kernel-ark/kernel/audit.c

858 lines
23 KiB
C
Raw Normal View History

/* audit.c -- Auditing support
* Gateway between the kernel (e.g., selinux) and the user-space audit daemon.
* System-call specific features have moved to auditsc.c
*
* Copyright 2003-2004 Red Hat Inc., Durham, North Carolina.
* All Rights Reserved.
*
* 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.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
*
* Written by Rickard E. (Rik) Faith <faith@redhat.com>
*
* Goals: 1) Integrate fully with SELinux.
* 2) Minimal run-time overhead:
* a) Minimal when syscall auditing is disabled (audit_enable=0).
* b) Small when syscall auditing is enabled and no audit record
* is generated (defer as much work as possible to record
* generation time):
* i) context is allocated,
* ii) names from getname are stored without a copy, and
* iii) inode information stored from path_lookup.
* 3) Ability to disable syscall auditing at boot time (audit=0).
* 4) Usable by other parts of the kernel (if audit_log* is called,
* then a syscall record will be generated automatically for the
* current syscall).
* 5) Netlink interface to user-space.
* 6) Support low-overhead kernel-based filtering to minimize the
* information that must be passed to user-space.
*
* Example user-space utilities: http://people.redhat.com/sgrubb/audit/
*/
#include <linux/init.h>
#include <asm/atomic.h>
#include <asm/types.h>
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/audit.h>
#include <net/sock.h>
#include <linux/skbuff.h>
#include <linux/netlink.h>
/* No auditing will take place until audit_initialized != 0.
* (Initialization happens after skb_init is called.) */
static int audit_initialized;
/* No syscall auditing will take place unless audit_enabled != 0. */
int audit_enabled;
/* Default state when kernel boots without any parameters. */
static int audit_default;
/* If auditing cannot proceed, audit_failure selects what happens. */
static int audit_failure = AUDIT_FAIL_PRINTK;
/* If audit records are to be written to the netlink socket, audit_pid
* contains the (non-zero) pid. */
int audit_pid;
/* If audit_limit is non-zero, limit the rate of sending audit records
* to that number per second. This prevents DoS attacks, but results in
* audit records being dropped. */
static int audit_rate_limit;
/* Number of outstanding audit_buffers allowed. */
static int audit_backlog_limit = 64;
static atomic_t audit_backlog = ATOMIC_INIT(0);
/* The identity of the user shutting down the audit system. */
uid_t audit_sig_uid = -1;
pid_t audit_sig_pid = -1;
/* Records can be lost in several ways:
0) [suppressed in audit_alloc]
1) out of memory in audit_log_start [kmalloc of struct audit_buffer]
2) out of memory in audit_log_move [alloc_skb]
3) suppressed due to audit_rate_limit
4) suppressed due to audit_backlog_limit
*/
static atomic_t audit_lost = ATOMIC_INIT(0);
/* The netlink socket. */
static struct sock *audit_sock;
/* There are two lists of audit buffers. The txlist contains audit
* buffers that cannot be sent immediately to the netlink device because
* we are in an irq context (these are sent later in a tasklet).
*
* The second list is a list of pre-allocated audit buffers (if more
* than AUDIT_MAXFREE are in use, the audit buffer is freed instead of
* being placed on the freelist). */
static DEFINE_SPINLOCK(audit_txlist_lock);
static DEFINE_SPINLOCK(audit_freelist_lock);
static int audit_freelist_count = 0;
static LIST_HEAD(audit_txlist);
static LIST_HEAD(audit_freelist);
/* There are three lists of rules -- one to search at task creation
* time, one to search at syscall entry time, and another to search at
* syscall exit time. */
static LIST_HEAD(audit_tsklist);
static LIST_HEAD(audit_entlist);
static LIST_HEAD(audit_extlist);
/* The netlink socket is only to be read by 1 CPU, which lets us assume
* that list additions and deletions never happen simultaneously in
* auditsc.c */
static DECLARE_MUTEX(audit_netlink_sem);
/* AUDIT_BUFSIZ is the size of the temporary buffer used for formatting
* audit records. Since printk uses a 1024 byte buffer, this buffer
* should be at least that large. */
#define AUDIT_BUFSIZ 1024
