kernel/bpf/core.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * Linux Socket Filter - Kernel level socket filtering
   4  *
   5  * Based on the design of the Berkeley Packet Filter. The new
   6  * internal format has been designed by PLUMgrid:
   7  *
   8  *      Copyright (c) 2011 - 2014 PLUMgrid, http://plumgrid.com
   9  *
  10  * Authors:
  11  *
  12  *      Jay Schulist <jschlst@samba.org>
  13  *      Alexei Starovoitov <ast@plumgrid.com>
  14  *      Daniel Borkmann <dborkman@redhat.com>
  15  *
  16  * Andi Kleen - Fix a few bad bugs and races.
  17  * Kris Katterjohn - Added many additional checks in bpf_check_classic()
  18  */
  19
  20 #include <uapi/linux/btf.h>
  21 #include <linux/filter.h>
  22 #include <linux/skbuff.h>
  23 #include <linux/vmalloc.h>
  24 #include <linux/random.h>
  25 #include <linux/moduleloader.h>
  26 #include <linux/bpf.h>
  27 #include <linux/btf.h>
  28 #include <linux/objtool.h>
  29 #include <linux/rbtree_latch.h>
  30 #include <linux/kallsyms.h>
  31 #include <linux/rcupdate.h>
  32 #include <linux/perf_event.h>
  33 #include <linux/extable.h>
  34 #include <linux/log2.h>
  35 #include <linux/bpf_verifier.h>
  36 #include <linux/nodemask.h>
  37 #include <linux/nospec.h>
  38 #include <linux/bpf_mem_alloc.h>
  39 #include <linux/memcontrol.h>
  40
  41 #include <asm/barrier.h>
  42 #include <asm/unaligned.h>
  43
  44 /* Registers */
  45 #define BPF_R0  regs[BPF_REG_0]
  46 #define BPF_R1  regs[BPF_REG_1]
  47 #define BPF_R2  regs[BPF_REG_2]
  48 #define BPF_R3  regs[BPF_REG_3]
  49 #define BPF_R4  regs[BPF_REG_4]
  50 #define BPF_R5  regs[BPF_REG_5]
  51 #define BPF_R6  regs[BPF_REG_6]
  52 #define BPF_R7  regs[BPF_REG_7]
  53 #define BPF_R8  regs[BPF_REG_8]
  54 #define BPF_R9  regs[BPF_REG_9]
  55 #define BPF_R10 regs[BPF_REG_10]
  56
  57 /* Named registers */
  58 #define DST     regs[insn->dst_reg]
  59 #define SRC     regs[insn->src_reg]
  60 #define FP      regs[BPF_REG_FP]
  61 #define AX      regs[BPF_REG_AX]
  62 #define ARG1    regs[BPF_REG_ARG1]
  63 #define CTX     regs[BPF_REG_CTX]
  64 #define OFF     insn->off
  65 #define IMM     insn->imm
  66
  67 struct bpf_mem_alloc bpf_global_ma;
  68 bool bpf_global_ma_set;
  69
  70 /* No hurry in this branch
  71  *
  72  * Exported for the bpf jit load helper.
  73  */
  74 void *bpf_internal_load_pointer_neg_helper(const struct sk_buff *skb, int k, unsigned int size)
  75 {
  76         u8 *ptr = NULL;
  77
  78         if (k >= SKF_NET_OFF) {
  79                 ptr = skb_network_header(skb) + k - SKF_NET_OFF;
  80         } else if (k >= SKF_LL_OFF) {
  81                 if (unlikely(!skb_mac_header_was_set(skb)))
  82                         return NULL;
  83                 ptr = skb_mac_header(skb) + k - SKF_LL_OFF;
  84         }
  85         if (ptr >= skb->head && ptr + size <= skb_tail_pointer(skb))
  86                 return ptr;
  87
  88         return NULL;
  89 }
  90
  91 struct bpf_prog *bpf_prog_alloc_no_stats(unsigned int size, gfp_t gfp_extra_flags)
  92 {
  93         gfp_t gfp_flags = bpf_memcg_flags(GFP_KERNEL | __GFP_ZERO | gfp_extra_flags);
  94         struct bpf_prog_aux *aux;
  95         struct bpf_prog *fp;
  96
  97         size = round_up(size, PAGE_SIZE);
  98         fp = __vmalloc(size, gfp_flags);
  99         if (fp == NULL)
 100                 return NULL;
 101
 102         aux = kzalloc(sizeof(*aux), bpf_memcg_flags(GFP_KERNEL | gfp_extra_flags));
 103         if (aux == NULL) {
 104                 vfree(fp);
 105                 return NULL;
 106         }
 107         fp->active = alloc_percpu_gfp(int, bpf_memcg_flags(GFP_KERNEL | gfp_extra_flags));
 108         if (!fp->active) {
 109                 vfree(fp);
 110                 kfree(aux);
 111                 return NULL;
 112         }
 113
 114         fp->pages = size / PAGE_SIZE;
 115         fp->aux = aux;
 116         fp->aux->prog = fp;
 117         fp->jit_requested = ebpf_jit_enabled();
 118         fp->blinding_requested = bpf_jit_blinding_enabled(fp);
 119 #ifdef CONFIG_CGROUP_BPF
 120         aux->cgroup_atype = CGROUP_BPF_ATTACH_TYPE_INVALID;
 121 #endif
 122
 123         INIT_LIST_HEAD_RCU(&fp->aux->ksym.lnode);
 124         mutex_init(&fp->aux->used_maps_mutex);
 125         mutex_init(&fp->aux->dst_mutex);
 126
 127         return fp;
 128 }
 129
 130 struct bpf_prog *bpf_prog_alloc(unsigned int size, gfp_t gfp_extra_flags)
 131 {
 132         gfp_t gfp_flags = bpf_memcg_flags(GFP_KERNEL | __GFP_ZERO | gfp_extra_flags);
 133         struct bpf_prog *prog;
 134         int cpu;
 135
 136         prog = bpf_prog_alloc_no_stats(size, gfp_extra_flags);
 137         if (!prog)
 138                 return NULL;
 139
 140         prog->stats = alloc_percpu_gfp(struct bpf_prog_stats, gfp_flags);
 141         if (!prog->stats) {
 142                 free_percpu(prog->active);
 143                 kfree(prog->aux);
 144                 vfree(prog);
 145                 return NULL;
 146         }
 147
 148         for_each_possible_cpu(cpu) {
 149                 struct bpf_prog_stats *pstats;
 150
 151                 pstats = per_cpu_ptr(prog->stats, cpu);
 152                 u64_stats_init(&pstats->syncp);
 153         }
 154         return prog;
 155 }
 156 EXPORT_SYMBOL_GPL(bpf_prog_alloc);
 157
 158 int bpf_prog_alloc_jited_linfo(struct bpf_prog *prog)
 159 {
 160         if (!prog->aux->nr_linfo || !prog->jit_requested)
 161                 return 0;
 162
 163         prog->aux->jited_linfo = kvcalloc(prog->aux->nr_linfo,
 164                                           sizeof(*prog->aux->jited_linfo),
 165                                           bpf_memcg_flags(GFP_KERNEL | __GFP_NOWARN));
 166         if (!prog->aux->jited_linfo)
 167                 return -ENOMEM;
 168
 169         return 0;
 170 }
 171
 172 void bpf_prog_jit_attempt_done(struct bpf_prog *prog)
 173 {
 174         if (prog->aux->jited_linfo &&
 175             (!prog->jited || !prog->aux->jited_linfo[0])) {
 176                 kvfree(prog->aux->jited_linfo);
 177                 prog->aux->jited_linfo = NULL;
 178         }
 179
 180         kfree(prog->aux->kfunc_tab);
 181         prog->aux->kfunc_tab = NULL;
 182 }
 183
 184 /* The jit engine is responsible to provide an array
 185  * for insn_off to the jited_off mapping (insn_to_jit_off).
 186  *
 187  * The idx to this array is the insn_off.  Hence, the insn_off
 188  * here is relative to the prog itself instead of the main prog.
 189  * This array has one entry for each xlated bpf insn.
 190  *
 191  * jited_off is the byte off to the end of the jited insn.
 192  *
 193  * Hence, with
 194  * insn_start:
 195  *      The first bpf insn off of the prog.  The insn off
 196  *      here is relative to the main prog.
 197  *      e.g. if prog is a subprog, insn_start > 0
 198  * linfo_idx:
 199  *      The prog's idx to prog->aux->linfo and jited_linfo
 200  *
 201  * jited_linfo[linfo_idx] = prog->bpf_func
 202  *
 203  * For i > linfo_idx,
 204  *
 205  * jited_linfo[i] = prog->bpf_func +
 206  *      insn_to_jit_off[linfo[i].insn_off - insn_start - 1]
 207  */
 208 void bpf_prog_fill_jited_linfo(struct bpf_prog *prog,
 209                                const u32 *insn_to_jit_off)
 210 {
 211         u32 linfo_idx, insn_start, insn_end, nr_linfo, i;
 212         const struct bpf_line_info *linfo;
 213         void **jited_linfo;
 214
 215         if (!prog->aux->jited_linfo)
 216                 /* Userspace did not provide linfo */
 217                 return;
 218
 219         linfo_idx = prog->aux->linfo_idx;
 220         linfo = &prog->aux->linfo[linfo_idx];
 221         insn_start = linfo[0].insn_off;
 222         insn_end = insn_start + prog->len;
 223
 224         jited_linfo = &prog->aux->jited_linfo[linfo_idx];
 225         jited_linfo[0] = prog->bpf_func;
 226
 227         nr_linfo = prog->aux->nr_linfo - linfo_idx;
 228
 229         for (i = 1; i < nr_linfo && linfo[i].insn_off < insn_end; i++)
 230                 /* The verifier ensures that linfo[i].insn_off is
 231                  * strictly increasing
 232                  */
 233                 jited_linfo[i] = prog->bpf_func +
 234                         insn_to_jit_off[linfo[i].insn_off - insn_start - 1];
 235 }
 236
 237 struct bpf_prog *bpf_prog_realloc(struct bpf_prog *fp_old, unsigned int size,
 238                                   gfp_t gfp_extra_flags)
 239 {
 240         gfp_t gfp_flags = bpf_memcg_flags(GFP_KERNEL | __GFP_ZERO | gfp_extra_flags);
 241         struct bpf_prog *fp;
 242         u32 pages;
 243
 244         size = round_up(size, PAGE_SIZE);
 245         pages = size / PAGE_SIZE;
 246         if (pages <= fp_old->pages)
 247                 return fp_old;
 248
 249         fp = __vmalloc(size, gfp_flags);
 250         if (fp) {
 251                 memcpy(fp, fp_old, fp_old->pages * PAGE_SIZE);
 252                 fp->pages = pages;
 253                 fp->aux->prog = fp;
 254
 255                 /* We keep fp->aux from fp_old around in the new
 256                  * reallocated structure.
 257                  */
 258                 fp_old->aux = NULL;
 259                 fp_old->stats = NULL;
 260                 fp_old->active = NULL;
 261                 __bpf_prog_free(fp_old);
 262         }
 263
 264         return fp;
 265 }
 266
 267 void __bpf_prog_free(struct bpf_prog *fp)
 268 {
 269         if (fp->aux) {
 270                 mutex_destroy(&fp->aux->used_maps_mutex);
 271                 mutex_destroy(&fp->aux->dst_mutex);
 272                 kfree(fp->aux->poke_tab);
 273                 kfree(fp->aux);
 274         }
 275         free_percpu(fp->stats);
 276         free_percpu(fp->active);
 277         vfree(fp);
 278 }
 279
 280 int bpf_prog_calc_tag(struct bpf_prog *fp)
 281 {
 282         const u32 bits_offset = SHA1_BLOCK_SIZE - sizeof(__be64);
 283         u32 raw_size = bpf_prog_tag_scratch_size(fp);
 284         u32 digest[SHA1_DIGEST_WORDS];
 285         u32 ws[SHA1_WORKSPACE_WORDS];
 286         u32 i, bsize, psize, blocks;
 287         struct bpf_insn *dst;
 288         bool was_ld_map;
 289         u8 *raw, *todo;
 290         __be32 *result;
 291         __be64 *bits;
 292
 293         raw = vmalloc(raw_size);
 294         if (!raw)
 295                 return -ENOMEM;
 296
 297         sha1_init(digest);
 298         memset(ws, 0, sizeof(ws));
 299
 300         /* We need to take out the map fd for the digest calculation
 301          * since they are unstable from user space side.
 302          */
 303         dst = (void *)raw;
 304         for (i = 0, was_ld_map = false; i < fp->len; i++) {
 305                 dst[i] = fp->insnsi[i];
 306                 if (!was_ld_map &&
 307                     dst[i].code == (BPF_LD | BPF_IMM | BPF_DW) &&
 308                     (dst[i].src_reg == BPF_PSEUDO_MAP_FD ||
 309                      dst[i].src_reg == BPF_PSEUDO_MAP_VALUE)) {
 310                         was_ld_map = true;
 311                         dst[i].imm = 0;
 312                 } else if (was_ld_map &&
 313                            dst[i].code == 0 &&
 314                            dst[i].dst_reg == 0 &&
 315                            dst[i].src_reg == 0 &&
 316                            dst[i].off == 0) {
 317                         was_ld_map = false;
 318                         dst[i].imm = 0;
 319                 } else {
 320                         was_ld_map = false;
 321                 }
 322         }
 323
 324         psize = bpf_prog_insn_size(fp);
 325         memset(&raw[psize], 0, raw_size - psize);
 326         raw[psize++] = 0x80;
 327
 328         bsize  = round_up(psize, SHA1_BLOCK_SIZE);
 329         blocks = bsize / SHA1_BLOCK_SIZE;
 330         todo   = raw;
 331         if (bsize - psize >= sizeof(__be64)) {
 332                 bits = (__be64 *)(todo + bsize - sizeof(__be64));
 333         } else {
 334                 bits = (__be64 *)(todo + bsize + bits_offset);
 335                 blocks++;
 336         }
 337         *bits = cpu_to_be64((psize - 1) << 3);
 338
 339         while (blocks--) {
 340                 sha1_transform(digest, todo, ws);
 341                 todo += SHA1_BLOCK_SIZE;
 342         }
 343
 344         result = (__force __be32 *)digest;
 345         for (i = 0; i < SHA1_DIGEST_WORDS; i++)
 346                 result[i] = cpu_to_be32(digest[i]);
 347         memcpy(fp->tag, result, sizeof(fp->tag));
 348
 349         vfree(raw);
 350         return 0;
 351 }
 352
 353 static int bpf_adj_delta_to_imm(struct bpf_insn *insn, u32 pos, s32 end_old,
 354                                 s32 end_new, s32 curr, const bool probe_pass)
 355 {
 356         const s64 imm_min = S32_MIN, imm_max = S32_MAX;
 357         s32 delta = end_new - end_old;
 358         s64 imm = insn->imm;
 359
 360         if (curr < pos && curr + imm + 1 >= end_old)
 361                 imm += delta;
 362         else if (curr >= end_new && curr + imm + 1 < end_new)
 363                 imm -= delta;
 364         if (imm < imm_min || imm > imm_max)
 365                 return -ERANGE;
 366         if (!probe_pass)
 367                 insn->imm = imm;
 368         return 0;
 369 }
 370
 371 static int bpf_adj_delta_to_off(struct bpf_insn *insn, u32 pos, s32 end_old,
 372                                 s32 end_new, s32 curr, const bool probe_pass)
 373 {
 374         const s32 off_min = S16_MIN, off_max = S16_MAX;
 375         s32 delta = end_new - end_old;
 376         s32 off;
 377
 378         if (insn->code == (BPF_JMP32 | BPF_JA))
 379                 off = insn->imm;
 380         else
 381                 off = insn->off;
 382
 383         if (curr < pos && curr + off + 1 >= end_old)
 384                 off += delta;
 385         else if (curr >= end_new && curr + off + 1 < end_new)
 386                 off -= delta;
 387         if (off < off_min || off > off_max)
 388                 return -ERANGE;
 389         if (!probe_pass) {
 390                 if (insn->code == (BPF_JMP32 | BPF_JA))
 391                         insn->imm = off;
 392                 else
 393                         insn->off = off;
 394         }
 395         return 0;
 396 }
 397
 398 static int bpf_adj_branches(struct bpf_prog *prog, u32 pos, s32 end_old,
 399                             s32 end_new, const bool probe_pass)
 400 {
 401         u32 i, insn_cnt = prog->len + (probe_pass ? end_new - end_old : 0);
 402         struct bpf_insn *insn = prog->insnsi;
 403         int ret = 0;
 404
 405         for (i = 0; i < insn_cnt; i++, insn++) {
 406                 u8 code;
 407
 408                 /* In the probing pass we still operate on the original,
 409                  * unpatched image in order to check overflows before we
 410                  * do any other adjustments. Therefore skip the patchlet.
