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 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
194 /* Use final bit of flags to indicate a control mbuf */
195 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
198 * Get the name of a RX offload flag
201 * The mask describing the flag.
203 * The name of this flag, or NULL if it's not a valid RX flag.
205 const char *rte_get_rx_ol_flag_name(uint64_t mask);
208 * Get the name of a TX offload flag
211 * The mask describing the flag. Usually only one bit must be set.
212 * Several bits can be given if they belong to the same mask.
213 * Ex: PKT_TX_L4_MASK.
215 * The name of this flag, or NULL if it's not a valid TX flag.
217 const char *rte_get_tx_ol_flag_name(uint64_t mask);
219 /* define a set of marker types that can be used to refer to set points in the
221 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
222 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
223 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
224 * with a single assignment */
227 * The generic rte_mbuf, containing a packet mbuf.
232 void *buf_addr; /**< Virtual address of segment buffer. */
233 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
235 uint16_t buf_len; /**< Length of segment buffer. */
237 /* next 6 bytes are initialised on RX descriptor rearm */
242 * 16-bit Reference counter.
243 * It should only be accessed using the following functions:
244 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
245 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
246 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
250 #ifdef RTE_MBUF_REFCNT
251 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
252 uint16_t refcnt; /**< Non-atomically accessed refcnt */
254 uint16_t refcnt_reserved; /**< Do not use this field */
256 uint8_t nb_segs; /**< Number of segments. */
257 uint8_t port; /**< Input port. */
259 uint64_t ol_flags; /**< Offload features. */
261 /* remaining bytes are set on RX when pulling packet from descriptor */
262 MARKER rx_descriptor_fields1;
265 * The packet type, which is used to indicate ordinary packet and also
266 * tunneled packet format, i.e. each number is represented a type of
269 uint16_t packet_type;
271 uint16_t data_len; /**< Amount of data in segment buffer. */
272 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
273 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
276 uint32_t rss; /**< RSS hash result if RSS enabled */
284 /**< Second 4 flexible bytes */
287 /**< First 4 flexible bytes or FD ID, dependent on
288 PKT_RX_FDIR_* flag in ol_flags. */
289 } fdir; /**< Filter identifier if FDIR enabled */
290 uint32_t sched; /**< Hierarchical scheduler */
291 uint32_t usr; /**< User defined tags. See @rte_distributor_process */
292 } hash; /**< hash information */
294 /* sequence number - field used in distributor and reorder library */
297 /* second cache line - fields only used in slow path or on TX */
298 MARKER cacheline1 __rte_cache_aligned;
301 void *userdata; /**< Can be used for external metadata */
302 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
305 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
306 struct rte_mbuf *next; /**< Next segment of scattered packet. */
308 /* fields to support TX offloads */
310 uint64_t tx_offload; /**< combined for easy fetch */
312 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
313 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
314 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
315 uint64_t tso_segsz:16; /**< TCP TSO segment size */
317 /* fields for TX offloading of tunnels */
318 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
319 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
321 /* uint64_t unused:8; */
324 } __rte_cache_aligned;
327 * Given the buf_addr returns the pointer to corresponding mbuf.
329 #define RTE_MBUF_FROM_BADDR(ba) (((struct rte_mbuf *)(ba)) - 1)
332 * Given the pointer to mbuf returns an address where it's buf_addr
335 #define RTE_MBUF_TO_BADDR(mb) (((struct rte_mbuf *)(mb)) + 1)
338 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
340 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
343 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
345 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
348 * Private data in case of pktmbuf pool.
350 * A structure that contains some pktmbuf_pool-specific data that are
351 * appended after the mempool structure (in private data).
