diff options
Diffstat (limited to 'grammar')
-rw-r--r-- | grammar/grammar.y | 77 | ||||
-rw-r--r-- | grammar/lexer.l | 21 | ||||
-rw-r--r-- | grammar/parserif.h | 2 | ||||
-rw-r--r-- | grammar/rainerscript.c | 1018 | ||||
-rw-r--r-- | grammar/rainerscript.h | 106 |
5 files changed, 966 insertions, 258 deletions
diff --git a/grammar/grammar.y b/grammar/grammar.y index cdb19c3d..ce749807 100644 --- a/grammar/grammar.y +++ b/grammar/grammar.y @@ -6,10 +6,9 @@ * of course, encouraged to use new constructs only. But it needs to be noted * that some of the legacy constructs (specifically the in-front-of-action * PRI filter) are very hard to beat in ease of use, at least for simpler - * cases. So while we hope that cfsysline support can be dropped some time in - * the future, we will probably keep these useful constructs. + * cases. * - * Copyright 2011 Rainer Gerhards and Adiscon GmbH. + * Copyright 2011-2012 Rainer Gerhards and Adiscon GmbH. * * This file is part of the rsyslog runtime library. * @@ -37,7 +36,7 @@ #define YYDEBUG 1 extern int yylineno; -/* keep compile rule cleam of errors */ +/* keep compile rule clean of errors */ extern int yylex(void); extern int yyerror(char*); %} @@ -48,11 +47,10 @@ extern int yyerror(char*); es_str_t *estr; enum cnfobjType objType; struct cnfobj *obj; + struct cnfstmt *stmt; struct nvlst *nvlst; struct objlst *objlst; - struct cnfactlst *actlst; struct cnfexpr *expr; - struct cnfrule *rule; struct cnffunc *func; struct cnffparamlst *fparams; } @@ -62,12 +60,13 @@ extern int yyerror(char*); %token <estr> FUNC %token <objType> BEGINOBJ %token ENDOBJ -%token <s> CFSYSLINE %token BEGIN_ACTION %token BEGIN_PROPERTY %token BEGIN_CONSTANT %token BEGIN_TPL %token STOP +%token SET +%token UNSET %token <s> LEGACY_ACTION %token <s> LEGACY_RULESET %token <s> PRIFILT @@ -76,6 +75,7 @@ extern int yyerror(char*); %token <s> BSD_HOST_SELECTOR %token IF %token THEN +%token ELSE %token OR %token AND %token NOT @@ -96,30 +96,27 @@ extern int yyerror(char*); %type <nvlst> nv nvlst %type <obj> obj property constant %type <objlst> propconst -%type <actlst> actlst +/*%type <actlst> actlst %type <actlst> act -%type <s> cfsysline %type <actlst> block +*/ %type <expr> expr +%type <stmt> stmt s_act actlst block script +/* %type <rule> rule %type <rule> scriptfilt +*/ %type <fparams> fparams %left AND OR %left CMP_EQ CMP_NE CMP_LE CMP_GE CMP_LT CMP_GT CMP_CONTAINS CMP_CONTAINSI CMP_STARTSWITH CMP_STARTSWITHI -%left '+' '-' +%left '+' '-' '&' %left '*' '/' '%' %nonassoc UMINUS NOT -%expect 3 -/* these shift/reduce conflicts are created by the CFSYSLINE construct, which we - * unfortunately can not avoid. The problem is that CFSYSLINE can occur both in - * global context as well as within an action. It's not permitted somewhere else, - * but this is suficient for conflicts. The "dangling else" built-in resolution - * works well to solve this issue, so we accept it (it's a wonder that our - * old style grammar doesn't work at all, so we better do not complain...). - * Use "bison -v rscript.y" if more conflicts arise and check rscript.out for - * were exactly these conflicts exits. +%expect 1 /* dangling else */ +/* If more erors show up, Use "bison -v grammar.y" if more conflicts arise and + * check grammar.output for were exactly these conflicts exits. */ %% /* note: we use left recursion below, because that saves stack space AND @@ -128,13 +125,11 @@ extern int yyerror(char*); */ conf: /* empty (to end recursion) */ | conf obj { cnfDoObj($2); } - | conf rule { cnfDoRule($2); } - | conf cfsysline { cnfDoCfsysline($2); } + | conf stmt { cnfDoScript($2); } | conf LEGACY_RULESET { cnfDoCfsysline($2); } | conf BSD_TAG_SELECTOR { cnfDoBSDTag($2); } | conf BSD_HOST_SELECTOR { cnfDoBSDHost($2); } obj: BEGINOBJ nvlst ENDOBJ { $$ = cnfobjNew($1, $2); } - | BEGIN_ACTION nvlst ENDOBJ { $$ = cnfobjNew(CNFOBJ_ACTION, $2); } | BEGIN_TPL nvlst ENDOBJ { $$ = cnfobjNew(CNFOBJ_TPL, $2); } | BEGIN_TPL nvlst ENDOBJ '{' propconst '}' { $$ = cnfobjNew(CNFOBJ_TPL, $2); @@ -145,24 +140,31 @@ propconst: { $$ = NULL; } | propconst constant { $$ = objlstAdd($1, $2); } property: BEGIN_PROPERTY nvlst ENDOBJ { $$ = cnfobjNew(CNFOBJ_PROPERTY, $2); } constant: BEGIN_CONSTANT nvlst ENDOBJ { $$ = cnfobjNew(CNFOBJ_CONSTANT, $2); } -cfsysline: CFSYSLINE { $$ = $1; } nvlst: { $$ = NULL; } | nvlst nv { $2->next = $1; $$ = $2; } nv: NAME '=' VALUE { $$ = nvlstNew($1, $3); } -rule: PRIFILT actlst { $$ = cnfruleNew(CNFFILT_PRI, $2); $$->filt.s = $1; } - | PROPFILT actlst { $$ = cnfruleNew(CNFFILT_PROP, $2); $$->filt.s = $1; } - | scriptfilt { $$ = $1; } - -scriptfilt: IF expr THEN actlst { $$ = cnfruleNew(CNFFILT_SCRIPT, $4); - $$->filt.expr = $2; } -block: actlst { $$ = $1; } - | block actlst { $2->next = $1; $$ = $2; } -actlst: act { $$=$1; } - | actlst '&' act { $3->next = $1; $$ = $3; } - | actlst cfsysline { $$ = cnfactlstAddSysline($1, $2); } - | '{' block '}' { $$ = $2; } -act: BEGIN_ACTION nvlst ENDOBJ { $$ = cnfactlstNew(CNFACT_V2, $2, NULL); } - | LEGACY_ACTION { $$ = cnfactlstNew(CNFACT_LEGACY, NULL, $1); } +script: stmt { $$ = $1; } + | script stmt { $$ = scriptAddStmt($1, $2); } +stmt: actlst { $$ = $1; } + | STOP { $$ = cnfstmtNew(S_STOP); } + | IF expr THEN block { $$ = cnfstmtNew(S_IF); + $$->d.s_if.expr = $2; + $$->d.s_if.t_then = $4; + $$->d.s_if.t_else = NULL; } + | IF expr THEN block ELSE block { $$ = cnfstmtNew(S_IF); + $$->d.s_if.expr = $2; + $$->d.s_if.t_then = $4; + $$->d.s_if.t_else = $6; } + | SET VAR '=' expr ';' { $$ = cnfstmtNewSet($2, $4); } + | UNSET VAR ';' { $$ = cnfstmtNewUnset($2); } + | PRIFILT block { $$ = cnfstmtNewPRIFILT($1, $2); } + | PROPFILT block { $$ = cnfstmtNewPROPFILT($1, $2); } +block: stmt { $$ = $1; } + | '{' script '}' { $$ = $2; } +actlst: s_act { $$ = $1; } + | actlst '&' s_act { $$ = scriptAddStmt($1, $3); } +s_act: BEGIN_ACTION nvlst ENDOBJ { $$ = cnfstmtNewAct($2); } + | LEGACY_ACTION { $$ = cnfstmtNewLegaAct($1); } expr: expr AND expr { $$ = cnfexprNew(AND, $1, $3); } | expr OR expr { $$ = cnfexprNew(OR, $1, $3); } | NOT expr { $$ = cnfexprNew(NOT, NULL, $2); } @@ -176,6 +178,7 @@ expr: expr AND expr { $$ = cnfexprNew(AND, $1, $3); } | expr CMP_CONTAINSI expr { $$ = cnfexprNew(CMP_CONTAINSI, $1, $3); } | expr CMP_STARTSWITH expr { $$ = cnfexprNew(CMP_STARTSWITH, $1, $3); } | expr CMP_STARTSWITHI expr { $$ = cnfexprNew(CMP_STARTSWITHI, $1, $3); } + | expr '&' expr { $$ = cnfexprNew('&', $1, $3); } | expr '+' expr { $$ = cnfexprNew('+', $1, $3); } | expr '-' expr { $$ = cnfexprNew('-', $1, $3); } | expr '*' expr { $$ = cnfexprNew('*', $1, $3); } diff --git a/grammar/lexer.l b/grammar/lexer.l index 4f79c9aa..38703f25 100644 --- a/grammar/lexer.l +++ b/grammar/lexer.l @@ -96,14 +96,17 @@ int fileno(FILE *stream); /* keywords */ "if" { BEGIN EXPR; return IF; } <EXPR>"then" { BEGIN INITIAL; return THEN; } +<EXPR>";" { BEGIN INITIAL; return ';'; } <EXPR>"or" { return OR; } <EXPR>"and" { return AND; } <EXPR>"not" { return NOT; } +<EXPR>"=" | <EXPR>"," | <EXPR>"*" | <EXPR>"/" | <EXPR>"%" | <EXPR>"+" | +<EXPR>"&" | <EXPR>"-" | <EXPR>"(" | <EXPR>")" { return yytext[0]; } @@ -121,7 +124,7 @@ int fileno(FILE *stream); <EXPR>0[0-7]+ | /* octal number */ <EXPR>0x[0-7a-f] | /* hex number, following rule is dec; strtoll handles all! */ <EXPR>([1-9][0-9]*|0) { yylval.n = strtoll(yytext, NULL, 0); return NUMBER; } -<EXPR>\$[$!]{0,1}[a-z][a-z0-9\-_\.]* { yylval.s = strdup(yytext); return VAR; } +<EXPR>\$[$!]{0,1}[a-z][!a-z0-9\-_\.]* { yylval.s = strdup(yytext); return VAR; } <EXPR>\'([^'\\]|\\['"\\$bntr]|\\x[0-9a-f][0-9a-f]|\\[0-7][0-7][0-7])*\' { yytext[yyleng-1] = '\0'; unescapeStr((uchar*)yytext+1, yyleng-2); @@ -143,7 +146,10 @@ int fileno(FILE *stream); /* line number support because the "preprocessor" combines lines and so needs * to tell us the real source line. */ -"stop" { dbgprintf("STOP\n"); return STOP; } +"stop" { return STOP; } +"else" { return ELSE; } +"set" { BEGIN EXPR; return SET; } +"unset" { BEGIN EXPR; return UNSET; } "preprocfilelinenumber(" { BEGIN LINENO; } <LINENO>[0-9]+ { yylineno = atoi(yytext) - 1; } <LINENO>")" { BEGIN INITIAL; } @@ -169,17 +175,15 @@ int fileno(FILE *stream); BEGIN INOBJ; return BEGINOBJ; } "action"[ \n\t]*"(" { BEGIN INOBJ; return BEGIN_ACTION; } ^[ \t]*:\$?[a-z\-]+[ ]*,[ ]*!?[a-z]+[ ]*,[ ]*\".*\" { - yylval.s = strdup(yytext); return PROPFILT; } -^[ \t]*[\*a-z][\*a-z]*[0-7]*[\.,][,!=;\.\*a-z0-7]+ { yylval.s = strdup(yytext); return PRIFILT; } + yylval.s = strdup(rmLeadingSpace(yytext)); return PROPFILT; } +^[ \t]*[\*a-z][\*a-z]*[0-7]*[\.,][,!=;\.\*a-z0-7]+ { yylval.s = strdup(rmLeadingSpace(yytext)); return PRIFILT; } "~" | "*" | \-\/[^*][^\n]* | \/[^*][^\n]* | :[a-z0-9]+:[^\n]* | [\|\.\-\@\^?~>][^\n]+ | -[a-z0-9_][a-z0-9_\-\+,;]* { yylval.s = strdup(yytext); - dbgprintf("lex: LEGA ACT: '%s'\n", yytext); - return LEGACY_ACTION; } +[a-z0-9_][a-z0-9_\-\+,;]* { yylval.s = yytext; return LEGACY_ACTION; } <INOBJ>")" { BEGIN INITIAL; return ENDOBJ; } <INOBJ>[a-z][a-z0-9_\.]* { yylval.estr = es_newStrFromCStr(yytext, yyleng); return NAME; } @@ -202,8 +206,7 @@ int fileno(FILE *stream); yylval.s = strdup(yytext); return LEGACY_RULESET; } else { - yylval.s = strdup(yytext); - return CFSYSLINE; + cnfDoCfsysline(strdup(yytext)); } } ![^ \t\n]+[ \t]*$ { yylval.s = strdup(yytext); return BSD_TAG_SELECTOR; } diff --git a/grammar/parserif.h b/grammar/parserif.h index 597cfe40..aa271ec4 100644 --- a/grammar/parserif.h +++ b/grammar/parserif.h @@ -15,7 +15,7 @@ extern int yylineno; * these functions. */ void cnfDoObj(struct cnfobj *o); -void cnfDoRule(struct cnfrule *rule); +void cnfDoScript(struct cnfstmt *script); void cnfDoCfsysline(char *ln); void cnfDoBSDTag(char *ln); void cnfDoBSDHost(char *ln); diff --git a/grammar/rainerscript.c b/grammar/rainerscript.c index 9e0d04c7..1a14b212 100644 --- a/grammar/rainerscript.c +++ b/grammar/rainerscript.c @@ -2,7 +2,7 @@ * * Module begun 2011-07-01 by Rainer Gerhards * - * Copyright 2011 Rainer Gerhards and Adiscon GmbH. + * Copyright 2011-2012 Rainer Gerhards and Adiscon GmbH. * * This file is part of the rsyslog runtime library. * @@ -37,16 +37,53 @@ #include <libestr.h> #include "rsyslog.h" #include "rainerscript.h" +#include "conf.h" #include "parserif.h" +#include "rsconf.h" #include "grammar.h" #include "queue.h" #include "srUtils.h" #include "regexp.h" #include "obj.h" +#include "modules.h" DEFobjCurrIf(obj) DEFobjCurrIf(regexp) +void cnfexprOptimize(struct cnfexpr *expr); +static void cnfstmtOptimizePRIFilt(struct cnfstmt *stmt); + +char* +getFIOPName(unsigned iFIOP) +{ + char *pRet; + switch(iFIOP) { + case FIOP_CONTAINS: + pRet = "contains"; + break; + case FIOP_ISEQUAL: + pRet = "isequal"; + break; + case FIOP_STARTSWITH: + pRet = "startswith"; + break; + case FIOP_REGEX: + pRet = "regex"; + break; + case FIOP_EREREGEX: + pRet = "ereregex"; + break; + case FIOP_ISEMPTY: + pRet = "isempty"; + break; + default: + pRet = "NOP"; + break; + } + return pRet; +} + + void readConfFile(FILE *fp, es_str_t **str) { @@ -107,7 +144,6 @@ objlstNew(struct cnfobj *o) lst->next = NULL; lst->obj = o; } -dbgprintf("AAAA: creating new objlst\n"); cnfobjPrint(o); return lst; @@ -131,6 +167,22 @@ objlstAdd(struct objlst *root, struct cnfobj *o) return root; } +/* add stmt to current script, always return root stmt pointer */ +struct cnfstmt* +scriptAddStmt(struct cnfstmt *root, struct cnfstmt *s) +{ + struct cnfstmt *l; + + if(root == NULL) { + root = s; + } else { /* find last, linear search ok, as only during config phase */ + for(l = root ; l->next != NULL ; l = l->next) + ; + l->next = s; + } + return root; +} + void objlstDestruct(struct objlst *lst) { @@ -748,116 +800,6 @@ cnfobjPrint(struct cnfobj *o) } -struct cnfactlst* -cnfactlstNew(enum cnfactType actType, struct nvlst *lst, char *actLine) -{ - struct cnfactlst *actlst; - - if((actlst = malloc(sizeof(struct cnfactlst))) != NULL) { - actlst->next = NULL; - actlst->syslines = NULL; - actlst->actType = actType; - actlst->lineno = yylineno; - actlst->cnfFile = strdup(cnfcurrfn); - if(actType == CNFACT_V2) - actlst->data.lst = lst; - else - actlst->data.legActLine = actLine; - } - return actlst; -} - -struct cnfactlst* -cnfactlstAddSysline(struct