4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved.
5 * Copyright 2014 6WIND S.A.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
12 * * Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * * Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in
16 * the documentation and/or other materials provided with the
18 * * Neither the name of Intel Corporation nor the names of its
19 * contributors may be used to endorse or promote products derived
20 * from this software without specific prior written permission.
22 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
23 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
24 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
25 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
26 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
27 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
28 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
29 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
30 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
31 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
32 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
42 * The mbuf library provides the ability to create and destroy buffers
43 * that may be used by the RTE application to store message
44 * buffers. The message buffers are stored in a mempool, using the
45 * RTE mempool library.
47 * This library provide an API to allocate/free packet mbufs, which are
48 * used to carry network packets.
50 * To understand the concepts of packet buffers or mbufs, you
51 * should read "TCP/IP Illustrated, Volume 2: The Implementation,
52 * Addison-Wesley, 1995, ISBN 0-201-63354-X from Richard Stevens"
53 * http://www.kohala.com/start/tcpipiv2.html
57 #include <rte_mempool.h>
58 #include <rte_memory.h>
59 #include <rte_atomic.h>
60 #include <rte_prefetch.h>
61 #include <rte_branch_prediction.h>
67 /* deprecated feature, renamed in RTE_MBUF_REFCNT */
68 #pragma GCC poison RTE_MBUF_SCATTER_GATHER
71 * Packet Offload Features Flags. It also carry packet type information.
72 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
74 * - RX flags start at bit position zero, and get added to the left of previous
76 * - The most-significant 8 bits are reserved for generic mbuf flags
77 * - TX flags therefore start at bit position 55 (i.e. 63-8), and new flags get
78 * added to the right of the previously defined flags
80 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
81 * rte_get_tx_ol_flag_name().
83 #define PKT_RX_VLAN_PKT (1ULL << 0) /**< RX packet is a 802.1q VLAN packet. */
84 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
85 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
86 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
87 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
88 #define PKT_RX_EIP_CKSUM_BAD (0ULL << 0) /**< External IP header checksum error. */
89 #define PKT_RX_OVERSIZE (0ULL << 0) /**< Num of desc of an RX pkt oversize. */
90 #define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
91 #define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
92 #define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
93 #define PKT_RX_IPV4_HDR (1ULL << 5) /**< RX packet with IPv4 header. */
94 #define PKT_RX_IPV4_HDR_EXT (1ULL << 6) /**< RX packet with extended IPv4 header. */
95 #define PKT_RX_IPV6_HDR (1ULL << 7) /**< RX packet with IPv6 header. */
96 #define PKT_RX_IPV6_HDR_EXT (1ULL << 8) /**< RX packet with extended IPv6 header. */
97 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
98 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
99 #define PKT_RX_TUNNEL_IPV4_HDR (1ULL << 11) /**< RX tunnel packet with IPv4 header.*/
100 #define PKT_RX_TUNNEL_IPV6_HDR (1ULL << 12) /**< RX tunnel packet with IPv6 header. */
101 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
102 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
103 /* add new RX flags here */
105 /* add new TX flags here */
108 * TCP segmentation offload. To enable this offload feature for a
109 * packet to be transmitted on hardware supporting TSO:
110 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
112 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
113 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
115 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
116 * - calculate the pseudo header checksum without taking ip_len in account,
117 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
118 * rte_ipv6_phdr_cksum() that can be used as helpers.
120 #define PKT_TX_TCP_SEG (1ULL << 50)
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. */
142 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
143 * also be set by the application, although a PMD will only check
145 * - set the IP checksum field in the packet to 0
146 * - fill the mbuf offload information: l2_len, l3_len
148 #define PKT_TX_IP_CKSUM (1ULL << 54)
151 * Packet is IPv4. This flag must be set when using any offload feature
152 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
153 * packet. If the packet is a tunneled packet, this flag is related to
156 #define PKT_TX_IPV4 (1ULL << 55)
159 * Packet is IPv6. This flag must be set when using an offload feature
160 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
161 * packet. If the packet is a tunneled packet, this flag is related to
164 #define PKT_TX_IPV6 (1ULL << 56)
166 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
169 * Offload the IP checksum of an external header in the hardware. The
170 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
171 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
172 * packet must be set to 0.