 411                  */
 412                 if (probe_pass && i == pos) {
 413                         i = end_new;
 414                         insn = prog->insnsi + end_old;
 415                 }
 416                 if (bpf_pseudo_func(insn)) {
 417                         ret = bpf_adj_delta_to_imm(insn, pos, end_old,
 418                                                    end_new, i, probe_pass);
 419                         if (ret)
 420                                 return ret;
 421                         continue;
 422                 }
 423                 code = insn->code;
 424                 if ((BPF_CLASS(code) != BPF_JMP &&
 425                      BPF_CLASS(code) != BPF_JMP32) ||
 426                     BPF_OP(code) == BPF_EXIT)
 427                         continue;
 428                 /* Adjust offset of jmps if we cross patch boundaries. */
 429                 if (BPF_OP(code) == BPF_CALL) {
 430                         if (insn->src_reg != BPF_PSEUDO_CALL)
 431                                 continue;
 432                         ret = bpf_adj_delta_to_imm(insn, pos, end_old,
 433                                                    end_new, i, probe_pass);
 434                 } else {
 435                         ret = bpf_adj_delta_to_off(insn, pos, end_old,
 436                                                    end_new, i, probe_pass);
 437                 }
 438                 if (ret)
 439                         break;
 440         }
 441
 442         return ret;
 443 }
 444
 445 static void bpf_adj_linfo(struct bpf_prog *prog, u32 off, u32 delta)
 446 {
 447         struct bpf_line_info *linfo;
 448         u32 i, nr_linfo;
 449
 450         nr_linfo = prog->aux->nr_linfo;
 451         if (!nr_linfo || !delta)
 452                 return;
 453
 454         linfo = prog->aux->linfo;
 455
 456         for (i = 0; i < nr_linfo; i++)
 457                 if (off < linfo[i].insn_off)
 458                         break;
 459
 460         /* Push all off < linfo[i].insn_off by delta */
 461         for (; i < nr_linfo; i++)
 462                 linfo[i].insn_off += delta;
 463 }
 464
 465 struct bpf_prog *bpf_patch_insn_single(struct bpf_prog *prog, u32 off,
 466                                        const struct bpf_insn *patch, u32 len)
 467 {
 468         u32 insn_adj_cnt, insn_rest, insn_delta = len - 1;
 469         const u32 cnt_max = S16_MAX;
 470         struct bpf_prog *prog_adj;
 471         int err;
 472
 473         /* Since our patchlet doesn't expand the image, we're done. */
 474         if (insn_delta == 0) {
 475                 memcpy(prog->insnsi + off, patch, sizeof(*patch));
 476                 return prog;
 477         }
 478
 479         insn_adj_cnt = prog->len + insn_delta;
 480
 481         /* Reject anything that would potentially let the insn->off
 482          * target overflow when we have excessive program expansions.
 483          * We need to probe here before we do any reallocation where
 484          * we afterwards may not fail anymore.
 485          */
 486         if (insn_adj_cnt > cnt_max &&
 487             (err = bpf_adj_branches(prog, off, off + 1, off + len, true)))
 488                 return ERR_PTR(err);
 489
 490         /* Several new instructions need to be inserted. Make room
 491          * for them. Likely, there's no need for a new allocation as
 492          * last page could have large enough tailroom.
 493          */
 494         prog_adj = bpf_prog_realloc(prog, bpf_prog_size(insn_adj_cnt),
 495                                     GFP_USER);
 496         if (!prog_adj)
 497                 return ERR_PTR(-ENOMEM);
 498
 499         prog_adj->len = insn_adj_cnt;
 500
 501         /* Patching happens in 3 steps:
 502          *
 503          * 1) Move over tail of insnsi from next instruction onwards,
 504          *    so we can patch the single target insn with one or more
 505          *    new ones (patching is always from 1 to n insns, n > 0).
 506          * 2) Inject new instructions at the target location.
 507          * 3) Adjust branch offsets if necessary.
 508          */
 509         insn_rest = insn_adj_cnt - off - len;
 510
 511         memmove(prog_adj->insnsi + off + len, prog_adj->insnsi + off + 1,
 512                 sizeof(*patch) * insn_rest);
 513         memcpy(prog_adj->insnsi + off, patch, sizeof(*patch) * len);
 514
 515         /* We are guaranteed to not fail at this point, otherwise
 516          * the ship has sailed to reverse to the original state. An
 517          * overflow cannot happen at this point.
 518          */
 519         BUG_ON(bpf_adj_branches(prog_adj, off, off + 1, off + len, false));
 520
 521         bpf_adj_linfo(prog_adj, off, insn_delta);
 522
 523         return prog_adj;
 524 }
 525
 526 int bpf_remove_insns(struct bpf_prog *prog, u32 off, u32 cnt)
 527 {
 528         /* Branch offsets can't overflow when program is shrinking, no need
 529          * to call bpf_adj_branches(..., true) here
 530          */
 531         memmove(prog->insnsi + off, prog->insnsi + off + cnt,
 532                 sizeof(struct bpf_insn) * (prog->len - off - cnt));
 533         prog->len -= cnt;
 534
 535         return WARN_ON_ONCE(bpf_adj_branches(prog, off, off + cnt, off, false));
 536 }
 537
 538 static void bpf_prog_kallsyms_del_subprogs(struct bpf_prog *fp)
 539 {
 540         int i;
 541
 542         for (i = 0; i < fp->aux->func_cnt; i++)
 543                 bpf_prog_kallsyms_del(fp->aux->func[i]);
 544 }
 545
 546 void bpf_prog_kallsyms_del_all(struct bpf_prog *fp)
 547 {
 548         bpf_prog_kallsyms_del_subprogs(fp);
 549         bpf_prog_kallsyms_del(fp);
 550 }
 551
 552 #ifdef CONFIG_BPF_JIT
 553 /* All BPF JIT sysctl knobs here. */
 554 int bpf_jit_enable   __read_mostly = IS_BUILTIN(CONFIG_BPF_JIT_DEFAULT_ON);
 555 int bpf_jit_kallsyms __read_mostly = IS_BUILTIN(CONFIG_BPF_JIT_DEFAULT_ON);
 556 int bpf_jit_harden   __read_mostly;
 557 long bpf_jit_limit   __read_mostly;
 558 long bpf_jit_limit_max __read_mostly;
 559
 560 static void
 561 bpf_prog_ksym_set_addr(struct bpf_prog *prog)
 562 {
 563         WARN_ON_ONCE(!bpf_prog_ebpf_jited(prog));
 564
 565         prog->aux->ksym.start = (unsigned long) prog->bpf_func;
 566         prog->aux->ksym.end   = prog->aux->ksym.start + prog->jited_len;
 567 }
 568
 569 static void
 570 bpf_prog_ksym_set_name(struct bpf_prog *prog)
 571 {
 572         char *sym = prog->aux->ksym.name;
 573         const char *end = sym + KSYM_NAME_LEN;
 574         const struct btf_type *type;
 575         const char *func_name;
 576
 577         BUILD_BUG_ON(sizeof("bpf_prog_") +
 578                      sizeof(prog->tag) * 2 +
 579                      /* name has been null terminated.
 580                       * We should need +1 for the '_' preceding
 581                       * the name.  However, the null character
 582                       * is double counted between the name and the
 583                       * sizeof("bpf_prog_") above, so we omit
 584                       * the +1 here.
 585                       */
 586                      sizeof(prog->aux->name) > KSYM_NAME_LEN);
 587
 588         sym += snprintf(sym, KSYM_NAME_LEN, "bpf_prog_");
 589         sym  = bin2hex(sym, prog->tag, sizeof(prog->tag));
 590
 591         /* prog->aux->name will be ignored if full btf name is available */
 592         if (prog->aux->func_info_cnt) {
 593                 type = btf_type_by_id(prog->aux->btf,
 594                                       prog->aux->func_info[prog->aux->func_idx].type_id);
 595                 func_name = btf_name_by_offset(prog->aux->btf, type->name_off);
 596                 snprintf(sym, (size_t)(end - sym), "_%s", func_name);
 597                 return;
 598         }
 599
 600         if (prog->aux->name[0])
 601                 snprintf(sym, (size_t)(end - sym), "_%s", prog->aux->name);
 602         else
 603                 *sym = 0;
 604 }
 605
 606 static unsigned long bpf_get_ksym_start(struct latch_tree_node *n)
 607 {
 608         return container_of(n, struct bpf_ksym, tnode)->start;
 609 }
 610
 611 static __always_inline bool bpf_tree_less(struct latch_tree_node *a,
 612                                           struct latch_tree_node *b)
 613 {
 614         return bpf_get_ksym_start(a) < bpf_get_ksym_start(b);
 615 }
 616
 617 static __always_inline int bpf_tree_comp(void *key, struct latch_tree_node *n)
 618 {
 619         unsigned long val = (unsigned long)key;
 620         const struct bpf_ksym *ksym;
 621
 622         ksym = container_of(n, struct bpf_ksym, tnode);
 623
 624         if (val < ksym->start)
 625                 return -1;
 626         /* Ensure that we detect return addresses as part of the program, when
 627          * the final instruction is a call for a program part of the stack
 628          * trace. Therefore, do val > ksym->end instead of val >= ksym->end.
 629          */
 630         if (val > ksym->end)
 631                 return  1;
 632
 633         return 0;
 634 }
 635
 636 static const struct latch_tree_ops bpf_tree_ops = {
 637         .less   = bpf_tree_less,
 638         .comp   = bpf_tree_comp,
 639 };
 640
 641 static DEFINE_SPINLOCK(bpf_lock);
 642 static LIST_HEAD(bpf_kallsyms);
 643 static struct latch_tree_root bpf_tree __cacheline_aligned;
 644
 645 void bpf_ksym_add(struct bpf_ksym *ksym)
 646 {
 647         spin_lock_bh(&bpf_lock);
 648         WARN_ON_ONCE(!list_empty(&ksym->lnode));
 649         list_add_tail_rcu(&ksym->lnode, &bpf_kallsyms);
 650         latch_tree_insert(&ksym->tnode, &bpf_tree, &bpf_tree_ops);
 651         spin_unlock_bh(&bpf_lock);
 652 }
 653
 654 static void __bpf_ksym_del(struct bpf_ksym *ksym)
 655 {
 656         if (list_empty(&ksym->lnode))
 657                 return;
 658
 659         latch_tree_erase(&ksym->tnode, &bpf_tree, &bpf_tree_ops);
 660         list_del_rcu(&ksym->lnode);
 661 }
 662
 663 void bpf_ksym_del(struct bpf_ksym *ksym)
 664 {
 665         spin_lock_bh(&bpf_lock);
 666         __bpf_ksym_del(ksym);
 667         spin_unlock_bh(&bpf_lock);
 668 }
 669
 670 static bool bpf_prog_kallsyms_candidate(const struct bpf_prog *fp)
 671 {
 672         return fp->jited && !bpf_prog_was_classic(fp);
 673 }
 674
 675 void bpf_prog_kallsyms_add(struct bpf_prog *fp)
 676 {
 677         if (!bpf_prog_kallsyms_candidate(fp) ||
 678             !bpf_capable())
 679                 return;
 680
 681         bpf_prog_ksym_set_addr(fp);
 682         bpf_prog_ksym_set_name(fp);
 683         fp->aux->ksym.prog = true;
 684
 685         bpf_ksym_add(&fp->aux->ksym);
 686 }
 687
 688 void bpf_prog_kallsyms_del(struct bpf_prog *fp)
 689 {
 690         if (!bpf_prog_kallsyms_candidate(fp))
 691                 return;
 692
 693         bpf_ksym_del(&fp->aux->ksym);
 694 }
 695
 696 static struct bpf_ksym *bpf_ksym_find(unsigned long addr)
 697 {
 698         struct latch_tree_node *n;
 699
 700         n = latch_tree_find((void *)addr, &bpf_tree, &bpf_tree_ops);
 701         return n ? container_of(n, struct bpf_ksym, tnode) : NULL;
 702 }
 703
 704 const char *__bpf_address_lookup(unsigned long addr, unsigned long *size,
 705                                  unsigned long *off, char *sym)
 706 {
 707         struct bpf_ksym *ksym;
 708         char *ret = NULL;
 709
 710         rcu_read_lock();
 711         ksym = bpf_ksym_find(addr);
 712         if (ksym) {
 713                 unsigned long symbol_start = ksym->start;
 714                 unsigned long symbol_end = ksym->end;
 715
 716                 strncpy(sym, ksym->name, KSYM_NAME_LEN);
 717
 718                 ret = sym;
 719                 if (size)
 720                         *size = symbol_end - symbol_start;
 721                 if (off)
 722                         *off  = addr - symbol_start;
 723         }
 724         rcu_read_unlock();
 725
 726         return ret;
 727 }
 728
 729 bool is_bpf_text_address(unsigned long addr)
 730 {
 731         bool ret;
 732
 733         rcu_read_lock();
 734         ret = bpf_ksym_find(addr) != NULL;
 735         rcu_read_unlock();
 736
 737         return ret;
 738 }
 739
 740 static struct bpf_prog *bpf_prog_ksym_find(unsigned long addr)
 741 {
 742         struct bpf_ksym *ksym = bpf_ksym_find(addr);
 743
 744         return ksym && ksym->prog ?
