diff --git a/sbin/pfctl/parse.y b/sbin/pfctl/parse.y index 9d95122f9826..ad25f1996d36 100644 --- a/sbin/pfctl/parse.y +++ b/sbin/pfctl/parse.y @@ -1,7313 +1,7313 @@ /* $OpenBSD: parse.y,v 1.554 2008/10/17 12:59:53 henning Exp $ */ /*- * SPDX-License-Identifier: BSD-2-Clause * * Copyright (c) 2001 Markus Friedl. All rights reserved. * Copyright (c) 2001 Daniel Hartmeier. All rights reserved. * Copyright (c) 2001 Theo de Raadt. All rights reserved. * Copyright (c) 2002,2003 Henning Brauer. All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * 1. Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * 2. Redistributions in binary form must reproduce the above copyright * notice, this list of conditions and the following disclaimer in the * documentation and/or other materials provided with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ %{ #include #define PFIOC_USE_LATEST #include #include #include #ifdef __FreeBSD__ #include #endif #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "pfctl_parser.h" #include "pfctl.h" static struct pfctl *pf = NULL; static int debug = 0; static int rulestate = 0; static u_int16_t returnicmpdefault = (ICMP_UNREACH << 8) | ICMP_UNREACH_PORT; static u_int16_t returnicmp6default = (ICMP6_DST_UNREACH << 8) | ICMP6_DST_UNREACH_NOPORT; static int blockpolicy = PFRULE_DROP; static int failpolicy = PFRULE_DROP; static int require_order = 1; static int default_statelock; static TAILQ_HEAD(files, file) files = TAILQ_HEAD_INITIALIZER(files); static struct file { TAILQ_ENTRY(file) entry; FILE *stream; char *name; int lineno; int errors; } *file; struct file *pushfile(const char *, int); int popfile(void); int check_file_secrecy(int, const char *); int yyparse(void); int yylex(void); int yyerror(const char *, ...); int kw_cmp(const void *, const void *); int lookup(char *); int lgetc(int); int lungetc(int); int findeol(void); static TAILQ_HEAD(symhead, sym) symhead = TAILQ_HEAD_INITIALIZER(symhead); struct sym { TAILQ_ENTRY(sym) entry; int used; int persist; char *nam; char *val; }; int symset(const char *, const char *, int); char *symget(const char *); int atoul(char *, u_long *); enum { PFCTL_STATE_NONE, PFCTL_STATE_OPTION, PFCTL_STATE_ETHER, PFCTL_STATE_SCRUB, PFCTL_STATE_QUEUE, PFCTL_STATE_NAT, PFCTL_STATE_FILTER }; struct node_etherproto { u_int16_t proto; struct node_etherproto *next; struct node_etherproto *tail; }; struct node_proto { u_int8_t proto; struct node_proto *next; struct node_proto *tail; }; struct node_port { u_int16_t port[2]; u_int8_t op; struct node_port *next; struct node_port *tail; }; struct node_uid { uid_t uid[2]; u_int8_t op; struct node_uid *next; struct node_uid *tail; }; struct node_gid { gid_t gid[2]; u_int8_t op; struct node_gid *next; struct node_gid *tail; }; struct node_icmp { u_int8_t code; u_int8_t type; u_int8_t proto; struct node_icmp *next; struct node_icmp *tail; }; enum { PF_STATE_OPT_MAX, PF_STATE_OPT_NOSYNC, PF_STATE_OPT_SRCTRACK, PF_STATE_OPT_MAX_SRC_STATES, PF_STATE_OPT_MAX_SRC_CONN, PF_STATE_OPT_MAX_SRC_CONN_RATE, PF_STATE_OPT_MAX_SRC_NODES, PF_STATE_OPT_OVERLOAD, PF_STATE_OPT_STATELOCK, PF_STATE_OPT_TIMEOUT, PF_STATE_OPT_SLOPPY, PF_STATE_OPT_PFLOW }; enum { PF_SRCTRACK_NONE, PF_SRCTRACK, PF_SRCTRACK_GLOBAL, PF_SRCTRACK_RULE }; struct node_state_opt { int type; union { u_int32_t max_states; u_int32_t max_src_states; u_int32_t max_src_conn; struct { u_int32_t limit; u_int32_t seconds; } max_src_conn_rate; struct { u_int8_t flush; char tblname[PF_TABLE_NAME_SIZE]; } overload; u_int32_t max_src_nodes; u_int8_t src_track; u_int32_t statelock; struct { int number; u_int32_t seconds; } timeout; } data; struct node_state_opt *next; struct node_state_opt *tail; }; struct peer { struct node_host *host; struct node_port *port; }; static struct node_queue { char queue[PF_QNAME_SIZE]; char parent[PF_QNAME_SIZE]; char ifname[IFNAMSIZ]; int scheduler; struct node_queue *next; struct node_queue *tail; } *queues = NULL; struct node_qassign { char *qname; char *pqname; }; static struct filter_opts { int marker; #define FOM_FLAGS 0x0001 #define FOM_ICMP 0x0002 #define FOM_TOS 0x0004 #define FOM_KEEP 0x0008 #define FOM_SRCTRACK 0x0010 #define FOM_MINTTL 0x0020 #define FOM_MAXMSS 0x0040 #define FOM_AFTO 0x0080 /* not yet implemmented */ #define FOM_SETTOS 0x0100 #define FOM_SCRUB_TCP 0x0200 #define FOM_SETPRIO 0x0400 #define FOM_ONCE 0x1000 /* not yet implemmented */ #define FOM_PRIO 0x2000 #define FOM_SETDELAY 0x4000 #define FOM_FRAGCACHE 0x8000 /* does not exist in OpenBSD */ struct node_uid *uid; struct node_gid *gid; struct node_if *rcv; struct { u_int8_t b1; u_int8_t b2; u_int16_t w; u_int16_t w2; } flags; struct node_icmp *icmpspec; u_int32_t tos; u_int32_t prob; u_int32_t ridentifier; struct { int action; struct node_state_opt *options; } keep; int fragment; int allowopts; char *label[PF_RULE_MAX_LABEL_COUNT]; int labelcount; struct node_qassign queues; char *tag; char *match_tag; u_int8_t match_tag_not; u_int16_t dnpipe; u_int16_t dnrpipe; u_int32_t free_flags; u_int rtableid; u_int8_t prio; u_int8_t set_prio[2]; struct { struct node_host *addr; u_int16_t port; } divert; /* new-style scrub opts */ int nodf; int minttl; int settos; int randomid; int max_mss; } filter_opts; static struct antispoof_opts { char *label[PF_RULE_MAX_LABEL_COUNT]; int labelcount; u_int32_t ridentifier; u_int rtableid; } antispoof_opts; static struct scrub_opts { int marker; int nodf; int minttl; int maxmss; int settos; int fragcache; int randomid; int reassemble_tcp; char *match_tag; u_int8_t match_tag_not; u_int rtableid; } scrub_opts; static struct queue_opts { int marker; #define QOM_BWSPEC 0x01 #define QOM_SCHEDULER 0x02 #define QOM_PRIORITY 0x04 #define QOM_TBRSIZE 0x08 #define QOM_QLIMIT 0x10 struct node_queue_bw queue_bwspec; struct node_queue_opt scheduler; int priority; unsigned int tbrsize; int qlimit; } queue_opts; static struct table_opts { int flags; int init_addr; struct node_tinithead init_nodes; } table_opts; static struct pool_opts { int marker; #define POM_TYPE 0x01 #define POM_STICKYADDRESS 0x02 #define POM_ENDPI 0x04 u_int8_t opts; int type; int staticport; struct pf_poolhashkey *key; struct pf_mape_portset mape; } pool_opts; static struct codel_opts codel_opts; static struct node_hfsc_opts hfsc_opts; static struct node_fairq_opts fairq_opts; static struct node_state_opt *keep_state_defaults = NULL; static struct pfctl_watermarks syncookie_opts; int disallow_table(struct node_host *, const char *); int disallow_urpf_failed(struct node_host *, const char *); int disallow_alias(struct node_host *, const char *); int rule_consistent(struct pfctl_rule *, int); int filter_consistent(struct pfctl_rule *, int); int nat_consistent(struct pfctl_rule *); int rdr_consistent(struct pfctl_rule *); int process_tabledef(char *, struct table_opts *); void expand_label_str(char *, size_t, const char *, const char *); void expand_label_if(const char *, char *, size_t, const char *); void expand_label_addr(const char *, char *, size_t, sa_family_t, struct pf_rule_addr *); void expand_label_port(const char *, char *, size_t, struct pf_rule_addr *); void expand_label_proto(const char *, char *, size_t, u_int8_t); void expand_label_nr(const char *, char *, size_t, struct pfctl_rule *); void expand_eth_rule(struct pfctl_eth_rule *, struct node_if *, struct node_etherproto *, struct node_mac *, struct node_mac *, struct node_host *, struct node_host *, const char *, const char *); void expand_rule(struct pfctl_rule *, struct node_if *, struct node_host *, struct node_proto *, struct node_os *, struct node_host *, struct node_port *, struct node_host *, struct node_port *, struct node_uid *, struct node_gid *, struct node_if *, struct node_icmp *, const char *); int expand_altq(struct pf_altq *, struct node_if *, struct node_queue *, struct node_queue_bw bwspec, struct node_queue_opt *); int expand_queue(struct pf_altq *, struct node_if *, struct node_queue *, struct node_queue_bw, struct node_queue_opt *); int expand_skip_interface(struct node_if *); int check_rulestate(int); int getservice(char *); int rule_label(struct pfctl_rule *, char *s[PF_RULE_MAX_LABEL_COUNT]); int eth_rule_label(struct pfctl_eth_rule *, char *s[PF_RULE_MAX_LABEL_COUNT]); int rt_tableid_max(void); void mv_rules(struct pfctl_ruleset *, struct pfctl_ruleset *); void mv_eth_rules(struct pfctl_eth_ruleset *, struct pfctl_eth_ruleset *); void decide_address_family(struct node_host *, sa_family_t *); void remove_invalid_hosts(struct node_host **, sa_family_t *); int invalid_redirect(struct node_host *, sa_family_t); u_int16_t parseicmpspec(char *, sa_family_t); int kw_casecmp(const void *, const void *); int map_tos(char *string, int *); struct node_mac* node_mac_from_string(const char *); struct node_mac* node_mac_from_string_masklen(const char *, int); struct node_mac* node_mac_from_string_mask(const char *, const char *); static TAILQ_HEAD(loadanchorshead, loadanchors) loadanchorshead = TAILQ_HEAD_INITIALIZER(loadanchorshead); struct loadanchors { TAILQ_ENTRY(loadanchors) entries; char *anchorname; char *filename; }; typedef struct { union { int64_t number; double probability; int i; char *string; u_int rtableid; struct { u_int8_t b1; u_int8_t b2; u_int16_t w; u_int16_t w2; } b; struct range { int a; int b; int t; } range; struct node_if *interface; struct node_proto *proto; struct node_etherproto *etherproto; struct node_icmp *icmp; struct node_host *host; struct node_os *os; struct node_port *port; struct node_uid *uid; struct node_gid *gid; struct node_state_opt *state_opt; struct peer peer; struct { struct peer src, dst; struct node_os *src_os; } fromto; struct { struct node_mac *src; struct node_mac *dst; } etherfromto; struct node_mac *mac; struct { struct node_mac *mac; } etheraddr; char *bridge_to; struct { struct node_host *host; u_int8_t rt; u_int8_t pool_opts; sa_family_t af; struct pf_poolhashkey *key; } route; struct redirection { struct node_host *host; struct range rport; } *redirection; struct { int action; struct node_state_opt *options; } keep_state; struct { u_int8_t log; u_int8_t logif; u_int8_t quick; } logquick; struct { int neg; char *name; } tagged; struct pf_poolhashkey *hashkey; struct node_queue *queue; struct node_queue_opt queue_options; struct node_queue_bw queue_bwspec; struct node_qassign qassign; struct filter_opts filter_opts; struct antispoof_opts antispoof_opts; struct queue_opts queue_opts; struct scrub_opts scrub_opts; struct table_opts table_opts; struct pool_opts pool_opts; struct node_hfsc_opts hfsc_opts; struct node_fairq_opts fairq_opts; struct codel_opts codel_opts; struct pfctl_watermarks *watermarks; } v; int lineno; } YYSTYPE; #define PPORT_RANGE 1 #define PPORT_STAR 2 int parseport(char *, struct range *r, int); #define DYNIF_MULTIADDR(addr) ((addr).type == PF_ADDR_DYNIFTL && \ (!((addr).iflags & PFI_AFLAG_NOALIAS) || \ !isdigit((addr).v.ifname[strlen((addr).v.ifname)-1]))) %} %token PASS BLOCK MATCH SCRUB RETURN IN OS OUT LOG QUICK ON FROM TO FLAGS %token RETURNRST RETURNICMP RETURNICMP6 PROTO INET INET6 ALL ANY ICMPTYPE %token ICMP6TYPE CODE KEEP MODULATE STATE PORT RDR NAT BINAT ARROW NODF %token MINTTL ERROR ALLOWOPTS FASTROUTE FILENAME ROUTETO DUPTO REPLYTO NO LABEL %token NOROUTE URPFFAILED FRAGMENT USER GROUP MAXMSS MAXIMUM TTL TOS DROP TABLE %token REASSEMBLE ANCHOR NATANCHOR RDRANCHOR BINATANCHOR %token SET OPTIMIZATION TIMEOUT LIMIT LOGINTERFACE BLOCKPOLICY FAILPOLICY %token RANDOMID REQUIREORDER SYNPROXY FINGERPRINTS NOSYNC DEBUG SKIP HOSTID %token ANTISPOOF FOR INCLUDE KEEPCOUNTERS SYNCOOKIES L3 MATCHES %token ETHER %token BITMASK RANDOM SOURCEHASH ROUNDROBIN STATICPORT PROBABILITY MAPEPORTSET %token ALTQ CBQ CODEL PRIQ HFSC FAIRQ BANDWIDTH TBRSIZE LINKSHARE REALTIME %token UPPERLIMIT QUEUE PRIORITY QLIMIT HOGS BUCKETS RTABLE TARGET INTERVAL %token DNPIPE DNQUEUE RIDENTIFIER %token LOAD RULESET_OPTIMIZATION PRIO %token STICKYADDRESS ENDPI MAXSRCSTATES MAXSRCNODES SOURCETRACK GLOBAL RULE %token MAXSRCCONN MAXSRCCONNRATE OVERLOAD FLUSH SLOPPY PFLOW %token TAGGED TAG IFBOUND FLOATING STATEPOLICY STATEDEFAULTS ROUTE SETTOS %token DIVERTTO DIVERTREPLY BRIDGE_TO RECEIVEDON NE LE GE %token STRING %token NUMBER %token PORTBINARY %type interface if_list if_item_not if_item %type number icmptype icmp6type uid gid %type tos not yesno optnodf %type probability %type no dir af fragcache optimizer syncookie_val %type sourcetrack flush unaryop statelock %type etherprotoval %type action nataction natpasslog scrubaction %type flags flag blockspec prio %type portplain portstar portrange %type hashkey %type proto proto_list proto_item %type protoval %type icmpspec %type icmp_list icmp_item %type icmp6_list icmp6_item %type reticmpspec reticmp6spec %type fromto l3fromto %type ipportspec from to %type ipspec toipspec xhost host dynaddr host_list %type redir_host_list redirspec %type route_host route_host_list routespec %type os xos os_list %type portspec port_list port_item %type uids uid_list uid_item %type gids gid_list gid_item %type route %type redirection redirpool %type label stringall tag anchorname %type string varstring numberstring %type keep %type state_opt_spec state_opt_list state_opt_item %type logquick quick log logopts logopt %type antispoof_ifspc antispoof_iflst antispoof_if %type qname etherqname %type qassign qassign_list qassign_item %type scheduler %type cbqflags_list cbqflags_item %type priqflags_list priqflags_item %type hfscopts_list hfscopts_item hfsc_opts %type fairqopts_list fairqopts_item fairq_opts %type codelopts_list codelopts_item codel_opts %type bandwidth %type filter_opts filter_opt filter_opts_l etherfilter_opts etherfilter_opt etherfilter_opts_l %type filter_sets filter_set filter_sets_l %type antispoof_opts antispoof_opt antispoof_opts_l %type queue_opts queue_opt queue_opts_l %type scrub_opts scrub_opt scrub_opts_l %type table_opts table_opt table_opts_l %type pool_opts pool_opt pool_opts_l %type tagged %type rtable %type syncookie_opts %type etherproto etherproto_list etherproto_item %type etherfromto %type etherfrom etherto %type bridge %type xmac mac mac_list macspec %% ruleset : /* empty */ | ruleset include '\n' | ruleset '\n' | ruleset option '\n' | ruleset etherrule '\n' | ruleset etheranchorrule '\n' | ruleset scrubrule '\n' | ruleset natrule '\n' | ruleset binatrule '\n' | ruleset pfrule '\n' | ruleset anchorrule '\n' | ruleset loadrule '\n' | ruleset altqif '\n' | ruleset queuespec '\n' | ruleset varset '\n' | ruleset antispoof '\n' | ruleset tabledef '\n' | '{' fakeanchor '}' '\n'; | ruleset error '\n' { file->errors++; } ; include : INCLUDE STRING { struct file *nfile; if ((nfile = pushfile($2, 0)) == NULL) { yyerror("failed to include file %s", $2); free($2); YYERROR; } free($2); file = nfile; lungetc('\n'); } ; /* * apply to previouslys specified rule: must be careful to note * what that is: pf or nat or binat or rdr */ fakeanchor : fakeanchor '\n' | fakeanchor anchorrule '\n' | fakeanchor binatrule '\n' | fakeanchor natrule '\n' | fakeanchor pfrule '\n' | fakeanchor error '\n' ; optimizer : string { if (!strcmp($1, "none")) $$ = 0; else if (!strcmp($1, "basic")) $$ = PF_OPTIMIZE_BASIC; else if (!strcmp($1, "profile")) $$ = PF_OPTIMIZE_BASIC | PF_OPTIMIZE_PROFILE; else { yyerror("unknown ruleset-optimization %s", $1); YYERROR; } } ; optnodf : /* empty */ { $$ = 0; } | NODF { $$ = 1; } ; option : SET REASSEMBLE yesno optnodf { if (check_rulestate(PFCTL_STATE_OPTION)) YYERROR; pfctl_set_reassembly(pf, $3, $4); } | SET OPTIMIZATION STRING { if (check_rulestate(PFCTL_STATE_OPTION)) { free($3); YYERROR; } if (pfctl_set_optimization(pf, $3) != 0) { yyerror("unknown optimization %s", $3); free($3); YYERROR; } free($3); } | SET RULESET_OPTIMIZATION optimizer { if (!(pf->opts & PF_OPT_OPTIMIZE)) { pf->opts |= PF_OPT_OPTIMIZE; pf->optimize = $3; } } | SET TIMEOUT timeout_spec | SET TIMEOUT '{' optnl timeout_list '}' | SET LIMIT limit_spec | SET LIMIT '{' optnl limit_list '}' | SET LOGINTERFACE stringall { if (check_rulestate(PFCTL_STATE_OPTION)) { free($3); YYERROR; } if (pfctl_set_logif(pf, $3) != 0) { yyerror("error setting loginterface %s", $3); free($3); YYERROR; } free($3); } | SET HOSTID number { if ($3 == 0 || $3 > UINT_MAX) { yyerror("hostid must be non-zero"); YYERROR; } if (pfctl_set_hostid(pf, $3) != 0) { yyerror("error setting hostid %08x", $3); YYERROR; } } | SET BLOCKPOLICY DROP { if (pf->opts & PF_OPT_VERBOSE) printf("set block-policy drop\n"); if (check_rulestate(PFCTL_STATE_OPTION)) YYERROR; blockpolicy = PFRULE_DROP; } | SET BLOCKPOLICY RETURN { if (pf->opts & PF_OPT_VERBOSE) printf("set block-policy return\n"); if (check_rulestate(PFCTL_STATE_OPTION)) YYERROR; blockpolicy = PFRULE_RETURN; } | SET FAILPOLICY DROP { if (pf->opts & PF_OPT_VERBOSE) printf("set fail-policy drop\n"); if (check_rulestate(PFCTL_STATE_OPTION)) YYERROR; failpolicy = PFRULE_DROP; } | SET FAILPOLICY RETURN { if (pf->opts & PF_OPT_VERBOSE) printf("set fail-policy return\n"); if (check_rulestate(PFCTL_STATE_OPTION)) YYERROR; failpolicy = PFRULE_RETURN; } | SET REQUIREORDER yesno { if (pf->opts & PF_OPT_VERBOSE) printf("set require-order %s\n", $3 == 1 ? "yes" : "no"); require_order = $3; } | SET FINGERPRINTS STRING { if (pf->opts & PF_OPT_VERBOSE) printf("set fingerprints \"%s\"\n", $3); if (check_rulestate(PFCTL_STATE_OPTION)) { free($3); YYERROR; } if (!pf->anchor->name[0]) { if (pfctl_file_fingerprints(pf->dev, pf->opts, $3)) { yyerror("error loading " "fingerprints %s", $3); free($3); YYERROR; } } free($3); } | SET STATEPOLICY statelock { if (pf->opts & PF_OPT_VERBOSE) switch ($3) { case 0: printf("set state-policy floating\n"); break; case PFRULE_IFBOUND: printf("set state-policy if-bound\n"); break; } default_statelock = $3; } | SET DEBUG STRING { if (check_rulestate(PFCTL_STATE_OPTION)) { free($3); YYERROR; } if (pfctl_do_set_debug(pf, $3) != 0) { yyerror("error setting debuglevel %s", $3); free($3); YYERROR; } free($3); } | SET SKIP interface { if (expand_skip_interface($3) != 0) { yyerror("error setting skip interface(s)"); YYERROR; } } | SET STATEDEFAULTS state_opt_list { if (keep_state_defaults != NULL) { yyerror("cannot redefine state-defaults"); YYERROR; } keep_state_defaults = $3; } | SET KEEPCOUNTERS { pf->keep_counters = true; } | SET SYNCOOKIES syncookie_val syncookie_opts { if (pfctl_cfg_syncookies(pf, $3, $4)) { yyerror("error setting syncookies"); YYERROR; } } ; syncookie_val : STRING { if (!strcmp($1, "never")) $$ = PFCTL_SYNCOOKIES_NEVER; else if (!strcmp($1, "adaptive")) $$ = PFCTL_SYNCOOKIES_ADAPTIVE; else if (!strcmp($1, "always")) $$ = PFCTL_SYNCOOKIES_ALWAYS; else { yyerror("illegal value for syncookies"); YYERROR; } } ; syncookie_opts : /* empty */ { $$ = NULL; } | { memset(&syncookie_opts, 0, sizeof(syncookie_opts)); } '(' syncookie_opt_l ')' { $$ = &syncookie_opts; } ; syncookie_opt_l : syncookie_opt_l comma syncookie_opt | syncookie_opt ; syncookie_opt : STRING STRING { double val; char *cp; val = strtod($2, &cp); if (cp == NULL || strcmp(cp, "%")) YYERROR; if (val <= 0 || val > 100) { yyerror("illegal percentage value"); YYERROR; } if (!strcmp($1, "start")) { syncookie_opts.hi = val; } else if (!strcmp($1, "end")) { syncookie_opts.lo = val; } else { yyerror("illegal syncookie option"); YYERROR; } } ; stringall : STRING { $$ = $1; } | ALL { if (($$ = strdup("all")) == NULL) { err(1, "stringall: strdup"); } } ; string : STRING string { if (asprintf(&$$, "%s %s", $1, $2) == -1) err(1, "string: asprintf"); free($1); free($2); } | STRING ; varstring : numberstring varstring { if (asprintf(&$$, "%s %s", $1, $2) == -1) err(1, "string: asprintf"); free($1); free($2); } | numberstring ; numberstring : NUMBER { char *s; if (asprintf(&s, "%lld", (long long)$1) == -1) { yyerror("string: asprintf"); YYERROR; } $$ = s; } | STRING ; varset : STRING '=' varstring { char *s = $1; if (pf->opts & PF_OPT_VERBOSE) printf("%s = \"%s\"\n", $1, $3); while (*s++) { if (isspace((unsigned char)*s)) { yyerror("macro name cannot contain " "whitespace"); YYERROR; } } if (symset($1, $3, 0) == -1) err(1, "cannot store variable %s", $1); free($1); free($3); } ; anchorname : STRING { $$ = $1; } | /* empty */ { $$ = NULL; } ; pfa_anchorlist : /* empty */ | pfa_anchorlist '\n' | pfa_anchorlist pfrule '\n' | pfa_anchorlist anchorrule '\n' ; pfa_anchor : '{' { char ta[PF_ANCHOR_NAME_SIZE]; struct pfctl_ruleset *rs; /* stepping into a brace anchor */ pf->asd++; pf->bn++; /* * Anchor contents are parsed before the anchor rule * production completes, so we don't know the real * location yet. Create a holding ruleset in the root; * contents will be moved afterwards. */ snprintf(ta, PF_ANCHOR_NAME_SIZE, "_%d", pf->bn); rs = pf_find_or_create_ruleset(ta); if (rs == NULL) err(1, "pfa_anchor: pf_find_or_create_ruleset"); pf->astack[pf->asd] = rs->anchor; pf->anchor = rs->anchor; } '\n' pfa_anchorlist '}' { pf->alast = pf->anchor; pf->asd--; pf->anchor = pf->astack[pf->asd]; } | /* empty */ ; anchorrule : ANCHOR anchorname dir quick interface af proto fromto filter_opts pfa_anchor { struct pfctl_rule r; struct node_proto *proto; if (check_rulestate(PFCTL_STATE_FILTER)) { if ($2) free($2); YYERROR; } if ($2 && ($2[0] == '_' || strstr($2, "/_") != NULL)) { free($2); yyerror("anchor names beginning with '_' " "are reserved for internal use"); YYERROR; } memset(&r, 0, sizeof(r)); if (pf->astack[pf->asd + 1]) { if ($2 && strchr($2, '/') != NULL) { free($2); yyerror("anchor paths containing '/' " "cannot be used for inline anchors."); YYERROR; } /* Move inline rules into relative location. */ pfctl_anchor_setup(&r, &pf->astack[pf->asd]->ruleset, $2 ? $2 : pf->alast->name); if (r.anchor == NULL) err(1, "anchorrule: unable to " "create ruleset"); if (pf->alast != r.anchor) { if (r.anchor->match) { yyerror("inline anchor '%s' " "already exists", r.anchor->name); YYERROR; } mv_rules(&pf->alast->ruleset, &r.anchor->ruleset); } pf_remove_if_empty_ruleset(&pf->alast->ruleset); pf->alast = r.anchor; } else { if (!$2) { yyerror("anchors without explicit " "rules must specify a name"); YYERROR; } } r.direction = $3; r.quick = $4.quick; r.af = $6; r.prob = $9.prob; r.rtableid = $9.rtableid; r.ridentifier = $9.ridentifier; if ($9.tag) if (strlcpy(r.tagname, $9.tag, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } if ($9.match_tag) if (strlcpy(r.match_tagname, $9.match_tag, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } r.match_tag_not = $9.match_tag_not; if (rule_label(&r, $9.label)) YYERROR; for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) free($9.label[i]); r.flags = $9.flags.b1; r.flagset = $9.flags.b2; if (($9.flags.b1 & $9.flags.b2) != $9.flags.b1) { yyerror("flags always false"); YYERROR; } if ($9.flags.b1 || $9.flags.b2 || $8.src_os) { for (proto = $7; proto != NULL && proto->proto != IPPROTO_TCP; proto = proto->next) ; /* nothing */ if (proto == NULL && $7 != NULL) { if ($9.flags.b1 || $9.flags.b2) yyerror( "flags only apply to tcp"); if ($8.src_os) yyerror( "OS fingerprinting only " "applies to tcp"); YYERROR; } } r.tos = $9.tos; if ($9.keep.action) { yyerror("cannot specify state handling " "on anchors"); YYERROR; } if ($9.match_tag) if (strlcpy(r.match_tagname, $9.match_tag, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } r.match_tag_not = $9.match_tag_not; if ($9.marker & FOM_PRIO) { if ($9.prio == 0) r.prio = PF_PRIO_ZERO; else r.prio = $9.prio; } if ($9.marker & FOM_SETPRIO) { r.set_prio[0] = $9.set_prio[0]; r.set_prio[1] = $9.set_prio[1]; r.scrub_flags |= PFSTATE_SETPRIO; } decide_address_family($8.src.host, &r.af); decide_address_family($8.dst.host, &r.af); expand_rule(&r, $5, NULL, $7, $8.src_os, $8.src.host, $8.src.port, $8.dst.host, $8.dst.port, $9.uid, $9.gid, $9.rcv, $9.icmpspec, pf->astack[pf->asd + 1] ? pf->alast->name : $2); free($2); pf->astack[pf->asd + 1] = NULL; } | NATANCHOR string interface af proto fromto rtable { struct pfctl_rule r; if (check_rulestate(PFCTL_STATE_NAT)) { free($2); YYERROR; } memset(&r, 0, sizeof(r)); r.action = PF_NAT; r.af = $4; r.rtableid = $7; decide_address_family($6.src.host, &r.af); decide_address_family($6.dst.host, &r.af); expand_rule(&r, $3, NULL, $5, $6.src_os, $6.src.host, $6.src.port, $6.dst.host, $6.dst.port, 0, 0, 0, 0, $2); free($2); } | RDRANCHOR string interface af proto fromto rtable { struct pfctl_rule r; if (check_rulestate(PFCTL_STATE_NAT)) { free($2); YYERROR; } memset(&r, 0, sizeof(r)); r.action = PF_RDR; r.af = $4; r.rtableid = $7; decide_address_family($6.src.host, &r.af); decide_address_family($6.dst.host, &r.af); if ($6.src.port != NULL) { yyerror("source port parameter not supported" " in rdr-anchor"); YYERROR; } if ($6.dst.port != NULL) { if ($6.dst.port->next != NULL) { yyerror("destination port list " "expansion not supported in " "rdr-anchor"); YYERROR; } else if ($6.dst.port->op != PF_OP_EQ) { yyerror("destination port operators" " not supported in rdr-anchor"); YYERROR; } r.dst.port[0] = $6.dst.port->port[0]; r.dst.port[1] = $6.dst.port->port[1]; r.dst.port_op = $6.dst.port->op; } expand_rule(&r, $3, NULL, $5, $6.src_os, $6.src.host, $6.src.port, $6.dst.host, $6.dst.port, 0, 0, 0, 0, $2); free($2); } | BINATANCHOR string interface af proto fromto rtable { struct pfctl_rule r; if (check_rulestate(PFCTL_STATE_NAT)) { free($2); YYERROR; } memset(&r, 0, sizeof(r)); r.action = PF_BINAT; r.af = $4; r.rtableid = $7; if ($5 != NULL) { if ($5->next != NULL) { yyerror("proto list expansion" " not supported in binat-anchor"); YYERROR; } r.proto = $5->proto; free($5); } if ($6.src.host != NULL || $6.src.port != NULL || $6.dst.host != NULL || $6.dst.port != NULL) { yyerror("fromto parameter not supported" " in binat-anchor"); YYERROR; } decide_address_family($6.src.host, &r.af); decide_address_family($6.dst.host, &r.af); pfctl_append_rule(pf, &r, $2); free($2); } ; loadrule : LOAD ANCHOR string FROM string { struct loadanchors *loadanchor; if (strlen(pf->anchor->name) + 1 + strlen($3) >= MAXPATHLEN) { yyerror("anchorname %s too long, max %u\n", $3, MAXPATHLEN - 1); free($3); YYERROR; } loadanchor = calloc(1, sizeof(struct loadanchors)); if (loadanchor == NULL) err(1, "loadrule: calloc"); if ((loadanchor->anchorname = malloc(MAXPATHLEN)) == NULL) err(1, "loadrule: malloc"); if (pf->anchor->name[0]) snprintf(loadanchor->anchorname, MAXPATHLEN, "%s/%s", pf->anchor->name, $3); else strlcpy(loadanchor->anchorname, $3, MAXPATHLEN); if ((loadanchor->filename = strdup($5)) == NULL) err(1, "loadrule: strdup"); TAILQ_INSERT_TAIL(&loadanchorshead, loadanchor, entries); free($3); free($5); }; scrubaction : no SCRUB { $$.b2 = $$.w = 0; if ($1) $$.b1 = PF_NOSCRUB; else $$.b1 = PF_SCRUB; } ; etherrule : ETHER action dir quick interface bridge etherproto etherfromto l3fromto etherfilter_opts { struct pfctl_eth_rule r; bzero(&r, sizeof(r)); if (check_rulestate(PFCTL_STATE_ETHER)) YYERROR; r.action = $2.b1; r.direction = $3; r.quick = $4.quick; if ($10.tag != NULL) strlcpy(r.tagname, $10.tag, sizeof(r.tagname)); if ($10.match_tag) if (strlcpy(r.match_tagname, $10.match_tag, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } r.match_tag_not = $10.match_tag_not; if ($10.queues.qname != NULL) strlcpy(r.qname, $10.queues.qname, sizeof(r.qname)); r.dnpipe = $10.dnpipe; r.dnflags = $10.free_flags; if (eth_rule_label(&r, $10.label)) YYERROR; for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) free($10.label[i]); r.ridentifier = $10.ridentifier; expand_eth_rule(&r, $5, $7, $8.src, $8.dst, $9.src.host, $9.dst.host, $6, ""); } ; etherpfa_anchorlist : /* empty */ | etherpfa_anchorlist '\n' | etherpfa_anchorlist etherrule '\n' | etherpfa_anchorlist etheranchorrule '\n' ; etherpfa_anchor : '{' { char ta[PF_ANCHOR_NAME_SIZE]; struct pfctl_eth_ruleset *rs; /* steping into a brace anchor */ pf->asd++; pf->bn++; /* create a holding ruleset in the root */ snprintf(ta, PF_ANCHOR_NAME_SIZE, "_%d", pf->bn); rs = pf_find_or_create_eth_ruleset(ta); if (rs == NULL) err(1, "etherpfa_anchor: pf_find_or_create_eth_ruleset"); pf->eastack[pf->asd] = rs->anchor; pf->eanchor = rs->anchor; } '\n' etherpfa_anchorlist '}' { pf->ealast = pf->eanchor; pf->asd--; pf->eanchor = pf->eastack[pf->asd]; } | /* empty */ ; etheranchorrule : ETHER ANCHOR anchorname dir quick interface etherproto etherfromto l3fromto etherpfa_anchor { struct pfctl_eth_rule r; if (check_rulestate(PFCTL_STATE_ETHER)) { free($3); YYERROR; } if ($3 && ($3[0] == '_' || strstr($3, "/_") != NULL)) { free($3); yyerror("anchor names beginning with '_' " "are reserved for internal use"); YYERROR; } memset(&r, 0, sizeof(r)); if (pf->eastack[pf->asd + 1]) { if ($3 && strchr($3, '/') != NULL) { free($3); yyerror("anchor paths containing '/' " "cannot be used for inline anchors."); YYERROR; } /* Move inline rules into relative location. */ pfctl_eth_anchor_setup(pf, &r, &pf->eastack[pf->asd]->ruleset, $3 ? $3 : pf->ealast->name); if (r.anchor == NULL) err(1, "etheranchorrule: unable to " "create ruleset"); if (pf->ealast != r.anchor) { if (r.anchor->match) { yyerror("inline anchor '%s' " "already exists", r.anchor->name); YYERROR; } mv_eth_rules(&pf->ealast->ruleset, &r.anchor->ruleset); } pf_remove_if_empty_eth_ruleset(&pf->ealast->ruleset); pf->ealast = r.anchor; } else { if (!$3) { yyerror("anchors without explicit " "rules must specify a name"); YYERROR; } } r.direction = $4; r.quick = $5.quick; expand_eth_rule(&r, $6, $7, $8.src, $8.dst, $9.src.host, $9.dst.host, NULL, pf->eastack[pf->asd + 1] ? pf->ealast->name : $3); free($3); pf->eastack[pf->asd + 1] = NULL; } ; etherfilter_opts : { bzero(&filter_opts, sizeof filter_opts); } etherfilter_opts_l { $$ = filter_opts; } | /* empty */ { bzero(&filter_opts, sizeof filter_opts); $$ = filter_opts; } ; etherfilter_opts_l : etherfilter_opts_l etherfilter_opt | etherfilter_opt etherfilter_opt : etherqname { if (filter_opts.queues.qname) { yyerror("queue cannot be redefined"); YYERROR; } filter_opts.queues = $1; } | RIDENTIFIER number { filter_opts.ridentifier = $2; } | label { if (filter_opts.labelcount >= PF_RULE_MAX_LABEL_COUNT) { yyerror("label can only be used %d times", PF_RULE_MAX_LABEL_COUNT); YYERROR; } filter_opts.label[filter_opts.labelcount++] = $1; } | TAG string { filter_opts.tag = $2; } | not TAGGED string { filter_opts.match_tag = $3; filter_opts.match_tag_not = $1; } | DNPIPE number { filter_opts.dnpipe = $2; filter_opts.free_flags |= PFRULE_DN_IS_PIPE; } | DNQUEUE number { filter_opts.dnpipe = $2; filter_opts.free_flags |= PFRULE_DN_IS_QUEUE; } ; bridge : /* empty */ { $$ = NULL; } | BRIDGE_TO STRING { $$ = strdup($2); } ; scrubrule : scrubaction dir logquick interface af proto fromto scrub_opts { struct pfctl_rule r; if (check_rulestate(PFCTL_STATE_SCRUB)) YYERROR; memset(&r, 0, sizeof(r)); r.action = $1.b1; r.direction = $2; r.log = $3.log; r.logif = $3.logif; if ($3.quick) { yyerror("scrub rules do not support 'quick'"); YYERROR; } r.af = $5; if ($8.nodf) r.rule_flag |= PFRULE_NODF; if ($8.randomid) r.rule_flag |= PFRULE_RANDOMID; if ($8.reassemble_tcp) { if (r.direction != PF_INOUT) { yyerror("reassemble tcp rules can not " "specify direction"); YYERROR; } r.rule_flag |= PFRULE_REASSEMBLE_TCP; } if ($8.minttl) r.min_ttl = $8.minttl; if ($8.maxmss) r.max_mss = $8.maxmss; if ($8.marker & FOM_SETTOS) { r.rule_flag |= PFRULE_SET_TOS; r.set_tos = $8.settos; } if ($8.fragcache) r.rule_flag |= $8.fragcache; if ($8.match_tag) if (strlcpy(r.match_tagname, $8.match_tag, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } r.match_tag_not = $8.match_tag_not; r.rtableid = $8.rtableid; expand_rule(&r, $4, NULL, $6, $7.src_os, $7.src.host, $7.src.port, $7.dst.host, $7.dst.port, NULL, NULL, NULL, NULL, ""); } ; scrub_opts : { bzero(&scrub_opts, sizeof scrub_opts); scrub_opts.rtableid = -1; } scrub_opts_l { $$ = scrub_opts; } | /* empty */ { bzero(&scrub_opts, sizeof scrub_opts); scrub_opts.rtableid = -1; $$ = scrub_opts; } ; scrub_opts_l : scrub_opts_l comma scrub_opt | scrub_opt ; scrub_opt : NODF { if (scrub_opts.nodf) { yyerror("no-df cannot be respecified"); YYERROR; } scrub_opts.nodf = 1; } | MINTTL NUMBER { if (scrub_opts.marker & FOM_MINTTL) { yyerror("min-ttl cannot be respecified"); YYERROR; } if ($2 < 0 || $2 > 255) { yyerror("illegal min-ttl value %d", $2); YYERROR; } scrub_opts.marker |= FOM_MINTTL; scrub_opts.minttl = $2; } | MAXMSS NUMBER { if (scrub_opts.marker & FOM_MAXMSS) { yyerror("max-mss cannot be respecified"); YYERROR; } if ($2 < 0 || $2 > 65535) { yyerror("illegal max-mss value %d", $2); YYERROR; } scrub_opts.marker |= FOM_MAXMSS; scrub_opts.maxmss = $2; } | SETTOS tos { if (scrub_opts.marker & FOM_SETTOS) { yyerror("set-tos cannot be respecified"); YYERROR; } scrub_opts.marker |= FOM_SETTOS; scrub_opts.settos = $2; } | fragcache { if (scrub_opts.marker & FOM_FRAGCACHE) { yyerror("fragcache cannot be respecified"); YYERROR; } scrub_opts.marker |= FOM_FRAGCACHE; scrub_opts.fragcache = $1; } | REASSEMBLE STRING { if (strcasecmp($2, "tcp") != 0) { yyerror("scrub reassemble supports only tcp, " "not '%s'", $2); free($2); YYERROR; } free($2); if (scrub_opts.reassemble_tcp) { yyerror("reassemble tcp cannot be respecified"); YYERROR; } scrub_opts.reassemble_tcp = 1; } | RANDOMID { if (scrub_opts.randomid) { yyerror("random-id cannot be respecified"); YYERROR; } scrub_opts.randomid = 1; } | RTABLE NUMBER { if ($2 < 0 || $2 > rt_tableid_max()) { yyerror("invalid rtable id"); YYERROR; } scrub_opts.rtableid = $2; } | not TAGGED string { scrub_opts.match_tag = $3; scrub_opts.match_tag_not = $1; } ; fragcache : FRAGMENT REASSEMBLE { $$ = 0; /* default */ } | FRAGMENT NO REASSEMBLE { $$ = PFRULE_FRAGMENT_NOREASS; } ; antispoof : ANTISPOOF logquick antispoof_ifspc af antispoof_opts { struct pfctl_rule r; struct node_host *h = NULL, *hh; struct node_if *i, *j; if (check_rulestate(PFCTL_STATE_FILTER)) YYERROR; for (i = $3; i; i = i->next) { bzero(&r, sizeof(r)); r.action = PF_DROP; r.direction = PF_IN; r.log = $2.log; r.logif = $2.logif; r.quick = $2.quick; r.af = $4; r.ridentifier = $5.ridentifier; if (rule_label(&r, $5.label)) YYERROR; r.rtableid = $5.rtableid; j = calloc(1, sizeof(struct node_if)); if (j == NULL) err(1, "antispoof: calloc"); if (strlcpy(j->ifname, i->ifname, sizeof(j->ifname)) >= sizeof(j->ifname)) { free(j); yyerror("interface name too long"); YYERROR; } j->not = 1; if (i->dynamic) { h = calloc(1, sizeof(*h)); if (h == NULL) err(1, "address: calloc"); h->addr.type = PF_ADDR_DYNIFTL; set_ipmask(h, 128); if (strlcpy(h->addr.v.ifname, i->ifname, sizeof(h->addr.v.ifname)) >= sizeof(h->addr.v.ifname)) { free(h); yyerror( "interface name too long"); YYERROR; } hh = malloc(sizeof(*hh)); if (hh == NULL) err(1, "address: malloc"); bcopy(h, hh, sizeof(*hh)); h->addr.iflags = PFI_AFLAG_NETWORK; } else { h = ifa_lookup(j->ifname, PFI_AFLAG_NETWORK); hh = NULL; } if (h != NULL) expand_rule(&r, j, NULL, NULL, NULL, h, NULL, NULL, NULL, NULL, NULL, NULL, NULL, ""); if ((i->ifa_flags & IFF_LOOPBACK) == 0) { bzero(&r, sizeof(r)); r.action = PF_DROP; r.direction = PF_IN; r.log = $2.log; r.logif = $2.logif; r.quick = $2.quick; r.af = $4; r.ridentifier = $5.ridentifier; if (rule_label(&r, $5.label)) YYERROR; r.rtableid = $5.rtableid; if (hh != NULL) h = hh; else h = ifa_lookup(i->ifname, 0); if (h != NULL) expand_rule(&r, NULL, NULL, NULL, NULL, h, NULL, NULL, NULL, NULL, NULL, NULL, NULL, ""); } else free(hh); } for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) free($5.label[i]); } ; antispoof_ifspc : FOR antispoof_if { $$ = $2; } | FOR '{' optnl antispoof_iflst '}' { $$ = $4; } ; antispoof_iflst : antispoof_if optnl { $$ = $1; } | antispoof_iflst comma antispoof_if optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; antispoof_if : if_item { $$ = $1; } | '(' if_item ')' { $2->dynamic = 1; $$ = $2; } ; antispoof_opts : { bzero(&antispoof_opts, sizeof antispoof_opts); antispoof_opts.rtableid = -1; } antispoof_opts_l { $$ = antispoof_opts; } | /* empty */ { bzero(&antispoof_opts, sizeof antispoof_opts); antispoof_opts.rtableid = -1; $$ = antispoof_opts; } ; antispoof_opts_l : antispoof_opts_l antispoof_opt | antispoof_opt ; antispoof_opt : label { if (antispoof_opts.labelcount >= PF_RULE_MAX_LABEL_COUNT) { yyerror("label can only be used %d times", PF_RULE_MAX_LABEL_COUNT); YYERROR; } antispoof_opts.label[antispoof_opts.labelcount++] = $1; } | RIDENTIFIER number { antispoof_opts.ridentifier = $2; } | RTABLE NUMBER { if ($2 < 0 || $2 > rt_tableid_max()) { yyerror("invalid rtable id"); YYERROR; } antispoof_opts.rtableid = $2; } ; not : '!' { $$ = 1; } | /* empty */ { $$ = 0; } ; tabledef : TABLE '<' STRING '>' table_opts { struct node_host *h, *nh; struct node_tinit *ti, *nti; if (strlen($3) >= PF_TABLE_NAME_SIZE) { yyerror("table name too long, max %d chars", PF_TABLE_NAME_SIZE - 1); free($3); YYERROR; } if (pf->loadopt & PFCTL_FLAG_TABLE) if (process_tabledef($3, &$5)) { free($3); YYERROR; } free($3); for (ti = SIMPLEQ_FIRST(&$5.init_nodes); ti != SIMPLEQ_END(&$5.init_nodes); ti = nti) { if (ti->file) free(ti->file); for (h = ti->host; h != NULL; h = nh) { nh = h->next; free(h); } nti = SIMPLEQ_NEXT(ti, entries); free(ti); } } ; table_opts : { bzero(&table_opts, sizeof table_opts); SIMPLEQ_INIT(&table_opts.init_nodes); } table_opts_l { $$ = table_opts; } | /* empty */ { bzero(&table_opts, sizeof table_opts); SIMPLEQ_INIT(&table_opts.init_nodes); $$ = table_opts; } ; table_opts_l : table_opts_l table_opt | table_opt ; table_opt : STRING { if (!strcmp($1, "const")) table_opts.flags |= PFR_TFLAG_CONST; else if (!strcmp($1, "persist")) table_opts.flags |= PFR_TFLAG_PERSIST; else if (!strcmp($1, "counters")) table_opts.flags |= PFR_TFLAG_COUNTERS; else { yyerror("invalid table option '%s'", $1); free($1); YYERROR; } free($1); } | '{' optnl '}' { table_opts.init_addr = 1; } | '{' optnl host_list '}' { struct node_host *n; struct node_tinit *ti; for (n = $3; n != NULL; n = n->next) { switch (n->addr.type) { case PF_ADDR_ADDRMASK: continue; /* ok */ case PF_ADDR_RANGE: yyerror("address ranges are not " "permitted inside tables"); break; case PF_ADDR_DYNIFTL: yyerror("dynamic addresses are not " "permitted inside tables"); break; case PF_ADDR_TABLE: yyerror("tables cannot contain tables"); break; case PF_ADDR_NOROUTE: yyerror("\"no-route\" is not permitted " "inside tables"); break; case PF_ADDR_URPFFAILED: yyerror("\"urpf-failed\" is not " "permitted inside tables"); break; default: yyerror("unknown address type %d", n->addr.type); } YYERROR; } if (!(ti = calloc(1, sizeof(*ti)))) err(1, "table_opt: calloc"); ti->host = $3; SIMPLEQ_INSERT_TAIL(&table_opts.init_nodes, ti, entries); table_opts.init_addr = 1; } | FILENAME STRING { struct node_tinit *ti; if (!(ti = calloc(1, sizeof(*ti)))) err(1, "table_opt: calloc"); ti->file = $2; SIMPLEQ_INSERT_TAIL(&table_opts.init_nodes, ti, entries); table_opts.init_addr = 1; } ; altqif : ALTQ interface queue_opts QUEUE qassign { struct pf_altq a; if (check_rulestate(PFCTL_STATE_QUEUE)) YYERROR; memset(&a, 0, sizeof(a)); if ($3.scheduler.qtype == ALTQT_NONE) { yyerror("no scheduler specified!"); YYERROR; } a.scheduler = $3.scheduler.qtype; a.qlimit = $3.qlimit; a.tbrsize = $3.tbrsize; if ($5 == NULL && $3.scheduler.qtype != ALTQT_CODEL) { yyerror("no child queues specified"); YYERROR; } if (expand_altq(&a, $2, $5, $3.queue_bwspec, &$3.scheduler)) YYERROR; } ; queuespec : QUEUE STRING interface queue_opts qassign { struct pf_altq a; if (check_rulestate(PFCTL_STATE_QUEUE)) { free($2); YYERROR; } memset(&a, 0, sizeof(a)); if (strlcpy(a.qname, $2, sizeof(a.qname)) >= sizeof(a.qname)) { yyerror("queue name too long (max " "%d chars)", PF_QNAME_SIZE-1); free($2); YYERROR; } free($2); if ($4.tbrsize) { yyerror("cannot specify tbrsize for queue"); YYERROR; } if ($4.priority > 255) { yyerror("priority out of range: max 255"); YYERROR; } a.priority = $4.priority; a.qlimit = $4.qlimit; a.scheduler = $4.scheduler.qtype; if (expand_queue(&a, $3, $5, $4.queue_bwspec, &$4.scheduler)) { yyerror("errors in queue definition"); YYERROR; } } ; queue_opts : { bzero(&queue_opts, sizeof queue_opts); queue_opts.priority = DEFAULT_PRIORITY; queue_opts.qlimit = DEFAULT_QLIMIT; queue_opts.scheduler.qtype = ALTQT_NONE; queue_opts.queue_bwspec.bw_percent = 100; } queue_opts_l { $$ = queue_opts; } | /* empty */ { bzero(&queue_opts, sizeof queue_opts); queue_opts.priority = DEFAULT_PRIORITY; queue_opts.qlimit = DEFAULT_QLIMIT; queue_opts.scheduler.qtype = ALTQT_NONE; queue_opts.queue_bwspec.bw_percent = 100; $$ = queue_opts; } ; queue_opts_l : queue_opts_l queue_opt | queue_opt ; queue_opt : BANDWIDTH bandwidth { if (queue_opts.marker & QOM_BWSPEC) { yyerror("bandwidth cannot be respecified"); YYERROR; } queue_opts.marker |= QOM_BWSPEC; queue_opts.queue_bwspec = $2; } | PRIORITY NUMBER { if (queue_opts.marker & QOM_PRIORITY) { yyerror("priority cannot be respecified"); YYERROR; } if ($2 < 0 || $2 > 255) { yyerror("priority out of range: max 255"); YYERROR; } queue_opts.marker |= QOM_PRIORITY; queue_opts.priority = $2; } | QLIMIT NUMBER { if (queue_opts.marker & QOM_QLIMIT) { yyerror("qlimit cannot be respecified"); YYERROR; } if ($2 < 0 || $2 > 65535) { yyerror("qlimit out of range: max 65535"); YYERROR; } queue_opts.marker |= QOM_QLIMIT; queue_opts.qlimit = $2; } | scheduler { if (queue_opts.marker & QOM_SCHEDULER) { yyerror("scheduler cannot be respecified"); YYERROR; } queue_opts.marker |= QOM_SCHEDULER; queue_opts.scheduler = $1; } | TBRSIZE NUMBER { if (queue_opts.marker & QOM_TBRSIZE) { yyerror("tbrsize cannot be respecified"); YYERROR; } if ($2 < 0 || $2 > UINT_MAX) { yyerror("tbrsize too big: max %u", UINT_MAX); YYERROR; } queue_opts.marker |= QOM_TBRSIZE; queue_opts.tbrsize = $2; } ; bandwidth : STRING { double bps; char *cp; $$.bw_percent = 0; bps = strtod($1, &cp); if (cp != NULL) { if (strlen(cp) > 1) { char *cu = cp + 1; if (!strcmp(cu, "Bit") || !strcmp(cu, "B") || !strcmp(cu, "bit") || !strcmp(cu, "b")) { *cu = 0; } } if (!strcmp(cp, "b")) ; /* nothing */ else if (!strcmp(cp, "K")) bps *= 1000; else if (!strcmp(cp, "M")) bps *= 1000 * 1000; else if (!strcmp(cp, "G")) bps *= 1000 * 1000 * 1000; else if (!strcmp(cp, "%")) { if (bps < 0 || bps > 100) { yyerror("bandwidth spec " "out of range"); free($1); YYERROR; } $$.bw_percent = bps; bps = 0; } else { yyerror("unknown unit %s", cp); free($1); YYERROR; } } free($1); $$.bw_absolute = (u_int64_t)bps; } | NUMBER { if ($1 < 0 || $1 >= LLONG_MAX) { yyerror("bandwidth number too big"); YYERROR; } $$.bw_percent = 0; $$.bw_absolute = $1; } ; scheduler : CBQ { $$.qtype = ALTQT_CBQ; $$.data.cbq_opts.flags = 0; } | CBQ '(' cbqflags_list ')' { $$.qtype = ALTQT_CBQ; $$.data.cbq_opts.flags = $3; } | PRIQ { $$.qtype = ALTQT_PRIQ; $$.data.priq_opts.flags = 0; } | PRIQ '(' priqflags_list ')' { $$.qtype = ALTQT_PRIQ; $$.data.priq_opts.flags = $3; } | HFSC { $$.qtype = ALTQT_HFSC; bzero(&$$.data.hfsc_opts, sizeof(struct node_hfsc_opts)); } | HFSC '(' hfsc_opts ')' { $$.qtype = ALTQT_HFSC; $$.data.hfsc_opts = $3; } | FAIRQ { $$.qtype = ALTQT_FAIRQ; bzero(&$$.data.fairq_opts, sizeof(struct node_fairq_opts)); } | FAIRQ '(' fairq_opts ')' { $$.qtype = ALTQT_FAIRQ; $$.data.fairq_opts = $3; } | CODEL { $$.qtype = ALTQT_CODEL; bzero(&$$.data.codel_opts, sizeof(struct codel_opts)); } | CODEL '(' codel_opts ')' { $$.qtype = ALTQT_CODEL; $$.data.codel_opts = $3; } ; cbqflags_list : cbqflags_item { $$ |= $1; } | cbqflags_list comma cbqflags_item { $$ |= $3; } ; cbqflags_item : STRING { if (!strcmp($1, "default")) $$ = CBQCLF_DEFCLASS; else if (!strcmp($1, "borrow")) $$ = CBQCLF_BORROW; else if (!strcmp($1, "red")) $$ = CBQCLF_RED; else if (!strcmp($1, "ecn")) $$ = CBQCLF_RED|CBQCLF_ECN; else if (!strcmp($1, "rio")) $$ = CBQCLF_RIO; else if (!strcmp($1, "codel")) $$ = CBQCLF_CODEL; else { yyerror("unknown cbq flag \"%s\"", $1); free($1); YYERROR; } free($1); } ; priqflags_list : priqflags_item { $$ |= $1; } | priqflags_list comma priqflags_item { $$ |= $3; } ; priqflags_item : STRING { if (!strcmp($1, "default")) $$ = PRCF_DEFAULTCLASS; else if (!strcmp($1, "red")) $$ = PRCF_RED; else if (!strcmp($1, "ecn")) $$ = PRCF_RED|PRCF_ECN; else if (!strcmp($1, "rio")) $$ = PRCF_RIO; else if (!strcmp($1, "codel")) $$ = PRCF_CODEL; else { yyerror("unknown priq flag \"%s\"", $1); free($1); YYERROR; } free($1); } ; hfsc_opts : { bzero(&hfsc_opts, sizeof(struct node_hfsc_opts)); } hfscopts_list { $$ = hfsc_opts; } ; hfscopts_list : hfscopts_item | hfscopts_list comma hfscopts_item ; hfscopts_item : LINKSHARE bandwidth { if (hfsc_opts.linkshare.used) { yyerror("linkshare already specified"); YYERROR; } hfsc_opts.linkshare.m2 = $2; hfsc_opts.linkshare.used = 1; } | LINKSHARE '(' bandwidth comma NUMBER comma bandwidth ')' { if ($5 < 0 || $5 > INT_MAX) { yyerror("timing in curve out of range"); YYERROR; } if (hfsc_opts.linkshare.used) { yyerror("linkshare already specified"); YYERROR; } hfsc_opts.linkshare.m1 = $3; hfsc_opts.linkshare.d = $5; hfsc_opts.linkshare.m2 = $7; hfsc_opts.linkshare.used = 1; } | REALTIME bandwidth { if (hfsc_opts.realtime.used) { yyerror("realtime already specified"); YYERROR; } hfsc_opts.realtime.m2 = $2; hfsc_opts.realtime.used = 1; } | REALTIME '(' bandwidth comma NUMBER comma bandwidth ')' { if ($5 < 0 || $5 > INT_MAX) { yyerror("timing in curve out of range"); YYERROR; } if (hfsc_opts.realtime.used) { yyerror("realtime already specified"); YYERROR; } hfsc_opts.realtime.m1 = $3; hfsc_opts.realtime.d = $5; hfsc_opts.realtime.m2 = $7; hfsc_opts.realtime.used = 1; } | UPPERLIMIT bandwidth { if (hfsc_opts.upperlimit.used) { yyerror("upperlimit already specified"); YYERROR; } hfsc_opts.upperlimit.m2 = $2; hfsc_opts.upperlimit.used = 1; } | UPPERLIMIT '(' bandwidth comma NUMBER comma bandwidth ')' { if ($5 < 0 || $5 > INT_MAX) { yyerror("timing in curve out of range"); YYERROR; } if (hfsc_opts.upperlimit.used) { yyerror("upperlimit already specified"); YYERROR; } hfsc_opts.upperlimit.m1 = $3; hfsc_opts.upperlimit.d = $5; hfsc_opts.upperlimit.m2 = $7; hfsc_opts.upperlimit.used = 1; } | STRING { if (!strcmp($1, "default")) hfsc_opts.flags |= HFCF_DEFAULTCLASS; else if (!strcmp($1, "red")) hfsc_opts.flags |= HFCF_RED; else if (!strcmp($1, "ecn")) hfsc_opts.flags |= HFCF_RED|HFCF_ECN; else if (!strcmp($1, "rio")) hfsc_opts.flags |= HFCF_RIO; else if (!strcmp($1, "codel")) hfsc_opts.flags |= HFCF_CODEL; else { yyerror("unknown hfsc flag \"%s\"", $1); free($1); YYERROR; } free($1); } ; fairq_opts : { bzero(&fairq_opts, sizeof(struct node_fairq_opts)); } fairqopts_list { $$ = fairq_opts; } ; fairqopts_list : fairqopts_item | fairqopts_list comma fairqopts_item ; fairqopts_item : LINKSHARE bandwidth { if (fairq_opts.linkshare.used) { yyerror("linkshare already specified"); YYERROR; } fairq_opts.linkshare.m2 = $2; fairq_opts.linkshare.used = 1; } | LINKSHARE '(' bandwidth number bandwidth ')' { if (fairq_opts.linkshare.used) { yyerror("linkshare already specified"); YYERROR; } fairq_opts.linkshare.m1 = $3; fairq_opts.linkshare.d = $4; fairq_opts.linkshare.m2 = $5; fairq_opts.linkshare.used = 1; } | HOGS bandwidth { fairq_opts.hogs_bw = $2; } | BUCKETS number { fairq_opts.nbuckets = $2; } | STRING { if (!strcmp($1, "default")) fairq_opts.flags |= FARF_DEFAULTCLASS; else if (!strcmp($1, "red")) fairq_opts.flags |= FARF_RED; else if (!strcmp($1, "ecn")) fairq_opts.flags |= FARF_RED|FARF_ECN; else if (!strcmp($1, "rio")) fairq_opts.flags |= FARF_RIO; else if (!strcmp($1, "codel")) fairq_opts.flags |= FARF_CODEL; else { yyerror("unknown fairq flag \"%s\"", $1); free($1); YYERROR; } free($1); } ; codel_opts : { bzero(&codel_opts, sizeof(struct codel_opts)); } codelopts_list { $$ = codel_opts; } ; codelopts_list : codelopts_item | codelopts_list comma codelopts_item ; codelopts_item : INTERVAL number { if (codel_opts.interval) { yyerror("interval already specified"); YYERROR; } codel_opts.interval = $2; } | TARGET number { if (codel_opts.target) { yyerror("target already specified"); YYERROR; } codel_opts.target = $2; } | STRING { if (!strcmp($1, "ecn")) codel_opts.ecn = 1; else { yyerror("unknown codel option \"%s\"", $1); free($1); YYERROR; } free($1); } ; qassign : /* empty */ { $$ = NULL; } | qassign_item { $$ = $1; } | '{' optnl qassign_list '}' { $$ = $3; } ; qassign_list : qassign_item optnl { $$ = $1; } | qassign_list comma qassign_item optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; qassign_item : STRING { $$ = calloc(1, sizeof(struct node_queue)); if ($$ == NULL) err(1, "qassign_item: calloc"); if (strlcpy($$->queue, $1, sizeof($$->queue)) >= sizeof($$->queue)) { yyerror("queue name '%s' too long (max " "%d chars)", $1, sizeof($$->queue)-1); free($1); free($$); YYERROR; } free($1); $$->next = NULL; $$->tail = $$; } ; pfrule : action dir logquick interface route af proto fromto filter_opts { struct pfctl_rule r; struct node_state_opt *o; struct node_proto *proto; int srctrack = 0; int statelock = 0; int adaptive = 0; int defaults = 0; if (check_rulestate(PFCTL_STATE_FILTER)) YYERROR; memset(&r, 0, sizeof(r)); r.action = $1.b1; switch ($1.b2) { case PFRULE_RETURNRST: r.rule_flag |= PFRULE_RETURNRST; r.return_ttl = $1.w; break; case PFRULE_RETURNICMP: r.rule_flag |= PFRULE_RETURNICMP; r.return_icmp = $1.w; r.return_icmp6 = $1.w2; break; case PFRULE_RETURN: r.rule_flag |= PFRULE_RETURN; r.return_icmp = $1.w; r.return_icmp6 = $1.w2; break; } r.direction = $2; r.log = $3.log; r.logif = $3.logif; r.quick = $3.quick; r.prob = $9.prob; r.rtableid = $9.rtableid; if ($9.nodf) r.scrub_flags |= PFSTATE_NODF; if ($9.randomid) r.scrub_flags |= PFSTATE_RANDOMID; if ($9.minttl) r.min_ttl = $9.minttl; if ($9.max_mss) r.max_mss = $9.max_mss; if ($9.marker & FOM_SETTOS) { r.scrub_flags |= PFSTATE_SETTOS; r.set_tos = $9.settos; } if ($9.marker & FOM_SCRUB_TCP) r.scrub_flags |= PFSTATE_SCRUB_TCP; if ($9.marker & FOM_PRIO) { if ($9.prio == 0) r.prio = PF_PRIO_ZERO; else r.prio = $9.prio; } if ($9.marker & FOM_SETPRIO) { r.set_prio[0] = $9.set_prio[0]; r.set_prio[1] = $9.set_prio[1]; r.scrub_flags |= PFSTATE_SETPRIO; } r.af = $6; if ($9.tag) if (strlcpy(r.tagname, $9.tag, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } if ($9.match_tag) if (strlcpy(r.match_tagname, $9.match_tag, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } r.match_tag_not = $9.match_tag_not; if (rule_label(&r, $9.label)) YYERROR; for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) free($9.label[i]); r.ridentifier = $9.ridentifier; r.flags = $9.flags.b1; r.flagset = $9.flags.b2; if (($9.flags.b1 & $9.flags.b2) != $9.flags.b1) { yyerror("flags always false"); YYERROR; } if ($9.flags.b1 || $9.flags.b2 || $8.src_os) { for (proto = $7; proto != NULL && proto->proto != IPPROTO_TCP; proto = proto->next) ; /* nothing */ if (proto == NULL && $7 != NULL) { if ($9.flags.b1 || $9.flags.b2) yyerror( "flags only apply to tcp"); if ($8.src_os) yyerror( "OS fingerprinting only " "apply to tcp"); YYERROR; } #if 0 if (($9.flags.b1 & parse_flags("S")) == 0 && $8.src_os) { yyerror("OS fingerprinting requires " "the SYN TCP flag (flags S/SA)"); YYERROR; } #endif } r.tos = $9.tos; r.keep_state = $9.keep.action; o = $9.keep.options; /* 'keep state' by default on pass rules. */ if (!r.keep_state && !r.action && !($9.marker & FOM_KEEP)) { r.keep_state = PF_STATE_NORMAL; o = keep_state_defaults; defaults = 1; } while (o) { struct node_state_opt *p = o; switch (o->type) { case PF_STATE_OPT_MAX: if (r.max_states) { yyerror("state option 'max' " "multiple definitions"); YYERROR; } r.max_states = o->data.max_states; break; case PF_STATE_OPT_NOSYNC: if (r.rule_flag & PFRULE_NOSYNC) { yyerror("state option 'sync' " "multiple definitions"); YYERROR; } r.rule_flag |= PFRULE_NOSYNC; break; case PF_STATE_OPT_SRCTRACK: if (srctrack) { yyerror("state option " "'source-track' " "multiple definitions"); YYERROR; } srctrack = o->data.src_track; r.rule_flag |= PFRULE_SRCTRACK; break; case PF_STATE_OPT_MAX_SRC_STATES: if (r.max_src_states) { yyerror("state option " "'max-src-states' " "multiple definitions"); YYERROR; } if (o->data.max_src_states == 0) { yyerror("'max-src-states' must " "be > 0"); YYERROR; } r.max_src_states = o->data.max_src_states; r.rule_flag |= PFRULE_SRCTRACK; break; case PF_STATE_OPT_OVERLOAD: if (r.overload_tblname[0]) { yyerror("multiple 'overload' " "table definitions"); YYERROR; } if (strlcpy(r.overload_tblname, o->data.overload.tblname, PF_TABLE_NAME_SIZE) >= PF_TABLE_NAME_SIZE) { yyerror("state option: " "strlcpy"); YYERROR; } r.flush = o->data.overload.flush; break; case PF_STATE_OPT_MAX_SRC_CONN: if (r.max_src_conn) { yyerror("state option " "'max-src-conn' " "multiple definitions"); YYERROR; } if (o->data.max_src_conn == 0) { yyerror("'max-src-conn' " "must be > 0"); YYERROR; } r.max_src_conn = o->data.max_src_conn; r.rule_flag |= PFRULE_SRCTRACK | PFRULE_RULESRCTRACK; break; case PF_STATE_OPT_MAX_SRC_CONN_RATE: if (r.max_src_conn_rate.limit) { yyerror("state option " "'max-src-conn-rate' " "multiple definitions"); YYERROR; } if (!o->data.max_src_conn_rate.limit || !o->data.max_src_conn_rate.seconds) { yyerror("'max-src-conn-rate' " "values must be > 0"); YYERROR; } if (o->data.max_src_conn_rate.limit > PF_THRESHOLD_MAX) { yyerror("'max-src-conn-rate' " "maximum rate must be < %u", PF_THRESHOLD_MAX); YYERROR; } r.max_src_conn_rate.limit = o->data.max_src_conn_rate.limit; r.max_src_conn_rate.seconds = o->data.max_src_conn_rate.seconds; r.rule_flag |= PFRULE_SRCTRACK | PFRULE_RULESRCTRACK; break; case PF_STATE_OPT_MAX_SRC_NODES: if (r.max_src_nodes) { yyerror("state option " "'max-src-nodes' " "multiple definitions"); YYERROR; } if (o->data.max_src_nodes == 0) { yyerror("'max-src-nodes' must " "be > 0"); YYERROR; } r.max_src_nodes = o->data.max_src_nodes; r.rule_flag |= PFRULE_SRCTRACK | PFRULE_RULESRCTRACK; break; case PF_STATE_OPT_STATELOCK: if (statelock) { yyerror("state locking option: " "multiple definitions"); YYERROR; } statelock = 1; r.rule_flag |= o->data.statelock; break; case PF_STATE_OPT_SLOPPY: if (r.rule_flag & PFRULE_STATESLOPPY) { yyerror("state sloppy option: " "multiple definitions"); YYERROR; } r.rule_flag |= PFRULE_STATESLOPPY; break; case PF_STATE_OPT_PFLOW: if (r.rule_flag & PFRULE_PFLOW) { yyerror("state pflow option: " "multiple definitions"); YYERROR; } r.rule_flag |= PFRULE_PFLOW; break; case PF_STATE_OPT_TIMEOUT: if (o->data.timeout.number == PFTM_ADAPTIVE_START || o->data.timeout.number == PFTM_ADAPTIVE_END) adaptive = 1; if (r.timeout[o->data.timeout.number]) { yyerror("state timeout %s " "multiple definitions", pf_timeouts[o->data. timeout.number].name); YYERROR; } r.timeout[o->data.timeout.number] = o->data.timeout.seconds; } o = o->next; if (!defaults) free(p); } /* 'flags S/SA' by default on stateful rules */ if (!r.action && !r.flags && !r.flagset && !$9.fragment && !($9.marker & FOM_FLAGS) && r.keep_state) { r.flags = parse_flags("S"); r.flagset = parse_flags("SA"); } if (!adaptive && r.max_states) { r.timeout[PFTM_ADAPTIVE_START] = (r.max_states / 10) * 6; r.timeout[PFTM_ADAPTIVE_END] = (r.max_states / 10) * 12; } if (r.rule_flag & PFRULE_SRCTRACK) { if (srctrack == PF_SRCTRACK_GLOBAL && r.max_src_nodes) { yyerror("'max-src-nodes' is " "incompatible with " "'source-track global'"); YYERROR; } if (srctrack == PF_SRCTRACK_GLOBAL && r.max_src_conn) { yyerror("'max-src-conn' is " "incompatible with " "'source-track global'"); YYERROR; } if (srctrack == PF_SRCTRACK_GLOBAL && r.max_src_conn_rate.seconds) { yyerror("'max-src-conn-rate' is " "incompatible with " "'source-track global'"); YYERROR; } if (r.timeout[PFTM_SRC_NODE] < r.max_src_conn_rate.seconds) r.timeout[PFTM_SRC_NODE] = r.max_src_conn_rate.seconds; r.rule_flag |= PFRULE_SRCTRACK; if (srctrack == PF_SRCTRACK_RULE) r.rule_flag |= PFRULE_RULESRCTRACK; } if (r.keep_state && !statelock) r.rule_flag |= default_statelock; if ($9.fragment) r.rule_flag |= PFRULE_FRAGMENT; r.allow_opts = $9.allowopts; decide_address_family($8.src.host, &r.af); decide_address_family($8.dst.host, &r.af); if ($5.rt) { if (!r.direction) { yyerror("direction must be explicit " "with rules that specify routing"); YYERROR; } r.rt = $5.rt; r.rpool.opts = $5.pool_opts; if ($5.key != NULL) memcpy(&r.rpool.key, $5.key, sizeof(struct pf_poolhashkey)); } if (r.rt) { decide_address_family($5.host, &r.af); remove_invalid_hosts(&$5.host, &r.af); if ($5.host == NULL) { yyerror("no routing address with " "matching address family found."); YYERROR; } if ((r.rpool.opts & PF_POOL_TYPEMASK) == PF_POOL_NONE && ($5.host->next != NULL || $5.host->addr.type == PF_ADDR_TABLE || DYNIF_MULTIADDR($5.host->addr))) r.rpool.opts |= PF_POOL_ROUNDROBIN; if ((r.rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN && disallow_table($5.host, "tables are only " "supported in round-robin routing pools")) YYERROR; if ((r.rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN && disallow_alias($5.host, "interface (%s) " "is only supported in round-robin " "routing pools")) YYERROR; if ($5.host->next != NULL) { if ((r.rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) { yyerror("r.rpool.opts must " "be PF_POOL_ROUNDROBIN"); YYERROR; } } } if ($9.queues.qname != NULL) { if (strlcpy(r.qname, $9.queues.qname, sizeof(r.qname)) >= sizeof(r.qname)) { yyerror("rule qname too long (max " "%d chars)", sizeof(r.qname)-1); YYERROR; } free($9.queues.qname); } if ($9.queues.pqname != NULL) { if (strlcpy(r.pqname, $9.queues.pqname, sizeof(r.pqname)) >= sizeof(r.pqname)) { yyerror("rule pqname too long (max " "%d chars)", sizeof(r.pqname)-1); YYERROR; } free($9.queues.pqname); } #ifdef __FreeBSD__ r.divert.port = $9.divert.port; #else if ((r.divert.port = $9.divert.port)) { if (r.direction == PF_OUT) { if ($9.divert.addr) { yyerror("address specified " "for outgoing divert"); YYERROR; } bzero(&r.divert.addr, sizeof(r.divert.addr)); } else { if (!$9.divert.addr) { yyerror("no address specified " "for incoming divert"); YYERROR; } if ($9.divert.addr->af != r.af) { yyerror("address family " "mismatch for divert"); YYERROR; } r.divert.addr = $9.divert.addr->addr.v.a.addr; } } #endif if ($9.dnpipe || $9.dnrpipe) { r.dnpipe = $9.dnpipe; r.dnrpipe = $9.dnrpipe; if ($9.free_flags & PFRULE_DN_IS_PIPE) r.free_flags |= PFRULE_DN_IS_PIPE; else r.free_flags |= PFRULE_DN_IS_QUEUE; } expand_rule(&r, $4, $5.host, $7, $8.src_os, $8.src.host, $8.src.port, $8.dst.host, $8.dst.port, $9.uid, $9.gid, $9.rcv, $9.icmpspec, ""); } ; filter_opts : { bzero(&filter_opts, sizeof filter_opts); filter_opts.rtableid = -1; } filter_opts_l { $$ = filter_opts; } | /* empty */ { bzero(&filter_opts, sizeof filter_opts); filter_opts.rtableid = -1; $$ = filter_opts; } ; filter_opts_l : filter_opts_l filter_opt | filter_opt ; filter_opt : USER uids { if (filter_opts.uid) $2->tail->next = filter_opts.uid; filter_opts.uid = $2; } | GROUP gids { if (filter_opts.gid) $2->tail->next = filter_opts.gid; filter_opts.gid = $2; } | flags { if (filter_opts.marker & FOM_FLAGS) { yyerror("flags cannot be redefined"); YYERROR; } filter_opts.marker |= FOM_FLAGS; filter_opts.flags.b1 |= $1.b1; filter_opts.flags.b2 |= $1.b2; filter_opts.flags.w |= $1.w; filter_opts.flags.w2 |= $1.w2; } | icmpspec { if (filter_opts.marker & FOM_ICMP) { yyerror("icmp-type cannot be redefined"); YYERROR; } filter_opts.marker |= FOM_ICMP; filter_opts.icmpspec = $1; } | PRIO NUMBER { if (filter_opts.marker & FOM_PRIO) { yyerror("prio cannot be redefined"); YYERROR; } if ($2 < 0 || $2 > PF_PRIO_MAX) { yyerror("prio must be 0 - %u", PF_PRIO_MAX); YYERROR; } filter_opts.marker |= FOM_PRIO; filter_opts.prio = $2; } | TOS tos { if (filter_opts.marker & FOM_TOS) { yyerror("tos cannot be redefined"); YYERROR; } filter_opts.marker |= FOM_TOS; filter_opts.tos = $2; } | keep { if (filter_opts.marker & FOM_KEEP) { yyerror("modulate or keep cannot be redefined"); YYERROR; } filter_opts.marker |= FOM_KEEP; filter_opts.keep.action = $1.action; filter_opts.keep.options = $1.options; } | RIDENTIFIER number { filter_opts.ridentifier = $2; } | FRAGMENT { filter_opts.fragment = 1; } | ALLOWOPTS { filter_opts.allowopts = 1; } | label { if (filter_opts.labelcount >= PF_RULE_MAX_LABEL_COUNT) { yyerror("label can only be used %d times", PF_RULE_MAX_LABEL_COUNT); YYERROR; } filter_opts.label[filter_opts.labelcount++] = $1; } | qname { if (filter_opts.queues.qname) { yyerror("queue cannot be redefined"); YYERROR; } filter_opts.queues = $1; } | DNPIPE number { filter_opts.dnpipe = $2; filter_opts.free_flags |= PFRULE_DN_IS_PIPE; } | DNPIPE '(' number ')' { filter_opts.dnpipe = $3; filter_opts.free_flags |= PFRULE_DN_IS_PIPE; } | DNPIPE '(' number comma number ')' { filter_opts.dnrpipe = $5; filter_opts.dnpipe = $3; filter_opts.free_flags |= PFRULE_DN_IS_PIPE; } | DNQUEUE number { filter_opts.dnpipe = $2; filter_opts.free_flags |= PFRULE_DN_IS_QUEUE; } | DNQUEUE '(' number comma number ')' { filter_opts.dnrpipe = $5; filter_opts.dnpipe = $3; filter_opts.free_flags |= PFRULE_DN_IS_QUEUE; } | DNQUEUE '(' number ')' { filter_opts.dnpipe = $3; filter_opts.free_flags |= PFRULE_DN_IS_QUEUE; } | TAG string { filter_opts.tag = $2; } | not TAGGED string { filter_opts.match_tag = $3; filter_opts.match_tag_not = $1; } | RECEIVEDON if_item { if (filter_opts.rcv) { yyerror("cannot respecify received-on"); YYERROR; } filter_opts.rcv = $2; } | PROBABILITY probability { double p; p = floor($2 * UINT_MAX + 0.5); if (p < 0.0 || p > UINT_MAX) { yyerror("invalid probability: %lf", p); YYERROR; } filter_opts.prob = (u_int32_t)p; if (filter_opts.prob == 0) filter_opts.prob = 1; } | RTABLE NUMBER { if ($2 < 0 || $2 > rt_tableid_max()) { yyerror("invalid rtable id"); YYERROR; } filter_opts.rtableid = $2; } | DIVERTTO portplain { #ifdef __FreeBSD__ filter_opts.divert.port = $2.a; if (!filter_opts.divert.port) { yyerror("invalid divert port: %u", ntohs($2.a)); YYERROR; } #endif } | DIVERTTO STRING PORT portplain { #ifndef __FreeBSD__ if ((filter_opts.divert.addr = host($2)) == NULL) { yyerror("could not parse divert address: %s", $2); free($2); YYERROR; } #else if ($2) #endif free($2); filter_opts.divert.port = $4.a; if (!filter_opts.divert.port) { yyerror("invalid divert port: %u", ntohs($4.a)); YYERROR; } } | DIVERTREPLY { #ifdef __FreeBSD__ yyerror("divert-reply has no meaning in FreeBSD pf(4)"); YYERROR; #else filter_opts.divert.port = 1; /* some random value */ #endif } | SCRUB '(' scrub_opts ')' { filter_opts.nodf = $3.nodf; filter_opts.minttl = $3.minttl; if ($3.marker & FOM_SETTOS) { /* Old style rules are "scrub set-tos 0x42" * New style are "set tos 0x42 scrub (...)" * What is in "scrub(...)"" is unfortunately the * original scrub syntax so it would overwrite * "set tos" of a pass/match rule. */ filter_opts.settos = $3.settos; } filter_opts.randomid = $3.randomid; filter_opts.max_mss = $3.maxmss; if ($3.reassemble_tcp) filter_opts.marker |= FOM_SCRUB_TCP; filter_opts.marker |= $3.marker; } | filter_sets ; filter_sets : SET '(' filter_sets_l ')' { $$ = filter_opts; } | SET filter_set { $$ = filter_opts; } ; filter_sets_l : filter_sets_l comma filter_set | filter_set ; filter_set : prio { if (filter_opts.marker & FOM_SETPRIO) { yyerror("prio cannot be redefined"); YYERROR; } filter_opts.marker |= FOM_SETPRIO; filter_opts.set_prio[0] = $1.b1; filter_opts.set_prio[1] = $1.b2; } | TOS tos { if (filter_opts.marker & FOM_SETTOS) { yyerror("tos cannot be respecified"); YYERROR; } filter_opts.marker |= FOM_SETTOS; filter_opts.settos = $2; } prio : PRIO NUMBER { if ($2 < 0 || $2 > PF_PRIO_MAX) { yyerror("prio must be 0 - %u", PF_PRIO_MAX); YYERROR; } $$.b1 = $$.b2 = $2; } | PRIO '(' NUMBER comma NUMBER ')' { if ($3 < 0 || $3 > PF_PRIO_MAX || $5 < 0 || $5 > PF_PRIO_MAX) { yyerror("prio must be 0 - %u", PF_PRIO_MAX); YYERROR; } $$.b1 = $3; $$.b2 = $5; } ; probability : STRING { char *e; double p = strtod($1, &e); if (*e == '%') { p *= 0.01; e++; } if (*e) { yyerror("invalid probability: %s", $1); free($1); YYERROR; } free($1); $$ = p; } | NUMBER { $$ = (double)$1; } ; action : PASS { $$.b1 = PF_PASS; $$.b2 = failpolicy; $$.w = returnicmpdefault; $$.w2 = returnicmp6default; } | MATCH { $$.b1 = PF_MATCH; $$.b2 = $$.w = 0; } | BLOCK blockspec { $$ = $2; $$.b1 = PF_DROP; } ; blockspec : /* empty */ { $$.b2 = blockpolicy; $$.w = returnicmpdefault; $$.w2 = returnicmp6default; } | DROP { $$.b2 = PFRULE_DROP; $$.w = 0; $$.w2 = 0; } | RETURNRST { $$.b2 = PFRULE_RETURNRST; $$.w = 0; $$.w2 = 0; } | RETURNRST '(' TTL NUMBER ')' { if ($4 < 0 || $4 > 255) { yyerror("illegal ttl value %d", $4); YYERROR; } $$.b2 = PFRULE_RETURNRST; $$.w = $4; $$.w2 = 0; } | RETURNICMP { $$.b2 = PFRULE_RETURNICMP; $$.w = returnicmpdefault; $$.w2 = returnicmp6default; } | RETURNICMP6 { $$.b2 = PFRULE_RETURNICMP; $$.w = returnicmpdefault; $$.w2 = returnicmp6default; } | RETURNICMP '(' reticmpspec ')' { $$.b2 = PFRULE_RETURNICMP; $$.w = $3; $$.w2 = returnicmpdefault; } | RETURNICMP6 '(' reticmp6spec ')' { $$.b2 = PFRULE_RETURNICMP; $$.w = returnicmpdefault; $$.w2 = $3; } | RETURNICMP '(' reticmpspec comma reticmp6spec ')' { $$.b2 = PFRULE_RETURNICMP; $$.w = $3; $$.w2 = $5; } | RETURN { $$.b2 = PFRULE_RETURN; $$.w = returnicmpdefault; $$.w2 = returnicmp6default; } ; reticmpspec : STRING { if (!