173 * - set the outer IP checksum field in the packet to 0
174 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
176 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
179 * Packet outer header is IPv4. This flag must be set when using any
180 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
181 * outer header of the tunneled packet is an IPv4 packet.
183 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
186 * Packet outer header is IPv6. This flag must be set when using any
187 * outer offload feature (L4 checksum) to tell the NIC that the outer
188 * header of the tunneled packet is an IPv6 packet.
190 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
192 /* Use final bit of flags to indicate a control mbuf */
193 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
196 * Get the name of a RX offload flag
199 * The mask describing the flag.
201 * The name of this flag, or NULL if it's not a valid RX flag.
203 const char *rte_get_rx_ol_flag_name(uint64_t mask);
206 * Get the name of a TX offload flag
209 * The mask describing the flag. Usually only one bit must be set.
210 * Several bits can be given if they belong to the same mask.
211 * Ex: PKT_TX_L4_MASK.
213 * The name of this flag, or NULL if it's not a valid TX flag.
215 const char *rte_get_tx_ol_flag_name(uint64_t mask);
217 /* define a set of marker types that can be used to refer to set points in the
219 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
220 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
221 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
222 * with a single assignment */
225 * The generic rte_mbuf, containing a packet mbuf.
230 void *buf_addr; /**< Virtual address of segment buffer. */
231 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
233 uint16_t buf_len; /**< Length of segment buffer. */
235 /* next 6 bytes are initialised on RX descriptor rearm */
240 * 16-bit Reference counter.
241 * It should only be accessed using the following functions:
242 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
243 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
244 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
248 #ifdef RTE_MBUF_REFCNT
249 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
250 uint16_t refcnt; /**< Non-atomically accessed refcnt */
252 uint16_t refcnt_reserved; /**< Do not use this field */
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 /* sequence number - field used in distributor and reorder library */
295 /* second cache line - fields only used in slow path or on TX */
296 MARKER cacheline1 __rte_cache_aligned;
299 void *userdata; /**< Can be used for external metadata */
300 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
303 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
304 struct rte_mbuf *next; /**< Next segment of scattered packet. */
306 /* fields to support TX offloads */
308 uint64_t tx_offload; /**< combined for easy fetch */
310 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
311 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
312 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
313 uint64_t tso_segsz:16; /**< TCP TSO segment size */
315 /* fields for TX offloading of tunnels */
316 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
317 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
319 /* uint64_t unused:8; */
322 } __rte_cache_aligned;
325 * Given the buf_addr returns the pointer to corresponding mbuf.
327 #define RTE_MBUF_FROM_BADDR(ba) (((struct rte_mbuf *)(ba)) - 1)
330 * Given the pointer to mbuf returns an address where it's buf_addr
333 #define RTE_MBUF_TO_BADDR(mb) (((struct rte_mbuf *)(mb)) + 1)
336 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
338 #define RTE_MBUF_INDIRECT(mb) (RTE_MBUF_FROM_BADDR((mb)->buf_addr) != (mb))
341 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
343 #define RTE_MBUF_DIRECT(mb) (RTE_MBUF_FROM_BADDR((mb)->buf_addr) == (mb))
347 * Private data in case of pktmbuf pool.
349 * A structure that contains some pktmbuf_pool-specific data that are
350 * appended after the mempool structure (in private data).