353 struct rte_pktmbuf_pool_private {
354 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf.*/
357 #ifdef RTE_LIBRTE_MBUF_DEBUG
359 /** check mbuf type in debug mode */
360 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
362 /** check mbuf type in debug mode if mbuf pointer is not null */
363 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
365 rte_mbuf_sanity_check(m, is_h); \
368 /** MBUF asserts in debug mode */
369 #define RTE_MBUF_ASSERT(exp) \
371 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
374 #else /* RTE_LIBRTE_MBUF_DEBUG */
376 /** check mbuf type in debug mode */
377 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
379 /** check mbuf type in debug mode if mbuf pointer is not null */
380 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
382 /** MBUF asserts in debug mode */
383 #define RTE_MBUF_ASSERT(exp) do { } while (0)
385 #endif /* RTE_LIBRTE_MBUF_DEBUG */
387 #ifdef RTE_MBUF_REFCNT
388 #ifdef RTE_MBUF_REFCNT_ATOMIC
391 * Adds given value to an mbuf's refcnt and returns its new value.
395 * Value to add/subtract
399 static inline uint16_t
400 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
402 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
406 * Reads the value of an mbuf's refcnt.
410 * Reference count number.
412 static inline uint16_t
413 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
415 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
419 * Sets an mbuf's refcnt to a defined value.
426 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
428 rte_atomic16_set(&m->refcnt_atomic, new_value);
431 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
434 * Adds given value to an mbuf's refcnt and returns its new value.
436 static inline uint16_t
437 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
439 m->refcnt = (uint16_t)(m->refcnt + value);
444 * Reads the value of an mbuf's refcnt.
446 static inline uint16_t
447 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
453 * Sets an mbuf's refcnt to the defined value.
456 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
458 m->refcnt = new_value;
461 #endif /* RTE_MBUF_REFCNT_ATOMIC */
464 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
469 #else /* ! RTE_MBUF_REFCNT */
472 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { } while(0)
474 #define rte_mbuf_refcnt_set(m,v) do { } while(0)
476 #endif /* RTE_MBUF_REFCNT */
480 * Sanity checks on an mbuf.
482 * Check the consistency of the given mbuf. The function will cause a
483 * panic if corruption is detected.
486 * The mbuf to be checked.
488 * True if the mbuf is a packet header, false if it is a sub-segment
489 * of a packet (in this case, some fields like nb_segs are not checked)
492 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
495 * @internal Allocate a new mbuf from mempool *mp*.
496 * The use of that function is reserved for RTE internal needs.
497 * Please use rte_pktmbuf_alloc().
500 * The mempool from which mbuf is allocated.
502 * - The pointer to the new mbuf on success.
503 * - NULL if allocation failed.
505 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
509 if (rte_mempool_get(mp, &mb) < 0)
511 m = (struct rte_mbuf *)mb;
512 #ifdef RTE_MBUF_REFCNT
513 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
514 rte_mbuf_refcnt_set(m, 1);
515 #endif /* RTE_MBUF_REFCNT */
520 * @internal Put mbuf back into its original mempool.
521 * The use of that function is reserved for RTE internal needs.
522 * Please use rte_pktmbuf_free().
525 * The mbuf to be freed.
527 static inline void __attribute__((always_inline))
528 __rte_mbuf_raw_free(struct rte_mbuf *m)
530 #ifdef RTE_MBUF_REFCNT
531 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
532 #endif /* RTE_MBUF_REFCNT */
533 rte_mempool_put(m->pool, m);
536 /* Operations on ctrl mbuf */
539 * The control mbuf constructor.
541 * This function initializes some fields in an mbuf structure that are
542 * not modified by the user once created (mbuf type, origin pool, buffer
543 * start address, and so on). This function is given as a callback function
544 * to rte_mempool_create() at pool creation time.
547 * The mempool from which the mbuf is allocated.
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_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
558 void *m, unsigned i);
561 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
563 * This new mbuf is initialized with data pointing to the beginning of
564 * buffer, and with a length of zero.
567 * The mempool from which the mbuf is allocated.
569 * - The pointer to the new mbuf on success.
570 * - NULL if allocation failed.
572 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
575 * Free a control mbuf back into its original mempool.
578 * The control mbuf to be freed.
580 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
583 * A macro that returns the pointer to the carried data.
