[linux-2.6-block.git] / lib / zlib_inflate / inffast.c

/* inffast.c -- fast decoding
 * Copyright (C) 1995-2004 Mark Adler
 * For conditions of distribution and use, see copyright notice in zlib.h
 */

#include <linux/zutil.h>
#include "inftrees.h"
#include "inflate.h"
#include "inffast.h"

#ifndef ASMINF

union uu {
	unsigned short us;
	unsigned char b[2];
};

/* Endian independent version */
static inline unsigned short
get_unaligned16(const unsigned short *p)
{
	union uu  mm;
	unsigned char *b = (unsigned char *)p;

	mm.b[0] = b[0];
	mm.b[1] = b[1];
	return mm.us;
}

/*
   Decode literal, length, and distance codes and write out the resulting
   literal and match bytes until either not enough input or output is
   available, an end-of-block is encountered, or a data error is encountered.
   When large enough input and output buffers are supplied to inflate(), for
   example, a 16K input buffer and a 64K output buffer, more than 95% of the
   inflate execution time is spent in this routine.

   Entry assumptions:

        state->mode == LEN
        strm->avail_in >= 6
        strm->avail_out >= 258
        start >= strm->avail_out
        state->bits < 8

   On return, state->mode is one of:

        LEN -- ran out of enough output space or enough available input
        TYPE -- reached end of block code, inflate() to interpret next block
        BAD -- error in block data

   Notes:

    - The maximum input bits used by a length/distance pair is 15 bits for the
      length code, 5 bits for the length extra, 15 bits for the distance code,
      and 13 bits for the distance extra.  This totals 48 bits, or six bytes.
      Therefore if strm->avail_in >= 6, then there is enough input to avoid
      checking for available input while decoding.

    - The maximum bytes that a single length/distance pair can output is 258
      bytes, which is the maximum length that can be coded.  inflate_fast()
      requires strm->avail_out >= 258 for each loop to avoid checking for
      output space.

    - @start:	inflate()'s starting value for strm->avail_out
 */
void inflate_fast(z_streamp strm, unsigned start)
{
    struct inflate_state *state;
    const unsigned char *in;    /* local strm->next_in */
    const unsigned char *last;  /* while in < last, enough input available */
    unsigned char *out;         /* local strm->next_out */
    unsigned char *beg;         /* inflate()'s initial strm->next_out */
    unsigned char *end;         /* while out < end, enough space available */
#ifdef INFLATE_STRICT
    unsigned dmax;              /* maximum distance from zlib header */
#endif
    unsigned wsize;             /* window size or zero if not using window */
    unsigned whave;             /* valid bytes in the window */
    unsigned write;             /* window write index */
    unsigned char *window;      /* allocated sliding window, if wsize != 0 */
    unsigned long hold;         /* local strm->hold */
    unsigned bits;              /* local strm->bits */
    code const *lcode;          /* local strm->lencode */
    code const *dcode;          /* local strm->distcode */
    unsigned lmask;             /* mask for first level of length codes */
    unsigned dmask;             /* mask for first level of distance codes */
    code this;                  /* retrieved table entry */
    unsigned op;                /* code bits, operation, extra bits, or */
                                /*  window position, window bytes to copy */
    unsigned len;               /* match length, unused bytes */
    unsigned dist;              /* match distance */
    unsigned char *from;        /* where to copy match from */

    /* copy state to local variables */
    state = (struct inflate_state *)strm->state;
    in = strm->next_in;
    last = in + (strm->avail_in - 5);
    out = strm->next_out;
    beg = out - (start - strm->avail_out);
    end = out + (strm->avail_out - 257);
#ifdef INFLATE_STRICT
    dmax = state->dmax;
#endif
    wsize = state->wsize;
    whave = state->whave;
    write = state->write;
    window = state->window;
    hold = state->hold;
    bits = state->bits;
    lcode = state->lencode;
    dcode = state->distcode;
    lmask = (1U << state->lenbits) - 1;
    dmask = (1U << state->distbits) - 1;

