diff -Nurp linux-2.4.30/Documentation/Configure.help linux-2.4.30-layer7/Documentation/Configure.help --- linux-2.4.30/Documentation/Configure.help 2005-04-03 20:42:19.000000000 -0500 +++ linux-2.4.30-layer7/Documentation/Configure.help 2005-05-03 18:37:03.000000000 -0500 @@ -29056,6 +29056,23 @@ CONFIG_SOUND_WM97XX If unsure, say N. +CONFIG_IP_NF_MATCH_LAYER7 + Say Y if you want to be able to classify connections (and their + packets) based on regular expression matching of their application + layer data. This is one way to classify applications such as + peer-to-peer filesharing systems that do not always use the same + port. + + To compile it as a module, choose M here. If unsure, say N. + +CONFIG_IP_NF_MATCH_LAYER7_DEBUG + Say Y to get lots of debugging output. + +CONFIG_IP_NF_MATCH_LAYER7_MAXDATALEN + Size of the buffer that the application layer data is stored in. + Unless you know what you're doing, leave it at the default of 2048 + Bytes. + # # A couple of things I keep forgetting: # capitalize: AppleTalk, Ethernet, DOS, DMA, FAT, FTP, Internet, diff -Nurp linux-2.4.30/include/linux/netfilter_ipv4/ip_conntrack.h linux-2.4.30-layer7/include/linux/netfilter_ipv4/ip_conntrack.h --- linux-2.4.30/include/linux/netfilter_ipv4/ip_conntrack.h 2005-04-03 20:42:20.000000000 -0500 +++ linux-2.4.30-layer7/include/linux/netfilter_ipv4/ip_conntrack.h 2005-05-03 18:37:03.000000000 -0500 @@ -207,6 +207,17 @@ struct ip_conntrack } nat; #endif /* CONFIG_IP_NF_NAT_NEEDED */ +#if defined(CONFIG_IP_NF_MATCH_LAYER7) || defined(CONFIG_IP_NF_MATCH_LAYER7_MODULE) + struct { + unsigned int numpackets; /* surely this is kept track of somewhere else, right? I can't find it... */ + char * app_proto; /* "http", "ftp", etc. NULL if unclassifed */ + + /* the application layer data so far. NULL if ->numpackets > numpackets */ + char * app_data; + + unsigned int app_data_len; + } layer7; +#endif }; /* get master conntrack via master expectation */ diff -Nurp linux-2.4.30/include/linux/netfilter_ipv4/ipt_layer7.h linux-2.4.30-layer7/include/linux/netfilter_ipv4/ipt_layer7.h --- linux-2.4.30/include/linux/netfilter_ipv4/ipt_layer7.h 1969-12-31 18:00:00.000000000 -0600 +++ linux-2.4.30-layer7/include/linux/netfilter_ipv4/ipt_layer7.h 2005-05-03 18:37:03.000000000 -0500 @@ -0,0 +1,26 @@ +/* + By Matthew Strait , Dec 2003. + http://l7-filter.sf.net + + This program is free software; you can redistribute it and/or + modify it under the terms of the GNU General Public License + as published by the Free Software Foundation; either version + 2 of the License, or (at your option) any later version. + http://www.gnu.org/licenses/gpl.txt +*/ + +#ifndef _IPT_LAYER7_H +#define _IPT_LAYER7_H + +#define MAX_PATTERN_LEN 8192 +#define MAX_PROTOCOL_LEN 256 + +typedef char *(*proc_ipt_search) (char *, char, char *); + +struct ipt_layer7_info { + char protocol[MAX_PROTOCOL_LEN]; + char invert:1; + char pattern[MAX_PATTERN_LEN]; +}; + +#endif /* _IPT_LAYER7_H */ diff -Nurp linux-2.4.30/net/ipv4/netfilter/Config.in linux-2.4.30-layer7/net/ipv4/netfilter/Config.in --- linux-2.4.30/net/ipv4/netfilter/Config.in 2005-01-19 08:10:13.000000000 -0600 +++ linux-2.4.30-layer7/net/ipv4/netfilter/Config.in 2005-05-03 18:37:03.000000000 -0500 @@ -43,6 +43,10 @@ if [ "$CONFIG_IP_NF_IPTABLES" != "n" ]; if [ "$CONFIG_EXPERIMENTAL" = "y" ]; then dep_tristate ' Unclean match support (EXPERIMENTAL)' CONFIG_IP_NF_MATCH_UNCLEAN $CONFIG_IP_NF_IPTABLES dep_tristate ' Owner match support (EXPERIMENTAL)' CONFIG_IP_NF_MATCH_OWNER $CONFIG_IP_NF_IPTABLES + dep_tristate ' Layer 7 match support (EXPERIMENTAL)' CONFIG_IP_NF_MATCH_LAYER7 $CONFIG_IP_NF_CONNTRACK + dep_mbool ' Layer 7 debugging output (EXPERIMENTAL)' CONFIG_IP_NF_MATCH_LAYER7_DEBUG $CONFIG_IP_NF_MATCH_LAYER7 + int ' Buffer size for application layer data (256-65536)' CONFIG_IP_NF_MATCH_LAYER7_MAXDATALEN 2048 + fi # The targets dep_tristate ' Packet filtering' CONFIG_IP_NF_FILTER $CONFIG_IP_NF_IPTABLES diff -Nurp linux-2.4.30/net/ipv4/netfilter/Makefile linux-2.4.30-layer7/net/ipv4/netfilter/Makefile --- linux-2.4.30/net/ipv4/netfilter/Makefile 2003-08-25 06:44:44.000000000 -0500 +++ linux-2.4.30-layer7/net/ipv4/netfilter/Makefile 2005-05-03 18:44:12.000000000 -0500 @@ -86,6 +86,7 @@ obj-$(CONFIG_IP_NF_MATCH_STATE) += ipt_s obj-$(CONFIG_IP_NF_MATCH_CONNTRACK) += ipt_conntrack.o obj-$(CONFIG_IP_NF_MATCH_UNCLEAN) += ipt_unclean.o obj-$(CONFIG_IP_NF_MATCH_TCPMSS) += ipt_tcpmss.o +obj-$(CONFIG_IP_NF_MATCH_LAYER7) += ipt_layer7.o # targets obj-$(CONFIG_IP_NF_TARGET_REJECT) += ipt_REJECT.o diff -Nurp linux-2.4.30/net/ipv4/netfilter/ip_conntrack_core.c linux-2.4.30-layer7/net/ipv4/netfilter/ip_conntrack_core.c --- linux-2.4.30/net/ipv4/netfilter/ip_conntrack_core.c 2005-04-03 20:42:20.000000000 -0500 +++ linux-2.4.30-layer7/net/ipv4/netfilter/ip_conntrack_core.c 2005-05-03 18:37:03.000000000 -0500 @@ -346,6 +346,14 @@ destroy_conntrack(struct nf_conntrack *n } kfree(ct->master); } + + #if defined(CONFIG_IP_NF_MATCH_LAYER7) || defined(CONFIG_IP_NF_MATCH_LAYER7_MODULE) + if(ct->layer7.app_proto) + kfree(ct->layer7.app_proto); + if(ct->layer7.app_data) + kfree(ct->layer7.app_data); + #endif + WRITE_UNLOCK(&ip_conntrack_lock); if (master) diff -Nurp linux-2.4.30/net/ipv4/netfilter/ip_conntrack_standalone.c linux-2.4.30-layer7/net/ipv4/netfilter/ip_conntrack_standalone.c --- linux-2.4.30/net/ipv4/netfilter/ip_conntrack_standalone.c 2005-04-03 20:42:20.000000000 -0500 +++ linux-2.4.30-layer7/net/ipv4/netfilter/ip_conntrack_standalone.c 2005-05-03 18:37:03.000000000 -0500 @@ -107,6 +107,13 @@ print_conntrack(char *buffer, struct ip_ len += sprintf(buffer + len, "[ASSURED] "); len += sprintf(buffer + len, "use=%u ", atomic_read(&conntrack->ct_general.use)); + + #if defined(CONFIG_IP_NF_MATCH_LAYER7) || defined(CONFIG_IP_NF_MATCH_LAYER7_MODULE) + if(conntrack->layer7.app_proto) + len += sprintf(buffer + len, "l7proto=%s ", + conntrack->layer7.app_proto); + #endif + len += sprintf(buffer + len, "\n"); return len; diff -Nurp linux-2.4.30/net/ipv4/netfilter/ipt_layer7.c linux-2.4.30-layer7/net/ipv4/netfilter/ipt_layer7.c --- linux-2.4.30/net/ipv4/netfilter/ipt_layer7.c 1969-12-31 18:00:00.000000000 -0600 +++ linux-2.4.30-layer7/net/ipv4/netfilter/ipt_layer7.c 2005-05-03 18:37:03.000000000 -0500 @@ -0,0 +1,557 @@ +/* + Kernel module to match application layer (OSI layer 7) + data in connections. + + http://l7-filter.sf.net + + By Matthew Strait and Ethan Sommer, 2003-2005. + + This program is free software; you can redistribute it and/or + modify it under the terms of the GNU General Public License + as published by the Free Software Foundation; either version + 2 of the License, or (at your option) any later version. + http://www.gnu.org/licenses/gpl.txt + + Based on ipt_string.c (C) 2000 Emmanuel Roger + and cls_layer7.c (C) 2003 Matthew Strait, Ethan Sommer, Justin Levandoski +*/ + +#include +#include +#include +#include +#include +#include +#include +#include + +#include "regexp/regexp.c" + +#include +#include + +MODULE_AUTHOR("Matthew Strait , Ethan Sommer "); +MODULE_LICENSE("GPL"); +MODULE_DESCRIPTION("iptables application layer match module"); + +#if defined(CONFIG_IP_NF_MATCH_LAYER7_DEBUG) + #define DPRINTK(format,args...) printk(format,##args) +#else + #define DPRINTK(format,args...) +#endif + +#define TOTAL_PACKETS master_conntrack->layer7.numpackets + +/* Number of packets whose data we look at. +This can be modified through /proc/net/layer7_numpackets */ +static int num_packets = 8; + +static struct pattern_cache { + char * regex_string; + regexp * pattern; + struct pattern_cache * next; +} * first_pattern_cache = NULL; + +/* I'm new to locking. Here are my assumptions: + +- No one will write to /proc/net/layer7_numpackets over and over very fast; + if they did, nothing awful would happen. + +- This code will never be processing the same packet twice at the same time, + because iptables rules are traversed in order. + +- It doesn't matter if two packets from different connections are in here at + the same time, because they don't share any data. + +- It _does_ matter if two packets from the same connection are here at the same + time. In this case, we have to protect the conntracks and the list of + compiled patterns. +*/ +DECLARE_RWLOCK(ct_lock); +DECLARE_LOCK(list_lock); + +#if CONFIG_IP_NF_MATCH_LAYER7_DEBUG +/* Converts an unfriendly string into a friendly one by +replacing unprintables with periods and all whitespace with " ". */ +static char * friendly_print(unsigned char * s) +{ + char * f = kmalloc(strlen(s) + 1, GFP_ATOMIC); + int i; + + if(!f) { + if (net_ratelimit()) + printk(KERN_ERR "layer7: out of memory in friendly_print, bailing.\n"); + return NULL; + } + + for(i = 0; i < strlen(s); i++){ + if(isprint(s[i]) && s[i] < 128) f[i] = s[i]; + else if(isspace(s[i])) f[i] = ' '; + else f[i] = '.'; + } + f[i] = '\0'; + return f; +} + +static char dec2hex(int i) +{ + switch (i) { + case 0 ... 9: + return (char)(i + '0'); + break; + case 10 ... 15: + return (char)(i - 10 + 'a'); + break; + default: + if (net_ratelimit()) + printk("Problem in dec2hex\n"); + return '\0'; + } +} + +static char * hex_print(unsigned char * s) +{ + char * g = kmalloc(strlen(s)*3 + 1, GFP_ATOMIC); + int i; + + if(!g) { + if (net_ratelimit()) + printk(KERN_ERR "layer7: out of memory in hex_print, bailing.\n"); + return NULL; + } + + for(i = 0; i < strlen(s); i++) { + g[i*3 ] = dec2hex(s[i]/16); + g[i*3 + 1] = dec2hex(s[i]%16); + g[i*3 + 2] = ' '; + } + g[i*3] = '\0'; + + return g; +} +#endif // DEBUG + +/* Use instead of regcomp. As we expect to be seeing the same regexps over and +over again, it make sense to cache the results. */ +static regexp * compile_and_cache(char * regex_string, char * protocol) +{ + struct pattern_cache * node = first_pattern_cache; + struct pattern_cache * last_pattern_cache = first_pattern_cache; + struct pattern_cache * tmp; + unsigned int len; + + while (node != NULL) { + if (!strcmp(node->regex_string, regex_string)) + return node->pattern; + + last_pattern_cache = node;/* points at the last non-NULL node */ + node = node->next; + } + + /* If we reach the end of the list, then we have not yet cached + the pattern for this regex. Let's do that now. + Be paranoid about running out of memory to avoid list corruption. */ + tmp = kmalloc(sizeof(struct pattern_cache), GFP_ATOMIC); + + if(!tmp) { + if (net_ratelimit()) + printk(KERN_ERR "layer7: out of memory in compile_and_cache, bailing.\n"); + return NULL; + } + + tmp->regex_string = kmalloc(strlen(regex_string) + 1, GFP_ATOMIC); + tmp->pattern = kmalloc(sizeof(struct regexp), GFP_ATOMIC); + tmp->next = NULL; + + if(!tmp->regex_string || !tmp->pattern) { + if (net_ratelimit()) + printk(KERN_ERR "layer7: out of memory in compile_and_cache, bailing.\n"); + kfree(tmp->regex_string); + kfree(tmp->pattern); + kfree(tmp); + return NULL; + } + + /* Ok. The new node is all ready now. */ + node = tmp; + + if(first_pattern_cache == NULL) /* list is empty */ + first_pattern_cache = node; /* make node the beginning */ + else + last_pattern_cache->next = node; /* attach node to the end */ + + /* copy the string and compile the regex */ + len = strlen(regex_string); + DPRINTK("About to compile this: \"%s\"\n", regex_string); + node->pattern = regcomp(regex_string, &len); + if ( !node->pattern ) { + if (net_ratelimit()) + printk(KERN_ERR "layer7: Error compiling regexp \"%s\" (%s)\n", regex_string, protocol); + /* pattern is now cached as NULL, so we won't try again. */ + } + + strcpy(node->regex_string, regex_string); + return node->pattern; +} + +static int can_handle(const struct sk_buff *skb) +{ + if(!skb->nh.iph) /* not IP */ + return 0; + if(skb->nh.iph->protocol != IPPROTO_TCP && + skb->nh.iph->protocol != IPPROTO_UDP && + skb->nh.iph->protocol != IPPROTO_ICMP) + return 0; + return 1; +} + +/* Returns offset the into the skb->data that the application data starts */ +static int app_data_offset(const struct sk_buff *skb) +{ + /* In case we are ported somewhere (ebtables?) where skb->nh.iph + isn't set, this can be gotten from 4*(skb->data[0] & 0x0f) as well. */ + int ip_hl = 4*skb->nh.iph->ihl; + + if( skb->nh.iph->protocol == IPPROTO_TCP ) { + /* 12 == offset into TCP header for the header length field. + Can't get this with skb->h.th->doff because the tcphdr + struct doesn't get set when routing (this is confirmed to be + true in Netfilter as well as QoS.) */ + int tcp_hl = 4*(skb->data[ip_hl + 12] >> 4); + + return ip_hl + tcp_hl; + } else if( skb->nh.iph->protocol == IPPROTO_UDP ) { + return ip_hl + 8; /* UDP header is always 8 bytes */ + } else if( skb->nh.iph->protocol == IPPROTO_ICMP ) { + return ip_hl + 8; /* ICMP header is 8 bytes */ + } else { + if (net_ratelimit()) + printk(KERN_ERR "layer7: tried to handle unknown protocol!\n"); + return ip_hl + 8; /* something reasonable */ + } +} + +/* handles whether there's a match when we aren't appending data anymore */ +static int match_no_append(struct ip_conntrack * conntrack, struct ip_conntrack * master_conntrack, + enum ip_conntrack_info ctinfo, enum ip_conntrack_info master_ctinfo, + struct ipt_layer7_info * info) +{ + /* If we're in here, throw the app data away */ + WRITE_LOCK(&ct_lock); + if(master_conntrack->layer7.app_data != NULL) { + + #ifdef CONFIG_IP_NF_MATCH_LAYER7_DEBUG + if(!master_conntrack->layer7.app_proto) { + char * f = friendly_print(master_conntrack->layer7.app_data); + char * g = hex_print(master_conntrack->layer7.app_data); + DPRINTK("\nl7-filter gave up after %d bytes (%d packets):\n%s\n", + strlen(f), + TOTAL_PACKETS, f); + kfree(f); + DPRINTK("In hex: %s\n", g); + kfree(g); + } + #endif + + kfree(master_conntrack->layer7.app_data); + master_conntrack->layer7.app_data = NULL; /* don't free again */ + } + WRITE_UNLOCK(&ct_lock); + + if(master_conntrack->layer7.app_proto){ + /* Here child connections set their .app_proto (for /proc/net/ip_conntrack) */ + WRITE_LOCK(&ct_lock); + if(!conntrack->layer7.app_proto) { + conntrack->layer7.app_proto = kmalloc(strlen(master_conntrack->layer7.app_proto)+1, GFP_ATOMIC); + if(!conntrack->layer7.app_proto){ + if (net_ratelimit()) + printk(KERN_ERR "layer7: out of memory in match_no_append, bailing.