585 * The value that can be read or assigned.
590 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
593 * A macro that returns the length of the carried data.
595 * The value that can be read or assigned.
600 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
603 * Tests if an mbuf is a control mbuf
606 * The mbuf to be tested
608 * - True (1) if the mbuf is a control mbuf
609 * - False(0) otherwise
612 rte_is_ctrlmbuf(struct rte_mbuf *m)
614 return (!!(m->ol_flags & CTRL_MBUF_FLAG));
617 /* Operations on pkt mbuf */
620 * The packet mbuf constructor.
622 * This function initializes some fields in the mbuf structure that are
623 * not modified by the user once created (origin pool, buffer start
624 * address, and so on). This function is given as a callback function to
625 * rte_mempool_create() at pool creation time.
628 * The mempool from which mbufs originate.
630 * A pointer that can be used by the user to retrieve useful information
631 * for mbuf initialization. This pointer comes from the ``init_arg``
632 * parameter of rte_mempool_create().
634 * The mbuf to initialize.
636 * The index of the mbuf in the pool table.
638 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
639 void *m, unsigned i);
643 * A packet mbuf pool constructor.
645 * This function initializes the mempool private data in the case of a
646 * pktmbuf pool. This private data is needed by the driver. The
647 * function is given as a callback function to rte_mempool_create() at
648 * pool creation. It can be extended by the user, for example, to
649 * provide another packet size.
652 * The mempool from which mbufs originate.
654 * A pointer that can be used by the user to retrieve useful information
655 * for mbuf initialization. This pointer comes from the ``init_arg``
656 * parameter of rte_mempool_create().
658 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
661 * Reset the fields of a packet mbuf to their default values.
663 * The given mbuf must have only one segment.
666 * The packet mbuf to be resetted.
668 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
679 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
680 RTE_PKTMBUF_HEADROOM : m->buf_len;
683 __rte_mbuf_sanity_check(m, 1);
687 * Allocate a new mbuf from a mempool.
689 * This new mbuf contains one segment, which has a length of 0. The pointer
690 * to data is initialized to have some bytes of headroom in the buffer
691 * (if buffer size allows).
694 * The mempool from which the mbuf is allocated.
696 * - The pointer to the new mbuf on success.
697 * - NULL if allocation failed.
699 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
702 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
703 rte_pktmbuf_reset(m);
707 #ifdef RTE_MBUF_REFCNT
710 * Attach packet mbuf to another packet mbuf.
711 * After attachment we refer the mbuf we attached as 'indirect',
712 * while mbuf we attached to as 'direct'.
713 * Right now, not supported:
714 * - attachment to indirect mbuf (e.g. - md has to be direct).
715 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
716 * - mbuf we trying to attach (mi) is used by someone else
717 * e.g. it's reference counter is greater then 1.
720 * The indirect packet mbuf.
722 * The direct packet mbuf.
725 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *md)
727 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(md) &&
728 RTE_MBUF_DIRECT(mi) &&
729 rte_mbuf_refcnt_read(mi) == 1);
731 rte_mbuf_refcnt_update(md, 1);
732 mi->buf_physaddr = md->buf_physaddr;
733 mi->buf_addr = md->buf_addr;
734 mi->buf_len = md->buf_len;
737 mi->data_off = md->data_off;
738 mi->data_len = md->data_len;
740 mi->vlan_tci = md->vlan_tci;
741 mi->tx_offload = md->tx_offload;
745 mi->pkt_len = mi->data_len;
747 mi->ol_flags = md->ol_flags | IND_ATTACHED_MBUF;
748 mi->packet_type = md->packet_type;
750 __rte_mbuf_sanity_check(mi, 1);
751 __rte_mbuf_sanity_check(md, 0);
755 * Detach an indirect packet mbuf -
756 * - restore original mbuf address and length values.
757 * - reset pktmbuf data and data_len to their default values.
758 * All other fields of the given packet mbuf will be left intact.
761 * The indirect attached packet mbuf.
