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 * The preferred way to create a mbuf pool is to use
48 * rte_pktmbuf_pool_create(). However, in some situations, an
49 * application may want to have more control (ex: populate the pool with
50 * specific memory), in this case it is possible to use functions from
51 * rte_mempool. See how rte_pktmbuf_pool_create() is implemented for
54 * This library provides an API to allocate/free packet mbufs, which are
55 * used to carry network packets.
57 * To understand the concepts of packet buffers or mbufs, you
58 * should read "TCP/IP Illustrated, Volume 2: The Implementation,
59 * Addison-Wesley, 1995, ISBN 0-201-63354-X from Richard Stevens"
60 * http://www.kohala.com/start/tcpipiv2.html
64 #include <rte_common.h>
65 #include <rte_mempool.h>
66 #include <rte_memory.h>
67 #include <rte_atomic.h>
68 #include <rte_prefetch.h>
69 #include <rte_branch_prediction.h>
70 #include <rte_mbuf_ptype.h>
77 * Packet Offload Features Flags. It also carry packet type information.
78 * Critical resources. Both rx/tx shared these bits. Be cautious on any change
80 * - RX flags start at bit position zero, and get added to the left of previous
82 * - The most-significant 3 bits are reserved for generic mbuf flags
83 * - TX flags therefore start at bit position 60 (i.e. 63-3), and new flags get
84 * added to the right of the previously defined flags i.e. they should count
85 * downwards, not upwards.
87 * Keep these flags synchronized with rte_get_rx_ol_flag_name() and
88 * rte_get_tx_ol_flag_name().
92 * RX packet is a 802.1q VLAN packet. This flag was set by PMDs when
93 * the packet is recognized as a VLAN, but the behavior between PMDs
94 * was not the same. This flag is kept for some time to avoid breaking
95 * applications and should be replaced by PKT_RX_VLAN_STRIPPED.
97 #define PKT_RX_VLAN_PKT (1ULL << 0)
99 #define PKT_RX_RSS_HASH (1ULL << 1) /**< RX packet with RSS hash result. */
100 #define PKT_RX_FDIR (1ULL << 2) /**< RX packet with FDIR match indicate. */
104 * Checking this flag alone is deprecated: check the 2 bits of
105 * PKT_RX_L4_CKSUM_MASK.
106 * This flag was set when the L4 checksum of a packet was detected as
107 * wrong by the hardware.
109 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
113 * Checking this flag alone is deprecated: check the 2 bits of
114 * PKT_RX_IP_CKSUM_MASK.
115 * This flag was set when the IP checksum of a packet was detected as
116 * wrong by the hardware.
118 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
120 #define PKT_RX_EIP_CKSUM_BAD (1ULL << 5) /**< External IP header checksum error. */
123 * A vlan has been stripped by the hardware and its tci is saved in
124 * mbuf->vlan_tci. This can only happen if vlan stripping is enabled
125 * in the RX configuration of the PMD.
127 #define PKT_RX_VLAN_STRIPPED (1ULL << 6)
130 * Mask of bits used to determine the status of RX IP checksum.
131 * - PKT_RX_IP_CKSUM_UNKNOWN: no information about the RX IP checksum
132 * - PKT_RX_IP_CKSUM_BAD: the IP checksum in the packet is wrong
133 * - PKT_RX_IP_CKSUM_GOOD: the IP checksum in the packet is valid
134 * - PKT_RX_IP_CKSUM_NONE: the IP checksum is not correct in the packet
135 * data, but the integrity of the IP header is verified.
137 #define PKT_RX_IP_CKSUM_MASK ((1ULL << 4) | (1ULL << 7))
139 #define PKT_RX_IP_CKSUM_UNKNOWN 0
140 #define PKT_RX_IP_CKSUM_BAD (1ULL << 4)
141 #define PKT_RX_IP_CKSUM_GOOD (1ULL << 7)
142 #define PKT_RX_IP_CKSUM_NONE ((1ULL << 4) | (1ULL << 7))
145 * Mask of bits used to determine the status of RX L4 checksum.
146 * - PKT_RX_L4_CKSUM_UNKNOWN: no information about the RX L4 checksum
147 * - PKT_RX_L4_CKSUM_BAD: the L4 checksum in the packet is wrong
148 * - PKT_RX_L4_CKSUM_GOOD: the L4 checksum in the packet is valid
149 * - PKT_RX_L4_CKSUM_NONE: the L4 checksum is not correct in the packet
150 * data, but the integrity of the L4 data is verified.
152 #define PKT_RX_L4_CKSUM_MASK ((1ULL << 3) | (1ULL << 8))
154 #define PKT_RX_L4_CKSUM_UNKNOWN 0
155 #define PKT_RX_L4_CKSUM_BAD (1ULL << 3)
156 #define PKT_RX_L4_CKSUM_GOOD (1ULL << 8)
157 #define PKT_RX_L4_CKSUM_NONE ((1ULL << 3) | (1ULL << 8))
159 #define PKT_RX_IEEE1588_PTP (1ULL << 9) /**< RX IEEE1588 L2 Ethernet PT Packet. */
160 #define PKT_RX_IEEE1588_TMST (1ULL << 10) /**< RX IEEE1588 L2/L4 timestamped packet.*/
161 #define PKT_RX_FDIR_ID (1ULL << 13) /**< FD id reported if FDIR match. */
162 #define PKT_RX_FDIR_FLX (1ULL << 14) /**< Flexible bytes reported if FDIR match. */
165 * The 2 vlans have been stripped by the hardware and their tci are
166 * saved in mbuf->vlan_tci (inner) and mbuf->vlan_tci_outer (outer).
167 * This can only happen if vlan stripping is enabled in the RX
168 * configuration of the PMD. If this flag is set, PKT_RX_VLAN_STRIPPED
171 #define PKT_RX_QINQ_STRIPPED (1ULL << 15)
175 * RX packet with double VLAN stripped.
176 * This flag is replaced by PKT_RX_QINQ_STRIPPED.
178 #define PKT_RX_QINQ_PKT PKT_RX_QINQ_STRIPPED
181 * When packets are coalesced by a hardware or virtual driver, this flag
182 * can be set in the RX mbuf, meaning that the m->tso_segsz field is
183 * valid and is set to the segment size of original packets.
185 #define PKT_RX_LRO (1ULL << 16)
188 * Indicate that the timestamp field in the mbuf is valid.
190 #define PKT_RX_TIMESTAMP (1ULL << 17)
192 /* add new RX flags here */
194 /* add new TX flags here */
197 * Offload the MACsec. This flag must be set by the application to enable
198 * this offload feature for a packet to be transmitted.
200 #define PKT_TX_MACSEC (1ULL << 44)
203 * Bits 45:48 used for the tunnel type.
204 * When doing Tx offload like TSO or checksum, the HW needs to configure the
205 * tunnel type into the HW descriptors.