    /* decode literals and length/distances until end-of-block or not enough
       input data or output space */
    do {
        if (bits < 15) {
            hold += (unsigned long)(*in++) << bits;
            bits += 8;
            hold += (unsigned long)(*in++) << bits;
            bits += 8;
        }
        this = lcode[hold & lmask];
      dolen:
        op = (unsigned)(this.bits);
        hold >>= op;
        bits -= op;
        op = (unsigned)(this.op);
        if (op == 0) {                          /* literal */
            *out++ = (unsigned char)(this.val);
        }
        else if (op & 16) {                     /* length base */
            len = (unsigned)(this.val);
            op &= 15;                           /* number of extra bits */
            if (op) {
                if (bits < op) {
                    hold += (unsigned long)(*in++) << bits;
                    bits += 8;
                }
                len += (unsigned)hold & ((1U << op) - 1);
                hold >>= op;
                bits -= op;
            }
            if (bits < 15) {
                hold += (unsigned long)(*in++) << bits;
                bits += 8;
                hold += (unsigned long)(*in++) << bits;
                bits += 8;
            }
            this = dcode[hold & dmask];
          dodist:
            op = (unsigned)(this.bits);
            hold >>= op;
            bits -= op;
            op = (unsigned)(this.op);
            if (op & 16) {                      /* distance base */
                dist = (unsigned)(this.val);
                op &= 15;                       /* number of extra bits */
                if (bits < op) {
                    hold += (unsigned long)(*in++) << bits;
                    bits += 8;
                    if (bits < op) {
                        hold += (unsigned long)(*in++) << bits;
                        bits += 8;
                    }
                }
                dist += (unsigned)hold & ((1U << op) - 1);
#ifdef INFLATE_STRICT
                if (dist > dmax) {
                    strm->msg = (char *)"invalid distance too far back";
                    state->mode = BAD;
                    break;
                }
#endif
                hold >>= op;
                bits -= op;
                op = (unsigned)(out - beg);     /* max distance in output */
                if (dist > op) {                /* see if copy from window */
                    op = dist - op;             /* distance back in window */
                    if (op > whave) {
                        strm->msg = (char *)"invalid distance too far back";
                        state->mode = BAD;
                        break;
                    }
                    from = window;
                    if (write == 0) {           /* very common case */
                        from += wsize - op;
                        if (op < len) {         /* some from window */
                            len -= op;
                            do {
                                *out++ = *from++;
                            } while (--op);
                            from = out - dist;  /* rest from output */
                        }
                    }
                    else if (write < op) {      /* wrap around window */
                        from += wsize + write - op;
                        op -= write;
                        if (op < len) {         /* some from end of window */
                            len -= op;
                            do {
                                *out++ = *from++;
                            } while (--op);
                            from = window;
                            if (write < len) {  /* some from start of window */
                                op = write;
                                len -= op;
                                do {
                                    *out++ = *from++;
                                } while (--op);
                                from = out - dist;      /* rest from output */
                            }
                        }
                    }
                    else {                      /* contiguous in window */
                        from += write - op;
                        if (op < len) {         /* some from window */
                            len -= op;
                            do {
                                *out++ = *from++;
                            } while (--op);
                            from = out - dist;  /* rest from output */
                        }
                    }
                    while (len > 2) {
                        *out++ = *from++;
                        *out++ = *from++;
                        *out++ = *from++;
                        len -= 3;
                    }
                    if (len) {
                        *out++ = *from++;
                        if (len > 1)
                            *out++ = *from++;
                    }
                }
                else {
		    unsigned short *sout;
		    unsigned long loops;

                    from = out - dist;          /* copy direct from output */
		    /* minimum length is three */
		    /* Align out addr */
		    if (!((long)(out - 1) & 1)) {
			*out++ = *from++;
			len--;
		    }
		    sout = (unsigned short *)(out);
		    if (dist > 2) {
			unsigned short *sfrom;

			sfrom = (unsigned short *)(from);
			loops = len >> 1;
			do {
			    if (IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS))
				*sout++ = *sfrom++;
			    else
				*sout++ = get_unaligned16(sfrom++);
			} while (--loops);
			out = (unsigned char *)sout;
			from = (unsigned char *)sfrom;
		    } else { /* dist == 1 or dist == 2 */
			unsigned short pat16;

