[NETFILTER]: Add new "nfnetlink_log" userspace packet logging facility
authorHarald Welte <laforge@netfilter.org>
Wed, 10 Aug 2005 02:58:39 +0000 (19:58 -0700)
committerDavid S. Miller <davem@sunset.davemloft.net>
Mon, 29 Aug 2005 22:38:12 +0000 (15:38 -0700)
This is a generic (layer3 independent) version of what ipt_ULOG is already
doing for IPv4 today.  ipt_ULOG, ebt_ulog and finally also ip[6]t_LOG will
be deprecated by this mechanism in the long term.

Signed-off-by: Harald Welte <laforge@netfilter.org>
Signed-off-by: David S. Miller <davem@davemloft.net>

include/linux/netfilter/nfnetlink_log.h [new file with mode: 0644]
net/netfilter/Kconfig
net/netfilter/Makefile
net/netfilter/nfnetlink_log.c [new file with mode: 0644]
net/netfilter/nfnetlink_queue.c

diff --git a/include/linux/netfilter/nfnetlink_log.h b/include/linux/netfilter/nfnetlink_log.h
new file mode 100644 (file)
index 0000000..420ff46
--- /dev/null
@@ -0,0 +1,85 @@
+#ifndef _NFNETLINK_LOG_H
+#define _NFNETLINK_LOG_H
+
+/* This file describes the netlink messages (i.e. 'protocol packets'),
+ * and not any kind of function definitions.  It is shared between kernel and
+ * userspace.  Don't put kernel specific stuff in here */
+
+#include <linux/netfilter/nfnetlink.h>
+
+enum nfulnl_msg_types {
+       NFULNL_MSG_PACKET,              /* packet from kernel to userspace */
+       NFULNL_MSG_CONFIG,              /* connect to a particular queue */
+
+       NFULNL_MSG_MAX
+};
+
+struct nfulnl_msg_packet_hdr {
+       u_int16_t       hw_protocol;    /* hw protocol (network order) */
+       u_int8_t        hook;           /* netfilter hook */
+       u_int8_t        _pad;
+} __attribute__ ((packed));
+
+struct nfulnl_msg_packet_hw {
+       u_int16_t       hw_addrlen;
+       u_int16_t       _pad;
+       u_int8_t        hw_addr[8];
+} __attribute__ ((packed));
+
+struct nfulnl_msg_packet_timestamp {
+       u_int64_t       sec;
+       u_int64_t       usec;
+} __attribute__ ((packed));
+
+#define NFULNL_PREFIXLEN       30      /* just like old log target */
+
+enum nfulnl_attr_type {
+       NFULA_UNSPEC,
+       NFULA_PACKET_HDR,
+       NFULA_MARK,                     /* u_int32_t nfmark */
+       NFULA_TIMESTAMP,                /* nfulnl_msg_packet_timestamp */
+       NFULA_IFINDEX_INDEV,            /* u_int32_t ifindex */
+       NFULA_IFINDEX_OUTDEV,           /* u_int32_t ifindex */
+       NFULA_HWADDR,                   /* nfulnl_msg_packet_hw */
+       NFULA_PAYLOAD,                  /* opaque data payload */
+       NFULA_PREFIX,                   /* string prefix */
+       NFULA_UID,                      /* user id of socket */
+
+       __NFULA_MAX
+};
+#define NFULA_MAX (__NFULA_MAX - 1)
+
+enum nfulnl_msg_config_cmds {
+       NFULNL_CFG_CMD_NONE,
+       NFULNL_CFG_CMD_BIND,
+       NFULNL_CFG_CMD_UNBIND,
+       NFULNL_CFG_CMD_PF_BIND,
+       NFULNL_CFG_CMD_PF_UNBIND,
+};
+
+struct nfulnl_msg_config_cmd {
+       u_int8_t        command;        /* nfulnl_msg_config_cmds */
+} __attribute__ ((packed));
+
+struct nfulnl_msg_config_mode {
+       u_int32_t       copy_range;
+       u_int8_t        copy_mode;
+       u_int8_t        _pad;
+} __attribute__ ((packed));
+
+enum nfulnl_attr_config {
+       NFULA_CFG_UNSPEC,
+       NFULA_CFG_CMD,                  /* nfulnl_msg_config_cmd */
+       NFULA_CFG_MODE,                 /* nfulnl_msg_config_mode */
+       NFULA_CFG_NLBUFSIZ,             /* u_int32_t buffer size */
+       NFULA_CFG_TIMEOUT,              /* u_int32_t in 1/100 s */
+       NFULA_CFG_QTHRESH,              /* u_int32_t */
+       __NFULA_CFG_MAX
+};
+#define NFULA_CFG_MAX (__NFULA_CFG_MAX -1)
+
+#define NFULNL_COPY_NONE       0x00
+#define NFULNL_COPY_META       0x01
+#define NFULNL_COPY_PACKET     0x02
+
+#endif /* _NFNETLINK_LOG_H */
index f0eb23e..8296b38 100644 (file)
@@ -11,3 +11,14 @@ config NETFILTER_NETLINK_QUEUE
          If this option isenabled, the kernel will include support
          for queueing packets via NFNETLINK.
