user message limits
by LC Bruzenak
I know I can go look at the code, however I figured I'd ask here first
about the limits on the user message in both audit_log_user_message and
ausearch.
With audit_log_user_message the maximum length allowed appears to be
around MAX_AUDIT_MESSAGE_LENGTH-100. I think it may depend on the
executable name length (and other stuff auto-pushed into the string)
which is why I say "around".
Even when I get a successful return value (from audit_log_user_message),
I don't get my string back out in "ausearch" unless it is WAY smaller -
~1K or less I think.
Any ideas/thoughts?
This is the latest (1.7.11-2) audit package.
Thx,
LCB.
--
LC (Lenny) Bruzenak
lenny(a)magitekltd.com
11 years, 2 months
AUDIT_SIGNAL_INFO
by Matthew Booth
Under what circumstances will the RHEL 4 kernel generate a message of
type AUDIT_SIGNAL_INFO? My understanding is that it should be sent when
a process sends a signal to the audit daemon, however I have not
observed that. Any ideas?
Thanks,
Matt
--
Matthew Booth, RHCA, RHCSS
Red Hat, Global Professional Services
M: +44 (0)7977 267231
GPG ID: D33C3490
GPG FPR: 3733 612D 2D05 5458 8A8A 1600 3441 EA19 D33C 3490
12 years, 6 months
Near Term Audit Road Map
by Steve Grubb
Hi,
With the proposals sent to the list, I wanted to talk about how this might
play out code-wise. With regard to the current code base, I am working on a
1.8 release. This would represent finishing the remote logging app and
nothing more. The 1.8 series would become just an update series just like the
1.0.x series did.
In parallel with finishing remote logging, I would release a 2.0 version.
Patches applied to 1.8 would also be applied to 2.0. A 2.1 release would
signify the completion of remote logging that branch. I would recommend this
branch for all distributions pulling new code in.
The 2.0 branch will also have a couple more changes. I want to split up the
audit source code a little bit. I want to drop the system-config-audit code
and let it become standalone package updated and distributed separately.
I also want to drop all audispd-plugins in the 2.0 branch and have them
released separately. They cause unnecessary build dependencies for the audit
package.
During the work for a 2.2 release, I would also like to pull the audispd
program inside auditd. In the past, I tried to keep auditd lean and single
purpose, but with adding remote logging and kerberos support, we already have
something that is hard to analyze. So, to improve performance and decrease
system load, the audit daemon will also do event dispatching.
Would this proposal impact anyone in a Bad Way?
Thanks,
-Steve
12 years, 6 months
[PATCH] Add auditd listener and remote audit protocol
by DJ Delorie
Second in a series, a bit bigger than the first one.
(http://www.redhat.com/archives/linux-audit/2008-August/msg00070.html)
The goal of this patch is to add the server side of the remote logging
feature. To this end, a new auditd-listener.c is added which listens
on a TCP port for connections from other systems' audisp-remote
plugins. A new (private) protocol is added which prepends each
message with a header, giving length, status, version, and sequence
information. Each message begets a reply from the server, so we can
pass along status like "disk full" or "ok". Currently, these call a
set of stub functions, as the details of performing appropriate
actions from the plugin are yet to be decided.
The remote plugin has a new option "format" for "ascii" or "managed"
to choose between the old protocol (ascii strings) and the new one
(the header with ACK, default).
The listener will accept either format. It has new options for the
listen port, accept queue size, and acceptable client-side ports.
Comments?
DJ
Proposed ChangeLog entry:
- Add TCP listener and managed remote protocol features.
diff -N -U 3 -x .svn -r pristine/audisp/plugins/remote/audisp-remote.c trunk/audisp/plugins/remote/audisp-remote.c
--- pristine/audisp/plugins/remote/audisp-remote.c 2008-08-04 12:47:28.000000000 -0400
+++ trunk/audisp/plugins/remote/audisp-remote.c 2008-08-14 14:10:48.000000000 -0400
@@ -30,7 +30,11 @@
#include <unistd.h>
#include <stdlib.h>
#include <errno.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <netinet/tcp.h>
#include "libaudit.h"
+#include "private.h"
#include "remote-config.h"
#define CONFIG_FILE "/etc/audisp/audisp-remote.conf"
@@ -68,6 +72,56 @@
hup = 0;
}
+/*
+ * Handlers for various events coming back from the remote server.
+ * Return -1 if the remote dispatcher should exit.