			pat16 = *(sout-1);
			if (dist == 1) {
				union uu mm;
				/* copy one char pattern to both bytes */
				mm.us = pat16;
				mm.b[0] = mm.b[1];
				pat16 = mm.us;
			}
			loops = len >> 1;
			do
			    *sout++ = pat16;
			while (--loops);
			out = (unsigned char *)sout;
		    }
		    if (len & 1)
			*out++ = *from++;
                }
            }
            else if ((op & 64) == 0) {          /* 2nd level distance code */
                this = dcode[this.val + (hold & ((1U << op) - 1))];
                goto dodist;
            }
            else {
                strm->msg = (char *)"invalid distance code";
                state->mode = BAD;
                break;
            }
        }
        else if ((op & 64) == 0) {              /* 2nd level length code */
            this = lcode[this.val + (hold & ((1U << op) - 1))];
            goto dolen;
        }
        else if (op & 32) {                     /* end-of-block */
            state->mode = TYPE;
            break;
        }
        else {
            strm->msg = (char *)"invalid literal/length code";
            state->mode = BAD;
            break;
        }
    } while (in < last && out < end);

    /* return unused bytes (on entry, bits < 8, so in won't go too far back) */
    len = bits >> 3;
    in -= len;
    bits -= len << 3;
    hold &= (1U << bits) - 1;

    /* update state and return */
    strm->next_in = in;
    strm->next_out = out;
    strm->avail_in = (unsigned)(in < last ? 5 + (last - in) : 5 - (in - last));
    strm->avail_out = (unsigned)(out < end ?
                                 257 + (end - out) : 257 - (out - end));
    state->hold = hold;
    state->bits = bits;
    return;
}

/*
   inflate_fast() speedups that turned out slower (on a PowerPC G3 750CXe):
   - Using bit fields for code structure
   - Different op definition to avoid & for extra bits (do & for table bits)
   - Three separate decoding do-loops for direct, window, and write == 0
   - Special case for distance > 1 copies to do overlapped load and store copy
   - Explicit branch predictions (based on measured branch probabilities)
   - Deferring match copy and interspersed it with decoding subsequent codes
   - Swapping literal/length else
   - Swapping window/direct else
   - Larger unrolled copy loops (three is about right)
   - Moving len -= 3 statement into middle of loop
 */

