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 options */
68 #pragma GCC poison RTE_MBUF_SCATTER_GATHER
69 #pragma GCC poison RTE_MBUF_REFCNT
72 * Packet Offload Features Flags. It also carry packet type information.
73 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
75 * - RX flags start at bit position zero, and get added to the left of previous
77 * - The most-significant 8 bits are reserved for generic mbuf flags
78 * - TX flags therefore start at bit position 55 (i.e. 63-8), and new flags get
79 * added to the right of the previously defined flags
81 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
82 * rte_get_tx_ol_flag_name().
84 #define PKT_RX_VLAN_PKT (1ULL << 0) /**< RX packet is a 802.1q VLAN packet. */
85 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
86 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
87 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3) /**< L4 cksum of RX pkt. is not OK. */
88 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4) /**< IP cksum of RX pkt. is not OK. */
89 #define PKT_RX_EIP_CKSUM_BAD (0ULL << 0) /**< External IP header checksum error. */
90 #define PKT_RX_OVERSIZE (0ULL << 0) /**< Num of desc of an RX pkt oversize. */
91 #define PKT_RX_HBUF_OVERFLOW (0ULL << 0) /**< Header buffer overflow. */
92 #define PKT_RX_RECIP_ERR (0ULL << 0) /**< Hardware processing error. */
93 #define PKT_RX_MAC_ERR (0ULL << 0) /**< MAC error. */
94 #define PKT_RX_IPV4_HDR (1ULL << 5) /**< RX packet with IPv4 header. */
95 #define PKT_RX_IPV4_HDR_EXT (1ULL << 6) /**< RX packet with extended IPv4 header. */
96 #define PKT_RX_IPV6_HDR (1ULL << 7) /**< RX packet with IPv6 header. */
97 #define PKT_RX_IPV6_HDR_EXT (1ULL << 8) /**< RX packet with extended IPv6 header. */
98 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
99 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
100 #define PKT_RX_TUNNEL_IPV4_HDR (1ULL << 11) /**< RX tunnel packet with IPv4 header.*/
101 #define PKT_RX_TUNNEL_IPV6_HDR (1ULL << 12) /**< RX tunnel packet with IPv6 header. */
102 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
103 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
104 /* add new RX flags here */
106 /* add new TX flags here */
109 * TCP segmentation offload. To enable this offload feature for a
110 * packet to be transmitted on hardware supporting TSO:
111 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
113 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
114 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
116 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
117 * - calculate the pseudo header checksum without taking ip_len in account,
118 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
119 * rte_ipv6_phdr_cksum() that can be used as helpers.
121 #define PKT_TX_TCP_SEG (1ULL << 50)
123 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
126 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
127 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
128 * L4 checksum offload, the user needs to:
129 * - fill l2_len and l3_len in mbuf
130 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
131 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
132 * - calculate the pseudo header checksum and set it in the L4 header (only
133 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
134 * For SCTP, set the crc field to 0.
136 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
137 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
138 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
139 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
140 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
143 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
144 * also be set by the application, although a PMD will only check
146 * - set the IP checksum field in the packet to 0
147 * - fill the mbuf offload information: l2_len, l3_len
149 #define PKT_TX_IP_CKSUM (1ULL << 54)
152 * Packet is IPv4. This flag must be set when using any offload feature
153 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
154 * packet. If the packet is a tunneled packet, this flag is related to
157 #define PKT_TX_IPV4 (1ULL << 55)
160 * Packet is IPv6. This flag must be set when using an offload feature
161 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
162 * packet. If the packet is a tunneled packet, this flag is related to
165 #define PKT_TX_IPV6 (1ULL << 56)
167 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
170 * Offload the IP checksum of an external header in the hardware. The
171 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
172 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
173 * packet must be set to 0.
174 * - set the outer IP checksum field in the packet to 0
175 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
177 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
180 * Packet outer header is IPv4. This flag must be set when using any
181 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
182 * outer header of the tunneled packet is an IPv4 packet.
184 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
187 * Packet outer header is IPv6. This flag must be set when using any
188 * outer offload feature (L4 checksum) to tell the NIC that the outer
189 * header of the tunneled packet is an IPv6 packet.
191 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
193 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
195 /* Use final bit of flags to indicate a control mbuf */
196 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
199 * Get the name of a RX offload flag
202 * The mask describing the flag.
