[linux-2.6-block.git] / lib / lzo / lzo1x_compress.c

// SPDX-License-Identifier: GPL-2.0-only
/*
 *  LZO1X Compressor from LZO
 *
 *  Copyright (C) 1996-2012 Markus F.X.J. Oberhumer <markus@oberhumer.com>
 *
 *  The full LZO package can be found at:
 *  http://www.oberhumer.com/opensource/lzo/
 *
 *  Changed for Linux kernel use by:
 *  Nitin Gupta <nitingupta910@gmail.com>
 *  Richard Purdie <rpurdie@openedhand.com>
 */

#include <linux/module.h>
#include <linux/kernel.h>
#include <asm/unaligned.h>
#include <linux/lzo.h>
#include "lzodefs.h"

static noinline size_t
lzo1x_1_do_compress(const unsigned char *in, size_t in_len,
		    unsigned char *out, size_t *out_len,
		    size_t ti, void *wrkmem, signed char *state_offset,
		    const unsigned char bitstream_version)
{
	const unsigned char *ip;
	unsigned char *op;
	const unsigned char * const in_end = in + in_len;
	const unsigned char * const ip_end = in + in_len - 20;
	const unsigned char *ii;
	lzo_dict_t * const dict = (lzo_dict_t *) wrkmem;

	op = out;
	ip = in;
	ii = ip;
	ip += ti < 4 ? 4 - ti : 0;

	for (;;) {
		const unsigned char *m_pos = NULL;
		size_t t, m_len, m_off;
		u32 dv;
		u32 run_length = 0;
literal:
		ip += 1 + ((ip - ii) >> 5);
next:
		if (unlikely(ip >= ip_end))
			break;
		dv = get_unaligned_le32(ip);

		if (dv == 0 && bitstream_version) {
			const unsigned char *ir = ip + 4;
			const unsigned char *limit = min(ip_end, ip + MAX_ZERO_RUN_LENGTH + 1);
#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && \
	defined(LZO_FAST_64BIT_MEMORY_ACCESS)
			u64 dv64;

			for (; (ir + 32) <= limit; ir += 32) {
				dv64 = get_unaligned((u64 *)ir);
				dv64 |= get_unaligned((u64 *)ir + 1);
				dv64 |= get_unaligned((u64 *)ir + 2);
				dv64 |= get_unaligned((u64 *)ir + 3);
				if (dv64)
					break;
			}
			for (; (ir + 8) <= limit; ir += 8) {
				dv64 = get_unaligned((u64 *)ir);
				if (dv64) {
#  if defined(__LITTLE_ENDIAN)
					ir += __builtin_ctzll(dv64) >> 3;
#  elif defined(__BIG_ENDIAN)
					ir += __builtin_clzll(dv64) >> 3;
#  else
#    error "missing endian definition"
#  endif
					break;
				}
			}
#else
			while ((ir < (const unsigned char *)
					ALIGN((uintptr_t)ir, 4)) &&
					(ir < limit) && (*ir == 0))
				ir++;
			if (IS_ALIGNED((uintptr_t)ir, 4)) {
				for (; (ir + 4) <= limit; ir += 4) {
					dv = *((u32 *)ir);
					if (dv) {
#  if defined(__LITTLE_ENDIAN)
						ir += __builtin_ctz(dv) >> 3;
#  elif defined(__BIG_ENDIAN)
						ir += __builtin_clz(dv) >> 3;
#  else
#    error "missing endian definition"
#  endif
						break;
					}
				}
			}
#endif
			while (likely(ir < limit) && unlikely(*ir == 0))
				ir++;
			run_length = ir - ip;
			if (run_length > MAX_ZERO_RUN_LENGTH)
				run_length = MAX_ZERO_RUN_LENGTH;
		} else {
			t = ((dv * 0x1824429d) >> (32 - D_BITS)) & D_MASK;
			m_pos = in + dict[t];
			dict[t] = (lzo_dict_t) (ip - in);
			if (unlikely(dv != get_unaligned_le32(m_pos)))
				goto literal;
		}