+ */
+
+/* Loss of sync - got an invalid response. */
+static int sync_error_handler (const char *why)
+{
+ /* "why" has human-readable details on why we've lost (or will
+ be losing) sync. */
+ syslog (LOG_ERR, "lost/losing sync, %s", why);
+ return -1;
+}
+
+static int remote_disk_low_handler (const char *message)
+{
+ syslog (LOG_WARNING, "remote disk low, %s", message);
+ return 0;
+}
+
+static int remote_disk_full_handler (const char *message)
+{
+ syslog (LOG_ERR, "remote disk full, %s", message);
+ return -1;
+}
+
+static int remote_disk_error_handler (const char *message)
+{
+ syslog (LOG_ERR, "remote disk error, %s", message);
+ return -1;
+}
+
+static int remote_server_ending_handler (const char *message)
+{
+ syslog (LOG_INFO, "remote server ending, %s", message);
+ return -1;
+}
+
+static int generic_remote_error_handler (const char *message)
+{
+ stop = 1;
+ syslog(LOG_INFO, "audisp-remote: remote error: %s", message);
+ return -1;
+}
+static int generic_remote_warning_handler (const char *message)
+{
+ syslog(LOG_INFO, "audisp-remote: remote warning: %s", message);
+ return 0;
+}
+
int main(int argc, char *argv[])
{
@@ -122,6 +176,7 @@
struct addrinfo *ai;
struct addrinfo hints;
char remote[BUF_SIZE];
+ int one=1;
memset(&hints, '\0', sizeof(hints));
hints.ai_flags = AI_ADDRCONFIG|AI_NUMERICSERV;
@@ -140,12 +195,35 @@
freeaddrinfo(ai);
return -1;
}
+
+ setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, (char *)&one, sizeof (int));
+
+ if (config.local_port != 0) {
+ struct sockaddr_in address;
+
+ memset (&address, 0, sizeof(address));
+ address.sin_family = htons(AF_INET);
+ address.sin_port = htons(config.local_port);
+ address.sin_addr.s_addr = htonl(INADDR_ANY);
+
+ if ( bind ( sock, (struct sockaddr *)&address, sizeof(address)) ) {
+ syslog(LOG_ERR, "Cannot bind local socket to port %d - exiting",
+ config.local_port);
+ close(sock);
+ return -1;
+ }
+
+ }
if (connect(sock, ai->ai_addr, ai->ai_addrlen)) {
syslog(LOG_ERR, "Error connecting to %s: %s - exiting",
config.remote_server, strerror(errno));
freeaddrinfo(ai);
return -1;
}
+
+ setsockopt(sock, SOL_SOCKET, SO_KEEPALIVE, (char *)&one, sizeof (int));
+ setsockopt(sock, IPPROTO_TCP, TCP_NODELAY, (char *)&one, sizeof (int));
+
freeaddrinfo(ai);
return 0;
}
@@ -166,16 +244,143 @@
return rc;
}
-static int relay_sock(const char *s, size_t len)
+static int ar_write (int sock, const void *buf, int len)
{
int rc;
-
do {
- rc = write(sock, s, len);
+ rc = write(sock, buf, len);
} while (rc < 0 && errno == EINTR);
- if (rc > 0)
- return 0;
- return -1;
+ return rc;
+}
+
+static int ar_read (int sock, void *buf, int len)
+{
+ unsigned char *obuf = buf;
+ int rc = 0, r;
+ while (len > 0) {
+ do {
+ r = read(sock, buf, len);
+ } while (r < 0 && errno == EINTR);
+ if (r < 0)
+ return r;
+ if (r == 0)
+ break;
+ rc += r;
+ buf = (void *)((char *)buf + r);
+ len -= r;
+ }
+ return rc;
+}
+
+static int relay_sock_ascii(const char *s, size_t len)
+{
+ int rc;
+
+ rc = ar_write(sock, s, len);
+ if (rc <= 0) {
+ stop = 1;
+ syslog(LOG_ERR, "connection to %s closed unexpectedly - exiting",
+ config.remote_server);
+ return -1;
+ }
+
+ return 0;
+}
+
+static int relay_sock_managed(const char *s, size_t len)
+{
+ static int sequence_id = 1;
+ int rc;
+ unsigned char header[AUDIT_RMW_HEADER_SIZE];
+ int hver, mver;
+ uint32_t magic, type, rlen, seq;
+ char msg[MAX_AUDIT_MESSAGE_LENGTH+1];
+
+ sequence_id ++;
+ AUDIT_RMW_PACK_HEADER (header, 0, 0, len, sequence_id);
+ rc = ar_write(sock, header, AUDIT_RMW_HEADER_SIZE);
+ if (rc <= 0) {
+ stop = 1;
+ syslog(LOG_ERR, "connection to %s closed unexpectedly - exiting",
+ config.remote_server);
+ return -1;
+ }
+
+ rc = ar_write(sock, s, len);
+ if (rc <= 0) {
+ stop = 1;
+ syslog(LOG_ERR, "connection to %s closed unexpectedly - exiting",
+ config.remote_server);
+ return -1;
+ }
+
+ rc = ar_read (sock, header, AUDIT_RMW_HEADER_SIZE);
+ if (rc < 16) {
+ stop = 1;
+ syslog(LOG_ERR, "connection to %s closed unexpectedly - exiting",
+ config.remote_server);
+ return -1;
+ }
+
+
+ if (! AUDIT_RMW_IS_MAGIC (header, AUDIT_RMW_HEADER_SIZE))
+ /* FIXME: the right thing to do here is close the socket and start a new one. */
+ return sync_error_handler ("bad magic number");
+
+ AUDIT_RMW_UNPACK_HEADER (header, hver, mver, type, rlen, seq);
+
+ if (rlen > MAX_AUDIT_MESSAGE_LENGTH)
+ return sync_error_handler ("message too long");
+
+ if (rlen > 0
+ && ar_read (sock, msg, rlen) < rlen)
+ return sync_error_handler ("ran out of data reading reply");
+ msg[rlen] = 0;
+
+ if (seq != sequence_id)
+ /* FIXME: should we read another header and
+ see if it matches? If so, we need to deal
+ with timeouts. */
+ return sync_error_handler ("mismatched response");
+
+ /* Specific errors we know how to deal with. */
+
+ if (type == AUDIT_RMW_TYPE_ENDING)
+ return remote_server_ending_handler (msg);
+ if (type == AUDIT_RMW_TYPE_DISKLOW)
+ return remote_disk_low_handler (msg);
+ if (type == AUDIT_RMW_TYPE_DISKFULL)
+ return remote_disk_full_handler (msg);
+ if (type == AUDIT_RMW_TYPE_DISKERROR)
+ return remote_disk_error_handler (msg);
+
+ /* Generic errors. */
+ if (type & AUDIT_RMW_TYPE_FATALMASK)
+ return generic_remote_error_handler (msg);
+ if (type & AUDIT_RMW_TYPE_WARNMASK)
+ return generic_remote_warning_handler (msg);
+
+ return 0;
+}
+
+static int relay_sock(const char *s, size_t len)
+{
+ int rc;
+
+ switch (config.format)
+ {
+ case F_MANAGED:
+ rc = relay_sock_managed (s, len);
+ break;
+ case F_ASCII:
+ rc = relay_sock_ascii (s, len);
+ break;
+ default:
+ rc = -1;
+ break;
+ }
+
+ return rc;
}
static int relay_event(const char *s, size_t len)
diff -N -U 3 -x .svn -r pristine/audisp/plugins/remote/audisp-remote.conf trunk/audisp/plugins/remote/audisp-remote.conf
--- pristine/audisp/plugins/remote/audisp-remote.conf 2008-08-04 12:47:28.000000000 -0400
+++ trunk/audisp/plugins/remote/audisp-remote.conf 2008-08-14 15:11:23.000000000 -0400
@@ -5,8 +5,9 @@
remote_server =
port = 60
+#local_port =
transport = tcp
mode = immediate
queue_depth = 20
fail_action = SYSLOG
-
+format = managed
diff -N -U 3 -x .svn -r pristine/audisp/plugins/remote/audisp-remote.conf.5 trunk/audisp/plugins/remote/audisp-remote.conf.5
--- pristine/audisp/plugins/remote/audisp-remote.conf.5 2008-08-04 12:47:28.000000000 -0400
+++ trunk/audisp/plugins/remote/audisp-remote.conf.5 2008-08-14 15:36:53.000000000 -0400
@@ -11,6 +11,13 @@
.I port
This option is an unsigned integer that indicates what port to connect to on the remote machine.