/* AUDIT_MAXFREE is the number of empty audit_buffers we keep on the
* audit_freelist. Doing so eliminates many kmalloc/kfree calls. */
#define AUDIT_MAXFREE (2*NR_CPUS)
/* The audit_buffer is used when formatting an audit record. The caller
* locks briefly to get the record off the freelist or to allocate the
* buffer, and locks briefly to send the buffer to the netlink layer or
* to place it on a transmit queue. Multiple audit_buffers can be in
* use simultaneously. */
struct audit_buffer {
struct list_head list;
struct sk_buff *skb; /* formatted skb ready to send */
struct audit_context *ctx; /* NULL or associated context */
};
static void audit_set_pid(struct audit_buffer *ab, pid_t pid)
{
struct nlmsghdr *nlh = (struct nlmsghdr *)ab->skb->data;
nlh->nlmsg_pid = pid;
}
struct audit_entry {
struct list_head list;
struct audit_rule rule;
};
static void audit_log_end_irq(struct audit_buffer *ab);
static void audit_log_end_fast(struct audit_buffer *ab);
static void audit_panic(const char *message)
{
switch (audit_failure)
{
case AUDIT_FAIL_SILENT:
break;
case AUDIT_FAIL_PRINTK:
printk(KERN_ERR "audit: %s\n", message);
break;
case AUDIT_FAIL_PANIC:
panic("audit: %s\n", message);
break;
}
}
static inline int audit_rate_check(void)
{
static unsigned long last_check = 0;
static int messages = 0;
static DEFINE_SPINLOCK(lock);
unsigned long flags;
unsigned long now;
unsigned long elapsed;
int retval = 0;
if (!audit_rate_limit) return 1;
spin_lock_irqsave(&lock, flags);
if (++messages < audit_rate_limit) {
retval = 1;
} else {
now = jiffies;
elapsed = now - last_check;
if (elapsed > HZ) {
last_check = now;
messages = 0;
retval = 1;
}
}
spin_unlock_irqrestore(&lock, flags);
return retval;
}
/* Emit at least 1 message per second, even if audit_rate_check is
* throttling. */
void audit_log_lost(const char *message)
{
static unsigned long last_msg = 0;
static DEFINE_SPINLOCK(lock);
unsigned long flags;
unsigned long now;
int print;
atomic_inc(&audit_lost);
print = (audit_failure == AUDIT_FAIL_PANIC || !audit_rate_limit);
if (!print) {
spin_lock_irqsave(&lock, flags);
now = jiffies;
if (now - last_msg > HZ) {
print = 1;
last_msg = now;
}
spin_unlock_irqrestore(&lock, flags);
}
if (print) {
printk(KERN_WARNING
"audit: audit_lost=%d audit_backlog=%d"
" audit_rate_limit=%d audit_backlog_limit=%d\n",
atomic_read(&audit_lost),
atomic_read(&audit_backlog),
audit_rate_limit,
audit_backlog_limit);
audit_panic(message);
}
}
static int audit_set_rate_limit(int limit, uid_t loginuid)
{
int old = audit_rate_limit;
audit_rate_limit = limit;
audit_log(NULL, AUDIT_CONFIG_CHANGE,
"audit_rate_limit=%d old=%d by auid %u",
audit_rate_limit, old, loginuid);
return old;
}
static int audit_set_backlog_limit(int limit, uid_t loginuid)
{
int old = audit_backlog_limit;
audit_backlog_limit = limit;
audit_log(NULL, AUDIT_CONFIG_CHANGE,
"audit_backlog_limit=%d old=%d by auid %u",
audit_backlog_limit, old, loginuid);
return old;
}
static int audit_set_enabled(int state, uid_t loginuid)
{
int old = audit_enabled;
if (state != 0 && state != 1)
return -EINVAL;
audit_enabled = state;
audit_log(NULL, AUDIT_CONFIG_CHANGE,
"audit_enabled=%d old=%d by auid %u",
audit_enabled, old, loginuid);
return old;
}
static int audit_set_failure(int state, uid_t loginuid)
{
int old = audit_failure;
if (state != AUDIT_FAIL_SILENT
&& state != AUDIT_FAIL_PRINTK
&& state != AUDIT_FAIL_PANIC)
return -EINVAL;
audit_failure = state;
audit_log(NULL, AUDIT_CONFIG_CHANGE,
"audit_failure=%d old=%d by auid %u",
audit_failure, old, loginuid);
return old;
}
void audit_send_reply(int pid, int seq, int type, int done, int multi,
void *payload, int size)
{
struct sk_buff *skb;
struct nlmsghdr *nlh;
int len = NLMSG_SPACE(size);
void *data;
int flags = multi ? NLM_F_MULTI : 0;
int t = done ? NLMSG_DONE : type;
skb = alloc_skb(len, GFP_KERNEL);
if (!skb)
goto nlmsg_failure;
nlh = NLMSG_PUT(skb, pid, seq, t, len - sizeof(*nlh));
nlh->nlmsg_flags = flags;
data = NLMSG_DATA(nlh);
memcpy(data, payload, size);
netlink_unicast(audit_sock, skb, pid, MSG_DONTWAIT);
return;
nlmsg_failure: /* Used by NLMSG_PUT */
if (skb)
kfree_skb(skb);
}
/*
* Check for appropriate CAP_AUDIT_ capabilities on incoming audit
* control messages.