		ii -= ti;
		ti = 0;
		t = ip - ii;
		if (t != 0) {
			if (t <= 3) {
				op[*state_offset] |= t;
				COPY4(op, ii);
				op += t;
			} else if (t <= 16) {
				*op++ = (t - 3);
				COPY8(op, ii);
				COPY8(op + 8, ii + 8);
				op += t;
			} else {
				if (t <= 18) {
					*op++ = (t - 3);
				} else {
					size_t tt = t - 18;
					*op++ = 0;
					while (unlikely(tt > 255)) {
						tt -= 255;
						*op++ = 0;
					}
					*op++ = tt;
				}
				do {
					COPY8(op, ii);
					COPY8(op + 8, ii + 8);
					op += 16;
					ii += 16;
					t -= 16;
				} while (t >= 16);
				if (t > 0) do {
					*op++ = *ii++;
				} while (--t > 0);
			}
		}

		if (unlikely(run_length)) {
			ip += run_length;
			run_length -= MIN_ZERO_RUN_LENGTH;
			put_unaligned_le32((run_length << 21) | 0xfffc18
					   | (run_length & 0x7), op);
			op += 4;
			run_length = 0;
			*state_offset = -3;
			goto finished_writing_instruction;
		}

		m_len = 4;
		{
#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && defined(LZO_USE_CTZ64)
		u64 v;
		v = get_unaligned((const u64 *) (ip + m_len)) ^
		    get_unaligned((const u64 *) (m_pos + m_len));
		if (unlikely(v == 0)) {
			do {
				m_len += 8;
				v = get_unaligned((const u64 *) (ip + m_len)) ^
				    get_unaligned((const u64 *) (m_pos + m_len));
				if (unlikely(ip + m_len >= ip_end))
					goto m_len_done;
			} while (v == 0);
		}
#  if defined(__LITTLE_ENDIAN)
		m_len += (unsigned) __builtin_ctzll(v) / 8;
#  elif defined(__BIG_ENDIAN)
		m_len += (unsigned) __builtin_clzll(v) / 8;
#  else
#    error "missing endian definition"
#  endif
#elif defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && defined(LZO_USE_CTZ32)
		u32 v;
		v = get_unaligned((const u32 *) (ip + m_len)) ^
		    get_unaligned((const u32 *) (m_pos + m_len));
		if (unlikely(v == 0)) {
			do {
				m_len += 4;
				v = get_unaligned((const u32 *) (ip + m_len)) ^
				    get_unaligned((const u32 *) (m_pos + m_len));
				if (v != 0)
					break;
				m_len += 4;
				v = get_unaligned((const u32 *) (ip + m_len)) ^
				    get_unaligned((const u32 *) (m_pos + m_len));
				if (unlikely(ip + m_len >= ip_end))
					goto m_len_done;
			} while (v == 0);
		}
#  if defined(__LITTLE_ENDIAN)
		m_len += (unsigned) __builtin_ctz(v) / 8;
#  elif defined(__BIG_ENDIAN)
		m_len += (unsigned) __builtin_clz(v) / 8;
#  else
#    error "missing endian definition"
#  endif
#else
		if (unlikely(ip[m_len] == m_pos[m_len])) {
			do {
				m_len += 1;
				if (ip[m_len] != m_pos[m_len])
					break;
				m_len += 1;
				if (ip[m_len] != m_pos[m_len])
					break;
				m_len += 1;
				if (ip[m_len] != m_pos[m_len])
					break;
				m_len += 1;
				if (ip[m_len] != m_pos[m_len])
					break;
				m_len += 1;
				if (ip[m_len] != m_pos[m_len])
					break;
				m_len += 1;
				if (ip[m_len] != m_pos[m_len])
					break;
				m_len += 1;
				if (ip[m_len] != m_pos[m_len])
					break;
				m_len += 1;
				if (unlikely(ip + m_len >= ip_end))
					goto m_len_done;
			} while (ip[m_len] == m_pos[m_len]);
		}
#endif
		}
m_len_done:

		m_off = ip - m_pos;
		ip += m_len;
		if (m_len <= M2_MAX_LEN && m_off <= M2_MAX_OFFSET) {
			m_off -= 1;
			*op++ = (((m_len - 1) << 5) | ((m_off & 7) << 2));
			*op++ = (m_off >> 3);
		} else if (m_off <= M3_MAX_OFFSET) {
			m_off -= 1;
			if (m_len <= M3_MAX_LEN)
				*op++ = (M3_MARKER | (m_len - 2));
			else {
				m_len -= M3_MAX_LEN;
				*op++ = M3_MARKER | 0;
				while (unlikely(m_len > 255)) {
					m_len -= 255;
					*op++ = 0;
				}
				*op++ = (m_len);
			}
			*op++ = (m_off << 2);
			*op++ = (m_off >> 6);
		} else {
			m_off -= 0x4000;
			if (m_len <= M4_MAX_LEN)
				*op++ = (M4_MARKER | ((m_off >> 11) & 8)
						| (m_len - 2));
			else {
				if (unlikely(((m_off & 0x403f) == 0x403f)
						&& (m_len >= 261)
						&& (m_len <= 264))
						&& likely(bitstream_version)) {
					// Under lzo-rle, block copies
					// for 261 <= length <= 264 and
					// (distance & 0x80f3) == 0x80f3
					// can result in ambiguous
					// output. Adjust length
					// to 260 to prevent ambiguity.
					ip -= m_len - 260;
					m_len = 260;
				}
				m_len -= M4_MAX_LEN;
				*op++ = (M4_MARKER | ((m_off >> 11) & 8));
				while (unlikely(m_len > 255)) {
					m_len -= 255;
					*op++ = 0;
				}
				*op++ = (m_len);
			}
			*op++ = (m_off << 2);
			*op++ = (m_off >> 6);
		}
		*state_offset = -2;
finished_writing_instruction:
		ii = ip;
		goto next;
	}
	*out_len = op - out;
	return in_end - (ii - ti);
}

static int lzogeneric1x_1_compress(const unsigned char *in, size_t in_len,
		     unsigned char *out, size_t *out_len,
		     void *wrkmem, const unsigned char bitstream_version)
{
	const unsigned char *ip = in;
	unsigned char *op = out;
	unsigned char *data_start;
	size_t l = in_len;
	size_t t = 0;
	signed char state_offset = -2;
	unsigned int m4_max_offset;

	// LZO v0 will never write 17 as first byte (except for zero-length
	// input), so this is used to version the bitstream
	if (bitstream_version > 0) {
		*op++ = 17;
		*op++ = bitstream_version;
		m4_max_offset = M4_MAX_OFFSET_V1;
	} else {
		m4_max_offset = M4_MAX_OFFSET_V0;
	}

	data_start = op;

	while (l > 20) {
		size_t ll = min_t(size_t, l, m4_max_offset + 1);
		uintptr_t ll_end = (uintptr_t) ip + ll;
		if ((ll_end + ((t + ll) >> 5)) <= ll_end)
			break;
		BUILD_BUG_ON(D_SIZE * sizeof(lzo_dict_t) > LZO1X_1_MEM_COMPRESS);
		memset(wrkmem, 0, D_SIZE * sizeof(lzo_dict_t));
		t = lzo1x_1_do_compress(ip, ll, op, out_len, t, wrkmem,
					&state_offset, bitstream_version);
		ip += ll;
		op += *out_len;
		l  -= ll;
	}
	t += l;

	if (t > 0) {
		const unsigned char *ii = in + in_len - t;