.TP
+.I local_port
+This option is an unsigned integer that indicates what local port to
+connect from on the local machine. If unspecified (the default) or
+set to the word
+.I any
+then any available unpriviledged port is used.
+.TP
.I transport
This parameter tells the remote logging app how to send events to the remote system. Valid values are
.IR tcp ", and " ssl ".
@@ -55,7 +62,22 @@
option will cause the remote logging app to put the computer system in single user mode.
.I halt
option will cause the remote logging app to shutdown the computer system.
+.TP
+.I format
+This parameter tells the remote logging app what data format will be
+used for the messages sent over the network. The default is
+.I managed
+which adds some overhead to ensure each message is properly handled on
+the remote end, and to receive status messages from the remote server.
+If
+.I ascii
+is given instead, each message is a simple ASCII text line with no
+overhead at all.
+.SH "NOTES"
+Specifying a local port may make it difficult to restart the audit
+subsystem due to the previous connection being in a TIME_WAIT state,
+if you're reconnecting to and from the same hosts and ports as before.
.SH "SEE ALSO"
.BR audispd (8),
.BR audisp-remote(8).
diff -N -U 3 -x .svn -r pristine/audisp/plugins/remote/remote-config.c trunk/audisp/plugins/remote/remote-config.c
--- pristine/audisp/plugins/remote/remote-config.c 2008-08-04 12:47:28.000000000 -0400
+++ trunk/audisp/plugins/remote/remote-config.c 2008-08-13 22:19:31.000000000 -0400
@@ -62,6 +62,8 @@
remote_conf_t *config);
static int port_parser(struct nv_pair *nv, int line,
remote_conf_t *config);
+static int local_port_parser(struct nv_pair *nv, int line,
+ remote_conf_t *config);
static int transport_parser(struct nv_pair *nv, int line,
remote_conf_t *config);
static int mode_parser(struct nv_pair *nv, int line,
@@ -70,16 +72,20 @@
remote_conf_t *config);
static int fail_action_parser(struct nv_pair *nv, int line,
remote_conf_t *config);
+static int format_parser(struct nv_pair *nv, int line,
+ remote_conf_t *config);
static int sanity_check(remote_conf_t *config, const char *file);
static const struct kw_pair keywords[] =
{
{"remote_server", server_parser, 0 },
{"port", port_parser, 0 },
+ {"local_port", local_port_parser, 0 },
{"transport", transport_parser, 0 },
{"mode", mode_parser, 0 },
{"queue_depth", depth_parser, 0 },
{"fail_action", fail_action_parser, 0 },
+ {"format", format_parser, 0 },
{ NULL, NULL }
};
@@ -107,6 +113,13 @@
{ NULL, 0 }
};
+static const struct nv_list format_words[] =
+{
+ {"ascii", F_ASCII },
+ {"managed", F_MANAGED },
+ { NULL, 0 }
+};
+
/*
* Set everything to its default value
*/
@@ -114,11 +127,13 @@
{
config->remote_server = NULL;
config->port = 60;
+ config->local_port = 0;
config->port = T_TCP;
config->mode = M_IMMEDIATE;
config->queue_depth = 20;
config->fail_action = F_SYSLOG;
config->fail_exe = NULL;
+ config->format = F_MANAGED;
}
int load_config(remote_conf_t *config, const char *file)
@@ -392,6 +407,46 @@
return 0;
}
+static int local_port_parser(struct nv_pair *nv, int line, remote_conf_t *config)
+{
+ const char *ptr = nv->value;
+ int i;
+
+ if (strcasecmp (ptr, "any") == 0) {
+ config->local_port = 0;
+ return 0;
+ }
+
+ /* check that all chars are numbers */
+ for (i=0; ptr[i]; i++) {
+ if (!isdigit(ptr[i])) {
+ syslog(LOG_ERR,
+ "Value %s should only be numbers - line %d",
+ nv->value, line);
+ return 1;
+ }
+ }
+
+ /* convert to unsigned int */
+ errno = 0;
+ i = strtoul(nv->value, NULL, 10);
+ if (errno) {
+ syslog(LOG_ERR,
+ "Error converting string to a number (%s) - line %d",
+ strerror(errno), line);
+ return 1;
+ }
+ /* Check its range */
+ if (i > INT_MAX) {
+ syslog(LOG_ERR,
+ "Error - converted number (%s) is too large - line %d",
+ nv->value, line);
+ return 1;
+ }
+ config->local_port = (unsigned int)i;
+ return 0;
+}
+
static int transport_parser(struct nv_pair *nv, int line, remote_conf_t *config)
{
int i;
@@ -477,6 +532,20 @@
return 1;
}
+static int format_parser(struct nv_pair *nv, int line,
+ remote_conf_t *config)
+{
+ int i;
+ for (i=0; format_words[i].name != NULL; i++) {
+ if (strcasecmp(nv->value, format_words[i].name) == 0) {
+ config->format = format_words[i].option;
+ return 0;
+ }
+ }
+ syslog(LOG_ERR, "Option %s not found - line %d", nv->value, line);
+ return 1;
+}
+
/*
* This function is where we do the integrated check of the audispd config
* options. At this point, all fields have been read. Returns 0 if no
