|
ADC Home > Reference Library > Reference > Darwin > Kernel Framework Reference
|
kpi_mbuf.h |
| Includes: |
<sys/kernel_types.h> <mach/vm_types.h> |
This header defines an API for interacting with mbufs. mbufs are the primary method of storing packets in the networking stack.
mbufs are used to store various items in the networking stack. The most common usage of an mbuf is to store a packet or data on a socket waiting to be sent or received. The mbuf is a contiguous structure with some header followed by some data. To store more data than would fit in an mbuf, external data is used. Most mbufs with external data use clusters to store the external data.
mbufs can be chained, contiguous data in a packet can be found by following the m_next chain. Packets may be bundled together using m_nextpacket. Many parts of the stack do not properly handle chains of packets. When in doubt, don't chain packets.
mbuf_adj |
void mbuf_adj( mbuf_t mbuf, int len);
mbufThe mbuf chain to trim.
lenThe number of bytes to trim from the mbuf chain.
Trims len bytes from the mbuf. If the length is greater than zero, the bytes are trimmed from the front of the mbuf. If the length is less than zero, the bytes are trimmed from the end of the mbuf chain.
mbuf_adjustlen |
errno_t mbuf_adjustlen( mbuf_t mbuf, int amount);
mbufThe mbuf to adjust.
amountThe number of bytes increment the length by.
0 upon success otherwise the errno error.
Adds amount to the mbuf len. Verifies that the new length is valid (greater than or equal to zero and less than maximum amount of data that may be stored in the mbuf). This function will not adjust the packet header length field or affect any other mbufs in a chain.
mbuf_align_32 |
errno_t mbuf_align_32( mbuf_t mbuf, size_t len);
mbufThe mbuf.
lenThe minimum length of space that should follow the new data location.
0 on success, errno error on failure.
mbuf_align_32 is a replacement for M_ALIGN and MH_ALIGN. mbuf_align_32 will set the data pointer to a location aligned on a four byte boundry with at least 'len' bytes between the data pointer and the end of the data block.
mbuf_alloccluster |
errno_t mbuf_alloccluster( mbuf_how_t how, size_t *size, caddr_t *addr);
howBlocking or non-blocking.
sizePointer to size of requested cluster. Sizes up to 2048 will be rounded up to 2048; sizes greater than 2048 and up to 4096 will be rounded up to 4096. Sizes greater than 4096 will be rounded up to 16384.
addrPointer to the address of the requested cluster.
0 on success or ENOMEM if failure. If the caller requests greater than 4096 bytes and the system is unable to fulfill the request due to the lack of jumbo clusters support based on the configuration, this routine will return ENOTSUP. In this case, the caller is advised to use 4096 bytes or smaller during subseqent requests.
Allocate a cluster that can be later attached to an mbuf by calling mbuf_attachcluster(). The allocated cluster can also be freed (without being attached to an mbuf) by calling mbuf_freecluster(). At the moment this routine will either return a cluster of 2048, 4096 or 16384 bytes depending on the requested size. Note that clusters greater than 4096 bytes might not be available in all configurations; the caller must additionally check for ENOTSUP (see below).
mbuf_allocpacket |
errno_t mbuf_allocpacket( mbuf_how_t how, size_t packetlen, unsigned int *maxchunks, mbuf_t *mbuf);
howBlocking or non-blocking
packetlenThe total length of the packet mbuf to be allocated. The length must be greater than zero.
maxchunksAn input/output pointer to the maximum number of mbufs segments making up the chain. On input, if maxchunks is NULL, or the value pointed to by maxchunks is zero, the packet will be made up of as few or as many buffer segments as necessary to fit the length. The allocation will fail with ENOBUFS if the number of segments requested is too small and the sum of the maximum size of each individual segment is less than the packet length. On output, if the allocation succeed and maxchunks is non-NULL, it will point to the actual number of segments allocated. Additional notes for packetlen greater than 4096 bytes: the caller may pass a non-NULL maxchunks and initialize it with zero such that upon success, it can find out whether or not the system configuration allows for larger than 4096 bytes cluster allocations, by checking on the value pointed to by maxchunks. E.g. a request for 9018 bytes may result in 1 chunk when jumbo clusters are available, or 3 chunks otherwise.
Uponsuccess, *mbuf will be a reference to the new mbuf.
Returns 0 upon success or the following error code: EINVAL - Invalid parameter ENOMEM - Not enough memory available ENOBUFS - Buffers not big enough for the maximum number of chunks requested
Allocate an mbuf chain to store a single packet of the requested length. According to the requested length, a chain of mbufs will be created. The mbuf type will be set to MBUF_TYPE_DATA. The caller may specify the maximum number of buffer.
mbuf_allocpacket_list |
errno_t mbuf_allocpacket_list( unsigned int numpkts, mbuf_how_t how, size_t packetlen, unsigned int *maxchunks, mbuf_t *mbuf);
numpktsNumber of packets to allocate
howBlocking or non-blocking
packetlenThe total length of the packet mbuf to be allocated. The length must be greater than zero.
maxchunksAn input/output pointer to the maximum number of mbufs segments making up the chain. On input, if maxchunks is zero, or the value pointed to by maxchunks is zero, the packet will be made of as few or as many buffer segments as necessary to fit the length. The allocation will fail with ENOBUFS if the number of segments requested is too small and the sum of the maximum size of each individual segment is less than the packet length. On output, if the allocation succeed and maxchunks is non zero, it will point to the actual number of segments allocated. Additional notes for packetlen greater than 4096 bytes: the caller may pass a non-NULL maxchunks and initialize it with zero such that upon success, it can find out whether or not the system configuration allows for larger than 4096 bytes cluster allocations, by checking on the value pointed to by maxchunks. E.g. a request for 9018 bytes may result in 1 chunk when jumbo clusters are available, or 3 chunks otherwise.