207 #define PKT_TX_TUNNEL_VXLAN (0x1ULL << 45)
208 #define PKT_TX_TUNNEL_GRE (0x2ULL << 45)
209 #define PKT_TX_TUNNEL_IPIP (0x3ULL << 45)
210 #define PKT_TX_TUNNEL_GENEVE (0x4ULL << 45)
211 /* add new TX TUNNEL type here */
212 #define PKT_TX_TUNNEL_MASK (0xFULL << 45)
215 * Second VLAN insertion (QinQ) flag.
217 #define PKT_TX_QINQ_PKT (1ULL << 49) /**< TX packet with double VLAN inserted. */
220 * TCP segmentation offload. To enable this offload feature for a
221 * packet to be transmitted on hardware supporting TSO:
222 * - set the PKT_TX_TCP_SEG flag in mbuf->ol_flags (this flag implies
224 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
225 * - if it's IPv4, set the PKT_TX_IP_CKSUM flag and write the IP checksum
227 * - fill the mbuf offload information: l2_len, l3_len, l4_len, tso_segsz
228 * - calculate the pseudo header checksum without taking ip_len in account,
229 * and set it in the TCP header. Refer to rte_ipv4_phdr_cksum() and
230 * rte_ipv6_phdr_cksum() that can be used as helpers.
232 #define PKT_TX_TCP_SEG (1ULL << 50)
234 #define PKT_TX_IEEE1588_TMST (1ULL << 51) /**< TX IEEE1588 packet to timestamp. */
237 * Bits 52+53 used for L4 packet type with checksum enabled: 00: Reserved,
238 * 01: TCP checksum, 10: SCTP checksum, 11: UDP checksum. To use hardware
239 * L4 checksum offload, the user needs to:
240 * - fill l2_len and l3_len in mbuf
241 * - set the flags PKT_TX_TCP_CKSUM, PKT_TX_SCTP_CKSUM or PKT_TX_UDP_CKSUM
242 * - set the flag PKT_TX_IPV4 or PKT_TX_IPV6
243 * - calculate the pseudo header checksum and set it in the L4 header (only
244 * for TCP or UDP). See rte_ipv4_phdr_cksum() and rte_ipv6_phdr_cksum().
245 * For SCTP, set the crc field to 0.
247 #define PKT_TX_L4_NO_CKSUM (0ULL << 52) /**< Disable L4 cksum of TX pkt. */
248 #define PKT_TX_TCP_CKSUM (1ULL << 52) /**< TCP cksum of TX pkt. computed by NIC. */
249 #define PKT_TX_SCTP_CKSUM (2ULL << 52) /**< SCTP cksum of TX pkt. computed by NIC. */
250 #define PKT_TX_UDP_CKSUM (3ULL << 52) /**< UDP cksum of TX pkt. computed by NIC. */
251 #define PKT_TX_L4_MASK (3ULL << 52) /**< Mask for L4 cksum offload request. */
254 * Offload the IP checksum in the hardware. The flag PKT_TX_IPV4 should
255 * also be set by the application, although a PMD will only check
257 * - set the IP checksum field in the packet to 0
258 * - fill the mbuf offload information: l2_len, l3_len
260 #define PKT_TX_IP_CKSUM (1ULL << 54)
263 * Packet is IPv4. This flag must be set when using any offload feature
264 * (TSO, L3 or L4 checksum) to tell the NIC that the packet is an IPv4
265 * packet. If the packet is a tunneled packet, this flag is related to
268 #define PKT_TX_IPV4 (1ULL << 55)
271 * Packet is IPv6. This flag must be set when using an offload feature
272 * (TSO or L4 checksum) to tell the NIC that the packet is an IPv6
273 * packet. If the packet is a tunneled packet, this flag is related to
276 #define PKT_TX_IPV6 (1ULL << 56)
278 #define PKT_TX_VLAN_PKT (1ULL << 57) /**< TX packet is a 802.1q VLAN packet. */
281 * Offload the IP checksum of an external header in the hardware. The
282 * flag PKT_TX_OUTER_IPV4 should also be set by the application, alto ugh
283 * a PMD will only check PKT_TX_IP_CKSUM. The IP checksum field in the
284 * packet must be set to 0.
285 * - set the outer IP checksum field in the packet to 0
286 * - fill the mbuf offload information: outer_l2_len, outer_l3_len
288 #define PKT_TX_OUTER_IP_CKSUM (1ULL << 58)
291 * Packet outer header is IPv4. This flag must be set when using any
292 * outer offload feature (L3 or L4 checksum) to tell the NIC that the
293 * outer header of the tunneled packet is an IPv4 packet.
295 #define PKT_TX_OUTER_IPV4 (1ULL << 59)
298 * Packet outer header is IPv6. This flag must be set when using any
299 * outer offload feature (L4 checksum) to tell the NIC that the outer
300 * header of the tunneled packet is an IPv6 packet.
302 #define PKT_TX_OUTER_IPV6 (1ULL << 60)
305 * Bitmask of all supported packet Tx offload features flags,
306 * which can be set for packet.
308 #define PKT_TX_OFFLOAD_MASK ( \
311 PKT_TX_OUTER_IP_CKSUM | \
313 PKT_TX_IEEE1588_TMST | \
316 PKT_TX_TUNNEL_MASK | \
319 #define __RESERVED (1ULL << 61) /**< reserved for future mbuf use */
321 #define IND_ATTACHED_MBUF (1ULL << 62) /**< Indirect attached mbuf */
323 /* Use final bit of flags to indicate a control mbuf */
324 #define CTRL_MBUF_FLAG (1ULL << 63) /**< Mbuf contains control data */
326 /** Alignment constraint of mbuf private area. */
327 #define RTE_MBUF_PRIV_ALIGN 8
330 * Get the name of a RX offload flag
333 * The mask describing the flag.
335 * The name of this flag, or NULL if it's not a valid RX flag.
337 const char *rte_get_rx_ol_flag_name(uint64_t mask);
340 * Dump the list of RX offload flags in a buffer
343 * The mask describing the RX flags.
347 * The length of the buffer.
349 * 0 on success, (-1) on error.
351 int rte_get_rx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
354 * Get the name of a TX offload flag
357 * The mask describing the flag. Usually only one bit must be set.
358 * Several bits can be given if they belong to the same mask.
359 * Ex: PKT_TX_L4_MASK.
361 * The name of this flag, or NULL if it's not a valid TX flag.
363 const char *rte_get_tx_ol_flag_name(uint64_t mask);
366 * Dump the list of TX offload flags in a buffer
369 * The mask describing the TX flags.
373 * The length of the buffer.
375 * 0 on success, (-1) on error.
377 int rte_get_tx_ol_flag_list(uint64_t mask, char *buf, size_t buflen);
380 * Some NICs need at least 2KB buffer to RX standard Ethernet frame without
381 * splitting it into multiple segments.
382 * So, for mbufs that planned to be involved into RX/TX, the recommended
383 * minimal buffer length is 2KB + RTE_PKTMBUF_HEADROOM.