		if (op == data_start && t <= 238) {
			*op++ = (17 + t);
		} else if (t <= 3) {
			op[state_offset] |= t;
		} else if (t <= 18) {
			*op++ = (t - 3);
		} else {
			size_t tt = t - 18;
			*op++ = 0;
			while (tt > 255) {
				tt -= 255;
				*op++ = 0;
			}
			*op++ = tt;
		}
		if (t >= 16) do {
			COPY8(op, ii);
			COPY8(op + 8, ii + 8);
			op += 16;
			ii += 16;
			t -= 16;
		} while (t >= 16);
		if (t > 0) do {
			*op++ = *ii++;
		} while (--t > 0);
	}

	*op++ = M4_MARKER | 1;
	*op++ = 0;
	*op++ = 0;

	*out_len = op - out;
	return LZO_E_OK;
}

int lzo1x_1_compress(const unsigned char *in, size_t in_len,
		     unsigned char *out, size_t *out_len,
		     void *wrkmem)
{
	return lzogeneric1x_1_compress(in, in_len, out, out_len, wrkmem, 0);
}

int lzorle1x_1_compress(const unsigned char *in, size_t in_len,
		     unsigned char *out, size_t *out_len,
		     void *wrkmem)
{
	return lzogeneric1x_1_compress(in, in_len, out, out_len,
				       wrkmem, LZO_VERSION);
}

EXPORT_SYMBOL_GPL(lzo1x_1_compress);
EXPORT_SYMBOL_GPL(lzorle1x_1_compress);