*/
static int audit_netlink_ok(kernel_cap_t eff_cap, u16 msg_type)
{
int err = 0;
switch (msg_type) {
case AUDIT_GET:
case AUDIT_LIST:
case AUDIT_SET:
case AUDIT_ADD:
case AUDIT_DEL:
case AUDIT_SIGNAL_INFO:
if (!cap_raised(eff_cap, CAP_AUDIT_CONTROL))
err = -EPERM;
break;
case AUDIT_USER:
case AUDIT_USER_AUTH:
case AUDIT_USER_ACCT:
case AUDIT_USER_MGMT:
case AUDIT_CRED_ACQ:
case AUDIT_CRED_DISP:
case AUDIT_USER_START:
case AUDIT_USER_END:
case AUDIT_USER_AVC:
if (!cap_raised(eff_cap, CAP_AUDIT_WRITE))
err = -EPERM;
break;
default: /* bad msg */
err = -EINVAL;
}
return err;
}
static int audit_receive_msg(struct sk_buff *skb, struct nlmsghdr *nlh)
{
u32 uid, pid, seq;
void *data;
struct audit_status *status_get, status_set;
int err;
struct audit_buffer *ab;
u16 msg_type = nlh->nlmsg_type;
uid_t loginuid; /* loginuid of sender */
struct audit_sig_info sig_data;
err = audit_netlink_ok(NETLINK_CB(skb).eff_cap, msg_type);
if (err)
return err;
pid = NETLINK_CREDS(skb)->pid;
uid = NETLINK_CREDS(skb)->uid;
loginuid = NETLINK_CB(skb).loginuid;
seq = nlh->nlmsg_seq;
data = NLMSG_DATA(nlh);
switch (msg_type) {
case AUDIT_GET:
status_set.enabled = audit_enabled;
status_set.failure = audit_failure;
status_set.pid = audit_pid;
status_set.rate_limit = audit_rate_limit;
status_set.backlog_limit = audit_backlog_limit;
status_set.lost = atomic_read(&audit_lost);
status_set.backlog = atomic_read(&audit_backlog);
audit_send_reply(NETLINK_CB(skb).pid, seq, AUDIT_GET, 0, 0,
&status_set, sizeof(status_set));
break;
case AUDIT_SET:
if (nlh->nlmsg_len < sizeof(struct audit_status))
return -EINVAL;
status_get = (struct audit_status *)data;
if (status_get->mask & AUDIT_STATUS_ENABLED) {
err = audit_set_enabled(status_get->enabled, loginuid);
if (err < 0) return err;
}
if (status_get->mask & AUDIT_STATUS_FAILURE) {
err = audit_set_failure(status_get->failure, loginuid);
if (err < 0) return err;
}
if (status_get->mask & AUDIT_STATUS_PID) {
int old = audit_pid;
audit_pid = status_get->pid;
audit_log(NULL, AUDIT_CONFIG_CHANGE,
"audit_pid=%d old=%d by auid %u",
audit_pid, old, loginuid);
}
if (status_get->mask & AUDIT_STATUS_RATE_LIMIT)
audit_set_rate_limit(status_get->rate_limit, loginuid);
if (status_get->mask & AUDIT_STATUS_BACKLOG_LIMIT)
audit_set_backlog_limit(status_get->backlog_limit,
loginuid);
break;
case AUDIT_USER:
case AUDIT_USER_AUTH:
case AUDIT_USER_ACCT:
case AUDIT_USER_MGMT:
case AUDIT_CRED_ACQ:
case AUDIT_CRED_DISP:
case AUDIT_USER_START:
case AUDIT_USER_END:
case AUDIT_USER_AVC:
ab = audit_log_start(NULL, msg_type);
if (!ab)
break; /* audit_panic has been called */
audit_log_format(ab,
"user pid=%d uid=%d length=%d loginuid=%u"
" msg='%.1024s'",
pid, uid,
(int)(nlh->nlmsg_len
- ((char *)data - (char *)nlh)),
loginuid, (char *)data);
audit_set_pid(ab, pid);
audit_log_end(ab);
break;
case AUDIT_ADD:
case AUDIT_DEL:
if (nlh->nlmsg_len < sizeof(struct audit_rule))
return -EINVAL;
/* fallthrough */
case AUDIT_LIST:
err = audit_receive_filter(nlh->nlmsg_type, NETLINK_CB(skb).pid,
uid, seq, data, loginuid);
break;
case AUDIT_SIGNAL_INFO:
sig_data.uid = audit_sig_uid;
sig_data.pid = audit_sig_pid;
audit_send_reply(NETLINK_CB(skb).pid, seq, AUDIT_SIGNAL_INFO,
0, 0, &sig_data, sizeof(sig_data));
break;
default:
err = -EINVAL;
break;
}
return err < 0 ? err : 0;
}
/* Get message from skb (based on rtnetlink_rcv_skb). Each message is
* processed by audit_receive_msg. Malformed skbs with wrong length are
* discarded silently. */
[NETLINK]: Synchronous message processing. Let's recap the problem. The current asynchronous netlink kernel message processing is vulnerable to these attacks: 1) Hit and run: Attacker sends one or more messages and then exits before they're processed. This may confuse/disable the next netlink user that gets the netlink address of the attacker since it may receive the responses to the attacker's messages. Proposed solutions: a) Synchronous processing. b) Stream mode socket. c) Restrict/prohibit binding. 