\n"); + WRITE_UNLOCK(&ct_lock); + return 1; + } + strcpy(conntrack->layer7.app_proto, master_conntrack->layer7.app_proto); + } + WRITE_UNLOCK(&ct_lock); + + return (!strcmp(master_conntrack->layer7.app_proto, info->protocol)); + } + else { + /* If not classified, set to "unknown" to distinguish from + connections that are still being tested. */ + WRITE_LOCK(&ct_lock); + master_conntrack->layer7.app_proto = kmalloc(strlen("unknown")+1, GFP_ATOMIC); + if(!master_conntrack->layer7.app_proto){ + if (net_ratelimit()) + printk(KERN_ERR "layer7: out of memory in match_no_append, bailing.\n"); + WRITE_UNLOCK(&ct_lock); + return 1; + } + strcpy(master_conntrack->layer7.app_proto, "unknown"); + WRITE_UNLOCK(&ct_lock); + return 0; + } +} + +/* add the new app data to the conntrack. Return number of bytes added. */ +static int add_data(struct ip_conntrack * master_conntrack, + char * app_data, int appdatalen) +{ + int length = 0, i; + int oldlength = master_conntrack->layer7.app_data_len; + + /* Strip nulls. Make everything lower case (our regex lib doesn't + do case insensitivity). Add it to the end of the current data. */ + for(i = 0; i < CONFIG_IP_NF_MATCH_LAYER7_MAXDATALEN-oldlength-1 && + i < appdatalen; i++) { + if(app_data[i] != '\0') { + master_conntrack->layer7.app_data[length+oldlength] = + /* the kernel version of tolower mungs 'upper ascii' */ + isascii(app_data[i])? tolower(app_data[i]) : app_data[i]; + length++; + } + } + + master_conntrack->layer7.app_data[length+oldlength] = '\0'; + master_conntrack->layer7.app_data_len = length + oldlength; + + return length; +} + +/* Returns true on match and false otherwise. */ +static int match(/* const */struct sk_buff *skb, const struct net_device *in, + const struct net_device *out, const void *matchinfo, + int offset, int *hotdrop) +{ + struct ipt_layer7_info * info = (struct ipt_layer7_info *)matchinfo; + enum ip_conntrack_info master_ctinfo, ctinfo; + struct ip_conntrack *master_conntrack, *conntrack; + unsigned char * app_data; + unsigned int pattern_result, appdatalen; + regexp * comppattern; + + if(!can_handle(skb)){ + DPRINTK("layer7: This is some protocol I can't handle.\n"); + return info->invert; + } + + /* Treat the parent and all its children together as one connection, + except for the purpose of setting conntrack->layer7.app_proto in the + actual connection. This makes /proc/net/ip_conntrack somewhat more + satisfying. */ + if(!(conntrack = ip_conntrack_get((struct sk_buff *)skb, &ctinfo)) || + !(master_conntrack = ip_conntrack_get((struct sk_buff *)skb, &master_ctinfo))) { + DPRINTK("layer7: packet is not from a known connection, giving up.\n"); + return info->invert; + } + + /* Try to get a master conntrack (and its master etc) for FTP, etc. */ + while (master_ct(master_conntrack) != NULL) + master_conntrack = master_ct(master_conntrack); + + if(!skb->cb[0]){ + WRITE_LOCK(&ct_lock); + master_conntrack->layer7.numpackets++;/*starts at 0 via memset*/ + WRITE_UNLOCK(&ct_lock); + } + + /* if we've classified it or seen too many packets */ + if(TOTAL_PACKETS > num_packets || + master_conntrack->layer7.app_proto) { + + pattern_result = match_no_append(conntrack, master_conntrack, ctinfo, master_ctinfo, info); + + /* skb->cb[0] == seen. Avoid doing things twice if there are two l7 + rules. I'm not sure that using cb for this purpose is correct, although + it says "put your private variables there". But it doesn't look like it + is being used for anything else in the skbs that make it here. How can + I write to cb without making the compiler angry? */ + skb->cb[0] = 1; /* marking it seen here is probably irrelevant, but consistant */ + + return (pattern_result ^ info->invert); + } + + if(skb_is_nonlinear(skb)){ + if(skb_linearize(skb, GFP_ATOMIC) != 0){ + if (net_ratelimit()) + printk(KERN_ERR "layer7: failed to linearize packet, bailing.\n"); + return info->invert; + } + } + + /* now that the skb is linearized, it's safe to set these. */ + app_data = skb->data + app_data_offset(skb); + appdatalen = skb->tail - app_data; + + LOCK_BH(&list_lock); + /* the return value gets checked later, when we're ready to use it */ + comppattern = compile_and_cache(info->pattern, info->protocol); + UNLOCK_BH(&list_lock); + + /* On the first packet of a connection, allocate space for app data */ + WRITE_LOCK(&ct_lock); + if(TOTAL_PACKETS == 1 && !skb->cb[0] && !master_conntrack->layer7.app_data) { + master_conntrack->layer7.app_data = kmalloc(CONFIG_IP_NF_MATCH_LAYER7_MAXDATALEN, GFP_ATOMIC); + if(!master_conntrack->layer7.app_data){ + if (net_ratelimit()) + printk(KERN_ERR "layer7: out of memory in match, bailing.\n"); + WRITE_UNLOCK(&ct_lock); + return info->invert; + } + + master_conntrack->layer7.app_data[0] = '\0'; + } + WRITE_UNLOCK(&ct_lock); + + /* Can be here, but unallocated, if numpackets is increased near + the beginning of a connection */ + if(master_conntrack->layer7.app_data == NULL) + return (info->invert); /* unmatched */ + + if(!skb->cb[0]){ + int newbytes; + WRITE_LOCK(&ct_lock); + newbytes = add_data(master_conntrack, app_data, appdatalen); + WRITE_UNLOCK(&ct_lock); + + if(newbytes == 0) { /* didn't add any data */ + skb->cb[0] = 1; + /* Didn't match before, not going to match now */ + return info->invert; + } + } + + /* If looking for "unknown", then never match. "Unknown" means that + we've given up; we're still trying with these packets. */ + if(!strcmp(info->protocol, "unknown")) { + pattern_result = 0; + /* If the regexp failed to compile, don't bother running it */ + } else if(comppattern && regexec(comppattern, master_conntrack->layer7.app_data)) { + DPRINTK("layer7: regexec positive: %s!\n", info->protocol); + pattern_result = 1; + } else pattern_result = 0; + + if(pattern_result) { + WRITE_LOCK(&ct_lock); + master_conntrack->layer7.app_proto = kmalloc(strlen(info->protocol)+1, GFP_ATOMIC); + if(!master_conntrack->layer7.app_proto){ + if (net_ratelimit()) + printk(KERN_ERR "layer7: out of memory in match, bailing.