($$ = parseicmpspec($1, AF_INET))) { free($1); YYERROR; } free($1); } | NUMBER { u_int8_t icmptype; if ($1 < 0 || $1 > 255) { yyerror("invalid icmp code %lu", $1); YYERROR; } icmptype = returnicmpdefault >> 8; $$ = (icmptype << 8 | $1); } ; reticmp6spec : STRING { if (!($$ = parseicmpspec($1, AF_INET6))) { free($1); YYERROR; } free($1); } | NUMBER { u_int8_t icmptype; if ($1 < 0 || $1 > 255) { yyerror("invalid icmp code %lu", $1); YYERROR; } icmptype = returnicmp6default >> 8; $$ = (icmptype << 8 | $1); } ; dir : /* empty */ { $$ = PF_INOUT; } | IN { $$ = PF_IN; } | OUT { $$ = PF_OUT; } ; quick : /* empty */ { $$.quick = 0; } | QUICK { $$.quick = 1; } ; logquick : /* empty */ { $$.log = 0; $$.quick = 0; $$.logif = 0; } | log { $$ = $1; $$.quick = 0; } | QUICK { $$.quick = 1; $$.log = 0; $$.logif = 0; } | log QUICK { $$ = $1; $$.quick = 1; } | QUICK log { $$ = $2; $$.quick = 1; } ; log : LOG { $$.log = PF_LOG; $$.logif = 0; } | LOG '(' logopts ')' { $$.log = PF_LOG | $3.log; $$.logif = $3.logif; } ; logopts : logopt { $$ = $1; } | logopts comma logopt { $$.log = $1.log | $3.log; $$.logif = $3.logif; if ($$.logif == 0) $$.logif = $1.logif; } ; logopt : ALL { $$.log = PF_LOG_ALL; $$.logif = 0; } | MATCHES { $$.log = PF_LOG_MATCHES; $$.logif = 0; } | USER { $$.log = PF_LOG_SOCKET_LOOKUP; $$.logif = 0; } | GROUP { $$.log = PF_LOG_SOCKET_LOOKUP; $$.logif = 0; } | TO string { const char *errstr; u_int i; $$.log = 0; if (strncmp($2, "pflog", 5)) { yyerror("%s: should be a pflog interface", $2); free($2); YYERROR; } i = strtonum($2 + 5, 0, 255, &errstr); if (errstr) { yyerror("%s: %s", $2, errstr); free($2); YYERROR; } free($2); $$.logif = i; } ; interface : /* empty */ { $$ = NULL; } | ON if_item_not { $$ = $2; } | ON '{' optnl if_list '}' { $$ = $4; } ; if_list : if_item_not optnl { $$ = $1; } | if_list comma if_item_not optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; if_item_not : not if_item { $$ = $2; $$->not = $1; } ; if_item : STRING { struct node_host *n; $$ = calloc(1, sizeof(struct node_if)); if ($$ == NULL) err(1, "if_item: calloc"); if (strlcpy($$->ifname, $1, sizeof($$->ifname)) >= sizeof($$->ifname)) { free($1); free($$); yyerror("interface name too long"); YYERROR; } if ((n = ifa_exists($1)) != NULL) $$->ifa_flags = n->ifa_flags; free($1); $$->not = 0; $$->next = NULL; $$->tail = $$; } ; af : /* empty */ { $$ = 0; } | INET { $$ = AF_INET; } | INET6 { $$ = AF_INET6; } ; etherproto : /* empty */ { $$ = NULL; } | PROTO etherproto_item { $$ = $2; } | PROTO '{' optnl etherproto_list '}' { $$ = $4; } ; etherproto_list : etherproto_item optnl { $$ = $1; } | etherproto_list comma etherproto_item optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; etherproto_item : etherprotoval { u_int16_t pr; pr = (u_int16_t)$1; if (pr == 0) { yyerror("proto 0 cannot be used"); YYERROR; } $$ = calloc(1, sizeof(struct node_proto)); if ($$ == NULL) err(1, "proto_item: calloc"); $$->proto = pr; $$->next = NULL; $$->tail = $$; } ; etherprotoval : NUMBER { if ($1 < 0 || $1 > 65565) { yyerror("protocol outside range"); YYERROR; } } | STRING { if (!strncmp($1, "0x", 2)) { if (sscanf($1, "0x%4x", &$$) != 1) { free($1); yyerror("invalid EtherType hex"); YYERROR; } } else { yyerror("Symbolic EtherType not yet supported"); } } ; proto : /* empty */ { $$ = NULL; } | PROTO proto_item { $$ = $2; } | PROTO '{' optnl proto_list '}' { $$ = $4; } ; proto_list : proto_item optnl { $$ = $1; } | proto_list comma proto_item optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; proto_item : protoval { u_int8_t pr; pr = (u_int8_t)$1; if (pr == 0) { yyerror("proto 0 cannot be used"); YYERROR; } $$ = calloc(1, sizeof(struct node_proto)); if ($$ == NULL) err(1, "proto_item: calloc"); $$->proto = pr; $$->next = NULL; $$->tail = $$; } ; protoval : STRING { struct protoent *p; p = getprotobyname($1); if (p == NULL) { yyerror("unknown protocol %s", $1); free($1); YYERROR; } $$ = p->p_proto; free($1); } | NUMBER { if ($1 < 0 || $1 > 255) { yyerror("protocol outside range"); YYERROR; } } ; l3fromto : /* empty */ { bzero(&$$, sizeof($$)); } | L3 fromto { if ($2.src.host != NULL && $2.src.host->addr.type != PF_ADDR_ADDRMASK && $2.src.host->addr.type != PF_ADDR_TABLE) { yyerror("from must be an address or table"); YYERROR; } if ($2.dst.host != NULL && $2.dst.host->addr.type != PF_ADDR_ADDRMASK && $2.dst.host->addr.type != PF_ADDR_TABLE) { yyerror("to must be an address or table"); YYERROR; } $$ = $2; } ; etherfromto : ALL { $$.src = NULL; $$.dst = NULL; } | etherfrom etherto { $$.src = $1.mac; $$.dst = $2.mac; } ; etherfrom : /* emtpy */ { bzero(&$$, sizeof($$)); } | FROM macspec { $$.mac = $2; } ; etherto : /* empty */ { bzero(&$$, sizeof($$)); } | TO macspec { $$.mac = $2; } ; mac : string '/' NUMBER { $$ = node_mac_from_string_masklen($1, $3); free($1); if ($$ == NULL) YYERROR; } | string { if (strchr($1, '&')) { /* mac&mask */ char *mac = strtok($1, "&"); char *mask = strtok(NULL, "&"); $$ = node_mac_from_string_mask(mac, mask); } else { $$ = node_mac_from_string($1); } free($1); if ($$ == NULL) YYERROR; } xmac : not mac { struct node_mac *n; for (n = $2; n != NULL; n = n->next) n->neg = $1; $$ = $2; } ; macspec : xmac { $$ = $1; } | '{' optnl mac_list '}' { $$ = $3; } ; mac_list : xmac optnl { $$ = $1; } | mac_list comma xmac { if ($3 == NULL) $$ = $1; else if ($1 == NULL) $$ = $3; else { $1->tail->next = $3; $1->tail = $3->tail; $$ = $1; } } fromto : ALL { $$.src.host = NULL; $$.src.port = NULL; $$.dst.host = NULL; $$.dst.port = NULL; $$.src_os = NULL; } | from os to { $$.src = $1; $$.src_os = $2; $$.dst = $3; } ; os : /* empty */ { $$ = NULL; } | OS xos { $$ = $2; } | OS '{' optnl os_list '}' { $$ = $4; } ; xos : STRING { $$ = calloc(1, sizeof(struct node_os)); if ($$ == NULL) err(1, "os: calloc"); $$->os = $1; $$->tail = $$; } ; os_list : xos optnl { $$ = $1; } | os_list comma xos optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; from : /* empty */ { $$.host = NULL; $$.port = NULL; } | FROM ipportspec { $$ = $2; } ; to : /* empty */ { $$.host = NULL; $$.port = NULL; } | TO ipportspec { if (disallow_urpf_failed($2.host, "\"urpf-failed\" is " "not permitted in a destination address")) YYERROR; $$ = $2; } ; ipportspec : ipspec { $$.host = $1; $$.port = NULL; } | ipspec PORT portspec { $$.host = $1; $$.port = $3; } | PORT portspec { $$.host = NULL; $$.port = $2; } ; optnl : '\n' optnl | ; ipspec : ANY { $$ = NULL; } | xhost { $$ = $1; } | '{' optnl host_list '}' { $$ = $3; } ; toipspec : TO ipspec { $$ = $2; } | /* empty */ { $$ = NULL; } ; host_list : ipspec optnl { $$ = $1; } | host_list comma ipspec optnl { if ($1 == NULL) { freehostlist($3); $$ = $1; } else if ($3 == NULL) { freehostlist($1); $$ = $3; } else { $1->tail->next = $3; $1->tail = $3->tail; $$ = $1; } } ; xhost : not host { struct node_host *n; for (n = $2; n != NULL; n = n->next) n->not = $1; $$ = $2; } | not NOROUTE { $$ = calloc(1, sizeof(struct node_host)); if ($$ == NULL) err(1, "xhost: calloc"); $$->addr.type = PF_ADDR_NOROUTE; $$->next = NULL; $$->not = $1; $$->tail = $$; } | not URPFFAILED { $$ = calloc(1, sizeof(struct node_host)); if ($$ == NULL) err(1, "xhost: calloc"); $$->addr.type = PF_ADDR_URPFFAILED; $$->next = NULL; $$->not = $1; $$->tail = $$; } ; host : STRING { if (($$ = host($1)) == NULL) { /* error. "any" is handled elsewhere */ free($1); yyerror("could not parse host specification"); YYERROR; } free($1); } | STRING '-' STRING { struct node_host *b, *e; if ((b = host($1)) == NULL || (e = host($3)) == NULL) { free($1); free($3); yyerror("could not parse host specification"); YYERROR; } if (b->af != e->af || b->addr.type != PF_ADDR_ADDRMASK || e->addr.type != PF_ADDR_ADDRMASK || unmask(&b->addr.v.a.mask, b->af) != (b->af == AF_INET ? 32 : 128) || unmask(&e->addr.v.a.mask, e->af) != (e->af == AF_INET ? 32 : 128) || b->next != NULL || b->not || e->next != NULL || e->not) { free(b); free(e); free($1); free($3); yyerror("invalid address range"); YYERROR; } memcpy(&b->addr.v.a.mask, &e->addr.v.a.addr, sizeof(b->addr.v.a.mask)); b->addr.type = PF_ADDR_RANGE; $$ = b; free(e); free($1); free($3); } | STRING '/' NUMBER { char *buf; if (asprintf(&buf, "%s/%lld", $1, (long long)$3) == -1) err(1, "host: asprintf"); free($1); if (($$ = host(buf)) == NULL) { /* error. "any" is handled elsewhere */ free(buf); yyerror("could not parse host specification"); YYERROR; } free(buf); } | NUMBER '/' NUMBER { char *buf; /* ie. for 10/8 parsing */ #ifdef __FreeBSD__ if (asprintf(&buf, "%lld/%lld", (long long)$1, (long long)$3) == -1) #else if (asprintf(&buf, "%lld/%lld", $1, $3) == -1) #endif err(1, "host: asprintf"); if (($$ = host(buf)) == NULL) { /* error. "any" is handled elsewhere */ free(buf); yyerror("could not parse host specification"); YYERROR; } free(buf); } | dynaddr | dynaddr '/' NUMBER { struct node_host *n; if ($3 < 0 || $3 > 128) { yyerror("bit number too big"); YYERROR; } $$ = $1; for (n = $1; n != NULL; n = n->next) set_ipmask(n, $3); } | '<' STRING '>' { if (strlen($2) >= PF_TABLE_NAME_SIZE) { yyerror("table name '%s' too long", $2); free($2); YYERROR; } $$ = calloc(1, sizeof(struct node_host)); if ($$ == NULL) err(1, "host: calloc"); $$->addr.type = PF_ADDR_TABLE; if (strlcpy($$->addr.v.tblname, $2, sizeof($$->addr.v.tblname)) >= sizeof($$->addr.v.tblname)) errx(1, "host: strlcpy"); free($2); $$->next = NULL; $$->tail = $$; } ; number : NUMBER | STRING { u_long ulval; if (atoul($1, &ulval) == -1) { yyerror("%s is not a number", $1); free($1); YYERROR; } else $$ = ulval; free($1); } ; dynaddr : '(' STRING ')' { int flags = 0; char *p, *op; op = $2; if (!isalpha(op[0])) { yyerror("invalid interface name '%s'", op); free(op); YYERROR; } while ((p = strrchr($2, ':')) != NULL) { if (!strcmp(p+1, "network")) flags |= PFI_AFLAG_NETWORK; else if (!strcmp(p+1, "broadcast")) flags |= PFI_AFLAG_BROADCAST; else if (!strcmp(p+1, "peer")) flags |= PFI_AFLAG_PEER; else if (!strcmp(p+1, "0")) flags |= PFI_AFLAG_NOALIAS; else { yyerror("interface %s has bad modifier", $2); free(op); YYERROR; } *p = '\0'; } if (flags & (flags - 1) & PFI_AFLAG_MODEMASK) { free(op); yyerror("illegal combination of " "interface modifiers"); YYERROR; } $$ = calloc(1, sizeof(struct node_host)); if ($$ == NULL) err(1, "address: calloc"); $$->af = 0; set_ipmask($$, 128); $$->addr.type = PF_ADDR_DYNIFTL; $$->addr.iflags = flags; if (strlcpy($$->addr.v.ifname, $2, sizeof($$->addr.v.ifname)) >= sizeof($$->addr.v.ifname)) { free(op); free($$); yyerror("interface name too long"); YYERROR; } free(op); $$->next = NULL; $$->tail = $$; } ; portspec : port_item { $$ = $1; } | '{' optnl port_list '}' { $$ = $3; } ; port_list : port_item optnl { $$ = $1; } | port_list comma port_item optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; port_item : portrange { $$ = calloc(1, sizeof(struct node_port)); if ($$ == NULL) err(1, "port_item: calloc"); $$->port[0] = $1.a; $$->port[1] = $1.b; if ($1.t) $$->op = PF_OP_RRG; else $$->op = PF_OP_EQ; $$->next = NULL; $$->tail = $$; } | unaryop portrange { if ($2.t) { yyerror("':' cannot be used with an other " "port operator"); YYERROR; } $$ = calloc(1, sizeof(struct node_port)); if ($$ == NULL) err(1, "port_item: calloc"); $$->port[0] = $2.a; $$->port[1] = $2.b; $$->op = $1; $$->next = NULL; $$->tail = $$; } | portrange PORTBINARY portrange { if ($1.t || $3.t) { yyerror("':' cannot be used with an other " "port operator"); YYERROR; } $$ = calloc(1, sizeof(struct node_port)); if ($$ == NULL) err(1, "port_item: calloc"); $$->port[0] = $1.a; $$->port[1] = $3.a; $$->op = $2; $$->next = NULL; $$->tail = $$; } ; portplain : numberstring { if (parseport($1, &$$, 0) == -1) { free($1); YYERROR; } free($1); } ; portrange : numberstring { if (parseport($1, &$$, PPORT_RANGE) == -1) { free($1); YYERROR; } free($1); } ; uids : uid_item { $$ = $1; } | '{' optnl uid_list '}' { $$ = $3; } ; uid_list : uid_item optnl { $$ = $1; } | uid_list comma uid_item optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; uid_item : uid { $$ = calloc(1, sizeof(struct node_uid)); if ($$ == NULL) err(1, "uid_item: calloc"); $$->uid[0] = $1; $$->uid[1] = $1; $$->op = PF_OP_EQ; $$->next = NULL; $$->tail = $$; } | unaryop uid { if ($2 == UID_MAX && $1 != PF_OP_EQ && $1 != PF_OP_NE) { yyerror("user unknown requires operator = or " "!="); YYERROR; } $$ = calloc(1, sizeof(struct node_uid)); if ($$ == NULL) err(1, "uid_item: calloc"); $$->uid[0] = $2; $$->uid[1] = $2; $$->op = $1; $$->next = NULL; $$->tail = $$; } | uid PORTBINARY uid { if ($1 == UID_MAX || $3 == UID_MAX) { yyerror("user unknown requires operator = or " "!="); YYERROR; } $$ = calloc(1, sizeof(struct node_uid)); if ($$ == NULL) err(1, "uid_item: calloc"); $$->uid[0] = $1; $$->uid[1] = $3; $$->op = $2; $$->next = NULL; $$->tail = $$; } ; uid : STRING { if (!strcmp($1, "unknown")) $$ = UID_MAX; else { struct passwd *pw; if ((pw = getpwnam($1)) == NULL) { yyerror("unknown user %s", $1); free($1); YYERROR; } $$ = pw->pw_uid; } free($1); } | NUMBER { if ($1 < 0 || $1 >= UID_MAX) { yyerror("illegal uid value %lu", $1); YYERROR; } $$ = $1; } ; gids : gid_item { $$ = $1; } | '{' optnl gid_list '}' { $$ = $3; } ; gid_list : gid_item optnl { $$ = $1; } | gid_list comma gid_item optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; gid_item : gid { $$ = calloc(1, sizeof(struct node_gid)); if ($$ == NULL) err(1, "gid_item: calloc"); $$->gid[0] = $1; $$->gid[1] = $1; $$->op = PF_OP_EQ; $$->next = NULL; $$->tail = $$; } | unaryop gid { if ($2 == GID_MAX && $1 != PF_OP_EQ && $1 != PF_OP_NE) { yyerror("group unknown requires operator = or " "!="); YYERROR; } $$ = calloc(1, sizeof(struct node_gid)); if ($$ == NULL) err(1, "gid_item: calloc"); $$->gid[0] = $2; $$->gid[1] = $2; $$->op = $1; $$->next = NULL; $$->tail = $$; } | gid PORTBINARY gid { if ($1 == GID_MAX || $3 == GID_MAX) { yyerror("group unknown requires operator = or " "!="); YYERROR; } $$ = calloc(1, sizeof(struct node_gid)); if ($$ == NULL) err(1, "gid_item: calloc"); $$->gid[0] = $1; $$->gid[1] = $3; $$->op = $2; $$->next = NULL; $$->tail = $$; } ; gid : STRING { if (!strcmp($1, "unknown")) $$ = GID_MAX; else { struct group *grp; if ((grp = getgrnam($1)) == NULL) { yyerror("unknown group %s", $1); free($1); YYERROR; } $$ = grp->gr_gid; } free($1); } | NUMBER { if ($1 < 0 || $1 >= GID_MAX) { yyerror("illegal gid value %lu", $1); YYERROR; } $$ = $1; } ; flag : STRING { int f; if ((f = parse_flags($1)) < 0) { yyerror("bad flags %s", $1); free($1); YYERROR; } free($1); $$.b1 = f; } ; flags : FLAGS flag '/' flag { $$.b1 = $2.b1; $$.b2 = $4.b1; } | FLAGS '/' flag { $$.b1 = 0; $$.b2 = $3.b1; } | FLAGS ANY { $$.b1 = 0; $$.b2 = 0; } ; icmpspec : ICMPTYPE icmp_item { $$ = $2; } | ICMPTYPE '{' optnl icmp_list '}' { $$ = $4; } | ICMP6TYPE icmp6_item { $$ = $2; } | ICMP6TYPE '{' optnl icmp6_list '}' { $$ = $4; } ; icmp_list : icmp_item optnl { $$ = $1; } | icmp_list comma icmp_item optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; icmp6_list : icmp6_item optnl { $$ = $1; } | icmp6_list comma icmp6_item optnl { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; icmp_item : icmptype { $$ = calloc(1, sizeof(struct node_icmp)); if ($$ == NULL) err(1, "icmp_item: calloc"); $$->type = $1; $$->code = 0; $$->proto = IPPROTO_ICMP; $$->next = NULL; $$->tail = $$; } | icmptype CODE STRING { const struct icmpcodeent *p; if ((p = geticmpcodebyname($1-1, $3, AF_INET)) == NULL) { yyerror("unknown icmp-code %s", $3); free($3); YYERROR; } free($3); $$ = calloc(1, sizeof(struct node_icmp)); if ($$ == NULL) err(1, "icmp_item: calloc"); $$->type = $1; $$->code = p->code + 1; $$->proto = IPPROTO_ICMP; $$->next = NULL; $$->tail = $$; } | icmptype CODE NUMBER { if ($3 < 0 || $3 > 255) { yyerror("illegal icmp-code %lu", $3); YYERROR; } $$ = calloc(1, sizeof(struct node_icmp)); if ($$ == NULL) err(1, "icmp_item: calloc"); $$->type = $1; $$->code = $3 + 1; $$->proto = IPPROTO_ICMP; $$->next = NULL; $$->tail = $$; } ; icmp6_item : icmp6type { $$ = calloc(1, sizeof(struct node_icmp)); if ($$ == NULL) err(1, "icmp_item: calloc"); $$->type = $1; $$->code = 0; $$->proto = IPPROTO_ICMPV6; $$->next = NULL; $$->tail = $$; } | icmp6type CODE STRING { const struct icmpcodeent *p; if ((p = geticmpcodebyname($1-1, $3, AF_INET6)) == NULL) { yyerror("unknown icmp6-code %s", $3); free($3); YYERROR; } free($3); $$ = calloc(1, sizeof(struct node_icmp)); if ($$ == NULL) err(1, "icmp_item: calloc"); $$->type = $1; $$->code = p->code + 1; $$->proto = IPPROTO_ICMPV6; $$->next = NULL; $$->tail = $$; } | icmp6type CODE NUMBER { if ($3 < 0 || $3 > 255) { yyerror("illegal icmp-code %lu", $3); YYERROR; } $$ = calloc(1, sizeof(struct node_icmp)); if ($$ == NULL) err(1, "icmp_item: calloc"); $$->type = $1; $$->code = $3 + 1; $$->proto = IPPROTO_ICMPV6; $$->next = NULL; $$->tail = $$; } ; icmptype : STRING { const struct icmptypeent *p; if ((p = geticmptypebyname($1, AF_INET)) == NULL) { yyerror("unknown icmp-type %s", $1); free($1); YYERROR; } $$ = p->type + 1; free($1); } | NUMBER { if ($1 < 0 || $1 > 255) { yyerror("illegal icmp-type %lu", $1); YYERROR; } $$ = $1 + 1; } ; icmp6type : STRING { const struct icmptypeent *p; if ((p = geticmptypebyname($1, AF_INET6)) == NULL) { yyerror("unknown icmp6-type %s", $1); free($1); YYERROR; } $$ = p->type + 1; free($1); } | NUMBER { if ($1 < 0 || $1 > 255) { yyerror("illegal icmp6-type %lu", $1); YYERROR; } $$ = $1 + 1; } ; tos : STRING { int val; char *end; if (map_tos($1, &val)) $$ = val; else if ($1[0] == '0' && $1[1] == 'x') { errno = 0; $$ = strtoul($1, &end, 16); if (errno || *end != '\0') $$ = 256; } else $$ = 256; /* flag bad argument */ if ($$ < 0 || $$ > 255) { yyerror("illegal tos value %s", $1); free($1); YYERROR; } free($1); } | NUMBER { $$ = $1; if ($$ < 0 || $$ > 255) { yyerror("illegal tos value %s", $1); YYERROR; } } ; sourcetrack : SOURCETRACK { $$ = PF_SRCTRACK; } | SOURCETRACK GLOBAL { $$ = PF_SRCTRACK_GLOBAL; } | SOURCETRACK RULE { $$ = PF_SRCTRACK_RULE; } ; statelock : IFBOUND { $$ = PFRULE_IFBOUND; } | FLOATING { $$ = 0; } ; keep : NO STATE { $$.action = 0; $$.options = NULL; } | KEEP STATE state_opt_spec { $$.action = PF_STATE_NORMAL; $$.options = $3; } | MODULATE STATE state_opt_spec { $$.action = PF_STATE_MODULATE; $$.options = $3; } | SYNPROXY STATE state_opt_spec { $$.action = PF_STATE_SYNPROXY; $$.options = $3; } ; flush : /* empty */ { $$ = 0; } | FLUSH { $$ = PF_FLUSH; } | FLUSH GLOBAL { $$ = PF_FLUSH | PF_FLUSH_GLOBAL; } ; state_opt_spec : '(' state_opt_list ')' { $$ = $2; } | /* empty */ { $$ = NULL; } ; state_opt_list : state_opt_item { $$ = $1; } | state_opt_list comma state_opt_item { $1->tail->next = $3; $1->tail = $3; $$ = $1; } ; state_opt_item : MAXIMUM NUMBER { if ($2 < 0 || $2 > UINT_MAX) { yyerror("only positive values permitted"); YYERROR; } $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_MAX; $$->data.max_states = $2; $$->next = NULL; $$->tail = $$; } | NOSYNC { $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_NOSYNC; $$->next = NULL; $$->tail = $$; } | MAXSRCSTATES NUMBER { if ($2 < 0 || $2 > UINT_MAX) { yyerror("only positive values permitted"); YYERROR; } $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_MAX_SRC_STATES; $$->data.max_src_states = $2; $$->next = NULL; $$->tail = $$; } | MAXSRCCONN NUMBER { if ($2 < 0 || $2 > UINT_MAX) { yyerror("only positive values permitted"); YYERROR; } $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_MAX_SRC_CONN; $$->data.max_src_conn = $2; $$->next = NULL; $$->tail = $$; } | MAXSRCCONNRATE NUMBER '/' NUMBER { if ($2 < 0 || $2 > UINT_MAX || $4 < 0 || $4 > UINT_MAX) { yyerror("only positive values permitted"); YYERROR; } $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_MAX_SRC_CONN_RATE; $$->data.max_src_conn_rate.limit = $2; $$->data.max_src_conn_rate.seconds = $4; $$->next = NULL; $$->tail = $$; } | OVERLOAD '<' STRING '>' flush { if (strlen($3) >= PF_TABLE_NAME_SIZE) { yyerror("table name '%s' too long", $3); free($3); YYERROR; } $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); if (strlcpy($$->data.overload.tblname, $3, PF_TABLE_NAME_SIZE) >= PF_TABLE_NAME_SIZE) errx(1, "state_opt_item: strlcpy"); free($3); $$->type = PF_STATE_OPT_OVERLOAD; $$->data.overload.flush = $5; $$->next = NULL; $$->tail = $$; } | MAXSRCNODES NUMBER { if ($2 < 0 || $2 > UINT_MAX) { yyerror("only positive values permitted"); YYERROR; } $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_MAX_SRC_NODES; $$->data.max_src_nodes = $2; $$->next = NULL; $$->tail = $$; } | sourcetrack { $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_SRCTRACK; $$->data.src_track = $1; $$->next = NULL; $$->tail = $$; } | statelock { $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_STATELOCK; $$->data.statelock = $1; $$->next = NULL; $$->tail = $$; } | SLOPPY { $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_SLOPPY; $$->next = NULL; $$->tail = $$; } | PFLOW { $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_PFLOW; $$->next = NULL; $$->tail = $$; } | STRING NUMBER { int i; if ($2 < 0 || $2 > UINT_MAX) { yyerror("only positive values permitted"); YYERROR; } for (i = 0; pf_timeouts[i].name && strcmp(pf_timeouts[i].name, $1); ++i) ; /* nothing */ if (!pf_timeouts[i].name) { yyerror("illegal timeout name %s", $1); free($1); YYERROR; } if (strchr(pf_timeouts[i].name, '.') == NULL) { yyerror("illegal state timeout %s", $1); free($1); YYERROR; } free($1); $$ = calloc(1, sizeof(struct node_state_opt)); if ($$ == NULL) err(1, "state_opt_item: calloc"); $$->type = PF_STATE_OPT_TIMEOUT; $$->data.timeout.number = pf_timeouts[i].timeout; $$->data.timeout.seconds = $2; $$->next = NULL; $$->tail = $$; } ; label : LABEL STRING { $$ = $2; } ; etherqname : QUEUE STRING { $$.qname = $2; } | QUEUE '(' STRING ')' { $$.qname = $3; } ; qname : QUEUE STRING { $$.qname = $2; $$.pqname = NULL; } | QUEUE '(' STRING ')' { $$.qname = $3; $$.pqname = NULL; } | QUEUE '(' STRING comma STRING ')' { $$.qname = $3; $$.pqname = $5; } ; no : /* empty */ { $$ = 0; } | NO { $$ = 1; } ; portstar : numberstring { if (parseport($1, &$$, PPORT_RANGE|PPORT_STAR) == -1) { free($1); YYERROR; } free($1); } ; redirspec : host { $$ = $1; } | '{' optnl redir_host_list '}' { $$ = $3; } ; redir_host_list : host optnl { $$ = $1; } | redir_host_list comma host optnl { $1->tail->next = $3; $1->tail = $3->tail; $$ = $1; } ; redirpool : /* empty */ { $$ = NULL; } | ARROW redirspec { $$ = calloc(1, sizeof(struct redirection)); if ($$ == NULL) err(1, "redirection: calloc"); $$->host = $2; $$->rport.a = $$->rport.b = $$->rport.t = 0; } | ARROW redirspec PORT portstar { $$ = calloc(1, sizeof(struct redirection)); if ($$ == NULL) err(1, "redirection: calloc"); $$->host = $2; $$->rport = $4; } ; hashkey : /* empty */ { $$ = calloc(1, sizeof(struct pf_poolhashkey)); if ($$ == NULL) err(1, "hashkey: calloc"); $$->key32[0] = arc4random(); $$->key32[1] = arc4random(); $$->key32[2] = arc4random(); $$->key32[3] = arc4random(); } | string { if (!strncmp($1, "0x", 2)) { if (strlen($1) != 34) { free($1); yyerror("hex key must be 128 bits " "(32 hex digits) long"); YYERROR; } $$ = calloc(1, sizeof(struct pf_poolhashkey)); if ($$ == NULL) err(1, "hashkey: calloc"); if (sscanf($1, "0x%8x%8x%8x%8x", &$$->key32[0], &$$->key32[1], &$$->key32[2], &$$->key32[3]) != 4) { free($$); free($1); yyerror("invalid hex key"); YYERROR; } } else { MD5_CTX context; $$ = calloc(1, sizeof(struct pf_poolhashkey)); if ($$ == NULL) err(1, "hashkey: calloc"); MD5Init(&context); MD5Update(&context, (unsigned char *)$1, strlen($1)); MD5Final((unsigned char *)$$, &context); HTONL($$->key32[0]); HTONL($$->key32[1]); HTONL($$->key32[2]); HTONL($$->key32[3]); } free($1); } ; pool_opts : { bzero(&pool_opts, sizeof pool_opts); } pool_opts_l { $$ = pool_opts; } | /* empty */ { bzero(&pool_opts, sizeof pool_opts); $$ = pool_opts; } ; pool_opts_l : pool_opts_l pool_opt | pool_opt ; pool_opt : BITMASK { if (pool_opts.type) { yyerror("pool type cannot be redefined"); YYERROR; } pool_opts.type = PF_POOL_BITMASK; } | RANDOM { if (pool_opts.type) { yyerror("pool type cannot be redefined"); YYERROR; } pool_opts.type = PF_POOL_RANDOM; } | SOURCEHASH hashkey { if (pool_opts.type) { yyerror("pool type cannot be redefined"); YYERROR; } pool_opts.type = PF_POOL_SRCHASH; pool_opts.key = $2; } | ROUNDROBIN { if (pool_opts.type) { yyerror("pool type cannot be redefined"); YYERROR; } pool_opts.type = PF_POOL_ROUNDROBIN; } | STATICPORT { if (pool_opts.staticport) { yyerror("static-port cannot be redefined"); YYERROR; } pool_opts.staticport = 1; } | STICKYADDRESS { if (pool_opts.marker & POM_STICKYADDRESS) { yyerror("sticky-address cannot be redefined"); YYERROR; } pool_opts.marker |= POM_STICKYADDRESS; pool_opts.opts |= PF_POOL_STICKYADDR; } | ENDPI { if (pool_opts.marker & POM_ENDPI) { yyerror("endpoint-independent cannot be redefined"); YYERROR; } pool_opts.marker |= POM_ENDPI; pool_opts.opts |= PF_POOL_ENDPI; } | MAPEPORTSET number '/' number '/' number { if (pool_opts.mape.offset) { yyerror("map-e-portset cannot be redefined"); YYERROR; } if (pool_opts.type) { yyerror("map-e-portset cannot be used with " "address pools"); YYERROR; } if ($2 <= 0 || $2 >= 16) { yyerror("MAP-E PSID offset must be 1-15"); YYERROR; } if ($4 < 0 || $4 >= 16 || $2 + $4 > 16) { yyerror("Invalid MAP-E PSID length"); YYERROR; } else if ($4 == 0) { yyerror("PSID Length = 0: this means" " you do not need MAP-E"); YYERROR; } if ($6 < 0 || $6 > 65535) { yyerror("Invalid MAP-E PSID"); YYERROR; } pool_opts.mape.offset = $2; pool_opts.mape.psidlen = $4; pool_opts.mape.psid = $6; } ; redirection : /* empty */ { $$ = NULL; } | ARROW host { $$ = calloc(1, sizeof(struct redirection)); if ($$ == NULL) err(1, "redirection: calloc"); $$->host = $2; $$->rport.a = $$->rport.b = $$->rport.t = 0; } | ARROW host PORT portstar { $$ = calloc(1, sizeof(struct redirection)); if ($$ == NULL) err(1, "redirection: calloc"); $$->host = $2; $$->rport = $4; } ; natpasslog : /* empty */ { $$.b1 = $$.b2 = 0; $$.w2 = 0; } | PASS { $$.b1 = 1; $$.b2 = 0; $$.w2 = 0; } | PASS log { $$.b1 = 1; $$.b2 = $2.log; $$.w2 = $2.logif; } | log { $$.b1 = 0; $$.b2 = $1.log; $$.w2 = $1.logif; } ; nataction : no NAT natpasslog { if ($1 && $3.b1) { yyerror("\"pass\" not valid with \"no\""); YYERROR; } if ($1) $$.b1 = PF_NONAT; else $$.b1 = PF_NAT; $$.b2 = $3.b1; $$.w = $3.b2; $$.w2 = $3.w2; } | no RDR natpasslog { if ($1 && $3.b1) { yyerror("\"pass\" not valid with \"no\""); YYERROR; } if ($1) $$.b1 = PF_NORDR; else $$.b1 = PF_RDR; $$.b2 = $3.b1; $$.w = $3.b2; $$.w2 = $3.w2; } ; natrule : nataction interface af proto fromto tag tagged rtable redirpool pool_opts { struct pfctl_rule r; struct node_state_opt *o; if (check_rulestate(PFCTL_STATE_NAT)) YYERROR; memset(&r, 0, sizeof(r)); r.action = $1.b1; r.natpass = $1.b2; r.log = $1.w; r.logif = $1.w2; r.af = $3; if (!r.af) { if ($5.src.host && $5.src.host->af && !$5.src.host->ifindex) r.af = $5.src.host->af; else if ($5.dst.host && $5.dst.host->af && !$5.dst.host->ifindex) r.af = $5.dst.host->af; } if ($6 != NULL) if (strlcpy(r.tagname, $6, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } if ($7.name) if (strlcpy(r.match_tagname, $7.name, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } r.match_tag_not = $7.neg; r.rtableid = $8; if (r.action == PF_NONAT || r.action == PF_NORDR) { if ($9 != NULL) { yyerror("translation rule with 'no' " "does not need '->'"); YYERROR; } } else { if ($9 == NULL || $9->host == NULL) { yyerror("translation rule requires '-> " "address'"); YYERROR; } if (!r.af && ! $9->host->ifindex) r.af = $9->host->af; remove_invalid_hosts(&$9->host, &r.af); if (invalid_redirect($9->host, r.af)) YYERROR; if ($9->host->addr.type == PF_ADDR_DYNIFTL) { if (($9->host = gen_dynnode($9->host, r.af)) == NULL) err(1, "calloc"); } if (check_netmask($9->host, r.af)) YYERROR; r.rpool.proxy_port[0] = ntohs($9->rport.a); switch (r.action) { case PF_RDR: if (!$9->rport.b && $9->rport.t && $5.dst.port != NULL) { r.rpool.proxy_port[1] = ntohs($9->rport.a) + (ntohs( $5.dst.port->port[1]) - ntohs( $5.dst.port->port[0])); } else r.rpool.proxy_port[1] = ntohs($9->rport.b); break; case PF_NAT: r.rpool.proxy_port[1] = ntohs($9->rport.b); if (!r.rpool.proxy_port[0] && !r.rpool.proxy_port[1]) { r.rpool.proxy_port[0] = PF_NAT_PROXY_PORT_LOW; r.rpool.proxy_port[1] = PF_NAT_PROXY_PORT_HIGH; } else if (!r.rpool.proxy_port[1]) r.rpool.proxy_port[1] = r.rpool.proxy_port[0]; break; default: break; } r.rpool.opts = $10.type; if ((r.rpool.opts & PF_POOL_TYPEMASK) == PF_POOL_NONE && ($9->host->next != NULL || $9->host->addr.type == PF_ADDR_TABLE || DYNIF_MULTIADDR($9->host->addr))) r.rpool.opts = PF_POOL_ROUNDROBIN; if ((r.rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN && disallow_table($9->host, "tables are only " "supported in round-robin redirection " "pools")) YYERROR; if ((r.rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN && disallow_alias($9->host, "interface (%s) " "is only supported in round-robin " "redirection pools")) YYERROR; if ($9->host->next != NULL) { if ((r.rpool.opts & PF_POOL_TYPEMASK) != PF_POOL_ROUNDROBIN) { yyerror("only round-robin " "valid for multiple " "redirection addresses"); YYERROR; } } } if ($10.key != NULL) memcpy(&r.rpool.key, $10.key, sizeof(struct pf_poolhashkey)); if ($10.opts) r.rpool.opts |= $10.opts; if ($10.staticport) { if (r.action != PF_NAT) { yyerror("the 'static-port' option is " "only valid with nat rules"); YYERROR; } if (r.rpool.proxy_port[0] != PF_NAT_PROXY_PORT_LOW && r.rpool.proxy_port[1] != PF_NAT_PROXY_PORT_HIGH) { yyerror("the 'static-port' option can't" " be used when specifying a port" " range"); YYERROR; } r.rpool.proxy_port[0] = 0; r.rpool.proxy_port[1] = 0; } if ($10.mape.offset) { if (r.action != PF_NAT) { yyerror("the 'map-e-portset' option is" " only valid with nat rules"); YYERROR; } if ($10.staticport) { yyerror("the 'map-e-portset' option" " can't be used 'static-port'"); YYERROR; } if (r.rpool.proxy_port[0] != PF_NAT_PROXY_PORT_LOW && r.rpool.proxy_port[1] != PF_NAT_PROXY_PORT_HIGH) { yyerror("the 'map-e-portset' option" " can't be used when specifying" " a port range"); YYERROR; } r.rpool.mape = $10.mape; } o = keep_state_defaults; while (o) { switch (o->type) { case PF_STATE_OPT_PFLOW: if (r.rule_flag & PFRULE_PFLOW) { yyerror("state pflow option: " "multiple definitions"); YYERROR; } r.rule_flag |= PFRULE_PFLOW; break; } o = o->next; } expand_rule(&r, $2, $9 == NULL ? NULL : $9->host, $4, $5.src_os, $5.src.host, $5.src.port, $5.dst.host, $5.dst.port, 0, 0, 0, 0, ""); free($9); } ; binatrule : no BINAT natpasslog interface af proto FROM ipspec toipspec tag tagged rtable redirection { struct pfctl_rule binat; struct pf_pooladdr *pa; if (check_rulestate(PFCTL_STATE_NAT)) YYERROR; if (disallow_urpf_failed($9, "\"urpf-failed\" is not " "permitted as a binat destination")) YYERROR; memset(&binat, 0, sizeof(binat)); if ($1 && $3.b1) { yyerror("\"pass\" not valid with \"no\""); YYERROR; } if ($1) binat.action = PF_NOBINAT; else binat.action = PF_BINAT; binat.natpass = $3.b1; binat.log = $3.b2; binat.logif = $3.w2; binat.af = $5; if (!binat.af && $8 != NULL && $8->af) binat.af = $8->af; if (!binat.af && $9 != NULL && $9->af) binat.af = $9->af; if (!binat.af && $13 != NULL && $13->host) binat.af = $13->host->af; if (!binat.af) { yyerror("address family (inet/inet6) " "undefined"); YYERROR; } if ($4 != NULL) { memcpy(binat.ifname, $4->ifname, sizeof(binat.ifname)); binat.ifnot = $4->not; free($4); } if ($10 != NULL) if (strlcpy(binat.tagname, $10, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } if ($11.name) if (strlcpy(binat.match_tagname, $11.name, PF_TAG_NAME_SIZE) >= PF_TAG_NAME_SIZE) { yyerror("tag too long, max %u chars", PF_TAG_NAME_SIZE - 1); YYERROR; } binat.match_tag_not = $11.neg; binat.rtableid = $12; if ($6 != NULL) { binat.proto = $6->proto; free($6); } if ($8 != NULL && disallow_table($8, "invalid use of " "table <%s> as the source address of a binat rule")) YYERROR; if ($8 != NULL && disallow_alias($8, "invalid use of " "interface (%s) as the source address of a binat " "rule")) YYERROR; if ($13 != NULL && $13->host != NULL && disallow_table( $13->host, "invalid use of table <%s> as the " "redirect address of a binat rule")) YYERROR; if ($13 != NULL && $13->host != NULL && disallow_alias( $13->host, "invalid use of interface (%s) as the " "redirect address of a binat rule")) YYERROR; if ($8 != NULL) { if ($8->next) { yyerror("multiple binat ip addresses"); YYERROR; } if ($8->addr.type == PF_ADDR_DYNIFTL) $8->af = binat.af; if ($8->af != binat.af) { yyerror("binat ip versions must match"); YYERROR; } if ($8->addr.type == PF_ADDR_DYNIFTL) { if (($8 = gen_dynnode($8, binat.af)) == NULL) err(1, "calloc"); } if (check_netmask($8, binat.af)) YYERROR; memcpy(&binat.src.addr, &$8->addr, sizeof(binat.src.addr)); free($8); } if ($9 != NULL) { if ($9->next) { yyerror("multiple binat ip addresses"); YYERROR; } if ($9->af != binat.af && $9->af) { yyerror("binat ip versions must match"); YYERROR; } if ($9->addr.type == PF_ADDR_DYNIFTL) { if (($9 = gen_dynnode($9, binat.af)) == NULL) err(1, "calloc"); } if (check_netmask($9, binat.af)) YYERROR; memcpy(&binat.dst.addr, &$9->addr, sizeof(binat.dst.addr)); binat.dst.neg = $9->not; free($9); } if (binat.action == PF_NOBINAT) { if ($13 != NULL) { yyerror("'no binat' rule does not need" " '->'"); YYERROR; } } else { if ($13 == NULL || $13->host == NULL) { yyerror("'binat' rule requires" " '-> address'"); YYERROR; } remove_invalid_hosts(&$13->host, &binat.af); if (invalid_redirect($13->host, binat.af)) YYERROR; if ($13->host->next != NULL) { yyerror("binat rule must redirect to " "a single address"); YYERROR; } if ($13->host->addr.type == PF_ADDR_DYNIFTL) { if (($13->host = gen_dynnode($13->host, binat.af)) == NULL) err(1, "calloc"); } if (check_netmask($13->host, binat.af)) YYERROR; if (!PF_AZERO(&binat.src.addr.v.a.mask, binat.af) && !PF_AEQ(&binat.src.addr.v.a.mask, &$13->host->addr.v.a.mask, binat.af)) { yyerror("'binat' source mask and " "redirect mask must be the same"); YYERROR; } TAILQ_INIT(&binat.rpool.list); pa = calloc(1, sizeof(struct pf_pooladdr)); if (pa == NULL) err(1, "binat: calloc"); pa->addr = $13->host->addr; pa->ifname[0] = 0; TAILQ_INSERT_TAIL(&binat.rpool.list, pa, entries); free($13); } pfctl_append_rule(pf, &binat, ""); } ; tag : /* empty */ { $$ = NULL; } | TAG STRING { $$ = $2; } ; tagged : /* empty */ { $$.neg = 0; $$.name = NULL; } | not TAGGED string { $$.neg = $1; $$.name = $3; } ; rtable : /* empty */ { $$ = -1; } | RTABLE NUMBER { if ($2 < 0 || $2 > rt_tableid_max()) { yyerror("invalid rtable id"); YYERROR; } $$ = $2; } ; route_host : STRING { $$ = calloc(1, sizeof(struct node_host)); if ($$ == NULL) err(1, "route_host: calloc"); if (strlen($1) >= IFNAMSIZ) { yyerror("interface name too long"); YYERROR; } $$->ifname = strdup($1); set_ipmask($$, 128); $$->next = NULL; $$->tail = $$; } | '(' STRING host ')' { struct node_host *n; $$ = $3; for (n = $3; n != NULL; n = n->next) { if (strlen($2) >= IFNAMSIZ) { yyerror("interface name too long"); YYERROR; } n->ifname = strdup($2); } } ; route_host_list : route_host optnl { $$ = $1; } | route_host_list comma route_host optnl { if ($1->af == 0) $1->af = $3->af; if ($1->af != $3->af) { yyerror("all pool addresses must be in the " "same address family"); YYERROR; } $1->tail->next = $3; $1->tail = $3->tail; $$ = $1; } ; routespec : route_host { $$ = $1; } | '{' optnl route_host_list '}' { $$ = $3; } ; route : /* empty */ { $$.host = NULL; $$.rt = 0; $$.pool_opts = 0; } | FASTROUTE { /* backwards-compat */ $$.host = NULL; $$.rt = 0; $$.pool_opts = 0; } | ROUTETO routespec pool_opts { $$.host = $2; $$.rt = PF_ROUTETO; $$.pool_opts = $3.type | $3.opts; if ($3.key != NULL) $$.key = $3.key; } | REPLYTO routespec pool_opts { $$.host = $2; $$.rt = PF_REPLYTO; $$.pool_opts = $3.type | $3.opts; if ($3.key != NULL) $$.key = $3.key; } | DUPTO routespec pool_opts { $$.host = $2; $$.rt = PF_DUPTO; $$.pool_opts = $3.type | $3.opts; if ($3.key != NULL) $$.key = $3.key; } ; timeout_spec : STRING NUMBER { if (check_rulestate(PFCTL_STATE_OPTION)) { free($1); YYERROR; } if ($2 < 0 || $2 > UINT_MAX) { yyerror("only positive values permitted"); YYERROR; } if (pfctl_apply_timeout(pf, $1, $2, 0) != 0) { yyerror("unknown timeout %s", $1); free($1); YYERROR; } free($1); } | INTERVAL NUMBER { if (check_rulestate(PFCTL_STATE_OPTION)) YYERROR; if ($2 < 0 || $2 > UINT_MAX) { yyerror("only positive values permitted"); YYERROR; } if (pfctl_apply_timeout(pf, "interval", $2, 0) != 0) YYERROR; } ; timeout_list : timeout_list comma timeout_spec optnl | timeout_spec optnl ; limit_spec : STRING NUMBER { if (check_rulestate(PFCTL_STATE_OPTION)) { free($1); YYERROR; } if ($2 < 0 || $2 > UINT_MAX) { yyerror("only positive values permitted"); YYERROR; } if (pfctl_apply_limit(pf, $1, $2) != 0) { yyerror("unable to set limit %s %u", $1, $2); free($1); YYERROR; } free($1); } ; limit_list : limit_list comma limit_spec optnl | limit_spec optnl ; comma : ',' | /* empty */ ; yesno : NO { $$ = 0; } | STRING { if (!strcmp($1, "yes")) $$ = 1; else { yyerror("invalid value '%s', expected 'yes' " "or 'no'", $1); free($1); YYERROR; } free($1); } ; unaryop : '=' { $$ = PF_OP_EQ; } | NE { $$ = PF_OP_NE; } | LE { $$ = PF_OP_LE; } | '<' { $$ = PF_OP_LT; } | GE { $$ = PF_OP_GE; } | '>' { $$ = PF_OP_GT; } ; %% int yyerror(const char *fmt, ...) { va_list ap; file->errors++; va_start(ap, fmt); fprintf(stderr, "%s:%d: ", file->name, yylval.lineno); vfprintf(stderr, fmt, ap); fprintf(stderr, "\n"); va_end(ap); return (0); } int disallow_table(struct node_host *h, const char *fmt) { for (; h != NULL; h = h->next) if (h->addr.type == PF_ADDR_TABLE) { yyerror(fmt, h->addr.v.tblname); return (1); } return (0); } int disallow_urpf_failed(struct node_host *h, const char *fmt) { for (; h != NULL; h = h->next) if (h->addr.type == PF_ADDR_URPFFAILED) { yyerror(fmt); return (1); } return (0); } int disallow_alias(struct node_host *h, const char *fmt) { for (; h != NULL; h = h->next) if (DYNIF_MULTIADDR(h->addr)) { yyerror(fmt, h->addr.v.tblname); return (1); } return (0); } int rule_consistent(struct pfctl_rule *r, int anchor_call) { int problems = 0; switch (r->action) { case PF_PASS: case PF_MATCH: case PF_DROP: case PF_SCRUB: case PF_NOSCRUB: problems = filter_consistent(r, anchor_call); break; case PF_NAT: case PF_NONAT: problems = nat_consistent(r); break; case PF_RDR: case PF_NORDR: problems = rdr_consistent(r); break; case PF_BINAT: case PF_NOBINAT: default: break; } return (problems); } int filter_consistent(struct pfctl_rule *r, int anchor_call) { int problems = 0; if (r->proto != IPPROTO_TCP && r->proto != IPPROTO_UDP && r->proto != IPPROTO_SCTP && (r->src.port_op || r->dst.port_op)) { yyerror("port only applies to tcp/udp/sctp"); problems++; } if (r->proto != IPPROTO_ICMP && r->proto != IPPROTO_ICMPV6 && (r->type || r->code)) { yyerror("icmp-type/code only applies to icmp"); problems++; } if (!r->af && (r->type || r->code)) { yyerror("must indicate address family with icmp-type/code"); problems++; } if (r->overload_tblname[0] && r->max_src_conn == 0 && r->max_src_conn_rate.seconds == 0) { yyerror("'overload' requires 'max-src-conn' " "or 'max-src-conn-rate'"); problems++; } if ((r->proto == IPPROTO_ICMP && r->af == AF_INET6) || (r->proto == IPPROTO_ICMPV6 && r->af == AF_INET)) { yyerror("proto %s doesn't match address family %s", r->proto == IPPROTO_ICMP ? "icmp" : "icmp6", r->af == AF_INET ? "inet" : "inet6"); problems++; } if (r->allow_opts && r->action != PF_PASS) { yyerror("allow-opts can only be specified for pass rules"); problems++; } if (r->rule_flag & PFRULE_FRAGMENT && (r->src.port_op || r->dst.port_op || r->flagset || r->type || r->code)) { yyerror("fragments can be filtered only on IP header fields"); problems++; } if (r->rule_flag & PFRULE_RETURNRST && r->proto != IPPROTO_TCP) { yyerror("return-rst can only be applied to TCP rules"); problems++; } if (r->max_src_nodes && !(r->rule_flag & PFRULE_RULESRCTRACK)) { yyerror("max-src-nodes requires 'source-track rule'"); problems++; } if (r->action != PF_PASS && r->keep_state) { yyerror("keep state is great, but only for pass rules"); problems++; } if (r->rule_flag & PFRULE_STATESLOPPY && (r->keep_state == PF_STATE_MODULATE || r->keep_state == PF_STATE_SYNPROXY)) { yyerror("sloppy state matching cannot be used with " "synproxy state or modulate state"); problems++; } /* match rules rules */ if (r->action == PF_MATCH) { if (r->divert.port) { yyerror("divert is not supported on match rules"); problems++; } if (r->rt) { yyerror("route-to, reply-to, dup-to and fastroute " "must not be used on match rules"); problems++; } } if (r->rpool.opts & PF_POOL_STICKYADDR && !r->keep_state) { yyerror("'sticky-address' requires 'keep state'"); problems++; } return (-problems); } int nat_consistent(struct pfctl_rule *r) { return (0); /* yeah! */ } int rdr_consistent(struct pfctl_rule *r) { int problems = 0; if (r->proto != IPPROTO_TCP && r->proto != IPPROTO_UDP && r->proto != IPPROTO_SCTP) { if (r->src.port_op) { yyerror("src port only applies to tcp/udp/sctp"); problems++; } if (r->dst.port_op) { yyerror("dst port only applies to tcp/udp/sctp"); problems++; } if (r->rpool.proxy_port[0]) { yyerror("rpool port only applies to tcp/udp/sctp"); problems++; } } if (r->dst.port_op && r->dst.port_op != PF_OP_EQ && r->dst.port_op != PF_OP_RRG) { yyerror("invalid port operator for rdr destination port"); problems++; } return (-problems); } int process_tabledef(char *name, struct table_opts *opts) { struct pfr_buffer ab; struct node_tinit *ti; unsigned long maxcount; size_t s = sizeof(maxcount); bzero(&ab, sizeof(ab)); ab.pfrb_type = PFRB_ADDRS; SIMPLEQ_FOREACH(ti, &opts->init_nodes, entries) { if (ti->file) if (pfr_buf_load(&ab, ti->file, 0, append_addr)) { if (errno) yyerror("cannot load \"%s\": %s", ti->file, strerror(errno)); else yyerror("file \"%s\" contains bad data", ti->file); goto _error; } if (ti->host) if (append_addr_host(&ab, ti->host, 0, 0)) { yyerror("cannot create address buffer: %s", strerror(errno)); goto _error; } } if (pf->opts & PF_OPT_VERBOSE) print_tabledef(name, opts->flags, opts->init_addr, &opts->init_nodes); if (!(pf->opts & PF_OPT_NOACTION) && pfctl_define_table(name, opts->flags, opts->init_addr, pf->anchor->name, &ab, pf->anchor->ruleset.tticket)) { if (sysctlbyname("net.pf.request_maxcount", &maxcount, &s, NULL, 0) == -1) maxcount = 65535; if (ab.pfrb_size > maxcount) yyerror("cannot define table %s: too many elements.\n" "Consider increasing net.pf.request_maxcount.", name); else yyerror("cannot define table %s: %s", name, pfr_strerror(errno)); goto _error; } pf->tdirty = 1; pfr_buf_clear(&ab); return (0); _error: pfr_buf_clear(&ab); return (-1); } struct keywords { const char *k_name; int k_val; }; /* macro gore, but you should've seen the prior indentation nightmare... */ #define FREE_LIST(T,r) \ do { \ T *p, *node = r; \ while (node != NULL) { \ p = node; \ node = node->next; \ free(p); \ } \ } while (0) #define LOOP_THROUGH(T,n,r,C) \ do { \ T *n; \ if (r == NULL) { \ r = calloc(1, sizeof(T)); \ if (r == NULL) \ err(1, "LOOP: calloc"); \ r->next = NULL; \ } \ n = r; \ while (n != NULL) { \ do { \ C; \ } while (0); \ n = n->next; \ } \ } while (0) void expand_label_str(char *label, size_t len, const char *srch, const char *repl) { char *tmp; char *p, *q; if ((tmp = calloc(1, len)) == NULL) err(1, "expand_label_str: calloc"); p = q = label; while ((q = strstr(p, srch)) != NULL) { *q = '\0'; if ((strlcat(tmp, p, len) >= len) || (strlcat(tmp, repl, len) >= len)) errx(1, "expand_label: label too long"); q += strlen(srch); p = q; } if (strlcat(tmp, p, len) >= len) errx(1, "expand_label: label too long"); strlcpy(label, tmp, len); /* always fits */ free(tmp); } void expand_label_if(const char *name, char *label, size_t len, const char *ifname) { if (strstr(label, name) != NULL) { if (!*ifname) expand_label_str(label, len, name, "any"); else expand_label_str(label, len, name, ifname); } } void expand_label_addr(const char *name, char *label, size_t len, sa_family_t af, struct pf_rule_addr *addr) { char tmp[64], tmp_not[66]; if (strstr(label, name) != NULL) { switch (addr->addr.type) { case PF_ADDR_DYNIFTL: snprintf(tmp, sizeof(tmp), "(%s)", addr->addr.v.ifname); break; case PF_ADDR_TABLE: snprintf(tmp, sizeof(tmp), "<%s>", addr->addr.v.tblname); break; case PF_ADDR_NOROUTE: snprintf(tmp, sizeof(tmp), "no-route"); break; case PF_ADDR_URPFFAILED: snprintf(tmp, sizeof(tmp), "urpf-failed"); break; case PF_ADDR_ADDRMASK: if (!af || (PF_AZERO(&addr->addr.v.a.addr, af) && PF_AZERO(&addr->addr.v.a.mask, af))) snprintf(tmp, sizeof(tmp), "any"); else { char a[48]; int bits; if (inet_ntop(af, &addr->addr.v.a.addr, a, sizeof(a)) == NULL) snprintf(tmp, sizeof(tmp), "?"); else { bits = unmask(&addr->addr.v.a.mask, af); if ((af == AF_INET && bits < 32) || (af == AF_INET6 && bits < 128)) snprintf(tmp, sizeof(tmp), "%s/%d", a, bits); else snprintf(tmp, sizeof(tmp), "%s", a); } } break; default: snprintf(tmp, sizeof(tmp), "?"); break; } if (addr->neg) { snprintf(tmp_not, sizeof(tmp_not), "! %s", tmp); expand_label_str(label, len, name, tmp_not); } else expand_label_str(label, len, name, tmp); } } void expand_label_port(const char *name, char *label, size_t len, struct pf_rule_addr *addr) { char a1[6], a2[6], op[13] = ""; if (strstr(label, name) != NULL) { snprintf(a1, sizeof(a1), "%u", ntohs(addr->port[0])); snprintf(a2, sizeof(a2), "%u", ntohs(addr->port[1])); if (!addr->port_op) ; else if (addr->port_op == PF_OP_IRG) snprintf(op, sizeof(op), "%s><%s", a1, a2); else if (addr->port_op == PF_OP_XRG) snprintf(op, sizeof(op), "%s<>%s", a1, a2); else if (addr->port_op == PF_OP_EQ) snprintf(op, sizeof(op), "%s", a1); else if (addr->port_op == PF_OP_NE) snprintf(op, sizeof(op), "!=%s", a1); else if (addr->port_op == PF_OP_LT) snprintf(op, sizeof(op), "<%s", a1); else if (addr->port_op == PF_OP_LE) snprintf(op, sizeof(op), "<=%s", a1); else if (addr->port_op == PF_OP_GT) snprintf(op, sizeof(op), ">%s", a1); else if (addr->port_op == PF_OP_GE) snprintf(op, sizeof(op), ">=%s", a1); expand_label_str(label, len, name, op); } } void expand_label_proto(const char *name, char *label, size_t len, u_int8_t proto) { const char *protoname; char n[4]; if (strstr(label, name) != NULL) { protoname = pfctl_proto2name(proto); if (protoname != NULL) expand_label_str(label, len, name, protoname); else { snprintf(n, sizeof(n), "%u", proto); expand_label_str(label, len, name, n); } } } void expand_label_nr(const char *name, char *label, size_t len, struct pfctl_rule *r) { char n[11]; if (strstr(label, name) != NULL) { snprintf(n, sizeof(n), "%u", r->nr); expand_label_str(label, len, name, n); } } void expand_label(char *label, size_t len, struct pfctl_rule *r) { expand_label_if("$if", label, len, r->ifname); expand_label_addr("$srcaddr", label, len, r->af, &r->src); expand_label_addr("$dstaddr", label, len, r->af, &r->dst); expand_label_port("$srcport", label, len, &r->src); expand_label_port("$dstport", label, len, &r->dst); expand_label_proto("$proto", label, len, r->proto); expand_label_nr("$nr", label, len, r); } int expand_altq(struct pf_altq *a, struct node_if *interfaces, struct node_queue *nqueues, struct node_queue_bw bwspec, struct node_queue_opt *opts) { struct pf_altq pa, pb; char qname[PF_QNAME_SIZE]; struct node_queue *n; struct node_queue_bw bw; int errs = 0; if ((pf->loadopt & PFCTL_FLAG_ALTQ) == 0) { FREE_LIST(struct node_if, interfaces); if (nqueues) FREE_LIST(struct node_queue, nqueues); return (0); } LOOP_THROUGH(struct node_if, interface, interfaces, memcpy(&pa, a, sizeof(struct pf_altq)); if (strlcpy(pa.ifname, interface->ifname, sizeof(pa.ifname)) >= sizeof(pa.ifname)) errx(1, "expand_altq: strlcpy"); if (interface->not) { yyerror("altq on ! is not supported"); errs++; } else { if (eval_pfaltq(pf, &pa, &bwspec, opts)) errs++; else if (pfctl_add_altq(pf, &pa)) errs++; if (pf->opts & PF_OPT_VERBOSE) { print_altq(&pf->paltq->altq, 0, &bwspec, opts); if (nqueues && nqueues->tail) { printf("queue { "); LOOP_THROUGH(struct node_queue, queue, nqueues, printf("%s ", queue->queue); ); printf("}"); } printf("\n"); } if (pa.scheduler == ALTQT_CBQ || pa.scheduler == ALTQT_HFSC || pa.scheduler == ALTQT_FAIRQ) { /* now create a root queue */ memset(&pb, 0, sizeof(struct pf_altq)); if (strlcpy(qname, "root_", sizeof(qname)) >= sizeof(qname)) errx(1, "expand_altq: strlcpy"); if (strlcat(qname, interface->ifname, sizeof(qname)) >= sizeof(qname)) errx(1, "expand_altq: strlcat"); if (strlcpy(pb.qname, qname, sizeof(pb.qname)) >= sizeof(pb.qname)) errx(1, "expand_altq: strlcpy"); if (strlcpy(pb.ifname, interface->ifname, sizeof(pb.ifname)) >= sizeof(pb.ifname)) errx(1, "expand_altq: strlcpy"); pb.qlimit = pa.qlimit; pb.scheduler = pa.scheduler; bw.bw_absolute = pa.ifbandwidth; bw.bw_percent = 0; if (eval_pfqueue(pf, &pb, &bw, opts)) errs++; else if (pfctl_add_altq(pf, &pb)) errs++; } LOOP_THROUGH(struct node_queue, queue, nqueues, n = calloc(1, sizeof(struct node_queue)); if (n == NULL) err(1, "expand_altq: calloc"); if (pa.scheduler == ALTQT_CBQ || pa.scheduler == ALTQT_HFSC || pa.scheduler == ALTQT_FAIRQ) if (strlcpy(n->parent, qname, sizeof(n->parent)) >= sizeof(n->parent)) errx(1, "expand_altq: strlcpy"); if (strlcpy(n->queue, queue->queue, sizeof(n->queue)) >= sizeof(n->queue)) errx(1, "expand_altq: strlcpy"); if (strlcpy(n->ifname, interface->ifname, sizeof(n->ifname)) >= sizeof(n->ifname)) errx(1, "expand_altq: strlcpy"); n->scheduler = pa.scheduler; n->next = NULL; n->tail = n; if (queues == NULL) queues = n; else { queues->tail->next = n; queues->tail = n; } ); } ); FREE_LIST(struct node_if, interfaces); if (nqueues) FREE_LIST(struct node_queue, nqueues); return (errs); } int expand_queue(struct pf_altq *a, struct node_if *interfaces, struct node_queue *nqueues, struct node_queue_bw bwspec, struct node_queue_opt *opts) { struct node_queue *n, *nq; struct pf_altq pa; u_int8_t found = 0; u_int8_t errs = 0; if ((pf->loadopt & PFCTL_FLAG_ALTQ) == 0) { FREE_LIST(struct node_queue, nqueues); return (0); } if (queues == NULL) { yyerror("queue %s has no parent", a->qname); FREE_LIST(struct node_queue, nqueues); return (1); } LOOP_THROUGH(struct node_if, interface, interfaces, LOOP_THROUGH(struct node_queue, tqueue, queues, if (!strncmp(a->qname, tqueue->queue, PF_QNAME_SIZE) && (interface->ifname[0] == 0 || (!interface->not && !strncmp(interface->ifname, tqueue->ifname, IFNAMSIZ)) || (interface->not && strncmp(interface->ifname, tqueue->ifname, IFNAMSIZ)))) { /* found ourself in queues */ found++; memcpy(&pa, a, sizeof(struct pf_altq)); if (pa.scheduler != ALTQT_NONE && pa.scheduler != tqueue->scheduler) { yyerror("exactly one scheduler type " "per interface allowed"); return (1); } pa.scheduler = tqueue->scheduler; /* scheduler dependent error checking */ switch (pa.scheduler) { case ALTQT_PRIQ: if (nqueues != NULL) { yyerror("priq queues cannot " "have child queues"); return (1); } if (bwspec.bw_absolute > 0 || bwspec.bw_percent < 100) { yyerror("priq doesn't take " "bandwidth"); return (1); } break; default: break; } if (strlcpy(pa.ifname, tqueue->ifname, sizeof(pa.ifname)) >= sizeof(pa.ifname)) errx(1, "expand_queue: strlcpy"); if (strlcpy(pa.parent, tqueue->parent, sizeof(pa.parent)) >= sizeof(pa.parent)) errx(1, "expand_queue: strlcpy"); if (eval_pfqueue(pf, &pa, &bwspec, opts)) errs++; else if (pfctl_add_altq(pf, &pa)) errs++; for (nq = nqueues; nq != NULL; nq = nq->next) { if (!strcmp(a->qname, nq->queue)) { yyerror("queue cannot have " "itself as child"); errs++; continue; } n = calloc(1, sizeof(struct node_queue)); if (n == NULL) err(1, "expand_queue: calloc"); if (strlcpy(n->parent, a->qname, sizeof(n->parent)) >= sizeof(n->parent)) errx(1, "expand_queue strlcpy"); if (strlcpy(n->queue, nq->queue, sizeof(n->queue)) >= sizeof(n->queue)) errx(1, "expand_queue strlcpy"); if (strlcpy(n->ifname, tqueue->ifname, sizeof(n->ifname)) >= sizeof(n->ifname)) errx(1, "expand_queue strlcpy"); n->scheduler = tqueue->scheduler; n->next = NULL; n->tail = n; if (queues == NULL) queues = n; else { queues->tail->next = n; queues->tail = n; } } if ((pf->opts & PF_OPT_VERBOSE) && ( (found == 1 && interface->ifname[0] == 0) || (found > 0 && interface->ifname[0] != 0))) { print_queue(&pf->paltq->altq, 0, &bwspec, interface->ifname[0] != 0, opts); if (nqueues && nqueues->tail) { printf("{ "); LOOP_THROUGH(struct node_queue, queue, nqueues, printf("%s ", queue->queue); ); printf("}"); } printf("\n"); } } ); ); FREE_LIST(struct node_queue, nqueues); FREE_LIST(struct node_if, interfaces); if (!found) { yyerror("queue %s has no parent", a->qname); errs++; } if (errs) return (1); else return (0); } static int pf_af_to_proto(sa_family_t af) { if (af == AF_INET) return (ETHERTYPE_IP); if (af == AF_INET6) return (ETHERTYPE_IPV6); return (0); } void expand_eth_rule(struct pfctl_eth_rule *r, struct node_if *interfaces, struct node_etherproto *protos, struct node_mac *srcs, struct node_mac *dsts, struct node_host *ipsrcs, struct node_host *ipdsts, const char *bridge_to, const char *anchor_call) { char tagname[PF_TAG_NAME_SIZE]; char match_tagname[PF_TAG_NAME_SIZE]; char qname[PF_QNAME_SIZE]; if (strlcpy(tagname, r->tagname, sizeof(tagname)) >= sizeof(tagname)) errx(1, "expand_eth_rule: tagname"); if (strlcpy(match_tagname, r->match_tagname, sizeof(match_tagname)) >= sizeof(match_tagname)) errx(1, "expand_eth_rule: match_tagname"); if (strlcpy(qname, r->qname, sizeof(qname)) >= sizeof(qname)) errx(1, "expand_eth_rule: qname"); LOOP_THROUGH(struct node_if, interface, interfaces, LOOP_THROUGH(struct node_etherproto, proto, protos, LOOP_THROUGH(struct node_mac, src, srcs, LOOP_THROUGH(struct node_mac, dst, dsts, LOOP_THROUGH(struct node_host, ipsrc, ipsrcs, LOOP_THROUGH(struct node_host, ipdst, ipdsts, strlcpy(r->ifname, interface->ifname, sizeof(r->ifname)); r->ifnot = interface->not; r->proto = proto->proto; if (!r->proto && ipsrc->af) r->proto = pf_af_to_proto(ipsrc->af); else if (!r->proto && ipdst->af) r->proto = pf_af_to_proto(ipdst->af); bcopy(src->mac, r->src.addr, ETHER_ADDR_LEN); bcopy(src->mask, r->src.mask, ETHER_ADDR_LEN); r->src.neg = src->neg; r->src.isset = src->isset; r->ipsrc.addr = ipsrc->addr; r->ipsrc.neg = ipsrc->not; r->ipdst.addr = ipdst->addr; r->ipdst.neg = ipdst->not; bcopy(dst->mac, r->dst.addr, ETHER_ADDR_LEN); bcopy(dst->mask, r->dst.mask, ETHER_ADDR_LEN); r->dst.neg = dst->neg; r->dst.isset = dst->isset; r->nr = pf->eastack[pf->asd]->match++; if (strlcpy(r->tagname, tagname, sizeof(r->tagname)) >= sizeof(r->tagname)) errx(1, "expand_eth_rule: r->tagname"); if (strlcpy(r->match_tagname, match_tagname, sizeof(r->match_tagname)) >= sizeof(r->match_tagname)) errx(1, "expand_eth_rule: r->match_tagname"); if (strlcpy(r->qname, qname, sizeof(r->qname)) >= sizeof(r->qname)) errx(1, "expand_eth_rule: r->qname"); if (bridge_to) strlcpy(r->bridge_to, bridge_to, sizeof(r->bridge_to)); pfctl_append_eth_rule(pf, r, anchor_call); )))))); FREE_LIST(struct node_if, interfaces); FREE_LIST(struct node_etherproto, protos); FREE_LIST(struct node_mac, srcs); FREE_LIST(struct node_mac, dsts); FREE_LIST(struct node_host, ipsrcs); FREE_LIST(struct node_host, ipdsts); } void expand_rule(struct pfctl_rule *r, struct node_if *interfaces, struct node_host *rpool_hosts, struct node_proto *protos, struct node_os *src_oses, struct node_host *src_hosts, struct node_port *src_ports, struct node_host *dst_hosts, struct node_port *dst_ports, struct node_uid *uids, struct node_gid *gids, struct node_if *rcv, struct node_icmp *icmp_types, const char *anchor_call) { sa_family_t af = r->af; int added = 0, error = 0; char ifname[IF_NAMESIZE]; char label[PF_RULE_MAX_LABEL_COUNT][PF_RULE_LABEL_SIZE]; char tagname[PF_TAG_NAME_SIZE]; char match_tagname[PF_TAG_NAME_SIZE]; struct pf_pooladdr *pa; struct node_host *h, *osrch, *odsth; u_int8_t flags, flagset, keep_state; memcpy(label, r->label, sizeof(r->label)); assert(sizeof(r->label) == sizeof(label)); if (strlcpy(tagname, r->tagname, sizeof(tagname)) >= sizeof(tagname)) errx(1, "expand_rule: strlcpy"); if (strlcpy(match_tagname, r->match_tagname, sizeof(match_tagname)) >= sizeof(match_tagname)) errx(1, "expand_rule: strlcpy"); flags = r->flags; flagset = r->flagset; keep_state = r->keep_state; LOOP_THROUGH(struct node_if, interface, interfaces, LOOP_THROUGH(struct node_proto, proto, protos, LOOP_THROUGH(struct node_icmp, icmp_type, icmp_types, LOOP_THROUGH(struct node_host, src_host, src_hosts, - LOOP_THROUGH(struct node_port, src_port, src_ports, - LOOP_THROUGH(struct node_os, src_os, src_oses, LOOP_THROUGH(struct node_host, dst_host, dst_hosts, + LOOP_THROUGH(struct node_port, src_port, src_ports, LOOP_THROUGH(struct node_port, dst_port, dst_ports, + LOOP_THROUGH(struct node_os, src_os, src_oses, LOOP_THROUGH(struct node_uid, uid, uids, LOOP_THROUGH(struct node_gid, gid, gids, r->af = af; /* for link-local IPv6 address, interface must match up */ if ((r->af && src_host->af && r->af != src_host->af) || (r->af && dst_host->af && r->af != dst_host->af) || (src_host->af && dst_host->af && src_host->af != dst_host->af) || (src_host->ifindex && dst_host->ifindex && src_host->ifindex != dst_host->ifindex) || (src_host->ifindex && *interface->ifname && src_host->ifindex != if_nametoindex(interface->ifname)) || (dst_host->ifindex && *interface->ifname && dst_host->ifindex != if_nametoindex(interface->ifname))) continue; if (!r->af && src_host->af) r->af = src_host->af; else if (!r->af && dst_host->af) r->af = dst_host->af; if (*interface->ifname) strlcpy(r->ifname, interface->ifname, sizeof(r->ifname)); else if (if_indextoname(src_host->ifindex, ifname)) strlcpy(r->ifname, ifname, sizeof(r->ifname)); else if (if_indextoname(dst_host->ifindex, ifname)) strlcpy(r->ifname, ifname, sizeof(r->ifname)); else memset(r->ifname, '\0', sizeof(r->ifname)); memcpy(r->label, label, sizeof(r->label)); if (strlcpy(r->tagname, tagname, sizeof(r->tagname)) >= sizeof(r->tagname)) errx(1, "expand_rule: strlcpy"); if (strlcpy(r->match_tagname, match_tagname, sizeof(r->match_tagname)) >= sizeof(r->match_tagname)) errx(1, "expand_rule: strlcpy"); osrch = odsth = NULL; if (src_host->addr.type == PF_ADDR_DYNIFTL) { osrch = src_host; if ((src_host = gen_dynnode(src_host, r->af)) == NULL) err(1, "expand_rule: calloc"); } if (dst_host->addr.type == PF_ADDR_DYNIFTL) { odsth = dst_host; if ((dst_host = gen_dynnode(dst_host, r->af)) == NULL) err(1, "expand_rule: calloc"); } error += check_netmask(src_host, r->af); error += check_netmask(dst_host, r->af); r->ifnot = interface->not; r->proto = proto->proto; r->src.addr = src_host->addr; r->src.neg = src_host->not; r->src.port[0] = src_port->port[0]; r->src.port[1] = src_port->port[1]; r->src.port_op = src_port->op; r->dst.addr = dst_host->addr; r->dst.neg = dst_host->not; r->dst.port[0] = dst_port->port[0]; r->dst.port[1] = dst_port->port[1]; r->dst.port_op = dst_port->op; r->uid.op = uid->op; r->uid.uid[0] = uid->uid[0]; r->uid.uid[1] = uid->uid[1]; r->gid.op = gid->op; r->gid.gid[0] = gid->gid[0]; r->gid.gid[1] = gid->gid[1]; if (rcv) { strlcpy(r->rcv_ifname, rcv->ifname, sizeof(r->rcv_ifname)); } r->type = icmp_type->type; r->code = icmp_type->code; if ((keep_state == PF_STATE_MODULATE || keep_state == PF_STATE_SYNPROXY) && r->proto && r->proto != IPPROTO_TCP) r->keep_state = PF_STATE_NORMAL; else r->keep_state = keep_state; if (r->proto && r->proto != IPPROTO_TCP) { r->flags = 0; r->flagset = 0; } else { r->flags = flags; r->flagset = flagset; } if (icmp_type->proto && r->proto != icmp_type->proto) { yyerror("icmp-type mismatch"); error++; } if (src_os && src_os->os) { r->os_fingerprint = pfctl_get_fingerprint(src_os->os); if ((pf->opts & PF_OPT_VERBOSE2) && r->os_fingerprint == PF_OSFP_NOMATCH) fprintf(stderr, "warning: unknown '%s' OS fingerprint\n", src_os->os); } else { r->os_fingerprint = PF_OSFP_ANY; } TAILQ_INIT(&r->rpool.list); for (h = rpool_hosts; h != NULL; h = h->next) { pa = calloc(1, sizeof(struct pf_pooladdr)); if (pa == NULL) err(1, "expand_rule: calloc"); pa->addr = h->addr; if (h->ifname != NULL) { if (strlcpy(pa->ifname, h->ifname, sizeof(pa->ifname)) >= sizeof(pa->ifname)) errx(1, "expand_rule: strlcpy"); } else pa->ifname[0] = 0; TAILQ_INSERT_TAIL(&r->rpool.list, pa, entries); } if (rule_consistent(r, anchor_call[0]) < 0 || error) yyerror("skipping rule due to errors"); else { r->nr = pf->astack[pf->asd]->match++; pfctl_append_rule(pf, r, anchor_call); added++; } if (osrch && src_host->addr.type == PF_ADDR_DYNIFTL) { free(src_host); src_host = osrch; } if (odsth && dst_host->addr.type == PF_ADDR_DYNIFTL) { free(dst_host); dst_host = odsth; } )))))))))); FREE_LIST(struct node_if, interfaces); FREE_LIST(struct node_proto, protos); FREE_LIST(struct node_host, src_hosts); FREE_LIST(struct node_port, src_ports); FREE_LIST(struct node_os, src_oses); FREE_LIST(struct node_host, dst_hosts); FREE_LIST(struct node_port, dst_ports); FREE_LIST(struct node_uid, uids); FREE_LIST(struct node_gid, gids); FREE_LIST(struct node_icmp, icmp_types); FREE_LIST(struct node_host, rpool_hosts); if (!added) yyerror("rule expands to no valid combination"); } int expand_skip_interface(struct node_if *interfaces) { int errs = 0; if (!interfaces || (!interfaces->next && !interfaces->not && !strcmp(interfaces->ifname, "none"))) { if (pf->opts & PF_OPT_VERBOSE) printf("set skip on none\n"); errs = pfctl_set_interface_flags(pf, "", PFI_IFLAG_SKIP, 0); return (errs); } if (pf->opts & PF_OPT_VERBOSE) printf("set skip on {"); LOOP_THROUGH(struct node_if, interface, interfaces, if (pf->opts & PF_OPT_VERBOSE) printf(" %s", interface->ifname); if (interface->not) { yyerror("skip on ! is not supported"); errs++; } else errs += pfctl_set_interface_flags(pf, interface->ifname, PFI_IFLAG_SKIP, 1); ); if (pf->opts & PF_OPT_VERBOSE) printf(" }\n"); FREE_LIST(struct node_if, interfaces); if (errs) return (1); else return (0); } void freehostlist(struct node_host *h) { FREE_LIST(struct node_host, h); } #undef FREE_LIST #undef LOOP_THROUGH int check_rulestate(int desired_state) { if (require_order && (rulestate > desired_state)) { yyerror("Rules must be in order: options, ethernet, " "normalization, queueing, translation, filtering"); return (1); } rulestate = desired_state; return (0); } int kw_cmp(const void *k, const void *e) { return (strcmp(k, ((const struct keywords *)e)->k_name)); } int lookup(char *s) { /* this has to be sorted always */ static const struct keywords