 745                container_of(ksym, struct bpf_prog_aux, ksym)->prog :
 746                NULL;
 747 }
 748
 749 const struct exception_table_entry *search_bpf_extables(unsigned long addr)
 750 {
 751         const struct exception_table_entry *e = NULL;
 752         struct bpf_prog *prog;
 753
 754         rcu_read_lock();
 755         prog = bpf_prog_ksym_find(addr);
 756         if (!prog)
 757                 goto out;
 758         if (!prog->aux->num_exentries)
 759                 goto out;
 760
 761         e = search_extable(prog->aux->extable, prog->aux->num_exentries, addr);
 762 out:
 763         rcu_read_unlock();
 764         return e;
 765 }
 766
 767 int bpf_get_kallsym(unsigned int symnum, unsigned long *value, char *type,
 768                     char *sym)
 769 {
 770         struct bpf_ksym *ksym;
 771         unsigned int it = 0;
 772         int ret = -ERANGE;
 773
 774         if (!bpf_jit_kallsyms_enabled())
 775                 return ret;
 776
 777         rcu_read_lock();
 778         list_for_each_entry_rcu(ksym, &bpf_kallsyms, lnode) {
 779                 if (it++ != symnum)
 780                         continue;
 781
 782                 strncpy(sym, ksym->name, KSYM_NAME_LEN);
 783
 784                 *value = ksym->start;
 785                 *type  = BPF_SYM_ELF_TYPE;
 786
 787                 ret = 0;
 788                 break;
 789         }
 790         rcu_read_unlock();
 791
 792         return ret;
 793 }
 794
 795 int bpf_jit_add_poke_descriptor(struct bpf_prog *prog,
 796                                 struct bpf_jit_poke_descriptor *poke)
 797 {
 798         struct bpf_jit_poke_descriptor *tab = prog->aux->poke_tab;
 799         static const u32 poke_tab_max = 1024;
 800         u32 slot = prog->aux->size_poke_tab;
 801         u32 size = slot + 1;
 802
 803         if (size > poke_tab_max)
 804                 return -ENOSPC;
 805         if (poke->tailcall_target || poke->tailcall_target_stable ||
 806             poke->tailcall_bypass || poke->adj_off || poke->bypass_addr)
 807                 return -EINVAL;
 808
 809         switch (poke->reason) {
 810         case BPF_POKE_REASON_TAIL_CALL:
 811                 if (!poke->tail_call.map)
 812                         return -EINVAL;
 813                 break;
 814         default:
 815                 return -EINVAL;
 816         }
 817
 818         tab = krealloc(tab, size * sizeof(*poke), GFP_KERNEL);
 819         if (!tab)
 820                 return -ENOMEM;
 821
 822         memcpy(&tab[slot], poke, sizeof(*poke));
 823         prog->aux->size_poke_tab = size;
 824         prog->aux->poke_tab = tab;
 825
 826         return slot;
 827 }
 828
 829 /*
 830  * BPF program pack allocator.
 831  *
 832  * Most BPF programs are pretty small. Allocating a hole page for each
 833  * program is sometime a waste. Many small bpf program also adds pressure
 834  * to instruction TLB. To solve this issue, we introduce a BPF program pack
 835  * allocator. The prog_pack allocator uses HPAGE_PMD_SIZE page (2MB on x86)
 836  * to host BPF programs.
 837  */
 838 #define BPF_PROG_CHUNK_SHIFT    6
 839 #define BPF_PROG_CHUNK_SIZE     (1 << BPF_PROG_CHUNK_SHIFT)
 840 #define BPF_PROG_CHUNK_MASK     (~(BPF_PROG_CHUNK_SIZE - 1))
 841
 842 struct bpf_prog_pack {
 843         struct list_head list;
 844         void *ptr;
 845         unsigned long bitmap[];
 846 };
 847
 848 void bpf_jit_fill_hole_with_zero(void *area, unsigned int size)
 849 {
 850         memset(area, 0, size);
 851 }
 852
 853 #define BPF_PROG_SIZE_TO_NBITS(size)    (round_up(size, BPF_PROG_CHUNK_SIZE) / BPF_PROG_CHUNK_SIZE)
 854
 855 static DEFINE_MUTEX(pack_mutex);
 856 static LIST_HEAD(pack_list);
 857
 858 /* PMD_SIZE is not available in some special config, e.g. ARCH=arm with
 859  * CONFIG_MMU=n. Use PAGE_SIZE in these cases.
 860  */
 861 #ifdef PMD_SIZE
 862 #define BPF_PROG_PACK_SIZE (PMD_SIZE * num_possible_nodes())
 863 #else
 864 #define BPF_PROG_PACK_SIZE PAGE_SIZE
 865 #endif
 866
 867 #define BPF_PROG_CHUNK_COUNT (BPF_PROG_PACK_SIZE / BPF_PROG_CHUNK_SIZE)
 868
 869 static struct bpf_prog_pack *alloc_new_pack(bpf_jit_fill_hole_t bpf_fill_ill_insns)
 870 {
 871         struct bpf_prog_pack *pack;
 872
 873         pack = kzalloc(struct_size(pack, bitmap, BITS_TO_LONGS(BPF_PROG_CHUNK_COUNT)),
 874                        GFP_KERNEL);
 875         if (!pack)
 876                 return NULL;
 877         pack->ptr = bpf_jit_alloc_exec(BPF_PROG_PACK_SIZE);
 878         if (!pack->ptr) {
 879                 kfree(pack);
 880                 return NULL;
 881         }
 882         bpf_fill_ill_insns(pack->ptr, BPF_PROG_PACK_SIZE);
 883         bitmap_zero(pack->bitmap, BPF_PROG_PACK_SIZE / BPF_PROG_CHUNK_SIZE);
 884         list_add_tail(&pack->list, &pack_list);
 885
 886         set_vm_flush_reset_perms(pack->ptr);
 887         set_memory_rox((unsigned long)pack->ptr, BPF_PROG_PACK_SIZE / PAGE_SIZE);
 888         return pack;
 889 }
 890
 891 void *bpf_prog_pack_alloc(u32 size, bpf_jit_fill_hole_t bpf_fill_ill_insns)
 892 {
 893         unsigned int nbits = BPF_PROG_SIZE_TO_NBITS(size);
 894         struct bpf_prog_pack *pack;
 895         unsigned long pos;
 896         void *ptr = NULL;
 897
 898         mutex_lock(&pack_mutex);
 899         if (size > BPF_PROG_PACK_SIZE) {
 900                 size = round_up(size, PAGE_SIZE);
 901                 ptr = bpf_jit_alloc_exec(size);
 902                 if (ptr) {
 903                         bpf_fill_ill_insns(ptr, size);
 904                         set_vm_flush_reset_perms(ptr);
 905                         set_memory_rox((unsigned long)ptr, size / PAGE_SIZE);
 906                 }
 907                 goto out;
 908         }
 909         list_for_each_entry(pack, &pack_list, list) {
 910                 pos = bitmap_find_next_zero_area(pack->bitmap, BPF_PROG_CHUNK_COUNT, 0,
 911                                                  nbits, 0);
 912                 if (pos < BPF_PROG_CHUNK_COUNT)
 913                         goto found_free_area;
 914         }
 915
 916         pack = alloc_new_pack(bpf_fill_ill_insns);
 917         if (!pack)
 918                 goto out;
 919
 920         pos = 0;
 921
 922 found_free_area:
 923         bitmap_set(pack->bitmap, pos, nbits);
 924         ptr = (void *)(pack->ptr) + (pos << BPF_PROG_CHUNK_SHIFT);
 925
 926 out:
 927         mutex_unlock(&pack_mutex);
 928         return ptr;
 929 }
 930
 931 void bpf_prog_pack_free(struct bpf_binary_header *hdr)
 932 {
 933         struct bpf_prog_pack *pack = NULL, *tmp;
 934         unsigned int nbits;
 935         unsigned long pos;
 936
 937         mutex_lock(&pack_mutex);
 938         if (hdr->size > BPF_PROG_PACK_SIZE) {
 939                 bpf_jit_free_exec(hdr);
 940                 goto out;
 941         }
 942
 943         list_for_each_entry(tmp, &pack_list, list) {
 944                 if ((void *)hdr >= tmp->ptr && (tmp->ptr + BPF_PROG_PACK_SIZE) > (void *)hdr) {
 945                         pack = tmp;
 946                         break;
 947                 }
 948         }
 949
 950         if (WARN_ONCE(!pack, "bpf_prog_pack bug\n"))
 951                 goto out;
 952
 953         nbits = BPF_PROG_SIZE_TO_NBITS(hdr->size);
 954         pos = ((unsigned long)hdr - (unsigned long)pack->ptr) >> BPF_PROG_CHUNK_SHIFT;
 955
 956         WARN_ONCE(bpf_arch_text_invalidate(hdr, hdr->size),
 957                   "bpf_prog_pack bug: missing bpf_arch_text_invalidate?\n");
 958
 959         bitmap_clear(pack->bitmap, pos, nbits);
 960         if (bitmap_find_next_zero_area(pack->bitmap, BPF_PROG_CHUNK_COUNT, 0,
 961                                        BPF_PROG_CHUNK_COUNT, 0) == 0) {
 962                 list_del(&pack->list);
 963                 bpf_jit_free_exec(pack->ptr);
 964                 kfree(pack);
 965         }
 966 out:
 967         mutex_unlock(&pack_mutex);
 968 }
 969
 970 static atomic_long_t bpf_jit_current;
 971
 972 /* Can be overridden by an arch's JIT compiler if it has a custom,
 973  * dedicated BPF backend memory area, or if neither of the two
 974  * below apply.
 975  */
 976 u64 __weak bpf_jit_alloc_exec_limit(void)
 977 {
 978 #if defined(MODULES_VADDR)
 979         return MODULES_END - MODULES_VADDR;
 980 #else
 981         return VMALLOC_END - VMALLOC_START;
 982 #endif
 983 }
 984
 985 static int __init bpf_jit_charge_init(void)
 986 {
 987         /* Only used as heuristic here to derive limit. */
 988         bpf_jit_limit_max = bpf_jit_alloc_exec_limit();
 989         bpf_jit_limit = min_t(u64, round_up(bpf_jit_limit_max >> 1,
 990                                             PAGE_SIZE), LONG_MAX);
 991         return 0;
 992 }
 993 pure_initcall(bpf_jit_charge_init);
 994
 995 int bpf_jit_charge_modmem(u32 size)
 996 {
 997         if (atomic_long_add_return(size, &bpf_jit_current) > READ_ONCE(bpf_jit_limit)) {
 998                 if (!bpf_capable()) {
 999                         atomic_long_sub(size, &bpf_jit_current);
1000                         return -EPERM;
1001                 }
1002         }
1003
1004         return 0;
1005 }
1006
1007 void bpf_jit_uncharge_modmem(u32 size)
1008 {
1009         atomic_long_sub(size, &bpf_jit_current);
1010 }
1011
1012 void *__weak bpf_jit_alloc_exec(unsigned long size)
1013 {
1014         return module_alloc(size);
1015 }
1016
1017 void __weak bpf_jit_free_exec(void *addr)
1018 {
1019         module_memfree(addr);
1020 }
1021
1022 struct bpf_binary_header *
1023 bpf_jit_binary_alloc(unsigned int proglen, u8 **image_ptr,
1024                      unsigned int alignment,
1025                      bpf_jit_fill_hole_t bpf_fill_ill_insns)
1026 {
1027         struct bpf_binary_header *hdr;
1028         u32 size, hole, start;
1029
1030         WARN_ON_ONCE(!is_power_of_2(alignment) ||
1031                      alignment > BPF_IMAGE_ALIGNMENT);
1032
1033         /* Most of BPF filters are really small, but if some of them
1034          * fill a page, allow at least 128 extra bytes to insert a
1035          * random section of illegal instructions.
1036          */
1037         size = round_up(proglen + sizeof(*hdr) + 128, PAGE_SIZE);
1038
1039         if (bpf_jit_charge_modmem(size))
1040                 return NULL;
1041         hdr = bpf_jit_alloc_exec(size);
1042         if (!hdr) {
1043                 bpf_jit_uncharge_modmem(size);
1044                 return NULL;
1045         }
1046
1047         /* Fill space with illegal/arch-dep instructions. */
1048         bpf_fill_ill_insns(hdr, size);
1049
1050         hdr->size = size;
1051         hole = min_t(unsigned int, size - (proglen + sizeof(*hdr)),
1052                      PAGE_SIZE - sizeof(*hdr));
1053         start = get_random_u32_below(hole) & ~(alignment - 1);
1054
1055         /* Leave a random number of instructions before BPF code. */
1056         *image_ptr = &hdr->image[start];
1057
1058         return hdr;
1059 }
1060
1061 void bpf_jit_binary_free(struct bpf_binary_header *hdr)
1062 {
1063         u32 size = hdr->size;
1064
1065         bpf_jit_free_exec(hdr);
1066         bpf_jit_uncharge_modmem(size);
1067 }
1068
1069 /* Allocate jit binary from bpf_prog_pack allocator.
1070  * Since the allocated memory is RO+X, the JIT engine cannot write directly
1071  * to the memory. To solve this problem, a RW buffer is also allocated at
1072  * as the same time. The JIT engine should calculate offsets based on the
1073  * RO memory address, but write JITed program to the RW buffer. Once the
1074  * JIT engine finishes, it calls bpf_jit_binary_pack_finalize, which copies
1075  * the JITed program to the RO memory.
1076  */
1077 struct bpf_binary_header *
1078 bpf_jit_binary_pack_alloc(unsigned int proglen, u8 **image_ptr,
1079                           unsigned int alignment,
1080                           struct bpf_binary_header **rw_header,
1081                           u8 **rw_image,
1082                           bpf_jit_fill_hole_t bpf_fill_ill_insns)
1083 {
1084         struct bpf_binary_header *ro_header;
1085         u32 size, hole, start;
1086
1087         WARN_ON_ONCE(!is_power_of_2(alignment) ||
1088                      alignment > BPF_IMAGE_ALIGNMENT);
1089
1090         /* add 16 bytes for a random section of illegal instructions */
1091         size = round_up(proglen + sizeof(*ro_header) + 16, BPF_PROG_CHUNK_SIZE);
1092
1093         if (bpf_jit_charge_modmem(size))
1094                 return NULL;
1095         ro_header = bpf_prog_pack_alloc(size, bpf_fill_ill_insns);
1096         if (!ro_header) {
1097                 bpf_jit_uncharge_modmem(size);
1098                 return NULL;
1099         }
1100
1101         *rw_header = kvmalloc(size, GFP_KERNEL);
1102         if (!*rw_header) {
1103                 bpf_arch_text_copy(&ro_header->size, &size, sizeof(size));
1104                 bpf_prog_pack_free(ro_header);
1105                 bpf_jit_uncharge_modmem(size);
1106                 return NULL;
1107         }
1108
1109         /* Fill space with illegal/arch-dep instructions. */
1110         bpf_fill_ill_insns(*rw_header, size);
1111         (*rw_header)->size = size;
1112
1113         hole = min_t(unsigned int, size - (proglen + sizeof(*ro_header)),
1114                      BPF_PROG_CHUNK_SIZE - sizeof(*ro_header));
1115         start = get_random_u32_below(hole) & ~(alignment - 1);
1116
1117         *image_ptr = &ro_header->image[start];
1118         *rw_image = &(*rw_header)->image[start];
1119
1120         return ro_header;
1121 }
1122
1123 /* Copy JITed text from rw_header to its final location, the ro_header. */
1124 int bpf_jit_binary_pack_finalize(struct bpf_prog *prog,
1125                                  struct bpf_binary_header *ro_header,
1126                                  struct bpf_binary_header *rw_header)
1127 {
1128         void *ptr;
1129
1130         ptr = bpf_arch_text_copy(ro_header, rw_header, rw_header->size);
1131
1132         kvfree(rw_header);
1133
1134         if (IS_ERR(ptr)) {
1135                 bpf_prog_pack_free(ro_header);
1136                 return PTR_ERR(ptr);
1137         }
1138         return 0;
1139 }
1140
1141 /* bpf_jit_binary_pack_free is called in two different scenarios:
1142  *   1) when the program is freed after;
1143  *   2) when the JIT engine fails (before bpf_jit_binary_pack_finalize).
1144  * For case 2), we need to free both the RO memory and the RW buffer.
1145  *
1146  * bpf_jit_binary_pack_free requires proper ro_header->size. However,
1147  * bpf_jit_binary_pack_alloc does not set it. Therefore, ro_header->size
1148  * must be set with either bpf_jit_binary_pack_finalize (normal path) or
1149  * bpf_arch_text_copy (when jit fails).
