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)
352 /* mi->buf_addr and mi->priv_size correspond to buffer and
353 * private size of the direct mbuf */
354 md = (struct rte_mbuf *)((char *)mi->buf_addr - sizeof(*mi) -
360 * Return the buffer address embedded in the given mbuf.
363 * The pointer to the mbuf.
365 * The address of the data buffer owned by the mbuf.
368 rte_mbuf_to_baddr(struct rte_mbuf *md)
371 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
376 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
378 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
381 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
383 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
386 * Private data in case of pktmbuf pool.
388 * A structure that contains some pktmbuf_pool-specific data that are
389 * appended after the mempool structure (in private data).
391 struct rte_pktmbuf_pool_private {
392 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
393 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
396 #ifdef RTE_LIBRTE_MBUF_DEBUG
398 /** check mbuf type in debug mode */
399 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
401 /** check mbuf type in debug mode if mbuf pointer is not null */
402 #define __rte_mbuf_sanity_check_raw(m, is_h) do { \
404 rte_mbuf_sanity_check(m, is_h); \
407 /** MBUF asserts in debug mode */
408 #define RTE_MBUF_ASSERT(exp) \
410 rte_panic("line%d\tassert \"" #exp "\" failed\n", __LINE__); \
413 #else /* RTE_LIBRTE_MBUF_DEBUG */
415 /** check mbuf type in debug mode */
416 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
418 /** check mbuf type in debug mode if mbuf pointer is not null */
419 #define __rte_mbuf_sanity_check_raw(m, is_h) do { } while (0)
421 /** MBUF asserts in debug mode */
422 #define RTE_MBUF_ASSERT(exp) do { } while (0)
424 #endif /* RTE_LIBRTE_MBUF_DEBUG */
426 #ifdef RTE_MBUF_REFCNT_ATOMIC
429 * Adds given value to an mbuf's refcnt and returns its new value.
433 * Value to add/subtract
437 static inline uint16_t
438 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
440 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
444 * Reads the value of an mbuf's refcnt.
448 * Reference count number.
450 static inline uint16_t
451 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
453 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
457 * Sets an mbuf's refcnt to a defined value.
464 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
466 rte_atomic16_set(&m->refcnt_atomic, new_value);
469 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
472 * Adds given value to an mbuf's refcnt and returns its new value.
474 static inline uint16_t
475 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
477 m->refcnt = (uint16_t)(m->refcnt + value);
482 * Reads the value of an mbuf's refcnt.
484 static inline uint16_t
485 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
491 * Sets an mbuf's refcnt to the defined value.
494 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
496 m->refcnt = new_value;
499 #endif /* RTE_MBUF_REFCNT_ATOMIC */
502 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
509 * Sanity checks on an mbuf.
511 * Check the consistency of the given mbuf. The function will cause a
512 * panic if corruption is detected.
515 * The mbuf to be checked.
517 * True if the mbuf is a packet header, false if it is a sub-segment
518 * of a packet (in this case, some fields like nb_segs are not checked)
521 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
524 * @internal Allocate a new mbuf from mempool *mp*.
525 * The use of that function is reserved for RTE internal needs.
526 * Please use rte_pktmbuf_alloc().
529 * The mempool from which mbuf is allocated.
531 * - The pointer to the new mbuf on success.
532 * - NULL if allocation failed.
534 static inline struct rte_mbuf *__rte_mbuf_raw_alloc(struct rte_mempool *mp)
538 if (rte_mempool_get(mp, &mb) < 0)
540 m = (struct rte_mbuf *)mb;
541 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
542 rte_mbuf_refcnt_set(m, 1);
547 * @internal Put mbuf back into its original mempool.
548 * The use of that function is reserved for RTE internal needs.
549 * Please use rte_pktmbuf_free().
552 * The mbuf to be freed.
