From 0c253ebdd7392e06045c0b1fb16f8aa55d8a7092 Mon Sep 17 00:00:00 2001 From: Rickard Green Date: Wed, 22 May 2019 19:02:21 +0200 Subject: Only validate subject once when global is used in re:run() --- erts/emulator/beam/bif.tab | 1 + erts/emulator/beam/erl_bif_re.c | 29 +++++++++++++++++++++++++---- 2 files changed, 26 insertions(+), 4 deletions(-) (limited to 'erts/emulator') diff --git a/erts/emulator/beam/bif.tab b/erts/emulator/beam/bif.tab index 10ca0b5066..db454a95ea 100644 --- a/erts/emulator/beam/bif.tab +++ b/erts/emulator/beam/bif.tab @@ -399,6 +399,7 @@ bif re:compile/1 bif re:compile/2 bif re:run/2 bif re:run/3 +bif re:internal_run/4 # # Bifs in lists module. diff --git a/erts/emulator/beam/erl_bif_re.c b/erts/emulator/beam/erl_bif_re.c index ad124fd979..9d99d24a61 100644 --- a/erts/emulator/beam/erl_bif_re.c +++ b/erts/emulator/beam/erl_bif_re.c @@ -46,7 +46,7 @@ static Export *urun_trap_exportp = NULL; static Export *ucompile_trap_exportp = NULL; static BIF_RETTYPE re_exec_trap(BIF_ALIST_3); -static BIF_RETTYPE re_run(Process *p, Eterm arg1, Eterm arg2, Eterm arg3); +static BIF_RETTYPE re_run(Process *p, Eterm arg1, Eterm arg2, Eterm arg3, int first); static void *erts_erts_pcre_malloc(size_t size) { return erts_alloc(ERTS_ALC_T_RE_HEAP,size); @@ -1110,7 +1110,7 @@ build_capture(Eterm capture_spec[CAPSPEC_SIZE], const pcre *code) * The actual re:run/2,3 BIFs */ static BIF_RETTYPE -re_run(Process *p, Eterm arg1, Eterm arg2, Eterm arg3) +re_run(Process *p, Eterm arg1, Eterm arg2, Eterm arg3, int first) { const pcre *code_tmp; RestartContext restart; @@ -1136,6 +1136,14 @@ re_run(Process *p, Eterm arg1, Eterm arg2, Eterm arg3) < 0) { BIF_ERROR(p,BADARG); } + if (!first) { + /* + * 'first' is false when re:grun() previously has called re:internal_run() + * with the same subject; i.e., no need to do yet another validation of + * the subject regarding utf8 encoding... + */ + options |= PCRE_NO_UTF8_CHECK; + } is_list_cap = ((pflags & PARSE_FLAG_CAPTURE_OPT) && (capture[CAPSPEC_TYPE] == am_list)); @@ -1386,16 +1394,29 @@ handle_iolist: BIF_RET(res); } +BIF_RETTYPE +re_internal_run_4(BIF_ALIST_4) +{ + int first; + if (BIF_ARG_4 == am_false) + first = 0; + else if (BIF_ARG_4 == am_true) + first = !0; + else + BIF_ERROR(BIF_P,BADARG); + return re_run(BIF_P,BIF_ARG_1, BIF_ARG_2, BIF_ARG_3, first); +} + BIF_RETTYPE re_run_3(BIF_ALIST_3) { - return re_run(BIF_P,BIF_ARG_1, BIF_ARG_2, BIF_ARG_3); + return re_run(BIF_P,BIF_ARG_1, BIF_ARG_2, BIF_ARG_3, !0); } BIF_RETTYPE re_run_2(BIF_ALIST_2) { - return re_run(BIF_P,BIF_ARG_1, BIF_ARG_2, NIL); + return re_run(BIF_P,BIF_ARG_1, BIF_ARG_2, NIL, !0); } /* -- cgit v1.2.3 From b4ee064b539958f6a13bc9554758a106a49f4bd6 Mon Sep 17 00:00:00 2001 From: Rickard Green Date: Wed, 22 May 2019 17:51:42 +0200 Subject: Yield when validating UTF8 for long subject in re:run() --- erts/emulator/beam/erl_bif_re.c | 1 + erts/emulator/pcre/pcre.h | 3 + erts/emulator/pcre/pcre_exec.c | 115 +++++++++++++++++++++++++---------- erts/emulator/pcre/pcre_internal.h | 11 ++++ erts/emulator/pcre/pcre_valid_utf8.c | 73 +++++++++++++++++++++- 5 files changed, 170 insertions(+), 33 deletions(-) (limited to 'erts/emulator') diff --git a/erts/emulator/beam/erl_bif_re.c