          
+config NETFILTER_NETLINK_LOG
+       tristate "Netfilter LOG over NFNETLINK interface"
+       depends on NETFILTER_NETLINK
+       help
+         If this option is enabled, the kernel will include support
+         for logging packets via NFNETLINK.
+
+         This obsoletes the existing ipt_ULOG and ebg_ulog mechanisms,
+         and is also scheduled to replace the old syslog-based ipt_LOG
+         and ip6t_LOG modules.
+
index 14a0b18..c41caeb 100644 (file)
@@ -1,2 +1,3 @@
 obj-$(CONFIG_NETFILTER_NETLINK) += nfnetlink.o
 obj-$(CONFIG_NETFILTER_NETLINK_QUEUE) += nfnetlink_queue.o
+obj-$(CONFIG_NETFILTER_NETLINK_LOG) += nfnetlink_log.o
diff --git a/net/netfilter/nfnetlink_log.c b/net/netfilter/nfnetlink_log.c
new file mode 100644 (file)
index 0000000..f41045e
--- /dev/null
@@ -0,0 +1,995 @@
+/*
+ * This is a module which is used for logging packets to userspace via
+ * nfetlink.
+ *
+ * (C) 2005 by Harald Welte <laforge@netfilter.org>
+ *
+ * Based on the old ipv4-only ipt_ULOG.c:
+ * (C) 2000-2004 by Harald Welte <laforge@netfilter.org>
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of the GNU General Public License version 2 as
+ * published by the Free Software Foundation.
+ *
+ */
+#include <linux/module.h>
+#include <linux/skbuff.h>
+#include <linux/init.h>
+#include <linux/ip.h>
+#include <linux/ipv6.h>
+#include <linux/netdevice.h>
+#include <linux/netfilter.h>
+#include <linux/netlink.h>
+#include <linux/netfilter/nfnetlink.h>
+#include <linux/netfilter/nfnetlink_log.h>
+#include <linux/spinlock.h>
+#include <linux/sysctl.h>
+#include <linux/proc_fs.h>
+#include <linux/security.h>
+#include <linux/list.h>
+#include <linux/jhash.h>
+#include <linux/random.h>
+#include <net/sock.h>
+
+#include <asm/atomic.h>
+
+#define NFULNL_NLBUFSIZ_DEFAULT        4096
+#define NFULNL_TIMEOUT_DEFAULT         100     /* every second */
+#define NFULNL_QTHRESH_DEFAULT         100     /* 100 packets */
+
+#define PRINTR(x, args...)     do { if (net_ratelimit()) \
+                                    printk(x, ## args); } while (0);
+
+#if 0
+#define UDEBUG(x, args ...)    printk(KERN_DEBUG "%s(%d):%s(): " x,       \
+                                       __FILE__, __LINE__, __FUNCTION__,  \
+                                       ## args)
+#else
+#define UDEBUG(x, ...)
+#endif
+
+struct nfulnl_instance {
+       struct hlist_node hlist;        /* global list of instances */
+       spinlock_t lock;
+       atomic_t use;                   /* use count */
+
+       unsigned int qlen;              /* number of nlmsgs in skb */
+       struct sk_buff *skb;            /* pre-allocatd skb */
+       struct nlmsghdr *lastnlh;       /* netlink header of last msg in skb */
+       struct timer_list timer;
+       int peer_pid;                   /* PID of the peer process */
+
+       /* configurable parameters */
+       unsigned int flushtimeout;      /* timeout until queue flush */
+       unsigned int nlbufsiz;          /* netlink buffer allocation size */
+       unsigned int qthreshold;        /* threshold of the queue */
+       u_int32_t copy_range;
+       u_int16_t group_num;            /* number of this queue */
+       u_int8_t copy_mode;     
+};
+
+static DEFINE_RWLOCK(instances_lock);
+
+#define INSTANCE_BUCKETS       16
+static struct hlist_head instance_table[INSTANCE_BUCKETS];
+static unsigned int hash_init;
+
+static inline u_int8_t instance_hashfn(u_int16_t group_num)
+{
+       return ((group_num & 0xff) % INSTANCE_BUCKETS);
+}
+
+static struct nfulnl_instance *
+__instance_lookup(u_int16_t group_num)
+{
+       struct hlist_head *head;
+       struct hlist_node *pos;
+       struct nfulnl_instance *inst;
+
+       UDEBUG("entering (group_num=%u)\n", group_num);
+
+       head = &instance_table[instance_hashfn(group_num)];
+       hlist_for_each_entry(inst, pos, head, hlist) {
+               if (inst->group_num == group_num)
+                       return inst;
+       }
+       return NULL;
+}
+
+static inline void
+instance_get(struct nfulnl_instance *inst)
+{
+       atomic_inc(&inst->use);
+}
+
+static struct nfulnl_instance *
+instance_lookup_get(u_int16_t group_num)
+{
+       struct nfulnl_instance *inst;
+
+       read_lock_bh(&instances_lock);
+       inst = __instance_lookup(group_num);
+       if (inst)
+               instance_get(inst);
+       read_unlock_bh(&instances_lock);
+
+       return inst;
+}
+
+static void
+instance_put(struct nfulnl_instance *inst)
+{
+       if (inst && atomic_dec_and_test(&inst->use)) {
+               UDEBUG("kfree(inst=%p)\n", inst);
+               kfree(inst);
+       }
