Dan Streetman | 2da572c | 2015-05-07 13:49:14 -0400 | [diff] [blame] | 1 | /* |
| 2 | * 842 Software Decompression |
| 3 | * |
| 4 | * Copyright (C) 2015 Dan Streetman, IBM Corp |
| 5 | * |
| 6 | * This program is free software; you can redistribute it and/or modify |
| 7 | * it under the terms of the GNU General Public License as published by |
| 8 | * the Free Software Foundation; either version 2 of the License, or |
| 9 | * (at your option) any later version. |
| 10 | * |
| 11 | * This program is distributed in the hope that it will be useful, |
| 12 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 13 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 14 | * GNU General Public License for more details. |
| 15 | * |
| 16 | * See 842.h for details of the 842 compressed format. |
| 17 | */ |
| 18 | |
| 19 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt |
| 20 | #define MODULE_NAME "842_decompress" |
| 21 | |
| 22 | #include "842.h" |
| 23 | #include "842_debugfs.h" |
| 24 | |
| 25 | /* rolling fifo sizes */ |
| 26 | #define I2_FIFO_SIZE (2 * (1 << I2_BITS)) |
| 27 | #define I4_FIFO_SIZE (4 * (1 << I4_BITS)) |
| 28 | #define I8_FIFO_SIZE (8 * (1 << I8_BITS)) |
| 29 | |
| 30 | static u8 decomp_ops[OPS_MAX][4] = { |
| 31 | { D8, N0, N0, N0 }, |
| 32 | { D4, D2, I2, N0 }, |
| 33 | { D4, I2, D2, N0 }, |
| 34 | { D4, I2, I2, N0 }, |
| 35 | { D4, I4, N0, N0 }, |
| 36 | { D2, I2, D4, N0 }, |
| 37 | { D2, I2, D2, I2 }, |
| 38 | { D2, I2, I2, D2 }, |
| 39 | { D2, I2, I2, I2 }, |
| 40 | { D2, I2, I4, N0 }, |
| 41 | { I2, D2, D4, N0 }, |
| 42 | { I2, D4, I2, N0 }, |
| 43 | { I2, D2, I2, D2 }, |
| 44 | { I2, D2, I2, I2 }, |
| 45 | { I2, D2, I4, N0 }, |
| 46 | { I2, I2, D4, N0 }, |
| 47 | { I2, I2, D2, I2 }, |
| 48 | { I2, I2, I2, D2 }, |
| 49 | { I2, I2, I2, I2 }, |
| 50 | { I2, I2, I4, N0 }, |
| 51 | { I4, D4, N0, N0 }, |
| 52 | { I4, D2, I2, N0 }, |
| 53 | { I4, I2, D2, N0 }, |
| 54 | { I4, I2, I2, N0 }, |
| 55 | { I4, I4, N0, N0 }, |
| 56 | { I8, N0, N0, N0 } |
| 57 | }; |
| 58 | |
| 59 | struct sw842_param { |
| 60 | u8 *in; |
| 61 | u8 bit; |
| 62 | u64 ilen; |
| 63 | u8 *out; |
| 64 | u8 *ostart; |
| 65 | u64 olen; |
| 66 | }; |
| 67 | |
| 68 | #define beN_to_cpu(d, s) \ |
| 69 | ((s) == 2 ? be16_to_cpu(get_unaligned((__be16 *)d)) : \ |
| 70 | (s) == 4 ? be32_to_cpu(get_unaligned((__be32 *)d)) : \ |
| 71 | (s) == 8 ? be64_to_cpu(get_unaligned((__be64 *)d)) : \ |
| 72 | WARN(1, "pr_debug param err invalid size %x\n", s)) |
| 73 | |
| 74 | static int next_bits(struct sw842_param *p, u64 *d, u8 n); |
| 75 | |
| 76 | static int __split_next_bits(struct sw842_param *p, u64 *d, u8 n, u8 s) |
| 77 | { |
| 78 | u64 tmp = 0; |
| 79 | int ret; |
| 80 | |
| 81 | if (n <= s) { |
| 82 | pr_debug("split_next_bits invalid n %u s %u\n", n, s); |
| 83 | return -EINVAL; |
| 84 | } |
| 85 | |
| 86 | ret = next_bits(p, &tmp, n - s); |
| 87 | if (ret) |
| 88 | return ret; |
| 89 | ret = next_bits(p, d, s); |
| 90 | if (ret) |
| 91 | return ret; |
| 92 | *d |= tmp << s; |
| 93 | return 0; |
| 94 | } |
| 95 | |