Uponsuccess, *mbuf will be a reference to the new mbuf.
Returns 0 upon success or the following error code: EINVAL - Invalid parameter ENOMEM - Not enough memory available ENOBUFS - Buffers not big enough for the maximum number of chunks requested
Allocate a linked list of packets. According to the requested length, each packet will made of a chain of one or more mbufs. The mbuf type will be set to MBUF_TYPE_DATA. The caller may specify the maximum number of element for each mbuf chain making up a packet.
mbuf_attachcluster |
errno_t mbuf_attachcluster( mbuf_how_t how, mbuf_type_t type, mbuf_t *mbuf, caddr_t extbuf, void (*extfree)( caddr_t, u_int, caddr_t), size_t extsize, caddr_t extarg);
howBlocking or non-blocking.
typeThe type of the mbuf if mbuf is non-NULL; otherwise ignored.
mbufPointer to the address of the mbuf; if NULL, an mbuf will be allocated, otherwise, it must point to a valid mbuf address. If the user-supplied mbuf is already attached to a cluster, the current cluster will be freed before the mbuf gets attached to the supplied external buffer. Note that this routine may return a different mbuf_t than the one you passed in.
extbufAddress of the external buffer.
extfreeFree routine for the external buffer; the caller is required to defined a routine that will be invoked when the mbuf is freed.
extsizeSize of the external buffer.
extargPrivate value that will be passed to the free routine when it is called at the time the mbuf is freed.
0 on success EINVAL - Invalid parameter ENOMEM - Not enough memory available
Attach an external buffer as a cluster for an mbuf. If mbuf points to a NULL mbuf_t, an mbuf will be allocated for you. If mbuf points to a non-NULL mbuf_t, the user-supplied mbuf will be used instead. The caller is responsible for allocating the external buffer by calling mbuf_alloccluster().
mbuf_clear_csum_performed |
errno_t mbuf_clear_csum_performed( mbuf_t mbuf);
mbufThe mbuf containing the packet.
0 upon success otherwise the errno error.
Clears the hardware checksum flags and values.
mbuf_clear_csum_requested |
errno_t mbuf_clear_csum_requested( mbuf_t mbuf);
mbufThe mbuf containing the packet.
0 upon success otherwise the errno error.
This function clears the checksum request flags.
mbuf_clear_vlan_tag |
errno_t mbuf_clear_vlan_tag( mbuf_t mbuf);
mbufThe mbuf containing the packet.
0 upon success otherwise the errno error.
This function will clear any vlan tag associated with the mbuf.
mbuf_copy_pkthdr |
errno_t mbuf_copy_pkthdr( mbuf_t dest, const mbuf_t src);
srcThe mbuf from which the packet header will be copied.
mbufThe mbuf to which the packet header will be copied.
0 upon success otherwise the errno error.
Copies the packet header from src to dest.
mbuf_copyback |
errno_t mbuf_copyback( mbuf_t mbuf, size_t offset, size_t length, const void *data, mbuf_how_t how);
mbufThe first mbuf in the chain to copy the data in to.
offsetOffset in bytes to skip before copying data.
lengthThe length, in bytes, of the data to copy in to the mbuf chain.
dataA pointer to data in the kernel's address space.
howBlocking or non-blocking.
0 upon success, EINVAL or ENOBUFS upon failure.
Copies data from a buffer to an mbuf chain. mbuf_copyback will grow the chain to fit the specified buffer.
If mbuf_copydata is unable to allocate enough mbufs to grow the chain, ENOBUFS will be returned. The mbuf chain will be shorter than expected but all of the data up to the end of the mbuf chain will be valid.
If an offset is specified, mbuf_copyback will skip that many bytes in the mbuf chain before starting to write the buffer in to the chain. If the mbuf chain does not contain this many bytes, mbufs will be allocated to create the space.
mbuf_copydata |
errno_t mbuf_copydata( const mbuf_t mbuf, size_t offset, size_t length, void* out_data);
mbufThe mbuf chain to copy data out of.
offsetThe offset in to the mbuf to start copying.
lengthThe number of bytes to copy.
out_dataA pointer to the location where the data will be copied.
0 upon success otherwise the errno error.
Copies data out of an mbuf in to a specified buffer. If the data is stored in a chain of mbufs, the data will be copied from each mbuf in the chain until length bytes have been copied.
mbuf_copym |
errno_t mbuf_copym( const mbuf_t src, size_t offset, size_t len, mbuf_how_t how, mbuf_t* new_mbuf);
srcThe source mbuf.
offsetThe offset in the mbuf to start copying from.
lenThe the number of bytes to copy.
howTo block or not to block, that is a question.
new_mbufUpon success, the newly allocated mbuf.
0 upon success otherwise the errno error.