+}
+
+static void nfulnl_timer(unsigned long data);
+
+static struct nfulnl_instance *
+instance_create(u_int16_t group_num, int pid)
+{
+       struct nfulnl_instance *inst;
+
+       UDEBUG("entering (group_num=%u, pid=%d)\n", group_num,
+               pid);
+
+       write_lock_bh(&instances_lock); 
+       if (__instance_lookup(group_num)) {
+               inst = NULL;
+               UDEBUG("aborting, instance already exists\n");
+               goto out_unlock;
+       }
+
+       inst = kmalloc(sizeof(*inst), GFP_ATOMIC);
+       if (!inst)
+               goto out_unlock;
+
+       memset(inst, 0, sizeof(*inst));
+       INIT_HLIST_NODE(&inst->hlist);
+       inst->lock = SPIN_LOCK_UNLOCKED;
+       /* needs to be two, since we _put() after creation */
+       atomic_set(&inst->use, 2);
+
+       init_timer(&inst->timer);
+       inst->timer.function = nfulnl_timer;
+       inst->timer.data = (unsigned long)inst;
+       /* don't start timer yet. (re)start it  with every packet */
+
+       inst->peer_pid = pid;
+       inst->group_num = group_num;
+
+       inst->qthreshold        = NFULNL_QTHRESH_DEFAULT;
+       inst->flushtimeout      = NFULNL_TIMEOUT_DEFAULT;
+       inst->nlbufsiz          = NFULNL_NLBUFSIZ_DEFAULT;
+       inst->copy_mode         = NFULNL_COPY_PACKET;
+       inst->copy_range        = 0xffff;
+
+       if (!try_module_get(THIS_MODULE))
+               goto out_free;
+
+       hlist_add_head(&inst->hlist, 
+                      &instance_table[instance_hashfn(group_num)]);
+
+       UDEBUG("newly added node: %p, next=%p\n", &inst->hlist, 
+               inst->hlist.next);
+
+       write_unlock_bh(&instances_lock);
+
+       return inst;
+
+out_free:
+       instance_put(inst);
+out_unlock:
+       write_unlock_bh(&instances_lock);
+       return NULL;
+}
+
+static int __nfulnl_send(struct nfulnl_instance *inst);
+
+static void
+_instance_destroy2(struct nfulnl_instance *inst, int lock)
+{
+       /* first pull it out of the global list */
+       if (lock)
+               write_lock_bh(&instances_lock);
+
+       UDEBUG("removing instance %p (queuenum=%u) from hash\n",
+               inst, inst->group_num);
+
+       hlist_del(&inst->hlist);
+
+       if (lock)
+               write_unlock_bh(&instances_lock);
+
+       /* then flush all pending packets from skb */
+
+       spin_lock_bh(&inst->lock);
+       if (inst->skb) {
+               if (inst->qlen)
+                       __nfulnl_send(inst);
+               if (inst->skb) {
+                       kfree_skb(inst->skb);
+                       inst->skb = NULL;
+               }
+       }
+       spin_unlock_bh(&inst->lock);
+
+       /* and finally put the refcount */
+       instance_put(inst);
+
+       module_put(THIS_MODULE);
+}
+
+static inline void
+__instance_destroy(struct nfulnl_instance *inst)
+{
+       _instance_destroy2(inst, 0);
+}
+
+static inline void
+instance_destroy(struct nfulnl_instance *inst)
+{
+       _instance_destroy2(inst, 1);
+}
+
+static int
+nfulnl_set_mode(struct nfulnl_instance *inst, u_int8_t mode,
+                 unsigned int range)
+{
+       int status = 0;
+
+       spin_lock_bh(&inst->lock);
+       
+       switch (mode) {
+       case NFULNL_COPY_NONE:
+       case NFULNL_COPY_META:
+               inst->copy_mode = mode;
+               inst->copy_range = 0;
+               break;
+               
+       case NFULNL_COPY_PACKET:
+               inst->copy_mode = mode;
+               /* we're using struct nfattr which has 16bit nfa_len */
+               if (range > 0xffff)
+                       inst->copy_range = 0xffff;
+               else
+                       inst->copy_range = range;
+               break;
+               
+       default:
+               status = -EINVAL;
+               break;
+       }
+
+       spin_unlock_bh(&inst->lock);
+
+       return status;
+}
+
+static int
+nfulnl_set_nlbufsiz(struct nfulnl_instance *inst, u_int32_t nlbufsiz)
+{
+       int status;
+
+       spin_lock_bh(&inst->lock);
+       if (nlbufsiz < NFULNL_NLBUFSIZ_DEFAULT)
+               status = -ERANGE;
+       else if (nlbufsiz > 131072)
+               status = -ERANGE;
+       else {
+               inst->nlbufsiz = nlbufsiz;
+               status = 0;
+       }
+       spin_unlock_bh(&inst->lock);
+
+       return status;
+}
+
+static int
+nfulnl_set_timeout(struct nfulnl_instance *inst, u_int32_t timeout)
+{
+       spin_lock_bh(&inst->lock);
+       inst->flushtimeout = timeout;
+       spin_unlock_bh(&inst->lock);
+
+       return 0;
+}
+
+static int
+nfulnl_set_qthresh(struct nfulnl_instance *inst, u_int32_t qthresh)
