freebsd-dev/sys/net/bpf.h

887 lines
26 KiB
C
Raw Normal View History

/*-
1994-05-24 10:09:53 +00:00
* Copyright (c) 1990, 1991, 1993
* The Regents of the University of California. All rights reserved.
*
* This code is derived from the Stanford/CMU enet packet filter,
* (net/enet.c) distributed as part of 4.3BSD, and code contributed
* to Berkeley by Steven McCanne and Van Jacobson both of Lawrence
* Berkeley Laboratory.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 4. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)bpf.h 8.1 (Berkeley) 6/10/93
* @(#)bpf.h 1.34 (LBL) 6/16/96
1994-05-24 10:09:53 +00:00
*
1999-08-28 01:08:13 +00:00
* $FreeBSD$
1994-05-24 10:09:53 +00:00
*/
1994-08-21 05:11:48 +00:00
#ifndef _NET_BPF_H_
#define _NET_BPF_H_
/* BSD style release date */
#define BPF_RELEASE 199606
typedef int32_t bpf_int32;
typedef u_int32_t bpf_u_int32;
1994-05-24 10:09:53 +00:00
/*
1995-05-30 08:16:23 +00:00
* Alignment macros. BPF_WORDALIGN rounds up to the next
* even multiple of BPF_ALIGNMENT.
1994-05-24 10:09:53 +00:00
*/
#define BPF_ALIGNMENT sizeof(long)
1994-05-24 10:09:53 +00:00
#define BPF_WORDALIGN(x) (((x)+(BPF_ALIGNMENT-1))&~(BPF_ALIGNMENT-1))
#define BPF_MAXINSNS 512
#define BPF_MAXBUFSIZE 0x80000
1994-05-24 10:09:53 +00:00
#define BPF_MINBUFSIZE 32
/*
* Structure for BIOCSETF.
*/
struct bpf_program {
u_int bf_len;
struct bpf_insn *bf_insns;
};
1995-05-30 08:16:23 +00:00
1994-05-24 10:09:53 +00:00
/*
* Struct returned by BIOCGSTATS.
*/
struct bpf_stat {
u_int bs_recv; /* number of packets received */
u_int bs_drop; /* number of packets dropped */
};
/*
1995-05-30 08:16:23 +00:00
* Struct return by BIOCVERSION. This represents the version number of
1994-05-24 10:09:53 +00:00
* the filter language described by the instruction encodings below.
* bpf understands a program iff kernel_major == filter_major &&
* kernel_minor >= filter_minor, that is, if the value returned by the
* running kernel has the same major number and a minor number equal
* equal to or less than the filter being downloaded. Otherwise, the
* results are undefined, meaning an error may be returned or packets
* may be accepted haphazardly.
* It has nothing to do with the source code version.
*/
struct bpf_version {
u_short bv_major;
u_short bv_minor;
};
/* Current version number of filter architecture. */
1994-05-24 10:09:53 +00:00
#define BPF_MAJOR_VERSION 1
#define BPF_MINOR_VERSION 1
Introduce support for zero-copy BPF buffering, which reduces the overhead of packet capture by allowing a user process to directly "loan" buffer memory to the kernel rather than using read(2) to explicitly copy data from kernel address space. The user process will issue new BPF ioctls to set the shared memory buffer mode and provide pointers to buffers and their size. The kernel then wires and maps the pages into kernel address space using sf_buf(9), which on supporting architectures will use the direct map region. The current "buffered" access mode remains the default, and support for zero-copy buffers must, for the time being, be explicitly enabled using a sysctl for the kernel to accept requests to use it. The kernel and user process synchronize use of the buffers with atomic operations, avoiding the need for system calls under load; the user process may use select()/poll()/kqueue() to manage blocking while waiting for network data if the user process is able to consume data faster than the kernel generates it. Patchs to libpcap are available to allow libpcap applications to transparently take advantage of this support. Detailed information on the new API may be found in bpf(4), including specific atomic operations and memory barriers required to synchronize buffer use safely. These changes modify the base BPF implementation to (roughly) abstrac the current buffer model, allowing the new shared memory model to be added, and add new monitoring statistics for netstat to print. The implementation, with the exception of some monitoring hanges that break the netstat monitoring ABI for BPF, will be MFC'd. Zerocopy bpf buffers are still considered experimental are disabled by default. To experiment with this new facility, adjust the net.bpf.zerocopy_enable sysctl variable to 1. Changes to libpcap will be made available as a patch for the time being, and further refinements to the implementation are expected. Sponsored by: Seccuris Inc. In collaboration with: rwatson Tested by: pwood, gallatin MFC after: 4 months [1] [1] Certain portions will probably not be MFCed, specifically things that can break the monitoring ABI.
2008-03-24 13:49:17 +00:00
/*
* Historically, BPF has supported a single buffering model, first using mbuf
* clusters in kernel, and later using malloc(9) buffers in kernel. We now
* support multiple buffering modes, which may be queried and set using
* BIOCGETBUFMODE and BIOCSETBUFMODE. So as to avoid handling the complexity
* of changing modes while sniffing packets, the mode becomes fixed once an
* interface has been attached to the BPF descriptor.
*/
#define BPF_BUFMODE_BUFFER 1 /* Kernel buffers with read(). */
#define BPF_BUFMODE_ZBUF 2 /* Zero-copy buffers. */
/*-
* Struct used by BIOCSETZBUF, BIOCROTZBUF: describes up to two zero-copy
* buffer as used by BPF.
