Undark (fork) - SQLite deleted and corrupted data recovery tool

undark

// Forked from http://pldaniels.com/undark/
// Build: gcc undark.c -o undark.exe -lws2_32 -s
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <stdint.h>
#include <sys/stat.h>
#include <fcntl.h>
#include <ctype.h>

#ifndef _WIN32
#include <sys/mman.h>
#else
// https://gist.github.com/r-lyeh-archived/bc29c8630dd778454001
#include <windows.h>

#define PROT_READ    0x1
#define PROT_WRITE   0x2
/* This flag is only available in WinXP+ */
#ifdef FILE_MAP_EXECUTE
#define PROT_EXEC    0x4
#else
#define PROT_EXEC        0x0
#define FILE_MAP_EXECUTE 0
#endif

#define MAP_SHARED    0x01
#define MAP_PRIVATE   0x02
#define MAP_ANONYMOUS 0x20
#define MAP_ANON      MAP_ANONYMOUS
#define MAP_FAILED    ((void *) -1)

#ifdef __USE_FILE_OFFSET64
# define DWORD_HI(x) (x >> 32)
# define DWORD_LO(x) ((x) & 0xffffffff)
#else
# define DWORD_HI(x) (0)
# define DWORD_LO(x) (x)
#endif

static void *mmap(void *start, size_t length, int prot, int flags, int fd, off_t offset) {
	if (prot & ~(PROT_READ | PROT_WRITE | PROT_EXEC))
		return MAP_FAILED;
	if (fd == -1) {
		if (!(flags & MAP_ANON) || offset)
			return MAP_FAILED;
	} else if (flags & MAP_ANON)
		return MAP_FAILED;

	DWORD flProtect;
	if (prot & PROT_WRITE) {
		if (prot & PROT_EXEC)
			flProtect = PAGE_EXECUTE_READWRITE;
		else
			flProtect = PAGE_READWRITE;
	} else if (prot & PROT_EXEC) {
		if (prot & PROT_READ)
			flProtect = PAGE_EXECUTE_READ;
		else if (prot & PROT_EXEC)
			flProtect = PAGE_EXECUTE;
	} else
		flProtect = PAGE_READONLY;

	off_t end = length + offset;
	HANDLE mmap_fd, h;
	if (fd == -1)
		mmap_fd = INVALID_HANDLE_VALUE;
	else
		mmap_fd = (HANDLE)_get_osfhandle(fd);
	h = CreateFileMapping(mmap_fd, NULL, flProtect, DWORD_HI(end), DWORD_LO(end), NULL);
	if (h == NULL)
		return MAP_FAILED;

	DWORD dwDesiredAccess;
	if (prot & PROT_WRITE)
		dwDesiredAccess = FILE_MAP_WRITE;
	else
		dwDesiredAccess = FILE_MAP_READ;
	if (prot & PROT_EXEC)
		dwDesiredAccess |= FILE_MAP_EXECUTE;
	if (flags & MAP_PRIVATE)
		dwDesiredAccess |= FILE_MAP_COPY;
	void *ret = MapViewOfFile(h, dwDesiredAccess, DWORD_HI(offset), DWORD_LO(offset), length);
	if (ret == NULL) {
		CloseHandle(h);
		ret = MAP_FAILED;
	}
	return ret;
}

static void munmap(void *addr, size_t length)
{
	UnmapViewOfFile(addr);
	/* ruh-ro, we leaked handle from CreateFileMapping() ... */
}

#undef DWORD_HI
#undef DWORD_LO

#endif

int varint_decode(uint64_t *result, char *varint_p, char **end) {
	char *p;
	int shift;
	int length;
	uint64_t value;

	p = varint_p;
	length = 0;
	value = 0;
	shift = 0;
	for (;;) {
		value <<= shift;
		value |= ((*p & 0x7f));
		length++;
		if ((*p & 0x80) == 0x0) {
			break;
		}
		p++;
		shift += 7;
	}

	if (end != NULL) {
		*end = ++p;
	}

	*result = value;

	return length;
}

char to_signed_byte(unsigned char value) {
	int signed_value = value;
	if (value >> 7) signed_value |= -1 << 7;
	return signed_value;
}

int to_signed_int( unsigned int value ) {
	int signed_value = value;

	if (value >> 15) signed_value |= -1 << 15;
	return signed_value;
}

long int to_signed_long( unsigned long int value ) {
	long int signed_value = value;

	if (value >> 31) signed_value |= -1 << 31;
	return signed_value;
}

uint64_t swap64(uint64_t x) {
	uint8_t i;
	uint64_t y ;
	uint8_t *px, *py;

	px = (uint8_t *)&x;
	py = (uint8_t *)&y;

	for (i=0; i<8; i++) {
		*(py+i) = *(px +(7-i));
	}
	return y;
}

uint64_t ntohll(uint64_t value) {
	return 1 == ntohl(1) ? value : swap64(value);
}

#define FL __FILE__,__LINE__
#define VERBOSE if (g->verbose)
#define DEBUG if (g->debug)

#define DECODE_MODE_FREESPACE 1
#define DECODE_MODE_NORMAL 0

#define PAYLOAD_SIZE_MINIMUM 10
#define PAYLOAD_CELLS_MAX 1000
#define OVERFLOW_PAGES_MAX 10000