keywords[] = { { "all", ALL}, { "allow-opts", ALLOWOPTS}, { "altq", ALTQ}, { "anchor", ANCHOR}, { "antispoof", ANTISPOOF}, { "any", ANY}, { "bandwidth", BANDWIDTH}, { "binat", BINAT}, { "binat-anchor", BINATANCHOR}, { "bitmask", BITMASK}, { "block", BLOCK}, { "block-policy", BLOCKPOLICY}, { "bridge-to", BRIDGE_TO}, { "buckets", BUCKETS}, { "cbq", CBQ}, { "code", CODE}, { "codelq", CODEL}, { "debug", DEBUG}, { "divert-reply", DIVERTREPLY}, { "divert-to", DIVERTTO}, { "dnpipe", DNPIPE}, { "dnqueue", DNQUEUE}, { "drop", DROP}, { "dup-to", DUPTO}, { "endpoint-independent", ENDPI}, { "ether", ETHER}, { "fail-policy", FAILPOLICY}, { "fairq", FAIRQ}, { "fastroute", FASTROUTE}, { "file", FILENAME}, { "fingerprints", FINGERPRINTS}, { "flags", FLAGS}, { "floating", FLOATING}, { "flush", FLUSH}, { "for", FOR}, { "fragment", FRAGMENT}, { "from", FROM}, { "global", GLOBAL}, { "group", GROUP}, { "hfsc", HFSC}, { "hogs", HOGS}, { "hostid", HOSTID}, { "icmp-type", ICMPTYPE}, { "icmp6-type", ICMP6TYPE}, { "if-bound", IFBOUND}, { "in", IN}, { "include", INCLUDE}, { "inet", INET}, { "inet6", INET6}, { "interval", INTERVAL}, { "keep", KEEP}, { "keepcounters", KEEPCOUNTERS}, { "l3", L3}, { "label", LABEL}, { "limit", LIMIT}, { "linkshare", LINKSHARE}, { "load", LOAD}, { "log", LOG}, { "loginterface", LOGINTERFACE}, { "map-e-portset", MAPEPORTSET}, { "match", MATCH}, { "matches", MATCHES}, { "max", MAXIMUM}, { "max-mss", MAXMSS}, { "max-src-conn", MAXSRCCONN}, { "max-src-conn-rate", MAXSRCCONNRATE}, { "max-src-nodes", MAXSRCNODES}, { "max-src-states", MAXSRCSTATES}, { "min-ttl", MINTTL}, { "modulate", MODULATE}, { "nat", NAT}, { "nat-anchor", NATANCHOR}, { "no", NO}, { "no-df", NODF}, { "no-route", NOROUTE}, { "no-sync", NOSYNC}, { "on", ON}, { "optimization", OPTIMIZATION}, { "os", OS}, { "out", OUT}, { "overload", OVERLOAD}, { "pass", PASS}, { "pflow", PFLOW}, { "port", PORT}, { "prio", PRIO}, { "priority", PRIORITY}, { "priq", PRIQ}, { "probability", PROBABILITY}, { "proto", PROTO}, { "qlimit", QLIMIT}, { "queue", QUEUE}, { "quick", QUICK}, { "random", RANDOM}, { "random-id", RANDOMID}, { "rdr", RDR}, { "rdr-anchor", RDRANCHOR}, { "realtime", REALTIME}, { "reassemble", REASSEMBLE}, { "received-on", RECEIVEDON}, { "reply-to", REPLYTO}, { "require-order", REQUIREORDER}, { "return", RETURN}, { "return-icmp", RETURNICMP}, { "return-icmp6", RETURNICMP6}, { "return-rst", RETURNRST}, { "ridentifier", RIDENTIFIER}, { "round-robin", ROUNDROBIN}, { "route", ROUTE}, { "route-to", ROUTETO}, { "rtable", RTABLE}, { "rule", RULE}, { "ruleset-optimization", RULESET_OPTIMIZATION}, { "scrub", SCRUB}, { "set", SET}, { "set-tos", SETTOS}, { "skip", SKIP}, { "sloppy", SLOPPY}, { "source-hash", SOURCEHASH}, { "source-track", SOURCETRACK}, { "state", STATE}, { "state-defaults", STATEDEFAULTS}, { "state-policy", STATEPOLICY}, { "static-port", STATICPORT}, { "sticky-address", STICKYADDRESS}, { "syncookies", SYNCOOKIES}, { "synproxy", SYNPROXY}, { "table", TABLE}, { "tag", TAG}, { "tagged", TAGGED}, { "target", TARGET}, { "tbrsize", TBRSIZE}, { "timeout", TIMEOUT}, { "to", TO}, { "tos", TOS}, { "ttl", TTL}, { "upperlimit", UPPERLIMIT}, { "urpf-failed", URPFFAILED}, { "user", USER}, }; const struct keywords *p; p = bsearch(s, keywords, sizeof(keywords)/sizeof(keywords[0]), sizeof(keywords[0]), kw_cmp); if (p) { if (debug > 1) fprintf(stderr, "%s: %d\n", s, p->k_val); return (p->k_val); } else { if (debug > 1) fprintf(stderr, "string: %s\n", s); return (STRING); } } #define MAXPUSHBACK 128 static char *parsebuf; static int parseindex; static char pushback_buffer[MAXPUSHBACK]; static int pushback_index = 0; int lgetc(int quotec) { int c, next; if (parsebuf) { /* Read character from the parsebuffer instead of input. */ if (parseindex >= 0) { c = parsebuf[parseindex++]; if (c != '\0') return (c); parsebuf = NULL; } else parseindex++; } if (pushback_index) return (pushback_buffer[--pushback_index]); if (quotec) { if ((c = getc(file->stream)) == EOF) { yyerror("reached end of file while parsing quoted string"); if (popfile() == EOF) return (EOF); return (quotec); } return (c); } while ((c = getc(file->stream)) == '\\') { next = getc(file->stream); if (next != '\n') { c = next; break; } yylval.lineno = file->lineno; file->lineno++; } while (c == EOF) { if (popfile() == EOF) return (EOF); c = getc(file->stream); } return (c); } int lungetc(int c) { if (c == EOF) return (EOF); if (parsebuf) { parseindex--; if (parseindex >= 0) return (c); } if (pushback_index < MAXPUSHBACK-1) return (pushback_buffer[pushback_index++] = c); else return (EOF); } int findeol(void) { int c; parsebuf = NULL; /* skip to either EOF or the first real EOL */ while (1) { if (pushback_index) c = pushback_buffer[--pushback_index]; else c = lgetc(0); if (c == '\n') { file->lineno++; break; } if (c == EOF) break; } return (ERROR); } int yylex(void) { char buf[8096]; char *p, *val; int quotec, next, c; int token; top: p = buf; while ((c = lgetc(0)) == ' ' || c == '\t') ; /* nothing */ yylval.lineno = file->lineno; if (c == '#') while ((c = lgetc(0)) != '\n' && c != EOF) ; /* nothing */ if (c == '$' && parsebuf == NULL) { while (1) { if ((c = lgetc(0)) == EOF) return (0); if (p + 1 >= buf + sizeof(buf) - 1) { yyerror("string too long"); return (findeol()); } if (isalnum(c) || c == '_') { *p++ = (char)c; continue; } *p = '\0'; lungetc(c); break; } val = symget(buf); if (val == NULL) { yyerror("macro '%s' not defined", buf); return (findeol()); } parsebuf = val; parseindex = 0; goto top; } switch (c) { case '\'': case '"': quotec = c; while (1) { if ((c = lgetc(quotec)) == EOF) return (0); if (c == '\n') { file->lineno++; continue; } else if (c == '\\') { if ((next = lgetc(quotec)) == EOF) return (0); if (next == quotec || c == ' ' || c == '\t') c = next; else if (next == '\n') { file->lineno++; continue; } else lungetc(next); } else if (c == quotec) { *p = '\0'; break; } if (p + 1 >= buf + sizeof(buf) - 1) { yyerror("string too long"); return (findeol()); } *p++ = (char)c; } yylval.v.string = strdup(buf); if (yylval.v.string == NULL) err(1, "yylex: strdup"); return (STRING); case '!': next = lgetc(0); if (next == '=') return (NE); lungetc(next); break; case '<': next = lgetc(0); if (next == '>') { yylval.v.i = PF_OP_XRG; return (PORTBINARY); } else if (next == '=') return (LE); lungetc(next); break; case '>': next = lgetc(0); if (next == '<') { yylval.v.i = PF_OP_IRG; return (PORTBINARY); } else if (next == '=') return (GE); lungetc(next); break; case '-': next = lgetc(0); if (next == '>') return (ARROW); lungetc(next); break; } #define allowed_to_end_number(x) \ (isspace(x) || x == ')' || x ==',' || x == '/' || x == '}' || x == '=') if (c == '-' || isdigit(c)) { do { *p++ = c; if ((unsigned)(p-buf) >= sizeof(buf)) { yyerror("string too long"); return (findeol()); } } while ((c = lgetc(0)) != EOF && isdigit(c)); lungetc(c); if (p == buf + 1 && buf[0] == '-') goto nodigits; if (c == EOF || allowed_to_end_number(c)) { const char *errstr = NULL; *p = '\0'; yylval.v.number = strtonum(buf, LLONG_MIN, LLONG_MAX, &errstr); if (errstr) { yyerror("\"%s\" invalid number: %s", buf, errstr); return (findeol()); } return (NUMBER); } else { nodigits: while (p > buf + 1) lungetc(*--p); c = *--p; if (c == '-') return (c); } } #define allowed_in_string(x) \ (isalnum(x) || (ispunct(x) && x != '(' && x != ')' && \ x != '{' && x != '}' && x != '<' && x != '>' && \ x != '!' && x != '=' && x != '/' && x != '#' && \ x != ',')) if (isalnum(c) || c == ':' || c == '_') { do { *p++ = c; if ((unsigned)(p-buf) >= sizeof(buf)) { yyerror("string too long"); return (findeol()); } } while ((c = lgetc(0)) != EOF && (allowed_in_string(c))); lungetc(c); *p = '\0'; if ((token = lookup(buf)) == STRING) if ((yylval.v.string = strdup(buf)) == NULL) err(1, "yylex: strdup"); return (token); } if (c == '\n') { yylval.lineno = file->lineno; file->lineno++; } if (c == EOF) return (0); return (c); } int check_file_secrecy(int fd, const char *fname) { struct stat st; if (fstat(fd, &st)) { warn("cannot stat %s", fname); return (-1); } if (st.st_uid != 0 && st.st_uid != getuid()) { warnx("%s: owner not root or current user", fname); return (-1); } if (st.st_mode & (S_IRWXG | S_IRWXO)) { warnx("%s: group/world readable/writeable", fname); return (-1); } return (0); } struct file * pushfile(const char *name, int secret) { struct file *nfile; if ((nfile = calloc(1, sizeof(struct file))) == NULL || (nfile->name = strdup(name)) == NULL) { warn("malloc"); return (NULL); } if (TAILQ_FIRST(&files) == NULL && strcmp(nfile->name, "-") == 0) { nfile->stream = stdin; free(nfile->name); if ((nfile->name = strdup("stdin")) == NULL) { warn("strdup"); free(nfile); return (NULL); } } else if ((nfile->stream = fopen(nfile->name, "r")) == NULL) { warn("%s", nfile->name); free(nfile->name); free(nfile); return (NULL); } else if (secret && check_file_secrecy(fileno(nfile->stream), nfile->name)) { fclose(nfile->stream); free(nfile->name); free(nfile); return (NULL); } nfile->lineno = 1; TAILQ_INSERT_TAIL(&files, nfile, entry); return (nfile); } int popfile(void) { struct file *prev; if ((prev = TAILQ_PREV(file, files, entry)) != NULL) { prev->errors += file->errors; TAILQ_REMOVE(&files, file, entry); fclose(file->stream); free(file->name); free(file); file = prev; return (0); } return (EOF); } int parse_config(char *filename, struct pfctl *xpf) { int errors = 0; struct sym *sym; pf = xpf; errors = 0; rulestate = PFCTL_STATE_NONE; returnicmpdefault = (ICMP_UNREACH << 8) | ICMP_UNREACH_PORT; returnicmp6default = (ICMP6_DST_UNREACH << 8) | ICMP6_DST_UNREACH_NOPORT; blockpolicy = PFRULE_DROP; failpolicy = PFRULE_DROP; require_order = 1; if ((file = pushfile(filename, 0)) == NULL) { warn("cannot open the main config file!"); return (-1); } yyparse(); errors = file->errors; popfile(); /* Free macros and check which have not been used. */ while ((sym = TAILQ_FIRST(&symhead))) { if ((pf->opts & PF_OPT_VERBOSE2) && !sym->used) fprintf(stderr, "warning: macro '%s' not " "used\n", sym->nam); free(sym->nam); free(sym->val); TAILQ_REMOVE(&symhead, sym, entry); free(sym); } return (errors ? -1 : 0); } int symset(const char *nam, const char *val, int persist) { struct sym *sym; for (sym = TAILQ_FIRST(&symhead); sym && strcmp(nam, sym->nam); sym = TAILQ_NEXT(sym, entry)) ; /* nothing */ if (sym != NULL) { if (sym->persist == 1) return (0); else { free(sym->nam); free(sym->val); TAILQ_REMOVE(&symhead, sym, entry); free(sym); } } if ((sym = calloc(1, sizeof(*sym))) == NULL) return (-1); sym->nam = strdup(nam); if (sym->nam == NULL) { free(sym); return (-1); } sym->val = strdup(val); if (sym->val == NULL) { free(sym->nam); free(sym); return (-1); } sym->used = 0; sym->persist = persist; TAILQ_INSERT_TAIL(&symhead, sym, entry); return (0); } int pfctl_cmdline_symset(char *s) { char *sym, *val; int ret; if ((val = strrchr(s, '=')) == NULL) return (-1); if ((sym = malloc(strlen(s) - strlen(val) + 1)) == NULL) err(1, "pfctl_cmdline_symset: malloc"); strlcpy(sym, s, strlen(s) - strlen(val) + 1); ret = symset(sym, val + 1, 1); free(sym); return (ret); } char * symget(const char *nam) { struct sym *sym; TAILQ_FOREACH(sym, &symhead, entry) if (strcmp(nam, sym->nam) == 0) { sym->used = 1; return (sym->val); } return (NULL); } void mv_rules(struct pfctl_ruleset *src, struct pfctl_ruleset *dst) { int i; struct pfctl_rule *r; for (i = 0; i < PF_RULESET_MAX; ++i) { while ((r = TAILQ_FIRST(src->rules[i].active.ptr)) != NULL) { TAILQ_REMOVE(src->rules[i].active.ptr, r, entries); TAILQ_INSERT_TAIL(dst->rules[i].active.ptr, r, entries); dst->anchor->match++; } src->anchor->match = 0; while ((r = TAILQ_FIRST(src->rules[i].inactive.ptr)) != NULL) { TAILQ_REMOVE(src->rules[i].inactive.ptr, r, entries); TAILQ_INSERT_TAIL(dst->rules[i].inactive.ptr, r, entries); } } } void mv_eth_rules(struct pfctl_eth_ruleset *src, struct pfctl_eth_ruleset *dst) { struct pfctl_eth_rule *r; while ((r = TAILQ_FIRST(&src->rules)) != NULL) { TAILQ_REMOVE(&src->rules, r, entries); TAILQ_INSERT_TAIL(&dst->rules, r, entries); dst->anchor->match++; } src->anchor->match = 0; } void decide_address_family(struct node_host *n, sa_family_t *af) { if (*af != 0 || n == NULL) return; *af = n->af; while ((n = n->next) != NULL) { if (n->af != *af) { *af = 0; return; } } } void remove_invalid_hosts(struct node_host **nh, sa_family_t *af) { struct node_host *n = *nh, *prev = NULL; while (n != NULL) { if (*af && n->af && n->af != *af) { /* unlink and free n */ struct node_host *next = n->next; /* adjust tail pointer */ if (n == (*nh)->tail) (*nh)->tail = prev; /* adjust previous node's next pointer */ if (prev == NULL) *nh = next; else prev->next = next; /* free node */ if (n->ifname != NULL) free(n->ifname); free(n); n = next; } else { if (n->af && !*af) *af = n->af; prev = n; n = n->next; } } } int invalid_redirect(struct node_host *nh, sa_family_t af) { if (!af) { struct node_host *n; /* tables and dyniftl are ok without an address family */ for (n = nh; n != NULL; n = n->next) { if (n->addr.type != PF_ADDR_TABLE && n->addr.type != PF_ADDR_DYNIFTL) { yyerror("address family not given and " "translation address expands to multiple " "address families"); return (1); } } } if (nh == NULL) { yyerror("no translation address with matching address family " "found."); return (1); } return (0); } int atoul(char *s, u_long *ulvalp) { u_long ulval; char *ep; errno = 0; ulval = strtoul(s, &ep, 0); if (s[0] == '\0' || *ep != '\0') return (-1); if (errno == ERANGE && ulval == ULONG_MAX) return (-1); *ulvalp = ulval; return (0); } int getservice(char *n) { struct servent *s; u_long ulval; if (atoul(n, &ulval) == 0) { if (ulval > 65535) { yyerror("illegal port value %lu", ulval); return (-1); } return (htons(ulval)); } else { s = getservbyname(n, "tcp"); if (s == NULL) s = getservbyname(n, "udp"); if (s == NULL) s = getservbyname(n, "sctp"); if (s == NULL) { yyerror("unknown port %s", n); return (-1); } return (s->s_port); } } int rule_label(struct pfctl_rule *r, char *s[PF_RULE_MAX_LABEL_COUNT]) { for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) { if (s[i] == NULL) return (0); if (strlcpy(r->label[i], s[i], sizeof(r->label[0])) >= sizeof(r->label[0])) { yyerror("rule label too long (max %d chars)", sizeof(r->label[0])-1); return (-1); } } return (0); } int eth_rule_label(struct pfctl_eth_rule *r, char *s[PF_RULE_MAX_LABEL_COUNT]) { for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) { if (s[i] == NULL) return (0); if (strlcpy(r->label[i], s[i], sizeof(r->label[0])) >= sizeof(r->label[0])) { yyerror("rule label too long (max %d chars)", sizeof(r->label[0])-1); return (-1); } } return (0); } u_int16_t parseicmpspec(char *w, sa_family_t af) { const struct icmpcodeent *p; u_long ulval; u_int8_t icmptype; if (af == AF_INET) icmptype = returnicmpdefault >> 8; else icmptype = returnicmp6default >> 8; if (atoul(w, &ulval) == -1) { if ((p = geticmpcodebyname(icmptype, w, af)) == NULL) { yyerror("unknown icmp code %s", w); return (0); } ulval = p->code; } if (ulval > 255) { yyerror("invalid icmp code %lu", ulval); return (0); } return (icmptype << 8 | ulval); } int parseport(char *port, struct range *r, int extensions) { char *p = strchr(port, ':'); if (p == NULL) { if ((r->a = getservice(port)) == -1) return (-1); r->b = 0; r->t = PF_OP_NONE; return (0); } if ((extensions & PPORT_STAR) && !strcmp(p+1, "*")) { *p = 0; if ((r->a = getservice(port)) == -1) return (-1); r->b = 0; r->t = PF_OP_IRG; return (0); } if ((extensions & PPORT_RANGE)) { *p++ = 0; if ((r->a = getservice(port)) == -1 || (r->b = getservice(p)) == -1) return (-1); if (r->a == r->b) { r->b = 0; r->t = PF_OP_NONE; } else r->t = PF_OP_RRG; return (0); } return (-1); } int pfctl_load_anchors(int dev, struct pfctl *pf, struct pfr_buffer *trans) { struct loadanchors *la; TAILQ_FOREACH(la, &loadanchorshead, entries) { if (pf->opts & PF_OPT_VERBOSE) fprintf(stderr, "\nLoading anchor %s from %s\n", la->anchorname, la->filename); if (pfctl_rules(dev, la->filename, pf->opts, pf->optimize, la->anchorname, trans) == -1) return (-1); } return (0); } int kw_casecmp(const void *k, const void *e) { return (strcasecmp(k, ((const struct keywords *)e)->k_name)); } int map_tos(char *s, int *val) { /* DiffServ Codepoints and other TOS mappings */ const struct keywords toswords[] = { { "af11", IPTOS_DSCP_AF11 }, { "af12", IPTOS_DSCP_AF12 }, { "af13", IPTOS_DSCP_AF13 }, { "af21", IPTOS_DSCP_AF21 }, { "af22", IPTOS_DSCP_AF22 }, { "af23", IPTOS_DSCP_AF23 }, { "af31", IPTOS_DSCP_AF31 }, { "af32", IPTOS_DSCP_AF32 }, { "af33", IPTOS_DSCP_AF33 }, { "af41", IPTOS_DSCP_AF41 }, { "af42", IPTOS_DSCP_AF42 }, { "af43", IPTOS_DSCP_AF43 }, { "critical", IPTOS_PREC_CRITIC_ECP }, { "cs0", IPTOS_DSCP_CS0 }, { "cs1", IPTOS_DSCP_CS1 }, { "cs2", IPTOS_DSCP_CS2 }, { "cs3", IPTOS_DSCP_CS3 }, { "cs4", IPTOS_DSCP_CS4 }, { "cs5", IPTOS_DSCP_CS5 }, { "cs6", IPTOS_DSCP_CS6 }, { "cs7", IPTOS_DSCP_CS7 }, { "ef", IPTOS_DSCP_EF }, { "inetcontrol", IPTOS_PREC_INTERNETCONTROL }, { "lowdelay", IPTOS_LOWDELAY }, { "netcontrol", IPTOS_PREC_NETCONTROL }, { "reliability", IPTOS_RELIABILITY }, { "throughput", IPTOS_THROUGHPUT }, { "va", IPTOS_DSCP_VA } }; const struct keywords *p; p = bsearch(s, toswords, sizeof(toswords)/sizeof(toswords[0]), sizeof(toswords[0]), kw_casecmp); if (p) { *val = p->k_val; return (1); } return (0); } int rt_tableid_max(void) { #ifdef __FreeBSD__ int fibs; size_t l = sizeof(fibs); if (sysctlbyname("net.fibs", &fibs, &l, NULL, 0) == -1) fibs = 16; /* XXX RT_MAXFIBS, at least limit it some. */ /* * As the OpenBSD code only compares > and not >= we need to adjust * here given we only accept values of 0..n and want to avoid #ifdefs * in the grammar. */ return (fibs - 1); #else return (RT_TABLEID_MAX); #endif } struct node_mac* node_mac_from_string(const char *str) { struct node_mac *m; m = calloc(1, sizeof(struct node_mac)); if (m == NULL) err(1, "mac: calloc"); if (sscanf(str, "%02hhx:%02hhx:%02hhx:%02hhx:%02hhx:%02hhx", &m->mac[0], &m->mac[1], &m->mac[2], &m->mac[3], &m->mac[4], &m->mac[5]) != 6) { free(m); yyerror("invalid MAC address"); return (NULL); } memset(m->mask, 0xff, ETHER_ADDR_LEN); m->isset = true; m->next = NULL; m->tail = m; return (m); } struct node_mac* node_mac_from_string_masklen(const char *str, int masklen) { struct node_mac *m; if (masklen < 0 || masklen > (ETHER_ADDR_LEN * 8)) { yyerror("invalid MAC mask length"); return (NULL); } m = node_mac_from_string(str); if (m == NULL) return (NULL); memset(m->mask, 0, ETHER_ADDR_LEN); for (int i = 0; i < masklen; i++) m->mask[i / 8] |= 1 << (i % 8); return (m); } struct node_mac* node_mac_from_string_mask(const char *str, const char *mask) { struct node_mac *m; m = node_mac_from_string(str); if (m == NULL) return (NULL); if (sscanf(mask, "%02hhx:%02hhx:%02hhx:%02hhx:%02hhx:%02hhx", &m->mask[0], &m->mask[1], &m->mask[2], &m->mask[3], &m->mask[4], &m->mask[5]) != 6) { free(m); yyerror("invalid MAC mask"); return (NULL); } return (m); } diff --git a/sbin/pfctl/pfctl_optimize.c b/sbin/pfctl/pfctl_optimize.c index 7440bf8a506d..9858f38b8671 100644 --- a/sbin/pfctl/pfctl_optimize.c +++ b/sbin/pfctl/pfctl_optimize.c @@ -1,1661 +1,1661 @@ /* $OpenBSD: pfctl_optimize.c,v 1.17 2008/05/06 03:45:21 mpf Exp $ */ /* * Copyright (c) 2004 Mike Frantzen * * Permission to use, copy, modify, and distribute this software for any * purpose with or without fee is hereby granted, provided that the above * copyright notice and this permission notice appear in all copies. * * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include "pfctl_parser.h" #include "pfctl.h" /* The size at which a table becomes faster than individual rules */ #define TABLE_THRESHOLD 6 /* #define OPT_DEBUG 1 */ #ifdef OPT_DEBUG # define DEBUG(str, v...) \ printf("%s: " str "\n", __FUNCTION__ , ## v) #else # define DEBUG(str, v...) ((void)0) #endif /* * A container that lets us sort a superblock to optimize the skip step jumps */ struct pf_skip_step { int ps_count; /* number of items */ TAILQ_HEAD( , pf_opt_rule) ps_rules; TAILQ_ENTRY(pf_skip_step) ps_entry; }; /* * A superblock is a block of adjacent rules of similar action. If there * are five PASS rules in a row, they all become members of a superblock. * Once we have a superblock, we are free to re-order any rules within it * in order to improve performance; if a packet is passed, it doesn't matter * who passed it. */ struct superblock { TAILQ_HEAD( , pf_opt_rule) sb_rules; TAILQ_ENTRY(superblock) sb_entry; struct superblock *sb_profiled_block; TAILQ_HEAD(skiplist, pf_skip_step) sb_skipsteps[PF_SKIP_COUNT]; }; TAILQ_HEAD(superblocks, superblock); /* * Description of the PF rule structure. */ enum { BARRIER, /* the presence of the field puts the rule in its own block */ BREAK, /* the field may not differ between rules in a superblock */ NOMERGE, /* the field may not differ between rules when combined */ COMBINED, /* the field may itself be combined with other rules */ DC, /* we just don't care about the field */ NEVER}; /* we should never see this field set?!? */ static struct pf_rule_field { const char *prf_name; int prf_type; size_t prf_offset; size_t prf_size; } pf_rule_desc[] = { #define PF_RULE_FIELD(field, ty) \ {#field, \ ty, \ offsetof(struct pfctl_rule, field), \ sizeof(((struct pfctl_rule *)0)->field)} /* * The presence of these fields in a rule put the rule in its own * superblock. Thus it will not be optimized. It also prevents the * rule from being re-ordered at all. */ PF_RULE_FIELD(label, BARRIER), PF_RULE_FIELD(prob, BARRIER), PF_RULE_FIELD(max_states, BARRIER), PF_RULE_FIELD(max_src_nodes, BARRIER), PF_RULE_FIELD(max_src_states, BARRIER), PF_RULE_FIELD(max_src_conn, BARRIER), PF_RULE_FIELD(max_src_conn_rate, BARRIER), PF_RULE_FIELD(anchor, BARRIER), /* for now */ /* * These fields must be the same between all rules in the same superblock. * These rules are allowed to be re-ordered but only among like rules. * For instance we can re-order all 'tag "foo"' rules because they have the * same tag. But we can not re-order between a 'tag "foo"' and a * 'tag "bar"' since that would change the meaning of the ruleset. */ PF_RULE_FIELD(tagname, BREAK), PF_RULE_FIELD(keep_state, BREAK), PF_RULE_FIELD(qname, BREAK), PF_RULE_FIELD(pqname, BREAK), PF_RULE_FIELD(rt, BREAK), PF_RULE_FIELD(allow_opts, BREAK), PF_RULE_FIELD(rule_flag, BREAK), PF_RULE_FIELD(action, BREAK), PF_RULE_FIELD(log, BREAK), PF_RULE_FIELD(quick, BREAK), PF_RULE_FIELD(return_ttl, BREAK), PF_RULE_FIELD(overload_tblname, BREAK), PF_RULE_FIELD(flush, BREAK), PF_RULE_FIELD(rpool, BREAK), PF_RULE_FIELD(logif, BREAK), /* * Any fields not listed in this structure act as BREAK fields */ /* * These fields must not differ when we merge two rules together but * their difference isn't enough to put the rules in different superblocks. * There are no problems re-ordering any rules with these fields. */ PF_RULE_FIELD(af, NOMERGE), PF_RULE_FIELD(ifnot, NOMERGE), PF_RULE_FIELD(ifname, NOMERGE), /* hack for IF groups */ PF_RULE_FIELD(match_tag_not, NOMERGE), PF_RULE_FIELD(match_tagname, NOMERGE), PF_RULE_FIELD(os_fingerprint, NOMERGE), PF_RULE_FIELD(timeout, NOMERGE), PF_RULE_FIELD(return_icmp, NOMERGE), PF_RULE_FIELD(return_icmp6, NOMERGE), PF_RULE_FIELD(uid, NOMERGE), PF_RULE_FIELD(gid, NOMERGE), PF_RULE_FIELD(direction, NOMERGE), PF_RULE_FIELD(proto, NOMERGE), PF_RULE_FIELD(type, NOMERGE), PF_RULE_FIELD(code, NOMERGE), PF_RULE_FIELD(flags, NOMERGE), PF_RULE_FIELD(flagset, NOMERGE), PF_RULE_FIELD(tos, NOMERGE), PF_RULE_FIELD(src.port, NOMERGE), PF_RULE_FIELD(dst.port, NOMERGE), PF_RULE_FIELD(src.port_op, NOMERGE), PF_RULE_FIELD(dst.port_op, NOMERGE), PF_RULE_FIELD(src.neg, NOMERGE), PF_RULE_FIELD(dst.neg, NOMERGE), /* These fields can be merged */ PF_RULE_FIELD(src.addr, COMBINED), PF_RULE_FIELD(dst.addr, COMBINED), /* We just don't care about these fields. They're set by the kernel */ PF_RULE_FIELD(skip, DC), PF_RULE_FIELD(evaluations, DC), PF_RULE_FIELD(packets, DC), PF_RULE_FIELD(bytes, DC), PF_RULE_FIELD(kif, DC), PF_RULE_FIELD(states_cur, DC), PF_RULE_FIELD(states_tot, DC), PF_RULE_FIELD(src_nodes, DC), PF_RULE_FIELD(nr, DC), PF_RULE_FIELD(entries, DC), PF_RULE_FIELD(qid, DC), PF_RULE_FIELD(pqid, DC), PF_RULE_FIELD(anchor_relative, DC), PF_RULE_FIELD(anchor_wildcard, DC), PF_RULE_FIELD(tag, DC), PF_RULE_FIELD(match_tag, DC), PF_RULE_FIELD(overload_tbl, DC), /* These fields should never be set in a PASS/BLOCK rule */ PF_RULE_FIELD(natpass, NEVER), PF_RULE_FIELD(max_mss, NEVER), PF_RULE_FIELD(min_ttl, NEVER), PF_RULE_FIELD(set_tos, NEVER), }; int add_opt_table(struct pfctl *, struct pf_opt_tbl **, sa_family_t, struct pf_rule_addr *); int addrs_combineable(struct pf_rule_addr *, struct pf_rule_addr *); int addrs_equal(struct pf_rule_addr *, struct pf_rule_addr *); int block_feedback(struct pfctl *, struct superblock *); int combine_rules(struct pfctl *, struct superblock *); void comparable_rule(struct pfctl_rule *, const struct pfctl_rule *, int); int construct_superblocks(struct pfctl *, struct pf_opt_queue *, struct superblocks *); void exclude_supersets(struct pfctl_rule *, struct pfctl_rule *); int interface_group(const char *); int load_feedback_profile(struct pfctl *, struct superblocks *); int optimize_superblock(struct pfctl *, struct superblock *); int pf_opt_create_table(struct pfctl *, struct pf_opt_tbl *); void remove_from_skipsteps(struct skiplist *, struct superblock *, struct pf_opt_rule *, struct pf_skip_step *); int remove_identical_rules(struct pfctl *, struct superblock *); int reorder_rules(struct pfctl *, struct superblock *, int); int rules_combineable(struct pfctl_rule *, struct pfctl_rule *); void skip_append(struct superblock *, int, struct pf_skip_step *, struct pf_opt_rule *); int skip_compare(int, struct pf_skip_step *, struct pf_opt_rule *); void skip_init(void); int skip_cmp_af(struct pfctl_rule *, struct pfctl_rule *); int skip_cmp_dir(struct pfctl_rule *, struct pfctl_rule *); int skip_cmp_dst_addr(struct pfctl_rule *, struct pfctl_rule *); int skip_cmp_dst_port(struct pfctl_rule *, struct pfctl_rule *); int skip_cmp_ifp(struct pfctl_rule *, struct pfctl_rule *); int skip_cmp_proto(struct pfctl_rule *, struct pfctl_rule *); int skip_cmp_src_addr(struct pfctl_rule *, struct pfctl_rule *); int skip_cmp_src_port(struct pfctl_rule *, struct pfctl_rule *); int superblock_inclusive(struct superblock *, struct pf_opt_rule *); void superblock_free(struct pfctl *, struct superblock *); static int (*skip_comparitors[PF_SKIP_COUNT])(struct pfctl_rule *, struct pfctl_rule *); static const char *skip_comparitors_names[PF_SKIP_COUNT]; #define PF_SKIP_COMPARITORS { \ { "ifp", PF_SKIP_IFP, skip_cmp_ifp }, \ { "dir", PF_SKIP_DIR, skip_cmp_dir }, \ { "af", PF_SKIP_AF, skip_cmp_af }, \ { "proto", PF_SKIP_PROTO, skip_cmp_proto }, \ { "saddr", PF_SKIP_SRC_ADDR, skip_cmp_src_addr }, \ - { "sport", PF_SKIP_SRC_PORT, skip_cmp_src_port }, \ { "daddr", PF_SKIP_DST_ADDR, skip_cmp_dst_addr }, \ + { "sport", PF_SKIP_SRC_PORT, skip_cmp_src_port }, \ { "dport", PF_SKIP_DST_PORT, skip_cmp_dst_port } \ } static struct pfr_buffer table_buffer; static int table_identifier; int pfctl_optimize_ruleset(struct pfctl *pf, struct pfctl_ruleset *rs) { struct superblocks superblocks; struct pf_opt_queue opt_queue; struct superblock *block; struct pf_opt_rule *por; struct pfctl_rule *r; struct pfctl_rulequeue *old_rules; DEBUG("optimizing ruleset"); memset(&table_buffer, 0, sizeof(table_buffer)); skip_init(); TAILQ_INIT(&opt_queue); old_rules = rs->rules[PF_RULESET_FILTER].active.ptr; rs->rules[PF_RULESET_FILTER].active.ptr = rs->rules[PF_RULESET_FILTER].inactive.ptr; rs->rules[PF_RULESET_FILTER].inactive.ptr = old_rules; /* * XXX expanding the pf_opt_rule format throughout pfctl might allow * us to avoid all this copying. */ while ((r = TAILQ_FIRST(rs->rules[PF_RULESET_FILTER].inactive.ptr)) != NULL) { TAILQ_REMOVE(rs->rules[PF_RULESET_FILTER].inactive.ptr, r, entries); if ((por = calloc(1, sizeof(*por))) == NULL) err(1, "calloc"); memcpy(&por->por_rule, r, sizeof(*r)); if (TAILQ_FIRST(&r->rpool.list) != NULL) { TAILQ_INIT(&por->por_rule.rpool.list); pfctl_move_pool(&r->rpool, &por->por_rule.rpool); } else bzero(&por->por_rule.rpool, sizeof(por->por_rule.rpool)); TAILQ_INSERT_TAIL(&opt_queue, por, por_entry); } TAILQ_INIT(&superblocks); if (construct_superblocks(pf, &opt_queue, &superblocks)) goto error; if (pf->optimize & PF_OPTIMIZE_PROFILE) { if (load_feedback_profile(pf, &superblocks)) goto error; } TAILQ_FOREACH(block, &superblocks, sb_entry) { if (optimize_superblock(pf, block)) goto error; } rs->anchor->refcnt = 0; while ((block = TAILQ_FIRST(&superblocks))) { TAILQ_REMOVE(&superblocks, block, sb_entry); while ((por = TAILQ_FIRST(&block->sb_rules))) { TAILQ_REMOVE(&block->sb_rules, por, por_entry); por->por_rule.nr = rs->anchor->refcnt++; if ((r = calloc(1, sizeof(*r))) == NULL) err(1, "calloc"); memcpy(r, &por->por_rule, sizeof(*r)); TAILQ_INIT(&r->rpool.list); pfctl_move_pool(&por->por_rule.rpool, &r->rpool); TAILQ_INSERT_TAIL( rs->rules[PF_RULESET_FILTER].active.ptr, r, entries); free(por); } free(block); } return (0); error: while ((por = TAILQ_FIRST(&opt_queue))) { TAILQ_REMOVE(&opt_queue, por, por_entry); if (por->por_src_tbl) { pfr_buf_clear(por->por_src_tbl->pt_buf); free(por->por_src_tbl->pt_buf); free(por->por_src_tbl); } if (por->por_dst_tbl) { pfr_buf_clear(por->por_dst_tbl->pt_buf); free(por->por_dst_tbl->pt_buf); free(por->por_dst_tbl); } free(por); } while ((block = TAILQ_FIRST(&superblocks))) { TAILQ_REMOVE(&superblocks, block, sb_entry); superblock_free(pf, block); } return (1); } /* * Go ahead and optimize a superblock */ int optimize_superblock(struct pfctl *pf, struct superblock *block) { #ifdef OPT_DEBUG struct pf_opt_rule *por; #endif /* OPT_DEBUG */ /* We have a few optimization passes: * 1) remove duplicate rules or rules that are a subset of other * rules * 2) combine otherwise identical rules with different IP addresses * into a single rule and put the addresses in a table. * 3) re-order the rules to improve kernel skip steps * 4) re-order the 'quick' rules based on feedback from the * active ruleset statistics * * XXX combine_rules() doesn't combine v4 and v6 rules. would just * have to keep af in the table container, make af 'COMBINE' and * twiddle the af on the merged rule * XXX maybe add a weighting to the metric on skipsteps when doing * reordering. sometimes two sequential tables will be better * that four consecutive interfaces. * XXX need to adjust the skipstep count of everything after PROTO, * since they aren't actually checked on a proto mismatch in * pf_test_{tcp, udp, icmp}() * XXX should i treat proto=0, af=0 or dir=0 special in skepstep * calculation since they are a DC? * XXX keep last skiplist of last superblock to influence this * superblock. '5 inet6 log' should make '3 inet6' come before '4 * inet' in the next superblock. * XXX would be useful to add tables for ports * XXX we can also re-order some mutually exclusive superblocks to * try merging superblocks before any of these optimization passes. * for instance a single 'log in' rule in the middle of non-logging * out rules. */ /* shortcut. there will be a lot of 1-rule superblocks */ if (!TAILQ_NEXT(TAILQ_FIRST(&block->sb_rules), por_entry)) return (0); #ifdef OPT_DEBUG printf("--- Superblock ---\n"); TAILQ_FOREACH(por, &block->sb_rules, por_entry) { printf(" "); print_rule(&por->por_rule, por->por_rule.anchor ? por->por_rule.anchor->name : "", 1, 0); } #endif /* OPT_DEBUG */ if (remove_identical_rules(pf, block)) return (1); if (combine_rules(pf, block)) return (1); if ((pf->optimize & PF_OPTIMIZE_PROFILE) && TAILQ_FIRST(&block->sb_rules)->por_rule.quick && block->sb_profiled_block) { if (block_feedback(pf, block)) return (1); } else if (reorder_rules(pf, block, 0)) { return (1); } /* * Don't add any optimization passes below reorder_rules(). It will * have divided superblocks into smaller blocks for further refinement * and doesn't put them back together again. What once was a true * superblock might have been split into multiple superblocks. */ #ifdef OPT_DEBUG printf("--- END Superblock ---\n"); #endif /* OPT_DEBUG */ return (0); } /* * Optimization pass #1: remove identical rules */ int remove_identical_rules(struct pfctl *pf, struct superblock *block) { struct pf_opt_rule *por1, *por2, *por_next, *por2_next; struct pfctl_rule a, a2, b, b2; for (por1 = TAILQ_FIRST(&block->sb_rules); por1; por1 = por_next) { por_next = TAILQ_NEXT(por1, por_entry); for (por2 = por_next; por2; por2 = por2_next) { por2_next = TAILQ_NEXT(por2, por_entry); comparable_rule(&a, &por1->por_rule, DC); comparable_rule(&b, &por2->por_rule, DC); memcpy(&a2, &a, sizeof(a2)); memcpy(&b2, &b, sizeof(b2)); exclude_supersets(&a, &b); exclude_supersets(&b2, &a2); if (memcmp(&a, &b, sizeof(a)) == 0) { DEBUG("removing identical rule nr%d = *nr%d*", por1->por_rule.nr, por2->por_rule.nr); TAILQ_REMOVE(&block->sb_rules, por2, por_entry); if (por_next == por2) por_next = TAILQ_NEXT(por1, por_entry); free(por2); } else if (memcmp(&a2, &b2, sizeof(a2)) == 0) { DEBUG("removing identical rule *nr%d* = nr%d", por1->por_rule.nr, por2->por_rule.nr); TAILQ_REMOVE(&block->sb_rules, por1, por_entry); free(por1); break; } } } return (0); } /* * Optimization pass #2: combine similar rules with different addresses * into a single rule and a table */ int combine_rules(struct pfctl *pf, struct superblock *block) { struct pf_opt_rule *p1, *p2, *por_next; int src_eq, dst_eq; if ((pf->loadopt & PFCTL_FLAG_TABLE) == 0) { warnx("Must enable table loading for optimizations"); return (1); } /* First we make a pass to combine the rules. O(n log n) */ TAILQ_FOREACH(p1, &block->sb_rules, por_entry) { for (p2 = TAILQ_NEXT(p1, por_entry); p2; p2 = por_next) { por_next = TAILQ_NEXT(p2, por_entry); src_eq = addrs_equal(&p1->por_rule.src, &p2->por_rule.src); dst_eq = addrs_equal(&p1->por_rule.dst, &p2->por_rule.dst); if (src_eq && !dst_eq && p1->por_src_tbl == NULL && p2->por_dst_tbl == NULL && p2->por_src_tbl == NULL && rules_combineable(&p1->por_rule, &p2->por_rule) && addrs_combineable(&p1->por_rule.dst, &p2->por_rule.dst)) { DEBUG("can combine rules nr%d = nr%d", p1->por_rule.nr, p2->por_rule.nr); if (p1->por_dst_tbl == NULL && add_opt_table(pf, &p1->por_dst_tbl, p1->por_rule.af, &p1->por_rule.dst)) return (1); if (add_opt_table(pf, &p1->por_dst_tbl, p1->por_rule.af, &p2->por_rule.dst)) return (1); p2->por_dst_tbl = p1->por_dst_tbl; if (p1->por_dst_tbl->pt_rulecount >= TABLE_THRESHOLD) { TAILQ_REMOVE(&block->sb_rules, p2, por_entry); free(p2); } } else if (!src_eq && dst_eq && p1->por_dst_tbl == NULL && p2->por_src_tbl == NULL && p2->por_dst_tbl == NULL && rules_combineable(&p1->por_rule, &p2->por_rule) && addrs_combineable(&p1->por_rule.src, &p2->por_rule.src)) { DEBUG("can combine rules nr%d = nr%d", p1->por_rule.nr, p2->por_rule.nr); if (p1->por_src_tbl == NULL && add_opt_table(pf, &p1->por_src_tbl, p1->por_rule.af, &p1->por_rule.src)) return (1); if (add_opt_table(pf, &p1->por_src_tbl, p1->por_rule.af, &p2->por_rule.src)) return (1); p2->por_src_tbl = p1->por_src_tbl; if (p1->por_src_tbl->pt_rulecount >= TABLE_THRESHOLD) { TAILQ_REMOVE(&block->sb_rules, p2, por_entry); free(p2); } } } } /* * Then we make a final pass to create a valid table name and * insert the name into the rules. */ for (p1 = TAILQ_FIRST(&block->sb_rules); p1; p1 = por_next) { por_next = TAILQ_NEXT(p1, por_entry); assert(p1->por_src_tbl == NULL || p1->por_dst_tbl == NULL); if (p1->por_src_tbl && p1->por_src_tbl->pt_rulecount >= TABLE_THRESHOLD) { if (p1->por_src_tbl->pt_generated) { /* This rule is included in a table */ TAILQ_REMOVE(&block->sb_rules, p1, por_entry); free(p1); continue; } p1->por_src_tbl->pt_generated = 1; if ((pf->opts & PF_OPT_NOACTION) == 0 && pf_opt_create_table(pf, p1->por_src_tbl)) return (1); pf->tdirty = 1; if (pf->opts & PF_OPT_VERBOSE) print_tabledef(p1->por_src_tbl->pt_name, PFR_TFLAG_CONST, 1, &p1->por_src_tbl->pt_nodes); memset(&p1->por_rule.src.addr, 0, sizeof(p1->por_rule.src.addr)); p1->por_rule.src.addr.type = PF_ADDR_TABLE; strlcpy(p1->por_rule.src.addr.v.tblname, p1->por_src_tbl->pt_name, sizeof(p1->por_rule.src.addr.v.tblname)); pfr_buf_clear(p1->por_src_tbl->pt_buf); free(p1->por_src_tbl->pt_buf); p1->por_src_tbl->pt_buf = NULL; } if (p1->por_dst_tbl && p1->por_dst_tbl->pt_rulecount >= TABLE_THRESHOLD) { if (p1->por_dst_tbl->pt_generated) { /* This rule is included in a table */ TAILQ_REMOVE(&block->sb_rules, p1, por_entry); free(p1); continue; } p1->por_dst_tbl->pt_generated = 1; if ((pf->opts & PF_OPT_NOACTION) == 0 && pf_opt_create_table(pf, p1->por_dst_tbl)) return (1); pf->tdirty = 1; if (pf->opts & PF_OPT_VERBOSE) print_tabledef(p1->por_dst_tbl->pt_name, PFR_TFLAG_CONST, 1, &p1->por_dst_tbl->pt_nodes); memset(&p1->por_rule.dst.addr, 0, sizeof(p1->por_rule.dst.addr)); p1->por_rule.dst.addr.type = PF_ADDR_TABLE; strlcpy(p1->por_rule.dst.addr.v.tblname, p1->por_dst_tbl->pt_name, sizeof(p1->por_rule.dst.addr.v.tblname)); pfr_buf_clear(p1->por_dst_tbl->pt_buf); free(p1->por_dst_tbl->pt_buf); p1->por_dst_tbl->pt_buf = NULL; } } return (0); } /* * Optimization pass #3: re-order rules to improve skip steps */ int reorder_rules(struct pfctl *pf, struct superblock *block, int depth) { struct superblock *newblock; struct pf_skip_step *skiplist; struct pf_opt_rule *por; int i, largest, largest_list, rule_count = 0; TAILQ_HEAD( , pf_opt_rule) head; /* * Calculate the best-case skip steps. We put each rule in a list * of other rules with common fields */ for (i = 0; i < PF_SKIP_COUNT; i++) { TAILQ_FOREACH(por, &block->sb_rules, por_entry) { TAILQ_FOREACH(skiplist, &block->sb_skipsteps[i], ps_entry) { if (skip_compare(i, skiplist, por) == 0) break; } if (skiplist == NULL) { if ((skiplist = calloc(1, sizeof(*skiplist))) == NULL) err(1, "calloc"); TAILQ_INIT(&skiplist->ps_rules); TAILQ_INSERT_TAIL(&block->sb_skipsteps[i], skiplist, ps_entry); } skip_append(block, i, skiplist, por); } } TAILQ_FOREACH(por, &block->sb_rules, por_entry) rule_count++; /* * Now we're going to ignore any fields that are identical between * all of the rules in the superblock and those fields which differ * between every rule in the superblock. */ largest = 0; for (i = 0; i < PF_SKIP_COUNT; i++) { skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]); if (skiplist->ps_count == rule_count) { DEBUG("(%d) original skipstep '%s' is all rules", depth, skip_comparitors_names[i]); skiplist->ps_count = 0; } else if (skiplist->ps_count == 1) { skiplist->ps_count = 0; } else { DEBUG("(%d) original skipstep '%s' largest jump is %d", depth, skip_comparitors_names[i], skiplist->ps_count); if (skiplist->ps_count > largest) largest = skiplist->ps_count; } } if (largest == 0) { /* Ugh. There is NO commonality in the superblock on which * optimize the skipsteps optimization. */ goto done; } /* * Now we're going to empty the superblock rule list and re-create * it based on a more optimal skipstep order. */ TAILQ_INIT(&head); while ((por = TAILQ_FIRST(&block->sb_rules))) { TAILQ_REMOVE(&block->sb_rules, por, por_entry); TAILQ_INSERT_TAIL(&head, por, por_entry); } while (!TAILQ_EMPTY(&head)) { largest = 1; /* * Find the most useful skip steps remaining */ for (i = 0; i < PF_SKIP_COUNT; i++) { skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]); if (skiplist->ps_count > largest) { largest = skiplist->ps_count; largest_list = i; } } if (largest <= 1) { /* * Nothing useful left. Leave remaining rules in order. */ DEBUG("(%d) no more commonality for skip steps", depth); while ((por = TAILQ_FIRST(&head))) { TAILQ_REMOVE(&head, por, por_entry); TAILQ_INSERT_TAIL(&block->sb_rules, por, por_entry); } } else { /* * There is commonality. Extract those common rules * and place them in the ruleset adjacent to each * other. */ skiplist = TAILQ_FIRST(&block->sb_skipsteps[ largest_list]); DEBUG("(%d) skipstep '%s' largest jump is %d @ #%d", depth, skip_comparitors_names[largest_list], largest, TAILQ_FIRST(&TAILQ_FIRST(&block-> sb_skipsteps [largest_list])->ps_rules)-> por_rule.nr); TAILQ_REMOVE(&block->sb_skipsteps[largest_list], skiplist, ps_entry); /* * There may be further commonality inside these * rules. So we'll split them off into they're own * superblock and pass it back into the optimizer. */ if (skiplist->ps_count > 2) { if ((newblock = calloc(1, sizeof(*newblock))) == NULL) { warn("calloc"); return (1); } TAILQ_INIT(&newblock->sb_rules); for (i = 0; i < PF_SKIP_COUNT; i++) TAILQ_INIT(&newblock->sb_skipsteps[i]); TAILQ_INSERT_BEFORE(block, newblock, sb_entry); DEBUG("(%d) splitting off %d rules from superblock @ #%d", depth, skiplist->ps_count, TAILQ_FIRST(&skiplist->ps_rules)-> por_rule.nr); } else { newblock = block; } while ((por = TAILQ_FIRST(&skiplist->ps_rules))) { TAILQ_REMOVE(&head, por, por_entry); TAILQ_REMOVE(&skiplist->ps_rules, por, por_skip_entry[largest_list]); TAILQ_INSERT_TAIL(&newblock->sb_rules, por, por_entry); /* Remove this rule from all other skiplists */ remove_from_skipsteps(&block->sb_skipsteps[ largest_list], block, por, skiplist); } free(skiplist); if (newblock != block) if (reorder_rules(pf, newblock, depth + 1)) return (1); } } done: for (i = 0; i < PF_SKIP_COUNT; i++) { while ((skiplist = TAILQ_FIRST(&block->sb_skipsteps[i]))) { TAILQ_REMOVE(&block->sb_skipsteps[i], skiplist, ps_entry); free(skiplist); } } return (0); } /* * Optimization pass #4: re-order 'quick' rules based on feedback from the * currently running ruleset */ int block_feedback(struct pfctl *pf, struct superblock *block) { TAILQ_HEAD( , pf_opt_rule) queue; struct pf_opt_rule *por1, *por2; struct pfctl_rule a, b; /* * Walk through all of the profiled superblock's rules and copy * the counters onto our rules. */ TAILQ_FOREACH(por1, &block->sb_profiled_block->sb_rules, por_entry) { comparable_rule(&a, &por1->por_rule, DC); TAILQ_FOREACH(por2, &block->sb_rules, por_entry) { if (por2->por_profile_count) continue; comparable_rule(&b, &por2->por_rule, DC); if (memcmp(&a, &b, sizeof(a)) == 0) { por2->por_profile_count = por1->por_rule.packets[0] + por1->por_rule.packets[1]; break; } } } superblock_free(pf, block->sb_profiled_block); block->sb_profiled_block = NULL; /* * Now we pull all of the rules off the superblock and re-insert them * in sorted order. */ TAILQ_INIT(&queue); while ((por1 = TAILQ_FIRST(&block->sb_rules)) != NULL) { TAILQ_REMOVE(&block->sb_rules, por1, por_entry); TAILQ_INSERT_TAIL(&queue, por1, por_entry); } while ((por1 = TAILQ_FIRST(&queue)) != NULL) { TAILQ_REMOVE(&queue, por1, por_entry); /* XXX I should sort all of the unused rules based on skip steps */ TAILQ_FOREACH(por2, &block->sb_rules, por_entry) { if (por1->por_profile_count > por2->por_profile_count) { TAILQ_INSERT_BEFORE(por2, por1, por_entry); break; } } #ifdef __FreeBSD__ if (por2 == NULL) #else if (por2 == TAILQ_END(&block->sb_rules)) #endif TAILQ_INSERT_TAIL(&block->sb_rules, por1, por_entry); } return (0); } /* * Load the current ruleset from the kernel and try to associate them with * the ruleset we're optimizing. */ int load_feedback_profile(struct pfctl *pf, struct superblocks *superblocks) { char anchor_call[MAXPATHLEN] = ""; struct superblock *block, *blockcur; struct superblocks prof_superblocks; struct pf_opt_rule *por; struct pf_opt_queue queue; struct pfctl_rules_info rules; struct pfctl_rule a, b, rule; int nr, mnr; TAILQ_INIT(&queue); TAILQ_INIT(&prof_superblocks); if (pfctl_get_rules_info_h(pf->h, &rules, PF_PASS, "")) { warn("DIOCGETRULES"); return (1); } mnr = rules.nr; DEBUG("Loading %d active rules for a feedback profile", mnr); for (nr = 0; nr < mnr; ++nr) { struct pfctl_ruleset *rs; if ((por = calloc(1, sizeof(*por))) == NULL) { warn("calloc"); return (1); } if (pfctl_get_rule_h(pf->h, nr, rules.ticket, "", PF_PASS, &rule, anchor_call)) { warn("DIOCGETRULENV"); return (1); } memcpy(&por->por_rule, &rule, sizeof(por->por_rule)); rs = pf_find_or_create_ruleset(anchor_call); por->por_rule.anchor = rs->anchor; if (TAILQ_EMPTY(&por->por_rule.rpool.list)) memset(&por->por_rule.rpool, 0, sizeof(por->por_rule.rpool)); TAILQ_INSERT_TAIL(&queue, por, por_entry); /* XXX pfctl_get_pool(pf->dev, &rule.rpool, nr, pr.ticket, * PF_PASS, pf->anchor) ??? * ... pfctl_clear_pool(&rule.rpool) */ } if (construct_superblocks(pf, &queue, &prof_superblocks)) return (1); /* * Now we try to associate the active ruleset's superblocks with * the superblocks we're compiling. */ block = TAILQ_FIRST(superblocks); blockcur = TAILQ_FIRST(&prof_superblocks); while (block && blockcur) { comparable_rule(&a, &TAILQ_FIRST(&block->sb_rules)->por_rule, BREAK); comparable_rule(&b, &TAILQ_FIRST(&blockcur->sb_rules)->por_rule, BREAK); if (memcmp(&a, &b, sizeof(a)) == 0) { /* The two superblocks lined up */ block->sb_profiled_block = blockcur; } else { DEBUG("superblocks don't line up between #%d and #%d", TAILQ_FIRST(&block->sb_rules)->por_rule.nr, TAILQ_FIRST(&blockcur->sb_rules)->por_rule.nr); break; } block = TAILQ_NEXT(block, sb_entry); blockcur = TAILQ_NEXT(blockcur, sb_entry); } /* Free any superblocks we couldn't link */ while (blockcur) { block = TAILQ_NEXT(blockcur, sb_entry); superblock_free(pf, blockcur); blockcur = block; } return (0); } /* * Compare a rule to a skiplist to see if the rule is a member */ int skip_compare(int skipnum, struct pf_skip_step *skiplist, struct pf_opt_rule *por) { struct pfctl_rule *a, *b; if (skipnum >= PF_SKIP_COUNT || skipnum < 0) errx(1, "skip_compare() out of bounds"); a = &por->por_rule; b = &TAILQ_FIRST(&skiplist->ps_rules)->por_rule; return ((skip_comparitors[skipnum])(a, b)); } /* * Add a rule to a skiplist */ void skip_append(struct superblock *superblock, int skipnum, struct pf_skip_step *skiplist, struct pf_opt_rule *por) { struct pf_skip_step *prev; skiplist->ps_count++; TAILQ_INSERT_TAIL(&skiplist->ps_rules, por, por_skip_entry[skipnum]); /* Keep the list of skiplists sorted by whichever is larger */ while ((prev = TAILQ_PREV(skiplist, skiplist, ps_entry)) && prev->ps_count < skiplist->ps_count) { TAILQ_REMOVE(&superblock->sb_skipsteps[skipnum], skiplist, ps_entry); TAILQ_INSERT_BEFORE(prev, skiplist, ps_entry); } } /* * Remove a rule from the other skiplist calculations. */ void remove_from_skipsteps(struct skiplist *head, struct superblock *block, struct pf_opt_rule *por, struct pf_skip_step *active_list) { struct pf_skip_step *sk, *next; struct pf_opt_rule *p2; int i, found; for (i = 0; i < PF_SKIP_COUNT; i++) { sk = TAILQ_FIRST(&block->sb_skipsteps[i]); if (sk == NULL || sk == active_list || sk->ps_count <= 1) continue; found = 0; do { TAILQ_FOREACH(p2, &sk->ps_rules, por_skip_entry[i]) if (p2 == por) { TAILQ_REMOVE(&sk->ps_rules, p2, por_skip_entry[i]); found = 1; sk->ps_count--; break; } } while (!found && (sk = TAILQ_NEXT(sk, ps_entry))); if (found && sk) { /* Does this change the sorting order? */ while ((next = TAILQ_NEXT(sk, ps_entry)) && next->ps_count > sk->ps_count) { TAILQ_REMOVE(head, sk, ps_entry); TAILQ_INSERT_AFTER(head, next, sk, ps_entry); } #ifdef OPT_DEBUG next = TAILQ_NEXT(sk, ps_entry); assert(next == NULL || next->ps_count <= sk->ps_count); #endif /* OPT_DEBUG */ } } } /* Compare two rules AF field for skiplist construction */ int skip_cmp_af(struct pfctl_rule *a, struct pfctl_rule *b) { if (a->af != b->af || a->af == 0) return (1); return (0); } /* Compare two rules DIRECTION field for skiplist construction */ int skip_cmp_dir(struct pfctl_rule *a, struct pfctl_rule *b) { if (a->direction == 0 || a->direction != b->direction) return (1); return (0); } /* Compare two rules DST Address field for skiplist construction */ int skip_cmp_dst_addr(struct pfctl_rule *a, struct pfctl_rule *b) { if (a->dst.neg != b->dst.neg || a->dst.addr.type != b->dst.addr.type) return (1); /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP || * a->proto == IPPROTO_ICMP * return (1); */ switch (a->dst.addr.type) { case PF_ADDR_ADDRMASK: if (memcmp(&a->dst.addr.v.a.addr, &b->dst.addr.v.a.addr, sizeof(a->dst.addr.v.a.addr)) || memcmp(&a->dst.addr.v.a.mask, &b->dst.addr.v.a.mask, sizeof(a->dst.addr.v.a.mask)) || (a->dst.addr.v.a.addr.addr32[0] == 0 && a->dst.addr.v.a.addr.addr32[1] == 0 && a->dst.addr.v.a.addr.addr32[2] == 0 && a->dst.addr.v.a.addr.addr32[3] == 0)) return (1); return (0); case PF_ADDR_DYNIFTL: if (strcmp(a->dst.addr.v.ifname, b->dst.addr.v.ifname) != 0 || a->dst.addr.iflags != b->dst.addr.iflags || memcmp(&a->dst.addr.v.a.mask, &b->dst.addr.v.a.mask, sizeof(a->dst.addr.v.a.mask))) return (1); return (0); case PF_ADDR_NOROUTE: case PF_ADDR_URPFFAILED: return (0); case PF_ADDR_TABLE: return (strcmp(a->dst.addr.v.tblname, b->dst.addr.v.tblname)); } return (1); } /* Compare two rules DST port field for skiplist construction */ int skip_cmp_dst_port(struct pfctl_rule *a, struct pfctl_rule *b) { /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP || * a->proto == IPPROTO_ICMP * return (1); */ if (a->dst.port_op == PF_OP_NONE || a->dst.port_op != b->dst.port_op || a->dst.port[0] != b->dst.port[0] || a->dst.port[1] != b->dst.port[1]) return (1); return (0); } /* Compare two rules IFP field for skiplist construction */ int skip_cmp_ifp(struct pfctl_rule *a, struct pfctl_rule *b) { if (strcmp(a->ifname, b->ifname) || a->ifname[0] == '\0') return (1); return (a->ifnot != b->ifnot); } /* Compare two rules PROTO field for skiplist construction */ int skip_cmp_proto(struct pfctl_rule *a, struct pfctl_rule *b) { return (a->proto != b->proto || a->proto == 0); } /* Compare two rules SRC addr field for skiplist construction */ int skip_cmp_src_addr(struct pfctl_rule *a, struct pfctl_rule *b) { if (a->src.neg != b->src.neg || a->src.addr.type != b->src.addr.type) return (1); /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP || * a->proto == IPPROTO_ICMP * return (1); */ switch (a->src.addr.type) { case PF_ADDR_ADDRMASK: if (memcmp(&a->src.addr.v.a.addr, &b->src.addr.v.a.addr, sizeof(a->src.addr.v.a.addr)) || memcmp(&a->src.addr.v.a.mask, &b->src.addr.v.a.mask, sizeof(a->src.addr.v.a.mask)) || (a->src.addr.v.a.addr.addr32[0] == 0 && a->src.addr.v.a.addr.addr32[1] == 0 && a->src.addr.v.a.addr.addr32[2] == 0 && a->src.addr.v.a.addr.addr32[3] == 0)) return (1); return (0); case PF_ADDR_DYNIFTL: if (strcmp(a->src.addr.v.ifname, b->src.addr.v.ifname) != 0 || a->src.addr.iflags != b->src.addr.iflags || memcmp(&a->src.addr.v.a.mask, &b->src.addr.v.a.mask, sizeof(a->src.addr.v.a.mask))) return (1); return (0); case PF_ADDR_NOROUTE: case PF_ADDR_URPFFAILED: return (0); case PF_ADDR_TABLE: return (strcmp(a->src.addr.v.tblname, b->src.addr.v.tblname)); } return (1); } /* Compare two rules SRC port field for skiplist construction */ int skip_cmp_src_port(struct pfctl_rule *a, struct pfctl_rule *b) { if (a->src.port_op == PF_OP_NONE || a->src.port_op != b->src.port_op || a->src.port[0] != b->src.port[0] || a->src.port[1] != b->src.port[1]) return (1); /* XXX if (a->proto != b->proto && a->proto != 0 && b->proto != 0 * && (a->proto == IPPROTO_TCP || a->proto == IPPROTO_UDP || * a->proto == IPPROTO_ICMP * return (1); */ return (0); } void skip_init(void) { struct { char *name; int skipnum; int (*func)(struct pfctl_rule *, struct pfctl_rule *); } comps[] = PF_SKIP_COMPARITORS; int skipnum, i; for (skipnum = 0; skipnum < PF_SKIP_COUNT; skipnum++) { for (i = 0; i < sizeof(comps)/sizeof(*comps); i++) if (comps[i].skipnum == skipnum) { skip_comparitors[skipnum] = comps[i].func; skip_comparitors_names[skipnum] = comps[i].name; } } for (skipnum = 0; skipnum < PF_SKIP_COUNT; skipnum++) if (skip_comparitors[skipnum] == NULL) errx(1, "Need to add skip step comparitor to pfctl?!"); } /* * Add a host/netmask to a table */ int add_opt_table(struct pfctl *pf, struct pf_opt_tbl **tbl, sa_family_t af, struct pf_rule_addr *addr) { #ifdef OPT_DEBUG char buf[128]; #endif /* OPT_DEBUG */ static int tablenum = 0; struct node_host node_host; if (*tbl == NULL) { if ((*tbl = calloc(1, sizeof(**tbl))) == NULL || ((*tbl)->pt_buf = calloc(1, sizeof(*(*tbl)->pt_buf))) == NULL) err(1, "calloc"); (*tbl)->pt_buf->pfrb_type = PFRB_ADDRS; SIMPLEQ_INIT(&(*tbl)->pt_nodes); /* This is just a temporary table name */ snprintf((*tbl)->pt_name, sizeof((*tbl)->pt_name), "%s%d", PF_OPT_TABLE_PREFIX, tablenum++); DEBUG("creating table <%s>", (*tbl)->pt_name); } memset(&node_host, 0, sizeof(node_host)); node_host.af = af; node_host.addr = addr->addr; #ifdef OPT_DEBUG DEBUG("<%s> adding %s/%d", (*tbl)->pt_name, inet_ntop(af, &node_host.addr.v.a.addr, buf, sizeof(buf)), unmask(&node_host.addr.v.a.mask, af)); #endif /* OPT_DEBUG */ if (append_addr_host((*tbl)->pt_buf, &node_host, 0, 0)) { warn("failed to add host"); return (1); } if (pf->opts & PF_OPT_VERBOSE) { struct node_tinit *ti; if ((ti = calloc(1, sizeof(*ti))) == NULL) err(1, "malloc"); if ((ti->host = malloc(sizeof(*ti->host))) == NULL) err(1, "malloc"); memcpy(ti->host, &node_host, sizeof(*ti->host)); SIMPLEQ_INSERT_TAIL(&(*tbl)->pt_nodes, ti, entries); } (*tbl)->pt_rulecount++; if ((*tbl)->pt_rulecount == TABLE_THRESHOLD) DEBUG("table <%s> now faster than skip steps", (*tbl)->pt_name); return (0); } /* * Do the dirty work of choosing an unused table name and creating it. * (be careful with the table name, it might already be used in another anchor) */ int pf_opt_create_table(struct pfctl *pf, struct pf_opt_tbl *tbl) { static int tablenum; struct pfr_table *t; if (table_buffer.pfrb_type == 0) { /* Initialize the list of tables */ table_buffer.pfrb_type = PFRB_TABLES; for (;;) { pfr_buf_grow(&table_buffer, table_buffer.pfrb_size); table_buffer.pfrb_size = table_buffer.pfrb_msize; if (pfr_get_tables(NULL, table_buffer.pfrb_caddr, &table_buffer.pfrb_size, PFR_FLAG_ALLRSETS)) err(1, "pfr_get_tables"); if (table_buffer.pfrb_size <= table_buffer.pfrb_msize) break; } table_identifier = arc4random(); } /* XXX would be *really* nice to avoid duplicating identical tables */ /* Now we have to pick a table name that isn't used */ again: DEBUG("translating temporary table <%s> to <%s%x_%d>", tbl->pt_name, PF_OPT_TABLE_PREFIX, table_identifier, tablenum); snprintf(tbl->pt_name, sizeof(tbl->pt_name), "%s%x_%d", PF_OPT_TABLE_PREFIX, table_identifier, tablenum); PFRB_FOREACH(t, &table_buffer) { if (strcasecmp(t->pfrt_name, tbl->pt_name) == 0) { /* Collision. Try again */ DEBUG("wow, table <%s> in use. trying again", tbl->pt_name); table_identifier = arc4random(); goto again; } } tablenum++; if (pfctl_define_table(tbl->pt_name, PFR_TFLAG_CONST, 1, pf->astack[0]->name, tbl->pt_buf, pf->astack[0]->ruleset.tticket)) { warn("failed to create table %s in %s", tbl->pt_name, pf->astack[0]->name); return (1); } return (0); } /* * Partition the flat ruleset into a list of distinct superblocks */ int construct_superblocks(struct pfctl *pf, struct pf_opt_queue *opt_queue, struct superblocks *superblocks) { struct superblock *block = NULL; struct pf_opt_rule *por; int i; while (!TAILQ_EMPTY(opt_queue)) { por = TAILQ_FIRST(opt_queue); TAILQ_REMOVE(opt_queue, por, por_entry); if (block == NULL || !superblock_inclusive(block, por)) { if ((block = calloc(1, sizeof(*block))) == NULL) { warn("calloc"); return (1); } TAILQ_INIT(&block->sb_rules); for (i = 0; i < PF_SKIP_COUNT; i++) TAILQ_INIT(&block->sb_skipsteps[i]); TAILQ_INSERT_TAIL(superblocks, block, sb_entry); } TAILQ_INSERT_TAIL(&block->sb_rules, por, por_entry); } return (0); } /* * Compare two rule addresses */ int addrs_equal(struct pf_rule_addr *a, struct pf_rule_addr *b) { if (a->neg != b->neg) return (0); return (memcmp(&a->addr, &b->addr, sizeof(a->addr)) == 0); } /* * The addresses are not equal, but can we combine them into one table? */ int addrs_combineable(struct pf_rule_addr *a, struct pf_rule_addr *b) { if (a->addr.type != PF_ADDR_ADDRMASK || b->addr.type != PF_ADDR_ADDRMASK) return (0); if (a->neg != b->neg || a->port_op != b->port_op || a->port[0] != b->port[0] || a->port[1] != b->port[1]) return (0); return (1); } /* * Are we allowed to combine these two rules */ int rules_combineable(struct pfctl_rule *p1, struct pfctl_rule *p2) { struct pfctl_rule a, b; comparable_rule(&a, p1, COMBINED); comparable_rule(&b, p2, COMBINED); return (memcmp(&a, &b, sizeof(a)) == 0); } /* * Can a rule be included inside a superblock */ int superblock_inclusive(struct superblock *block, struct pf_opt_rule *por) { struct pfctl_rule a, b; int i, j; /* First check for hard breaks */ for (i = 0; i < sizeof(pf_rule_desc)/sizeof(*pf_rule_desc); i++) { if (pf_rule_desc[i].prf_type == BARRIER) { for (j = 0; j < pf_rule_desc[i].prf_size; j++) if (((char *)&por->por_rule)[j + pf_rule_desc[i].prf_offset] != 0) return (0); } } /* per-rule src-track is also a hard break */ if (por->por_rule.rule_flag & PFRULE_RULESRCTRACK) return (0); /* * Have to handle interface groups separately. Consider the following * rules: * block on EXTIFS to any port 22 * pass on em0 to any port 22 * (where EXTIFS is an arbitrary interface group) * The optimizer may decide to re-order the pass rule in front of the * block rule. But what if EXTIFS includes em0??? Such a reordering * would change the meaning of the ruleset. * We can't just lookup the EXTIFS group and check if em0 is a member * because the user is allowed to add interfaces to a group during * runtime. * Ergo interface groups become a defacto superblock break :-( */ if (interface_group(por->por_rule.ifname) || interface_group(TAILQ_FIRST(&block->sb_rules)->por_rule.ifname)) { if (strcasecmp(por->por_rule.ifname, TAILQ_FIRST(&block->sb_rules)->por_rule.ifname) != 0) return (0); } comparable_rule(&a, &TAILQ_FIRST(&block->sb_rules)->por_rule, NOMERGE); comparable_rule(&b, &por->por_rule, NOMERGE); if (memcmp(&a, &b, sizeof(a)) == 0) return (1); #ifdef OPT_DEBUG for (i = 0; i < sizeof(por->por_rule); i++) { int closest = -1; if (((u_int8_t *)&a)[i] != ((u_int8_t *)&b)[i]) { for (j = 0; j < sizeof(pf_rule_desc) / sizeof(*pf_rule_desc); j++) { if (i >= pf_rule_desc[j].prf_offset && i < pf_rule_desc[j].prf_offset + pf_rule_desc[j].prf_size) { DEBUG("superblock break @ %d due to %s", por->por_rule.nr, pf_rule_desc[j].prf_name); return (0); } if (i > pf_rule_desc[j].prf_offset) { if (closest == -1 || i-pf_rule_desc[j].prf_offset < i-pf_rule_desc[closest].prf_offset) closest = j; } } if (closest >= 0) DEBUG("superblock break @ %d on %s+%zxh", por->por_rule.nr, pf_rule_desc[closest].prf_name, i - pf_rule_desc[closest].prf_offset - pf_rule_desc[closest].prf_size); else DEBUG("superblock break @ %d on field @ %d", por->por_rule.nr, i); return (0); } } #endif /* OPT_DEBUG */ return (0); } /* * Figure out if an interface name is an actual interface or actually a * group of interfaces. */ int interface_group(const char *ifname) { int s; struct ifgroupreq ifgr; if (ifname == NULL || !ifname[0]) return (0); s = get_query_socket(); memset(&ifgr, 0, sizeof(ifgr)); strlcpy(ifgr.ifgr_name, ifname, IFNAMSIZ); if (ioctl(s, SIOCGIFGMEMB, (caddr_t)&ifgr) == -1) { if (errno == ENOENT) return (0); else err(1, "SIOCGIFGMEMB"); } return (1); } /* * Make a rule that can directly compared by memcmp() */ void comparable_rule(struct pfctl_rule *dst, const struct pfctl_rule *src, int type) { int i; /* * To simplify the comparison, we just zero out the fields that are * allowed to be different and then do a simple memcmp() */ memcpy(dst, src, sizeof(*dst)); for (i = 0; i < sizeof(pf_rule_desc)/sizeof(*pf_rule_desc); i++) if (pf_rule_desc[i].prf_type >= type) { #ifdef OPT_DEBUG assert(pf_rule_desc[i].prf_type != NEVER || *(((char *)dst) + pf_rule_desc[i].prf_offset) == 0); #endif /* OPT_DEBUG */ memset(((char *)dst) + pf_rule_desc[i].prf_offset, 0, pf_rule_desc[i].prf_size); } } /* * Remove superset information from two rules so we can directly compare them * with memcmp() */ void exclude_supersets(struct pfctl_rule *super, struct pfctl_rule *sub) { if (super->ifname[0] == '\0') memset(sub->ifname, 0, sizeof(sub->ifname)); if (super->direction == PF_INOUT) sub->direction = PF_INOUT; if ((super->proto == 0 || super->proto == sub->proto) && super->flags == 0 && super->flagset == 0 && (sub->flags || sub->flagset)) { sub->flags = super->flags; sub->flagset = super->flagset; } if (super->proto == 0) sub->proto = 0; if (super->src.port_op == 0) { sub->src.port_op = 0; sub->src.port[0] = 0; sub->src.port[1] = 0; } if (super->dst.port_op == 0) { sub->dst.port_op = 0; sub->dst.port[0] = 0; sub->dst.port[1] = 0; } if (super->src.addr.type == PF_ADDR_ADDRMASK && !super->src.neg && !sub->src.neg && super->src.addr.v.a.mask.addr32[0] == 0 && super->src.addr.v.a.mask.addr32[1] == 0 && super->src.addr.v.a.mask.addr32[2] == 0 && super->src.addr.v.a.mask.addr32[3] == 0) memset(&sub->src.addr, 0, sizeof(sub->src.addr)); else if (super->src.addr.type == PF_ADDR_ADDRMASK && sub->src.addr.type == PF_ADDR_ADDRMASK && super->src.neg == sub->src.neg && super->af == sub->af && unmask(&super->src.addr.v.a.mask, super->af) < unmask(&sub->src.addr.v.a.mask, sub->af) && super->src.addr.v.a.addr.addr32[0] == (sub->src.addr.v.a.addr.addr32[0] & super->src.addr.v.a.mask.addr32[0]) && super->src.addr.v.a.addr.addr32[1] == (sub->src.addr.v.a.addr.addr32[1] & super->src.addr.v.a.mask.addr32[1]) && super->src.addr.v.a.addr.addr32[2] == (sub->src.addr.v.a.addr.addr32[2] & super->src.addr.v.a.mask.addr32[2]) && super->src.addr.v.a.addr.addr32[3] == (sub->src.addr.v.a.addr.addr32[3] & super->src.addr.v.a.mask.addr32[3])) { /* sub->src.addr is a subset of super->src.addr/mask */ memcpy(&sub->src.addr, &super->src.addr, sizeof(sub->src.addr)); } if (super->dst.addr.type == PF_ADDR_ADDRMASK && !super->dst.neg && !sub->dst.neg && super->dst.addr.v.a.mask.addr32[0] == 0 && super->dst.addr.v.a.mask.addr32[1] == 0 && super->dst.addr.v.a.mask.addr32[2] == 0 && super->dst.addr.v.a.mask.addr32[3] == 0) memset(&sub->dst.addr, 0, sizeof(sub->dst.addr)); else if (super->dst.addr.type == PF_ADDR_ADDRMASK && sub->dst.addr.type == PF_ADDR_ADDRMASK && super->dst.neg == sub->dst.neg && super->af == sub->af && unmask(&super->dst.addr.v.a.mask, super->af) < unmask(&sub->dst.addr.v.a.mask, sub->af) && super->dst.addr.v.a.addr.addr32[0] == (sub->dst.addr.v.a.addr.addr32[0] & super->dst.addr.v.a.mask.addr32[0]) && super->dst.addr.v.a.addr.addr32[1] == (sub->dst.addr.v.a.addr.addr32[1] & super->dst.addr.v.a.mask.addr32[1]) && super->dst.addr.v.a.addr.addr32[2] == (sub->dst.addr.v.a.addr.addr32[2] & super->dst.addr.v.a.mask.addr32[2]) && super->dst.addr.v.a.addr.addr32[3] == (sub->dst.addr.v.a.addr.addr32[3] & super->dst.addr.v.a.mask.addr32[3])) { /* sub->dst.addr is a subset of super->dst.addr/mask */ memcpy(&sub->dst.addr, &super->dst.addr, sizeof(sub->dst.addr)); } if (super->af == 0) sub->af = 0; } void superblock_free(struct pfctl *pf, struct superblock *block) { struct pf_opt_rule *por; while ((por = TAILQ_FIRST(&block->sb_rules))) { TAILQ_REMOVE(&block->sb_rules, por, por_entry); if (por->por_src_tbl) { if (por->por_src_tbl->pt_buf) { pfr_buf_clear(por->por_src_tbl->pt_buf); free(por->por_src_tbl->pt_buf); } free(por->por_src_tbl); } if (por->por_dst_tbl) { if (por->por_dst_tbl->pt_buf) { pfr_buf_clear(por->por_dst_tbl->pt_buf); free(por->por_dst_tbl->pt_buf); } free(por->por_dst_tbl); } free(por); } if (block->sb_profiled_block) superblock_free(pf, block->sb_profiled_block); free(block); } diff --git a/sbin/pfctl/tests/files/pf0004.ok b/sbin/pfctl/tests/files/pf0004.ok index 5fca4a50f7b1..87b71cdeff3d 100644 --- a/sbin/pfctl/tests/files/pf0004.ok +++ b/sbin/pfctl/tests/files/pf0004.ok @@ -1,62 +1,62 @@ block drop in all block drop in proto tcp all block drop in proto tcp all block drop in proto udp all block drop in all block drop in inet from 10.0.0.0/8 to any block drop in inet from ! 