Copies len bytes from offset from src to a new mbuf. If the original mbuf contains a packet header, the new mbuf will contain similar packet header except for any tags which may be associated with the original mbuf. mbuf_dup() should be used instead if tags are to be copied to the new mbuf.
mbuf_data |
void* mbuf_data( mbuf_t mbuf);
mbufThe mbuf.
A pointer to the data in the mbuf.
Returns a pointer to the start of data in this mbuf. There may be additional data on chained mbufs. The data you're looking for may not be virtually contiguous if it spans more than one mbuf. In addition, data that is virtually contiguous might not be represented by physically contiguous pages; see further comments in mbuf_data_to_physical. Use mbuf_len to determine the lenght of data available in this mbuf. If a data structure you want to access stradles two mbufs in a chain, either use mbuf_pullup to get the data contiguous in one mbuf or copy the pieces of data from each mbuf in to a contiguous buffer. Using mbuf_pullup has the advantage of not having to copy the data. On the other hand, if you don't make sure there is space in the mbuf, mbuf_pullup may fail and free the mbuf.
mbuf_data_to_physical |
addr64_t mbuf_data_to_physical( void *ptr);
ptrA pointer to data stored in an mbuf.
The 64 bit physical address of the mbuf data or NULL if ptr does not point to data stored in an mbuf.
mbuf_data_to_physical is a replacement for mcl_to_paddr. Given a pointer returned from mbuf_data of mbuf_datastart, mbuf_data_to_physical will return the phyical address for that block of data. Note that even though the data is in virtually contiguous span, the underlying physical pages might not be physically contiguous. Because of this, callers must ensure to call this routine for each page boundary. Device drivers that deal with DMA are strongly encouraged to utilize the IOMbufNaturalMemoryCursor and walk down the list of vectors instead of using this interface to obtain the physical address. Use of this routine is therefore discouraged.
mbuf_datastart |
void* mbuf_datastart( mbuf_t mbuf);
mbufThe mbuf.
A pointer to smallest possible value for data.
Returns the start of the space set aside for storing data in an mbuf. An mbuf's data may come from a cluster or be embedded in the mbuf structure itself. The data pointer retrieved by mbuf_data may not be at the start of the data (mbuf_leadingspace will be non-zero). This function will return to you a pointer that matches mbuf_data() - mbuf_leadingspace().
mbuf_dup |
errno_t mbuf_dup( const mbuf_t src, mbuf_how_t how, mbuf_t* new_mbuf);
srcThe source mbuf.
howBlocking or non-blocking.
new_mbufUpon success, the newly allocated mbuf.
0 upon success otherwise the errno error.
Exactly duplicates an mbuf chain. If the original mbuf contains a packet header (including tags), the new mbuf will have the same packet header contents and a copy of each tag associated with the original mbuf.
mbuf_flags |
mbuf_flags_t mbuf_flags( const mbuf_t mbuf);
mbufThe mbuf.
The flags.
Returns the set flags.
mbuf_free |
mbuf_t mbuf_free( mbuf_t mbuf);
mbufThe mbuf to free.
The next mbuf in the chain.
Frees a single mbuf. Not commonly used because it doesn't touch the rest of the mbufs on the chain.
mbuf_freecluster |
void mbuf_freecluster( caddr_t addr, size_t size);
addrThe address of the cluster.
sizeThe actual size of the cluster.
Free a cluster that was previously allocated by a call to mbuf_alloccluster(). The caller must pass the actual size of the cluster as returned by mbuf_alloccluster(), which at this point must be either 2048, 4096 or 16384 bytes.
mbuf_freem |
void mbuf_freem( mbuf_t mbuf);
mbufThe first mbuf in the chain to free.
Frees a chain of mbufs link through mnext.
mbuf_freem_list |
int mbuf_freem_list( mbuf_t mbuf);
mbufThe first mbuf in the linked list to free.
The number of mbufs freed.
Frees linked list of mbuf chains. Walks through mnextpackt and does the equivalent of mbuf_freem to each.
mbuf_get |
errno_t mbuf_get( mbuf_how_t how, mbuf_type_t type, mbuf_t *mbuf);
howBlocking or non-blocking.
typeThe type of the mbuf.
mbufThe mbuf.
0 on success, errno error on failure.
Allocates an mbuf without a cluster for external data.
mbuf_get_csum_requested |
errno_t mbuf_get_csum_requested( mbuf_t mbuf, mbuf_csum_request_flags_t *request, u_int32_t *value);
mbufThe mbuf containing the packet.
requestFlags indicating which checksums are being requested for this packet.
valueThis parameter is currently unsupported.
0 upon success otherwise the errno error.
This function is used by the driver to determine which checksum operations should be performed in hardware.
mbuf_get_vlan_tag |
errno_t mbuf_get_vlan_tag( mbuf_t mbuf, u_int16_t *vlan);
mbufThe mbuf containing the packet.
vlanThe protocol family of the aux data to add.
0 upon success otherwise the errno error. ENXIO indicates that the vlan tag is not set.
This function is used by drivers that support hardware vlan tagging to determine which vlan this packet belongs to. To differentiate between the case where the vlan tag is zero and the case where there is no vlan tag, this function will return ENXIO when there is no vlan.
mbuf_getcluster |
errno_t mbuf_getcluster( mbuf_how_t how, mbuf_type_t type, size_t size, mbuf_t *mbuf);
howBlocking or non-blocking.
typeThe type of the mbuf.
sizeThe size of the cluster to be allocated. Supported sizes for a cluster are be 2048, 4096, or 16384. Any other value with return EINVAL. Note that clusters greater than 4096 bytes might not be available in all configurations; the caller must additionally check for ENOTSUP (see below).
mbufThe mbuf the cluster will be attached to.