| 96 | static int next_bits(struct sw842_param *p, u64 *d, u8 n) |
| 97 | { |
| 98 | u8 *in = p->in, b = p->bit, bits = b + n; |
| 99 | |
| 100 | if (n > 64) { |
| 101 | pr_debug("next_bits invalid n %u\n", n); |
| 102 | return -EINVAL; |
| 103 | } |
| 104 | |
| 105 | /* split this up if reading > 8 bytes, or if we're at the end of |
| 106 | * the input buffer and would read past the end |
| 107 | */ |
| 108 | if (bits > 64) |
| 109 | return __split_next_bits(p, d, n, 32); |
| 110 | else if (p->ilen < 8 && bits > 32 && bits <= 56) |
| 111 | return __split_next_bits(p, d, n, 16); |
| 112 | else if (p->ilen < 4 && bits > 16 && bits <= 24) |
| 113 | return __split_next_bits(p, d, n, 8); |
| 114 | |
| 115 | if (DIV_ROUND_UP(bits, 8) > p->ilen) |
| 116 | return -EOVERFLOW; |
| 117 | |
| 118 | if (bits <= 8) |
| 119 | *d = *in >> (8 - bits); |
| 120 | else if (bits <= 16) |
| 121 | *d = be16_to_cpu(get_unaligned((__be16 *)in)) >> (16 - bits); |
| 122 | else if (bits <= 32) |
| 123 | *d = be32_to_cpu(get_unaligned((__be32 *)in)) >> (32 - bits); |
| 124 | else |
| 125 | *d = be64_to_cpu(get_unaligned((__be64 *)in)) >> (64 - bits); |
| 126 | |
| 127 | *d &= GENMASK_ULL(n - 1, 0); |
| 128 | |
| 129 | p->bit += n; |
| 130 | |
| 131 | if (p->bit > 7) { |
| 132 | p->in += p->bit / 8; |
| 133 | p->ilen -= p->bit / 8; |
| 134 | p->bit %= 8; |
| 135 | } |
| 136 | |
| 137 | return 0; |
| 138 | } |
| 139 | |
| 140 | static int do_data(struct sw842_param *p, u8 n) |
| 141 | { |
| 142 | u64 v; |
| 143 | int ret; |
| 144 | |
| 145 | if (n > p->olen) |
| 146 | return -ENOSPC; |
| 147 | |
| 148 | ret = next_bits(p, &v, n * 8); |
| 149 | if (ret) |
| 150 | return ret; |
| 151 | |
| 152 | switch (n) { |
| 153 | case 2: |
| 154 | put_unaligned(cpu_to_be16((u16)v), (__be16 *)p->out); |
| 155 | break; |
| 156 | case 4: |
| 157 | put_unaligned(cpu_to_be32((u32)v), (__be32 *)p->out); |
| 158 | break; |
| 159 | case 8: |
| 160 | put_unaligned(cpu_to_be64((u64)v), (__be64 *)p->out); |
| 161 | break; |
| 162 | default: |
| 163 | return -EINVAL; |
| 164 | } |
| 165 | |
| 166 | p->out += n; |
| 167 | p->olen -= n; |
| 168 | |
| 169 | return 0; |
| 170 | } |
| 171 | |
| 172 | static int __do_index(struct sw842_param *p, u8 size, u8 bits, u64 fsize) |
| 173 | { |
| 174 | u64 index, offset, total = round_down(p->out - p->ostart, 8); |
| 175 | int ret; |
| 176 | |
| 177 | ret = next_bits(p, &index, bits); |
| 178 | if (ret) |
| 179 | return ret; |
| 180 | |
| 181 | offset = index * size; |
| 182 | |
| 183 | /* a ring buffer of fsize is used; correct the offset */ |
| 184 | if (total > fsize) { |
| 185 | /* this is where the current fifo is */ |
| 186 | u64 section = round_down(total, fsize); |
| 187 | /* the current pos in the fifo */ |
Dan Streetman | ca7fc7e | 2015-05-11 18:53:36 -0400 | [diff] [blame] | 188 | u64 pos = total - section; |
Dan Streetman | 2da572c | 2015-05-07 13:49:14 -0400 | [diff] [blame] | 189 | |
| 190 | /* if the offset is past/at the pos, we need to |
| 191 | * go back to the last fifo section |
| 192 | */ |
| 193 | if (offset >= pos) |