10.0.0.0/8 to any block drop in inet from 10.0.0.0/8 to any block drop in inet from 172.16.0.0/12 to any block drop in proto tcp from any port = ssh to any block drop in proto tcp from any port = ssh to any port 1024:2048 block drop in proto tcp from any port 21 >< 2048 to any port 1024:2048 block drop in proto tcp from any port != 1234 to any port 1024:2048 block drop in proto tcp from any port >= 80 to any port 1024:2048 block drop in inet proto tcp from 10.0.0.0/8 port = ssh to 192.168.0.0/16 port = ircd block drop in inet proto tcp from 10.0.0.0/8 port = ssh to 192.168.0.0/16 port = 6668 block drop in inet proto tcp from 10.0.0.0/8 port = ssh to 192.168.0.0/16 port 6669:65535 -block drop in inet proto tcp from 10.0.0.0/8 port = ssh to 12.34.56.78 port = ircd -block drop in inet proto tcp from 10.0.0.0/8 port = ssh to 12.34.56.78 port = 6668 -block drop in inet proto tcp from 10.0.0.0/8 port = ssh to 12.34.56.78 port 6669:65535 block drop in inet proto tcp from 10.0.0.0/8 port = ftp to 192.168.0.0/16 port = ircd block drop in inet proto tcp from 10.0.0.0/8 port = ftp to 192.168.0.0/16 port = 6668 block drop in inet proto tcp from 10.0.0.0/8 port = ftp to 192.168.0.0/16 port 6669:65535 +block drop in inet proto tcp from 10.0.0.0/8 port = ssh to 12.34.56.78 port = ircd +block drop in inet proto tcp from 10.0.0.0/8 port = ssh to 12.34.56.78 port = 6668 +block drop in inet proto tcp from 10.0.0.0/8 port = ssh to 12.34.56.78 port 6669:65535 block drop in inet proto tcp from 10.0.0.0/8 port = ftp to 12.34.56.78 port = ircd block drop in inet proto tcp from 10.0.0.0/8 port = ftp to 12.34.56.78 port = 6668 block drop in inet proto tcp from 10.0.0.0/8 port = ftp to 12.34.56.78 port 6669:65535 block drop in inet proto tcp from 172.16.0.0/12 port = ssh to 192.168.0.0/16 port = ircd block drop in inet proto tcp from 172.16.0.0/12 port = ssh to 192.168.0.0/16 port = 6668 block drop in inet proto tcp from 172.16.0.0/12 port = ssh to 192.168.0.0/16 port 6669:65535 -block drop in inet proto tcp from 172.16.0.0/12 port = ssh to 12.34.56.78 port = ircd -block drop in inet proto tcp from 172.16.0.0/12 port = ssh to 12.34.56.78 port = 6668 -block drop in inet proto tcp from 172.16.0.0/12 port = ssh to 12.34.56.78 port 6669:65535 block drop in inet proto tcp from 172.16.0.0/12 port = ftp to 192.168.0.0/16 port = ircd block drop in inet proto tcp from 172.16.0.0/12 port = ftp to 192.168.0.0/16 port = 6668 block drop in inet proto tcp from 172.16.0.0/12 port = ftp to 192.168.0.0/16 port 6669:65535 +block drop in inet proto tcp from 172.16.0.0/12 port = ssh to 12.34.56.78 port = ircd +block drop in inet proto tcp from 172.16.0.0/12 port = ssh to 12.34.56.78 port = 6668 +block drop in inet proto tcp from 172.16.0.0/12 port = ssh to 12.34.56.78 port 6669:65535 block drop in inet proto tcp from 172.16.0.0/12 port = ftp to 12.34.56.78 port = ircd block drop in inet proto tcp from 172.16.0.0/12 port = ftp to 12.34.56.78 port = 6668 block drop in inet proto tcp from 172.16.0.0/12 port = ftp to 12.34.56.78 port 6669:65535 block drop in inet proto udp from 10.0.0.0/8 port = ssh to 192.168.0.0/16 port = 6667 block drop in inet proto udp from 10.0.0.0/8 port = ssh to 192.168.0.0/16 port = 6668 block drop in inet proto udp from 10.0.0.0/8 port = ssh to 192.168.0.0/16 port 6669:65535 -block drop in inet proto udp from 10.0.0.0/8 port = ssh to 12.34.56.78 port = 6667 -block drop in inet proto udp from 10.0.0.0/8 port = ssh to 12.34.56.78 port = 6668 -block drop in inet proto udp from 10.0.0.0/8 port = ssh to 12.34.56.78 port 6669:65535 block drop in inet proto udp from 10.0.0.0/8 port = ftp to 192.168.0.0/16 port = 6667 block drop in inet proto udp from 10.0.0.0/8 port = ftp to 192.168.0.0/16 port = 6668 block drop in inet proto udp from 10.0.0.0/8 port = ftp to 192.168.0.0/16 port 6669:65535 +block drop in inet proto udp from 10.0.0.0/8 port = ssh to 12.34.56.78 port = 6667 +block drop in inet proto udp from 10.0.0.0/8 port = ssh to 12.34.56.78 port = 6668 +block drop in inet proto udp from 10.0.0.0/8 port = ssh to 12.34.56.78 port 6669:65535 block drop in inet proto udp from 10.0.0.0/8 port = ftp to 12.34.56.78 port = 6667 block drop in inet proto udp from 10.0.0.0/8 port = ftp to 12.34.56.78 port = 6668 block drop in inet proto udp from 10.0.0.0/8 port = ftp to 12.34.56.78 port 6669:65535 block drop in inet proto udp from 172.16.0.0/12 port = ssh to 192.168.0.0/16 port = 6667 block drop in inet proto udp from 172.16.0.0/12 port = ssh to 192.168.0.0/16 port = 6668 block drop in inet proto udp from 172.16.0.0/12 port = ssh to 192.168.0.0/16 port 6669:65535 -block drop in inet proto udp from 172.16.0.0/12 port = ssh to 12.34.56.78 port = 6667 -block drop in inet proto udp from 172.16.0.0/12 port = ssh to 12.34.56.78 port = 6668 -block drop in inet proto udp from 172.16.0.0/12 port = ssh to 12.34.56.78 port 6669:65535 block drop in inet proto udp from 172.16.0.0/12 port = ftp to 192.168.0.0/16 port = 6667 block drop in inet proto udp from 172.16.0.0/12 port = ftp to 192.168.0.0/16 port = 6668 block drop in inet proto udp from 172.16.0.0/12 port = ftp to 192.168.0.0/16 port 6669:65535 +block drop in inet proto udp from 172.16.0.0/12 port = ssh to 12.34.56.78 port = 6667 +block drop in inet proto udp from 172.16.0.0/12 port = ssh to 12.34.56.78 port = 6668 +block drop in inet proto udp from 172.16.0.0/12 port = ssh to 12.34.56.78 port 6669:65535 block drop in inet proto udp from 172.16.0.0/12 port = ftp to 12.34.56.78 port = 6667 block drop in inet proto udp from 172.16.0.0/12 port = ftp to 12.34.56.78 port = 6668 block drop in inet proto udp from 172.16.0.0/12 port = ftp to 12.34.56.78 port 6669:65535 diff --git a/sys/net/pfvar.h b/sys/net/pfvar.h index dfc42c16547f..79dcd0d65985 100644 --- a/sys/net/pfvar.h +++ b/sys/net/pfvar.h @@ -1,2651 +1,2653 @@ /*- * SPDX-License-Identifier: BSD-2-Clause * * Copyright (c) 2001 Daniel Hartmeier * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * - Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * - Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials provided * with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. * * $OpenBSD: pfvar.h,v 1.282 2009/01/29 15:12:28 pyr Exp $ */ #ifndef _NET_PFVAR_H_ #define _NET_PFVAR_H_ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef _KERNEL #include #include #include #include #include #include #endif #include #include #include #ifdef _KERNEL #if defined(__arm__) #define PF_WANT_32_TO_64_COUNTER #endif /* * A hybrid of 32-bit and 64-bit counters which can be used on platforms where * counter(9) is very expensive. * * As 32-bit counters are expected to overflow, a periodic job sums them up to * a saved 64-bit state. Fetching the value still walks all CPUs to get the most * current snapshot. */ #ifdef PF_WANT_32_TO_64_COUNTER struct pf_counter_u64_pcpu { u_int32_t current; u_int32_t snapshot; }; struct pf_counter_u64 { struct pf_counter_u64_pcpu *pfcu64_pcpu; u_int64_t pfcu64_value; seqc_t pfcu64_seqc; }; static inline int pf_counter_u64_init(struct pf_counter_u64 *pfcu64, int flags) { pfcu64->pfcu64_value = 0; pfcu64->pfcu64_seqc = 0; pfcu64->pfcu64_pcpu = uma_zalloc_pcpu(pcpu_zone_8, flags | M_ZERO); if (__predict_false(pfcu64->pfcu64_pcpu == NULL)) return (ENOMEM); return (0); } static inline void pf_counter_u64_deinit(struct pf_counter_u64 *pfcu64) { uma_zfree_pcpu(pcpu_zone_8, pfcu64->pfcu64_pcpu); } static inline void pf_counter_u64_critical_enter(void) { critical_enter(); } static inline void pf_counter_u64_critical_exit(void) { critical_exit(); } static inline void pf_counter_u64_rollup_protected(struct pf_counter_u64 *pfcu64, uint64_t n) { MPASS(curthread->td_critnest > 0); pfcu64->pfcu64_value += n; } static inline void pf_counter_u64_add_protected(struct pf_counter_u64 *pfcu64, uint32_t n) { struct pf_counter_u64_pcpu *pcpu; u_int32_t val; MPASS(curthread->td_critnest > 0); pcpu = zpcpu_get(pfcu64->pfcu64_pcpu); val = atomic_load_int(&pcpu->current); atomic_store_int(&pcpu->current, val + n); } static inline void pf_counter_u64_add(struct pf_counter_u64 *pfcu64, uint32_t n) { critical_enter(); pf_counter_u64_add_protected(pfcu64, n); critical_exit(); } static inline u_int64_t pf_counter_u64_periodic(struct pf_counter_u64 *pfcu64) { struct pf_counter_u64_pcpu *pcpu; u_int64_t sum; u_int32_t val; int cpu; MPASS(curthread->td_critnest > 0); seqc_write_begin(&pfcu64->pfcu64_seqc); sum = pfcu64->pfcu64_value; CPU_FOREACH(cpu) { pcpu = zpcpu_get_cpu(pfcu64->pfcu64_pcpu, cpu); val = atomic_load_int(&pcpu->current); sum += (uint32_t)(val - pcpu->snapshot); pcpu->snapshot = val; } pfcu64->pfcu64_value = sum; seqc_write_end(&pfcu64->pfcu64_seqc); return (sum); } static inline u_int64_t pf_counter_u64_fetch(const struct pf_counter_u64 *pfcu64) { struct pf_counter_u64_pcpu *pcpu; u_int64_t sum; seqc_t seqc; int cpu; for (;;) { seqc = seqc_read(&pfcu64->pfcu64_seqc); sum = 0; CPU_FOREACH(cpu) { pcpu = zpcpu_get_cpu(pfcu64->pfcu64_pcpu, cpu); sum += (uint32_t)(atomic_load_int(&pcpu->current) -pcpu->snapshot); } sum += pfcu64->pfcu64_value; if (seqc_consistent(&pfcu64->pfcu64_seqc, seqc)) break; } return (sum); } static inline void pf_counter_u64_zero_protected(struct pf_counter_u64 *pfcu64) { struct pf_counter_u64_pcpu *pcpu; int cpu; MPASS(curthread->td_critnest > 0); seqc_write_begin(&pfcu64->pfcu64_seqc); CPU_FOREACH(cpu) { pcpu = zpcpu_get_cpu(pfcu64->pfcu64_pcpu, cpu); pcpu->snapshot = atomic_load_int(&pcpu->current); } pfcu64->pfcu64_value = 0; seqc_write_end(&pfcu64->pfcu64_seqc); } static inline void pf_counter_u64_zero(struct pf_counter_u64 *pfcu64) { critical_enter(); pf_counter_u64_zero_protected(pfcu64); critical_exit(); } #else struct pf_counter_u64 { counter_u64_t counter; }; static inline int pf_counter_u64_init(struct pf_counter_u64 *pfcu64, int flags) { pfcu64->counter = counter_u64_alloc(flags); if (__predict_false(pfcu64->counter == NULL)) return (ENOMEM); return (0); } static inline void pf_counter_u64_deinit(struct pf_counter_u64 *pfcu64) { counter_u64_free(pfcu64->counter); } static inline void pf_counter_u64_critical_enter(void) { } static inline void pf_counter_u64_critical_exit(void) { } static inline void pf_counter_u64_rollup_protected(struct pf_counter_u64 *pfcu64, uint64_t n) { counter_u64_add(pfcu64->counter, n); } static inline void pf_counter_u64_add_protected(struct pf_counter_u64 *pfcu64, uint32_t n) { counter_u64_add(pfcu64->counter, n); } static inline void pf_counter_u64_add(struct pf_counter_u64 *pfcu64, uint32_t n) { pf_counter_u64_add_protected(pfcu64, n); } static inline u_int64_t pf_counter_u64_fetch(const struct pf_counter_u64 *pfcu64) { return (counter_u64_fetch(pfcu64->counter)); } static inline void pf_counter_u64_zero_protected(struct pf_counter_u64 *pfcu64) { counter_u64_zero(pfcu64->counter); } static inline void pf_counter_u64_zero(struct pf_counter_u64 *pfcu64) { pf_counter_u64_zero_protected(pfcu64); } #endif #define pf_get_timestamp(prule)({ \ uint32_t _ts = 0; \ uint32_t __ts; \ int cpu; \ CPU_FOREACH(cpu) { \ __ts = *zpcpu_get_cpu(prule->timestamp, cpu); \ if (__ts > _ts) \ _ts = __ts; \ } \ _ts; \ }) #define pf_update_timestamp(prule) \ do { \ critical_enter(); \ *zpcpu_get((prule)->timestamp) = time_second; \ critical_exit(); \ } while (0) #define pf_timestamp_pcpu_zone (sizeof(time_t) == 4 ? pcpu_zone_4 : pcpu_zone_8) _Static_assert(sizeof(time_t) == 4 || sizeof(time_t) == 8, "unexpected time_t size"); SYSCTL_DECL(_net_pf); MALLOC_DECLARE(M_PFHASH); MALLOC_DECLARE(M_PF_RULE_ITEM); SDT_PROVIDER_DECLARE(pf); struct pfi_dynaddr { TAILQ_ENTRY(pfi_dynaddr) entry; struct pf_addr pfid_addr4; struct pf_addr pfid_mask4; struct pf_addr pfid_addr6; struct pf_addr pfid_mask6; struct pfr_ktable *pfid_kt; struct pfi_kkif *pfid_kif; int pfid_net; /* mask or 128 */ int pfid_acnt4; /* address count IPv4 */ int pfid_acnt6; /* address count IPv6 */ sa_family_t pfid_af; /* rule af */ u_int8_t pfid_iflags; /* PFI_AFLAG_* */ }; /* * Address manipulation macros */ #define HTONL(x) (x) = htonl((__uint32_t)(x)) #define HTONS(x) (x) = htons((__uint16_t)(x)) #define NTOHL(x) (x) = ntohl((__uint32_t)(x)) #define NTOHS(x) (x) = ntohs((__uint16_t)(x)) #define PF_NAME "pf" #define PF_HASHROW_ASSERT(h) mtx_assert(&(h)->lock, MA_OWNED) #define PF_HASHROW_LOCK(h) mtx_lock(&(h)->lock) #define PF_HASHROW_UNLOCK(h) mtx_unlock(&(h)->lock) #ifdef INVARIANTS #define PF_STATE_LOCK(s) \ do { \ struct pf_kstate *_s = (s); \ struct pf_idhash *_ih = &V_pf_idhash[PF_IDHASH(_s)]; \ MPASS(_s->lock == &_ih->lock); \ mtx_lock(_s->lock); \ } while (0) #define PF_STATE_UNLOCK(s) \ do { \ struct pf_kstate *_s = (s); \ struct pf_idhash *_ih = &V_pf_idhash[PF_IDHASH(_s)]; \ MPASS(_s->lock == &_ih->lock); \ mtx_unlock(_s->lock); \ } while (0) #else #define PF_STATE_LOCK(s) mtx_lock((s)->lock) #define PF_STATE_UNLOCK(s) mtx_unlock((s)->lock) #endif #ifdef INVARIANTS #define PF_STATE_LOCK_ASSERT(s) \ do { \ struct pf_kstate *_s = (s); \ struct pf_idhash *_ih = &V_pf_idhash[PF_IDHASH(_s)]; \ MPASS(_s->lock == &_ih->lock); \ PF_HASHROW_ASSERT(_ih); \ } while (0) #else /* !INVARIANTS */ #define PF_STATE_LOCK_ASSERT(s) do {} while (0) #endif /* INVARIANTS */ #ifdef INVARIANTS #define PF_SRC_NODE_LOCK(sn) \ do { \ struct pf_ksrc_node *_sn = (sn); \ struct pf_srchash *_sh = &V_pf_srchash[ \ pf_hashsrc(&_sn->addr, _sn->af)]; \ MPASS(_sn->lock == &_sh->lock); \ mtx_lock(_sn->lock); \ } while (0) #define PF_SRC_NODE_UNLOCK(sn) \ do { \ struct pf_ksrc_node *_sn = (sn); \ struct pf_srchash *_sh = &V_pf_srchash[ \ pf_hashsrc(&_sn->addr, _sn->af)]; \ MPASS(_sn->lock == &_sh->lock); \ mtx_unlock(_sn->lock); \ } while (0) #else #define PF_SRC_NODE_LOCK(sn) mtx_lock((sn)->lock) #define PF_SRC_NODE_UNLOCK(sn) mtx_unlock((sn)->lock) #endif #ifdef INVARIANTS #define PF_SRC_NODE_LOCK_ASSERT(sn) \ do { \ struct pf_ksrc_node *_sn = (sn); \ struct pf_srchash *_sh = &V_pf_srchash[ \ pf_hashsrc(&_sn->addr, _sn->af)]; \ MPASS(_sn->lock == &_sh->lock); \ PF_HASHROW_ASSERT(_sh); \ } while (0) #else /* !INVARIANTS */ #define PF_SRC_NODE_LOCK_ASSERT(sn) do {} while (0) #endif /* INVARIANTS */ extern struct mtx_padalign pf_unlnkdrules_mtx; #define PF_UNLNKDRULES_LOCK() mtx_lock(&pf_unlnkdrules_mtx) #define PF_UNLNKDRULES_UNLOCK() mtx_unlock(&pf_unlnkdrules_mtx) #define PF_UNLNKDRULES_ASSERT() mtx_assert(&pf_unlnkdrules_mtx, MA_OWNED) extern struct sx pf_config_lock; #define PF_CONFIG_LOCK() sx_xlock(&pf_config_lock) #define PF_CONFIG_UNLOCK() sx_xunlock(&pf_config_lock) #define PF_CONFIG_ASSERT() sx_assert(&pf_config_lock, SA_XLOCKED) VNET_DECLARE(struct rmlock, pf_rules_lock); #define V_pf_rules_lock VNET(pf_rules_lock) #define PF_RULES_RLOCK_TRACKER struct rm_priotracker _pf_rules_tracker #define PF_RULES_RLOCK() rm_rlock(&V_pf_rules_lock, &_pf_rules_tracker) #define PF_RULES_RUNLOCK() rm_runlock(&V_pf_rules_lock, &_pf_rules_tracker) #define PF_RULES_WLOCK() rm_wlock(&V_pf_rules_lock) #define PF_RULES_WUNLOCK() rm_wunlock(&V_pf_rules_lock) #define PF_RULES_WOWNED() rm_wowned(&V_pf_rules_lock) #define PF_RULES_ASSERT() rm_assert(&V_pf_rules_lock, RA_LOCKED) #define PF_RULES_RASSERT() rm_assert(&V_pf_rules_lock, RA_RLOCKED) #define PF_RULES_WASSERT() rm_assert(&V_pf_rules_lock, RA_WLOCKED) extern struct mtx_padalign pf_table_stats_lock; #define PF_TABLE_STATS_LOCK() mtx_lock(&pf_table_stats_lock) #define PF_TABLE_STATS_UNLOCK() mtx_unlock(&pf_table_stats_lock) #define PF_TABLE_STATS_OWNED() mtx_owned(&pf_table_stats_lock) #define PF_TABLE_STATS_ASSERT() mtx_assert(&pf_table_stats_lock, MA_OWNED) extern struct sx pf_end_lock; #define PF_MODVER 1 #define PFLOG_MODVER 1 #define PFSYNC_MODVER 1 #define PFLOG_MINVER 1 #define PFLOG_PREFVER PFLOG_MODVER #define PFLOG_MAXVER 1 #define PFSYNC_MINVER 1 #define PFSYNC_PREFVER PFSYNC_MODVER #define PFSYNC_MAXVER 1 #ifdef INET #ifndef INET6 #define PF_INET_ONLY #endif /* ! INET6 */ #endif /* INET */ #ifdef INET6 #ifndef INET #define PF_INET6_ONLY #endif /* ! INET */ #endif /* INET6 */ #ifdef INET #ifdef INET6 #define PF_INET_INET6 #endif /* INET6 */ #endif /* INET */ #else #define PF_INET_INET6 #endif /* _KERNEL */ /* Both IPv4 and IPv6 */ #ifdef PF_INET_INET6 #define PF_AEQ(a, b, c) \ ((c == AF_INET && (a)->addr32[0] == (b)->addr32[0]) || \ (c == AF_INET6 && (a)->addr32[3] == (b)->addr32[3] && \ (a)->addr32[2] == (b)->addr32[2] && \ (a)->addr32[1] == (b)->addr32[1] && \ (a)->addr32[0] == (b)->addr32[0])) \ #define PF_ANEQ(a, b, c) \ ((c == AF_INET && (a)->addr32[0] != (b)->addr32[0]) || \ (c == AF_INET6 && ((a)->addr32[0] != (b)->addr32[0] || \ (a)->addr32[1] != (b)->addr32[1] || \ (a)->addr32[2] != (b)->addr32[2] || \ (a)->addr32[3] != (b)->addr32[3]))) \ #define PF_AZERO(a, c) \ ((c == AF_INET && !(a)->addr32[0]) || \ (c == AF_INET6 && !(a)->addr32[0] && !(a)->addr32[1] && \ !(a)->addr32[2] && !(a)->addr32[3] )) \ #define PF_MATCHA(n, a, m, b, f) \ pf_match_addr(n, a, m, b, f) #define PF_ACPY(a, b, f) \ pf_addrcpy(a, b, f) #define PF_AINC(a, f) \ pf_addr_inc(a, f) #define PF_POOLMASK(a, b, c, d, f) \ pf_poolmask(a, b, c, d, f) #else /* Just IPv6 */ #ifdef PF_INET6_ONLY #define PF_AEQ(a, b, c) \ ((a)->addr32[3] == (b)->addr32[3] && \ (a)->addr32[2] == (b)->addr32[2] && \ (a)->addr32[1] == (b)->addr32[1] && \ (a)->addr32[0] == (b)->addr32[0]) \ #define PF_ANEQ(a, b, c) \ ((a)->addr32[3] != (b)->addr32[3] || \ (a)->addr32[2] != (b)->addr32[2] || \ (a)->addr32[1] != (b)->addr32[1] || \ (a)->addr32[0] != (b)->addr32[0]) \ #define PF_AZERO(a, c) \ (!(a)->addr32[0] && \ !(a)->addr32[1] && \ !(a)->addr32[2] && \ !(a)->addr32[3] ) \ #define PF_MATCHA(n, a, m, b, f) \ pf_match_addr(n, a, m, b, f) #define PF_ACPY(a, b, f) \ pf_addrcpy(a, b, f) #define PF_AINC(a, f) \ pf_addr_inc(a, f) #define PF_POOLMASK(a, b, c, d, f) \ pf_poolmask(a, b, c, d, f) #else /* Just IPv4 */ #ifdef PF_INET_ONLY #define PF_AEQ(a, b, c) \ ((a)->addr32[0] == (b)->addr32[0]) #define PF_ANEQ(a, b, c) \ ((a)->addr32[0] != (b)->addr32[0]) #define PF_AZERO(a, c) \ (!(a)->addr32[0]) #define PF_MATCHA(n, a, m, b, f) \ pf_match_addr(n, a, m, b, f) #define PF_ACPY(a, b, f) \ (a)->v4.s_addr = (b)->v4.s_addr #define PF_AINC(a, f) \ do { \ (a)->addr32[0] = htonl(ntohl((a)->addr32[0]) + 1); \ } while (0) #define PF_POOLMASK(a, b, c, d, f) \ do { \ (a)->addr32[0] = ((b)->addr32[0] & (c)->addr32[0]) | \ (((c)->addr32[0] ^ 0xffffffff ) & (d)->addr32[0]); \ } while (0) #endif /* PF_INET_ONLY */ #endif /* PF_INET6_ONLY */ #endif /* PF_INET_INET6 */ /* * XXX callers not FIB-aware in our version of pf yet. * OpenBSD fixed it later it seems, 2010/05/07 13:33:16 claudio. */ #define PF_MISMATCHAW(aw, x, af, neg, ifp, rtid) \ ( \ (((aw)->type == PF_ADDR_NOROUTE && \ pf_routable((x), (af), NULL, (rtid))) || \ (((aw)->type == PF_ADDR_URPFFAILED && (ifp) != NULL && \ pf_routable((x), (af), (ifp), (rtid))) || \ ((aw)->type == PF_ADDR_TABLE && \ !pfr_match_addr((aw)->p.tbl, (x), (af))) || \ ((aw)->type == PF_ADDR_DYNIFTL && \ !pfi_match_addr((aw)->p.dyn, (x), (af))) || \ ((aw)->type == PF_ADDR_RANGE && \ !pf_match_addr_range(&(aw)->v.a.addr, \ &(aw)->v.a.mask, (x), (af))) || \ ((aw)->type == PF_ADDR_ADDRMASK && \ !PF_AZERO(&(aw)->v.a.mask, (af)) && \ !PF_MATCHA(0, &(aw)->v.a.addr, \ &(aw)->v.a.mask, (x), (af))))) != \ (neg) \ ) #define PF_ALGNMNT(off) (((off) % 2) == 0) #ifdef _KERNEL struct pf_kpooladdr { struct pf_addr_wrap addr; TAILQ_ENTRY(pf_kpooladdr) entries; char ifname[IFNAMSIZ]; struct pfi_kkif *kif; }; TAILQ_HEAD(pf_kpalist, pf_kpooladdr); struct pf_kpool { struct mtx mtx; struct pf_kpalist list; struct pf_kpooladdr *cur; struct pf_poolhashkey key; struct pf_addr counter; struct pf_mape_portset mape; int tblidx; u_int16_t proxy_port[2]; u_int8_t opts; }; struct pf_rule_actions { int32_t rtableid; uint16_t qid; uint16_t pqid; uint16_t max_mss; uint8_t log; uint8_t set_tos; uint8_t min_ttl; uint16_t dnpipe; uint16_t dnrpipe; /* Reverse direction pipe */ uint32_t flags; uint8_t set_prio[2]; }; union pf_keth_rule_ptr { struct pf_keth_rule *ptr; uint32_t nr; }; struct pf_keth_rule_addr { uint8_t addr[ETHER_ADDR_LEN]; uint8_t mask[ETHER_ADDR_LEN]; bool neg; uint8_t isset; }; struct pf_keth_anchor; TAILQ_HEAD(pf_keth_ruleq, pf_keth_rule); struct pf_keth_ruleset { struct pf_keth_ruleq rules[2]; struct pf_keth_rules { struct pf_keth_ruleq *rules; int open; uint32_t ticket; } active, inactive; struct epoch_context epoch_ctx; struct vnet *vnet; struct pf_keth_anchor *anchor; }; RB_HEAD(pf_keth_anchor_global, pf_keth_anchor); RB_HEAD(pf_keth_anchor_node, pf_keth_anchor); struct pf_keth_anchor { RB_ENTRY(pf_keth_anchor) entry_node; RB_ENTRY(pf_keth_anchor) entry_global; struct pf_keth_anchor *parent; struct pf_keth_anchor_node children; char name[PF_ANCHOR_NAME_SIZE]; char path[MAXPATHLEN]; struct pf_keth_ruleset ruleset; int refcnt; /* anchor rules */ uint8_t anchor_relative; uint8_t anchor_wildcard; }; RB_PROTOTYPE(pf_keth_anchor_node, pf_keth_anchor, entry_node, pf_keth_anchor_compare); RB_PROTOTYPE(pf_keth_anchor_global, pf_keth_anchor, entry_global, pf_keth_anchor_compare); struct pf_keth_rule { #define PFE_SKIP_IFP 0 #define PFE_SKIP_DIR 1 #define PFE_SKIP_PROTO 2 #define PFE_SKIP_SRC_ADDR 3 #define PFE_SKIP_DST_ADDR 4 #define PFE_SKIP_SRC_IP_ADDR 5 #define PFE_SKIP_DST_IP_ADDR 6 #define PFE_SKIP_COUNT 7 union pf_keth_rule_ptr skip[PFE_SKIP_COUNT]; TAILQ_ENTRY(pf_keth_rule) entries; struct pf_keth_anchor *anchor; u_int8_t anchor_relative; u_int8_t anchor_wildcard; uint32_t nr; bool quick; /* Filter */ char ifname[IFNAMSIZ]; struct pfi_kkif *kif; bool ifnot; uint8_t direction; uint16_t proto; struct pf_keth_rule_addr src, dst; struct pf_rule_addr ipsrc, ipdst; char match_tagname[PF_TAG_NAME_SIZE]; uint16_t match_tag; bool match_tag_not; /* Stats */ counter_u64_t evaluations; counter_u64_t packets[2]; counter_u64_t bytes[2]; time_t *timestamp; /* Action */ char qname[PF_QNAME_SIZE]; int qid; char tagname[PF_TAG_NAME_SIZE]; uint16_t tag; char bridge_to_name[IFNAMSIZ]; struct pfi_kkif *bridge_to; uint8_t action; uint16_t dnpipe; uint32_t dnflags; char label[PF_RULE_MAX_LABEL_COUNT][PF_RULE_LABEL_SIZE]; uint32_t ridentifier; }; union pf_krule_ptr { struct pf_krule *ptr; u_int32_t nr; }; RB_HEAD(pf_krule_global, pf_krule); RB_PROTOTYPE(pf_krule_global, pf_krule, entry_global, pf_krule_compare); struct pf_krule { struct pf_rule_addr src; struct pf_rule_addr dst; union pf_krule_ptr skip[PF_SKIP_COUNT]; char label[PF_RULE_MAX_LABEL_COUNT][PF_RULE_LABEL_SIZE]; uint32_t ridentifier; char ifname[IFNAMSIZ]; char rcv_ifname[IFNAMSIZ]; char qname[PF_QNAME_SIZE]; char pqname[PF_QNAME_SIZE]; char tagname[PF_TAG_NAME_SIZE]; char match_tagname[PF_TAG_NAME_SIZE]; char overload_tblname[PF_TABLE_NAME_SIZE]; TAILQ_ENTRY(pf_krule) entries; struct pf_kpool rpool; struct pf_counter_u64 evaluations; struct pf_counter_u64 packets[2]; struct pf_counter_u64 bytes[2]; time_t *timestamp; struct pfi_kkif *kif; struct pfi_kkif *rcv_kif; struct pf_kanchor *anchor; struct pfr_ktable *overload_tbl; pf_osfp_t os_fingerprint; int32_t rtableid; u_int32_t timeout[PFTM_MAX]; u_int32_t max_states; u_int32_t max_src_nodes; u_int32_t max_src_states; u_int32_t max_src_conn; struct { u_int32_t limit; u_int32_t seconds; } max_src_conn_rate; u_int16_t qid; u_int16_t pqid; u_int16_t dnpipe; u_int16_t dnrpipe; u_int32_t free_flags; u_int32_t nr; u_int32_t prob; uid_t cuid; pid_t cpid; counter_u64_t states_cur; counter_u64_t states_tot; counter_u64_t src_nodes; u_int16_t return_icmp; u_int16_t return_icmp6; u_int16_t max_mss; u_int16_t tag; u_int16_t match_tag; u_int16_t scrub_flags; struct pf_rule_uid uid; struct pf_rule_gid gid; u_int32_t rule_flag; uint32_t rule_ref; u_int8_t action; u_int8_t direction; u_int8_t log; u_int8_t logif; u_int8_t quick; u_int8_t ifnot; u_int8_t match_tag_not; u_int8_t natpass; u_int8_t keep_state; sa_family_t af; u_int8_t proto; u_int8_t type; u_int8_t code; u_int8_t flags; u_int8_t flagset; u_int8_t min_ttl; u_int8_t allow_opts; u_int8_t rt; u_int8_t return_ttl; u_int8_t tos; u_int8_t set_tos; u_int8_t anchor_relative; u_int8_t anchor_wildcard; u_int8_t flush; u_int8_t prio; u_int8_t set_prio[2]; struct { struct pf_addr addr; u_int16_t port; } divert; u_int8_t md5sum[PF_MD5_DIGEST_LENGTH]; RB_ENTRY(pf_krule) entry_global; #ifdef PF_WANT_32_TO_64_COUNTER LIST_ENTRY(pf_krule) allrulelist; bool allrulelinked; #endif }; struct pf_krule_item { SLIST_ENTRY(pf_krule_item) entry; struct pf_krule *r; }; SLIST_HEAD(pf_krule_slist, pf_krule_item); struct pf_ksrc_node { LIST_ENTRY(pf_ksrc_node) entry; struct pf_addr addr; struct pf_addr raddr; struct pf_krule_slist match_rules; union pf_krule_ptr rule; struct pfi_kkif *rkif; counter_u64_t bytes[2]; counter_u64_t packets[2]; u_int32_t states; u_int32_t conn; struct pf_threshold conn_rate; u_int32_t creation; u_int32_t expire; sa_family_t af; u_int8_t ruletype; struct mtx *lock; }; #endif struct pf_state_scrub { struct timeval pfss_last; /* time received last packet */ u_int32_t pfss_tsecr; /* last echoed timestamp */ u_int32_t pfss_tsval; /* largest timestamp */ u_int32_t pfss_tsval0; /* original timestamp */ u_int16_t pfss_flags; #define PFSS_TIMESTAMP 0x0001 /* modulate timestamp */ #define PFSS_PAWS 0x0010 /* stricter PAWS checks */ #define PFSS_PAWS_IDLED 0x0020 /* was idle too long. no PAWS */ #define PFSS_DATA_TS 0x0040 /* timestamp on data packets */ #define PFSS_DATA_NOTS 0x0080 /* no timestamp on data packets */ u_int8_t pfss_ttl; /* stashed TTL */ u_int8_t pad; union { u_int32_t pfss_ts_mod; /* timestamp modulation */ u_int32_t pfss_v_tag; /* SCTP verification tag */ }; }; struct pf_state_host { struct pf_addr addr; u_int16_t port; u_int16_t pad; }; struct pf_state_peer { struct pf_state_scrub *scrub; /* state is scrubbed */ u_int32_t seqlo; /* Max sequence number sent */ u_int32_t seqhi; /* Max the other end ACKd + win */ u_int32_t seqdiff; /* Sequence number modulator */ u_int16_t max_win; /* largest window (pre scaling) */ u_int16_t mss; /* Maximum segment size option */ u_int8_t state; /* active state level */ u_int8_t wscale; /* window scaling factor */ u_int8_t tcp_est; /* Did we reach TCPS_ESTABLISHED */ u_int8_t pad[1]; }; /* Keep synced with struct pf_udp_endpoint. */ struct pf_udp_endpoint_cmp { struct pf_addr addr; uint16_t port; sa_family_t af; uint8_t pad[1]; }; struct pf_udp_endpoint { struct pf_addr addr; uint16_t port; sa_family_t af; uint8_t pad[1]; struct pf_udp_mapping *mapping; LIST_ENTRY(pf_udp_endpoint) entry; }; struct pf_udp_mapping { struct pf_udp_endpoint endpoints[2]; u_int refs; }; /* Keep synced with struct pf_state_key. */ struct pf_state_key_cmp { struct pf_addr addr[2]; u_int16_t port[2]; sa_family_t af; u_int8_t proto; u_int8_t pad[2]; }; struct pf_state_key { struct pf_addr addr[2]; u_int16_t port[2]; sa_family_t af; u_int8_t proto; u_int8_t pad[2]; LIST_ENTRY(pf_state_key) entry; TAILQ_HEAD(, pf_kstate) states[2]; }; /* Keep synced with struct pf_kstate. */ struct pf_state_cmp { u_int64_t id; u_int32_t creatorid; u_int8_t direction; u_int8_t pad[3]; }; struct pf_state_scrub_export { uint16_t pfss_flags; uint8_t pfss_ttl; /* stashed TTL */ #define PF_SCRUB_FLAG_VALID 0x01 uint8_t scrub_flag; uint32_t pfss_ts_mod; /* timestamp modulation */ }; struct pf_state_key_export { struct pf_addr addr[2]; uint16_t port[2]; }; struct pf_state_peer_export { struct pf_state_scrub_export scrub; /* state is scrubbed */ uint32_t seqlo; /* Max sequence number sent */ uint32_t seqhi; /* Max the other end ACKd + win */ uint32_t seqdiff; /* Sequence number modulator */ uint16_t max_win; /* largest window (pre scaling) */ uint16_t mss; /* Maximum segment size option */ uint8_t state; /* active state level */ uint8_t wscale; /* window scaling factor */ uint8_t dummy[6]; }; _Static_assert(sizeof(struct pf_state_peer_export) == 32, "size incorrect"); struct pf_state_export { uint64_t version; #define PF_STATE_VERSION 20230404 uint64_t id; char ifname[IFNAMSIZ]; char orig_ifname[IFNAMSIZ]; struct pf_state_key_export key[2]; struct pf_state_peer_export src; struct pf_state_peer_export dst; struct pf_addr rt_addr; uint32_t rule; uint32_t anchor; uint32_t nat_rule; uint32_t creation; uint32_t expire; uint32_t spare0; uint64_t packets[2]; uint64_t bytes[2]; uint32_t creatorid; uint32_t spare1; sa_family_t af; uint8_t proto; uint8_t direction; uint8_t log; uint8_t state_flags_compat; uint8_t timeout; uint8_t sync_flags; uint8_t updates; uint16_t state_flags; uint16_t qid; uint16_t pqid; uint16_t dnpipe; uint16_t dnrpipe; int32_t rtableid; uint8_t min_ttl; uint8_t set_tos; uint16_t max_mss; uint8_t set_prio[2]; uint8_t rt; char rt_ifname[IFNAMSIZ]; uint8_t spare[72]; }; _Static_assert(sizeof(struct pf_state_export) == 384, "size incorrect"); #ifdef _KERNEL struct pf_kstate { /* * Area shared with pf_state_cmp */ u_int64_t id; u_int32_t creatorid; u_int8_t direction; u_int8_t pad[3]; /* * end of the area */ u_int16_t state_flags; u_int8_t timeout; u_int8_t sync_state; /* PFSYNC_S_x */ u_int8_t sync_updates; /* XXX */ u_int refs; struct mtx *lock; TAILQ_ENTRY(pf_kstate) sync_list; TAILQ_ENTRY(pf_kstate) key_list[2]; LIST_ENTRY(pf_kstate) entry; struct pf_state_peer src; struct pf_state_peer dst; struct pf_krule_slist match_rules; union pf_krule_ptr rule; union pf_krule_ptr anchor; union pf_krule_ptr nat_rule; struct pf_addr rt_addr; struct pf_state_key *key[2]; /* addresses stack and wire */ struct pf_udp_mapping *udp_mapping; struct pfi_kkif *kif; struct pfi_kkif *orig_kif; /* The real kif, even if we're a floating state (i.e. if == V_pfi_all). */ struct pfi_kkif *rt_kif; struct pf_ksrc_node *src_node; struct pf_ksrc_node *nat_src_node; u_int64_t packets[2]; u_int64_t bytes[2]; u_int64_t creation; u_int64_t expire; u_int32_t pfsync_time; struct pf_rule_actions act; u_int16_t tag; u_int8_t rt; u_int16_t if_index_in; u_int16_t if_index_out; }; /* * Size <= fits 11 objects per page on LP64. Try to not grow the struct beyond that. */ _Static_assert(sizeof(struct pf_kstate) <= 372, "pf_kstate size crosses 372 bytes"); #endif /* * Unified state structures for pulling states out of the kernel * used by pfsync(4) and the pf(4) ioctl. */ struct pfsync_state_scrub { u_int16_t pfss_flags; u_int8_t pfss_ttl; /* stashed TTL */ #define PFSYNC_SCRUB_FLAG_VALID 0x01 u_int8_t scrub_flag; u_int32_t pfss_ts_mod; /* timestamp modulation */ } __packed; struct pfsync_state_peer { struct pfsync_state_scrub scrub; /* state is scrubbed */ u_int32_t seqlo; /* Max sequence number sent */ u_int32_t seqhi; /* Max the other end ACKd + win */ u_int32_t seqdiff; /* Sequence number modulator */ u_int16_t max_win; /* largest window (pre scaling) */ u_int16_t mss; /* Maximum segment size option */ u_int8_t state; /* active state level */ u_int8_t wscale; /* window scaling factor */ u_int8_t pad[6]; } __packed; struct pfsync_state_key { struct pf_addr addr[2]; u_int16_t port[2]; }; struct pfsync_state_1301 { u_int64_t id; char ifname[IFNAMSIZ]; struct pfsync_state_key key[2]; struct pfsync_state_peer src; struct pfsync_state_peer dst; struct pf_addr rt_addr; u_int32_t rule; u_int32_t anchor; u_int32_t nat_rule; u_int32_t creation; u_int32_t expire; u_int32_t packets[2][2]; u_int32_t bytes[2][2]; u_int32_t creatorid; sa_family_t af; u_int8_t proto; u_int8_t direction; u_int8_t __spare[2]; u_int8_t log; u_int8_t state_flags; u_int8_t timeout; u_int8_t sync_flags; u_int8_t updates; } __packed; struct pfsync_state_1400 { /* The beginning of the struct is compatible with previous versions */ u_int64_t id; char ifname[IFNAMSIZ]; struct pfsync_state_key key[2]; struct pfsync_state_peer src; struct pfsync_state_peer dst; struct pf_addr rt_addr; u_int32_t rule; u_int32_t anchor; u_int32_t nat_rule; u_int32_t creation; u_int32_t expire; u_int32_t packets[2][2]; u_int32_t bytes[2][2]; u_int32_t creatorid; sa_family_t af; u_int8_t proto; u_int8_t direction; u_int16_t state_flags; u_int8_t log; u_int8_t __spare; u_int8_t timeout; u_int8_t sync_flags; u_int8_t updates; /* The rest is not */ u_int16_t qid; u_int16_t pqid; u_int16_t dnpipe; u_int16_t dnrpipe; int32_t rtableid; u_int8_t min_ttl; u_int8_t set_tos; u_int16_t max_mss; u_int8_t set_prio[2]; u_int8_t rt; char rt_ifname[IFNAMSIZ]; } __packed; union pfsync_state_union { struct pfsync_state_1301 pfs_1301; struct pfsync_state_1400 pfs_1400; } __packed; #ifdef _KERNEL /* pfsync */ typedef int pfsync_state_import_t(union pfsync_state_union *, int, int); typedef void pfsync_insert_state_t(struct pf_kstate *); typedef void pfsync_update_state_t(struct pf_kstate *); typedef void pfsync_delete_state_t(struct pf_kstate *); typedef void pfsync_clear_states_t(u_int32_t, const char *); typedef int pfsync_defer_t(struct pf_kstate *, struct mbuf *); typedef void pfsync_detach_ifnet_t(struct ifnet *); typedef void pflow_export_state_t(const struct pf_kstate *); typedef bool pf_addr_filter_func_t(const sa_family_t, const struct pf_addr *); VNET_DECLARE(pfsync_state_import_t *, pfsync_state_import_ptr); #define V_pfsync_state_import_ptr VNET(pfsync_state_import_ptr) VNET_DECLARE(pfsync_insert_state_t *, pfsync_insert_state_ptr); #define V_pfsync_insert_state_ptr VNET(pfsync_insert_state_ptr) VNET_DECLARE(pfsync_update_state_t *, pfsync_update_state_ptr); #define V_pfsync_update_state_ptr VNET(pfsync_update_state_ptr) VNET_DECLARE(pfsync_delete_state_t *, pfsync_delete_state_ptr); #define V_pfsync_delete_state_ptr VNET(pfsync_delete_state_ptr) VNET_DECLARE(pfsync_clear_states_t *, pfsync_clear_states_ptr); #define V_pfsync_clear_states_ptr VNET(pfsync_clear_states_ptr) VNET_DECLARE(pfsync_defer_t *, pfsync_defer_ptr); #define V_pfsync_defer_ptr VNET(pfsync_defer_ptr) VNET_DECLARE(pflow_export_state_t *, pflow_export_state_ptr); #define V_pflow_export_state_ptr VNET(pflow_export_state_ptr) extern pfsync_detach_ifnet_t *pfsync_detach_ifnet_ptr; void pfsync_state_export(union pfsync_state_union *, struct pf_kstate *, int); void pf_state_export(struct pf_state_export *, struct pf_kstate *); /* pflog */ struct pf_kruleset; struct pf_pdesc; typedef int pflog_packet_t(struct pfi_kkif *, struct mbuf *, uint8_t, u_int8_t, struct pf_krule *, struct pf_krule *, struct pf_kruleset *, struct pf_pdesc *, int); extern pflog_packet_t *pflog_packet_ptr; #endif /* _KERNEL */ #define PFSYNC_FLAG_SRCNODE 0x04 #define PFSYNC_FLAG_NATSRCNODE 0x08 /* for copies to/from network byte order */ /* ioctl interface also uses network byte order */ #define pf_state_peer_hton(s,d) do { \ (d)->seqlo = htonl((s)->seqlo); \ (d)->seqhi = htonl((s)->seqhi); \ (d)->seqdiff = htonl((s)->seqdiff); \ (d)->max_win = htons((s)->max_win); \ (d)->mss = htons((s)->mss); \ (d)->state = (s)->state; \ (d)->wscale = (s)->wscale; \ if ((s)->scrub) { \ (d)->scrub.pfss_flags = \ htons((s)->scrub->pfss_flags & PFSS_TIMESTAMP); \ (d)->scrub.pfss_ttl = (s)->scrub->pfss_ttl; \ (d)->scrub.pfss_ts_mod = htonl((s)->scrub->pfss_ts_mod);\ (d)->scrub.scrub_flag = PFSYNC_SCRUB_FLAG_VALID; \ } \ } while (0) #define pf_state_peer_ntoh(s,d) do { \ (d)->seqlo = ntohl((s)->seqlo); \ (d)->seqhi = ntohl((s)->seqhi); \ (d)->seqdiff = ntohl((s)->seqdiff); \ (d)->max_win = ntohs((s)->max_win); \ (d)->mss = ntohs((s)->mss); \ (d)->state = (s)->state; \ (d)->wscale = (s)->wscale; \ if ((s)->scrub.scrub_flag == PFSYNC_SCRUB_FLAG_VALID && \ (d)->scrub != NULL) { \ (d)->scrub->pfss_flags = \ ntohs((s)->scrub.pfss_flags) & PFSS_TIMESTAMP; \ (d)->scrub->pfss_ttl = (s)->scrub.pfss_ttl; \ (d)->scrub->pfss_ts_mod = ntohl((s)->scrub.pfss_ts_mod);\ } \ } while (0) #define pf_state_counter_hton(s,d) do { \ d[0] = htonl((s>>32)&0xffffffff); \ d[1] = htonl(s&0xffffffff); \ } while (0) #define pf_state_counter_from_pfsync(s) \ (((u_int64_t)(s[0])<<32) | (u_int64_t)(s[1])) #define pf_state_counter_ntoh(s,d) do { \ d = ntohl(s[0]); \ d = d<<32; \ d += ntohl(s[1]); \ } while (0) TAILQ_HEAD(pf_krulequeue, pf_krule); struct pf_kanchor; struct pf_kruleset { struct { struct pf_krulequeue queues[2]; struct { struct pf_krulequeue *ptr; struct pf_krule **ptr_array; u_int32_t rcount; u_int32_t ticket; int open; struct pf_krule_global *tree; } active, inactive; } rules[PF_RULESET_MAX]; struct pf_kanchor *anchor; u_int32_t tticket; int tables; int topen; }; RB_HEAD(pf_kanchor_global, pf_kanchor); RB_HEAD(pf_kanchor_node, pf_kanchor); struct pf_kanchor { RB_ENTRY(pf_kanchor) entry_global; RB_ENTRY(pf_kanchor) entry_node; struct pf_kanchor *parent; struct pf_kanchor_node children; char name[PF_ANCHOR_NAME_SIZE]; char path[MAXPATHLEN]; struct pf_kruleset ruleset; int refcnt; /* anchor rules */ }; RB_PROTOTYPE(pf_kanchor_global, pf_kanchor, entry_global, pf_anchor_compare); RB_PROTOTYPE(pf_kanchor_node, pf_kanchor, entry_node, pf_kanchor_compare); #define PF_RESERVED_ANCHOR "_pf" #define PFR_TFLAG_PERSIST 0x00000001 #define PFR_TFLAG_CONST 0x00000002 #define PFR_TFLAG_ACTIVE 0x00000004 #define PFR_TFLAG_INACTIVE 0x00000008 #define PFR_TFLAG_REFERENCED 0x00000010 #define PFR_TFLAG_REFDANCHOR 0x00000020 #define PFR_TFLAG_COUNTERS 0x00000040 /* Adjust masks below when adding flags. */ #define PFR_TFLAG_USRMASK (PFR_TFLAG_PERSIST | \ PFR_TFLAG_CONST | \ PFR_TFLAG_COUNTERS) #define PFR_TFLAG_SETMASK (PFR_TFLAG_ACTIVE | \ PFR_TFLAG_INACTIVE | \ PFR_TFLAG_REFERENCED | \ PFR_TFLAG_REFDANCHOR) #define PFR_TFLAG_ALLMASK (PFR_TFLAG_PERSIST | \ PFR_TFLAG_CONST | \ PFR_TFLAG_ACTIVE | \ PFR_TFLAG_INACTIVE | \ PFR_TFLAG_REFERENCED | \ PFR_TFLAG_REFDANCHOR | \ PFR_TFLAG_COUNTERS) struct pf_kanchor_stackframe; struct pf_keth_anchor_stackframe; struct pfr_table { char pfrt_anchor[MAXPATHLEN]; char pfrt_name[PF_TABLE_NAME_SIZE]; u_int32_t pfrt_flags; u_int8_t pfrt_fback; }; enum { PFR_FB_NONE, PFR_FB_MATCH, PFR_FB_ADDED, PFR_FB_DELETED, PFR_FB_CHANGED, PFR_FB_CLEARED, PFR_FB_DUPLICATE, PFR_FB_NOTMATCH, PFR_FB_CONFLICT, PFR_FB_NOCOUNT, PFR_FB_MAX }; struct pfr_addr { union { struct in_addr _pfra_ip4addr; struct in6_addr _pfra_ip6addr; } pfra_u; u_int8_t pfra_af; u_int8_t pfra_net; u_int8_t pfra_not; u_int8_t pfra_fback; }; #define pfra_ip4addr pfra_u._pfra_ip4addr #define pfra_ip6addr pfra_u._pfra_ip6addr enum { PFR_DIR_IN, PFR_DIR_OUT, PFR_DIR_MAX }; enum { PFR_OP_BLOCK, PFR_OP_PASS, PFR_OP_ADDR_MAX, PFR_OP_TABLE_MAX }; enum { PFR_TYPE_PACKETS, PFR_TYPE_BYTES, PFR_TYPE_MAX }; #define PFR_NUM_COUNTERS (PFR_DIR_MAX * PFR_OP_ADDR_MAX * PFR_TYPE_MAX) #define PFR_OP_XPASS PFR_OP_ADDR_MAX struct pfr_astats { struct pfr_addr pfras_a; u_int64_t pfras_packets[PFR_DIR_MAX][PFR_OP_ADDR_MAX]; u_int64_t pfras_bytes[PFR_DIR_MAX][PFR_OP_ADDR_MAX]; long pfras_tzero; }; enum { PFR_REFCNT_RULE, PFR_REFCNT_ANCHOR, PFR_REFCNT_MAX }; struct pfr_tstats { struct pfr_table pfrts_t; u_int64_t pfrts_packets[PFR_DIR_MAX][PFR_OP_TABLE_MAX]; u_int64_t pfrts_bytes[PFR_DIR_MAX][PFR_OP_TABLE_MAX]; u_int64_t pfrts_match; u_int64_t pfrts_nomatch; long pfrts_tzero; int pfrts_cnt; int pfrts_refcnt[PFR_REFCNT_MAX]; }; #ifdef _KERNEL struct pfr_kstate_counter { counter_u64_t pkc_pcpu; u_int64_t pkc_zero; }; static inline int pfr_kstate_counter_init(struct pfr_kstate_counter *pfrc, int flags) { pfrc->pkc_zero = 0; pfrc->pkc_pcpu = counter_u64_alloc(flags); if (pfrc->pkc_pcpu == NULL) return (ENOMEM); return (0); } static inline void pfr_kstate_counter_deinit(struct pfr_kstate_counter *pfrc) { counter_u64_free(pfrc->pkc_pcpu); } static inline u_int64_t pfr_kstate_counter_fetch(struct pfr_kstate_counter *pfrc) { u_int64_t c; c = counter_u64_fetch(pfrc->pkc_pcpu); c -= pfrc->pkc_zero; return (c); } static inline void pfr_kstate_counter_zero(struct pfr_kstate_counter *pfrc) { u_int64_t c; c = counter_u64_fetch(pfrc->pkc_pcpu); pfrc->pkc_zero = c; } static inline void pfr_kstate_counter_add(struct pfr_kstate_counter *pfrc, int64_t n) { counter_u64_add(pfrc->pkc_pcpu, n); } struct pfr_ktstats { struct pfr_table pfrts_t; struct pfr_kstate_counter pfrkts_packets[PFR_DIR_MAX][PFR_OP_TABLE_MAX]; struct pfr_kstate_counter pfrkts_bytes[PFR_DIR_MAX][PFR_OP_TABLE_MAX]; struct pfr_kstate_counter pfrkts_match; struct pfr_kstate_counter pfrkts_nomatch; long pfrkts_tzero; int pfrkts_cnt; int pfrkts_refcnt[PFR_REFCNT_MAX]; }; #endif /* _KERNEL */ #define pfrts_name pfrts_t.pfrt_name #define pfrts_flags pfrts_t.pfrt_flags #ifndef _SOCKADDR_UNION_DEFINED #define _SOCKADDR_UNION_DEFINED union sockaddr_union { struct sockaddr sa; struct sockaddr_in sin; struct sockaddr_in6 sin6; }; #endif /* _SOCKADDR_UNION_DEFINED */ struct pfr_kcounters { counter_u64_t pfrkc_counters; long pfrkc_tzero; }; #define pfr_kentry_counter(kc, dir, op, t) \ ((kc)->pfrkc_counters + \ (dir) * PFR_OP_ADDR_MAX * PFR_TYPE_MAX + (op) * PFR_TYPE_MAX + (t)) #ifdef _KERNEL SLIST_HEAD(pfr_kentryworkq, pfr_kentry); struct pfr_kentry { struct radix_node pfrke_node[2]; union sockaddr_union pfrke_sa; SLIST_ENTRY(pfr_kentry) pfrke_workq; struct pfr_kcounters pfrke_counters; u_int8_t pfrke_af; u_int8_t pfrke_net; u_int8_t pfrke_not; u_int8_t pfrke_mark; }; SLIST_HEAD(pfr_ktableworkq, pfr_ktable); RB_HEAD(pfr_ktablehead, pfr_ktable); struct pfr_ktable { struct pfr_ktstats pfrkt_kts; RB_ENTRY(pfr_ktable) pfrkt_tree; SLIST_ENTRY(pfr_ktable) pfrkt_workq; struct radix_node_head *pfrkt_ip4; struct radix_node_head *pfrkt_ip6; struct pfr_ktable *pfrkt_shadow; struct pfr_ktable *pfrkt_root; struct pf_kruleset *pfrkt_rs; long pfrkt_larg; int pfrkt_nflags; }; #define pfrkt_t pfrkt_kts.pfrts_t #define pfrkt_name pfrkt_t.pfrt_name #define pfrkt_anchor pfrkt_t.pfrt_anchor #define pfrkt_ruleset pfrkt_t.pfrt_ruleset #define pfrkt_flags pfrkt_t.pfrt_flags #define pfrkt_cnt pfrkt_kts.pfrkts_cnt #define pfrkt_refcnt pfrkt_kts.pfrkts_refcnt #define pfrkt_packets pfrkt_kts.pfrkts_packets #define pfrkt_bytes pfrkt_kts.pfrkts_bytes #define pfrkt_match pfrkt_kts.pfrkts_match #define pfrkt_nomatch pfrkt_kts.pfrkts_nomatch #define pfrkt_tzero pfrkt_kts.pfrkts_tzero #endif #ifdef _KERNEL struct pfi_kkif { char pfik_name[IFNAMSIZ]; union { RB_ENTRY(pfi_kkif) _pfik_tree; LIST_ENTRY(pfi_kkif) _pfik_list; } _pfik_glue; #define pfik_tree _pfik_glue._pfik_tree #define pfik_list _pfik_glue._pfik_list struct pf_counter_u64 pfik_packets[2][2][2]; struct pf_counter_u64 pfik_bytes[2][2][2]; u_int32_t pfik_tzero; u_int pfik_flags; struct ifnet *pfik_ifp; struct ifg_group *pfik_group; u_int pfik_rulerefs; TAILQ_HEAD(, pfi_dynaddr) pfik_dynaddrs; #ifdef PF_WANT_32_TO_64_COUNTER LIST_ENTRY(pfi_kkif) pfik_allkiflist; #endif }; #endif #define PFI_IFLAG_REFS 0x0001 /* has state references */ #define PFI_IFLAG_SKIP 0x0100 /* skip filtering on interface */ #ifdef _KERNEL struct pf_sctp_multihome_job; TAILQ_HEAD(pf_sctp_multihome_jobs, pf_sctp_multihome_job); struct pf_pdesc { struct { int done; uid_t uid; gid_t gid; } lookup; u_int64_t tot_len; /* Make Mickey money */ union pf_headers { struct tcphdr tcp; struct udphdr udp; struct sctphdr sctp; struct icmp icmp; #ifdef INET6 struct icmp6_hdr icmp6; #endif /* INET6 */ char any[0]; } hdr; struct pf_krule *nat_rule; /* nat/rdr rule applied to packet */ struct pf_addr *src; /* src address */ struct pf_addr *dst; /* dst address */ u_int16_t *sport; u_int16_t *dport; struct pf_mtag *pf_mtag; struct pf_rule_actions act; u_int32_t p_len; /* total length of payload */ u_int32_t badopts; /* v4 options or v6 routing headers */ u_int16_t *ip_sum; u_int16_t *proto_sum; u_int16_t flags; /* Let SCRUB trigger behavior in * state code. Easier than tags */ #define PFDESC_TCP_NORM 0x0001 /* TCP shall be statefully scrubbed */ #define PFDESC_IP_REAS 0x0002 /* IP frags would've been reassembled */ + u_int16_t virtual_proto; +#define PF_VPROTO_FRAGMENT 256 sa_family_t af; u_int8_t proto; u_int8_t tos; u_int8_t dir; /* direction */ u_int8_t sidx; /* key index for source */ u_int8_t didx; /* key index for destination */ #define PFDESC_SCTP_INIT 0x0001 #define PFDESC_SCTP_INIT_ACK 0x0002 #define PFDESC_SCTP_COOKIE 0x0004 #define PFDESC_SCTP_COOKIE_ACK 0x0008 #define PFDESC_SCTP_ABORT 0x0010 #define PFDESC_SCTP_SHUTDOWN 0x0020 #define PFDESC_SCTP_SHUTDOWN_COMPLETE 0x0040 #define PFDESC_SCTP_DATA 0x0080 #define PFDESC_SCTP_ASCONF 0x0100 #define PFDESC_SCTP_HEARTBEAT 0x0200 #define PFDESC_SCTP_HEARTBEAT_ACK 0x0400 #define PFDESC_SCTP_OTHER 0x0800 #define PFDESC_SCTP_ADD_IP 0x1000 u_int16_t sctp_flags; u_int32_t sctp_initiate_tag; struct pf_sctp_multihome_jobs sctp_multihome_jobs; }; struct pf_sctp_multihome_job { TAILQ_ENTRY(pf_sctp_multihome_job) next; struct pf_pdesc pd; struct pf_addr src; struct pf_addr dst; struct mbuf *m; int op; }; #endif /* flags for RDR options */ #define PF_DPORT_RANGE 0x01 /* Dest port uses range */ #define PF_RPORT_RANGE 0x02 /* RDR'ed port uses range */ /* UDP state enumeration */ #define PFUDPS_NO_TRAFFIC 0 #define PFUDPS_SINGLE 1 #define PFUDPS_MULTIPLE 2 #define PFUDPS_NSTATES 3 /* number of state levels */ #define PFUDPS_NAMES { \ "NO_TRAFFIC", \ "SINGLE", \ "MULTIPLE", \ NULL \ } /* Other protocol state enumeration */ #define PFOTHERS_NO_TRAFFIC 0 #define PFOTHERS_SINGLE 1 #define PFOTHERS_MULTIPLE 2 #define PFOTHERS_NSTATES 3 /* number of state levels */ #define PFOTHERS_NAMES { \ "NO_TRAFFIC", \ "SINGLE", \ "MULTIPLE", \ NULL \ } #define ACTION_SET(a, x) \ do { \ if ((a) != NULL) \ *(a) = (x); \ } while (0) #define REASON_SET(a, x) \ do { \ if ((a) != NULL) \ *(a) = (x); \ if (x < PFRES_MAX) \ counter_u64_add(V_pf_status.counters[x], 1); \ } while (0) enum pf_syncookies_mode { PF_SYNCOOKIES_NEVER = 0, PF_SYNCOOKIES_ALWAYS = 1, PF_SYNCOOKIES_ADAPTIVE = 2, PF_SYNCOOKIES_MODE_MAX = PF_SYNCOOKIES_ADAPTIVE }; #define PF_SYNCOOKIES_HIWATPCT 25 #define PF_SYNCOOKIES_LOWATPCT (PF_SYNCOOKIES_HIWATPCT / 2) #ifdef _KERNEL struct pf_kstatus { counter_u64_t counters[PFRES_MAX]; /* reason for passing/dropping */ counter_u64_t lcounters[KLCNT_MAX]; /* limit counters */ struct pf_counter_u64 fcounters[FCNT_MAX]; /* state operation counters */ counter_u64_t scounters[SCNT_MAX]; /* src_node operation counters */ uint32_t states; uint32_t src_nodes; uint32_t running; uint32_t since; uint32_t debug; uint32_t hostid; char ifname[IFNAMSIZ]; uint8_t pf_chksum[PF_MD5_DIGEST_LENGTH]; bool keep_counters; enum pf_syncookies_mode syncookies_mode; bool syncookies_active; uint64_t syncookies_inflight[2]; uint32_t states_halfopen; uint32_t reass; }; #endif struct pf_divert { union { struct in_addr ipv4; struct in6_addr ipv6; } addr; u_int16_t port; }; #define PFFRAG_FRENT_HIWAT 5000 /* Number of fragment entries */ #define PFR_KENTRY_HIWAT 200000 /* Number of table entries */ /* * Limit the length of the fragment queue traversal. Remember * search entry points based on the fragment offset. */ #define PF_FRAG_ENTRY_POINTS 16 /* * The number of entries in the fragment queue must be limited * to avoid DoS by linear searching. Instead of a global limit, * use a limit per entry point. For large packets these sum up. */ #define PF_FRAG_ENTRY_LIMIT 64 /* * ioctl parameter structures */ struct pfioc_pooladdr { u_int32_t action; u_int32_t ticket; u_int32_t nr; u_int32_t r_num; u_int8_t r_action; u_int8_t r_last; u_int8_t af; char anchor[MAXPATHLEN]; struct pf_pooladdr addr; }; struct pfioc_rule { u_int32_t action; u_int32_t ticket; u_int32_t pool_ticket; u_int32_t nr; char anchor[MAXPATHLEN]; char anchor_call[MAXPATHLEN]; struct pf_rule rule; }; struct pfioc_natlook { struct pf_addr saddr; struct pf_addr daddr; struct pf_addr rsaddr; struct pf_addr rdaddr; u_int16_t sport; u_int16_t dport; u_int16_t rsport; u_int16_t rdport; sa_family_t af; u_int8_t proto; u_int8_t direction; }; struct pfioc_state { struct pfsync_state_1301 state; }; struct pfioc_src_node_kill { sa_family_t psnk_af; struct pf_rule_addr psnk_src; struct pf_rule_addr psnk_dst; u_int psnk_killed; }; #ifdef _KERNEL struct pf_kstate_kill { struct pf_state_cmp psk_pfcmp; sa_family_t psk_af; int psk_proto; struct pf_rule_addr psk_src; struct pf_rule_addr psk_dst; struct pf_rule_addr psk_rt_addr; char psk_ifname[IFNAMSIZ]; char psk_label[PF_RULE_LABEL_SIZE]; u_int psk_killed; bool psk_kill_match; bool psk_nat; }; #endif struct pfioc_state_kill { struct pf_state_cmp psk_pfcmp; sa_family_t psk_af; int psk_proto; struct pf_rule_addr psk_src; struct pf_rule_addr psk_dst; char psk_ifname[IFNAMSIZ]; char psk_label[PF_RULE_LABEL_SIZE]; u_int psk_killed; }; struct pfioc_states { int ps_len; union { void *ps_buf; struct pfsync_state_1301 *ps_states; }; }; struct pfioc_states_v2 { int ps_len; uint64_t ps_req_version; union { void *ps_buf; struct pf_state_export *ps_states; }; }; struct pfioc_src_nodes { int psn_len; union { void *psn_buf; struct pf_src_node *psn_src_nodes; }; }; struct pfioc_if { char ifname[IFNAMSIZ]; }; struct pfioc_tm { int timeout; int seconds; }; struct pfioc_limit { int index; unsigned limit; }; struct pfioc_altq_v0 { u_int32_t action; u_int32_t ticket; u_int32_t nr; struct pf_altq_v0 altq; }; struct pfioc_altq_v1 { u_int32_t action; u_int32_t ticket; u_int32_t nr; /* * Placed here so code that only uses the above parameters can be * written entirely in terms of the v0 or v1 type. */ u_int32_t version; struct pf_altq_v1 altq; }; /* * Latest version of struct pfioc_altq_vX. This must move in lock-step with * the latest version of struct pf_altq_vX as it has that struct as a * member. */ #define PFIOC_ALTQ_VERSION PF_ALTQ_VERSION struct pfioc_qstats_v0 { u_int32_t ticket; u_int32_t nr; void *buf; int nbytes; u_int8_t scheduler; }; struct pfioc_qstats_v1 { u_int32_t ticket; u_int32_t nr; void *buf; int nbytes; u_int8_t scheduler; /* * Placed here so code that only uses the above parameters can be * written entirely in terms of the v0 or v1 type. */ u_int32_t version; /* Requested version of stats struct */ }; /* Latest version of struct pfioc_qstats_vX */ #define PFIOC_QSTATS_VERSION 1 struct pfioc_ruleset { u_int32_t nr; char path[MAXPATHLEN]; char name[PF_ANCHOR_NAME_SIZE]; }; #define PF_RULESET_ALTQ (PF_RULESET_MAX) #define PF_RULESET_TABLE (PF_RULESET_MAX+1) #define PF_RULESET_ETH (PF_RULESET_MAX+2) struct pfioc_trans { int size; /* number of elements */ int esize; /* size of each element in bytes */ struct pfioc_trans_e { int rs_num; char anchor[MAXPATHLEN]; u_int32_t ticket; } *array; }; #define PFR_FLAG_ATOMIC 0x00000001 /* unused */ #define PFR_FLAG_DUMMY 0x00000002 #define PFR_FLAG_FEEDBACK 0x00000004 #define PFR_FLAG_CLSTATS 0x00000008 #define PFR_FLAG_ADDRSTOO 0x00000010 #define PFR_FLAG_REPLACE 0x00000020 #define PFR_FLAG_ALLRSETS 0x00000040 #define PFR_FLAG_ALLMASK 0x0000007F #ifdef _KERNEL #define PFR_FLAG_USERIOCTL 0x10000000 #endif struct pfioc_table { struct pfr_table pfrio_table; void *pfrio_buffer; int pfrio_esize; int pfrio_size; int pfrio_size2; int pfrio_nadd; int pfrio_ndel; int pfrio_nchange; int pfrio_flags; u_int32_t pfrio_ticket; }; #define pfrio_exists pfrio_nadd #define pfrio_nzero pfrio_nadd #define pfrio_nmatch pfrio_nadd #define pfrio_naddr pfrio_size2 #define pfrio_setflag pfrio_size2 #define pfrio_clrflag pfrio_nadd struct pfioc_iface { char pfiio_name[IFNAMSIZ]; void *pfiio_buffer; int pfiio_esize; int pfiio_size; int pfiio_nzero; int pfiio_flags; }; /* * ioctl operations */ #define DIOCSTART _IO ('D', 1) #define DIOCSTOP _IO ('D', 2) #define DIOCADDRULE _IOWR('D', 4, struct pfioc_rule) #define DIOCADDRULENV _IOWR('D', 4, struct pfioc_nv) #define DIOCGETRULES _IOWR('D', 6, struct pfioc_rule) #define DIOCGETRULENV _IOWR('D', 7, struct pfioc_nv) /* XXX cut 8 - 17 */ #define DIOCCLRSTATESNV _IOWR('D', 18, struct pfioc_nv) #define DIOCGETSTATE _IOWR('D', 19, struct pfioc_state) #define DIOCGETSTATENV _IOWR('D', 19, struct pfioc_nv) #define DIOCSETSTATUSIF _IOWR('D', 20, struct pfioc_if) #define DIOCGETSTATUSNV _IOWR('D', 21, struct pfioc_nv) #define DIOCCLRSTATUS _IO ('D', 22) #define DIOCNATLOOK _IOWR('D', 23, struct pfioc_natlook) #define DIOCSETDEBUG _IOWR('D', 24, u_int32_t) #ifdef COMPAT_FREEBSD14 #define DIOCGETSTATES _IOWR('D', 25, struct pfioc_states) #endif #define DIOCCHANGERULE _IOWR('D', 26, struct pfioc_rule) /* XXX cut 26 - 28 */ #define DIOCSETTIMEOUT _IOWR('D', 29, struct pfioc_tm) #define DIOCGETTIMEOUT _IOWR('D', 30, struct pfioc_tm) #define DIOCADDSTATE _IOWR('D', 37, struct pfioc_state) #define DIOCCLRRULECTRS _IO ('D', 38) #define DIOCGETLIMIT _IOWR('D', 39, struct pfioc_limit) #define DIOCSETLIMIT _IOWR('D', 40, struct pfioc_limit) #define DIOCKILLSTATESNV _IOWR('D', 41, struct pfioc_nv) #define DIOCSTARTALTQ _IO ('D', 42) #define DIOCSTOPALTQ _IO ('D', 43) #define DIOCADDALTQV0 _IOWR('D', 45, struct pfioc_altq_v0) #define DIOCADDALTQV1 _IOWR('D', 45, struct pfioc_altq_v1) #define DIOCGETALTQSV0 _IOWR('D', 47, struct pfioc_altq_v0) #define DIOCGETALTQSV1 _IOWR('D', 47, struct pfioc_altq_v1) #define DIOCGETALTQV0 _IOWR('D', 48, struct pfioc_altq_v0) #define DIOCGETALTQV1 _IOWR('D', 48, struct pfioc_altq_v1) #define DIOCCHANGEALTQV0 _IOWR('D', 49, struct pfioc_altq_v0) #define DIOCCHANGEALTQV1 _IOWR('D', 49, struct pfioc_altq_v1) #define DIOCGETQSTATSV0 _IOWR('D', 50, struct pfioc_qstats_v0) #define DIOCGETQSTATSV1 _IOWR('D', 50, struct pfioc_qstats_v1) #define DIOCBEGINADDRS _IOWR('D', 51, struct pfioc_pooladdr) #define DIOCADDADDR _IOWR('D', 52, struct pfioc_pooladdr) #define DIOCGETADDRS _IOWR('D', 53, struct pfioc_pooladdr) #define DIOCGETADDR _IOWR('D', 54, struct pfioc_pooladdr) #define DIOCCHANGEADDR _IOWR('D', 55, struct pfioc_pooladdr) /* XXX cut 55 - 57 */ #define DIOCGETRULESETS _IOWR('D', 58, struct pfioc_ruleset) #define DIOCGETRULESET _IOWR('D', 59, struct pfioc_ruleset) #define DIOCRCLRTABLES _IOWR('D', 60, struct pfioc_table) #define DIOCRADDTABLES _IOWR('D', 61, struct pfioc_table) #define DIOCRDELTABLES _IOWR('D', 62, struct pfioc_table) #define DIOCRGETTABLES _IOWR('D', 63, struct pfioc_table) #define DIOCRGETTSTATS _IOWR('D', 64, struct pfioc_table) #define DIOCRCLRTSTATS _IOWR('D', 65, struct pfioc_table) #define DIOCRCLRADDRS _IOWR('D', 66, struct pfioc_table) #define DIOCRADDADDRS _IOWR('D', 67, struct pfioc_table) #define DIOCRDELADDRS _IOWR('D', 68, struct pfioc_table) #define DIOCRSETADDRS _IOWR('D', 69, struct pfioc_table) #define DIOCRGETADDRS _IOWR('D', 70, struct pfioc_table) #define DIOCRGETASTATS _IOWR('D', 71, struct pfioc_table) #define DIOCRCLRASTATS _IOWR('D', 72, struct pfioc_table) #define DIOCRTSTADDRS _IOWR('D', 73, struct pfioc_table) #define DIOCRSETTFLAGS _IOWR('D', 74, struct pfioc_table) #define DIOCRINADEFINE _IOWR('D', 77, struct pfioc_table) #define DIOCOSFPFLUSH _IO('D', 78) #define DIOCOSFPADD _IOWR('D', 79, struct pf_osfp_ioctl) #define DIOCOSFPGET _IOWR('D', 80, struct pf_osfp_ioctl) #define DIOCXBEGIN _IOWR('D', 81, struct pfioc_trans) #define DIOCXCOMMIT _IOWR('D', 82, struct pfioc_trans) #define DIOCXROLLBACK _IOWR('D', 83, struct pfioc_trans) #define DIOCGETSRCNODES _IOWR('D', 84, struct pfioc_src_nodes) #define DIOCCLRSRCNODES _IO('D', 85) #define DIOCSETHOSTID _IOWR('D', 86, u_int32_t) #define DIOCIGETIFACES _IOWR('D', 87, struct pfioc_iface) #define DIOCSETIFFLAG _IOWR('D', 89, struct pfioc_iface) #define DIOCCLRIFFLAG _IOWR('D', 90, struct pfioc_iface) #define DIOCKILLSRCNODES _IOWR('D', 91, struct pfioc_src_node_kill) #define DIOCGIFSPEEDV0 _IOWR('D', 92, struct pf_ifspeed_v0) #define DIOCGIFSPEEDV1 _IOWR('D', 92, struct pf_ifspeed_v1) #ifdef COMPAT_FREEBSD14 #define DIOCGETSTATESV2 _IOWR('D', 93, struct pfioc_states_v2) #endif #define DIOCGETSYNCOOKIES _IOWR('D', 94, struct pfioc_nv) #define DIOCSETSYNCOOKIES _IOWR('D', 95, struct pfioc_nv) #define DIOCKEEPCOUNTERS _IOWR('D', 96, struct pfioc_nv) #define DIOCKEEPCOUNTERS_FREEBSD13 _IOWR('D', 92, struct pfioc_nv) #define DIOCADDETHRULE _IOWR('D', 97, struct pfioc_nv) #define DIOCGETETHRULE _IOWR('D', 98, struct pfioc_nv) #define DIOCGETETHRULES _IOWR('D', 99, struct pfioc_nv) #define DIOCGETETHRULESETS _IOWR('D', 100, struct pfioc_nv) #define DIOCGETETHRULESET _IOWR('D', 101, struct pfioc_nv) #define DIOCSETREASS _IOWR('D', 102, u_int32_t) struct pf_ifspeed_v0 { char ifname[IFNAMSIZ]; u_int32_t baudrate; }; struct pf_ifspeed_v1 { char ifname[IFNAMSIZ]; u_int32_t baudrate32; /* layout identical to struct pf_ifspeed_v0 up to this point */ u_int64_t baudrate; }; /* Latest version of struct pf_ifspeed_vX */ #define PF_IFSPEED_VERSION 1 /* * Compatibility and convenience macros */ #ifndef _KERNEL #ifdef PFIOC_USE_LATEST /* * Maintaining in-tree consumers of the ioctl interface is easier when that * code can be written in terms old names that refer to the latest interface * version as that reduces the required changes in the consumers to those * that are functionally necessary to accommodate a new interface version. */ #define pfioc_altq __CONCAT(pfioc_altq_v, PFIOC_ALTQ_VERSION) #define pfioc_qstats __CONCAT(pfioc_qstats_v, PFIOC_QSTATS_VERSION) #define pf_ifspeed __CONCAT(pf_ifspeed_v, PF_IFSPEED_VERSION) #define DIOCADDALTQ __CONCAT(DIOCADDALTQV, PFIOC_ALTQ_VERSION) #define DIOCGETALTQS __CONCAT(DIOCGETALTQSV, PFIOC_ALTQ_VERSION) #define DIOCGETALTQ __CONCAT(DIOCGETALTQV, PFIOC_ALTQ_VERSION) #define DIOCCHANGEALTQ __CONCAT(DIOCCHANGEALTQV, PFIOC_ALTQ_VERSION) #define DIOCGETQSTATS __CONCAT(DIOCGETQSTATSV, PFIOC_QSTATS_VERSION) #define DIOCGIFSPEED __CONCAT(DIOCGIFSPEEDV, PF_IFSPEED_VERSION) #else /* * When building out-of-tree code that is written for the old interface, * such as may exist in ports for example, resolve the old struct tags and * ioctl command names to the v0 versions. */ #define pfioc_altq __CONCAT(pfioc_altq_v, 0) #define pfioc_qstats __CONCAT(pfioc_qstats_v, 0) #define pf_ifspeed __CONCAT(pf_ifspeed_v, 0) #define DIOCADDALTQ __CONCAT(DIOCADDALTQV, 0) #define DIOCGETALTQS __CONCAT(DIOCGETALTQSV, 0) #define DIOCGETALTQ __CONCAT(DIOCGETALTQV, 0) #define DIOCCHANGEALTQ __CONCAT(DIOCCHANGEALTQV, 0) #define DIOCGETQSTATS __CONCAT(DIOCGETQSTATSV, 0) #define DIOCGIFSPEED __CONCAT(DIOCGIFSPEEDV, 0) #endif /* PFIOC_USE_LATEST */ #endif /* _KERNEL */ #ifdef _KERNEL LIST_HEAD(pf_ksrc_node_list, pf_ksrc_node); struct pf_srchash { struct pf_ksrc_node_list nodes; struct mtx lock; }; struct pf_keyhash { LIST_HEAD(, pf_state_key) keys; struct mtx lock; }; struct pf_idhash { LIST_HEAD(, pf_kstate) states; struct mtx lock; }; struct pf_udpendpointhash { LIST_HEAD(, pf_udp_endpoint) endpoints; /* refcont is synchronized on the source endpoint's row lock */ struct mtx lock; }; extern u_long pf_ioctl_maxcount; VNET_DECLARE(u_long, pf_hashmask); #define V_pf_hashmask VNET(pf_hashmask) VNET_DECLARE(u_long, pf_srchashmask); #define V_pf_srchashmask VNET(pf_srchashmask) VNET_DECLARE(u_long, pf_udpendpointhashmask); #define V_pf_udpendpointhashmask VNET(pf_udpendpointhashmask) #define PF_HASHSIZ (131072) #define PF_SRCHASHSIZ (PF_HASHSIZ/4) #define PF_UDPENDHASHSIZ (PF_HASHSIZ/4) VNET_DECLARE(struct pf_keyhash *, pf_keyhash); VNET_DECLARE(struct pf_idhash *, pf_idhash); VNET_DECLARE(struct pf_udpendpointhash *, pf_udpendpointhash); #define V_pf_keyhash VNET(pf_keyhash) #define V_pf_idhash VNET(pf_idhash) #define V_pf_udpendpointhash VNET(pf_udpendpointhash) VNET_DECLARE(struct pf_srchash *, pf_srchash); #define V_pf_srchash VNET(pf_srchash) #define PF_IDHASH(s) (be64toh((s)->id) % (V_pf_hashmask + 1)) VNET_DECLARE(void *, pf_swi_cookie); #define V_pf_swi_cookie VNET(pf_swi_cookie) VNET_DECLARE(struct intr_event *, pf_swi_ie); #define V_pf_swi_ie VNET(pf_swi_ie) VNET_DECLARE(struct unrhdr64, pf_stateid); #define V_pf_stateid VNET(pf_stateid) TAILQ_HEAD(pf_altqqueue, pf_altq); VNET_DECLARE(struct pf_altqqueue, pf_altqs[4]); #define V_pf_altqs VNET(pf_altqs) VNET_DECLARE(struct pf_kpalist, pf_pabuf); #define V_pf_pabuf VNET(pf_pabuf) VNET_DECLARE(u_int32_t, ticket_altqs_active); #define V_ticket_altqs_active VNET(ticket_altqs_active) VNET_DECLARE(u_int32_t, ticket_altqs_inactive); #define V_ticket_altqs_inactive VNET(ticket_altqs_inactive) VNET_DECLARE(int, altqs_inactive_open); #define V_altqs_inactive_open VNET(altqs_inactive_open) VNET_DECLARE(u_int32_t, ticket_pabuf); #define V_ticket_pabuf VNET(ticket_pabuf) VNET_DECLARE(struct pf_altqqueue *, pf_altqs_active); #define V_pf_altqs_active VNET(pf_altqs_active) VNET_DECLARE(struct pf_altqqueue *, pf_altq_ifs_active); #define V_pf_altq_ifs_active VNET(pf_altq_ifs_active) VNET_DECLARE(struct pf_altqqueue *, pf_altqs_inactive); #define V_pf_altqs_inactive VNET(pf_altqs_inactive) VNET_DECLARE(struct pf_altqqueue *, pf_altq_ifs_inactive); #define V_pf_altq_ifs_inactive VNET(pf_altq_ifs_inactive) VNET_DECLARE(struct pf_krulequeue, pf_unlinked_rules); #define V_pf_unlinked_rules VNET(pf_unlinked_rules) #ifdef PF_WANT_32_TO_64_COUNTER LIST_HEAD(allkiflist_head, pfi_kkif); VNET_DECLARE(struct allkiflist_head, pf_allkiflist); #define V_pf_allkiflist VNET(pf_allkiflist) VNET_DECLARE(size_t, pf_allkifcount); #define V_pf_allkifcount VNET(pf_allkifcount) VNET_DECLARE(struct pfi_kkif *, pf_kifmarker); #define V_pf_kifmarker VNET(pf_kifmarker) LIST_HEAD(allrulelist_head, pf_krule); VNET_DECLARE(struct allrulelist_head, pf_allrulelist); #define V_pf_allrulelist VNET(pf_allrulelist) VNET_DECLARE(size_t, pf_allrulecount); #define V_pf_allrulecount VNET(pf_allrulecount) VNET_DECLARE(struct pf_krule *, pf_rulemarker); #define V_pf_rulemarker VNET(pf_rulemarker) #endif int pf_start(void); int pf_stop(void); void pf_initialize(void); void pf_mtag_initialize(void); void pf_mtag_cleanup(void); void pf_cleanup(void); struct pf_mtag *pf_get_mtag(struct mbuf *); extern void pf_calc_skip_steps(struct pf_krulequeue *); #ifdef ALTQ extern void pf_altq_ifnet_event(struct ifnet *, int); #endif VNET_DECLARE(uma_zone_t, pf_state_z); #define V_pf_state_z VNET(pf_state_z) VNET_DECLARE(uma_zone_t, pf_state_key_z); #define V_pf_state_key_z VNET(pf_state_key_z) VNET_DECLARE(uma_zone_t, pf_udp_mapping_z); #define V_pf_udp_mapping_z VNET(pf_udp_mapping_z) VNET_DECLARE(uma_zone_t, pf_state_scrub_z); #define V_pf_state_scrub_z VNET(pf_state_scrub_z) extern void pf_purge_thread(void *); extern void pf_unload_vnet_purge(void); extern void pf_intr(void *); extern void pf_purge_expired_src_nodes(void); extern int pf_unlink_state(struct pf_kstate *); extern int pf_state_insert(struct pfi_kkif *, struct pfi_kkif *, struct pf_state_key *, struct pf_state_key *, struct pf_kstate *); extern struct pf_kstate *pf_alloc_state(int); extern void pf_free_state(struct pf_kstate *); extern void pf_killstates(struct pf_kstate_kill *, unsigned int *); extern unsigned int pf_clear_states(const struct pf_kstate_kill *); static __inline void pf_ref_state(struct pf_kstate *s) { refcount_acquire(&s->refs); } static __inline int pf_release_state(struct pf_kstate *s) { if (refcount_release(&s->refs)) { pf_free_state(s); return (1); } else return (0); } static __inline int pf_release_staten(struct pf_kstate *s, u_int n) { if (refcount_releasen(&s->refs, n)) { pf_free_state(s); return (1); } else return (0); } static __inline uint64_t pf_get_uptime(void) { struct timeval t; microuptime(&t); return ((t.tv_sec * 1000) + (t.tv_usec / 1000)); } static __inline uint64_t pf_get_time(void) { struct timeval t; microtime(&t); return ((t.tv_sec * 1000) + (t.tv_usec / 1000)); } extern struct pf_kstate *pf_find_state_byid(uint64_t, uint32_t); extern struct pf_kstate *pf_find_state_all( const struct pf_state_key_cmp *, u_int, int *); extern bool pf_find_state_all_exists( const struct pf_state_key_cmp *, u_int); extern struct pf_udp_mapping *pf_udp_mapping_find(struct pf_udp_endpoint_cmp *endpoint); extern struct pf_udp_mapping *pf_udp_mapping_create(sa_family_t af, struct pf_addr *src_addr, uint16_t src_port, struct pf_addr *nat_addr, uint16_t nat_port); extern int pf_udp_mapping_insert(struct pf_udp_mapping *mapping); extern void pf_udp_mapping_release(struct pf_udp_mapping *mapping); extern struct pf_ksrc_node *pf_find_src_node(struct pf_addr *, struct pf_krule *, sa_family_t, struct pf_srchash **, bool); extern void pf_unlink_src_node(struct pf_ksrc_node *); extern u_int pf_free_src_nodes(struct pf_ksrc_node_list *); extern void pf_print_state(struct pf_kstate *); extern void pf_print_flags(u_int8_t); extern int pf_addr_wrap_neq(struct pf_addr_wrap *, struct pf_addr_wrap *); extern u_int16_t pf_cksum_fixup(u_int16_t, u_int16_t, u_int16_t, u_int8_t); extern u_int16_t pf_proto_cksum_fixup(struct mbuf *, u_int16_t, u_int16_t, u_int16_t, u_int8_t); VNET_DECLARE(struct ifnet *, sync_ifp); #define V_sync_ifp VNET(sync_ifp); VNET_DECLARE(struct pf_krule, pf_default_rule); #define V_pf_default_rule VNET(pf_default_rule) extern void pf_addrcpy(struct pf_addr *, struct pf_addr *, sa_family_t); void pf_free_rule(struct pf_krule *); int pf_test_eth(int, int, struct ifnet *, struct mbuf **, struct inpcb *); int pf_scan_sctp(struct mbuf *, int, struct pf_pdesc *, struct pfi_kkif *); #if defined(INET) || defined(INET6) int pf_test(sa_family_t, int, int, struct ifnet *, struct mbuf **, struct inpcb *, struct pf_rule_actions *); #endif #ifdef INET int pf_normalize_ip(struct mbuf **, struct pfi_kkif *, u_short *, struct pf_pdesc *); #endif /* INET */ #ifdef INET6 int pf_normalize_ip6(struct mbuf **, struct pfi_kkif *, u_short *, struct pf_pdesc *); void pf_poolmask(struct pf_addr *, struct pf_addr*, struct pf_addr *, struct pf_addr *, sa_family_t); void pf_addr_inc(struct pf_addr *, sa_family_t); int pf_max_frag_size(struct mbuf *); int pf_refragment6(struct ifnet *, struct mbuf **, struct m_tag *, bool); #endif /* INET6 */ int pf_multihome_scan_init(struct mbuf *, int, int, struct pf_pdesc *, struct pfi_kkif *); int pf_multihome_scan_asconf(struct mbuf *, int, int, struct pf_pdesc *, struct pfi_kkif *); u_int32_t pf_new_isn(struct pf_kstate *); void *pf_pull_hdr(const struct mbuf *, int, void *, int, u_short *, u_short *, sa_family_t); void pf_change_a(void *, u_int16_t *, u_int32_t, u_int8_t); void pf_change_proto_a(struct mbuf *, void *, u_int16_t *, u_int32_t, u_int8_t); void pf_change_tcp_a(struct mbuf *, void *, u_int16_t *, u_int32_t); void pf_patch_16_unaligned(struct mbuf *, u_int16_t *, void *, u_int16_t, bool, u_int8_t); void pf_patch_32_unaligned(struct mbuf *, u_int16_t *, void *, u_int32_t, bool, u_int8_t); void pf_send_deferred_syn(struct pf_kstate *); int pf_match_addr(u_int8_t, struct pf_addr *, struct pf_addr *, struct pf_addr *, sa_family_t); int pf_match_addr_range(struct pf_addr *, struct pf_addr *, struct pf_addr *, sa_family_t); int pf_match_port(u_int8_t, u_int16_t, u_int16_t, u_int16_t); void pf_normalize_init(void); void pf_normalize_cleanup(void); int pf_normalize_tcp(struct pfi_kkif *, struct mbuf *, int, int, struct pf_pdesc *); void pf_normalize_tcp_cleanup(struct pf_kstate *); int pf_normalize_tcp_init(struct mbuf *, int, struct pf_pdesc *, struct tcphdr *, struct pf_state_peer *, struct pf_state_peer *); int pf_normalize_tcp_stateful(struct mbuf *, int, struct pf_pdesc *, u_short *, struct tcphdr *, struct pf_kstate *, struct pf_state_peer *, struct pf_state_peer *, int *); int pf_normalize_sctp_init(struct mbuf *, int, struct pf_pdesc *, struct pf_state_peer *, struct pf_state_peer *); int pf_normalize_sctp(int, struct pfi_kkif *, struct mbuf *, int, int, struct pf_pdesc *); u_int32_t pf_state_expires(const struct pf_kstate *); void pf_purge_expired_fragments(void); void pf_purge_fragments(uint32_t); int pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kkif *, int); int pf_socket_lookup(struct pf_pdesc *, struct mbuf *); struct pf_state_key *pf_alloc_state_key(int); void pfr_initialize(void); void pfr_cleanup(void); int