MODULE_LICENSE("GPL");
MODULE_DESCRIPTION("LZO1X-1 Compressor");
Commit	Line	Data
09c434b8	1	// SPDX-License-Identifier: GPL-2.0-only
64c70b1c	2	/*
8b975bd3	3	* LZO1X Compressor from LZO
64c70b1c	4	*
8b975bd3	5	* Copyright (C) 1996-2012 Markus F.X.J. Oberhumer <markus@oberhumer.com>
64c70b1c RP	6	*
	7	* The full LZO package can be found at:
	8	* http://www.oberhumer.com/opensource/lzo/
	9	*
8b975bd3	10	* Changed for Linux kernel use by:
64c70b1c RP	11	* Nitin Gupta <nitingupta910@gmail.com>
	12	* Richard Purdie <rpurdie@openedhand.com>
	13	*/
	14
	15	#include <linux/module.h>
	16	#include <linux/kernel.h>
64c70b1c	17	#include <asm/unaligned.h>
8b975bd3	18	#include <linux/lzo.h>
64c70b1c RP	19	#include "lzodefs.h"
	20
	21	static noinline size_t
8b975bd3 MO	22	lzo1x_1_do_compress(const unsigned char *in, size_t in_len,
8b975bd3 MO	23	unsigned char out, size_t out_len,
45ec975e DR	24	size_t ti, void wrkmem, signed char state_offset,
45ec975e DR	25	const unsigned char bitstream_version)
64c70b1c	26	{
8b975bd3 MO	27	const unsigned char *ip;
8b975bd3 MO	28	unsigned char *op;
64c70b1c	29	const unsigned char * const in_end = in + in_len;
8b975bd3 MO	30	const unsigned char * const ip_end = in + in_len - 20;
	31	const unsigned char *ii;
	32	lzo_dict_t * const dict = (lzo_dict_t *) wrkmem;
64c70b1c	33
8b975bd3 MO	34	op = out;
	35	ip = in;
	36	ii = ip;
	37	ip += ti < 4 ? 4 - ti : 0;
64c70b1c RP	38
64c70b1c RP	39	for (;;) {
5ee4014a	40	const unsigned char *m_pos = NULL;
8b975bd3 MO	41	size_t t, m_len, m_off;
8b975bd3 MO	42	u32 dv;
5ee4014a	43	u32 run_length = 0;
64c70b1c	44	literal:
8b975bd3 MO	45	ip += 1 + ((ip - ii) >> 5);
8b975bd3 MO	46	next:
64c70b1c RP	47	if (unlikely(ip >= ip_end))
64c70b1c RP	48	break;
8b975bd3	49	dv = get_unaligned_le32(ip);
5ee4014a	50
45ec975e	51	if (dv == 0 && bitstream_version) {
5ee4014a	52	const unsigned char *ir = ip + 4;
a10c9ede	53	const unsigned char *limit = min(ip_end, ip + MAX_ZERO_RUN_LENGTH + 1);
5ee4014a DR	54	#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && \
	55	defined(LZO_FAST_64BIT_MEMORY_ACCESS)
	56	u64 dv64;
	57
	58	for (; (ir + 32) <= limit; ir += 32) {
	59	dv64 = get_unaligned((u64 *)ir);
	60	dv64 \|= get_unaligned((u64 *)ir + 1);
	61	dv64 \|= get_unaligned((u64 *)ir + 2);
	62	dv64 \|= get_unaligned((u64 *)ir + 3);
	63	if (dv64)
	64	break;
	65	}
	66	for (; (ir + 8) <= limit; ir += 8) {
	67	dv64 = get_unaligned((u64 *)ir);
	68	if (dv64) {
	69	# if defined(__LITTLE_ENDIAN)
	70	ir += __builtin_ctzll(dv64) >> 3;
	71	# elif defined(__BIG_ENDIAN)
	72	ir += __builtin_clzll(dv64) >> 3;
	73	# else
	74	# error "missing endian definition"
	75	# endif
	76	break;
	77	}
	78	}
	79	#else
	80	while ((ir < (const unsigned char *)
	81	ALIGN((uintptr_t)ir, 4)) &&
	82	(ir < limit) && (*ir == 0))
	83	ir++;
09b35b41 DR	84	if (IS_ALIGNED((uintptr_t)ir, 4)) {
	85	for (; (ir + 4) <= limit; ir += 4) {
	86	dv = ((u32 )ir);
	87	if (dv) {
5ee4014a	88	# if defined(__LITTLE_ENDIAN)