554 static inline void __attribute__((always_inline))
555 __rte_mbuf_raw_free(struct rte_mbuf *m)
557 RTE_MBUF_ASSERT(rte_mbuf_refcnt_read(m) == 0);
558 rte_mempool_put(m->pool, m);
561 /* Operations on ctrl mbuf */
564 * The control mbuf constructor.
566 * This function initializes some fields in an mbuf structure that are
567 * not modified by the user once created (mbuf type, origin pool, buffer
568 * start address, and so on). This function is given as a callback function
569 * to rte_mempool_create() at pool creation time.
572 * The mempool from which the mbuf is allocated.
574 * A pointer that can be used by the user to retrieve useful information
575 * for mbuf initialization. This pointer comes from the ``init_arg``
576 * parameter of rte_mempool_create().
578 * The mbuf to initialize.
580 * The index of the mbuf in the pool table.
582 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
583 void *m, unsigned i);
586 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
588 * This new mbuf is initialized with data pointing to the beginning of
589 * buffer, and with a length of zero.
592 * The mempool from which the mbuf is allocated.
594 * - The pointer to the new mbuf on success.
595 * - NULL if allocation failed.
597 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
600 * Free a control mbuf back into its original mempool.
603 * The control mbuf to be freed.
605 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
608 * A macro that returns the pointer to the carried data.
610 * The value that can be read or assigned.
615 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
618 * A macro that returns the length of the carried data.
620 * The value that can be read or assigned.
625 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
628 * Tests if an mbuf is a control mbuf
631 * The mbuf to be tested
633 * - True (1) if the mbuf is a control mbuf
634 * - False(0) otherwise
637 rte_is_ctrlmbuf(struct rte_mbuf *m)
639 return (!!(m->ol_flags & CTRL_MBUF_FLAG));
642 /* Operations on pkt mbuf */
645 * The packet mbuf constructor.
647 * This function initializes some fields in the mbuf structure that are
648 * not modified by the user once created (origin pool, buffer start
649 * address, and so on). This function is given as a callback function to
650 * rte_mempool_create() at pool creation time.
653 * The mempool from which mbufs originate.
655 * A pointer that can be used by the user to retrieve useful information
656 * for mbuf initialization. This pointer comes from the ``init_arg``
657 * parameter of rte_mempool_create().
659 * The mbuf to initialize.
661 * The index of the mbuf in the pool table.
663 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
664 void *m, unsigned i);
668 * A packet mbuf pool constructor.
670 * This function initializes the mempool private data in the case of a
671 * pktmbuf pool. This private data is needed by the driver. The
672 * function is given as a callback function to rte_mempool_create() at
673 * pool creation. It can be extended by the user, for example, to
674 * provide another packet size.
677 * The mempool from which mbufs originate.
679 * A pointer that can be used by the user to retrieve useful information
680 * for mbuf initialization. This pointer comes from the ``init_arg``
681 * parameter of rte_mempool_create().
683 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
686 * Create a mbuf pool.
688 * This function creates and initializes a packet mbuf pool. It is
689 * a wrapper to rte_mempool_create() with the proper packet constructor
690 * and mempool constructor.
693 * The name of the mbuf pool.
695 * The number of elements in the mbuf pool. The optimum size (in terms
696 * of memory usage) for a mempool is when n is a power of two minus one:
699 * Size of the per-core object cache. See rte_mempool_create() for
702 * Size of application private are between the rte_mbuf structure
703 * and the data buffer.
704 * @param data_room_size
705 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
707 * The socket identifier where the memory should be allocated. The
708 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
711 * The pointer to the new allocated mempool, on success. NULL on error
712 * with rte_errno set appropriately. Possible rte_errno values include:
713 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
714 * - E_RTE_SECONDARY - function was called from a secondary process instance
715 * - EINVAL - cache size provided is too large
716 * - ENOSPC - the maximum number of memzones has already been allocated
717 * - EEXIST - a memzone with the same name already exists
718 * - ENOMEM - no appropriate memory area found in which to create memzone
721 rte_pktmbuf_pool_create(const char *name, unsigned n,
722 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
726 * Get the data room size of mbufs stored in a pktmbuf_pool
728 * The data room size is the amount of data that can be stored in a
729 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
732 * The packet mbuf pool.