0 on success, errno error on failure. If you specified NULL for the mbuf, any intermediate mbuf that may have been allocated will be freed. If you specify an mbuf value in *mbuf, mbuf_mclget will not free it. EINVAL - Invalid parameter ENOMEM - Not enough memory available ENOTSUP - The caller had requested greater than 4096 bytes cluster and the system is unable to fulfill it due to the lack of jumbo clusters support based on the configuration. In this case, the caller is advised to use 4096 bytes or smaller during subsequent requests.
Allocate a cluster of the requested size and attach it to an mbuf for use as external data. If mbuf points to a NULL mbuf_t, an mbuf will be allocated for you. If mbuf points to a non-NULL mbuf_t, mbuf_getcluster may return a different mbuf_t than the one you passed in.
mbuf_gethdr |
errno_t mbuf_gethdr( mbuf_how_t how, mbuf_type_t type, mbuf_t *mbuf);
howBlocking or non-blocking.
typeThe type of the mbuf.
mbufThe mbuf.
0 on success, errno error on failure.
Allocates an mbuf without a cluster for external data. Sets a flag to indicate there is a packet header and initializes the packet header.
mbuf_getpacket |
errno_t mbuf_getpacket( mbuf_how_t how, mbuf_t *mbuf);
howBlocking or non-blocking.
mbufUpon success, *mbuf will be a reference to the new mbuf.
0 on success, errno error on failure.
Allocate an mbuf, allocate and attach a cluster, and set the packet header flag.
mbuf_inbound_modified |
void mbuf_inbound_modified( mbuf_t mbuf);
mbufThe mbuf that has been modified.
This function will clear the checksum flags to indicate that a hardware checksum should not be used. Any filter modifying data should call this function on an mbuf before passing the packet up the stack. If a filter modifies a packet in a way that affects any checksum, the filter is responsible for either modifying the checksum to compensate for the changes or verifying the checksum before making the changes and then modifying the data and calculating a new checksum only if the original checksum was valid.
mbuf_inet6_cksum |
errno_t mbuf_inet6_cksum( mbuf_t mbuf, int protocol, u_int32_t offset, u_int32_t length, u_int16_t *csum);
mbufThe mbuf (or chain of mbufs) containing the packet.
protocolA zero or non-zero value. A non-zero value specifies the transport protocol used for pseudo header checksum.
offsetA zero or non-zero value; if the latter, it specifies the offset of the transport header from the beginning of mbuf.
lengthThe total (non-zero) length of the transport segment.
csumPointer to the checksum variable; upon success, this routine will return the calculated Internet checksum through this variable. The caller must set it to a non-NULL value.
0 upon success otherwise the errno error.
@discussions Calculates 16-bit 1's complement Internet checksum of the transport segment with or without the pseudo header checksum of a given IPv6 packet. If the caller specifies a non-zero transport protocol, the checksum returned will also include the pseudo header checksum for the corresponding transport header. Otherwise, no header parsing will be done and the caller may use this to calculate the Internet checksum of an arbitrary span of data.
This routine does not modify the contents of the packet. If the caller specifies a non-zero protocol and/or offset, the routine expects the complete protocol header(s) to be present at the beginning of the first mbuf.
mbuf_inet_cksum |
errno_t mbuf_inet_cksum( mbuf_t mbuf, int protocol, u_int32_t offset, u_int32_t length, u_int16_t *csum);
mbufThe mbuf (or chain of mbufs) containing the packet.
protocolA zero or non-zero value. A non-zero value specifies the transport protocol used for pseudo header checksum.
offsetA zero or non-zero value; if the latter, it specifies the offset of the transport header from the beginning of mbuf.
lengthThe total (non-zero) length of the transport segment.
csumPointer to the checksum variable; upon success, this routine will return the calculated Internet checksum through this variable. The caller must set it to a non-NULL value.
0 upon success otherwise the errno error.
@discussions Calculates 16-bit 1's complement Internet checksum of the transport segment with or without the pseudo header checksum of a given IPv4 packet. If the caller specifies a non-zero transport protocol, the checksum returned will also include the pseudo header checksum for the corresponding transport header. Otherwise, no header parsing will be done and the caller may use this to calculate the Internet checksum of an arbitrary span of data.
This routine does not modify the contents of the packet. If the caller specifies a non-zero protocol and/or offset, the routine expects the complete protocol header to be present at the beginning of the first mbuf.
mbuf_leadingspace |
size_t mbuf_leadingspace( const mbuf_t mbuf);
mbufThe mbuf.
The number of unused bytes at the start of the mbuf.
Determines the space available in the mbuf proceeding the current data.
mbuf_len |
size_t mbuf_len( const mbuf_t mbuf);
mbufThe mbuf.
The length.
Gets the length of data in this mbuf.
mbuf_maxlen |
size_t mbuf_maxlen( const mbuf_t mbuf);
mbufThe mbuf.
The maximum lenght of data for this mbuf.
Retrieves the maximum length of data that may be stored in this mbuf. This value assumes that the data pointer was set to the start of the possible range for that pointer (mbuf_data_start).
mbuf_mclget |
errno_t mbuf_mclget( mbuf_how_t how, mbuf_type_t type, mbuf_t *mbuf);
howBlocking or non-blocking.
typeThe type of the mbuf.
mbufThe mbuf the cluster will be attached to.