diff -N -U 3 -x .svn -r pristine/audisp/plugins/remote/remote-config.h trunk/audisp/plugins/remote/remote-config.h
--- pristine/audisp/plugins/remote/remote-config.h 2008-08-04 12:47:28.000000000 -0400
+++ trunk/audisp/plugins/remote/remote-config.h 2008-08-13 22:17:19.000000000 -0400
@@ -27,16 +27,19 @@
typedef enum { M_IMMEDIATE, M_STORE_AND_FORWARD } mode_t;
typedef enum { T_TCP, T_SSL, T_GSSAPI, T_LABELED } transport_t;
typedef enum { F_IGNORE, F_SYSLOG, F_EXEC, F_SUSPEND, F_SINGLE, F_HALT } fail_t;
+typedef enum { F_ASCII, F_MANAGED } format_t;
typedef struct remote_conf
{
const char *remote_server;
unsigned int port;
+ unsigned int local_port;
transport_t transport;
mode_t mode;
unsigned int queue_depth;
fail_t fail_action;
const char *fail_exe;
+ format_t format;
} remote_conf_t;
void clear_config(remote_conf_t *config);
diff -N -U 3 -x .svn -r pristine/docs/auditd.conf.5 trunk/docs/auditd.conf.5
--- pristine/docs/auditd.conf.5 2008-08-04 12:47:31.000000000 -0400
+++ trunk/docs/auditd.conf.5 2008-08-14 15:35:29.000000000 -0400
@@ -212,6 +212,26 @@
option will cause the audit daemon to put the computer system in single user mode.
.I halt
option will cause the audit daemon to shutdown the computer system.
+.TP
+.I tcp_listen_port
+This is a numeric value in the range 1..65535 which, if specified,
+causes auditd to listen on the corresponding TCP port for audit
+records from remote systems.
+.TP
+.I tcp_listen_queue
+This is a numeric value which indicates how many pending (requested
+but unaccepted) connections are allowed. The default is 5. Setting
+this too small may cause connections to be rejected if too many hosts
+start up at exactly the same time, such as after a power failure.
+.TP
+.I tcp_client_ports
+This parameter may be a single numeric value or two values separated
+by a dash (no spaces allowed). It indicates which client ports are
+allowed for incoming connections. If not specified, any port is
+allowed. Allowed values are 1..65535. For example, to require the
+client use a priviledged port, specify
+.I 1-1023
+for this parameter.
.SH NOTES
In a CAPP environment, the audit trail is considered so important that access to system resources must be denied if an audit trail cannot be created. In this environment, it would be suggested that /var/log/audit be on its own partition. This is to ensure that space detection is accurate and that no other process comes along and consumes part of it.
@@ -227,6 +247,11 @@
The disk_full_action is triggered when no more room exists on the partition. All access should be terminated since no more audit capability exists. This can be set to either single or halt.
.PP
The disk_error_action should be set to syslog, single, or halt depending on your local policies regarding handling of hardware malfunctions.
+.PP
+Specifying a single allowed client port may make it difficult for the
+client to restart their audit subsystem, as it will be unable to
+recreate a connection with the same host addresses and ports until the
+connection closure TIME_WAIT state times out.
.SH FILES
.TP
diff -N -U 3 -x .svn -r pristine/init.d/auditd.conf trunk/init.d/auditd.conf
--- pristine/init.d/auditd.conf 2008-08-04 12:47:28.000000000 -0400
+++ trunk/init.d/auditd.conf 2008-08-14 15:22:08.000000000 -0400
@@ -22,4 +22,6 @@
admin_space_left_action = SUSPEND
disk_full_action = SUSPEND
disk_error_action = SUSPEND
-
+#tcp_listen_port =
+#tcp_listen_queue = 5
+#tcp_client_ports = 1024-65535
diff -N -U 3 -x .svn -r pristine/lib/libaudit.h trunk/lib/libaudit.h
--- pristine/lib/libaudit.h 2008-08-04 12:47:34.000000000 -0400
+++ trunk/lib/libaudit.h 2008-08-14 10:58:46.000000000 -0400
@@ -413,10 +413,6 @@
};
};
-struct auditd_reply_list {
- struct audit_reply reply;
- struct auditd_reply_list *next;
-};
//
// End of ABI control
//////////////////////////////////////////////////////
diff -N -U 3 -x .svn -r pristine/lib/private.h trunk/lib/private.h
--- pristine/lib/private.h 2008-08-04 12:47:34.000000000 -0400
+++ trunk/lib/private.h 2008-08-14 10:58:46.000000000 -0400
@@ -55,6 +55,83 @@
;
#endif
+/* This structure is for protocol reference only. All fields are
+ packed and in network order (LSB first). */
+struct auditd_remote_message_wrapper {
+ /* The magic number shall never have LF (0x0a) as one of its bytes. */
+ uint32_t magic;
+ /* Bumped when the layout of this structure changes. */
+ uint8_t header_version;
+ /* The minimum support needed to understand this message type.