#define PARAM_VERSION "--version"
#define PARAM_HELP "--help"
#define PARAM_FINE_SEARCH "--fine-search"
#define PARAM_FREESPACE_ONLY "--freespace"
#define PARAM_FREESPACE_MINIMUM "--freespace-minimum="
#define PARAM_NO_BLOBS "--no-blobs"
#define PARAM_BLOB_SIZE_LIMIT "--blob-size-limit="
#define PARAM_CELLCOUNT_MIN "--cellcount-min="
#define PARAM_CELLCOUNT_MAX "--cellcount-max="
#define PARAM_ROWSIZE_MIN "--rowsize-min="
#define PARAM_ROWSIZE_MAX "--rowsize-max="
#define PARAM_PAGE_SIZE "--page-size="
#define PARAM_PAGE_START "--page-start="  // add to 0.5
#define PARAM_PAGE_END "--page-end=" // add to 0.5
#define PARAM_REMOVED_ONLY "--removed-only"

struct globals {
	uint8_t debug;
	uint8_t verbose;

	char *input_file; // actual file name
	char *db_origin;	// the mmap'd file origin
	char *db_end; // the computed end of the mmap'd file ( based on file size )
	char *db_cfp; // current file position
	char *db_cpp; // current page position
	char *db_cpp_limit; // end of the current page
	size_t db_size;

	uint32_t page_size, page_count, page_number;
	uint32_t page_start, page_end;

	uint32_t freelist_first_page, freelist_page_count;
	uint32_t *freelist_pages;
	uint32_t freelist_pages_current_index;
	int freelist_space_only;
	int removed_only;
	size_t freespace_minimum;

	time_t date_upper, date_lower; // deprecated - now that Undark has become a generic tool
	int cc_min, cc_max;  // cell count limits
	size_t rs_min, rs_max; // row/payload limits
	int report_blobs; // do we even handle blob data
	size_t blob_size_limit; // at which point do we cut over to dumping to *.blob files?

	int blob_count;
	int fine_search;
};

struct cell {
	int t; // serial
	int o; // offset
	int s; // size
};

struct sql_payload {
	uint64_t prefix_length;
	uint64_t length;
	uint64_t rowid;
	uint64_t header_size;
	int cell_count;
	int cell_page;
	int cell_page_offset;
	struct cell cells[PAYLOAD_CELLS_MAX+1];
	uint32_t overflow_pages[OVERFLOW_PAGES_MAX+1];
	char *mapped_data, *mapped_data_endpoint;
};

struct sqlite_leaf_header {
	int page_number;
	int page_byte;
	uint16_t freeblock_offset;
	uint16_t freeblock_size;
	uint16_t freeblock_next;
	int cellcount;
	int cell_offset;
	int freebytes;
};

char version[] = "undark version 0.8, origin by Paul L Daniels ( pldaniels@pldaniels.com )\n";
char help[] = "-i <sqlite DB> [-d] [-v] [-V|--version] [--cellcount-min=<count>] [--cellcount-max=<count>] [--rowsize-min=<bytes>] [--rowsize-max=<bytes>] [--no-blobs] [--blob-size-limit=<bytes>] [--page-size=<bytes>] [--page-start=<number>] [--page-end=<number>] [--freespace] [--freespace-minimum=<bytes>]\n"
"\t-i: input SQLite3 format database\n"
"\t-d: enable debugging output (very large dumps)\n"
"\t-v: enable verbose output\n"
"\t-V|--version: show version of software\n"
"\t-h|--help: show this help\n"
"\t--cellcount-min: define the minimum number of cells a row must have to be extracted\n"
"\t--cellcount-max: define the maximum number of cells a row must have to be extracted\n"
"\t--rowsize-min: define the minimum number of bytes a row must have to be extracted\n"
"\t--rowsize-max: define the maximum number of bytes a row must have to be extracted\n"
"\t--no-blobs: disable the dumping of blob data\n"
"\t--blob-size-limit: all blobs larger than this size are dumped to .blob files\n"
"\t--fine-search: search DB shifting one byte at a time, rather than records\n"
"\t--page-size: hard code the page size for the DB (useful when header is damaged)\n"
"\t--removed-only: Dumps rows that have their key set to -1\n"
"\t--freespace: search for rows in the freespace\n";

int UNDARK_init( struct globals *g ) {
	g->page_size = 0;
	g->page_count = 0;
	g->page_number = 1;
	g->debug = 0;
	g->verbose = 0;
	g->input_file = NULL;
	g->date_lower = 0;
	g->date_upper = 0;
	g->cc_max = PAYLOAD_CELLS_MAX;
	g->cc_min = 2;
	g->rs_max = SIZE_MAX;
	g->rs_min = 10;
	g->blob_count = 0;
	g->report_blobs = 1;
	g->blob_size_limit = SIZE_MAX; // C99
	g->fine_search = 0;
	g->freelist_space_only = 0;
	g->removed_only = 0;
	g->freespace_minimum = SIZE_MAX; // C99
	g->page_start = 0;
	g->page_end = 0;

	g->db_cfp = NULL;
	g->db_cpp = NULL;

	return 0;
}

int UNDARK_parse_parameters( int argc, char **argv, struct globals *g ) {
	int param;

	if (argc < 2) {
		fprintf(stderr,"%s", help);
		fprintf(stderr,"Sizeof double = %ld,  long double = %ld\n", sizeof(double), sizeof(long double));
		exit(1);
	}

	for (param = 1; param < argc; param++) {
		char *p = argv[param];

		if (strcmp(p, "-V") == 0) { fprintf(stdout,"%s", version); exit(0); }
		if (strcmp(p, "-h") == 0) { fprintf(stdout,"%s %s", argv[0], help); exit(0); }
		if (strcmp(p, "-d") == 0) g->debug = 1;
		if (strcmp(p, "-v") == 0) g->verbose = 1;
		if (strcmp(p, "-i") == 0) {
			param++;
			if (param < argc) {
				g->input_file = argv[param];
			} else {
				fprintf(stderr,"Not enough paramters\n");
				exit(1);
			}
		} else if (strncmp(p,"--", 2) == 0) {

			DEBUG fprintf(stderr,"Parameter: '%s' %d\n", p, (int)strlen(PARAM_BLOB_SIZE_LIMIT));
			// extended parameters

			if (strncmp(p,PARAM_VERSION, strlen(PARAM_VERSION))==0) {
				fprintf(stderr,"%s", version);
				exit(0);

			} else if (strncmp(p,PARAM_HELP, strlen(PARAM_HELP))==0) {
				fprintf(stderr,"%s %s", argv[0], help);
				exit(0);