0 on success, errno error on failure. If you specified NULL for the mbuf, any intermediate mbuf that may have been allocated will be freed. If you specify an mbuf value in *mbuf, mbuf_mclget will not free it.
Allocate a cluster and attach it to an mbuf for use as external data. If mbuf points to a NULL mbuf_t, an mbuf will be allocated for you. If mbuf points to a non-NULL mbuf_t, mbuf_mclget may return a different mbuf_t than the one you passed in.
mbuf_mclhasreference |
int mbuf_mclhasreference( mbuf_t mbuf);
mbufThe mbuf with the cluster to test.
0 if there is no reference by another mbuf, 1 otherwise.
Check if a cluster of an mbuf is referenced by another mbuf. References may be taken, for example, as a result of a call to mbuf_split or mbuf_copym
mbuf_next |
mbuf_t mbuf_next( const mbuf_t mbuf);
mbufThe mbuf.
The next mbuf in the chain.
Returns the next mbuf in the chain.
mbuf_nextpkt |
mbuf_t mbuf_nextpkt( const mbuf_t mbuf);
mbufThe mbuf.
The nextpkt.
Gets the next packet from the mbuf.
mbuf_outbound_finalize |
void mbuf_outbound_finalize( mbuf_t mbuf, u_long protocol_family, size_t protocol_offset);
mbufThe mbuf that should be finalized.
protocol_familyThe protocol family of the packet in the mbuf.
protocol_offsetThe offset from the start of the mbuf to the protocol header. For an IP packet with an ethernet header, this would be the length of an ethernet header.
This function will "finalize" the packet allowing your code to inspect the final packet.
There are a number of operations that are performed in hardware, such as calculating checksums. This function will perform in software the various opterations that were scheduled to be done in hardware. Future operations may include IPSec processing or vlan support. If you are redirecting a packet to a new interface which may not have the same hardware support or encapsulating the packet, you should call this function to force the stack to calculate and fill out the checksums. This will bypass hardware checksums but give you a complete packet to work with. If you need to inspect aspects of the packet which may be generated by hardware, you must call this function to get an aproximate final packet. If you plan to modify the packet in any way, you should call this function.
This function should be called before modifying any outbound packets.
This function may be called at various levels, in some cases additional headers may have already been prepended, such as the case of a packet seen by an interface filter. To handle this, the caller must pass the protocol family of the packet as well as the offset from the start of the packet to the protocol header.
mbuf_pkthdr_adjustlen |
void mbuf_pkthdr_adjustlen( mbuf_t mbuf, int amount);
mbufThe mbuf containing the packet header.
amountThe number of bytes to adjust the packet header length field by.
Adjusts the length of the packet in the packet header.
mbuf_pkthdr_header |
void* mbuf_pkthdr_header( const mbuf_t mbuf);
mbufThe mbuf containing the packet header.
A pointer to the packet header.
Returns a pointer to the packet header.
mbuf_pkthdr_len |
size_t mbuf_pkthdr_len( const mbuf_t mbuf);
mbufThe mbuf containing the packet header with the length to be changed.
The length, in bytes, of the packet.
Returns the length as reported by the packet header.
mbuf_pkthdr_rcvif |
ifnet_t mbuf_pkthdr_rcvif( const mbuf_t mbuf);
mbufThe mbuf containing the packet header.
A reference to the interface.
Returns the interface the packet was received on. This funciton does not modify the reference count of the interface. The interface is only valid for as long as the mbuf is not freed and the rcvif for the mbuf is not changed. Take a reference on the interface that you will release later before doing any of the following: free the mbuf, change the rcvif, pass the mbuf to any function that may free the mbuf or change the rcvif.
mbuf_pkthdr_setheader |
void mbuf_pkthdr_setheader( mbuf_t mbuf, void *header);
mbufThe mbuf containing the packet header.
ifnetA pointer to the header.
0 upon success otherwise the errno error.
Sets the pointer to the packet header.
mbuf_pkthdr_setlen |
void mbuf_pkthdr_setlen( mbuf_t mbuf, size_t len);
mbufThe mbuf containing the packet header.
lenThe new length of the packet.
Sets the length of the packet in the packet header.
mbuf_pkthdr_setrcvif |
errno_t mbuf_pkthdr_setrcvif( mbuf_t mbuf, ifnet_t ifp);
mbufThe mbuf containing the packet header.
ifnetA reference to an interface.
0 upon success otherwise the errno error.
Sets the interface the packet was received on.
mbuf_prepend |
errno_t mbuf_prepend( mbuf_t *mbuf, size_t len, mbuf_how_t how);
mbufThe mbuf to prepend data to. This may change if a new mbuf must be allocated or may be NULL if the operation fails.
lenThe length, in bytes, to be prepended to the mbuf.
howBlocking or non-blocking.
0 upon success otherwise the errno error.
Prepend len bytes to an mbuf. If there is space (mbuf_leadingspace >= len), the mbuf's data ptr is changed and the same mbuf is returned. If there is no space, a new mbuf may be allocated and prepended to the mbuf chain. If the operation fails, the mbuf may be freed (*mbuf will be NULL).
mbuf_pulldown |
errno_t mbuf_pulldown( mbuf_t src, size_t *offset, size_t length, mbuf_t *location);
srcThe start of the mbuf chain.
offsetPass in a pointer to a value with the offset of the data you're interested in making contiguous. Upon success, this will be overwritten with the offset from the mbuf returned in location.
lengthThe length of data that should be made contiguous.
locationUpon success, *location will be the mbuf the data is in.