pfr_match_addr(struct pfr_ktable *, struct pf_addr *, sa_family_t); void pfr_update_stats(struct pfr_ktable *, struct pf_addr *, sa_family_t, u_int64_t, int, int, int); int pfr_pool_get(struct pfr_ktable *, int *, struct pf_addr *, sa_family_t, pf_addr_filter_func_t); void pfr_dynaddr_update(struct pfr_ktable *, struct pfi_dynaddr *); struct pfr_ktable * pfr_attach_table(struct pf_kruleset *, char *); struct pfr_ktable * pfr_eth_attach_table(struct pf_keth_ruleset *, char *); void pfr_detach_table(struct pfr_ktable *); int pfr_clr_tables(struct pfr_table *, int *, int); int pfr_add_tables(struct pfr_table *, int, int *, int); int pfr_del_tables(struct pfr_table *, int, int *, int); int pfr_table_count(struct pfr_table *, int); int pfr_get_tables(struct pfr_table *, struct pfr_table *, int *, int); int pfr_get_tstats(struct pfr_table *, struct pfr_tstats *, int *, int); int pfr_clr_tstats(struct pfr_table *, int, int *, int); int pfr_set_tflags(struct pfr_table *, int, int, int, int *, int *, int); int pfr_clr_addrs(struct pfr_table *, int *, int); int pfr_insert_kentry(struct pfr_ktable *, struct pfr_addr *, long); int pfr_add_addrs(struct pfr_table *, struct pfr_addr *, int, int *, int); int pfr_del_addrs(struct pfr_table *, struct pfr_addr *, int, int *, int); int pfr_set_addrs(struct pfr_table *, struct pfr_addr *, int, int *, int *, int *, int *, int, u_int32_t); int pfr_get_addrs(struct pfr_table *, struct pfr_addr *, int *, int); int pfr_get_astats(struct pfr_table *, struct pfr_astats *, int *, int); int pfr_clr_astats(struct pfr_table *, struct pfr_addr *, int, int *, int); int pfr_tst_addrs(struct pfr_table *, struct pfr_addr *, int, int *, int); int pfr_ina_begin(struct pfr_table *, u_int32_t *, int *, int); int pfr_ina_rollback(struct pfr_table *, u_int32_t, int *, int); int pfr_ina_commit(struct pfr_table *, u_int32_t, int *, int *, int); int pfr_ina_define(struct pfr_table *, struct pfr_addr *, int, int *, int *, u_int32_t, int); MALLOC_DECLARE(PFI_MTYPE); VNET_DECLARE(struct pfi_kkif *, pfi_all); #define V_pfi_all VNET(pfi_all) void pfi_initialize(void); void pfi_initialize_vnet(void); void pfi_cleanup(void); void pfi_cleanup_vnet(void); void pfi_kkif_ref(struct pfi_kkif *); void pfi_kkif_unref(struct pfi_kkif *); struct pfi_kkif *pfi_kkif_find(const char *); struct pfi_kkif *pfi_kkif_attach(struct pfi_kkif *, const char *); int pfi_kkif_match(struct pfi_kkif *, struct pfi_kkif *); void pfi_kkif_purge(void); int pfi_match_addr(struct pfi_dynaddr *, struct pf_addr *, sa_family_t); int pfi_dynaddr_setup(struct pf_addr_wrap *, sa_family_t); void pfi_dynaddr_remove(struct pfi_dynaddr *); void pfi_dynaddr_copyout(struct pf_addr_wrap *); void pfi_update_status(const char *, struct pf_status *); void pfi_get_ifaces(const char *, struct pfi_kif *, int *); int pfi_set_flags(const char *, int); int pfi_clear_flags(const char *, int); int pf_match_tag(struct mbuf *, struct pf_krule *, int *, int); int pf_tag_packet(struct mbuf *, struct pf_pdesc *, int); int pf_addr_cmp(struct pf_addr *, struct pf_addr *, sa_family_t); u_int16_t pf_get_mss(struct mbuf *, int, u_int16_t, sa_family_t); u_int8_t pf_get_wscale(struct mbuf *, int, u_int16_t, sa_family_t); struct mbuf *pf_build_tcp(const struct pf_krule *, sa_family_t, const struct pf_addr *, const struct pf_addr *, u_int16_t, u_int16_t, u_int32_t, u_int32_t, u_int8_t, u_int16_t, u_int16_t, u_int8_t, bool, u_int16_t, u_int16_t, int); void pf_send_tcp(const struct pf_krule *, sa_family_t, const struct pf_addr *, const struct pf_addr *, u_int16_t, u_int16_t, u_int32_t, u_int32_t, u_int8_t, u_int16_t, u_int16_t, u_int8_t, bool, u_int16_t, u_int16_t, int); void pf_syncookies_init(void); void pf_syncookies_cleanup(void); int pf_get_syncookies(struct pfioc_nv *); int pf_set_syncookies(struct pfioc_nv *); int pf_synflood_check(struct pf_pdesc *); void pf_syncookie_send(struct mbuf *m, int off, struct pf_pdesc *); bool pf_syncookie_check(struct pf_pdesc *); u_int8_t pf_syncookie_validate(struct pf_pdesc *); struct mbuf * pf_syncookie_recreate_syn(uint8_t, int, struct pf_pdesc *); VNET_DECLARE(struct pf_kstatus, pf_status); #define V_pf_status VNET(pf_status) struct pf_limit { uma_zone_t zone; u_int limit; }; VNET_DECLARE(struct pf_limit, pf_limits[PF_LIMIT_MAX]); #define V_pf_limits VNET(pf_limits) #endif /* _KERNEL */ #ifdef _KERNEL VNET_DECLARE(struct pf_kanchor_global, pf_anchors); #define V_pf_anchors VNET(pf_anchors) VNET_DECLARE(struct pf_kanchor, pf_main_anchor); #define V_pf_main_anchor VNET(pf_main_anchor) VNET_DECLARE(struct pf_keth_anchor_global, pf_keth_anchors); #define V_pf_keth_anchors VNET(pf_keth_anchors) #define pf_main_ruleset V_pf_main_anchor.ruleset VNET_DECLARE(struct pf_keth_anchor, pf_main_keth_anchor); #define V_pf_main_keth_anchor VNET(pf_main_keth_anchor) VNET_DECLARE(struct pf_keth_ruleset*, pf_keth); #define V_pf_keth VNET(pf_keth) void pf_init_kruleset(struct pf_kruleset *); void pf_init_keth(struct pf_keth_ruleset *); int pf_kanchor_setup(struct pf_krule *, const struct pf_kruleset *, const char *); int pf_kanchor_copyout(const struct pf_kruleset *, const struct pf_krule *, char *, size_t); int pf_kanchor_nvcopyout(const struct pf_kruleset *, const struct pf_krule *, nvlist_t *); void pf_kanchor_remove(struct pf_krule *); void pf_remove_if_empty_kruleset(struct pf_kruleset *); struct pf_kruleset *pf_find_kruleset(const char *); struct pf_kruleset *pf_find_or_create_kruleset(const char *); void pf_rs_initialize(void); struct pf_krule *pf_krule_alloc(void); void pf_remove_if_empty_keth_ruleset( struct pf_keth_ruleset *); struct pf_keth_ruleset *pf_find_keth_ruleset(const char *); struct pf_keth_anchor *pf_find_keth_anchor(const char *); int pf_keth_anchor_setup(struct pf_keth_rule *, const struct pf_keth_ruleset *, const char *); int pf_keth_anchor_nvcopyout( const struct pf_keth_ruleset *, const struct pf_keth_rule *, nvlist_t *); struct pf_keth_ruleset *pf_find_or_create_keth_ruleset(const char *); void pf_keth_anchor_remove(struct pf_keth_rule *); int pf_ioctl_getrules(struct pfioc_rule *); int pf_ioctl_addrule(struct pf_krule *, uint32_t, uint32_t, const char *, const char *, uid_t uid, pid_t); void pf_ioctl_clear_status(void); int pf_ioctl_get_timeout(int, int *); int pf_ioctl_set_timeout(int, int, int *); int pf_ioctl_get_limit(int, unsigned int *); int pf_ioctl_set_limit(int, unsigned int, unsigned int *); int pf_ioctl_begin_addrs(uint32_t *); int pf_ioctl_add_addr(struct pfioc_pooladdr *); int pf_ioctl_get_addrs(struct pfioc_pooladdr *); int pf_ioctl_get_addr(struct pfioc_pooladdr *); void pf_krule_free(struct pf_krule *); void pf_krule_clear_counters(struct pf_krule *); void pf_addr_copyout(struct pf_addr_wrap *); #endif /* The fingerprint functions can be linked into userland programs (tcpdump) */ int pf_osfp_add(struct pf_osfp_ioctl *); #ifdef _KERNEL struct pf_osfp_enlist * pf_osfp_fingerprint(struct pf_pdesc *, struct mbuf *, int, const struct tcphdr *); #endif /* _KERNEL */ void pf_osfp_flush(void); int pf_osfp_get(struct pf_osfp_ioctl *); int pf_osfp_match(struct pf_osfp_enlist *, pf_osfp_t); #ifdef _KERNEL void pf_print_host(struct pf_addr *, u_int16_t, sa_family_t); void pf_step_into_anchor(struct pf_kanchor_stackframe *, int *, struct pf_kruleset **, int, struct pf_krule **, struct pf_krule **, int *); int pf_step_out_of_anchor(struct pf_kanchor_stackframe *, int *, struct pf_kruleset **, int, struct pf_krule **, struct pf_krule **, int *); void pf_step_into_keth_anchor(struct pf_keth_anchor_stackframe *, int *, struct pf_keth_ruleset **, struct pf_keth_rule **, struct pf_keth_rule **, int *); int pf_step_out_of_keth_anchor(struct pf_keth_anchor_stackframe *, int *, struct pf_keth_ruleset **, struct pf_keth_rule **, struct pf_keth_rule **, int *); u_short pf_map_addr(u_int8_t, struct pf_krule *, struct pf_addr *, struct pf_addr *, struct pfi_kkif **nkif, struct pf_addr *); u_short pf_map_addr_sn(u_int8_t, struct pf_krule *, struct pf_addr *, struct pf_addr *, struct pfi_kkif **nkif, struct pf_addr *, struct pf_ksrc_node **); u_short pf_get_translation(struct pf_pdesc *, struct mbuf *, int, struct pfi_kkif *, struct pf_ksrc_node **, struct pf_state_key **, struct pf_state_key **, struct pf_addr *, struct pf_addr *, uint16_t, uint16_t, struct pf_kanchor_stackframe *, struct pf_krule **, struct pf_udp_mapping **udp_mapping); struct pf_state_key *pf_state_key_setup(struct pf_pdesc *, struct mbuf *, int, struct pf_addr *, struct pf_addr *, u_int16_t, u_int16_t); struct pf_state_key *pf_state_key_clone(const struct pf_state_key *); void pf_rule_to_actions(struct pf_krule *, struct pf_rule_actions *); int pf_normalize_mss(struct mbuf *m, int off, struct pf_pdesc *pd); #if defined(INET) || defined(INET6) void pf_scrub(struct mbuf *, struct pf_pdesc *); #endif struct pfi_kkif *pf_kkif_create(int); void pf_kkif_free(struct pfi_kkif *); void pf_kkif_zero(struct pfi_kkif *); #endif /* _KERNEL */ #endif /* _NET_PFVAR_H_ */ diff --git a/sys/netpfil/pf/pf.c b/sys/netpfil/pf/pf.c index 19e0014ce4eb..ccfe1a0fcd96 100644 --- a/sys/netpfil/pf/pf.c +++ b/sys/netpfil/pf/pf.c @@ -1,9450 +1,9358 @@ /*- * SPDX-License-Identifier: BSD-2-Clause * * Copyright (c) 2001 Daniel Hartmeier * Copyright (c) 2002 - 2008 Henning Brauer * Copyright (c) 2012 Gleb Smirnoff * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * - Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * - Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials provided * with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. * * Effort sponsored in part by the Defense Advanced Research Projects * Agency (DARPA) and Air Force Research Laboratory, Air Force * Materiel Command, USAF, under agreement number F30602-01-2-0537. * * $OpenBSD: pf.c,v 1.634 2009/02/27 12:37:45 henning Exp $ */ #include #include "opt_bpf.h" #include "opt_inet.h" #include "opt_inet6.h" #include "opt_pf.h" #include "opt_sctp.h" #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* dummynet */ #include #include #include #include #include #ifdef INET6 #include #include #include #include #include #include #include #endif /* INET6 */ #include #include #include #include #define DPFPRINTF(n, x) if (V_pf_status.debug >= (n)) printf x SDT_PROVIDER_DEFINE(pf); SDT_PROBE_DEFINE4(pf, ip, test, done, "int", "int", "struct pf_krule *", "struct pf_kstate *"); SDT_PROBE_DEFINE5(pf, ip, state, lookup, "struct pfi_kkif *", "struct pf_state_key_cmp *", "int", "struct pf_pdesc *", "struct pf_kstate *"); SDT_PROBE_DEFINE2(pf, ip, , bound_iface, "struct pf_kstate *", "struct pfi_kkif *"); SDT_PROBE_DEFINE4(pf, sctp, multihome, test, "struct pfi_kkif *", "struct pf_krule *", "struct mbuf *", "int"); SDT_PROBE_DEFINE2(pf, sctp, multihome, add, "uint32_t", "struct pf_sctp_source *"); SDT_PROBE_DEFINE3(pf, sctp, multihome, remove, "uint32_t", "struct pf_kstate *", "struct pf_sctp_source *"); SDT_PROBE_DEFINE3(pf, eth, test_rule, entry, "int", "struct ifnet *", "struct mbuf *"); SDT_PROBE_DEFINE2(pf, eth, test_rule, test, "int", "struct pf_keth_rule *"); SDT_PROBE_DEFINE3(pf, eth, test_rule, mismatch, "int", "struct pf_keth_rule *", "char *"); SDT_PROBE_DEFINE2(pf, eth, test_rule, match, "int", "struct pf_keth_rule *"); SDT_PROBE_DEFINE2(pf, eth, test_rule, final_match, "int", "struct pf_keth_rule *"); SDT_PROBE_DEFINE2(pf, purge, state, rowcount, "int", "size_t"); /* * Global variables */ /* state tables */ VNET_DEFINE(struct pf_altqqueue, pf_altqs[4]); VNET_DEFINE(struct pf_kpalist, pf_pabuf); VNET_DEFINE(struct pf_altqqueue *, pf_altqs_active); VNET_DEFINE(struct pf_altqqueue *, pf_altq_ifs_active); VNET_DEFINE(struct pf_altqqueue *, pf_altqs_inactive); VNET_DEFINE(struct pf_altqqueue *, pf_altq_ifs_inactive); VNET_DEFINE(struct pf_kstatus, pf_status); VNET_DEFINE(u_int32_t, ticket_altqs_active); VNET_DEFINE(u_int32_t, ticket_altqs_inactive); VNET_DEFINE(int, altqs_inactive_open); VNET_DEFINE(u_int32_t, ticket_pabuf); VNET_DEFINE(MD5_CTX, pf_tcp_secret_ctx); #define V_pf_tcp_secret_ctx VNET(pf_tcp_secret_ctx) VNET_DEFINE(u_char, pf_tcp_secret[16]); #define V_pf_tcp_secret VNET(pf_tcp_secret) VNET_DEFINE(int, pf_tcp_secret_init); #define V_pf_tcp_secret_init VNET(pf_tcp_secret_init) VNET_DEFINE(int, pf_tcp_iss_off); #define V_pf_tcp_iss_off VNET(pf_tcp_iss_off) VNET_DECLARE(int, pf_vnet_active); #define V_pf_vnet_active VNET(pf_vnet_active) VNET_DEFINE_STATIC(uint32_t, pf_purge_idx); #define V_pf_purge_idx VNET(pf_purge_idx) #ifdef PF_WANT_32_TO_64_COUNTER VNET_DEFINE_STATIC(uint32_t, pf_counter_periodic_iter); #define V_pf_counter_periodic_iter VNET(pf_counter_periodic_iter) VNET_DEFINE(struct allrulelist_head, pf_allrulelist); VNET_DEFINE(size_t, pf_allrulecount); VNET_DEFINE(struct pf_krule *, pf_rulemarker); #endif struct pf_sctp_endpoint; RB_HEAD(pf_sctp_endpoints, pf_sctp_endpoint); struct pf_sctp_source { sa_family_t af; struct pf_addr addr; TAILQ_ENTRY(pf_sctp_source) entry; }; TAILQ_HEAD(pf_sctp_sources, pf_sctp_source); struct pf_sctp_endpoint { uint32_t v_tag; struct pf_sctp_sources sources; RB_ENTRY(pf_sctp_endpoint) entry; }; static int pf_sctp_endpoint_compare(struct pf_sctp_endpoint *a, struct pf_sctp_endpoint *b) { return (a->v_tag - b->v_tag); } RB_PROTOTYPE(pf_sctp_endpoints, pf_sctp_endpoint, entry, pf_sctp_endpoint_compare); RB_GENERATE(pf_sctp_endpoints, pf_sctp_endpoint, entry, pf_sctp_endpoint_compare); VNET_DEFINE_STATIC(struct pf_sctp_endpoints, pf_sctp_endpoints); #define V_pf_sctp_endpoints VNET(pf_sctp_endpoints) static struct mtx_padalign pf_sctp_endpoints_mtx; MTX_SYSINIT(pf_sctp_endpoints_mtx, &pf_sctp_endpoints_mtx, "SCTP endpoints", MTX_DEF); #define PF_SCTP_ENDPOINTS_LOCK() mtx_lock(&pf_sctp_endpoints_mtx) #define PF_SCTP_ENDPOINTS_UNLOCK() mtx_unlock(&pf_sctp_endpoints_mtx) /* * Queue for pf_intr() sends. */ static MALLOC_DEFINE(M_PFTEMP, "pf_temp", "pf(4) temporary allocations"); struct pf_send_entry { STAILQ_ENTRY(pf_send_entry) pfse_next; struct mbuf *pfse_m; enum { PFSE_IP, PFSE_IP6, PFSE_ICMP, PFSE_ICMP6, } pfse_type; struct { int type; int code; int mtu; } icmpopts; }; STAILQ_HEAD(pf_send_head, pf_send_entry); VNET_DEFINE_STATIC(struct pf_send_head, pf_sendqueue); #define V_pf_sendqueue VNET(pf_sendqueue) static struct mtx_padalign pf_sendqueue_mtx; MTX_SYSINIT(pf_sendqueue_mtx, &pf_sendqueue_mtx, "pf send queue", MTX_DEF); #define PF_SENDQ_LOCK() mtx_lock(&pf_sendqueue_mtx) #define PF_SENDQ_UNLOCK() mtx_unlock(&pf_sendqueue_mtx) /* * Queue for pf_overload_task() tasks. */ struct pf_overload_entry { SLIST_ENTRY(pf_overload_entry) next; struct pf_addr addr; sa_family_t af; uint8_t dir; struct pf_krule *rule; }; SLIST_HEAD(pf_overload_head, pf_overload_entry); VNET_DEFINE_STATIC(struct pf_overload_head, pf_overloadqueue); #define V_pf_overloadqueue VNET(pf_overloadqueue) VNET_DEFINE_STATIC(struct task, pf_overloadtask); #define V_pf_overloadtask VNET(pf_overloadtask) static struct mtx_padalign pf_overloadqueue_mtx; MTX_SYSINIT(pf_overloadqueue_mtx, &pf_overloadqueue_mtx, "pf overload/flush queue", MTX_DEF); #define PF_OVERLOADQ_LOCK() mtx_lock(&pf_overloadqueue_mtx) #define PF_OVERLOADQ_UNLOCK() mtx_unlock(&pf_overloadqueue_mtx) VNET_DEFINE(struct pf_krulequeue, pf_unlinked_rules); struct mtx_padalign pf_unlnkdrules_mtx; MTX_SYSINIT(pf_unlnkdrules_mtx, &pf_unlnkdrules_mtx, "pf unlinked rules", MTX_DEF); struct sx pf_config_lock; SX_SYSINIT(pf_config_lock, &pf_config_lock, "pf config"); struct mtx_padalign pf_table_stats_lock; MTX_SYSINIT(pf_table_stats_lock, &pf_table_stats_lock, "pf table stats", MTX_DEF); VNET_DEFINE_STATIC(uma_zone_t, pf_sources_z); #define V_pf_sources_z VNET(pf_sources_z) uma_zone_t pf_mtag_z; VNET_DEFINE(uma_zone_t, pf_state_z); VNET_DEFINE(uma_zone_t, pf_state_key_z); VNET_DEFINE(uma_zone_t, pf_udp_mapping_z); VNET_DEFINE(struct unrhdr64, pf_stateid); static void pf_src_tree_remove_state(struct pf_kstate *); static void pf_init_threshold(struct pf_threshold *, u_int32_t, u_int32_t); static void pf_add_threshold(struct pf_threshold *); static int pf_check_threshold(struct pf_threshold *); static void pf_change_ap(struct mbuf *, struct pf_addr *, u_int16_t *, u_int16_t *, u_int16_t *, struct pf_addr *, u_int16_t, u_int8_t, sa_family_t); static int pf_modulate_sack(struct mbuf *, int, struct pf_pdesc *, struct tcphdr *, struct pf_state_peer *); int pf_icmp_mapping(struct pf_pdesc *, u_int8_t, int *, int *, u_int16_t *, u_int16_t *); static void pf_change_icmp(struct pf_addr *, u_int16_t *, struct pf_addr *, struct pf_addr *, u_int16_t, u_int16_t *, u_int16_t *, u_int16_t *, u_int16_t *, u_int8_t, sa_family_t); static void pf_send_icmp(struct mbuf *, u_int8_t, u_int8_t, sa_family_t, struct pf_krule *, int); static void pf_detach_state(struct pf_kstate *); static int pf_state_key_attach(struct pf_state_key *, struct pf_state_key *, struct pf_kstate *); static void pf_state_key_detach(struct pf_kstate *, int); static int pf_state_key_ctor(void *, int, void *, int); static u_int32_t pf_tcp_iss(struct pf_pdesc *); static __inline void pf_dummynet_flag_remove(struct mbuf *m, struct pf_mtag *pf_mtag); static int pf_dummynet(struct pf_pdesc *, struct pf_kstate *, struct pf_krule *, struct mbuf **); static int pf_dummynet_route(struct pf_pdesc *, struct pf_kstate *, struct pf_krule *, struct ifnet *, struct sockaddr *, struct mbuf **); static int pf_test_eth_rule(int, struct pfi_kkif *, struct mbuf **); static int pf_test_rule(struct pf_krule **, struct pf_kstate **, struct pfi_kkif *, struct mbuf *, int, struct pf_pdesc *, struct pf_krule **, struct pf_kruleset **, struct inpcb *, int); static int pf_create_state(struct pf_krule *, struct pf_krule *, struct pf_krule *, struct pf_pdesc *, struct pf_ksrc_node *, struct pf_state_key *, struct pf_state_key *, struct mbuf *, int, u_int16_t, u_int16_t, int *, struct pfi_kkif *, struct pf_kstate **, int, u_int16_t, u_int16_t, int, struct pf_krule_slist *, struct pf_udp_mapping *); static int pf_state_key_addr_setup(struct pf_pdesc *, struct mbuf *, int, struct pf_state_key_cmp *, int, struct pf_addr *, int, struct pf_addr *, int); -static int pf_test_fragment(struct pf_krule **, struct pfi_kkif *, - struct mbuf *, struct pf_pdesc *, - struct pf_krule **, struct pf_kruleset **); static int pf_tcp_track_full(struct pf_kstate **, struct pfi_kkif *, struct mbuf *, int, struct pf_pdesc *, u_short *, int *); static int pf_tcp_track_sloppy(struct pf_kstate **, struct pf_pdesc *, u_short *); static int pf_test_state_tcp(struct pf_kstate **, struct pfi_kkif *, struct mbuf *, int, struct pf_pdesc *, u_short *); static int pf_test_state_udp(struct pf_kstate **, struct pfi_kkif *, struct mbuf *, int, struct pf_pdesc *); int pf_icmp_state_lookup(struct pf_state_key_cmp *, struct pf_pdesc *, struct pf_kstate **, struct mbuf *, int, int, struct pfi_kkif *, u_int16_t, u_int16_t, int, int *, int, int); static int pf_test_state_icmp(struct pf_kstate **, struct pfi_kkif *, struct mbuf *, int, struct pf_pdesc *, u_short *); static void pf_sctp_multihome_detach_addr(const struct pf_kstate *); static void pf_sctp_multihome_delayed(struct pf_pdesc *, int, struct pfi_kkif *, struct pf_kstate *, int); static int pf_test_state_sctp(struct pf_kstate **, struct pfi_kkif *, struct mbuf *, int, struct pf_pdesc *, u_short *); static int pf_test_state_other(struct pf_kstate **, struct pfi_kkif *, struct mbuf *, struct pf_pdesc *); static u_int16_t pf_calc_mss(struct pf_addr *, sa_family_t, int, u_int16_t); static int pf_check_proto_cksum(struct mbuf *, int, int, u_int8_t, sa_family_t); static void pf_print_state_parts(struct pf_kstate *, struct pf_state_key *, struct pf_state_key *); static void pf_patch_8(struct mbuf *, u_int16_t *, u_int8_t *, u_int8_t, bool, u_int8_t); static struct pf_kstate *pf_find_state(struct pfi_kkif *, const struct pf_state_key_cmp *, u_int); static int pf_src_connlimit(struct pf_kstate **); static int pf_match_rcvif(struct mbuf *, struct pf_krule *); static void pf_counters_inc(int, struct pf_pdesc *, struct pfi_kkif *, struct pf_kstate *, struct pf_krule *, struct pf_krule *); static void pf_overload_task(void *v, int pending); static u_short pf_insert_src_node(struct pf_ksrc_node **, struct pf_krule *, struct pf_addr *, sa_family_t); static u_int pf_purge_expired_states(u_int, int); static void pf_purge_unlinked_rules(void); static int pf_mtag_uminit(void *, int, int); static void pf_mtag_free(struct m_tag *); static void pf_packet_rework_nat(struct mbuf *, struct pf_pdesc *, int, struct pf_state_key *); #ifdef INET static void pf_route(struct mbuf **, struct pf_krule *, struct ifnet *, struct pf_kstate *, struct pf_pdesc *, struct inpcb *); #endif /* INET */ #ifdef INET6 static void pf_change_a6(struct pf_addr *, u_int16_t *, struct pf_addr *, u_int8_t); static void pf_route6(struct mbuf **, struct pf_krule *, struct ifnet *, struct pf_kstate *, struct pf_pdesc *, struct inpcb *); #endif /* INET6 */ static __inline void pf_set_protostate(struct pf_kstate *, int, u_int8_t); int in4_cksum(struct mbuf *m, u_int8_t nxt, int off, int len); extern int pf_end_threads; extern struct proc *pf_purge_proc; VNET_DEFINE(struct pf_limit, pf_limits[PF_LIMIT_MAX]); enum { PF_ICMP_MULTI_NONE, PF_ICMP_MULTI_LINK }; #define PACKET_UNDO_NAT(_m, _pd, _off, _s) \ do { \ struct pf_state_key *nk; \ if ((pd->dir) == PF_OUT) \ nk = (_s)->key[PF_SK_STACK]; \ else \ nk = (_s)->key[PF_SK_WIRE]; \ pf_packet_rework_nat(_m, _pd, _off, nk); \ } while (0) #define PACKET_LOOPED(pd) ((pd)->pf_mtag && \ (pd)->pf_mtag->flags & PF_MTAG_FLAG_PACKET_LOOPED) #define STATE_LOOKUP(i, k, s, pd) \ do { \ (s) = pf_find_state((i), (k), (pd->dir)); \ SDT_PROBE5(pf, ip, state, lookup, i, k, (pd->dir), pd, (s)); \ if ((s) == NULL) \ return (PF_DROP); \ if (PACKET_LOOPED(pd)) \ return (PF_PASS); \ } while (0) static struct pfi_kkif * BOUND_IFACE(struct pf_kstate *st, struct pfi_kkif *k) { SDT_PROBE2(pf, ip, , bound_iface, st, k); /* Floating unless otherwise specified. */ if (! (st->rule.ptr->rule_flag & PFRULE_IFBOUND)) return (V_pfi_all); /* * Initially set to all, because we don't know what interface we'll be * sending this out when we create the state. */ if (st->rule.ptr->rt == PF_REPLYTO) return (V_pfi_all); /* Don't overrule the interface for states created on incoming packets. */ if (st->direction == PF_IN) return (k); /* No route-to, so don't overrule. */ if (st->rt != PF_ROUTETO) return (k); /* Bind to the route-to interface. */ return (st->rt_kif); } #define STATE_INC_COUNTERS(s) \ do { \ struct pf_krule_item *mrm; \ counter_u64_add(s->rule.ptr->states_cur, 1); \ counter_u64_add(s->rule.ptr->states_tot, 1); \ if (s->anchor.ptr != NULL) { \ counter_u64_add(s->anchor.ptr->states_cur, 1); \ counter_u64_add(s->anchor.ptr->states_tot, 1); \ } \ if (s->nat_rule.ptr != NULL) { \ counter_u64_add(s->nat_rule.ptr->states_cur, 1);\ counter_u64_add(s->nat_rule.ptr->states_tot, 1);\ } \ SLIST_FOREACH(mrm, &s->match_rules, entry) { \ counter_u64_add(mrm->r->states_cur, 1); \ counter_u64_add(mrm->r->states_tot, 1); \ } \ } while (0) #define STATE_DEC_COUNTERS(s) \ do { \ struct pf_krule_item *mrm; \ if (s->nat_rule.ptr != NULL) \ counter_u64_add(s->nat_rule.ptr->states_cur, -1);\ if (s->anchor.ptr != NULL) \ counter_u64_add(s->anchor.ptr->states_cur, -1); \ counter_u64_add(s->rule.ptr->states_cur, -1); \ SLIST_FOREACH(mrm, &s->match_rules, entry) \ counter_u64_add(mrm->r->states_cur, -1); \ } while (0) MALLOC_DEFINE(M_PFHASH, "pf_hash", "pf(4) hash header structures"); MALLOC_DEFINE(M_PF_RULE_ITEM, "pf_krule_item", "pf(4) rule items"); VNET_DEFINE(struct pf_keyhash *, pf_keyhash); VNET_DEFINE(struct pf_idhash *, pf_idhash); VNET_DEFINE(struct pf_srchash *, pf_srchash); VNET_DEFINE(struct pf_udpendpointhash *, pf_udpendpointhash); VNET_DEFINE(struct pf_udpendpointmapping *, pf_udpendpointmapping); SYSCTL_NODE(_net, OID_AUTO, pf, CTLFLAG_RW | CTLFLAG_MPSAFE, 0, "pf(4)"); VNET_DEFINE(u_long, pf_hashmask); VNET_DEFINE(u_long, pf_srchashmask); VNET_DEFINE(u_long, pf_udpendpointhashmask); VNET_DEFINE_STATIC(u_long, pf_hashsize); #define V_pf_hashsize VNET(pf_hashsize) VNET_DEFINE_STATIC(u_long, pf_srchashsize); #define V_pf_srchashsize VNET(pf_srchashsize) VNET_DEFINE_STATIC(u_long, pf_udpendpointhashsize); #define V_pf_udpendpointhashsize VNET(pf_udpendpointhashsize) u_long pf_ioctl_maxcount = 65535; SYSCTL_ULONG(_net_pf, OID_AUTO, states_hashsize, CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(pf_hashsize), 0, "Size of pf(4) states hashtable"); SYSCTL_ULONG(_net_pf, OID_AUTO, source_nodes_hashsize, CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(pf_srchashsize), 0, "Size of pf(4) source nodes hashtable"); SYSCTL_ULONG(_net_pf, OID_AUTO, udpendpoint_hashsize, CTLFLAG_VNET | CTLFLAG_RDTUN, &VNET_NAME(pf_udpendpointhashsize), 0, "Size of pf(4) endpoint hashtable"); SYSCTL_ULONG(_net_pf, OID_AUTO, request_maxcount, CTLFLAG_RWTUN, &pf_ioctl_maxcount, 0, "Maximum number of tables, addresses, ... in a single ioctl() call"); VNET_DEFINE(void *, pf_swi_cookie); VNET_DEFINE(struct intr_event *, pf_swi_ie); VNET_DEFINE(uint32_t, pf_hashseed); #define V_pf_hashseed VNET(pf_hashseed) static void pf_sctp_checksum(struct mbuf *m, int off) { uint32_t sum = 0; /* Zero out the checksum, to enable recalculation. */ m_copyback(m, off + offsetof(struct sctphdr, checksum), sizeof(sum), (caddr_t)&sum); sum = sctp_calculate_cksum(m, off); m_copyback(m, off + offsetof(struct sctphdr, checksum), sizeof(sum), (caddr_t)&sum); } int pf_addr_cmp(struct pf_addr *a, struct pf_addr *b, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: if (a->addr32[0] > b->addr32[0]) return (1); if (a->addr32[0] < b->addr32[0]) return (-1); break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (a->addr32[3] > b->addr32[3]) return (1); if (a->addr32[3] < b->addr32[3]) return (-1); if (a->addr32[2] > b->addr32[2]) return (1); if (a->addr32[2] < b->addr32[2]) return (-1); if (a->addr32[1] > b->addr32[1]) return (1); if (a->addr32[1] < b->addr32[1]) return (-1); if (a->addr32[0] > b->addr32[0]) return (1); if (a->addr32[0] < b->addr32[0]) return (-1); break; #endif /* INET6 */ default: panic("%s: unknown address family %u", __func__, af); } return (0); } static bool pf_is_loopback(sa_family_t af, struct pf_addr *addr) { switch (af) { case AF_INET: return IN_LOOPBACK(ntohl(addr->v4.s_addr)); case AF_INET6: return IN6_IS_ADDR_LOOPBACK(&addr->v6); default: panic("Unknown af %d", af); } } static void pf_packet_rework_nat(struct mbuf *m, struct pf_pdesc *pd, int off, struct pf_state_key *nk) { switch (pd->proto) { case IPPROTO_TCP: { struct tcphdr *th = &pd->hdr.tcp; if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af)) pf_change_ap(m, pd->src, &th->th_sport, pd->ip_sum, &th->th_sum, &nk->addr[pd->sidx], nk->port[pd->sidx], 0, pd->af); if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af)) pf_change_ap(m, pd->dst, &th->th_dport, pd->ip_sum, &th->th_sum, &nk->addr[pd->didx], nk->port[pd->didx], 0, pd->af); m_copyback(m, off, sizeof(*th), (caddr_t)th); break; } case IPPROTO_UDP: { struct udphdr *uh = &pd->hdr.udp; if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af)) pf_change_ap(m, pd->src, &uh->uh_sport, pd->ip_sum, &uh->uh_sum, &nk->addr[pd->sidx], nk->port[pd->sidx], 1, pd->af); if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af)) pf_change_ap(m, pd->dst, &uh->uh_dport, pd->ip_sum, &uh->uh_sum, &nk->addr[pd->didx], nk->port[pd->didx], 1, pd->af); m_copyback(m, off, sizeof(*uh), (caddr_t)uh); break; } case IPPROTO_SCTP: { struct sctphdr *sh = &pd->hdr.sctp; uint16_t checksum = 0; if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af)) { pf_change_ap(m, pd->src, &sh->src_port, pd->ip_sum, &checksum, &nk->addr[pd->sidx], nk->port[pd->sidx], 1, pd->af); } if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af)) { pf_change_ap(m, pd->dst, &sh->dest_port, pd->ip_sum, &checksum, &nk->addr[pd->didx], nk->port[pd->didx], 1, pd->af); } break; } case IPPROTO_ICMP: { struct icmp *ih = &pd->hdr.icmp; if (nk->port[pd->sidx] != ih->icmp_id) { pd->hdr.icmp.icmp_cksum = pf_cksum_fixup( ih->icmp_cksum, ih->icmp_id, nk->port[pd->sidx], 0); ih->icmp_id = nk->port[pd->sidx]; pd->sport = &ih->icmp_id; m_copyback(m, off, ICMP_MINLEN, (caddr_t)ih); } /* FALLTHROUGH */ } default: if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af)) { switch (pd->af) { case AF_INET: pf_change_a(&pd->src->v4.s_addr, pd->ip_sum, nk->addr[pd->sidx].v4.s_addr, 0); break; case AF_INET6: PF_ACPY(pd->src, &nk->addr[pd->sidx], pd->af); break; } } if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af)) { switch (pd->af) { case AF_INET: pf_change_a(&pd->dst->v4.s_addr, pd->ip_sum, nk->addr[pd->didx].v4.s_addr, 0); break; case AF_INET6: PF_ACPY(pd->dst, &nk->addr[pd->didx], pd->af); break; } } break; } } static __inline uint32_t pf_hashkey(const struct pf_state_key *sk) { uint32_t h; h = murmur3_32_hash32((const uint32_t *)sk, sizeof(struct pf_state_key_cmp)/sizeof(uint32_t), V_pf_hashseed); return (h & V_pf_hashmask); } static __inline uint32_t pf_hashsrc(struct pf_addr *addr, sa_family_t af) { uint32_t h; switch (af) { case AF_INET: h = murmur3_32_hash32((uint32_t *)&addr->v4, sizeof(addr->v4)/sizeof(uint32_t), V_pf_hashseed); break; case AF_INET6: h = murmur3_32_hash32((uint32_t *)&addr->v6, sizeof(addr->v6)/sizeof(uint32_t), V_pf_hashseed); break; default: panic("%s: unknown address family %u", __func__, af); } return (h & V_pf_srchashmask); } static inline uint32_t pf_hashudpendpoint(struct pf_udp_endpoint *endpoint) { uint32_t h; h = murmur3_32_hash32((uint32_t *)endpoint, sizeof(struct pf_udp_endpoint_cmp)/sizeof(uint32_t), V_pf_hashseed); return (h & V_pf_udpendpointhashmask); } #ifdef ALTQ static int pf_state_hash(struct pf_kstate *s) { u_int32_t hv = (intptr_t)s / sizeof(*s); hv ^= crc32(&s->src, sizeof(s->src)); hv ^= crc32(&s->dst, sizeof(s->dst)); if (hv == 0) hv = 1; return (hv); } #endif static __inline void pf_set_protostate(struct pf_kstate *s, int which, u_int8_t newstate) { if (which == PF_PEER_DST || which == PF_PEER_BOTH) s->dst.state = newstate; if (which == PF_PEER_DST) return; if (s->src.state == newstate) return; if (s->creatorid == V_pf_status.hostid && s->key[PF_SK_STACK] != NULL && s->key[PF_SK_STACK]->proto == IPPROTO_TCP && !(TCPS_HAVEESTABLISHED(s->src.state) || s->src.state == TCPS_CLOSED) && (TCPS_HAVEESTABLISHED(newstate) || newstate == TCPS_CLOSED)) atomic_add_32(&V_pf_status.states_halfopen, -1); s->src.state = newstate; } #ifdef INET6 void pf_addrcpy(struct pf_addr *dst, struct pf_addr *src, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: memcpy(&dst->v4, &src->v4, sizeof(dst->v4)); break; #endif /* INET */ case AF_INET6: memcpy(&dst->v6, &src->v6, sizeof(dst->v6)); break; } } #endif /* INET6 */ static void pf_init_threshold(struct pf_threshold *threshold, u_int32_t limit, u_int32_t seconds) { threshold->limit = limit * PF_THRESHOLD_MULT; threshold->seconds = seconds; threshold->count = 0; threshold->last = time_uptime; } static void pf_add_threshold(struct pf_threshold *threshold) { u_int32_t t = time_uptime, diff = t - threshold->last; if (diff >= threshold->seconds) threshold->count = 0; else threshold->count -= threshold->count * diff / threshold->seconds; threshold->count += PF_THRESHOLD_MULT; threshold->last = t; } static int pf_check_threshold(struct pf_threshold *threshold) { return (threshold->count > threshold->limit); } static int pf_src_connlimit(struct pf_kstate **state) { struct pf_overload_entry *pfoe; int bad = 0; PF_STATE_LOCK_ASSERT(*state); /* * XXXKS: The src node is accessed unlocked! * PF_SRC_NODE_LOCK_ASSERT((*state)->src_node); */ (*state)->src_node->conn++; (*state)->src.tcp_est = 1; pf_add_threshold(&(*state)->src_node->conn_rate); if ((*state)->rule.ptr->max_src_conn && (*state)->rule.ptr->max_src_conn < (*state)->src_node->conn) { counter_u64_add(V_pf_status.lcounters[LCNT_SRCCONN], 1); bad++; } if ((*state)->rule.ptr->max_src_conn_rate.limit && pf_check_threshold(&(*state)->src_node->conn_rate)) { counter_u64_add(V_pf_status.lcounters[LCNT_SRCCONNRATE], 1); bad++; } if (!bad) return (0); /* Kill this state. */ (*state)->timeout = PFTM_PURGE; pf_set_protostate(*state, PF_PEER_BOTH, TCPS_CLOSED); if ((*state)->rule.ptr->overload_tbl == NULL) return (1); /* Schedule overloading and flushing task. */ pfoe = malloc(sizeof(*pfoe), M_PFTEMP, M_NOWAIT); if (pfoe == NULL) return (1); /* too bad :( */ bcopy(&(*state)->src_node->addr, &pfoe->addr, sizeof(pfoe->addr)); pfoe->af = (*state)->key[PF_SK_WIRE]->af; pfoe->rule = (*state)->rule.ptr; pfoe->dir = (*state)->direction; PF_OVERLOADQ_LOCK(); SLIST_INSERT_HEAD(&V_pf_overloadqueue, pfoe, next); PF_OVERLOADQ_UNLOCK(); taskqueue_enqueue(taskqueue_swi, &V_pf_overloadtask); return (1); } static void pf_overload_task(void *v, int pending) { struct pf_overload_head queue; struct pfr_addr p; struct pf_overload_entry *pfoe, *pfoe1; uint32_t killed = 0; CURVNET_SET((struct vnet *)v); PF_OVERLOADQ_LOCK(); queue = V_pf_overloadqueue; SLIST_INIT(&V_pf_overloadqueue); PF_OVERLOADQ_UNLOCK(); bzero(&p, sizeof(p)); SLIST_FOREACH(pfoe, &queue, next) { counter_u64_add(V_pf_status.lcounters[LCNT_OVERLOAD_TABLE], 1); if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("%s: blocking address ", __func__); pf_print_host(&pfoe->addr, 0, pfoe->af); printf("\n"); } p.pfra_af = pfoe->af; switch (pfoe->af) { #ifdef INET case AF_INET: p.pfra_net = 32; p.pfra_ip4addr = pfoe->addr.v4; break; #endif #ifdef INET6 case AF_INET6: p.pfra_net = 128; p.pfra_ip6addr = pfoe->addr.v6; break; #endif } PF_RULES_WLOCK(); pfr_insert_kentry(pfoe->rule->overload_tbl, &p, time_second); PF_RULES_WUNLOCK(); } /* * Remove those entries, that don't need flushing. */ SLIST_FOREACH_SAFE(pfoe, &queue, next, pfoe1) if (pfoe->rule->flush == 0) { SLIST_REMOVE(&queue, pfoe, pf_overload_entry, next); free(pfoe, M_PFTEMP); } else counter_u64_add( V_pf_status.lcounters[LCNT_OVERLOAD_FLUSH], 1); /* If nothing to flush, return. */ if (SLIST_EMPTY(&queue)) { CURVNET_RESTORE(); return; } for (int i = 0; i <= V_pf_hashmask; i++) { struct pf_idhash *ih = &V_pf_idhash[i]; struct pf_state_key *sk; struct pf_kstate *s; PF_HASHROW_LOCK(ih); LIST_FOREACH(s, &ih->states, entry) { sk = s->key[PF_SK_WIRE]; SLIST_FOREACH(pfoe, &queue, next) if (sk->af == pfoe->af && ((pfoe->rule->flush & PF_FLUSH_GLOBAL) || pfoe->rule == s->rule.ptr) && ((pfoe->dir == PF_OUT && PF_AEQ(&pfoe->addr, &sk->addr[1], sk->af)) || (pfoe->dir == PF_IN && PF_AEQ(&pfoe->addr, &sk->addr[0], sk->af)))) { s->timeout = PFTM_PURGE; pf_set_protostate(s, PF_PEER_BOTH, TCPS_CLOSED); killed++; } } PF_HASHROW_UNLOCK(ih); } SLIST_FOREACH_SAFE(pfoe, &queue, next, pfoe1) free(pfoe, M_PFTEMP); if (V_pf_status.debug >= PF_DEBUG_MISC) printf("%s: %u states killed", __func__, killed); CURVNET_RESTORE(); } /* * Can return locked on failure, so that we can consistently * allocate and insert a new one. */ struct pf_ksrc_node * pf_find_src_node(struct pf_addr *src, struct pf_krule *rule, sa_family_t af, struct pf_srchash **sh, bool returnlocked) { struct pf_ksrc_node *n; counter_u64_add(V_pf_status.scounters[SCNT_SRC_NODE_SEARCH], 1); *sh = &V_pf_srchash[pf_hashsrc(src, af)]; PF_HASHROW_LOCK(*sh); LIST_FOREACH(n, &(*sh)->nodes, entry) if (n->rule.ptr == rule && n->af == af && ((af == AF_INET && n->addr.v4.s_addr == src->v4.s_addr) || (af == AF_INET6 && bcmp(&n->addr, src, sizeof(*src)) == 0))) break; if (n != NULL) { n->states++; PF_HASHROW_UNLOCK(*sh); } else if (returnlocked == false) PF_HASHROW_UNLOCK(*sh); return (n); } static void pf_free_src_node(struct pf_ksrc_node *sn) { for (int i = 0; i < 2; i++) { counter_u64_free(sn->bytes[i]); counter_u64_free(sn->packets[i]); } uma_zfree(V_pf_sources_z, sn); } static u_short pf_insert_src_node(struct pf_ksrc_node **sn, struct pf_krule *rule, struct pf_addr *src, sa_family_t af) { u_short reason = 0; struct pf_srchash *sh = NULL; KASSERT((rule->rule_flag & PFRULE_SRCTRACK || rule->rpool.opts & PF_POOL_STICKYADDR), ("%s for non-tracking rule %p", __func__, rule)); if (*sn == NULL) *sn = pf_find_src_node(src, rule, af, &sh, true); if (*sn == NULL) { PF_HASHROW_ASSERT(sh); if (rule->max_src_nodes && counter_u64_fetch(rule->src_nodes) >= rule->max_src_nodes) { counter_u64_add(V_pf_status.lcounters[LCNT_SRCNODES], 1); PF_HASHROW_UNLOCK(sh); reason = PFRES_SRCLIMIT; goto done; } (*sn) = uma_zalloc(V_pf_sources_z, M_NOWAIT | M_ZERO); if ((*sn) == NULL) { PF_HASHROW_UNLOCK(sh); reason = PFRES_MEMORY; goto done; } for (int i = 0; i < 2; i++) { (*sn)->bytes[i] = counter_u64_alloc(M_NOWAIT); (*sn)->packets[i] = counter_u64_alloc(M_NOWAIT); if ((*sn)->bytes[i] == NULL || (*sn)->packets[i] == NULL) { pf_free_src_node(*sn); PF_HASHROW_UNLOCK(sh); reason = PFRES_MEMORY; goto done; } } pf_init_threshold(&(*sn)->conn_rate, rule->max_src_conn_rate.limit, rule->max_src_conn_rate.seconds); MPASS((*sn)->lock == NULL); (*sn)->lock = &sh->lock; (*sn)->af = af; (*sn)->rule.ptr = rule; PF_ACPY(&(*sn)->addr, src, af); LIST_INSERT_HEAD(&sh->nodes, *sn, entry); (*sn)->creation = time_uptime; (*sn)->ruletype = rule->action; (*sn)->states = 1; if ((*sn)->rule.ptr != NULL) counter_u64_add((*sn)->rule.ptr->src_nodes, 1); PF_HASHROW_UNLOCK(sh); counter_u64_add(V_pf_status.scounters[SCNT_SRC_NODE_INSERT], 1); } else { if (rule->max_src_states && (*sn)->states >= rule->max_src_states) { counter_u64_add(V_pf_status.lcounters[LCNT_SRCSTATES], 1); reason = PFRES_SRCLIMIT; goto done; } } done: return (reason); } void pf_unlink_src_node(struct pf_ksrc_node *src) { PF_SRC_NODE_LOCK_ASSERT(src); LIST_REMOVE(src, entry); if (src->rule.ptr) counter_u64_add(src->rule.ptr->src_nodes, -1); } u_int pf_free_src_nodes(struct pf_ksrc_node_list *head) { struct pf_ksrc_node *sn, *tmp; u_int count = 0; LIST_FOREACH_SAFE(sn, head, entry, tmp) { pf_free_src_node(sn); count++; } counter_u64_add(V_pf_status.scounters[SCNT_SRC_NODE_REMOVALS], count); return (count); } void pf_mtag_initialize(void) { pf_mtag_z = uma_zcreate("pf mtags", sizeof(struct m_tag) + sizeof(struct pf_mtag), NULL, NULL, pf_mtag_uminit, NULL, UMA_ALIGN_PTR, 0); } /* Per-vnet data storage structures initialization. */ void pf_initialize(void) { struct pf_keyhash *kh; struct pf_idhash *ih; struct pf_srchash *sh; struct pf_udpendpointhash *uh; u_int i; if (V_pf_hashsize == 0 || !powerof2(V_pf_hashsize)) V_pf_hashsize = PF_HASHSIZ; if (V_pf_srchashsize == 0 || !powerof2(V_pf_srchashsize)) V_pf_srchashsize = PF_SRCHASHSIZ; if (V_pf_udpendpointhashsize == 0 || !powerof2(V_pf_udpendpointhashsize)) V_pf_udpendpointhashsize = PF_UDPENDHASHSIZ; V_pf_hashseed = arc4random(); /* States and state keys storage. */ V_pf_state_z = uma_zcreate("pf states", sizeof(struct pf_kstate), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); V_pf_limits[PF_LIMIT_STATES].zone = V_pf_state_z; uma_zone_set_max(V_pf_state_z, PFSTATE_HIWAT); uma_zone_set_warning(V_pf_state_z, "PF states limit reached"); V_pf_state_key_z = uma_zcreate("pf state keys", sizeof(struct pf_state_key), pf_state_key_ctor, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); V_pf_keyhash = mallocarray(V_pf_hashsize, sizeof(struct pf_keyhash), M_PFHASH, M_NOWAIT | M_ZERO); V_pf_idhash = mallocarray(V_pf_hashsize, sizeof(struct pf_idhash), M_PFHASH, M_NOWAIT | M_ZERO); if (V_pf_keyhash == NULL || V_pf_idhash == NULL) { printf("pf: Unable to allocate memory for " "state_hashsize %lu.\n", V_pf_hashsize); free(V_pf_keyhash, M_PFHASH); free(V_pf_idhash, M_PFHASH); V_pf_hashsize = PF_HASHSIZ; V_pf_keyhash = mallocarray(V_pf_hashsize, sizeof(struct pf_keyhash), M_PFHASH, M_WAITOK | M_ZERO); V_pf_idhash = mallocarray(V_pf_hashsize, sizeof(struct pf_idhash), M_PFHASH, M_WAITOK | M_ZERO); } V_pf_hashmask = V_pf_hashsize - 1; for (i = 0, kh = V_pf_keyhash, ih = V_pf_idhash; i <= V_pf_hashmask; i++, kh++, ih++) { mtx_init(&kh->lock, "pf_keyhash", NULL, MTX_DEF | MTX_DUPOK); mtx_init(&ih->lock, "pf_idhash", NULL, MTX_DEF); } /* Source nodes. */ V_pf_sources_z = uma_zcreate("pf source nodes", sizeof(struct pf_ksrc_node), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); V_pf_limits[PF_LIMIT_SRC_NODES].zone = V_pf_sources_z; uma_zone_set_max(V_pf_sources_z, PFSNODE_HIWAT); uma_zone_set_warning(V_pf_sources_z, "PF source nodes limit reached"); V_pf_srchash = mallocarray(V_pf_srchashsize, sizeof(struct pf_srchash), M_PFHASH, M_NOWAIT | M_ZERO); if (V_pf_srchash == NULL) { printf("pf: Unable to allocate memory for " "source_hashsize %lu.\n", V_pf_srchashsize); V_pf_srchashsize = PF_SRCHASHSIZ; V_pf_srchash = mallocarray(V_pf_srchashsize, sizeof(struct pf_srchash), M_PFHASH, M_WAITOK | M_ZERO); } V_pf_srchashmask = V_pf_srchashsize - 1; for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; i++, sh++) mtx_init(&sh->lock, "pf_srchash", NULL, MTX_DEF); /* UDP endpoint mappings. */ V_pf_udp_mapping_z = uma_zcreate("pf UDP mappings", sizeof(struct pf_udp_mapping), NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0); V_pf_udpendpointhash = mallocarray(V_pf_udpendpointhashsize, sizeof(struct pf_udpendpointhash), M_PFHASH, M_NOWAIT | M_ZERO); if (V_pf_udpendpointhash == NULL) { printf("pf: Unable to allocate memory for " "udpendpoint_hashsize %lu.\n", V_pf_udpendpointhashsize); V_pf_udpendpointhashsize = PF_UDPENDHASHSIZ; V_pf_udpendpointhash = mallocarray(V_pf_udpendpointhashsize, sizeof(struct pf_udpendpointhash), M_PFHASH, M_WAITOK | M_ZERO); } V_pf_udpendpointhashmask = V_pf_udpendpointhashsize - 1; for (i = 0, uh = V_pf_udpendpointhash; i <= V_pf_udpendpointhashmask; i++, uh++) { mtx_init(&uh->lock, "pf_udpendpointhash", NULL, MTX_DEF | MTX_DUPOK); } /* ALTQ */ TAILQ_INIT(&V_pf_altqs[0]); TAILQ_INIT(&V_pf_altqs[1]); TAILQ_INIT(&V_pf_altqs[2]); TAILQ_INIT(&V_pf_altqs[3]); TAILQ_INIT(&V_pf_pabuf); V_pf_altqs_active = &V_pf_altqs[0]; V_pf_altq_ifs_active = &V_pf_altqs[1]; V_pf_altqs_inactive = &V_pf_altqs[2]; V_pf_altq_ifs_inactive = &V_pf_altqs[3]; /* Send & overload+flush queues. */ STAILQ_INIT(&V_pf_sendqueue); SLIST_INIT(&V_pf_overloadqueue); TASK_INIT(&V_pf_overloadtask, 0, pf_overload_task, curvnet); /* Unlinked, but may be referenced rules. */ TAILQ_INIT(&V_pf_unlinked_rules); } void pf_mtag_cleanup(void) { uma_zdestroy(pf_mtag_z); } void pf_cleanup(void) { struct pf_keyhash *kh; struct pf_idhash *ih; struct pf_srchash *sh; struct pf_udpendpointhash *uh; struct pf_send_entry *pfse, *next; u_int i; for (i = 0, kh = V_pf_keyhash, ih = V_pf_idhash; i <= V_pf_hashmask; i++, kh++, ih++) { KASSERT(LIST_EMPTY(&kh->keys), ("%s: key hash not empty", __func__)); KASSERT(LIST_EMPTY(&ih->states), ("%s: id hash not empty", __func__)); mtx_destroy(&kh->lock); mtx_destroy(&ih->lock); } free(V_pf_keyhash, M_PFHASH); free(V_pf_idhash, M_PFHASH); for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; i++, sh++) { KASSERT(LIST_EMPTY(&sh->nodes), ("%s: source node hash not empty", __func__)); mtx_destroy(&sh->lock); } free(V_pf_srchash, M_PFHASH); for (i = 0, uh = V_pf_udpendpointhash; i <= V_pf_udpendpointhashmask; i++, uh++) { KASSERT(LIST_EMPTY(&uh->endpoints), ("%s: udp endpoint hash not empty", __func__)); mtx_destroy(&uh->lock); } free(V_pf_udpendpointhash, M_PFHASH); STAILQ_FOREACH_SAFE(pfse, &V_pf_sendqueue, pfse_next, next) { m_freem(pfse->pfse_m); free(pfse, M_PFTEMP); } MPASS(RB_EMPTY(&V_pf_sctp_endpoints)); uma_zdestroy(V_pf_sources_z); uma_zdestroy(V_pf_state_z); uma_zdestroy(V_pf_state_key_z); uma_zdestroy(V_pf_udp_mapping_z); } static int pf_mtag_uminit(void *mem, int size, int how) { struct m_tag *t; t = (struct m_tag *)mem; t->m_tag_cookie = MTAG_ABI_COMPAT; t->m_tag_id = PACKET_TAG_PF; t->m_tag_len = sizeof(struct pf_mtag); t->m_tag_free = pf_mtag_free; return (0); } static void pf_mtag_free(struct m_tag *t) { uma_zfree(pf_mtag_z, t); } struct pf_mtag * pf_get_mtag(struct mbuf *m) { struct m_tag *mtag; if ((mtag = m_tag_find(m, PACKET_TAG_PF, NULL)) != NULL) return ((struct pf_mtag *)(mtag + 1)); mtag = uma_zalloc(pf_mtag_z, M_NOWAIT); if (mtag == NULL) return (NULL); bzero(mtag + 1, sizeof(struct pf_mtag)); m_tag_prepend(m, mtag); return ((struct pf_mtag *)(mtag + 1)); } static int pf_state_key_attach(struct pf_state_key *skw, struct pf_state_key *sks, struct pf_kstate *s) { struct pf_keyhash *khs, *khw, *kh; struct pf_state_key *sk, *cur; struct pf_kstate *si, *olds = NULL; int idx; NET_EPOCH_ASSERT(); KASSERT(s->refs == 0, ("%s: state not pristine", __func__)); KASSERT(s->key[PF_SK_WIRE] == NULL, ("%s: state has key", __func__)); KASSERT(s->key[PF_SK_STACK] == NULL, ("%s: state has key", __func__)); /* * We need to lock hash slots of both keys. To avoid deadlock * we always lock the slot with lower address first. Unlock order * isn't important. * * We also need to lock ID hash slot before dropping key * locks. On success we return with ID hash slot locked. */ if (skw == sks) { khs = khw = &V_pf_keyhash[pf_hashkey(skw)]; PF_HASHROW_LOCK(khs); } else { khs = &V_pf_keyhash[pf_hashkey(sks)]; khw = &V_pf_keyhash[pf_hashkey(skw)]; if (khs == khw) { PF_HASHROW_LOCK(khs); } else if (khs < khw) { PF_HASHROW_LOCK(khs); PF_HASHROW_LOCK(khw); } else { PF_HASHROW_LOCK(khw); PF_HASHROW_LOCK(khs); } } #define KEYS_UNLOCK() do { \ if (khs != khw) { \ PF_HASHROW_UNLOCK(khs); \ PF_HASHROW_UNLOCK(khw); \ } else \ PF_HASHROW_UNLOCK(khs); \ } while (0) /* * First run: start with wire key. */ sk = skw; kh = khw; idx = PF_SK_WIRE; MPASS(s->lock == NULL); s->lock = &V_pf_idhash[PF_IDHASH(s)].lock; keyattach: LIST_FOREACH(cur, &kh->keys, entry) if (bcmp(cur, sk, sizeof(struct pf_state_key_cmp)) == 0) break; if (cur != NULL) { /* Key exists. Check for same kif, if none, add to key. */ TAILQ_FOREACH(si, &cur->states[idx], key_list[idx]) { struct pf_idhash *ih = &V_pf_idhash[PF_IDHASH(si)]; PF_HASHROW_LOCK(ih); if (si->kif == s->kif && si->direction == s->direction) { if (sk->proto == IPPROTO_TCP && si->src.state >= TCPS_FIN_WAIT_2 && si->dst.state >= TCPS_FIN_WAIT_2) { /* * New state matches an old >FIN_WAIT_2 * state. We can't drop key hash locks, * thus we can't unlink it properly. * * As a workaround we drop it into * TCPS_CLOSED state, schedule purge * ASAP and push it into the very end * of the slot TAILQ, so that it won't * conflict with our new state. */ pf_set_protostate(si, PF_PEER_BOTH, TCPS_CLOSED); si->timeout = PFTM_PURGE; olds = si; } else { if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pf: %s key attach " "failed on %s: ", (idx == PF_SK_WIRE) ? "wire" : "stack", s->kif->pfik_name); pf_print_state_parts(s, (idx == PF_SK_WIRE) ? sk : NULL, (idx == PF_SK_STACK) ? sk : NULL); printf(", existing: "); pf_print_state_parts(si, (idx == PF_SK_WIRE) ? sk : NULL, (idx == PF_SK_STACK) ? sk : NULL); printf("\n"); } s->timeout = PFTM_UNLINKED; PF_HASHROW_UNLOCK(ih); KEYS_UNLOCK(); uma_zfree(V_pf_state_key_z, sk); if (idx == PF_SK_STACK) pf_detach_state(s); return (EEXIST); /* collision! */ } } PF_HASHROW_UNLOCK(ih); } uma_zfree(V_pf_state_key_z, sk); s->key[idx] = cur; } else { LIST_INSERT_HEAD(&kh->keys, sk, entry); s->key[idx] = sk; } stateattach: /* List is sorted, if-bound states before floating. */ if (s->kif == V_pfi_all) TAILQ_INSERT_TAIL(&s->key[idx]->states[idx], s, key_list[idx]); else TAILQ_INSERT_HEAD(&s->key[idx]->states[idx], s, key_list[idx]); if (olds) { TAILQ_REMOVE(&s->key[idx]->states[idx], olds, key_list[idx]); TAILQ_INSERT_TAIL(&s->key[idx]->states[idx], olds, key_list[idx]); olds = NULL; } /* * Attach done. See how should we (or should not?) * attach a second key. */ if (sks == skw) { s->key[PF_SK_STACK] = s->key[PF_SK_WIRE]; idx = PF_SK_STACK; sks = NULL; goto stateattach; } else if (sks != NULL) { /* * Continue attaching with stack key. */ sk = sks; kh = khs; idx = PF_SK_STACK; sks = NULL; goto keyattach; } PF_STATE_LOCK(s); KEYS_UNLOCK(); KASSERT(s->key[PF_SK_WIRE] != NULL && s->key[PF_SK_STACK] != NULL, ("%s failure", __func__)); return (0); #undef KEYS_UNLOCK } static void pf_detach_state(struct pf_kstate *s) { struct pf_state_key *sks = s->key[PF_SK_STACK]; struct pf_keyhash *kh; NET_EPOCH_ASSERT(); MPASS(s->timeout >= PFTM_MAX); pf_sctp_multihome_detach_addr(s); if ((s->state_flags & PFSTATE_PFLOW) && V_pflow_export_state_ptr) V_pflow_export_state_ptr(s); if (sks != NULL) { kh = &V_pf_keyhash[pf_hashkey(sks)]; PF_HASHROW_LOCK(kh); if (s->key[PF_SK_STACK] != NULL) pf_state_key_detach(s, PF_SK_STACK); /* * If both point to same key, then we are done. */ if (sks == s->key[PF_SK_WIRE]) { pf_state_key_detach(s, PF_SK_WIRE); PF_HASHROW_UNLOCK(kh); return; } PF_HASHROW_UNLOCK(kh); } if (s->key[PF_SK_WIRE] != NULL) { kh = &V_pf_keyhash[pf_hashkey(s->key[PF_SK_WIRE])]; PF_HASHROW_LOCK(kh); if (s->key[PF_SK_WIRE] != NULL) pf_state_key_detach(s, PF_SK_WIRE); PF_HASHROW_UNLOCK(kh); } } static void pf_state_key_detach(struct pf_kstate *s, int idx) { struct pf_state_key *sk = s->key[idx]; #ifdef INVARIANTS struct pf_keyhash *kh = &V_pf_keyhash[pf_hashkey(sk)]; PF_HASHROW_ASSERT(kh); #endif TAILQ_REMOVE(&sk->states[idx], s, key_list[idx]); s->key[idx] = NULL; if (TAILQ_EMPTY(&sk->states[0]) && TAILQ_EMPTY(&sk->states[1])) { LIST_REMOVE(sk, entry); uma_zfree(V_pf_state_key_z, sk); } } static int pf_state_key_ctor(void *mem, int size, void *arg, int flags) { struct pf_state_key *sk = mem; bzero(sk, sizeof(struct pf_state_key_cmp)); TAILQ_INIT(&sk->states[PF_SK_WIRE]); TAILQ_INIT(&sk->states[PF_SK_STACK]); return (0); } static int pf_state_key_addr_setup(struct pf_pdesc *pd, struct mbuf *m, int off, struct pf_state_key_cmp *key, int sidx, struct pf_addr *saddr, int didx, struct pf_addr *daddr, int multi) { #ifdef INET6 struct nd_neighbor_solicit nd; struct pf_addr *target; u_short action, reason; if (pd->af == AF_INET || pd->proto != IPPROTO_ICMPV6) goto copy; switch (pd->hdr.icmp6.icmp6_type) { case ND_NEIGHBOR_SOLICIT: if (multi) return (-1); if (!pf_pull_hdr(m, off, &nd, sizeof(nd), &action, &reason, pd->af)) return (-1); target = (struct pf_addr *)&nd.nd_ns_target; daddr = target; break; case ND_NEIGHBOR_ADVERT: if (multi) return (-1); if (!pf_pull_hdr(m, off, &nd, sizeof(nd), &action, &reason, pd->af)) return (-1); target = (struct pf_addr *)&nd.nd_ns_target; saddr = target; if (IN6_IS_ADDR_MULTICAST(&pd->dst->v6)) { key->addr[didx].addr32[0] = 0; key->addr[didx].addr32[1] = 0; key->addr[didx].addr32[2] = 0; key->addr[didx].addr32[3] = 0; daddr = NULL; /* overwritten */ } break; default: if (multi == PF_ICMP_MULTI_LINK) { key->addr[sidx].addr32[0] = IPV6_ADDR_INT32_MLL; key->addr[sidx].addr32[1] = 0; key->addr[sidx].addr32[2] = 0; key->addr[sidx].addr32[3] = IPV6_ADDR_INT32_ONE; saddr = NULL; /* overwritten */ } } copy: #endif if (saddr) PF_ACPY(&key->addr[sidx], saddr, pd->af); if (daddr) PF_ACPY(&key->addr[didx], daddr, pd->af); return (0); } struct pf_state_key * pf_state_key_setup(struct pf_pdesc *pd, struct mbuf *m, int off, struct pf_addr *saddr, struct pf_addr *daddr, u_int16_t sport, u_int16_t dport) { struct pf_state_key *sk; sk = uma_zalloc(V_pf_state_key_z, M_NOWAIT); if (sk == NULL) return (NULL); if (pf_state_key_addr_setup(pd, m, off, (struct pf_state_key_cmp *)sk, pd->sidx, pd->src, pd->didx, pd->dst, 0)) { uma_zfree(V_pf_state_key_z, sk); return (NULL); } sk->port[pd->sidx] = sport; sk->port[pd->didx] = dport; sk->proto = pd->proto; sk->af = pd->af; return (sk); } struct pf_state_key * pf_state_key_clone(const struct pf_state_key *orig) { struct pf_state_key *sk; sk = uma_zalloc(V_pf_state_key_z, M_NOWAIT); if (sk == NULL) return (NULL); bcopy(orig, sk, sizeof(struct pf_state_key_cmp)); return (sk); } int pf_state_insert(struct pfi_kkif *kif, struct pfi_kkif *orig_kif, struct pf_state_key *skw, struct pf_state_key *sks, struct pf_kstate *s) { struct pf_idhash *ih; struct pf_kstate *cur; int error; NET_EPOCH_ASSERT(); KASSERT(TAILQ_EMPTY(&sks->states[0]) && TAILQ_EMPTY(&sks->states[1]), ("%s: sks not pristine", __func__)); KASSERT(TAILQ_EMPTY(&skw->states[0]) && TAILQ_EMPTY(&skw->states[1]), ("%s: skw not pristine", __func__)); KASSERT(s->refs == 0, ("%s: state not pristine", __func__)); s->kif = kif; s->orig_kif = orig_kif; if (s->id == 0 && s->creatorid == 0) { s->id = alloc_unr64(&V_pf_stateid); s->id = htobe64(s->id); s->creatorid = V_pf_status.hostid; } /* Returns with ID locked on success. */ if ((error = pf_state_key_attach(skw, sks, s)) != 0) return (error); ih = &V_pf_idhash[PF_IDHASH(s)]; PF_HASHROW_ASSERT(ih); LIST_FOREACH(cur, &ih->states, entry) if (cur->id == s->id && cur->creatorid == s->creatorid) break; if (cur != NULL) { s->timeout = PFTM_UNLINKED; PF_HASHROW_UNLOCK(ih); if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pf: state ID collision: " "id: %016llx creatorid: %08x\n", (unsigned long long)be64toh(s->id), ntohl(s->creatorid)); } pf_detach_state(s); return (EEXIST); } LIST_INSERT_HEAD(&ih->states, s, entry); /* One for keys, one for ID hash. */ refcount_init(&s->refs, 2); pf_counter_u64_add(&V_pf_status.fcounters[FCNT_STATE_INSERT], 1); if (V_pfsync_insert_state_ptr != NULL) V_pfsync_insert_state_ptr(s); /* Returns locked. */ return (0); } /* * Find state by ID: returns with locked row on success. */ struct pf_kstate * pf_find_state_byid(uint64_t id, uint32_t creatorid) { struct pf_idhash *ih; struct pf_kstate *s; pf_counter_u64_add(&V_pf_status.fcounters[FCNT_STATE_SEARCH], 1); ih = &V_pf_idhash[(be64toh(id) % (V_pf_hashmask + 1))]; PF_HASHROW_LOCK(ih); LIST_FOREACH(s, &ih->states, entry) if (s->id == id && s->creatorid == creatorid) break; if (s == NULL) PF_HASHROW_UNLOCK(ih); return (s); } /* * Find state by key. * Returns with ID hash slot locked on success. */ static struct pf_kstate * pf_find_state(struct pfi_kkif *kif, const struct pf_state_key_cmp *key, u_int dir) { struct pf_keyhash *kh; struct pf_state_key *sk; struct pf_kstate *s; int idx; pf_counter_u64_add(&V_pf_status.fcounters[FCNT_STATE_SEARCH], 1); kh = &V_pf_keyhash[pf_hashkey((const struct pf_state_key *)key)]; PF_HASHROW_LOCK(kh); LIST_FOREACH(sk, &kh->keys, entry) if (bcmp(sk, key, sizeof(struct pf_state_key_cmp)) == 0) break; if (sk == NULL) { PF_HASHROW_UNLOCK(kh); return (NULL); } idx = (dir == PF_IN ? PF_SK_WIRE : PF_SK_STACK); /* List is sorted, if-bound states before floating ones. */ TAILQ_FOREACH(s, &sk->states[idx], key_list[idx]) if (s->kif == V_pfi_all || s->kif == kif || s->orig_kif == kif) { PF_STATE_LOCK(s); PF_HASHROW_UNLOCK(kh); if (__predict_false(s->timeout >= PFTM_MAX)) { /* * State is either being processed by * pf_unlink_state() in an other thread, or * is scheduled for immediate expiry. */ PF_STATE_UNLOCK(s); return (NULL); } return (s); } PF_HASHROW_UNLOCK(kh); return (NULL); } /* * Returns with ID hash slot locked on success. */ struct pf_kstate * pf_find_state_all(const struct pf_state_key_cmp *key, u_int dir, int *more) { struct pf_keyhash *kh; struct pf_state_key *sk; struct pf_kstate *s, *ret = NULL; int idx, inout = 0; pf_counter_u64_add(&V_pf_status.fcounters[FCNT_STATE_SEARCH], 1); kh = &V_pf_keyhash[pf_hashkey((const struct pf_state_key *)key)]; PF_HASHROW_LOCK(kh); LIST_FOREACH(sk, &kh->keys, entry) if (bcmp(sk, key, sizeof(struct pf_state_key_cmp)) == 0) break; if (sk == NULL) { PF_HASHROW_UNLOCK(kh); return (NULL); } switch (dir) { case PF_IN: idx = PF_SK_WIRE; break; case PF_OUT: idx = PF_SK_STACK; break; case PF_INOUT: idx = PF_SK_WIRE; inout = 1; break; default: panic("%s: dir %u", __func__, dir); } second_run: TAILQ_FOREACH(s, &sk->states[idx], key_list[idx]) { if (more == NULL) { PF_STATE_LOCK(s); PF_HASHROW_UNLOCK(kh); return (s); } if (ret) (*more)++; else { ret = s; PF_STATE_LOCK(s); } } if (inout == 1) { inout = 0; idx = PF_SK_STACK; goto second_run; } PF_HASHROW_UNLOCK(kh); return (ret); } /* * FIXME * This routine is inefficient -- locks the state only to unlock immediately on * return. * It is racy -- after the state is unlocked nothing stops other threads from * removing it. */ bool pf_find_state_all_exists(const struct pf_state_key_cmp *key, u_int dir) { struct pf_kstate *s; s = pf_find_state_all(key, dir, NULL); if (s != NULL) { PF_STATE_UNLOCK(s); return (true); } return (false); } struct pf_udp_mapping * pf_udp_mapping_create(sa_family_t af, struct pf_addr *src_addr, uint16_t src_port, struct pf_addr *nat_addr, uint16_t nat_port) { struct pf_udp_mapping *mapping; mapping = uma_zalloc(V_pf_udp_mapping_z, M_NOWAIT | M_ZERO); if (mapping == NULL) return (NULL); PF_ACPY(&mapping->endpoints[0].addr, src_addr, af); mapping->endpoints[0].port = src_port; mapping->endpoints[0].af = af; mapping->endpoints[0].mapping = mapping; PF_ACPY(&mapping->endpoints[1].addr, nat_addr, af); mapping->endpoints[1].port = nat_port; mapping->endpoints[1].af = af; mapping->endpoints[1].mapping = mapping; refcount_init(&mapping->refs, 1); return (mapping); } int pf_udp_mapping_insert(struct pf_udp_mapping *mapping) { struct pf_udpendpointhash *h0, *h1; struct pf_udp_endpoint *endpoint; int ret = EEXIST; h0 = &V_pf_udpendpointhash[pf_hashudpendpoint(&mapping->endpoints[0])]; h1 = &V_pf_udpendpointhash[pf_hashudpendpoint(&mapping->endpoints[1])]; if (h0 == h1) { PF_HASHROW_LOCK(h0); } else if (h0 < h1) { PF_HASHROW_LOCK(h0); PF_HASHROW_LOCK(h1); } else { PF_HASHROW_LOCK(h1); PF_HASHROW_LOCK(h0); } LIST_FOREACH(endpoint, &h0->endpoints, entry) { if (bcmp(endpoint, &mapping->endpoints[0], sizeof(struct pf_udp_endpoint_cmp)) == 0) break; } if (endpoint != NULL) goto cleanup; LIST_FOREACH(endpoint, &h1->endpoints, entry) { if (bcmp(endpoint, &mapping->endpoints[1], sizeof(struct pf_udp_endpoint_cmp)) == 0) break; } if (endpoint != NULL) goto cleanup; LIST_INSERT_HEAD(&h0->endpoints, &mapping->endpoints[0], entry); LIST_INSERT_HEAD(&h1->endpoints, &mapping->endpoints[1], entry); ret = 0; cleanup: if (h0 != h1) { PF_HASHROW_UNLOCK(h0); PF_HASHROW_UNLOCK(h1); } else { PF_HASHROW_UNLOCK(h0); } return (ret); } void pf_udp_mapping_release(struct pf_udp_mapping *mapping) { /* refcount is synchronized on the source endpoint's row lock */ struct pf_udpendpointhash *h0, *h1; if (mapping == NULL) return; h0 = &V_pf_udpendpointhash[pf_hashudpendpoint(&mapping->endpoints[0])]; PF_HASHROW_LOCK(h0); if (refcount_release(&mapping->refs)) { LIST_REMOVE(&mapping->endpoints[0], entry); PF_HASHROW_UNLOCK(h0); h1 = &V_pf_udpendpointhash[pf_hashudpendpoint(&mapping->endpoints[1])]; PF_HASHROW_LOCK(h1); LIST_REMOVE(&mapping->endpoints[1], entry); PF_HASHROW_UNLOCK(h1); uma_zfree(V_pf_udp_mapping_z, mapping); } else { PF_HASHROW_UNLOCK(h0); } } struct pf_udp_mapping * pf_udp_mapping_find(struct pf_udp_endpoint_cmp *key) { struct pf_udpendpointhash *uh; struct pf_udp_endpoint *endpoint; uh = &V_pf_udpendpointhash[pf_hashudpendpoint((struct pf_udp_endpoint*)key)]; PF_HASHROW_LOCK(uh); LIST_FOREACH(endpoint, &uh->endpoints, entry) { if (bcmp(endpoint, key, sizeof(struct pf_udp_endpoint_cmp)) == 0 && bcmp(endpoint, &endpoint->mapping->endpoints[0], sizeof(struct pf_udp_endpoint_cmp)) == 0) break; } if (endpoint == NULL) { PF_HASHROW_UNLOCK(uh); return (NULL); } refcount_acquire(&endpoint->mapping->refs); PF_HASHROW_UNLOCK(uh); return (endpoint->mapping); } /* END state table stuff */ static void pf_send(struct pf_send_entry *pfse) { PF_SENDQ_LOCK(); STAILQ_INSERT_TAIL(&V_pf_sendqueue, pfse, pfse_next); PF_SENDQ_UNLOCK(); swi_sched(V_pf_swi_cookie, 0); } static bool pf_isforlocal(struct mbuf *m, int af) { switch (af) { #ifdef INET case AF_INET: { struct ip *ip = mtod(m, struct ip *); return (in_localip(ip->ip_dst)); } #endif #ifdef INET6 case AF_INET6: { struct ip6_hdr *ip6; struct in6_ifaddr *ia; ip6 = mtod(m, struct ip6_hdr *); ia = in6ifa_ifwithaddr(&ip6->ip6_dst, 0 /* XXX */, false); if (ia == NULL) return (false); return (! (ia->ia6_flags & IN6_IFF_NOTREADY)); } #endif default: panic("Unsupported af %d", af); } return (false); } int pf_icmp_mapping(struct pf_pdesc *pd, u_int8_t type, int *icmp_dir, int *multi, u_int16_t *virtual_id, u_int16_t *virtual_type) { /* * ICMP types marked with PF_OUT are typically responses to * PF_IN, and will match states in the opposite direction. * PF_IN ICMP types need to match a state with that type. */ *icmp_dir = PF_OUT; *multi = PF_ICMP_MULTI_LINK; /* Queries (and responses) */ switch (pd->af) { #ifdef INET case AF_INET: switch (type) { case ICMP_ECHO: *icmp_dir = PF_IN; case ICMP_ECHOREPLY: *virtual_type = ICMP_ECHO; *virtual_id = pd->hdr.icmp.icmp_id; break; case ICMP_TSTAMP: *icmp_dir = PF_IN; case ICMP_TSTAMPREPLY: *virtual_type = ICMP_TSTAMP; *virtual_id = pd->hdr.icmp.icmp_id; break; case ICMP_IREQ: *icmp_dir = PF_IN; case ICMP_IREQREPLY: *virtual_type = ICMP_IREQ; *virtual_id = pd->hdr.icmp.icmp_id; break; case ICMP_MASKREQ: *icmp_dir = PF_IN; case ICMP_MASKREPLY: *virtual_type = ICMP_MASKREQ; *virtual_id = pd->hdr.icmp.icmp_id; break; case ICMP_IPV6_WHEREAREYOU: *icmp_dir = PF_IN; case ICMP_IPV6_IAMHERE: *virtual_type = ICMP_IPV6_WHEREAREYOU; *virtual_id = 0; /* Nothing sane to match on! */ break; case ICMP_MOBILE_REGREQUEST: *icmp_dir = PF_IN; case ICMP_MOBILE_REGREPLY: *virtual_type = ICMP_MOBILE_REGREQUEST; *virtual_id = 0; /* Nothing sane to match on! */ break; case ICMP_ROUTERSOLICIT: *icmp_dir = PF_IN; case ICMP_ROUTERADVERT: *virtual_type = ICMP_ROUTERSOLICIT; *virtual_id = 0; /* Nothing sane to match on! */ break; /* These ICMP types map to other connections */ case ICMP_UNREACH: case ICMP_SOURCEQUENCH: case ICMP_REDIRECT: case ICMP_TIMXCEED: case ICMP_PARAMPROB: /* These will not be used, but set them anyway */ *icmp_dir = PF_IN; *virtual_type = type; *virtual_id = 0; HTONS(*virtual_type); return (1); /* These types match to another state */ /* * All remaining ICMP types get their own states, * and will only match in one direction. */ default: *icmp_dir = PF_IN; *virtual_type = type; *virtual_id = 0; break; } break; #endif /* INET */ #ifdef INET6 case AF_INET6: switch (type) { case ICMP6_ECHO_REQUEST: *icmp_dir = PF_IN; case ICMP6_ECHO_REPLY: *virtual_type = ICMP6_ECHO_REQUEST; *virtual_id = pd->hdr.icmp6.icmp6_id; break; case MLD_LISTENER_QUERY: case MLD_LISTENER_REPORT: { /* * Listener Report can be sent by clients * without an associated Listener Query. * In addition to that, when Report is sent as a * reply to a Query its source and destination * address are different. */ *icmp_dir = PF_IN; *virtual_type = MLD_LISTENER_QUERY; *virtual_id = 0; break; } case MLD_MTRACE: *icmp_dir = PF_IN; case MLD_MTRACE_RESP: *virtual_type = MLD_MTRACE; *virtual_id = 0; /* Nothing sane to match on! */ break; case ND_NEIGHBOR_SOLICIT: *icmp_dir = PF_IN; case ND_NEIGHBOR_ADVERT: { *virtual_type = ND_NEIGHBOR_SOLICIT; *virtual_id = 0; break; } /* * These ICMP types map to other connections. * ND_REDIRECT can't be in this list because the triggering * packet header is optional. */ case ICMP6_DST_UNREACH: case ICMP6_PACKET_TOO_BIG: case ICMP6_TIME_EXCEEDED: case ICMP6_PARAM_PROB: /* These will not be used, but set them anyway */ *icmp_dir = PF_IN; *virtual_type = type; *virtual_id = 0; HTONS(*virtual_type); return (1); /* These types match to another state */ /* * All remaining ICMP6 types get their own states, * and will only match in one direction. */ default: *icmp_dir = PF_IN; *virtual_type = type; *virtual_id = 0; break; } break; #endif /* INET6 */ default: *icmp_dir = PF_IN; *virtual_type = type; *virtual_id = 0; break; } HTONS(*virtual_type); return (0); /* These types match to their own state */ } void pf_intr(void *v) { struct epoch_tracker et; struct pf_send_head queue; struct pf_send_entry *pfse, *next; CURVNET_SET((struct vnet *)v); PF_SENDQ_LOCK(); queue = V_pf_sendqueue; STAILQ_INIT(&V_pf_sendqueue); PF_SENDQ_UNLOCK(); NET_EPOCH_ENTER(et); STAILQ_FOREACH_SAFE(pfse, &queue, pfse_next, next) { switch (pfse->pfse_type) { #ifdef INET case PFSE_IP: { if (pf_isforlocal(pfse->pfse_m, AF_INET)) { pfse->pfse_m->m_flags |= M_SKIP_FIREWALL; pfse->pfse_m->m_pkthdr.csum_flags |= CSUM_IP_VALID | CSUM_IP_CHECKED; ip_input(pfse->pfse_m); } else { ip_output(pfse->pfse_m, NULL, NULL, 0, NULL, NULL); } break; } case PFSE_ICMP: icmp_error(pfse->pfse_m, pfse->icmpopts.type, pfse->icmpopts.code, 0, pfse->icmpopts.mtu); break; #endif /* INET */ #ifdef INET6 case PFSE_IP6: if (pf_isforlocal(pfse->pfse_m, AF_INET6)) { pfse->pfse_m->m_flags |= M_SKIP_FIREWALL; ip6_input(pfse->pfse_m); } else { ip6_output(pfse->pfse_m, NULL, NULL, 0, NULL, NULL, NULL); } break; case PFSE_ICMP6: icmp6_error(pfse->pfse_m, pfse->icmpopts.type, pfse->icmpopts.code, pfse->icmpopts.mtu); break; #endif /* INET6 */ default: panic("%s: unknown type", __func__); } free(pfse, M_PFTEMP); } NET_EPOCH_EXIT(et); CURVNET_RESTORE(); } #define pf_purge_thread_period (hz / 10) #ifdef PF_WANT_32_TO_64_COUNTER static void pf_status_counter_u64_periodic(void) { PF_RULES_RASSERT(); if ((V_pf_counter_periodic_iter % (pf_purge_thread_period * 10 * 60)) != 0) { return; } for (int i = 0; i < FCNT_MAX; i++) { pf_counter_u64_periodic(&V_pf_status.fcounters[i]); } } static void pf_kif_counter_u64_periodic(void) { struct pfi_kkif *kif; size_t r, run; PF_RULES_RASSERT(); if (__predict_false(V_pf_allkifcount == 0)) { return; } if ((V_pf_counter_periodic_iter % (pf_purge_thread_period * 10 * 300)) != 0) { return; } run = V_pf_allkifcount / 10; if (run < 5) run = 5; for (r = 0; r < run; r++) { kif = LIST_NEXT(V_pf_kifmarker, pfik_allkiflist); if (kif == NULL) { LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist); LIST_INSERT_HEAD(&V_pf_allkiflist, V_pf_kifmarker, pfik_allkiflist); break; } LIST_REMOVE(V_pf_kifmarker, pfik_allkiflist); LIST_INSERT_AFTER(kif, V_pf_kifmarker, pfik_allkiflist); for (int i = 0; i < 2; i++) { for (int j = 0; j < 2; j++) { for (int k = 0; k < 2; k++) { pf_counter_u64_periodic(&kif->pfik_packets[i][j][k]); pf_counter_u64_periodic(&kif->pfik_bytes[i][j][k]); } } } } } static void pf_rule_counter_u64_periodic(void) { struct pf_krule *rule; size_t r, run; PF_RULES_RASSERT(); if (__predict_false(V_pf_allrulecount == 0)) { return; } if ((V_pf_counter_periodic_iter % (pf_purge_thread_period * 10 * 300)) != 0) { return; } run = V_pf_allrulecount / 10; if (run < 5) run = 5; for (r = 0; r < run; r++) { rule = LIST_NEXT(V_pf_rulemarker, allrulelist); if (rule == NULL) { LIST_REMOVE(V_pf_rulemarker, allrulelist); LIST_INSERT_HEAD(&V_pf_allrulelist, V_pf_rulemarker, allrulelist); break; } LIST_REMOVE(V_pf_rulemarker, allrulelist); LIST_INSERT_AFTER(rule, V_pf_rulemarker, allrulelist); pf_counter_u64_periodic(&rule->evaluations); for (int i = 0; i < 2; i++) { pf_counter_u64_periodic(&rule->packets[i]); pf_counter_u64_periodic(&rule->bytes[i]); } } } static void pf_counter_u64_periodic_main(void) { PF_RULES_RLOCK_TRACKER; V_pf_counter_periodic_iter++; PF_RULES_RLOCK(); pf_counter_u64_critical_enter(); pf_status_counter_u64_periodic(); pf_kif_counter_u64_periodic(); pf_rule_counter_u64_periodic(); pf_counter_u64_critical_exit(); PF_RULES_RUNLOCK(); } #else #define pf_counter_u64_periodic_main() do { } while (0) #endif void pf_purge_thread(void *unused __unused) { struct epoch_tracker et; VNET_ITERATOR_DECL(vnet_iter); sx_xlock(&pf_end_lock); while (pf_end_threads == 0) { sx_sleep(pf_purge_thread, &pf_end_lock, 0, "pftm", pf_purge_thread_period); VNET_LIST_RLOCK(); NET_EPOCH_ENTER(et); VNET_FOREACH(vnet_iter) { CURVNET_SET(vnet_iter); /* Wait until V_pf_default_rule is initialized. */ if (V_pf_vnet_active == 0) { CURVNET_RESTORE(); continue; } pf_counter_u64_periodic_main(); /* * Process 1/interval fraction of the state * table every run. */ V_pf_purge_idx = pf_purge_expired_states(V_pf_purge_idx, V_pf_hashmask / (V_pf_default_rule.timeout[PFTM_INTERVAL] * 10)); /* * Purge other expired types every * PFTM_INTERVAL seconds. */ if (V_pf_purge_idx == 0) { /* * Order is important: * - states and src nodes reference rules * - states and rules reference kifs */ pf_purge_expired_fragments(); pf_purge_expired_src_nodes(); pf_purge_unlinked_rules(); pfi_kkif_purge(); } CURVNET_RESTORE(); } NET_EPOCH_EXIT(et); VNET_LIST_RUNLOCK(); } pf_end_threads++; sx_xunlock(&pf_end_lock); kproc_exit(0); } void pf_unload_vnet_purge(void) { /* * To cleanse up all kifs and rules we need * two runs: first one clears reference flags, * then pf_purge_expired_states() doesn't * raise them, and then second run frees. */ pf_purge_unlinked_rules(); pfi_kkif_purge(); /* * Now purge everything. */ pf_purge_expired_states(0, V_pf_hashmask); pf_purge_fragments(UINT_MAX); pf_purge_expired_src_nodes(); /* * Now all kifs & rules should be unreferenced, * thus should be successfully freed. */ pf_purge_unlinked_rules(); pfi_kkif_purge(); } u_int32_t pf_state_expires(const struct pf_kstate *state) { u_int32_t timeout; u_int32_t start; u_int32_t end; u_int32_t states; /* handle all PFTM_* > PFTM_MAX here */ if (state->timeout == PFTM_PURGE) return (time_uptime); KASSERT(state->timeout != PFTM_UNLINKED, ("pf_state_expires: timeout == PFTM_UNLINKED")); KASSERT((state->timeout < PFTM_MAX), ("pf_state_expires: timeout > PFTM_MAX")); timeout = state->rule.ptr->timeout[state->timeout]; if (!timeout) timeout = V_pf_default_rule.timeout[state->timeout]; start = state->rule.ptr->timeout[PFTM_ADAPTIVE_START]; if (start && state->rule.ptr != &V_pf_default_rule) { end = state->rule.ptr->timeout[PFTM_ADAPTIVE_END]; states = counter_u64_fetch(state->rule.ptr->states_cur); } else { start = V_pf_default_rule.timeout[PFTM_ADAPTIVE_START]; end = V_pf_default_rule.timeout[PFTM_ADAPTIVE_END]; states = V_pf_status.states; } if (end && states > start && start < end) { if (states < end) { timeout = (u_int64_t)timeout * (end - states) / (end - start); return ((state->expire / 1000) + timeout); } else return (time_uptime); } return ((state->expire / 1000) + timeout); } void pf_purge_expired_src_nodes(void) { struct pf_ksrc_node_list freelist; struct pf_srchash *sh; struct pf_ksrc_node *cur, *next; int i; LIST_INIT(&freelist); for (i = 0, sh = V_pf_srchash; i <= V_pf_srchashmask; i++, sh++) { PF_HASHROW_LOCK(sh); LIST_FOREACH_SAFE(cur, &sh->nodes, entry, next) if (cur->states == 0 && cur->expire <= time_uptime) { pf_unlink_src_node(cur); LIST_INSERT_HEAD(&freelist, cur, entry); } else if (cur->rule.ptr != NULL) cur->rule.ptr->rule_ref |= PFRULE_REFS; PF_HASHROW_UNLOCK(sh); } pf_free_src_nodes(&freelist); V_pf_status.src_nodes = uma_zone_get_cur(V_pf_sources_z); } static void pf_src_tree_remove_state(struct pf_kstate *s) { struct pf_ksrc_node *sn; uint32_t timeout; timeout = s->rule.ptr->timeout[PFTM_SRC_NODE] ? s->rule.ptr->timeout[PFTM_SRC_NODE] : V_pf_default_rule.timeout[PFTM_SRC_NODE]; if (s->src_node != NULL) { sn = s->src_node; PF_SRC_NODE_LOCK(sn); if (s->src.tcp_est) --sn->conn; if (--sn->states == 0) sn->expire = time_uptime + timeout; PF_SRC_NODE_UNLOCK(sn); } if (s->nat_src_node != s->src_node && s->nat_src_node != NULL) { sn = s->nat_src_node; PF_SRC_NODE_LOCK(sn); if (--sn->states == 0) sn->expire = time_uptime + timeout; PF_SRC_NODE_UNLOCK(sn); } s->src_node = s->nat_src_node = NULL; } /* * Unlink and potentilly free a state. Function may be * called with ID hash row locked, but always returns * unlocked, since it needs to go through key hash locking. */ int pf_unlink_state(struct pf_kstate *s) { struct pf_idhash *ih = &V_pf_idhash[PF_IDHASH(s)]; NET_EPOCH_ASSERT(); PF_HASHROW_ASSERT(ih); if (s->timeout == PFTM_UNLINKED) { /* * State is being processed * by pf_unlink_state() in * an other thread. */ PF_HASHROW_UNLOCK(ih); return (0); /* XXXGL: undefined actually */ } if (s->src.state == PF_TCPS_PROXY_DST) { /* XXX wire key the right one? */ pf_send_tcp(s->rule.ptr, s->key[PF_SK_WIRE]->af, &s->key[PF_SK_WIRE]->addr[1], &s->key[PF_SK_WIRE]->addr[0], s->key[PF_SK_WIRE]->port[1], s->key[PF_SK_WIRE]->port[0], s->src.seqhi, s->src.seqlo + 1, TH_RST|TH_ACK, 0, 0, 0, true, s->tag, 0, s->act.rtableid); } LIST_REMOVE(s, entry); pf_src_tree_remove_state(s); if (V_pfsync_delete_state_ptr != NULL) V_pfsync_delete_state_ptr(s); STATE_DEC_COUNTERS(s); s->timeout = PFTM_UNLINKED; /* Ensure we remove it from the list of halfopen states, if needed. */ if (s->key[PF_SK_STACK] != NULL && s->key[PF_SK_STACK]->proto == IPPROTO_TCP) pf_set_protostate(s, PF_PEER_BOTH, TCPS_CLOSED); PF_HASHROW_UNLOCK(ih); pf_detach_state(s); pf_udp_mapping_release(s->udp_mapping); /* pf_state_insert() initialises refs to 2 */ return (pf_release_staten(s, 2)); } struct pf_kstate * pf_alloc_state(int flags) { return (uma_zalloc(V_pf_state_z, flags | M_ZERO)); } void pf_free_state(struct pf_kstate *cur) { struct pf_krule_item *ri; KASSERT(cur->refs == 0, ("%s: %p has refs", __func__, cur)); KASSERT(cur->timeout == PFTM_UNLINKED, ("%s: timeout %u", __func__, cur->timeout)); while ((ri = SLIST_FIRST(&cur->match_rules))) { SLIST_REMOVE_HEAD(&cur->match_rules, entry); free(ri, M_PF_RULE_ITEM); } pf_normalize_tcp_cleanup(cur); uma_zfree(V_pf_state_z, cur); pf_counter_u64_add(&V_pf_status.fcounters[FCNT_STATE_REMOVALS], 1); } /* * Called only from pf_purge_thread(), thus serialized. */ static u_int pf_purge_expired_states(u_int i, int maxcheck) { struct pf_idhash *ih; struct pf_kstate *s; struct pf_krule_item *mrm; size_t count __unused; V_pf_status.states = uma_zone_get_cur(V_pf_state_z); /* * Go through hash and unlink states that expire now. */ while (maxcheck > 0) { count = 0; ih = &V_pf_idhash[i]; /* only take the lock if we expect to do work */ if (!LIST_EMPTY(&ih->states)) { relock: PF_HASHROW_LOCK(ih); LIST_FOREACH(s, &ih->states, entry) { if (pf_state_expires(s) <= time_uptime) { V_pf_status.states -= pf_unlink_state(s); goto relock; } s->rule.ptr->rule_ref |= PFRULE_REFS; if (s->nat_rule.ptr != NULL) s->nat_rule.ptr->rule_ref |= PFRULE_REFS; if (s->anchor.ptr != NULL) s->anchor.ptr->rule_ref |= PFRULE_REFS; s->kif->pfik_flags |= PFI_IFLAG_REFS; SLIST_FOREACH(mrm, &s->match_rules, entry) mrm->r->rule_ref |= PFRULE_REFS; if (s->rt_kif) s->rt_kif->pfik_flags |= PFI_IFLAG_REFS; count++; } PF_HASHROW_UNLOCK(ih); } SDT_PROBE2(pf, purge, state, rowcount, i, count); /* Return when we hit end of hash. */ if (++i > V_pf_hashmask) { V_pf_status.states = uma_zone_get_cur(V_pf_state_z); return (0); } maxcheck--; } V_pf_status.states = uma_zone_get_cur(V_pf_state_z); return (i); } static void pf_purge_unlinked_rules(void) { struct pf_krulequeue tmpq; struct pf_krule *r, *r1; /* * If we have overloading task pending, then we'd * better skip purging this time. There is a tiny * probability that overloading task references * an already unlinked rule. */ PF_OVERLOADQ_LOCK(); if (!SLIST_EMPTY(&V_pf_overloadqueue)) { PF_OVERLOADQ_UNLOCK(); return; } PF_OVERLOADQ_UNLOCK(); /* * Do naive mark-and-sweep garbage collecting of old rules. * Reference flag is raised by pf_purge_expired_states() * and pf_purge_expired_src_nodes(). * * To avoid LOR between PF_UNLNKDRULES_LOCK/PF_RULES_WLOCK, * use a temporary queue. */ TAILQ_INIT(&tmpq); PF_UNLNKDRULES_LOCK(); TAILQ_FOREACH_SAFE(r, &V_pf_unlinked_rules, entries, r1) { if (!(r->rule_ref & PFRULE_REFS)) { TAILQ_REMOVE(&V_pf_unlinked_rules, r, entries); TAILQ_INSERT_TAIL(&tmpq, r, entries); } else r->rule_ref &= ~PFRULE_REFS; } PF_UNLNKDRULES_UNLOCK(); if (!TAILQ_EMPTY(&tmpq)) { PF_CONFIG_LOCK(); PF_RULES_WLOCK(); TAILQ_FOREACH_SAFE(r, &tmpq, entries, r1) { TAILQ_REMOVE(&tmpq, r, entries); pf_free_rule(r); } PF_RULES_WUNLOCK(); PF_CONFIG_UNLOCK(); } } void pf_print_host(struct pf_addr *addr, u_int16_t p, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: { u_int32_t a = ntohl(addr->addr32[0]); printf("%u.%u.%u.%u", (a>>24)&255, (a>>16)&255, (a>>8)&255, a&255); if (p) { p = ntohs(p); printf(":%u", p); } break; } #endif /* INET */ #ifdef INET6 case AF_INET6: { u_int16_t b; u_int8_t i, curstart, curend, maxstart, maxend; curstart = curend = maxstart = maxend = 255; for (i = 0; i < 8; i++) { if (!addr->addr16[i]) { if (curstart == 255) curstart = i; curend = i; } else { if ((curend - curstart) > (maxend - maxstart)) { maxstart = curstart; maxend = curend; } curstart = curend = 255; } } if ((curend - curstart) > (maxend - maxstart)) { maxstart = curstart; maxend = curend; } for (i = 0; i < 8; i++) { if (i >= maxstart && i <= maxend) { if (i == 0) printf(":"); if (i == maxend) printf(":"); } else { b = ntohs(addr->addr16[i]); printf("%x", b); if (i < 7) printf(":"); } } if (p) { p = ntohs(p); printf("[%u]", p); } break; } #endif /* INET6 */ } } void pf_print_state(struct pf_kstate *s) { pf_print_state_parts(s, NULL, NULL); } static void pf_print_state_parts(struct pf_kstate *s, struct pf_state_key *skwp, struct pf_state_key *sksp) { struct pf_state_key *skw, *sks; u_int8_t proto, dir; /* Do our best to fill these, but they're skipped if NULL */ skw = skwp ? skwp : (s ? s->key[PF_SK_WIRE] : NULL); sks = sksp ? sksp : (s ? s->key[PF_SK_STACK] : NULL); proto = skw ? skw->proto : (sks ? sks->proto : 0); dir = s ? s->direction : 0; switch (proto) { case IPPROTO_IPV4: printf("IPv4"); break; case IPPROTO_IPV6: printf("IPv6"); break; case IPPROTO_TCP: printf("TCP"); break; case IPPROTO_UDP: printf("UDP"); break; case IPPROTO_ICMP: printf("ICMP"); break; case IPPROTO_ICMPV6: printf("ICMPv6"); break; default: printf("%u", proto); break; } switch (dir) { case PF_IN: printf(" in"); break; case PF_OUT: printf(" out"); break; } if (skw) { printf(" wire: "); pf_print_host(&skw->addr[0], skw->port[0], skw->af); printf(" "); pf_print_host(&skw->addr[1], skw->port[1], skw->af); } if (sks) { printf(" stack: "); if (sks != skw) { pf_print_host(&sks->addr[0], sks->port[0], sks->af); printf(" "); pf_print_host(&sks->addr[1], sks->port[1], sks->af); } else printf("-"); } if (s) { if (proto == IPPROTO_TCP) { printf(" [lo=%u high=%u win=%u modulator=%u", s->src.seqlo, s->src.seqhi, s->src.max_win, s->src.seqdiff); if (s->src.wscale && s->dst.wscale) printf(" wscale=%u", s->src.wscale & PF_WSCALE_MASK); printf("]"); printf(" [lo=%u high=%u win=%u modulator=%u", s->dst.seqlo, s->dst.seqhi, s->dst.max_win, s->dst.seqdiff); if (s->src.wscale && s->dst.wscale) printf(" wscale=%u", s->dst.wscale & PF_WSCALE_MASK); printf("]"); } printf(" %u:%u", s->src.state, s->dst.state); if (s->rule.ptr) printf(" @%d", s->rule.ptr->nr); } } void pf_print_flags(u_int8_t f) { if (f) printf(" "); if (f & TH_FIN) printf("F"); if (f & TH_SYN) printf("S"); if (f & TH_RST) printf("R"); if (f & TH_PUSH) printf("P"); if (f & TH_ACK) printf("A"); if (f & TH_URG) printf("U"); if (f & TH_ECE) printf("E"); if (f & TH_CWR) printf("W"); } #define PF_SET_SKIP_STEPS(i) \ do { \ while (head[i] != cur) { \ head[i]->skip[i].ptr = cur; \ head[i] = TAILQ_NEXT(head[i], entries); \ } \ } while (0) void pf_calc_skip_steps(struct pf_krulequeue *rules) { struct pf_krule *cur, *prev, *head[PF_SKIP_COUNT]; int i; cur = TAILQ_FIRST(rules); prev = cur; for (i = 0; i < PF_SKIP_COUNT; ++i) head[i] = cur; while (cur != NULL) { if (cur->kif != prev->kif || cur->ifnot != prev->ifnot) PF_SET_SKIP_STEPS(PF_SKIP_IFP); if (cur->direction != prev->direction) PF_SET_SKIP_STEPS(PF_SKIP_DIR); if (cur->af != prev->af) PF_SET_SKIP_STEPS(PF_SKIP_AF); if (cur->proto != prev->proto) PF_SET_SKIP_STEPS(PF_SKIP_PROTO); if (cur->src.neg != prev->src.neg || pf_addr_wrap_neq(&cur->src.addr, &prev->src.addr)) PF_SET_SKIP_STEPS(PF_SKIP_SRC_ADDR); + if (cur->dst.neg != prev->dst.neg || + pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr)) + PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR); if (cur->src.port[0] != prev->src.port[0] || cur->src.port[1] != prev->src.port[1] || cur->src.port_op != prev->src.port_op) PF_SET_SKIP_STEPS(PF_SKIP_SRC_PORT); - if (cur->dst.neg != prev->dst.neg || - pf_addr_wrap_neq(&cur->dst.addr, &prev->dst.addr)) - PF_SET_SKIP_STEPS(PF_SKIP_DST_ADDR); if (cur->dst.port[0] != prev->dst.port[0] || cur->dst.port[1] != prev->dst.port[1] || cur->dst.port_op != prev->dst.port_op) PF_SET_SKIP_STEPS(PF_SKIP_DST_PORT); prev = cur; cur = TAILQ_NEXT(cur, entries); } for (i = 0; i < PF_SKIP_COUNT; ++i) PF_SET_SKIP_STEPS(i); } int pf_addr_wrap_neq(struct pf_addr_wrap *aw1, struct pf_addr_wrap *aw2) { if (aw1->type != aw2->type) return (1); switch (aw1->type) { case PF_ADDR_ADDRMASK: case PF_ADDR_RANGE: if (PF_ANEQ(&aw1->v.a.addr, &aw2->v.a.addr, AF_INET6)) return (1); if (PF_ANEQ(&aw1->v.a.mask, &aw2->v.a.mask, AF_INET6)) return (1); return (0); case PF_ADDR_DYNIFTL: return (aw1->p.dyn->pfid_kt != aw2->p.dyn->pfid_kt); case PF_ADDR_NOROUTE: case PF_ADDR_URPFFAILED: return (0); case PF_ADDR_TABLE: return (aw1->p.tbl != aw2->p.tbl); default: printf("invalid address type: %d\n", aw1->type); return (1); } } /** * Checksum updates are a little complicated because the checksum in the TCP/UDP * header isn't always a full checksum. In some cases (i.e. output) it's a * pseudo-header checksum, which is a partial checksum over src/dst IP * addresses, protocol number and length. * * That means we have the following cases: * * Input or forwarding: we don't have TSO, the checksum fields are full * checksums, we need to update the checksum whenever we change anything. * * Output (i.e. the checksum is a pseudo-header checksum): * x The field being updated is src/dst address or affects the length of * the packet. We need to update the pseudo-header checksum (note that this * checksum is not ones' complement). * x Some other field is being modified (e.g. src/dst port numbers): We * don't have to update anything. **/ u_int16_t pf_cksum_fixup(u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp) { u_int32_t x; x = cksum + old - new; x = (x + (x >> 16)) & 0xffff; /* optimise: eliminate a branch when not udp */ if (udp && cksum == 0x0000) return cksum; if (udp && x == 0x0000) x = 0xffff; return (u_int16_t)(x); } static void pf_patch_8(struct mbuf *m, u_int16_t *cksum, u_int8_t *f, u_int8_t v, bool hi, u_int8_t udp) { u_int16_t old = htons(hi ? (*f << 8) : *f); u_int16_t new = htons(hi ? ( v << 8) : v); if (*f == v) return; *f = v; if (m->m_pkthdr.csum_flags & (CSUM_DELAY_DATA | CSUM_DELAY_DATA_IPV6)) return; *cksum = pf_cksum_fixup(*cksum, old, new, udp); } void pf_patch_16_unaligned(struct mbuf *m, u_int16_t *cksum, void *f, u_int16_t v, bool hi, u_int8_t udp) { u_int8_t *fb = (u_int8_t *)f; u_int8_t *vb = (u_int8_t *)&v; pf_patch_8(m, cksum, fb++, *vb++, hi, udp); pf_patch_8(m, cksum, fb++, *vb++, !hi, udp); } void pf_patch_32_unaligned(struct mbuf *m, u_int16_t *cksum, void *f, u_int32_t v, bool hi, u_int8_t udp) { u_int8_t *fb = (u_int8_t *)f; u_int8_t *vb = (u_int8_t *)&v; pf_patch_8(m, cksum, fb++, *vb++, hi, udp); pf_patch_8(m, cksum, fb++, *vb++, !hi, udp); pf_patch_8(m, cksum, fb++, *vb++, hi, udp); pf_patch_8(m, cksum, fb++, *vb++, !hi, udp); } u_int16_t pf_proto_cksum_fixup(struct mbuf *m, u_int16_t cksum, u_int16_t old, u_int16_t new, u_int8_t udp) { if (m->m_pkthdr.csum_flags & (CSUM_DELAY_DATA | CSUM_DELAY_DATA_IPV6)) return (cksum); return (pf_cksum_fixup(cksum, old, new, udp)); } static void pf_change_ap(struct mbuf *m, struct pf_addr *a, u_int16_t *p, u_int16_t *ic, u_int16_t *pc, struct pf_addr *an, u_int16_t pn, u_int8_t u, sa_family_t af) { struct pf_addr ao; u_int16_t po = *p; PF_ACPY(&ao, a, af); PF_ACPY(a, an, af); if (m->m_pkthdr.csum_flags & (CSUM_DELAY_DATA | CSUM_DELAY_DATA_IPV6)) *pc = ~*pc; *p = pn; switch (af) { #ifdef INET case AF_INET: *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, ao.addr16[0], an->addr16[0], 0), ao.addr16[1], an->addr16[1], 0); *p = pn; *pc = pf_cksum_fixup(pf_cksum_fixup(*pc, ao.addr16[0], an->addr16[0], u), ao.addr16[1], an->addr16[1], u); *pc = pf_proto_cksum_fixup(m, *pc, po, pn, u); break; #endif /* INET */ #ifdef INET6 case AF_INET6: *pc = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(*pc, ao.addr16[0], an->addr16[0], u), ao.addr16[1], an->addr16[1], u), ao.addr16[2], an->addr16[2], u), ao.addr16[3], an->addr16[3], u), ao.addr16[4], an->addr16[4], u), ao.addr16[5], an->addr16[5], u), ao.addr16[6], an->addr16[6], u), ao.addr16[7], an->addr16[7], u); *pc = pf_proto_cksum_fixup(m, *pc, po, pn, u); break; #endif /* INET6 */ } if (m->m_pkthdr.csum_flags & (CSUM_DELAY_DATA | CSUM_DELAY_DATA_IPV6)) { *pc = ~*pc; if (! *pc) *pc = 0xffff; } } /* Changes a u_int32_t. Uses a void * so there are no align restrictions */ void pf_change_a(void *a, u_int16_t *c, u_int32_t an, u_int8_t u) { u_int32_t ao; memcpy(&ao, a, sizeof(ao)); memcpy(a, &an, sizeof(u_int32_t)); *c = pf_cksum_fixup(pf_cksum_fixup(*c, ao / 65536, an / 65536, u), ao % 65536, an % 65536, u); } void pf_change_proto_a(struct mbuf *m, void *a, u_int16_t *c, u_int32_t an, u_int8_t udp) { u_int32_t ao; memcpy(&ao, a, sizeof(ao)); memcpy(a, &an, sizeof(u_int32_t)); *c = pf_proto_cksum_fixup(m, pf_proto_cksum_fixup(m, *c, ao / 65536, an / 65536, udp), ao % 65536, an % 65536, udp); } #ifdef INET6 static void pf_change_a6(struct pf_addr *a, u_int16_t *c, struct pf_addr *an, u_int8_t u) { struct pf_addr ao; PF_ACPY(&ao, a, AF_INET6); PF_ACPY(a, an, AF_INET6); *c = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(*c, ao.addr16[0], an->addr16[0], u), ao.addr16[1], an->addr16[1], u), ao.addr16[2], an->addr16[2], u), ao.addr16[3], an->addr16[3], u), ao.addr16[4], an->addr16[4], u), ao.addr16[5], an->addr16[5], u), ao.addr16[6], an->addr16[6], u), ao.addr16[7], an->addr16[7], u); } #endif /* INET6 */ static void pf_change_icmp(struct pf_addr *ia, u_int16_t *ip, struct pf_addr *oa, struct pf_addr *na, u_int16_t np, u_int16_t *pc, u_int16_t *h2c, u_int16_t *ic, u_int16_t *hc, u_int8_t u, sa_family_t af) { struct pf_addr oia, ooa; PF_ACPY(&oia, ia, af); if (oa) PF_ACPY(&ooa, oa, af); /* Change inner protocol port, fix inner protocol checksum. */ if (ip != NULL) { u_int16_t oip = *ip; u_int32_t opc; if (pc != NULL) opc = *pc; *ip = np; if (pc != NULL) *pc = pf_cksum_fixup(*pc, oip, *ip, u); *ic = pf_cksum_fixup(*ic, oip, *ip, 0); if (pc != NULL) *ic = pf_cksum_fixup(*ic, opc, *pc, 0); } /* Change inner ip address, fix inner ip and icmp checksums. */ PF_ACPY(ia, na, af); switch (af) { #ifdef INET case AF_INET: { u_int32_t oh2c = *h2c; *h2c = pf_cksum_fixup(pf_cksum_fixup(*h2c, oia.addr16[0], ia->addr16[0], 0), oia.addr16[1], ia->addr16[1], 0); *ic = pf_cksum_fixup(pf_cksum_fixup(*ic, oia.addr16[0], ia->addr16[0], 0), oia.addr16[1], ia->addr16[1], 0); *ic = pf_cksum_fixup(*ic, oh2c, *h2c, 0); break; } #endif /* INET */ #ifdef INET6 case AF_INET6: *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(*ic, oia.addr16[0], ia->addr16[0], u), oia.addr16[1], ia->addr16[1], u), oia.addr16[2], ia->addr16[2], u), oia.addr16[3], ia->addr16[3], u), oia.addr16[4], ia->addr16[4], u), oia.addr16[5], ia->addr16[5], u), oia.addr16[6], ia->addr16[6], u), oia.addr16[7], ia->addr16[7], u); break; #endif /* INET6 */ } /* Outer ip address, fix outer ip or icmpv6 checksum, if necessary. */ if (oa) { PF_ACPY(oa, na, af); switch (af) { #ifdef INET case AF_INET: *hc = pf_cksum_fixup(pf_cksum_fixup(*hc, ooa.addr16[0], oa->addr16[0], 0), ooa.addr16[1], oa->addr16[1], 0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: *ic = pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(pf_cksum_fixup( pf_cksum_fixup(pf_cksum_fixup(*ic, ooa.addr16[0], oa->addr16[0], u), ooa.addr16[1], oa->addr16[1], u), ooa.addr16[2], oa->addr16[2], u), ooa.addr16[3], oa->addr16[3], u), ooa.addr16[4], oa->addr16[4], u), ooa.addr16[5], oa->addr16[5], u), ooa.addr16[6], oa->addr16[6], u), ooa.addr16[7], oa->addr16[7], u); break; #endif /* INET6 */ } } } /* * Need to modulate the sequence numbers in the TCP SACK option * (credits to Krzysztof Pfaff for report and patch) */ static int pf_modulate_sack(struct mbuf *m, int off, struct pf_pdesc *pd, struct tcphdr *th, struct pf_state_peer *dst) { int hlen = (th->th_off << 2) - sizeof(*th), thoptlen = hlen; u_int8_t opts[TCP_MAXOLEN], *opt = opts; int copyback = 0, i, olen; struct sackblk sack; #define TCPOLEN_SACKLEN (TCPOLEN_SACK + 2) if (hlen < TCPOLEN_SACKLEN || !pf_pull_hdr(m, off + sizeof(*th), opts, hlen, NULL, NULL, pd->af)) return 0; while (hlen >= TCPOLEN_SACKLEN) { size_t startoff = opt - opts; olen = opt[1]; switch (*opt) { case TCPOPT_EOL: /* FALLTHROUGH */ case TCPOPT_NOP: opt++; hlen--; break; case TCPOPT_SACK: if (olen > hlen) olen = hlen; if (olen >= TCPOLEN_SACKLEN) { for (i = 2; i + TCPOLEN_SACK <= olen; i += TCPOLEN_SACK) { memcpy(&sack, &opt[i], sizeof(sack)); pf_patch_32_unaligned(m, &th->th_sum, &sack.start, htonl(ntohl(sack.start) - dst->seqdiff), PF_ALGNMNT(startoff), 0); pf_patch_32_unaligned(m, &th->th_sum, &sack.end, htonl(ntohl(sack.end) - dst->seqdiff), PF_ALGNMNT(startoff), 0); memcpy(&opt[i], &sack, sizeof(sack)); } copyback = 1; } /* FALLTHROUGH */ default: if (olen < 2) olen = 2; hlen -= olen; opt += olen; } } if (copyback) m_copyback(m, off + sizeof(*th), thoptlen, (caddr_t)opts); return (copyback); } struct mbuf * pf_build_tcp(const struct pf_krule *r, sa_family_t af, const struct pf_addr *saddr, const struct pf_addr *daddr, u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack, u_int8_t tcp_flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, bool skip_firewall, u_int16_t mtag_tag, u_int16_t mtag_flags, int rtableid) { struct mbuf *m; int len, tlen; #ifdef INET struct ip *h = NULL; #endif /* INET */ #ifdef INET6 struct ip6_hdr *h6 = NULL; #endif /* INET6 */ struct tcphdr *th; char *opt; struct pf_mtag *pf_mtag; len = 0; th = NULL; /* maximum segment size tcp option */ tlen = sizeof(struct tcphdr); if (mss) tlen += 4; switch (af) { #ifdef INET case AF_INET: len = sizeof(struct ip) + tlen; break; #endif /* INET */ #ifdef INET6 case AF_INET6: len = sizeof(struct ip6_hdr) + tlen; break; #endif /* INET6 */ default: panic("%s: unsupported af %d", __func__, af); } m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) return (NULL); #ifdef MAC mac_netinet_firewall_send(m); #endif if ((pf_mtag = pf_get_mtag(m)) == NULL) { m_freem(m); return (NULL); } if (skip_firewall) m->m_flags |= M_SKIP_FIREWALL; pf_mtag->tag = mtag_tag; pf_mtag->flags = mtag_flags; if (rtableid >= 0) M_SETFIB(m, rtableid); #ifdef ALTQ if (r != NULL && r->qid) { pf_mtag->qid = r->qid; /* add hints for ecn */ pf_mtag->hdr = mtod(m, struct ip *); } #endif /* ALTQ */ m->m_data += max_linkhdr; m->m_pkthdr.len = m->m_len = len; /* The rest of the stack assumes a rcvif, so provide one. * This is a locally generated packet, so .. close enough. */ m->m_pkthdr.rcvif = V_loif; bzero(m->m_data, len); switch (af) { #ifdef INET case AF_INET: h = mtod(m, struct ip *); /* IP header fields included in the TCP checksum */ h->ip_p = IPPROTO_TCP; h->ip_len = htons(tlen); h->ip_src.s_addr = saddr->v4.s_addr; h->ip_dst.s_addr = daddr->v4.s_addr; th = (struct tcphdr *)((caddr_t)h + sizeof(struct ip)); break; #endif /* INET */ #ifdef INET6 case AF_INET6: h6 = mtod(m, struct ip6_hdr *); /* IP header fields included in the TCP checksum */ h6->ip6_nxt = IPPROTO_TCP; h6->ip6_plen = htons(tlen); memcpy(&h6->ip6_src, &saddr->v6, sizeof(struct in6_addr)); memcpy(&h6->ip6_dst, &daddr->v6, sizeof(struct in6_addr)); th = (struct tcphdr *)((caddr_t)h6 + sizeof(struct ip6_hdr)); break; #endif /* INET6 */ } /* TCP header */ th->th_sport = sport; th->th_dport = dport; th->th_seq = htonl(seq); th->th_ack = htonl(ack); th->th_off = tlen >> 2; th->th_flags = tcp_flags; th->th_win = htons(win); if (mss) { opt = (char *)(th + 1); opt[0] = TCPOPT_MAXSEG; opt[1] = 4; HTONS(mss); bcopy((caddr_t)&mss, (caddr_t)(opt + 2), 2); } switch (af) { #ifdef INET case AF_INET: /* TCP checksum */ th->th_sum = in_cksum(m, len); /* Finish the IP header */ h->ip_v = 4; h->ip_hl = sizeof(*h) >> 2; h->ip_tos = IPTOS_LOWDELAY; h->ip_off = htons(V_path_mtu_discovery ? IP_DF : 0); h->ip_len = htons(len); h->ip_ttl = ttl ? ttl : V_ip_defttl; h->ip_sum = 0; break; #endif /* INET */ #ifdef INET6 case AF_INET6: /* TCP checksum */ th->th_sum = in6_cksum(m, IPPROTO_TCP, sizeof(struct ip6_hdr), tlen); h6->ip6_vfc |= IPV6_VERSION; h6->ip6_hlim = IPV6_DEFHLIM; break; #endif /* INET6 */ } return (m); } static void pf_send_sctp_abort(sa_family_t af, struct pf_pdesc *pd, uint8_t ttl, int rtableid) { struct mbuf *m; #ifdef INET struct ip *h = NULL; #endif /* INET */ #ifdef INET6 struct ip6_hdr *h6 = NULL; #endif /* INET6 */ struct sctphdr *hdr; struct sctp_chunkhdr *chunk; struct pf_send_entry *pfse; int off = 0; MPASS(af == pd->af); m = m_gethdr(M_NOWAIT, MT_DATA); if (m == NULL) return; m->m_data += max_linkhdr; m->m_flags |= M_SKIP_FIREWALL; /* The rest of the stack assumes a rcvif, so provide one. * This is a locally generated packet, so .. close enough. */ m->m_pkthdr.rcvif = V_loif; /* IPv4|6 header */ switch (af) { #ifdef INET case AF_INET: bzero(m->m_data, sizeof(struct ip) + sizeof(*hdr) + sizeof(*chunk)); h = mtod(m, struct ip *); /* IP header fields included in the TCP checksum */ h->ip_p = IPPROTO_SCTP; h->ip_len = htons(sizeof(*h) + sizeof(*hdr) + sizeof(*chunk)); h->ip_ttl = ttl ? ttl : V_ip_defttl; h->ip_src = pd->dst->v4; h->ip_dst = pd->src->v4; off += sizeof(struct ip); break; #endif /* INET */ #ifdef INET6 case AF_INET6: bzero(m->m_data, sizeof(struct ip6_hdr) + sizeof(*hdr) + sizeof(*chunk)); h6 = mtod(m, struct ip6_hdr *); /* IP header fields included in the TCP checksum */ h6->ip6_vfc |= IPV6_VERSION; h6->ip6_nxt = IPPROTO_SCTP; h6->ip6_plen = htons(sizeof(*h6) + sizeof(*hdr) + sizeof(*chunk)); h6->ip6_hlim = ttl ? ttl : V_ip6_defhlim; memcpy(&h6->ip6_src, &pd->dst->v6, sizeof(struct in6_addr)); memcpy(&h6->ip6_dst, &pd->src->v6, sizeof(struct in6_addr)); off += sizeof(struct ip6_hdr); break; #endif /* INET6 */ } /* SCTP header */ hdr = mtodo(m, off); hdr->src_port = pd->hdr.sctp.dest_port; hdr->dest_port = pd->hdr.sctp.src_port; hdr->v_tag = pd->sctp_initiate_tag; hdr->checksum = 0; /* Abort chunk. */ off += sizeof(struct sctphdr); chunk = mtodo(m, off); chunk->chunk_type = SCTP_ABORT_ASSOCIATION; chunk->chunk_length = htons(sizeof(*chunk)); /* SCTP checksum */ off += sizeof(*chunk); m->m_pkthdr.len = m->m_len = off; pf_sctp_checksum(m, off - sizeof(*hdr) - sizeof(*chunk)); if (rtableid >= 0) M_SETFIB(m, rtableid); /* Allocate outgoing queue entry, mbuf and mbuf tag. */ pfse = malloc(sizeof(*pfse), M_PFTEMP, M_NOWAIT); if (pfse == NULL) { m_freem(m); return; } switch (af) { #ifdef INET case AF_INET: pfse->pfse_type = PFSE_IP; break; #endif /* INET */ #ifdef INET6 case AF_INET6: pfse->pfse_type = PFSE_IP6; break; #endif /* INET6 */ } pfse->pfse_m = m; pf_send(pfse); } void pf_send_tcp(const struct pf_krule *r, sa_family_t af, const struct pf_addr *saddr, const struct pf_addr *daddr, u_int16_t sport, u_int16_t dport, u_int32_t seq, u_int32_t ack, u_int8_t tcp_flags, u_int16_t win, u_int16_t mss, u_int8_t ttl, bool skip_firewall, u_int16_t mtag_tag, u_int16_t mtag_flags, int rtableid) { struct pf_send_entry *pfse; struct mbuf *m; m = pf_build_tcp(r, af, saddr, daddr, sport, dport, seq, ack, tcp_flags, win, mss, ttl, skip_firewall, mtag_tag, mtag_flags, rtableid); if (m == NULL) return; /* Allocate outgoing queue entry, mbuf and mbuf tag. */ pfse = malloc(sizeof(*pfse), M_PFTEMP, M_NOWAIT); if (pfse == NULL) { m_freem(m); return; } switch (af) { #ifdef INET case AF_INET: pfse->pfse_type = PFSE_IP; break; #endif /* INET */ #ifdef INET6 case AF_INET6: pfse->pfse_type = PFSE_IP6; break; #endif /* INET6 */ } pfse->pfse_m = m; pf_send(pfse); } static void pf_return(struct pf_krule *r, struct pf_krule *nr, struct pf_pdesc *pd, struct pf_state_key *sk, int off, struct mbuf *m, struct tcphdr *th, struct pfi_kkif *kif, u_int16_t bproto_sum, u_int16_t bip_sum, int hdrlen, u_short *reason, int rtableid) { struct pf_addr * const saddr = pd->src; struct pf_addr * const daddr = pd->dst; sa_family_t af = pd->af; /* undo NAT changes, if they have taken place */ if (nr != NULL) { PF_ACPY(saddr, &sk->addr[pd->sidx], af); PF_ACPY(daddr, &sk->addr[pd->didx], af); if (pd->sport) *pd->sport = sk->port[pd->sidx]; if (pd->dport) *pd->dport = sk->port[pd->didx]; if (pd->proto_sum) *pd->proto_sum = bproto_sum; if (pd->ip_sum) *pd->ip_sum = bip_sum; m_copyback(m, off, hdrlen, pd->hdr.any); } if (pd->proto == IPPROTO_TCP && ((r->rule_flag & PFRULE_RETURNRST) || (r->rule_flag & PFRULE_RETURN)) && !(th->th_flags & TH_RST)) { u_int32_t ack = ntohl(th->th_seq) + pd->p_len; int len = 0; #ifdef INET struct ip *h4; #endif #ifdef INET6 struct ip6_hdr *h6; #endif switch (af) { #ifdef INET case AF_INET: h4 = mtod(m, struct ip *); len = ntohs(h4->ip_len) - off; break; #endif #ifdef INET6 case AF_INET6: h6 = mtod(m, struct ip6_hdr *); len = ntohs(h6->ip6_plen) - (off - sizeof(*h6)); break; #endif } if (pf_check_proto_cksum(m, off, len, IPPROTO_TCP, af)) REASON_SET(reason, PFRES_PROTCKSUM); else { if (th->th_flags & TH_SYN) ack++; if (th->th_flags & TH_FIN) ack++; pf_send_tcp(r, af, pd->dst, pd->src, th->th_dport, th->th_sport, ntohl(th->th_ack), ack, TH_RST|TH_ACK, 0, 0, r->return_ttl, true, 0, 0, rtableid); } } else if (pd->proto == IPPROTO_SCTP && (r->rule_flag & PFRULE_RETURN)) { pf_send_sctp_abort(af, pd, r->return_ttl, rtableid); } else if (pd->proto != IPPROTO_ICMP && af == AF_INET && r->return_icmp) pf_send_icmp(m, r->return_icmp >> 8, r->return_icmp & 255, af, r, rtableid); else if (pd->proto != IPPROTO_ICMPV6 && af == AF_INET6 && r->return_icmp6) pf_send_icmp(m, r->return_icmp6 >> 8, r->return_icmp6 & 255, af, r, rtableid); } static int pf_match_ieee8021q_pcp(u_int8_t prio, struct mbuf *m) { struct m_tag *mtag; u_int8_t mpcp; mtag = m_tag_locate(m, MTAG_8021Q, MTAG_8021Q_PCP_IN, NULL); if (mtag == NULL) return (0); if (prio == PF_PRIO_ZERO) prio = 0; mpcp = *(uint8_t *)(mtag + 1); return (mpcp == prio); } static int pf_icmp_to_bandlim(uint8_t type) { switch (type) { case ICMP_ECHO: case ICMP_ECHOREPLY: return (BANDLIM_ICMP_ECHO); case ICMP_TSTAMP: case ICMP_TSTAMPREPLY: return (BANDLIM_ICMP_TSTAMP); case ICMP_UNREACH: default: return (BANDLIM_ICMP_UNREACH); } } static void pf_send_icmp(struct mbuf *m, u_int8_t type, u_int8_t code, sa_family_t af, struct pf_krule *r, int rtableid) { struct pf_send_entry *pfse; struct mbuf *m0; struct pf_mtag *pf_mtag; /* ICMP packet rate limitation. */ #ifdef INET6 if (af == AF_INET6) { if (icmp6_ratelimit(NULL, type, code)) return; } #endif #ifdef INET if (af == AF_INET) { if (badport_bandlim(pf_icmp_to_bandlim(type)) != 0) return; } #endif /* Allocate outgoing queue entry, mbuf and mbuf tag. */ pfse = malloc(sizeof(*pfse), M_PFTEMP, M_NOWAIT); if (pfse == NULL) return; if ((m0 = m_copypacket(m, M_NOWAIT)) == NULL) { free(pfse, M_PFTEMP); return; } if ((pf_mtag = pf_get_mtag(m0)) == NULL) { free(pfse, M_PFTEMP); return; } /* XXX: revisit */ m0->m_flags |= M_SKIP_FIREWALL; if (rtableid >= 0) M_SETFIB(m0, rtableid); #ifdef ALTQ if (r->qid) { pf_mtag->qid = r->qid; /* add hints for ecn */ pf_mtag->hdr = mtod(m0, struct ip *); } #endif /* ALTQ */ switch (af) { #ifdef INET case AF_INET: pfse->pfse_type = PFSE_ICMP; break; #endif /* INET */ #ifdef INET6 case AF_INET6: pfse->pfse_type = PFSE_ICMP6; break; #endif /* INET6 */ } pfse->pfse_m = m0; pfse->icmpopts.type = type; pfse->icmpopts.code = code; pf_send(pfse); } /* * Return 1 if the addresses a and b match (with mask m), otherwise return 0. * If n is 0, they match if they are equal. If n is != 0, they match if they * are different. */ int pf_match_addr(u_int8_t n, struct pf_addr *a, struct pf_addr *m, struct pf_addr *b, sa_family_t af) { int match = 0; switch (af) { #ifdef INET case AF_INET: if (IN_ARE_MASKED_ADDR_EQUAL(a->v4, b->v4, m->v4)) match++; break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (IN6_ARE_MASKED_ADDR_EQUAL(&a->v6, &b->v6, &m->v6)) match++; break; #endif /* INET6 */ } if (match) { if (n) return (0); else return (1); } else { if (n) return (1); else return (0); } } /* * Return 1 if b <= a <= e, otherwise return 0. */ int pf_match_addr_range(struct pf_addr *b, struct pf_addr *e, struct pf_addr *a, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: if ((ntohl(a->addr32[0]) < ntohl(b->addr32[0])) || (ntohl(a->addr32[0]) > ntohl(e->addr32[0]))) return (0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: { int i; /* check a >= b */ for (i = 0; i < 4; ++i) if (ntohl(a->addr32[i]) > ntohl(b->addr32[i])) break; else if (ntohl(a->addr32[i]) < ntohl(b->addr32[i])) return (0); /* check a <= e */ for (i = 0; i < 4; ++i) if (ntohl(a->addr32[i]) < ntohl(e->addr32[i])) break; else if (ntohl(a->addr32[i]) > ntohl(e->addr32[i])) return (0); break; } #endif /* INET6 */ } return (1); } static int pf_match(u_int8_t op, u_int32_t a1, u_int32_t a2, u_int32_t p) { switch (op) { case PF_OP_IRG: return ((p > a1) && (p < a2)); case PF_OP_XRG: return ((p < a1) || (p > a2)); case PF_OP_RRG: return ((p >= a1) && (p <= a2)); case PF_OP_EQ: return (p == a1); case PF_OP_NE: return (p != a1); case PF_OP_LT: return (p < a1); case PF_OP_LE: return (p <= a1); case PF_OP_GT: return (p > a1); case PF_OP_GE: return (p >= a1); } return (0); /* never reached */ } int pf_match_port(u_int8_t op, u_int16_t a1, u_int16_t a2, u_int16_t p) { NTOHS(a1); NTOHS(a2); NTOHS(p); return (pf_match(op, a1, a2, p)); } static int pf_match_uid(u_int8_t op, uid_t a1, uid_t a2, uid_t u) { if (u == UID_MAX && op != PF_OP_EQ && op != PF_OP_NE) return (0); return (pf_match(op, a1, a2, u)); } static int pf_match_gid(u_int8_t op, gid_t a1, gid_t a2, gid_t g) { if (g == GID_MAX && op != PF_OP_EQ && op != PF_OP_NE) return (0); return (pf_match(op, a1, a2, g)); } int pf_match_tag(struct mbuf *m, struct pf_krule *r, int *tag, int mtag) { if (*tag == -1) *tag = mtag; return ((!r->match_tag_not && r->match_tag == *tag) || (r->match_tag_not && r->match_tag != *tag)); } static int pf_match_rcvif(struct mbuf *m, struct pf_krule *r) { struct ifnet *ifp = m->m_pkthdr.rcvif; struct pfi_kkif *kif; if (ifp == NULL) return (0); kif = (struct pfi_kkif *)ifp->if_pf_kif; if (kif == NULL) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_test_via: kif == NULL, @%d via %s\n", r->nr, r->rcv_ifname)); return (0); } return (pfi_kkif_match(r->rcv_kif, kif)); } int pf_tag_packet(struct mbuf *m, struct pf_pdesc *pd, int tag) { KASSERT(tag > 0, ("%s: tag %d", __func__, tag)); if (pd->pf_mtag == NULL && ((pd->pf_mtag = pf_get_mtag(m)) == NULL)) return (ENOMEM); pd->pf_mtag->tag = tag; return (0); } #define PF_ANCHOR_STACKSIZE 32 struct pf_kanchor_stackframe { struct pf_kruleset *rs; struct pf_krule *r; /* XXX: + match bit */ struct pf_kanchor *child; }; /* * XXX: We rely on malloc(9) returning pointer aligned addresses. */ #define PF_ANCHORSTACK_MATCH 0x00000001 #define PF_ANCHORSTACK_MASK (PF_ANCHORSTACK_MATCH) #define PF_ANCHOR_MATCH(f) ((uintptr_t)(f)->r & PF_ANCHORSTACK_MATCH) #define PF_ANCHOR_RULE(f) (struct pf_krule *) \ ((uintptr_t)(f)->r & ~PF_ANCHORSTACK_MASK) #define PF_ANCHOR_SET_MATCH(f) do { (f)->r = (void *) \ ((uintptr_t)(f)->r | PF_ANCHORSTACK_MATCH); \ } while (0) void pf_step_into_anchor(struct pf_kanchor_stackframe *stack, int *depth, struct pf_kruleset **rs, int n, struct pf_krule **r, struct pf_krule **a, int *match) { struct pf_kanchor_stackframe *f; PF_RULES_RASSERT(); if (match) *match = 0; if (*depth >= PF_ANCHOR_STACKSIZE) { printf("%s: anchor stack overflow on %s\n", __func__, (*r)->anchor->name); *r = TAILQ_NEXT(*r, entries); return; } else if (*depth == 0 && a != NULL) *a = *r; f = stack + (*depth)++; f->rs = *rs; f->r = *r; if ((*r)->anchor_wildcard) { struct pf_kanchor_node *parent = &(*r)->anchor->children; if ((f->child = RB_MIN(pf_kanchor_node, parent)) == NULL) { *r = NULL; return; } *rs = &f->child->ruleset; } else { f->child = NULL; *rs = &(*r)->anchor->ruleset; } *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); } int pf_step_out_of_anchor(struct pf_kanchor_stackframe *stack, int *depth, struct pf_kruleset **rs, int n, struct pf_krule **r, struct pf_krule **a, int *match) { struct pf_kanchor_stackframe *f; struct pf_krule *fr; int quick = 0; PF_RULES_RASSERT(); do { if (*depth <= 0) break; f = stack + *depth - 1; fr = PF_ANCHOR_RULE(f); if (f->child != NULL) { /* * This block traverses through * a wildcard anchor. */ if (match != NULL && *match) { /* * If any of "*" matched, then * "foo/ *" matched, mark frame * appropriately. */ PF_ANCHOR_SET_MATCH(f); *match = 0; } f->child = RB_NEXT(pf_kanchor_node, &fr->anchor->children, f->child); if (f->child != NULL) { *rs = &f->child->ruleset; *r = TAILQ_FIRST((*rs)->rules[n].active.ptr); if (*r == NULL) continue; else break; } } (*depth)--; if (*depth == 0 && a != NULL) *a = NULL; *rs = f->rs; if (PF_ANCHOR_MATCH(f) || (match != NULL && *match)) quick = fr->quick; *r = TAILQ_NEXT(fr, entries); } while (*r == NULL); return (quick); } struct pf_keth_anchor_stackframe { struct pf_keth_ruleset *rs; struct pf_keth_rule *r; /* XXX: + match bit */ struct pf_keth_anchor *child; }; #define PF_ETH_ANCHOR_MATCH(f) ((uintptr_t)(f)->r & PF_ANCHORSTACK_MATCH) #define PF_ETH_ANCHOR_RULE(f) (struct pf_keth_rule *) \ ((uintptr_t)(f)->r & ~PF_ANCHORSTACK_MASK) #define PF_ETH_ANCHOR_SET_MATCH(f) do { (f)->r = (void *) \ ((uintptr_t)(f)->r | PF_ANCHORSTACK_MATCH); \ } while (0) void pf_step_into_keth_anchor(struct pf_keth_anchor_stackframe *stack, int *depth, struct pf_keth_ruleset **rs, struct pf_keth_rule **r, struct pf_keth_rule **a, int *match) { struct pf_keth_anchor_stackframe *f; NET_EPOCH_ASSERT(); if (match) *match = 0; if (*depth >= PF_ANCHOR_STACKSIZE) { printf("%s: anchor stack overflow on %s\n", __func__, (*r)->anchor->name); *r = TAILQ_NEXT(*r, entries); return; } else if (*depth == 0 && a != NULL) *a = *r; f = stack + (*depth)++; f->rs = *rs; f->r = *r; if ((*r)->anchor_wildcard) { struct pf_keth_anchor_node *parent = &(*r)->anchor->children; if ((f->child = RB_MIN(pf_keth_anchor_node, parent)) == NULL) { *r = NULL; return; } *rs = &f->child->ruleset; } else { f->child = NULL; *rs = &(*r)->anchor->ruleset; } *r = TAILQ_FIRST((*rs)->active.rules); } int pf_step_out_of_keth_anchor(struct pf_keth_anchor_stackframe *stack, int *depth, struct pf_keth_ruleset **rs, struct pf_keth_rule **r, struct pf_keth_rule **a, int *match) { struct pf_keth_anchor_stackframe *f; struct pf_keth_rule *fr; int quick = 0; NET_EPOCH_ASSERT(); do { if (*depth <= 0) break; f = stack + *depth - 1; fr = PF_ETH_ANCHOR_RULE(f); if (f->child != NULL) { /* * This block traverses through * a wildcard anchor. */ if (match != NULL && *match) { /* * If any of "*" matched, then * "foo/ *" matched, mark frame * appropriately. */ PF_ETH_ANCHOR_SET_MATCH(f); *match = 0; } f->child = RB_NEXT(pf_keth_anchor_node, &fr->anchor->children, f->child); if (f->child != NULL) { *rs = &f->child->ruleset; *r = TAILQ_FIRST((*rs)->active.rules); if (*r == NULL) continue; else break; } } (*depth)--; if (*depth == 0 && a != NULL) *a = NULL; *rs = f->rs; if (PF_ETH_ANCHOR_MATCH(f) || (match != NULL && *match)) quick = fr->quick; *r = TAILQ_NEXT(fr, entries); } while (*r == NULL); return (quick); } #ifdef INET6 void pf_poolmask(struct pf_addr *naddr, struct pf_addr *raddr, struct pf_addr *rmask, struct pf_addr *saddr, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); break; #endif /* INET */ case AF_INET6: naddr->addr32[0] = (raddr->addr32[0] & rmask->addr32[0]) | ((rmask->addr32[0] ^ 0xffffffff ) & saddr->addr32[0]); naddr->addr32[1] = (raddr->addr32[1] & rmask->addr32[1]) | ((rmask->addr32[1] ^ 0xffffffff ) & saddr->addr32[1]); naddr->addr32[2] = (raddr->addr32[2] & rmask->addr32[2]) | ((rmask->addr32[2] ^ 0xffffffff ) & saddr->addr32[2]); naddr->addr32[3] = (raddr->addr32[3] & rmask->addr32[3]) | ((rmask->addr32[3] ^ 0xffffffff ) & saddr->addr32[3]); break; } } void pf_addr_inc(struct pf_addr *addr, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); break; #endif /* INET */ case AF_INET6: if (addr->addr32[3] == 0xffffffff) { addr->addr32[3] = 0; if (addr->addr32[2] == 0xffffffff) { addr->addr32[2] = 0; if (addr->addr32[1] == 0xffffffff) { addr->addr32[1] = 0; addr->addr32[0] = htonl(ntohl(addr->addr32[0]) + 1); } else addr->addr32[1] = htonl(ntohl(addr->addr32[1]) + 1); } else addr->addr32[2] = htonl(ntohl(addr->addr32[2]) + 1); } else addr->addr32[3] = htonl(ntohl(addr->addr32[3]) + 1); break; } } #endif /* INET6 */ void pf_rule_to_actions(struct pf_krule *r, struct pf_rule_actions *a) { /* * Modern rules use the same flags in rules as they do in states. */ a->flags |= (r->scrub_flags & (PFSTATE_NODF|PFSTATE_RANDOMID| PFSTATE_SCRUB_TCP|PFSTATE_SETPRIO)); /* * Old-style scrub rules have different flags which need to be translated. */ if (r->rule_flag & PFRULE_RANDOMID) a->flags |= PFSTATE_RANDOMID; if (r->scrub_flags & PFSTATE_SETTOS || r->rule_flag & PFRULE_SET_TOS ) { a->flags |= PFSTATE_SETTOS; a->set_tos = r->set_tos; } if (r->qid) a->qid = r->qid; if (r->pqid) a->pqid = r->pqid; if (r->rtableid >= 0) a->rtableid = r->rtableid; a->log |= r->log; if (r->min_ttl) a->min_ttl = r->min_ttl; if (r->max_mss) a->max_mss = r->max_mss; if (r->dnpipe) a->dnpipe = r->dnpipe; if (r->dnrpipe) a->dnrpipe = r->dnrpipe; if (r->dnpipe || r->dnrpipe) { if (r->free_flags & PFRULE_DN_IS_PIPE) a->flags |= PFSTATE_DN_IS_PIPE; else a->flags &= ~PFSTATE_DN_IS_PIPE; } if (r->scrub_flags & PFSTATE_SETPRIO) { a->set_prio[0] = r->set_prio[0]; a->set_prio[1] = r->set_prio[1]; } } int pf_socket_lookup(struct pf_pdesc *pd, struct mbuf *m) { struct pf_addr *saddr, *daddr; u_int16_t sport, dport; struct inpcbinfo *pi; struct inpcb *inp; pd->lookup.uid = UID_MAX; pd->lookup.gid = GID_MAX; switch (pd->proto) { case IPPROTO_TCP: sport = pd->hdr.tcp.th_sport; dport = pd->hdr.tcp.th_dport; pi = &V_tcbinfo; break; case IPPROTO_UDP: sport = pd->hdr.udp.uh_sport; dport = pd->hdr.udp.uh_dport; pi = &V_udbinfo; break; default: return (-1); } if (pd->dir == PF_IN) { saddr = pd->src; daddr = pd->dst; } else { u_int16_t p; p = sport; sport = dport; dport = p; saddr = pd->dst; daddr = pd->src; } switch (pd->af) { #ifdef INET case AF_INET: inp = in_pcblookup_mbuf(pi, saddr->v4, sport, daddr->v4, dport, INPLOOKUP_RLOCKPCB, NULL, m); if (inp == NULL) { inp = in_pcblookup_mbuf(pi, saddr->v4, sport, daddr->v4, dport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL, m); if (inp == NULL) return (-1); } break; #endif /* INET */ #ifdef INET6 case AF_INET6: inp = in6_pcblookup_mbuf(pi, &saddr->v6, sport, &daddr->v6, dport, INPLOOKUP_RLOCKPCB, NULL, m); if (inp == NULL) { inp = in6_pcblookup_mbuf(pi, &saddr->v6, sport, &daddr->v6, dport, INPLOOKUP_WILDCARD | INPLOOKUP_RLOCKPCB, NULL, m); if (inp == NULL) return (-1); } break; #endif /* INET6 */ default: return (-1); } INP_RLOCK_ASSERT(inp); pd->lookup.uid = inp->inp_cred->cr_uid; pd->lookup.gid = inp->inp_cred->cr_groups[0]; INP_RUNLOCK(inp); return (1); } u_int8_t pf_get_wscale(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) { int hlen; u_int8_t hdr[60]; u_int8_t *opt, optlen; u_int8_t wscale = 0; hlen = th_off << 2; /* hlen <= sizeof(hdr) */ if (hlen <= sizeof(struct tcphdr)) return (0); if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) return (0); opt = hdr + sizeof(struct tcphdr); hlen -= sizeof(struct tcphdr); while (hlen >= 3) { switch (*opt) { case TCPOPT_EOL: case TCPOPT_NOP: ++opt; --hlen; break; case TCPOPT_WINDOW: wscale = opt[2]; if (wscale > TCP_MAX_WINSHIFT) wscale = TCP_MAX_WINSHIFT; wscale |= PF_WSCALE_FLAG; /* FALLTHROUGH */ default: optlen = opt[1]; if (optlen < 2) optlen = 2; hlen -= optlen; opt += optlen; break; } } return (wscale); } u_int16_t pf_get_mss(struct mbuf *m, int off, u_int16_t th_off, sa_family_t af) { int hlen; u_int8_t hdr[60]; u_int8_t *opt, optlen; u_int16_t mss = V_tcp_mssdflt; hlen = th_off << 2; /* hlen <= sizeof(hdr) */ if (hlen <= sizeof(struct tcphdr)) return (0); if (!pf_pull_hdr(m, off, hdr, hlen, NULL, NULL, af)) return (0); opt = hdr + sizeof(struct tcphdr); hlen -= sizeof(struct tcphdr); while (hlen >= TCPOLEN_MAXSEG) { switch (*opt) { case TCPOPT_EOL: case TCPOPT_NOP: ++opt; --hlen; break; case TCPOPT_MAXSEG: bcopy((caddr_t)(opt + 2), (caddr_t)&mss, 2); NTOHS(mss); /* FALLTHROUGH */ default: optlen = opt[1]; if (optlen < 2) optlen = 2; hlen -= optlen; opt += optlen; break; } } return (mss); } static u_int16_t pf_calc_mss(struct pf_addr *addr, sa_family_t af, int rtableid, u_int16_t offer) { struct nhop_object *nh; #ifdef INET6 struct in6_addr dst6; uint32_t scopeid; #endif /* INET6 */ int hlen = 0; uint16_t mss = 0; NET_EPOCH_ASSERT(); switch (af) { #ifdef INET case AF_INET: hlen = sizeof(struct ip); nh = fib4_lookup(rtableid, addr->v4, 0, 0, 0); if (nh != NULL) mss = nh->nh_mtu - hlen - sizeof(struct tcphdr); break; #endif /* INET */ #ifdef INET6 case AF_INET6: hlen = sizeof(struct ip6_hdr); in6_splitscope(&addr->v6, &dst6, &scopeid); nh = fib6_lookup(rtableid, &dst6, scopeid, 0, 0); if (nh != NULL) mss = nh->nh_mtu - hlen - sizeof(struct tcphdr); break; #endif /* INET6 */ } mss = max(V_tcp_mssdflt, mss); mss = min(mss, offer); mss = max(mss, 64); /* sanity - at least max opt space */ return (mss); } static u_int32_t pf_tcp_iss(struct pf_pdesc *pd) { MD5_CTX ctx; u_int32_t digest[4]; if (V_pf_tcp_secret_init == 0) { arc4random_buf(&V_pf_tcp_secret, sizeof(V_pf_tcp_secret)); MD5Init(&V_pf_tcp_secret_ctx); MD5Update(&V_pf_tcp_secret_ctx, V_pf_tcp_secret, sizeof(V_pf_tcp_secret)); V_pf_tcp_secret_init = 1; } ctx = V_pf_tcp_secret_ctx; MD5Update(&ctx, (char *)&pd->hdr.tcp.th_sport, sizeof(u_short)); MD5Update(&ctx, (char *)&pd->hdr.tcp.th_dport, sizeof(u_short)); if (pd->af == AF_INET6) { MD5Update(&ctx, (char *)&pd->src->v6, sizeof(struct in6_addr)); MD5Update(&ctx, (char *)&pd->dst->v6, sizeof(struct in6_addr)); } else { MD5Update(&ctx, (char *)&pd->src->v4, sizeof(struct in_addr)); MD5Update(&ctx, (char *)&pd->dst->v4, sizeof(struct in_addr)); } MD5Final((u_char *)digest, &ctx); V_pf_tcp_iss_off += 4096; #define ISN_RANDOM_INCREMENT (4096 - 1) return (digest[0] + (arc4random() & ISN_RANDOM_INCREMENT) + V_pf_tcp_iss_off); #undef ISN_RANDOM_INCREMENT } static bool pf_match_eth_addr(const uint8_t *a, const struct pf_keth_rule_addr *r) { bool match = true; /* Always matches if not set */ if (! r->isset) return (!r->neg); for (int i = 0; i < ETHER_ADDR_LEN; i++) { if ((a[i] & r->mask[i]) != (r->addr[i] & r->mask[i])) { match = false; break; } } return (match ^ r->neg); } static int pf_match_eth_tag(struct mbuf *m, struct pf_keth_rule *r, int *tag, int mtag) { if (*tag == -1) *tag = mtag; return ((!r->match_tag_not && r->match_tag == *tag) || (r->match_tag_not && r->match_tag != *tag)); } static void pf_bridge_to(struct ifnet *ifp, struct mbuf *m) { /* If we don't have the interface drop the packet. */ if (ifp == NULL) { m_freem(m); return; } switch (ifp->if_type) { case IFT_ETHER: case IFT_XETHER: case IFT_L2VLAN: case IFT_BRIDGE: case IFT_IEEE8023ADLAG: break; default: m_freem(m); return; } ifp->if_transmit(ifp, m); } static int pf_test_eth_rule(int dir, struct pfi_kkif *kif, struct mbuf **m0) { #ifdef INET struct ip ip; #endif #ifdef INET6 struct ip6_hdr ip6; #endif struct mbuf *m = *m0; struct ether_header *e; struct pf_keth_rule *r, *rm, *a = NULL; struct pf_keth_ruleset *ruleset = NULL; struct pf_mtag *mtag; struct pf_keth_ruleq *rules; struct pf_addr *src = NULL, *dst = NULL; struct pfi_kkif *bridge_to; sa_family_t af = 0; uint16_t proto; int asd = 0, match = 0; int tag = -1; uint8_t action; struct pf_keth_anchor_stackframe anchor_stack[PF_ANCHOR_STACKSIZE]; MPASS(kif->pfik_ifp->if_vnet == curvnet); NET_EPOCH_ASSERT(); PF_RULES_RLOCK_TRACKER; SDT_PROBE3(pf, eth, test_rule, entry, dir, kif->pfik_ifp, m); mtag = pf_find_mtag(m); if (mtag != NULL && mtag->flags & PF_MTAG_FLAG_DUMMYNET) { /* Dummynet re-injects packets after they've * completed their delay. We've already * processed them, so pass unconditionally. */ /* But only once. We may see the packet multiple times (e.g. * PFIL_IN/PFIL_OUT). */ pf_dummynet_flag_remove(m, mtag); return (PF_PASS); } ruleset = V_pf_keth; rules = ck_pr_load_ptr(&ruleset->active.rules); r = TAILQ_FIRST(rules); rm = NULL; if (__predict_false(m->m_len < sizeof(struct ether_header)) && (m = *m0 = m_pullup(*m0, sizeof(struct ether_header))) == NULL) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_test_eth_rule: m_len < sizeof(struct ether_header)" ", pullup failed\n")); return (PF_DROP); } e = mtod(m, struct ether_header *); proto = ntohs(e->ether_type); switch (proto) { #ifdef INET case ETHERTYPE_IP: { if (m_length(m, NULL) < (sizeof(struct ether_header) + sizeof(ip))) return (PF_DROP); af = AF_INET; m_copydata(m, sizeof(struct ether_header), sizeof(ip), (caddr_t)&ip); src = (struct pf_addr *)&ip.ip_src; dst = (struct pf_addr *)&ip.ip_dst; break; } #endif /* INET */ #ifdef INET6 case ETHERTYPE_IPV6: { if (m_length(m, NULL) < (sizeof(struct ether_header) + sizeof(ip6))) return (PF_DROP); af = AF_INET6; m_copydata(m, sizeof(struct ether_header), sizeof(ip6), (caddr_t)&ip6); src = (struct pf_addr *)&ip6.ip6_src; dst = (struct pf_addr *)&ip6.ip6_dst; break; } #endif /* INET6 */ } PF_RULES_RLOCK(); while (r != NULL) { counter_u64_add(r->evaluations, 1); SDT_PROBE2(pf, eth, test_rule, test, r->nr, r); if (pfi_kkif_match(r->kif, kif) == r->ifnot) { SDT_PROBE3(pf, eth, test_rule, mismatch, r->nr, r, "kif"); r = r->skip[PFE_SKIP_IFP].ptr; } else if (r->direction && r->direction != dir) { SDT_PROBE3(pf, eth, test_rule, mismatch, r->nr, r, "dir"); r = r->skip[PFE_SKIP_DIR].ptr; } else if (r->proto && r->proto != proto) { SDT_PROBE3(pf, eth, test_rule, mismatch, r->nr, r, "proto"); r = r->skip[PFE_SKIP_PROTO].ptr; } else if (! pf_match_eth_addr(e->ether_shost, &r->src)) { SDT_PROBE3(pf, eth, test_rule, mismatch, r->nr, r, "src"); r = r->skip[PFE_SKIP_SRC_ADDR].ptr; } else if (! pf_match_eth_addr(e->ether_dhost, &r->dst)) { SDT_PROBE3(pf, eth, test_rule, mismatch, r->nr, r, "dst"); r = r->skip[PFE_SKIP_DST_ADDR].ptr; } else if (src != NULL && PF_MISMATCHAW(&r->ipsrc.addr, src, af, r->ipsrc.neg, kif, M_GETFIB(m))) { SDT_PROBE3(pf, eth, test_rule, mismatch, r->nr, r, "ip_src"); r = r->skip[PFE_SKIP_SRC_IP_ADDR].ptr; } else if (dst != NULL && PF_MISMATCHAW(&r->ipdst.addr, dst, af, r->ipdst.neg, kif, M_GETFIB(m))) { SDT_PROBE3(pf, eth, test_rule, mismatch, r->nr, r, "ip_dst"); r = r->skip[PFE_SKIP_DST_IP_ADDR].ptr; } else if (r->match_tag && !pf_match_eth_tag(m, r, &tag, mtag ? mtag->tag : 0)) { SDT_PROBE3(pf, eth, test_rule, mismatch, r->nr, r, "match_tag"); r = TAILQ_NEXT(r, entries); } else { if (r->tag) tag = r->tag; if (r->anchor == NULL) { /* Rule matches */ rm = r; SDT_PROBE2(pf, eth, test_rule, match, r->nr, r); if (r->quick) break; r = TAILQ_NEXT(r, entries); } else { pf_step_into_keth_anchor(anchor_stack, &asd, &ruleset, &r, &a, &match); } } if (r == NULL && pf_step_out_of_keth_anchor(anchor_stack, &asd, &ruleset, &r, &a, &match)) break; } r = rm; SDT_PROBE2(pf, eth, test_rule, final_match, (r != NULL ? r->nr : -1), r); /* Default to pass. */ if (r == NULL) { PF_RULES_RUNLOCK(); return (PF_PASS); } /* Execute action. */ counter_u64_add(r->packets[dir == PF_OUT], 1); counter_u64_add(r->bytes[dir == PF_OUT], m_length(m, NULL)); pf_update_timestamp(r); /* Shortcut. Don't tag if we're just going to drop anyway. */ if (r->action == PF_DROP) { PF_RULES_RUNLOCK(); return (PF_DROP); } if (tag > 0) { if (mtag == NULL) mtag = pf_get_mtag(m); if (mtag == NULL) { PF_RULES_RUNLOCK(); counter_u64_add(V_pf_status.counters[PFRES_MEMORY], 1); return (PF_DROP); } mtag->tag = tag; } if (r->qid != 0) { if (mtag == NULL) mtag = pf_get_mtag(m); if (mtag == NULL) { PF_RULES_RUNLOCK(); counter_u64_add(V_pf_status.counters[PFRES_MEMORY], 1); return (PF_DROP); } mtag->qid = r->qid; } action = r->action; bridge_to = r->bridge_to; /* Dummynet */ if (r->dnpipe) { struct ip_fw_args dnflow; /* Drop packet if dummynet is not loaded. */ if (ip_dn_io_ptr == NULL) { PF_RULES_RUNLOCK(); m_freem(m); counter_u64_add(V_pf_status.counters[PFRES_MEMORY], 1); return (PF_DROP); } if (mtag == NULL) mtag = pf_get_mtag(m); if (mtag == NULL) { PF_RULES_RUNLOCK(); counter_u64_add(V_pf_status.counters[PFRES_MEMORY], 1); return (PF_DROP); } bzero(&dnflow, sizeof(dnflow)); /* We don't have port numbers here, so we set 0. That means * that we'll be somewhat limited in distinguishing flows (i.e. * only based on IP addresses, not based on port numbers), but * it's better than nothing. */ dnflow.f_id.dst_port = 0; dnflow.f_id.src_port = 0; dnflow.f_id.proto = 0; dnflow.rule.info = r->dnpipe; dnflow.rule.info |= IPFW_IS_DUMMYNET; if (r->dnflags & PFRULE_DN_IS_PIPE) dnflow.rule.info |= IPFW_IS_PIPE; dnflow.f_id.extra = dnflow.rule.info; dnflow.flags = dir == PF_IN ? IPFW_ARGS_IN : IPFW_ARGS_OUT; dnflow.flags |= IPFW_ARGS_ETHER; dnflow.ifp = kif->pfik_ifp; switch (af) { case AF_INET: dnflow.f_id.addr_type = 4; dnflow.f_id.src_ip = src->v4.s_addr; dnflow.f_id.dst_ip = dst->v4.s_addr; break; case AF_INET6: dnflow.flags |= IPFW_ARGS_IP6; dnflow.f_id.addr_type = 6; dnflow.f_id.src_ip6 = src->v6; dnflow.f_id.dst_ip6 = dst->v6; break; } PF_RULES_RUNLOCK(); mtag->flags |= PF_MTAG_FLAG_DUMMYNET; ip_dn_io_ptr(m0, &dnflow); if (*m0 != NULL) pf_dummynet_flag_remove(m, mtag); } else { PF_RULES_RUNLOCK(); } if (action == PF_PASS && bridge_to) { pf_bridge_to(bridge_to->pfik_ifp, *m0); *m0 = NULL; /* We've eaten the packet. */ } return (action); } +#define PF_TEST_ATTRIB(t, a)\ + do { \ + if (t) { \ + r = a; \ + goto nextrule; \ + } \ + } while (0) + static int pf_test_rule(struct pf_krule **rm, struct pf_kstate **sm, struct pfi_kkif *kif, struct mbuf *m, int off, struct pf_pdesc *pd, struct pf_krule **am, struct pf_kruleset **rsm, struct inpcb *inp, int hdrlen) { struct pf_krule *nr = NULL; struct pf_addr * const saddr = pd->src; struct pf_addr * const daddr = pd->dst; sa_family_t af = pd->af; struct pf_krule *r, *a = NULL; struct pf_kruleset *ruleset = NULL; struct pf_krule_slist match_rules; struct pf_krule_item *ri; struct pf_ksrc_node *nsn = NULL; struct tcphdr *th = &pd->hdr.tcp; struct pf_state_key *sk = NULL, *nk = NULL; u_short reason, transerror; int rewrite = 0; int tag = -1; int asd = 0; int match = 0; int state_icmp = 0, icmp_dir, multi; u_int16_t sport = 0, dport = 0, virtual_type, virtual_id; u_int16_t bproto_sum = 0, bip_sum = 0; u_int8_t icmptype = 0, icmpcode = 0; struct pf_kanchor_stackframe anchor_stack[PF_ANCHOR_STACKSIZE]; struct pf_udp_mapping *udp_mapping = NULL; PF_RULES_RASSERT(); SLIST_INIT(&match_rules); if (inp != NULL) { INP_LOCK_ASSERT(inp); pd->lookup.uid = inp->inp_cred->cr_uid; pd->lookup.gid = inp->inp_cred->cr_groups[0]; pd->lookup.done = 1; } - switch (pd->proto) { + switch (pd->virtual_proto) { case IPPROTO_TCP: sport = th->th_sport; dport = th->th_dport; break; case IPPROTO_UDP: sport = pd->hdr.udp.uh_sport; dport = pd->hdr.udp.uh_dport; break; case IPPROTO_SCTP: sport = pd->hdr.sctp.src_port; dport = pd->hdr.sctp.dest_port; break; #ifdef INET case IPPROTO_ICMP: MPASS(af == AF_INET); icmptype = pd->hdr.icmp.icmp_type; icmpcode = pd->hdr.icmp.icmp_code; state_icmp = pf_icmp_mapping(pd, icmptype, &icmp_dir, &multi, &virtual_id, &virtual_type); if (icmp_dir == PF_IN) { sport = virtual_id; dport = virtual_type; } else { sport = virtual_type; dport = virtual_id; } break; #endif /* INET */ #ifdef INET6 case IPPROTO_ICMPV6: MPASS(af == AF_INET6); icmptype = pd->hdr.icmp6.icmp6_type; icmpcode = pd->hdr.icmp6.icmp6_code; state_icmp = pf_icmp_mapping(pd, icmptype, &icmp_dir, &multi, &virtual_id, &virtual_type); if (icmp_dir == PF_IN) { sport = virtual_id; dport = virtual_type; } else { sport = virtual_type; dport = virtual_id; } break; #endif /* INET6 */ default: sport = dport = 0; break; } r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); /* check packet for BINAT/NAT/RDR */ transerror = pf_get_translation(pd, m, off, kif, &nsn, &sk, &nk, saddr, daddr, sport, dport, anchor_stack, &nr, &udp_mapping); switch (transerror) { default: /* A translation error occurred. */ REASON_SET(&reason, transerror); goto cleanup; case PFRES_MAX: /* No match. */ break; case PFRES_MATCH: KASSERT(sk != NULL, ("%s: null sk", __func__)); KASSERT(nk != NULL, ("%s: null nk", __func__)); if (nr->log) { PFLOG_PACKET(kif, m, PF_PASS, PFRES_MATCH, nr, a, ruleset, pd, 1); } if (pd->ip_sum) bip_sum = *pd->ip_sum; switch (pd->proto) { case IPPROTO_TCP: bproto_sum = th->th_sum; pd->proto_sum = &th->th_sum; if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) || nk->port[pd->sidx] != sport) { pf_change_ap(m, saddr, &th->th_sport, pd->ip_sum, &th->th_sum, &nk->addr[pd->sidx], nk->port[pd->sidx], 0, af); pd->sport = &th->th_sport; sport = th->th_sport; } if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) || nk->port[pd->didx] != dport) { pf_change_ap(m, daddr, &th->th_dport, pd->ip_sum, &th->th_sum, &nk->addr[pd->didx], nk->port[pd->didx], 0, af); dport = th->th_dport; pd->dport = &th->th_dport; } rewrite++; break; case IPPROTO_UDP: bproto_sum = pd->hdr.udp.uh_sum; pd->proto_sum = &pd->hdr.udp.uh_sum; if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) || nk->port[pd->sidx] != sport) { pf_change_ap(m, saddr, &pd->hdr.udp.uh_sport, pd->ip_sum, &pd->hdr.udp.uh_sum, &nk->addr[pd->sidx], nk->port[pd->sidx], 1, af); sport = pd->hdr.udp.uh_sport; pd->sport = &pd->hdr.udp.uh_sport; } if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) || nk->port[pd->didx] != dport) { pf_change_ap(m, daddr, &pd->hdr.udp.uh_dport, pd->ip_sum, &pd->hdr.udp.uh_sum, &nk->addr[pd->didx], nk->port[pd->didx], 1, af); dport = pd->hdr.udp.uh_dport; pd->dport = &pd->hdr.udp.uh_dport; } rewrite++; break; case IPPROTO_SCTP: { uint16_t checksum = 0; if (PF_ANEQ(saddr, &nk->addr[pd->sidx], af) || nk->port[pd->sidx] != sport) { pf_change_ap(m, saddr, &pd->hdr.sctp.src_port, pd->ip_sum, &checksum, &nk->addr[pd->sidx], nk->port[pd->sidx], 1, af); } if (PF_ANEQ(daddr, &nk->addr[pd->didx], af) || nk->port[pd->didx] != dport) { pf_change_ap(m, daddr, &pd->hdr.sctp.dest_port, pd->ip_sum, &checksum, &nk->addr[pd->didx], nk->port[pd->didx], 1, af); } break; } #ifdef INET case IPPROTO_ICMP: if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET)) pf_change_a(&saddr->v4.s_addr, pd->ip_sum, nk->addr[pd->sidx].v4.s_addr, 0); if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET)) pf_change_a(&daddr->v4.s_addr, pd->ip_sum, nk->addr[pd->didx].v4.s_addr, 0); if (virtual_type == htons(ICMP_ECHO) && nk->port[pd->sidx] != pd->hdr.icmp.icmp_id) { pd->hdr.icmp.icmp_cksum = pf_cksum_fixup( pd->hdr.icmp.icmp_cksum, sport, nk->port[pd->sidx], 0); pd->hdr.icmp.icmp_id = nk->port[pd->sidx]; pd->sport = &pd->hdr.icmp.icmp_id; } m_copyback(m, off, ICMP_MINLEN, (caddr_t)&pd->hdr.icmp); break; #endif /* INET */ #ifdef INET6 case IPPROTO_ICMPV6: if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6)) pf_change_a6(saddr, &pd->hdr.icmp6.icmp6_cksum, &nk->addr[pd->sidx], 0); if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6)) pf_change_a6(daddr, &pd->hdr.icmp6.icmp6_cksum, &nk->addr[pd->didx], 0); rewrite++; break; #endif /* INET */ default: switch (af) { #ifdef INET case AF_INET: if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET)) pf_change_a(&saddr->v4.s_addr, pd->ip_sum, nk->addr[pd->sidx].v4.s_addr, 0); if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET)) pf_change_a(&daddr->v4.s_addr, pd->ip_sum, nk->addr[pd->didx].v4.s_addr, 0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (PF_ANEQ(saddr, &nk->addr[pd->sidx], AF_INET6)) PF_ACPY(saddr, &nk->addr[pd->sidx], af); if (PF_ANEQ(daddr, &nk->addr[pd->didx], AF_INET6)) PF_ACPY(daddr, &nk->addr[pd->didx], af); break; #endif /* INET */ } break; } if (nr->natpass) r = NULL; pd->nat_rule = nr; } while (r != NULL) { pf_counter_u64_add(&r->evaluations, 1); - if (pfi_kkif_match(r->kif, kif) == r->ifnot) - r = r->skip[PF_SKIP_IFP].ptr; - else if (r->direction && r->direction != pd->dir) - r = r->skip[PF_SKIP_DIR].ptr; - else if (r->af && r->af != af) - r = r->skip[PF_SKIP_AF].ptr; - else if (r->proto && r->proto != pd->proto) - r = r->skip[PF_SKIP_PROTO].ptr; - else if (PF_MISMATCHAW(&r->src.addr, saddr, af, - r->src.neg, kif, M_GETFIB(m))) - r = r->skip[PF_SKIP_SRC_ADDR].ptr; - /* tcp/udp only. port_op always 0 in other cases */ - else if (r->src.port_op && !pf_match_port(r->src.port_op, - r->src.port[0], r->src.port[1], sport)) - r = r->skip[PF_SKIP_SRC_PORT].ptr; - else if (PF_MISMATCHAW(&r->dst.addr, daddr, af, - r->dst.neg, NULL, M_GETFIB(m))) - r = r->skip[PF_SKIP_DST_ADDR].ptr; - /* tcp/udp only. port_op always 0 in other cases */ - else if (r->dst.port_op && !pf_match_port(r->dst.port_op, - r->dst.port[0], r->dst.port[1], dport)) - r = r->skip[PF_SKIP_DST_PORT].ptr; - /* icmp only. type always 0 in other cases */ - else if (r->type && r->type != icmptype + 1) - r = TAILQ_NEXT(r, entries); - /* icmp only. type always 0 in other cases */ - else if (r->code && r->code != icmpcode + 1) - r = TAILQ_NEXT(r, entries); - else if (r->tos && !(r->tos == pd->tos)) - r = TAILQ_NEXT(r, entries); - else if (r->rule_flag & PFRULE_FRAGMENT) - r = TAILQ_NEXT(r, entries); - else if (pd->proto == IPPROTO_TCP && - (r->flagset & th->th_flags) != r->flags) - r = TAILQ_NEXT(r, entries); - /* tcp/udp only. uid.op always 0 in other cases */ - else if (r->uid.op && (pd->lookup.done || (pd->lookup.done = - pf_socket_lookup(pd, m), 1)) && - !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], - pd->lookup.uid)) - r = TAILQ_NEXT(r, entries); - /* tcp/udp only. gid.op always 0 in other cases */ - else if (r->gid.op && (pd->lookup.done || (pd->lookup.done = - pf_socket_lookup(pd, m), 1)) && - !