352 struct rte_pktmbuf_pool_private {
353 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf.*/
356 #ifdef RTE_LIBRTE_MBUF_DEBUG
358 /** check mbuf type in debug mode */
359 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
361 /** check mbuf type in debug mode if mbuf pointer is not null */
362 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
364 rte_mbuf_sanity_check(m, is_h); \
367 /** MBUF asserts in debug mode */
368 #define RTE_MBUF_ASSERT(exp) \
370 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
373 #else /* RTE_LIBRTE_MBUF_DEBUG */
375 /** check mbuf type in debug mode */
376 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
378 /** check mbuf type in debug mode if mbuf pointer is not null */
379 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
381 /** MBUF asserts in debug mode */
382 #define RTE_MBUF_ASSERT(exp) do { } while (0)
384 #endif /* RTE_LIBRTE_MBUF_DEBUG */
386 #ifdef RTE_MBUF_REFCNT
387 #ifdef RTE_MBUF_REFCNT_ATOMIC
390 * Adds given value to an mbuf's refcnt and returns its new value.
394 * Value to add/subtract
398 static inline uint16_t
399 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
401 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
405 * Reads the value of an mbuf's refcnt.
409 * Reference count number.
411 static inline uint16_t
412 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
414 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
418 * Sets an mbuf's refcnt to a defined value.
425 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
427 rte_atomic16_set(&m->refcnt_atomic, new_value);
430 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
433 * Adds given value to an mbuf's refcnt and returns its new value.
435 static inline uint16_t
436 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
438 m->refcnt = (uint16_t)(m->refcnt + value);
443 * Reads the value of an mbuf's refcnt.
445 static inline uint16_t
446 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
452 * Sets an mbuf's refcnt to the defined value.
455 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
457 m->refcnt = new_value;
460 #endif /* RTE_MBUF_REFCNT_ATOMIC */
463 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
468 #else /* ! RTE_MBUF_REFCNT */
471 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { } while(0)
473 #define rte_mbuf_refcnt_set(m,v) do { } while(0)
475 #endif /* RTE_MBUF_REFCNT */
479 * Sanity checks on an mbuf.
481 * Check the consistency of the given mbuf. The function will cause a
482 * panic if corruption is detected.
485 * The mbuf to be checked.
487 * True if the mbuf is a packet header, false if it is a sub-segment
488 * of a packet (in this case, some fields like nb_segs are not checked)
491 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
494 * @internal Allocate a new mbuf from mempool *mp*.
495 * The use of that function is reserved for RTE internal needs.
496 * Please use rte_pktmbuf_alloc().
499 * The mempool from which mbuf is allocated.
501 * - The pointer to the new mbuf on success.
502 * - NULL if allocation failed.
504 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
508 if (rte_mempool_get(mp, &mb) < 0)
510 m = (struct rte_mbuf *)mb;
511 #ifdef RTE_MBUF_REFCNT
512 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
513 rte_mbuf_refcnt_set(m, 1);
514 #endif /* RTE_MBUF_REFCNT */
519 * @internal Put mbuf back into its original mempool.
520 * The use of that function is reserved for RTE internal needs.
521 * Please use rte_pktmbuf_free().
524 * The mbuf to be freed.
526 static inline void __attribute__((always_inline))
527 __rte_mbuf_raw_free(struct rte_mbuf *m)
529 #ifdef RTE_MBUF_REFCNT
530 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
531 #endif /* RTE_MBUF_REFCNT */
532 rte_mempool_put(m->pool, m);
535 /* Operations on ctrl mbuf */
538 * The control mbuf constructor.
540 * This function initializes some fields in an mbuf structure that are
541 * not modified by the user once created (mbuf type, origin pool, buffer
542 * start address, and so on). This function is given as a callback function
543 * to rte_mempool_create() at pool creation time.
546 * The mempool from which the mbuf is allocated.
548 * A pointer that can be used by the user to retrieve useful information
549 * for mbuf initialization. This pointer comes from the ``init_arg``
550 * parameter of rte_mempool_create().
552 * The mbuf to initialize.
554 * The index of the mbuf in the pool table.
556 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
557 void *m, unsigned i);
560 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
562 * This new mbuf is initialized with data pointing to the beginning of
563 * buffer, and with a length of zero.
566 * The mempool from which the mbuf is allocated.
568 * - The pointer to the new mbuf on success.
569 * - NULL if allocation failed.
