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_common.h>
58 #include <rte_mempool.h>
59 #include <rte_memory.h>
60 #include <rte_atomic.h>
61 #include <rte_prefetch.h>
62 #include <rte_branch_prediction.h>
68 /* deprecated options */
69 #pragma GCC poison RTE_MBUF_SCATTER_GATHER
70 #pragma GCC poison RTE_MBUF_REFCNT
73 * Packet Offload Features Flags. It also carry packet type information.
74 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
76 * - RX flags start at bit position zero, and get added to the left of previous
78 * - The most-significant 8 bits are reserved for generic mbuf flags
79 * - TX flags therefore start at bit position 55 (i.e. 63-8), and new flags get
80 * added to the right of the previously defined flags
82 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
83 * rte_get_tx_ol_flag_name().
85 #define PKT_RX_VLAN_PKT (1ULL << 0) /**< RX packet is a 802.1q VLAN packet. */
86 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
87 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
88 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
89 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
90 #define PKT_RX_EIP_CKSUM_BAD (0ULL << 0) /**< External IP header checksum error. */
91 #define PKT_RX_OVERSIZE (0ULL << 0) /**< Num of desc of an RX pkt oversize. */
92 #define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
93 #define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
94 #define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
95 #define PKT_RX_IPV4_HDR (1ULL << 5) /**< RX packet with IPv4 header. */
96 #define PKT_RX_IPV4_HDR_EXT (1ULL << 6) /**< RX packet with extended IPv4 header. */
97 #define PKT_RX_IPV6_HDR (1ULL << 7) /**< RX packet with IPv6 header. */
98 #define PKT_RX_IPV6_HDR_EXT (1ULL << 8) /**< RX packet with extended IPv6 header. */
99 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
100 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
101 #define PKT_RX_TUNNEL_IPV4_HDR (1ULL << 11) /**< RX tunnel packet with IPv4 header.*/
102 #define PKT_RX_TUNNEL_IPV6_HDR (1ULL << 12) /**< RX tunnel packet with IPv6 header. */
103 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
104 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
105 #define PKT_RX_QINQ_PKT (1ULL << 15) /**< RX packet with double VLAN stripped. */
106 /* add new RX flags here */
108 /* add new TX flags here */
111 * Second VLAN insertion (QinQ) flag.
113 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
116 * TCP segmentation offload. To enable this offload feature for a
117 * packet to be transmitted on hardware supporting TSO:
118 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
120 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
121 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
123 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
124 * - calculate the pseudo header checksum without taking ip_len in account,
125 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
126 * rte_ipv6_phdr_cksum() that can be used as helpers.
128 #define PKT_TX_TCP_SEG (1ULL << 50)
130 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
133 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
134 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
135 * L4 checksum offload, the user needs to:
136 * - fill l2_len and l3_len in mbuf
137 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
138 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
139 * - calculate the pseudo header checksum and set it in the L4 header (only
140 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
141 * For SCTP, set the crc field to 0.
143 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
144 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
145 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
146 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
147 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
150 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
151 * also be set by the application, although a PMD will only check
153 * - set the IP checksum field in the packet to 0
154 * - fill the mbuf offload information: l2_len, l3_len
156 #define PKT_TX_IP_CKSUM (1ULL << 54)
159 * Packet is IPv4. This flag must be set when using any offload feature
160 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
161 * packet. If the packet is a tunneled packet, this flag is related to
164 #define PKT_TX_IPV4 (1ULL << 55)
167 * Packet is IPv6. This flag must be set when using an offload feature
168 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
169 * packet. If the packet is a tunneled packet, this flag is related to
172 #define PKT_TX_IPV6 (1ULL << 56)
174 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
177 * Offload the IP checksum of an external header in the hardware. The
178 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
179 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
180 * packet must be set to 0.
181 * - set the outer IP checksum field in the packet to 0
182 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
184 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
187 * Packet outer header is IPv4. This flag must be set when using any
188 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
189 * outer header of the tunneled packet is an IPv4 packet.
