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 /* add new RX flags here */
107 /* add new TX flags here */
110 * TCP segmentation offload. To enable this offload feature for a
111 * packet to be transmitted on hardware supporting TSO:
112 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
114 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
115 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
117 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
118 * - calculate the pseudo header checksum without taking ip_len in account,
119 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
120 * rte_ipv6_phdr_cksum() that can be used as helpers.
122 #define PKT_TX_TCP_SEG (1ULL << 50)
124 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
127 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
128 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
129 * L4 checksum offload, the user needs to:
130 * - fill l2_len and l3_len in mbuf
131 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
132 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
133 * - calculate the pseudo header checksum and set it in the L4 header (only
134 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
135 * For SCTP, set the crc field to 0.
137 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
138 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
139 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
140 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
141 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
144 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
145 * also be set by the application, although a PMD will only check
147 * - set the IP checksum field in the packet to 0
148 * - fill the mbuf offload information: l2_len, l3_len
150 #define PKT_TX_IP_CKSUM (1ULL << 54)
153 * Packet is IPv4. This flag must be set when using any offload feature
154 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
155 * packet. If the packet is a tunneled packet, this flag is related to
158 #define PKT_TX_IPV4 (1ULL << 55)
161 * Packet is IPv6. This flag must be set when using an offload feature
162 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
163 * packet. If the packet is a tunneled packet, this flag is related to
166 #define PKT_TX_IPV6 (1ULL << 56)
168 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
171 * Offload the IP checksum of an external header in the hardware. The
172 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
173 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
174 * packet must be set to 0.
175 * - set the outer IP checksum field in the packet to 0
176 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
178 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
181 * Packet outer header is IPv4. This flag must be set when using any
182 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
183 * outer header of the tunneled packet is an IPv4 packet.
185 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
188 * Packet outer header is IPv6. This flag must be set when using any
189 * outer offload feature (L4 checksum) to tell the NIC that the outer
190 * header of the tunneled packet is an IPv6 packet.
192 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
194 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
196 /* Use final bit of flags to indicate a control mbuf */
197 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
200 * Get the name of a RX offload flag
203 * The mask describing the flag.
205 * The name of this flag, or NULL if it's not a valid RX flag.
207 const char *rte_get_rx_ol_flag_name(uint64_t mask);
210 * Get the name of a TX offload flag
213 * The mask describing the flag. Usually only one bit must be set.
214 * Several bits can be given if they belong to the same mask.
215 * Ex: PKT_TX_L4_MASK.
217 * The name of this flag, or NULL if it's not a valid TX flag.
219 const char *rte_get_tx_ol_flag_name(uint64_t mask);
222 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
223 * splitting it into multiple segments.
224 * So, for mbufs that planned to be involved into RX/TX, the recommended
225 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
227 #define RTE_MBUF_DEFAULT_DATAROOM 2048
228 #define RTE_MBUF_DEFAULT_BUF_SIZE \
229 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
231 /* define a set of marker types that can be used to refer to set points in the
233 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
234 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
235 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
236 * with a single assignment */
239 * The generic rte_mbuf, containing a packet mbuf.
244 void *buf_addr; /**< Virtual address of segment buffer. */
245 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
247 uint16_t buf_len; /**< Length of segment buffer. */
249 /* next 6 bytes are initialised on RX descriptor rearm */
254 * 16-bit Reference counter.