+{
+       spin_lock_bh(&inst->lock);
+       inst->qthreshold = qthresh;
+       spin_unlock_bh(&inst->lock);
+
+       return 0;
+}
+
+static struct sk_buff *nfulnl_alloc_skb(unsigned int inst_size, 
+                                       unsigned int pkt_size)
+{
+       struct sk_buff *skb;
+
+       UDEBUG("entered (%u, %u)\n", inst_size, pkt_size);
+
+       /* alloc skb which should be big enough for a whole multipart
+        * message.  WARNING: has to be <= 128k due to slab restrictions */
+
+       skb = alloc_skb(inst_size, GFP_ATOMIC);
+       if (!skb) {
+               PRINTR("nfnetlink_log: can't alloc whole buffer (%u bytes)\n",
+                       inst_size);
+
+               /* try to allocate only as much as we need for current
+                * packet */
+
+               skb = alloc_skb(pkt_size, GFP_ATOMIC);
+               if (!skb)
+                       PRINTR("nfnetlink_log: can't even alloc %u bytes\n",
+                               pkt_size);
+       }
+
+       return skb;
+}
+
+static int
+__nfulnl_send(struct nfulnl_instance *inst)
+{
+       int status;
+
+       if (timer_pending(&inst->timer))
+               del_timer(&inst->timer);
+
+       if (inst->qlen > 1)
+               inst->lastnlh->nlmsg_type = NLMSG_DONE;
+
+       status = nfnetlink_unicast(inst->skb, inst->peer_pid, MSG_DONTWAIT);
+       if (status < 0) {
+               UDEBUG("netlink_unicast() failed\n");
+               /* FIXME: statistics */
+       }
+
+       inst->qlen = 0;
+       inst->skb = NULL;
+       inst->lastnlh = NULL;
+
+       return status;
+}
+
+static void nfulnl_timer(unsigned long data)
+{
+       struct nfulnl_instance *inst = (struct nfulnl_instance *)data; 
+
+       UDEBUG("timer function called, flushing buffer\n");
+
+       spin_lock_bh(&inst->lock);
+       __nfulnl_send(inst);
+       instance_put(inst);
+       spin_unlock_bh(&inst->lock);
+}
+
+static inline int 
+__build_packet_message(struct nfulnl_instance *inst,
+                       const struct sk_buff *skb, 
+                       unsigned int data_len,
+                       unsigned int pf,
+                       unsigned int hooknum,
+                       const struct net_device *indev,
+                       const struct net_device *outdev,
+                       const struct nf_loginfo *li,
+                       const char *prefix)
+{
+       unsigned char *old_tail;
+       struct nfulnl_msg_packet_hdr pmsg;
+       struct nlmsghdr *nlh;
+       struct nfgenmsg *nfmsg;
+       u_int32_t tmp_uint;
+
+       UDEBUG("entered\n");
+               
+       old_tail = inst->skb->tail;
+       nlh = NLMSG_PUT(inst->skb, 0, 0, 
+                       NFNL_SUBSYS_ULOG << 8 | NFULNL_MSG_PACKET,
+                       sizeof(struct nfgenmsg));
+       nfmsg = NLMSG_DATA(nlh);
+       nfmsg->nfgen_family = pf;
+       nfmsg->version = NFNETLINK_V0;
+       nfmsg->res_id = htons(inst->group_num);
+
+       pmsg.hw_protocol        = htons(skb->protocol);
+       pmsg.hook               = hooknum;
+
+       NFA_PUT(inst->skb, NFULA_PACKET_HDR, sizeof(pmsg), &pmsg);
+
+       if (prefix) {
+               int slen = strlen(prefix);
+               if (slen > NFULNL_PREFIXLEN)
+                       slen = NFULNL_PREFIXLEN;
+               NFA_PUT(inst->skb, NFULA_PREFIX, slen, prefix);
+       }
+
+       if (indev) {
+               tmp_uint = htonl(indev->ifindex);
+               NFA_PUT(inst->skb, NFULA_IFINDEX_INDEV, sizeof(tmp_uint),
+                       &tmp_uint);
+       }
+
+       if (outdev) {
+               tmp_uint = htonl(outdev->ifindex);
+               NFA_PUT(inst->skb, NFULA_IFINDEX_OUTDEV, sizeof(tmp_uint),
+                       &tmp_uint);
+       }
+
+       if (skb->nfmark) {
+               tmp_uint = htonl(skb->nfmark);
+               NFA_PUT(inst->skb, NFULA_MARK, sizeof(tmp_uint), &tmp_uint);
+       }
+
+       if (indev && skb->dev && skb->dev->hard_header_parse) {
+               struct nfulnl_msg_packet_hw phw;
+
+               phw.hw_addrlen = 
+                       skb->dev->hard_header_parse((struct sk_buff *)skb, 
+                                                   phw.hw_addr);
+               phw.hw_addrlen = htons(phw.hw_addrlen);
+               NFA_PUT(inst->skb, NFULA_HWADDR, sizeof(phw), &phw);
+       }
+
+       if (skb->stamp.tv_sec) {
+               struct nfulnl_msg_packet_timestamp ts;
+
+               ts.sec = cpu_to_be64(skb->stamp.tv_sec);
+               ts.usec = cpu_to_be64(skb->stamp.tv_usec);
+
+               NFA_PUT(inst->skb, NFULA_TIMESTAMP, sizeof(ts), &ts);