cnfactlst* actlst, char *line) -{ - struct cnfcfsyslinelst *cflst; - - if((cflst = malloc(sizeof(struct cnfcfsyslinelst))) != NULL) { - cflst->line = line; - if(actlst->syslines == NULL) { - cflst->next = NULL; - } else { - cflst->next = actlst->syslines; - } - actlst->syslines = cflst; - } - return actlst; -} - - -void -cnfactlstDestruct(struct cnfactlst *actlst) -{ - struct cnfactlst *toDel; - - while(actlst != NULL) { - toDel = actlst; - actlst = actlst->next; - free(toDel->cnfFile); - cnfcfsyslinelstDestruct(toDel->syslines); - if(toDel->actType == CNFACT_V2) - nvlstDestruct(toDel->data.lst); - else - free(toDel->data.legActLine); - free(toDel); - } - -} - -static inline struct cnfcfsyslinelst* -cnfcfsyslinelstReverse(struct cnfcfsyslinelst *lst) -{ - struct cnfcfsyslinelst *curr, *prev; - if(lst == NULL) - return NULL; - prev = NULL; - while(lst != NULL) { - curr = lst; - lst = lst->next; - curr->next = prev; - prev = curr; - } - return prev; -} - -struct cnfactlst* -cnfactlstReverse(struct cnfactlst *actlst) -{ - struct cnfactlst *curr, *prev; - - prev = NULL; - while(actlst != NULL) { - curr = actlst; - actlst = actlst->next; - curr->syslines = cnfcfsyslinelstReverse(curr->syslines); - curr->next = prev; - prev = curr; - } - return prev; -} - -void -cnfactlstPrint(struct cnfactlst *actlst) -{ - struct cnfcfsyslinelst *cflst; - - while(actlst != NULL) { - dbgprintf("aclst %p: ", actlst); - if(actlst->actType == CNFACT_V2) { - dbgprintf("V2 action type: "); - nvlstPrint(actlst->data.lst); - } else { - dbgprintf("legacy action line: '%s'\n", - actlst->data.legActLine); - } - for( cflst = actlst->syslines - ; cflst != NULL ; cflst = cflst->next) { - dbgprintf("action:cfsysline: '%s'\n", cflst->line); - } - actlst = actlst->next; - } -} - struct cnfexpr* cnfexprNew(unsigned nodetype, struct cnfexpr *l, struct cnfexpr *r) { @@ -884,15 +826,19 @@ done: * try to convert it to one. The semantics from es_str2num() * are used (bSuccess tells if the conversion went well or not). */ -static inline long long +static long long var2Number(struct var *r, int *bSuccess) { long long n; if(r->datatype == 'S') { n = es_str2num(r->d.estr, bSuccess); } else { - n = r->d.n; - if(bSuccess) + if(r->datatype == 'J') { + n = (r->d.json == NULL) ? 0 : json_object_get_int(r->d.json); + } else { + n = r->d.n; + } + if(bSuccess != NULL) *bSuccess = 1; } return n; @@ -903,12 +849,88 @@ var2Number(struct var *r, int *bSuccess) static inline es_str_t * var2String(struct var *r, int *bMustFree) { + es_str_t *estr; + char *cstr; + rs_size_t lenstr; if(r->datatype == 'N') { *bMustFree = 1; - return es_newStrFromNumber(r->d.n); + estr = es_newStrFromNumber(r->d.n); + } else if(r->datatype == 'J') { + *bMustFree = 1; + if(r->d.json == NULL) { + cstr = "", + lenstr = 0; + } else { + cstr = (char*)json_object_get_string(r->d.json); + lenstr = strlen(cstr); + } + estr = es_newStrFromCStr(cstr, lenstr); + } else { + *bMustFree = 0; + estr = r->d.estr; } - *bMustFree = 0; - return r->d.estr; + return estr; +} + +static uchar* +var2CString(struct var *r, int *bMustFree) +{ + uchar *cstr; + es_str_t *estr; + estr = var2String(r, bMustFree); + cstr = (uchar*) es_str2cstr(estr, NULL); + if(*bMustFree) + es_deleteStr(estr); + *bMustFree = 1; + return cstr; +} + +rsRetVal +doExtractField(uchar *str, uchar delim, int matchnbr, uchar **resstr) +{ + int iCurrFld; + int iLen; + uchar *pBuf; + uchar *pFld; + uchar *pFldEnd; + DEFiRet; + + /* first, skip to the field in question */ + iCurrFld = 1; + pFld = str; + while(*pFld && iCurrFld < matchnbr) { + /* skip fields until the requested field or end of string is found */ + while(*pFld && (uchar) *pFld != delim) + ++pFld; /* skip to field terminator */ + if(*pFld == delim) { + ++pFld; /* eat it */ + ++iCurrFld; + } + } + dbgprintf("field() field requested %d, field found %d\n", matchnbr, iCurrFld); + + if(iCurrFld == matchnbr) { + /* field found, now extract it */ + /* first of all, we need to find the end */ + pFldEnd = pFld; + while(*pFldEnd && *pFldEnd != delim) + ++pFldEnd; + --pFldEnd; /* we are already at the delimiter - so we need to + * step back a little not to copy it as part of the field. */ + /* we got our end pointer, now do the copy */ + iLen = pFldEnd - pFld + 1; /* the +1 is for an actual char, NOT \0! */ + CHKmalloc(pBuf = MALLOC((iLen + 1) * sizeof(char))); + /* now copy */ + memcpy(pBuf, pFld, iLen); + pBuf[iLen] = '\0'; /* terminate it */ + if(*(pFldEnd+1) != '\0') + ++pFldEnd; /* OK, skip again over delimiter char */ + *resstr = pBuf; + } else { + ABORT_FINALIZE(RS_RET_FIELD_NOT_FOUND); + } +finalize_it: + RETiRet; } /* Perform a function call. This has been moved out of cnfExprEval in order @@ -922,8 +944,13 @@ doFuncCall(struct cnffunc *func, struct var *ret, void* usrptr) int bMustFree; es_str_t *estr; char *str; + uchar *resStr; int retval; struct var r[CNFFUNC_MAX_ARGS]; + int delim; + int matchnbr; + struct funcData_prifilt *pPrifilt; + rsRetVal localRet; dbgprintf("rainerscript: executing function id %d\n", func->fID); switch(func->fID) { @@ -991,8 +1018,7 @@ doFuncCall(struct cnffunc *func, struct var *ret, void* usrptr) break; case CNFFUNC_RE_MATCH: cnfexprEval(func->expr[0], &r[0], usrptr); - estr = var2String(&r[0], &bMustFree); - str = es_str2cstr(estr, NULL); + str = (char*) var2CString(&r[0], &bMustFree); retval = regexp.regexec(func->funcdata, str, 0, NULL, 0); if(retval == 0) ret->d.n = 1; @@ -1003,10 +1029,44 @@ doFuncCall(struct cnffunc *func, struct var *ret, void* usrptr) } } ret->datatype = 'N'; - if(bMustFree) es_deleteStr(estr); + if(bMustFree) free(str); free(str); if(r[0].datatype == 'S') es_deleteStr(r[0].d.estr); break; + case CNFFUNC_FIELD: + cnfexprEval(func->expr[0], &r[0], usrptr); + cnfexprEval(func->expr[1], &r[1], usrptr); + cnfexprEval(func->expr[2], &r[2], usrptr); + str = (char*) var2CString(&r[0], &bMustFree); + delim = var2Number(&r[1], NULL); + matchnbr = var2Number(&r[2], NULL); + localRet = doExtractField((uchar*)str, (char) delim, matchnbr, &resStr); + if(localRet == RS_RET_OK) { + ret->d.estr = es_newStrFromCStr((char*)resStr, strlen((char*)resStr)); + free(resStr); + } else if(localRet == RS_RET_OK) { + ret->d.estr = es_newStrFromCStr("***FIELD NOT FOUND***", + sizeof("***FIELD NOT FOUND***")-1); + } else { + ret->d.estr = es_newStrFromCStr("***ERROR in field() FUNCTION***", + sizeof("***ERROR in field() FUNCTION***")-1); + } + ret->datatype = 'S'; + if(bMustFree) free(str); + if(r[0].datatype == 'S') es_deleteStr(r[0].d.estr); + if(r[1].datatype == 'S') es_deleteStr(r[1].d.estr); + if(r[2].datatype == 'S') es_deleteStr(r[2].d.estr); + break; + case CNFFUNC_PRIFILT: + pPrifilt = (struct funcData_prifilt*) func->funcdata; + if( (pPrifilt->pmask[((msg_t*)usrptr)->iFacility] == TABLE_NOPRI) || + ((pPrifilt->pmask[((msg_t*)usrptr)->iFacility] + & (1<<((msg_t*)usrptr)->iSeverity)) == 0) ) + ret->d.n = 0; + else + ret->d.n = 1; + ret->datatype = 'N'; + break; default: if(Debug) { fname = es_str2cstr(func->fname, NULL); @@ -1019,6 +1079,27 @@ doFuncCall(struct cnffunc *func, struct var *ret, void* usrptr) } } +static inline void +evalVar(struct cnfvar *var, void *usrptr, struct var *ret) +{ + rsRetVal localRet; + es_str_t *estr; + struct json_object *json; + + if(var->name[0] == '$' && var->name[1] == '!') { + /* TODO: unify string libs */ + estr = es_newStrFromBuf(var->name+1, strlen(var->name)-1); + localRet = msgGetCEEPropJSON((msg_t*)usrptr, estr, &json); + es_deleteStr(estr); + ret->datatype = 'J'; + ret->d.json = (localRet == RS_RET_OK) ? json : NULL; + } else { + ret->datatype = 'S'; + ret->d.estr = cnfGetVar(var->name, usrptr); + } + +} + #define FREE_BOTH_RET \ if(r.datatype == 'S') es_deleteStr(r.d.estr); \ if(l.datatype == 'S') es_deleteStr(l.d.estr) @@ -1033,13 +1114,19 @@ doFuncCall(struct cnffunc *func, struct var *ret, void* usrptr) #define PREP_TWO_STRINGS \ cnfexprEval(expr->l, &l, usrptr); \ estr_l = var2String(&l, &bMustFree2); \ - cnfexprEval(expr->r, &r, usrptr); \ - estr_r = var2String(&r, &bMustFree) + if(expr->r->nodetype == 'S') { \ + estr_r = ((struct cnfstringval*)expr->r)->estr;\ + bMustFree = 0; \ + } else { \ + cnfexprEval(expr->r, &r, usrptr); \ + estr_r = var2String(&r, &bMustFree); \ + } #define FREE_TWO_STRINGS \ if(bMustFree) es_deleteStr(estr_r); \ + if(expr->r->nodetype != 'S' && r.datatype == 'S') es_deleteStr(r.d.estr); \ if(bMustFree2) es_deleteStr(estr_l); \ - FREE_BOTH_RET + if(l.datatype == 'S') es_deleteStr(l.d.estr) /* evaluate an expression. * Note that we try to avoid malloc whenever possible (because of @@ -1066,23 +1153,32 @@ cnfexprEval(struct cnfexpr *expr, struct var *ret, void* usrptr) * places flagged with "CMP" need to be changed. */ case CMP_EQ: + /* this is optimized in regard to right param as a PoC for all compOps + * So this is a NOT yet the copy template! + */ cnfexprEval(expr->l, &l, usrptr); - cnfexprEval(expr->r, &r, usrptr); ret->datatype = 'N'; if(l.datatype == 'S') { - if(r.datatype == 'S') { - ret->d.n = !es_strcmp(l.d.estr, r.d.estr); /*CMP*/ + if(expr->r->nodetype == 'S') { + ret->d.n = !es_strcmp(l.d.estr, ((struct cnfstringval*)expr->r)->estr); /*CMP*/ } else { - n_l = var2Number(&l, &convok_l); - if(convok_l) { - ret->d.n = (n_l == r.d.n); /*CMP*/ + cnfexprEval(expr->r, &r, usrptr); + if(r.datatype == 'S') { + ret->d.n = !es_strcmp(l.d.estr, r.d.estr); /*CMP*/ } else { - estr_r = var2String(&r, &bMustFree); - ret->d.n = !es_strcmp(l.d.estr, estr_r); /*CMP*/ - if(bMustFree) es_deleteStr(estr_r); + n_l = var2Number(&l, &convok_l); + if(convok_l) { + ret->d.n = (n_l == r.d.n); /*CMP*/ + } else { + estr_r = var2String(&r, &bMustFree); + ret->d.n = !es_strcmp(l.d.estr, estr_r); /*CMP*/ + if(bMustFree) es_deleteStr(estr_r); + } } + if(r.datatype == 'S') es_deleteStr(r.d.estr); } } else { + cnfexprEval(expr->r, &r, usrptr); if(r.datatype == 'S') { n_r = var2Number(&r, &convok_r); if(convok_r) { @@ -1095,8 +1191,9 @@ cnfexprEval(struct cnfexpr *expr, struct var *ret, void* usrptr) } else { ret->d.n = (l.d.n == r.d.n); /*CMP*/ } + if(r.datatype == 'S') es_deleteStr(r.d.estr); } - FREE_BOTH_RET; + if(l.datatype == 'S') es_deleteStr(l.d.estr); break; case CMP_NE: cnfexprEval(expr->l, &l, usrptr); @@ -1332,8 +1429,15 @@ cnfexprEval(struct cnfexpr *expr, struct var *ret, void* usrptr) ret->d.estr = es_strdup(((struct cnfstringval*)expr)->estr); break; case 'V': + evalVar((struct cnfvar*)expr, usrptr, ret); + break; + case '&': + /* TODO: think about optimization, should be possible ;) */ + PREP_TWO_STRINGS; ret->datatype = 'S'; - ret->d.estr = cnfGetVar(((struct cnfvar*)expr)->name, usrptr); + ret->d.estr = es_strdup(estr_l); + es_addStr(&ret->d.estr, estr_r); + FREE_TWO_STRINGS; break; case '+': COMP_NUM_BINOP(+); @@ -1381,12 +1485,13 @@ cnffuncDestruct(struct cnffunc *func) /* some functions require special destruction */ switch(func->fID) { case CNFFUNC_RE_MATCH: - regexp.regfree(func->funcdata); - free(func->funcdata); - free(func->fname); + if(func->funcdata != NULL) + regexp.regfree(func->funcdata); break; default:break; } + free(func->funcdata); + free(func->fname); } /* Destruct an expression and all sub-expressions contained in it. @@ -1409,6 +1514,7 @@ cnfexprDestruct(struct cnfexpr *expr) case CMP_CONTAINSI: case OR: case AND: + case '&': case '+': case '-': case '*': @@ -1460,6 +1566,22 @@ doIndent(int indent) for(i = 0 ; i < indent ; ++i) dbgprintf(" "); } + +static void +pmaskPrint(uchar *pmask, int indent) +{ + int i; + doIndent(indent); + dbgprintf("pmask: "); + for (i = 0; i <= LOG_NFACILITIES; i++) + if (pmask[i] == TABLE_NOPRI) + dbgprintf(" X "); + else + dbgprintf("%2X ", pmask[i]); + dbgprintf("\n"); +} + + void cnfexprPrint(struct cnfexpr *expr, int indent) { @@ -1563,10 +1685,16 @@ cnfexprPrint(struct cnfexpr *expr, int indent) func = (struct cnffunc*) expr; cstrPrint("function '", func->fname); dbgprintf("' (id:%d, params:%hu)\n", func->fID, func->nParams); + if(func->fID == CNFFUNC_PRIFILT) { + struct funcData_prifilt *pD; + pD = (struct funcData_prifilt*) func->funcdata; + pmaskPrint(pD->pmask, indent+1); + } for(i = 0 ; i < func->nParams ; ++i) { cnfexprPrint(func->expr[i], indent+1); } break; + case '&': case '+': case '-': case '*': @@ -1580,11 +1708,88 @@ cnfexprPrint(struct cnfexpr *expr, int indent) cnfexprPrint(expr->r, indent+1); break; default: - dbgprintf("error: unknown nodetype %u\n", - (unsigned) expr->nodetype); + dbgprintf("error: unknown nodetype %u['%c']\n", + (unsigned) expr->nodetype, (char) expr->nodetype); break; } } +void +cnfstmtPrint(struct cnfstmt *root, int indent) +{ + struct cnfstmt *stmt; + //dbgprintf("stmt %p, indent %d, type '%c'\n", expr, indent, expr->nodetype); + for(stmt = root ; stmt != NULL ; stmt = stmt->next) { + switch(stmt->nodetype) { + case S_NOP: + doIndent(indent); dbgprintf("NOP\n"); + break; + case S_STOP: + doIndent(indent); dbgprintf("STOP\n"); + break; + case S_ACT: + doIndent(indent); dbgprintf("ACTION %p [%s]\n", stmt->d.act, stmt->printable); + break; + case S_IF: + doIndent(indent); dbgprintf("IF\n"); + cnfexprPrint(stmt->d.s_if.expr, indent+1); + doIndent(indent); dbgprintf("THEN\n"); + cnfstmtPrint(stmt->d.s_if.t_then, indent+1); + if(stmt->d.s_if.t_else != NULL) { + doIndent(indent); dbgprintf("ELSE\n"); + cnfstmtPrint(stmt->d.s_if.t_else, indent+1); + } + doIndent(indent); dbgprintf("END IF\n"); + break; + case S_SET: + doIndent(indent); dbgprintf("SET %s =\n", + stmt->d.s_set.varname); + cnfexprPrint(stmt->d.s_set.expr, indent+1); + doIndent(indent); dbgprintf("END SET\n"); + break; + case S_UNSET: + doIndent(indent); dbgprintf("UNSET %s\n", + stmt->d.s_unset.varname); + break; + case S_PRIFILT: + doIndent(indent); dbgprintf("PRIFILT '%s'\n", stmt->printable); + pmaskPrint(stmt->d.s_prifilt.pmask, indent); + cnfstmtPrint(stmt->d.s_prifilt.t_then, indent+1); + if(stmt->d.s_prifilt.t_else != NULL) { + doIndent(indent); dbgprintf("ELSE\n"); + cnfstmtPrint(stmt->d.s_prifilt.t_else, indent+1); + } + doIndent(indent); dbgprintf("END PRIFILT\n"); + break; + case S_PROPFILT: + doIndent(indent); dbgprintf("PROPFILT\n"); + doIndent(indent); dbgprintf("\tProperty.: '%s'\n", + propIDToName(stmt->d.s_propfilt.propID)); + if(stmt->d.s_propfilt.propName != NULL) { + char *cstr; + cstr = es_str2cstr(stmt->d.s_propfilt.propName, NULL); + doIndent(indent); + dbgprintf("\tCEE-Prop.: '%s'\n", cstr); + free(cstr); + } + doIndent(indent); dbgprintf("\tOperation: "); + if(stmt->d.s_propfilt.isNegated) + dbgprintf("NOT "); + dbgprintf("'%s'\n", getFIOPName(stmt->d.s_propfilt.operation)); + if(stmt->d.s_propfilt.pCSCompValue != NULL) { + doIndent(indent); dbgprintf("\tValue....: '%s'\n", + rsCStrGetSzStrNoNULL(stmt->d.s_propfilt.pCSCompValue)); + } + doIndent(indent); dbgprintf("THEN\n"); + cnfstmtPrint(stmt->d.s_propfilt.t_then, indent+1); + doIndent(indent); dbgprintf("END PROPFILT\n"); + break; + default: + dbgprintf("error: unknown stmt type %u\n", + (unsigned) stmt->nodetype); + break; + } + } +} struct cnfnumval* cnfnumvalNew(long long val) @@ -1608,6 +1813,7 @@ cnfstringvalNew(es_str_t *estr) return strval; } + struct cnfvar* cnfvarNew(char *name) { @@ -1619,63 +1825,471 @@ cnfvarNew(char *name) return var; } -struct cnfrule * -cnfruleNew(enum cnfFiltType filttype, struct cnfactlst *actlst) +struct cnfstmt * +cnfstmtNew(unsigned s_type) { - struct cnfrule* cnfrule; - if((cnfrule = malloc(sizeof(struct cnfrule))) != NULL) { - cnfrule->nodetype = 'R'; - cnfrule->filttype = filttype; - cnfrule->actlst = cnfactlstReverse(actlst); + struct cnfstmt* cnfstmt; + if((cnfstmt = malloc(sizeof(struct cnfstmt))) != NULL) { + cnfstmt->nodetype = s_type; + cnfstmt->printable = NULL; + cnfstmt->next = NULL; } - return cnfrule; + return cnfstmt; } void -cnfrulePrint(struct cnfrule *rule) +cnfstmtDestruct(struct cnfstmt *root) +{ + struct cnfstmt *stmt, *todel; + for(stmt = root ; stmt != NULL ; ) { + switch(stmt->nodetype) { + case S_NOP: + case S_STOP: + break; + case S_ACT: + actionDestruct(stmt->d.act); + break; + case S_IF: + cnfexprDestruct(stmt->d.s_if.expr); + if(stmt->d.s_if.t_then != NULL) { + cnfstmtDestruct(stmt->d.s_if.t_then); + } + if(stmt->d.s_if.t_else != NULL) { + cnfstmtDestruct(stmt->d.s_if.t_else); + } + break; + case S_SET: + free(stmt->d.s_set.varname); + cnfexprDestruct(stmt->d.s_set.expr); + break; + case S_UNSET: + free(stmt->d.s_set.varname); + break; + case S_PRIFILT: + cnfstmtDestruct(stmt->d.s_prifilt.t_then); + cnfstmtDestruct(stmt->d.s_prifilt.t_else); + break; + case S_PROPFILT: + if(stmt->d.s_propfilt.propName != NULL) + es_deleteStr(stmt->d.s_propfilt.propName); + if(stmt->d.s_propfilt.regex_cache != NULL) + rsCStrRegexDestruct(&stmt->d.s_propfilt.regex_cache); + if(stmt->d.s_propfilt.pCSCompValue != NULL) + cstrDestruct(&stmt->d.s_propfilt.pCSCompValue); + cnfstmtDestruct(stmt->d.s_propfilt.t_then); + break; + default: + dbgprintf("error: unknown stmt type during destruct %u\n", + (unsigned) stmt->nodetype); + break; + } + free(stmt->printable); + todel = stmt; + stmt = stmt->next; + free(todel); + } +} + +struct cnfstmt * +cnfstmtNewSet(char *var, struct cnfexpr *expr) +{ + struct cnfstmt* cnfstmt; + if((cnfstmt = cnfstmtNew(S_SET)) != NULL) { + cnfstmt->d.s_set.varname = (uchar*) var; + cnfstmt->d.s_set.expr = expr; + } + return cnfstmt; +} + +struct cnfstmt * +cnfstmtNewUnset(char *var) +{ + struct cnfstmt* cnfstmt; + if((cnfstmt = cnfstmtNew(S_UNSET)) != NULL) { + cnfstmt->d.s_unset.varname = (uchar*) var; + } + return cnfstmt; +} + +struct cnfstmt * +cnfstmtNewPRIFILT(char *prifilt, struct cnfstmt *t_then) +{ + struct cnfstmt* cnfstmt; + if((cnfstmt = cnfstmtNew(S_PRIFILT)) != NULL) { + cnfstmt->printable = (uchar*)prifilt; + cnfstmt->d.s_prifilt.t_then = t_then; + cnfstmt->d.s_prifilt.t_else = NULL; + DecodePRIFilter((uchar*)prifilt, cnfstmt->d.s_prifilt.pmask); + } + return cnfstmt; +} + +struct cnfstmt * +cnfstmtNewPROPFILT(char *propfilt, struct cnfstmt *t_then) +{ + struct cnfstmt* cnfstmt; + rsRetVal lRet; + if((cnfstmt = cnfstmtNew(S_PROPFILT)) != NULL) { + cnfstmt->printable = (uchar*)propfilt; + cnfstmt->d.s_propfilt.t_then = t_then; + cnfstmt->d.s_propfilt.propName = NULL; + cnfstmt->d.s_propfilt.regex_cache = NULL; + cnfstmt->d.s_propfilt.pCSCompValue = NULL; + lRet = DecodePropFilter((uchar*)propfilt, cnfstmt); + } + return cnfstmt; +} + +struct