/* pppdump.c
 *
 * Copyright (c) 2000 by Gilbert Ramirez <gram@alumni.rice.edu>
 *
 * SPDX-License-Identifier: GPL-2.0-or-later
 */

#include "config.h"
#include "pppdump.h"
#include "wtap-int.h"
#include "file_wrappers.h"

#include <stdlib.h>
#include <errno.h>
#include <string.h>

#include <wsutil/array.h>
#include <wsutil/ws_assert.h>
#include <wsutil/pint.h>

/*
pppdump records
Daniel Thompson (STMicroelectronics) <daniel.thompson@st.com>

+------+
| 0x07 |                              Reset time
+------+------+------+------+
|  t3  |  t2  |  t1  |  t0  |         t = time_t
+------+------+------+------+

+------+
| 0x06 |                              Time step (short)
+------+
|  ts  |                              ts = time step (tenths of seconds)
+------+

+------+
| 0x05 |                              Time step (long)
+------+------+------+------+
| ts3  | ts2  | ts1  | ts0  |         ts = time step (tenths of seconds)
+------+------+------+------+

+------+
| 0x04 |                              Receive delimiter (not seen in practice)
+------+

+------+
| 0x03 |                              Send delimiter (not seen in practice)
+------+

+------+
| 0x02 |                              Received data
+------+------+
|  n1  |  n0  |                       n = number of bytes following
+------+------+
|    data     |
|             |

+------+
| 0x01 |                              Sent data
+------+------+
|  n1  |  n0  |                       n = number of bytes following
+------+------+
|    data     |
|             |
*/

#define PPPD_SENT_DATA		0x01
#define PPPD_RECV_DATA		0x02
#define PPPD_SEND_DELIM		0x03
#define PPPD_RECV_DELIM		0x04
#define PPPD_TIME_STEP_LONG	0x05
#define PPPD_TIME_STEP_SHORT	0x06
#define PPPD_RESET_TIME		0x07

/* this buffer must be at least (2*PPPD_MTU) + sizeof(ppp_header) +
 * sizeof(lcp_header) + sizeof(ipcp_header).  PPPD_MTU is *very* rarely
 * larger than 1500 so this value is fine.
 *
 * It's less than WTAP_MAX_PACKET_SIZE_STANDARD, so we don't have to worry about
 * too-large packets.
 */
#define PPPD_BUF_SIZE		8192

typedef enum {
	DIRECTION_SENT,
	DIRECTION_RECV
} direction_enum;

static bool pppdump_read(wtap *wth, wtap_rec *rec,
	int *err, char **err_info, int64_t *data_offset);
static bool pppdump_seek_read(wtap *wth, int64_t seek_off, wtap_rec *rec,
        int *err, char **err_info);

static int pppdump_file_type_subtype = -1;

void register_pppdump(void);

/*
 * Information saved about a packet, during the initial sequential pass
 * through the file, to allow us to later re-read it when randomly
 * reading packets.
 *
 * "offset" is the offset in the file of the first data chunk containing data
 * from that packet; note that it may also contain data from previous
 * packets.
 *
 * "num_bytes_to_skip" is the number of bytes from previous packets in that
 * first data chunk.
 *
 * "dir" is the direction of the packet.
 */
typedef struct {
	int64_t		offset;
	int64_t		num_bytes_to_skip;
	direction_enum	dir;
} pkt_id;

/*
 * Information about a packet currently being processed.  There is one of
 * these for the sent packet being processed and one of these for the
 * received packet being processed, as we could be in the middle of
 * processing both a received packet and a sent packet.
 *
 * "dir" is the direction of the packet.
 *
 * "cnt" is the number of bytes of packet data we've accumulated.
 *
 * "esc" is true if the next byte we see is escaped (and thus must be XORed
 * with 0x20 before saving it), false otherwise.
 *
 * "buf" is a buffer containing the packet data we've accumulated.
 *
 * "id_offset" is the offset in the file of the first data chunk
 * containing data from the packet we're processing.
 *
 * "sd_offset" is the offset in the file of the first data byte from
 * the packet we're processing - which isn't necessarily right after
 * the header of the first data chunk, as we may already have assembled
 * packets from that chunk.
 *
 * "cd_offset" is the offset in the file of the current data chunk we're
 * processing.
 */
typedef struct {
	direction_enum	dir;
	int		cnt;
	bool	esc;
	uint8_t		buf[PPPD_BUF_SIZE];
	int64_t		id_offset;
	int64_t		sd_offset;
	int64_t		cd_offset;
} pkt_t;