571 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
574 * Free a control mbuf back into its original mempool.
577 * The control mbuf to be freed.
579 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
582 * A macro that returns the pointer to the carried data.
584 * The value that can be read or assigned.
589 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
592 * A macro that returns the length of the carried data.
594 * The value that can be read or assigned.
599 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
602 * Tests if an mbuf is a control mbuf
605 * The mbuf to be tested
607 * - True (1) if the mbuf is a control mbuf
608 * - False(0) otherwise
611 rte_is_ctrlmbuf(struct rte_mbuf *m)
613 return (!!(m->ol_flags & CTRL_MBUF_FLAG));
616 /* Operations on pkt mbuf */
619 * The packet mbuf constructor.
621 * This function initializes some fields in the mbuf structure that are
622 * not modified by the user once created (origin pool, buffer start
623 * address, and so on). This function is given as a callback function to
624 * rte_mempool_create() at pool creation time.
627 * The mempool from which mbufs originate.
629 * A pointer that can be used by the user to retrieve useful information
630 * for mbuf initialization. This pointer comes from the ``init_arg``
631 * parameter of rte_mempool_create().
633 * The mbuf to initialize.
635 * The index of the mbuf in the pool table.
637 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
638 void *m, unsigned i);
642 * A packet mbuf pool constructor.
644 * This function initializes the mempool private data in the case of a
645 * pktmbuf pool. This private data is needed by the driver. The
646 * function is given as a callback function to rte_mempool_create() at
647 * pool creation. It can be extended by the user, for example, to
648 * provide another packet size.
651 * The mempool from which mbufs originate.
653 * A pointer that can be used by the user to retrieve useful information
654 * for mbuf initialization. This pointer comes from the ``init_arg``
655 * parameter of rte_mempool_create().
657 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
660 * Reset the fields of a packet mbuf to their default values.
662 * The given mbuf must have only one segment.
665 * The packet mbuf to be resetted.
667 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
678 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
679 RTE_PKTMBUF_HEADROOM : m->buf_len;
682 __rte_mbuf_sanity_check(m, 1);
686 * Allocate a new mbuf from a mempool.
688 * This new mbuf contains one segment, which has a length of 0. The pointer
689 * to data is initialized to have some bytes of headroom in the buffer
690 * (if buffer size allows).
693 * The mempool from which the mbuf is allocated.
695 * - The pointer to the new mbuf on success.
696 * - NULL if allocation failed.
698 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
701 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
702 rte_pktmbuf_reset(m);
706 #ifdef RTE_MBUF_REFCNT
709 * Attach packet mbuf to another packet mbuf.
710 * After attachment we refer the mbuf we attached as 'indirect',
711 * while mbuf we attached to as 'direct'.
712 * Right now, not supported:
713 * - attachment to indirect mbuf (e.g. - md has to be direct).
714 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
715 * - mbuf we trying to attach (mi) is used by someone else
716 * e.g. it's reference counter is greater then 1.
719 * The indirect packet mbuf.
721 * The direct packet mbuf.
724 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *md)
726 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(md) &&
727 RTE_MBUF_DIRECT(mi) &&
728 rte_mbuf_refcnt_read(mi) == 1);
730 rte_mbuf_refcnt_update(md, 1);
731 mi->buf_physaddr = md->buf_physaddr;
732 mi->buf_addr = md->buf_addr;
733 mi->buf_len = md->buf_len;
736 mi->data_off = md->data_off;
737 mi->data_len = md->data_len;
739 mi->vlan_tci = md->vlan_tci;
740 mi->tx_offload = md->tx_offload;
744 mi->pkt_len = mi->data_len;
746 mi->ol_flags = md->ol_flags;
747 mi->packet_type = md->packet_type;
749 __rte_mbuf_sanity_check(mi, 1);
750 __rte_mbuf_sanity_check(md, 0);
754 * Detach an indirect packet mbuf -
755 * - restore original mbuf address and length values.
756 * - reset pktmbuf data and data_len to their default values.
757 * All other fields of the given packet mbuf will be left intact.