*/
struct bpf_zbuf {
void *bz_bufa; /* Location of 'a' zero-copy buffer. */
void *bz_bufb; /* Location of 'b' zero-copy buffer. */
size_t bz_buflen; /* Size of zero-copy buffers. */
};
1994-05-24 10:09:53 +00:00
#define BIOCGBLEN _IOR('B',102, u_int)
#define BIOCSBLEN _IOWR('B',102, u_int)
#define BIOCSETF _IOW('B',103, struct bpf_program)
#define BIOCFLUSH _IO('B',104)
#define BIOCPROMISC _IO('B',105)
#define BIOCGDLT _IOR('B',106, u_int)
#define BIOCGETIF _IOR('B',107, struct ifreq)
#define BIOCSETIF _IOW('B',108, struct ifreq)
#define BIOCSRTIMEOUT _IOW('B',109, struct timeval)
#define BIOCGRTIMEOUT _IOR('B',110, struct timeval)
#define BIOCGSTATS _IOR('B',111, struct bpf_stat)
#define BIOCIMMEDIATE _IOW('B',112, u_int)
#define BIOCVERSION _IOR('B',113, struct bpf_version)
#define BIOCGRSIG _IOR('B',114, u_int)
#define BIOCSRSIG _IOW('B',115, u_int)
#define BIOCGHDRCMPLT _IOR('B',116, u_int)
#define BIOCSHDRCMPLT _IOW('B',117, u_int)
#define BIOCGDIRECTION _IOR('B',118, u_int)
#define BIOCSDIRECTION _IOW('B',119, u_int)
#define BIOCSDLT _IOW('B',120, u_int)
#define BIOCGDLTLIST _IOWR('B',121, struct bpf_dltlist)
Introduce two new ioctl(2) commands, BIOCLOCK and BIOCSETWF. These commands enhance the security of bpf(4) by further relinquishing the privilege of the bpf(4) consumer (assuming the ioctl commands are being implemented). Once BIOCLOCK is executed, the device becomes locked which prevents the execution of ioctl(2) commands which can change the underly parameters of the bpf(4) device. An example might be the setting of bpf(4) filter programs or attaching to different network interfaces. BIOCSETWF can be used to set write filters for outgoing packets. Currently if a bpf(4) consumer is compromised, the bpf(4) descriptor can essentially be used as a raw socket, regardless of consumer's UID. Write filters give users the ability to constrain which packets can be sent through the bpf(4) descriptor. These features are currently implemented by a couple programs which came from OpenBSD, such as the new dhclient and pflogd. -Modify bpf_setf(9) to accept a "cmd" parameter. This will be used to specify whether a read or write filter is to be set. -Add a bpf(4) filter program as a parameter to bpf_movein(9) as we will run the filter program on the mbuf data once we move the packet in from user-space. -Rather than execute two uiomove operations, (one for the link header and the other for the packet data), execute one and manually copy the linker header into the sockaddr structure via bcopy. -Restructure bpf_setf to compensate for write filters, as well as read. -Adjust bpf(4) stats structures to include a bd_locked member. It should be noted that the FreeBSD and OpenBSD implementations differ a bit in the sense that we unconditionally enforce the lock, where OpenBSD enforces it only if the calling credential is not root. Idea from: OpenBSD Reviewed by: mlaier
2005-08-22 19:35:48 +00:00
#define BIOCLOCK _IO('B', 122)
#define BIOCSETWF _IOW('B',123, struct bpf_program)
#define BIOCFEEDBACK _IOW('B',124, u_int)
Introduce support for zero-copy BPF buffering, which reduces the overhead of packet capture by allowing a user process to directly "loan" buffer memory to the kernel rather than using read(2) to explicitly copy data from kernel address space. The user process will issue new BPF ioctls to set the shared memory buffer mode and provide pointers to buffers and their size. The kernel then wires and maps the pages into kernel address space using sf_buf(9), which on supporting architectures will use the direct map region. The current "buffered" access mode remains the default, and support for zero-copy buffers must, for the time being, be explicitly enabled using a sysctl for the kernel to accept requests to use it. The kernel and user process synchronize use of the buffers with atomic operations, avoiding the need for system calls under load; the user process may use select()/poll()/kqueue() to manage blocking while waiting for network data if the user process is able to consume data faster than the kernel generates it. Patchs to libpcap are available to allow libpcap applications to transparently take advantage of this support. Detailed information on the new API may be found in bpf(4), including specific atomic operations and memory barriers required to synchronize buffer use safely. These changes modify the base BPF implementation to (roughly) abstrac the current buffer model, allowing the new shared memory model to be added, and add new monitoring statistics for netstat to print. The implementation, with the exception of some monitoring hanges that break the netstat monitoring ABI for BPF, will be MFC'd. Zerocopy bpf buffers are still considered experimental are disabled by default. To experiment with this new facility, adjust the net.bpf.zerocopy_enable sysctl variable to 1. Changes to libpcap will be made available as a patch for the time being, and further refinements to the implementation are expected. Sponsored by: Seccuris Inc. In collaboration with: rwatson Tested by: pwood, gallatin MFC after: 4 months [1] [1] Certain portions will probably not be MFCed, specifically things that can break the monitoring ABI.
2008-03-24 13:49:17 +00:00
#define BIOCGETBUFMODE _IOR('B',125, u_int)
#define BIOCSETBUFMODE _IOW('B',126, u_int)
#define BIOCGETZMAX _IOR('B',127, size_t)
#define BIOCROTZBUF _IOR('B',128, struct bpf_zbuf)
#define BIOCSETZBUF _IOW('B',129, struct bpf_zbuf)
#define BIOCSETFNR _IOW('B',130, struct bpf_program)
/* Obsolete */
#define BIOCGSEESENT BIOCGDIRECTION
#define BIOCSSEESENT BIOCSDIRECTION
/* Packet directions */
enum bpf_direction {
BPF_D_IN, /* See incoming packets */
BPF_D_INOUT, /* See incoming and outgoing packets */
BPF_D_OUT /* See outgoing packets */
};
1994-05-24 10:09:53 +00:00
/*
* Structure prepended to each packet.
*/
struct bpf_hdr {
struct timeval bh_tstamp; /* time stamp */
bpf_u_int32 bh_caplen; /* length of captured portion */
bpf_u_int32 bh_datalen; /* original length of packet */
1994-05-24 10:09:53 +00:00
u_short bh_hdrlen; /* length of bpf header (this struct
plus alignment padding) */
};
/*
* Because the structure above is not a multiple of 4 bytes, some compilers
* will insist on inserting padding; hence, sizeof(struct bpf_hdr) won't work.
* Only the kernel needs to know about it; applications use bh_hdrlen.
*/
#ifdef _KERNEL
#define SIZEOF_BPF_HDR (sizeof(struct bpf_hdr) <= 20 ? 18 : \
sizeof(struct bpf_hdr))
1994-05-24 10:09:53 +00:00
#endif
Introduce support for zero-copy BPF buffering, which reduces the overhead of packet capture by allowing a user process to directly "loan" buffer memory to the kernel rather than using read(2) to explicitly copy data from kernel address space. The user process will issue new BPF ioctls to set the shared memory buffer mode and provide pointers to buffers and their size. The kernel then wires and maps the pages into kernel address space using sf_buf(9), which on supporting architectures will use the direct map region. The current "buffered" access mode remains the default, and support for zero-copy buffers must, for the time being, be explicitly enabled using a sysctl for the kernel to accept requests to use it. The kernel and user process synchronize use of the buffers with atomic operations, avoiding the need for system calls under load; the user process may use select()/poll()/kqueue() to manage blocking while waiting for network data if the user process is able to consume data faster than the kernel generates it. Patchs to libpcap are available to allow libpcap applications to transparently take advantage of this support. Detailed information on the new API may be found in bpf(4), including specific atomic operations and memory barriers required to synchronize buffer use safely. These changes modify the base BPF implementation to (roughly) abstrac the current buffer model, allowing the new shared memory model to be added, and add new monitoring statistics for netstat to print. The implementation, with the exception of some monitoring hanges that break the netstat monitoring ABI for BPF, will be MFC'd. Zerocopy bpf buffers are still considered experimental are disabled by default. To experiment with this new facility, adjust the net.bpf.zerocopy_enable sysctl variable to 1. Changes to libpcap will be made available as a patch for the time being, and further refinements to the implementation are expected. Sponsored by: Seccuris Inc. In collaboration with: rwatson Tested by: pwood, gallatin MFC after: 4 months [1] [1] Certain portions will probably not be MFCed, specifically things that can break the monitoring ABI.