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], - pd->lookup.gid)) - r = TAILQ_NEXT(r, entries); - else if (r->prio && - !pf_match_ieee8021q_pcp(r->prio, m)) - r = TAILQ_NEXT(r, entries); - else if (r->prob && - r->prob <= arc4random()) - r = TAILQ_NEXT(r, entries); - else if (r->match_tag && !pf_match_tag(m, r, &tag, - pd->pf_mtag ? pd->pf_mtag->tag : 0)) - r = TAILQ_NEXT(r, entries); - else if (r->rcv_kif && !pf_match_rcvif(m, r)) - r = TAILQ_NEXT(r, entries); - else if (r->os_fingerprint != PF_OSFP_ANY && - (pd->proto != IPPROTO_TCP || !pf_osfp_match( + PF_TEST_ATTRIB(pfi_kkif_match(r->kif, kif) == r->ifnot, + r->skip[PF_SKIP_IFP].ptr); + PF_TEST_ATTRIB(r->direction && r->direction != pd->dir, + r->skip[PF_SKIP_DIR].ptr); + PF_TEST_ATTRIB(r->af && r->af != af, + r->skip[PF_SKIP_AF].ptr); + PF_TEST_ATTRIB(r->proto && r->proto != pd->proto, + r->skip[PF_SKIP_PROTO].ptr); + PF_TEST_ATTRIB(PF_MISMATCHAW(&r->src.addr, saddr, af, + r->src.neg, kif, M_GETFIB(m)), + r->skip[PF_SKIP_SRC_ADDR].ptr); + PF_TEST_ATTRIB(PF_MISMATCHAW(&r->dst.addr, daddr, af, + r->dst.neg, NULL, M_GETFIB(m)), + r->skip[PF_SKIP_DST_ADDR].ptr); + switch (pd->virtual_proto) { + case PF_VPROTO_FRAGMENT: + /* tcp/udp only. port_op always 0 in other cases */ + PF_TEST_ATTRIB((r->src.port_op || r->dst.port_op), + TAILQ_NEXT(r, entries)); + PF_TEST_ATTRIB((pd->proto == IPPROTO_TCP && r->flagset), + TAILQ_NEXT(r, entries)); + /* icmp only. type/code always 0 in other cases */ + PF_TEST_ATTRIB((r->type || r->code), + TAILQ_NEXT(r, entries)); + /* tcp/udp only. {uid|gid}.op always 0 in other cases */ + PF_TEST_ATTRIB((r->gid.op || r->uid.op), + TAILQ_NEXT(r, entries)); + break; + + case IPPROTO_TCP: + PF_TEST_ATTRIB((r->flagset & th->th_flags) != r->flags, + TAILQ_NEXT(r, entries)); + /* FALLTHROUGH */ + case IPPROTO_SCTP: + case IPPROTO_UDP: + /* tcp/udp only. port_op always 0 in other cases */ + PF_TEST_ATTRIB(r->src.port_op && !pf_match_port(r->src.port_op, + r->src.port[0], r->src.port[1], sport), + r->skip[PF_SKIP_SRC_PORT].ptr); + /* tcp/udp only. port_op always 0 in other cases */ + PF_TEST_ATTRIB(r->dst.port_op && !pf_match_port(r->dst.port_op, + r->dst.port[0], r->dst.port[1], dport), + r->skip[PF_SKIP_DST_PORT].ptr); + /* tcp/udp only. uid.op always 0 in other cases */ + PF_TEST_ATTRIB(r->uid.op && (pd->lookup.done || (pd->lookup.done = + pf_socket_lookup(pd, m), 1)) && + !pf_match_uid(r->uid.op, r->uid.uid[0], r->uid.uid[1], + pd->lookup.uid), + TAILQ_NEXT(r, entries)); + /* tcp/udp only. gid.op always 0 in other cases */ + PF_TEST_ATTRIB(r->gid.op && (pd->lookup.done || (pd->lookup.done = + pf_socket_lookup(pd, m), 1)) && + !pf_match_gid(r->gid.op, r->gid.gid[0], r->gid.gid[1], + pd->lookup.gid), + TAILQ_NEXT(r, entries)); + break; + + case IPPROTO_ICMP: + case IPPROTO_ICMPV6: + /* icmp only. type always 0 in other cases */ + PF_TEST_ATTRIB(r->type && r->type != icmptype + 1, + TAILQ_NEXT(r, entries)); + /* icmp only. type always 0 in other cases */ + PF_TEST_ATTRIB(r->code && r->code != icmpcode + 1, + TAILQ_NEXT(r, entries)); + break; + + default: + break; + } + PF_TEST_ATTRIB(r->tos && !(r->tos == pd->tos), + TAILQ_NEXT(r, entries)); + PF_TEST_ATTRIB(r->prio && + !pf_match_ieee8021q_pcp(r->prio, m), + TAILQ_NEXT(r, entries)); + PF_TEST_ATTRIB(r->prob && + r->prob <= arc4random(), + TAILQ_NEXT(r, entries)); + PF_TEST_ATTRIB(r->match_tag && !pf_match_tag(m, r, &tag, + pd->pf_mtag ? pd->pf_mtag->tag : 0), + TAILQ_NEXT(r, entries)); + PF_TEST_ATTRIB(r->rcv_kif && !pf_match_rcvif(m, r), + TAILQ_NEXT(r, entries)); + PF_TEST_ATTRIB((r->rule_flag & PFRULE_FRAGMENT && + pd->virtual_proto != PF_VPROTO_FRAGMENT), + TAILQ_NEXT(r, entries)); + PF_TEST_ATTRIB(r->os_fingerprint != PF_OSFP_ANY && + (pd->virtual_proto != IPPROTO_TCP || !pf_osfp_match( pf_osfp_fingerprint(pd, m, off, th), - r->os_fingerprint))) - r = TAILQ_NEXT(r, entries); - else { - if (r->tag) - tag = r->tag; - if (r->anchor == NULL) { - if (r->action == PF_MATCH) { - ri = malloc(sizeof(struct pf_krule_item), M_PF_RULE_ITEM, M_NOWAIT | M_ZERO); - if (ri == NULL) { - REASON_SET(&reason, PFRES_MEMORY); - goto cleanup; - } - ri->r = r; - SLIST_INSERT_HEAD(&match_rules, ri, entry); - pf_counter_u64_critical_enter(); - pf_counter_u64_add_protected(&r->packets[pd->dir == PF_OUT], 1); - pf_counter_u64_add_protected(&r->bytes[pd->dir == PF_OUT], pd->tot_len); - pf_counter_u64_critical_exit(); - pf_rule_to_actions(r, &pd->act); - if (r->log || pd->act.log & PF_LOG_MATCHES) - PFLOG_PACKET(kif, m, - r->action, PFRES_MATCH, r, - a, ruleset, pd, 1); - } else { - match = 1; - *rm = r; - *am = a; - *rsm = ruleset; - if (pd->act.log & PF_LOG_MATCHES) - PFLOG_PACKET(kif, m, - r->action, PFRES_MATCH, r, - a, ruleset, pd, 1); + r->os_fingerprint)), + TAILQ_NEXT(r, entries)); + /* FALLTHROUGH */ + if (r->tag) + tag = r->tag; + if (r->anchor == NULL) { + if (r->action == PF_MATCH) { + ri = malloc(sizeof(struct pf_krule_item), M_PF_RULE_ITEM, M_NOWAIT | M_ZERO); + if (ri == NULL) { + REASON_SET(&reason, PFRES_MEMORY); + goto cleanup; } - if ((*rm)->quick) - break; - r = TAILQ_NEXT(r, entries); - } else - pf_step_into_anchor(anchor_stack, &asd, - &ruleset, PF_RULESET_FILTER, &r, &a, - &match); - } + ri->r = r; + SLIST_INSERT_HEAD(&match_rules, ri, entry); + pf_counter_u64_critical_enter(); + pf_counter_u64_add_protected(&r->packets[pd->dir == PF_OUT], 1); + pf_counter_u64_add_protected(&r->bytes[pd->dir == PF_OUT], pd->tot_len); + pf_counter_u64_critical_exit(); + pf_rule_to_actions(r, &pd->act); + if (r->log || pd->act.log & PF_LOG_MATCHES) + PFLOG_PACKET(kif, m, + r->action, PFRES_MATCH, r, + a, ruleset, pd, 1); + } else { + match = 1; + *rm = r; + *am = a; + *rsm = ruleset; + if (pd->act.log & PF_LOG_MATCHES) + PFLOG_PACKET(kif, m, + r->action, PFRES_MATCH, r, + a, ruleset, pd, 1); + } + if ((*rm)->quick) + break; + r = TAILQ_NEXT(r, entries); + } else + pf_step_into_anchor(anchor_stack, &asd, + &ruleset, PF_RULESET_FILTER, &r, &a, + &match); +nextrule: if (r == NULL && pf_step_out_of_anchor(anchor_stack, &asd, &ruleset, PF_RULESET_FILTER, &r, &a, &match)) break; } r = *rm; a = *am; ruleset = *rsm; REASON_SET(&reason, PFRES_MATCH); /* apply actions for last matching pass/block rule */ pf_rule_to_actions(r, &pd->act); if (r->log || pd->act.log & PF_LOG_MATCHES) { if (rewrite) m_copyback(m, off, hdrlen, pd->hdr.any); PFLOG_PACKET(kif, m, r->action, reason, r, a, ruleset, pd, 1); } - if ((r->action == PF_DROP) && + if (pd->virtual_proto != PF_VPROTO_FRAGMENT && + (r->action == PF_DROP) && ((r->rule_flag & PFRULE_RETURNRST) || (r->rule_flag & PFRULE_RETURNICMP) || (r->rule_flag & PFRULE_RETURN))) { pf_return(r, nr, pd, sk, off, m, th, kif, bproto_sum, bip_sum, hdrlen, &reason, r->rtableid); } if (r->action == PF_DROP) goto cleanup; if (tag > 0 && pf_tag_packet(m, pd, tag)) { REASON_SET(&reason, PFRES_MEMORY); goto cleanup; } if (pd->act.rtableid >= 0) M_SETFIB(m, pd->act.rtableid); - if (!state_icmp && (r->keep_state || nr != NULL || - (pd->flags & PFDESC_TCP_NORM))) { + if (pd->virtual_proto != PF_VPROTO_FRAGMENT && + (!state_icmp && (r->keep_state || nr != NULL || + (pd->flags & PFDESC_TCP_NORM)))) { int action; action = pf_create_state(r, nr, a, pd, nsn, nk, sk, m, off, sport, dport, &rewrite, kif, sm, tag, bproto_sum, bip_sum, hdrlen, &match_rules, udp_mapping); if (action != PF_PASS) { pf_udp_mapping_release(udp_mapping); if (action == PF_DROP && (r->rule_flag & PFRULE_RETURN)) pf_return(r, nr, pd, sk, off, m, th, kif, bproto_sum, bip_sum, hdrlen, &reason, pd->act.rtableid); return (action); } } else { while ((ri = SLIST_FIRST(&match_rules))) { SLIST_REMOVE_HEAD(&match_rules, entry); free(ri, M_PF_RULE_ITEM); } uma_zfree(V_pf_state_key_z, sk); uma_zfree(V_pf_state_key_z, nk); pf_udp_mapping_release(udp_mapping); } /* copy back packet headers if we performed NAT operations */ if (rewrite) m_copyback(m, off, hdrlen, pd->hdr.any); if (*sm != NULL && !((*sm)->state_flags & PFSTATE_NOSYNC) && pd->dir == PF_OUT && V_pfsync_defer_ptr != NULL && V_pfsync_defer_ptr(*sm, m)) /* * We want the state created, but we dont * want to send this in case a partner * firewall has to know about it to allow * replies through it. */ return (PF_DEFER); return (PF_PASS); cleanup: while ((ri = SLIST_FIRST(&match_rules))) { SLIST_REMOVE_HEAD(&match_rules, entry); free(ri, M_PF_RULE_ITEM); } uma_zfree(V_pf_state_key_z, sk); uma_zfree(V_pf_state_key_z, nk); pf_udp_mapping_release(udp_mapping); return (PF_DROP); } static int pf_create_state(struct pf_krule *r, struct pf_krule *nr, struct pf_krule *a, struct pf_pdesc *pd, struct pf_ksrc_node *nsn, struct pf_state_key *nk, struct pf_state_key *sk, struct mbuf *m, int off, u_int16_t sport, u_int16_t dport, int *rewrite, struct pfi_kkif *kif, struct pf_kstate **sm, int tag, u_int16_t bproto_sum, u_int16_t bip_sum, int hdrlen, struct pf_krule_slist *match_rules, struct pf_udp_mapping *udp_mapping) { struct pf_kstate *s = NULL; struct pf_ksrc_node *sn = NULL; struct tcphdr *th = &pd->hdr.tcp; u_int16_t mss = V_tcp_mssdflt; u_short reason, sn_reason; struct pf_krule_item *ri; /* check maximums */ if (r->max_states && (counter_u64_fetch(r->states_cur) >= r->max_states)) { counter_u64_add(V_pf_status.lcounters[LCNT_STATES], 1); REASON_SET(&reason, PFRES_MAXSTATES); goto csfailed; } /* src node for filter rule */ if ((r->rule_flag & PFRULE_SRCTRACK || r->rpool.opts & PF_POOL_STICKYADDR) && (sn_reason = pf_insert_src_node(&sn, r, pd->src, pd->af)) != 0) { REASON_SET(&reason, sn_reason); goto csfailed; } /* src node for translation rule */ if (nr != NULL && (nr->rpool.opts & PF_POOL_STICKYADDR) && (sn_reason = pf_insert_src_node(&nsn, nr, &sk->addr[pd->sidx], pd->af)) != 0 ) { REASON_SET(&reason, sn_reason); goto csfailed; } s = pf_alloc_state(M_NOWAIT); if (s == NULL) { REASON_SET(&reason, PFRES_MEMORY); goto csfailed; } s->rule.ptr = r; s->nat_rule.ptr = nr; s->anchor.ptr = a; bcopy(match_rules, &s->match_rules, sizeof(s->match_rules)); memcpy(&s->act, &pd->act, sizeof(struct pf_rule_actions)); STATE_INC_COUNTERS(s); if (r->allow_opts) s->state_flags |= PFSTATE_ALLOWOPTS; if (r->rule_flag & PFRULE_STATESLOPPY) s->state_flags |= PFSTATE_SLOPPY; if (pd->flags & PFDESC_TCP_NORM) /* Set by old-style scrub rules */ s->state_flags |= PFSTATE_SCRUB_TCP; if ((r->rule_flag & PFRULE_PFLOW) || (nr != NULL && nr->rule_flag & PFRULE_PFLOW)) s->state_flags |= PFSTATE_PFLOW; s->act.log = pd->act.log & PF_LOG_ALL; s->sync_state = PFSYNC_S_NONE; s->state_flags |= pd->act.flags; /* Only needed for pfsync and state export */ if (nr != NULL) s->act.log |= nr->log & PF_LOG_ALL; switch (pd->proto) { case IPPROTO_TCP: s->src.seqlo = ntohl(th->th_seq); s->src.seqhi = s->src.seqlo + pd->p_len + 1; if ((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && r->keep_state == PF_STATE_MODULATE) { /* Generate sequence number modulator */ if ((s->src.seqdiff = pf_tcp_iss(pd) - s->src.seqlo) == 0) s->src.seqdiff = 1; pf_change_proto_a(m, &th->th_seq, &th->th_sum, htonl(s->src.seqlo + s->src.seqdiff), 0); *rewrite = 1; } else s->src.seqdiff = 0; if (th->th_flags & TH_SYN) { s->src.seqhi++; s->src.wscale = pf_get_wscale(m, off, th->th_off, pd->af); } s->src.max_win = MAX(ntohs(th->th_win), 1); if (s->src.wscale & PF_WSCALE_MASK) { /* Remove scale factor from initial window */ int win = s->src.max_win; win += 1 << (s->src.wscale & PF_WSCALE_MASK); s->src.max_win = (win - 1) >> (s->src.wscale & PF_WSCALE_MASK); } if (th->th_flags & TH_FIN) s->src.seqhi++; s->dst.seqhi = 1; s->dst.max_win = 1; pf_set_protostate(s, PF_PEER_SRC, TCPS_SYN_SENT); pf_set_protostate(s, PF_PEER_DST, TCPS_CLOSED); s->timeout = PFTM_TCP_FIRST_PACKET; atomic_add_32(&V_pf_status.states_halfopen, 1); break; case IPPROTO_UDP: pf_set_protostate(s, PF_PEER_SRC, PFUDPS_SINGLE); pf_set_protostate(s, PF_PEER_DST, PFUDPS_NO_TRAFFIC); s->timeout = PFTM_UDP_FIRST_PACKET; break; case IPPROTO_SCTP: pf_set_protostate(s, PF_PEER_SRC, SCTP_COOKIE_WAIT); pf_set_protostate(s, PF_PEER_DST, SCTP_CLOSED); s->timeout = PFTM_SCTP_FIRST_PACKET; break; case IPPROTO_ICMP: #ifdef INET6 case IPPROTO_ICMPV6: #endif s->timeout = PFTM_ICMP_FIRST_PACKET; break; default: pf_set_protostate(s, PF_PEER_SRC, PFOTHERS_SINGLE); pf_set_protostate(s, PF_PEER_DST, PFOTHERS_NO_TRAFFIC); s->timeout = PFTM_OTHER_FIRST_PACKET; } if (r->rt) { /* pf_map_addr increases the reason counters */ if ((reason = pf_map_addr_sn(pd->af, r, pd->src, &s->rt_addr, &s->rt_kif, NULL, &sn)) != 0) goto csfailed; s->rt = r->rt; } s->creation = s->expire = pf_get_uptime(); if (sn != NULL) s->src_node = sn; if (nsn != NULL) { /* XXX We only modify one side for now. */ PF_ACPY(&nsn->raddr, &nk->addr[1], pd->af); s->nat_src_node = nsn; } if (pd->proto == IPPROTO_TCP) { if (s->state_flags & PFSTATE_SCRUB_TCP && pf_normalize_tcp_init(m, off, pd, th, &s->src, &s->dst)) { REASON_SET(&reason, PFRES_MEMORY); goto csfailed; } if (s->state_flags & PFSTATE_SCRUB_TCP && s->src.scrub && pf_normalize_tcp_stateful(m, off, pd, &reason, th, s, &s->src, &s->dst, rewrite)) { /* This really shouldn't happen!!! */ DPFPRINTF(PF_DEBUG_URGENT, ("pf_normalize_tcp_stateful failed on first " "pkt\n")); goto csfailed; } } else if (pd->proto == IPPROTO_SCTP) { if (pf_normalize_sctp_init(m, off, pd, &s->src, &s->dst)) goto csfailed; if (! (pd->sctp_flags & (PFDESC_SCTP_INIT | PFDESC_SCTP_ADD_IP))) goto csfailed; } s->direction = pd->dir; /* * sk/nk could already been setup by pf_get_translation(). */ if (nr == NULL) { KASSERT((sk == NULL && nk == NULL), ("%s: nr %p sk %p, nk %p", __func__, nr, sk, nk)); sk = pf_state_key_setup(pd, m, off, pd->src, pd->dst, sport, dport); if (sk == NULL) goto csfailed; nk = sk; } else KASSERT((sk != NULL && nk != NULL), ("%s: nr %p sk %p, nk %p", __func__, nr, sk, nk)); /* Swap sk/nk for PF_OUT. */ if (pf_state_insert(BOUND_IFACE(s, kif), kif, (pd->dir == PF_IN) ? sk : nk, (pd->dir == PF_IN) ? nk : sk, s)) { REASON_SET(&reason, PFRES_STATEINS); goto drop; } else *sm = s; if (tag > 0) s->tag = tag; if (pd->proto == IPPROTO_TCP && (th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN && r->keep_state == PF_STATE_SYNPROXY) { pf_set_protostate(s, PF_PEER_SRC, PF_TCPS_PROXY_SRC); /* undo NAT changes, if they have taken place */ if (nr != NULL) { struct pf_state_key *skt = s->key[PF_SK_WIRE]; if (pd->dir == PF_OUT) skt = s->key[PF_SK_STACK]; PF_ACPY(pd->src, &skt->addr[pd->sidx], pd->af); PF_ACPY(pd->dst, &skt->addr[pd->didx], pd->af); if (pd->sport) *pd->sport = skt->port[pd->sidx]; if (pd->dport) *pd->dport = skt->port[pd->didx]; if (pd->proto_sum) *pd->proto_sum = bproto_sum; if (pd->ip_sum) *pd->ip_sum = bip_sum; m_copyback(m, off, hdrlen, pd->hdr.any); } s->src.seqhi = htonl(arc4random()); /* Find mss option */ int rtid = M_GETFIB(m); mss = pf_get_mss(m, off, th->th_off, pd->af); mss = pf_calc_mss(pd->src, pd->af, rtid, mss); mss = pf_calc_mss(pd->dst, pd->af, rtid, mss); s->src.mss = mss; pf_send_tcp(r, pd->af, pd->dst, pd->src, th->th_dport, th->th_sport, s->src.seqhi, ntohl(th->th_seq) + 1, TH_SYN|TH_ACK, 0, s->src.mss, 0, true, 0, 0, pd->act.rtableid); REASON_SET(&reason, PFRES_SYNPROXY); return (PF_SYNPROXY_DROP); } s->udp_mapping = udp_mapping; return (PF_PASS); csfailed: while ((ri = SLIST_FIRST(match_rules))) { SLIST_REMOVE_HEAD(match_rules, entry); free(ri, M_PF_RULE_ITEM); } uma_zfree(V_pf_state_key_z, sk); uma_zfree(V_pf_state_key_z, nk); if (sn != NULL) { PF_SRC_NODE_LOCK(sn); if (--sn->states == 0 && sn->expire == 0) { pf_unlink_src_node(sn); uma_zfree(V_pf_sources_z, sn); counter_u64_add( V_pf_status.scounters[SCNT_SRC_NODE_REMOVALS], 1); } PF_SRC_NODE_UNLOCK(sn); } if (nsn != sn && nsn != NULL) { PF_SRC_NODE_LOCK(nsn); if (--nsn->states == 0 && nsn->expire == 0) { pf_unlink_src_node(nsn); uma_zfree(V_pf_sources_z, nsn); counter_u64_add( V_pf_status.scounters[SCNT_SRC_NODE_REMOVALS], 1); } PF_SRC_NODE_UNLOCK(nsn); } drop: if (s != NULL) { pf_src_tree_remove_state(s); s->timeout = PFTM_UNLINKED; STATE_DEC_COUNTERS(s); pf_free_state(s); } return (PF_DROP); } -static int -pf_test_fragment(struct pf_krule **rm, struct pfi_kkif *kif, - struct mbuf *m, struct pf_pdesc *pd, struct pf_krule **am, - struct pf_kruleset **rsm) -{ - struct pf_krule *r, *a = NULL; - struct pf_kruleset *ruleset = NULL; - struct pf_krule_slist match_rules; - struct pf_krule_item *ri; - sa_family_t af = pd->af; - u_short reason; - int tag = -1; - int asd = 0; - int match = 0; - struct pf_kanchor_stackframe anchor_stack[PF_ANCHOR_STACKSIZE]; - - PF_RULES_RASSERT(); - - r = TAILQ_FIRST(pf_main_ruleset.rules[PF_RULESET_FILTER].active.ptr); - SLIST_INIT(&match_rules); - while (r != NULL) { - pf_counter_u64_add(&r->evaluations, 1); - if (pfi_kkif_match(r->kif, kif) == r->ifnot) - r = r->skip[PF_SKIP_IFP].ptr; - else if (r->direction && r->direction != pd->dir) - r = r->skip[PF_SKIP_DIR].ptr; - else if (r->af && r->af != af) - r = r->skip[PF_SKIP_AF].ptr; - else if (r->proto && r->proto != pd->proto) - r = r->skip[PF_SKIP_PROTO].ptr; - else if (PF_MISMATCHAW(&r->src.addr, pd->src, af, - r->src.neg, kif, M_GETFIB(m))) - r = r->skip[PF_SKIP_SRC_ADDR].ptr; - else if (PF_MISMATCHAW(&r->dst.addr, pd->dst, af, - r->dst.neg, NULL, M_GETFIB(m))) - r = r->skip[PF_SKIP_DST_ADDR].ptr; - else if (r->tos && !(r->tos == pd->tos)) - r = TAILQ_NEXT(r, entries); - else if (r->os_fingerprint != PF_OSFP_ANY) - r = TAILQ_NEXT(r, entries); - else if (pd->proto == IPPROTO_UDP && - (r->src.port_op || r->dst.port_op)) - r = TAILQ_NEXT(r, entries); - else if (pd->proto == IPPROTO_TCP && - (r->src.port_op || r->dst.port_op || r->flagset)) - r = TAILQ_NEXT(r, entries); - else if ((pd->proto == IPPROTO_ICMP || - pd->proto == IPPROTO_ICMPV6) && - (r->type || r->code)) - r = TAILQ_NEXT(r, entries); - else if (r->prio && - !pf_match_ieee8021q_pcp(r->prio, m)) - r = TAILQ_NEXT(r, entries); - else if (r->prob && r->prob <= - (arc4random() % (UINT_MAX - 1) + 1)) - r = TAILQ_NEXT(r, entries); - else if (r->match_tag && !pf_match_tag(m, r, &tag, - pd->pf_mtag ? pd->pf_mtag->tag : 0)) - r = TAILQ_NEXT(r, entries); - else { - if (r->anchor == NULL) { - if (r->action == PF_MATCH) { - ri = malloc(sizeof(struct pf_krule_item), M_PF_RULE_ITEM, M_NOWAIT | M_ZERO); - if (ri == NULL) { - REASON_SET(&reason, PFRES_MEMORY); - goto cleanup; - } - ri->r = r; - SLIST_INSERT_HEAD(&match_rules, ri, entry); - pf_counter_u64_critical_enter(); - pf_counter_u64_add_protected(&r->packets[pd->dir == PF_OUT], 1); - pf_counter_u64_add_protected(&r->bytes[pd->dir == PF_OUT], pd->tot_len); - pf_counter_u64_critical_exit(); - pf_rule_to_actions(r, &pd->act); - if (r->log) - PFLOG_PACKET(kif, m, - r->action, PFRES_MATCH, r, - a, ruleset, pd, 1); - } else { - match = 1; - *rm = r; - *am = a; - *rsm = ruleset; - } - if ((*rm)->quick) - break; - r = TAILQ_NEXT(r, entries); - } else - pf_step_into_anchor(anchor_stack, &asd, - &ruleset, PF_RULESET_FILTER, &r, &a, - &match); - } - if (r == NULL && pf_step_out_of_anchor(anchor_stack, &asd, - &ruleset, PF_RULESET_FILTER, &r, &a, &match)) - break; - } - r = *rm; - a = *am; - ruleset = *rsm; - - REASON_SET(&reason, PFRES_MATCH); - - /* apply actions for last matching pass/block rule */ - pf_rule_to_actions(r, &pd->act); - - if (r->log) - PFLOG_PACKET(kif, m, r->action, reason, r, a, ruleset, pd, 1); - - if (r->action != PF_PASS) - return (PF_DROP); - - if (tag > 0 && pf_tag_packet(m, pd, tag)) { - REASON_SET(&reason, PFRES_MEMORY); - goto cleanup; - } - - return (PF_PASS); - -cleanup: - while ((ri = SLIST_FIRST(&match_rules))) { - SLIST_REMOVE_HEAD(&match_rules, entry); - free(ri, M_PF_RULE_ITEM); - } - - return (PF_DROP); -} - static int pf_tcp_track_full(struct pf_kstate **state, struct pfi_kkif *kif, struct mbuf *m, int off, struct pf_pdesc *pd, u_short *reason, int *copyback) { struct tcphdr *th = &pd->hdr.tcp; struct pf_state_peer *src, *dst; u_int16_t win = ntohs(th->th_win); u_int32_t ack, end, data_end, seq, orig_seq; u_int8_t sws, dws, psrc, pdst; int ackskew; if (pd->dir == (*state)->direction) { src = &(*state)->src; dst = &(*state)->dst; psrc = PF_PEER_SRC; pdst = PF_PEER_DST; } else { src = &(*state)->dst; dst = &(*state)->src; psrc = PF_PEER_DST; pdst = PF_PEER_SRC; } if (src->wscale && dst->wscale && !(th->th_flags & TH_SYN)) { sws = src->wscale & PF_WSCALE_MASK; dws = dst->wscale & PF_WSCALE_MASK; } else sws = dws = 0; /* * Sequence tracking algorithm from Guido van Rooij's paper: * http://www.madison-gurkha.com/publications/tcp_filtering/ * tcp_filtering.ps */ orig_seq = seq = ntohl(th->th_seq); if (src->seqlo == 0) { /* First packet from this end. Set its state */ if (((*state)->state_flags & PFSTATE_SCRUB_TCP || dst->scrub) && src->scrub == NULL) { if (pf_normalize_tcp_init(m, off, pd, th, src, dst)) { REASON_SET(reason, PFRES_MEMORY); return (PF_DROP); } } /* Deferred generation of sequence number modulator */ if (dst->seqdiff && !src->seqdiff) { /* use random iss for the TCP server */ while ((src->seqdiff = arc4random() - seq) == 0) ; ack = ntohl(th->th_ack) - dst->seqdiff; pf_change_proto_a(m, &th->th_seq, &th->th_sum, htonl(seq + src->seqdiff), 0); pf_change_proto_a(m, &th->th_ack, &th->th_sum, htonl(ack), 0); *copyback = 1; } else { ack = ntohl(th->th_ack); } end = seq + pd->p_len; if (th->th_flags & TH_SYN) { end++; if (dst->wscale & PF_WSCALE_FLAG) { src->wscale = pf_get_wscale(m, off, th->th_off, pd->af); if (src->wscale & PF_WSCALE_FLAG) { /* Remove scale factor from initial * window */ sws = src->wscale & PF_WSCALE_MASK; win = ((u_int32_t)win + (1 << sws) - 1) >> sws; dws = dst->wscale & PF_WSCALE_MASK; } else { /* fixup other window */ dst->max_win = MIN(TCP_MAXWIN, (u_int32_t)dst->max_win << (dst->wscale & PF_WSCALE_MASK)); /* in case of a retrans SYN|ACK */ dst->wscale = 0; } } } data_end = end; if (th->th_flags & TH_FIN) end++; src->seqlo = seq; if (src->state < TCPS_SYN_SENT) pf_set_protostate(*state, psrc, TCPS_SYN_SENT); /* * May need to slide the window (seqhi may have been set by * the crappy stack check or if we picked up the connection * after establishment) */ if (src->seqhi == 1 || SEQ_GEQ(end + MAX(1, dst->max_win << dws), src->seqhi)) src->seqhi = end + MAX(1, dst->max_win << dws); if (win > src->max_win) src->max_win = win; } else { ack = ntohl(th->th_ack) - dst->seqdiff; if (src->seqdiff) { /* Modulate sequence numbers */ pf_change_proto_a(m, &th->th_seq, &th->th_sum, htonl(seq + src->seqdiff), 0); pf_change_proto_a(m, &th->th_ack, &th->th_sum, htonl(ack), 0); *copyback = 1; } end = seq + pd->p_len; if (th->th_flags & TH_SYN) end++; data_end = end; if (th->th_flags & TH_FIN) end++; } if ((th->th_flags & TH_ACK) == 0) { /* Let it pass through the ack skew check */ ack = dst->seqlo; } else if ((ack == 0 && (th->th_flags & (TH_ACK|TH_RST)) == (TH_ACK|TH_RST)) || /* broken tcp stacks do not set ack */ (dst->state < TCPS_SYN_SENT)) { /* * Many stacks (ours included) will set the ACK number in an * FIN|ACK if the SYN times out -- no sequence to ACK. */ ack = dst->seqlo; } if (seq == end) { /* Ease sequencing restrictions on no data packets */ seq = src->seqlo; data_end = end = seq; } ackskew = dst->seqlo - ack; /* * Need to demodulate the sequence numbers in any TCP SACK options * (Selective ACK). We could optionally validate the SACK values * against the current ACK window, either forwards or backwards, but * I'm not confident that SACK has been implemented properly * everywhere. It wouldn't surprise me if several stacks accidentally * SACK too far backwards of previously ACKed data. There really aren't * any security implications of bad SACKing unless the target stack * doesn't validate the option length correctly. Someone trying to * spoof into a TCP connection won't bother blindly sending SACK * options anyway. */ if (dst->seqdiff && (th->th_off << 2) > sizeof(struct tcphdr)) { if (pf_modulate_sack(m, off, pd, th, dst)) *copyback = 1; } #define MAXACKWINDOW (0xffff + 1500) /* 1500 is an arbitrary fudge factor */ if (SEQ_GEQ(src->seqhi, data_end) && /* Last octet inside other's window space */ SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) && /* Retrans: not more than one window back */ (ackskew >= -MAXACKWINDOW) && /* Acking not more than one reassembled fragment backwards */ (ackskew <= (MAXACKWINDOW << sws)) && /* Acking not more than one window forward */ ((th->th_flags & TH_RST) == 0 || orig_seq == src->seqlo || (orig_seq == src->seqlo + 1) || (orig_seq + 1 == src->seqlo))) { /* Require an exact/+1 sequence match on resets when possible */ if (dst->scrub || src->scrub) { if (pf_normalize_tcp_stateful(m, off, pd, reason, th, *state, src, dst, copyback)) return (PF_DROP); } /* update max window */ if (src->max_win < win) src->max_win = win; /* synchronize sequencing */ if (SEQ_GT(end, src->seqlo)) src->seqlo = end; /* slide the window of what the other end can send */ if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) dst->seqhi = ack + MAX((win << sws), 1); /* update states */ if (th->th_flags & TH_SYN) if (src->state < TCPS_SYN_SENT) pf_set_protostate(*state, psrc, TCPS_SYN_SENT); if (th->th_flags & TH_FIN) if (src->state < TCPS_CLOSING) pf_set_protostate(*state, psrc, TCPS_CLOSING); if (th->th_flags & TH_ACK) { if (dst->state == TCPS_SYN_SENT) { pf_set_protostate(*state, pdst, TCPS_ESTABLISHED); if (src->state == TCPS_ESTABLISHED && (*state)->src_node != NULL && pf_src_connlimit(state)) { REASON_SET(reason, PFRES_SRCLIMIT); return (PF_DROP); } } else if (dst->state == TCPS_CLOSING) pf_set_protostate(*state, pdst, TCPS_FIN_WAIT_2); } if (th->th_flags & TH_RST) pf_set_protostate(*state, PF_PEER_BOTH, TCPS_TIME_WAIT); /* update expire time */ (*state)->expire = pf_get_uptime(); if (src->state >= TCPS_FIN_WAIT_2 && dst->state >= TCPS_FIN_WAIT_2) (*state)->timeout = PFTM_TCP_CLOSED; else if (src->state >= TCPS_CLOSING && dst->state >= TCPS_CLOSING) (*state)->timeout = PFTM_TCP_FIN_WAIT; else if (src->state < TCPS_ESTABLISHED || dst->state < TCPS_ESTABLISHED) (*state)->timeout = PFTM_TCP_OPENING; else if (src->state >= TCPS_CLOSING || dst->state >= TCPS_CLOSING) (*state)->timeout = PFTM_TCP_CLOSING; else (*state)->timeout = PFTM_TCP_ESTABLISHED; /* Fall through to PASS packet */ } else if ((dst->state < TCPS_SYN_SENT || dst->state >= TCPS_FIN_WAIT_2 || src->state >= TCPS_FIN_WAIT_2) && SEQ_GEQ(src->seqhi + MAXACKWINDOW, data_end) && /* Within a window forward of the originating packet */ SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW)) { /* Within a window backward of the originating packet */ /* * This currently handles three situations: * 1) Stupid stacks will shotgun SYNs before their peer * replies. * 2) When PF catches an already established stream (the * firewall rebooted, the state table was flushed, routes * changed...) * 3) Packets get funky immediately after the connection * closes (this should catch Solaris spurious ACK|FINs * that web servers like to spew after a close) * * This must be a little more careful than the above code * since packet floods will also be caught here. We don't * update the TTL here to mitigate the damage of a packet * flood and so the same code can handle awkward establishment * and a loosened connection close. * In the establishment case, a correct peer response will * validate the connection, go through the normal state code * and keep updating the state TTL. */ if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pf: loose state match: "); pf_print_state(*state); pf_print_flags(th->th_flags); printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack, pd->p_len, ackskew, (unsigned long long)(*state)->packets[0], (unsigned long long)(*state)->packets[1], pd->dir == PF_IN ? "in" : "out", pd->dir == (*state)->direction ? "fwd" : "rev"); } if (dst->scrub || src->scrub) { if (pf_normalize_tcp_stateful(m, off, pd, reason, th, *state, src, dst, copyback)) return (PF_DROP); } /* update max window */ if (src->max_win < win) src->max_win = win; /* synchronize sequencing */ if (SEQ_GT(end, src->seqlo)) src->seqlo = end; /* slide the window of what the other end can send */ if (SEQ_GEQ(ack + (win << sws), dst->seqhi)) dst->seqhi = ack + MAX((win << sws), 1); /* * Cannot set dst->seqhi here since this could be a shotgunned * SYN and not an already established connection. */ if (th->th_flags & TH_FIN) if (src->state < TCPS_CLOSING) pf_set_protostate(*state, psrc, TCPS_CLOSING); if (th->th_flags & TH_RST) pf_set_protostate(*state, PF_PEER_BOTH, TCPS_TIME_WAIT); /* Fall through to PASS packet */ } else { if ((*state)->dst.state == TCPS_SYN_SENT && (*state)->src.state == TCPS_SYN_SENT) { /* Send RST for state mismatches during handshake */ if (!(th->th_flags & TH_RST)) pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, pd->src, th->th_dport, th->th_sport, ntohl(th->th_ack), 0, TH_RST, 0, 0, (*state)->rule.ptr->return_ttl, true, 0, 0, (*state)->act.rtableid); src->seqlo = 0; src->seqhi = 1; src->max_win = 1; } else if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pf: BAD state: "); pf_print_state(*state); pf_print_flags(th->th_flags); printf(" seq=%u (%u) ack=%u len=%u ackskew=%d " "pkts=%llu:%llu dir=%s,%s\n", seq, orig_seq, ack, pd->p_len, ackskew, (unsigned long long)(*state)->packets[0], (unsigned long long)(*state)->packets[1], pd->dir == PF_IN ? "in" : "out", pd->dir == (*state)->direction ? "fwd" : "rev"); printf("pf: State failure on: %c %c %c %c | %c %c\n", SEQ_GEQ(src->seqhi, data_end) ? ' ' : '1', SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)) ? ' ': '2', (ackskew >= -MAXACKWINDOW) ? ' ' : '3', (ackskew <= (MAXACKWINDOW << sws)) ? ' ' : '4', SEQ_GEQ(src->seqhi + MAXACKWINDOW, data_end) ?' ' :'5', SEQ_GEQ(seq, src->seqlo - MAXACKWINDOW) ?' ' :'6'); } REASON_SET(reason, PFRES_BADSTATE); return (PF_DROP); } return (PF_PASS); } static int pf_tcp_track_sloppy(struct pf_kstate **state, struct pf_pdesc *pd, u_short *reason) { struct tcphdr *th = &pd->hdr.tcp; struct pf_state_peer *src, *dst; u_int8_t psrc, pdst; if (pd->dir == (*state)->direction) { src = &(*state)->src; dst = &(*state)->dst; psrc = PF_PEER_SRC; pdst = PF_PEER_DST; } else { src = &(*state)->dst; dst = &(*state)->src; psrc = PF_PEER_DST; pdst = PF_PEER_SRC; } if (th->th_flags & TH_SYN) if (src->state < TCPS_SYN_SENT) pf_set_protostate(*state, psrc, TCPS_SYN_SENT); if (th->th_flags & TH_FIN) if (src->state < TCPS_CLOSING) pf_set_protostate(*state, psrc, TCPS_CLOSING); if (th->th_flags & TH_ACK) { if (dst->state == TCPS_SYN_SENT) { pf_set_protostate(*state, pdst, TCPS_ESTABLISHED); if (src->state == TCPS_ESTABLISHED && (*state)->src_node != NULL && pf_src_connlimit(state)) { REASON_SET(reason, PFRES_SRCLIMIT); return (PF_DROP); } } else if (dst->state == TCPS_CLOSING) { pf_set_protostate(*state, pdst, TCPS_FIN_WAIT_2); } else if (src->state == TCPS_SYN_SENT && dst->state < TCPS_SYN_SENT) { /* * Handle a special sloppy case where we only see one * half of the connection. If there is a ACK after * the initial SYN without ever seeing a packet from * the destination, set the connection to established. */ pf_set_protostate(*state, PF_PEER_BOTH, TCPS_ESTABLISHED); dst->state = src->state = TCPS_ESTABLISHED; if ((*state)->src_node != NULL && pf_src_connlimit(state)) { REASON_SET(reason, PFRES_SRCLIMIT); return (PF_DROP); } } else if (src->state == TCPS_CLOSING && dst->state == TCPS_ESTABLISHED && dst->seqlo == 0) { /* * Handle the closing of half connections where we * don't see the full bidirectional FIN/ACK+ACK * handshake. */ pf_set_protostate(*state, pdst, TCPS_CLOSING); } } if (th->th_flags & TH_RST) pf_set_protostate(*state, PF_PEER_BOTH, TCPS_TIME_WAIT); /* update expire time */ (*state)->expire = pf_get_uptime(); if (src->state >= TCPS_FIN_WAIT_2 && dst->state >= TCPS_FIN_WAIT_2) (*state)->timeout = PFTM_TCP_CLOSED; else if (src->state >= TCPS_CLOSING && dst->state >= TCPS_CLOSING) (*state)->timeout = PFTM_TCP_FIN_WAIT; else if (src->state < TCPS_ESTABLISHED || dst->state < TCPS_ESTABLISHED) (*state)->timeout = PFTM_TCP_OPENING; else if (src->state >= TCPS_CLOSING || dst->state >= TCPS_CLOSING) (*state)->timeout = PFTM_TCP_CLOSING; else (*state)->timeout = PFTM_TCP_ESTABLISHED; return (PF_PASS); } static int pf_synproxy(struct pf_pdesc *pd, struct pf_kstate **state, u_short *reason) { struct pf_state_key *sk = (*state)->key[pd->didx]; struct tcphdr *th = &pd->hdr.tcp; if ((*state)->src.state == PF_TCPS_PROXY_SRC) { if (pd->dir != (*state)->direction) { REASON_SET(reason, PFRES_SYNPROXY); return (PF_SYNPROXY_DROP); } if (th->th_flags & TH_SYN) { if (ntohl(th->th_seq) != (*state)->src.seqlo) { REASON_SET(reason, PFRES_SYNPROXY); return (PF_DROP); } pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, pd->src, th->th_dport, th->th_sport, (*state)->src.seqhi, ntohl(th->th_seq) + 1, TH_SYN|TH_ACK, 0, (*state)->src.mss, 0, true, 0, 0, (*state)->act.rtableid); REASON_SET(reason, PFRES_SYNPROXY); return (PF_SYNPROXY_DROP); } else if ((th->th_flags & (TH_ACK|TH_RST|TH_FIN)) != TH_ACK || (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { REASON_SET(reason, PFRES_SYNPROXY); return (PF_DROP); } else if ((*state)->src_node != NULL && pf_src_connlimit(state)) { REASON_SET(reason, PFRES_SRCLIMIT); return (PF_DROP); } else pf_set_protostate(*state, PF_PEER_SRC, PF_TCPS_PROXY_DST); } if ((*state)->src.state == PF_TCPS_PROXY_DST) { if (pd->dir == (*state)->direction) { if (((th->th_flags & (TH_SYN|TH_ACK)) != TH_ACK) || (ntohl(th->th_ack) != (*state)->src.seqhi + 1) || (ntohl(th->th_seq) != (*state)->src.seqlo + 1)) { REASON_SET(reason, PFRES_SYNPROXY); return (PF_DROP); } (*state)->src.max_win = MAX(ntohs(th->th_win), 1); if ((*state)->dst.seqhi == 1) (*state)->dst.seqhi = htonl(arc4random()); pf_send_tcp((*state)->rule.ptr, pd->af, &sk->addr[pd->sidx], &sk->addr[pd->didx], sk->port[pd->sidx], sk->port[pd->didx], (*state)->dst.seqhi, 0, TH_SYN, 0, (*state)->src.mss, 0, false, (*state)->tag, 0, (*state)->act.rtableid); REASON_SET(reason, PFRES_SYNPROXY); return (PF_SYNPROXY_DROP); } else if (((th->th_flags & (TH_SYN|TH_ACK)) != (TH_SYN|TH_ACK)) || (ntohl(th->th_ack) != (*state)->dst.seqhi + 1)) { REASON_SET(reason, PFRES_SYNPROXY); return (PF_DROP); } else { (*state)->dst.max_win = MAX(ntohs(th->th_win), 1); (*state)->dst.seqlo = ntohl(th->th_seq); pf_send_tcp((*state)->rule.ptr, pd->af, pd->dst, pd->src, th->th_dport, th->th_sport, ntohl(th->th_ack), ntohl(th->th_seq) + 1, TH_ACK, (*state)->src.max_win, 0, 0, false, (*state)->tag, 0, (*state)->act.rtableid); pf_send_tcp((*state)->rule.ptr, pd->af, &sk->addr[pd->sidx], &sk->addr[pd->didx], sk->port[pd->sidx], sk->port[pd->didx], (*state)->src.seqhi + 1, (*state)->src.seqlo + 1, TH_ACK, (*state)->dst.max_win, 0, 0, true, 0, 0, (*state)->act.rtableid); (*state)->src.seqdiff = (*state)->dst.seqhi - (*state)->src.seqlo; (*state)->dst.seqdiff = (*state)->src.seqhi - (*state)->dst.seqlo; (*state)->src.seqhi = (*state)->src.seqlo + (*state)->dst.max_win; (*state)->dst.seqhi = (*state)->dst.seqlo + (*state)->src.max_win; (*state)->src.wscale = (*state)->dst.wscale = 0; pf_set_protostate(*state, PF_PEER_BOTH, TCPS_ESTABLISHED); REASON_SET(reason, PFRES_SYNPROXY); return (PF_SYNPROXY_DROP); } } return (PF_PASS); } static int pf_test_state_tcp(struct pf_kstate **state, struct pfi_kkif *kif, struct mbuf *m, int off, struct pf_pdesc *pd, u_short *reason) { struct pf_state_key_cmp key; struct tcphdr *th = &pd->hdr.tcp; int copyback = 0; int action; struct pf_state_peer *src, *dst; bzero(&key, sizeof(key)); key.af = pd->af; key.proto = IPPROTO_TCP; if (pd->dir == PF_IN) { /* wire side, straight */ PF_ACPY(&key.addr[0], pd->src, key.af); PF_ACPY(&key.addr[1], pd->dst, key.af); key.port[0] = th->th_sport; key.port[1] = th->th_dport; } else { /* stack side, reverse */ PF_ACPY(&key.addr[1], pd->src, key.af); PF_ACPY(&key.addr[0], pd->dst, key.af); key.port[1] = th->th_sport; key.port[0] = th->th_dport; } STATE_LOOKUP(kif, &key, *state, pd); if (pd->dir == (*state)->direction) { src = &(*state)->src; dst = &(*state)->dst; } else { src = &(*state)->dst; dst = &(*state)->src; } if ((action = pf_synproxy(pd, state, reason)) != PF_PASS) return (action); if (dst->state >= TCPS_FIN_WAIT_2 && src->state >= TCPS_FIN_WAIT_2 && (((th->th_flags & (TH_SYN|TH_ACK)) == TH_SYN) || ((th->th_flags & (TH_SYN|TH_ACK|TH_RST)) == TH_ACK && pf_syncookie_check(pd) && pd->dir == PF_IN))) { if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pf: state reuse "); pf_print_state(*state); pf_print_flags(th->th_flags); printf("\n"); } /* XXX make sure it's the same direction ?? */ pf_set_protostate(*state, PF_PEER_BOTH, TCPS_CLOSED); pf_unlink_state(*state); *state = NULL; return (PF_DROP); } if ((*state)->state_flags & PFSTATE_SLOPPY) { if (pf_tcp_track_sloppy(state, pd, reason) == PF_DROP) return (PF_DROP); } else { if (pf_tcp_track_full(state, kif, m, off, pd, reason, ©back) == PF_DROP) return (PF_DROP); } /* translate source/destination address, if necessary */ if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { struct pf_state_key *nk = (*state)->key[pd->didx]; if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) || nk->port[pd->sidx] != th->th_sport) pf_change_ap(m, pd->src, &th->th_sport, pd->ip_sum, &th->th_sum, &nk->addr[pd->sidx], nk->port[pd->sidx], 0, pd->af); if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) || nk->port[pd->didx] != th->th_dport) pf_change_ap(m, pd->dst, &th->th_dport, pd->ip_sum, &th->th_sum, &nk->addr[pd->didx], nk->port[pd->didx], 0, pd->af); copyback = 1; } /* Copyback sequence modulation or stateful scrub changes if needed */ if (copyback) m_copyback(m, off, sizeof(*th), (caddr_t)th); return (PF_PASS); } static int pf_test_state_udp(struct pf_kstate **state, struct pfi_kkif *kif, struct mbuf *m, int off, struct pf_pdesc *pd) { struct pf_state_peer *src, *dst; struct pf_state_key_cmp key; struct udphdr *uh = &pd->hdr.udp; uint8_t psrc, pdst; bzero(&key, sizeof(key)); key.af = pd->af; key.proto = IPPROTO_UDP; if (pd->dir == PF_IN) { /* wire side, straight */ PF_ACPY(&key.addr[0], pd->src, key.af); PF_ACPY(&key.addr[1], pd->dst, key.af); key.port[0] = uh->uh_sport; key.port[1] = uh->uh_dport; } else { /* stack side, reverse */ PF_ACPY(&key.addr[1], pd->src, key.af); PF_ACPY(&key.addr[0], pd->dst, key.af); key.port[1] = uh->uh_sport; key.port[0] = uh->uh_dport; } STATE_LOOKUP(kif, &key, *state, pd); if (pd->dir == (*state)->direction) { src = &(*state)->src; dst = &(*state)->dst; psrc = PF_PEER_SRC; pdst = PF_PEER_DST; } else { src = &(*state)->dst; dst = &(*state)->src; psrc = PF_PEER_DST; pdst = PF_PEER_SRC; } /* update states */ if (src->state < PFUDPS_SINGLE) pf_set_protostate(*state, psrc, PFUDPS_SINGLE); if (dst->state == PFUDPS_SINGLE) pf_set_protostate(*state, pdst, PFUDPS_MULTIPLE); /* update expire time */ (*state)->expire = pf_get_uptime(); if (src->state == PFUDPS_MULTIPLE && dst->state == PFUDPS_MULTIPLE) (*state)->timeout = PFTM_UDP_MULTIPLE; else (*state)->timeout = PFTM_UDP_SINGLE; /* translate source/destination address, if necessary */ if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { struct pf_state_key *nk = (*state)->key[pd->didx]; if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) || nk->port[pd->sidx] != uh->uh_sport) pf_change_ap(m, pd->src, &uh->uh_sport, pd->ip_sum, &uh->uh_sum, &nk->addr[pd->sidx], nk->port[pd->sidx], 1, pd->af); if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) || nk->port[pd->didx] != uh->uh_dport) pf_change_ap(m, pd->dst, &uh->uh_dport, pd->ip_sum, &uh->uh_sum, &nk->addr[pd->didx], nk->port[pd->didx], 1, pd->af); m_copyback(m, off, sizeof(*uh), (caddr_t)uh); } return (PF_PASS); } static int pf_test_state_sctp(struct pf_kstate **state, struct pfi_kkif *kif, struct mbuf *m, int off, struct pf_pdesc *pd, u_short *reason) { struct pf_state_key_cmp key; struct pf_state_peer *src, *dst; struct sctphdr *sh = &pd->hdr.sctp; u_int8_t psrc; //, pdst; bzero(&key, sizeof(key)); key.af = pd->af; key.proto = IPPROTO_SCTP; if (pd->dir == PF_IN) { /* wire side, straight */ PF_ACPY(&key.addr[0], pd->src, key.af); PF_ACPY(&key.addr[1], pd->dst, key.af); key.port[0] = sh->src_port; key.port[1] = sh->dest_port; } else { /* stack side, reverse */ PF_ACPY(&key.addr[1], pd->src, key.af); PF_ACPY(&key.addr[0], pd->dst, key.af); key.port[1] = sh->src_port; key.port[0] = sh->dest_port; } STATE_LOOKUP(kif, &key, *state, pd); if (pd->dir == (*state)->direction) { src = &(*state)->src; dst = &(*state)->dst; psrc = PF_PEER_SRC; } else { src = &(*state)->dst; dst = &(*state)->src; psrc = PF_PEER_DST; } if ((src->state >= SCTP_SHUTDOWN_SENT || src->state == SCTP_CLOSED) && (dst->state >= SCTP_SHUTDOWN_SENT || dst->state == SCTP_CLOSED) && pd->sctp_flags & PFDESC_SCTP_INIT) { pf_set_protostate(*state, PF_PEER_BOTH, SCTP_CLOSED); pf_unlink_state(*state); *state = NULL; return (PF_DROP); } /* Track state. */ if (pd->sctp_flags & PFDESC_SCTP_INIT) { if (src->state < SCTP_COOKIE_WAIT) { pf_set_protostate(*state, psrc, SCTP_COOKIE_WAIT); (*state)->timeout = PFTM_SCTP_OPENING; } } if (pd->sctp_flags & PFDESC_SCTP_INIT_ACK) { MPASS(dst->scrub != NULL); if (dst->scrub->pfss_v_tag == 0) dst->scrub->pfss_v_tag = pd->sctp_initiate_tag; } if (pd->sctp_flags & (PFDESC_SCTP_COOKIE | PFDESC_SCTP_HEARTBEAT_ACK)) { if (src->state < SCTP_ESTABLISHED) { pf_set_protostate(*state, psrc, SCTP_ESTABLISHED); (*state)->timeout = PFTM_SCTP_ESTABLISHED; } } if (pd->sctp_flags & (PFDESC_SCTP_SHUTDOWN | PFDESC_SCTP_ABORT | PFDESC_SCTP_SHUTDOWN_COMPLETE)) { if (src->state < SCTP_SHUTDOWN_PENDING) { pf_set_protostate(*state, psrc, SCTP_SHUTDOWN_PENDING); (*state)->timeout = PFTM_SCTP_CLOSING; } } if (pd->sctp_flags & (PFDESC_SCTP_SHUTDOWN_COMPLETE)) { pf_set_protostate(*state, psrc, SCTP_CLOSED); (*state)->timeout = PFTM_SCTP_CLOSED; } if (src->scrub != NULL) { if (src->scrub->pfss_v_tag == 0) { src->scrub->pfss_v_tag = pd->hdr.sctp.v_tag; } else if (src->scrub->pfss_v_tag != pd->hdr.sctp.v_tag) return (PF_DROP); } (*state)->expire = pf_get_uptime(); /* translate source/destination address, if necessary */ if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { uint16_t checksum = 0; struct pf_state_key *nk = (*state)->key[pd->didx]; if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], pd->af) || nk->port[pd->sidx] != pd->hdr.sctp.src_port) { pf_change_ap(m, pd->src, &pd->hdr.sctp.src_port, pd->ip_sum, &checksum, &nk->addr[pd->sidx], nk->port[pd->sidx], 1, pd->af); } if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], pd->af) || nk->port[pd->didx] != pd->hdr.sctp.dest_port) { pf_change_ap(m, pd->dst, &pd->hdr.sctp.dest_port, pd->ip_sum, &checksum, &nk->addr[pd->didx], nk->port[pd->didx], 1, pd->af); } } return (PF_PASS); } static void pf_sctp_multihome_detach_addr(const struct pf_kstate *s) { struct pf_sctp_endpoint key; struct pf_sctp_endpoint *ep; struct pf_state_key *sks = s->key[PF_SK_STACK]; struct pf_sctp_source *i, *tmp; if (sks == NULL || sks->proto != IPPROTO_SCTP || s->dst.scrub == NULL) return; PF_SCTP_ENDPOINTS_LOCK(); key.v_tag = s->dst.scrub->pfss_v_tag; ep = RB_FIND(pf_sctp_endpoints, &V_pf_sctp_endpoints, &key); if (ep != NULL) { TAILQ_FOREACH_SAFE(i, &ep->sources, entry, tmp) { if (pf_addr_cmp(&i->addr, &s->key[PF_SK_WIRE]->addr[s->direction == PF_OUT], s->key[PF_SK_WIRE]->af) == 0) { SDT_PROBE3(pf, sctp, multihome, remove, key.v_tag, s, i); TAILQ_REMOVE(&ep->sources, i, entry); free(i, M_PFTEMP); break; } } if (TAILQ_EMPTY(&ep->sources)) { RB_REMOVE(pf_sctp_endpoints, &V_pf_sctp_endpoints, ep); free(ep, M_PFTEMP); } } /* Other direction. */ key.v_tag = s->src.scrub->pfss_v_tag; ep = RB_FIND(pf_sctp_endpoints, &V_pf_sctp_endpoints, &key); if (ep != NULL) { TAILQ_FOREACH_SAFE(i, &ep->sources, entry, tmp) { if (pf_addr_cmp(&i->addr, &s->key[PF_SK_WIRE]->addr[s->direction == PF_IN], s->key[PF_SK_WIRE]->af) == 0) { SDT_PROBE3(pf, sctp, multihome, remove, key.v_tag, s, i); TAILQ_REMOVE(&ep->sources, i, entry); free(i, M_PFTEMP); break; } } if (TAILQ_EMPTY(&ep->sources)) { RB_REMOVE(pf_sctp_endpoints, &V_pf_sctp_endpoints, ep); free(ep, M_PFTEMP); } } PF_SCTP_ENDPOINTS_UNLOCK(); } static void pf_sctp_multihome_add_addr(struct pf_pdesc *pd, struct pf_addr *a, uint32_t v_tag) { struct pf_sctp_endpoint key = { .v_tag = v_tag, }; struct pf_sctp_source *i; struct pf_sctp_endpoint *ep; PF_SCTP_ENDPOINTS_LOCK(); ep = RB_FIND(pf_sctp_endpoints, &V_pf_sctp_endpoints, &key); if (ep == NULL) { ep = malloc(sizeof(struct pf_sctp_endpoint), M_PFTEMP, M_NOWAIT); if (ep == NULL) { PF_SCTP_ENDPOINTS_UNLOCK(); return; } ep->v_tag = v_tag; TAILQ_INIT(&ep->sources); RB_INSERT(pf_sctp_endpoints, &V_pf_sctp_endpoints, ep); } /* Avoid inserting duplicates. */ TAILQ_FOREACH(i, &ep->sources, entry) { if (pf_addr_cmp(&i->addr, a, pd->af) == 0) { PF_SCTP_ENDPOINTS_UNLOCK(); return; } } i = malloc(sizeof(*i), M_PFTEMP, M_NOWAIT); if (i == NULL) { PF_SCTP_ENDPOINTS_UNLOCK(); return; } i->af = pd->af; memcpy(&i->addr, a, sizeof(*a)); TAILQ_INSERT_TAIL(&ep->sources, i, entry); SDT_PROBE2(pf, sctp, multihome, add, v_tag, i); PF_SCTP_ENDPOINTS_UNLOCK(); } static void pf_sctp_multihome_delayed(struct pf_pdesc *pd, int off, struct pfi_kkif *kif, struct pf_kstate *s, int action) { struct pf_sctp_multihome_job *j, *tmp; struct pf_sctp_source *i; int ret __unused; struct pf_kstate *sm = NULL; struct pf_krule *ra = NULL; struct pf_krule *r = &V_pf_default_rule; struct pf_kruleset *rs = NULL; bool do_extra = true; PF_RULES_RLOCK_TRACKER; again: TAILQ_FOREACH_SAFE(j, &pd->sctp_multihome_jobs, next, tmp) { if (s == NULL || action != PF_PASS) goto free; /* Confirm we don't recurse here. */ MPASS(! (pd->sctp_flags & PFDESC_SCTP_ADD_IP)); switch (j->op) { case SCTP_ADD_IP_ADDRESS: { uint32_t v_tag = pd->sctp_initiate_tag; if (v_tag == 0) { if (s->direction == pd->dir) v_tag = s->src.scrub->pfss_v_tag; else v_tag = s->dst.scrub->pfss_v_tag; } /* * Avoid duplicating states. We'll already have * created a state based on the source address of * the packet, but SCTP endpoints may also list this * address again in the INIT(_ACK) parameters. */ if (pf_addr_cmp(&j->src, pd->src, pd->af) == 0) { break; } j->pd.sctp_flags |= PFDESC_SCTP_ADD_IP; PF_RULES_RLOCK(); sm = NULL; /* * New connections need to be floating, because * we cannot know what interfaces it will use. * That's why we pass V_pfi_all rather than kif. */ ret = pf_test_rule(&r, &sm, V_pfi_all, j->m, off, &j->pd, &ra, &rs, NULL, sizeof(j->pd.hdr.sctp)); PF_RULES_RUNLOCK(); SDT_PROBE4(pf, sctp, multihome, test, kif, r, j->m, ret); if (ret != PF_DROP && sm != NULL) { /* Inherit v_tag values. */ if (sm->direction == s->direction) { sm->src.scrub->pfss_v_tag = s->src.scrub->pfss_v_tag; sm->dst.scrub->pfss_v_tag = s->dst.scrub->pfss_v_tag; } else { sm->src.scrub->pfss_v_tag = s->dst.scrub->pfss_v_tag; sm->dst.scrub->pfss_v_tag = s->src.scrub->pfss_v_tag; } PF_STATE_UNLOCK(sm); } else { /* If we try duplicate inserts? */ break; } /* Only add the address if we've actually allowed the state. */ pf_sctp_multihome_add_addr(pd, &j->src, v_tag); if (! do_extra) { break; } /* * We need to do this for each of our source addresses. * Find those based on the verification tag. */ struct pf_sctp_endpoint key = { .v_tag = pd->hdr.sctp.v_tag, }; struct pf_sctp_endpoint *ep; PF_SCTP_ENDPOINTS_LOCK(); ep = RB_FIND(pf_sctp_endpoints, &V_pf_sctp_endpoints, &key); if (ep == NULL) { PF_SCTP_ENDPOINTS_UNLOCK(); break; } MPASS(ep != NULL); TAILQ_FOREACH(i, &ep->sources, entry) { struct pf_sctp_multihome_job *nj; /* SCTP can intermingle IPv4 and IPv6. */ if (i->af != pd->af) continue; nj = malloc(sizeof(*nj), M_PFTEMP, M_NOWAIT | M_ZERO); if (! nj) { continue; } memcpy(&nj->pd, &j->pd, sizeof(j->pd)); memcpy(&nj->src, &j->src, sizeof(nj->src)); nj->pd.src = &nj->src; // New destination address! memcpy(&nj->dst, &i->addr, sizeof(nj->dst)); nj->pd.dst = &nj->dst; nj->m = j->m; nj->op = j->op; TAILQ_INSERT_TAIL(&pd->sctp_multihome_jobs, nj, next); } PF_SCTP_ENDPOINTS_UNLOCK(); break; } case SCTP_DEL_IP_ADDRESS: { struct pf_state_key_cmp key; uint8_t psrc; bzero(&key, sizeof(key)); key.af = j->pd.af; key.proto = IPPROTO_SCTP; if (j->pd.dir == PF_IN) { /* wire side, straight */ PF_ACPY(&key.addr[0], j->pd.src, key.af); PF_ACPY(&key.addr[1], j->pd.dst, key.af); key.port[0] = j->pd.hdr.sctp.src_port; key.port[1] = j->pd.hdr.sctp.dest_port; } else { /* stack side, reverse */ PF_ACPY(&key.addr[1], j->pd.src, key.af); PF_ACPY(&key.addr[0], j->pd.dst, key.af); key.port[1] = j->pd.hdr.sctp.src_port; key.port[0] = j->pd.hdr.sctp.dest_port; } sm = pf_find_state(kif, &key, j->pd.dir); if (sm != NULL) { PF_STATE_LOCK_ASSERT(sm); if (j->pd.dir == sm->direction) { psrc = PF_PEER_SRC; } else { psrc = PF_PEER_DST; } pf_set_protostate(sm, psrc, SCTP_SHUTDOWN_PENDING); sm->timeout = PFTM_SCTP_CLOSING; PF_STATE_UNLOCK(sm); } break; default: panic("Unknown op %#x", j->op); } } free: TAILQ_REMOVE(&pd->sctp_multihome_jobs, j, next); free(j, M_PFTEMP); } /* We may have inserted extra work while processing the list. */ if (! TAILQ_EMPTY(&pd->sctp_multihome_jobs)) { do_extra = false; goto again; } } static int pf_multihome_scan(struct mbuf *m, int start, int len, struct pf_pdesc *pd, struct pfi_kkif *kif, int op) { int off = 0; struct pf_sctp_multihome_job *job; while (off < len) { struct sctp_paramhdr h; if (!pf_pull_hdr(m, start + off, &h, sizeof(h), NULL, NULL, pd->af)) return (PF_DROP); /* Parameters are at least 4 bytes. */ if (ntohs(h.param_length) < 4) return (PF_DROP); switch (ntohs(h.param_type)) { case SCTP_IPV4_ADDRESS: { struct in_addr t; if (ntohs(h.param_length) != (sizeof(struct sctp_paramhdr) + sizeof(t))) return (PF_DROP); if (!pf_pull_hdr(m, start + off + sizeof(h), &t, sizeof(t), NULL, NULL, pd->af)) return (PF_DROP); if (in_nullhost(t)) t.s_addr = pd->src->v4.s_addr; /* * We hold the state lock (idhash) here, which means * that we can't acquire the keyhash, or we'll get a * LOR (and potentially double-lock things too). We also * can't release the state lock here, so instead we'll * enqueue this for async handling. * There's a relatively small race here, in that a * packet using the new addresses could arrive already, * but that's just though luck for it. */ job = malloc(sizeof(*job), M_PFTEMP, M_NOWAIT | M_ZERO); if (! job) return (PF_DROP); memcpy(&job->pd, pd, sizeof(*pd)); // New source address! memcpy(&job->src, &t, sizeof(t)); job->pd.src = &job->src; memcpy(&job->dst, pd->dst, sizeof(job->dst)); job->pd.dst = &job->dst; job->m = m; job->op = op; TAILQ_INSERT_TAIL(&pd->sctp_multihome_jobs, job, next); break; } #ifdef INET6 case SCTP_IPV6_ADDRESS: { struct in6_addr t; if (ntohs(h.param_length) != (sizeof(struct sctp_paramhdr) + sizeof(t))) return (PF_DROP); if (!pf_pull_hdr(m, start + off + sizeof(h), &t, sizeof(t), NULL, NULL, pd->af)) return (PF_DROP); if (memcmp(&t, &pd->src->v6, sizeof(t)) == 0) break; if (memcmp(&t, &in6addr_any, sizeof(t)) == 0) memcpy(&t, &pd->src->v6, sizeof(t)); job = malloc(sizeof(*job), M_PFTEMP, M_NOWAIT | M_ZERO); if (! job) return (PF_DROP); memcpy(&job->pd, pd, sizeof(*pd)); memcpy(&job->src, &t, sizeof(t)); job->pd.src = &job->src; memcpy(&job->dst, pd->dst, sizeof(job->dst)); job->pd.dst = &job->dst; job->m = m; job->op = op; TAILQ_INSERT_TAIL(&pd->sctp_multihome_jobs, job, next); break; } #endif case SCTP_ADD_IP_ADDRESS: { int ret; struct sctp_asconf_paramhdr ah; if (!pf_pull_hdr(m, start + off, &ah, sizeof(ah), NULL, NULL, pd->af)) return (PF_DROP); ret = pf_multihome_scan(m, start + off + sizeof(ah), ntohs(ah.ph.param_length) - sizeof(ah), pd, kif, SCTP_ADD_IP_ADDRESS); if (ret != PF_PASS) return (ret); break; } case SCTP_DEL_IP_ADDRESS: { int ret; struct sctp_asconf_paramhdr ah; if (!pf_pull_hdr(m, start + off, &ah, sizeof(ah), NULL, NULL, pd->af)) return (PF_DROP); ret = pf_multihome_scan(m, start + off + sizeof(ah), ntohs(ah.ph.param_length) - sizeof(ah), pd, kif, SCTP_DEL_IP_ADDRESS); if (ret != PF_PASS) return (ret); break; } default: break; } off += roundup(ntohs(h.param_length), 4); } return (PF_PASS); } int pf_multihome_scan_init(struct mbuf *m, int start, int len, struct pf_pdesc *pd, struct pfi_kkif *kif) { start += sizeof(struct sctp_init_chunk); len -= sizeof(struct sctp_init_chunk); return (pf_multihome_scan(m, start, len, pd, kif, SCTP_ADD_IP_ADDRESS)); } int pf_multihome_scan_asconf(struct mbuf *m, int start, int len, struct pf_pdesc *pd, struct pfi_kkif *kif) { start += sizeof(struct sctp_asconf_chunk); len -= sizeof(struct sctp_asconf_chunk); return (pf_multihome_scan(m, start, len, pd, kif, SCTP_ADD_IP_ADDRESS)); } int pf_icmp_state_lookup(struct pf_state_key_cmp *key, struct pf_pdesc *pd, struct pf_kstate **state, struct mbuf *m, int off, int direction, struct pfi_kkif *kif, u_int16_t icmpid, u_int16_t type, int icmp_dir, int *iidx, int multi, int inner) { key->af = pd->af; key->proto = pd->proto; if (icmp_dir == PF_IN) { *iidx = pd->sidx; key->port[pd->sidx] = icmpid; key->port[pd->didx] = type; } else { *iidx = pd->didx; key->port[pd->sidx] = type; key->port[pd->didx] = icmpid; } if (pf_state_key_addr_setup(pd, m, off, key, pd->sidx, pd->src, pd->didx, pd->dst, multi)) return (PF_DROP); STATE_LOOKUP(kif, key, *state, pd); if ((*state)->state_flags & PFSTATE_SLOPPY) return (-1); /* Is this ICMP message flowing in right direction? */ if ((*state)->rule.ptr->type && (((!inner && (*state)->direction == direction) || (inner && (*state)->direction != direction)) ? PF_IN : PF_OUT) != icmp_dir) { if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pf: icmp type %d in wrong direction (%d): ", ntohs(type), icmp_dir); pf_print_state(*state); printf("\n"); } PF_STATE_UNLOCK(*state); *state = NULL; return (PF_DROP); } return (-1); } static int pf_test_state_icmp(struct pf_kstate **state, struct pfi_kkif *kif, struct mbuf *m, int off, struct pf_pdesc *pd, u_short *reason) { struct pf_addr *saddr = pd->src, *daddr = pd->dst; u_int16_t *icmpsum, virtual_id, virtual_type; u_int8_t icmptype, icmpcode; int icmp_dir, iidx, ret, multi; struct pf_state_key_cmp key; #ifdef INET u_int16_t icmpid; #endif MPASS(*state == NULL); bzero(&key, sizeof(key)); switch (pd->proto) { #ifdef INET case IPPROTO_ICMP: icmptype = pd->hdr.icmp.icmp_type; icmpcode = pd->hdr.icmp.icmp_code; icmpid = pd->hdr.icmp.icmp_id; icmpsum = &pd->hdr.icmp.icmp_cksum; break; #endif /* INET */ #ifdef INET6 case IPPROTO_ICMPV6: icmptype = pd->hdr.icmp6.icmp6_type; icmpcode = pd->hdr.icmp6.icmp6_code; #ifdef INET icmpid = pd->hdr.icmp6.icmp6_id; #endif icmpsum = &pd->hdr.icmp6.icmp6_cksum; break; #endif /* INET6 */ } if (pf_icmp_mapping(pd, icmptype, &icmp_dir, &multi, &virtual_id, &virtual_type) == 0) { /* * ICMP query/reply message not related to a TCP/UDP packet. * Search for an ICMP state. */ ret = pf_icmp_state_lookup(&key, pd, state, m, off, pd->dir, kif, virtual_id, virtual_type, icmp_dir, &iidx, PF_ICMP_MULTI_NONE, 0); if (ret >= 0) { MPASS(*state == NULL); if (ret == PF_DROP && pd->af == AF_INET6 && icmp_dir == PF_OUT) { ret = pf_icmp_state_lookup(&key, pd, state, m, off, pd->dir, kif, virtual_id, virtual_type, icmp_dir, &iidx, multi, 0); if (ret >= 0) { MPASS(*state == NULL); return (ret); } } else return (ret); } (*state)->expire = pf_get_uptime(); (*state)->timeout = PFTM_ICMP_ERROR_REPLY; /* translate source/destination address, if necessary */ if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { struct pf_state_key *nk = (*state)->key[pd->didx]; switch (pd->af) { #ifdef INET case AF_INET: if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET)) pf_change_a(&saddr->v4.s_addr, pd->ip_sum, nk->addr[pd->sidx].v4.s_addr, 0); if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET)) pf_change_a(&daddr->v4.s_addr, pd->ip_sum, nk->addr[pd->didx].v4.s_addr, 0); if (nk->port[iidx] != pd->hdr.icmp.icmp_id) { pd->hdr.icmp.icmp_cksum = pf_cksum_fixup( pd->hdr.icmp.icmp_cksum, icmpid, nk->port[iidx], 0); pd->hdr.icmp.icmp_id = nk->port[iidx]; } m_copyback(m, off, ICMP_MINLEN, (caddr_t )&pd->hdr.icmp); break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET6)) pf_change_a6(saddr, &pd->hdr.icmp6.icmp6_cksum, &nk->addr[pd->sidx], 0); if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET6)) pf_change_a6(daddr, &pd->hdr.icmp6.icmp6_cksum, &nk->addr[pd->didx], 0); m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t )&pd->hdr.icmp6); break; #endif /* INET6 */ } } return (PF_PASS); } else { /* * ICMP error message in response to a TCP/UDP packet. * Extract the inner TCP/UDP header and search for that state. */ struct pf_pdesc pd2; bzero(&pd2, sizeof pd2); #ifdef INET struct ip h2; #endif /* INET */ #ifdef INET6 struct ip6_hdr h2_6; int terminal = 0; #endif /* INET6 */ int ipoff2 = 0; int off2 = 0; pd2.af = pd->af; pd2.dir = pd->dir; /* Payload packet is from the opposite direction. */ pd2.sidx = (pd->dir == PF_IN) ? 1 : 0; pd2.didx = (pd->dir == PF_IN) ? 0 : 1; switch (pd->af) { #ifdef INET case AF_INET: /* offset of h2 in mbuf chain */ ipoff2 = off + ICMP_MINLEN; if (!pf_pull_hdr(m, ipoff2, &h2, sizeof(h2), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short " "(ip)\n")); return (PF_DROP); } /* * ICMP error messages don't refer to non-first * fragments */ if (h2.ip_off & htons(IP_OFFMASK)) { REASON_SET(reason, PFRES_FRAG); return (PF_DROP); } /* offset of protocol header that follows h2 */ off2 = ipoff2 + (h2.ip_hl << 2); pd2.proto = h2.ip_p; pd2.src = (struct pf_addr *)&h2.ip_src; pd2.dst = (struct pf_addr *)&h2.ip_dst; pd2.ip_sum = &h2.ip_sum; break; #endif /* INET */ #ifdef INET6 case AF_INET6: ipoff2 = off + sizeof(struct icmp6_hdr); if (!pf_pull_hdr(m, ipoff2, &h2_6, sizeof(h2_6), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short " "(ip6)\n")); return (PF_DROP); } pd2.proto = h2_6.ip6_nxt; pd2.src = (struct pf_addr *)&h2_6.ip6_src; pd2.dst = (struct pf_addr *)&h2_6.ip6_dst; pd2.ip_sum = NULL; off2 = ipoff2 + sizeof(h2_6); do { switch (pd2.proto) { case IPPROTO_FRAGMENT: /* * ICMPv6 error messages for * non-first fragments */ REASON_SET(reason, PFRES_FRAG); return (PF_DROP); case IPPROTO_AH: case IPPROTO_HOPOPTS: case IPPROTO_ROUTING: case IPPROTO_DSTOPTS: { /* get next header and header length */ struct ip6_ext opt6; if (!pf_pull_hdr(m, off2, &opt6, sizeof(opt6), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMPv6 short opt\n")); return (PF_DROP); } if (pd2.proto == IPPROTO_AH) off2 += (opt6.ip6e_len + 2) * 4; else off2 += (opt6.ip6e_len + 1) * 8; pd2.proto = opt6.ip6e_nxt; /* goto the next header */ break; } default: terminal++; break; } } while (!terminal); break; #endif /* INET6 */ } if (PF_ANEQ(pd->dst, pd2.src, pd->af)) { if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pf: BAD ICMP %d:%d outer dst: ", icmptype, icmpcode); pf_print_host(pd->src, 0, pd->af); printf(" -> "); pf_print_host(pd->dst, 0, pd->af); printf(" inner src: "); pf_print_host(pd2.src, 0, pd2.af); printf(" -> "); pf_print_host(pd2.dst, 0, pd2.af); printf("\n"); } REASON_SET(reason, PFRES_BADSTATE); return (PF_DROP); } switch (pd2.proto) { case IPPROTO_TCP: { struct tcphdr th; u_int32_t seq; struct pf_state_peer *src, *dst; u_int8_t dws; int copyback = 0; /* * Only the first 8 bytes of the TCP header can be * expected. Don't access any TCP header fields after * th_seq, an ackskew test is not possible. */ if (!pf_pull_hdr(m, off2, &th, 8, NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short " "(tcp)\n")); return (PF_DROP); } key.af = pd2.af; key.proto = IPPROTO_TCP; PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af); PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af); key.port[pd2.sidx] = th.th_sport; key.port[pd2.didx] = th.th_dport; STATE_LOOKUP(kif, &key, *state, pd); if (pd->dir == (*state)->direction) { src = &(*state)->dst; dst = &(*state)->src; } else { src = &(*state)->src; dst = &(*state)->dst; } if (src->wscale && dst->wscale) dws = dst->wscale & PF_WSCALE_MASK; else dws = 0; /* Demodulate sequence number */ seq = ntohl(th.th_seq) - src->seqdiff; if (src->seqdiff) { pf_change_a(&th.th_seq, icmpsum, htonl(seq), 0); copyback = 1; } if (!