0 upon success otherwise the errno error.
Make length bytes at offset in the mbuf chain contiguous. Nothing before offset bytes in the chain will be modified. Upon return, location will be the mbuf the data is contiguous in and offset will be the offset in that mbuf at which the data is located. In the case of a failure, the mbuf chain will be freed.
mbuf_pullup |
errno_t mbuf_pullup( mbuf_t *mbuf, size_t len);
mbufThe mbuf in the chain the data should be contiguous in.
lenThe number of bytes to pull from the next mbuf(s).
0 upon success otherwise the errno error. In the case of an error, the mbuf chain has been freed.
Move the next len bytes in to mbuf from other mbufs in the chain. This is commonly used to get the IP and TCP or UDP header contiguous in the first mbuf. If mbuf_pullup fails, the entire mbuf chain will be freed.
mbuf_set_csum_performed |
errno_t mbuf_set_csum_performed( mbuf_t mbuf, mbuf_csum_performed_flags_t flags, u_int32_t value);
mbufThe mbuf containing the packet.
flagsFlags indicating which hardware checksum operations were performed.
valueIf the MBUF_CSUM_DID_DATA flag is set, value should be set to the value of the TCP or UDP header as calculated by the hardware.
0 upon success otherwise the errno error.
This is used by the driver to indicate to the stack which checksum operations were performed in hardware.
mbuf_set_vlan_tag |
errno_t mbuf_set_vlan_tag( mbuf_t mbuf, u_int16_t vlan);
mbufThe mbuf containing the packet.
vlanThe protocol family of the aux data to add.
0 upon success otherwise the errno error.
This function is used by interfaces that support vlan tagging in hardware. This function will set properties in the mbuf to indicate which vlan the packet was received for.
mbuf_setdata |
errno_t mbuf_setdata( mbuf_t mbuf, void *data, size_t len);
mbufThe mbuf.
dataThe new pointer value for data.
lenThe new length of data in the mbuf.
0 on success, errno error on failure.
Sets the data and length values for an mbuf. The data value must be in a valid range. In the case of an mbuf with a cluster, the data value must point to a location in the cluster and the data value plus the length, must be less than the end of the cluster. For data embedded directly in an mbuf (no cluster), the data value must fall somewhere between the start and end of the data area in the mbuf and the data + length must also be in the same range.
mbuf_setflags |
errno_t mbuf_setflags( mbuf_t mbuf, mbuf_flags_t flags);
mbufThe mbuf.
flagsThe flags that should be set, all other flags will be cleared.
0 upon success otherwise the errno error.
Sets the set of set flags.
mbuf_setflags_mask |
errno_t mbuf_setflags_mask( mbuf_t mbuf, mbuf_flags_t flags, mbuf_flags_t mask);
mbufThe mbuf.
flagsThe flags that should be set or cleared.
maskThe mask controlling which flags will be modified.
0 upon success otherwise the errno error.
Useful for setting or clearing individual flags. Easier than calling mbuf_setflags(m, mbuf_flags(m) | M_FLAG).
mbuf_setlen |
void mbuf_setlen( mbuf_t mbuf, size_t len);
mbufThe mbuf.
lenThe new length.
0 upon success otherwise the errno error.
Sets the length of data in this packet. Be careful to not set the length over the space available in the mbuf.
mbuf_setnext |
errno_t mbuf_setnext( mbuf_t mbuf, mbuf_t next);
mbufThe mbuf.
nextThe new next mbuf.
0 upon success otherwise the errno error.
Sets the next mbuf in the chain.
mbuf_setnextpkt |
void mbuf_setnextpkt( mbuf_t mbuf, mbuf_t nextpkt);
mbufThe mbuf.
nextpktThe new next packet.
Sets the next packet attached to this mbuf.
mbuf_settype |
errno_t mbuf_settype( mbuf_t mbuf, mbuf_type_t new_type);
mbufThe mbuf.
new_typeThe new type.
0 upon success otherwise the errno error.
Sets the type of mbuf.
mbuf_split |
errno_t mbuf_split( mbuf_t src, size_t offset, mbuf_how_t how, mbuf_t *new_mbuf);
srcThe mbuf to be split.
offsetThe offset in the buffer where the mbuf should be split.
howBlocking or non-blocking.
new_mbufUpon success, the second half of the split mbuf chain.
0 upon success otherwise the errno error. In the case of failure, the original mbuf chain passed in to src will be preserved.
Split an mbuf chain at a specific offset.
mbuf_stats |
void mbuf_stats( struct mbuf_stat *stats);
statsStorage to copy the stats in to.
Get the mbuf statistics.
mbuf_tag_allocate |
errno_t mbuf_tag_allocate( mbuf_t mbuf, mbuf_tag_id_t module_id, mbuf_tag_type_t type, size_t length, mbuf_how_t how, void **data_p);
mbufThe mbuf to attach this tag to.
module_idA module identifier returned by mbuf_tag_id_find.
typeA 16 bit type value. For a given module_id, you can use a number of different tag types.
lengthThe length, in bytes, to allocate for storage that will be associated with this tag on this mbuf.
howIndicate whether you want to block and wait for memory if memory is not immediately available.
data_pUpon successful return, *data_p will point to the buffer allocated for the mtag.