1150  */
1151 void bpf_jit_binary_pack_free(struct bpf_binary_header *ro_header,
1152                               struct bpf_binary_header *rw_header)
1153 {
1154         u32 size = ro_header->size;
1155
1156         bpf_prog_pack_free(ro_header);
1157         kvfree(rw_header);
1158         bpf_jit_uncharge_modmem(size);
1159 }
1160
1161 struct bpf_binary_header *
1162 bpf_jit_binary_pack_hdr(const struct bpf_prog *fp)
1163 {
1164         unsigned long real_start = (unsigned long)fp->bpf_func;
1165         unsigned long addr;
1166
1167         addr = real_start & BPF_PROG_CHUNK_MASK;
1168         return (void *)addr;
1169 }
1170
1171 static inline struct bpf_binary_header *
1172 bpf_jit_binary_hdr(const struct bpf_prog *fp)
1173 {
1174         unsigned long real_start = (unsigned long)fp->bpf_func;
1175         unsigned long addr;
1176
1177         addr = real_start & PAGE_MASK;
1178         return (void *)addr;
1179 }
1180
1181 /* This symbol is only overridden by archs that have different
1182  * requirements than the usual eBPF JITs, f.e. when they only
1183  * implement cBPF JIT, do not set images read-only, etc.
1184  */
1185 void __weak bpf_jit_free(struct bpf_prog *fp)
1186 {
1187         if (fp->jited) {
1188                 struct bpf_binary_header *hdr = bpf_jit_binary_hdr(fp);
1189
1190                 bpf_jit_binary_free(hdr);
1191                 WARN_ON_ONCE(!bpf_prog_kallsyms_verify_off(fp));
1192         }
1193
1194         bpf_prog_unlock_free(fp);
1195 }
1196
1197 int bpf_jit_get_func_addr(const struct bpf_prog *prog,
1198                           const struct bpf_insn *insn, bool extra_pass,
1199                           u64 *func_addr, bool *func_addr_fixed)
1200 {
1201         s16 off = insn->off;
1202         s32 imm = insn->imm;
1203         u8 *addr;
1204         int err;
1205
1206         *func_addr_fixed = insn->src_reg != BPF_PSEUDO_CALL;
1207         if (!*func_addr_fixed) {
1208                 /* Place-holder address till the last pass has collected
1209                  * all addresses for JITed subprograms in which case we
1210                  * can pick them up from prog->aux.
1211                  */
1212                 if (!extra_pass)
1213                         addr = NULL;
1214                 else if (prog->aux->func &&
1215                          off >= 0 && off < prog->aux->func_cnt)
1216                         addr = (u8 *)prog->aux->func[off]->bpf_func;
1217                 else
1218                         return -EINVAL;
1219         } else if (insn->src_reg == BPF_PSEUDO_KFUNC_CALL &&
1220                    bpf_jit_supports_far_kfunc_call()) {
1221                 err = bpf_get_kfunc_addr(prog, insn->imm, insn->off, &addr);
1222                 if (err)
1223                         return err;
1224         } else {
1225                 /* Address of a BPF helper call. Since part of the core
1226                  * kernel, it's always at a fixed location. __bpf_call_base
1227                  * and the helper with imm relative to it are both in core
1228                  * kernel.
1229                  */
1230                 addr = (u8 *)__bpf_call_base + imm;
1231         }
1232
1233         *func_addr = (unsigned long)addr;
1234         return 0;
1235 }
1236
1237 static int bpf_jit_blind_insn(const struct bpf_insn *from,
1238                               const struct bpf_insn *aux,
1239                               struct bpf_insn *to_buff,
1240                               bool emit_zext)
1241 {
1242         struct bpf_insn *to = to_buff;
1243         u32 imm_rnd = get_random_u32();
1244         s16 off;
1245
1246         BUILD_BUG_ON(BPF_REG_AX  + 1 != MAX_BPF_JIT_REG);
1247         BUILD_BUG_ON(MAX_BPF_REG + 1 != MAX_BPF_JIT_REG);
1248
1249         /* Constraints on AX register:
1250          *
1251          * AX register is inaccessible from user space. It is mapped in
1252          * all JITs, and used here for constant blinding rewrites. It is
1253          * typically "stateless" meaning its contents are only valid within
1254          * the executed instruction, but not across several instructions.
1255          * There are a few exceptions however which are further detailed
1256          * below.
1257          *
1258          * Constant blinding is only used by JITs, not in the interpreter.
1259          * The interpreter uses AX in some occasions as a local temporary
1260          * register e.g. in DIV or MOD instructions.
1261          *
1262          * In restricted circumstances, the verifier can also use the AX
1263          * register for rewrites as long as they do not interfere with
1264          * the above cases!
1265          */
1266         if (from->dst_reg == BPF_REG_AX || from->src_reg == BPF_REG_AX)
1267                 goto out;
1268
1269         if (from->imm == 0 &&
1270             (from->code == (BPF_ALU   | BPF_MOV | BPF_K) ||
1271              from->code == (BPF_ALU64 | BPF_MOV | BPF_K))) {
1272                 *to++ = BPF_ALU64_REG(BPF_XOR, from->dst_reg, from->dst_reg);
1273                 goto out;
1274         }
1275
1276         switch (from->code) {
1277         case BPF_ALU | BPF_ADD | BPF_K:
1278         case BPF_ALU | BPF_SUB | BPF_K:
1279         case BPF_ALU | BPF_AND | BPF_K:
1280         case BPF_ALU | BPF_OR  | BPF_K:
1281         case BPF_ALU | BPF_XOR | BPF_K:
1282         case BPF_ALU | BPF_MUL | BPF_K:
1283         case BPF_ALU | BPF_MOV | BPF_K:
1284         case BPF_ALU | BPF_DIV | BPF_K:
1285         case BPF_ALU | BPF_MOD | BPF_K:
1286                 *to++ = BPF_ALU32_IMM(BPF_MOV, BPF_REG_AX, imm_rnd ^ from->imm);
1287                 *to++ = BPF_ALU32_IMM(BPF_XOR, BPF_REG_AX, imm_rnd);
1288                 *to++ = BPF_ALU32_REG_OFF(from->code, from->dst_reg, BPF_REG_AX, from->off);
1289                 break;
1290
1291         case BPF_ALU64 | BPF_ADD | BPF_K:
1292         case BPF_ALU64 | BPF_SUB | BPF_K:
1293         case BPF_ALU64 | BPF_AND | BPF_K:
1294         case BPF_ALU64 | BPF_OR  | BPF_K:
1295         case BPF_ALU64 | BPF_XOR | BPF_K:
1296         case BPF_ALU64 | BPF_MUL | BPF_K:
1297         case BPF_ALU64 | BPF_MOV | BPF_K:
1298         case BPF_ALU64 | BPF_DIV | BPF_K:
1299         case BPF_ALU64 | BPF_MOD | BPF_K:
1300                 *to++ = BPF_ALU64_IMM(BPF_MOV, BPF_REG_AX, imm_rnd ^ from->imm);
1301                 *to++ = BPF_ALU64_IMM(BPF_XOR, BPF_REG_AX, imm_rnd);
1302                 *to++ = BPF_ALU64_REG_OFF(from->code, from->dst_reg, BPF_REG_AX, from->off);
1303                 break;
1304
1305         case BPF_JMP | BPF_JEQ  | BPF_K:
1306         case BPF_JMP | BPF_JNE  | BPF_K:
1307         case BPF_JMP | BPF_JGT  | BPF_K:
1308         case BPF_JMP | BPF_JLT  | BPF_K:
1309         case BPF_JMP | BPF_JGE  | BPF_K:
1310         case BPF_JMP | BPF_JLE  | BPF_K:
1311         case BPF_JMP | BPF_JSGT | BPF_K:
1312         case BPF_JMP | BPF_JSLT | BPF_K:
1313         case BPF_JMP | BPF_JSGE | BPF_K:
1314         case BPF_JMP | BPF_JSLE | BPF_K:
1315         case BPF_JMP | BPF_JSET | BPF_K:
1316                 /* Accommodate for extra offset in case of a backjump. */
1317                 off = from->off;
1318                 if (off < 0)
1319                         off -= 2;
1320                 *to++ = BPF_ALU64_IMM(BPF_MOV, BPF_REG_AX, imm_rnd ^ from->imm);
1321                 *to++ = BPF_ALU64_IMM(BPF_XOR, BPF_REG_AX, imm_rnd);
1322                 *to++ = BPF_JMP_REG(from->code, from->dst_reg, BPF_REG_AX, off);
1323                 break;
1324
1325         case BPF_JMP32 | BPF_JEQ  | BPF_K:
1326         case BPF_JMP32 | BPF_JNE  | BPF_K:
1327         case BPF_JMP32 | BPF_JGT  | BPF_K:
1328         case BPF_JMP32 | BPF_JLT  | BPF_K:
1329         case BPF_JMP32 | BPF_JGE  | BPF_K:
1330         case BPF_JMP32 | BPF_JLE  | BPF_K:
1331         case BPF_JMP32 | BPF_JSGT | BPF_K:
1332         case BPF_JMP32 | BPF_JSLT | BPF_K:
1333         case BPF_JMP32 | BPF_JSGE | BPF_K:
1334         case BPF_JMP32 | BPF_JSLE | BPF_K:
1335         case BPF_JMP32 | BPF_JSET | BPF_K:
1336                 /* Accommodate for extra offset in case of a backjump. */
1337                 off = from->off;
1338                 if (off < 0)
1339                         off -= 2;
1340                 *to++ = BPF_ALU32_IMM(BPF_MOV, BPF_REG_AX, imm_rnd ^ from->imm);
1341                 *to++ = BPF_ALU32_IMM(BPF_XOR, BPF_REG_AX, imm_rnd);
1342                 *to++ = BPF_JMP32_REG(from->code, from->dst_reg, BPF_REG_AX,
1343                                       off);
1344                 break;
1345
1346         case BPF_LD | BPF_IMM | BPF_DW:
1347                 *to++ = BPF_ALU64_IMM(BPF_MOV, BPF_REG_AX, imm_rnd ^ aux[1].imm);
1348                 *to++ = BPF_ALU64_IMM(BPF_XOR, BPF_REG_AX, imm_rnd);
1349                 *to++ = BPF_ALU64_IMM(BPF_LSH, BPF_REG_AX, 32);
1350                 *to++ = BPF_ALU64_REG(BPF_MOV, aux[0].dst_reg, BPF_REG_AX);
1351                 break;
1352         case 0: /* Part 2 of BPF_LD | BPF_IMM | BPF_DW. */
1353                 *to++ = BPF_ALU32_IMM(BPF_MOV, BPF_REG_AX, imm_rnd ^ aux[0].imm);
1354                 *to++ = BPF_ALU32_IMM(BPF_XOR, BPF_REG_AX, imm_rnd);
1355                 if (emit_zext)
1356                         *to++ = BPF_ZEXT_REG(BPF_REG_AX);
1357                 *to++ = BPF_ALU64_REG(BPF_OR,  aux[0].dst_reg, BPF_REG_AX);
1358                 break;
1359
1360         case BPF_ST | BPF_MEM | BPF_DW:
1361         case BPF_ST | BPF_MEM | BPF_W:
1362         case BPF_ST | BPF_MEM | BPF_H:
1363         case BPF_ST | BPF_MEM | BPF_B:
1364                 *to++ = BPF_ALU64_IMM(BPF_MOV, BPF_REG_AX, imm_rnd ^ from->imm);
1365                 *to++ = BPF_ALU64_IMM(BPF_XOR, BPF_REG_AX, imm_rnd);
1366                 *to++ = BPF_STX_MEM(from->code, from->dst_reg, BPF_REG_AX, from->off);
1367                 break;
1368         }
1369 out:
1370         return to - to_buff;
1371 }
1372
1373 static struct bpf_prog *bpf_prog_clone_create(struct bpf_prog *fp_other,
1374                                               gfp_t gfp_extra_flags)
1375 {
1376         gfp_t gfp_flags = GFP_KERNEL | __GFP_ZERO | gfp_extra_flags;
1377         struct bpf_prog *fp;
1378
1379         fp = __vmalloc(fp_other->pages * PAGE_SIZE, gfp_flags);
1380         if (fp != NULL) {
1381                 /* aux->prog still points to the fp_other one, so
1382                  * when promoting the clone to the real program,
1383                  * this still needs to be adapted.
1384                  */
1385                 memcpy(fp, fp_other, fp_other->pages * PAGE_SIZE);
1386         }
1387
1388         return fp;
1389 }
1390
1391 static void bpf_prog_clone_free(struct bpf_prog *fp)
1392 {
1393         /* aux was stolen by the other clone, so we cannot free
1394          * it from this path! It will be freed eventually by the
1395          * other program on release.
1396          *
1397          * At this point, we don't need a deferred release since
1398          * clone is guaranteed to not be locked.
1399          */
1400         fp->aux = NULL;
1401         fp->stats = NULL;
1402         fp->active = NULL;
1403         __bpf_prog_free(fp);
1404 }
1405
1406 void bpf_jit_prog_release_other(struct bpf_prog *fp, struct bpf_prog *fp_other)
1407 {
1408         /* We have to repoint aux->prog to self, as we don't
1409          * know whether fp here is the clone or the original.
1410          */
1411         fp->aux->prog = fp;
1412         bpf_prog_clone_free(fp_other);
1413 }
1414
1415 struct bpf_prog *bpf_jit_blind_constants(struct bpf_prog *prog)
1416 {
1417         struct bpf_insn insn_buff[16], aux[2];
1418         struct bpf_prog *clone, *tmp;
1419         int insn_delta, insn_cnt;
1420         struct bpf_insn *insn;
1421         int i, rewritten;
1422
1423         if (!prog->blinding_requested || prog->blinded)
1424                 return prog;
1425
1426         clone = bpf_prog_clone_create(prog, GFP_USER);
1427         if (!clone)
1428                 return ERR_PTR(-ENOMEM);
1429
1430         insn_cnt = clone->len;
1431         insn = clone->insnsi;
1432
1433         for (i = 0; i < insn_cnt; i++, insn++) {
1434                 if (bpf_pseudo_func(insn)) {
1435                         /* ld_imm64 with an address of bpf subprog is not
1436                          * a user controlled constant. Don't randomize it,
1437                          * since it will conflict with jit_subprogs() logic.
1438                          */
1439                         insn++;
1440                         i++;
1441                         continue;
1442                 }
1443
1444                 /* We temporarily need to hold the original ld64 insn
1445                  * so that we can still access the first part in the
1446                  * second blinding run.
1447                  */
1448                 if (insn[0].code == (BPF_LD | BPF_IMM | BPF_DW) &&
1449                     insn[1].code == 0)
1450                         memcpy(aux, insn, sizeof(aux));
1451
1452                 rewritten = bpf_jit_blind_insn(insn, aux, insn_buff,
1453                                                 clone->aux->verifier_zext);
1454                 if (!rewritten)
1455                         continue;
1456
1457                 tmp = bpf_patch_insn_single(clone, i, insn_buff, rewritten);
1458                 if (IS_ERR(tmp)) {
1459                         /* Patching may have repointed aux->prog during
1460                          * realloc from the original one, so we need to
1461                          * fix it up here on error.