734 * The data room size of mbufs stored in this mempool.
736 static inline uint16_t
737 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
739 struct rte_pktmbuf_pool_private *mbp_priv;
741 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
742 return mbp_priv->mbuf_data_room_size;
746 * Get the application private size of mbufs stored in a pktmbuf_pool
748 * The private size of mbuf is a zone located between the rte_mbuf
749 * structure and the data buffer where an application can store data
750 * associated to a packet.
753 * The packet mbuf pool.
755 * The private size of mbufs stored in this mempool.
757 static inline uint16_t
758 rte_pktmbuf_priv_size(struct rte_mempool *mp)
760 struct rte_pktmbuf_pool_private *mbp_priv;
762 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
763 return mbp_priv->mbuf_priv_size;
767 * Reset the fields of a packet mbuf to their default values.
769 * The given mbuf must have only one segment.
772 * The packet mbuf to be resetted.
774 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
785 m->data_off = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
786 RTE_PKTMBUF_HEADROOM : m->buf_len;
789 __rte_mbuf_sanity_check(m, 1);
793 * Allocate a new mbuf from a mempool.
795 * This new mbuf contains one segment, which has a length of 0. The pointer
796 * to data is initialized to have some bytes of headroom in the buffer
797 * (if buffer size allows).
800 * The mempool from which the mbuf is allocated.
802 * - The pointer to the new mbuf on success.
803 * - NULL if allocation failed.
805 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
808 if ((m = __rte_mbuf_raw_alloc(mp)) != NULL)
809 rte_pktmbuf_reset(m);
814 * Attach packet mbuf to another packet mbuf.
816 * After attachment we refer the mbuf we attached as 'indirect',
817 * while mbuf we attached to as 'direct'.
818 * Right now, not supported:
819 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
820 * - mbuf we trying to attach (mi) is used by someone else
821 * e.g. it's reference counter is greater then 1.
824 * The indirect packet mbuf.
826 * The packet mbuf we're attaching to.
828 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
832 RTE_MBUF_ASSERT(RTE_MBUF_DIRECT(mi) &&
833 rte_mbuf_refcnt_read(mi) == 1);
835 /* if m is not direct, get the mbuf that embeds the data */
836 if (RTE_MBUF_DIRECT(m))
839 md = rte_mbuf_from_indirect(m);
841 rte_mbuf_refcnt_update(md, 1);
842 mi->priv_size = m->priv_size;
843 mi->buf_physaddr = m->buf_physaddr;
844 mi->buf_addr = m->buf_addr;
845 mi->buf_len = m->buf_len;
848 mi->data_off = m->data_off;
849 mi->data_len = m->data_len;
851 mi->vlan_tci = m->vlan_tci;
852 mi->tx_offload = m->tx_offload;
856 mi->pkt_len = mi->data_len;
858 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
859 mi->packet_type = m->packet_type;
861 __rte_mbuf_sanity_check(mi, 1);
862 __rte_mbuf_sanity_check(m, 0);
866 * Detach an indirect packet mbuf.
868 * - restore original mbuf address and length values.
869 * - reset pktmbuf data and data_len to their default values.
870 * All other fields of the given packet mbuf will be left intact.
873 * The indirect attached packet mbuf.
875 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
877 struct rte_mempool *mp = m->pool;
878 uint32_t mbuf_size, buf_len, priv_size;
880 priv_size = rte_pktmbuf_priv_size(mp);
881 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
882 buf_len = rte_pktmbuf_data_room_size(mp);
884 m->priv_size = priv_size;
885 m->buf_addr = (char *)m + mbuf_size;
886 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
887 m->buf_len = (uint16_t)buf_len;
888 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
893 static inline struct rte_mbuf* __attribute__((always_inline))
894 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
896 __rte_mbuf_sanity_check(m, 0);
899 * Check to see if this is the last reference to the mbuf.