2008-03-24 13:49:17 +00:00
/*
* When using zero-copy BPF buffers, a shared memory header is present
* allowing the kernel BPF implementation and user process to synchronize
* without using system calls. This structure defines that header. When
* accessing these fields, appropriate atomic operation and memory barriers
* are required in order not to see stale or out-of-order data; see bpf(4)
* for reference code to access these fields from userspace.
*
* The layout of this structure is critical, and must not be changed; if must
* fit in a single page on all architectures.
*/
struct bpf_zbuf_header {
volatile u_int bzh_kernel_gen; /* Kernel generation number. */
volatile u_int bzh_kernel_len; /* Length of data in the buffer. */
volatile u_int bzh_user_gen; /* User generation number. */
u_int _bzh_pad[5];
};
1994-05-24 10:09:53 +00:00
/*
* Data-link level type codes.
*/
#define DLT_NULL 0 /* BSD loopback encapsulation */
1994-05-24 10:09:53 +00:00
#define DLT_EN10MB 1 /* Ethernet (10Mb) */
#define DLT_EN3MB 2 /* Experimental Ethernet (3Mb) */
#define DLT_AX25 3 /* Amateur Radio AX.25 */
#define DLT_PRONET 4 /* Proteon ProNET Token Ring */
#define DLT_CHAOS 5 /* Chaos */
#define DLT_IEEE802 6 /* IEEE 802 Networks */
#define DLT_ARCNET 7 /* ARCNET */
#define DLT_SLIP 8 /* Serial Line IP */
#define DLT_PPP 9 /* Point-to-point Protocol */
#define DLT_FDDI 10 /* FDDI */
#define DLT_ATM_RFC1483 11 /* LLC/SNAP encapsulated atm */
1998-08-18 10:13:11 +00:00
#define DLT_RAW 12 /* raw IP */
/*
* These are values from BSD/OS's "bpf.h".
* These are not the same as the values from the traditional libpcap
* "bpf.h"; however, these values shouldn't be generated by any
* OS other than BSD/OS, so the correct values to use here are the
* BSD/OS values.
*
* Platforms that have already assigned these values to other
* DLT_ codes, however, should give these codes the values
* from that platform, so that programs that use these codes will
* continue to compile - even though they won't correctly read
* files of these types.
*/
#define DLT_SLIP_BSDOS 15 /* BSD/OS Serial Line IP */
#define DLT_PPP_BSDOS 16 /* BSD/OS Point-to-point Protocol */
#define DLT_ATM_CLIP 19 /* Linux Classical-IP over ATM */
/*
* These values are defined by NetBSD; other platforms should refrain from
* using them for other purposes, so that NetBSD savefiles with link
* types of 50 or 51 can be read as this type on all platforms.
*/
#define DLT_PPP_SERIAL 50 /* PPP over serial with HDLC encapsulation */
#define DLT_PPP_ETHER 51 /* PPP over Ethernet */
/*
* Reserved for the Symantec Enterprise Firewall.
*/
#define DLT_SYMANTEC_FIREWALL 99
/*
* This value was defined by libpcap 0.5; platforms that have defined
* it with a different value should define it here with that value -
* a link type of 104 in a save file will be mapped to DLT_C_HDLC,
* whatever value that happens to be, so programs will correctly
* handle files with that link type regardless of the value of
* DLT_C_HDLC.
*
* The name DLT_C_HDLC was used by BSD/OS; we use that name for source
* compatibility with programs written for BSD/OS.
*
* libpcap 0.5 defined it as DLT_CHDLC; we define DLT_CHDLC as well,
* for source compatibility with programs written for libpcap 0.5.
*/
#define DLT_C_HDLC 104 /* Cisco HDLC */
#define DLT_CHDLC DLT_C_HDLC
#define DLT_IEEE802_11 105 /* IEEE 802.11 wireless */
/*
* Values between 106 and 107 are used in capture file headers as
* link-layer types corresponding to DLT_ types that might differ
* between platforms; don't use those values for new DLT_ new types.
*/
/*
* Frame Relay; BSD/OS has a DLT_FR with a value of 11, but that collides
* with other values.
* DLT_FR and DLT_FRELAY packets start with the Q.922 Frame Relay header
* (DLCI, etc.).
*/
#define DLT_FRELAY 107
/*
* OpenBSD DLT_LOOP, for loopback devices; it's like DLT_NULL, except
* that the AF_ type in the link-layer header is in network byte order.
*
* OpenBSD defines it as 12, but that collides with DLT_RAW, so we
* define it as 108 here. If OpenBSD picks up this file, it should
* define DLT_LOOP as 12 in its version, as per the comment above -
* and should not use 108 as a DLT_ value.
*/
#define DLT_LOOP 108
/*
* Values between 109 and 112 are used in capture file headers as
* link-layer types corresponding to DLT_ types that might differ
* between platforms; don't use those values for new DLT_ new types.
*/
/*
* Encapsulated packets for IPsec; DLT_ENC is 13 in OpenBSD, but that's
* DLT_SLIP_BSDOS in NetBSD, so we don't use 13 for it in OSes other
* than OpenBSD.
*/
#define DLT_ENC 109
/*
* This is for Linux cooked sockets.
*/
#define DLT_LINUX_SLL 113
1994-05-24 10:09:53 +00:00
/*
* Apple LocalTalk hardware.
*/
#define DLT_LTALK 114
/*
* Acorn Econet.
*/
#define DLT_ECONET 115
/*
* Reserved for use with OpenBSD ipfilter.
*/
#define DLT_IPFILTER 116
/*
* Reserved for use in capture-file headers as a link-layer type
* corresponding to OpenBSD DLT_PFLOG; DLT_PFLOG is 17 in OpenBSD,
* but that's DLT_LANE8023 in SuSE 6.3, so we can't use 17 for it
* in capture-file headers.
*/
#define DLT_PFLOG 117
/*
* Registered for Cisco-internal use.
*/
#define DLT_CISCO_IOS 118
/*
* Reserved for 802.11 cards using the Prism II chips, with a link-layer
* header including Prism monitor mode information plus an 802.11
* header.
*/
#define DLT_PRISM_HEADER 119
/*
* Reserved for Aironet 802.11 cards, with an Aironet link-layer header
* (see Doug Ambrisko's FreeBSD patches).
*/
#define DLT_AIRONET_HEADER 120
/*
* Reserved for use by OpenBSD's pfsync device.