204 * The name of this flag, or NULL if it's not a valid RX flag.
206 const char *rte_get_rx_ol_flag_name(uint64_t mask);
209 * Get the name of a TX offload flag
212 * The mask describing the flag. Usually only one bit must be set.
213 * Several bits can be given if they belong to the same mask.
214 * Ex: PKT_TX_L4_MASK.
216 * The name of this flag, or NULL if it's not a valid TX flag.
218 const char *rte_get_tx_ol_flag_name(uint64_t mask);
221 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
222 * splitting it into multiple segments.
223 * So, for mbufs that planned to be involved into RX/TX, the recommended
224 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
226 #define RTE_MBUF_DEFAULT_DATAROOM 2048
227 #define RTE_MBUF_DEFAULT_BUF_SIZE \
228 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
230 /* define a set of marker types that can be used to refer to set points in the
232 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
233 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
234 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
235 * with a single assignment */
238 * The generic rte_mbuf, containing a packet mbuf.
243 void *buf_addr; /**< Virtual address of segment buffer. */
244 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
246 uint16_t buf_len; /**< Length of segment buffer. */
248 /* next 6 bytes are initialised on RX descriptor rearm */
253 * 16-bit Reference counter.
254 * It should only be accessed using the following functions:
255 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
256 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
257 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
261 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
262 uint16_t refcnt; /**< Non-atomically accessed refcnt */
264 uint8_t nb_segs; /**< Number of segments. */
265 uint8_t port; /**< Input port. */
267 uint64_t ol_flags; /**< Offload features. */
269 /* remaining bytes are set on RX when pulling packet from descriptor */
270 MARKER rx_descriptor_fields1;
273 * The packet type, which is used to indicate ordinary packet and also
274 * tunneled packet format, i.e. each number is represented a type of
277 uint16_t packet_type;
279 uint16_t data_len; /**< Amount of data in segment buffer. */
280 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
281 uint16_t vlan_tci; /**< VLAN Tag Control Identifier (CPU order) */
284 uint32_t rss; /**< RSS hash result if RSS enabled */
292 /**< Second 4 flexible bytes */
295 /**< First 4 flexible bytes or FD ID, dependent on
296 PKT_RX_FDIR_* flag in ol_flags. */
297 } fdir; /**< Filter identifier if FDIR enabled */
298 uint32_t sched; /**< Hierarchical scheduler */
299 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
300 } hash; /**< hash information */
302 uint32_t seqn; /**< Sequence number. See also rte_reorder_insert() */
304 /* second cache line - fields only used in slow path or on TX */
305 MARKER cacheline1 __rte_cache_aligned;
308 void *userdata; /**< Can be used for external metadata */
309 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
312 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
313 struct rte_mbuf *next; /**< Next segment of scattered packet. */
315 /* fields to support TX offloads */
317 uint64_t tx_offload; /**< combined for easy fetch */
319 uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
320 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
321 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
322 uint64_t tso_segsz:16; /**< TCP TSO segment size */
324 /* fields for TX offloading of tunnels */
325 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
326 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
328 /* uint64_t unused:8; */
332 /** Size of the application private data. In case of an indirect
333 * mbuf, it stores the direct mbuf private data size. */
335 } __rte_cache_aligned;
337 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
340 * Return the mbuf owning the data buffer address of an indirect mbuf.
343 * The pointer to the indirect mbuf.
345 * The address of the direct mbuf corresponding to buffer_addr.
347 static inline struct rte_mbuf *
348 rte_mbuf_from_indirect(struct rte_mbuf *mi)
350 return RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
354 * Return the buffer address embedded in the given mbuf.
357 * The pointer to the mbuf.
359 * The address of the data buffer owned by the mbuf.
362 rte_mbuf_to_baddr(struct rte_mbuf *md)
365 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
370 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
372 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
375 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
377 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
380 * Private data in case of pktmbuf pool.
382 * A structure that contains some pktmbuf_pool-specific data that are
383 * appended after the mempool structure (in private data).