191 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
194 * Packet outer header is IPv6. This flag must be set when using any
195 * outer offload feature (L4 checksum) to tell the NIC that the outer
196 * header of the tunneled packet is an IPv6 packet.
198 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
200 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
202 /* Use final bit of flags to indicate a control mbuf */
203 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
206 * Get the name of a RX offload flag
209 * The mask describing the flag.
211 * The name of this flag, or NULL if it's not a valid RX flag.
213 const char *rte_get_rx_ol_flag_name(uint64_t mask);
216 * Get the name of a TX offload flag
219 * The mask describing the flag. Usually only one bit must be set.
220 * Several bits can be given if they belong to the same mask.
221 * Ex: PKT_TX_L4_MASK.
223 * The name of this flag, or NULL if it's not a valid TX flag.
225 const char *rte_get_tx_ol_flag_name(uint64_t mask);
228 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
229 * splitting it into multiple segments.
230 * So, for mbufs that planned to be involved into RX/TX, the recommended
231 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
233 #define RTE_MBUF_DEFAULT_DATAROOM 2048
234 #define RTE_MBUF_DEFAULT_BUF_SIZE \
235 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
237 /* define a set of marker types that can be used to refer to set points in the
239 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
240 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
241 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
242 * with a single assignment */
245 * The generic rte_mbuf, containing a packet mbuf.
250 void *buf_addr; /**< Virtual address of segment buffer. */
251 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
253 uint16_t buf_len; /**< Length of segment buffer. */
255 /* next 6 bytes are initialised on RX descriptor rearm */
260 * 16-bit Reference counter.
261 * It should only be accessed using the following functions:
262 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
263 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
264 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
268 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
269 uint16_t refcnt; /**< Non-atomically accessed refcnt */
271 uint8_t nb_segs; /**< Number of segments. */
272 uint8_t port; /**< Input port. */
274 uint64_t ol_flags; /**< Offload features. */
276 /* remaining bytes are set on RX when pulling packet from descriptor */
277 MARKER rx_descriptor_fields1;
281 * The packet type, which is the combination of outer/inner L2, L3, L4
285 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
287 uint32_t l2_type:4; /**< (Outer) L2 type. */
288 uint32_t l3_type:4; /**< (Outer) L3 type. */
289 uint32_t l4_type:4; /**< (Outer) L4 type. */
290 uint32_t tun_type:4; /**< Tunnel type. */
291 uint32_t inner_l2_type:4; /**< Inner L2 type. */
292 uint32_t inner_l3_type:4; /**< Inner L3 type. */
293 uint32_t inner_l4_type:4; /**< Inner L4 type. */
297 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
298 uint16_t data_len; /**< Amount of data in segment buffer. */
299 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
300 #else /* RTE_NEXT_ABI */
302 * The packet type, which is used to indicate ordinary packet and also
303 * tunneled packet format, i.e. each number is represented a type of
306 uint16_t packet_type;
308 uint16_t data_len; /**< Amount of data in segment buffer. */
309 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
310 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
311 uint16_t vlan_tci_outer; /**< Outer VLAN Tag Control Identifier (CPU order) */
312 #endif /* RTE_NEXT_ABI */
314 uint32_t rss; /**< RSS hash result if RSS enabled */
322 /**< Second 4 flexible bytes */
325 /**< First 4 flexible bytes or FD ID, dependent on
326 PKT_RX_FDIR_* flag in ol_flags. */
327 } fdir; /**< Filter identifier if FDIR enabled */
328 uint32_t sched; /**< Hierarchical scheduler */
329 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
330 } hash; /**< hash information */
332 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
334 uint16_t vlan_tci_outer; /**< Outer VLAN Tag Control Identifier (CPU order) */
335 #endif /* RTE_NEXT_ABI */
337 /* second cache line - fields only used in slow path or on TX */
338 MARKER cacheline1 __rte_cache_aligned;
341 void *userdata; /**< Can be used for external metadata */
342 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
345 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
346 struct rte_mbuf *next; /**< Next segment of scattered packet. */
348 /* fields to support TX offloads */
350 uint64_t tx_offload; /**< combined for easy fetch */
352 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
353 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
354 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
355 uint64_t tso_segsz:16; /**< TCP TSO segment size */
357 /* fields for TX offloading of tunnels */
358 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
359 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
361 /* uint64_t unused:8; */
365 /** Size of the application private data. In case of an indirect
366 * mbuf, it stores the direct mbuf private data size. */
369 /** Timesync flags for use with IEEE1588. */
371 } __rte_cache_aligned;
373 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
376 * Return the mbuf owning the data buffer address of an indirect mbuf.