#endif /* !ASMINF */
Commit	Line	Data
4f3865fb RP	1	/* inffast.c -- fast decoding
	2	* Copyright (C) 1995-2004 Mark Adler
	3	* For conditions of distribution and use, see copyright notice in zlib.h
1da177e4 LT	4	*/
	5
	6	#include <linux/zutil.h>
	7	#include "inftrees.h"
4f3865fb	8	#include "inflate.h"
1da177e4 LT	9	#include "inffast.h"
1da177e4 LT	10
4f3865fb RP	11	#ifndef ASMINF
4f3865fb RP	12
e69eae65 JT	13	union uu {
	14	unsigned short us;
	15	unsigned char b[2];
	16	};
	17
05911c5d	18	/* Endian independent version */
e69eae65 JT	19	static inline unsigned short
	20	get_unaligned16(const unsigned short *p)
	21	{
	22	union uu mm;
	23	unsigned char b = (unsigned char )p;
	24
	25	mm.b[0] = b[0];
	26	mm.b[1] = b[1];
	27	return mm.us;
	28	}
	29
4f3865fb RP	30	/*
	31	Decode literal, length, and distance codes and write out the resulting
	32	literal and match bytes until either not enough input or output is
	33	available, an end-of-block is encountered, or a data error is encountered.
	34	When large enough input and output buffers are supplied to inflate(), for
	35	example, a 16K input buffer and a 64K output buffer, more than 95% of the
	36	inflate execution time is spent in this routine.
	37
	38	Entry assumptions:
	39
	40	state->mode == LEN
	41	strm->avail_in >= 6
	42	strm->avail_out >= 258
	43	start >= strm->avail_out
	44	state->bits < 8
	45
	46	On return, state->mode is one of:
	47
	48	LEN -- ran out of enough output space or enough available input
	49	TYPE -- reached end of block code, inflate() to interpret next block
	50	BAD -- error in block data
	51
	52	Notes:
	53
	54	- The maximum input bits used by a length/distance pair is 15 bits for the
	55	length code, 5 bits for the length extra, 15 bits for the distance code,
	56	and 13 bits for the distance extra. This totals 48 bits, or six bytes.
	57	Therefore if strm->avail_in >= 6, then there is enough input to avoid
	58	checking for available input while decoding.
	59
	60	- The maximum bytes that a single length/distance pair can output is 258
	61	bytes, which is the maximum length that can be coded. inflate_fast()
	62	requires strm->avail_out >= 258 for each loop to avoid checking for
	63	output space.
b762450e RD	64
b762450e RD	65	- @start: inflate()'s starting value for strm->avail_out
4f3865fb	66	*/
b762450e	67	void inflate_fast(z_streamp strm, unsigned start)
1da177e4	68	{
4f3865fb	69	struct inflate_state *state;
8336793b DV	70	const unsigned char in; / local strm->next_in */
	71	const unsigned char last; / while in < last, enough input available */
	72	unsigned char out; / local strm->next_out */
	73	unsigned char beg; / inflate()'s initial strm->next_out */
	74	unsigned char end; / while out < end, enough space available */
4f3865fb RP	75	#ifdef INFLATE_STRICT
	76	unsigned dmax; /* maximum distance from zlib header */
	77	#endif
	78	unsigned wsize; /* window size or zero if not using window */
	79	unsigned whave; /* valid bytes in the window */
	80	unsigned write; /* window write index */
8336793b	81	unsigned char window; / allocated sliding window, if wsize != 0 */
4f3865fb RP	82	unsigned long hold; /* local strm->hold */
4f3865fb RP	83	unsigned bits; /* local strm->bits */
8336793b DV	84	code const lcode; / local strm->lencode */
8336793b DV	85	code const dcode; / local strm->distcode */
4f3865fb RP	86	unsigned lmask; /* mask for first level of length codes */
	87	unsigned dmask; /* mask for first level of distance codes */
	88	code this; /* retrieved table entry */
	89	unsigned op; /* code bits, operation, extra bits, or */
	90	/* window position, window bytes to copy */
	91	unsigned len; /* match length, unused bytes */
	92	unsigned dist; /* match distance */
8336793b	93	unsigned char from; / where to copy match from */