385 #define RTE_MBUF_DEFAULT_DATAROOM 2048
386 #define RTE_MBUF_DEFAULT_BUF_SIZE \
387 (RTE_MBUF_DEFAULT_DATAROOM + RTE_PKTMBUF_HEADROOM)
389 /* define a set of marker types that can be used to refer to set points in the
392 typedef void *MARKER[0]; /**< generic marker for a point in a structure */
394 typedef uint8_t MARKER8[0]; /**< generic marker with 1B alignment */
396 typedef uint64_t MARKER64[0]; /**< marker that allows us to overwrite 8 bytes
397 * with a single assignment */
400 * The generic rte_mbuf, containing a packet mbuf.
405 void *buf_addr; /**< Virtual address of segment buffer. */
406 phys_addr_t buf_physaddr; /**< Physical address of segment buffer. */
408 /* next 8 bytes are initialised on RX descriptor rearm */
413 * Reference counter. Its size should at least equal to the size
414 * of port field (16 bits), to support zero-copy broadcast.
415 * It should only be accessed using the following functions:
416 * rte_mbuf_refcnt_update(), rte_mbuf_refcnt_read(), and
417 * rte_mbuf_refcnt_set(). The functionality of these functions (atomic,
418 * or non-atomic) is controlled by the CONFIG_RTE_MBUF_REFCNT_ATOMIC
423 rte_atomic16_t refcnt_atomic; /**< Atomically accessed refcnt */
424 uint16_t refcnt; /**< Non-atomically accessed refcnt */
426 uint16_t nb_segs; /**< Number of segments. */
428 /** Input port (16 bits to support more than 256 virtual ports). */
431 uint64_t ol_flags; /**< Offload features. */
433 /* remaining bytes are set on RX when pulling packet from descriptor */
434 MARKER rx_descriptor_fields1;
437 * The packet type, which is the combination of outer/inner L2, L3, L4
438 * and tunnel types. The packet_type is about data really present in the
439 * mbuf. Example: if vlan stripping is enabled, a received vlan packet
440 * would have RTE_PTYPE_L2_ETHER and not RTE_PTYPE_L2_VLAN because the
441 * vlan is stripped from the data.
445 uint32_t packet_type; /**< L2/L3/L4 and tunnel information. */
447 uint32_t l2_type:4; /**< (Outer) L2 type. */
448 uint32_t l3_type:4; /**< (Outer) L3 type. */
449 uint32_t l4_type:4; /**< (Outer) L4 type. */
450 uint32_t tun_type:4; /**< Tunnel type. */
451 uint32_t inner_l2_type:4; /**< Inner L2 type. */
452 uint32_t inner_l3_type:4; /**< Inner L3 type. */
453 uint32_t inner_l4_type:4; /**< Inner L4 type. */
457 uint32_t pkt_len; /**< Total pkt len: sum of all segments. */
458 uint16_t data_len; /**< Amount of data in segment buffer. */
459 /** VLAN TCI (CPU order), valid if PKT_RX_VLAN_STRIPPED is set. */
463 uint32_t rss; /**< RSS hash result if RSS enabled */
472 /**< Second 4 flexible bytes */
475 /**< First 4 flexible bytes or FD ID, dependent on
476 PKT_RX_FDIR_* flag in ol_flags. */
477 } fdir; /**< Filter identifier if FDIR enabled */
481 } sched; /**< Hierarchical scheduler */
482 uint32_t usr; /**< User defined tags. See rte_distributor_process() */
483 } hash; /**< hash information */
485 /** Outer VLAN TCI (CPU order), valid if PKT_RX_QINQ_STRIPPED is set. */
486 uint16_t vlan_tci_outer;
488 uint16_t buf_len; /**< Length of segment buffer. */
490 /** Valid if PKT_RX_TIMESTAMP is set. The unit and time reference
491 * are not normalized but are always the same for a given port.
495 /* second cache line - fields only used in slow path or on TX */
496 MARKER cacheline1 __rte_cache_min_aligned;
500 void *userdata; /**< Can be used for external metadata */
501 uint64_t udata64; /**< Allow 8-byte userdata on 32-bit */
504 struct rte_mempool *pool; /**< Pool from which mbuf was allocated. */
505 struct rte_mbuf *next; /**< Next segment of scattered packet. */
507 /* fields to support TX offloads */
510 uint64_t tx_offload; /**< combined for easy fetch */
514 /**< L2 (MAC) Header Length for non-tunneling pkt.
515 * Outer_L4_len + ... + Inner_L2_len for tunneling pkt.
517 uint64_t l3_len:9; /**< L3 (IP) Header Length. */
518 uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
519 uint64_t tso_segsz:16; /**< TCP TSO segment size */
521 /* fields for TX offloading of tunnels */
522 uint64_t outer_l3_len:9; /**< Outer L3 (IP) Hdr Length. */
523 uint64_t outer_l2_len:7; /**< Outer L2 (MAC) Hdr Length. */
525 /* uint64_t unused:8; */
529 /** Size of the application private data. In case of an indirect
530 * mbuf, it stores the direct mbuf private data size. */
533 /** Timesync flags for use with IEEE1588. */
536 /** Sequence number. See also rte_reorder_insert(). */
539 } __rte_cache_aligned;
542 * Prefetch the first part of the mbuf
544 * The first 64 bytes of the mbuf corresponds to fields that are used early
545 * in the receive path. If the cache line of the architecture is higher than
546 * 64B, the second part will also be prefetched.
549 * The pointer to the mbuf.
552 rte_mbuf_prefetch_part1(struct rte_mbuf *m)
554 rte_prefetch0(&m->cacheline0);
558 * Prefetch the second part of the mbuf
560 * The next 64 bytes of the mbuf corresponds to fields that are used in the
561 * transmit path. If the cache line of the architecture is higher than 64B,
562 * this function does nothing as it is expected that the full mbuf is
566 * The pointer to the mbuf.
569 rte_mbuf_prefetch_part2(struct rte_mbuf *m)
571 #if RTE_CACHE_LINE_SIZE == 64
572 rte_prefetch0(&m->cacheline1);
579 static inline uint16_t rte_pktmbuf_priv_size(struct rte_mempool *mp);
582 * Return the DMA address of the beginning of the mbuf data
585 * The pointer to the mbuf.
587 * The physical address of the beginning of the mbuf data
589 static inline phys_addr_t
590 rte_mbuf_data_dma_addr(const struct rte_mbuf *mb)
592 return mb->buf_physaddr + mb->data_off;
596 * Return the default DMA address of the beginning of the mbuf data
598 * This function is used by drivers in their receive function, as it
599 * returns the location where data should be written by the NIC, taking
600 * the default headroom in account.
603 * The pointer to the mbuf.
605 * The physical address of the beginning of the mbuf data
607 static inline phys_addr_t
608 rte_mbuf_data_dma_addr_default(const struct rte_mbuf *mb)
610 return mb->buf_physaddr + RTE_PKTMBUF_HEADROOM;
614 * Return the mbuf owning the data buffer address of an indirect mbuf.