/*
 * This keeps state used while processing records.
 *
 * "timestamp" is the seconds portion of the current time stamp value,
 * as updated from PPPD_RESET_TIME, PPPD_TIME_STEP_LONG, and
 * PPPD_TIME_STEP_SHORT records.  "tenths" is the tenths-of-seconds
 * portion.
 *
 * "spkt" and "rpkt" are "pkt_t" structures for the sent and received
 * packets we're currently working on.
 *
 * "offset" is the current offset in the file.
 *
 * "num_bytes" and "pkt" are information saved when we finish accumulating
 * the data for a packet, if the data chunk we're working on still has more
 * data in it:
 *
 *	"num_bytes" is the number of bytes of additional data remaining
 *	in the chunk after we've finished accumulating the data for the
 *	packet.
 *
 *	"pkt" is the "pkt_t" for the type of packet the data chunk is for
 *	(sent or received packet).
 *
 * "seek_state" is another state structure used while processing records
 * when doing a seek-and-read.  (That structure doesn't itself have a
 * "seek_state" structure.)
 *
 * "pids" is a GPtrArray of pointers to "pkt_id" structures for all the
 * packets we've seen during the initial sequential pass, to allow us to
 * later retrieve them with random accesses.
 *
 * "pkt_cnt" is the number of packets we've seen up to this point in the
 * sequential pass.
 */
typedef struct _pppdump_t {
	time_t			timestamp;
	unsigned			tenths;
	pkt_t			spkt;
	pkt_t			rpkt;
	int64_t			offset;
	int			num_bytes;
	pkt_t			*pkt;
	struct _pppdump_t	*seek_state;
	GPtrArray		*pids;
	unsigned			pkt_cnt;
} pppdump_t;

static int
process_data(pppdump_t *state, FILE_T fh, pkt_t *pkt, int n, uint8_t *pd,
    int *err, char **err_info, pkt_id *pid);

static bool
collate(pppdump_t *state, FILE_T fh, int *err, char **err_info, uint8_t *pd,
		int *num_bytes, direction_enum *direction, pkt_id *pid,
		int64_t num_bytes_to_skip);

static void
pppdump_close(wtap *wth);

static void
init_state(pppdump_t *state)
{

	state->num_bytes = 0;
	state->pkt = NULL;

	state->spkt.dir = DIRECTION_SENT;
	state->spkt.cnt = 0;
	state->spkt.esc = false;
	state->spkt.id_offset = 0;
	state->spkt.sd_offset = 0;
	state->spkt.cd_offset = 0;

	state->rpkt.dir = DIRECTION_RECV;
	state->rpkt.cnt = 0;
	state->rpkt.esc = false;
	state->rpkt.id_offset = 0;
	state->rpkt.sd_offset = 0;
	state->rpkt.cd_offset = 0;

	state->seek_state = NULL;
	state->offset = 0x100000; /* to detect errors during development */
}


wtap_open_return_val
pppdump_open(wtap *wth, int *err, char **err_info)
{
	uint8_t		buffer[6];	/* Looking for: 0x07 t3 t2 t1 t0 ID */
	pppdump_t	*state;

	/* There is no file header, only packet records. Fortunately for us,
	* timestamp records are separated from packet records, so we should
	* find an "initial time stamp" (i.e., a "reset time" record, or
	* record type 0x07) at the beginning of the file. We'll check for
	* that, plus a valid record following the 0x07 and the four bytes
	* representing the timestamp.
	*/

	if (!wtap_read_bytes(wth->fh, buffer, sizeof(buffer),
	    err, err_info)) {
		if (*err != WTAP_ERR_SHORT_READ)
			return WTAP_OPEN_ERROR;
		return WTAP_OPEN_NOT_MINE;
	}

