 * Prepend len bytes to an mbuf data area.
967 * Returns a pointer to the new
968 * data start address. If there is not enough headroom in the first
969 * segment, the function will return NULL, without modifying the mbuf.
974 * The amount of data to prepend (in bytes).
976 * A pointer to the start of the newly prepended data, or
977 * NULL if there is not enough headroom space in the first segment
979 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
982 __rte_mbuf_sanity_check(m, 1);
984 if (unlikely(len > rte_pktmbuf_headroom(m)))
988 m->data_len = (uint16_t)(m->data_len + len);
989 m->pkt_len = (m->pkt_len + len);
991 return (char *)m->buf_addr + m->data_off;
995 * Append len bytes to an mbuf.
997 * Append len bytes to an mbuf and return a pointer to the start address
998 * of the added data. If there is not enough tailroom in the last
999 * segment, the function will return NULL, without modifying the mbuf.
1004 * The amount of data to append (in bytes).
1006 * A pointer to the start of the newly appended data, or
1007 * NULL if there is not enough tailroom space in the last segment
1009 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1012 struct rte_mbuf *m_last;
1014 __rte_mbuf_sanity_check(m, 1);
1016 m_last = rte_pktmbuf_lastseg(m);
1017 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1020 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1021 m_last->data_len = (uint16_t)(m_last->data_len + len);
1022 m->pkt_len = (m->pkt_len + len);
1023 return (char*) tail;
1027 * Remove len bytes at the beginning of an mbuf.
1029 * Returns a pointer to the start address of the new data area. If the
1030 * length is greater than the length of the first segment, then the
1031 * function will fail and return NULL, without modifying the mbuf.
1036 * The amount of data to remove (in bytes).
1038 * A pointer to the new start of the data.
1040 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1042 __rte_mbuf_sanity_check(m, 1);
1044 if (unlikely(len > m->data_len))
1047 m->data_len = (uint16_t)(m->data_len - len);
1049 m->pkt_len = (m->pkt_len - len);
1050 return (char *)m->buf_addr + m->data_off;
1054 * Remove len bytes of data at the end of the mbuf.
1056 * If the length is greater than the length of the last segment, the
1057 * function will fail and return -1 without modifying the mbuf.
1062 * The amount of data to remove (in bytes).
1067 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1069 struct rte_mbuf *m_last;
1071 __rte_mbuf_sanity_check(m, 1);
1073 m_last = rte_pktmbuf_lastseg(m);
1074 if (unlikely(len > m_last->data_len))
1077 m_last->data_len = (uint16_t)(m_last->data_len - len);
1078 m->pkt_len = (m->pkt_len - len);
1083 * Test if mbuf data is contiguous.
1088 * - 1, if all data is contiguous (one segment).
1089 * - 0, if there is several segments.
1091 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1093 __rte_mbuf_sanity_check(m, 1);
1094 return !!(m->nb_segs == 1);
1098 * Dump an mbuf structure to the console.
1100 * Dump all fields for the given packet mbuf and all its associated
1101 * segments (in the case of a chained buffer).
1104 * A pointer to a file for output
1108 * If dump_len != 0, also dump the "dump_len" first data bytes of
1111 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1117 #endif /* _RTE_MBUF_H_ */