*/
#define DLT_PFSYNC 121
/*
* Reserved for Siemens HiPath HDLC. XXX
*/
#define DLT_HHDLC 121
/*
* Reserved for RFC 2625 IP-over-Fibre Channel.
*/
#define DLT_IP_OVER_FC 122
/*
* Reserved for Full Frontal ATM on Solaris.
*/
#define DLT_SUNATM 123
/*
* Reserved as per request from Kent Dahlgren <kent@praesum.com>
* for private use.
*/
#define DLT_RIO 124 /* RapidIO */
#define DLT_PCI_EXP 125 /* PCI Express */
#define DLT_AURORA 126 /* Xilinx Aurora link layer */
/*
* BSD header for 802.11 plus a number of bits of link-layer information
* including radio information.
*/
#ifndef DLT_IEEE802_11_RADIO
#define DLT_IEEE802_11_RADIO 127
#endif
/*
* Reserved for TZSP encapsulation.
*/
#define DLT_TZSP 128 /* Tazmen Sniffer Protocol */
/*
* Reserved for Linux ARCNET.
*/
#define DLT_ARCNET_LINUX 129
/*
* Juniper-private data link types.
*/
#define DLT_JUNIPER_MLPPP 130
#define DLT_JUNIPER_MLFR 131
#define DLT_JUNIPER_ES 132
#define DLT_JUNIPER_GGSN 133
#define DLT_JUNIPER_MFR 134
#define DLT_JUNIPER_ATM2 135
#define DLT_JUNIPER_SERVICES 136
#define DLT_JUNIPER_ATM1 137
/*
* Apple IP-over-IEEE 1394, as per a request from Dieter Siegmund
* <dieter@apple.com>. The header that's presented is an Ethernet-like
* header:
*
* #define FIREWIRE_EUI64_LEN 8
* struct firewire_header {
* u_char firewire_dhost[FIREWIRE_EUI64_LEN];
* u_char firewire_shost[FIREWIRE_EUI64_LEN];
* u_short firewire_type;
* };
*
* with "firewire_type" being an Ethernet type value, rather than,
* for example, raw GASP frames being handed up.
*/
#define DLT_APPLE_IP_OVER_IEEE1394 138
/*
* Various SS7 encapsulations, as per a request from Jeff Morriss
* <jeff.morriss[AT]ulticom.com> and subsequent discussions.
*/
#define DLT_MTP2_WITH_PHDR 139 /* pseudo-header with various info, followed by MTP2 */
#define DLT_MTP2 140 /* MTP2, without pseudo-header */
#define DLT_MTP3 141 /* MTP3, without pseudo-header or MTP2 */
#define DLT_SCCP 142 /* SCCP, without pseudo-header or MTP2 or MTP3 */
/*
* Reserved for DOCSIS.
*/
#define DLT_DOCSIS 143
/*
* Reserved for Linux IrDA.
*/
#define DLT_LINUX_IRDA 144
/*
* Reserved for IBM SP switch and IBM Next Federation switch.
*/
#define DLT_IBM_SP 145
#define DLT_IBM_SN 146
/*
* Reserved for private use. If you have some link-layer header type
* that you want to use within your organization, with the capture files
* using that link-layer header type not ever be sent outside your
* organization, you can use these values.
*
* No libpcap release will use these for any purpose, nor will any
* tcpdump release use them, either.
*
* Do *NOT* use these in capture files that you expect anybody not using
* your private versions of capture-file-reading tools to read; in
* particular, do *NOT* use them in products, otherwise you may find that
* people won't be able to use tcpdump, or snort, or Ethereal, or... to
* read capture files from your firewall/intrusion detection/traffic
* monitoring/etc. appliance, or whatever product uses that DLT_ value,
* and you may also find that the developers of those applications will
* not accept patches to let them read those files.
*
* Also, do not use them if somebody might send you a capture using them
* for *their* private type and tools using them for *your* private type
* would have to read them.
*
* Instead, ask "tcpdump-workers@tcpdump.org" for a new DLT_ value,
* as per the comment above, and use the type you're given.
*/
#define DLT_USER0 147
#define DLT_USER1 148
#define DLT_USER2 149
#define DLT_USER3 150
#define DLT_USER4 151
#define DLT_USER5 152
#define DLT_USER6 153
#define DLT_USER7 154
#define DLT_USER8 155
#define DLT_USER9 156
#define DLT_USER10 157
#define DLT_USER11 158
#define DLT_USER12 159
#define DLT_USER13 160
#define DLT_USER14 161
#define DLT_USER15 162
/*
* For future use with 802.11 captures - defined by AbsoluteValue
* Systems to store a number of bits of link-layer information
* including radio information:
*
* http://www.shaftnet.org/~pizza/software/capturefrm.txt
*
* but it might be used by some non-AVS drivers now or in the
* future.
*/
#define DLT_IEEE802_11_RADIO_AVS 163 /* 802.11 plus AVS radio header */
/*
* Juniper-private data link type, as per request from
* Hannes Gredler <hannes@juniper.net>. The DLT_s are used
* for passing on chassis-internal metainformation such as
* QOS profiles, etc..
*/
#define DLT_JUNIPER_MONITOR 164
/*
* Reserved for BACnet MS/TP.
*/
#define DLT_BACNET_MS_TP 165
/*
* Another PPP variant as per request from Karsten Keil <kkeil@suse.de>.
*
* This is used in some OSes to allow a kernel socket filter to distinguish
* between incoming and outgoing packets, on a socket intended to
* supply pppd with outgoing packets so it can do dial-on-demand and
* hangup-on-lack-of-demand; incoming packets are filtered out so they
* don't cause pppd to hold the connection up (you don't want random
* input packets such as port scans, packets from old lost connections,
* etc. to force the connection to stay up).
*
* The first byte of the PPP header (0xff03) is modified to accomodate
* the direction - 0x00 = IN, 0x01 = OUT.
*/
#define DLT_PPP_PPPD 166
/*
* Names for backwards compatibility with older versions of some PPP
* software; new software should use DLT_PPP_PPPD.
*/
#define DLT_PPP_WITH_DIRECTION DLT_PPP_PPPD
#define DLT_LINUX_PPP_WITHDIRECTION DLT_PPP_PPPD
/*
* Juniper-private data link type, as per request from
* Hannes Gredler <hannes@juniper.net>. The DLT_s are used
* for passing on chassis-internal metainformation such as
* QOS profiles, cookies, etc..
*/
#define DLT_JUNIPER_PPPOE 167
#define DLT_JUNIPER_PPPOE_ATM 168
#define DLT_GPRS_LLC 169 /* GPRS LLC */
#define DLT_GPF_T 170 /* GPF-T (ITU-T G.7041/Y.1303) */
#define DLT_GPF_F 171 /* GPF-F (ITU-T G.7041/Y.1303) */
/*
* Requested by Oolan Zimmer <oz@gcom.com> for use in Gcom's T1/E1 line
* monitoring equipment.
