Re: pf label $nr macro expand reproducable bug
- Reply: Özkan KIRIK : "Re: pf label $nr macro expand reproducable bug"
- In reply to: Özkan KIRIK : "Re: pf label $nr macro expand reproducable bug"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Sat, 02 Oct 2021 16:03:31 UTC
On 22 Sep 2021, at 11:47, Özkan KIRIK wrote: > Hi Kristof, > > I tried many things and I found the real problem to reproduce the bug. > Tested with the latest stable/12. > And also tested with Live CD without installing > (https://download.freebsd.org/ftp/snapshots/ISO-IMAGES/12.2/FreeBSD-12.2-STABLE-amd64-20210916-r370608-disc1.iso). > The result is same. > > My determination is the problem in the rule optimizer of pf. You can > see the difference with / without ruleset optimization. > Without ruleset optimization, $nr macro expanding is true. otherwise > false. > > if the interface used in the rule, have multiple IP addresses that > rule optimizer removes lines then the rule number expanding fails. ie: > Yeah, that makes sense. It looks like the problem is that we expand the $nr tag before the ruleset optimisation, so we end up using the wrong rule number. As a first attempt this also seems to work: diff --git a/sbin/pfctl/parse.y b/sbin/pfctl/parse.y index 89d5f330da47..eb38fd9e344e 100644 --- a/sbin/pfctl/parse.y +++ b/sbin/pfctl/parse.y @@ -330,14 +330,12 @@ int filter_consistent(struct pfctl_rule *, int); int nat_consistent(struct pfctl_rule *); int rdr_consistent(struct pfctl_rule *); int process_tabledef(char *, struct table_opts *); -void expand_label_str(char *, size_t, const char *, const char *); void expand_label_if(const char *, char *, size_t, const char *); void expand_label_addr(const char *, char *, size_t, u_int8_t, struct node_host *); void expand_label_port(const char *, char *, size_t, struct node_port *); void expand_label_proto(const char *, char *, size_t, u_int8_t); -void expand_label_nr(const char *, char *, size_t); void expand_label(char *, size_t, const char *, u_int8_t, struct node_host *, struct node_port *, struct node_host *, struct node_port *, u_int8_t); @@ -5125,17 +5123,6 @@ expand_label_proto(const char *name, char *label, size_t len, u_int8_t proto) } } -void -expand_label_nr(const char *name, char *label, size_t len) -{ - char n[11]; - - if (strstr(label, name) != NULL) { - snprintf(n, sizeof(n), "%u", pf->anchor->match); - expand_label_str(label, len, name, n); - } -} - void expand_label(char *label, size_t len, const char *ifname, sa_family_t af, struct node_host *src_host, struct node_port *src_port, @@ -5148,7 +5135,6 @@ expand_label(char *label, size_t len, const char *ifname, sa_family_t af, expand_label_port("$srcport", label, len, src_port); expand_label_port("$dstport", label, len, dst_port); expand_label_proto("$proto", label, len, proto); - expand_label_nr("$nr", label, len); } int diff --git a/sbin/pfctl/pfctl.c b/sbin/pfctl/pfctl.c index d7bde0012e9b..749d73705f45 100644 --- a/sbin/pfctl/pfctl.c +++ b/sbin/pfctl/pfctl.c @@ -1528,6 +1528,15 @@ pfctl_load_ruleset(struct pfctl *pf, char *path, struct pfctl_ruleset *rs, while ((r = TAILQ_FIRST(rs->rules[rs_num].active.ptr)) != NULL) { TAILQ_REMOVE(rs->rules[rs_num].active.ptr, r, entries); + char n[11]; + snprintf(n, sizeof(n), "%u", r->nr); + for (int i = 0; i < PF_RULE_MAX_LABEL_COUNT; i++) { + expand_label_str(r->label[i], PF_RULE_LABEL_SIZE, + "$nr", n); + } + expand_label_str(r->tagname, PF_TAG_NAME_SIZE, "$nr", n); + expand_label_str(r->match_tagname, PF_TAG_NAME_SIZE, "$nr", n); + if ((error = pfctl_load_rule(pf, path, r, depth))) goto error; if (r->anchor) { diff --git a/sbin/pfctl/pfctl.h b/sbin/pfctl/pfctl.h index 80ef184fa90f..d725341d1cd7 100644 --- a/sbin/pfctl/pfctl.h +++ b/sbin/pfctl/pfctl.h @@ -139,5 +139,6 @@ struct pfctl_ruleset *pf_find_ruleset(const char *); struct pfctl_ruleset *pf_find_or_create_ruleset(const char *); const char *pfctl_proto2name(int); +void expand_label_str(char *, size_t, const char *, const char *); #endif /* _PFCTL_H_ */ I think I want to restructure that a bit, but this should already work. Best regards, Kristof