			} else if (strncmp(p,PARAM_NO_BLOBS, strlen(PARAM_NO_BLOBS))==0) {
				g->report_blobs = 0;

			} else if (strncmp(p,PARAM_BLOB_SIZE_LIMIT, strlen(PARAM_BLOB_SIZE_LIMIT))==0) {
				p = p +strlen(PARAM_BLOB_SIZE_LIMIT);
				g->blob_size_limit = strtol( p, NULL, 10 );

			} else if (strncmp(p,PARAM_PAGE_START, strlen(PARAM_PAGE_START))==0) {
				p = p +strlen(PARAM_PAGE_START);
				g->page_start = strtol( p, NULL, 10 );

			} else if (strncmp(p,PARAM_PAGE_END, strlen(PARAM_PAGE_END))==0) {
				p = p +strlen(PARAM_PAGE_END);
				g->page_end = strtol( p, NULL, 10 );

			} else if (strncmp(p,PARAM_PAGE_SIZE, strlen(PARAM_PAGE_SIZE))==0) {
				p = p +strlen(PARAM_PAGE_SIZE);
				g->page_size = strtol( p, NULL, 10 );

			} else if (strncmp(p,PARAM_FREESPACE_MINIMUM, strlen(PARAM_FREESPACE_MINIMUM))==0) {
				p = p +strlen(PARAM_FREESPACE_MINIMUM);
				g->freespace_minimum = strtol( p, NULL, 10 );

			} else if (strncmp(p,PARAM_CELLCOUNT_MIN, strlen(PARAM_CELLCOUNT_MIN))==0) {
				p = p +strlen(PARAM_CELLCOUNT_MIN);
				g->cc_min = strtol( p, NULL, 10 );

			} else if (strncmp(p,PARAM_CELLCOUNT_MAX, strlen(PARAM_CELLCOUNT_MAX))==0) {
				p = p +strlen(PARAM_CELLCOUNT_MAX);
				g->cc_max = strtol( p, NULL, 10 );

			} else if (strncmp(p,PARAM_ROWSIZE_MIN, strlen(PARAM_ROWSIZE_MIN))==0) {
				p = p +strlen(PARAM_ROWSIZE_MIN);
				g->rs_min = strtol( p, NULL, 10 );

			} else if (strncmp(p,PARAM_ROWSIZE_MAX, strlen(PARAM_ROWSIZE_MAX))==0) {
				p = p +strlen(PARAM_ROWSIZE_MAX);
				g->rs_max = strtol( p, NULL, 10 );

			} else if (strncmp(p,PARAM_FINE_SEARCH, strlen(PARAM_FINE_SEARCH))==0) {
				g->fine_search = 1;

			} else if (strncmp(p,PARAM_FREESPACE_ONLY, strlen(PARAM_FREESPACE_ONLY))==0) {
				g->freelist_space_only = 1;

			} else if (strncmp(p,PARAM_REMOVED_ONLY, strlen(PARAM_REMOVED_ONLY))==0) {
				g->removed_only = 1;

			} else {
				fprintf(stderr,"Cannot interpret extended parameter: \"%s\"\n",p);
				exit(1);
			}

		}
	}

	if (g->input_file == NULL) {
		fprintf(stderr,"ERROR: Need input file\n");
		exit(1);
	}

	return 0;
}

int tdump( char *p, uint16_t l ) {
	while (l--) {
		if (isprint(*p)) fprintf(stdout,"%c", *p); else fprintf(stdout,".");
		p++;
	}

	return 0;
}

// Dumps text in a SQL friendly format ( doubling of single quotes )
int sqltdump( char *p, uint16_t l ) {
	fprintf(stdout,"\"");
	while (l--) {
		if (*p == '\"') fprintf(stdout,"\"");
		if (isprint(*p)) fprintf(stdout,"%c", *p); else fprintf(stdout,".");
		p++;
	}
	fprintf(stdout,"\"");

	return 0;
}

int blob_dump( unsigned char *p, uint16_t l ) {
	fprintf(stdout,"x'");
	while (l--) {
		fprintf(stdout,"%02X", ( unsigned char)*p);
		p++;
	}
	fprintf(stdout,"'");

	return 0;
}

// Combo hex + text dump, 16 byte wide rows
int hdump( unsigned char *p, uint16_t length, char *msg ) {
	int oc = 0;
	int ll = length;

	fprintf(stdout,"%s: Hexdumping %d bytes from %p\n", msg, ll, p);
	uint16_t c = 0;

	if (p == NULL) {
		fprintf(stdout,"ERROR: NULL passed.\n");
		//		exit(1);
	}

	while (ll > 0) {
		int br;
		unsigned char *op;

		fprintf(stdout,"%04X [%06d] ", oc, ll);
		oc+=16;

		br = ll;
		op = p;
		while (ll--) {
			fprintf(stdout, "%02X ", *p);
			c++;
			p++;
			if (c%16 == 0) break;
		}

		ll = br;
		p = op;
		c = 0;

		fprintf(stdout, "  [%06d]", ll );
		while (ll--) {
			fprintf(stdout,"%c", isprint(*p)?*p:'.');
			c++;
			p++;
			if (c%16 == 0)  break;
		}

		fprintf(stdout," %d\n",ll);
	}

	fprintf(stdout,"\n");

	return 0;
}

int blob_dump_to_file( struct globals *g, char *p, size_t l ) {
	int f;
	ssize_t written;
	char fn[1024];

	snprintf(fn, sizeof(fn), "%d.blob", g->blob_count);
	DEBUG fprintf(stdout,"%s:%d:DEBUG: Writing %ld bytes to %s\n", FL , l, fn );
	f = open(fn, O_WRONLY|O_CREAT|O_TRUNC, S_IRUSR|S_IWUSR );
	if (!f) { fprintf(stderr,"Cannot open %s (%s)\n", fn, strerror(errno)); return 1; }
	written = write(f, p, l);
	if ( written != l ) {
		fprintf(stderr,"Wrote %ld of %ld bytes to %s ( %s )\n", written, l, fn, strerror(errno));
		close(f);
		return 1;
	}
	close(f);