*/
#define DLT_GCOM_T1E1 172
#define DLT_GCOM_SERIAL 173
/*
* Juniper-private data link type, as per request from
* Hannes Gredler <hannes@juniper.net>. The DLT_ is used
* for internal communication to Physical Interface Cards (PIC)
*/
#define DLT_JUNIPER_PIC_PEER 174
/*
* Link types requested by Gregor Maier <gregor@endace.com> of Endace
* Measurement Systems. They add an ERF header (see
* http://www.endace.com/support/EndaceRecordFormat.pdf) in front of
* the link-layer header.
*/
#define DLT_ERF_ETH 175 /* Ethernet */
#define DLT_ERF_POS 176 /* Packet-over-SONET */
/*
* Requested by Daniele Orlandi <daniele@orlandi.com> for raw LAPD
* for vISDN (http://www.orlandi.com/visdn/). Its link-layer header
* includes additional information before the LAPD header, so it's
* not necessarily a generic LAPD header.
*/
#define DLT_LINUX_LAPD 177
2006-09-04 19:24:34 +00:00
/*
* Juniper-private data link type, as per request from
* Hannes Gredler <hannes@juniper.net>.
2006-09-04 19:24:34 +00:00
* The DLT_ are used for prepending meta-information
* like interface index, interface name
* before standard Ethernet, PPP, Frelay & C-HDLC Frames
*/
#define DLT_JUNIPER_ETHER 178
#define DLT_JUNIPER_PPP 179
#define DLT_JUNIPER_FRELAY 180
#define DLT_JUNIPER_CHDLC 181
/*
* Multi Link Frame Relay (FRF.16)
*/
#define DLT_MFR 182
/*
* Juniper-private data link type, as per request from
* Hannes Gredler <hannes@juniper.net>.
* The DLT_ is used for internal communication with a
* voice Adapter Card (PIC)
*/
#define DLT_JUNIPER_VP 183
/*
* Arinc 429 frames.
* DLT_ requested by Gianluca Varenni <gianluca.varenni@cacetech.com>.
* Every frame contains a 32bit A429 label.
* More documentation on Arinc 429 can be found at
* http://www.condoreng.com/support/downloads/tutorials/ARINCTutorial.pdf
*/
#define DLT_A429 184
/*
* Arinc 653 Interpartition Communication messages.
* DLT_ requested by Gianluca Varenni <gianluca.varenni@cacetech.com>.
* Please refer to the A653-1 standard for more information.
*/
#define DLT_A653_ICM 185
/*
* USB packets, beginning with a USB setup header; requested by
* Paolo Abeni <paolo.abeni@email.it>.
*/
#define DLT_USB 186
/*
* Bluetooth HCI UART transport layer (part H:4); requested by
* Paolo Abeni.
*/
#define DLT_BLUETOOTH_HCI_H4 187
/*
* IEEE 802.16 MAC Common Part Sublayer; requested by Maria Cruz
* <cruz_petagay@bah.com>.
*/
#define DLT_IEEE802_16_MAC_CPS 188
/*
* USB packets, beginning with a Linux USB header; requested by
* Paolo Abeni <paolo.abeni@email.it>.
*/
#define DLT_USB_LINUX 189
/*
* Controller Area Network (CAN) v. 2.0B packets.
* DLT_ requested by Gianluca Varenni <gianluca.varenni@cacetech.com>.
* Used to dump CAN packets coming from a CAN Vector board.
* More documentation on the CAN v2.0B frames can be found at
* http://www.can-cia.org/downloads/?269
*/
#define DLT_CAN20B 190
/*
* IEEE 802.15.4, with address fields padded, as is done by Linux
* drivers; requested by Juergen Schimmer.
*/
#define DLT_IEEE802_15_4_LINUX 191
/*
* Per Packet Information encapsulated packets.
* DLT_ requested by Gianluca Varenni <gianluca.varenni@cacetech.com>.
*/
#define DLT_PPI 192
/*
* Header for 802.16 MAC Common Part Sublayer plus a radiotap radio header;
* requested by Charles Clancy.
*/
#define DLT_IEEE802_16_MAC_CPS_RADIO 193
/*
* Juniper-private data link type, as per request from
* Hannes Gredler <hannes@juniper.net>.
* The DLT_ is used for internal communication with a
* integrated service module (ISM).
*/
#define DLT_JUNIPER_ISM 194
/*
* IEEE 802.15.4, exactly as it appears in the spec (no padding, no
* nothing); requested by Mikko Saarnivala <mikko.saarnivala@sensinode.com>.
*/
#define DLT_IEEE802_15_4 195
/*
* Various link-layer types, with a pseudo-header, for SITA
* (http://www.sita.aero/); requested by Fulko Hew (fulko.hew@gmail.com).
*/
#define DLT_SITA 196
/*
* Various link-layer types, with a pseudo-header, for Endace DAG cards;
* encapsulates Endace ERF records. Requested by Stephen Donnelly
* <stephen@endace.com>.
*/
#define DLT_ERF 197
/*
* Special header prepended to Ethernet packets when capturing from a
* u10 Networks board. Requested by Phil Mulholland
* <phil@u10networks.com>.
*/
#define DLT_RAIF1 198
/*
* IPMB packet for IPMI, beginning with the I2C slave address, followed
* by the netFn and LUN, etc.. Requested by Chanthy Toeung
* <chanthy.toeung@ca.kontron.com>.
*/
#define DLT_IPMB 199
/*
* Juniper-private data link type, as per request from
* Hannes Gredler <hannes@juniper.net>.
* The DLT_ is used for capturing data on a secure tunnel interface.
*/
#define DLT_JUNIPER_ST 200
/*
* Bluetooth HCI UART transport layer (part H:4), with pseudo-header
* that includes direction information; requested by Paolo Abeni.
*/
#define DLT_BLUETOOTH_HCI_H4_WITH_PHDR 201
1994-05-24 10:09:53 +00:00
/*
* The instruction encodings.