379 * The pointer to the indirect mbuf.
381 * The address of the direct mbuf corresponding to buffer_addr.
383 static inline struct rte_mbuf *
384 rte_mbuf_from_indirect(struct rte_mbuf *mi)
386 return RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
390 * Return the buffer address embedded in the given mbuf.
393 * The pointer to the mbuf.
395 * The address of the data buffer owned by the mbuf.
398 rte_mbuf_to_baddr(struct rte_mbuf *md)
401 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
406 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
408 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
411 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
413 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
416 * Private data in case of pktmbuf pool.
418 * A structure that contains some pktmbuf_pool-specific data that are
419 * appended after the mempool structure (in private data).
421 struct rte_pktmbuf_pool_private {
422 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
423 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
426 #ifdef RTE_LIBRTE_MBUF_DEBUG
428 /** check mbuf type in debug mode */
429 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
431 /** check mbuf type in debug mode if mbuf pointer is not null */
432 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
434 rte_mbuf_sanity_check(m, is_h); \
437 /** MBUF asserts in debug mode */
438 #define RTE_MBUF_ASSERT(exp) \
440 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
443 #else /* RTE_LIBRTE_MBUF_DEBUG */
445 /** check mbuf type in debug mode */
446 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
448 /** check mbuf type in debug mode if mbuf pointer is not null */
449 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
451 /** MBUF asserts in debug mode */
452 #define RTE_MBUF_ASSERT(exp) do { } while (0)
454 #endif /* RTE_LIBRTE_MBUF_DEBUG */
456 #ifdef RTE_MBUF_REFCNT_ATOMIC
459 * Reads the value of an mbuf's refcnt.
463 * Reference count number.
465 static inline uint16_t
466 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
468 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
472 * Sets an mbuf's refcnt to a defined value.
479 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
481 rte_atomic16_set(&m->refcnt_atomic, new_value);
485 * Adds given value to an mbuf's refcnt and returns its new value.
489 * Value to add/subtract
493 static inline uint16_t
494 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
497 * The atomic_add is an expensive operation, so we don't want to
498 * call it in the case where we know we are the uniq holder of
499 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
500 * operation has to be used because concurrent accesses on the
501 * reference counter can occur.
503 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
504 rte_mbuf_refcnt_set(m, 1 + value);
508 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
511 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
514 * Adds given value to an mbuf's refcnt and returns its new value.
516 static inline uint16_t
517 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
519 m->refcnt = (uint16_t)(m->refcnt + value);
524 * Reads the value of an mbuf's refcnt.
526 static inline uint16_t
527 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
533 * Sets an mbuf's refcnt to the defined value.
536 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
538 m->refcnt = new_value;
541 #endif /* RTE_MBUF_REFCNT_ATOMIC */
544 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
551 * Sanity checks on an mbuf.
553 * Check the consistency of the given mbuf. The function will cause a
554 * panic if corruption is detected.
557 * The mbuf to be checked.
559 * True if the mbuf is a packet header, false if it is a sub-segment
560 * of a packet (in this case, some fields like nb_segs are not checked)
563 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
566 * @internal Allocate a new mbuf from mempool *mp*.