	return 0;
}

// Searches for the needle among a haystack possibly containing \0 delimeted data.
char *bstrstr( char *haystack, char *needle, char *limit ) {
	char *p;

	if ((!needle)||(*needle == '\0')) return NULL;
	if (!haystack) return NULL;
	if ((limit == NULL)||(limit <=haystack)) return NULL;

	p = haystack;
	while (p < limit) {
		char *tp;
		char *tn;

		tn = needle;
		tp = p;
		while ((*tn) && (tp < limit) && (*tp == *tn)) {
			tn++;
			tp++;
			if (*tn == '\0') return p;
		}
		p++;
	}

	return NULL;
}

// Decodes the payload header data so that we can then later pull the actual data from the file.
int decode_row( struct globals *g, char *p, char *data_endpoint, struct sql_payload *payload, int mode, size_t forced_length ) {
	int t = 0, offset;
	char *plh_ep; // payload header end point
	char *base = p;

	DEBUG {
		fprintf(stdout,"%s:%d:DEBUG:DECODING ROW-------------------------MODE:%s\n", FL, (mode?"Freespace":"Standard"));
		hdump((unsigned char *)p, 16, "Decode_row start data");
	}

	payload->overflow_pages[0] = 0;
	payload->cell_count = 0;

	if ( mode == DECODE_MODE_FREESPACE ) {
		payload->length = forced_length -4; // and we still have to deduct the payload header size
	} else {
		varint_decode( &(payload->length), p, &p );
	}

	if (payload->length > g->db_size) return 0;
	if (payload->length < g->rs_min) return 0;
	if (payload->length > g->rs_max) return 0;

	DEBUG fprintf(stdout,"%s:%d:DEBUG:Payload size: %lu\n", FL, (unsigned long int)payload->length);

	if (mode == DECODE_MODE_FREESPACE) {
		payload->rowid = 1;
	} else {
		varint_decode(&(payload->rowid), p, &p);
	}

	if (payload->rowid < 1) return 0;

	payload->prefix_length = p -base; // store this so we know how many bytes the length + Row ID took up.

	plh_ep = p; // first set up the beginning of the payload header array size.
	varint_decode( &(payload->header_size), p, &p );
	if (payload->header_size > g->page_size) return 0;

	if (mode == DECODE_MODE_FREESPACE) {
		payload->length -= payload->header_size;
		DEBUG fprintf(stdout,"%s:%d:DEBUG: Looking for %lu bytes of data after the payload header\n", FL , (long unsigned int)payload->length);
		fflush(stdout);
	}

	// If the payload size exceeds the page_size, then we have to do some more checking
	if (payload->length > (g->page_size -35)) {
		uint32_t tmp, ovp;
		int ovpi = 1;

		// get the FIRST overflow page
		memcpy(&tmp, data_endpoint -4, 4);
		ovp = payload->overflow_pages[0] = ntohl(tmp);

		// if the page is beyond the file range, then we've just got defective input data
		if (ovp > g->page_count) return 0;
		DEBUG fprintf(stdout,"%s:%d:DEBUG: First overflow page = %lu\n", FL , (long unsigned int)ovp);
		DEBUG hdump((unsigned char *)(data_endpoint -16), 16, "First overflow page start data");


		while (ovp > 0) {

			void *calculated_address;

			calculated_address = g->db_origin +( (ovp -1) *g->page_size);
			DEBUG fprintf(stdout,"%s:%d:DEBUG: Calculated address: %p\n", FL, calculated_address);

			// test for seeking beyond the db limit
			//if ((g->db_origin +((ovp -1) *g->page_size)) > (g->db_end -4)) {

			if ( calculated_address > (void *)(g->db_end -4)) { //PLD:20141220-0000
				DEBUG	fprintf(stdout,"%s:%d:ERROR: Seek beyond end of data looking for overflow page (%p > %p)\n", FL, calculated_address, g->db_end);
				break;
			}

			if ( calculated_address < (void *)(g->db_origin)) { //PLD:20141220-0000
				DEBUG	fprintf(stdout,"%s:%d:ERROR: Seek before DB starts (%p < %p)\n", FL, calculated_address, g->db_origin);
				break;
			}

			memcpy(&tmp, calculated_address, 4);
			ovp = payload->overflow_pages[ovpi] = ntohl(tmp);
			DEBUG fprintf(stdout,"%s:%d:DEBUG: overflow page[%d] = %d\n", FL , ovpi, ovp);
			DEBUG fflush(stdout);
			ovpi++;
			if (ovpi > OVERFLOW_PAGES_MAX) {
				fprintf(stdout,"ERROR: No more space for overflow pages\n");
				fflush(stdout);
				payload->overflow_pages[0] = 0;
				break;
			}
			payload->overflow_pages[ovpi] = 0;
		}


		DEBUG {
			fprintf(stdout,"DEBUG: Total of %d overflow pages\n",ovpi);
			ovpi = 0;
			while (payload->overflow_pages[ovpi]) {
				fprintf(stdout,"DEBUG: Overflow %d->%d\n", ovpi, payload->overflow_pages[ovpi]);
				ovpi++;
			}
		}
	}  // overflow handling

	if (payload->header_size > g->page_size) return 0; // sorry, no can do with the way we're playing this decoding game.
	if (payload->header_size < 2) return 0; // need at least 2 bytes

	plh_ep += payload->header_size; // if we got a sane value, then we can use this for the full decode size ( includes the size of the first varint telling us the size )

	DEBUG { fprintf(stdout,"[L:%lld][id:%lld][PLHz:%lld]",(long long int) payload->length, (long long int)payload->rowid, (long long int)payload->header_size); }

	t = 0;
	offset = 0;

	while (1) {
		uint64_t s;
		int vil;

		vil = varint_decode( &s, p, &p );

		if (vil > 8) return 0; // no var int should be bigger than 8 bytes.