+ * Normally zero. */
+ uint8_t message_version;
+ /* Upper 8 bits are generic type, see below. */
+ uint32_t type;
+ /* Number of bytes that follow this header Must be 0..MAX_AUDIT_MESSAGE_LENGTH. */
+ uint16_t length;
+ /* Copied from message to its reply. */
+ uint32_t sequence_id;
+ /* The message follows for LENGTH bytes. */
+};
+
+#define AUDIT_RMW_HEADER_SIZE 16
+/* The magic number shall never have LF (0x0a) as one of its bytes. */
+#define AUDIT_RMW_MAGIC 0xff0000feUL
+
+#define AUDIT_RMW_HEADER_VERSION 0
+
+/* If set, this is a reply. */
+#define AUDIT_RMW_TYPE_REPLYMASK 0x40000000
+/* If set, this reply indicates a fatal error of some sort. */
+#define AUDIT_RMW_TYPE_FATALMASK 0x20000000
+/* If set, this reply indicates success but with some warnings. */
+#define AUDIT_RMW_TYPE_WARNMASK 0x10000000
+/* This part of the message type is the details for the above. */
+#define AUDIT_RMW_TYPE_DETAILMASK 0x0fffffff
+
+/* Version 0 messages. */
+#define AUDIT_RMW_TYPE_MESSAGE 0x00000000
+#define AUDIT_RMW_TYPE_ACK 0x40000000
+#define AUDIT_RMW_TYPE_ENDING 0x40000001
+#define AUDIT_RMW_TYPE_DISKLOW 0x50000001
+#define AUDIT_RMW_TYPE_DISKFULL 0x60000001
+#define AUDIT_RMW_TYPE_DISKERROR 0x60000002
+
+/* These next four should not be called directly. */
+#define _AUDIT_RMW_PUTN32(header,i,v) \
+ header[i] = v & 0xff; \
+ header[i+1] = (v>>8) & 0xff; \
+ header[i+2] = (v>>16) & 0xff; \
+ header[i+3] = (v>>24) & 0xff;
+#define _AUDIT_RMW_PUTN16(header,i,v) \
+ header[i] = v & 0xff; \
+ header[i+1] = (v>>8) & 0xff;
+#define _AUDIT_RMW_GETN32(header,i) \
+ (header[i] | (header[i+1]<<8) | (header[i+2]<<16) | (header[i+3]<<24))
+#define _AUDIT_RMW_GETN16(header,i) \
+ (header[i] | (header[i+1]<<8))
+
+/* For these, HEADER must by of type "unsigned char *" or "unsigned
+ char []" */
+
+#define AUDIT_RMW_PACK_HEADER(header,mver,type,len,seq) \
+ _AUDIT_RMW_PUTN32 (header,0, AUDIT_RMW_MAGIC); \
+ header[4] = AUDIT_RMW_HEADER_VERSION; \
+ header[5] = mver; \
+ _AUDIT_RMW_PUTN32 (header,6, type); \
+ _AUDIT_RMW_PUTN16 (header,10, len); \
+ _AUDIT_RMW_PUTN32 (header,12, seq);
+
+#define AUDIT_RMW_IS_MAGIC(header,length) \
+ (length >= 4 && _AUDIT_RMW_GETN32 (header,0) == AUDIT_RMW_MAGIC)
+
+#define AUDIT_RMW_UNPACK_HEADER(header,hver,mver,type,len,seq) \
+ hver = header[4]; \
+ mver = header[5]; \
+ type = _AUDIT_RMW_GETN32 (header,6); \
+ len = _AUDIT_RMW_GETN16 (header,10); \
+ seq = _AUDIT_RMW_GETN32 (header,12);
+
/* General */
extern int audit_send(int fd, int type, const void *data, unsigned int size)
hidden;
diff -N -U 3 -x .svn -r pristine/src/Makefile.am trunk/src/Makefile.am
--- pristine/src/Makefile.am 2008-08-07 21:18:16.000000000 -0400
+++ trunk/src/Makefile.am 2008-08-08 16:41:07.000000000 -0400
@@ -27,9 +27,9 @@
LIBS = -Lmt -lauditmt -Llibev -lev -lrt -lm
LDADD = -lpthread
AM_CFLAGS = -D_REENTRANT -D_GNU_SOURCE
-noinst_HEADERS = auditd-config.h auditd-event.h ausearch-llist.h ausearch-options.h auditctl-llist.h aureport-options.h ausearch-parse.h aureport-scan.h ausearch-lookup.h ausearch-int.h auditd-dispatch.h ausearch-string.h ausearch-nvpair.h ausearch-common.h ausearch-avc.h ausearch-time.h
+noinst_HEADERS = auditd-config.h auditd-event.h auditd-listen.h ausearch-llist.h ausearch-options.h auditctl-llist.h aureport-options.h ausearch-parse.h aureport-scan.h ausearch-lookup.h ausearch-int.h auditd-dispatch.h ausearch-string.h ausearch-nvpair.h ausearch-common.h ausearch-avc.h ausearch-time.h
-auditd_SOURCES = auditd.c auditd-event.c auditd-config.c auditd-reconfig.c auditd-sendmail.c auditd-dispatch.c
+auditd_SOURCES = auditd.c auditd-event.c auditd-config.c auditd-reconfig.c auditd-sendmail.c auditd-dispatch.c auditd-listen.c
auditd_CFLAGS = -fPIE -DPIE -g -D_REENTRANT -D_GNU_SOURCE
auditd_LDFLAGS = -pie -Wl,-z,relro
auditd_DEPENDENCIES = mt/libauditmt.a libev/libev.a
diff -N -U 3 -x .svn -r pristine/src/auditd-config.c trunk/src/auditd-config.c
--- pristine/src/auditd-config.c 2008-08-07 17:32:20.000000000 -0400
+++ trunk/src/auditd-config.c 2008-08-08 15:56:13.000000000 -0400
@@ -39,6 +39,8 @@
#include "libaudit.h"
#include "private.h"
+#define TCP_PORT_MAX 65535
+
/* Local prototypes */
struct nv_pair
{
@@ -103,6 +105,12 @@
struct daemon_conf *config);
static int priority_boost_parser(struct nv_pair *nv, int line,
struct daemon_conf *config);
+static int tcp_listen_port_parser(struct nv_pair *nv, int line,
+ struct daemon_conf *config);