617 * The pointer to the indirect mbuf.
619 * The address of the direct mbuf corresponding to buffer_addr.
621 static inline struct rte_mbuf *
622 rte_mbuf_from_indirect(struct rte_mbuf *mi)
624 return (struct rte_mbuf *)RTE_PTR_SUB(mi->buf_addr, sizeof(*mi) + mi->priv_size);
628 * Return the buffer address embedded in the given mbuf.
631 * The pointer to the mbuf.
633 * The address of the data buffer owned by the mbuf.
636 rte_mbuf_to_baddr(struct rte_mbuf *md)
639 buffer_addr = (char *)md + sizeof(*md) + rte_pktmbuf_priv_size(md->pool);
644 * Returns TRUE if given mbuf is indirect, or FALSE otherwise.
646 #define RTE_MBUF_INDIRECT(mb) ((mb)->ol_flags & IND_ATTACHED_MBUF)
649 * Returns TRUE if given mbuf is direct, or FALSE otherwise.
651 #define RTE_MBUF_DIRECT(mb) (!RTE_MBUF_INDIRECT(mb))
654 * Private data in case of pktmbuf pool.
656 * A structure that contains some pktmbuf_pool-specific data that are
657 * appended after the mempool structure (in private data).
659 struct rte_pktmbuf_pool_private {
660 uint16_t mbuf_data_room_size; /**< Size of data space in each mbuf. */
661 uint16_t mbuf_priv_size; /**< Size of private area in each mbuf. */
664 #ifdef RTE_LIBRTE_MBUF_DEBUG
666 /** check mbuf type in debug mode */
667 #define __rte_mbuf_sanity_check(m, is_h) rte_mbuf_sanity_check(m, is_h)
669 #else /* RTE_LIBRTE_MBUF_DEBUG */
671 /** check mbuf type in debug mode */
672 #define __rte_mbuf_sanity_check(m, is_h) do { } while (0)
674 #endif /* RTE_LIBRTE_MBUF_DEBUG */
676 #ifdef RTE_MBUF_REFCNT_ATOMIC
679 * Reads the value of an mbuf's refcnt.
683 * Reference count number.
685 static inline uint16_t
686 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
688 return (uint16_t)(rte_atomic16_read(&m->refcnt_atomic));
692 * Sets an mbuf's refcnt to a defined value.
699 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
701 rte_atomic16_set(&m->refcnt_atomic, new_value);
705 * Adds given value to an mbuf's refcnt and returns its new value.
709 * Value to add/subtract
713 static inline uint16_t
714 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
717 * The atomic_add is an expensive operation, so we don't want to
718 * call it in the case where we know we are the uniq holder of
719 * this mbuf (i.e. ref_cnt == 1). Otherwise, an atomic
720 * operation has to be used because concurrent accesses on the
721 * reference counter can occur.
723 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
724 rte_mbuf_refcnt_set(m, 1 + value);
728 return (uint16_t)(rte_atomic16_add_return(&m->refcnt_atomic, value));
731 #else /* ! RTE_MBUF_REFCNT_ATOMIC */
734 * Adds given value to an mbuf's refcnt and returns its new value.
736 static inline uint16_t
737 rte_mbuf_refcnt_update(struct rte_mbuf *m, int16_t value)
739 m->refcnt = (uint16_t)(m->refcnt + value);
744 * Reads the value of an mbuf's refcnt.
746 static inline uint16_t
747 rte_mbuf_refcnt_read(const struct rte_mbuf *m)
753 * Sets an mbuf's refcnt to the defined value.
756 rte_mbuf_refcnt_set(struct rte_mbuf *m, uint16_t new_value)
758 m->refcnt = new_value;
761 #endif /* RTE_MBUF_REFCNT_ATOMIC */
764 #define RTE_MBUF_PREFETCH_TO_FREE(m) do { \
771 * Sanity checks on an mbuf.
773 * Check the consistency of the given mbuf. The function will cause a
774 * panic if corruption is detected.
777 * The mbuf to be checked.
779 * True if the mbuf is a packet header, false if it is a sub-segment
780 * of a packet (in this case, some fields like nb_segs are not checked)
783 rte_mbuf_sanity_check(const struct rte_mbuf *m, int is_header);
786 * Allocate an unitialized mbuf from mempool *mp*.
788 * This function can be used by PMDs (especially in RX functions) to
789 * allocate an unitialized mbuf. The driver is responsible of
790 * initializing all the required fields. See rte_pktmbuf_reset().
791 * For standard needs, prefer rte_pktmbuf_alloc().
793 * The caller can expect that the following fields of the mbuf structure
794 * are initialized: buf_addr, buf_physaddr, buf_len, refcnt=1, nb_segs=1,
795 * next=NULL, pool, priv_size. The other fields must be initialized
799 * The mempool from which mbuf is allocated.
801 * - The pointer to the new mbuf on success.
802 * - NULL if allocation failed.
804 static inline struct rte_mbuf *rte_mbuf_raw_alloc(struct rte_mempool *mp)
809 if (rte_mempool_get(mp, &mb) < 0)
811 m = (struct rte_mbuf *)mb;
812 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
813 RTE_ASSERT(m->next == NULL);
814 RTE_ASSERT(m->nb_segs == 1);
815 __rte_mbuf_sanity_check(m, 0);
821 * Put mbuf back into its original mempool.
823 * The caller must ensure that the mbuf is direct and properly
824 * reinitialized (refcnt=1, next=NULL, nb_segs=1), as done by
825 * rte_pktmbuf_prefree_seg().
827 * This function should be used with care, when optimization is
828 * required. For standard needs, prefer rte_pktmbuf_free() or
829 * rte_pktmbuf_free_seg().
832 * The mbuf to be freed.
834 static inline void __attribute__((always_inline))
835 rte_mbuf_raw_free(struct rte_mbuf *m)
837 RTE_ASSERT(RTE_MBUF_DIRECT(m));
838 RTE_ASSERT(rte_mbuf_refcnt_read(m) == 1);
839 RTE_ASSERT(m->next == NULL);
840 RTE_ASSERT(m->nb_segs == 1);
841 __rte_mbuf_sanity_check(m, 0);
842 rte_mempool_put(m->pool, m);
845 /* compat with older versions */
848 __rte_mbuf_raw_free(struct rte_mbuf *m)
850 rte_mbuf_raw_free(m);
853 /* Operations on ctrl mbuf */
856 * The control mbuf constructor.
858 * This function initializes some fields in an mbuf structure that are
859 * not modified by the user once created (mbuf type, origin pool, buffer
860 * start address, and so on). This function is given as a callback function
861 * to rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
864 * The mempool from which the mbuf is allocated.
866 * A pointer that can be used by the user to retrieve useful information
867 * for mbuf initialization. This pointer is the opaque argument passed to
868 * rte_mempool_obj_iter() or rte_mempool_create().