1462                          */
1463                         bpf_jit_prog_release_other(prog, clone);
1464                         return tmp;
1465                 }
1466
1467                 clone = tmp;
1468                 insn_delta = rewritten - 1;
1469
1470                 /* Walk new program and skip insns we just inserted. */
1471                 insn = clone->insnsi + i + insn_delta;
1472                 insn_cnt += insn_delta;
1473                 i        += insn_delta;
1474         }
1475
1476         clone->blinded = 1;
1477         return clone;
1478 }
1479 #endif /* CONFIG_BPF_JIT */
1480
1481 /* Base function for offset calculation. Needs to go into .text section,
1482  * therefore keeping it non-static as well; will also be used by JITs
1483  * anyway later on, so do not let the compiler omit it. This also needs
1484  * to go into kallsyms for correlation from e.g. bpftool, so naming
1485  * must not change.
1486  */
1487 noinline u64 __bpf_call_base(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5)
1488 {
1489         return 0;
1490 }
1491 EXPORT_SYMBOL_GPL(__bpf_call_base);
1492
1493 /* All UAPI available opcodes. */
1494 #define BPF_INSN_MAP(INSN_2, INSN_3)            \
1495         /* 32 bit ALU operations. */            \
1496         /*   Register based. */                 \
1497         INSN_3(ALU, ADD,  X),                   \
1498         INSN_3(ALU, SUB,  X),                   \
1499         INSN_3(ALU, AND,  X),                   \
1500         INSN_3(ALU, OR,   X),                   \
1501         INSN_3(ALU, LSH,  X),                   \
1502         INSN_3(ALU, RSH,  X),                   \
1503         INSN_3(ALU, XOR,  X),                   \
1504         INSN_3(ALU, MUL,  X),                   \
1505         INSN_3(ALU, MOV,  X),                   \
1506         INSN_3(ALU, ARSH, X),                   \
1507         INSN_3(ALU, DIV,  X),                   \
1508         INSN_3(ALU, MOD,  X),                   \
1509         INSN_2(ALU, NEG),                       \
1510         INSN_3(ALU, END, TO_BE),                \
1511         INSN_3(ALU, END, TO_LE),                \
1512         /*   Immediate based. */                \
1513         INSN_3(ALU, ADD,  K),                   \
1514         INSN_3(ALU, SUB,  K),                   \
1515         INSN_3(ALU, AND,  K),                   \
1516         INSN_3(ALU, OR,   K),                   \
1517         INSN_3(ALU, LSH,  K),                   \
1518         INSN_3(ALU, RSH,  K),                   \
1519         INSN_3(ALU, XOR,  K),                   \
1520         INSN_3(ALU, MUL,  K),                   \
1521         INSN_3(ALU, MOV,  K),                   \
1522         INSN_3(ALU, ARSH, K),                   \
1523         INSN_3(ALU, DIV,  K),                   \
1524         INSN_3(ALU, MOD,  K),                   \
1525         /* 64 bit ALU operations. */            \
1526         /*   Register based. */                 \
1527         INSN_3(ALU64, ADD,  X),                 \
1528         INSN_3(ALU64, SUB,  X),                 \
1529         INSN_3(ALU64, AND,  X),                 \
1530         INSN_3(ALU64, OR,   X),                 \
1531         INSN_3(ALU64, LSH,  X),                 \
1532         INSN_3(ALU64, RSH,  X),                 \
1533         INSN_3(ALU64, XOR,  X),                 \
1534         INSN_3(ALU64, MUL,  X),                 \
1535         INSN_3(ALU64, MOV,  X),                 \
1536         INSN_3(ALU64, ARSH, X),                 \
1537         INSN_3(ALU64, DIV,  X),                 \
1538         INSN_3(ALU64, MOD,  X),                 \
1539         INSN_2(ALU64, NEG),                     \
1540         INSN_3(ALU64, END, TO_LE),              \
1541         /*   Immediate based. */                \
1542         INSN_3(ALU64, ADD,  K),                 \
1543         INSN_3(ALU64, SUB,  K),                 \
1544         INSN_3(ALU64, AND,  K),                 \
1545         INSN_3(ALU64, OR,   K),                 \
1546         INSN_3(ALU64, LSH,  K),                 \
1547         INSN_3(ALU64, RSH,  K),                 \
1548         INSN_3(ALU64, XOR,  K),                 \
1549         INSN_3(ALU64, MUL,  K),                 \
1550         INSN_3(ALU64, MOV,  K),                 \
1551         INSN_3(ALU64, ARSH, K),                 \
1552         INSN_3(ALU64, DIV,  K),                 \
1553         INSN_3(ALU64, MOD,  K),                 \
1554         /* Call instruction. */                 \
1555         INSN_2(JMP, CALL),                      \
1556         /* Exit instruction. */                 \
1557         INSN_2(JMP, EXIT),                      \
1558         /* 32-bit Jump instructions. */         \
1559         /*   Register based. */                 \
1560         INSN_3(JMP32, JEQ,  X),                 \
1561         INSN_3(JMP32, JNE,  X),                 \
1562         INSN_3(JMP32, JGT,  X),                 \
1563         INSN_3(JMP32, JLT,  X),                 \
1564         INSN_3(JMP32, JGE,  X),                 \
1565         INSN_3(JMP32, JLE,  X),                 \
1566         INSN_3(JMP32, JSGT, X),                 \
1567         INSN_3(JMP32, JSLT, X),                 \
1568         INSN_3(JMP32, JSGE, X),                 \
1569         INSN_3(JMP32, JSLE, X),                 \
1570         INSN_3(JMP32, JSET, X),                 \
1571         /*   Immediate based. */                \
1572         INSN_3(JMP32, JEQ,  K),                 \
1573         INSN_3(JMP32, JNE,  K),                 \
1574         INSN_3(JMP32, JGT,  K),                 \
1575         INSN_3(JMP32, JLT,  K),                 \
1576         INSN_3(JMP32, JGE,  K),                 \
1577         INSN_3(JMP32, JLE,  K),                 \
1578         INSN_3(JMP32, JSGT, K),                 \
1579         INSN_3(JMP32, JSLT, K),                 \
1580         INSN_3(JMP32, JSGE, K),                 \
1581         INSN_3(JMP32, JSLE, K),                 \
1582         INSN_3(JMP32, JSET, K),                 \
1583         /* Jump instructions. */                \
1584         /*   Register based. */                 \
1585         INSN_3(JMP, JEQ,  X),                   \
1586         INSN_3(JMP, JNE,  X),                   \
1587         INSN_3(JMP, JGT,  X),                   \
1588         INSN_3(JMP, JLT,  X),                   \
1589         INSN_3(JMP, JGE,  X),                   \
1590         INSN_3(JMP, JLE,  X),                   \
1591         INSN_3(JMP, JSGT, X),                   \
1592         INSN_3(JMP, JSLT, X),                   \
1593         INSN_3(JMP, JSGE, X),                   \
1594         INSN_3(JMP, JSLE, X),                   \
1595         INSN_3(JMP, JSET, X),                   \
1596         /*   Immediate based. */                \
1597         INSN_3(JMP, JEQ,  K),                   \
1598         INSN_3(JMP, JNE,  K),                   \
1599         INSN_3(JMP, JGT,  K),                   \
1600         INSN_3(JMP, JLT,  K),                   \
1601         INSN_3(JMP, JGE,  K),                   \
1602         INSN_3(JMP, JLE,  K),                   \
1603         INSN_3(JMP, JSGT, K),                   \
1604         INSN_3(JMP, JSLT, K),                   \
1605         INSN_3(JMP, JSGE, K),                   \
1606         INSN_3(JMP, JSLE, K),                   \
1607         INSN_3(JMP, JSET, K),                   \
1608         INSN_2(JMP, JA),                        \
1609         INSN_2(JMP32, JA),                      \
1610         /* Store instructions. */               \
1611         /*   Register based. */                 \
1612         INSN_3(STX, MEM,  B),                   \
1613         INSN_3(STX, MEM,  H),                   \
1614         INSN_3(STX, MEM,  W),                   \
1615         INSN_3(STX, MEM,  DW),                  \
1616         INSN_3(STX, ATOMIC, W),                 \
1617         INSN_3(STX, ATOMIC, DW),                \
1618         /*   Immediate based. */                \
1619         INSN_3(ST, MEM, B),                     \
1620         INSN_3(ST, MEM, H),                     \
1621         INSN_3(ST, MEM, W),                     \
1622         INSN_3(ST, MEM, DW),                    \
1623         /* Load instructions. */                \
1624         /*   Register based. */                 \
1625         INSN_3(LDX, MEM, B),                    \
1626         INSN_3(LDX, MEM, H),                    \
1627         INSN_3(LDX, MEM, W),                    \
1628         INSN_3(LDX, MEM, DW),                   \
1629         INSN_3(LDX, MEMSX, B),                  \
1630         INSN_3(LDX, MEMSX, H),                  \
1631         INSN_3(LDX, MEMSX, W),                  \
1632         /*   Immediate based. */                \
1633         INSN_3(LD, IMM, DW)
1634
1635 bool bpf_opcode_in_insntable(u8 code)
1636 {
1637 #define BPF_INSN_2_TBL(x, y)    [BPF_##x | BPF_##y] = true
1638 #define BPF_INSN_3_TBL(x, y, z) [BPF_##x | BPF_##y | BPF_##z] = true
1639         static const bool public_insntable[256] = {
1640                 [0 ... 255] = false,
1641                 /* Now overwrite non-defaults ... */
1642                 BPF_INSN_MAP(BPF_INSN_2_TBL, BPF_INSN_3_TBL),
1643                 /* UAPI exposed, but rewritten opcodes. cBPF carry-over. */
1644                 [BPF_LD | BPF_ABS | BPF_B] = true,
1645                 [BPF_LD | BPF_ABS | BPF_H] = true,
1646                 [BPF_LD | BPF_ABS | BPF_W] = true,
1647                 [BPF_LD | BPF_IND | BPF_B] = true,
1648                 [BPF_LD | BPF_IND | BPF_H] = true,
1649                 [BPF_LD | BPF_IND | BPF_W] = true,
1650         };
1651 #undef BPF_INSN_3_TBL
1652 #undef BPF_INSN_2_TBL
1653         return public_insntable[code];
1654 }
1655
1656 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
1657 /**
1658  *      ___bpf_prog_run - run eBPF program on a given context
1659  *      @regs: is the array of MAX_BPF_EXT_REG eBPF pseudo-registers
1660  *      @insn: is the array of eBPF instructions
1661  *
1662  * Decode and execute eBPF instructions.
1663  *
1664  * Return: whatever value is in %BPF_R0 at program exit
1665  */
1666 static u64 ___bpf_prog_run(u64 *regs, const struct bpf_insn *insn)
1667 {
1668 #define BPF_INSN_2_LBL(x, y)    [BPF_##x | BPF_##y] = &&x##_##y
1669 #define BPF_INSN_3_LBL(x, y, z) [BPF_##x | BPF_##y | BPF_##z] = &&x##_##y##_##z
1670         static const void * const jumptable[256] __annotate_jump_table = {
1671                 [0 ... 255] = &&default_label,
1672                 /* Now overwrite non-defaults ... */
1673                 BPF_INSN_MAP(BPF_INSN_2_LBL, BPF_INSN_3_LBL),
1674                 /* Non-UAPI available opcodes. */
1675                 [BPF_JMP | BPF_CALL_ARGS] = &&JMP_CALL_ARGS,
1676                 [BPF_JMP | BPF_TAIL_CALL] = &&JMP_TAIL_CALL,
1677                 [BPF_ST  | BPF_NOSPEC] = &&ST_NOSPEC,
1678                 [BPF_LDX | BPF_PROBE_MEM | BPF_B] = &&LDX_PROBE_MEM_B,
1679                 [BPF_LDX | BPF_PROBE_MEM | BPF_H] = &&LDX_PROBE_MEM_H,
1680                 [BPF_LDX | BPF_PROBE_MEM | BPF_W] = &&LDX_PROBE_MEM_W,
1681                 [BPF_LDX | BPF_PROBE_MEM | BPF_DW] = &&LDX_PROBE_MEM_DW,
1682                 [BPF_LDX | BPF_PROBE_MEMSX | BPF_B] = &&LDX_PROBE_MEMSX_B,
1683                 [BPF_LDX | BPF_PROBE_MEMSX | BPF_H] = &&LDX_PROBE_MEMSX_H,
1684                 [BPF_LDX | BPF_PROBE_MEMSX | BPF_W] = &&LDX_PROBE_MEMSX_W,
1685         };
1686 #undef BPF_INSN_3_LBL
1687 #undef BPF_INSN_2_LBL
1688         u32 tail_call_cnt = 0;
1689
1690 #define CONT     ({ insn++; goto select_insn; })
1691 #define CONT_JMP ({ insn++; goto select_insn; })
1692
1693 select_insn:
1694         goto *jumptable[insn->code];
1695
1696         /* Explicitly mask the register-based shift amounts with 63 or 31
1697          * to avoid undefined behavior. Normally this won't affect the
1698          * generated code, for example, in case of native 64 bit archs such
1699          * as x86-64 or arm64, the compiler is optimizing the AND away for
1700          * the interpreter. In case of JITs, each of the JIT backends compiles
1701          * the BPF shift operations to machine instructions which produce
1702          * implementation-defined results in such a case; the resulting
1703          * contents of the register may be arbitrary, but program behaviour
1704          * as a whole remains defined. In other words, in case of JIT backends,
1705          * the AND must /not/ be added to the emitted LSH/RSH/ARSH translation.