\n"); + WRITE_UNLOCK(&ct_lock); + return (pattern_result ^ info->invert); + } + strcpy(master_conntrack->layer7.app_proto, info->protocol); + WRITE_UNLOCK(&ct_lock); + } + + /* mark the packet seen */ + skb->cb[0] = 1; + + return (pattern_result ^ info->invert); +} + +static int checkentry(const char *tablename, const struct ipt_ip *ip, + void *matchinfo, unsigned int matchsize, unsigned int hook_mask) +{ + if (matchsize != IPT_ALIGN(sizeof(struct ipt_layer7_info))) + return 0; + return 1; +} + +static struct ipt_match layer7_match = { + .name = "layer7", + .match = &match, + .checkentry = &checkentry, + .me = THIS_MODULE +}; + +/* taken from drivers/video/modedb.c */ +static int my_atoi(const char *s) +{ + int val = 0; + + for (;; s++) { + switch (*s) { + case '0'...'9': + val = 10*val+(*s-'0'); + break; + default: + return val; + } + } +} + +/* write out num_packets to userland. */ +static int layer7_read_proc(char* page, char ** start, off_t off, int count, + int* eof, void * data) +{ + if(num_packets > 99 && net_ratelimit()) + printk(KERN_ERR "layer7: NOT REACHED. num_packets too big\n"); + + page[0] = num_packets/10 + '0'; + page[1] = num_packets%10 + '0'; + page[2] = '\n'; + page[3] = '\0'; + + *eof=1; + + return 3; +} + +/* Read in num_packets from userland */ +static int layer7_write_proc(struct file* file, const char* buffer, + unsigned long count, void *data) +{ + char * foo = kmalloc(count, GFP_ATOMIC); + + if(!foo){ + if (net_ratelimit()) + printk(KERN_ERR "layer7: out of memory, bailing. num_packets unchanged.\n"); + return count; + } + + copy_from_user(foo, buffer, count); + + num_packets = my_atoi(foo); + kfree (foo); + + /* This has an arbitrary limit to make the math easier. I'm lazy. + But anyway, 99 is a LOT! If you want more, you're doing it wrong! */ + if(num_packets > 99) { + printk(KERN_WARNING "layer7: num_packets can't be > 99.\n"); + num_packets = 99; + } else if(num_packets < 1) { + printk(KERN_WARNING "layer7: num_packets can't be < 1.\n"); + num_packets = 1; + } + + return count; +} + +/* register the proc file */ +static void layer7_init_proc(void) +{ + struct proc_dir_entry* entry; + entry = create_proc_entry("layer7_numpackets", 0644, proc_net); + entry->read_proc = layer7_read_proc; + entry->write_proc = layer7_write_proc; +} + +static void layer7_cleanup_proc(void) +{ + remove_proc_entry("layer7_numpackets", proc_net); +} + +static int __init init(void) +{ + layer7_init_proc(); + return ipt_register_match(&layer7_match); +} + +static void __exit fini(void) +{ + layer7_cleanup_proc(); + ipt_unregister_match(&layer7_match); +} + +module_init(init); +module_exit(fini); diff -Nurp linux-2.4.30/net/ipv4/netfilter/regexp/regexp.c linux-2.4.30-layer7/net/ipv4/netfilter/regexp/regexp.c --- linux-2.4.30/net/ipv4/netfilter/regexp/regexp.c 1969-12-31 18:00:00.000000000 -0600 +++ linux-2.4.30-layer7/net/ipv4/netfilter/regexp/regexp.c 2005-05-03 18:37:03.000000000 -0500 @@ -0,0 +1,1195 @@ +/* + * regcomp and regexec -- regsub and regerror are elsewhere + * @(#)regexp.c 1.3 of 18 April 87 + * + * Copyright (c) 1986 by University of Toronto. + * Written by Henry Spencer. Not derived from licensed software. + * + * Permission is granted to anyone to use this software for any + * purpose on any computer system, and to redistribute it freely, + * subject to the following restrictions: + * + * 1. The author is not responsible for the consequences of use of + * this software, no matter how awful, even if they arise + * from defects in it. + * + * 2. The origin of this software must not be misrepresented, either + * by explicit claim or by omission. + * + * 3. Altered versions must be plainly marked as such, and must not + * be misrepresented as being the original software. + * + * Beware that some of this code is subtly aware of the way operator + * precedence is structured in regular expressions. Serious changes in + * regular-expression syntax might require a total rethink. + * + * This code was modified by Ethan Sommer to work within the kernel + * (it now uses kmalloc etc..) + * + * Modified slightly by Matthew Strait to use more modern C. + */ + +#include "regexp.h" +#include "regmagic.h" + +/* added by ethan and matt. Lets it work in both kernel and user space. +(So iptables can use it, for instance.) Yea, it goes both ways... */ +#if __KERNEL__ + #define malloc(foo) kmalloc(foo,GFP_ATOMIC) +#else + #define printk(format,args...) printf(format,##args) +#endif + +void regerror(char * s) +{ + printk("<3>Regexp: %s\n", s); + /* NOTREACHED */ +} + +/* + * The "internal use only" fields in regexp.h are present to pass info from + * compile to execute that permits the execute phase to run lots faster on + * simple cases. They are: + * + * regstart char that must begin a match; '\0' if none obvious + * reganch is the match anchored (at beginning-of-line only)? + * regmust string (pointer into program) that match must include, or NULL + * regmlen length of regmust string + * + * Regstart and reganch permit very fast decisions on suitable starting points + * for a match, cutting down the work a lot. Regmust permits fast rejection + * of lines that cannot possibly match. The regmust tests are costly enough + * that regcomp() supplies a regmust only if the r.e. contains something + * potentially expensive (at present, the only such thing detected is * or + + * at the start of the r.e., which can involve a lot of backup). Regmlen is + * supplied because the test in regexec() needs it and regcomp() is computing + * it anyway. + */ + +/* + * Structure for regexp "program". This is essentially a linear encoding + * of a nondeterministic finite-state machine (aka syntax charts or + * "railroad normal form" in parsing technology). Each node is an opcode + * plus a "next" pointer, possibly plus an operand. "Next" pointers of + * all nodes except BRANCH implement concatenation; a "next" pointer with + * a BRANCH on both ends of it is connecting two alternatives. (Here we + * have one of the subtle syntax dependencies: an individual BRANCH (as + * opposed to a collection of them) is never concatenated with anything + * because of operator precedence.) The operand of some types of node is + * a literal string; for others, it is a node leading into a sub-FSM. In + * particular, the operand of a BRANCH node is the first node of the branch. + * (NB this is *not* a tree structure: the tail of the branch connects + * to the thing following the set of BRANCHes.) The opcodes are: + */ + +/* definition number opnd? meaning */ +#define END 0 /* no End of program. */ +#define BOL 1 /* no Match "" at beginning of line. */ +#define EOL 2 /* no Match "" at end of line. */ +#define ANY 3 /* no Match any one character. */ +#define ANYOF 4 /* str Match any character in this string. */ +#define ANYBUT 5 /* str Match any character not in this string. */ +#define BRANCH 6 /* node Match this alternative, or the next... */ +#define BACK 7 /* no Match "", "next" ptr points backward. */ +#define EXACTLY 8 /* str Match this string. */ +#define NOTHING 9 /* no Match empty string. */ +#define STAR 10 /* node Match this (simple) thing 0 or more times. */ +#define PLUS 11 /* node Match this (simple) thing 1 or more times. */ +#define OPEN 20 /* no Mark this point in input as start of #n. */ + /* OPEN+1 is number 1, etc. */ +#define CLOSE 30 /* no Analogous to OPEN. */ + +/* + * Opcode notes: + * + * BRANCH The set of branches constituting a single choice are hooked + * together with their "next" pointers, since precedence prevents + * anything being concatenated to any individual branch. The + * "next" pointer of the last BRANCH in a choice points to the + * thing