870 * The mbuf to initialize.
872 * The index of the mbuf in the pool table.
874 void rte_ctrlmbuf_init(struct rte_mempool *mp, void *opaque_arg,
875 void *m, unsigned i);
878 * Allocate a new mbuf (type is ctrl) from mempool *mp*.
880 * This new mbuf is initialized with data pointing to the beginning of
881 * buffer, and with a length of zero.
884 * The mempool from which the mbuf is allocated.
886 * - The pointer to the new mbuf on success.
887 * - NULL if allocation failed.
889 #define rte_ctrlmbuf_alloc(mp) rte_pktmbuf_alloc(mp)
892 * Free a control mbuf back into its original mempool.
895 * The control mbuf to be freed.
897 #define rte_ctrlmbuf_free(m) rte_pktmbuf_free(m)
900 * A macro that returns the pointer to the carried data.
902 * The value that can be read or assigned.
907 #define rte_ctrlmbuf_data(m) ((char *)((m)->buf_addr) + (m)->data_off)
910 * A macro that returns the length of the carried data.
912 * The value that can be read or assigned.
917 #define rte_ctrlmbuf_len(m) rte_pktmbuf_data_len(m)
920 * Tests if an mbuf is a control mbuf
923 * The mbuf to be tested
925 * - True (1) if the mbuf is a control mbuf
926 * - False(0) otherwise
929 rte_is_ctrlmbuf(struct rte_mbuf *m)
931 return !!(m->ol_flags & CTRL_MBUF_FLAG);
934 /* Operations on pkt mbuf */
937 * The packet mbuf constructor.
939 * This function initializes some fields in the mbuf structure that are
940 * not modified by the user once created (origin pool, buffer start
941 * address, and so on). This function is given as a callback function to
942 * rte_mempool_obj_iter() or rte_mempool_create() at pool creation time.
945 * The mempool from which mbufs originate.
947 * A pointer that can be used by the user to retrieve useful information
948 * for mbuf initialization. This pointer is the opaque argument passed to
949 * rte_mempool_obj_iter() or rte_mempool_create().
951 * The mbuf to initialize.
953 * The index of the mbuf in the pool table.
955 void rte_pktmbuf_init(struct rte_mempool *mp, void *opaque_arg,
956 void *m, unsigned i);
960 * A packet mbuf pool constructor.
962 * This function initializes the mempool private data in the case of a
963 * pktmbuf pool. This private data is needed by the driver. The
964 * function must be called on the mempool before it is used, or it
965 * can be given as a callback function to rte_mempool_create() at
966 * pool creation. It can be extended by the user, for example, to
967 * provide another packet size.
970 * The mempool from which mbufs originate.
972 * A pointer that can be used by the user to retrieve useful information
973 * for mbuf initialization. This pointer is the opaque argument passed to
974 * rte_mempool_create().
976 void rte_pktmbuf_pool_init(struct rte_mempool *mp, void *opaque_arg);
979 * Create a mbuf pool.
981 * This function creates and initializes a packet mbuf pool. It is
982 * a wrapper to rte_mempool functions.
985 * The name of the mbuf pool.
987 * The number of elements in the mbuf pool. The optimum size (in terms
988 * of memory usage) for a mempool is when n is a power of two minus one:
991 * Size of the per-core object cache. See rte_mempool_create() for
994 * Size of application private are between the rte_mbuf structure
995 * and the data buffer. This value must be aligned to RTE_MBUF_PRIV_ALIGN.
996 * @param data_room_size
997 * Size of data buffer in each mbuf, including RTE_PKTMBUF_HEADROOM.
999 * The socket identifier where the memory should be allocated. The
1000 * value can be *SOCKET_ID_ANY* if there is no NUMA constraint for the
1003 * The pointer to the new allocated mempool, on success. NULL on error
1004 * with rte_errno set appropriately. Possible rte_errno values include:
1005 * - E_RTE_NO_CONFIG - function could not get pointer to rte_config structure
1006 * - E_RTE_SECONDARY - function was called from a secondary process instance
1007 * - EINVAL - cache size provided is too large, or priv_size is not aligned.
1008 * - ENOSPC - the maximum number of memzones has already been allocated
1009 * - EEXIST - a memzone with the same name already exists
1010 * - ENOMEM - no appropriate memory area found in which to create memzone
1012 struct rte_mempool *
1013 rte_pktmbuf_pool_create(const char *name, unsigned n,
1014 unsigned cache_size, uint16_t priv_size, uint16_t data_room_size,
1018 * Get the data room size of mbufs stored in a pktmbuf_pool
1020 * The data room size is the amount of data that can be stored in a
1021 * mbuf including the headroom (RTE_PKTMBUF_HEADROOM).
1024 * The packet mbuf pool.
1026 * The data room size of mbufs stored in this mempool.
1028 static inline uint16_t
1029 rte_pktmbuf_data_room_size(struct rte_mempool *mp)
1031 struct rte_pktmbuf_pool_private *mbp_priv;
1033 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1034 return mbp_priv->mbuf_data_room_size;
1038 * Get the application private size of mbufs stored in a pktmbuf_pool
1040 * The private size of mbuf is a zone located between the rte_mbuf
1041 * structure and the data buffer where an application can store data
1042 * associated to a packet.
1045 * The packet mbuf pool.
1047 * The private size of mbufs stored in this mempool.
1049 static inline uint16_t
1050 rte_pktmbuf_priv_size(struct rte_mempool *mp)
1052 struct rte_pktmbuf_pool_private *mbp_priv;
1054 mbp_priv = (struct rte_pktmbuf_pool_private *)rte_mempool_get_priv(mp);
1055 return mbp_priv->mbuf_priv_size;
1059 * Reset the data_off field of a packet mbuf to its default value.
1061 * The given mbuf must have only one segment, which should be empty.
1064 * The packet mbuf's data_off field has to be reset.
1066 static inline void rte_pktmbuf_reset_headroom(struct rte_mbuf *m)
1068 m->data_off = RTE_MIN(RTE_PKTMBUF_HEADROOM, (uint16_t)m->buf_len);
1072 * Reset the fields of a packet mbuf to their default values.
1074 * The given mbuf must have only one segment.
1077 * The packet mbuf to be resetted.
1079 static inline void rte_pktmbuf_reset(struct rte_mbuf *m)
1085 m->vlan_tci_outer = 0;
1091 rte_pktmbuf_reset_headroom(m);
1094 __rte_mbuf_sanity_check(m, 1);
1098 * Allocate a new mbuf from a mempool.
1100 * This new mbuf contains one segment, which has a length of 0. The pointer
1101 * to data is initialized to have some bytes of headroom in the buffer
1102 * (if buffer size allows).
1105 * The mempool from which the mbuf is allocated.
1107 * - The pointer to the new mbuf on success.
1108 * - NULL if allocation failed.