255 * It should only be accessed using the following functions:
256 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
257 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
258 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
262 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
263 uint16_t refcnt; /**< Non-atomically accessed refcnt */
265 uint8_t nb_segs; /**< Number of segments. */
266 uint8_t port; /**< Input port. */
268 uint64_t ol_flags; /**< Offload features. */
270 /* remaining bytes are set on RX when pulling packet from descriptor */
271 MARKER rx_descriptor_fields1;
274 * The packet type, which is used to indicate ordinary packet and also
275 * tunneled packet format, i.e. each number is represented a type of
278 uint16_t packet_type;
280 uint16_t data_len; /**< Amount of data in segment buffer. */
281 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
282 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
285 uint32_t rss; /**< RSS hash result if RSS enabled */
293 /**< Second 4 flexible bytes */
296 /**< First 4 flexible bytes or FD ID, dependent on
297 PKT_RX_FDIR_* flag in ol_flags. */
298 } fdir; /**< Filter identifier if FDIR enabled */
299 uint32_t sched; /**< Hierarchical scheduler */
300 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
301 } hash; /**< hash information */
303 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
305 /* second cache line - fields only used in slow path or on TX */
306 MARKER cacheline1 __rte_cache_aligned;
309 void *userdata; /**< Can be used for external metadata */
310 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
313 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
314 struct rte_mbuf *next; /**< Next segment of scattered packet. */
316 /* fields to support TX offloads */
318 uint64_t tx_offload; /**< combined for easy fetch */
320 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
321 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
322 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
323 uint64_t tso_segsz:16; /**< TCP TSO segment size */
325 /* fields for TX offloading of tunnels */
326 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
327 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
329 /* uint64_t unused:8; */
333 /** Size of the application private data. In case of an indirect
334 * mbuf, it stores the direct mbuf private data size. */
336 } __rte_cache_aligned;
338 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
341 * Return the mbuf owning the data buffer address of an indirect mbuf.
344 * The pointer to the indirect mbuf.
346 * The address of the direct mbuf corresponding to buffer_addr.
348 static inline struct rte_mbuf *
349 rte_mbuf_from_indirect(struct rte_mbuf *mi)
351 return RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
355 * Return the buffer address embedded in the given mbuf.
358 * The pointer to the mbuf.
360 * The address of the data buffer owned by the mbuf.
363 rte_mbuf_to_baddr(struct rte_mbuf *md)
366 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
371 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
373 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
376 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
378 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
381 * Private data in case of pktmbuf pool.
383 * A structure that contains some pktmbuf_pool-specific data that are
384 * appended after the mempool structure (in private data).
386 struct rte_pktmbuf_pool_private {
387 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
388 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
391 #ifdef RTE_LIBRTE_MBUF_DEBUG
393 /** check mbuf type in debug mode */
394 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
396 /** check mbuf type in debug mode if mbuf pointer is not null */
397 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
399 rte_mbuf_sanity_check(m, is_h); \
402 /** MBUF asserts in debug mode */
403 #define RTE_MBUF_ASSERT(exp) \
405 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
408 #else /* RTE_LIBRTE_MBUF_DEBUG */
410 /** check mbuf type in debug mode */
411 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
413 /** check mbuf type in debug mode if mbuf pointer is not null */
414 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
416 /** MBUF asserts in debug mode */
417 #define RTE_MBUF_ASSERT(exp) do { } while (0)
419 #endif /* RTE_LIBRTE_MBUF_DEBUG */
421 #ifdef RTE_MBUF_REFCNT_ATOMIC
424 * Reads the value of an mbuf's refcnt.
428 * Reference count number.
430 static inline uint16_t
431 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
433 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
437 * Sets an mbuf's refcnt to a defined value.
444 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
446 rte_atomic16_set(&m->refcnt_atomic, new_value);
450 * Adds given value to an mbuf's refcnt and returns its new value.
454 * Value to add/subtract
458 static inline uint16_t
459 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
462 * The atomic_add is an expensive operation, so we don't want to
463 * call it in the case where we know we are the uniq holder of
464 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
465 * operation has to be used because concurrent accesses on the
466 * reference counter can occur.
468 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
469 rte_mbuf_refcnt_set(m, 1 + value);
473 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
476 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
479 * Adds given value to an mbuf's refcnt and returns its new value.
481 static inline uint16_t
482 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
484 m->refcnt = (uint16_t)(m->refcnt + value);
489 * Reads the value of an mbuf's refcnt.
491 static inline uint16_t
492 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
498 * Sets an mbuf's refcnt to the defined value.