1706          */
1707         /* ALU (shifts) */
1708 #define SHT(OPCODE, OP)                                 \
1709         ALU64_##OPCODE##_X:                             \
1710                 DST = DST OP (SRC & 63);                \
1711                 CONT;                                   \
1712         ALU_##OPCODE##_X:                               \
1713                 DST = (u32) DST OP ((u32) SRC & 31);    \
1714                 CONT;                                   \
1715         ALU64_##OPCODE##_K:                             \
1716                 DST = DST OP IMM;                       \
1717                 CONT;                                   \
1718         ALU_##OPCODE##_K:                               \
1719                 DST = (u32) DST OP (u32) IMM;           \
1720                 CONT;
1721         /* ALU (rest) */
1722 #define ALU(OPCODE, OP)                                 \
1723         ALU64_##OPCODE##_X:                             \
1724                 DST = DST OP SRC;                       \
1725                 CONT;                                   \
1726         ALU_##OPCODE##_X:                               \
1727                 DST = (u32) DST OP (u32) SRC;           \
1728                 CONT;                                   \
1729         ALU64_##OPCODE##_K:                             \
1730                 DST = DST OP IMM;                       \
1731                 CONT;                                   \
1732         ALU_##OPCODE##_K:                               \
1733                 DST = (u32) DST OP (u32) IMM;           \
1734                 CONT;
1735         ALU(ADD,  +)
1736         ALU(SUB,  -)
1737         ALU(AND,  &)
1738         ALU(OR,   |)
1739         ALU(XOR,  ^)
1740         ALU(MUL,  *)
1741         SHT(LSH, <<)
1742         SHT(RSH, >>)
1743 #undef SHT
1744 #undef ALU
1745         ALU_NEG:
1746                 DST = (u32) -DST;
1747                 CONT;
1748         ALU64_NEG:
1749                 DST = -DST;
1750                 CONT;
1751         ALU_MOV_X:
1752                 switch (OFF) {
1753                 case 0:
1754                         DST = (u32) SRC;
1755                         break;
1756                 case 8:
1757                         DST = (u32)(s8) SRC;
1758                         break;
1759                 case 16:
1760                         DST = (u32)(s16) SRC;
1761                         break;
1762                 }
1763                 CONT;
1764         ALU_MOV_K:
1765                 DST = (u32) IMM;
1766                 CONT;
1767         ALU64_MOV_X:
1768                 switch (OFF) {
1769                 case 0:
1770                         DST = SRC;
1771                         break;
1772                 case 8:
1773                         DST = (s8) SRC;
1774                         break;
1775                 case 16:
1776                         DST = (s16) SRC;
1777                         break;
1778                 case 32:
1779                         DST = (s32) SRC;
1780                         break;
1781                 }
1782                 CONT;
1783         ALU64_MOV_K:
1784                 DST = IMM;
1785                 CONT;
1786         LD_IMM_DW:
1787                 DST = (u64) (u32) insn[0].imm | ((u64) (u32) insn[1].imm) << 32;
1788                 insn++;
1789                 CONT;
1790         ALU_ARSH_X:
1791                 DST = (u64) (u32) (((s32) DST) >> (SRC & 31));
1792                 CONT;
1793         ALU_ARSH_K:
1794                 DST = (u64) (u32) (((s32) DST) >> IMM);
1795                 CONT;
1796         ALU64_ARSH_X:
1797                 (*(s64 *) &DST) >>= (SRC & 63);
1798                 CONT;
1799         ALU64_ARSH_K:
1800                 (*(s64 *) &DST) >>= IMM;
1801                 CONT;
1802         ALU64_MOD_X:
1803                 switch (OFF) {
1804                 case 0:
1805                         div64_u64_rem(DST, SRC, &AX);
1806                         DST = AX;
1807                         break;
1808                 case 1:
1809                         AX = div64_s64(DST, SRC);
1810                         DST = DST - AX * SRC;
1811                         break;
1812                 }
1813                 CONT;
1814         ALU_MOD_X:
1815                 switch (OFF) {
1816                 case 0:
1817                         AX = (u32) DST;
1818                         DST = do_div(AX, (u32) SRC);
1819                         break;
1820                 case 1:
1821                         AX = abs((s32)DST);
1822                         AX = do_div(AX, abs((s32)SRC));
1823                         if ((s32)DST < 0)
1824                                 DST = (u32)-AX;
1825                         else
1826                                 DST = (u32)AX;
1827                         break;
1828                 }
1829                 CONT;
1830         ALU64_MOD_K:
1831                 switch (OFF) {
1832                 case 0:
1833                         div64_u64_rem(DST, IMM, &AX);
1834                         DST = AX;
1835                         break;
1836                 case 1:
1837                         AX = div64_s64(DST, IMM);
1838                         DST = DST - AX * IMM;
1839                         break;
1840                 }
1841                 CONT;
1842         ALU_MOD_K:
1843                 switch (OFF) {
1844                 case 0:
1845                         AX = (u32) DST;
1846                         DST = do_div(AX, (u32) IMM);
1847                         break;
1848                 case 1:
1849                         AX = abs((s32)DST);
1850                         AX = do_div(AX, abs((s32)IMM));
1851                         if ((s32)DST < 0)
1852                                 DST = (u32)-AX;
1853                         else
1854                                 DST = (u32)AX;
1855                         break;
1856                 }
1857                 CONT;
1858         ALU64_DIV_X:
1859                 switch (OFF) {
1860                 case 0:
1861                         DST = div64_u64(DST, SRC);
1862                         break;
1863                 case 1:
1864                         DST = div64_s64(DST, SRC);
1865                         break;
1866                 }
1867                 CONT;
1868         ALU_DIV_X:
1869                 switch (OFF) {
1870                 case 0:
1871                         AX = (u32) DST;
1872                         do_div(AX, (u32) SRC);
1873                         DST = (u32) AX;
1874                         break;
1875                 case 1:
1876                         AX = abs((s32)DST);
1877                         do_div(AX, abs((s32)SRC));
1878                         if (((s32)DST < 0) == ((s32)SRC < 0))
1879                                 DST = (u32)AX;
1880                         else
1881                                 DST = (u32)-AX;
1882                         break;
1883                 }
1884                 CONT;
1885         ALU64_DIV_K:
1886                 switch (OFF) {
1887                 case 0:
1888                         DST = div64_u64(DST, IMM);
1889                         break;
1890                 case 1:
1891                         DST = div64_s64(DST, IMM);
1892                         break;
1893                 }
1894                 CONT;
1895         ALU_DIV_K:
1896                 switch (OFF) {
1897                 case 0:
1898                         AX = (u32) DST;
1899                         do_div(AX, (u32) IMM);
1900                         DST = (u32) AX;
1901                         break;
1902                 case 1:
1903                         AX = abs((s32)DST);
1904                         do_div(AX, abs((s32)IMM));
1905                         if (((s32)DST < 0) == ((s32)IMM < 0))
1906                                 DST = (u32)AX;
1907                         else
1908                                 DST = (u32)-AX;
1909                         break;
1910                 }
1911                 CONT;
1912         ALU_END_TO_BE:
1913                 switch (IMM) {
1914                 case 16:
1915                         DST = (__force u16) cpu_to_be16(DST);
1916                         break;
1917                 case 32:
1918                         DST = (__force u32) cpu_to_be32(DST);
1919                         break;
1920                 case 64:
1921                         DST = (__force u64) cpu_to_be64(DST);
1922                         break;
1923                 }
1924                 CONT;
1925         ALU_END_TO_LE:
1926                 switch (IMM) {
1927                 case 16:
1928                         DST = (__force u16) cpu_to_le16(DST);
1929                         break;
1930                 case 32:
1931                         DST = (__force u32) cpu_to_le32(DST);
1932                         break;
1933                 case 64:
1934                         DST = (__force u64) cpu_to_le64(DST);
1935                         break;
1936                 }
1937                 CONT;
1938         ALU64_END_TO_LE:
1939                 switch (IMM) {
1940                 case 16:
1941                         DST = (__force u16) __swab16(DST);
1942                         break;
1943                 case 32:
1944                         DST = (__force u32) __swab32(DST);
1945                         break;
1946                 case 64:
1947                         DST = (__force u64) __swab64(DST);
1948                         break;
1949                 }
1950                 CONT;
1951
1952         /* CALL */
1953         JMP_CALL:
1954                 /* Function call scratches BPF_R1-BPF_R5 registers,
1955                  * preserves BPF_R6-BPF_R9, and stores return value
1956                  * into BPF_R0.
1957                  */
1958                 BPF_R0 = (__bpf_call_base + insn->imm)(BPF_R1, BPF_R2, BPF_R3,
1959                                                        BPF_R4, BPF_R5);
1960                 CONT;
1961
1962         JMP_CALL_ARGS:
1963                 BPF_R0 = (__bpf_call_base_args + insn->imm)(BPF_R1, BPF_R2,
1964                                                             BPF_R3, BPF_R4,
1965                                                             BPF_R5,
1966                                                             insn + insn->off + 1);
1967                 CONT;
1968
1969         JMP_TAIL_CALL: {
1970                 struct bpf_map *map = (struct bpf_map *) (unsigned long) BPF_R2;
1971                 struct bpf_array *array = container_of(map, struct bpf_array, map);
1972                 struct bpf_prog *prog;
1973                 u32 index = BPF_R3;
1974
1975                 if (unlikely(index >= array->map.max_entries))
1976                         goto out;
1977
1978                 if (unlikely(tail_call_cnt >= MAX_TAIL_CALL_CNT))
1979                         goto out;
1980
1981                 tail_call_cnt++;
1982
1983                 prog = READ_ONCE(array->ptrs[index]);
1984                 if (!prog)
1985                         goto out;
1986
1987                 /* ARG1 at this point is guaranteed to point to CTX from
1988                  * the verifier side due to the fact that the tail call is
1989                  * handled like a helper, that is, bpf_tail_call_proto,
1990                  * where arg1_type is ARG_PTR_TO_CTX.
1991                  */
1992                 insn = prog->insnsi;
1993                 goto select_insn;
1994 out:
1995                 CONT;
1996         }
1997         JMP_JA:
1998                 insn += insn->off;
1999                 CONT;
2000         JMP32_JA:
2001                 insn += insn->imm;
2002                 CONT;
2003         JMP_EXIT:
2004                 return BPF_R0;
2005         /* JMP */
2006 #define COND_JMP(SIGN, OPCODE, CMP_OP)                          \
2007         JMP_##OPCODE##_X:                                       \
2008                 if ((SIGN##64) DST CMP_OP (SIGN##64) SRC) {     \
2009                         insn += insn->off;                      \
2010                         CONT_JMP;                               \
2011                 }                                               \
2012                 CONT;                                           \
2013         JMP32_##OPCODE##_X:                                     \
2014                 if ((SIGN##32) DST CMP_OP (SIGN##32) SRC) {     \
2015                         insn += insn->off;                      \
2016                         CONT_JMP;                               \
2017                 }                                               \
2018                 CONT;                                           \
2019         JMP_##OPCODE##_K:                                       \
2020                 if ((SIGN##64) DST CMP_OP (SIGN##64) IMM) {     \
2021                         insn += insn->off;                      \
2022                         CONT_JMP;                               \
2023                 }                                               \
2024                 CONT;                                           \
2025         JMP32_##OPCODE##_K:                                     \
2026                 if ((SIGN##32) DST CMP_OP (SIGN##32) IMM) {     \
2027                         insn += insn->off;                      \
2028                         CONT_JMP;                               \
2029                 }                                               \
2030                 CONT;
2031         COND_JMP(u, JEQ, ==)
2032         COND_JMP(u, JNE, !=)
2033         COND_JMP(u, JGT, >)
2034         COND_JMP(u, JLT, <)
2035         COND_JMP(u, JGE, >=)
2036         COND_JMP(u, JLE, <=)
2037         COND_JMP(u, JSET, &)
2038         COND_JMP(s, JSGT, >)
2039         COND_JMP(s, JSLT, <)
2040         COND_JMP(s, JSGE, >=)
2041         COND_JMP(s, JSLE, <=)
2042 #undef COND_JMP
2043         /* ST, STX and LDX*/
2044         ST_NOSPEC:
2045                 /* Speculation barrier for mitigating Speculative Store Bypass.
2046                  * In case of arm64, we rely on the firmware mitigation as
2047                  * controlled via the ssbd kernel parameter. Whenever the
2048                  * mitigation is enabled, it works for all of the kernel code
2049                  * with no need to provide any additional instructions here.
2050                  * In case of x86, we use 'lfence' insn for mitigation. We
2051                  * reuse preexisting logic from Spectre v1 mitigation that
2052                  * happens to produce the required code on x86 for v4 as well.
2053                  */
2054                 barrier_nospec();
2055                 CONT;
2056 #define LDST(SIZEOP, SIZE)                                              \
2057         STX_MEM_##SIZEOP:                                               \
2058                 *(SIZE *)(unsigned long) (DST + insn->off) = SRC;       \
2059                 CONT;                                                   \
2060         ST_MEM_##SIZEOP:                                                \
2061                 *(SIZE *)(unsigned long) (DST + insn->off) = IMM;       \
2062                 CONT;                                                   \
2063         LDX_MEM_##SIZEOP:                                               \
2064                 DST = *(SIZE *)(unsigned long) (SRC + insn->off);       \
2065                 CONT;                                                   \
2066         LDX_PROBE_MEM_##SIZEOP:                                         \
2067                 bpf_probe_read_kernel_common(&DST, sizeof(SIZE),        \
2068                               (const void *)(long) (SRC + insn->off));  \
2069                 DST = *((SIZE *)&DST);                                  \
2070                 CONT;
2071
2072         LDST(B,   u8)
2073         LDST(H,  u16)
2074         LDST(W,  u32)
2075         LDST(DW, u64)
2076 #undef LDST
2077
2078 #define LDSX(SIZEOP, SIZE)                                              \
2079         LDX_MEMSX_##SIZEOP:                                             \
2080                 DST = *(SIZE *)(unsigned long) (SRC + insn->off);       \
2081                 CONT;                                                   \
2082         LDX_PROBE_MEMSX_##SIZEOP:                                       \
2083                 bpf_probe_read_kernel_common(&DST, sizeof(SIZE),                \
2084                                       (const void *)(long) (SRC + insn->off));  \
2085                 DST = *((SIZE *)&DST);                                  \
2086                 CONT;
2087
2088         LDSX(B,   s8)
2089         LDSX(H,  s16)
2090         LDSX(W,  s32)
2091 #undef LDSX
2092
2093 #define ATOMIC_ALU_OP(BOP, KOP)                                         \
2094                 case BOP:                                               \
2095                         if (BPF_SIZE(insn->code) == BPF_W)              \
2096                                 atomic_##KOP((u32) SRC, (atomic_t *)(unsigned long) \
2097                                              (DST + insn->off));        \
2098                         else                                            \
2099                                 atomic64_##KOP((u64) SRC, (atomic64_t *)(unsigned long) \
2100                                                (DST + insn->off));      \
2101                         break;                                          \
2102                 case BOP | BPF_FETCH:                                   \
2103                         if (BPF_SIZE(insn->code) == BPF_W)              \
2104                                 SRC = (u32) atomic_fetch_##KOP(         \
2105                                         (u32) SRC,                      \
2106                                         (atomic_t *)(unsigned long) (DST + insn->off)); \
2107                         else                                            \
2108                                 SRC = (u64) atomic64_fetch_##KOP(       \
2109                                         (u64) SRC,                      \
2110                                         (atomic64_t *)(unsigned long) (DST + insn->off)); \
2111                         break;
2112
2113         STX_ATOMIC_DW:
2114         STX_ATOMIC_W:
2115                 switch (IMM) {
2116                 ATOMIC_ALU_OP(BPF_ADD, add)
2117                 ATOMIC_ALU_OP(BPF_AND, and)
2118                 ATOMIC_ALU_OP(BPF_OR, or)
2119                 ATOMIC_ALU_OP(BPF_XOR, xor)
2120 #undef ATOMIC_ALU_OP
2121
2122                 case BPF_XCHG:
2123                         if (BPF_SIZE(insn->code) == BPF_W)
2124                                 SRC = (u32) atomic_xchg(
2125                                         (atomic_t *)(unsigned long) (DST + insn->off),
2126                                         (u32) SRC);
2127                         else
2128                                 SRC = (u64) atomic64_xchg(
2129                                         (atomic64_t *)(unsigned long) (DST + insn->off),
2130                                         (u64) SRC);
2131                         break;
2132                 case BPF_CMPXCHG:
2133                         if (BPF_SIZE(insn->code) == BPF_W)
2134                                 BPF_R0 = (u32) atomic_cmpxchg(
2135                                         (atomic_t *)(unsigned long) (DST + insn->off),
2136                                         (u32) BPF_R0, (u32) SRC);
2137                         else
2138                                 BPF_R0 = (u64) atomic64_cmpxchg(
2139                                         (atomic64_t *)(unsigned long) (DST + insn->off),
2140                                         (u64) BPF_R0, (u64) SRC);
2141                         break;
2142
2143                 default:
2144                         goto default_label;
2145                 }
2146                 CONT;
2147
2148         default_label:
2149                 /* If we ever reach this, we have a bug somewhere. Die hard here
2150                  * instead of just returning 0; we could be somewhere in a subprog,
2151                  * so execution could continue otherwise which we do /not/ want.
2152                  *
2153                  * Note, verifier whitelists all opcodes in bpf_opcode_in_insntable().