2) Starvation: Because various netlink rcv functions were written to not return until all messages have been processed on a socket, it is possible for these functions to execute for an arbitrarily long period of time. If this is successfully exploited it could also be used to hold rtnl forever. Proposed solutions: a) Synchronous processing. b) Stream mode socket. Firstly let's cross off solution c). It only solves the first problem and it has user-visible impacts. In particular, it'll break user space applications that expect to bind or communicate with specific netlink addresses (pid's). So we're left with a choice of synchronous processing versus SOCK_STREAM for netlink. For the moment I'm sticking with the synchronous approach as suggested by Alexey since it's simpler and I'd rather spend my time working on other things. However, it does have a number of deficiencies compared to the stream mode solution: 1) User-space to user-space netlink communication is still vulnerable. 2) Inefficient use of resources. This is especially true for rtnetlink since the lock is shared with other users such as networking drivers. The latter could hold the rtnl while communicating with hardware which causes the rtnetlink user to wait when it could be doing other things. 3) It is still possible to DoS all netlink users by flooding the kernel netlink receive queue. The attacker simply fills the receive socket with a single netlink message that fills up the entire queue. The attacker then continues to call sendmsg with the same message in a loop. Point 3) can be countered by retransmissions in user-space code, however it is pretty messy. In light of these problems (in particular, point 3), we should implement stream mode netlink at some point. In the mean time, here is a patch that implements synchronous processing. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2005-05-03 21:55:09 +00:00
static void audit_receive_skb(struct sk_buff *skb)
{
int err;
struct nlmsghdr *nlh;
u32 rlen;
while (skb->len >= NLMSG_SPACE(0)) {
nlh = (struct nlmsghdr *)skb->data;
if (nlh->nlmsg_len < sizeof(*nlh) || skb->len < nlh->nlmsg_len)
[NETLINK]: Synchronous message processing. Let's recap the problem. The current asynchronous netlink kernel message processing is vulnerable to these attacks: 1) Hit and run: Attacker sends one or more messages and then exits before they're processed. This may confuse/disable the next netlink user that gets the netlink address of the attacker since it may receive the responses to the attacker's messages. Proposed solutions: a) Synchronous processing. b) Stream mode socket. c) Restrict/prohibit binding. 2) Starvation: Because various netlink rcv functions were written to not return until all messages have been processed on a socket, it is possible for these functions to execute for an arbitrarily long period of time. If this is successfully exploited it could also be used to hold rtnl forever. Proposed solutions: a) Synchronous processing. b) Stream mode socket. Firstly let's cross off solution c). It only solves the first problem and it has user-visible impacts. In particular, it'll break user space applications that expect to bind or communicate with specific netlink addresses (pid's). So we're left with a choice of synchronous processing versus SOCK_STREAM for netlink. For the moment I'm sticking with the synchronous approach as suggested by Alexey since it's simpler and I'd rather spend my time working on other things. However, it does have a number of deficiencies compared to the stream mode solution: 1) User-space to user-space netlink communication is still vulnerable. 2) Inefficient use of resources. This is especially true for rtnetlink since the lock is shared with other users such as networking drivers. The latter could hold the rtnl while communicating with hardware which causes the rtnetlink user to wait when it could be doing other things. 3) It is still possible to DoS all netlink users by flooding the kernel netlink receive queue. The attacker simply fills the receive socket with a single netlink message that fills up the entire queue. The attacker then continues to call sendmsg with the same message in a loop. Point 3) can be countered by retransmissions in user-space code, however it is pretty messy. In light of these problems (in particular, point 3), we should implement stream mode netlink at some point. In the mean time, here is a patch that implements synchronous processing. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2005-05-03 21:55:09 +00:00
return;
rlen = NLMSG_ALIGN(nlh->nlmsg_len);
if (rlen > skb->len)
rlen = skb->len;
if ((err = audit_receive_msg(skb, nlh))) {
netlink_ack(skb, nlh, err);
} else if (nlh->nlmsg_flags & NLM_F_ACK)
netlink_ack(skb, nlh, 0);
skb_pull(skb, rlen);
}
}
/* Receive messages from netlink socket. */