501 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
503 m->refcnt = new_value;
506 #endif /* RTE_MBUF_REFCNT_ATOMIC */
509 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
516 * Sanity checks on an mbuf.
518 * Check the consistency of the given mbuf. The function will cause a
519 * panic if corruption is detected.
522 * The mbuf to be checked.
524 * True if the mbuf is a packet header, false if it is a sub-segment
525 * of a packet (in this case, some fields like nb_segs are not checked)
528 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
531 * @internal Allocate a new mbuf from mempool *mp*.
532 * The use of that function is reserved for RTE internal needs.
533 * Please use rte_pktmbuf_alloc().
536 * The mempool from which mbuf is allocated.
538 * - The pointer to the new mbuf on success.
539 * - NULL if allocation failed.
541 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
545 if (rte_mempool_get(mp, &mb) < 0)
547 m = (struct rte_mbuf *)mb;
548 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
549 rte_mbuf_refcnt_set(m, 1);
554 * @internal Put mbuf back into its original mempool.
555 * The use of that function is reserved for RTE internal needs.
556 * Please use rte_pktmbuf_free().
559 * The mbuf to be freed.
561 static inline void __attribute__((always_inline))
562 __rte_mbuf_raw_free(struct rte_mbuf *m)
564 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
565 rte_mempool_put(m->pool, m);
568 /* Operations on ctrl mbuf */
571 * The control mbuf constructor.
573 * This function initializes some fields in an mbuf structure that are
574 * not modified by the user once created (mbuf type, origin pool, buffer
575 * start address, and so on). This function is given as a callback function
576 * to rte_mempool_create() at pool creation time.
579 * The mempool from which the mbuf is allocated.
581 * A pointer that can be used by the user to retrieve useful information
582 * for mbuf initialization. This pointer comes from the ``init_arg``
583 * parameter of rte_mempool_create().
585 * The mbuf to initialize.
587 * The index of the mbuf in the pool table.
589 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
590 void *m, unsigned i);
593 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
595 * This new mbuf is initialized with data pointing to the beginning of
596 * buffer, and with a length of zero.
599 * The mempool from which the mbuf is allocated.
601 * - The pointer to the new mbuf on success.
602 * - NULL if allocation failed.
604 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
607 * Free a control mbuf back into its original mempool.
610 * The control mbuf to be freed.
612 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
615 * A macro that returns the pointer to the carried data.
617 * The value that can be read or assigned.
622 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
625 * A macro that returns the length of the carried data.
627 * The value that can be read or assigned.
632 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
635 * Tests if an mbuf is a control mbuf
638 * The mbuf to be tested
640 * - True (1) if the mbuf is a control mbuf
641 * - False(0) otherwise
644 rte_is_ctrlmbuf(struct rte_mbuf *m)
646 return !!(m->ol_flags & CTRL_MBUF_FLAG);
649 /* Operations on pkt mbuf */
652 * The packet mbuf constructor.
654 * This function initializes some fields in the mbuf structure that are
655 * not modified by the user once created (origin pool, buffer start
656 * address, and so on). This function is given as a callback function to
657 * rte_mempool_create() at pool creation time.
660 * The mempool from which mbufs originate.
662 * A pointer that can be used by the user to retrieve useful information
663 * for mbuf initialization. This pointer comes from the ``init_arg``
664 * parameter of rte_mempool_create().
666 * The mbuf to initialize.
668 * The index of the mbuf in the pool table.
670 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
671 void *m, unsigned i);
675 * A packet mbuf pool constructor.
677 * This function initializes the mempool private data in the case of a
678 * pktmbuf pool. This private data is needed by the driver. The
679 * function is given as a callback function to rte_mempool_create() at
680 * pool creation. It can be extended by the user, for example, to
681 * provide another packet size.
684 * The mempool from which mbufs originate.
686 * A pointer that can be used by the user to retrieve useful information
687 * for mbuf initialization. This pointer comes from the ``init_arg``
688 * parameter of rte_mempool_create().
690 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
693 * Create a mbuf pool.