+       }
+
+       /* UID */
+       if (skb->sk) {
+               read_lock_bh(&skb->sk->sk_callback_lock);
+               if (skb->sk->sk_socket && skb->sk->sk_socket->file) {
+                       u_int32_t uid = htonl(skb->sk->sk_socket->file->f_uid);
+                       /* need to unlock here since NFA_PUT may goto */
+                       read_unlock_bh(&skb->sk->sk_callback_lock);
+                       NFA_PUT(inst->skb, NFULA_UID, sizeof(uid), &uid);
+               } else
+                       read_unlock_bh(&skb->sk->sk_callback_lock);
+       }
+
+       if (data_len) {
+               struct nfattr *nfa;
+               int size = NFA_LENGTH(data_len);
+
+               if (skb_tailroom(inst->skb) < (int)NFA_SPACE(data_len)) {
+                       printk(KERN_WARNING "nfnetlink_log: no tailroom!\n");
+                       goto nlmsg_failure;
+               }
+
+               nfa = (struct nfattr *)skb_put(inst->skb, NFA_ALIGN(size));
+               nfa->nfa_type = NFULA_PAYLOAD;
+               nfa->nfa_len = size;
+
+               if (skb_copy_bits(skb, 0, NFA_DATA(nfa), data_len))
+                       BUG();
+       }
+               
+       nlh->nlmsg_len = inst->skb->tail - old_tail;
+       return 0;
+
+nlmsg_failure:
+       UDEBUG("nlmsg_failure\n");
+nfattr_failure:
+       PRINTR(KERN_ERR "nfnetlink_log: error creating log nlmsg\n");
+       return -1;
+}
+
+#define RCV_SKB_FAIL(err) do { netlink_ack(skb, nlh, (err)); return; } while (0)
+
+static struct nf_loginfo default_loginfo = {
+       .type =         NF_LOG_TYPE_ULOG,
+       .u = {
+               .ulog = {
+                       .copy_len       = 0xffff,
+                       .group          = 0,
+                       .qthreshold     = 1,
+               },
+       },
+};
+
+/* log handler for internal netfilter logging api */
+static void
+nfulnl_log_packet(unsigned int pf,
+                 unsigned int hooknum,
+                 const struct sk_buff *skb,
+                 const struct net_device *in,
+                 const struct net_device *out,
+                 const struct nf_loginfo *li_user,
+                 const char *prefix)
+{
+       unsigned int size, data_len;
+       struct nfulnl_instance *inst;
+       const struct nf_loginfo *li;
+       unsigned int qthreshold;
+       unsigned int nlbufsiz;
+
+       if (li_user && li_user->type == NF_LOG_TYPE_ULOG) 
+               li = li_user;
+       else
+               li = &default_loginfo;
+
+       inst = instance_lookup_get(li->u.ulog.group);
+       if (!inst)
+               inst = instance_lookup_get(0);
+       if (!inst) {
+               PRINTR("nfnetlink_log: trying to log packet, "
+                       "but no instance for group %u\n", li->u.ulog.group);
+               return;
+       }
+
+       /* all macros expand to constant values at compile time */
+       /* FIXME: do we want to make the size calculation conditional based on
+        * what is actually present?  way more branches and checks, but more
+        * memory efficient... */
+       size =    NLMSG_SPACE(sizeof(struct nfgenmsg))
+               + NFA_SPACE(sizeof(struct nfulnl_msg_packet_hdr))
+               + NFA_SPACE(sizeof(u_int32_t))  /* ifindex */
+               + NFA_SPACE(sizeof(u_int32_t))  /* ifindex */
+               + NFA_SPACE(sizeof(u_int32_t))  /* mark */
+               + NFA_SPACE(sizeof(u_int32_t))  /* uid */
+               + NFA_SPACE(NFULNL_PREFIXLEN)   /* prefix */
+               + NFA_SPACE(sizeof(struct nfulnl_msg_packet_hw))
+               + NFA_SPACE(sizeof(struct nfulnl_msg_packet_timestamp));
+
+       UDEBUG("initial size=%u\n", size);
+
+       spin_lock_bh(&inst->lock);
+
+       qthreshold = inst->qthreshold;
+       /* per-rule qthreshold overrides per-instance */
+       if (qthreshold > li->u.ulog.qthreshold)
+               qthreshold = li->u.ulog.qthreshold;
+       
+       switch (inst->copy_mode) {
+       case NFULNL_COPY_META:
+       case NFULNL_COPY_NONE:
+               data_len = 0;
+               break;
+       
+       case NFULNL_COPY_PACKET:
+               if (inst->copy_range == 0 
+                   || inst->copy_range > skb->len)
+                       data_len = skb->len;
+               else
+                       data_len = inst->copy_range;
+               
+               size += NFA_SPACE(data_len);
+               UDEBUG("copy_packet, therefore size now %u\n", size);
+               break;
+       
+       default:
+               spin_unlock_bh(&inst->lock);
+               instance_put(inst);