b/erts/emulator/beam/erl_bif_re.c index ad124fd979..4067d42216 100644 --- a/erts/emulator/beam/erl_bif_re.c +++ b/erts/emulator/beam/erl_bif_re.c @@ -1434,6 +1434,7 @@ static BIF_RETTYPE re_exec_trap(BIF_ALIST_3) loop_count = 0xFFFFFFFF; #endif rc = erts_pcre_exec(NULL, &(restartp->extra), NULL, 0, 0, 0, NULL, 0); + ASSERT(loop_count != 0xFFFFFFFF); BUMP_REDS(BIF_P, loop_count / LOOP_FACTOR); if (rc == PCRE_ERROR_LOOP_LIMIT) { diff --git a/erts/emulator/pcre/pcre.h b/erts/emulator/pcre/pcre.h index ab8f40cfc1..686925bf04 100644 --- a/erts/emulator/pcre/pcre.h +++ b/erts/emulator/pcre/pcre.h @@ -240,6 +240,9 @@ with J. */ #define PCRE_UTF8_ERR20 20 #define PCRE_UTF8_ERR21 21 #define PCRE_UTF8_ERR22 22 /* Unused (was non-character) */ +#if defined(ERLANG_INTEGRATION) +#define PCRE_UTF8_YIELD 23 +#endif /* Specific error codes for UTF-16 validity checks */ diff --git a/erts/emulator/pcre/pcre_exec.c b/erts/emulator/pcre/pcre_exec.c index 6708ba92a6..3602c3aaab 100644 --- a/erts/emulator/pcre/pcre_exec.c +++ b/erts/emulator/pcre/pcre_exec.c @@ -6642,10 +6642,16 @@ typedef struct { REAL_PCRE *Xre; heapframe Xframe_zero; /* Always NO_RECURSE */ + /* for yield in valid_utf() */ + + struct PRIV(valid_utf_ystate) valid_utf_ystate; + /* Original function parameters that need be saved */ int Xstart_offset; int Xoffsetcount; int *Xoffsets; + int Xlength; + PCRE_SPTR Xsubject; } PcreExecContext; #endif @@ -6675,6 +6681,7 @@ pcre32_exec(const pcre32 *argument_re, const pcre32_extra *extra_data, #endif { #ifndef ERLANG_INTEGRATION +#define ERTS_UPDATE_CONSUMED(X, MD) int rc, ocount, arg_offset_max; int newline; BOOL using_temporary_offsets = FALSE; @@ -6736,6 +6743,8 @@ heapframe frame_zero; start_offset = exec_context->Xstart_offset; \ offsetcount = exec_context->Xoffsetcount; \ offsets = exec_context->Xoffsets; \ + length = exec_context->Xlength; \ + subject = exec_context->Xsubject; \ } while (0) #define SWAPOUT() do { \ @@ -6750,8 +6759,30 @@ heapframe frame_zero; exec_context->Xstart_offset = start_offset; \ exec_context->Xoffsetcount = offsetcount; \ exec_context->Xoffsets = offsets; \ + exec_context->Xlength = length; \ + exec_context->Xsubject = subject; \ } while (0) +#define ERTS_UPDATE_CONSUMED(X, MD) \ +do { \ + if (((X)->flags & PCRE_EXTRA_LOOP_LIMIT) != 0) { \ + unsigned long consumed__; \ + if (!(X)->restart_data) { \ + consumed__ = 0; \ + } \ + else { \ + PcreExecContext *ctx__ = (PcreExecContext *) \ + (*(X)->restart_data); \ + consumed__ = ctx__->valid_utf_ystate.cnt; \ + ctx__->valid_utf_ystate.cnt = 0; \ + } \ + if ((MD)) { \ + match_data *md__ = (MD); \ + consumed__ += (X)->loop_limit - md__->loop_limit; \ + } \ + *((X)->loop_counter_return) = consumed__; \ + } \ +} while (0) PcreExecContext *exec_context; PcreExecContext internal_context; @@ -6776,15 +6807,21 @@ pcre_uchar req_char; /* we are restarting, every initialization is skipped and we jump directly into the loop */ exec_context = (PcreExecContext *) *(extra_data->restart_data); SWAPIN(); - + if (exec_context->valid_utf_ystate.yielded) + goto restart_valid_utf; goto RESTART_INTERRUPTED; } else { if (extra_data != NULL && (extra_data->flags & PCRE_EXTRA_LOOP_LIMIT)) { exec_context = (PcreExecContext *) (erts_pcre_malloc)(sizeof(PcreExecContext)); - *(extra_data->restart_data) = (void *) exec_context; + *(extra_data->restart_data) = (void *) exec_context; + exec_context->valid_utf_ystate.yielded = 0; /* need freeing by special routine from client */ } else { +#if defined(ERLANG_INTEGRATION) + fprintf(stderr, "Unexpected execution path\n"); + abort(); +#endif exec_context = &internal_context; } @@ -6865,9 +6902,38 @@ code for an invalid string if a results vector is available. */ if (utf && (options & PCRE_NO_UTF8_CHECK) == 0) { int erroroffset; - int errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length, &erroroffset); + int errorcode; + +#if !defined(ERLANG_INTEGRATION) + errorcode = PRIV(valid_utf)((PCRE_PUCHAR)subject, length); +#else + struct PRIV(valid_utf_ystate) *ystate; + + if (!extra_data || !extra_data->restart_data) { + ystate = NULL; + } + else if (!(extra_data->flags & PCRE_EXTRA_LOOP_LIMIT)) { + exec_context->valid_utf_ystate.cnt = 10; + ystate = NULL; + } + else { + exec_context->valid_utf_ystate.yielded = 0; + restart_valid_utf: + ystate = &exec_context->valid_utf_ystate; + ystate->cnt = (int) extra_data->loop_limit; + } + errorcode = PRIV(yielding_valid_utf)((PCRE_PUCHAR)subject, length, + &erroroffset, ystate); +#endif if (errorcode != 0) { +#if defined(ERLANG_INTEGRATION) + if (ystate && ystate->yielded) { + ERTS_UPDATE_CONSUMED(extra_data, NULL); + SWAPOUT(); + return PCRE_ERROR_LOOP_LIMIT; + } +#endif if (offsetcount >= 2) { offsets[0] = erroroffset; @@ -6890,6 +6956,11 @@ if (utf && (options & PCRE_NO_UTF8_CHECK) == 0) return PCRE_ERROR_BADUTF8_OFFSET; #endif } +#if defined(ERLANG_INTEGRATION) +else { + exec_context->valid_utf_ystate.cnt = 0; +} +#endif #endif /* If the pattern was successfully studied with JIT support, run the JIT @@ -6950,7 +7021,11 @@ if (extra_data != NULL) #ifdef ERLANG_INTEGRATION if ((flags & PCRE_EXTRA_LOOP_LIMIT) != 0) { - md->loop_limit = extra_data->loop_limit; + md->loop_limit = extra_data->loop_limit; + if (extra_data->restart_data) + md->loop_limit -= extra_data->loop_limit - exec_context->valid_utf_ystate.cnt; + if (md->loop_limit < 10) + md->loop_limit = 10; /* At least do something if we've come this far... */ } #endif } @@ -7266,14 +7341,8 @@ for(;;) #endif if ((start_bits[c/8] & (1 << (c&7))) != 0) { -#ifdef ERLANG_INTEGRATION - if ((extra_data->flags & PCRE_EXTRA_LOOP_LIMIT) != 0) - { - *extra_data->loop_counter_return = - (extra_data->loop_limit - md->loop_limit); - } -#endif - break; + ERTS_UPDATE_CONSUMED(extra_data, md); + break; } start_match++; } @@ -7298,13 +7367,7 @@ for(;;) (pcre_uint32)(end_subject - start_match) < study->minlength) { rc = MATCH_NOMATCH; -#ifdef ERLANG_INTEGRATION - if ((extra_data->flags & PCRE_EXTRA_LOOP_LIMIT) != 0) - { - *extra_data->loop_counter_return = - (extra_data->loop_limit - md->loop_limit); - } -#endif + ERTS_UPDATE_CONSUMED(extra_data, md); break; } @@ -7353,13 +7416,7 @@ for(;;) if (p >= end_subject) { rc = MATCH_NOMATCH; -#ifdef ERLANG_INTEGRATION - if ((extra_data->flags & PCRE_EXTRA_LOOP_LIMIT) != 0) - { - *extra_data->loop_counter_return = - (extra_data->loop_limit - md->loop_limit); - } -#endif + ERTS_UPDATE_CONSUMED(extra_data, md); break; } @@ -7390,11 +7447,7 @@ for(;;) EDEBUGF(("Calling match...")); rc = match(start_match, md->start_code, start_match, 2, md, NULL, 0); #ifdef ERLANG_INTEGRATION - if ((extra_data->flags & PCRE_EXTRA_LOOP_LIMIT) != 0) - { - *extra_data->loop_counter_return = - (extra_data->loop_limit - md->loop_limit); - } + ERTS_UPDATE_CONSUMED(extra_data, md); SWAPOUT(); while(rc == PCRE_ERROR_LOOP_LIMIT) { EDEBUGF(("Loop limit break detected")); diff --git a/erts/emulator/pcre/pcre_internal.h b/erts/emulator/pcre/pcre_internal.h index c84dcb5a38..71f473e86f 100644 --- a/erts/emulator/pcre/pcre_internal.h +++ b/erts/emulator/pcre/pcre_internal.h @@ -2756,6 +2756,17 @@ extern int PRIV(strcmp_uc_c8_utf)(const pcre_uchar *, #endif /* COMPILE_PCRE[8|16|32] */ +#if defined(ERLANG_INTEGRATION) +struct PRIV(valid_utf_ystate) { + unsigned int cnt; + int length; + int yielded; + PCRE_PUCHAR p; +}; +extern int PRIV(yielding_valid_utf)(PCRE_PUCHAR, int, int *, + struct PRIV(valid_utf_ystate) *); +#endif + extern const pcre_uchar *PRIV(find_bracket)(const pcre_uchar *, BOOL, int); extern BOOL PRIV(is_newline)(PCRE_PUCHAR, int, PCRE_PUCHAR, int *, BOOL); diff --git a/erts/emulator/pcre/pcre_valid_utf8.c b/erts/emulator/pcre/pcre_valid_utf8.c index 516d8f4725..1dc1f9ba0c 100644 --- a/erts/emulator/pcre/pcre_valid_utf8.c +++ b/erts/emulator/pcre/pcre_valid_utf8.c @@ -107,19 +107,80 @@ Returns: = 0 if the string is a valid UTF-8 string int PRIV(valid_utf)(PCRE_PUCHAR string, int length, int *erroroffset) { + +#if defined(ERLANG_INTEGRATION) + return PRIV(yielding_valid_utf)(string, length, erroroffset, NULL); +} + +int +PRIV(yielding_valid_utf)(PCRE_PUCHAR string, int length, int *erroroffset, struct PRIV(valid_utf_ystate) *ystate) +{ +#endif + #ifdef SUPPORT_UTF register PCRE_PUCHAR p; +#if defined(ERLANG_INTEGRATION) +register long cnt; + +if (!ystate) { + cnt = -1; +} +else { + cnt = ystate->cnt; + if (ystate->yielded) { + p = ystate->p; + length = ystate->length; + if (length < 0) + goto restart_length; + else + goto restart_validate; + } +} +#endif + if (length < 0) { - for (p = string; *p != 0; p++); - length = (int)(p - string); + for (p = string; *p != 0; p++) { +#if defined(ERLANG_INTEGRATION) + if (cnt > 0 && --cnt == 0) { + /* + * Return with cnt set to amount consumed; + * i.e. same amount as at start... + */ + ystate->yielded = !0; + ystate->length = length; + ystate->p = p; + return PCRE_UTF8_YIELD; + } + restart_length: + (void) !0; +#endif + } + length = (int)(p - string); } for (p = string; length-- > 0; p++) { register pcre_uchar ab, c, d; +#if defined(ERLANG_INTEGRATION) + + if (cnt > 0 && --cnt == 0) { + /* + * Return with cnt set to amount consumed; + * i.e. same amount as at start... + */ + ystate->yielded = !0; + ystate->length = length; + ystate->p = p; + return PCRE_UTF8_YIELD; + } + + restart_validate: + +#endif + c = *p; if (c < 128) continue; /* ASCII character */ @@ -290,6 +351,14 @@ for (p = string; length-- > 0; p++) } } +#if defined(ERLANG_INTEGRATION) +if (ystate) { + /* Return with cnt set to amount consumed... */ + ystate->cnt -= cnt; + ystate->yielded = 0; +} +#endif + #else /* Not SUPPORT_UTF */ (void)(string); /* Keep picky compilers happy */ (void)(length); -- cgit v1.2.3