695 * This function creates and initializes a packet mbuf pool. It is
696 * a wrapper to rte_mempool_create() with the proper packet constructor
697 * and mempool constructor.
700 * The name of the mbuf pool.
702 * The number of elements in the mbuf pool. The optimum size (in terms
703 * of memory usage) for a mempool is when n is a power of two minus one:
706 * Size of the per-core object cache. See rte_mempool_create() for
709 * Size of application private are between the rte_mbuf structure
710 * and the data buffer.
711 * @param data_room_size
712 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
714 * The socket identifier where the memory should be allocated. The
715 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
718 * The pointer to the new allocated mempool, on success. NULL on error
719 * with rte_errno set appropriately. Possible rte_errno values include:
720 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
721 * - E_RTE_SECONDARY - function was called from a secondary process instance
722 * - EINVAL - cache size provided is too large
723 * - ENOSPC - the maximum number of memzones has already been allocated
724 * - EEXIST - a memzone with the same name already exists
725 * - ENOMEM - no appropriate memory area found in which to create memzone
728 rte_pktmbuf_pool_create(const char *name, unsigned n,
729 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
733 * Get the data room size of mbufs stored in a pktmbuf_pool
735 * The data room size is the amount of data that can be stored in a
736 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
739 * The packet mbuf pool.
741 * The data room size of mbufs stored in this mempool.
743 static inline uint16_t
744 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
746 struct rte_pktmbuf_pool_private *mbp_priv;
748 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
749 return mbp_priv->mbuf_data_room_size;
753 * Get the application private size of mbufs stored in a pktmbuf_pool
755 * The private size of mbuf is a zone located between the rte_mbuf
756 * structure and the data buffer where an application can store data
757 * associated to a packet.
760 * The packet mbuf pool.
762 * The private size of mbufs stored in this mempool.
764 static inline uint16_t
765 rte_pktmbuf_priv_size(struct rte_mempool *mp)
767 struct rte_pktmbuf_pool_private *mbp_priv;
769 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
770 return mbp_priv->mbuf_priv_size;
774 * Reset the fields of a packet mbuf to their default values.
776 * The given mbuf must have only one segment.
779 * The packet mbuf to be resetted.
781 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
792 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
793 RTE_PKTMBUF_HEADROOM : m->buf_len;
796 __rte_mbuf_sanity_check(m, 1);
800 * Allocate a new mbuf from a mempool.
802 * This new mbuf contains one segment, which has a length of 0. The pointer
803 * to data is initialized to have some bytes of headroom in the buffer
804 * (if buffer size allows).
807 * The mempool from which the mbuf is allocated.
809 * - The pointer to the new mbuf on success.
810 * - NULL if allocation failed.
812 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
815 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
816 rte_pktmbuf_reset(m);
821 * Attach packet mbuf to another packet mbuf.
823 * After attachment we refer the mbuf we attached as 'indirect',
824 * while mbuf we attached to as 'direct'.
825 * Right now, not supported:
826 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
827 * - mbuf we trying to attach (mi) is used by someone else
828 * e.g. it's reference counter is greater then 1.
831 * The indirect packet mbuf.
833 * The packet mbuf we're attaching to.
835 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
839 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(mi) &&
840 rte_mbuf_refcnt_read(mi) == 1);
842 /* if m is not direct, get the mbuf that embeds the data */
843 if (RTE_MBUF_DIRECT(m))
846 md = rte_mbuf_from_indirect(m);
848 rte_mbuf_refcnt_update(md, 1);
849 mi->priv_size = m->priv_size;
850 mi->buf_physaddr = m->buf_physaddr;
851 mi->buf_addr = m->buf_addr;
852 mi->buf_len = m->buf_len;
855 mi->data_off = m->data_off;
856 mi->data_len = m->data_len;
858 mi->vlan_tci = m->vlan_tci;
859 mi->tx_offload = m->tx_offload;
863 mi->pkt_len = mi->data_len;
865 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
866 mi->packet_type = m->packet_type;
868 __rte_mbuf_sanity_check(mi, 1);
869 __rte_mbuf_sanity_check(m, 0);
873 * Detach an indirect packet mbuf.