900 * Note: the double check here is deliberate. If the ref_cnt is "atomic"
901 * the call to "refcnt_update" is a very expensive operation, so we
902 * don't want to call it in the case where we know we are the holder
903 * of the last reference to this mbuf i.e. ref_cnt == 1.
904 * If however, ref_cnt != 1, it's still possible that we may still be
905 * the final decrementer of the count, so we need to check that
906 * result also, to make sure the mbuf is freed properly.
908 if (likely (rte_mbuf_refcnt_read(m) == 1) ||
909 likely (rte_mbuf_refcnt_update(m, -1) == 0)) {
911 rte_mbuf_refcnt_set(m, 0);
913 /* if this is an indirect mbuf, then
915 * - free attached mbuf segment
917 if (RTE_MBUF_INDIRECT(m)) {
918 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
919 rte_pktmbuf_detach(m);
920 if (rte_mbuf_refcnt_update(md, -1) == 0)
921 __rte_mbuf_raw_free(md);
929 * Free a segment of a packet mbuf into its original mempool.
931 * Free an mbuf, without parsing other segments in case of chained
935 * The packet mbuf segment to be freed.
937 static inline void __attribute__((always_inline))
938 rte_pktmbuf_free_seg(struct rte_mbuf *m)
940 if (likely(NULL != (m = __rte_pktmbuf_prefree_seg(m)))) {
942 __rte_mbuf_raw_free(m);
947 * Free a packet mbuf back into its original mempool.
949 * Free an mbuf, and all its segments in case of chained buffers. Each
950 * segment is added back into its original mempool.
953 * The packet mbuf to be freed.
955 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
957 struct rte_mbuf *m_next;
959 __rte_mbuf_sanity_check(m, 1);
963 rte_pktmbuf_free_seg(m);
969 * Creates a "clone" of the given packet mbuf.
971 * Walks through all segments of the given packet mbuf, and for each of them:
972 * - Creates a new packet mbuf from the given pool.
973 * - Attaches newly created mbuf to the segment.
974 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
975 * from the original packet mbuf.
978 * The packet mbuf to be cloned.
980 * The mempool from which the "clone" mbufs are allocated.
982 * - The pointer to the new "clone" mbuf on success.
983 * - NULL if allocation fails.
985 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
986 struct rte_mempool *mp)
988 struct rte_mbuf *mc, *mi, **prev;
992 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
997 pktlen = md->pkt_len;
1002 rte_pktmbuf_attach(mi, md);
1005 } while ((md = md->next) != NULL &&
1006 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1010 mc->pkt_len = pktlen;
1012 /* Allocation of new indirect segment failed */
1013 if (unlikely (mi == NULL)) {
1014 rte_pktmbuf_free(mc);
1018 __rte_mbuf_sanity_check(mc, 1);
1023 * Adds given value to the refcnt of all packet mbuf segments.
1025 * Walks through all segments of given packet mbuf and for each of them
1026 * invokes rte_mbuf_refcnt_update().
1029 * The packet mbuf whose refcnt to be updated.
1031 * The value to add to the mbuf's segments refcnt.
1033 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1035 __rte_mbuf_sanity_check(m, 1);
1038 rte_mbuf_refcnt_update(m, v);
1039 } while ((m = m->next) != NULL);
1043 * Get the headroom in a packet mbuf.
1048 * The length of the headroom.
1050 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1052 __rte_mbuf_sanity_check(m, 1);
1057 * Get the tailroom of a packet mbuf.
1062 * The length of the tailroom.
1064 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1066 __rte_mbuf_sanity_check(m, 1);
1067 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1072 * Get the last segment of the packet.
1077 * The last segment of the given mbuf.