567 * The use of that function is reserved for RTE internal needs.
568 * Please use rte_pktmbuf_alloc().
571 * The mempool from which mbuf is allocated.
573 * - The pointer to the new mbuf on success.
574 * - NULL if allocation failed.
576 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
580 if (rte_mempool_get(mp, &mb) < 0)
582 m = (struct rte_mbuf *)mb;
583 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
584 rte_mbuf_refcnt_set(m, 1);
589 * @internal Put mbuf back into its original mempool.
590 * The use of that function is reserved for RTE internal needs.
591 * Please use rte_pktmbuf_free().
594 * The mbuf to be freed.
596 static inline void __attribute__((always_inline))
597 __rte_mbuf_raw_free(struct rte_mbuf *m)
599 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
600 rte_mempool_put(m->pool, m);
603 /* Operations on ctrl mbuf */
606 * The control mbuf constructor.
608 * This function initializes some fields in an mbuf structure that are
609 * not modified by the user once created (mbuf type, origin pool, buffer
610 * start address, and so on). This function is given as a callback function
611 * to rte_mempool_create() at pool creation time.
614 * The mempool from which the mbuf is allocated.
616 * A pointer that can be used by the user to retrieve useful information
617 * for mbuf initialization. This pointer comes from the ``init_arg``
618 * parameter of rte_mempool_create().
620 * The mbuf to initialize.
622 * The index of the mbuf in the pool table.
624 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
625 void *m, unsigned i);
628 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
630 * This new mbuf is initialized with data pointing to the beginning of
631 * buffer, and with a length of zero.
634 * The mempool from which the mbuf is allocated.
636 * - The pointer to the new mbuf on success.
637 * - NULL if allocation failed.
639 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
642 * Free a control mbuf back into its original mempool.
645 * The control mbuf to be freed.
647 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
650 * A macro that returns the pointer to the carried data.
652 * The value that can be read or assigned.
657 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
660 * A macro that returns the length of the carried data.
662 * The value that can be read or assigned.
667 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
670 * Tests if an mbuf is a control mbuf
673 * The mbuf to be tested
675 * - True (1) if the mbuf is a control mbuf
676 * - False(0) otherwise
679 rte_is_ctrlmbuf(struct rte_mbuf *m)
681 return !!(m->ol_flags & CTRL_MBUF_FLAG);
684 /* Operations on pkt mbuf */
687 * The packet mbuf constructor.
689 * This function initializes some fields in the mbuf structure that are
690 * not modified by the user once created (origin pool, buffer start
691 * address, and so on). This function is given as a callback function to
692 * rte_mempool_create() at pool creation time.
695 * The mempool from which mbufs originate.
697 * A pointer that can be used by the user to retrieve useful information
698 * for mbuf initialization. This pointer comes from the ``init_arg``
699 * parameter of rte_mempool_create().
701 * The mbuf to initialize.
703 * The index of the mbuf in the pool table.
705 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
706 void *m, unsigned i);
710 * A packet mbuf pool constructor.
712 * This function initializes the mempool private data in the case of a
713 * pktmbuf pool. This private data is needed by the driver. The
714 * function is given as a callback function to rte_mempool_create() at
715 * pool creation. It can be extended by the user, for example, to
716 * provide another packet size.
719 * The mempool from which mbufs originate.
721 * A pointer that can be used by the user to retrieve useful information
722 * for mbuf initialization. This pointer comes from the ``init_arg``
723 * parameter of rte_mempool_create().
725 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
728 * Create a mbuf pool.
730 * This function creates and initializes a packet mbuf pool. It is
731 * a wrapper to rte_mempool_create() with the proper packet constructor
732 * and mempool constructor.
735 * The name of the mbuf pool.
737 * The number of elements in the mbuf pool. The optimum size (in terms
738 * of memory usage) for a mempool is when n is a power of two minus one:
741 * Size of the per-core object cache. See rte_mempool_create() for
744 * Size of application private are between the rte_mbuf structure
745 * and the data buffer.