09b35b41	89	ir += __builtin_ctz(dv) >> 3;
5ee4014a	90	# elif defined(__BIG_ENDIAN)
09b35b41	91	ir += __builtin_clz(dv) >> 3;
5ee4014a DR	92	# else
	93	# error "missing endian definition"
	94	# endif
09b35b41 DR	95	break;
09b35b41 DR	96	}
5ee4014a DR	97	}
	98	}
	99	#endif
	100	while (likely(ir < limit) && unlikely(*ir == 0))
	101	ir++;
	102	run_length = ir - ip;
	103	if (run_length > MAX_ZERO_RUN_LENGTH)
	104	run_length = MAX_ZERO_RUN_LENGTH;
	105	} else {
	106	t = ((dv * 0x1824429d) >> (32 - D_BITS)) & D_MASK;
	107	m_pos = in + dict[t];
	108	dict[t] = (lzo_dict_t) (ip - in);
	109	if (unlikely(dv != get_unaligned_le32(m_pos)))
	110	goto literal;
	111	}
64c70b1c	112
8b975bd3 MO	113	ii -= ti;
	114	ti = 0;
	115	t = ip - ii;
	116	if (t != 0) {
64c70b1c	117	if (t <= 3) {
5ee4014a	118	op[*state_offset] \|= t;
8b975bd3 MO	119	COPY4(op, ii);
	120	op += t;
	121	} else if (t <= 16) {
64c70b1c	122	*op++ = (t - 3);
8b975bd3 MO	123	COPY8(op, ii);
	124	COPY8(op + 8, ii + 8);
	125	op += t;
64c70b1c	126	} else {
8b975bd3 MO	127	if (t <= 18) {
	128	*op++ = (t - 3);
	129	} else {
	130	size_t tt = t - 18;
64c70b1c	131	*op++ = 0;
8b975bd3 MO	132	while (unlikely(tt > 255)) {
	133	tt -= 255;
	134	*op++ = 0;
	135	}
	136	*op++ = tt;
64c70b1c	137	}
8b975bd3 MO	138	do {
	139	COPY8(op, ii);
	140	COPY8(op + 8, ii + 8);
	141	op += 16;
	142	ii += 16;
	143	t -= 16;
	144	} while (t >= 16);
	145	if (t > 0) do {
	146	op++ = ii++;
	147	} while (--t > 0);
64c70b1c	148	}
64c70b1c RP	149	}
64c70b1c RP	150
5ee4014a DR	151	if (unlikely(run_length)) {
	152	ip += run_length;
	153	run_length -= MIN_ZERO_RUN_LENGTH;
	154	put_unaligned_le32((run_length << 21) \| 0xfffc18
	155	\| (run_length & 0x7), op);
	156	op += 4;
	157	run_length = 0;
	158	*state_offset = -3;
	159	goto finished_writing_instruction;
	160	}
	161
8b975bd3 MO	162	m_len = 4;
	163	{
	164	#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && defined(LZO_USE_CTZ64)
	165	u64 v;
	166	v = get_unaligned((const u64 *) (ip + m_len)) ^
	167	get_unaligned((const u64 *) (m_pos + m_len));
	168	if (unlikely(v == 0)) {
	169	do {
	170	m_len += 8;
	171	v = get_unaligned((const u64 *) (ip + m_len)) ^
	172	get_unaligned((const u64 *) (m_pos + m_len));
	173	if (unlikely(ip + m_len >= ip_end))
	174	goto m_len_done;
	175	} while (v == 0);
	176	}
	177	# if defined(__LITTLE_ENDIAN)
	178	m_len += (unsigned) __builtin_ctzll(v) / 8;
	179	# elif defined(__BIG_ENDIAN)
	180	m_len += (unsigned) __builtin_clzll(v) / 8;
	181	# else
	182	# error "missing endian definition"
	183	# endif
	184	#elif defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && defined(LZO_USE_CTZ32)
	185	u32 v;
	186	v = get_unaligned((const u32 *) (ip + m_len)) ^
	187	get_unaligned((const u32 *) (m_pos + m_len));
	188	if (unlikely(v == 0)) {
	189	do {
	190	m_len += 4;
	191	v = get_unaligned((const u32 *) (ip + m_len)) ^
	192	get_unaligned((const u32 *) (m_pos + m_len));
	193	if (v != 0)
	194	break;
	195	m_len += 4;