		payload->cells[t].t = s; // set the type
		switch (s) {
			case 0: s = 0; break;
			case 1: s = 1; break;
			case 2: s = 2; break;
			case 3: s = 3; break;
			case 4: s = 4; break;
			case 5: s = 6; break;
			case 6: case 7: s = 8; break;
			case 8: case 9: s = 0; break;
			case 10: case 11: DEBUG fprintf(stdout,"%s:%d:DEBUG: celltype 10/11 reserved, aborting row.\n",FL); s = 0; return 0; break;
			default:
				if ((s >= 12) && ((s & 0x01) == 0)) {
					payload->cells[t].t = 12;
					s = (s - 12)/2;
				} else if ((s >= 13) && ((s&0x01) == 1)) {
					payload->cells[t].t = 13;
					s = (s - 13)/2;
				}
				break;
		}

		payload->cells[t].s = s; // set the size/length
		payload->cells[t].o = (plh_ep +offset) -base;
		offset += payload->cells[t].s;
		if (offset > payload->length) return 0;

		DEBUG { fprintf(stdout,"[%d:%d:%d-%d(%ld)]", t, payload->cells[t].t, payload->cells[t].s, payload->cells[t].o, plh_ep -p ); }

		if (p >= plh_ep) break;
		t++;
		payload->cell_count++;
		if ( t > g->cc_max ) return 0;
	} // while decoding the cells

	if (p == plh_ep) {
		DEBUG {
			fprintf(stdout,"DEBUG: Payload head size match. (%ld =? %ld)\n ", p -base,plh_ep -base);
			fprintf(stdout,"DEBUG: Data size by cell meta sum = %d\n ", offset );
		}
	} else {
		DEBUG {
			fprintf(stdout,"DEBUG: Payload scan end point, and predicted end point didn't match, difference %ld \n", p -plh_ep );
		}
	}

	if ( t < g->cc_min )  {
		DEBUG fprintf(stdout,"%s:%d:DEBUG: cell count under the minimum, so aborting\n", FL );
		return 0;
	}

	DEBUG fprintf(stdout,"Offset [%u] + headersize [%lu] = length check [%lu]... \n", offset, (unsigned long int)payload->header_size, (unsigned long int)payload->length);

	if (mode == DECODE_MODE_FREESPACE) {
		/** there can often be multiple entries within freespace, so we have to be
		 * a little looser with our acceptance criterion
		 */
		if (offset <= payload->length) {
			DEBUG fprintf(stdout,"%s:%d:DEBUG: FREESPACE SUBMATCH FOUND ( %u of %lu used )\n", FL , offset, (long unsigned int) payload->length);
			return (offset +payload->header_size +4);
		}
	}

	if (offset + payload->header_size  == payload->length) {
		DEBUG fprintf(stdout,"\nMATCH FOUND!\n");
		return 1;
	}

	return 0;
}

int dump_row( struct globals *g, char *base, char *data_endpoint, struct sql_payload *payload, int mode ) {
	int t = 0;
	int ovpi;
	void *addr;

	DEBUG fprintf(stdout,"\n-DUMPING ROW------------------\n");
	DEBUG hdump((unsigned char *)base, 16, "Dump_row starting data");

	if ( payload->length > g->db_size ) {
		DEBUG fprintf(stdout,"%s:%d:ERROR: Nonsensical payload length of %ld requested, ignoring.\n", FL, (long int)payload->length);
		return -1;
	}

	if (payload->overflow_pages[0] == 0) {
		payload->mapped_data = base;
		payload->mapped_data_endpoint = data_endpoint;

	} else {
		payload->mapped_data = malloc( (payload->length +100) *sizeof(char) );
		if ( !payload->mapped_data ) {
			fprintf(stderr,"%s:%d:ERROR: Cannot allocate %ld bytes for mapped data\n", FL, (long int)payload->length +100);
			return -1;
		}
		DEBUG fprintf(stdout,"ALLOCATED %d bytes to mapped data\n", (int)(payload->length +100) );
		if (!payload->mapped_data){ fprintf(stderr,"ERROR: Cannot allocate %d bytes for payload\n", (int)(payload->length +1)); return 0; }
		memset( payload->mapped_data, 'X', payload->length +1 );

		// load in the first, default page.
		DEBUG fprintf(stdout,"Copying data for initial page\n");
		memcpy(payload->mapped_data, base, data_endpoint -base );
		payload->mapped_data_endpoint = payload->mapped_data +(data_endpoint -base -4);
		//		DEBUG hdump( (unsigned char *)payload->mapped_data, payload->mapped_data_endpoint -payload->mapped_data +4  );

		// Load in the overflow pages (if any)
		ovpi = 0;
		while (payload->overflow_pages[ovpi]) {
			DEBUG fprintf(stdout,"Copying data from file to memory for page %d to offset [%d]\n", payload->overflow_pages[ovpi], (int)(payload->mapped_data_endpoint -payload->mapped_data));

			addr = g->db_origin +((payload->overflow_pages[ovpi]-1) *g->page_size) +4; //PLD:20141221-2240 segfault fix
			if (( addr < (void *)g->db_origin) || ( addr+4 > (void *)g->db_end)) {
				DEBUG fprintf(stdout,"%s:%d:dump_row:ERROR: page seek request outside of boundaries of file (%p < %p > %p)\n", FL, g->db_origin, addr, g->db_end);
				return -1;
			}

			memcpy(payload->mapped_data_endpoint, addr, g->page_size -4);
			payload->mapped_data_endpoint += g->page_size -4;

			//	DEBUG hdump( (unsigned char *)payload->mapped_data, payload->mapped_data_endpoint -payload->mapped_data );

			ovpi++;
		}
	}

	DEBUG hdump((unsigned char *)payload->mapped_data, payload->mapped_data_endpoint -payload->mapped_data, "Payload mapped data" );

	if (mode == DECODE_MODE_FREESPACE) {
		t = 0;
		fprintf(stdout,"-1");