	if (buffer[0] == PPPD_RESET_TIME &&
			(buffer[5] == PPPD_SENT_DATA ||
			 buffer[5] == PPPD_RECV_DATA ||
			 buffer[5] == PPPD_TIME_STEP_LONG ||
			 buffer[5] == PPPD_TIME_STEP_SHORT ||
			 buffer[5] == PPPD_RESET_TIME)) {

		goto my_file_type;
	}
	else {
		return WTAP_OPEN_NOT_MINE;
	}

  my_file_type:

	if (file_seek(wth->fh, 5, SEEK_SET, err) == -1)
		return WTAP_OPEN_ERROR;

	state = g_new(pppdump_t, 1);
	wth->priv = (void *)state;
	state->timestamp = pntohu32(&buffer[1]);
	state->tenths = 0;

	init_state(state);

	state->offset = 5;
	wth->file_encap = WTAP_ENCAP_PPP_WITH_PHDR;
	wth->file_type_subtype = pppdump_file_type_subtype;

	wth->snapshot_length = PPPD_BUF_SIZE; /* just guessing */
	wth->subtype_read = pppdump_read;
	wth->subtype_seek_read = pppdump_seek_read;
	wth->subtype_close = pppdump_close;
	wth->file_tsprec = WTAP_TSPREC_100_MSEC;

	state->seek_state = g_new(pppdump_t,1);

	/* If we have a random stream open, we're going to be reading
	   the file randomly; set up a GPtrArray of pointers to
	   information about how to retrieve the data for each packet. */
	if (wth->random_fh != NULL)
		state->pids = g_ptr_array_new();
	else
		state->pids = NULL;
	state->pkt_cnt = 0;

	/*
	 * Add an IDB; we don't know how many interfaces were
	 * involved, so we just say one interface, about which
	 * we only know the link-layer type, snapshot length,
	 * and time stamp resolution.
	 */
	wtap_add_generated_idb(wth);

	return WTAP_OPEN_MINE;
}

/* Set part of the struct wtap_rec. */
static void
pppdump_set_phdr(wtap *wth, wtap_rec *rec, int num_bytes, direction_enum direction)
{
	wtap_setup_packet_rec(rec, wth->file_encap);
	rec->block = wtap_block_create(WTAP_BLOCK_PACKET);
	rec->rec_header.packet_header.len = num_bytes;
	rec->rec_header.packet_header.caplen = num_bytes;

	rec->rec_header.packet_header.pseudo_header.p2p.sent = (direction == DIRECTION_SENT ? true : false);
}

/* Find the next packet and parse it; called from wtap_read(). */
static bool
pppdump_read(wtap *wth, wtap_rec *rec, int *err, char **err_info,
    int64_t *data_offset)
{
	int		num_bytes;
	direction_enum	direction;
	pppdump_t	*state;
	pkt_id		*pid;

	state = (pppdump_t *)wth->priv;

	/* If we have a random stream open, allocate a structure to hold
	   the information needed to read this packet's data again. */
	if (wth->random_fh != NULL) {
		pid = g_new(pkt_id, 1);
		if (!pid) {
			*err = errno;	/* assume a malloc failed and set "errno" */
			return false;
		}
		pid->offset = 0;
	} else
		pid = NULL;	/* sequential only */

	ws_buffer_assure_space(&rec->data, PPPD_BUF_SIZE);
	if (!collate(state, wth->fh, err, err_info, ws_buffer_start_ptr(&rec->data),
	    &num_bytes, &direction, pid, 0)) {
		g_free(pid);
		return false;
	}

	if (pid != NULL)
		pid->dir = direction;

	if (pid != NULL)
		g_ptr_array_add(state->pids, pid);
	/* The user's data_offset is not really an offset, but a packet number. */
	*data_offset = state->pkt_cnt;
	state->pkt_cnt++;

	pppdump_set_phdr(wth, rec, num_bytes, direction);
	rec->presence_flags = WTAP_HAS_TS;
	rec->ts.secs = state->timestamp;
	rec->ts.nsecs = state->tenths * 100000000;

	return true;
}

/* Returns number of bytes copied for record, -1 if failure.
 *
 * This is modeled after pppdump.c, the utility to parse pppd log files; it
 * comes with the ppp distribution.
 */
static int
process_data(pppdump_t *state, FILE_T fh, pkt_t *pkt, int n, uint8_t *pd,
    int *err, char **err_info, pkt_id *pid)
{
	int	c;
	int	num_bytes = n;
	int	num_written;