2154                  */
2155                 pr_warn("BPF interpreter: unknown opcode %02x (imm: 0x%x)\n",
2156                         insn->code, insn->imm);
2157                 BUG_ON(1);
2158                 return 0;
2159 }
2160
2161 #define PROG_NAME(stack_size) __bpf_prog_run##stack_size
2162 #define DEFINE_BPF_PROG_RUN(stack_size) \
2163 static unsigned int PROG_NAME(stack_size)(const void *ctx, const struct bpf_insn *insn) \
2164 { \
2165         u64 stack[stack_size / sizeof(u64)]; \
2166         u64 regs[MAX_BPF_EXT_REG] = {}; \
2167 \
2168         FP = (u64) (unsigned long) &stack[ARRAY_SIZE(stack)]; \
2169         ARG1 = (u64) (unsigned long) ctx; \
2170         return ___bpf_prog_run(regs, insn); \
2171 }
2172
2173 #define PROG_NAME_ARGS(stack_size) __bpf_prog_run_args##stack_size
2174 #define DEFINE_BPF_PROG_RUN_ARGS(stack_size) \
2175 static u64 PROG_NAME_ARGS(stack_size)(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5, \
2176                                       const struct bpf_insn *insn) \
2177 { \
2178         u64 stack[stack_size / sizeof(u64)]; \
2179         u64 regs[MAX_BPF_EXT_REG]; \
2180 \
2181         FP = (u64) (unsigned long) &stack[ARRAY_SIZE(stack)]; \
2182         BPF_R1 = r1; \
2183         BPF_R2 = r2; \
2184         BPF_R3 = r3; \
2185         BPF_R4 = r4; \
2186         BPF_R5 = r5; \
2187         return ___bpf_prog_run(regs, insn); \
2188 }
2189
2190 #define EVAL1(FN, X) FN(X)
2191 #define EVAL2(FN, X, Y...) FN(X) EVAL1(FN, Y)
2192 #define EVAL3(FN, X, Y...) FN(X) EVAL2(FN, Y)
2193 #define EVAL4(FN, X, Y...) FN(X) EVAL3(FN, Y)
2194 #define EVAL5(FN, X, Y...) FN(X) EVAL4(FN, Y)
2195 #define EVAL6(FN, X, Y...) FN(X) EVAL5(FN, Y)
2196
2197 EVAL6(DEFINE_BPF_PROG_RUN, 32, 64, 96, 128, 160, 192);
2198 EVAL6(DEFINE_BPF_PROG_RUN, 224, 256, 288, 320, 352, 384);
2199 EVAL4(DEFINE_BPF_PROG_RUN, 416, 448, 480, 512);
2200
2201 EVAL6(DEFINE_BPF_PROG_RUN_ARGS, 32, 64, 96, 128, 160, 192);
2202 EVAL6(DEFINE_BPF_PROG_RUN_ARGS, 224, 256, 288, 320, 352, 384);
2203 EVAL4(DEFINE_BPF_PROG_RUN_ARGS, 416, 448, 480, 512);
2204
2205 #define PROG_NAME_LIST(stack_size) PROG_NAME(stack_size),
2206
2207 static unsigned int (*interpreters[])(const void *ctx,
2208                                       const struct bpf_insn *insn) = {
2209 EVAL6(PROG_NAME_LIST, 32, 64, 96, 128, 160, 192)
2210 EVAL6(PROG_NAME_LIST, 224, 256, 288, 320, 352, 384)
2211 EVAL4(PROG_NAME_LIST, 416, 448, 480, 512)
2212 };
2213 #undef PROG_NAME_LIST
2214 #define PROG_NAME_LIST(stack_size) PROG_NAME_ARGS(stack_size),
2215 static __maybe_unused
2216 u64 (*interpreters_args[])(u64 r1, u64 r2, u64 r3, u64 r4, u64 r5,
2217                            const struct bpf_insn *insn) = {
2218 EVAL6(PROG_NAME_LIST, 32, 64, 96, 128, 160, 192)
2219 EVAL6(PROG_NAME_LIST, 224, 256, 288, 320, 352, 384)
2220 EVAL4(PROG_NAME_LIST, 416, 448, 480, 512)
2221 };
2222 #undef PROG_NAME_LIST
2223
2224 #ifdef CONFIG_BPF_SYSCALL
2225 void bpf_patch_call_args(struct bpf_insn *insn, u32 stack_depth)
2226 {
2227         stack_depth = max_t(u32, stack_depth, 1);
2228         insn->off = (s16) insn->imm;
2229         insn->imm = interpreters_args[(round_up(stack_depth, 32) / 32) - 1] -
2230                 __bpf_call_base_args;
2231         insn->code = BPF_JMP | BPF_CALL_ARGS;
2232 }
2233 #endif
2234 #else
2235 static unsigned int __bpf_prog_ret0_warn(const void *ctx,
2236                                          const struct bpf_insn *insn)
2237 {
2238         /* If this handler ever gets executed, then BPF_JIT_ALWAYS_ON
2239          * is not working properly, so warn about it!
2240          */
2241         WARN_ON_ONCE(1);
2242         return 0;
2243 }
2244 #endif
2245
2246 bool bpf_prog_map_compatible(struct bpf_map *map,
2247                              const struct bpf_prog *fp)
2248 {
2249         enum bpf_prog_type prog_type = resolve_prog_type(fp);
2250         bool ret;
2251
2252         if (fp->kprobe_override)
2253                 return false;
2254
2255         /* XDP programs inserted into maps are not guaranteed to run on
2256          * a particular netdev (and can run outside driver context entirely
2257          * in the case of devmap and cpumap). Until device checks
2258          * are implemented, prohibit adding dev-bound programs to program maps.
2259          */
2260         if (bpf_prog_is_dev_bound(fp->aux))
2261                 return false;
2262
2263         spin_lock(&map->owner.lock);
2264         if (!map->owner.type) {
2265                 /* There's no owner yet where we could check for
2266                  * compatibility.
2267                  */
2268                 map->owner.type  = prog_type;
2269                 map->owner.jited = fp->jited;
2270                 map->owner.xdp_has_frags = fp->aux->xdp_has_frags;
2271                 ret = true;
2272         } else {
2273                 ret = map->owner.type  == prog_type &&
2274                       map->owner.jited == fp->jited &&
2275                       map->owner.xdp_has_frags == fp->aux->xdp_has_frags;
2276         }
2277         spin_unlock(&map->owner.lock);
2278
2279         return ret;
2280 }
2281
2282 static int bpf_check_tail_call(const struct bpf_prog *fp)
2283 {
2284         struct bpf_prog_aux *aux = fp->aux;
2285         int i, ret = 0;
2286
2287         mutex_lock(&aux->used_maps_mutex);
2288         for (i = 0; i < aux->used_map_cnt; i++) {
2289                 struct bpf_map *map = aux->used_maps[i];
2290
2291                 if (!map_type_contains_progs(map))
2292                         continue;
2293
2294                 if (!bpf_prog_map_compatible(map, fp)) {
2295                         ret = -EINVAL;
2296                         goto out;
2297                 }
2298         }
2299
2300 out:
2301         mutex_unlock(&aux->used_maps_mutex);
2302         return ret;
2303 }
2304
2305 static void bpf_prog_select_func(struct bpf_prog *fp)
2306 {
2307 #ifndef CONFIG_BPF_JIT_ALWAYS_ON
2308         u32 stack_depth = max_t(u32, fp->aux->stack_depth, 1);
2309
2310         fp->bpf_func = interpreters[(round_up(stack_depth, 32) / 32) - 1];
2311 #else
2312         fp->bpf_func = __bpf_prog_ret0_warn;
2313 #endif
2314 }
2315
2316 /**
2317  *      bpf_prog_select_runtime - select exec runtime for BPF program
2318  *      @fp: bpf_prog populated with BPF program
2319  *      @err: pointer to error variable
2320  *
2321  * Try to JIT eBPF program, if JIT is not available, use interpreter.
2322  * The BPF program will be executed via bpf_prog_run() function.
2323  *
2324  * Return: the &fp argument along with &err set to 0 for success or
2325  * a negative errno code on failure
2326  */
2327 struct bpf_prog *bpf_prog_select_runtime(struct bpf_prog *fp, int *err)
2328 {
2329         /* In case of BPF to BPF calls, verifier did all the prep
2330          * work with regards to JITing, etc.
2331          */
2332         bool jit_needed = false;
2333
2334         if (fp->bpf_func)
2335                 goto finalize;
2336
2337         if (IS_ENABLED(CONFIG_BPF_JIT_ALWAYS_ON) ||
2338             bpf_prog_has_kfunc_call(fp))
2339                 jit_needed = true;
2340
2341         bpf_prog_select_func(fp);
2342
2343         /* eBPF JITs can rewrite the program in case constant
2344          * blinding is active. However, in case of error during
2345          * blinding, bpf_int_jit_compile() must always return a
2346          * valid program, which in this case would simply not
2347          * be JITed, but falls back to the interpreter.
2348          */
2349         if (!bpf_prog_is_offloaded(fp->aux)) {
2350                 *err = bpf_prog_alloc_jited_linfo(fp);
2351                 if (*err)
2352                         return fp;
2353
2354                 fp = bpf_int_jit_compile(fp);
2355                 bpf_prog_jit_attempt_done(fp);
2356                 if (!fp->jited && jit_needed) {
2357                         *err = -ENOTSUPP;
2358                         return fp;
2359                 }
2360         } else {
2361                 *err = bpf_prog_offload_compile(fp);
2362                 if (*err)
2363                         return fp;
2364         }
2365
2366 finalize:
2367         bpf_prog_lock_ro(fp);
2368
2369         /* The tail call compatibility check can only be done at
2370          * this late stage as we need to determine, if we deal
2371          * with JITed or non JITed program concatenations and not
2372          * all eBPF JITs might immediately support all features.
2373          */
2374         *err = bpf_check_tail_call(fp);
2375
2376         return fp;
2377 }
2378 EXPORT_SYMBOL_GPL(bpf_prog_select_runtime);
2379
2380 static unsigned int __bpf_prog_ret1(const void *ctx,
2381                                     const struct bpf_insn *insn)
2382 {
2383         return 1;
2384 }
2385
2386 static struct bpf_prog_dummy {
2387         struct bpf_prog prog;
2388 } dummy_bpf_prog = {
2389         .prog = {
2390                 .bpf_func = __bpf_prog_ret1,
2391         },
2392 };
2393
2394 struct bpf_empty_prog_array bpf_empty_prog_array = {
2395         .null_prog = NULL,
2396 };
2397 EXPORT_SYMBOL(bpf_empty_prog_array);
2398
2399 struct bpf_prog_array *bpf_prog_array_alloc(u32 prog_cnt, gfp_t flags)
2400 {
2401         if (prog_cnt)
2402                 return kzalloc(sizeof(struct bpf_prog_array) +
2403                                sizeof(struct bpf_prog_array_item) *
2404                                (prog_cnt + 1),
2405                                flags);
2406
2407         return &bpf_empty_prog_array.hdr;
2408 }
2409
2410 void bpf_prog_array_free(struct bpf_prog_array *progs)
2411 {
2412         if (!progs || progs == &bpf_empty_prog_array.hdr)
2413                 return;
2414         kfree_rcu(progs, rcu);
2415 }
2416
2417 static void __bpf_prog_array_free_sleepable_cb(struct rcu_head *rcu)
2418 {
2419         struct bpf_prog_array *progs;
2420
2421         /* If RCU Tasks Trace grace period implies RCU grace period, there is
2422          * no need to call kfree_rcu(), just call kfree() directly.
2423          */
2424         progs = container_of(rcu, struct bpf_prog_array, rcu);
2425         if (rcu_trace_implies_rcu_gp())
2426                 kfree(progs);
2427         else
2428                 kfree_rcu(progs, rcu);
2429 }
2430
2431 void bpf_prog_array_free_sleepable(struct bpf_prog_array *progs)
2432 {
2433         if (!progs || progs == &bpf_empty_prog_array.hdr)
2434                 return;
2435         call_rcu_tasks_trace(&progs->rcu, __bpf_prog_array_free_sleepable_cb);
2436 }
2437
2438 int bpf_prog_array_length(struct bpf_prog_array *array)
2439 {
2440         struct bpf_prog_array_item *item;
2441         u32 cnt = 0;
2442
2443         for (item = array->items; item->prog; item++)
2444                 if (item->prog != &dummy_bpf_prog.prog)
2445                         cnt++;
2446         return cnt;
2447 }
2448
2449 bool bpf_prog_array_is_empty(struct bpf_prog_array *array)
2450 {
2451         struct bpf_prog_array_item *item;
2452
2453         for (item = array->items; item->prog; item++)
2454                 if (item->prog != &dummy_bpf_prog.prog)
2455                         return false;
2456         return true;
2457 }
2458
2459 static bool bpf_prog_array_copy_core(struct bpf_prog_array *array,
2460                                      u32 *prog_ids,
2461                                      u32 request_cnt)
2462 {
2463         struct bpf_prog_array_item *item;
2464         int i = 0;
2465
2466         for (item = array->items; item->prog; item++) {
2467                 if (item->prog == &dummy_bpf_prog.prog)
2468                         continue;
2469                 prog_ids[i] = item->prog->aux->id;
2470                 if (++i == request_cnt) {
2471                         item++;
2472                         break;
2473                 }
2474         }
2475
2476         return !!(item->prog);
2477 }
2478
2479 int bpf_prog_array_copy_to_user(struct bpf_prog_array *array,
2480                                 __u32 __user *prog_ids, u32 cnt)
2481 {
2482         unsigned long err = 0;
2483         bool nospc;
2484         u32 *ids;
2485
2486         /* users of this function are doing:
2487          * cnt = bpf_prog_array_length();
2488          * if (cnt > 0)
2489          *     bpf_prog_array_copy_to_user(..., cnt);
2490          * so below kcalloc doesn't need extra cnt > 0 check.
2491          */
2492         ids = kcalloc(cnt, sizeof(u32), GFP_USER | __GFP_NOWARN);
2493         if (!ids)
2494                 return -ENOMEM;
2495         nospc = bpf_prog_array_copy_core(array, ids, cnt);
2496         err = copy_to_user(prog_ids, ids, cnt * sizeof(u32));
2497         kfree(ids);
2498         if (err)
2499                 return -EFAULT;
2500         if (nospc)
2501                 return -ENOSPC;
2502         return 0;
2503 }
2504
2505 void bpf_prog_array_delete_safe(struct bpf_prog_array *array,
2506                                 struct bpf_prog *old_prog)
2507 {
2508         struct bpf_prog_array_item *item;
2509
2510         for (item = array->items; item->prog; item++)
2511                 if (item->prog == old_prog) {
2512                         WRITE_ONCE(item->prog, &dummy_bpf_prog.prog);
2513                         break;
2514                 }
2515 }
2516
2517 /**
2518  * bpf_prog_array_delete_safe_at() - Replaces the program at the given
2519  *                                   index into the program array with
2520  *                                   a dummy no-op program.
2521  * @array: a bpf_prog_array
2522  * @index: the index of the program to replace
2523  *
2524  * Skips over dummy programs, by not counting them, when calculating
2525  * the position of the program to replace.
2526  *
2527  * Return:
2528  * * 0          - Success
2529  * * -EINVAL    - Invalid index value. Must be a non-negative integer.
2530  * * -ENOENT    - Index out of range
2531  */
2532 int bpf_prog_array_delete_safe_at(struct bpf_prog_array *array, int index)
2533 {
2534         return bpf_prog_array_update_at(array, index, &dummy_bpf_prog.prog);
2535 }
2536
2537 /**
2538  * bpf_prog_array_update_at() - Updates the program at the given index
2539  *                              into the program array.
2540  * @array: a bpf_prog_array
2541  * @index: the index of the program to update
2542  * @prog: the program to insert into the array
2543  *
2544  * Skips over dummy programs, by not counting them, when calculating
2545  * the position of the program to update.
2546  *
2547  * Return:
2548  * * 0          - Success
2549  * * -EINVAL    - Invalid index value. Must be a non-negative integer.