static void audit_receive(struct sock *sk, int length)
{
struct sk_buff *skb;
[NETLINK]: Synchronous message processing. Let's recap the problem. The current asynchronous netlink kernel message processing is vulnerable to these attacks: 1) Hit and run: Attacker sends one or more messages and then exits before they're processed. This may confuse/disable the next netlink user that gets the netlink address of the attacker since it may receive the responses to the attacker's messages. Proposed solutions: a) Synchronous processing. b) Stream mode socket. c) Restrict/prohibit binding. 2) Starvation: Because various netlink rcv functions were written to not return until all messages have been processed on a socket, it is possible for these functions to execute for an arbitrarily long period of time. If this is successfully exploited it could also be used to hold rtnl forever. Proposed solutions: a) Synchronous processing. b) Stream mode socket. Firstly let's cross off solution c). It only solves the first problem and it has user-visible impacts. In particular, it'll break user space applications that expect to bind or communicate with specific netlink addresses (pid's). So we're left with a choice of synchronous processing versus SOCK_STREAM for netlink. For the moment I'm sticking with the synchronous approach as suggested by Alexey since it's simpler and I'd rather spend my time working on other things. However, it does have a number of deficiencies compared to the stream mode solution: 1) User-space to user-space netlink communication is still vulnerable. 2) Inefficient use of resources. This is especially true for rtnetlink since the lock is shared with other users such as networking drivers. The latter could hold the rtnl while communicating with hardware which causes the rtnetlink user to wait when it could be doing other things. 3) It is still possible to DoS all netlink users by flooding the kernel netlink receive queue. The attacker simply fills the receive socket with a single netlink message that fills up the entire queue. The attacker then continues to call sendmsg with the same message in a loop. Point 3) can be countered by retransmissions in user-space code, however it is pretty messy. In light of these problems (in particular, point 3), we should implement stream mode netlink at some point. In the mean time, here is a patch that implements synchronous processing. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2005-05-03 21:55:09 +00:00
unsigned int qlen;
[NETLINK]: Synchronous message processing. Let's recap the problem. The current asynchronous netlink kernel message processing is vulnerable to these attacks: 1) Hit and run: Attacker sends one or more messages and then exits before they're processed. This may confuse/disable the next netlink user that gets the netlink address of the attacker since it may receive the responses to the attacker's messages. Proposed solutions: a) Synchronous processing. b) Stream mode socket. c) Restrict/prohibit binding. 2) Starvation: Because various netlink rcv functions were written to not return until all messages have been processed on a socket, it is possible for these functions to execute for an arbitrarily long period of time. If this is successfully exploited it could also be used to hold rtnl forever. Proposed solutions: a) Synchronous processing. b) Stream mode socket. Firstly let's cross off solution c). It only solves the first problem and it has user-visible impacts. In particular, it'll break user space applications that expect to bind or communicate with specific netlink addresses (pid's). So we're left with a choice of synchronous processing versus SOCK_STREAM for netlink. For the moment I'm sticking with the synchronous approach as suggested by Alexey since it's simpler and I'd rather spend my time working on other things. However, it does have a number of deficiencies compared to the stream mode solution: 1) User-space to user-space netlink communication is still vulnerable. 2) Inefficient use of resources. This is especially true for rtnetlink since the lock is shared with other users such as networking drivers. The latter could hold the rtnl while communicating with hardware which causes the rtnetlink user to wait when it could be doing other things. 3) It is still possible to DoS all netlink users by flooding the kernel netlink receive queue. The attacker simply fills the receive socket with a single netlink message that fills up the entire queue. The attacker then continues to call sendmsg with the same message in a loop. Point 3) can be countered by retransmissions in user-space code, however it is pretty messy. In light of these problems (in particular, point 3), we should implement stream mode netlink at some point. In the mean time, here is a patch that implements synchronous processing. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2005-05-03 21:55:09 +00:00