	} else t = -1;

	while (t <= payload->cell_count) {
		DEBUG fprintf(stdout,"%s:%d:DEBUG: Cell[%d], Type:%d, size:%d, offset:%d\n", FL , t, payload->cells[t].t, payload->cells[t].s, payload->cells[t].o);
		if (t == -1) fprintf(stdout,"%ld", (long unsigned int) payload->rowid);
		if (t>=0) { fprintf(stdout,",");
			switch (payload->cells[t].t) {
				case 0: fprintf(stdout,"NULL"); break;
				case 1: fprintf(stdout,"x%d", to_signed_byte(*(payload->mapped_data +payload->cells[t].o)) ); break;
				case 2: {
							  uint16_t n;
							  memcpy(&n, payload->mapped_data +payload->cells[t].o, 2 );
							  fprintf(stdout,"%d" , to_signed_int(ntohs(n)));
						  }
						  break;

				case 3: {
							  uint32_t n;
							  memcpy(&n, payload->mapped_data +payload->cells[t].o, 3 );
							  fprintf(stdout,"%ld", to_signed_long(ntohl(n)));
						  }
						  break;

				case 4: {
							  uint32_t n;
							  memcpy(&n, payload->mapped_data +payload->cells[t].o, 4 );
							  fprintf(stdout,"%ld", to_signed_long(ntohl(n)));
						  }
						  break;

				case 5: fprintf(stdout,"%d", ntohl(*(payload->mapped_data +payload->cells[t].o))); break;
				case 6: fprintf(stdout,"%d", ntohl(*(payload->mapped_data +payload->cells[t].o))); break;
				case 7:
						  {
								  uint64_t n;
								  uint64_t nn;
								  double *zz;
								  memcpy(&n, payload->mapped_data +payload->cells[t].o, 8 );
								  nn = (double) ntohll(n);
//									hdump( &nn, 8, "\nFPPP: ");
									zz = (double *)&nn;
								  fprintf(stdout,"%f",*zz);
						  }
						  break;

				case 8: fprintf(stdout,"0" ); break;
				case 9: fprintf(stdout,"1" ); break;
				case 12:
						  if ( g->report_blobs) {
							  if (payload->cells[t].s < g->blob_size_limit) {
								  DEBUG fprintf(stdout,"%s:%d:DEBUG:Not Dumping data to blob file, keeping in CSV\n", FL );
								  blob_dump((unsigned char *) (payload->mapped_data +payload->cells[t].o), payload->cells[t].s );
							  } else {
								  // dump the blob to a file.
								  DEBUG fprintf(stdout,"%s:%d:DEBUG:Dumping data to %d.blob [%d bytes]\n", FL ,g->blob_count, payload->cells[t].s);
								  blob_dump_to_file( g, (payload->mapped_data +payload->cells[t].o), payload->cells[t].s );
								  DEBUG fprintf(stdout,"\"%d.blob\"", g->blob_count);
							  }
						  }
						  g->blob_count++;
						  break;

				case 13:
						  DEBUG fprintf(stdout,"%s:%d:DEBUG: Dumping text-13\n", FL );
						  sqltdump( payload->mapped_data +payload->cells[t].o, payload->cells[t].s );
						  break;
				default:
						  fprintf(stderr,"Invalid cell type '%d'", payload->cells[t].t);
						  DEBUG fprintf(stdout,"%s:%d:DEBUG: Invalid cell type '%d'", FL, payload->cells[t].t);
						  DEBUG hdump( (unsigned char *) base, 128, "Invalid cell type" );
						  return 0;
						  break;
			} // switch cell type
		}

		t++;

	} // while decoding the cells

	fprintf(stdout,"\n");
	fflush(stdout);
	if (payload->overflow_pages[0] != 0) {
		free( payload->mapped_data );
	}

	return 0;
}

// Finds rows within a block.
char *find_next_row( struct globals *g, char *s, char *end_point, char *global_start, int mode, size_t forced_length ) {
	char *p;
	struct sql_payload sql;

	DEBUG fprintf(stdout,"find_next_row: MODE: %d\n", mode );
	if (s == NULL) fprintf(stdout,"ERROR: NULL passed as search-space parameter\n");
	p = s;
	do {
		int row;

		row = decode_row( g, p, end_point, &sql, mode, forced_length );
		if (row) {
			DEBUG fprintf(stdout,"ROWID: %ld found [+%ld] record size: %d bytes\n", (unsigned long int)sql.rowid, p -global_start, (unsigned int)( sql.length+sql.prefix_length ));
			fflush(stdout);

			/** If we're only wanting the removed, no-key-value rows, then
			  * continue to the next row
			  */
			if ((g->removed_only)&&(row >= 0)) {
				p++;
				continue;
			}



			if ((mode == DECODE_MODE_NORMAL)&&( g->freelist_space_only == 1)) {
				// do nothing
			} else  {
				dump_row( g, p, end_point, &sql, mode );
			}

			fflush(stdout);
			if (mode == DECODE_MODE_NORMAL) {
				if (g->fine_search) p++;
				else p+= sql.length;
			} else {
				if (row >= forced_length) {
					DEBUG fprintf(stdout,"%s:%d:DEBUG: No more data left in freespace block to examine\n", FL);
					p = end_point;
					break;
				} else {
					p+=row; forced_length -= row;
					DEBUG hdump((unsigned char *)p,64, "After freespace decode");
				}
			}
		} else {
			p++;
		}

	} while (p < end_point -PAYLOAD_SIZE_MINIMUM);

	return NULL;

}

int main( int argc, char **argv ) {
	int fd;
	struct globals globo, *g;
	struct stat st;
	char *p;
	int stat_result;

	g = &globo;