2550  * * -ENOENT    - Index out of range
2551  */
2552 int bpf_prog_array_update_at(struct bpf_prog_array *array, int index,
2553                              struct bpf_prog *prog)
2554 {
2555         struct bpf_prog_array_item *item;
2556
2557         if (unlikely(index < 0))
2558                 return -EINVAL;
2559
2560         for (item = array->items; item->prog; item++) {
2561                 if (item->prog == &dummy_bpf_prog.prog)
2562                         continue;
2563                 if (!index) {
2564                         WRITE_ONCE(item->prog, prog);
2565                         return 0;
2566                 }
2567                 index--;
2568         }
2569         return -ENOENT;
2570 }
2571
2572 int bpf_prog_array_copy(struct bpf_prog_array *old_array,
2573                         struct bpf_prog *exclude_prog,
2574                         struct bpf_prog *include_prog,
2575                         u64 bpf_cookie,
2576                         struct bpf_prog_array **new_array)
2577 {
2578         int new_prog_cnt, carry_prog_cnt = 0;
2579         struct bpf_prog_array_item *existing, *new;
2580         struct bpf_prog_array *array;
2581         bool found_exclude = false;
2582
2583         /* Figure out how many existing progs we need to carry over to
2584          * the new array.
2585          */
2586         if (old_array) {
2587                 existing = old_array->items;
2588                 for (; existing->prog; existing++) {
2589                         if (existing->prog == exclude_prog) {
2590                                 found_exclude = true;
2591                                 continue;
2592                         }
2593                         if (existing->prog != &dummy_bpf_prog.prog)
2594                                 carry_prog_cnt++;
2595                         if (existing->prog == include_prog)
2596                                 return -EEXIST;
2597                 }
2598         }
2599
2600         if (exclude_prog && !found_exclude)
2601                 return -ENOENT;
2602
2603         /* How many progs (not NULL) will be in the new array? */
2604         new_prog_cnt = carry_prog_cnt;
2605         if (include_prog)
2606                 new_prog_cnt += 1;
2607
2608         /* Do we have any prog (not NULL) in the new array? */
2609         if (!new_prog_cnt) {
2610                 *new_array = NULL;
2611                 return 0;
2612         }
2613
2614         /* +1 as the end of prog_array is marked with NULL */
2615         array = bpf_prog_array_alloc(new_prog_cnt + 1, GFP_KERNEL);
2616         if (!array)
2617                 return -ENOMEM;
2618         new = array->items;
2619
2620         /* Fill in the new prog array */
2621         if (carry_prog_cnt) {
2622                 existing = old_array->items;
2623                 for (; existing->prog; existing++) {
2624                         if (existing->prog == exclude_prog ||
2625                             existing->prog == &dummy_bpf_prog.prog)
2626                                 continue;
2627
2628                         new->prog = existing->prog;
2629                         new->bpf_cookie = existing->bpf_cookie;
2630                         new++;
2631                 }
2632         }
2633         if (include_prog) {
2634                 new->prog = include_prog;
2635                 new->bpf_cookie = bpf_cookie;
2636                 new++;
2637         }
2638         new->prog = NULL;
2639         *new_array = array;
2640         return 0;
2641 }
2642
2643 int bpf_prog_array_copy_info(struct bpf_prog_array *array,
2644                              u32 *prog_ids, u32 request_cnt,
2645                              u32 *prog_cnt)
2646 {
2647         u32 cnt = 0;
2648
2649         if (array)
2650                 cnt = bpf_prog_array_length(array);
2651
2652         *prog_cnt = cnt;
2653
2654         /* return early if user requested only program count or nothing to copy */
2655         if (!request_cnt || !cnt)
2656                 return 0;
2657
2658         /* this function is called under trace/bpf_trace.c: bpf_event_mutex */
2659         return bpf_prog_array_copy_core(array, prog_ids, request_cnt) ? -ENOSPC
2660                                                                      : 0;
2661 }
2662
2663 void __bpf_free_used_maps(struct bpf_prog_aux *aux,
2664                           struct bpf_map **used_maps, u32 len)
2665 {
2666         struct bpf_map *map;
2667         u32 i;
2668
2669         for (i = 0; i < len; i++) {
2670                 map = used_maps[i];
2671                 if (map->ops->map_poke_untrack)
2672                         map->ops->map_poke_untrack(map, aux);
2673                 bpf_map_put(map);
2674         }
2675 }
2676
2677 static void bpf_free_used_maps(struct bpf_prog_aux *aux)
2678 {
2679         __bpf_free_used_maps(aux, aux->used_maps, aux->used_map_cnt);
2680         kfree(aux->used_maps);
2681 }
2682
2683 void __bpf_free_used_btfs(struct bpf_prog_aux *aux,
2684                           struct btf_mod_pair *used_btfs, u32 len)
2685 {
2686 #ifdef CONFIG_BPF_SYSCALL
2687         struct btf_mod_pair *btf_mod;
2688         u32 i;
2689
2690         for (i = 0; i < len; i++) {
2691                 btf_mod = &used_btfs[i];
2692                 if (btf_mod->module)
2693                         module_put(btf_mod->module);
2694                 btf_put(btf_mod->btf);
2695         }
2696 #endif
2697 }
2698
2699 static void bpf_free_used_btfs(struct bpf_prog_aux *aux)
2700 {
2701         __bpf_free_used_btfs(aux, aux->used_btfs, aux->used_btf_cnt);
2702         kfree(aux->used_btfs);
2703 }
2704
2705 static void bpf_prog_free_deferred(struct work_struct *work)
2706 {
2707         struct bpf_prog_aux *aux;
2708         int i;
2709
2710         aux = container_of(work, struct bpf_prog_aux, work);
2711 #ifdef CONFIG_BPF_SYSCALL
2712         bpf_free_kfunc_btf_tab(aux->kfunc_btf_tab);
2713 #endif
2714 #ifdef CONFIG_CGROUP_BPF
2715         if (aux->cgroup_atype != CGROUP_BPF_ATTACH_TYPE_INVALID)
2716                 bpf_cgroup_atype_put(aux->cgroup_atype);
2717 #endif
2718         bpf_free_used_maps(aux);
2719         bpf_free_used_btfs(aux);
2720         if (bpf_prog_is_dev_bound(aux))
2721                 bpf_prog_dev_bound_destroy(aux->prog);
2722 #ifdef CONFIG_PERF_EVENTS
2723         if (aux->prog->has_callchain_buf)
2724                 put_callchain_buffers();
2725 #endif
2726         if (aux->dst_trampoline)
2727                 bpf_trampoline_put(aux->dst_trampoline);
2728         for (i = 0; i < aux->func_cnt; i++) {
2729                 /* We can just unlink the subprog poke descriptor table as
2730                  * it was originally linked to the main program and is also
2731                  * released along with it.
2732                  */
2733                 aux->func[i]->aux->poke_tab = NULL;
2734                 bpf_jit_free(aux->func[i]);
2735         }
2736         if (aux->func_cnt) {
2737                 kfree(aux->func);
2738                 bpf_prog_unlock_free(aux->prog);
2739         } else {
2740                 bpf_jit_free(aux->prog);
2741         }
2742 }
2743
2744 void bpf_prog_free(struct bpf_prog *fp)
2745 {
2746         struct bpf_prog_aux *aux = fp->aux;
2747
2748         if (aux->dst_prog)
2749                 bpf_prog_put(aux->dst_prog);
2750         INIT_WORK(&aux->work, bpf_prog_free_deferred);
2751         schedule_work(&aux->work);
2752 }
2753 EXPORT_SYMBOL_GPL(bpf_prog_free);
2754
2755 /* RNG for unpriviledged user space with separated state from prandom_u32(). */
2756 static DEFINE_PER_CPU(struct rnd_state, bpf_user_rnd_state);
2757
2758 void bpf_user_rnd_init_once(void)
2759 {
2760         prandom_init_once(&bpf_user_rnd_state);
2761 }
2762
2763 BPF_CALL_0(bpf_user_rnd_u32)
2764 {
2765         /* Should someone ever have the rather unwise idea to use some
2766          * of the registers passed into this function, then note that
2767          * this function is called from native eBPF and classic-to-eBPF
2768          * transformations. Register assignments from both sides are
2769          * different, f.e. classic always sets fn(ctx, A, X) here.
2770          */
2771         struct rnd_state *state;
2772         u32 res;
2773
2774         state = &get_cpu_var(bpf_user_rnd_state);
2775         res = prandom_u32_state(state);
2776         put_cpu_var(bpf_user_rnd_state);
2777
2778         return res;
2779 }
2780
2781 BPF_CALL_0(bpf_get_raw_cpu_id)
2782 {
2783         return raw_smp_processor_id();
2784 }
2785
2786 /* Weak definitions of helper functions in case we don't have bpf syscall. */
2787 const struct bpf_func_proto bpf_map_lookup_elem_proto __weak;
2788 const struct bpf_func_proto bpf_map_update_elem_proto __weak;
2789 const struct bpf_func_proto bpf_map_delete_elem_proto __weak;
2790 const struct bpf_func_proto bpf_map_push_elem_proto __weak;
2791 const struct bpf_func_proto bpf_map_pop_elem_proto __weak;
2792 const struct bpf_func_proto bpf_map_peek_elem_proto __weak;
2793 const struct bpf_func_proto bpf_map_lookup_percpu_elem_proto __weak;
2794 const struct bpf_func_proto bpf_spin_lock_proto __weak;
2795 const struct bpf_func_proto bpf_spin_unlock_proto __weak;
2796 const struct bpf_func_proto bpf_jiffies64_proto __weak;
2797
2798 const struct bpf_func_proto bpf_get_prandom_u32_proto __weak;
2799 const struct bpf_func_proto bpf_get_smp_processor_id_proto __weak;
2800 const struct bpf_func_proto bpf_get_numa_node_id_proto __weak;
2801 const struct bpf_func_proto bpf_ktime_get_ns_proto __weak;
2802 const struct bpf_func_proto bpf_ktime_get_boot_ns_proto __weak;
2803 const struct bpf_func_proto bpf_ktime_get_coarse_ns_proto __weak;
2804 const struct bpf_func_proto bpf_ktime_get_tai_ns_proto __weak;
2805
2806 const struct bpf_func_proto bpf_get_current_pid_tgid_proto __weak;
2807 const struct bpf_func_proto bpf_get_current_uid_gid_proto __weak;
2808 const struct bpf_func_proto bpf_get_current_comm_proto __weak;
2809 const struct bpf_func_proto bpf_get_current_cgroup_id_proto __weak;
2810 const struct bpf_func_proto bpf_get_current_ancestor_cgroup_id_proto __weak;
2811 const struct bpf_func_proto bpf_get_local_storage_proto __weak;
2812 const struct bpf_func_proto bpf_get_ns_current_pid_tgid_proto __weak;
2813 const struct bpf_func_proto bpf_snprintf_btf_proto __weak;
2814 const struct bpf_func_proto bpf_seq_printf_btf_proto __weak;
2815 const struct bpf_func_proto bpf_set_retval_proto __weak;
2816 const struct bpf_func_proto bpf_get_retval_proto __weak;
2817
2818 const struct bpf_func_proto * __weak bpf_get_trace_printk_proto(void)
2819 {
2820         return NULL;
2821 }
2822
2823 const struct bpf_func_proto * __weak bpf_get_trace_vprintk_proto(void)
2824 {
2825         return NULL;
2826 }
2827
2828 u64 __weak
2829 bpf_event_output(struct bpf_map *map, u64 flags, void *meta, u64 meta_size,
2830                  void *ctx, u64 ctx_size, bpf_ctx_copy_t ctx_copy)
2831 {
2832         return -ENOTSUPP;
2833 }
2834 EXPORT_SYMBOL_GPL(bpf_event_output);
2835
2836 /* Always built-in helper functions. */
2837 const struct bpf_func_proto bpf_tail_call_proto = {
2838         .func           = NULL,
2839         .gpl_only       = false,
2840         .ret_type       = RET_VOID,
2841         .arg1_type      = ARG_PTR_TO_CTX,
2842         .arg2_type      = ARG_CONST_MAP_PTR,
2843         .arg3_type      = ARG_ANYTHING,
2844 };
2845
2846 /* Stub for JITs that only support cBPF. eBPF programs are interpreted.
2847  * It is encouraged to implement bpf_int_jit_compile() instead, so that
2848  * eBPF and implicitly also cBPF can get JITed!
2849  */
2850 struct bpf_prog * __weak bpf_int_jit_compile(struct bpf_prog *prog)
2851 {
2852         return prog;
2853 }
2854
2855 /* Stub for JITs that support eBPF. All cBPF code gets transformed into
2856  * eBPF by the kernel and is later compiled by bpf_int_jit_compile().
2857  */
2858 void __weak bpf_jit_compile(struct bpf_prog *prog)
2859 {
2860 }
2861
2862 bool __weak bpf_helper_changes_pkt_data(void *func)
2863 {
2864         return false;
2865 }
2866
2867 /* Return TRUE if the JIT backend wants verifier to enable sub-register usage
2868  * analysis code and wants explicit zero extension inserted by verifier.
2869  * Otherwise, return FALSE.
2870  *
2871  * The verifier inserts an explicit zero extension after BPF_CMPXCHGs even if
2872  * you don't override this. JITs that don't want these extra insns can detect
2873  * them using insn_is_zext.
2874  */
2875 bool __weak bpf_jit_needs_zext(void)
2876 {
2877         return false;
2878 }
2879
2880 /* Return TRUE if the JIT backend supports mixing bpf2bpf and tailcalls. */
2881 bool __weak bpf_jit_supports_subprog_tailcalls(void)
2882 {
2883         return false;
2884 }
2885
2886 bool __weak bpf_jit_supports_kfunc_call(void)
2887 {
2888         return false;
2889 }
2890
2891 bool __weak bpf_jit_supports_far_kfunc_call(void)
2892 {
2893         return false;
2894 }
2895
2896 /* To execute LD_ABS/LD_IND instructions __bpf_prog_run() may call
2897  * skb_copy_bits(), so provide a weak definition of it for NET-less config.
2898  */
2899 int __weak skb_copy_bits(const struct sk_buff *skb, int offset, void *to,
2900                          int len)
2901 {
2902         return -EFAULT;
2903 }
2904
2905 int __weak bpf_arch_text_poke(void *ip, enum bpf_text_poke_type t,
2906                               void *addr1, void *addr2)
2907 {
2908         return -ENOTSUPP;
2909 }
2910
2911 void * __weak bpf_arch_text_copy(void *dst, void *src, size_t len)
2912 {
2913         return ERR_PTR(-ENOTSUPP);
2914 }
2915
2916 int __weak bpf_arch_text_invalidate(void *dst, size_t len)
2917 {
2918         return -ENOTSUPP;
2919 }
2920
2921 #ifdef CONFIG_BPF_SYSCALL
2922 static int __init bpf_global_ma_init(void)
2923 {
2924         int ret;
2925
2926         ret = bpf_mem_alloc_init(&bpf_global_ma, 0, false);
2927         bpf_global_ma_set = !ret;
2928         return ret;
2929 }
2930 late_initcall(bpf_global_ma_init);
2931 #endif
2932
2933 DEFINE_STATIC_KEY_FALSE(bpf_stats_enabled_key);
2934 EXPORT_SYMBOL(bpf_stats_enabled_key);
2935
2936 /* All definitions of tracepoints related to BPF. */
2937 #define CREATE_TRACE_POINTS
2938 #include <linux/bpf_trace.h>
2939
2940 EXPORT_TRACEPOINT_SYMBOL_GPL(xdp_exception);
2941 EXPORT_TRACEPOINT_SYMBOL_GPL(xdp_bulk_tx);