down(&audit_netlink_sem);
[NETLINK]: Synchronous message processing. Let's recap the problem. The current asynchronous netlink kernel message processing is vulnerable to these attacks: 1) Hit and run: Attacker sends one or more messages and then exits before they're processed. This may confuse/disable the next netlink user that gets the netlink address of the attacker since it may receive the responses to the attacker's messages. Proposed solutions: a) Synchronous processing. b) Stream mode socket. c) Restrict/prohibit binding. 2) Starvation: Because various netlink rcv functions were written to not return until all messages have been processed on a socket, it is possible for these functions to execute for an arbitrarily long period of time. If this is successfully exploited it could also be used to hold rtnl forever. Proposed solutions: a) Synchronous processing. b) Stream mode socket. Firstly let's cross off solution c). It only solves the first problem and it has user-visible impacts. In particular, it'll break user space applications that expect to bind or communicate with specific netlink addresses (pid's). So we're left with a choice of synchronous processing versus SOCK_STREAM for netlink. For the moment I'm sticking with the synchronous approach as suggested by Alexey since it's simpler and I'd rather spend my time working on other things. However, it does have a number of deficiencies compared to the stream mode solution: 1) User-space to user-space netlink communication is still vulnerable. 2) Inefficient use of resources. This is especially true for rtnetlink since the lock is shared with other users such as networking drivers. The latter could hold the rtnl while communicating with hardware which causes the rtnetlink user to wait when it could be doing other things. 3) It is still possible to DoS all netlink users by flooding the kernel netlink receive queue. The attacker simply fills the receive socket with a single netlink message that fills up the entire queue. The attacker then continues to call sendmsg with the same message in a loop. Point 3) can be countered by retransmissions in user-space code, however it is pretty messy. In light of these problems (in particular, point 3), we should implement stream mode netlink at some point. In the mean time, here is a patch that implements synchronous processing. Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au> Signed-off-by: David S. Miller <davem@davemloft.net>
2005-05-03 21:55:09 +00:00
for (qlen = skb_queue_len(&sk->sk_receive_queue); qlen; qlen--) {
skb = skb_dequeue(&sk->sk_receive_queue);
audit_receive_skb(skb);
kfree_skb(skb);
}
up(&audit_netlink_sem);
}
/* Grab skbuff from the audit_buffer and send to user space. */
static inline int audit_log_drain(struct audit_buffer *ab)
{
struct sk_buff *skb = ab->skb;
if (skb) {
int retval = 0;
if (audit_pid) {
struct nlmsghdr *nlh = (struct nlmsghdr *)skb->data;
nlh->nlmsg_len = skb->len - NLMSG_SPACE(0);
skb_get(skb); /* because netlink_* frees */
retval = netlink_unicast(audit_sock, skb, audit_pid,
MSG_DONTWAIT);
}
if (retval == -EAGAIN &&
(atomic_read(&audit_backlog)) < audit_backlog_limit) {
audit_log_end_irq(ab);
return 1;
}
if (retval < 0) {
if (retval == -ECONNREFUSED) {
printk(KERN_ERR
"audit: *NO* daemon at audit_pid=%d\n",
audit_pid);
audit_pid = 0;
} else
audit_log_lost("netlink socket too busy");
}
if (!audit_pid) { /* No daemon */
int offset = NLMSG_SPACE(0);
int len = skb->len - offset;
skb->data[offset + len] = '\0';
printk(KERN_ERR "%s\n", skb->data + offset);
}
}
return 0;
}
/* Initialize audit support at boot time. */
static int __init audit_init(void)
{
printk(KERN_INFO "audit: initializing netlink socket (%s)\n",
audit_default ? "enabled" : "disabled");
audit_sock = netlink_kernel_create(NETLINK_AUDIT, audit_receive);
if (!audit_sock)
audit_panic("cannot initialize netlink socket");
audit_initialized = 1;