	for (; num_bytes > 0; --num_bytes) {
		c = file_getc(fh);
		if (c == EOF) {
			*err = file_error(fh, err_info);
			if (*err == 0) {
				*err = WTAP_ERR_SHORT_READ;
			}
			return -1;
		}
		state->offset++;
		switch (c) {
			case 0x7e:
				/*
				 * Flag Sequence for RFC 1662 HDLC-like
				 * framing.
				 *
				 * As this is a raw trace of octets going
				 * over the wire, and that might include
				 * the login sequence, there is no
				 * guarantee that *only* PPP traffic
				 * appears in this file, so there is no
				 * guarantee that the first 0x7e we see is
				 * a start flag sequence, and therefore we
				 * cannot safely ignore all characters up
				 * to the first 0x7e, and therefore we
				 * might end up with some bogus PPP
				 * packets.
				 */
				if (pkt->cnt > 0) {
					/*
					 * We've seen stuff before this,
					 * so this is the end of a frame.
					 * Make a frame out of that stuff.
					 */
					pkt->esc = false;

					num_written = pkt->cnt;
					pkt->cnt = 0;
					if (num_written <= 0) {
						return 0;
					}

					if (num_written > PPPD_BUF_SIZE) {
						*err = WTAP_ERR_BAD_FILE;
						*err_info = ws_strdup_printf("pppdump: File has %u-byte packet, bigger than maximum of %u",
						    num_written, PPPD_BUF_SIZE);
						return -1;
					}

					memcpy(pd, pkt->buf, num_written);

					/*
					 * Remember the offset of the
					 * first record containing data
					 * for this packet, and how far
					 * into that record to skip to
					 * get to the beginning of the
					 * data for this packet; the number
					 * of bytes to skip into that record
					 * is the file offset of the first
					 * byte of this packet minus the
					 * file offset of the first byte of
					 * this record, minus 3 bytes for the
					 * header of this record (which, if
					 * we re-read this record, we will
					 * process, not skip).
					 */
					if (pid) {
						pid->offset = pkt->id_offset;
						pid->num_bytes_to_skip =
						    pkt->sd_offset - pkt->id_offset - 3;
						ws_assert(pid->num_bytes_to_skip >= 0);
					}

					num_bytes--;
					if (num_bytes > 0) {
						/*
						 * There's more data in this
						 * record.
						 * Set the initial data offset
						 * for the next packet.
						 */
						pkt->id_offset = pkt->cd_offset;
						pkt->sd_offset = state->offset;
					} else {
						/*
						 * There is no more data in
						 * this record.
						 * Thus, we don't have the
						 * initial data offset for
						 * the next packet.
						 */
						pkt->id_offset = 0;
						pkt->sd_offset = 0;
					}
					state->num_bytes = num_bytes;
					state->pkt = pkt;
					return num_written;
				}
				break;

			case 0x7d:
				/*
				 * Control Escape octet for octet-stuffed
				 * RFC 1662 HDLC-like framing.
				 */
				if (!pkt->esc) {
					/*
					 * Control Escape not preceded by
					 * Control Escape; discard it
					 * but XOR the next octet with
					 * 0x20.
					 */
					pkt->esc = true;
					break;
				}
				/*
				 * Control Escape preceded by Control Escape;
				 * treat it as an ordinary character,
				 * by falling through.
				 */

			/* FALL THROUGH */
			default:
				if (pkt->esc) {
					/*
					 * This character was preceded by
					 * Control Escape, so XOR it with
					 * 0x20, as per RFC 1662's octet-
					 * stuffed framing, and clear
					 * the flag saying that the
					 * character should be escaped.
					 */
					c ^= 0x20;
					pkt->esc = false;
				}

				if (pkt->cnt >= PPPD_BUF_SIZE) {
					*err = WTAP_ERR_BAD_FILE;
					*err_info = ws_strdup_printf("pppdump: File has %u-byte packet, bigger than maximum of %u",
					    pkt->cnt - 1, PPPD_BUF_SIZE);
					return -1;
				}
				pkt->buf[pkt->cnt++] = c;
				break;
		}
	}