875 * - restore original mbuf address and length values.
876 * - reset pktmbuf data and data_len to their default values.
877 * All other fields of the given packet mbuf will be left intact.
880 * The indirect attached packet mbuf.
882 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
884 struct rte_mempool *mp = m->pool;
885 uint32_t mbuf_size, buf_len, priv_size;
887 priv_size = rte_pktmbuf_priv_size(mp);
888 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
889 buf_len = rte_pktmbuf_data_room_size(mp);
891 m->priv_size = priv_size;
892 m->buf_addr = (char *)m + mbuf_size;
893 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
894 m->buf_len = (uint16_t)buf_len;
895 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
900 static inline struct rte_mbuf* __attribute__((always_inline))
901 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
903 __rte_mbuf_sanity_check(m, 0);
905 if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
907 /* if this is an indirect mbuf, then
909 * - free attached mbuf segment
911 if (RTE_MBUF_INDIRECT(m)) {
912 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
913 rte_pktmbuf_detach(m);
914 if (rte_mbuf_refcnt_update(md, -1) == 0)
915 __rte_mbuf_raw_free(md);
923 * Free a segment of a packet mbuf into its original mempool.
925 * Free an mbuf, without parsing other segments in case of chained
929 * The packet mbuf segment to be freed.
931 static inline void __attribute__((always_inline))
932 rte_pktmbuf_free_seg(struct rte_mbuf *m)
934 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
936 __rte_mbuf_raw_free(m);
941 * Free a packet mbuf back into its original mempool.
943 * Free an mbuf, and all its segments in case of chained buffers. Each
944 * segment is added back into its original mempool.
947 * The packet mbuf to be freed.
949 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
951 struct rte_mbuf *m_next;
953 __rte_mbuf_sanity_check(m, 1);
957 rte_pktmbuf_free_seg(m);
963 * Creates a "clone" of the given packet mbuf.
965 * Walks through all segments of the given packet mbuf, and for each of them:
966 * - Creates a new packet mbuf from the given pool.
967 * - Attaches newly created mbuf to the segment.
968 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
969 * from the original packet mbuf.
972 * The packet mbuf to be cloned.
974 * The mempool from which the "clone" mbufs are allocated.
976 * - The pointer to the new "clone" mbuf on success.
977 * - NULL if allocation fails.
979 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
980 struct rte_mempool *mp)
982 struct rte_mbuf *mc, *mi, **prev;
986 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
991 pktlen = md->pkt_len;
996 rte_pktmbuf_attach(mi, md);
999 } while ((md = md->next) != NULL &&
1000 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1004 mc->pkt_len = pktlen;
1006 /* Allocation of new indirect segment failed */
1007 if (unlikely (mi == NULL)) {
1008 rte_pktmbuf_free(mc);
1012 __rte_mbuf_sanity_check(mc, 1);
1017 * Adds given value to the refcnt of all packet mbuf segments.
1019 * Walks through all segments of given packet mbuf and for each of them
1020 * invokes rte_mbuf_refcnt_update().
1023 * The packet mbuf whose refcnt to be updated.
1025 * The value to add to the mbuf's segments refcnt.
1027 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1029 __rte_mbuf_sanity_check(m, 1);
1032 rte_mbuf_refcnt_update(m, v);
1033 } while ((m = m->next) != NULL);
1037 * Get the headroom in a packet mbuf.
1042 * The length of the headroom.
1044 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1046 __rte_mbuf_sanity_check(m, 1);
1051 * Get the tailroom of a packet mbuf.
1056 * The length of the tailroom.
1058 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1060 __rte_mbuf_sanity_check(m, 1);
1061 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1066 * Get the last segment of the packet.
1071 * The last segment of the given mbuf.
1073 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1075 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1077 __rte_mbuf_sanity_check(m, 1);
1078 while (m2->next != NULL)
1084 * A macro that points to an offset into the data in the mbuf.