746 * @param data_room_size
747 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
749 * The socket identifier where the memory should be allocated. The
750 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
753 * The pointer to the new allocated mempool, on success. NULL on error
754 * with rte_errno set appropriately. Possible rte_errno values include:
755 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
756 * - E_RTE_SECONDARY - function was called from a secondary process instance
757 * - EINVAL - cache size provided is too large
758 * - ENOSPC - the maximum number of memzones has already been allocated
759 * - EEXIST - a memzone with the same name already exists
760 * - ENOMEM - no appropriate memory area found in which to create memzone
763 rte_pktmbuf_pool_create(const char *name, unsigned n,
764 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
768 * Get the data room size of mbufs stored in a pktmbuf_pool
770 * The data room size is the amount of data that can be stored in a
771 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
774 * The packet mbuf pool.
776 * The data room size of mbufs stored in this mempool.
778 static inline uint16_t
779 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
781 struct rte_pktmbuf_pool_private *mbp_priv;
783 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
784 return mbp_priv->mbuf_data_room_size;
788 * Get the application private size of mbufs stored in a pktmbuf_pool
790 * The private size of mbuf is a zone located between the rte_mbuf
791 * structure and the data buffer where an application can store data
792 * associated to a packet.
795 * The packet mbuf pool.
797 * The private size of mbufs stored in this mempool.
799 static inline uint16_t
800 rte_pktmbuf_priv_size(struct rte_mempool *mp)
802 struct rte_pktmbuf_pool_private *mbp_priv;
804 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
805 return mbp_priv->mbuf_priv_size;
809 * Reset the fields of a packet mbuf to their default values.
811 * The given mbuf must have only one segment.
814 * The packet mbuf to be resetted.
816 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
822 m->vlan_tci_outer = 0;
828 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
829 RTE_PKTMBUF_HEADROOM : m->buf_len;
832 __rte_mbuf_sanity_check(m, 1);
836 * Allocate a new mbuf from a mempool.
838 * This new mbuf contains one segment, which has a length of 0. The pointer
839 * to data is initialized to have some bytes of headroom in the buffer
840 * (if buffer size allows).
843 * The mempool from which the mbuf is allocated.
845 * - The pointer to the new mbuf on success.
846 * - NULL if allocation failed.
848 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
851 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
852 rte_pktmbuf_reset(m);
857 * Attach packet mbuf to another packet mbuf.
859 * After attachment we refer the mbuf we attached as 'indirect',
860 * while mbuf we attached to as 'direct'.
861 * Right now, not supported:
862 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
863 * - mbuf we trying to attach (mi) is used by someone else
864 * e.g. it's reference counter is greater then 1.
867 * The indirect packet mbuf.
869 * The packet mbuf we're attaching to.
871 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
875 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(mi) &&
876 rte_mbuf_refcnt_read(mi) == 1);
878 /* if m is not direct, get the mbuf that embeds the data */
879 if (RTE_MBUF_DIRECT(m))
882 md = rte_mbuf_from_indirect(m);
884 rte_mbuf_refcnt_update(md, 1);
885 mi->priv_size = m->priv_size;
886 mi->buf_physaddr = m->buf_physaddr;
887 mi->buf_addr = m->buf_addr;
888 mi->buf_len = m->buf_len;
891 mi->data_off = m->data_off;
892 mi->data_len = m->data_len;
894 mi->vlan_tci = m->vlan_tci;
895 mi->vlan_tci_outer = m->vlan_tci_outer;
896 mi->tx_offload = m->tx_offload;
900 mi->pkt_len = mi->data_len;
902 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
903 mi->packet_type = m->packet_type;
905 __rte_mbuf_sanity_check(mi, 1);
906 __rte_mbuf_sanity_check(m, 0);
910 * Detach an indirect packet mbuf.
912 * - restore original mbuf address and length values.
913 * - reset pktmbuf data and data_len to their default values.