+static int tcp_listen_queue_parser(struct nv_pair *nv, int line,
+ struct daemon_conf *config);
+static int tcp_client_ports_parser(struct nv_pair *nv, int line,
+ struct daemon_conf *config);
static int sanity_check(struct daemon_conf *config);
static const struct kw_pair keywords[] =
@@ -127,6 +135,9 @@
{"disk_full_action", disk_full_action_parser, 1 },
{"disk_error_action", disk_error_action_parser, 1 },
{"priority_boost", priority_boost_parser, 0 },
+ {"tcp_listen_port", tcp_listen_port_parser, 0 },
+ {"tcp_listen_queue", tcp_listen_queue_parser, 0 },
+ {"tcp_client_ports", tcp_client_ports_parser, 0 },
{ NULL, NULL }
};
@@ -227,6 +238,10 @@
config->disk_full_exe = NULL;
config->disk_error_action = FA_SYSLOG;
config->disk_error_exe = NULL;
+ config->tcp_listen_port = 0;
+ config->tcp_listen_queue = 5;
+ config->tcp_client_min_port = 0;
+ config->tcp_client_max_port = TCP_PORT_MAX;
}
static log_test_t log_test = TEST_AUDITD;
@@ -1109,6 +1124,172 @@
return 0;
}
+static int tcp_listen_port_parser(struct nv_pair *nv, int line,
+ struct daemon_conf *config)
+{
+ const char *ptr = nv->value;
+ unsigned long i;
+
+ audit_msg(LOG_DEBUG, "tcp_listen_port_parser called with: %s",
+ nv->value);
+
+ /* check that all chars are numbers */
+ for (i=0; ptr[i]; i++) {
+ if (!isdigit(ptr[i])) {
+ audit_msg(LOG_ERR,
+ "Value %s should only be numbers - line %d",
+ nv->value, line);
+ return 1;
+ }
+ }
+
+ /* convert to unsigned int */
+ errno = 0;
+ i = strtoul(nv->value, NULL, 10);
+ if (errno) {
+ audit_msg(LOG_ERR,
+ "Error converting string to a number (%s) - line %d",
+ strerror(errno), line);
+ return 1;
+ }
+ /* Check its range */
+ if (i > TCP_PORT_MAX) {
+ audit_msg(LOG_ERR,
+ "Error - converted number (%s) is too large - line %d",
+ nv->value, line);
+ return 1;
+ }
+ if (i < 1) {
+ audit_msg(LOG_ERR,
+ "Error - converted number (%s) is too small - line %d",
+ nv->value, line);
+ return 1;
+ }
+ config->tcp_listen_port = (unsigned int)i;
+ return 0;
+}
+
+static int tcp_listen_queue_parser(struct nv_pair *nv, int line,
+ struct daemon_conf *config)
+{
+ const char *ptr = nv->value;
+ unsigned long i;
+
+ audit_msg(LOG_DEBUG, "tcp_listen_queue_parser called with: %s",
+ nv->value);
+
+ /* check that all chars are numbers */
+ for (i=0; ptr[i]; i++) {
+ if (!isdigit(ptr[i])) {
+ audit_msg(LOG_ERR,
+ "Value %s should only be numbers - line %d",
+ nv->value, line);
+ return 1;
+ }
+ }
+
+ /* convert to unsigned int */
+ errno = 0;
+ i = strtoul(nv->value, NULL, 10);
+ if (errno) {
+ audit_msg(LOG_ERR,
+ "Error converting string to a number (%s) - line %d",
+ strerror(errno), line);
+ return 1;
+ }
+ /* Check its range. While this value is technically
+ unlimited, it's limited by the kernel, and we limit it here
+ for sanity. */
+ if (i > TCP_PORT_MAX) {
+ audit_msg(LOG_ERR,
+ "Error - converted number (%s) is too large - line %d",
+ nv->value, line);
+ return 1;
+ }
+ if (i < 1) {
+ audit_msg(LOG_ERR,
+ "Error - converted number (%s) is too small - line %d",
+ nv->value, line);
+ return 1;
+ }
+ config->tcp_listen_queue = (unsigned int)i;
+ return 0;
+}
+
+static int tcp_client_ports_parser(struct nv_pair *nv, int line,
+ struct daemon_conf *config)
+{
+ const char *ptr = nv->value;
+ unsigned long i, minv, maxv;
+ const char *saw_dash = NULL;
+
+ audit_msg(LOG_DEBUG, "tcp_listen_queue_parser called with: %s",
+ nv->value);
+
+ /* check that all chars are numbers, with an optional inclusive '-'. */
+ for (i=0; ptr[i]; i++) {
+ if (i > 0 && ptr[i] == '-' && ptr[i+1] != '\0') {
+ saw_dash = ptr + i;
+ continue;
+ }
+ if (!isdigit(ptr[i])) {
+ audit_msg(LOG_ERR,
+ "Value %s should only be numbers, or two numbers separated by a dash - line %d",
+ nv->value, line);
+ return 1;
+ }
+ }
+ for (; ptr[i]; i++) {
+ if (!isdigit(ptr[i])) {
+ audit_msg(LOG_ERR,
+ "Value %s should only be numbers, or two numbers separated by a dash - line %d",
+ nv->value, line);
+ return 1;
+ }
+ }
+
+ /* convert to unsigned int */
+ errno = 0;
+ maxv = minv = strtoul(nv->value, NULL, 10);
+ if (errno) {
+ audit_msg(LOG_ERR,
+ "Error converting string to a number (%s) - line %d",
+ strerror(errno), line);
+ return 1;
+ }
+ if (saw_dash) {
+ maxv = strtoul(saw_dash + 1, NULL, 10);
+ if (errno) {
+ audit_msg(LOG_ERR,
+ "Error converting string to a number (%s) - line %d",
+ strerror(errno), line);
+ return 1;
+ }
+ }
+ /* Check their ranges. */
+ if (minv > TCP_PORT_MAX) {