	196	v = get_unaligned((const u32 *) (ip + m_len)) ^
	197	get_unaligned((const u32 *) (m_pos + m_len));
	198	if (unlikely(ip + m_len >= ip_end))
	199	goto m_len_done;
	200	} while (v == 0);
	201	}
	202	# if defined(__LITTLE_ENDIAN)
	203	m_len += (unsigned) __builtin_ctz(v) / 8;
	204	# elif defined(__BIG_ENDIAN)
	205	m_len += (unsigned) __builtin_clz(v) / 8;
	206	# else
	207	# error "missing endian definition"
	208	# endif
	209	#else
	210	if (unlikely(ip[m_len] == m_pos[m_len])) {
	211	do {
	212	m_len += 1;
	213	if (ip[m_len] != m_pos[m_len])
	214	break;
	215	m_len += 1;
	216	if (ip[m_len] != m_pos[m_len])
	217	break;
	218	m_len += 1;
	219	if (ip[m_len] != m_pos[m_len])
	220	break;
	221	m_len += 1;
	222	if (ip[m_len] != m_pos[m_len])
	223	break;
	224	m_len += 1;
	225	if (ip[m_len] != m_pos[m_len])
226	break;
227	m_len += 1;
228	if (ip[m_len] != m_pos[m_len])
229	break;
230	m_len += 1;
231	if (ip[m_len] != m_pos[m_len])
232	break;
233	m_len += 1;
234	if (unlikely(ip + m_len >= ip_end))
235	goto m_len_done;
236	} while (ip[m_len] == m_pos[m_len]);
237	}
238	#endif
239	}
240	m_len_done:
64c70b1c	241
8b975bd3 MO	242	m_off = ip - m_pos;
8b975bd3 MO	243	ip += m_len;
8b975bd3 MO	244	if (m_len <= M2_MAX_LEN && m_off <= M2_MAX_OFFSET) {
	245	m_off -= 1;
	246	*op++ = (((m_len - 1) << 5) \| ((m_off & 7) << 2));
	247	*op++ = (m_off >> 3);
	248	} else if (m_off <= M3_MAX_OFFSET) {
	249	m_off -= 1;
	250	if (m_len <= M3_MAX_LEN)
64c70b1c	251	*op++ = (M3_MARKER \| (m_len - 2));
8b975bd3 MO	252	else {
	253	m_len -= M3_MAX_LEN;
	254	*op++ = M3_MARKER \| 0;
	255	while (unlikely(m_len > 255)) {
	256	m_len -= 255;
	257	*op++ = 0;
	258	}
	259	*op++ = (m_len);
64c70b1c	260	}
8b975bd3 MO	261	*op++ = (m_off << 2);
8b975bd3 MO	262	*op++ = (m_off >> 6);
64c70b1c	263	} else {
8b975bd3 MO	264	m_off -= 0x4000;
	265	if (m_len <= M4_MAX_LEN)
	266	*op++ = (M4_MARKER \| ((m_off >> 11) & 8)
64c70b1c	267	\| (m_len - 2));
8b975bd3	268	else {
b5265c81 DR	269	if (unlikely(((m_off & 0x403f) == 0x403f)
	270	&& (m_len >= 261)
	271	&& (m_len <= 264))
	272	&& likely(bitstream_version)) {
	273	// Under lzo-rle, block copies
	274	// for 261 <= length <= 264 and
	275	// (distance & 0x80f3) == 0x80f3
	276	// can result in ambiguous
	277	// output. Adjust length
	278	// to 260 to prevent ambiguity.
	279	ip -= m_len - 260;
	280	m_len = 260;
	281	}
8b975bd3 MO	282	m_len -= M4_MAX_LEN;
	283	*op++ = (M4_MARKER \| ((m_off >> 11) & 8));
	284	while (unlikely(m_len > 255)) {
	285	m_len -= 255;
	286	*op++ = 0;
64c70b1c	287	}
8b975bd3	288	*op++ = (m_len);
64c70b1c	289	}
8b975bd3	290	*op++ = (m_off << 2);
64c70b1c RP	291	*op++ = (m_off >> 6);
64c70b1c RP	292	}
5ee4014a DR	293	*state_offset = -2;
	294	finished_writing_instruction:
	295	ii = ip;
8b975bd3	296	goto next;
64c70b1c	297	}
64c70b1c	298	*out_len = op - out;
8b975bd3	299	return in_end - (ii - ti);
64c70b1c RP	300	}
64c70b1c RP	301
33fb626b	302	static int lzogeneric1x_1_compress(const unsigned char *in, size_t in_len,