914 * All other fields of the given packet mbuf will be left intact.
917 * The indirect attached packet mbuf.
919 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
921 struct rte_mempool *mp = m->pool;
922 uint32_t mbuf_size, buf_len, priv_size;
924 priv_size = rte_pktmbuf_priv_size(mp);
925 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
926 buf_len = rte_pktmbuf_data_room_size(mp);
928 m->priv_size = priv_size;
929 m->buf_addr = (char *)m + mbuf_size;
930 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
931 m->buf_len = (uint16_t)buf_len;
932 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
937 static inline struct rte_mbuf* __attribute__((always_inline))
938 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
940 __rte_mbuf_sanity_check(m, 0);
942 if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
944 /* if this is an indirect mbuf, then
946 * - free attached mbuf segment
948 if (RTE_MBUF_INDIRECT(m)) {
949 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
950 rte_pktmbuf_detach(m);
951 if (rte_mbuf_refcnt_update(md, -1) == 0)
952 __rte_mbuf_raw_free(md);
960 * Free a segment of a packet mbuf into its original mempool.
962 * Free an mbuf, without parsing other segments in case of chained
966 * The packet mbuf segment to be freed.
968 static inline void __attribute__((always_inline))
969 rte_pktmbuf_free_seg(struct rte_mbuf *m)
971 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
973 __rte_mbuf_raw_free(m);
978 * Free a packet mbuf back into its original mempool.
980 * Free an mbuf, and all its segments in case of chained buffers. Each
981 * segment is added back into its original mempool.
984 * The packet mbuf to be freed.
986 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
988 struct rte_mbuf *m_next;
990 __rte_mbuf_sanity_check(m, 1);
994 rte_pktmbuf_free_seg(m);
1000 * Creates a "clone" of the given packet mbuf.
1002 * Walks through all segments of the given packet mbuf, and for each of them:
1003 * - Creates a new packet mbuf from the given pool.
1004 * - Attaches newly created mbuf to the segment.
1005 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1006 * from the original packet mbuf.
1009 * The packet mbuf to be cloned.
1011 * The mempool from which the "clone" mbufs are allocated.
1013 * - The pointer to the new "clone" mbuf on success.
1014 * - NULL if allocation fails.
1016 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1017 struct rte_mempool *mp)
1019 struct rte_mbuf *mc, *mi, **prev;
1023 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1028 pktlen = md->pkt_len;
1033 rte_pktmbuf_attach(mi, md);
1036 } while ((md = md->next) != NULL &&
1037 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1041 mc->pkt_len = pktlen;
1043 /* Allocation of new indirect segment failed */
1044 if (unlikely (mi == NULL)) {
1045 rte_pktmbuf_free(mc);
1049 __rte_mbuf_sanity_check(mc, 1);
1054 * Adds given value to the refcnt of all packet mbuf segments.
1056 * Walks through all segments of given packet mbuf and for each of them
1057 * invokes rte_mbuf_refcnt_update().
1060 * The packet mbuf whose refcnt to be updated.
1062 * The value to add to the mbuf's segments refcnt.
1064 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1066 __rte_mbuf_sanity_check(m, 1);
1069 rte_mbuf_refcnt_update(m, v);
1070 } while ((m = m->next) != NULL);
1074 * Get the headroom in a packet mbuf.
1079 * The length of the headroom.
1081 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1083 __rte_mbuf_sanity_check(m, 1);
1088 * Get the tailroom of a packet mbuf.
1093 * The length of the tailroom.
1095 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1097 __rte_mbuf_sanity_check(m, 1);
1098 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1103 * Get the last segment of the packet.
1108 * The last segment of the given mbuf.
1110 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1112 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1114 __rte_mbuf_sanity_check(m, 1);
1115 while (m2->next != NULL)
1121 * A macro that points to an offset into the data in the mbuf.
1123 * The returned pointer is cast to type t. Before using this
1124 * function, the user must ensure that the first segment is large
1125 * enough to accommodate its data.