+               return;
+       }
+
+       if (size > inst->nlbufsiz)
+               nlbufsiz = size;
+       else
+               nlbufsiz = inst->nlbufsiz;
+
+       if (!inst->skb) {
+               if (!(inst->skb = nfulnl_alloc_skb(nlbufsiz, size))) {
+                       UDEBUG("error in nfulnl_alloc_skb(%u, %u)\n",
+                               inst->nlbufsiz, size);
+                       goto alloc_failure;
+               }
+       } else if (inst->qlen >= qthreshold ||
+                  size > skb_tailroom(inst->skb)) {
+               /* either the queue len is too high or we don't have
+                * enough room in the skb left. flush to userspace. */
+               UDEBUG("flushing old skb\n");
+
+               __nfulnl_send(inst);
+
+               if (!(inst->skb = nfulnl_alloc_skb(nlbufsiz, size))) {
+                       UDEBUG("error in nfulnl_alloc_skb(%u, %u)\n",
+                               inst->nlbufsiz, size);
+                       goto alloc_failure;
+               }
+       }
+
+       UDEBUG("qlen %d, qthreshold %d\n", inst->qlen, qthreshold);
+       inst->qlen++;
+
+       __build_packet_message(inst, skb, data_len, pf,
+                               hooknum, in, out, li, prefix);
+
+       /* timer_pending always called within inst->lock, so there
+        * is no chance of a race here */
+       if (!timer_pending(&inst->timer)) {
+               instance_get(inst);
+               inst->timer.expires = jiffies + (inst->flushtimeout*HZ/100);
+               add_timer(&inst->timer);
+       }
+       spin_unlock_bh(&inst->lock);
+
+       return;
+
+alloc_failure:
+       spin_unlock_bh(&inst->lock);
+       instance_put(inst);
+       UDEBUG("error allocating skb\n");
+       /* FIXME: statistics */
+}
+
+static int
+nfulnl_rcv_nl_event(struct notifier_block *this,
+                  unsigned long event, void *ptr)
+{
+       struct netlink_notify *n = ptr;
+
+       if (event == NETLINK_URELEASE &&
+           n->protocol == NETLINK_NETFILTER && n->pid) {
+               int i;
+
+               /* destroy all instances for this pid */
+               write_lock_bh(&instances_lock);
+               for  (i = 0; i < INSTANCE_BUCKETS; i++) {
+                       struct hlist_node *tmp, *t2;
+                       struct nfulnl_instance *inst;
+                       struct hlist_head *head = &instance_table[i];
+
+                       hlist_for_each_entry_safe(inst, tmp, t2, head, hlist) {
+                               UDEBUG("node = %p\n", inst);
+                               if (n->pid == inst->peer_pid)
+                                       __instance_destroy(inst);
+                       }
+               }
+               write_unlock_bh(&instances_lock);
+       }
+       return NOTIFY_DONE;
+}
+
+static struct notifier_block nfulnl_rtnl_notifier = {
+       .notifier_call  = nfulnl_rcv_nl_event,
+};
+
+static int
+nfulnl_recv_unsupp(struct sock *ctnl, struct sk_buff *skb,
+                 struct nlmsghdr *nlh, struct nfattr *nfqa[], int *errp)
+{
+       return -ENOTSUPP;
+}
+
+static struct nf_logger nfulnl_logger = {
+       .name   = "nfnetlink_log",
+       .logfn  = &nfulnl_log_packet,
+       .me     = THIS_MODULE,
+};
+
+static const int nfula_min[NFULA_MAX] = {
+       [NFULA_PACKET_HDR-1]    = sizeof(struct nfulnl_msg_packet_hdr),
+       [NFULA_MARK-1]          = sizeof(u_int32_t),
+       [NFULA_TIMESTAMP-1]     = sizeof(struct nfulnl_msg_packet_timestamp),
+       [NFULA_IFINDEX_INDEV-1] = sizeof(u_int32_t),
+       [NFULA_IFINDEX_OUTDEV-1]= sizeof(u_int32_t),
+       [NFULA_HWADDR-1]        = sizeof(struct nfulnl_msg_packet_hw),
+       [NFULA_PAYLOAD-1]       = 0,
+       [NFULA_PREFIX-1]        = 0,
+       [NFULA_UID-1]           = sizeof(u_int32_t),
+};
+
+static const int nfula_cfg_min[NFULA_CFG_MAX] = {
+       [NFULA_CFG_CMD-1]       = sizeof(struct nfulnl_msg_config_cmd),
+       [NFULA_CFG_MODE-1]      = sizeof(struct nfulnl_msg_config_mode),
+       [NFULA_CFG_TIMEOUT-1]   = sizeof(u_int32_t),
+       [NFULA_CFG_QTHRESH-1]   = sizeof(u_int32_t),
+       [NFULA_CFG_NLBUFSIZ-1]  = sizeof(u_int32_t),
+};
+
+static int
+nfulnl_recv_config(struct sock *ctnl, struct sk_buff *skb,
+                  struct nlmsghdr *nlh, struct nfattr *nfula[], int *errp)
+{
+       struct nfgenmsg *nfmsg = NLMSG_DATA(nlh);
+       u_int16_t group_num = ntohs(nfmsg->res_id);
+       struct nfulnl_instance *inst;
+       int ret = 0;
+
+       UDEBUG("entering for msg %u\n", NFNL_MSG_TYPE(nlh->nlmsg_type));
+
+       if (nfattr_bad_size(nfula, NFULA_CFG_MAX, nfula_cfg_min)) {