4f3865fb RP	94
	95	/* copy state to local variables */
	96	state = (struct inflate_state *)strm->state;
acaab733	97	in = strm->next_in;
4f3865fb	98	last = in + (strm->avail_in - 5);
acaab733	99	out = strm->next_out;
4f3865fb RP	100	beg = out - (start - strm->avail_out);
	101	end = out + (strm->avail_out - 257);
	102	#ifdef INFLATE_STRICT
	103	dmax = state->dmax;
	104	#endif
	105	wsize = state->wsize;
	106	whave = state->whave;
	107	write = state->write;
	108	window = state->window;
	109	hold = state->hold;
	110	bits = state->bits;
	111	lcode = state->lencode;
	112	dcode = state->distcode;
	113	lmask = (1U << state->lenbits) - 1;
	114	dmask = (1U << state->distbits) - 1;
	115
	116	/* decode literals and length/distances until end-of-block or not enough
	117	input data or output space */
1da177e4	118	do {
4f3865fb	119	if (bits < 15) {
acaab733	120	hold += (unsigned long)(*in++) << bits;
4f3865fb	121	bits += 8;
acaab733	122	hold += (unsigned long)(*in++) << bits;
4f3865fb RP	123	bits += 8;
	124	}
	125	this = lcode[hold & lmask];
	126	dolen:
	127	op = (unsigned)(this.bits);
	128	hold >>= op;
	129	bits -= op;
	130	op = (unsigned)(this.op);
	131	if (op == 0) { /* literal */
acaab733	132	*out++ = (unsigned char)(this.val);
4f3865fb RP	133	}
	134	else if (op & 16) { /* length base */
	135	len = (unsigned)(this.val);
	136	op &= 15; /* number of extra bits */
	137	if (op) {
	138	if (bits < op) {
acaab733	139	hold += (unsigned long)(*in++) << bits;
4f3865fb RP	140	bits += 8;
	141	}
	142	len += (unsigned)hold & ((1U << op) - 1);
	143	hold >>= op;
	144	bits -= op;
	145	}
	146	if (bits < 15) {
acaab733	147	hold += (unsigned long)(*in++) << bits;
4f3865fb	148	bits += 8;
acaab733	149	hold += (unsigned long)(*in++) << bits;
4f3865fb RP	150	bits += 8;
	151	}
	152	this = dcode[hold & dmask];
	153	dodist:
	154	op = (unsigned)(this.bits);
	155	hold >>= op;
	156	bits -= op;
	157	op = (unsigned)(this.op);
	158	if (op & 16) { /* distance base */
	159	dist = (unsigned)(this.val);
	160	op &= 15; /* number of extra bits */
	161	if (bits < op) {
acaab733	162	hold += (unsigned long)(*in++) << bits;
4f3865fb RP	163	bits += 8;
4f3865fb RP	164	if (bits < op) {
acaab733	165	hold += (unsigned long)(*in++) << bits;
4f3865fb RP	166	bits += 8;
	167	}
	168	}
	169	dist += (unsigned)hold & ((1U << op) - 1);
	170	#ifdef INFLATE_STRICT
	171	if (dist > dmax) {
	172	strm->msg = (char *)"invalid distance too far back";
	173	state->mode = BAD;
	174	break;
	175	}
	176	#endif
	177	hold >>= op;
	178	bits -= op;
	179	op = (unsigned)(out - beg); /* max distance in output */
	180	if (dist > op) { /* see if copy from window */
	181	op = dist - op; /* distance back in window */
	182	if (op > whave) {
	183	strm->msg = (char *)"invalid distance too far back";
	184	state->mode = BAD;
	185	break;
	186	}
acaab733	187	from = window;
4f3865fb RP	188	if (write == 0) { /* very common case */
	189	from += wsize - op;
	190	if (op < len) { /* some from window */
	191	len -= op;
	192	do {
acaab733	193	out++ = from++;
4f3865fb RP	194	} while (--op);
	195	from = out - dist; /* rest from output */
	196	}
	197	}
	198	else if (write < op) { /* wrap around window */
	199	from += wsize + write - op;
	200	op -= write;
	201	if (op < len) { /* some from end of window */
	202	len -= op;
	203	do {
acaab733	204	out++ = from++;
4f3865fb	205	} while (--op);
acaab733	206	from = window;
4f3865fb RP	207	if (write < len) { /* some from start of window */
	208	op = write;
	209	len -= op;
	210	do {
acaab733	211	out++ = from++;
4f3865fb RP	212	} while (--op);
	213	from = out - dist; /* rest from output */
	214	}
	215	}
	216	}
	217	else { /* contiguous in window */
	218	from += write - op;
	219	if (op < len) { /* some from window */
	220	len -= op;
	221	do {
