IPL(4) Device Drivers Manual IPL(4)

ipl - IP packet log device

The ipl pseudo device's purpose is to provide an easy way to gather packet headers of packets you wish to log. If a packet header is to be logged, the entire header is logged (including any IP options - TCP/UDP options are not included when it calculates header size) or not at all. The packet contents is also logged after the header.

Prepending every packet header logged is a structure containing information relevant to the packet following and why it was logged. The structure's format is as follows:

struct ipl_ci   {


        u_long  sec;      /* time when the packet was logged */


        u_long  usec;


        u_long  plen;     /* length of packet data logged */


        u_short hlen;     /* length of headers logged */


        u_short rule;     /* rule number (for log ...) or 0 if result = log */


        u_long  flags:24; /* XXX FIXME do we care about the extra bytes? */
#if (defined(OpenBSD) && (OpenBSD <= 1991011) && (OpenBSD >= 199606))


        u_long  filler:8; /* XXX FIXME do we care? */


        u_char  ifname[IFNAMSIZ];
#else


        u_long  unit:8;


        u_char  ifname[4];
#endif
};



In the case of the header causing the buffer to finish on a non-32bit
boundary, padding will be `appended' to ensure that the next log entry
is aligned to a 32bit boundary.
If the packet contents is more then 128 bytes, then only 128 bytes of the
packet contents is logged. Should the packet contents finish on a non-32bit
boundary, then the last few bytes are not logged to ensure the log entry
is aligned to a 32bit boundary.
ipl is a read-only (sequential) character pseudo-device.
The ioctls which are loaded with this device can be found under ipf(4).
The only ioctl which is used for logging and doesn't affect the filter is:




        ioctl(fd, SIOCIPFFB, int *)

This ioctl flushes the log buffer and returns the number of bytes flushed.

There is currently no support for non-blocking IO with this device, meaning all read operations should be considered blocking in nature (if there is no data to read, it will sleep until some is made available).

ipf(4)

Packet headers are dropped when the internal buffer (static size) fills.

/dev/ipl0