1110 static inline struct rte_mbuf *rte_pktmbuf_alloc(struct rte_mempool *mp)
1113 if ((m = rte_mbuf_raw_alloc(mp)) != NULL)
1114 rte_pktmbuf_reset(m);
1119 * Allocate a bulk of mbufs, initialize refcnt and reset the fields to default
1123 * The mempool from which mbufs are allocated.
1125 * Array of pointers to mbufs
1131 static inline int rte_pktmbuf_alloc_bulk(struct rte_mempool *pool,
1132 struct rte_mbuf **mbufs, unsigned count)
1137 rc = rte_mempool_get_bulk(pool, (void **)mbufs, count);
1141 /* To understand duff's device on loop unwinding optimization, see
1142 * https://en.wikipedia.org/wiki/Duff's_device.
1143 * Here while() loop is used rather than do() while{} to avoid extra
1144 * check if count is zero.
1146 switch (count % 4) {
1148 while (idx != count) {
1149 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1150 rte_mbuf_refcnt_set(mbufs[idx], 1);
1151 rte_pktmbuf_reset(mbufs[idx]);
1154 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1155 rte_mbuf_refcnt_set(mbufs[idx], 1);
1156 rte_pktmbuf_reset(mbufs[idx]);
1159 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1160 rte_mbuf_refcnt_set(mbufs[idx], 1);
1161 rte_pktmbuf_reset(mbufs[idx]);
1164 RTE_ASSERT(rte_mbuf_refcnt_read(mbufs[idx]) == 0);
1165 rte_mbuf_refcnt_set(mbufs[idx], 1);
1166 rte_pktmbuf_reset(mbufs[idx]);
1174 * Attach packet mbuf to another packet mbuf.
1176 * After attachment we refer the mbuf we attached as 'indirect',
1177 * while mbuf we attached to as 'direct'.
1178 * The direct mbuf's reference counter is incremented.
1180 * Right now, not supported:
1181 * - attachment for already indirect mbuf (e.g. - mi has to be direct).
1182 * - mbuf we trying to attach (mi) is used by someone else
1183 * e.g. it's reference counter is greater then 1.
1186 * The indirect packet mbuf.
1188 * The packet mbuf we're attaching to.
1190 static inline void rte_pktmbuf_attach(struct rte_mbuf *mi, struct rte_mbuf *m)
1192 struct rte_mbuf *md;
1194 RTE_ASSERT(RTE_MBUF_DIRECT(mi) &&
1195 rte_mbuf_refcnt_read(mi) == 1);
1197 /* if m is not direct, get the mbuf that embeds the data */
1198 if (RTE_MBUF_DIRECT(m))
1201 md = rte_mbuf_from_indirect(m);
1203 rte_mbuf_refcnt_update(md, 1);
1204 mi->priv_size = m->priv_size;
1205 mi->buf_physaddr = m->buf_physaddr;
1206 mi->buf_addr = m->buf_addr;
1207 mi->buf_len = m->buf_len;
1209 mi->data_off = m->data_off;
1210 mi->data_len = m->data_len;
1212 mi->vlan_tci = m->vlan_tci;
1213 mi->vlan_tci_outer = m->vlan_tci_outer;
1214 mi->tx_offload = m->tx_offload;
1218 mi->pkt_len = mi->data_len;
1220 mi->ol_flags = m->ol_flags | IND_ATTACHED_MBUF;
1221 mi->packet_type = m->packet_type;
1222 mi->timestamp = m->timestamp;
1224 __rte_mbuf_sanity_check(mi, 1);
1225 __rte_mbuf_sanity_check(m, 0);
1229 * Detach an indirect packet mbuf.
1231 * - restore original mbuf address and length values.
1232 * - reset pktmbuf data and data_len to their default values.
1233 * - decrement the direct mbuf's reference counter. When the
1234 * reference counter becomes 0, the direct mbuf is freed.
1236 * All other fields of the given packet mbuf will be left intact.
1239 * The indirect attached packet mbuf.
1241 static inline void rte_pktmbuf_detach(struct rte_mbuf *m)
1243 struct rte_mbuf *md = rte_mbuf_from_indirect(m);
1244 struct rte_mempool *mp = m->pool;
1245 uint32_t mbuf_size, buf_len, priv_size;
1247 priv_size = rte_pktmbuf_priv_size(mp);
1248 mbuf_size = sizeof(struct rte_mbuf) + priv_size;
1249 buf_len = rte_pktmbuf_data_room_size(mp);
1251 m->priv_size = priv_size;
1252 m->buf_addr = (char *)m + mbuf_size;
1253 m->buf_physaddr = rte_mempool_virt2phy(mp, m) + mbuf_size;
1254 m->buf_len = (uint16_t)buf_len;
1255 rte_pktmbuf_reset_headroom(m);
1259 if (rte_mbuf_refcnt_update(md, -1) == 0) {
1262 rte_mbuf_refcnt_set(md, 1);
1263 rte_mbuf_raw_free(md);
1268 * Decrease reference counter and unlink a mbuf segment
1270 * This function does the same than a free, except that it does not
1271 * return the segment to its pool.
1272 * It decreases the reference counter, and if it reaches 0, it is
1273 * detached from its parent for an indirect mbuf.
1276 * The mbuf to be unlinked
1278 * - (m) if it is the last reference. It can be recycled or freed.
1279 * - (NULL) if the mbuf still has remaining references on it.
1281 __attribute__((always_inline))
1282 static inline struct rte_mbuf *
1283 rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1285 __rte_mbuf_sanity_check(m, 0);
1287 if (likely(rte_mbuf_refcnt_read(m) == 1)) {
1289 if (RTE_MBUF_INDIRECT(m))
1290 rte_pktmbuf_detach(m);
1292 if (m->next != NULL) {
1299 } else if (rte_atomic16_add_return(&m->refcnt_atomic, -1) == 0) {
1302 if (RTE_MBUF_INDIRECT(m))
1303 rte_pktmbuf_detach(m);
1305 if (m->next != NULL) {
1309 rte_mbuf_refcnt_set(m, 1);
1316 /* deprecated, replaced by rte_pktmbuf_prefree_seg() */
1318 static inline struct rte_mbuf *
1319 __rte_pktmbuf_prefree_seg(struct rte_mbuf *m)
1321 return rte_pktmbuf_prefree_seg(m);
1325 * Free a segment of a packet mbuf into its original mempool.
1327 * Free an mbuf, without parsing other segments in case of chained
1331 * The packet mbuf segment to be freed.
1333 static inline void __attribute__((always_inline))
1334 rte_pktmbuf_free_seg(struct rte_mbuf *m)
1336 m = rte_pktmbuf_prefree_seg(m);
1337 if (likely(m != NULL))
1338 rte_mbuf_raw_free(m);
1342 * Free a packet mbuf back into its original mempool.
1344 * Free an mbuf, and all its segments in case of chained buffers. Each
1345 * segment is added back into its original mempool.
1348 * The packet mbuf to be freed.