following the whole choice. This is also where the + * final "next" pointer of each individual branch points; each + * branch starts with the operand node of a BRANCH node. + * + * BACK Normal "next" pointers all implicitly point forward; BACK + * exists to make loop structures possible. + * + * STAR,PLUS '?', and complex '*' and '+', are implemented as circular + * BRANCH structures using BACK. Simple cases (one character + * per match) are implemented with STAR and PLUS for speed + * and to minimize recursive plunges. + * + * OPEN,CLOSE ...are numbered at compile time. + */ + +/* + * A node is one char of opcode followed by two chars of "next" pointer. + * "Next" pointers are stored as two 8-bit pieces, high order first. The + * value is a positive offset from the opcode of the node containing it. + * An operand, if any, simply follows the node. (Note that much of the + * code generation knows about this implicit relationship.) + * + * Using two bytes for the "next" pointer is vast overkill for most things, + * but allows patterns to get big without disasters. + */ +#define OP(p) (*(p)) +#define NEXT(p) (((*((p)+1)&0377)<<8) + (*((p)+2)&0377)) +#define OPERAND(p) ((p) + 3) + +/* + * See regmagic.h for one further detail of program structure. + */ + + +/* + * Utility definitions. + */ +#ifndef CHARBITS +#define UCHARAT(p) ((int)*(unsigned char *)(p)) +#else +#define UCHARAT(p) ((int)*(p)&CHARBITS) +#endif + +#define FAIL(m) { regerror(m); return(NULL); } +#define ISMULT(c) ((c) == '*' || (c) == '+' || (c) == '?') +#define META "^$.[()|?+*\\" + +/* + * Flags to be passed up and down. + */ +#define HASWIDTH 01 /* Known never to match null string. */ +#define SIMPLE 02 /* Simple enough to be STAR/PLUS operand. */ +#define SPSTART 04 /* Starts with * or +. */ +#define WORST 0 /* Worst case. */ + +/* + * Global work variables for regcomp(). + */ +static char *regparse; /* Input-scan pointer. */ +static int regnpar; /* () count. */ +static char regdummy; +static char *regcode; /* Code-emit pointer; ®dummy = don't. */ +static long regsize; /* Code size. */ + +/* + * Forward declarations for regcomp()'s friends. + */ +#ifndef STATIC +#define STATIC static +#endif +STATIC char *reg(int paren,int *flagp); +STATIC char *regbranch(int *flagp); +STATIC char *regpiece(int *flagp); +STATIC char *regatom(int *flagp); +STATIC char *regnode(char op); +STATIC char *regnext(char *p); +STATIC void regc(char b); +STATIC void reginsert(char op, char *opnd); +STATIC void regtail(char *p, char *val); +STATIC void regoptail(char *p, char *val); + + +__kernel_size_t my_strcspn(const char *s1,const char *s2) +{ + char *scan1; + char *scan2; + int count; + + count = 0; + for (scan1 = (char *)s1; *scan1 != '\0'; scan1++) { + for (scan2 = (char *)s2; *scan2 != '\0';) /* ++ moved down. */ + if (*scan1 == *scan2++) + return(count); + count++; + } + return(count); +} + +/* + - regcomp - compile a regular expression into internal code + * + * We can't allocate space until we know how big the compiled form will be, + * but we can't compile it (and thus know how big it is) until we've got a + * place to put the code. So we cheat: we compile it twice, once with code + * generation turned off and size counting turned on, and once "for real". + * This also means that we don't allocate space until we are sure that the + * thing really will compile successfully, and we never have to move the + * code and thus invalidate pointers into it. (Note that it has to be in + * one piece because free() must be able to free it all.) + * + * Beware that the optimization-preparation code in here knows about some + * of the structure of the compiled regexp. + */ +regexp * +regcomp(char *exp,int *patternsize) +{ + register regexp *r; + register char *scan; + register char *longest; + register int len; + int flags; + /* commented out by ethan + extern char *malloc(); + */ + + if (exp == NULL) + FAIL("NULL argument"); + + /* First pass: determine size, legality. */ + regparse = exp; + regnpar = 1; + regsize = 0L; + regcode = ®dummy; + regc(MAGIC); + if (reg(0, &flags) == NULL) + return(NULL); + + /* Small enough for pointer-storage convention? */ + if (regsize >= 32767L) /* Probably could be 65535L. */ + FAIL("regexp too big"); + + /* Allocate space. */ + *patternsize=sizeof(regexp) + (unsigned)regsize; + r = (regexp *)malloc(sizeof(regexp) + (unsigned)regsize); + if (r == NULL) + FAIL("out of space"); + + /* Second pass: emit code. */ + regparse = exp; + regnpar = 1; + regcode = r->program; + regc(MAGIC); + if (reg(0, &flags) == NULL) + return(NULL); + + /* Dig out information for optimizations. */ + r->regstart = '\0'; /* Worst-case defaults. */ + r->reganch = 0; + r->regmust = NULL; + r->regmlen = 0; + scan = r->program+1; /* First BRANCH. */ + if (OP(regnext(scan)) == END) { /* Only one top-level choice. */ + scan = OPERAND(scan); + + /* Starting-point info. */ + if (OP(scan) == EXACTLY) + r->regstart = *OPERAND(scan); + else if (OP(scan) == BOL) + r->reganch++; + + /* + * If there's something expensive in the r.e., find the + * longest literal string that must appear and make it the + * regmust. Resolve ties in favor of later strings, since + * the regstart check works with the beginning of the r.e. + * and avoiding duplication strengthens checking. Not a + * strong reason, but sufficient in the absence of others. + */ + if (flags&SPSTART) { + longest = NULL; + len = 0; + for (; scan != NULL; scan = regnext(scan)) + if (OP(scan) == EXACTLY && strlen(OPERAND(scan)) >= len) { + longest = OPERAND(scan); + len = strlen(OPERAND(scan)); + } + r->regmust = longest; + r->regmlen = len; + } + } + + return(r); +} + +/* + - reg - regular expression, i.e. main body or parenthesized thing + * + * Caller must absorb opening parenthesis. + * + * Combining parenthesis handling with the base level of regular expression + * is a trifle forced, but the need to tie the tails of the branches to what + * follows makes it hard to avoid. + */ +static char * +reg(int paren, int *flagp /* Parenthesized? */ ) +{ + register char *ret; + register char *br; + register char *ender; + register int parno = 0; /* 0 makes gcc happy */ + int flags; + + *flagp = HASWIDTH; /* Tentatively. */ + + /* Make an OPEN node, if parenthesized. */ + if (paren) { + if (regnpar >= NSUBEXP) + FAIL("too many ()"); + parno = regnpar; + regnpar++; + ret = regnode(OPEN+parno); + } else + ret = NULL; + + /* Pick up the branches, linking them together. */ + br = regbranch(&flags); + if (br == NULL) + return(NULL); + if (ret != NULL) + regtail(ret, br); /* OPEN -> first. */ + else + ret = br; + if (!(flags&HASWIDTH)) + *flagp &= ~HASWIDTH; + *flagp |= flags&SPSTART; + while (*regparse == '|') { + regparse++; + br = regbranch(&flags); + if (br == NULL) + return(NULL); + regtail(ret, br); /* BRANCH -> BRANCH. */ + if (!