audit_enabled = audit_default;
audit_log(NULL, AUDIT_KERNEL, "initialized");
return 0;
}
__initcall(audit_init);
/* Process kernel command-line parameter at boot time. audit=0 or audit=1. */
static int __init audit_enable(char *str)
{
audit_default = !!simple_strtol(str, NULL, 0);
printk(KERN_INFO "audit: %s%s\n",
audit_default ? "enabled" : "disabled",
audit_initialized ? "" : " (after initialization)");
if (audit_initialized)
audit_enabled = audit_default;
return 0;
}
__setup("audit=", audit_enable);
static void audit_buffer_free(struct audit_buffer *ab)
{
unsigned long flags;
if (!ab)
return;
if (ab->skb)
kfree_skb(ab->skb);
atomic_dec(&audit_backlog);
spin_lock_irqsave(&audit_freelist_lock, flags);
if (++audit_freelist_count > AUDIT_MAXFREE)
kfree(ab);
else
list_add(&ab->list, &audit_freelist);
spin_unlock_irqrestore(&audit_freelist_lock, flags);
}
static struct audit_buffer * audit_buffer_alloc(struct audit_context *ctx,
int gfp_mask, int type)
{
unsigned long flags;
struct audit_buffer *ab = NULL;
struct nlmsghdr *nlh;
spin_lock_irqsave(&audit_freelist_lock, flags);
if (!list_empty(&audit_freelist)) {
ab = list_entry(audit_freelist.next,
struct audit_buffer, list);
list_del(&ab->list);
--audit_freelist_count;
}
spin_unlock_irqrestore(&audit_freelist_lock, flags);
if (!ab) {
ab = kmalloc(sizeof(*ab), gfp_mask);
if (!ab)
goto err;
}
atomic_inc(&audit_backlog);
ab->skb = alloc_skb(AUDIT_BUFSIZ, gfp_mask);
if (!ab->skb)
goto err;
ab->ctx = ctx;
nlh = (struct nlmsghdr *)skb_put(ab->skb, NLMSG_SPACE(0));
nlh->nlmsg_type = type;
nlh->nlmsg_flags = 0;
nlh->nlmsg_pid = 0;
nlh->nlmsg_seq = 0;
return ab;
err:
audit_buffer_free(ab);
return NULL;
}
/* Obtain an audit buffer. This routine does locking to obtain the
* audit buffer, but then no locking is required for calls to
* audit_log_*format. If the tsk is a task that is currently in a
* syscall, then the syscall is marked as auditable and an audit record
* will be written at syscall exit. If there is no associated task, tsk
* should be NULL. */
struct audit_buffer *audit_log_start(struct audit_context *ctx, int type)
{
struct audit_buffer *ab = NULL;
struct timespec t;
unsigned int serial;
if (!audit_initialized)
return NULL;
if (audit_backlog_limit
&& atomic_read(&audit_backlog) > audit_backlog_limit) {
if (audit_rate_check())
printk(KERN_WARNING
"audit: audit_backlog=%d > "
"audit_backlog_limit=%d\n",
atomic_read(&audit_backlog),
audit_backlog_limit);
audit_log_lost("backlog limit exceeded");
return NULL;
}
ab = audit_buffer_alloc(ctx, GFP_ATOMIC, type);
if (!ab) {
audit_log_lost("out of memory in audit_log_start");
return NULL;
}
if (!audit_get_stamp(ab->ctx, &t, &serial)) {
t = CURRENT_TIME;
serial = 0;
}
audit_log_format(ab, "audit(%lu.%03lu:%u): ",
t.tv_sec, t.tv_nsec/1000000, serial);
return ab;
}
/**
* audit_expand - expand skb in the audit buffer
* @ab: audit_buffer
*
* Returns 0 (no space) on failed expansion, or available space if
* successful.
*/
static inline int audit_expand(struct audit_buffer *ab, int extra)
{
struct sk_buff *skb = ab->skb;
int ret = pskb_expand_head(skb, skb_headroom(skb), extra,
GFP_ATOMIC);
if (ret < 0) {
audit_log_lost("out of memory in audit_expand");
return 0;
}
return skb_tailroom(skb);
}
/* Format an audit message into the audit buffer. If there isn't enough
* room in the audit buffer, more room will be allocated and vsnprint
* will be called a second time. Currently, we assume that a printk
* can't format message larger than 1024 bytes, so we don't either. */
static void audit_log_vformat(struct audit_buffer *ab, const char *fmt,
va_list args)
{
int len, avail;
struct sk_buff *skb;
va_list args2;
if (!ab)
return;
BUG_ON(!ab->skb);
skb = ab->skb;
avail = skb_tailroom(skb);
if (avail == 0) {
avail = audit_expand(ab, AUDIT_BUFSIZ);
if (!avail)
goto out;
}
va_copy(args2, args);
len = vsnprintf(skb->tail, avail, fmt, args);
if (len >= avail) {
/* The printk buffer is 1024 bytes long, so if we get
* here and AUDIT_BUFSIZ is at least 1024, then we can
* log everything that printk could have logged. */
avail = audit_expand(ab, max_t(unsigned, AUDIT_BUFSIZ, 1+len-avail));
if (!avail)
goto out;
len = vsnprintf(skb->tail, avail, fmt, args2);
}
skb_put(skb, (len < avail) ? len : avail);
out:
return;
}
/* Format a message into the audit buffer. All the work is done in
* audit_log_vformat. */
void audit_log_format(struct audit_buffer *ab, const char *fmt, ...)