764 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
766 const struct rte_mempool *mp = m->pool;
767 void *buf = RTE_MBUF_TO_BADDR(m);
768 uint32_t buf_len = mp->elt_size - sizeof(*m);
769 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + sizeof (*m);
772 m->buf_len = (uint16_t)buf_len;
774 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
775 RTE_PKTMBUF_HEADROOM : m->buf_len;
782 #endif /* RTE_MBUF_REFCNT */
785 static inline struct rte_mbuf* __attribute__((always_inline))
786 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
788 __rte_mbuf_sanity_check(m, 0);
790 #ifdef RTE_MBUF_REFCNT
791 if (likely (rte_mbuf_refcnt_read(m) == 1) ||
792 likely (rte_mbuf_refcnt_update(m, -1) == 0)) {
794 rte_mbuf_refcnt_set(m, 0);
796 /* if this is an indirect mbuf, then
798 * - free attached mbuf segment
800 if (RTE_MBUF_INDIRECT(m)) {
801 struct rte_mbuf *md = RTE_MBUF_FROM_BADDR(m->buf_addr);
802 rte_pktmbuf_detach(m);
803 if (rte_mbuf_refcnt_update(md, -1) == 0)
804 __rte_mbuf_raw_free(md);
808 #ifdef RTE_MBUF_REFCNT
815 * Free a segment of a packet mbuf into its original mempool.
817 * Free an mbuf, without parsing other segments in case of chained
821 * The packet mbuf segment to be freed.
823 static inline void __attribute__((always_inline))
824 rte_pktmbuf_free_seg(struct rte_mbuf *m)
826 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
828 __rte_mbuf_raw_free(m);
833 * Free a packet mbuf back into its original mempool.
835 * Free an mbuf, and all its segments in case of chained buffers. Each
836 * segment is added back into its original mempool.
839 * The packet mbuf to be freed.
841 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
843 struct rte_mbuf *m_next;
845 __rte_mbuf_sanity_check(m, 1);
849 rte_pktmbuf_free_seg(m);
854 #ifdef RTE_MBUF_REFCNT
857 * Creates a "clone" of the given packet mbuf.
859 * Walks through all segments of the given packet mbuf, and for each of them:
860 * - Creates a new packet mbuf from the given pool.
861 * - Attaches newly created mbuf to the segment.
862 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
863 * from the original packet mbuf.
866 * The packet mbuf to be cloned.
868 * The mempool from which the "clone" mbufs are allocated.
870 * - The pointer to the new "clone" mbuf on success.
871 * - NULL if allocation fails.
873 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
874 struct rte_mempool *mp)
876 struct rte_mbuf *mc, *mi, **prev;
880 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
885 pktlen = md->pkt_len;
890 rte_pktmbuf_attach(mi, md);
893 } while ((md = md->next) != NULL &&
894 (mi = rte_pktmbuf_alloc(mp)) != NULL);
898 mc->pkt_len = pktlen;
900 /* Allocation of new indirect segment failed */
901 if (unlikely (mi == NULL)) {
902 rte_pktmbuf_free(mc);
906 __rte_mbuf_sanity_check(mc, 1);
911 * Adds given value to the refcnt of all packet mbuf segments.
913 * Walks through all segments of given packet mbuf and for each of them
914 * invokes rte_mbuf_refcnt_update().
917 * The packet mbuf whose refcnt to be updated.
919 * The value to add to the mbuf's segments refcnt.
921 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
923 __rte_mbuf_sanity_check(m, 1);
926 rte_mbuf_refcnt_update(m, v);
927 } while ((m = m->next) != NULL);
930 #endif /* RTE_MBUF_REFCNT */
933 * Get the headroom in a packet mbuf.
938 * The length of the headroom.
940 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
942 __rte_mbuf_sanity_check(m, 1);
947 * Get the tailroom of a packet mbuf.
952 * The length of the tailroom.
954 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
956 __rte_mbuf_sanity_check(m, 1);
957 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
962 * Get the last segment of the packet.