385 struct rte_pktmbuf_pool_private {
386 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
387 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
390 #ifdef RTE_LIBRTE_MBUF_DEBUG
392 /** check mbuf type in debug mode */
393 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
395 /** check mbuf type in debug mode if mbuf pointer is not null */
396 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
398 rte_mbuf_sanity_check(m, is_h); \
401 /** MBUF asserts in debug mode */
402 #define RTE_MBUF_ASSERT(exp) \
404 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
407 #else /* RTE_LIBRTE_MBUF_DEBUG */
409 /** check mbuf type in debug mode */
410 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
412 /** check mbuf type in debug mode if mbuf pointer is not null */
413 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
415 /** MBUF asserts in debug mode */
416 #define RTE_MBUF_ASSERT(exp) do { } while (0)
418 #endif /* RTE_LIBRTE_MBUF_DEBUG */
420 #ifdef RTE_MBUF_REFCNT_ATOMIC
423 * Reads the value of an mbuf's refcnt.
427 * Reference count number.
429 static inline uint16_t
430 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
432 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
436 * Sets an mbuf's refcnt to a defined value.
443 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
445 rte_atomic16_set(&m->refcnt_atomic, new_value);
449 * Adds given value to an mbuf's refcnt and returns its new value.
453 * Value to add/subtract
457 static inline uint16_t
458 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
461 * The atomic_add is an expensive operation, so we don't want to
462 * call it in the case where we know we are the uniq holder of
463 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
464 * operation has to be used because concurrent accesses on the
465 * reference counter can occur.
467 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
468 rte_mbuf_refcnt_set(m, 1 + value);
472 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
475 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
478 * Adds given value to an mbuf's refcnt and returns its new value.
480 static inline uint16_t
481 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
483 m->refcnt = (uint16_t)(m->refcnt + value);
488 * Reads the value of an mbuf's refcnt.
490 static inline uint16_t
491 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
497 * Sets an mbuf's refcnt to the defined value.
500 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
502 m->refcnt = new_value;
505 #endif /* RTE_MBUF_REFCNT_ATOMIC */
508 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
515 * Sanity checks on an mbuf.
517 * Check the consistency of the given mbuf. The function will cause a
518 * panic if corruption is detected.
521 * The mbuf to be checked.
523 * True if the mbuf is a packet header, false if it is a sub-segment
524 * of a packet (in this case, some fields like nb_segs are not checked)
527 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
530 * @internal Allocate a new mbuf from mempool *mp*.
531 * The use of that function is reserved for RTE internal needs.
532 * Please use rte_pktmbuf_alloc().
535 * The mempool from which mbuf is allocated.
537 * - The pointer to the new mbuf on success.
538 * - NULL if allocation failed.
540 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
544 if (rte_mempool_get(mp, &mb) < 0)
546 m = (struct rte_mbuf *)mb;
547 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
548 rte_mbuf_refcnt_set(m, 1);
553 * @internal Put mbuf back into its original mempool.
554 * The use of that function is reserved for RTE internal needs.
555 * Please use rte_pktmbuf_free().
558 * The mbuf to be freed.
560 static inline void __attribute__((always_inline))
561 __rte_mbuf_raw_free(struct rte_mbuf *m)
563 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
564 rte_mempool_put(m->pool, m);
567 /* Operations on ctrl mbuf */
570 * The control mbuf constructor.
572 * This function initializes some fields in an mbuf structure that are
573 * not modified by the user once created (mbuf type, origin pool, buffer
574 * start address, and so on). This function is given as a callback function
575 * to rte_mempool_create() at pool creation time.
578 * The mempool from which the mbuf is allocated.
580 * A pointer that can be used by the user to retrieve useful information
581 * for mbuf initialization. This pointer comes from the ``init_arg``
582 * parameter of rte_mempool_create().
584 * The mbuf to initialize.
586 * The index of the mbuf in the pool table.
588 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
589 void *m, unsigned i);
592 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
594 * This new mbuf is initialized with data pointing to the beginning of
595 * buffer, and with a length of zero.
598 * The mempool from which the mbuf is allocated.
600 * - The pointer to the new mbuf on success.
601 * - NULL if allocation failed.
603 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
606 * Free a control mbuf back into its original mempool.
609 * The control mbuf to be freed.
611 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
614 * A macro that returns the pointer to the carried data.
616 * The value that can be read or assigned.
621 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
624 * A macro that returns the length of the carried data.
626 * The value that can be read or assigned.
631 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
634 * Tests if an mbuf is a control mbuf
637 * The mbuf to be tested
639 * - True (1) if the mbuf is a control mbuf
640 * - False(0) otherwise
643 rte_is_ctrlmbuf(struct rte_mbuf *m)
645 return !!(m->ol_flags & CTRL_MBUF_FLAG);
648 /* Operations on pkt mbuf */
651 * The packet mbuf constructor.