1350 static inline void rte_pktmbuf_free(struct rte_mbuf *m)
1352 struct rte_mbuf *m_next;
1354 __rte_mbuf_sanity_check(m, 1);
1358 rte_pktmbuf_free_seg(m);
1364 * Creates a "clone" of the given packet mbuf.
1366 * Walks through all segments of the given packet mbuf, and for each of them:
1367 * - Creates a new packet mbuf from the given pool.
1368 * - Attaches newly created mbuf to the segment.
1369 * Then updates pkt_len and nb_segs of the "clone" packet mbuf to match values
1370 * from the original packet mbuf.
1373 * The packet mbuf to be cloned.
1375 * The mempool from which the "clone" mbufs are allocated.
1377 * - The pointer to the new "clone" mbuf on success.
1378 * - NULL if allocation fails.
1380 static inline struct rte_mbuf *rte_pktmbuf_clone(struct rte_mbuf *md,
1381 struct rte_mempool *mp)
1383 struct rte_mbuf *mc, *mi, **prev;
1387 if (unlikely ((mc = rte_pktmbuf_alloc(mp)) == NULL))
1392 pktlen = md->pkt_len;
1397 rte_pktmbuf_attach(mi, md);
1400 } while ((md = md->next) != NULL &&
1401 (mi = rte_pktmbuf_alloc(mp)) != NULL);
1405 mc->pkt_len = pktlen;
1407 /* Allocation of new indirect segment failed */
1408 if (unlikely (mi == NULL)) {
1409 rte_pktmbuf_free(mc);
1413 __rte_mbuf_sanity_check(mc, 1);
1418 * Adds given value to the refcnt of all packet mbuf segments.
1420 * Walks through all segments of given packet mbuf and for each of them
1421 * invokes rte_mbuf_refcnt_update().
1424 * The packet mbuf whose refcnt to be updated.
1426 * The value to add to the mbuf's segments refcnt.
1428 static inline void rte_pktmbuf_refcnt_update(struct rte_mbuf *m, int16_t v)
1430 __rte_mbuf_sanity_check(m, 1);
1433 rte_mbuf_refcnt_update(m, v);
1434 } while ((m = m->next) != NULL);
1438 * Get the headroom in a packet mbuf.
1443 * The length of the headroom.
1445 static inline uint16_t rte_pktmbuf_headroom(const struct rte_mbuf *m)
1447 __rte_mbuf_sanity_check(m, 1);
1452 * Get the tailroom of a packet mbuf.
1457 * The length of the tailroom.
1459 static inline uint16_t rte_pktmbuf_tailroom(const struct rte_mbuf *m)
1461 __rte_mbuf_sanity_check(m, 1);
1462 return (uint16_t)(m->buf_len - rte_pktmbuf_headroom(m) -
1467 * Get the last segment of the packet.
1472 * The last segment of the given mbuf.
1474 static inline struct rte_mbuf *rte_pktmbuf_lastseg(struct rte_mbuf *m)
1476 struct rte_mbuf *m2 = (struct rte_mbuf *)m;
1478 __rte_mbuf_sanity_check(m, 1);
1479 while (m2->next != NULL)
1485 * A macro that points to an offset into the data in the mbuf.
1487 * The returned pointer is cast to type t. Before using this
1488 * function, the user must ensure that the first segment is large
1489 * enough to accommodate its data.
1494 * The offset into the mbuf data.
1496 * The type to cast the result into.
1498 #define rte_pktmbuf_mtod_offset(m, t, o) \
1499 ((t)((char *)(m)->buf_addr + (m)->data_off + (o)))
1502 * A macro that points to the start of the data in the mbuf.
1504 * The returned pointer is cast to type t. Before using this
1505 * function, the user must ensure that the first segment is large
1506 * enough to accommodate its data.
1511 * The type to cast the result into.
1513 #define rte_pktmbuf_mtod(m, t) rte_pktmbuf_mtod_offset(m, t, 0)
1516 * A macro that returns the physical address that points to an offset of the
1517 * start of the data in the mbuf
1522 * The offset into the data to calculate address from.
1524 #define rte_pktmbuf_mtophys_offset(m, o) \
1525 (phys_addr_t)((m)->buf_physaddr + (m)->data_off + (o))
1528 * A macro that returns the physical address that points to the start of the
1534 #define rte_pktmbuf_mtophys(m) rte_pktmbuf_mtophys_offset(m, 0)
1537 * A macro that returns the length of the packet.
1539 * The value can be read or assigned.
1544 #define rte_pktmbuf_pkt_len(m) ((m)->pkt_len)
1547 * A macro that returns the length of the segment.
1549 * The value can be read or assigned.
1554 #define rte_pktmbuf_data_len(m) ((m)->data_len)
1557 * Prepend len bytes to an mbuf data area.
1559 * Returns a pointer to the new
1560 * data start address. If there is not enough headroom in the first
1561 * segment, the function will return NULL, without modifying the mbuf.
1566 * The amount of data to prepend (in bytes).
1568 * A pointer to the start of the newly prepended data, or
1569 * NULL if there is not enough headroom space in the first segment
1571 static inline char *rte_pktmbuf_prepend(struct rte_mbuf *m,
1574 __rte_mbuf_sanity_check(m, 1);
1576 if (unlikely(len > rte_pktmbuf_headroom(m)))
1580 m->data_len = (uint16_t)(m->data_len + len);
1581 m->pkt_len = (m->pkt_len + len);
1583 return (char *)m->buf_addr + m->data_off;
1587 * Append len bytes to an mbuf.
1589 * Append len bytes to an mbuf and return a pointer to the start address
1590 * of the added data. If there is not enough tailroom in the last
1591 * segment, the function will return NULL, without modifying the mbuf.
1596 * The amount of data to append (in bytes).
1598 * A pointer to the start of the newly appended data, or
1599 * NULL if there is not enough tailroom space in the last segment
1601 static inline char *rte_pktmbuf_append(struct rte_mbuf *m, uint16_t len)
1604 struct rte_mbuf *m_last;
1606 __rte_mbuf_sanity_check(m, 1);
1608 m_last = rte_pktmbuf_lastseg(m);
1609 if (unlikely(len > rte_pktmbuf_tailroom(m_last)))
1612 tail = (char *)m_last->buf_addr + m_last->data_off + m_last->data_len;
1613 m_last->data_len = (uint16_t)(m_last->data_len + len);
1614 m->pkt_len = (m->pkt_len + len);
1615 return (char*) tail;
1619 * Remove len bytes at the beginning of an mbuf.
1621 * Returns a pointer to the start address of the new data area. If the
1622 * length is greater than the length of the first segment, then the
1623 * function will fail and return NULL, without modifying the mbuf.
1628 * The amount of data to remove (in bytes).
1630 * A pointer to the new start of the data.