{
va_list args;
if (!ab)
return;
va_start(args, fmt);
audit_log_vformat(ab, fmt, args);
va_end(args);
}
void audit_log_hex(struct audit_buffer *ab, const unsigned char *buf, size_t len)
{
int i;
for (i=0; i<len; i++)
audit_log_format(ab, "%02x", buf[i]);
}
void audit_log_untrustedstring(struct audit_buffer *ab, const char *string)
{
const unsigned char *p = string;
while (*p) {
if (*p == '"' || *p == ' ' || *p < 0x20 || *p > 0x7f) {
audit_log_hex(ab, string, strlen(string));
return;
}
p++;
}
audit_log_format(ab, "\"%s\"", string);
}
/* This is a helper-function to print the d_path without using a static
* buffer or allocating another buffer in addition to the one in
* audit_buffer. */
void audit_log_d_path(struct audit_buffer *ab, const char *prefix,
struct dentry *dentry, struct vfsmount *vfsmnt)
{
char *p;
struct sk_buff *skb = ab->skb;
int len, avail;
if (prefix)
audit_log_format(ab, " %s", prefix);
avail = skb_tailroom(skb);
p = d_path(dentry, vfsmnt, skb->tail, avail);
if (IS_ERR(p)) {
/* FIXME: can we save some information here? */
audit_log_format(ab, "<toolong>");
} else {
/* path isn't at start of buffer */
len = ((char *)skb->tail + avail - 1) - p;
memmove(skb->tail, p, len);
skb_put(skb, len);
}
}
/* Remove queued messages from the audit_txlist and send them to user space. */
static void audit_tasklet_handler(unsigned long arg)
{
LIST_HEAD(list);
struct audit_buffer *ab;
unsigned long flags;
spin_lock_irqsave(&audit_txlist_lock, flags);
list_splice_init(&audit_txlist, &list);
spin_unlock_irqrestore(&audit_txlist_lock, flags);
while (!list_empty(&list)) {
ab = list_entry(list.next, struct audit_buffer, list);
list_del(&ab->list);
audit_log_end_fast(ab);
}
}
static DECLARE_TASKLET(audit_tasklet, audit_tasklet_handler, 0);
/* The netlink_* functions cannot be called inside an irq context, so
* the audit buffer is places on a queue and a tasklet is scheduled to
* remove them from the queue outside the irq context. May be called in
* any context. */
static void audit_log_end_irq(struct audit_buffer *ab)
{
unsigned long flags;
if (!ab)
return;
spin_lock_irqsave(&audit_txlist_lock, flags);
list_add_tail(&ab->list, &audit_txlist);
spin_unlock_irqrestore(&audit_txlist_lock, flags);
tasklet_schedule(&audit_tasklet);
}
/* Send the message in the audit buffer directly to user space. May not
* be called in an irq context. */
static void audit_log_end_fast(struct audit_buffer *ab)
{
BUG_ON(in_irq());
if (!ab)
return;
if (!audit_rate_check()) {
audit_log_lost("rate limit exceeded");
} else {
if (audit_log_drain(ab))
return;
}
audit_buffer_free(ab);
}
/* Send or queue the message in the audit buffer, depending on the
* current context. (A convenience function that may be called in any
* context.) */
void audit_log_end(struct audit_buffer *ab)
{
if (in_irq())
audit_log_end_irq(ab);
else
audit_log_end_fast(ab);
}
/* Log an audit record. This is a convenience function that calls
* audit_log_start, audit_log_vformat, and audit_log_end. It may be
* called in any context. */
void audit_log(struct audit_context *ctx, int type, const char *fmt, ...)
{
struct audit_buffer *ab;
va_list args;
ab = audit_log_start(ctx, type);
if (ab) {
va_start(args, fmt);
audit_log_vformat(ab, fmt, args);
va_end(args);
audit_log_end(ab);
}
}