653 * This function initializes some fields in the mbuf structure that are
654 * not modified by the user once created (origin pool, buffer start
655 * address, and so on). This function is given as a callback function to
656 * rte_mempool_create() at pool creation time.
659 * The mempool from which mbufs originate.
661 * A pointer that can be used by the user to retrieve useful information
662 * for mbuf initialization. This pointer comes from the ``init_arg``
663 * parameter of rte_mempool_create().
665 * The mbuf to initialize.
667 * The index of the mbuf in the pool table.
669 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
670 void *m, unsigned i);
674 * A packet mbuf pool constructor.
676 * This function initializes the mempool private data in the case of a
677 * pktmbuf pool. This private data is needed by the driver. The
678 * function is given as a callback function to rte_mempool_create() at
679 * pool creation. It can be extended by the user, for example, to
680 * provide another packet size.
683 * The mempool from which mbufs originate.
685 * A pointer that can be used by the user to retrieve useful information
686 * for mbuf initialization. This pointer comes from the ``init_arg``
687 * parameter of rte_mempool_create().
689 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
692 * Create a mbuf pool.
694 * This function creates and initializes a packet mbuf pool. It is
695 * a wrapper to rte_mempool_create() with the proper packet constructor
696 * and mempool constructor.
699 * The name of the mbuf pool.
701 * The number of elements in the mbuf pool. The optimum size (in terms
702 * of memory usage) for a mempool is when n is a power of two minus one:
705 * Size of the per-core object cache. See rte_mempool_create() for
708 * Size of application private are between the rte_mbuf structure
709 * and the data buffer.
710 * @param data_room_size
711 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
713 * The socket identifier where the memory should be allocated. The
714 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
717 * The pointer to the new allocated mempool, on success. NULL on error
718 * with rte_errno set appropriately. Possible rte_errno values include:
719 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
720 * - E_RTE_SECONDARY - function was called from a secondary process instance
721 * - EINVAL - cache size provided is too large
722 * - ENOSPC - the maximum number of memzones has already been allocated
723 * - EEXIST - a memzone with the same name already exists
724 * - ENOMEM - no appropriate memory area found in which to create memzone
727 rte_pktmbuf_pool_create(const char *name, unsigned n,
728 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
732 * Get the data room size of mbufs stored in a pktmbuf_pool
734 * The data room size is the amount of data that can be stored in a
735 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
738 * The packet mbuf pool.
740 * The data room size of mbufs stored in this mempool.
742 static inline uint16_t
743 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
745 struct rte_pktmbuf_pool_private *mbp_priv;
747 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
748 return mbp_priv->mbuf_data_room_size;
752 * Get the application private size of mbufs stored in a pktmbuf_pool
754 * The private size of mbuf is a zone located between the rte_mbuf
755 * structure and the data buffer where an application can store data
756 * associated to a packet.
759 * The packet mbuf pool.
761 * The private size of mbufs stored in this mempool.
763 static inline uint16_t
764 rte_pktmbuf_priv_size(struct rte_mempool *mp)
766 struct rte_pktmbuf_pool_private *mbp_priv;
768 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
769 return mbp_priv->mbuf_priv_size;
773 * Reset the fields of a packet mbuf to their default values.
775 * The given mbuf must have only one segment.
778 * The packet mbuf to be resetted.
780 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
791 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
792 RTE_PKTMBUF_HEADROOM : m->buf_len;
795 __rte_mbuf_sanity_check(m, 1);
799 * Allocate a new mbuf from a mempool.
801 * This new mbuf contains one segment, which has a length of 0. The pointer
802 * to data is initialized to have some bytes of headroom in the buffer
803 * (if buffer size allows).
806 * The mempool from which the mbuf is allocated.
808 * - The pointer to the new mbuf on success.
809 * - NULL if allocation failed.
811 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
814 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
815 rte_pktmbuf_reset(m);
820 * Attach packet mbuf to another packet mbuf.
822 * After attachment we refer the mbuf we attached as 'indirect',
823 * while mbuf we attached to as 'direct'.
824 * Right now, not supported:
825 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
826 * - mbuf we trying to attach (mi) is used by someone else
827 * e.g. it's reference counter is greater then 1.
830 * The indirect packet mbuf.