1632 static inline char *rte_pktmbuf_adj(struct rte_mbuf *m, uint16_t len)
1634 __rte_mbuf_sanity_check(m, 1);
1636 if (unlikely(len > m->data_len))
1639 m->data_len = (uint16_t)(m->data_len - len);
1641 m->pkt_len = (m->pkt_len - len);
1642 return (char *)m->buf_addr + m->data_off;
1646 * Remove len bytes of data at the end of the mbuf.
1648 * If the length is greater than the length of the last segment, the
1649 * function will fail and return -1 without modifying the mbuf.
1654 * The amount of data to remove (in bytes).
1659 static inline int rte_pktmbuf_trim(struct rte_mbuf *m, uint16_t len)
1661 struct rte_mbuf *m_last;
1663 __rte_mbuf_sanity_check(m, 1);
1665 m_last = rte_pktmbuf_lastseg(m);
1666 if (unlikely(len > m_last->data_len))
1669 m_last->data_len = (uint16_t)(m_last->data_len - len);
1670 m->pkt_len = (m->pkt_len - len);
1675 * Test if mbuf data is contiguous.
1680 * - 1, if all data is contiguous (one segment).
1681 * - 0, if there is several segments.
1683 static inline int rte_pktmbuf_is_contiguous(const struct rte_mbuf *m)
1685 __rte_mbuf_sanity_check(m, 1);
1686 return !!(m->nb_segs == 1);
1690 * @internal used by rte_pktmbuf_read().
1692 const void *__rte_pktmbuf_read(const struct rte_mbuf *m, uint32_t off,
1693 uint32_t len, void *buf);
1696 * Read len data bytes in a mbuf at specified offset.
1698 * If the data is contiguous, return the pointer in the mbuf data, else
1699 * copy the data in the buffer provided by the user and return its
1703 * The pointer to the mbuf.
1705 * The offset of the data in the mbuf.
1707 * The amount of bytes to read.
1709 * The buffer where data is copied if it is not contigous in mbuf
1710 * data. Its length should be at least equal to the len parameter.
1712 * The pointer to the data, either in the mbuf if it is contiguous,
1713 * or in the user buffer. If mbuf is too small, NULL is returned.
1715 static inline const void *rte_pktmbuf_read(const struct rte_mbuf *m,
1716 uint32_t off, uint32_t len, void *buf)
1718 if (likely(off + len <= rte_pktmbuf_data_len(m)))
1719 return rte_pktmbuf_mtod_offset(m, char *, off);
1721 return __rte_pktmbuf_read(m, off, len, buf);
1725 * Chain an mbuf to another, thereby creating a segmented packet.
1727 * Note: The implementation will do a linear walk over the segments to find
1728 * the tail entry. For cases when there are many segments, it's better to
1729 * chain the entries manually.
1732 * The head of the mbuf chain (the first packet)
1734 * The mbuf to put last in the chain
1738 * - -EOVERFLOW, if the chain is full (256 entries)
1740 static inline int rte_pktmbuf_chain(struct rte_mbuf *head, struct rte_mbuf *tail)
1742 struct rte_mbuf *cur_tail;
1744 /* Check for number-of-segments-overflow */
1745 if (head->nb_segs + tail->nb_segs >= 1 << (sizeof(head->nb_segs) * 8))
1748 /* Chain 'tail' onto the old tail */
1749 cur_tail = rte_pktmbuf_lastseg(head);
1750 cur_tail->next = tail;
1752 /* accumulate number of segments and total length. */
1753 head->nb_segs = (uint8_t)(head->nb_segs + tail->nb_segs);
1754 head->pkt_len += tail->pkt_len;
1756 /* pkt_len is only set in the head */
1757 tail->pkt_len = tail->data_len;
1763 * Validate general requirements for Tx offload in mbuf.
1765 * This function checks correctness and completeness of Tx offload settings.
1768 * The packet mbuf to be validated.
1770 * 0 if packet is valid
1773 rte_validate_tx_offload(const struct rte_mbuf *m)
1775 uint64_t ol_flags = m->ol_flags;
1776 uint64_t inner_l3_offset = m->l2_len;
1778 /* Does packet set any of available offloads? */
1779 if (!(ol_flags & PKT_TX_OFFLOAD_MASK))
1782 if (ol_flags & PKT_TX_OUTER_IP_CKSUM)
1783 inner_l3_offset += m->outer_l2_len + m->outer_l3_len;
1785 /* Headers are fragmented */
1786 if (rte_pktmbuf_data_len(m) < inner_l3_offset + m->l3_len + m->l4_len)
1789 /* IP checksum can be counted only for IPv4 packet */
1790 if ((ol_flags & PKT_TX_IP_CKSUM) && (ol_flags & PKT_TX_IPV6))
1793 /* IP type not set when required */
1794 if (ol_flags & (PKT_TX_L4_MASK | PKT_TX_TCP_SEG))
1795 if (!(ol_flags & (PKT_TX_IPV4 | PKT_TX_IPV6)))
1798 /* Check requirements for TSO packet */
1799 if (ol_flags & PKT_TX_TCP_SEG)
1800 if ((m->tso_segsz == 0) ||
1801 ((ol_flags & PKT_TX_IPV4) &&
1802 !(ol_flags & PKT_TX_IP_CKSUM)))
1805 /* PKT_TX_OUTER_IP_CKSUM set for non outer IPv4 packet. */
1806 if ((ol_flags & PKT_TX_OUTER_IP_CKSUM) &&
1807 !(ol_flags & PKT_TX_OUTER_IPV4))
1814 * Linearize data in mbuf.
1816 * This function moves the mbuf data in the first segment if there is enough
1817 * tailroom. The subsequent segments are unchained and freed.
1826 rte_pktmbuf_linearize(struct rte_mbuf *mbuf)
1828 int seg_len, copy_len;
1830 struct rte_mbuf *m_next;
1833 if (rte_pktmbuf_is_contiguous(mbuf))
1836 /* Extend first segment to the total packet length */
1837 copy_len = rte_pktmbuf_pkt_len(mbuf) - rte_pktmbuf_data_len(mbuf);
1839 if (unlikely(copy_len > rte_pktmbuf_tailroom(mbuf)))
1842 buffer = rte_pktmbuf_mtod_offset(mbuf, char *, mbuf->data_len);
1843 mbuf->data_len = (uint16_t)(mbuf->pkt_len);
1845 /* Append data from next segments to the first one */
1850 seg_len = rte_pktmbuf_data_len(m);
1851 rte_memcpy(buffer, rte_pktmbuf_mtod(m, char *), seg_len);
1854 rte_pktmbuf_free_seg(m);
1865 * Dump an mbuf structure to a file.
1867 * Dump all fields for the given packet mbuf and all its associated
1868 * segments (in the case of a chained buffer).
1871 * A pointer to a file for output
1875 * If dump_len != 0, also dump the "dump_len" first data bytes of
1878 void rte_pktmbuf_dump(FILE *f, const struct rte_mbuf *m, unsigned dump_len);
1884 #endif /* _RTE_MBUF_H_ */