760 * The indirect attached packet mbuf.
763 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
765 const struct rte_mempool *mp = m->pool;
766 void *buf = RTE_MBUF_TO_BADDR(m);
767 uint32_t buf_len = mp->elt_size - sizeof(*m);
768 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + sizeof (*m);
771 m->buf_len = (uint16_t)buf_len;
773 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
774 RTE_PKTMBUF_HEADROOM : m->buf_len;
779 #endif /* RTE_MBUF_REFCNT */
782 static inline struct rte_mbuf* __attribute__((always_inline))
783 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
785 __rte_mbuf_sanity_check(m, 0);
787 #ifdef RTE_MBUF_REFCNT
788 if (likely (rte_mbuf_refcnt_read(m) == 1) ||
789 likely (rte_mbuf_refcnt_update(m, -1) == 0)) {
790 struct rte_mbuf *md = RTE_MBUF_FROM_BADDR(m->buf_addr);
792 rte_mbuf_refcnt_set(m, 0);
794 /* if this is an indirect mbuf, then
796 * - free attached mbuf segment
798 if (unlikely (md != m)) {
799 rte_pktmbuf_detach(m);
800 if (rte_mbuf_refcnt_update(md, -1) == 0)
801 __rte_mbuf_raw_free(md);
805 #ifdef RTE_MBUF_REFCNT
812 * Free a segment of a packet mbuf into its original mempool.
814 * Free an mbuf, without parsing other segments in case of chained
818 * The packet mbuf segment to be freed.
820 static inline void __attribute__((always_inline))
821 rte_pktmbuf_free_seg(struct rte_mbuf *m)
823 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
825 __rte_mbuf_raw_free(m);
830 * Free a packet mbuf back into its original mempool.
832 * Free an mbuf, and all its segments in case of chained buffers. Each
833 * segment is added back into its original mempool.
836 * The packet mbuf to be freed.
838 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
840 struct rte_mbuf *m_next;
842 __rte_mbuf_sanity_check(m, 1);
846 rte_pktmbuf_free_seg(m);
851 #ifdef RTE_MBUF_REFCNT
854 * Creates a "clone" of the given packet mbuf.
856 * Walks through all segments of the given packet mbuf, and for each of them:
857 * - Creates a new packet mbuf from the given pool.
858 * - Attaches newly created mbuf to the segment.
859 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
860 * from the original packet mbuf.
863 * The packet mbuf to be cloned.
865 * The mempool from which the "clone" mbufs are allocated.
867 * - The pointer to the new "clone" mbuf on success.
868 * - NULL if allocation fails.
870 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
871 struct rte_mempool *mp)
873 struct rte_mbuf *mc, *mi, **prev;
877 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
882 pktlen = md->pkt_len;
887 rte_pktmbuf_attach(mi, md);
890 } while ((md = md->next) != NULL &&
891 (mi = rte_pktmbuf_alloc(mp)) != NULL);
895 mc->pkt_len = pktlen;
897 /* Allocation of new indirect segment failed */
898 if (unlikely (mi == NULL)) {
899 rte_pktmbuf_free(mc);
903 __rte_mbuf_sanity_check(mc, 1);
908 * Adds given value to the refcnt of all packet mbuf segments.
910 * Walks through all segments of given packet mbuf and for each of them
911 * invokes rte_mbuf_refcnt_update().
914 * The packet mbuf whose refcnt to be updated.
916 * The value to add to the mbuf's segments refcnt.
918 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
920 __rte_mbuf_sanity_check(m, 1);
923 rte_mbuf_refcnt_update(m, v);
924 } while ((m = m->next) != NULL);
927 #endif /* RTE_MBUF_REFCNT */
930 * Get the headroom in a packet mbuf.
935 * The length of the headroom.
937 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
939 __rte_mbuf_sanity_check(m, 1);
944 * Get the tailroom of a packet mbuf.
949 * The length of the tailroom.
951 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
953 __rte_mbuf_sanity_check(m, 1);
954 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
959 * Get the last segment of the packet.