967 * The last segment of the given mbuf.
969 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
971 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
973 __rte_mbuf_sanity_check(m, 1);
974 while (m2->next != NULL)
980 * A macro that points to the start of the data in the mbuf.
982 * The returned pointer is cast to type t. Before using this
983 * function, the user must ensure that m_headlen(m) is large enough to
989 * The type to cast the result into.
991 #define rte_pktmbuf_mtod(m, t) ((t)((char *)(m)->buf_addr + (m)->data_off))
994 * A macro that returns the length of the packet.
996 * The value can be read or assigned.
1001 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1004 * A macro that returns the length of the segment.
1006 * The value can be read or assigned.
1011 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1014 * Prepend len bytes to an mbuf data area.
1016 * Returns a pointer to the new
1017 * data start address. If there is not enough headroom in the first
1018 * segment, the function will return NULL, without modifying the mbuf.
1023 * The amount of data to prepend (in bytes).
1025 * A pointer to the start of the newly prepended data, or
1026 * NULL if there is not enough headroom space in the first segment
1028 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1031 __rte_mbuf_sanity_check(m, 1);
1033 if (unlikely(len > rte_pktmbuf_headroom(m)))
1037 m->data_len = (uint16_t)(m->data_len + len);
1038 m->pkt_len = (m->pkt_len + len);
1040 return (char *)m->buf_addr + m->data_off;
1044 * Append len bytes to an mbuf.
1046 * Append len bytes to an mbuf and return a pointer to the start address
1047 * of the added data. If there is not enough tailroom in the last
1048 * segment, the function will return NULL, without modifying the mbuf.
1053 * The amount of data to append (in bytes).
1055 * A pointer to the start of the newly appended data, or
1056 * NULL if there is not enough tailroom space in the last segment
1058 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1061 struct rte_mbuf *m_last;
1063 __rte_mbuf_sanity_check(m, 1);
1065 m_last = rte_pktmbuf_lastseg(m);
1066 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1069 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1070 m_last->data_len = (uint16_t)(m_last->data_len + len);
1071 m->pkt_len = (m->pkt_len + len);
1072 return (char*) tail;
1076 * Remove len bytes at the beginning of an mbuf.
1078 * Returns a pointer to the start address of the new data area. If the
1079 * length is greater than the length of the first segment, then the
1080 * function will fail and return NULL, without modifying the mbuf.
1085 * The amount of data to remove (in bytes).
1087 * A pointer to the new start of the data.
1089 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1091 __rte_mbuf_sanity_check(m, 1);
1093 if (unlikely(len > m->data_len))
1096 m->data_len = (uint16_t)(m->data_len - len);
1098 m->pkt_len = (m->pkt_len - len);
1099 return (char *)m->buf_addr + m->data_off;
1103 * Remove len bytes of data at the end of the mbuf.
1105 * If the length is greater than the length of the last segment, the
1106 * function will fail and return -1 without modifying the mbuf.
1111 * The amount of data to remove (in bytes).
1116 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1118 struct rte_mbuf *m_last;
1120 __rte_mbuf_sanity_check(m, 1);
1122 m_last = rte_pktmbuf_lastseg(m);
1123 if (unlikely(len > m_last->data_len))
1126 m_last->data_len = (uint16_t)(m_last->data_len - len);
1127 m->pkt_len = (m->pkt_len - len);
1132 * Test if mbuf data is contiguous.
1137 * - 1, if all data is contiguous (one segment).
1138 * - 0, if there is several segments.
1140 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1142 __rte_mbuf_sanity_check(m, 1);
1143 return !!(m->nb_segs == 1);
1147 * Dump an mbuf structure to the console.
1149 * Dump all fields for the given packet mbuf and all its associated
1150 * segments (in the case of a chained buffer).
1153 * A pointer to a file for output
1157 * If dump_len != 0, also dump the "dump_len" first data bytes of
1160 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1166 #endif /* _RTE_MBUF_H_ */