1086 * The returned pointer is cast to type t. Before using this
1087 * function, the user must ensure that the first segment is large
1088 * enough to accommodate its data.
1093 * The offset into the mbuf data.
1095 * The type to cast the result into.
1097 #define rte_pktmbuf_mtod_offset(m, t, o) \
1098 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1101 * A macro that points to the start of the data in the mbuf.
1103 * The returned pointer is cast to type t. Before using this
1104 * function, the user must ensure that the first segment is large
1105 * enough to accommodate its data.
1110 * The type to cast the result into.
1112 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1115 * A macro that returns the length of the packet.
1117 * The value can be read or assigned.
1122 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1125 * A macro that returns the length of the segment.
1127 * The value can be read or assigned.
1132 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1135 * Prepend len bytes to an mbuf data area.
1137 * Returns a pointer to the new
1138 * data start address. If there is not enough headroom in the first
1139 * segment, the function will return NULL, without modifying the mbuf.
1144 * The amount of data to prepend (in bytes).
1146 * A pointer to the start of the newly prepended data, or
1147 * NULL if there is not enough headroom space in the first segment
1149 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1152 __rte_mbuf_sanity_check(m, 1);
1154 if (unlikely(len > rte_pktmbuf_headroom(m)))
1158 m->data_len = (uint16_t)(m->data_len + len);
1159 m->pkt_len = (m->pkt_len + len);
1161 return (char *)m->buf_addr + m->data_off;
1165 * Append len bytes to an mbuf.
1167 * Append len bytes to an mbuf and return a pointer to the start address
1168 * of the added data. If there is not enough tailroom in the last
1169 * segment, the function will return NULL, without modifying the mbuf.
1174 * The amount of data to append (in bytes).
1176 * A pointer to the start of the newly appended data, or
1177 * NULL if there is not enough tailroom space in the last segment
1179 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1182 struct rte_mbuf *m_last;
1184 __rte_mbuf_sanity_check(m, 1);
1186 m_last = rte_pktmbuf_lastseg(m);
1187 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1190 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1191 m_last->data_len = (uint16_t)(m_last->data_len + len);
1192 m->pkt_len = (m->pkt_len + len);
1193 return (char*) tail;
1197 * Remove len bytes at the beginning of an mbuf.
1199 * Returns a pointer to the start address of the new data area. If the
1200 * length is greater than the length of the first segment, then the
1201 * function will fail and return NULL, without modifying the mbuf.
1206 * The amount of data to remove (in bytes).
1208 * A pointer to the new start of the data.
1210 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1212 __rte_mbuf_sanity_check(m, 1);
1214 if (unlikely(len > m->data_len))
1217 m->data_len = (uint16_t)(m->data_len - len);
1219 m->pkt_len = (m->pkt_len - len);
1220 return (char *)m->buf_addr + m->data_off;
1224 * Remove len bytes of data at the end of the mbuf.
1226 * If the length is greater than the length of the last segment, the
1227 * function will fail and return -1 without modifying the mbuf.
1232 * The amount of data to remove (in bytes).
1237 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1239 struct rte_mbuf *m_last;
1241 __rte_mbuf_sanity_check(m, 1);
1243 m_last = rte_pktmbuf_lastseg(m);
1244 if (unlikely(len > m_last->data_len))
1247 m_last->data_len = (uint16_t)(m_last->data_len - len);
1248 m->pkt_len = (m->pkt_len - len);
1253 * Test if mbuf data is contiguous.
1258 * - 1, if all data is contiguous (one segment).
1259 * - 0, if there is several segments.
1261 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1263 __rte_mbuf_sanity_check(m, 1);
1264 return !!(m->nb_segs == 1);
1268 * Dump an mbuf structure to the console.
1270 * Dump all fields for the given packet mbuf and all its associated
1271 * segments (in the case of a chained buffer).
1274 * A pointer to a file for output
1278 * If dump_len != 0, also dump the "dump_len" first data bytes of
1281 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1287 #endif /* _RTE_MBUF_H_ */