+               UDEBUG("bad attribute size\n");
+               return -EINVAL;
+       }
+
+       inst = instance_lookup_get(group_num);
+       if (nfula[NFULA_CFG_CMD-1]) {
+               u_int8_t pf = nfmsg->nfgen_family;
+               struct nfulnl_msg_config_cmd *cmd;
+               cmd = NFA_DATA(nfula[NFULA_CFG_CMD-1]);
+               UDEBUG("found CFG_CMD for\n");
+
+               switch (cmd->command) {
+               case NFULNL_CFG_CMD_BIND:
+                       if (inst) {
+                               ret = -EBUSY;
+                               goto out_put;
+                       }
+
+                       inst = instance_create(group_num,
+                                              NETLINK_CB(skb).pid);
+                       if (!inst) {
+                               ret = -EINVAL;
+                               goto out_put;
+                       }
+                       break;
+               case NFULNL_CFG_CMD_UNBIND:
+                       if (!inst) {
+                               ret = -ENODEV;
+                               goto out_put;
+                       }
+
+                       if (inst->peer_pid != NETLINK_CB(skb).pid) {
+                               ret = -EPERM;
+                               goto out_put;
+                       }
+
+                       instance_destroy(inst);
+                       break;
+               case NFULNL_CFG_CMD_PF_BIND:
+                       UDEBUG("registering log handler for pf=%u\n", pf);
+                       ret = nf_log_register(pf, &nfulnl_logger);
+                       break;
+               case NFULNL_CFG_CMD_PF_UNBIND:
+                       UDEBUG("unregistering log handler for pf=%u\n", pf);
+                       /* This is a bug and a feature.  We cannot unregister
+                        * other handlers, like nfnetlink_inst can */
+                       nf_log_unregister_pf(pf);
+                       break;
+               default:
+                       ret = -EINVAL;
+                       break;
+               }
+       } else {
+               if (!inst) {
+                       UDEBUG("no config command, and no instance for "
+                               "group=%u pid=%u =>ENOENT\n",
+                               group_num, NETLINK_CB(skb).pid);
+                       ret = -ENOENT;
+                       goto out_put;
+               }
+
+               if (inst->peer_pid != NETLINK_CB(skb).pid) {
+                       UDEBUG("no config command, and wrong pid\n");
+                       ret = -EPERM;
+                       goto out_put;
+               }
+       }
+
+       if (nfula[NFULA_CFG_MODE-1]) {
+               struct nfulnl_msg_config_mode *params;
+               params = NFA_DATA(nfula[NFULA_CFG_MODE-1]);
+
+               nfulnl_set_mode(inst, params->copy_mode,
+                               ntohs(params->copy_range));
+       }
+
+       if (nfula[NFULA_CFG_TIMEOUT-1]) {
+               u_int32_t timeout = 
+                       *(u_int32_t *)NFA_DATA(nfula[NFULA_CFG_TIMEOUT-1]);
+
+               nfulnl_set_timeout(inst, ntohl(timeout));
+       }
+
+       if (nfula[NFULA_CFG_NLBUFSIZ-1]) {
+               u_int32_t nlbufsiz = 
+                       *(u_int32_t *)NFA_DATA(nfula[NFULA_CFG_NLBUFSIZ-1]);
+
+               nfulnl_set_nlbufsiz(inst, ntohl(nlbufsiz));
+       }
+
+       if (nfula[NFULA_CFG_QTHRESH-1]) {
+               u_int32_t qthresh = 
+                       *(u_int16_t *)NFA_DATA(nfula[NFULA_CFG_QTHRESH-1]);
+
+               nfulnl_set_qthresh(inst, ntohl(qthresh));
+       }
+
+out_put:
+       instance_put(inst);
+       return ret;
+}
+
+static struct nfnl_callback nfulnl_cb[NFULNL_MSG_MAX] = {
+       [NFULNL_MSG_PACKET]     = { .call = nfulnl_recv_unsupp,
+                                   .cap_required = CAP_NET_ADMIN },
+       [NFULNL_MSG_CONFIG]     = { .call = nfulnl_recv_config,
+                                   .cap_required = CAP_NET_ADMIN },
+};
+
+static struct nfnetlink_subsystem nfulnl_subsys = {
+       .name           = "log",
+       .subsys_id      = NFNL_SUBSYS_ULOG,
+       .cb_count       = NFULNL_MSG_MAX,
+       .attr_count     = NFULA_MAX,
+       .cb             = nfulnl_cb,
+};
+
+#ifdef CONFIG_PROC_FS
+struct iter_state {
+       unsigned int bucket;
+};
+
+static struct hlist_node *get_first(struct seq_file *seq)
+{
+       struct iter_state *st = seq->private;
+
+       if (!st)
+               return NULL;
+
+       for (st->bucket = 0; st->bucket < INSTANCE_BUCKETS; st->bucket++) {