| 194 | section -= fsize; |
| 195 | |
| 196 | offset += section; |
| 197 | } |
| 198 | |
| 199 | if (offset + size > total) { |
| 200 | pr_debug("index%x %lx points past end %lx\n", size, |
| 201 | (unsigned long)offset, (unsigned long)total); |
| 202 | return -EINVAL; |
| 203 | } |
| 204 | |
| 205 | pr_debug("index%x to %lx off %lx adjoff %lx tot %lx data %lx\n", |
| 206 | size, (unsigned long)index, (unsigned long)(index * size), |
| 207 | (unsigned long)offset, (unsigned long)total, |
| 208 | (unsigned long)beN_to_cpu(&p->ostart[offset], size)); |
| 209 | |
| 210 | memcpy(p->out, &p->ostart[offset], size); |
| 211 | p->out += size; |
| 212 | p->olen -= size; |
| 213 | |
| 214 | return 0; |
| 215 | } |
| 216 | |
Dan Streetman | f7ead7b | 2015-05-11 07:22:35 -0400 | [diff] [blame] | 217 | static int do_index(struct sw842_param *p, u8 n) |
Dan Streetman | 2da572c | 2015-05-07 13:49:14 -0400 | [diff] [blame] | 218 | { |
| 219 | switch (n) { |
| 220 | case 2: |
| 221 | return __do_index(p, 2, I2_BITS, I2_FIFO_SIZE); |
| 222 | case 4: |
| 223 | return __do_index(p, 4, I4_BITS, I4_FIFO_SIZE); |
| 224 | case 8: |
| 225 | return __do_index(p, 8, I8_BITS, I8_FIFO_SIZE); |
| 226 | default: |
| 227 | return -EINVAL; |
| 228 | } |
| 229 | } |
| 230 | |
Dan Streetman | f7ead7b | 2015-05-11 07:22:35 -0400 | [diff] [blame] | 231 | static int do_op(struct sw842_param *p, u8 o) |
Dan Streetman | 2da572c | 2015-05-07 13:49:14 -0400 | [diff] [blame] | 232 | { |
| 233 | int i, ret = 0; |
| 234 | |
| 235 | if (o >= OPS_MAX) |
| 236 | return -EINVAL; |
| 237 | |
| 238 | for (i = 0; i < 4; i++) { |
| 239 | u8 op = decomp_ops[o][i]; |
| 240 | |
| 241 | pr_debug("op is %x\n", op); |
| 242 | |
| 243 | switch (op & OP_ACTION) { |
| 244 | case OP_ACTION_DATA: |
| 245 | ret = do_data(p, op & OP_AMOUNT); |
| 246 | break; |
| 247 | case OP_ACTION_INDEX: |
| 248 | ret = do_index(p, op & OP_AMOUNT); |
| 249 | break; |
| 250 | case OP_ACTION_NOOP: |
| 251 | break; |
| 252 | default: |
| 253 | pr_err("Interal error, invalid op %x\n", op); |
| 254 | return -EINVAL; |
| 255 | } |
| 256 | |
| 257 | if (ret) |
| 258 | return ret; |
| 259 | } |
| 260 | |
| 261 | if (sw842_template_counts) |
| 262 | atomic_inc(&template_count[o]); |
| 263 | |
| 264 | return 0; |
| 265 | } |
| 266 | |
| 267 | /** |
| 268 | * sw842_decompress |
| 269 | * |
| 270 | * Decompress the 842-compressed buffer of length @ilen at @in |
| 271 | * to the output buffer @out, using no more than @olen bytes. |
| 272 | * |
| 273 | * The compressed buffer must be only a single 842-compressed buffer, |
| 274 | * with the standard format described in the comments in 842.h |
| 275 | * Processing will stop when the 842 "END" template is detected, |
| 276 | * not the end of the buffer. |
| 277 | * |
| 278 | * Returns: 0 on success, error on failure. The @olen parameter |
| 279 | * will contain the number of output bytes written on success, or |
| 280 | * 0 on error. |
| 281 | */ |
| 282 | int sw842_decompress(const u8 *in, unsigned int ilen, |
| 283 | u8 *out, unsigned int *olen) |