(flags&HASWIDTH)) + *flagp &= ~HASWIDTH; + *flagp |= flags&SPSTART; + } + + /* Make a closing node, and hook it on the end. */ + ender = regnode((paren) ? CLOSE+parno : END); + regtail(ret, ender); + + /* Hook the tails of the branches to the closing node. */ + for (br = ret; br != NULL; br = regnext(br)) + regoptail(br, ender); + + /* Check for proper termination. */ + if (paren && *regparse++ != ')') { + FAIL("unmatched ()"); + } else if (!paren && *regparse != '\0') { + if (*regparse == ')') { + FAIL("unmatched ()"); + } else + FAIL("junk on end"); /* "Can't happen". */ + /* NOTREACHED */ + } + + return(ret); +} + +/* + - regbranch - one alternative of an | operator + * + * Implements the concatenation operator. + */ +static char * +regbranch(int *flagp) +{ + register char *ret; + register char *chain; + register char *latest; + int flags; + + *flagp = WORST; /* Tentatively. */ + + ret = regnode(BRANCH); + chain = NULL; + while (*regparse != '\0' && *regparse != '|' && *regparse != ')') { + latest = regpiece(&flags); + if (latest == NULL) + return(NULL); + *flagp |= flags&HASWIDTH; + if (chain == NULL) /* First piece. */ + *flagp |= flags&SPSTART; + else + regtail(chain, latest); + chain = latest; + } + if (chain == NULL) /* Loop ran zero times. */ + (void) regnode(NOTHING); + + return(ret); +} + +/* + - regpiece - something followed by possible [*+?] + * + * Note that the branching code sequences used for ? and the general cases + * of * and + are somewhat optimized: they use the same NOTHING node as + * both the endmarker for their branch list and the body of the last branch. + * It might seem that this node could be dispensed with entirely, but the + * endmarker role is not redundant. + */ +static char * +regpiece(int *flagp) +{ + register char *ret; + register char op; + register char *next; + int flags; + + ret = regatom(&flags); + if (ret == NULL) + return(NULL); + + op = *regparse; + if (!ISMULT(op)) { + *flagp = flags; + return(ret); + } + + if (!(flags&HASWIDTH) && op != '?') + FAIL("*+ operand could be empty"); + *flagp = (op != '+') ? (WORST|SPSTART) : (WORST|HASWIDTH); + + if (op == '*' && (flags&SIMPLE)) + reginsert(STAR, ret); + else if (op == '*') { + /* Emit x* as (x&|), where & means "self". */ + reginsert(BRANCH, ret); /* Either x */ + regoptail(ret, regnode(BACK)); /* and loop */ + regoptail(ret, ret); /* back */ + regtail(ret, regnode(BRANCH)); /* or */ + regtail(ret, regnode(NOTHING)); /* null. */ + } else if (op == '+' && (flags&SIMPLE)) + reginsert(PLUS, ret); + else if (op == '+') { + /* Emit x+ as x(&|), where & means "self". */ + next = regnode(BRANCH); /* Either */ + regtail(ret, next); + regtail(regnode(BACK), ret); /* loop back */ + regtail(next, regnode(BRANCH)); /* or */ + regtail(ret, regnode(NOTHING)); /* null. */ + } else if (op == '?') { + /* Emit x? as (x|) */ + reginsert(BRANCH, ret); /* Either x */ + regtail(ret, regnode(BRANCH)); /* or */ + next = regnode(NOTHING); /* null. */ + regtail(ret, next); + regoptail(ret, next); + } + regparse++; + if (ISMULT(*regparse)) + FAIL("nested *?+"); + + return(ret); +} + +/* + - regatom - the lowest level + * + * Optimization: gobbles an entire sequence of ordinary characters so that + * it can turn them into a single node, which is smaller to store and + * faster to run. Backslashed characters are exceptions, each becoming a + * separate node; the code is simpler that way and it's not worth fixing. + */ +static char * +regatom(int *flagp) +{ + register char *ret; + int flags; + + *flagp = WORST; /* Tentatively. */ + + switch (*regparse++) { + case '^': + ret = regnode(BOL); + break; + case '$': + ret = regnode(EOL); + break; + case '.': + ret = regnode(ANY); + *flagp |= HASWIDTH|SIMPLE; + break; + case '[': { + register int class; + register int classend; + + if (*regparse == '^') { /* Complement of range. */ + ret = regnode(ANYBUT); + regparse++; + } else + ret = regnode(ANYOF); + if (*regparse == ']' || *regparse == '-') + regc(*regparse++); + while (*regparse != '\0' && *regparse != ']') { + if (*regparse == '-') { + regparse++; + if (*regparse == ']' || *regparse == '\0') + regc('-'); + else { + class = UCHARAT(regparse-2)+1; + classend = UCHARAT(regparse); + if (class > classend+1) + FAIL("invalid [] range"); + for (; class <= classend; class++) + regc(class); + regparse++; + } + } else + regc(*regparse++); + } + regc('\0'); + if (*regparse != ']') + FAIL("unmatched []"); + regparse++; + *flagp |= HASWIDTH|SIMPLE; + } + break; + case '(': + ret = reg(1, &flags); + if (ret == NULL) + return(NULL); + *flagp |= flags&(HASWIDTH|SPSTART); + break; + case '\0': + case '|': + case ')': + FAIL("internal urp"); /* Supposed to be caught earlier. */ + break; + case '?': + case '+': + case '*': + FAIL("?+* follows nothing"); + break; + case '\\': + if (*regparse == '\0') + FAIL("trailing \\"); + ret = regnode(EXACTLY); + regc(*regparse++); + regc('\0'); + *flagp |= HASWIDTH|SIMPLE; + break; + default: { + register int len; + register char ender; + + regparse--; + len = my_strcspn((const char *)regparse, (const char *)META); + if (len <= 0) + FAIL("internal disaster"); + ender = *(regparse+len); + if (len > 1 && ISMULT(ender)) + len--; /* Back off clear of ?+* operand. */ + *flagp |= HASWIDTH; + if (len == 1) + *flagp |= SIMPLE; + ret = regnode(EXACTLY); + while (len > 0) { + regc(*regparse++); + len--; + } + regc('\0'); + } + break; + } + + return(ret); +} + +/* + - regnode - emit a node + */ +static char * /* Location. */ +regnode(char op) +{ + register char *ret; + register char *ptr; + + ret = regcode; + if (ret == ®dummy) { + regsize += 3; + return(ret); + } + + ptr = ret; + *ptr++ = op; + *ptr++ = '\0'; /* Null "next" pointer. */ + *ptr++ = '\0'; + regcode = ptr; + + return(ret); +} + +/* + - regc - emit (if appropriate) a byte of code + */ +static void +regc(char b) +{ + if (regcode != ®dummy) + *regcode++ = b; + else + regsize++; +} + +/* + - reginsert - insert an operator in front of already-emitted operand + * + * Means relocating the operand. + */ +static void +reginsert(char op, char* opnd) +{ + register char *src; + register char *dst; + register char *place; + + if (regcode == ®dummy) { + regsize += 3; + return; + } + + src = regcode; + regcode += 3; + dst = regcode; + while (src > opnd) + *--dst = *--src; + + place = opnd; /* Op node, where operand used to be. */ + *place++ = op; + *place++ = '\0'; + *place++ = '\0'; +} + +/* + - regtail - set the next-pointer at the end of a node chain + */ +static void +regtail(char *p, char *val) +{ + register char *scan; + register char *temp; + register int offset; + + if (p == ®dummy) + return; + + /* Find last node. */ + scan = p; + for (;;) { + temp = regnext(scan); + if (temp == NULL) + break; + scan = temp; + } + + if (OP(scan) == BACK) + offset = scan - val; + else + offset = val - scan; + *(scan+1) = (offset>>8)&0377; + *(scan+2) = offset&0377; +} + +/* + - regoptail - regtail on operand of first argument; nop if operandless + */ +static void +regoptail(char *p, char *val) +{ + /* "Operandless" and "op != BRANCH" are synonymous in practice. */ + if (p == NULL || p == ®dummy || OP(p) != BRANCH) + return; + regtail(OPERAND(p), val); +} + +/* + * regexec and friends + */ + +/* + * Global work variables for regexec(). + */ +static char *reginput; /* String-input pointer. */ +static char *regbol; /* Beginning of input, for ^ check. */ +static char **regstartp; /* Pointer to startp array. */ +static char **regendp; /* Ditto for endp. */ + +/* + * Forwards. + */ +STATIC int regtry(regexp *prog, char *string); +STATIC int regmatch(char *prog); +STATIC int regrepeat(char *p); + +#ifdef DEBUG +int regnarrate = 0; +void regdump(); +STATIC char *regprop(char *op); +#endif + +/* + - regexec - match a regexp against a string + */ +int +regexec(regexp *prog, char *string) +{ + register char *s; + + /* Be paranoid... */ + if (prog == NULL || string == NULL) { + printk("<3>Regexp: NULL parameter\n"); + return(0); + } + + /* Check validity of program. */ + if (UCHARAT(prog->program) != MAGIC) { + printk("<3>Regexp: corrupted program\n"); + return(0); + } + + /* If there is a "must appear" string, look for it. */ + if (prog->regmust != NULL) { + s = string; + while ((s = strchr(s, prog->regmust[0])) != NULL) { + if (strncmp(s, prog->regmust, prog->regmlen) == 0) + break; /* Found it. */ + s++; + } + if (s == NULL) /* Not present. */ + return(0); + } + + /* Mark beginning of line for ^ . */ + regbol = string; + + /* Simplest case: anchored match need be tried only once. */ + if (prog->reganch) + return(regtry(prog, string)); + + /* Messy cases: unanchored match. */ + s = string; + if (prog->regstart != '\0') + /* We know what char it must start with. */ + while ((s = strchr(s, prog->regstart)) != NULL) { + if (regtry(prog, s)) + return(1); + s++; + } + else + /* We don't -- general case. */ + do { + if (regtry(prog, s)) + return(1); + } while (*s++ != '\0'); + + /* Failure. */ + return(0); +} + +/* + - regtry - try match at specific point + */ +static int /* 0 failure, 1 success */ +regtry(regexp *prog, char *string) +{ + register int i; + register char **sp; + register char **ep; + + reginput = string; + regstartp = prog->startp; + regendp = prog->endp; + + sp = prog->startp; + ep = prog->endp; + for (i = NSUBEXP; i > 0; i--) { + *sp++ = NULL; + *ep++ = NULL; + } + if (regmatch(prog->program + 1)) { + prog->startp[0] = string; + prog->endp[0] = reginput; + return(1); + } else + return(0); +} + +/* + - regmatch - main matching routine + * + * Conceptually the strategy is simple: check to see whether the current + * node matches, call self recursively to see whether the rest matches, + * and then act accordingly. In practice we make some effort to avoid + * recursion, in particular by going through "ordinary" nodes (that don't + * need to know whether the rest of the match failed) by a loop instead of + * by recursion. + */ +static int /* 0 failure, 1 success */ +regmatch(char *prog) +{ + register char *scan = prog; /* Current node. */ + char *next; /* Next node. */ + +#ifdef DEBUG + if (scan != NULL && regnarrate) + fprintf(stderr, "%s(\n", regprop(scan)); +#endif + while (scan != NULL) { +#ifdef DEBUG + if (regnarrate) + fprintf(stderr, "%s...\n", regprop(scan)); +#endif + next = regnext(scan); + + switch (OP(scan)) { + case BOL: + if (reginput != regbol) + return(0); + break; + case EOL: + if (*reginput != '\0') + return(0); + break; + case ANY: + if (*reginput == '\0') + return(0); + reginput++; + break; + case EXACTLY: { + register int len; + register char *opnd; + + opnd = OPERAND(scan); + /* Inline the first character, for speed. */ + if (*opnd != *reginput) + return(0); + len = strlen(opnd); + if (len > 1 && strncmp(opnd, reginput, len) != 0) + return(0); + reginput += len; + } + break; + case ANYOF: + if (*reginput == '\0' || strchr(OPERAND(scan), *reginput) == NULL) + return(0); + reginput++; + break; + case ANYBUT: + if (*reginput == '\0' || strchr(OPERAND(scan), *reginput) != NULL) + return(0); + reginput++; + break; + case NOTHING: + case BACK: + break; + case OPEN+1: + case OPEN+2: + case OPEN+3: + case OPEN+4: + case OPEN+5: + case OPEN+6: + case OPEN+7: + case OPEN+8: + case OPEN+9: { + register int no; + register char *save; + + no = OP(scan) - OPEN; + save = reginput; + + if (regmatch(next)) { + /* + * Don't set startp if some later + * invocation of the same parentheses + * already has. + */ + if (regstartp[no] == NULL) + regstartp[no] = save; + return(1); + } else + return(0); + } + break; + case CLOSE+1: + case CLOSE+2: + case CLOSE+3: + case CLOSE+4: + case CLOSE+5: + case CLOSE+6: + case CLOSE+7: + case CLOSE+8: + case CLOSE+9: + { + register int no; + register char *save; + + no = OP(scan) - CLOSE; + save = reginput; + + if (regmatch(next)) { + /* + * Don't set endp if some later + * invocation of the same parentheses + * already has. + */ + if (regendp[no] == NULL) + regendp[no] = save; + return(1); + } else + return(0); + } + break; + case BRANCH: { + register char *save; + + if (OP(next) != BRANCH) /* No choice. */ + next = OPERAND(scan); /* Avoid recursion. */ + else { + do { + save = reginput; + if (regmatch(OPERAND(scan))) + return(1); + reginput = save; + scan = regnext(scan); + } while (scan != NULL && OP(scan) == BRANCH); + return(0); + /* NOTREACHED */ + } + } + break; + case STAR: + case PLUS: { + register char nextch; + register int no; + register char *save; + register int min; + + /* + * Lookahead to avoid useless match attempts + * when we know what character comes next. + */ + nextch = '\0'; + if (OP(next) == EXACTLY) + nextch = *OPERAND(next); + min = (OP(scan) == STAR) ? 0 : 1; + save = reginput; + no = regrepeat(OPERAND(scan)); + while (no >= min) { + /* If it could work, try it. */ + if (nextch == '\0' || *reginput == nextch) + if (regmatch(next)) + return(1); + /* Couldn't or didn't -- back up. */ + no--; + reginput = save + no; + } + return(0); + } + break; + case END: + return(1); /* Success! */ + break; + default: + printk("<3>Regexp: memory corruption\n"); + return(0); + break; + } + + scan = next; + } + + /* + * We get here only if there's trouble -- normally "case END" is + * the terminating point. + */ + printk("<3>Regexp: corrupted pointers\n"); + return(0); +} + +/* + - regrepeat - repeatedly match something simple, report how many + */ +static int +regrepeat(char *p) +{ + register int count = 0; + register char *scan; + register char *opnd; + + scan = reginput; + opnd = OPERAND(p); + switch (OP(p)) { + case ANY: + count = strlen(scan); + scan += count; + break; + case EXACTLY: + while (*opnd == *scan) { + count++; + scan++; + } + break; + case ANYOF: + while (*scan != '\0' && strchr(opnd, *scan) != NULL) { + count++; + scan++; + } + break; + case ANYBUT: + while (*scan != '\0' && strchr(opnd, *scan) == NULL) { + count++; + scan++; + } + break; + default: /* Oh dear. Called inappropriately. */ + printk("<3>Regexp: internal foulup\n"); + count = 0; /* Best compromise. */ + break; + } + reginput = scan; + + return(count); +} + +/* + - regnext - dig the "next" pointer out of a node + */ +static char* +regnext(char *p) +{ + register int offset; + + if (p == ®dummy) + return(NULL); + + offset = NEXT(p); + if (offset == 0) + return(NULL); + + if (OP(p) == BACK) + return(p-offset); + else + return(p+offset); +} + +#ifdef DEBUG + +STATIC char *regprop(); + +/* + - regdump - dump a regexp onto stdout in vaguely comprehensible form + */ +void +regdump(regexp *r) +{ + register char *s; + register char op = EXACTLY; /* Arbitrary non-END op. */ + register char *next; + /* extern char *strchr(); */ + + + s = r->program + 1; + while (op != END) { /* While that wasn't END last time... */ + op = OP(s); + printf("%2d%s", s-r->program, regprop(s)); /* Where, what. */ + next = regnext(s); + if (next == NULL) /* Next ptr. */ + printf("(0)"); + else + printf("(%d)", (s-r->program)+(next-s)); + s += 3; + if (op == ANYOF || op == ANYBUT || op == EXACTLY) { + /* Literal string, where present. */ + while (*s != '\0') { + putchar(*s); + s++; + } + s++; + } + putchar('\n'); + } + + /* Header fields of interest. */ + if (r->regstart != '\0') + printf("start `%c' ", r->regstart); + if (r->reganch) + printf("anchored "); + if (r->regmust != NULL) + printf("must have \"%s\"", r->regmust); + printf("\n"); +} + +/* + - regprop - printable representation of opcode + */ +static char * +regprop(char *op) +{ +#define BUFLEN 50 + register char *p; + static char buf[BUFLEN]; + + strcpy(buf, ":"); + + switch (OP(op)) { + case BOL: + p = "BOL"; + break; + case EOL: + p = "EOL"; + break; + case ANY: + p = "ANY"; + break; + case ANYOF: + p = "ANYOF"; + break; + case ANYBUT: + p = "ANYBUT"; + break; + case BRANCH: + p = "BRANCH"; + break; + case EXACTLY: + p = "EXACTLY"; + break; + case NOTHING: + p = "NOTHING"; + break; + case BACK: + p = "BACK"; + break; + case END: + p = "END"; + break; + case OPEN+1: + case OPEN+2: + case OPEN+3: + case OPEN+4: + case OPEN+5: + case OPEN+6: + case OPEN+7: + case OPEN+8: + case OPEN+9: + snprintf(buf+strlen(buf),BUFLEN-strlen(buf), "OPEN%d", OP(op)-OPEN); + p = NULL; + break; + case CLOSE+1: + case CLOSE+2: + case CLOSE+3: + case CLOSE+4: + case CLOSE+5: + case CLOSE+6: + case CLOSE+7: + case CLOSE+8: + case CLOSE+9: + snprintf(buf+strlen(buf),BUFLEN-strlen(buf), "CLOSE%d", OP(op)-CLOSE); + p = NULL; + break; + case STAR: + p = "STAR"; + break; + case PLUS: + p = "PLUS"; + break; + default: + printk("<3>Regexp: corrupted opcode\n"); + break; + } + if (p != NULL) + strncat(buf, p, BUFLEN-strlen(buf)); + return(buf); +} +#endif + + diff -Nurp linux-2.4.30/net/ipv4/netfilter/regexp/regexp.h linux-2.4.30-layer7/net/ipv4/netfilter/regexp/regexp.h --- linux-2.4.30/net/ipv4/netfilter/regexp/regexp.h 1969-12-31 18:00:00.000000000 -0600 +++ linux-2.4.30-layer7/net/ipv4/netfilter/regexp/regexp.h 2005-05-03 18:37:03.000000000 -0500 @@ -0,0 +1,40 @@ +/* + * Definitions etc. for regexp(3) routines. + * + * Caveat: this is V8 regexp(3) [actually, a reimplementation thereof], + * not the System V one. + */ + +#ifndef REGEXP_H +#define REGEXP_H + +/* +http://www.opensource.apple.com/darwinsource/10.3/expect-1/expect/expect.h , +which contains a version of this library, says: + + * + * NSUBEXP must be at least 10, and no greater than 117 or the parser + * will not work properly. + * + +However, it looks rather like this library is limited to 10. If you think +otherwise, let us know. +*/ + +#define NSUBEXP 10 +typedef struct regexp { + char *startp[NSUBEXP]; + char *endp[NSUBEXP]; + char regstart; /* Internal use only. */ + char reganch; /* Internal use only. */ + char *regmust; /* Internal use only. */ + int regmlen; /* Internal use only. */ + char program[1]; /* Unwarranted chumminess with compiler. */ +} regexp; + +regexp * regcomp(char *exp, int *patternsize); +int regexec(regexp *prog, char *string); +void regsub(regexp *prog, char *source, char *dest); +void regerror(char *s); + +#endif diff -Nurp linux-2.4.30/net/ipv4/netfilter/regexp/regmagic.h linux-2.4.30-layer7/net/ipv4/netfilter/regexp/regmagic.h --- linux-2.4.30/net/ipv4/netfilter/regexp/regmagic.h 1969-12-31 18:00:00.000000000 -0600 +++ linux-2.4.30-layer7/net/ipv4/netfilter/regexp/regmagic.h 2005-05-03 18:37:03.000000000 -0500 @@ -0,0 +1,5 @@ +/* + * The first byte of the regexp internal "program" is actually this magic + * number; the start node begins in the second byte. + */ +#define MAGIC 0234 diff -Nurp linux-2.4.30/net/ipv4/netfilter/regexp/regsub.c linux-2.4.30-layer7/net/ipv4/netfilter/regexp/regsub.c --- linux-2.4.30/net/ipv4/netfilter/regexp/regsub.c 1969-12-31 18:00:00.000000000 -0600 +++ linux-2.4.30-layer7/net/ipv4/netfilter/regexp/regsub.c 2005-05-03 18:37:03.000000000 -0500 @@ -0,0 +1,95 @@ +/* + * regsub + * @(#)regsub.c 1.3 of 2 April 86 + * + * Copyright (c) 1986 by University of Toronto. + * Written by Henry Spencer. Not derived from licensed software. + * + * Permission is granted to anyone to use this software for any + * purpose on any computer system, and to redistribute it freely, + * subject to the following restrictions: + * + * 1. The author is not responsible for the consequences of use of + * this software, no matter how awful, even if they arise + * from defects in it. + * + * 2. The origin of this software must not be misrepresented, either + * by explicit claim or by omission. + * + * 3. Altered versions must be plainly marked as such, and must not + * be misrepresented as being the original software. + * + * + * This code was modified by Ethan Sommer to work within the kernel + * (it now uses kmalloc etc..) + * + */ +#include "regexp.h" +#include "regmagic.h" +#include + + +#ifndef CHARBITS +#define UCHARAT(p) ((int)*(unsigned char *)(p)) +#else +#define UCHARAT(p) ((int)*(p)&CHARBITS) +#endif + +#if 0 +//void regerror(char * s) +//{ +// printk("regexp(3): %s", s); +// /* NOTREACHED */ +//} +#endif + +/* + - regsub - perform substitutions after a regexp match + */ +void +regsub(regexp * prog, char * source, char * dest) +{ + register char *src; + register char *dst; + register char c; + register int no; + register int len; + + /* Not necessary and gcc doesn't like it -MLS */ + /*extern char *strncpy();*/ + + if (prog == NULL || source == NULL || dest == NULL) { + regerror("NULL parm to regsub"); + return; + } + if (UCHARAT(prog->program) != MAGIC) { + regerror("damaged regexp fed to regsub"); + return; + } + + src = source; + dst = dest; + while ((c = *src++) != '\0') { + if (c == '&') + no = 0; + else if (c == '\\' && '0' <= *src && *src <= '9') + no = *src++ - '0'; + else + no = -1; + + if (no < 0) { /* Ordinary character. */ + if (c == '\\' && (*src == '\\' || *src == '&')) + c = *src++; + *dst++ = c; + } else if (prog->startp[no] != NULL && prog->endp[no] != NULL) { + len = prog->endp[no] - prog->startp[no]; + (void) strncpy(dst, prog->startp[no], len); + dst += len; + if (len != 0 && *(dst-1) == '\0') { /* strncpy hit NUL. */ + regerror("damaged match string"); + return; + } + } + } + *dst++ = '\0'; +}