+               if (!hlist_empty(&instance_table[st->bucket]))
+                       return instance_table[st->bucket].first;
+       }
+       return NULL;
+}
+
+static struct hlist_node *get_next(struct seq_file *seq, struct hlist_node *h)
+{
+       struct iter_state *st = seq->private;
+
+       h = h->next;
+       while (!h) {
+               if (++st->bucket >= INSTANCE_BUCKETS)
+                       return NULL;
+
+               h = instance_table[st->bucket].first;
+       }
+       return h;
+}
+
+static struct hlist_node *get_idx(struct seq_file *seq, loff_t pos)
+{
+       struct hlist_node *head;
+       head = get_first(seq);
+
+       if (head)
+               while (pos && (head = get_next(seq, head)))
+                       pos--;
+       return pos ? NULL : head;
+}
+
+static void *seq_start(struct seq_file *seq, loff_t *pos)
+{
+       read_lock_bh(&instances_lock);
+       return get_idx(seq, *pos);
+}
+
+static void *seq_next(struct seq_file *s, void *v, loff_t *pos)
+{
+       (*pos)++;
+       return get_next(s, v);
+}
+
+static void seq_stop(struct seq_file *s, void *v)
+{
+       read_unlock_bh(&instances_lock);
+}
+
+static int seq_show(struct seq_file *s, void *v)
+{
+       const struct nfulnl_instance *inst = v;
+
+       return seq_printf(s, "%5d %6d %5d %1d %5d %6d %2d\n", 
+                         inst->group_num,
+                         inst->peer_pid, inst->qlen, 
+                         inst->copy_mode, inst->copy_range,
+                         inst->flushtimeout, atomic_read(&inst->use));
+}
+
+static struct seq_operations nful_seq_ops = {
+       .start  = seq_start,
+       .next   = seq_next,
+       .stop   = seq_stop,
+       .show   = seq_show,
+};
+
+static int nful_open(struct inode *inode, struct file *file)
+{
+       struct seq_file *seq;
+       struct iter_state *is;
+       int ret;
+
+       is = kmalloc(sizeof(*is), GFP_KERNEL);
+       if (!is)
+               return -ENOMEM;
+       memset(is, 0, sizeof(*is));
+       ret = seq_open(file, &nful_seq_ops);
+       if (ret < 0)
+               goto out_free;
+       seq = file->private_data;
+       seq->private = is;
+       return ret;
+out_free:
+       kfree(is);
+       return ret;
+}
+
+static struct file_operations nful_file_ops = {
+       .owner   = THIS_MODULE,
+       .open    = nful_open,
+       .read    = seq_read,
+       .llseek  = seq_lseek,
+       .release = seq_release_private,
+};
+
+#endif /* PROC_FS */
+
+static int
+init_or_cleanup(int init)
+{
+       int i, status = -ENOMEM;
+#ifdef CONFIG_PROC_FS
+       struct proc_dir_entry *proc_nful;
+#endif
+       
+       if (!init)
+               goto cleanup;
+
+       for (i = 0; i < INSTANCE_BUCKETS; i++)
+               INIT_HLIST_HEAD(&instance_table[i]);
+       
+       /* it's not really all that important to have a random value, so
+        * we can do this from the init function, even if there hasn't
+        * been that much entropy yet */
+       get_random_bytes(&hash_init, sizeof(hash_init));
+
+       netlink_register_notifier(&nfulnl_rtnl_notifier);
+       status = nfnetlink_subsys_register(&nfulnl_subsys);
+       if (status < 0) {
+               printk(KERN_ERR "log: failed to create netlink socket\n");
+               goto cleanup_netlink_notifier;
+       }
+
+#ifdef CONFIG_PROC_FS
+       proc_nful = create_proc_entry("nfnetlink_log", 0440,
+                                     proc_net_netfilter);
+       if (!proc_nful)
+               goto cleanup_subsys;
+       proc_nful->proc_fops = &nful_file_ops;
+#endif
+
+       return status;
+
+cleanup:
+       nf_log_unregister_logger(&nfulnl_logger);
+#ifdef CONFIG_PROC_FS
+       remove_proc_entry("nfnetlink_log", proc_net_netfilter);
+cleanup_subsys:
+#endif
+       nfnetlink_subsys_unregister(&nfulnl_subsys);
+cleanup_netlink_notifier:
+       netlink_unregister_notifier(&nfulnl_rtnl_notifier);
+       return status;
+}
+
+static int __init init(void)
+{
+       
+       return init_or_cleanup(1);
+}
+
+static void __exit fini(void)
+{
+       init_or_cleanup(0);
+}
+
+MODULE_DESCRIPTION("netfilter userspace logging");
+MODULE_AUTHOR("Harald Welte <laforge@netfilter.org>");
+MODULE_LICENSE("GPL");
+
+module_init(init);
+module_exit(fini);
index eab309e..d7b0330 100644 (file)
@@ -1041,6 +1041,7 @@ cleanup:
        nf_unregister_queue_handlers(nfqnl_enqueue_packet);
        unregister_netdevice_notifier(&nfqnl_dev_notifier);
 #ifdef CONFIG_PROC_FS
+       remove_proc_entry("nfnetlink_queue", proc_net_netfilter);
 cleanup_subsys:
 #endif 
        nfnetlink_subsys_unregister(&nfqnl_subsys);