+ audit_msg(LOG_ERR,
+ "Error - converted number (%d) is too large - line %d",
+ minv, line);
+ return 1;
+ }
+ if (maxv > TCP_PORT_MAX) {
+ audit_msg(LOG_ERR,
+ "Error - converted number (%d) is too large - line %d",
+ maxv, line);
+ return 1;
+ }
+ if (minv > maxv) {
+ audit_msg(LOG_ERR,
+ "Error - converted range (%d-%d) is reversed - line %d",
+ minv, maxv, line);
+ return 1;
+ }
+ config->tcp_client_min_port = (unsigned int)minv;
+ config->tcp_client_max_port = (unsigned int)maxv;
+ return 0;
+}
+
/*
* This function is where we do the integrated check of the audit config
* options. At this point, all fields have been read. Returns 0 if no
diff -N -U 3 -x .svn -r pristine/src/auditd-config.h trunk/src/auditd-config.h
--- pristine/src/auditd-config.h 2008-08-04 12:47:30.000000000 -0400
+++ trunk/src/auditd-config.h 2008-08-14 15:33:28.000000000 -0400
@@ -70,6 +70,10 @@
const char *disk_full_exe;
failure_action_t disk_error_action;
const char *disk_error_exe;
+ unsigned long tcp_listen_port;
+ unsigned long tcp_listen_queue;
+ unsigned long tcp_client_min_port;
+ unsigned long tcp_client_max_port;
};
void set_allow_links(int allow);
@@ -80,7 +84,9 @@
int resolve_node(struct daemon_conf *config);
void init_config_manager(void);
+#ifdef AUDITD_EVENT_H
int start_config_manager(struct auditd_reply_list *rep);
+#endif
void shutdown_config(void);
void free_config(struct daemon_conf *config);
diff -N -U 3 -x .svn -r pristine/src/auditd-event.c trunk/src/auditd-event.c
--- pristine/src/auditd-event.c 2008-08-04 12:47:30.000000000 -0400
+++ trunk/src/auditd-event.c 2008-08-13 21:56:24.000000000 -0400
@@ -148,6 +148,36 @@
dequeue'r is responsible for freeing the memory. */
void enqueue_event(struct auditd_reply_list *rep)
{
+ char *buf;
+ int len;
+
+ rep->ack_socket = 0;
+ rep->sequence_id = 0;
+
+ switch (consumer_data.config->log_format)
+ {
+ case LF_RAW:
+ buf = format_raw(&rep->reply, consumer_data.config);
+ break;
+ case LF_NOLOG:
+ return;
+ default:
+ audit_msg(LOG_ERR,
+ "Illegal log format detected %d",
+ consumer_data.config->log_format);
+ return;
+ }
+
+ len = strlen (buf);
+ if (len < MAX_AUDIT_MESSAGE_LENGTH - 1)
+ memcpy (rep->reply.msg.data, buf, len+1);
+ else
+ {
+ /* FIXME: is truncation the right thing to do? */
+ memcpy (rep->reply.msg.data, buf, MAX_AUDIT_MESSAGE_LENGTH-1);
+ rep->reply.msg.data[MAX_AUDIT_MESSAGE_LENGTH-1] = 0;
+ }
+
rep->next = NULL; /* new packet goes at end - so zero this */
pthread_mutex_lock(&consumer_data.queue_lock);
@@ -164,6 +194,45 @@
pthread_mutex_unlock(&consumer_data.queue_lock);
}
+/* This function takes a preformatted message and places it on the
+ queue. The dequeue'r is responsible for freeing the memory. */
+void enqueue_formatted_event(char *msg, int ack_socket, uint32_t sequence_id)
+{
+ int len;
+ struct auditd_reply_list *rep;
+
+ rep = (struct auditd_reply_list *) calloc (1, sizeof (*rep));
+ if (rep == NULL) {
+ audit_msg(LOG_ERR, "Cannot allocate audit reply");
+ return;
+ }
+
+ rep->ack_socket = ack_socket;
+ rep->sequence_id = sequence_id;
+
+ len = strlen (msg);
+ if (len < MAX_AUDIT_MESSAGE_LENGTH - 1)
+ memcpy (rep->reply.msg.data, msg, len+1);
+ else {
+ /* FIXME: is truncation the right thing to do? */
+ memcpy (rep->reply.msg.data, msg, MAX_AUDIT_MESSAGE_LENGTH-1);
+ rep->reply.msg.data[MAX_AUDIT_MESSAGE_LENGTH-1] = 0;
+ }
+
+ pthread_mutex_lock(&consumer_data.queue_lock);
+ if (consumer_data.head == NULL) {
+ consumer_data.head = consumer_data.tail = rep;
+ pthread_cond_signal(&consumer_data.queue_nonempty);
+ } else {
+ /* FIXME: wait for room on the queue */
+
+ /* OK there's room...add it in */
+ consumer_data.tail->next = rep; /* link in at end */
+ consumer_data.tail = rep; /* move end to newest */
+ }
+ pthread_mutex_unlock(&consumer_data.queue_lock);
+}
+
void resume_logging(void)
{
logging_suspended = 0;
@@ -233,30 +302,8 @@
rotate_logs_now(data);
}
if (!logging_suspended) {
- char *buf = NULL;
-
- switch (data->config->log_format)
- {
- case LF_RAW:
- buf = format_raw(&data->head->reply,
- data->config);
- break;
- case LF_NOLOG:
- return;
- default:
- audit_msg(LOG_ERR,
- "Illegal log format detected %d",
- data->config->log_format);
- break;
- }
- /* The only way buf is NULL is if there is an
- * unidentified format...which is impossible since
- * start up would have failed. */
- if (buf) {
- write_to_log(buf, data);
- buf[0] = 0;
- }
+ write_to_log(data->head->reply.msg.data, data);