	/* we could have run out of bytes to read */
	return 0;
}

/* Returns true if packet data copied, false if error occurred or EOF (no more records). */
static bool
collate(pppdump_t* state, FILE_T fh, int *err, char **err_info, uint8_t *pd,
		int *num_bytes, direction_enum *direction, pkt_id *pid,
		int64_t num_bytes_to_skip)
{
	int		id;
	pkt_t		*pkt = NULL;
	int		byte0, byte1;
	int		n, num_written = 0;
	int64_t		start_offset;
	uint32_t		time_long;
	uint8_t		time_short;

	/*
	 * Process any data left over in the current record when doing
	 * sequential processing.
	 */
	if (state->num_bytes > 0) {
		ws_assert(num_bytes_to_skip == 0);
		pkt = state->pkt;
		num_written = process_data(state, fh, pkt, state->num_bytes,
		    pd, err, err_info, pid);

		if (num_written < 0) {
			return false;
		}
		else if (num_written > 0) {
			*num_bytes = num_written;
			*direction = pkt->dir;
			return true;
		}
		/* if 0 bytes written, keep processing */
	} else {
		/*
		 * We didn't have any data left over, so the packet will
		 * start at the beginning of a record.
		 */
		if (pid)
			pid->num_bytes_to_skip = 0;
	}

	/*
	 * That didn't get all the data for this packet, so process
	 * subsequent records.
	 */
	start_offset = state->offset;
	while ((id = file_getc(fh)) != EOF) {
		state->offset++;
		switch (id) {
			case PPPD_SENT_DATA:
			case PPPD_RECV_DATA:
				pkt = id == PPPD_SENT_DATA ? &state->spkt : &state->rpkt;

				/*
				 * Save the offset of the beginning of
				 * the current record.
				 */
				pkt->cd_offset = state->offset - 1;

				/*
				 * Get the length of the record.
				 */
				byte0 = file_getc(fh);
				if (byte0 == EOF)
					goto done;
				state->offset++;
				byte1 = file_getc(fh);
				if (byte1 == EOF)
					goto done;
				state->offset++;
				n = (byte0 << 8) | byte1;

				if (pkt->id_offset == 0) {
					/*
					 * We don't have the initial data
					 * offset for this packet, which
					 * means this is the first
					 * data record for that packet.
					 * Save the offset of the
					 * beginning of that record and
					 * the offset of the first data
					 * byte in the packet, which is
					 * the first data byte in the
					 * record.
					 */
					pkt->id_offset = pkt->cd_offset;
					pkt->sd_offset = state->offset;
				}

				if (n == 0)
					continue;

				ws_assert(num_bytes_to_skip < n);
				while (num_bytes_to_skip) {
					if (file_getc(fh) == EOF)
						goto done;
					state->offset++;
					num_bytes_to_skip--;
					n--;
				}
				num_written = process_data(state, fh, pkt, n,
				    pd, err, err_info, pid);

				if (num_written < 0) {
					return false;
				}
				else if (num_written > 0) {
					*num_bytes = num_written;
					*direction = pkt->dir;
					return true;
				}
				/* if 0 bytes written, keep looping */
				break;

			case PPPD_SEND_DELIM:
			case PPPD_RECV_DELIM:
				/* What can we do? */
				break;

			case PPPD_RESET_TIME:
				if (!wtap_read_bytes(fh, &time_long, sizeof(uint32_t), err, err_info))
					return false;
				state->offset += sizeof(uint32_t);
				state->timestamp = pntohu32(&time_long);
				state->tenths = 0;
				break;

			case PPPD_TIME_STEP_LONG:
				if (!wtap_read_bytes(fh, &time_long, sizeof(uint32_t), err, err_info))
					return false;
				state->offset += sizeof(uint32_t);
				state->tenths += pntohu32(&time_long);

				if (state->tenths >= 10) {
					state->timestamp += state->tenths / 10;
					state->tenths = state->tenths % 10;
				}

				break;

			case PPPD_TIME_STEP_SHORT:
				if (!wtap_read_bytes(fh, &time_short, sizeof(uint8_t), err, err_info))
					return false;
				state->offset += sizeof(uint8_t);
				state->tenths += time_short;

				if (state->tenths >= 10) {
					state->timestamp += state->tenths / 10;
					state->tenths = state->tenths % 10;
				}

				break;

			default:
				/* XXX - bad file */
				*err = WTAP_ERR_BAD_FILE;
				*err_info = ws_strdup_printf("pppdump: bad ID byte 0x%02x", id);
				return false;
		}