	UNDARK_init( g );
	UNDARK_parse_parameters( argc, argv, g );

	stat_result = stat( g->input_file, &st );
	if (stat_result != 0) {
		fprintf(stderr,"ERROR: Cannot access input file '%s' ( %s )\n", g->input_file, strerror(errno));
		exit(1);
	}

	fd = open( g->input_file, O_RDONLY );
	g->db_size = st.st_size;
	g->db_origin = mmap( NULL, st.st_size, PROT_READ, MAP_PRIVATE, fd, 0 );
	g->db_end = g->db_origin +st.st_size -1;

	//fprintf(stderr,"DB origin: %p\nDB end: %p\n", g->db_origin, g->db_end );

	// If the page size is already set via parameter, then skip
	if (g->page_size == 0) {
		p = g->db_origin +16;
		g->page_size =	(*(p+1)) | ((*p)<<8);
	}

	// Get the number of pages that are supposed to be in the database
	p = g->db_origin +28;
	memcpy( &g->page_count, g->db_origin +28, 4 ); // copy the page count from the header
	g->page_count = ntohl( g->page_count ); // convert to local format

	DEBUG fprintf(stdout,"Pagesize: %u, Pagecount: %u\n", g->page_size, g->page_count);

	// Get the free list meta data
	memcpy( &g->freelist_first_page, g->db_origin +32, 4 ); // copy the page count from the header
	g->freelist_first_page = ntohl( g->freelist_first_page );
	DEBUG fprintf(stdout,"First page of freelist trunk: %d\n", g->freelist_first_page );

	memcpy(&g->freelist_page_count, g->db_origin +36, 4); // copy the page count from the header
	g->freelist_page_count = ntohl( g->freelist_page_count );
	DEBUG fprintf(stdout,"Freelist page count: %d\n", g->freelist_page_count );

	// Get the actual free list pages
	if (0) {
		if (g->freelist_page_count) {
			g->freelist_pages = malloc( (g->freelist_page_count +1) * sizeof(uint32_t) );
			if (!g->freelist_pages) {
				fprintf(stderr,"ERROR: Cannot allocate memory to build page free list\n");
				exit(1);
			} else {
				uint32_t next_page;
				uint32_t pli;

				next_page = g->freelist_first_page;
				g->freelist_pages[0] = next_page;
				g->freelist_pages[1] = 0;
				pli = 1;
				if ( pli < g->freelist_page_count ) {
					do {
						uint32_t tmp_page, leaf_page_count;
						char *fp, *current_page_endpoint;
						uint32_t jump;

						jump = ((next_page-2) *g->page_size);
						fp = g->db_origin +jump;
						current_page_endpoint = fp +g->page_size;
						fprintf(stdout,"Freelist - current trunk page = %d [ offset: %X ]\n", next_page, jump);
						hdump((unsigned char*)fp, g->page_size, "Current trunk page");
						DEBUG fflush(stdout);

						memcpy( &tmp_page, fp, sizeof(uint32_t));
						tmp_page = ntohl(tmp_page);
						fp += sizeof(uint32_t);
						DEBUG fprintf(stdout,"Next trunk page (if any): %d\n",tmp_page);
						DEBUG fflush(stdout);

						memcpy( &leaf_page_count, fp, sizeof(uint32_t));
						leaf_page_count = ntohl(leaf_page_count);
						fp += sizeof(uint32_t);
						DEBUG fprintf(stdout,"Leaf page count: %d\n",leaf_page_count);
						DEBUG fflush(stdout);

						//while ((pli <= g->freelist_page_count)&&( fp < current_page_endpoint )) {
						while (( fp < current_page_endpoint )&&( leaf_page_count-- )) {
							hdump((unsigned char*)fp, 16, "Next free page possible");
							memcpy( &(g->freelist_pages[pli]), fp, sizeof(uint32_t));
							g->freelist_pages[pli] = ntohl( g->freelist_pages[pli] );
							DEBUG fprintf(stdout, "Next free page[%d]: %d\n", pli, g->freelist_pages[pli]);
							if (g->freelist_pages[pli] == 0) {
								fprintf(stdout,"End of freelist detected\n");
								fflush(stdout);
								break;
							}
							fflush(stdout);
							pli++;
							fp+= sizeof(uint32_t);
						}

						next_page = tmp_page;
					} while (next_page > 0);
						fprintf(stdout,"Freepages - END\n");
						fflush(stdout);
					}
				} // if there were more than one page
			}
		}



		g->db_cfp = g->db_cpp = g->db_origin;

		DEBUG fprintf(stdout,"%s:%d:DEBUG: Commence decoding data\n", FL );
		fflush(stdout);

		while (g->db_cfp < g->db_end ) {
			struct sqlite_leaf_header leaf;
			int freeblock_mode = 0;

			/* load the next page from the file in to the scratch pad */
			g->db_cfp = g->db_cpp;
			g->db_cpp_limit = g->db_cpp +g->page_size ; // was -1 ?

			DEBUG fprintf(stdout,"\n\n%s:%d:-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=START.\n", FL);

			// process the block, mostly this is just removing any 0-bytes from the block
			// so our strstr() calls aren't prematurely terminated.
			DEBUG {
				char *p;
				size_t l;
				int bc = 0;

				fprintf(stdout,"%s:%d:Dumping main block in RAW... [ Page No: %lu, Offset: %lu (0x%X),  size : %d ]\n"
						, FL
						, (long unsigned int)g->page_number
						, (long unsigned int)(g->db_cpp -g->db_origin)
						, (unsigned int)(g->db_cpp -g->db_origin)
						,  g->page_size
						);

				p = g->db_cfp;
				l = g->page_size;
					while ((l--)&&(p)) {
					{ if (isprint(*p)) { fprintf(stdout,"%c", *p); } else fprintf(stdout,"_");}
					p++;
					bc++;
					if (bc%128 == 0) fprintf(stdout,"\n");
				}
				fprintf(stdout,"\n");
				fflush(stdout);
			} // debug

			leaf.freeblock_offset = 0;
			leaf.freeblock_size = 0;
			leaf.freeblock_next = 0;
			leaf.page_number = g->page_number;