1994-05-24 10:09:53 +00:00
*/
/* instruction classes */
#define BPF_CLASS(code) ((code) & 0x07)
#define BPF_LD 0x00
#define BPF_LDX 0x01
#define BPF_ST 0x02
#define BPF_STX 0x03
#define BPF_ALU 0x04
#define BPF_JMP 0x05
#define BPF_RET 0x06
#define BPF_MISC 0x07
/* ld/ldx fields */
#define BPF_SIZE(code) ((code) & 0x18)
#define BPF_W 0x00
#define BPF_H 0x08
#define BPF_B 0x10
#define BPF_MODE(code) ((code) & 0xe0)
#define BPF_IMM 0x00
#define BPF_ABS 0x20
#define BPF_IND 0x40
#define BPF_MEM 0x60
#define BPF_LEN 0x80
#define BPF_MSH 0xa0
/* alu/jmp fields */
#define BPF_OP(code) ((code) & 0xf0)
#define BPF_ADD 0x00
#define BPF_SUB 0x10
#define BPF_MUL 0x20
#define BPF_DIV 0x30
#define BPF_OR 0x40
#define BPF_AND 0x50
#define BPF_LSH 0x60
#define BPF_RSH 0x70
#define BPF_NEG 0x80
#define BPF_JA 0x00
#define BPF_JEQ 0x10
#define BPF_JGT 0x20
#define BPF_JGE 0x30
#define BPF_JSET 0x40
#define BPF_SRC(code) ((code) & 0x08)
#define BPF_K 0x00
#define BPF_X 0x08
/* ret - BPF_K and BPF_X also apply */
#define BPF_RVAL(code) ((code) & 0x18)
#define BPF_A 0x10
/* misc */
#define BPF_MISCOP(code) ((code) & 0xf8)
#define BPF_TAX 0x00
#define BPF_TXA 0x80
/*
* The instruction data structure.
*/
struct bpf_insn {
u_short code;
u_char jt;
u_char jf;
bpf_u_int32 k;
1994-05-24 10:09:53 +00:00
};
/*
* Macros for insn array initializers.
*/
#define BPF_STMT(code, k) { (u_short)(code), 0, 0, k }
#define BPF_JUMP(code, k, jt, jf) { (u_short)(code), jt, jf, k }
/*
* Structure to retrieve available DLTs for the interface.
*/
struct bpf_dltlist {
u_int bfl_len; /* number of bfd_list array */
u_int *bfl_list; /* array of DLTs */
};
#ifdef _KERNEL
Introduce support for zero-copy BPF buffering, which reduces the overhead of packet capture by allowing a user process to directly "loan" buffer memory to the kernel rather than using read(2) to explicitly copy data from kernel address space. The user process will issue new BPF ioctls to set the shared memory buffer mode and provide pointers to buffers and their size. The kernel then wires and maps the pages into kernel address space using sf_buf(9), which on supporting architectures will use the direct map region. The current "buffered" access mode remains the default, and support for zero-copy buffers must, for the time being, be explicitly enabled using a sysctl for the kernel to accept requests to use it. The kernel and user process synchronize use of the buffers with atomic operations, avoiding the need for system calls under load; the user process may use select()/poll()/kqueue() to manage blocking while waiting for network data if the user process is able to consume data faster than the kernel generates it. Patchs to libpcap are available to allow libpcap applications to transparently take advantage of this support. Detailed information on the new API may be found in bpf(4), including specific atomic operations and memory barriers required to synchronize buffer use safely. These changes modify the base BPF implementation to (roughly) abstrac the current buffer model, allowing the new shared memory model to be added, and add new monitoring statistics for netstat to print. The implementation, with the exception of some monitoring hanges that break the netstat monitoring ABI for BPF, will be MFC'd. Zerocopy bpf buffers are still considered experimental are disabled by default. To experiment with this new facility, adjust the net.bpf.zerocopy_enable sysctl variable to 1. Changes to libpcap will be made available as a patch for the time being, and further refinements to the implementation are expected. Sponsored by: Seccuris Inc. In collaboration with: rwatson Tested by: pwood, gallatin MFC after: 4 months [1] [1] Certain portions will probably not be MFCed, specifically things that can break the monitoring ABI.
2008-03-24 13:49:17 +00:00
#ifdef MALLOC_DECLARE
MALLOC_DECLARE(M_BPF);
#endif
#ifdef SYSCTL_DECL
SYSCTL_DECL(_net_bpf);
#endif
/*
* Rotate the packet buffers in descriptor d. Move the store buffer into the
* hold slot, and the free buffer ino the store slot. Zero the length of the
* new store buffer. Descriptor lock should be held.
*/
#define ROTATE_BUFFERS(d) do { \
(d)->bd_hbuf = (d)->bd_sbuf; \
(d)->bd_hlen = (d)->bd_slen; \
(d)->bd_sbuf = (d)->bd_fbuf; \
(d)->bd_slen = 0; \
(d)->bd_fbuf = NULL; \
bpf_bufheld(d); \
} while (0)
Fix the following bpf(4) race condition which can result in a panic: (1) bpf peer attaches to interface netif0 (2) Packet is received by netif0 (3) ifp->if_bpf pointer is checked and handed off to bpf (4) bpf peer detaches from netif0 resulting in ifp->if_bpf being initialized to NULL. (5) ifp->if_bpf is dereferenced by bpf machinery (6) Kaboom This race condition likely explains the various different kernel panics reported around sending SIGINT to tcpdump or dhclient processes. But really this race can result in kernel panics anywhere you have frequent bpf attach and detach operations with high packet per second load. Summary of changes: - Remove the bpf interface's "driverp" member - When we attach bpf interfaces, we now set the ifp->if_bpf member to the bpf interface structure. Once this is done, ifp->if_bpf should never be NULL. [1] - Introduce bpf_peers_present function, an inline operation which will do a lockless read bpf peer list associated with the interface. It should be noted that the bpf code will pickup the bpf_interface lock before adding or removing bpf peers. This should serialize the access to the bpf descriptor list, removing the race. - Expose the bpf_if structure in bpf.h so that the bpf_peers_present function can use it. This also removes the struct bpf_if; hack that was there. - Adjust all consumers of the raw if_bpf structure to use bpf_peers_present Now what happens is: (1) Packet is received by netif0 (2) Check to see if bpf descriptor list is empty (3) Pickup the bpf interface lock (4) Hand packet off to process From the attach/detach side: (1) Pickup the bpf interface lock (2) Add/remove from bpf descriptor list Now that we are storing the bpf interface structure with the ifnet, there is is no need to walk the bpf interface list to locate the correct bpf interface. We now simply look up the interface, and initialize the pointer. This has a nice side effect of changing a bpf interface attach operation from O(N) (where N is the number of bpf interfaces), to O(1). [1] From now on, we can no longer check ifp->if_bpf to tell us whether or not we have any bpf peers that might be interested in receiving packets. In collaboration with: sam@ MFC after: 1 month
2006-06-02 19:59:33 +00:00
/*
* Descriptor associated with each attached hardware interface.