| 284 | { |
| 285 | struct sw842_param p; |
| 286 | int ret; |
| 287 | u64 op, rep, tmp, bytes, total; |
| 288 | |
| 289 | p.in = (u8 *)in; |
| 290 | p.bit = 0; |
| 291 | p.ilen = ilen; |
| 292 | p.out = out; |
| 293 | p.ostart = out; |
| 294 | p.olen = *olen; |
| 295 | |
| 296 | total = p.olen; |
| 297 | |
| 298 | *olen = 0; |
| 299 | |
| 300 | do { |
| 301 | ret = next_bits(&p, &op, OP_BITS); |
| 302 | if (ret) |
| 303 | return ret; |
| 304 | |
| 305 | pr_debug("template is %lx\n", (unsigned long)op); |
| 306 | |
| 307 | switch (op) { |
| 308 | case OP_REPEAT: |
| 309 | ret = next_bits(&p, &rep, REPEAT_BITS); |
| 310 | if (ret) |
| 311 | return ret; |
| 312 | |
| 313 | if (p.out == out) /* no previous bytes */ |
| 314 | return -EINVAL; |
| 315 | |
| 316 | /* copy rep + 1 */ |
| 317 | rep++; |
| 318 | |
| 319 | if (rep * 8 > p.olen) |
| 320 | return -ENOSPC; |
| 321 | |
| 322 | while (rep-- > 0) { |
| 323 | memcpy(p.out, p.out - 8, 8); |
| 324 | p.out += 8; |
| 325 | p.olen -= 8; |
| 326 | } |
| 327 | |
| 328 | if (sw842_template_counts) |
| 329 | atomic_inc(&template_repeat_count); |
| 330 | |
| 331 | break; |
| 332 | case OP_ZEROS: |
| 333 | if (8 > p.olen) |
| 334 | return -ENOSPC; |
| 335 | |
| 336 | memset(p.out, 0, 8); |
| 337 | p.out += 8; |
| 338 | p.olen -= 8; |
| 339 | |
| 340 | if (sw842_template_counts) |
| 341 | atomic_inc(&template_zeros_count); |
| 342 | |
| 343 | break; |
| 344 | case OP_SHORT_DATA: |
| 345 | ret = next_bits(&p, &bytes, SHORT_DATA_BITS); |
| 346 | if (ret) |
| 347 | return ret; |
| 348 | |
| 349 | if (!bytes || bytes > SHORT_DATA_BITS_MAX) |
| 350 | return -EINVAL; |
| 351 | |
| 352 | while (bytes-- > 0) { |
| 353 | ret = next_bits(&p, &tmp, 8); |
| 354 | if (ret) |
| 355 | return ret; |
| 356 | *p.out = (u8)tmp; |
| 357 | p.out++; |
| 358 | p.olen--; |
| 359 | } |
| 360 | |
| 361 | if (sw842_template_counts) |
| 362 | atomic_inc(&template_short_data_count); |
| 363 | |
| 364 | break; |
| 365 | case OP_END: |
| 366 | if (sw842_template_counts) |
| 367 | atomic_inc(&template_end_count); |
| 368 | |
| 369 | break; |
| 370 | default: /* use template */ |
| 371 | ret = do_op(&p, op); |
| 372 | if (ret) |
| 373 | return ret; |
| 374 | break; |
| 375 | } |
| 376 | } while (op != OP_END); |
| 377 | |
| 378 | if (unlikely((total - p.olen) > UINT_MAX)) |
| 379 | return -ENOSPC; |
| 380 | |
| 381 | *olen = total - p.olen; |
| 382 | |
| 383 | return 0; |
| 384 | } |
| 385 | EXPORT_SYMBOL_GPL(sw842_decompress); |
| 386 | |
| 387 | static int __init sw842_init(void) |
| 388 | { |
| 389 | if (sw842_template_counts) |
| 390 | sw842_debugfs_create(); |
| 391 | |
| 392 | return 0; |
| 393 | } |
| 394 | module_init(sw842_init); |
| 395 | |
| 396 | static void __exit sw842_exit(void) |
| 397 | { |
| 398 | if (sw842_template_counts) |
| 399 | sw842_debugfs_remove(); |
| 400 | } |
| 401 | module_exit(sw842_exit); |
| 402 | |
| 403 | MODULE_LICENSE("GPL"); |
| 404 | MODULE_DESCRIPTION("Software 842 Decompressor"); |
| 405 | MODULE_AUTHOR("Dan Streetman <ddstreet@ieee.org>"); |