832 * The packet mbuf we're attaching to.
834 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
838 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(mi) &&
839 rte_mbuf_refcnt_read(mi) == 1);
841 /* if m is not direct, get the mbuf that embeds the data */
842 if (RTE_MBUF_DIRECT(m))
845 md = rte_mbuf_from_indirect(m);
847 rte_mbuf_refcnt_update(md, 1);
848 mi->priv_size = m->priv_size;
849 mi->buf_physaddr = m->buf_physaddr;
850 mi->buf_addr = m->buf_addr;
851 mi->buf_len = m->buf_len;
854 mi->data_off = m->data_off;
855 mi->data_len = m->data_len;
857 mi->vlan_tci = m->vlan_tci;
858 mi->tx_offload = m->tx_offload;
862 mi->pkt_len = mi->data_len;
864 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
865 mi->packet_type = m->packet_type;
867 __rte_mbuf_sanity_check(mi, 1);
868 __rte_mbuf_sanity_check(m, 0);
872 * Detach an indirect packet mbuf.
874 * - restore original mbuf address and length values.
875 * - reset pktmbuf data and data_len to their default values.
876 * All other fields of the given packet mbuf will be left intact.
879 * The indirect attached packet mbuf.
881 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
883 struct rte_mempool *mp = m->pool;
884 uint32_t mbuf_size, buf_len, priv_size;
886 priv_size = rte_pktmbuf_priv_size(mp);
887 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
888 buf_len = rte_pktmbuf_data_room_size(mp);
890 m->priv_size = priv_size;
891 m->buf_addr = (char *)m + mbuf_size;
892 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
893 m->buf_len = (uint16_t)buf_len;
894 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
899 static inline struct rte_mbuf* __attribute__((always_inline))
900 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
902 __rte_mbuf_sanity_check(m, 0);
904 if (likely(rte_mbuf_refcnt_update(m, -1) == 0)) {
906 /* if this is an indirect mbuf, then
908 * - free attached mbuf segment
910 if (RTE_MBUF_INDIRECT(m)) {
911 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
912 rte_pktmbuf_detach(m);
913 if (rte_mbuf_refcnt_update(md, -1) == 0)
914 __rte_mbuf_raw_free(md);
922 * Free a segment of a packet mbuf into its original mempool.
924 * Free an mbuf, without parsing other segments in case of chained
928 * The packet mbuf segment to be freed.
930 static inline void __attribute__((always_inline))
931 rte_pktmbuf_free_seg(struct rte_mbuf *m)
933 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
935 __rte_mbuf_raw_free(m);
940 * Free a packet mbuf back into its original mempool.
942 * Free an mbuf, and all its segments in case of chained buffers. Each
943 * segment is added back into its original mempool.
946 * The packet mbuf to be freed.
948 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
950 struct rte_mbuf *m_next;
952 __rte_mbuf_sanity_check(m, 1);
956 rte_pktmbuf_free_seg(m);
962 * Creates a "clone" of the given packet mbuf.
964 * Walks through all segments of the given packet mbuf, and for each of them:
965 * - Creates a new packet mbuf from the given pool.
966 * - Attaches newly created mbuf to the segment.
967 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
968 * from the original packet mbuf.
971 * The packet mbuf to be cloned.
973 * The mempool from which the "clone" mbufs are allocated.
975 * - The pointer to the new "clone" mbuf on success.
976 * - NULL if allocation fails.
978 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
979 struct rte_mempool *mp)
981 struct rte_mbuf *mc, *mi, **prev;
985 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
990 pktlen = md->pkt_len;
995 rte_pktmbuf_attach(mi, md);
998 } while ((md = md->next) != NULL &&
999 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1003 mc->pkt_len = pktlen;
1005 /* Allocation of new indirect segment failed */
1006 if (unlikely (mi == NULL)) {
1007 rte_pktmbuf_free(mc);
1011 __rte_mbuf_sanity_check(mc, 1);
1016 * Adds given value to the refcnt of all packet mbuf segments.
1018 * Walks through all segments of given packet mbuf and for each of them
1019 * invokes rte_mbuf_refcnt_update().
1022 * The packet mbuf whose refcnt to be updated.
1024 * The value to add to the mbuf's segments refcnt.
1026 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1028 __rte_mbuf_sanity_check(m, 1);
1031 rte_mbuf_refcnt_update(m, v);
1032 } while ((m = m->next) != NULL);
1036 * Get the headroom in a packet mbuf.