*/
struct bpf_if {
LIST_ENTRY(bpf_if) bif_next; /* list of all interfaces */
LIST_HEAD(, bpf_d) bif_dlist; /* descriptor list */
u_int bif_dlt; /* link layer type */
u_int bif_hdrlen; /* length of header (with padding) */
struct ifnet *bif_ifp; /* corresponding interface */
struct mtx bif_mtx; /* mutex for interface */
};
Introduce support for zero-copy BPF buffering, which reduces the overhead of packet capture by allowing a user process to directly "loan" buffer memory to the kernel rather than using read(2) to explicitly copy data from kernel address space. The user process will issue new BPF ioctls to set the shared memory buffer mode and provide pointers to buffers and their size. The kernel then wires and maps the pages into kernel address space using sf_buf(9), which on supporting architectures will use the direct map region. The current "buffered" access mode remains the default, and support for zero-copy buffers must, for the time being, be explicitly enabled using a sysctl for the kernel to accept requests to use it. The kernel and user process synchronize use of the buffers with atomic operations, avoiding the need for system calls under load; the user process may use select()/poll()/kqueue() to manage blocking while waiting for network data if the user process is able to consume data faster than the kernel generates it. Patchs to libpcap are available to allow libpcap applications to transparently take advantage of this support. Detailed information on the new API may be found in bpf(4), including specific atomic operations and memory barriers required to synchronize buffer use safely. These changes modify the base BPF implementation to (roughly) abstrac the current buffer model, allowing the new shared memory model to be added, and add new monitoring statistics for netstat to print. The implementation, with the exception of some monitoring hanges that break the netstat monitoring ABI for BPF, will be MFC'd. Zerocopy bpf buffers are still considered experimental are disabled by default. To experiment with this new facility, adjust the net.bpf.zerocopy_enable sysctl variable to 1. Changes to libpcap will be made available as a patch for the time being, and further refinements to the implementation are expected. Sponsored by: Seccuris Inc. In collaboration with: rwatson Tested by: pwood, gallatin MFC after: 4 months [1] [1] Certain portions will probably not be MFCed, specifically things that can break the monitoring ABI.
2008-03-24 13:49:17 +00:00
void bpf_bufheld(struct bpf_d *d);
2002-03-19 21:54:18 +00:00
int bpf_validate(const struct bpf_insn *, int);
void bpf_tap(struct bpf_if *, u_char *, u_int);
void bpf_mtap(struct bpf_if *, struct mbuf *);
void bpf_mtap2(struct bpf_if *, void *, u_int, struct mbuf *);
2002-03-19 21:54:18 +00:00
void bpfattach(struct ifnet *, u_int, u_int);
void bpfattach2(struct ifnet *, u_int, u_int, struct bpf_if **);
2002-03-19 21:54:18 +00:00
void bpfdetach(struct ifnet *);
2002-03-19 21:54:18 +00:00
void bpfilterattach(int);
u_int bpf_filter(const struct bpf_insn *, u_char *, u_int, u_int);
Fix the following bpf(4) race condition which can result in a panic: (1) bpf peer attaches to interface netif0 (2) Packet is received by netif0 (3) ifp->if_bpf pointer is checked and handed off to bpf (4) bpf peer detaches from netif0 resulting in ifp->if_bpf being initialized to NULL. (5) ifp->if_bpf is dereferenced by bpf machinery (6) Kaboom This race condition likely explains the various different kernel panics reported around sending SIGINT to tcpdump or dhclient processes. But really this race can result in kernel panics anywhere you have frequent bpf attach and detach operations with high packet per second load. Summary of changes: - Remove the bpf interface's "driverp" member - When we attach bpf interfaces, we now set the ifp->if_bpf member to the bpf interface structure. Once this is done, ifp->if_bpf should never be NULL. [1] - Introduce bpf_peers_present function, an inline operation which will do a lockless read bpf peer list associated with the interface. It should be noted that the bpf code will pickup the bpf_interface lock before adding or removing bpf peers. This should serialize the access to the bpf descriptor list, removing the race. - Expose the bpf_if structure in bpf.h so that the bpf_peers_present function can use it. This also removes the struct bpf_if; hack that was there. - Adjust all consumers of the raw if_bpf structure to use bpf_peers_present Now what happens is: (1) Packet is received by netif0 (2) Check to see if bpf descriptor list is empty (3) Pickup the bpf interface lock (4) Hand packet off to process From the attach/detach side: (1) Pickup the bpf interface lock (2) Add/remove from bpf descriptor list Now that we are storing the bpf interface structure with the ifnet, there is is no need to walk the bpf interface list to locate the correct bpf interface. We now simply look up the interface, and initialize the pointer. This has a nice side effect of changing a bpf interface attach operation from O(N) (where N is the number of bpf interfaces), to O(1). [1] From now on, we can no longer check ifp->if_bpf to tell us whether or not we have any bpf peers that might be interested in receiving packets. In collaboration with: sam@ MFC after: 1 month
2006-06-02 19:59:33 +00:00
static __inline int
bpf_peers_present(struct bpf_if *bpf)
{
if (!LIST_EMPTY(&bpf->bif_dlist))
return (1);
return (0);
Fix the following bpf(4) race condition which can result in a panic: (1) bpf peer attaches to interface netif0 (2) Packet is received by netif0 (3) ifp->if_bpf pointer is checked and handed off to bpf (4) bpf peer detaches from netif0 resulting in ifp->if_bpf being initialized to NULL. (5) ifp->if_bpf is dereferenced by bpf machinery (6) Kaboom This race condition likely explains the various different kernel panics reported around sending SIGINT to tcpdump or dhclient processes. But really this race can result in kernel panics anywhere you have frequent bpf attach and detach operations with high packet per second load. Summary of changes: - Remove the bpf interface's "driverp" member - When we attach bpf interfaces, we now set the ifp->if_bpf member to the bpf interface structure. Once this is done, ifp->if_bpf should never be NULL. [1] - Introduce bpf_peers_present function, an inline operation which will do a lockless read bpf peer list associated with the interface. It should be noted that the bpf code will pickup the bpf_interface lock before adding or removing bpf peers. This should serialize the access to the bpf descriptor list, removing the race. - Expose the bpf_if structure in bpf.h so that the bpf_peers_present function can use it. This also removes the struct bpf_if; hack that was there. - Adjust all consumers of the raw if_bpf structure to use bpf_peers_present Now what happens is: (1) Packet is received by netif0 (2) Check to see if bpf descriptor list is empty (3) Pickup the bpf interface lock (4) Hand packet off to process From the attach/detach side: (1) Pickup the bpf interface lock (2) Add/remove from bpf descriptor list Now that we are storing the bpf interface structure with the ifnet, there is is no need to walk the bpf interface list to locate the correct bpf interface. We now simply look up the interface, and initialize the pointer. This has a nice side effect of changing a bpf interface attach operation from O(N) (where N is the number of bpf interfaces), to O(1). [1] From now on, we can no longer check ifp->if_bpf to tell us whether or not we have any bpf peers that might be interested in receiving packets. In collaboration with: sam@ MFC after: 1 month