	}

done:
	*err = file_error(fh, err_info);
	if (*err == 0) {
		if (state->offset != start_offset) {
			/*
			 * We read at least one byte, so we were working
			 * on a record; an EOF means that record was
			 * cut short.
			 */
			*err = WTAP_ERR_SHORT_READ;
		}
	}
	return false;
}



/* Used to read packets in random-access fashion */
static bool
pppdump_seek_read(wtap *wth,
		 int64_t seek_off,
		 wtap_rec *rec,
		 int *err,
		 char **err_info)
{
	int		num_bytes;
	uint8_t		*pd;
	direction_enum	direction;
	pppdump_t	*state;
	pkt_id		*pid;
	int64_t		num_bytes_to_skip;

	state = (pppdump_t *)wth->priv;

	pid = (pkt_id *)g_ptr_array_index(state->pids, seek_off);
	if (!pid) {
		*err = WTAP_ERR_BAD_FILE;	/* XXX - better error? */
		*err_info = g_strdup("pppdump: PID not found for record");
		return false;
	}

	if (file_seek(wth->random_fh, pid->offset, SEEK_SET, err) == -1)
		return false;

	init_state(state->seek_state);
	state->seek_state->offset = pid->offset;

	ws_buffer_assure_space(&rec->data, PPPD_BUF_SIZE);
	pd = ws_buffer_start_ptr(&rec->data);

	/*
	 * We'll start reading at the first record containing data from
	 * this packet; however, that doesn't mean "collate()" will
	 * stop only when we've read that packet, as there might be
	 * data for packets going in the other direction as well, and
	 * we might finish processing one of those packets before we
	 * finish processing the packet we're reading.
	 *
	 * Therefore, we keep reading until we get a packet that's
	 * going in the direction we want.
	 */
	num_bytes_to_skip = pid->num_bytes_to_skip;
	do {
		if (!collate(state->seek_state, wth->random_fh, err, err_info,
		    pd, &num_bytes, &direction, NULL, num_bytes_to_skip))
			return false;
		num_bytes_to_skip = 0;
	} while (direction != pid->dir);

	pppdump_set_phdr(wth, rec, num_bytes, pid->dir);

	return true;
}

static void
pppdump_close(wtap *wth)
{
	pppdump_t	*state;

	state = (pppdump_t *)wth->priv;

	if (state->seek_state) { /* should always be true */
		g_free(state->seek_state);
	}

	if (state->pids) {
		unsigned int i;
		for (i = 0; i < g_ptr_array_len(state->pids); i++) {
			g_free(g_ptr_array_index(state->pids, i));
		}
		g_ptr_array_free(state->pids, true);
	}
}

static const struct supported_block_type pppdump_blocks_supported[] = {
	/*
	 * We support packet blocks, with no comments or other options.
	 */
	{ WTAP_BLOCK_PACKET, MULTIPLE_BLOCKS_SUPPORTED, NO_OPTIONS_SUPPORTED }
};

static const struct file_type_subtype_info pppdump_info = {
	"pppd log (pppdump format)", "pppd", NULL, NULL,
	false, BLOCKS_SUPPORTED(pppdump_blocks_supported),
	NULL, NULL, NULL
};

void register_pppdump(void)
{
	pppdump_file_type_subtype = wtap_register_file_type_subtype(&pppdump_info);

	/*
	 * Register name for backwards compatibility with the
	 * wtap_filetypes table in Lua.
	 */
	wtap_register_backwards_compatibility_lua_name("PPPDUMP",
	    pppdump_file_type_subtype);
}

/*
 * Editor modelines  -  https://www.wireshark.org/tools/modelines.html
 *
 * Local variables:
 * c-basic-offset: 8
 * tab-width: 8
 * indent-tabs-mode: t
 * End:
 *
 * vi: set shiftwidth=8 tabstop=8 noexpandtab:
 * :indentSize=8:tabSize=8:noTabs=false:
 */