0 upon success otherwise the errno error.
Allocate an mbuf tag. Mbuf tags allow various portions of the stack to tag mbufs with data that will travel with the mbuf through the stack.
Tags may only be added to mbufs with packet headers (MBUF_PKTHDR flag is set). Mbuf tags are freed when the mbuf is freed or when mbuf_tag_free is called.
mbuf_tag_find |
errno_t mbuf_tag_find( mbuf_t mbuf, mbuf_tag_id_t module_id, mbuf_tag_type_t type, size_t *length, void **data_p);
mbufThe mbuf the tag is attached to.
module_idA module identifier returned by mbuf_tag_id_find.
typeThe 16 bit type of the tag to find.
lengthUpon success, the length of data will be store in *length.
data_pUpon successful return, *data_p will point to the buffer allocated for the mtag.
0 upon success otherwise the errno error.
Find the data associated with an mbuf tag.
mbuf_tag_free |
void mbuf_tag_free( mbuf_t mbuf, mbuf_tag_id_t module_id, mbuf_tag_type_t type);
mbufThe mbuf the tag was allocated on.
module_idThe ID of the tag to free.
typeThe type of the tag to free.
Frees a previously allocated mbuf tag.
mbuf_tag_id_find |
errno_t mbuf_tag_id_find( const char *module_string, mbuf_tag_id_t *module_id);
module_stringA unique string identifying your module. Example: com.apple.nke.SharedIP.
module_idUpon return, a unique identifier for use with mbuf_tag_* functions. This identifier is valid until the machine is rebooted.
0 upon success otherwise the errno error.
Lookup the module id for a string. If there is no module id assigned to this string, a new module id will be assigned. The string should be the bundle id of the kext. In the case of a tag that will be shared across multiple kexts, a common bundle id style string should be used.
The lookup operation is not optimized. A module should call this function once during startup and chache the module id. The module id will not be resassigned until the machine reboots.
mbuf_trailingspace |
size_t mbuf_trailingspace( const mbuf_t mbuf);
mbufThe mbuf.
The number of unused bytes following the current data.
Determines the space available in the mbuf following the current data.
mbuf_type |
mbuf_type_t mbuf_type( const mbuf_t mbuf);
mbufThe mbuf.
The type.
Gets the type of mbuf.
mbuf_stat |
struct mbuf_stat { u_long mbufs; /* mbufs obtained from page pool */ u_long clusters; /* clusters obtained from page pool */ u_long clfree; /* free clusters */ u_long drops; /* times failed to find space */ u_long wait; /* times waited for space */ u_long drain; /* times drained protocols for space */ u_short mtypes[256]; /* type specific mbuf allocations */ u_long mcfail; /* times m_copym failed */ u_long mpfail; /* times m_pullup failed */ u_long msize; /* length of an mbuf */ u_long mclbytes; /* length of an mbuf cluster */ u_long minclsize; /* min length of data to allocate a cluster */ u_long mlen; /* length of data in an mbuf */ u_long mhlen; /* length of data in a header mbuf */ u_long bigclusters; /* number of big clusters */ u_long bigclfree; /* number of big clustser free */ u_long bigmclbytes; /* length of data in a big cluster */ };
mbufsNumber of mbufs (free or otherwise).
clustersNumber of clusters (free or otherwise).
clfreeNumber of free clusters.
dropsNumber of times allocation failed.
waitNumber of times allocation blocked.
drainNumber of times protocol drain functions were called.
mtypesAn array of counts of each type of mbuf allocated.
mcfailNumber of times m_copym failed.
mpfailNumber of times m_pullup failed.
msizeLength of an mbuf.
mclbytesLength of an mbuf cluster.
minclsizeMinimum length of data to allocate a cluster. Anything smaller than this should be placed in chained mbufs.
mlenLength of data in an mbuf.
mhlenLength of data in an mbuf with a packet header.
bigclustersNumber of big clusters.
bigclfreeNumber of unused big clusters.
bigmclbytesLength of a big mbuf cluster.
The mbuf_stat contains mbuf statistics.
mbuf_csum_performed_flags_t |
Checksum performed/requested flags.
enum { MBUF_CSUM_DID_IP = 0x0100, MBUF_CSUM_IP_GOOD = 0x0200, MBUF_CSUM_DID_DATA = 0x0400, MBUF_CSUM_PSEUDO_HDR = 0x0800 };
MBUF_CSUM_DID_IPIndicates that the driver/hardware verified the IP checksum in hardware.
MBUF_CSUM_IP_GOODIndicates whether or not the IP checksum was good or bad. Only valid when MBUF_CSUM_DID_IP is set.
MBUF_CSUM_DID_DATAIndicates that the TCP or UDP checksum was calculated. The value for the checksum calculated in hardware should be passed as the second parameter of mbuf_set_csum_performed. The hardware calculated checksum value can be retrieved using the second parameter passed to mbuf_get_csum_performed.
MBUF_CSUM_PSEUDO_HDRIf set, this indicates that the checksum value for MBUF_CSUM_DID_DATA includes the pseudo header value. If this is not set, the stack will calculate the pseudo header value and add that to the checksum. The value of this bit is only valid when MBUF_CSUM_DID_DATA is set.