1130 * The offset into the mbuf data.
1132 * The type to cast the result into.
1134 #define rte_pktmbuf_mtod_offset(m, t, o) \
1135 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1138 * A macro that points to the start of the data in the mbuf.
1140 * The returned pointer is cast to type t. Before using this
1141 * function, the user must ensure that the first segment is large
1142 * enough to accommodate its data.
1147 * The type to cast the result into.
1149 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1152 * A macro that returns the length of the packet.
1154 * The value can be read or assigned.
1159 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1162 * A macro that returns the length of the segment.
1164 * The value can be read or assigned.
1169 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1172 * Prepend len bytes to an mbuf data area.
1174 * Returns a pointer to the new
1175 * data start address. If there is not enough headroom in the first
1176 * segment, the function will return NULL, without modifying the mbuf.
1181 * The amount of data to prepend (in bytes).
1183 * A pointer to the start of the newly prepended data, or
1184 * NULL if there is not enough headroom space in the first segment
1186 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1189 __rte_mbuf_sanity_check(m, 1);
1191 if (unlikely(len > rte_pktmbuf_headroom(m)))
1195 m->data_len = (uint16_t)(m->data_len + len);
1196 m->pkt_len = (m->pkt_len + len);
1198 return (char *)m->buf_addr + m->data_off;
1202 * Append len bytes to an mbuf.
1204 * Append len bytes to an mbuf and return a pointer to the start address
1205 * of the added data. If there is not enough tailroom in the last
1206 * segment, the function will return NULL, without modifying the mbuf.
1211 * The amount of data to append (in bytes).
1213 * A pointer to the start of the newly appended data, or
1214 * NULL if there is not enough tailroom space in the last segment
1216 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1219 struct rte_mbuf *m_last;
1221 __rte_mbuf_sanity_check(m, 1);
1223 m_last = rte_pktmbuf_lastseg(m);
1224 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1227 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1228 m_last->data_len = (uint16_t)(m_last->data_len + len);
1229 m->pkt_len = (m->pkt_len + len);
1230 return (char*) tail;
1234 * Remove len bytes at the beginning of an mbuf.
1236 * Returns a pointer to the start address of the new data area. If the
1237 * length is greater than the length of the first segment, then the
1238 * function will fail and return NULL, without modifying the mbuf.
1243 * The amount of data to remove (in bytes).
1245 * A pointer to the new start of the data.
1247 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1249 __rte_mbuf_sanity_check(m, 1);
1251 if (unlikely(len > m->data_len))
1254 m->data_len = (uint16_t)(m->data_len - len);
1256 m->pkt_len = (m->pkt_len - len);
1257 return (char *)m->buf_addr + m->data_off;
1261 * Remove len bytes of data at the end of the mbuf.
1263 * If the length is greater than the length of the last segment, the
1264 * function will fail and return -1 without modifying the mbuf.
1269 * The amount of data to remove (in bytes).
1274 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1276 struct rte_mbuf *m_last;
1278 __rte_mbuf_sanity_check(m, 1);
1280 m_last = rte_pktmbuf_lastseg(m);
1281 if (unlikely(len > m_last->data_len))
1284 m_last->data_len = (uint16_t)(m_last->data_len - len);
1285 m->pkt_len = (m->pkt_len - len);
1290 * Test if mbuf data is contiguous.
1295 * - 1, if all data is contiguous (one segment).
1296 * - 0, if there is several segments.
1298 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1300 __rte_mbuf_sanity_check(m, 1);
1301 return !!(m->nb_segs == 1);
1305 * Dump an mbuf structure to the console.
1307 * Dump all fields for the given packet mbuf and all its associated
1308 * segments (in the case of a chained buffer).
1311 * A pointer to a file for output
1315 * If dump_len != 0, also dump the "dump_len" first data bytes of
1318 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1324 #endif /* _RTE_MBUF_H_ */