2006-06-02 19:59:33 +00:00
}
#define BPF_TAP(_ifp,_pkt,_pktlen) do { \
Fix the following bpf(4) race condition which can result in a panic: (1) bpf peer attaches to interface netif0 (2) Packet is received by netif0 (3) ifp->if_bpf pointer is checked and handed off to bpf (4) bpf peer detaches from netif0 resulting in ifp->if_bpf being initialized to NULL. (5) ifp->if_bpf is dereferenced by bpf machinery (6) Kaboom This race condition likely explains the various different kernel panics reported around sending SIGINT to tcpdump or dhclient processes. But really this race can result in kernel panics anywhere you have frequent bpf attach and detach operations with high packet per second load. Summary of changes: - Remove the bpf interface's "driverp" member - When we attach bpf interfaces, we now set the ifp->if_bpf member to the bpf interface structure. Once this is done, ifp->if_bpf should never be NULL. [1] - Introduce bpf_peers_present function, an inline operation which will do a lockless read bpf peer list associated with the interface. It should be noted that the bpf code will pickup the bpf_interface lock before adding or removing bpf peers. This should serialize the access to the bpf descriptor list, removing the race. - Expose the bpf_if structure in bpf.h so that the bpf_peers_present function can use it. This also removes the struct bpf_if; hack that was there. - Adjust all consumers of the raw if_bpf structure to use bpf_peers_present Now what happens is: (1) Packet is received by netif0 (2) Check to see if bpf descriptor list is empty (3) Pickup the bpf interface lock (4) Hand packet off to process From the attach/detach side: (1) Pickup the bpf interface lock (2) Add/remove from bpf descriptor list Now that we are storing the bpf interface structure with the ifnet, there is is no need to walk the bpf interface list to locate the correct bpf interface. We now simply look up the interface, and initialize the pointer. This has a nice side effect of changing a bpf interface attach operation from O(N) (where N is the number of bpf interfaces), to O(1). [1] From now on, we can no longer check ifp->if_bpf to tell us whether or not we have any bpf peers that might be interested in receiving packets. In collaboration with: sam@ MFC after: 1 month
2006-06-02 19:59:33 +00:00
if (bpf_peers_present((_ifp)->if_bpf)) \
bpf_tap((_ifp)->if_bpf, (_pkt), (_pktlen)); \
} while (0)
#define BPF_MTAP(_ifp,_m) do { \
Fix the following bpf(4) race condition which can result in a panic: (1) bpf peer attaches to interface netif0 (2) Packet is received by netif0 (3) ifp->if_bpf pointer is checked and handed off to bpf (4) bpf peer detaches from netif0 resulting in ifp->if_bpf being initialized to NULL. (5) ifp->if_bpf is dereferenced by bpf machinery (6) Kaboom This race condition likely explains the various different kernel panics reported around sending SIGINT to tcpdump or dhclient processes. But really this race can result in kernel panics anywhere you have frequent bpf attach and detach operations with high packet per second load. Summary of changes: - Remove the bpf interface's "driverp" member - When we attach bpf interfaces, we now set the ifp->if_bpf member to the bpf interface structure. Once this is done, ifp->if_bpf should never be NULL. [1] - Introduce bpf_peers_present function, an inline operation which will do a lockless read bpf peer list associated with the interface. It should be noted that the bpf code will pickup the bpf_interface lock before adding or removing bpf peers. This should serialize the access to the bpf descriptor list, removing the race. - Expose the bpf_if structure in bpf.h so that the bpf_peers_present function can use it. This also removes the struct bpf_if; hack that was there. - Adjust all consumers of the raw if_bpf structure to use bpf_peers_present Now what happens is: (1) Packet is received by netif0 (2) Check to see if bpf descriptor list is empty (3) Pickup the bpf interface lock (4) Hand packet off to process From the attach/detach side: (1) Pickup the bpf interface lock (2) Add/remove from bpf descriptor list Now that we are storing the bpf interface structure with the ifnet, there is is no need to walk the bpf interface list to locate the correct bpf interface. We now simply look up the interface, and initialize the pointer. This has a nice side effect of changing a bpf interface attach operation from O(N) (where N is the number of bpf interfaces), to O(1). [1] From now on, we can no longer check ifp->if_bpf to tell us whether or not we have any bpf peers that might be interested in receiving packets. In collaboration with: sam@ MFC after: 1 month
2006-06-02 19:59:33 +00:00
if (bpf_peers_present((_ifp)->if_bpf)) { \
M_ASSERTVALID(_m); \
bpf_mtap((_ifp)->if_bpf, (_m)); \
} \
} while (0)
#define BPF_MTAP2(_ifp,_data,_dlen,_m) do { \
Fix the following bpf(4) race condition which can result in a panic: (1) bpf peer attaches to interface netif0 (2) Packet is received by netif0 (3) ifp->if_bpf pointer is checked and handed off to bpf (4) bpf peer detaches from netif0 resulting in ifp->if_bpf being initialized to NULL. (5) ifp->if_bpf is dereferenced by bpf machinery (6) Kaboom This race condition likely explains the various different kernel panics reported around sending SIGINT to tcpdump or dhclient processes. But really this race can result in kernel panics anywhere you have frequent bpf attach and detach operations with high packet per second load. Summary of changes: - Remove the bpf interface's "driverp" member - When we attach bpf interfaces, we now set the ifp->if_bpf member to the bpf interface structure. Once this is done, ifp->if_bpf should never be NULL. [1] - Introduce bpf_peers_present function, an inline operation which will do a lockless read bpf peer list associated with the interface. It should be noted that the bpf code will pickup the bpf_interface lock before adding or removing bpf peers. This should serialize the access to the bpf descriptor list, removing the race. - Expose the bpf_if structure in bpf.h so that the bpf_peers_present function can use it. This also removes the struct bpf_if; hack that was there. - Adjust all consumers of the raw if_bpf structure to use bpf_peers_present Now what happens is: (1) Packet is received by netif0 (2) Check to see if bpf descriptor list is empty (3) Pickup the bpf interface lock (4) Hand packet off to process From the attach/detach side: (1) Pickup the bpf interface lock (2) Add/remove from bpf descriptor list Now that we are storing the bpf interface structure with the ifnet, there is is no need to walk the bpf interface list to locate the correct bpf interface. We now simply look up the interface, and initialize the pointer. This has a nice side effect of changing a bpf interface attach operation from O(N) (where N is the number of bpf interfaces), to O(1). [1] From now on, we can no longer check ifp->if_bpf to tell us whether or not we have any bpf peers that might be interested in receiving packets. In collaboration with: sam@ MFC after: 1 month
2006-06-02 19:59:33 +00:00
if (bpf_peers_present((_ifp)->if_bpf)) { \
M_ASSERTVALID(_m); \
bpf_mtap2((_ifp)->if_bpf,(_data),(_dlen),(_m)); \
} \
} while (0)
1994-05-24 10:09:53 +00:00
#endif
/*
* Number of scratch memory words (for BPF_LD|BPF_MEM and BPF_ST).
*/
#define BPF_MEMWORDS 16
#endif /* _NET_BPF_H_ */