Mbufs often contain packets. Some hardware supports performing checksums in hardware. The driver uses these flags to communicate to the stack the checksums that were calculated in hardware.
mbuf_csum_request_flags_t |
Checksum performed/requested flags.
enum { MBUF_CSUM_REQ_IP = 0x0001, MBUF_CSUM_REQ_TCP = 0x0002, MBUF_CSUM_REQ_UDP = 0x0004 };
MBUF_CSUM_REQ_IPIndicates the IP checksum has not been calculated yet.
MBUF_CSUM_REQ_TCPIndicates the TCP checksum has not been calculated yet.
MBUF_CSUM_REQ_UDPIndicates the UDP checksum has not been calculated yet.
Mbufs often contain packets. Some hardware supports performing checksums in hardware. The stack uses these flags to indicate to the driver what sort of checksumming should be handled in by the driver/hardware. These flags will only be set if the driver indicates that it supports the corresponding checksums using ifnet_set_offload.
mbuf_flags_t |
Constants defining mbuf flags. Only the flags listed below can be set or retrieved.
enum { MBUF_EXT = 0x0001, /* has associated external storage */ MBUF_PKTHDR = 0x0002, /* start of record */ MBUF_EOR = 0x0004, /* end of record */ MBUF_BCAST = 0x0100, /* send/received as link-level broadcast */ MBUF_MCAST = 0x0200, /* send/received as link-level multicast */ MBUF_FRAG = 0x0400, /* packet is a fragment of a larger packet */ MBUF_FIRSTFRAG = 0x0800, /* packet is first fragment */ MBUF_LASTFRAG = 0x1000, /* packet is last fragment */ MBUF_PROMISC = 0x2000 /* packet is promiscuous */ };
MBUF_EXTIndicates this mbuf has external data.
MBUF_PKTHDRIndicates this mbuf has a packet header.
MBUF_EORIndicates this mbuf is the end of a record.
MBUF_BCASTIndicates this packet will be sent or was received as a brodcast.
MBUF_MCASTIndicates this packet will be sent or was received as a multicast.
MBUF_FRAGIndicates this packet is a fragment of a larger packet.
MBUF_FIRSTFRAGIndicates this packet is the first fragment.
MBUF_LASTFRAGIndicates this packet is the last fragment.
MBUF_PROMISCIndicates this packet was only received because the interface is in promiscuous mode. This should be set by the demux function. These packets will be discarded after being passed to any interface filters.
mbuf_how_t |
Method of allocating an mbuf.
enum { MBUF_WAITOK = 0, /* Ok to block to get memory */ MBUF_DONTWAIT = 1 /* Don't block, fail if blocking would be required */ };
MBUF_WAITOKAllow a call to allocate an mbuf to block.
MBUF_DONTWAITDon't allow the mbuf allocation call to block, if blocking is necessary fail and return immediately.
Blocking will cause the funnel to be dropped. If the funnel is dropped, other threads may make changes to networking data structures. This can lead to very bad things happening. Blocking on the input our output path can also impact performance. There are some cases where making a blocking call is acceptable. When in doubt, use MBUF_DONTWAIT.
mbuf_type_t |
Types of mbufs.
enum { MBUF_TYPE_FREE = 0, /* should be on free list */ MBUF_TYPE_DATA = 1, /* dynamic (data) allocation */ MBUF_TYPE_HEADER = 2, /* packet header */ MBUF_TYPE_SOCKET = 3, /* socket structure */ MBUF_TYPE_PCB = 4, /* protocol control block */ MBUF_TYPE_RTABLE = 5, /* routing tables */ MBUF_TYPE_HTABLE = 6, /* IMP host tables */ MBUF_TYPE_ATABLE = 7, /* address resolution tables */ MBUF_TYPE_SONAME = 8, /* socket name */ MBUF_TYPE_SOOPTS = 10, /* socket options */ MBUF_TYPE_FTABLE = 11, /* fragment reassembly header */ MBUF_TYPE_RIGHTS = 12, /* access rights */ MBUF_TYPE_IFADDR = 13, /* interface address */ MBUF_TYPE_CONTROL = 14, /* extra-data protocol message */ MBUF_TYPE_OOBDATA = 15 /* expedited data */ };
MBUF_MT_FREEIndicates the mbuf is free and is sitting on the queue of free mbufs. If you find that an mbuf you have a reference to has this type, something has gone terribly wrong.
MBUF_MT_DATAIndicates this mbuf is being used to store data.
MBUF_MT_HEADERIndicates this mbuf has a packet header, this is probably a packet.
MBUF_MT_SOCKETSocket structure.
MBUF_MT_PCBProtocol control block.
MBUF_MT_RTABLERouting table entry.
MBUF_MT_HTABLEIMP host tables???.
MBUF_MT_ATABLEAddress resolution table data.
MBUF_MT_SONAMESocket name, usually a sockaddr of some sort.
MBUF_MT_FTABLEFragment reassembly header.
MBUF_MT_RIGHTSAccess rights.
MBUF_MT_IFADDRInterface address.
MBUF_MT_CONTROLExtra-data protocol message (control message).
MBUF_MT_OOBDATAOut of band data.
Some mbufs represent packets, some represnt data waiting on sockets. Other mbufs store control data or other various structures. The mbuf type is used to store what sort of data the mbuf contains.
|
Last Updated: 2008-12-19