964 * The last segment of the given mbuf.
966 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
968 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
970 __rte_mbuf_sanity_check(m, 1);
971 while (m2->next != NULL)
977 * A macro that points to the start of the data in the mbuf.
979 * The returned pointer is cast to type t. Before using this
980 * function, the user must ensure that m_headlen(m) is large enough to
986 * The type to cast the result into.
988 #define rte_pktmbuf_mtod(m, t) ((t)((char *)(m)->buf_addr + (m)->data_off))
991 * A macro that returns the length of the packet.
993 * The value can be read or assigned.
998 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1001 * A macro that returns the length of the segment.
1003 * The value can be read or assigned.
1008 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1011 * Prepend len bytes to an mbuf data area.
1013 * Returns a pointer to the new
1014 * data start address. If there is not enough headroom in the first
1015 * segment, the function will return NULL, without modifying the mbuf.
1020 * The amount of data to prepend (in bytes).
1022 * A pointer to the start of the newly prepended data, or
1023 * NULL if there is not enough headroom space in the first segment
1025 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1028 __rte_mbuf_sanity_check(m, 1);
1030 if (unlikely(len > rte_pktmbuf_headroom(m)))
1034 m->data_len = (uint16_t)(m->data_len + len);
1035 m->pkt_len = (m->pkt_len + len);
1037 return (char *)m->buf_addr + m->data_off;
1041 * Append len bytes to an mbuf.
1043 * Append len bytes to an mbuf and return a pointer to the start address
1044 * of the added data. If there is not enough tailroom in the last
1045 * segment, the function will return NULL, without modifying the mbuf.
1050 * The amount of data to append (in bytes).
1052 * A pointer to the start of the newly appended data, or
1053 * NULL if there is not enough tailroom space in the last segment
1055 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1058 struct rte_mbuf *m_last;
1060 __rte_mbuf_sanity_check(m, 1);
1062 m_last = rte_pktmbuf_lastseg(m);
1063 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1066 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1067 m_last->data_len = (uint16_t)(m_last->data_len + len);
1068 m->pkt_len = (m->pkt_len + len);
1069 return (char*) tail;
1073 * Remove len bytes at the beginning of an mbuf.
1075 * Returns a pointer to the start address of the new data area. If the
1076 * length is greater than the length of the first segment, then the
1077 * function will fail and return NULL, without modifying the mbuf.
1082 * The amount of data to remove (in bytes).
1084 * A pointer to the new start of the data.
1086 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1088 __rte_mbuf_sanity_check(m, 1);
1090 if (unlikely(len > m->data_len))
1093 m->data_len = (uint16_t)(m->data_len - len);
1095 m->pkt_len = (m->pkt_len - len);
1096 return (char *)m->buf_addr + m->data_off;
1100 * Remove len bytes of data at the end of the mbuf.
1102 * If the length is greater than the length of the last segment, the
1103 * function will fail and return -1 without modifying the mbuf.
1108 * The amount of data to remove (in bytes).
1113 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1115 struct rte_mbuf *m_last;
1117 __rte_mbuf_sanity_check(m, 1);
1119 m_last = rte_pktmbuf_lastseg(m);
1120 if (unlikely(len > m_last->data_len))
1123 m_last->data_len = (uint16_t)(m_last->data_len - len);
1124 m->pkt_len = (m->pkt_len - len);
1129 * Test if mbuf data is contiguous.
1134 * - 1, if all data is contiguous (one segment).
1135 * - 0, if there is several segments.
1137 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1139 __rte_mbuf_sanity_check(m, 1);
1140 return !!(m->nb_segs == 1);
1144 * Dump an mbuf structure to the console.
1146 * Dump all fields for the given packet mbuf and all its associated
1147 * segments (in the case of a chained buffer).
1150 * A pointer to a file for output
1154 * If dump_len != 0, also dump the "dump_len" first data bytes of
1157 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1163 #endif /* _RTE_MBUF_H_ */