cnfstmt * +cnfstmtNewAct(struct nvlst *lst) { - dbgprintf("------ start rule %p:\n", rule); - dbgprintf("%s: ", cnfFiltType2str(rule->filttype)); - switch(rule->filttype) { - case CNFFILT_NONE: + struct cnfstmt* cnfstmt; + char namebuf[256]; + if((cnfstmt = cnfstmtNew(S_ACT)) == NULL) + goto done; + if(actionNewInst(lst, &cnfstmt->d.act) != RS_RET_OK) { + // TODO:RS_RET_WARN? + parser_errmsg("errors occured in file '%s' around line %d", + cnfcurrfn, yylineno); + cnfstmt->nodetype = S_NOP; /* disable action! */ + goto done; + } + snprintf(namebuf, sizeof(namebuf)-1, "action(type=\"%s\" ...)", + modGetName(cnfstmt->d.act->pMod)); + namebuf[255] = '\0'; /* be on safe side */ + cnfstmt->printable = (uchar*)strdup(namebuf); +done: return cnfstmt; +} + +struct cnfstmt * +cnfstmtNewLegaAct(char *actline) +{ + struct cnfstmt* cnfstmt; + rsRetVal localRet; + if((cnfstmt = cnfstmtNew(S_ACT)) == NULL) + goto done; + cnfstmt->printable = (uchar*)strdup((char*)actline); + localRet = cflineDoAction(loadConf, (uchar**)&actline, &cnfstmt->d.act); + if(localRet != RS_RET_OK && localRet != RS_RET_OK_WARN) { + parser_errmsg("%s occured in file '%s' around line %d", + (localRet == RS_RET_OK_WARN) ? "warnings" : "errors", + cnfcurrfn, yylineno); + if(localRet != RS_RET_OK_WARN) { + cnfstmt->nodetype = S_NOP; /* disable action! */ + goto done; + } + } +done: return cnfstmt; +} + + +/* returns 1 if the two expressions are constants, 0 otherwise + * if both are constants, the expression subtrees are destructed + * (this is an aid for constant folding optimizing) + */ +static int +getConstNumber(struct cnfexpr *expr, long long *l, long long *r) +{ + int ret = 0; + cnfexprOptimize(expr->l); + cnfexprOptimize(expr->r); + if(expr->l->nodetype == 'N') { + if(expr->r->nodetype == 'N') { + ret = 1; + *l = ((struct cnfnumval*)expr->l)->val; + *r = ((struct cnfnumval*)expr->r)->val; + cnfexprDestruct(expr->l); + cnfexprDestruct(expr->r); + } else if(expr->r->nodetype == 'S') { + ret = 1; + *l = ((struct cnfnumval*)expr->l)->val; + *r = es_str2num(((struct cnfstringval*)expr->r)->estr, NULL); + cnfexprDestruct(expr->l); + cnfexprDestruct(expr->r); + } + } else if(expr->l->nodetype == 'S') { + if(expr->r->nodetype == 'N') { + ret = 1; + *l = es_str2num(((struct cnfstringval*)expr->l)->estr, NULL); + *r = ((struct cnfnumval*)expr->r)->val; + cnfexprDestruct(expr->l); + cnfexprDestruct(expr->r); + } else if(expr->r->nodetype == 'S') { + ret = 1; + *l = es_str2num(((struct cnfstringval*)expr->l)->estr, NULL); + *r = es_str2num(((struct cnfstringval*)expr->r)->estr, NULL); + cnfexprDestruct(expr->l); + cnfexprDestruct(expr->r); + } + } + return ret; +} + + +/* constant folding for string concatenation */ +static inline void +constFoldConcat(struct cnfexpr *expr) +{ + es_str_t *estr; + cnfexprOptimize(expr->l); + cnfexprOptimize(expr->r); + if(expr->l->nodetype == 'S') { + if(expr->r->nodetype == 'S') { + estr = ((struct cnfstringval*)expr->l)->estr; + ((struct cnfstringval*)expr->l)->estr = NULL; + es_addStr(&estr, ((struct cnfstringval*)expr->r)->estr); + cnfexprDestruct(expr->l); + cnfexprDestruct(expr->r); + expr->nodetype = 'S'; + ((struct cnfstringval*)expr)->estr = estr; + } else if(expr->r->nodetype == 'N') { + es_str_t *numstr; + estr = ((struct cnfstringval*)expr->l)->estr; + ((struct cnfstringval*)expr->l)->estr = NULL; + numstr = es_newStrFromNumber(((struct cnfnumval*)expr->r)->val); + es_addStr(&estr, numstr); + es_deleteStr(numstr); + cnfexprDestruct(expr->l); + cnfexprDestruct(expr->r); + expr->nodetype = 'S'; + ((struct cnfstringval*)expr)->estr = estr; + } + } else if(expr->l->nodetype == 'N') { + if(expr->r->nodetype == 'S') { + estr = es_newStrFromNumber(((struct cnfnumval*)expr->l)->val); + es_addStr(&estr, ((struct cnfstringval*)expr->r)->estr); + cnfexprDestruct(expr->l); + cnfexprDestruct(expr->r); + expr->nodetype = 'S'; + ((struct cnfstringval*)expr)->estr = estr; + } else if(expr->r->nodetype == 'S') { + es_str_t *numstr; + estr = es_newStrFromNumber(((struct cnfnumval*)expr->l)->val); + numstr = es_newStrFromNumber(((struct cnfnumval*)expr->r)->val); + es_addStr(&estr, numstr); + es_deleteStr(numstr); + cnfexprDestruct(expr->l); + cnfexprDestruct(expr->r); + expr->nodetype = 'S'; + ((struct cnfstringval*)expr)->estr = estr; + } + } +} + + +/* (recursively) optimize an expression */ +void +cnfexprOptimize(struct cnfexpr *expr) +{ + long long ln, rn; + + dbgprintf("optimize expr %p, type '%c'(%u)\n", expr, expr->nodetype, expr->nodetype); + switch(expr->nodetype) { + case '&': + constFoldConcat(expr); break; - case CNFFILT_PRI: - case CNFFILT_PROP: - dbgprintf("%s\n", rule->filt.s); + case '+': + if(getConstNumber(expr, &ln, &rn)) { + expr->nodetype = 'N'; + ((struct cnfnumval*)expr)->val = ln + rn; + } break; - case CNFFILT_SCRIPT: - dbgprintf("\n"); - cnfexprPrint(rule->filt.expr, 0); + case '-': + if(getConstNumber(expr, &ln, &rn)) { + expr->nodetype = 'N'; + ((struct cnfnumval*)expr)->val = ln - rn; + } + break; + case '*': + if(getConstNumber(expr, &ln, &rn)) { + expr->nodetype = 'N'; + ((struct cnfnumval*)expr)->val = ln * rn; + } + break; + case '/': + if(getConstNumber(expr, &ln, &rn)) { + expr->nodetype = 'N'; + ((struct cnfnumval*)expr)->val = ln / rn; + } + break; + case '%': + if(getConstNumber(expr, &ln, &rn)) { + expr->nodetype = 'N'; + ((struct cnfnumval*)expr)->val = ln % rn; + } + break; + default:/* nodetype we cannot optimize */ break; } - cnfactlstPrint(rule->actlst); - dbgprintf("------ end rule %p\n", rule); + } -/* note: the sysline itself was already freed during processing - * and as such MUST NOT be freed again! +/* removes NOPs from a statement list and returns the + * first non-NOP entry. */ -void -cnfcfsyslinelstDestruct(struct cnfcfsyslinelst *cfslst) +static inline struct cnfstmt * +removeNOPs(struct cnfstmt *root) { - struct cnfcfsyslinelst *toDel; - while(cfslst != NULL) { - toDel = cfslst; - cfslst = cfslst->next; - free(toDel); + struct cnfstmt *stmt, *toDel, *prevstmt = NULL; + struct cnfstmt *newRoot = NULL; + + if(root == NULL) goto done; + stmt = root; + while(stmt != NULL) { + if(stmt->nodetype == S_NOP) { + if(prevstmt != NULL) + /* end chain, is rebuild if more non-NOPs follow */ + prevstmt->next = NULL; + toDel = stmt; + stmt = stmt->next; + cnfstmtDestruct(toDel); + } else { + if(newRoot == NULL) + newRoot = stmt; + if(prevstmt != NULL) + prevstmt->next = stmt; + prevstmt = stmt; + stmt = stmt->next; + } } +done: return newRoot; } -void -cnfruleDestruct(struct cnfrule *rule) + +static inline void +cnfstmtOptimizeIf(struct cnfstmt *stmt) { - if( rule->filttype == CNFFILT_PRI - || rule->filttype == CNFFILT_PROP) - free(rule->filt.s); - cnfactlstDestruct(rule->actlst); - free(rule); + struct cnfstmt *t_then, *t_else; + struct cnfexpr *expr; + struct cnffunc *func; + struct funcData_prifilt *prifilt; + + expr = stmt->d.s_if.expr; + cnfexprOptimize(expr); + stmt->d.s_if.t_then = removeNOPs(stmt->d.s_if.t_then); + stmt->d.s_if.t_else = removeNOPs(stmt->d.s_if.t_else); + cnfstmtOptimize(stmt->d.s_if.t_then); + cnfstmtOptimize(stmt->d.s_if.t_else); + + if(stmt->d.s_if.expr->nodetype == 'F') { + func = (struct cnffunc*)expr; + if(func->fID == CNFFUNC_PRIFILT) { + DBGPRINTF("optimizer: change IF to PRIFILT\n"); + t_then = stmt->d.s_if.t_then; + t_else = stmt->d.s_if.t_else; + stmt->nodetype = S_PRIFILT; + prifilt = (struct funcData_prifilt*) func->funcdata; + memcpy(stmt->d.s_prifilt.pmask, prifilt->pmask, + sizeof(prifilt->pmask)); + stmt->d.s_prifilt.t_then = t_then; + stmt->d.s_prifilt.t_else = t_else; + stmt->printable = (uchar*) + es_str2cstr(((struct cnfstringval*)func->expr[0])->estr, NULL); + cnfexprDestruct(expr); + cnfstmtOptimizePRIFilt(stmt); + } + } +} + +static inline void +cnfstmtOptimizeAct(struct cnfstmt *stmt) +{ + action_t *pAct; + + pAct = stmt->d.act; + if(!strcmp((char*)modGetName(stmt->d.act->pMod), "builtin:omdiscard")) { + DBGPRINTF("optimizer: replacing omdiscard by STOP\n"); + actionDestruct(stmt->d.act); + stmt->nodetype = S_STOP; + } +} + +static void +cnfstmtOptimizePRIFilt(struct cnfstmt *stmt) +{ + int i; + int isAlways = 1; + struct cnfstmt *subroot, *last; + + stmt->d.s_prifilt.t_then = removeNOPs(stmt->d.s_prifilt.t_then); + cnfstmtOptimize(stmt->d.s_prifilt.t_then); + + for(i = 0; i <= LOG_NFACILITIES; i++) + if(stmt->d.s_prifilt.pmask[i] != 0xff) { + isAlways = 0; + break; + } + if(!isAlways) + goto done; + + DBGPRINTF("optimizer: removing always-true PRIFILT %p\n", stmt); + if(stmt->d.s_prifilt.t_else != NULL) { + parser_errmsg("error: always-true PRI filter has else part!\n"); + cnfstmtDestruct(stmt->d.s_prifilt.t_else); + } + free(stmt->printable); + stmt->printable = NULL; + subroot = stmt->d.s_prifilt.t_then; + if(subroot == NULL) { + /* very strange, we set it to NOP, best we can do + * This case is NOT expected in practice + */ + stmt->nodetype = S_NOP; + goto done; + } + for(last = subroot ; last->next != NULL ; last = last->next) + /* find last node in subtree */; + last->next = stmt->next; + memcpy(stmt, subroot, sizeof(struct cnfstmt)); + free(subroot); + +done: return; +} + +/* (recursively) optimize a statement */ +void +cnfstmtOptimize(struct cnfstmt *root) +{ + struct cnfstmt *stmt; + if(root == NULL) goto done; + for(stmt = root ; stmt != NULL ; stmt = stmt->next) { +dbgprintf("RRRR: stmtOptimize: stmt %p, nodetype %u\n", stmt, stmt->nodetype); + switch(stmt->nodetype) { + case S_IF: + cnfstmtOptimizeIf(stmt); + break; + case S_PRIFILT: + cnfstmtOptimizePRIFilt(stmt); + break; + case S_PROPFILT: + stmt->d.s_propfilt.t_then = removeNOPs(stmt->d.s_propfilt.t_then); + cnfstmtOptimize(stmt->d.s_propfilt.t_then); + break; + case S_SET: + cnfexprOptimize(stmt->d.s_set.expr); + break; + case S_ACT: + cnfstmtOptimizeAct(stmt); + break; + case S_STOP: + if(stmt->next != NULL) + parser_errmsg("STOP is followed by unreachable statements!\n"); + break; + case S_UNSET: /* nothing to do */ + break; + case S_NOP: + DBGPRINTF("optimizer error: we see a NOP, how come?\n"); + break; + default: + dbgprintf("error: unknown stmt type %u during optimizer run\n", + (unsigned) stmt->nodetype); + break; + } + } +done: return; } + struct cnffparamlst * cnffparamlstNew(struct cnfexpr *expr, struct cnffparamlst *next) { @@ -1736,6 +2350,20 @@ funcName2ID(es_str_t *fname, unsigned short nParams) return CNFFUNC_INVALID; } return CNFFUNC_RE_MATCH; + } else if(!es_strbufcmp(fname, (unsigned char*)"field", sizeof("field") - 1)) { + if(nParams != 3) { + parser_errmsg("number of parameters for field() must be three " + "but is %d.", nParams); + return CNFFUNC_INVALID; + } + return CNFFUNC_FIELD; + } else if(!es_strbufcmp(fname, (unsigned char*)"prifilt", sizeof("prifilt") - 1)) { + if(nParams != 1) { + parser_errmsg("number of parameters for prifilt() must be one " + "but is %d.", nParams); + return CNFFUNC_INVALID; + } + return CNFFUNC_PRIFILT; } else { return CNFFUNC_INVALID; } @@ -1776,6 +2404,29 @@ finalize_it: RETiRet; } + +static inline rsRetVal +initFunc_prifilt(struct cnffunc *func) +{ + struct funcData_prifilt *pData; + uchar *cstr; + DEFiRet; + + func->funcdata = NULL; + if(func->expr[0]->nodetype != 'S') { + parser_errmsg("param 1 of prifilt() must be a constant string"); + FINALIZE; + } + + CHKmalloc(pData = calloc(1, sizeof(struct funcData_prifilt))); + func->funcdata = pData; + cstr = (uchar*)es_str2cstr(((struct cnfstringval*) func->expr[0])->estr, NULL); + CHKiRet(DecodePRIFilter(cstr, pData->pmask)); + free(cstr); +finalize_it: + RETiRet; +} + struct cnffunc * cnffuncNew(es_str_t *fname, struct cnffparamlst* paramlst) { @@ -1793,6 +2444,7 @@ cnffuncNew(es_str_t *fname, struct cnffparamlst* paramlst) func->nodetype = 'F'; func->fname = fname; func->nParams = nParams; + func->funcdata = NULL; func->fID = funcName2ID(fname, nParams); /* shuffle params over to array (access speed!) */ param = paramlst; @@ -1808,6 +2460,9 @@ cnffuncNew(es_str_t *fname, struct cnffparamlst* paramlst) /* need to compile the regexp in param 2, so this MUST be a constant */ initFunc_re_match(func); break; + case CNFFUNC_PRIFILT: + initFunc_prifilt(func); + break; default:break; } } @@ -1908,6 +2563,15 @@ cstrPrint(char *text, es_str_t *estr) free(str); } +char * +rmLeadingSpace(char *s) +{ + char *p; + for(p = s ; *p && isspace(*p) ; ++p) + ; + return(p); +} + /* init must be called once before any parsing of the script files start */ rsRetVal initRainerscript(void) diff --git a/grammar/rainerscript.h b/grammar/rainerscript.h index f5627597..902ff4c5 100644 --- a/grammar/rainerscript.h +++ b/grammar/rainerscript.h @@ -3,7 +3,14 @@ #include <stdio.h> #include <libestr.h> #include <typedefs.h> +#include <sys/types.h> +#include <regex.h> +//#include "stringbuf.h" +/* TODO: make this hack cleaner... we have circular definitions, so we need: */ + + +#define LOG_NFACILITIES 24 /* current number of syslog facilities */ #define CNFFUNC_MAX_ARGS 32 /**< maximum number of arguments that any function can have (among * others, this is used to size data structures). @@ -65,8 +72,9 @@ struct var { es_str_t *estr; struct cnfexpr *expr; long long n; + struct json_object *json; } d; - char datatype; /* 'N' number, 'S' string, 'E' expression */ + char datatype; /* 'N' number, 'S' string, 'E' expression, 'J' JSON */ }; struct cnfobj { @@ -91,23 +99,6 @@ struct nvlst { */ }; -struct cnfcfsyslinelst { - struct cnfcfsyslinelst *next; - char *line; -}; - -struct cnfactlst { - struct cnfactlst *next; - struct cnfcfsyslinelst *syslines; - enum cnfactType actType; - union { - struct nvlst *lst; - char *legActLine; - } data; - char *cnfFile; - int lineno; -}; - /* the following structures support expressions, and may (very much later * be the sole foundation for the AST. * @@ -118,7 +109,17 @@ struct cnfactlst { * R - rule * S - string * V - var + * ... plus the S_* #define's below: */ +#define S_STOP 4000 +#define S_PRIFILT 4001 +#define S_PROPFILT 4002 +#define S_IF 4003 +#define S_ACT 4004 +#define S_NOP 4005 /* usually used to disable some statement */ +#define S_SET 4006 +#define S_UNSET 4007 + enum cnfFiltType { CNFFILT_NONE, CNFFILT_PRI, CNFFILT_PROP, CNFFILT_SCRIPT }; static inline char* cnfFiltType2str(enum cnfFiltType filttype) @@ -137,14 +138,40 @@ cnfFiltType2str(enum cnfFiltType filttype) } -struct cnfrule { +struct cnfstmt { unsigned nodetype; - enum cnfFiltType filttype; + struct cnfstmt *next; + uchar *printable; /* printable text for debugging */ union { - char *s; - struct cnfexpr *expr; - } filt; - struct cnfactlst *actlst; + struct { + struct cnfexpr *expr; + struct cnfstmt *t_then; + struct cnfstmt *t_else; + } s_if; + struct { + uchar *varname; + struct cnfexpr *expr; + } s_set; + struct { + uchar *varname; + } s_unset; + struct { + uchar pmask[LOG_NFACILITIES+1]; /* priority mask */ + struct cnfstmt *t_then; + struct cnfstmt *t_else; + } s_prifilt; + struct { + fiop_t operation; + regex_t *regex_cache;/* cache for compiled REs, if used */ + struct cstr_s *pCSCompValue;/* value to "compare" against */ + sbool isNegated; + uintTiny propID;/* ID of the requested property */ + es_str_t *propName;/* name of property for CEE-based filters */ + struct cnfstmt *t_then; + struct cnfstmt *t_else; + } s_propfilt; + struct action_s *act; + } d; }; struct cnfexpr { @@ -182,7 +209,9 @@ enum cnffuncid { CNFFUNC_TOLOWER, CNFFUNC_CSTR, CNFFUNC_CNUM, - CNFFUNC_RE_MATCH + CNFFUNC_RE_MATCH, + CNFFUNC_FIELD, + CNFFUNC_PRIFILT }; struct cnffunc { @@ -233,6 +262,10 @@ struct cnfparamvals { /* the values we obtained for param descr. */ unsigned char bUsed; }; +struct funcData_prifilt { + uchar pmask[LOG_NFACILITIES+1]; /* priority mask */ +}; + int cnfParseBuffer(char *buf, unsigned lenBuf); void readConfFile(FILE *fp, es_str_t **str); @@ -247,11 +280,6 @@ struct nvlst* nvlstFindName(struct nvlst *lst, es_str_t *name); struct cnfobj* cnfobjNew(enum cnfobjType objType, struct nvlst *lst); void cnfobjDestruct(struct cnfobj *o); void cnfobjPrint(struct cnfobj *o); -struct cnfactlst* cnfactlstNew(enum cnfactType actType, struct nvlst *lst, char *actLine); -void cnfactlstDestruct(struct cnfactlst *actlst); -void cnfactlstPrint(struct cnfactlst *actlst); -struct cnfactlst* cnfactlstAddSysline(struct cnfactlst* actlst, char *line); -struct cnfactlst* cnfactlstReverse(struct cnfactlst *actlst); struct cnfexpr* cnfexprNew(unsigned nodetype, struct cnfexpr *l, struct cnfexpr *r); void cnfexprPrint(struct cnfexpr *expr, int indent); void cnfexprEval(struct cnfexpr *expr, struct var *ret, void *pusr); @@ -259,9 +287,6 @@ int cnfexprEvalBool(struct cnfexpr *expr, void *usrptr); void cnfexprDestruct(struct cnfexpr *expr); struct cnfnumval* cnfnumvalNew(long long val); struct cnfstringval* cnfstringvalNew(es_str_t *estr); -struct cnfrule * cnfruleNew(enum cnfFiltType filttype, struct cnfactlst *actlst); -void cnfruleDestruct(struct cnfrule *rule); -void cnfrulePrint(struct cnfrule *rule); struct cnfvar* cnfvarNew(char *name); struct cnffunc * cnffuncNew(es_str_t *fname, struct cnffparamlst* paramlst); struct cnffparamlst * cnffparamlstNew(struct cnfexpr *expr, struct cnffparamlst *next); @@ -272,7 +297,20 @@ struct cnfparamvals* nvlstGetParams(struct nvlst *lst, struct cnfparamblk *param void cnfparamsPrint(struct cnfparamblk *params, struct cnfparamvals *vals); void varDelete(struct var *v); void cnfparamvalsDestruct(struct cnfparamvals *paramvals, struct cnfparamblk *blk); -void cnfcfsyslinelstDestruct(struct cnfcfsyslinelst *cfslst); +struct cnfstmt * cnfstmtNew(unsigned s_type); +void cnfstmtPrint(struct cnfstmt *stmt, int indent); +struct cnfstmt* scriptAddStmt(struct cnfstmt *root, struct cnfstmt *s); +struct objlst* objlstAdd(struct objlst *root, struct cnfobj *o); +char *rmLeadingSpace(char *s); +struct cnfstmt * cnfstmtNewPRIFILT(char *prifilt, struct cnfstmt *t_then); +struct cnfstmt * cnfstmtNewPROPFILT(char *propfilt, struct cnfstmt *t_then); +struct cnfstmt * cnfstmtNewAct(struct nvlst *lst); +struct cnfstmt * cnfstmtNewLegaAct(char *actline); +struct cnfstmt * cnfstmtNewSet(char *var, struct cnfexpr *expr); +struct cnfstmt * cnfstmtNewUnset(char *var); +void cnfstmtDestruct(struct cnfstmt *root); +void cnfstmtOptimize(struct cnfstmt *root); +char* getFIOPName(unsigned iFIOP); rsRetVal initRainerscript(void); void unescapeStr(uchar *s, int len); |