1079 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1081 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1083 __rte_mbuf_sanity_check(m, 1);
1084 while (m2->next != NULL)
1090 * A macro that points to the start of the data in the mbuf.
1092 * The returned pointer is cast to type t. Before using this
1093 * function, the user must ensure that m_headlen(m) is large enough to
1099 * The type to cast the result into.
1101 #define rte_pktmbuf_mtod(m, t) ((t)((char *)(m)->buf_addr + (m)->data_off))
1104 * A macro that returns the length of the packet.
1106 * The value can be read or assigned.
1111 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1114 * A macro that returns the length of the segment.
1116 * The value can be read or assigned.
1121 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1124 * Prepend len bytes to an mbuf data area.
1126 * Returns a pointer to the new
1127 * data start address. If there is not enough headroom in the first
1128 * segment, the function will return NULL, without modifying the mbuf.
1133 * The amount of data to prepend (in bytes).
1135 * A pointer to the start of the newly prepended data, or
1136 * NULL if there is not enough headroom space in the first segment
1138 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1141 __rte_mbuf_sanity_check(m, 1);
1143 if (unlikely(len > rte_pktmbuf_headroom(m)))
1147 m->data_len = (uint16_t)(m->data_len + len);
1148 m->pkt_len = (m->pkt_len + len);
1150 return (char *)m->buf_addr + m->data_off;
1154 * Append len bytes to an mbuf.
1156 * Append len bytes to an mbuf and return a pointer to the start address
1157 * of the added data. If there is not enough tailroom in the last
1158 * segment, the function will return NULL, without modifying the mbuf.
1163 * The amount of data to append (in bytes).
1165 * A pointer to the start of the newly appended data, or
1166 * NULL if there is not enough tailroom space in the last segment
1168 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1171 struct rte_mbuf *m_last;
1173 __rte_mbuf_sanity_check(m, 1);
1175 m_last = rte_pktmbuf_lastseg(m);
1176 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1179 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1180 m_last->data_len = (uint16_t)(m_last->data_len + len);
1181 m->pkt_len = (m->pkt_len + len);
1182 return (char*) tail;
1186 * Remove len bytes at the beginning of an mbuf.
1188 * Returns a pointer to the start address of the new data area. If the
1189 * length is greater than the length of the first segment, then the
1190 * function will fail and return NULL, without modifying the mbuf.
1195 * The amount of data to remove (in bytes).
1197 * A pointer to the new start of the data.
1199 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1201 __rte_mbuf_sanity_check(m, 1);
1203 if (unlikely(len > m->data_len))
1206 m->data_len = (uint16_t)(m->data_len - len);
1208 m->pkt_len = (m->pkt_len - len);
1209 return (char *)m->buf_addr + m->data_off;
1213 * Remove len bytes of data at the end of the mbuf.
1215 * If the length is greater than the length of the last segment, the
1216 * function will fail and return -1 without modifying the mbuf.
1221 * The amount of data to remove (in bytes).
1226 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1228 struct rte_mbuf *m_last;
1230 __rte_mbuf_sanity_check(m, 1);
1232 m_last = rte_pktmbuf_lastseg(m);
1233 if (unlikely(len > m_last->data_len))
1236 m_last->data_len = (uint16_t)(m_last->data_len - len);
1237 m->pkt_len = (m->pkt_len - len);
1242 * Test if mbuf data is contiguous.
1247 * - 1, if all data is contiguous (one segment).
1248 * - 0, if there is several segments.
1250 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1252 __rte_mbuf_sanity_check(m, 1);
1253 return !!(m->nb_segs == 1);
1257 * Dump an mbuf structure to the console.
1259 * Dump all fields for the given packet mbuf and all its associated
1260 * segments (in the case of a chained buffer).
1263 * A pointer to a file for output
1267 * If dump_len != 0, also dump the "dump_len" first data bytes of
1270 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1276 #endif /* _RTE_MBUF_H_ */