acaab733	222	out++ = from++;
4f3865fb RP	223	} while (--op);
	224	from = out - dist; /* rest from output */
	225	}
	226	}
	227	while (len > 2) {
acaab733 JH	228	out++ = from++;
	229	out++ = from++;
	230	out++ = from++;
4f3865fb RP	231	len -= 3;
	232	}
	233	if (len) {
acaab733	234	out++ = from++;
4f3865fb	235	if (len > 1)
acaab733	236	out++ = from++;
4f3865fb RP	237	}
	238	}
	239	else {
ac4c2a3b JT	240	unsigned short *sout;
	241	unsigned long loops;
	242
4f3865fb	243	from = out - dist; /* copy direct from output */
ac4c2a3b JT	244	/* minimum length is three */
ac4c2a3b JT	245	/* Align out addr */
acaab733 JH	246	if (!((long)(out - 1) & 1)) {
acaab733 JH	247	out++ = from++;
ac4c2a3b JT	248	len--;
ac4c2a3b JT	249	}
acaab733	250	sout = (unsigned short *)(out);
ac4c2a3b JT	251	if (dist > 2) {
	252	unsigned short *sfrom;
	253
acaab733	254	sfrom = (unsigned short *)(from);
ac4c2a3b	255	loops = len >> 1;
b7cd9fa5 PM	256	do {
	257	if (IS_ENABLED(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS))
	258	sout++ = sfrom++;
	259	else
	260	*sout++ = get_unaligned16(sfrom++);
	261	} while (--loops);
acaab733 JH	262	out = (unsigned char *)sout;
acaab733 JH	263	from = (unsigned char *)sfrom;
ac4c2a3b JT	264	} else { /* dist == 1 or dist == 2 */
	265	unsigned short pat16;
	266
acaab733	267	pat16 = *(sout-1);
e69eae65 JT	268	if (dist == 1) {
	269	union uu mm;
	270	/* copy one char pattern to both bytes */
	271	mm.us = pat16;
	272	mm.b[0] = mm.b[1];
	273	pat16 = mm.us;
	274	}
ac4c2a3b JT	275	loops = len >> 1;
ac4c2a3b JT	276	do
acaab733	277	*sout++ = pat16;
ac4c2a3b	278	while (--loops);
acaab733	279	out = (unsigned char *)sout;
ac4c2a3b JT	280	}
ac4c2a3b JT	281	if (len & 1)
acaab733	282	out++ = from++;
4f3865fb RP	283	}
	284	}
	285	else if ((op & 64) == 0) { /* 2nd level distance code */
	286	this = dcode[this.val + (hold & ((1U << op) - 1))];
	287	goto dodist;
1da177e4	288	}
4f3865fb RP	289	else {
	290	strm->msg = (char *)"invalid distance code";
	291	state->mode = BAD;
	292	break;
1da177e4	293	}
4f3865fb RP	294	}
	295	else if ((op & 64) == 0) { /* 2nd level length code */
	296	this = lcode[this.val + (hold & ((1U << op) - 1))];
	297	goto dolen;
	298	}
	299	else if (op & 32) { /* end-of-block */
	300	state->mode = TYPE;
1da177e4	301	break;
1da177e4	302	}
4f3865fb RP	303	else {
	304	strm->msg = (char *)"invalid literal/length code";
	305	state->mode = BAD;
	306	break;
	307	}
	308	} while (in < last && out < end);
	309
	310	/* return unused bytes (on entry, bits < 8, so in won't go too far back) */
	311	len = bits >> 3;
	312	in -= len;
	313	bits -= len << 3;
	314	hold &= (1U << bits) - 1;
	315
	316	/* update state and return */
acaab733 JH	317	strm->next_in = in;
acaab733 JH	318	strm->next_out = out;
4f3865fb RP	319	strm->avail_in = (unsigned)(in < last ? 5 + (last - in) : 5 - (in - last));
	320	strm->avail_out = (unsigned)(out < end ?
	321	257 + (end - out) : 257 - (out - end));
	322	state->hold = hold;
	323	state->bits = bits;
	324	return;
1da177e4	325	}
4f3865fb RP	326
	327	/*
	328	inflate_fast() speedups that turned out slower (on a PowerPC G3 750CXe):
	329	- Using bit fields for code structure
	330	- Different op definition to avoid & for extra bits (do & for table bits)
	331	- Three separate decoding do-loops for direct, window, and write == 0
	332	- Special case for distance > 1 copies to do overlapped load and store copy
	333	- Explicit branch predictions (based on measured branch probabilities)
	334	- Deferring match copy and interspersed it with decoding subsequent codes
	335	- Swapping literal/length else
	336	- Swapping window/direct else
	337	- Larger unrolled copy loops (three is about right)
	338	- Moving len -= 3 statement into middle of loop
	339	*/
	340
	341	#endif /* !ASMINF */