			/* Decode the page header */
			if ((g->db_cfp < g->db_end) && *(g->db_cfp) == 13) {
				DEBUG fprintf(stdout,"%s:%d:DEBUG: Decoding page header for page %d\n", FL , g->page_number );
				fflush(stdout);
				leaf.page_byte = 13;

				/**
				 * Get freeblock offset and determine if we have a free block in this
				 * page that needs to be inspected.  This is one of the more commonly
				 * needed parts of data for our row recovery
				 *
				 */
				memcpy( &(leaf.freeblock_offset), (g->db_cfp +1), 2 );
				leaf.freeblock_offset = ntohs( leaf.freeblock_offset );
				if (leaf.freeblock_offset > 0) {
					uint16_t next, sz, off;

					freeblock_mode = 1;
					off = leaf.freeblock_offset;

					DEBUG fprintf(stdout,"%s:%d:DEBUG: FREEBLOCK mode ON: header decode [offset=%u]\n", FL , leaf.freeblock_offset);

					do {
						DEBUG hdump((unsigned char *)(g->db_cfp +off), 16, "Freeblock header data");

						memcpy( &next, ( g->db_cfp +off ), 2 );
						next = ntohs( next );
						memcpy( &sz, ( g->db_cfp +off +2 ), 2 );
						sz = ntohs( sz );

						DEBUG fprintf(stdout,"%s:%d:DEBUG: Freeblock size = %u, next position = %u\n", FL, sz, next );

						if (next) off = next;
					} while (next);
					DEBUG fprintf(stdout,"%s:%d:DEBUG: END OF FREEBLOCK TRACE\n", FL);

					memcpy( &(leaf.freeblock_next), ( g->db_cfp +leaf.freeblock_offset ), 2 );
					leaf.freeblock_next = ntohs( leaf.freeblock_next );
					memcpy( &(leaf.freeblock_size), ( g->db_cfp +leaf.freeblock_offset +2 ), 2 );
					leaf.freeblock_size = ntohs( leaf.freeblock_size );
				}

				DEBUG fprintf(stdout,"%s:%d:DEBUG: Freeblock offset = %u, size = %u, next block = %u \n", FL , leaf.freeblock_offset, leaf.freeblock_size, leaf.freeblock_next );
				if (leaf.freeblock_size > 0) {
					DEBUG fprintf(stdout,"%s:%d:DEBUG: Freeblock data [ %d bytes total [4 bytes for header] ]\n", FL, leaf.freeblock_size );
					DEBUG hdump( (unsigned char *)(g->db_cfp +leaf.freeblock_offset+4), leaf.freeblock_size-4, "Actual data in free block" );
				}
				fflush(stdout);
				//				leaf.freeblock_offset = ntohs( ta );
				leaf.cellcount = ntohs(*(g->db_cfp+3));
				leaf.cell_offset = ntohs(*(g->db_cfp+5));
				leaf.freebytes = (*(g->db_cfp+7));

				DEBUG fprintf(stdout,"%s:%d:DEBUG: PAGEHEADER:%d pagebyte: %d, freeblock offset: %d, cell count: %d, first cell offset %d, free bytes %d\n", FL
						, leaf.page_number
						, leaf.page_byte
						, leaf.freeblock_offset
						, leaf.cellcount
						, leaf.cell_offset
						, leaf.freebytes
						);

				/**
				 * If we're wanting free block sourced data, then simply jump
				 * to the start of the free block space and commence the searching
				 * in the next section ( find_next_row ).
				 *
				 * After this the g->db_cfp pointer should be sitting on the first
				 * varint of the payload header which defines the header length
				 * (inclusive)
				 *
				 * Detecting rows in the freeblocks is done differently to the
				 * normal data, so
				 *
				 */
				if (g->freelist_space_only) {

					if ((leaf.freeblock_offset > 0) && (leaf.freeblock_size > 0)) {

						DEBUG fprintf(stdout,"%s:%d:DEBUG: Shifting to freespace at %d from page start\n", FL , leaf.freeblock_offset);
						g->db_cfp = g->db_cfp + leaf.freeblock_offset +4;

						DEBUG fprintf(stdout,"%s:%d:DEBUG: New position = %p\n", FL , g->db_cfp);
						DEBUG hdump((unsigned char *)g->db_cfp -4,32, "Scratch pointer at freespace data start (including 4 byte header)");
						DEBUG fflush(stdout);
					}
				}

				fflush(stdout);
			} // if we have a leaf page, which we can decode the header on.

			//if ((leaf.page_byte == 13)) {
			if (1) {
				char *row;
				row = g->db_cfp;

				DEBUG fprintf(stdout,"%s:%d:DEBUG: g->db_cfp search at = %p\n", FL , g->db_cfp);
				do {
					if ((row > g->db_origin)&&(row < g->db_end)) {

						row = find_next_row( g, row, g->db_cpp_limit, g->db_cfp, freeblock_mode, leaf.freeblock_size );

						//if (row > g->db_end) fprintf(stdout,"ERROR: beyond end point\n");
						if (row > g->db_cpp_limit) fprintf(stdout,"ERROR: beyond end point\n");
						if (row < g->db_cfp) DEBUG fprintf(stdout,"%s:%d:DEBUG: Row location not in g->db_cfp page\n", FL );
						if (row == NULL) DEBUG fprintf(stdout,"%s:%d:DEBUG: Row has been returned as NULL\n", FL );
						DEBUG fprintf(stdout,"%s:%d:DEBUG: ROW found at offset: %ld\n", FL, row-g->db_cfp);
					} else {

						break;
					}

				} while (row && (row < g->db_cpp_limit ));
				//} while (row && (row < g->db_cpp_limit ) && (row < g->db_end) );

				DEBUG fprintf(stdout,"%s:%d:DEBUG: Finished searching for rows in DB page %d\n", FL , g->page_number);
		}

		g->db_cpp += g->page_size;
		g->page_number++;
	} // while (data < endpoint)

	close(fd);

	return 0;
}