1041 * The length of the headroom.
1043 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1045 __rte_mbuf_sanity_check(m, 1);
1050 * Get the tailroom of a packet mbuf.
1055 * The length of the tailroom.
1057 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1059 __rte_mbuf_sanity_check(m, 1);
1060 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1065 * Get the last segment of the packet.
1070 * The last segment of the given mbuf.
1072 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1074 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1076 __rte_mbuf_sanity_check(m, 1);
1077 while (m2->next != NULL)
1083 * A macro that points to the start of the data in the mbuf.
1085 * The returned pointer is cast to type t. Before using this
1086 * function, the user must ensure that m_headlen(m) is large enough to
1092 * The type to cast the result into.
1094 #define rte_pktmbuf_mtod(m, t) ((t)((char *)(m)->buf_addr + (m)->data_off))
1097 * A macro that returns the length of the packet.
1099 * The value can be read or assigned.
1104 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1107 * A macro that returns the length of the segment.
1109 * The value can be read or assigned.
1114 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1117 * Prepend len bytes to an mbuf data area.
1119 * Returns a pointer to the new
1120 * data start address. If there is not enough headroom in the first
1121 * segment, the function will return NULL, without modifying the mbuf.
1126 * The amount of data to prepend (in bytes).
1128 * A pointer to the start of the newly prepended data, or
1129 * NULL if there is not enough headroom space in the first segment
1131 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1134 __rte_mbuf_sanity_check(m, 1);
1136 if (unlikely(len > rte_pktmbuf_headroom(m)))
1140 m->data_len = (uint16_t)(m->data_len + len);
1141 m->pkt_len = (m->pkt_len + len);
1143 return (char *)m->buf_addr + m->data_off;
1147 * Append len bytes to an mbuf.
1149 * Append len bytes to an mbuf and return a pointer to the start address
1150 * of the added data. If there is not enough tailroom in the last
1151 * segment, the function will return NULL, without modifying the mbuf.
1156 * The amount of data to append (in bytes).
1158 * A pointer to the start of the newly appended data, or
1159 * NULL if there is not enough tailroom space in the last segment
1161 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1164 struct rte_mbuf *m_last;
1166 __rte_mbuf_sanity_check(m, 1);
1168 m_last = rte_pktmbuf_lastseg(m);
1169 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1172 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1173 m_last->data_len = (uint16_t)(m_last->data_len + len);
1174 m->pkt_len = (m->pkt_len + len);
1175 return (char*) tail;
1179 * Remove len bytes at the beginning of an mbuf.
1181 * Returns a pointer to the start address of the new data area. If the
1182 * length is greater than the length of the first segment, then the
1183 * function will fail and return NULL, without modifying the mbuf.
1188 * The amount of data to remove (in bytes).
1190 * A pointer to the new start of the data.
1192 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1194 __rte_mbuf_sanity_check(m, 1);
1196 if (unlikely(len > m->data_len))
1199 m->data_len = (uint16_t)(m->data_len - len);
1201 m->pkt_len = (m->pkt_len - len);
1202 return (char *)m->buf_addr + m->data_off;
1206 * Remove len bytes of data at the end of the mbuf.
1208 * If the length is greater than the length of the last segment, the
1209 * function will fail and return -1 without modifying the mbuf.
1214 * The amount of data to remove (in bytes).
1219 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1221 struct rte_mbuf *m_last;
1223 __rte_mbuf_sanity_check(m, 1);
1225 m_last = rte_pktmbuf_lastseg(m);
1226 if (unlikely(len > m_last->data_len))
1229 m_last->data_len = (uint16_t)(m_last->data_len - len);
1230 m->pkt_len = (m->pkt_len - len);
1235 * Test if mbuf data is contiguous.
1240 * - 1, if all data is contiguous (one segment).
1241 * - 0, if there is several segments.
1243 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1245 __rte_mbuf_sanity_check(m, 1);
1246 return !!(m->nb_segs == 1);
1250 * Dump an mbuf structure to the console.
1252 * Dump all fields for the given packet mbuf and all its associated
1253 * segments (in the case of a chained buffer).
1256 * A pointer to a file for output
1260 * If dump_len != 0, also dump the "dump_len" first data bytes of
1263 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1269 #endif /* _RTE_MBUF_H_ */