((*state)->state_flags & PFSTATE_SLOPPY) && (!SEQ_GEQ(src->seqhi, seq) || !SEQ_GEQ(seq, src->seqlo - (dst->max_win << dws)))) { if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pf: BAD ICMP %d:%d ", icmptype, icmpcode); pf_print_host(pd->src, 0, pd->af); printf(" -> "); pf_print_host(pd->dst, 0, pd->af); printf(" state: "); pf_print_state(*state); printf(" seq=%u\n", seq); } REASON_SET(reason, PFRES_BADSTATE); return (PF_DROP); } else { if (V_pf_status.debug >= PF_DEBUG_MISC) { printf("pf: OK ICMP %d:%d ", icmptype, icmpcode); pf_print_host(pd->src, 0, pd->af); printf(" -> "); pf_print_host(pd->dst, 0, pd->af); printf(" state: "); pf_print_state(*state); printf(" seq=%u\n", seq); } } /* translate source/destination address, if necessary */ if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { struct pf_state_key *nk = (*state)->key[pd->didx]; if (PF_ANEQ(pd2.src, &nk->addr[pd2.sidx], pd2.af) || nk->port[pd2.sidx] != th.th_sport) pf_change_icmp(pd2.src, &th.th_sport, daddr, &nk->addr[pd2.sidx], nk->port[pd2.sidx], NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, pd2.af); if (PF_ANEQ(pd2.dst, &nk->addr[pd2.didx], pd2.af) || nk->port[pd2.didx] != th.th_dport) pf_change_icmp(pd2.dst, &th.th_dport, saddr, &nk->addr[pd2.didx], nk->port[pd2.didx], NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, pd2.af); copyback = 1; } if (copyback) { switch (pd2.af) { #ifdef INET case AF_INET: m_copyback(m, off, ICMP_MINLEN, (caddr_t )&pd->hdr.icmp); m_copyback(m, ipoff2, sizeof(h2), (caddr_t )&h2); break; #endif /* INET */ #ifdef INET6 case AF_INET6: m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t )&pd->hdr.icmp6); m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t )&h2_6); break; #endif /* INET6 */ } m_copyback(m, off2, 8, (caddr_t)&th); } return (PF_PASS); break; } case IPPROTO_UDP: { struct udphdr uh; if (!pf_pull_hdr(m, off2, &uh, sizeof(uh), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short " "(udp)\n")); return (PF_DROP); } key.af = pd2.af; key.proto = IPPROTO_UDP; PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af); PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af); key.port[pd2.sidx] = uh.uh_sport; key.port[pd2.didx] = uh.uh_dport; STATE_LOOKUP(kif, &key, *state, pd); /* translate source/destination address, if necessary */ if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { struct pf_state_key *nk = (*state)->key[pd->didx]; if (PF_ANEQ(pd2.src, &nk->addr[pd2.sidx], pd2.af) || nk->port[pd2.sidx] != uh.uh_sport) pf_change_icmp(pd2.src, &uh.uh_sport, daddr, &nk->addr[pd2.sidx], nk->port[pd2.sidx], &uh.uh_sum, pd2.ip_sum, icmpsum, pd->ip_sum, 1, pd2.af); if (PF_ANEQ(pd2.dst, &nk->addr[pd2.didx], pd2.af) || nk->port[pd2.didx] != uh.uh_dport) pf_change_icmp(pd2.dst, &uh.uh_dport, saddr, &nk->addr[pd2.didx], nk->port[pd2.didx], &uh.uh_sum, pd2.ip_sum, icmpsum, pd->ip_sum, 1, pd2.af); switch (pd2.af) { #ifdef INET case AF_INET: m_copyback(m, off, ICMP_MINLEN, (caddr_t )&pd->hdr.icmp); m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2); break; #endif /* INET */ #ifdef INET6 case AF_INET6: m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t )&pd->hdr.icmp6); m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t )&h2_6); break; #endif /* INET6 */ } m_copyback(m, off2, sizeof(uh), (caddr_t)&uh); } return (PF_PASS); break; } #ifdef INET case IPPROTO_ICMP: { struct icmp *iih = &pd2.hdr.icmp; if (!pf_pull_hdr(m, off2, iih, ICMP_MINLEN, NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short i" "(icmp)\n")); return (PF_DROP); } icmpid = iih->icmp_id; pf_icmp_mapping(&pd2, iih->icmp_type, &icmp_dir, &multi, &virtual_id, &virtual_type); ret = pf_icmp_state_lookup(&key, &pd2, state, m, off, pd2.dir, kif, virtual_id, virtual_type, icmp_dir, &iidx, PF_ICMP_MULTI_NONE, 1); if (ret >= 0) { MPASS(*state == NULL); return (ret); } /* translate source/destination address, if necessary */ if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { struct pf_state_key *nk = (*state)->key[pd->didx]; if (PF_ANEQ(pd2.src, &nk->addr[pd2.sidx], pd2.af) || (virtual_type == htons(ICMP_ECHO) && nk->port[iidx] != iih->icmp_id)) pf_change_icmp(pd2.src, (virtual_type == htons(ICMP_ECHO)) ? &iih->icmp_id : NULL, daddr, &nk->addr[pd2.sidx], (virtual_type == htons(ICMP_ECHO)) ? nk->port[iidx] : 0, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, AF_INET); if (PF_ANEQ(pd2.dst, &nk->addr[pd2.didx], pd2.af)) pf_change_icmp(pd2.dst, NULL, NULL, &nk->addr[pd2.didx], 0, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, AF_INET); m_copyback(m, off, ICMP_MINLEN, (caddr_t)&pd->hdr.icmp); m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2); m_copyback(m, off2, ICMP_MINLEN, (caddr_t)iih); } return (PF_PASS); break; } #endif /* INET */ #ifdef INET6 case IPPROTO_ICMPV6: { struct icmp6_hdr *iih = &pd2.hdr.icmp6; if (!pf_pull_hdr(m, off2, iih, sizeof(struct icmp6_hdr), NULL, reason, pd2.af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: ICMP error message too short " "(icmp6)\n")); return (PF_DROP); } pf_icmp_mapping(&pd2, iih->icmp6_type, &icmp_dir, &multi, &virtual_id, &virtual_type); ret = pf_icmp_state_lookup(&key, &pd2, state, m, off, pd->dir, kif, virtual_id, virtual_type, icmp_dir, &iidx, PF_ICMP_MULTI_NONE, 1); if (ret >= 0) { MPASS(*state == NULL); if (ret == PF_DROP && pd2.af == AF_INET6 && icmp_dir == PF_OUT) { ret = pf_icmp_state_lookup(&key, &pd2, state, m, off, pd->dir, kif, virtual_id, virtual_type, icmp_dir, &iidx, multi, 1); if (ret >= 0) { MPASS(*state == NULL); return (ret); } } else return (ret); } /* translate source/destination address, if necessary */ if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { struct pf_state_key *nk = (*state)->key[pd->didx]; if (PF_ANEQ(pd2.src, &nk->addr[pd2.sidx], pd2.af) || ((virtual_type == htons(ICMP6_ECHO_REQUEST)) && nk->port[pd2.sidx] != iih->icmp6_id)) pf_change_icmp(pd2.src, (virtual_type == htons(ICMP6_ECHO_REQUEST)) ? &iih->icmp6_id : NULL, daddr, &nk->addr[pd2.sidx], (virtual_type == htons(ICMP6_ECHO_REQUEST)) ? nk->port[iidx] : 0, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, AF_INET6); if (PF_ANEQ(pd2.dst, &nk->addr[pd2.didx], pd2.af)) pf_change_icmp(pd2.dst, NULL, NULL, &nk->addr[pd2.didx], 0, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, AF_INET6); m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t)&pd->hdr.icmp6); m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t)&h2_6); m_copyback(m, off2, sizeof(struct icmp6_hdr), (caddr_t)iih); } return (PF_PASS); break; } #endif /* INET6 */ default: { key.af = pd2.af; key.proto = pd2.proto; PF_ACPY(&key.addr[pd2.sidx], pd2.src, key.af); PF_ACPY(&key.addr[pd2.didx], pd2.dst, key.af); key.port[0] = key.port[1] = 0; STATE_LOOKUP(kif, &key, *state, pd); /* translate source/destination address, if necessary */ if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { struct pf_state_key *nk = (*state)->key[pd->didx]; if (PF_ANEQ(pd2.src, &nk->addr[pd2.sidx], pd2.af)) pf_change_icmp(pd2.src, NULL, daddr, &nk->addr[pd2.sidx], 0, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, pd2.af); if (PF_ANEQ(pd2.dst, &nk->addr[pd2.didx], pd2.af)) pf_change_icmp(pd2.dst, NULL, saddr, &nk->addr[pd2.didx], 0, NULL, pd2.ip_sum, icmpsum, pd->ip_sum, 0, pd2.af); switch (pd2.af) { #ifdef INET case AF_INET: m_copyback(m, off, ICMP_MINLEN, (caddr_t)&pd->hdr.icmp); m_copyback(m, ipoff2, sizeof(h2), (caddr_t)&h2); break; #endif /* INET */ #ifdef INET6 case AF_INET6: m_copyback(m, off, sizeof(struct icmp6_hdr), (caddr_t )&pd->hdr.icmp6); m_copyback(m, ipoff2, sizeof(h2_6), (caddr_t )&h2_6); break; #endif /* INET6 */ } } return (PF_PASS); break; } } } } static int pf_test_state_other(struct pf_kstate **state, struct pfi_kkif *kif, struct mbuf *m, struct pf_pdesc *pd) { struct pf_state_peer *src, *dst; struct pf_state_key_cmp key; uint8_t psrc, pdst; bzero(&key, sizeof(key)); key.af = pd->af; key.proto = pd->proto; if (pd->dir == PF_IN) { PF_ACPY(&key.addr[0], pd->src, key.af); PF_ACPY(&key.addr[1], pd->dst, key.af); key.port[0] = key.port[1] = 0; } else { PF_ACPY(&key.addr[1], pd->src, key.af); PF_ACPY(&key.addr[0], pd->dst, key.af); key.port[1] = key.port[0] = 0; } STATE_LOOKUP(kif, &key, *state, pd); if (pd->dir == (*state)->direction) { src = &(*state)->src; dst = &(*state)->dst; psrc = PF_PEER_SRC; pdst = PF_PEER_DST; } else { src = &(*state)->dst; dst = &(*state)->src; psrc = PF_PEER_DST; pdst = PF_PEER_SRC; } /* update states */ if (src->state < PFOTHERS_SINGLE) pf_set_protostate(*state, psrc, PFOTHERS_SINGLE); if (dst->state == PFOTHERS_SINGLE) pf_set_protostate(*state, pdst, PFOTHERS_MULTIPLE); /* update expire time */ (*state)->expire = pf_get_uptime(); if (src->state == PFOTHERS_MULTIPLE && dst->state == PFOTHERS_MULTIPLE) (*state)->timeout = PFTM_OTHER_MULTIPLE; else (*state)->timeout = PFTM_OTHER_SINGLE; /* translate source/destination address, if necessary */ if ((*state)->key[PF_SK_WIRE] != (*state)->key[PF_SK_STACK]) { struct pf_state_key *nk = (*state)->key[pd->didx]; KASSERT(nk, ("%s: nk is null", __func__)); KASSERT(pd, ("%s: pd is null", __func__)); KASSERT(pd->src, ("%s: pd->src is null", __func__)); KASSERT(pd->dst, ("%s: pd->dst is null", __func__)); switch (pd->af) { #ifdef INET case AF_INET: if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET)) pf_change_a(&pd->src->v4.s_addr, pd->ip_sum, nk->addr[pd->sidx].v4.s_addr, 0); if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET)) pf_change_a(&pd->dst->v4.s_addr, pd->ip_sum, nk->addr[pd->didx].v4.s_addr, 0); break; #endif /* INET */ #ifdef INET6 case AF_INET6: if (PF_ANEQ(pd->src, &nk->addr[pd->sidx], AF_INET6)) PF_ACPY(pd->src, &nk->addr[pd->sidx], pd->af); if (PF_ANEQ(pd->dst, &nk->addr[pd->didx], AF_INET6)) PF_ACPY(pd->dst, &nk->addr[pd->didx], pd->af); #endif /* INET6 */ } } return (PF_PASS); } /* * ipoff and off are measured from the start of the mbuf chain. * h must be at "ipoff" on the mbuf chain. */ void * pf_pull_hdr(const struct mbuf *m, int off, void *p, int len, u_short *actionp, u_short *reasonp, sa_family_t af) { switch (af) { #ifdef INET case AF_INET: { const struct ip *h = mtod(m, struct ip *); u_int16_t fragoff = (ntohs(h->ip_off) & IP_OFFMASK) << 3; if (fragoff) { if (fragoff >= len) ACTION_SET(actionp, PF_PASS); else { ACTION_SET(actionp, PF_DROP); REASON_SET(reasonp, PFRES_FRAG); } return (NULL); } if (m->m_pkthdr.len < off + len || ntohs(h->ip_len) < off + len) { ACTION_SET(actionp, PF_DROP); REASON_SET(reasonp, PFRES_SHORT); return (NULL); } break; } #endif /* INET */ #ifdef INET6 case AF_INET6: { const struct ip6_hdr *h = mtod(m, struct ip6_hdr *); if (m->m_pkthdr.len < off + len || (ntohs(h->ip6_plen) + sizeof(struct ip6_hdr)) < (unsigned)(off + len)) { ACTION_SET(actionp, PF_DROP); REASON_SET(reasonp, PFRES_SHORT); return (NULL); } break; } #endif /* INET6 */ } m_copydata(m, off, len, p); return (p); } int pf_routable(struct pf_addr *addr, sa_family_t af, struct pfi_kkif *kif, int rtableid) { struct ifnet *ifp; /* * Skip check for addresses with embedded interface scope, * as they would always match anyway. */ if (af == AF_INET6 && IN6_IS_SCOPE_EMBED(&addr->v6)) return (1); if (af != AF_INET && af != AF_INET6) return (0); if (kif == V_pfi_all) return (1); /* Skip checks for ipsec interfaces */ if (kif != NULL && kif->pfik_ifp->if_type == IFT_ENC) return (1); ifp = (kif != NULL) ? kif->pfik_ifp : NULL; switch (af) { #ifdef INET6 case AF_INET6: return (fib6_check_urpf(rtableid, &addr->v6, 0, NHR_NONE, ifp)); #endif #ifdef INET case AF_INET: return (fib4_check_urpf(rtableid, addr->v4, 0, NHR_NONE, ifp)); #endif } return (0); } #ifdef INET static void pf_route(struct mbuf **m, struct pf_krule *r, struct ifnet *oifp, struct pf_kstate *s, struct pf_pdesc *pd, struct inpcb *inp) { struct mbuf *m0, *m1, *md; struct sockaddr_in dst; struct ip *ip; struct pfi_kkif *nkif = NULL; struct ifnet *ifp = NULL; struct pf_addr naddr; int error = 0; uint16_t ip_len, ip_off; uint16_t tmp; int r_rt, r_dir; KASSERT(m && *m && r && oifp, ("%s: invalid parameters", __func__)); if (s) { r_rt = s->rt; r_dir = s->direction; } else { r_rt = r->rt; r_dir = r->direction; } KASSERT(pd->dir == PF_IN || pd->dir == PF_OUT || r_dir == PF_IN || r_dir == PF_OUT, ("%s: invalid direction", __func__)); if ((pd->pf_mtag == NULL && ((pd->pf_mtag = pf_get_mtag(*m)) == NULL)) || pd->pf_mtag->routed++ > 3) { m0 = *m; *m = NULL; goto bad_locked; } if (r_rt == PF_DUPTO) { if ((pd->pf_mtag->flags & PF_MTAG_FLAG_DUPLICATED)) { if (s == NULL) { ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; } else { ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; /* If pfsync'd */ if (ifp == NULL && r->rpool.cur != NULL) ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; PF_STATE_UNLOCK(s); } if (ifp == oifp) { /* When the 2nd interface is not skipped */ return; } else { m0 = *m; *m = NULL; goto bad; } } else { pd->pf_mtag->flags |= PF_MTAG_FLAG_DUPLICATED; if (((m0 = m_dup(*m, M_NOWAIT)) == NULL)) { if (s) PF_STATE_UNLOCK(s); return; } } } else { if ((r_rt == PF_REPLYTO) == (r_dir == pd->dir)) { pf_dummynet(pd, s, r, m); if (s) PF_STATE_UNLOCK(s); return; } m0 = *m; } ip = mtod(m0, struct ip *); bzero(&dst, sizeof(dst)); dst.sin_family = AF_INET; dst.sin_len = sizeof(dst); dst.sin_addr = ip->ip_dst; bzero(&naddr, sizeof(naddr)); if (s == NULL) { if (TAILQ_EMPTY(&r->rpool.list)) { DPFPRINTF(PF_DEBUG_URGENT, ("%s: TAILQ_EMPTY(&r->rpool.list)\n", __func__)); goto bad_locked; } pf_map_addr(AF_INET, r, (struct pf_addr *)&ip->ip_src, &naddr, &nkif, NULL); if (!PF_AZERO(&naddr, AF_INET)) dst.sin_addr.s_addr = naddr.v4.s_addr; ifp = nkif ? nkif->pfik_ifp : NULL; } else { struct pfi_kkif *kif; if (!PF_AZERO(&s->rt_addr, AF_INET)) dst.sin_addr.s_addr = s->rt_addr.v4.s_addr; ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; kif = s->rt_kif; /* If pfsync'd */ if (ifp == NULL && r->rpool.cur != NULL) { ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; kif = r->rpool.cur->kif; } if (ifp != NULL && kif != NULL && r->rule_flag & PFRULE_IFBOUND && r->rt == PF_REPLYTO && s->kif == V_pfi_all) { s->kif = kif; s->orig_kif = oifp->if_pf_kif; } PF_STATE_UNLOCK(s); } if (ifp == NULL) goto bad; if (pd->dir == PF_IN) { if (pf_test(AF_INET, PF_OUT, PFIL_FWD, ifp, &m0, inp, &pd->act) != PF_PASS) goto bad; else if (m0 == NULL) goto done; if (m0->m_len < sizeof(struct ip)) { DPFPRINTF(PF_DEBUG_URGENT, ("%s: m0->m_len < sizeof(struct ip)\n", __func__)); goto bad; } ip = mtod(m0, struct ip *); } if (ifp->if_flags & IFF_LOOPBACK) m0->m_flags |= M_SKIP_FIREWALL; ip_len = ntohs(ip->ip_len); ip_off = ntohs(ip->ip_off); /* Copied from FreeBSD 10.0-CURRENT ip_output. */ m0->m_pkthdr.csum_flags |= CSUM_IP; if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA & ~ifp->if_hwassist) { in_delayed_cksum(m0); m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; } if (m0->m_pkthdr.csum_flags & CSUM_SCTP & ~ifp->if_hwassist) { pf_sctp_checksum(m0, (uint32_t)(ip->ip_hl << 2)); m0->m_pkthdr.csum_flags &= ~CSUM_SCTP; } if (pd->dir == PF_IN) { /* * Make sure dummynet gets the correct direction, in case it needs to * re-inject later. */ pd->dir = PF_OUT; /* * The following processing is actually the rest of the inbound processing, even * though we've marked it as outbound (so we don't look through dummynet) and it * happens after the outbound processing (pf_test(PF_OUT) above). * Swap the dummynet pipe numbers, because it's going to come to the wrong * conclusion about what direction it's processing, and we can't fix it or it * will re-inject incorrectly. Swapping the pipe numbers means that its incorrect * decision will pick the right pipe, and everything will mostly work as expected. */ tmp = pd->act.dnrpipe; pd->act.dnrpipe = pd->act.dnpipe; pd->act.dnpipe = tmp; } /* * If small enough for interface, or the interface will take * care of the fragmentation for us, we can just send directly. */ if (ip_len <= ifp->if_mtu || (m0->m_pkthdr.csum_flags & ifp->if_hwassist & CSUM_TSO) != 0) { ip->ip_sum = 0; if (m0->m_pkthdr.csum_flags & CSUM_IP & ~ifp->if_hwassist) { ip->ip_sum = in_cksum(m0, ip->ip_hl << 2); m0->m_pkthdr.csum_flags &= ~CSUM_IP; } m_clrprotoflags(m0); /* Avoid confusing lower layers. */ md = m0; error = pf_dummynet_route(pd, s, r, ifp, sintosa(&dst), &md); if (md != NULL) error = (*ifp->if_output)(ifp, md, sintosa(&dst), NULL); goto done; } /* Balk when DF bit is set or the interface didn't support TSO. */ if ((ip_off & IP_DF) || (m0->m_pkthdr.csum_flags & CSUM_TSO)) { error = EMSGSIZE; KMOD_IPSTAT_INC(ips_cantfrag); if (r_rt != PF_DUPTO) { if (s && pd->nat_rule != NULL) PACKET_UNDO_NAT(m0, pd, (ip->ip_hl << 2) + (ip_off & IP_OFFMASK), s); icmp_error(m0, ICMP_UNREACH, ICMP_UNREACH_NEEDFRAG, 0, ifp->if_mtu); goto done; } else goto bad; } error = ip_fragment(ip, &m0, ifp->if_mtu, ifp->if_hwassist); if (error) goto bad; for (; m0; m0 = m1) { m1 = m0->m_nextpkt; m0->m_nextpkt = NULL; if (error == 0) { m_clrprotoflags(m0); md = m0; pd->pf_mtag = pf_find_mtag(md); error = pf_dummynet_route(pd, s, r, ifp, sintosa(&dst), &md); if (md != NULL) error = (*ifp->if_output)(ifp, md, sintosa(&dst), NULL); } else m_freem(m0); } if (error == 0) KMOD_IPSTAT_INC(ips_fragmented); done: if (r_rt != PF_DUPTO) *m = NULL; return; bad_locked: if (s) PF_STATE_UNLOCK(s); bad: m_freem(m0); goto done; } #endif /* INET */ #ifdef INET6 static void pf_route6(struct mbuf **m, struct pf_krule *r, struct ifnet *oifp, struct pf_kstate *s, struct pf_pdesc *pd, struct inpcb *inp) { struct mbuf *m0, *md; struct sockaddr_in6 dst; struct ip6_hdr *ip6; struct pfi_kkif *nkif = NULL; struct ifnet *ifp = NULL; struct pf_addr naddr; int r_rt, r_dir; KASSERT(m && *m && r && oifp, ("%s: invalid parameters", __func__)); if (s) { r_rt = s->rt; r_dir = s->direction; } else { r_rt = r->rt; r_dir = r->direction; } KASSERT(pd->dir == PF_IN || pd->dir == PF_OUT || r_dir == PF_IN || r_dir == PF_OUT, ("%s: invalid direction", __func__)); if ((pd->pf_mtag == NULL && ((pd->pf_mtag = pf_get_mtag(*m)) == NULL)) || pd->pf_mtag->routed++ > 3) { m0 = *m; *m = NULL; goto bad_locked; } if (r_rt == PF_DUPTO) { if ((pd->pf_mtag->flags & PF_MTAG_FLAG_DUPLICATED)) { if (s == NULL) { ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; } else { ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; /* If pfsync'd */ if (ifp == NULL && r->rpool.cur != NULL) ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; PF_STATE_UNLOCK(s); } if (ifp == oifp) { /* When the 2nd interface is not skipped */ return; } else { m0 = *m; *m = NULL; goto bad; } } else { pd->pf_mtag->flags |= PF_MTAG_FLAG_DUPLICATED; if (((m0 = m_dup(*m, M_NOWAIT)) == NULL)) { if (s) PF_STATE_UNLOCK(s); return; } } } else { if ((r_rt == PF_REPLYTO) == (r_dir == pd->dir)) { pf_dummynet(pd, s, r, m); if (s) PF_STATE_UNLOCK(s); return; } m0 = *m; } ip6 = mtod(m0, struct ip6_hdr *); bzero(&dst, sizeof(dst)); dst.sin6_family = AF_INET6; dst.sin6_len = sizeof(dst); dst.sin6_addr = ip6->ip6_dst; bzero(&naddr, sizeof(naddr)); if (s == NULL) { if (TAILQ_EMPTY(&r->rpool.list)) { DPFPRINTF(PF_DEBUG_URGENT, ("%s: TAILQ_EMPTY(&r->rpool.list)\n", __func__)); goto bad_locked; } pf_map_addr(AF_INET6, r, (struct pf_addr *)&ip6->ip6_src, &naddr, &nkif, NULL); if (!PF_AZERO(&naddr, AF_INET6)) PF_ACPY((struct pf_addr *)&dst.sin6_addr, &naddr, AF_INET6); ifp = nkif ? nkif->pfik_ifp : NULL; } else { struct pfi_kkif *kif; if (!PF_AZERO(&s->rt_addr, AF_INET6)) PF_ACPY((struct pf_addr *)&dst.sin6_addr, &s->rt_addr, AF_INET6); ifp = s->rt_kif ? s->rt_kif->pfik_ifp : NULL; kif = s->rt_kif; /* If pfsync'd */ if (ifp == NULL && r->rpool.cur != NULL) { ifp = r->rpool.cur->kif ? r->rpool.cur->kif->pfik_ifp : NULL; kif = r->rpool.cur->kif; } if (ifp != NULL && kif != NULL && r->rule_flag & PFRULE_IFBOUND && r->rt == PF_REPLYTO && s->kif == V_pfi_all) { s->kif = kif; s->orig_kif = oifp->if_pf_kif; } } if (s) PF_STATE_UNLOCK(s); if (ifp == NULL) goto bad; if (pd->dir == PF_IN) { if (pf_test(AF_INET6, PF_OUT, PFIL_FWD, ifp, &m0, inp, &pd->act) != PF_PASS) goto bad; else if (m0 == NULL) goto done; if (m0->m_len < sizeof(struct ip6_hdr)) { DPFPRINTF(PF_DEBUG_URGENT, ("%s: m0->m_len < sizeof(struct ip6_hdr)\n", __func__)); goto bad; } ip6 = mtod(m0, struct ip6_hdr *); } if (ifp->if_flags & IFF_LOOPBACK) m0->m_flags |= M_SKIP_FIREWALL; if (m0->m_pkthdr.csum_flags & CSUM_DELAY_DATA_IPV6 & ~ifp->if_hwassist) { uint32_t plen = m0->m_pkthdr.len - sizeof(*ip6); in6_delayed_cksum(m0, plen, sizeof(struct ip6_hdr)); m0->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA_IPV6; } /* * If the packet is too large for the outgoing interface, * send back an icmp6 error. */ if (IN6_IS_SCOPE_EMBED(&dst.sin6_addr)) dst.sin6_addr.s6_addr16[1] = htons(ifp->if_index); if ((u_long)m0->m_pkthdr.len <= ifp->if_mtu) { md = m0; pf_dummynet_route(pd, s, r, ifp, sintosa(&dst), &md); if (md != NULL) nd6_output_ifp(ifp, ifp, md, &dst, NULL); } else { in6_ifstat_inc(ifp, ifs6_in_toobig); if (r_rt != PF_DUPTO) { if (s && pd->nat_rule != NULL) PACKET_UNDO_NAT(m0, pd, ((caddr_t)ip6 - m0->m_data) + sizeof(struct ip6_hdr), s); icmp6_error(m0, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu); } else goto bad; } done: if (r_rt != PF_DUPTO) *m = NULL; return; bad_locked: if (s) PF_STATE_UNLOCK(s); bad: m_freem(m0); goto done; } #endif /* INET6 */ /* * FreeBSD supports cksum offloads for the following drivers. * em(4), fxp(4), lge(4), nge(4), re(4), ti(4), txp(4), xl(4) * * CSUM_DATA_VALID | CSUM_PSEUDO_HDR : * network driver performed cksum including pseudo header, need to verify * csum_data * CSUM_DATA_VALID : * network driver performed cksum, needs to additional pseudo header * cksum computation with partial csum_data(i.e. lack of H/W support for * pseudo header, for instance sk(4) and possibly gem(4)) * * After validating the cksum of packet, set both flag CSUM_DATA_VALID and * CSUM_PSEUDO_HDR in order to avoid recomputation of the cksum in upper * TCP/UDP layer. * Also, set csum_data to 0xffff to force cksum validation. */ static int pf_check_proto_cksum(struct mbuf *m, int off, int len, u_int8_t p, sa_family_t af) { u_int16_t sum = 0; int hw_assist = 0; struct ip *ip; if (off < sizeof(struct ip) || len < sizeof(struct udphdr)) return (1); if (m->m_pkthdr.len < off + len) return (1); switch (p) { case IPPROTO_TCP: if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { sum = m->m_pkthdr.csum_data; } else { ip = mtod(m, struct ip *); sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl((u_short)len + m->m_pkthdr.csum_data + IPPROTO_TCP)); } sum ^= 0xffff; ++hw_assist; } break; case IPPROTO_UDP: if (m->m_pkthdr.csum_flags & CSUM_DATA_VALID) { if (m->m_pkthdr.csum_flags & CSUM_PSEUDO_HDR) { sum = m->m_pkthdr.csum_data; } else { ip = mtod(m, struct ip *); sum = in_pseudo(ip->ip_src.s_addr, ip->ip_dst.s_addr, htonl((u_short)len + m->m_pkthdr.csum_data + IPPROTO_UDP)); } sum ^= 0xffff; ++hw_assist; } break; case IPPROTO_ICMP: #ifdef INET6 case IPPROTO_ICMPV6: #endif /* INET6 */ break; default: return (1); } if (!hw_assist) { switch (af) { case AF_INET: if (p == IPPROTO_ICMP) { if (m->m_len < off) return (1); m->m_data += off; m->m_len -= off; sum = in_cksum(m, len); m->m_data -= off; m->m_len += off; } else { if (m->m_len < sizeof(struct ip)) return (1); sum = in4_cksum(m, p, off, len); } break; #ifdef INET6 case AF_INET6: if (m->m_len < sizeof(struct ip6_hdr)) return (1); sum = in6_cksum(m, p, off, len); break; #endif /* INET6 */ default: return (1); } } if (sum) { switch (p) { case IPPROTO_TCP: { KMOD_TCPSTAT_INC(tcps_rcvbadsum); break; } case IPPROTO_UDP: { KMOD_UDPSTAT_INC(udps_badsum); break; } #ifdef INET case IPPROTO_ICMP: { KMOD_ICMPSTAT_INC(icps_checksum); break; } #endif #ifdef INET6 case IPPROTO_ICMPV6: { KMOD_ICMP6STAT_INC(icp6s_checksum); break; } #endif /* INET6 */ } return (1); } else { if (p == IPPROTO_TCP || p == IPPROTO_UDP) { m->m_pkthdr.csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR); m->m_pkthdr.csum_data = 0xffff; } } return (0); } static bool pf_pdesc_to_dnflow(const struct pf_pdesc *pd, const struct pf_krule *r, const struct pf_kstate *s, struct ip_fw_args *dnflow) { int dndir = r->direction; if (s && dndir == PF_INOUT) { dndir = s->direction; } else if (dndir == PF_INOUT) { /* Assume primary direction. Happens when we've set dnpipe in * the ethernet level code. */ dndir = pd->dir; } if (pd->pf_mtag->flags & PF_MTAG_FLAG_DUMMYNETED) return (false); memset(dnflow, 0, sizeof(*dnflow)); if (pd->dport != NULL) dnflow->f_id.dst_port = ntohs(*pd->dport); if (pd->sport != NULL) dnflow->f_id.src_port = ntohs(*pd->sport); if (pd->dir == PF_IN) dnflow->flags |= IPFW_ARGS_IN; else dnflow->flags |= IPFW_ARGS_OUT; if (pd->dir != dndir && pd->act.dnrpipe) { dnflow->rule.info = pd->act.dnrpipe; } else if (pd->dir == dndir && pd->act.dnpipe) { dnflow->rule.info = pd->act.dnpipe; } else { return (false); } dnflow->rule.info |= IPFW_IS_DUMMYNET; if (r->free_flags & PFRULE_DN_IS_PIPE || pd->act.flags & PFSTATE_DN_IS_PIPE) dnflow->rule.info |= IPFW_IS_PIPE; dnflow->f_id.proto = pd->proto; dnflow->f_id.extra = dnflow->rule.info; switch (pd->af) { case AF_INET: dnflow->f_id.addr_type = 4; dnflow->f_id.src_ip = ntohl(pd->src->v4.s_addr); dnflow->f_id.dst_ip = ntohl(pd->dst->v4.s_addr); break; case AF_INET6: dnflow->flags |= IPFW_ARGS_IP6; dnflow->f_id.addr_type = 6; dnflow->f_id.src_ip6 = pd->src->v6; dnflow->f_id.dst_ip6 = pd->dst->v6; break; default: panic("Invalid AF"); break; } return (true); } int pf_test_eth(int dir, int pflags, struct ifnet *ifp, struct mbuf **m0, struct inpcb *inp) { struct pfi_kkif *kif; struct mbuf *m = *m0; M_ASSERTPKTHDR(m); MPASS(ifp->if_vnet == curvnet); NET_EPOCH_ASSERT(); if (!V_pf_status.running) return (PF_PASS); kif = (struct pfi_kkif *)ifp->if_pf_kif; if (kif == NULL) { DPFPRINTF(PF_DEBUG_URGENT, ("%s: kif == NULL, if_xname %s\n", __func__, ifp->if_xname)); return (PF_DROP); } if (kif->pfik_flags & PFI_IFLAG_SKIP) return (PF_PASS); if (m->m_flags & M_SKIP_FIREWALL) return (PF_PASS); if (__predict_false(! M_WRITABLE(*m0))) { m = *m0 = m_unshare(*m0, M_NOWAIT); if (*m0 == NULL) return (PF_DROP); } /* Stateless! */ return (pf_test_eth_rule(dir, kif, m0)); } static __inline void pf_dummynet_flag_remove(struct mbuf *m, struct pf_mtag *pf_mtag) { struct m_tag *mtag; pf_mtag->flags &= ~PF_MTAG_FLAG_DUMMYNET; /* dummynet adds this tag, but pf does not need it, * and keeping it creates unexpected behavior, * e.g. in case of divert(4) usage right after dummynet. */ mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL); if (mtag != NULL) m_tag_delete(m, mtag); } static int pf_dummynet(struct pf_pdesc *pd, struct pf_kstate *s, struct pf_krule *r, struct mbuf **m0) { return (pf_dummynet_route(pd, s, r, NULL, NULL, m0)); } static int pf_dummynet_route(struct pf_pdesc *pd, struct pf_kstate *s, struct pf_krule *r, struct ifnet *ifp, struct sockaddr *sa, struct mbuf **m0) { NET_EPOCH_ASSERT(); if (pd->act.dnpipe || pd->act.dnrpipe) { struct ip_fw_args dnflow; if (ip_dn_io_ptr == NULL) { m_freem(*m0); *m0 = NULL; return (ENOMEM); } if (pd->pf_mtag == NULL && ((pd->pf_mtag = pf_get_mtag(*m0)) == NULL)) { m_freem(*m0); *m0 = NULL; return (ENOMEM); } if (ifp != NULL) { pd->pf_mtag->flags |= PF_MTAG_FLAG_ROUTE_TO; pd->pf_mtag->if_index = ifp->if_index; pd->pf_mtag->if_idxgen = ifp->if_idxgen; MPASS(sa != NULL); if (pd->af == AF_INET) memcpy(&pd->pf_mtag->dst, sa, sizeof(struct sockaddr_in)); else memcpy(&pd->pf_mtag->dst, sa, sizeof(struct sockaddr_in6)); } if (s != NULL && s->nat_rule.ptr != NULL && s->nat_rule.ptr->action == PF_RDR && ( #ifdef INET (pd->af == AF_INET && IN_LOOPBACK(ntohl(pd->dst->v4.s_addr))) || #endif (pd->af == AF_INET6 && IN6_IS_ADDR_LOOPBACK(&pd->dst->v6)))) { /* * If we're redirecting to loopback mark this packet * as being local. Otherwise it might get dropped * if dummynet re-injects. */ (*m0)->m_pkthdr.rcvif = V_loif; } if (pf_pdesc_to_dnflow(pd, r, s, &dnflow)) { pd->pf_mtag->flags |= PF_MTAG_FLAG_DUMMYNET; pd->pf_mtag->flags |= PF_MTAG_FLAG_DUMMYNETED; ip_dn_io_ptr(m0, &dnflow); if (*m0 != NULL) { pd->pf_mtag->flags &= ~PF_MTAG_FLAG_ROUTE_TO; pf_dummynet_flag_remove(*m0, pd->pf_mtag); } } } return (0); } static int pf_setup_pdesc(sa_family_t af, int dir, struct pf_pdesc *pd, struct mbuf *m, u_short *action, u_short *reason, struct pfi_kkif *kif, struct pf_krule **a, - struct pf_krule **r, struct pf_kruleset **ruleset, int *off, int *hdrlen, + struct pf_krule **r, struct pf_kstate **s, struct pf_kruleset **ruleset, + int *off, int *hdrlen, struct inpcb *inp, struct pf_rule_actions *default_actions) { TAILQ_INIT(&pd->sctp_multihome_jobs); if (default_actions != NULL) memcpy(&pd->act, default_actions, sizeof(pd->act)); pd->pf_mtag = pf_find_mtag(m); if (pd->pf_mtag && pd->pf_mtag->dnpipe) { pd->act.dnpipe = pd->pf_mtag->dnpipe; pd->act.flags = pd->pf_mtag->dnflags; } pd->af = af; switch (af) { #ifdef INET case AF_INET: { struct ip *h; h = mtod(m, struct ip *); *off = h->ip_hl << 2; if (*off < (int)sizeof(*h)) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } pd->src = (struct pf_addr *)&h->ip_src; pd->dst = (struct pf_addr *)&h->ip_dst; pd->sport = pd->dport = NULL; pd->ip_sum = &h->ip_sum; pd->proto_sum = NULL; - pd->proto = h->ip_p; + pd->virtual_proto = pd->proto = h->ip_p; pd->dir = dir; pd->sidx = (dir == PF_IN) ? 0 : 1; pd->didx = (dir == PF_IN) ? 1 : 0; pd->tos = h->ip_tos; pd->tot_len = ntohs(h->ip_len); pd->act.rtableid = -1; if (h->ip_hl > 5) /* has options */ pd->badopts++; - /* fragments not reassembled handled later */ - if (h->ip_off & htons(IP_MF | IP_OFFMASK)) - return (0); + if (h->ip_off & htons(IP_MF | IP_OFFMASK)) { + /* + * handle fragments that aren't reassembled by + * normalization + */ + pd->virtual_proto = PF_VPROTO_FRAGMENT; + if (kif == NULL || r == NULL) /* pflog */ + *action = PF_DROP; + else + *action = pf_test_rule(r, s, kif, m, *off, + pd, a, ruleset, inp, *hdrlen); + if (*action != PF_PASS) + REASON_SET(reason, PFRES_FRAG); + return (-1); + } break; } #endif #ifdef INET6 case AF_INET6: { struct ip6_hdr *h; int terminal = 0; h = mtod(m, struct ip6_hdr *); pd->src = (struct pf_addr *)&h->ip6_src; pd->dst = (struct pf_addr *)&h->ip6_dst; pd->sport = pd->dport = NULL; pd->ip_sum = NULL; pd->proto_sum = NULL; pd->dir = dir; pd->sidx = (dir == PF_IN) ? 0 : 1; pd->didx = (dir == PF_IN) ? 1 : 0; pd->tos = IPV6_DSCP(h); pd->tot_len = ntohs(h->ip6_plen) + sizeof(struct ip6_hdr); *off = ((caddr_t)h - m->m_data) + sizeof(struct ip6_hdr); - pd->proto = h->ip6_nxt; + pd->virtual_proto = pd->proto = h->ip6_nxt; pd->act.rtableid = -1; do { switch (pd->proto) { case IPPROTO_FRAGMENT: + pd->virtual_proto = PF_VPROTO_FRAGMENT; if (kif == NULL || r == NULL) /* pflog */ *action = PF_DROP; else - *action = pf_test_fragment(r, kif, - m, pd, a, ruleset); + *action = pf_test_rule(r, s, kif, m, *off, + pd, a, ruleset, inp, + *hdrlen); if (*action == PF_DROP) REASON_SET(reason, PFRES_FRAG); return (-1); case IPPROTO_ROUTING: { struct ip6_rthdr rthdr; if (pd->badopts++) { DPFPRINTF(PF_DEBUG_MISC, ("pf: IPv6 more than one rthdr")); *action = PF_DROP; REASON_SET(reason, PFRES_IPOPTIONS); return (-1); } if (!pf_pull_hdr(m, *off, &rthdr, sizeof(rthdr), NULL, reason, pd->af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: IPv6 short rthdr")); *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } if (rthdr.ip6r_type == IPV6_RTHDR_TYPE_0) { DPFPRINTF(PF_DEBUG_MISC, ("pf: IPv6 rthdr0")); *action = PF_DROP; REASON_SET(reason, PFRES_IPOPTIONS); return (-1); } /* FALLTHROUGH */ } case IPPROTO_AH: case IPPROTO_HOPOPTS: case IPPROTO_DSTOPTS: { /* get next header and header length */ struct ip6_ext opt6; if (!pf_pull_hdr(m, *off, &opt6, sizeof(opt6), NULL, reason, pd->af)) { DPFPRINTF(PF_DEBUG_MISC, ("pf: IPv6 short opt")); *action = PF_DROP; return (-1); } if (pd->proto == IPPROTO_AH) *off += (opt6.ip6e_len + 2) * 4; else *off += (opt6.ip6e_len + 1) * 8; - pd->proto = opt6.ip6e_nxt; + pd->virtual_proto = pd->proto = opt6.ip6e_nxt; /* goto the next header */ break; } default: terminal++; break; } } while (!terminal); break; } #endif default: panic("pf_setup_pdesc called with illegal af %u", af); } switch (pd->proto) { case IPPROTO_TCP: { struct tcphdr *th = &pd->hdr.tcp; if (!pf_pull_hdr(m, *off, th, sizeof(*th), action, reason, af)) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } *hdrlen = sizeof(*th); pd->p_len = pd->tot_len - *off - (th->th_off << 2); pd->sport = &th->th_sport; pd->dport = &th->th_dport; break; } case IPPROTO_UDP: { struct udphdr *uh = &pd->hdr.udp; if (!pf_pull_hdr(m, *off, uh, sizeof(*uh), action, reason, af)) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } *hdrlen = sizeof(*uh); if (uh->uh_dport == 0 || ntohs(uh->uh_ulen) > m->m_pkthdr.len - *off || ntohs(uh->uh_ulen) < sizeof(struct udphdr)) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } pd->sport = &uh->uh_sport; pd->dport = &uh->uh_dport; break; } case IPPROTO_SCTP: { if (!pf_pull_hdr(m, *off, &pd->hdr.sctp, sizeof(pd->hdr.sctp), action, reason, af)) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } *hdrlen = sizeof(pd->hdr.sctp); pd->p_len = pd->tot_len - *off; pd->sport = &pd->hdr.sctp.src_port; pd->dport = &pd->hdr.sctp.dest_port; if (pd->hdr.sctp.src_port == 0 || pd->hdr.sctp.dest_port == 0) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } if (pf_scan_sctp(m, *off, pd, kif) != PF_PASS) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } break; } case IPPROTO_ICMP: { if (!pf_pull_hdr(m, *off, &pd->hdr.icmp, ICMP_MINLEN, action, reason, af)) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } *hdrlen = ICMP_MINLEN; break; } #ifdef INET6 case IPPROTO_ICMPV6: { size_t icmp_hlen = sizeof(struct icmp6_hdr); if (!pf_pull_hdr(m, *off, &pd->hdr.icmp6, icmp_hlen, action, reason, af)) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } /* ICMP headers we look further into to match state */ switch (pd->hdr.icmp6.icmp6_type) { case MLD_LISTENER_QUERY: case MLD_LISTENER_REPORT: icmp_hlen = sizeof(struct mld_hdr); break; case ND_NEIGHBOR_SOLICIT: case ND_NEIGHBOR_ADVERT: icmp_hlen = sizeof(struct nd_neighbor_solicit); break; } if (icmp_hlen > sizeof(struct icmp6_hdr) && !pf_pull_hdr(m, *off, &pd->hdr.icmp6, icmp_hlen, action, reason, af)) { *action = PF_DROP; REASON_SET(reason, PFRES_SHORT); return (-1); } *hdrlen = icmp_hlen; break; } #endif } return (0); } static void pf_counters_inc(int action, struct pf_pdesc *pd, struct pfi_kkif *kif, struct pf_kstate *s, struct pf_krule *r, struct pf_krule *a) { struct pf_krule *tr, *nr; int dir = pd->dir; int dirndx; pf_counter_u64_critical_enter(); pf_counter_u64_add_protected( &kif->pfik_bytes[pd->af == AF_INET6][dir == PF_OUT][action != PF_PASS], pd->tot_len); pf_counter_u64_add_protected( &kif->pfik_packets[pd->af == AF_INET6][dir == PF_OUT][action != PF_PASS], 1); if (action == PF_PASS || r->action == PF_DROP) { dirndx = (dir == PF_OUT); pf_counter_u64_add_protected(&r->packets[dirndx], 1); pf_counter_u64_add_protected(&r->bytes[dirndx], pd->tot_len); pf_update_timestamp(r); if (a != NULL) { pf_counter_u64_add_protected(&a->packets[dirndx], 1); pf_counter_u64_add_protected(&a->bytes[dirndx], pd->tot_len); } if (s != NULL) { struct pf_krule_item *ri; if (s->nat_rule.ptr != NULL) { pf_counter_u64_add_protected(&s->nat_rule.ptr->packets[dirndx], 1); pf_counter_u64_add_protected(&s->nat_rule.ptr->bytes[dirndx], pd->tot_len); } if (s->src_node != NULL) { counter_u64_add(s->src_node->packets[dirndx], 1); counter_u64_add(s->src_node->bytes[dirndx], pd->tot_len); } if (s->nat_src_node != NULL) { counter_u64_add(s->nat_src_node->packets[dirndx], 1); counter_u64_add(s->nat_src_node->bytes[dirndx], pd->tot_len); } dirndx = (dir == s->direction) ? 0 : 1; s->packets[dirndx]++; s->bytes[dirndx] += pd->tot_len; SLIST_FOREACH(ri, &s->match_rules, entry) { pf_counter_u64_add_protected(&ri->r->packets[dirndx], 1); pf_counter_u64_add_protected(&ri->r->bytes[dirndx], pd->tot_len); } } tr = r; nr = (s != NULL) ? s->nat_rule.ptr : pd->nat_rule; if (nr != NULL && r == &V_pf_default_rule) tr = nr; if (tr->src.addr.type == PF_ADDR_TABLE) pfr_update_stats(tr->src.addr.p.tbl, (s == NULL) ? pd->src : &s->key[(s->direction == PF_IN)]-> addr[(s->direction == PF_OUT)], pd->af, pd->tot_len, dir == PF_OUT, r->action == PF_PASS, tr->src.neg); if (tr->dst.addr.type == PF_ADDR_TABLE) pfr_update_stats(tr->dst.addr.p.tbl, (s == NULL) ? pd->dst : &s->key[(s->direction == PF_IN)]-> addr[(s->direction == PF_IN)], pd->af, pd->tot_len, dir == PF_OUT, r->action == PF_PASS, tr->dst.neg); } pf_counter_u64_critical_exit(); } #if defined(INET) || defined(INET6) int pf_test(sa_family_t af, int dir, int pflags, struct ifnet *ifp, struct mbuf **m0, struct inpcb *inp, struct pf_rule_actions *default_actions) { struct pfi_kkif *kif; u_short action, reason = 0; struct mbuf *m = *m0; struct ip *h = NULL; struct ip6_hdr *h6 = NULL; struct m_tag *mtag; struct pf_krule *a = NULL, *r = &V_pf_default_rule; struct pf_kstate *s = NULL; struct pf_kruleset *ruleset = NULL; struct pf_pdesc pd; int off, hdrlen, use_2nd_queue = 0; uint16_t tag; uint8_t rt; uint8_t ttl; PF_RULES_RLOCK_TRACKER; KASSERT(dir == PF_IN || dir == PF_OUT, ("%s: bad direction %d\n", __func__, dir)); M_ASSERTPKTHDR(m); if (!V_pf_status.running) return (PF_PASS); PF_RULES_RLOCK(); kif = (struct pfi_kkif *)ifp->if_pf_kif; if (__predict_false(kif == NULL)) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_test: kif == NULL, if_xname %s\n", ifp->if_xname)); PF_RULES_RUNLOCK(); return (PF_DROP); } if (kif->pfik_flags & PFI_IFLAG_SKIP) { PF_RULES_RUNLOCK(); return (PF_PASS); } if (m->m_flags & M_SKIP_FIREWALL) { PF_RULES_RUNLOCK(); return (PF_PASS); } if (__predict_false(! M_WRITABLE(*m0))) { m = *m0 = m_unshare(*m0, M_NOWAIT); if (*m0 == NULL) return (PF_DROP); } memset(&pd, 0, sizeof(pd)); pd.dir = dir; switch (af) { #ifdef INET case AF_INET: if (__predict_false(m->m_len < sizeof(struct ip)) && (m = *m0 = m_pullup(*m0, sizeof(struct ip))) == NULL) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_test: m_len < sizeof(struct ip), pullup failed\n")); PF_RULES_RUNLOCK(); return (PF_DROP); } if (pf_normalize_ip(m0, kif, &reason, &pd) != PF_PASS) { /* We do IP header normalization and packet reassembly here */ action = PF_DROP; goto done; } m = *m0; /* pf_normalize messes with m0 */ h = mtod(m, struct ip *); ttl = h->ip_ttl; break; #endif #ifdef INET6 case AF_INET6: /* * If we end up changing IP addresses (e.g. binat) the stack may get * confused and fail to send the icmp6 packet too big error. Just send * it here, before we do any NAT. */ if (dir == PF_OUT && pflags & PFIL_FWD && IN6_LINKMTU(ifp) < pf_max_frag_size(m)) { PF_RULES_RUNLOCK(); *m0 = NULL; icmp6_error(m, ICMP6_PACKET_TOO_BIG, 0, IN6_LINKMTU(ifp)); return (PF_DROP); } if (__predict_false(m->m_len < sizeof(struct ip6_hdr)) && (m = *m0 = m_pullup(*m0, sizeof(struct ip6_hdr))) == NULL) { DPFPRINTF(PF_DEBUG_URGENT, ("pf_test6: m_len < sizeof(struct ip6_hdr)" ", pullup failed\n")); PF_RULES_RUNLOCK(); return (PF_DROP); } /* We do IP header normalization and packet reassembly here */ if (pf_normalize_ip6(m0, kif, &reason, &pd) != PF_PASS) { action = PF_DROP; goto done; } m = *m0; /* pf_normalize messes with m0 */ h6 = mtod(m, struct ip6_hdr *); ttl = h6->ip6_hlim; break; #endif default: panic("Unknown af %d", af); } if (pf_setup_pdesc(af, dir, &pd, m, &action, &reason, kif, &a, &r, - &ruleset, &off, &hdrlen, default_actions) == -1) { + &s, &ruleset, &off, &hdrlen, inp, default_actions) == -1) { if (action != PF_PASS) pd.act.log |= PF_LOG_FORCE; goto done; } if (pd.pf_mtag != NULL && (pd.pf_mtag->flags & PF_MTAG_FLAG_ROUTE_TO)) { pd.pf_mtag->flags &= ~PF_MTAG_FLAG_ROUTE_TO; ifp = ifnet_byindexgen(pd.pf_mtag->if_index, pd.pf_mtag->if_idxgen); if (ifp == NULL || ifp->if_flags & IFF_DYING) { PF_RULES_RUNLOCK(); m_freem(*m0); *m0 = NULL; return (PF_PASS); } PF_RULES_RUNLOCK(); (ifp->if_output)(ifp, m, sintosa(&pd.pf_mtag->dst), NULL); *m0 = NULL; return (PF_PASS); } if (ip_dn_io_ptr != NULL && pd.pf_mtag != NULL && pd.pf_mtag->flags & PF_MTAG_FLAG_DUMMYNET) { /* Dummynet re-injects packets after they've * completed their delay. We've already * processed them, so pass unconditionally. */ /* But only once. We may see the packet multiple times (e.g. * PFIL_IN/PFIL_OUT). */ pf_dummynet_flag_remove(m, pd.pf_mtag); PF_RULES_RUNLOCK(); return (PF_PASS); } if (__predict_false(ip_divert_ptr != NULL) && ((mtag = m_tag_locate(m, MTAG_PF_DIVERT, 0, NULL)) != NULL)) { struct pf_divert_mtag *dt = (struct pf_divert_mtag *)(mtag+1); if ((dt->idir == PF_DIVERT_MTAG_DIR_IN && dir == PF_IN) || (dt->idir == PF_DIVERT_MTAG_DIR_OUT && dir == PF_OUT)) { if (pd.pf_mtag == NULL && ((pd.pf_mtag = pf_get_mtag(m)) == NULL)) { action = PF_DROP; goto done; } pd.pf_mtag->flags |= PF_MTAG_FLAG_PACKET_LOOPED; } if (pd.pf_mtag && pd.pf_mtag->flags & PF_MTAG_FLAG_FASTFWD_OURS_PRESENT) { m->m_flags |= M_FASTFWD_OURS; pd.pf_mtag->flags &= ~PF_MTAG_FLAG_FASTFWD_OURS_PRESENT; } m_tag_delete(m, mtag); mtag = m_tag_locate(m, MTAG_IPFW_RULE, 0, NULL); if (mtag != NULL) m_tag_delete(m, mtag); } - switch (af) { -#ifdef INET - case AF_INET: - /* handle fragments that didn't get reassembled by normalization */ - if (h->ip_off & htons(IP_MF | IP_OFFMASK)) { - action = pf_test_fragment(&r, kif, m, &pd, &a, &ruleset); - goto done; - } - break; -#endif #ifdef INET6 - case AF_INET6: - /* - * we do not support jumbogram. if we keep going, zero ip6_plen - * will do something bad, so drop the packet for now. - */ - if (htons(h6->ip6_plen) == 0) { - action = PF_DROP; - REASON_SET(&reason, PFRES_NORM); /*XXX*/ - goto done; - } - break; -#endif - default: - panic("Unknown af %d", af); + /* + * we do not support jumbogram. if we keep going, zero ip6_plen + * will do something bad, so drop the packet for now. + */ + if (af == AF_INET6 && htons(h6->ip6_plen) == 0) { + action = PF_DROP; + REASON_SET(&reason, PFRES_NORM); /*XXX*/ + goto done; } +#endif switch (pd.proto) { case IPPROTO_TCP: { /* Respond to SYN with a syncookie. */ if ((pd.hdr.tcp.th_flags & (TH_SYN|TH_ACK|TH_RST)) == TH_SYN && pd.dir == PF_IN && pf_synflood_check(&pd)) { pf_syncookie_send(m, off, &pd); action = PF_DROP; break; } if ((pd.hdr.tcp.th_flags & TH_ACK) && pd.p_len == 0) use_2nd_queue = 1; action = pf_normalize_tcp(kif, m, 0, off, &pd); if (action == PF_DROP) goto done; action = pf_test_state_tcp(&s, kif, m, off, &pd, &reason); if (action == PF_PASS) { if (V_pfsync_update_state_ptr != NULL) V_pfsync_update_state_ptr(s); r = s->rule.ptr; a = s->anchor.ptr; } else if (s == NULL) { /* Validate remote SYN|ACK, re-create original SYN if * valid. */ if ((pd.hdr.tcp.th_flags & (TH_SYN|TH_ACK|TH_RST)) == TH_ACK && pf_syncookie_validate(&pd) && pd.dir == PF_IN) { struct mbuf *msyn; msyn = pf_syncookie_recreate_syn(ttl, off, &pd); if (msyn == NULL) { action = PF_DROP; break; } action = pf_test(af, dir, pflags, ifp, &msyn, inp, &pd.act); m_freem(msyn); if (action != PF_PASS) break; action = pf_test_state_tcp(&s, kif, m, off, &pd, &reason); if (action != PF_PASS || s == NULL) { action = PF_DROP; break; } s->src.seqhi = ntohl(pd.hdr.tcp.th_ack) - 1; s->src.seqlo = ntohl(pd.hdr.tcp.th_seq) - 1; pf_set_protostate(s, PF_PEER_SRC, PF_TCPS_PROXY_DST); action = pf_synproxy(&pd, &s, &reason); break; } else { action = pf_test_rule(&r, &s, kif, m, off, &pd, &a, &ruleset, inp, hdrlen); } } break; } case IPPROTO_UDP: { action = pf_test_state_udp(&s, kif, m, off, &pd); if (action == PF_PASS) { if (V_pfsync_update_state_ptr != NULL) V_pfsync_update_state_ptr(s); r = s->rule.ptr; a = s->anchor.ptr; } else if (s == NULL) action = pf_test_rule(&r, &s, kif, m, off, &pd, &a, &ruleset, inp, hdrlen); break; } case IPPROTO_SCTP: { action = pf_normalize_sctp(dir, kif, m, 0, off, &pd); if (action == PF_DROP) goto done; action = pf_test_state_sctp(&s, kif, m, off, &pd, &reason); if (action == PF_PASS) { if (V_pfsync_update_state_ptr != NULL) V_pfsync_update_state_ptr(s); r = s->rule.ptr; a = s->anchor.ptr; } else if (s == NULL) { action = pf_test_rule(&r, &s, kif, m, off, &pd, &a, &ruleset, inp, hdrlen); } break; } case IPPROTO_ICMP: { if (af != AF_INET) { action = PF_DROP; DPFPRINTF(PF_DEBUG_MISC, ("dropping IPv6 packet with ICMPv4 payload")); goto done; } action = pf_test_state_icmp(&s, kif, m, off, &pd, &reason); if (action == PF_PASS) { if (V_pfsync_update_state_ptr != NULL) V_pfsync_update_state_ptr(s); r = s->rule.ptr; a = s->anchor.ptr; } else if (s == NULL) action = pf_test_rule(&r, &s, kif, m, off, &pd, &a, &ruleset, inp, hdrlen); break; } case IPPROTO_ICMPV6: { if (af != AF_INET6) { action = PF_DROP; DPFPRINTF(PF_DEBUG_MISC, ("pf: dropping IPv4 packet with ICMPv6 payload\n")); goto done; } action = pf_test_state_icmp(&s, kif, m, off, &pd, &reason); if (action == PF_PASS) { if (V_pfsync_update_state_ptr != NULL) V_pfsync_update_state_ptr(s); r = s->rule.ptr; a = s->anchor.ptr; } else if (s == NULL) action = pf_test_rule(&r, &s, kif, m, off, &pd, &a, &ruleset, inp, hdrlen); break; } default: action = pf_test_state_other(&s, kif, m, &pd); if (action == PF_PASS) { if (V_pfsync_update_state_ptr != NULL) V_pfsync_update_state_ptr(s); r = s->rule.ptr; a = s->anchor.ptr; } else if (s == NULL) action = pf_test_rule(&r, &s, kif, m, off, &pd, &a, &ruleset, inp, hdrlen); break; } done: PF_RULES_RUNLOCK(); if (action == PF_PASS && pd.badopts && !((s && s->state_flags & PFSTATE_ALLOWOPTS) || r->allow_opts)) { action = PF_DROP; REASON_SET(&reason, PFRES_IPOPTIONS); pd.act.log = PF_LOG_FORCE; DPFPRINTF(PF_DEBUG_MISC, ("pf: dropping packet with dangerous headers\n")); } if (s) { uint8_t log = pd.act.log; memcpy(&pd.act, &s->act, sizeof(struct pf_rule_actions)); pd.act.log |= log; tag = s->tag; rt = s->rt; } else { tag = r->tag; rt = r->rt; } if (tag > 0 && pf_tag_packet(m, &pd, tag)) { action = PF_DROP; REASON_SET(&reason, PFRES_MEMORY); } pf_scrub(m, &pd); if (pd.proto == IPPROTO_TCP && pd.act.max_mss) pf_normalize_mss(m, off, &pd); if (pd.act.rtableid >= 0) M_SETFIB(m, pd.act.rtableid); if (pd.act.flags & PFSTATE_SETPRIO) { if (pd.tos & IPTOS_LOWDELAY) use_2nd_queue = 1; if (vlan_set_pcp(m, pd.act.set_prio[use_2nd_queue])) { action = PF_DROP; REASON_SET(&reason, PFRES_MEMORY); pd.act.log = PF_LOG_FORCE; DPFPRINTF(PF_DEBUG_MISC, ("pf: failed to allocate 802.1q mtag\n")); } } #ifdef ALTQ if (action == PF_PASS && pd.act.qid) { if (pd.pf_mtag == NULL && ((pd.pf_mtag = pf_get_mtag(m)) == NULL)) { action = PF_DROP; REASON_SET(&reason, PFRES_MEMORY); } else { if (s != NULL) pd.pf_mtag->qid_hash = pf_state_hash(s); if (use_2nd_queue || (pd.tos & IPTOS_LOWDELAY)) pd.pf_mtag->qid = pd.act.pqid; else pd.pf_mtag->qid = pd.act.qid; /* Add hints for ecn. */ if (af == AF_INET) pd.pf_mtag->hdr = h; else pd.pf_mtag->hdr = h6; } } #endif /* ALTQ */ /* * connections redirected to loopback should not match sockets * bound specifically to loopback due to security implications, * see tcp_input() and in_pcblookup_listen(). */ if (dir == PF_IN && action == PF_PASS && (pd.proto == IPPROTO_TCP || pd.proto == IPPROTO_UDP) && s != NULL && s->nat_rule.ptr != NULL && (s->nat_rule.ptr->action == PF_RDR || s->nat_rule.ptr->action == PF_BINAT) && pf_is_loopback(af, pd.dst)) m->m_flags |= M_SKIP_FIREWALL; if (af == AF_INET && __predict_false(ip_divert_ptr != NULL) && action == PF_PASS && r->divert.port && !PACKET_LOOPED(&pd)) { mtag = m_tag_alloc(MTAG_PF_DIVERT, 0, sizeof(struct pf_divert_mtag), M_NOWAIT | M_ZERO); if (mtag != NULL) { ((struct pf_divert_mtag *)(mtag+1))->port = ntohs(r->divert.port); ((struct pf_divert_mtag *)(mtag+1))->idir = (dir == PF_IN) ? PF_DIVERT_MTAG_DIR_IN : PF_DIVERT_MTAG_DIR_OUT; if (s) PF_STATE_UNLOCK(s); m_tag_prepend(m, mtag); if (m->m_flags & M_FASTFWD_OURS) { if (pd.pf_mtag == NULL && ((pd.pf_mtag = pf_get_mtag(m)) == NULL)) { action = PF_DROP; REASON_SET(&reason, PFRES_MEMORY); pd.act.log = PF_LOG_FORCE; DPFPRINTF(PF_DEBUG_MISC, ("pf: failed to allocate tag\n")); } else { pd.pf_mtag->flags |= PF_MTAG_FLAG_FASTFWD_OURS_PRESENT; m->m_flags &= ~M_FASTFWD_OURS; } } ip_divert_ptr(*m0, dir == PF_IN); *m0 = NULL; return (action); } else { /* XXX: ipfw has the same behaviour! */ action = PF_DROP; REASON_SET(&reason, PFRES_MEMORY); pd.act.log = PF_LOG_FORCE; DPFPRINTF(PF_DEBUG_MISC, ("pf: failed to allocate divert tag\n")); } } /* XXX: Anybody working on it?! */ if (af == AF_INET6 && r->divert.port) printf("pf: divert(9) is not supported for IPv6\n"); /* this flag will need revising if the pkt is forwarded */ if (pd.pf_mtag) pd.pf_mtag->flags &= ~PF_MTAG_FLAG_PACKET_LOOPED; if (pd.act.log) { struct pf_krule *lr; struct pf_krule_item *ri; if (s != NULL && s->nat_rule.ptr != NULL && s->nat_rule.ptr->log & PF_LOG_ALL) lr = s->nat_rule.ptr; else lr = r; if (pd.act.log & PF_LOG_FORCE || lr->log & PF_LOG_ALL) PFLOG_PACKET(kif, m, action, reason, lr, a, ruleset, &pd, (s == NULL)); if (s) { SLIST_FOREACH(ri, &s->match_rules, entry) if (ri->r->log & PF_LOG_ALL) PFLOG_PACKET(kif, m, action, reason, ri->r, a, ruleset, &pd, 0); } } pf_counters_inc(action, &pd, kif, s, r, a); switch (action) { case PF_SYNPROXY_DROP: m_freem(*m0); case PF_DEFER: *m0 = NULL; action = PF_PASS; break; case PF_DROP: m_freem(*m0); *m0 = NULL; break; default: if (rt) { switch (af) { #ifdef INET case AF_INET: /* pf_route() returns unlocked. */ pf_route(m0, r, kif->pfik_ifp, s, &pd, inp); break; #endif #ifdef INET6 case AF_INET6: /* pf_route6() returns unlocked. */ pf_route6(m0, r, kif->pfik_ifp, s, &pd, inp); break; #endif default: panic("Unknown af %d", af); } goto out; } if (pf_dummynet(&pd, s, r, m0) != 0) { action = PF_DROP; REASON_SET(&reason, PFRES_MEMORY); } break; } SDT_PROBE4(pf, ip, test, done, action, reason, r, s); if (s && action != PF_DROP) { if (!s->if_index_in && dir == PF_IN) s->if_index_in = ifp->if_index; else if (!s->if_index_out && dir == PF_OUT) s->if_index_out = ifp->if_index; } if (s) PF_STATE_UNLOCK(s); #ifdef INET6 /* If reassembled packet passed, create new fragments. */ if (af == AF_INET6 && action == PF_PASS && *m0 && dir == PF_OUT && (mtag = m_tag_find(m, PACKET_TAG_PF_REASSEMBLED, NULL)) != NULL) action = pf_refragment6(ifp, m0, mtag, pflags & PFIL_FWD); #endif out: pf_sctp_multihome_delayed(&pd, off, kif, s, action); return (action); } #endif /* INET || INET6 */ diff --git a/sys/netpfil/pf/pf.h b/sys/netpfil/pf/pf.h index 063abe83274b..77d4ea6c5e70 100644 --- a/sys/netpfil/pf/pf.h +++ b/sys/netpfil/pf/pf.h @@ -1,720 +1,720 @@ /*- * SPDX-License-Identifier: BSD-2-Clause * * Copyright (c) 2001 Daniel Hartmeier * All rights reserved. * * Redistribution and use in source and binary forms, with or without * modification, are permitted provided that the following conditions * are met: * * - Redistributions of source code must retain the above copyright * notice, this list of conditions and the following disclaimer. * - Redistributions in binary form must reproduce the above * copyright notice, this list of conditions and the following * disclaimer in the documentation and/or other materials provided * with the distribution. * * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER * CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE * POSSIBILITY OF SUCH DAMAGE. * * $OpenBSD: pfvar.h,v 1.282 2009/01/29 15:12:28 pyr Exp $ */ #ifndef _NET_PF_H_ #define _NET_PF_H_ #include #define PF_TCPS_PROXY_SRC ((TCP_NSTATES)+0) #define PF_TCPS_PROXY_DST ((TCP_NSTATES)+1) #define PF_MD5_DIGEST_LENGTH 16 #ifdef MD5_DIGEST_LENGTH #if PF_MD5_DIGEST_LENGTH != MD5_DIGEST_LENGTH #error #endif #endif enum { PF_INOUT, PF_IN, PF_OUT }; enum { PF_PASS, PF_DROP, PF_SCRUB, PF_NOSCRUB, PF_NAT, PF_NONAT, PF_BINAT, PF_NOBINAT, PF_RDR, PF_NORDR, PF_SYNPROXY_DROP, PF_DEFER, PF_MATCH }; enum { PF_RULESET_SCRUB, PF_RULESET_FILTER, PF_RULESET_NAT, PF_RULESET_BINAT, PF_RULESET_RDR, PF_RULESET_MAX }; enum { PF_OP_NONE, PF_OP_IRG, PF_OP_EQ, PF_OP_NE, PF_OP_LT, PF_OP_LE, PF_OP_GT, PF_OP_GE, PF_OP_XRG, PF_OP_RRG }; enum { PF_DEBUG_NONE, PF_DEBUG_URGENT, PF_DEBUG_MISC, PF_DEBUG_NOISY }; enum { PF_CHANGE_NONE, PF_CHANGE_ADD_HEAD, PF_CHANGE_ADD_TAIL, PF_CHANGE_ADD_BEFORE, PF_CHANGE_ADD_AFTER, PF_CHANGE_REMOVE, PF_CHANGE_GET_TICKET }; enum { PF_GET_NONE, PF_GET_CLR_CNTR }; enum { PF_SK_WIRE, PF_SK_STACK, PF_SK_BOTH }; enum { PF_PEER_SRC, PF_PEER_DST, PF_PEER_BOTH }; /* * Note about PFTM_*: real indices into pf_rule.timeout[] come before * PFTM_MAX, special cases afterwards. See pf_state_expires(). */ enum { PFTM_TCP_FIRST_PACKET = 0, PFTM_TCP_OPENING = 1, PFTM_TCP_ESTABLISHED = 2, PFTM_TCP_CLOSING = 3, PFTM_TCP_FIN_WAIT = 4, PFTM_TCP_CLOSED = 5, PFTM_UDP_FIRST_PACKET = 6, PFTM_UDP_SINGLE = 7, PFTM_UDP_MULTIPLE = 8, PFTM_ICMP_FIRST_PACKET = 9, PFTM_ICMP_ERROR_REPLY = 10, PFTM_OTHER_FIRST_PACKET = 11, PFTM_OTHER_SINGLE = 12, PFTM_OTHER_MULTIPLE = 13, PFTM_FRAG = 14, PFTM_INTERVAL = 15, PFTM_ADAPTIVE_START = 16, PFTM_ADAPTIVE_END = 17, PFTM_SRC_NODE = 18, PFTM_TS_DIFF = 19, PFTM_OLD_MAX = 20, /* Legacy limit, for binary compatibility with old kernels. */ PFTM_SCTP_FIRST_PACKET = 20, PFTM_SCTP_OPENING = 21, PFTM_SCTP_ESTABLISHED = 22, PFTM_SCTP_CLOSING = 23, PFTM_SCTP_CLOSED = 24, PFTM_MAX = 25, PFTM_PURGE = 26, PFTM_UNLINKED = 27, }; /* PFTM default values */ #define PFTM_TCP_FIRST_PACKET_VAL 120 /* First TCP packet */ #define PFTM_TCP_OPENING_VAL 30 /* No response yet */ #define PFTM_TCP_ESTABLISHED_VAL 24*60*60/* Established */ #define PFTM_TCP_CLOSING_VAL 15 * 60 /* Half closed */ #define PFTM_TCP_FIN_WAIT_VAL 45 /* Got both FINs */ #define PFTM_TCP_CLOSED_VAL 90 /* Got a RST */ #define PFTM_UDP_FIRST_PACKET_VAL 60 /* First UDP packet */ #define PFTM_UDP_SINGLE_VAL 30 /* Unidirectional */ #define PFTM_UDP_MULTIPLE_VAL 60 /* Bidirectional */ #define PFTM_ICMP_FIRST_PACKET_VAL 20 /* First ICMP packet */ #define PFTM_ICMP_ERROR_REPLY_VAL 10 /* Got error response */ #define PFTM_OTHER_FIRST_PACKET_VAL 60 /* First packet */ #define PFTM_OTHER_SINGLE_VAL 30 /* Unidirectional */ #define PFTM_OTHER_MULTIPLE_VAL 60 /* Bidirectional */ #define PFTM_FRAG_VAL 30 /* Fragment expire */ #define PFTM_INTERVAL_VAL 10 /* Expire interval */ #define PFTM_SRC_NODE_VAL 0 /* Source tracking */ #define PFTM_TS_DIFF_VAL 30 /* Allowed TS diff */ enum { PF_NOPFROUTE, PF_FASTROUTE, PF_ROUTETO, PF_DUPTO, PF_REPLYTO }; enum { PF_LIMIT_STATES, PF_LIMIT_SRC_NODES, PF_LIMIT_FRAGS, PF_LIMIT_TABLE_ENTRIES, PF_LIMIT_MAX }; #define PF_POOL_IDMASK 0x0f enum { PF_POOL_NONE, PF_POOL_BITMASK, PF_POOL_RANDOM, PF_POOL_SRCHASH, PF_POOL_ROUNDROBIN }; enum { PF_ADDR_ADDRMASK, PF_ADDR_NOROUTE, PF_ADDR_DYNIFTL, PF_ADDR_TABLE, PF_ADDR_URPFFAILED, PF_ADDR_RANGE }; #define PF_POOL_TYPEMASK 0x0f #define PF_POOL_STICKYADDR 0x20 #define PF_POOL_ENDPI 0x40 #define PF_WSCALE_FLAG 0x80 #define PF_WSCALE_MASK 0x0f #define PF_LOG 0x01 #define PF_LOG_ALL 0x02 #define PF_LOG_SOCKET_LOOKUP 0x04 #define PF_LOG_FORCE 0x08 #define PF_LOG_MATCHES 0x10 /* Reasons code for passing/dropping a packet */ #define PFRES_MATCH 0 /* Explicit match of a rule */ #define PFRES_BADOFF 1 /* Bad offset for pull_hdr */ #define PFRES_FRAG 2 /* Dropping following fragment */ #define PFRES_SHORT 3 /* Dropping short packet */ #define PFRES_NORM 4 /* Dropping by normalizer */ #define PFRES_MEMORY 5 /* Dropped due to lacking mem */ #define PFRES_TS 6 /* Bad TCP Timestamp (RFC1323) */ #define PFRES_CONGEST 7 /* Congestion (of ipintrq) */ #define PFRES_IPOPTIONS 8 /* IP option */ #define PFRES_PROTCKSUM 9 /* Protocol checksum invalid */ #define PFRES_BADSTATE 10 /* State mismatch */ #define PFRES_STATEINS 11 /* State insertion failure */ #define PFRES_MAXSTATES 12 /* State limit */ #define PFRES_SRCLIMIT 13 /* Source node/conn limit */ #define PFRES_SYNPROXY 14 /* SYN proxy */ #define PFRES_MAPFAILED 15 /* pf_map_addr() failed */ #define PFRES_MAX 16 /* total+1 */ #define PFRES_NAMES { \ "match", \ "bad-offset", \ "fragment", \ "short", \ "normalize", \ "memory", \ "bad-timestamp", \ "congestion", \ "ip-option", \ "proto-cksum", \ "state-mismatch", \ "state-insert", \ "state-limit", \ "src-limit", \ "synproxy", \ "map-failed", \ NULL \ } /* Counters for other things we want to keep track of */ #define LCNT_STATES 0 /* states */ #define LCNT_SRCSTATES 1 /* max-src-states */ #define LCNT_SRCNODES 2 /* max-src-nodes */ #define LCNT_SRCCONN 3 /* max-src-conn */ #define LCNT_SRCCONNRATE 4 /* max-src-conn-rate */ #define LCNT_OVERLOAD_TABLE 5 /* entry added to overload table */ #define LCNT_OVERLOAD_FLUSH 6 /* state entries flushed */ #define LCNT_MAX 7 /* total+1 */ /* Only available via the nvlist-based API */ #define KLCNT_SYNFLOODS 7 /* synfloods detected */ #define KLCNT_SYNCOOKIES_SENT 8 /* syncookies sent */ #define KLCNT_SYNCOOKIES_VALID 9 /* syncookies validated */ #define KLCNT_MAX 10 /* total+1 */ #define LCNT_NAMES { \ "max states per rule", \ "max-src-states", \ "max-src-nodes", \ "max-src-conn", \ "max-src-conn-rate", \ "overload table insertion", \ "overload flush states", \ NULL \ } #define KLCNT_NAMES { \ "max states per rule", \ "max-src-states", \ "max-src-nodes", \ "max-src-conn", \ "max-src-conn-rate", \ "overload table insertion", \ "overload flush states", \ "synfloods detected", \ "syncookies sent", \ "syncookies validated", \ NULL \ } /* state operation counters */ #define FCNT_STATE_SEARCH 0 #define FCNT_STATE_INSERT 1 #define FCNT_STATE_REMOVALS 2 #define FCNT_MAX 3 #ifdef _KERNEL #define FCNT_NAMES { \ "searches", \ "inserts", \ "removals", \ NULL \ } #endif /* src_node operation counters */ #define SCNT_SRC_NODE_SEARCH 0 #define SCNT_SRC_NODE_INSERT 1 #define SCNT_SRC_NODE_REMOVALS 2 #define SCNT_MAX 3 #define PF_TABLE_NAME_SIZE 32 #define PF_QNAME_SIZE 64 struct pfioc_nv { void *data; size_t len; /* The length of the nvlist data. */ size_t size; /* The total size of the data buffer. */ }; struct pf_rule; /* keep synced with pfi_kif, used in RB_FIND */ struct pfi_kif_cmp { char pfik_name[IFNAMSIZ]; }; struct pfi_kif { char pfik_name[IFNAMSIZ]; union { RB_ENTRY(pfi_kif) pfik_tree; LIST_ENTRY(pfi_kif) pfik_list; }; u_int64_t pfik_packets[2][2][2]; u_int64_t pfik_bytes[2][2][2]; u_int32_t pfik_tzero; u_int pfik_flags; struct ifnet *pfik_ifp; struct ifg_group *pfik_group; u_int pfik_rulerefs; TAILQ_HEAD(, pfi_dynaddr) pfik_dynaddrs; }; struct pf_status { uint64_t counters[PFRES_MAX]; uint64_t lcounters[LCNT_MAX]; uint64_t fcounters[FCNT_MAX]; uint64_t scounters[SCNT_MAX]; uint64_t pcounters[2][2][3]; uint64_t bcounters[2][2]; uint32_t running; uint32_t states; uint32_t src_nodes; uint32_t since; uint32_t debug; uint32_t hostid; char ifname[IFNAMSIZ]; uint8_t pf_chksum[PF_MD5_DIGEST_LENGTH]; }; #define PF_REASS_ENABLED 0x01 #define PF_REASS_NODF 0x02 struct pf_addr { union { struct in_addr v4; struct in6_addr v6; u_int8_t addr8[16]; u_int16_t addr16[8]; u_int32_t addr32[4]; }; /* 128-bit address */ }; #define PFI_AFLAG_NETWORK 0x01 #define PFI_AFLAG_BROADCAST 0x02 #define PFI_AFLAG_PEER 0x04 #define PFI_AFLAG_MODEMASK 0x07 #define PFI_AFLAG_NOALIAS 0x08 struct pf_addr_wrap { union { struct { struct pf_addr addr; struct pf_addr mask; } a; char ifname[IFNAMSIZ]; char tblname[PF_TABLE_NAME_SIZE]; } v; union { struct pfi_dynaddr *dyn; struct pfr_ktable *tbl; int dyncnt; int tblcnt; } p; u_int8_t type; /* PF_ADDR_* */ u_int8_t iflags; /* PFI_AFLAG_* */ }; union pf_rule_ptr { struct pf_rule *ptr; u_int32_t nr; }; struct pf_rule_uid { uid_t uid[2]; u_int8_t op; }; struct pf_rule_gid { uid_t gid[2]; u_int8_t op; }; struct pf_rule_addr { struct pf_addr_wrap addr; u_int16_t port[2]; u_int8_t neg; u_int8_t port_op; }; struct pf_pooladdr { struct pf_addr_wrap addr; TAILQ_ENTRY(pf_pooladdr) entries; char ifname[IFNAMSIZ]; struct pfi_kif *kif; }; TAILQ_HEAD(pf_palist, pf_pooladdr); struct pf_poolhashkey { union { u_int8_t key8[16]; u_int16_t key16[8]; u_int32_t key32[4]; }; /* 128-bit hash key */ }; struct pf_mape_portset { u_int8_t offset; u_int8_t psidlen; u_int16_t psid; }; struct pf_pool { struct pf_palist list; struct pf_pooladdr *cur; struct pf_poolhashkey key; struct pf_addr counter; int tblidx; u_int16_t proxy_port[2]; u_int8_t opts; }; /* A packed Operating System description for fingerprinting */ typedef u_int32_t pf_osfp_t; #define PF_OSFP_ANY ((pf_osfp_t)0) #define PF_OSFP_UNKNOWN ((pf_osfp_t)-1) #define PF_OSFP_NOMATCH ((pf_osfp_t)-2) struct pf_osfp_entry { SLIST_ENTRY(pf_osfp_entry) fp_entry; pf_osfp_t fp_os; int fp_enflags; #define PF_OSFP_EXPANDED 0x001 /* expanded entry */ #define PF_OSFP_GENERIC 0x002 /* generic signature */ #define PF_OSFP_NODETAIL 0x004 /* no p0f details */ #define PF_OSFP_LEN 32 char fp_class_nm[PF_OSFP_LEN]; char fp_version_nm[PF_OSFP_LEN]; char fp_subtype_nm[PF_OSFP_LEN]; }; #define PF_OSFP_ENTRY_EQ(a, b) \ ((a)->fp_os == (b)->fp_os && \ memcmp((a)->fp_class_nm, (b)->fp_class_nm, PF_OSFP_LEN) == 0 && \ memcmp((a)->fp_version_nm, (b)->fp_version_nm, PF_OSFP_LEN) == 0 && \ memcmp((a)->fp_subtype_nm, (b)->fp_subtype_nm, PF_OSFP_LEN) == 0) /* handle pf_osfp_t packing */ #define _FP_RESERVED_BIT 1 /* For the special negative #defines */ #define _FP_UNUSED_BITS 1 #define _FP_CLASS_BITS 10 /* OS Class (Windows, Linux) */ #define _FP_VERSION_BITS 10 /* OS version (95, 98, NT, 2.4.54, 3.2) */ #define _FP_SUBTYPE_BITS 10 /* patch level (NT SP4, SP3, ECN patch) */ #define PF_OSFP_UNPACK(osfp, class, version, subtype) do { \ (class) = ((osfp) >> (_FP_VERSION_BITS+_FP_SUBTYPE_BITS)) & \ ((1 << _FP_CLASS_BITS) - 1); \ (version) = ((osfp) >> _FP_SUBTYPE_BITS) & \ ((1 << _FP_VERSION_BITS) - 1);\ (subtype) = (osfp) & ((1 << _FP_SUBTYPE_BITS) - 1); \ } while(0) #define PF_OSFP_PACK(osfp, class, version, subtype) do { \ (osfp) = ((class) & ((1 << _FP_CLASS_BITS) - 1)) << (_FP_VERSION_BITS \ + _FP_SUBTYPE_BITS); \ (osfp) |= ((version) & ((1 << _FP_VERSION_BITS) - 1)) << \ _FP_SUBTYPE_BITS; \ (osfp) |= (subtype) & ((1 << _FP_SUBTYPE_BITS) - 1); \ } while(0) /* the fingerprint of an OSes TCP SYN packet */ typedef u_int64_t pf_tcpopts_t; struct pf_os_fingerprint { SLIST_HEAD(pf_osfp_enlist, pf_osfp_entry) fp_oses; /* list of matches */ pf_tcpopts_t fp_tcpopts; /* packed TCP options */ u_int16_t fp_wsize; /* TCP window size */ u_int16_t fp_psize; /* ip->ip_len */ u_int16_t fp_mss; /* TCP MSS */ u_int16_t fp_flags; #define PF_OSFP_WSIZE_MOD 0x0001 /* Window modulus */ #define PF_OSFP_WSIZE_DC 0x0002 /* Window don't care */ #define PF_OSFP_WSIZE_MSS 0x0004 /* Window multiple of MSS */ #define PF_OSFP_WSIZE_MTU 0x0008 /* Window multiple of MTU */ #define PF_OSFP_PSIZE_MOD 0x0010 /* packet size modulus */ #define PF_OSFP_PSIZE_DC 0x0020 /* packet size don't care */ #define PF_OSFP_WSCALE 0x0040 /* TCP window scaling */ #define PF_OSFP_WSCALE_MOD 0x0080 /* TCP window scale modulus */ #define PF_OSFP_WSCALE_DC 0x0100 /* TCP window scale dont-care */ #define PF_OSFP_MSS 0x0200 /* TCP MSS */ #define PF_OSFP_MSS_MOD 0x0400 /* TCP MSS modulus */ #define PF_OSFP_MSS_DC 0x0800 /* TCP MSS dont-care */ #define PF_OSFP_DF 0x1000 /* IPv4 don't fragment bit */ #define PF_OSFP_TS0 0x2000 /* Zero timestamp */ #define PF_OSFP_INET6 0x4000 /* IPv6 */ u_int8_t fp_optcnt; /* TCP option count */ u_int8_t fp_wscale; /* TCP window scaling */ u_int8_t fp_ttl; /* IPv4 TTL */ #define PF_OSFP_MAXTTL_OFFSET 40 /* TCP options packing */ #define PF_OSFP_TCPOPT_NOP 0x0 /* TCP NOP option */ #define PF_OSFP_TCPOPT_WSCALE 0x1 /* TCP window scaling option */ #define PF_OSFP_TCPOPT_MSS 0x2 /* TCP max segment size opt */ #define PF_OSFP_TCPOPT_SACK 0x3 /* TCP SACK OK option */ #define PF_OSFP_TCPOPT_TS 0x4 /* TCP timestamp option */ #define PF_OSFP_TCPOPT_BITS 3 /* bits used by each option */ #define PF_OSFP_MAX_OPTS \ (sizeof(((struct pf_os_fingerprint *)0)->fp_tcpopts) * 8) \ / PF_OSFP_TCPOPT_BITS SLIST_ENTRY(pf_os_fingerprint) fp_next; }; struct pf_osfp_ioctl { struct pf_osfp_entry fp_os; pf_tcpopts_t fp_tcpopts; /* packed TCP options */ u_int16_t fp_wsize; /* TCP window size */ u_int16_t fp_psize; /* ip->ip_len */ u_int16_t fp_mss; /* TCP MSS */ u_int16_t fp_flags; u_int8_t fp_optcnt; /* TCP option count */ u_int8_t fp_wscale; /* TCP window scaling */ u_int8_t fp_ttl; /* IPv4 TTL */ int fp_getnum; /* DIOCOSFPGET number */ }; #define PF_ANCHOR_NAME_SIZE 64 struct pf_rule { struct pf_rule_addr src; struct pf_rule_addr dst; #define PF_SKIP_IFP 0 #define PF_SKIP_DIR 1 #define PF_SKIP_AF 2 #define PF_SKIP_PROTO 3 #define PF_SKIP_SRC_ADDR 4 -#define PF_SKIP_SRC_PORT 5 -#define PF_SKIP_DST_ADDR 6 +#define PF_SKIP_DST_ADDR 5 +#define PF_SKIP_SRC_PORT 6 #define PF_SKIP_DST_PORT 7 #define PF_SKIP_COUNT 8 union pf_rule_ptr skip[PF_SKIP_COUNT]; #define PF_RULE_LABEL_SIZE 64 #define PF_RULE_MAX_LABEL_COUNT 5 char label[PF_RULE_LABEL_SIZE]; char ifname[IFNAMSIZ]; char qname[PF_QNAME_SIZE]; char pqname[PF_QNAME_SIZE]; #define PF_TAG_NAME_SIZE 64 char tagname[PF_TAG_NAME_SIZE]; char match_tagname[PF_TAG_NAME_SIZE]; char overload_tblname[PF_TABLE_NAME_SIZE]; TAILQ_ENTRY(pf_rule) entries; struct pf_pool rpool; u_int64_t evaluations; u_int64_t packets[2]; u_int64_t bytes[2]; struct pfi_kif *kif; struct pf_anchor *anchor; struct pfr_ktable *overload_tbl; pf_osfp_t os_fingerprint; int rtableid; u_int32_t timeout[PFTM_OLD_MAX]; u_int32_t max_states; u_int32_t max_src_nodes; u_int32_t max_src_states; u_int32_t max_src_conn; struct { u_int32_t limit; u_int32_t seconds; } max_src_conn_rate; u_int32_t qid; u_int32_t pqid; u_int32_t rt_listid; u_int32_t nr; u_int32_t prob; uid_t cuid; pid_t cpid; counter_u64_t states_cur; counter_u64_t states_tot; counter_u64_t src_nodes; u_int16_t return_icmp; u_int16_t return_icmp6; u_int16_t max_mss; u_int16_t tag; u_int16_t match_tag; u_int16_t scrub_flags; struct pf_rule_uid uid; struct pf_rule_gid gid; u_int32_t rule_flag; u_int8_t action; u_int8_t direction; u_int8_t log; u_int8_t logif; u_int8_t quick; u_int8_t ifnot; u_int8_t match_tag_not; u_int8_t natpass; #define PF_STATE_NORMAL 0x1 #define PF_STATE_MODULATE 0x2 #define PF_STATE_SYNPROXY 0x3 u_int8_t keep_state; sa_family_t af; u_int8_t proto; u_int8_t type; u_int8_t code; u_int8_t flags; u_int8_t flagset; u_int8_t min_ttl; u_int8_t allow_opts; u_int8_t rt; u_int8_t return_ttl; u_int8_t tos; u_int8_t set_tos; u_int8_t anchor_relative; u_int8_t anchor_wildcard; #define PF_FLUSH 0x01 #define PF_FLUSH_GLOBAL 0x02 u_int8_t flush; #define PF_PRIO_ZERO 0xff /* match "prio 0" packets */ #define PF_PRIO_MAX 7 u_int8_t prio; u_int8_t set_prio[2]; struct { struct pf_addr addr; u_int16_t port; } divert; uint64_t u_states_cur; uint64_t u_states_tot; uint64_t u_src_nodes; }; /* pf_krule->rule_flag and old-style scrub flags */ #define PFRULE_DROP 0x00000000 #define PFRULE_RETURNRST 0x00000001 #define PFRULE_FRAGMENT 0x00000002 #define PFRULE_RETURNICMP 0x00000004 #define PFRULE_RETURN 0x00000008 #define PFRULE_NOSYNC 0x00000010 #define PFRULE_SRCTRACK 0x00000020 /* track source states */ #define PFRULE_RULESRCTRACK 0x00000040 /* per rule */ #define PFRULE_NODF 0x00000100 #define PFRULE_FRAGMENT_NOREASS 0x00000200 #define PFRULE_RANDOMID 0x00000800 #define PFRULE_REASSEMBLE_TCP 0x00001000 #define PFRULE_SET_TOS 0x00002000 #define PFRULE_IFBOUND 0x00010000 /* if-bound */ #define PFRULE_STATESLOPPY 0x00020000 /* sloppy state tracking */ #define PFRULE_PFLOW 0x00040000 #ifdef _KERNEL #define PFRULE_REFS 0x0080 /* rule has references */ #endif /* pf_rule_actions->dnflags */ #define PFRULE_DN_IS_PIPE 0x0040 #define PFRULE_DN_IS_QUEUE 0x0080 /* pf_state->state_flags, pf_rule_actions->flags, pf_krule->scrub_flags */ #define PFSTATE_ALLOWOPTS 0x0001 #define PFSTATE_SLOPPY 0x0002 #define PFSTATE_PFLOW 0x0004 #define PFSTATE_NOSYNC 0x0008 #define PFSTATE_ACK 0x0010 #define PFSTATE_NODF 0x0020 #define PFSTATE_SETTOS 0x0040 #define PFSTATE_RANDOMID 0x0080 #define PFSTATE_SCRUB_TCP 0x0100 #define PFSTATE_SETPRIO 0x0200 /* was PFSTATE_INP_UNLINKED 0x0400 */ /* FreeBSD-specific flags are added from the end to keep space for porting * flags from OpenBSD */ #define PFSTATE_DN_IS_PIPE 0x4000 #define PFSTATE_DN_IS_QUEUE 0x8000 #define PFSTATE_SCRUBMASK (PFSTATE_NODF|PFSTATE_RANDOMID|PFSTATE_SCRUB_TCP) #define PFSTATE_SETMASK (PFSTATE_SETTOS|PFSTATE_SETPRIO) #define PFSTATE_HIWAT 100000 /* default state table size */ #define PFSTATE_ADAPT_START 60000 /* default adaptive timeout start */ #define PFSTATE_ADAPT_END 120000 /* default adaptive timeout end */ struct pf_threshold { u_int32_t limit; #define PF_THRESHOLD_MULT 1000 #define PF_THRESHOLD_MAX 0xffffffff / PF_THRESHOLD_MULT u_int32_t seconds; u_int32_t count; u_int32_t last; }; struct pf_src_node { LIST_ENTRY(pf_src_node) entry; struct pf_addr addr; struct pf_addr raddr; union pf_rule_ptr rule; struct pfi_kif *kif; u_int64_t bytes[2]; u_int64_t packets[2]; u_int32_t states; u_int32_t conn; struct pf_threshold conn_rate; u_int32_t creation; u_int32_t expire; sa_family_t af; u_int8_t ruletype; }; #define PFSNODE_HIWAT 10000 /* default source node table size */ TAILQ_HEAD(pf_rulequeue, pf_rule); struct pf_anchor; struct pf_ruleset { struct { struct pf_rulequeue queues[2]; struct { struct pf_rulequeue *ptr; struct pf_rule **ptr_array; u_int32_t rcount; u_int32_t ticket; int open; } active, inactive; } rules[PF_RULESET_MAX]; struct pf_anchor *anchor; u_int32_t tticket; int tables; int topen; }; RB_HEAD(pf_anchor_global, pf_anchor); RB_HEAD(pf_anchor_node, pf_anchor); struct pf_anchor { RB_ENTRY(pf_anchor) entry_global; RB_ENTRY(pf_anchor) entry_node; struct pf_anchor *parent; struct pf_anchor_node children; char name[PF_ANCHOR_NAME_SIZE]; char path[MAXPATHLEN]; struct pf_ruleset ruleset; int refcnt; /* anchor rules */ int match; /* XXX: used for pfctl black magic */ }; RB_PROTOTYPE(pf_anchor_global, pf_anchor, entry_global, pf_anchor_compare); RB_PROTOTYPE(pf_anchor_node, pf_anchor, entry_node, pf_anchor_compare); int pf_get_ruleset_number(u_int8_t); #endif /* _NET_PF_H_ */