8b975bd3	303	unsigned char out, size_t out_len,
45ec975e	304	void *wrkmem, const unsigned char bitstream_version)
64c70b1c	305	{
8b975bd3	306	const unsigned char *ip = in;
64c70b1c	307	unsigned char *op = out;
b11ed18e	308	unsigned char *data_start;
8b975bd3 MO	309	size_t l = in_len;
8b975bd3 MO	310	size_t t = 0;
5ee4014a	311	signed char state_offset = -2;
45ec975e	312	unsigned int m4_max_offset;
5ee4014a	313
b11ed18e DR	314	// LZO v0 will never write 17 as first byte (except for zero-length
b11ed18e DR	315	// input), so this is used to version the bitstream
45ec975e DR	316	if (bitstream_version > 0) {
	317	*op++ = 17;
	318	*op++ = bitstream_version;
	319	m4_max_offset = M4_MAX_OFFSET_V1;
	320	} else {
	321	m4_max_offset = M4_MAX_OFFSET_V0;
	322	}
64c70b1c	323
b11ed18e DR	324	data_start = op;
b11ed18e DR	325
8b975bd3	326	while (l > 20) {
a10c9ede	327	size_t ll = min_t(size_t, l, m4_max_offset + 1);
8b975bd3 MO	328	uintptr_t ll_end = (uintptr_t) ip + ll;
	329	if ((ll_end + ((t + ll) >> 5)) <= ll_end)
	330	break;
	331	BUILD_BUG_ON(D_SIZE * sizeof(lzo_dict_t) > LZO1X_1_MEM_COMPRESS);
	332	memset(wrkmem, 0, D_SIZE * sizeof(lzo_dict_t));
45ec975e DR	333	t = lzo1x_1_do_compress(ip, ll, op, out_len, t, wrkmem,
45ec975e DR	334	&state_offset, bitstream_version);
8b975bd3	335	ip += ll;
64c70b1c	336	op += *out_len;
8b975bd3	337	l -= ll;
64c70b1c	338	}
8b975bd3	339	t += l;
64c70b1c RP	340
64c70b1c RP	341	if (t > 0) {
8b975bd3	342	const unsigned char *ii = in + in_len - t;
64c70b1c	343
b11ed18e	344	if (op == data_start && t <= 238) {
64c70b1c RP	345	*op++ = (17 + t);
64c70b1c RP	346	} else if (t <= 3) {
5ee4014a	347	op[state_offset] \|= t;
64c70b1c RP	348	} else if (t <= 18) {
	349	*op++ = (t - 3);
	350	} else {
	351	size_t tt = t - 18;
64c70b1c RP	352	*op++ = 0;
	353	while (tt > 255) {
	354	tt -= 255;
	355	*op++ = 0;
	356	}
64c70b1c RP	357	*op++ = tt;
64c70b1c RP	358	}
8b975bd3 MO	359	if (t >= 16) do {
	360	COPY8(op, ii);
	361	COPY8(op + 8, ii + 8);
	362	op += 16;
	363	ii += 16;
	364	t -= 16;
	365	} while (t >= 16);
	366	if (t > 0) do {
64c70b1c RP	367	op++ = ii++;
	368	} while (--t > 0);
	369	}
	370
	371	*op++ = M4_MARKER \| 1;
	372	*op++ = 0;
	373	*op++ = 0;
	374
	375	*out_len = op - out;
	376	return LZO_E_OK;
	377	}
45ec975e DR	378
	379	int lzo1x_1_compress(const unsigned char *in, size_t in_len,
	380	unsigned char out, size_t out_len,
	381	void *wrkmem)
	382	{
	383	return lzogeneric1x_1_compress(in, in_len, out, out_len, wrkmem, 0);
	384	}
	385
	386	int lzorle1x_1_compress(const unsigned char *in, size_t in_len,
	387	unsigned char out, size_t out_len,
	388	void *wrkmem)
	389	{
	390	return lzogeneric1x_1_compress(in, in_len, out, out_len,
	391	wrkmem, LZO_VERSION);
	392	}
	393
64c70b1c	394	EXPORT_SYMBOL_GPL(lzo1x_1_compress);
45ec975e	395	EXPORT_SYMBOL_GPL(lzorle1x_1_compress);
64c70b1c RP	396
	397	MODULE_LICENSE("GPL");
	398	MODULE_DESCRIPTION("LZO1X-1 Compressor");