/* See if we need to flush to disk manually */
if (data->config->flush == FT_INCREMENTAL) {
@@ -289,12 +336,32 @@
}
}
+static int ar_write (int sock, const void *buf, int len)
+{
+ int rc = 0, w;
+ while (len > 0) {
+ do {
+ w = write(sock, buf, len);
+ } while (w < 0 && errno == EINTR);
+ if (w < 0)
+ return w;
+ if (w == 0)
+ break;
+ rc += w;
+ len -= w;
+ buf = (const void *)((const char *)buf + w);
+ }
+ return rc;
+}
+
/* This function writes the given buf to the current log file */
static void write_to_log(const char *buf, struct auditd_consumer_data *data)
{
int rc;
FILE *f = data->log_file;
struct daemon_conf *config = data->config;
+ int ack_type = AUDIT_RMW_TYPE_ACK;
+ const char *msg = "";
/* write it to disk */
rc = fprintf(f, "%s\n", buf);
@@ -307,20 +374,35 @@
if (saved_errno == ENOSPC && fs_space_left == 1) {
fs_space_left = 0;
do_disk_full_action(config);
- } else
+ ack_type = AUDIT_RMW_TYPE_DISKFULL;
+ msg = "disk full";
+ } else {
do_disk_error_action("write", config);
+ ack_type = AUDIT_RMW_TYPE_DISKERROR;
+ msg = "disk write error";
+ }
+
+ } else {
- return;
+ /* check log file size & space left on partition */
+ if (config->daemonize == D_BACKGROUND) {
+ // If either of these fail, I consider it an inconvenience
+ // as opposed to something that is actionable. There may be
+ // some temporary condition that the system recovers from.
+ // The real error occurs on write.
+ check_log_file_size(data->log_fd, data);
+ check_space_left(data->log_fd, config);
+ }
}
- /* check log file size & space left on partition */
- if (config->daemonize == D_BACKGROUND) {
- // If either of these fail, I consider it an inconvenience
- // as opposed to something that is actionable. There may be
- // some temporary condition that the system recovers from.
- // The real error occurs on write.
- check_log_file_size(data->log_fd, data);
- check_space_left(data->log_fd, config);
+ if (data->head->ack_socket) {
+ unsigned char header[AUDIT_RMW_HEADER_SIZE];
+
+ AUDIT_RMW_PACK_HEADER (header, 0, ack_type, strlen(msg), data->head->sequence_id);
+
+ ar_write (data->head->ack_socket, header, AUDIT_RMW_HEADER_SIZE);
+ if (msg[0])
+ ar_write (data->head->ack_socket, msg, strlen(msg));
}
}
diff -N -U 3 -x .svn -r pristine/src/auditd-event.h trunk/src/auditd-event.h
--- pristine/src/auditd-event.h 2008-08-04 12:47:30.000000000 -0400
+++ trunk/src/auditd-event.h 2008-08-13 20:41:41.000000000 -0400
@@ -25,12 +25,21 @@
#define AUDITD_EVENT_H
#include "libaudit.h"
+
+struct auditd_reply_list {
+ struct audit_reply reply;
+ struct auditd_reply_list *next;
+ int ack_socket;
+ unsigned long sequence_id;
+};
+
#include "auditd-config.h"
void shutdown_events(void);
int init_event(struct daemon_conf *config);
void resume_logging(void);
void enqueue_event(struct auditd_reply_list *rep);
+void enqueue_formatted_event(char *msg, int ack_socket, uint32_t sequence_id);
void *consumer_thread_main(void *arg);
#endif
diff -N -U 3 -x .svn -r pristine/src/auditd-listen.c trunk/src/auditd-listen.c
--- pristine/src/auditd-listen.c 1969-12-31 19:00:00.000000000 -0500
+++ trunk/src/auditd-listen.c 2008-08-13 22:29:22.000000000 -0400
@@ -0,0 +1,335 @@
+/* auditd-listen.c --
+ * Copyright 2008 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
+ *
+ * Authors:
+ * DJ Delorie <dj(a)redhat.com>
+ *
+ */
+
+#include "config.h"
+#include <stdio.h>
+#include <unistd.h>
+#include <sys/stat.h>
+#include <errno.h>
+#include <string.h>
+#include <dirent.h>
+#include <ctype.h>
+#include <stdlib.h>
+#include <netdb.h>
+#include <fcntl.h> /* O_NOFOLLOW needs gnu defined */
+#include <libgen.h>
+#include <arpa/inet.h>
+#include <limits.h> /* INT_MAX */
+#include <sys/types.h>
+#include <sys/socket.h>
+#include <netinet/in.h>
+#include <netinet/tcp.h>
+#include "libaudit.h"
+#include "auditd-event.h"
+#include "auditd-config.h"
+#include "private.h"
+
+#include "ev.h"
+
+extern volatile int stop;
+
+typedef struct ev_tcp {
+ struct ev_io io;
+ struct sockaddr_in addr;
+ struct ev_tcp *next, *prev;
+ int bufptr;
+ unsigned char buffer [MAX_AUDIT_MESSAGE_LENGTH + 17];
+} ev_tcp;
+
+static int listen_socket;
+static struct ev_io tcp_listen_watcher;
+static int min_port, max_port;
+
+static struct ev_tcp *client_chain = 0;
+
+static char *sockaddr_to_ip (struct sockaddr_in *addr)
+{
+ unsigned char *uaddr = (unsigned char *)&(addr->sin_addr);
+ static char buf[40];
+
+ sprintf (buf, "%d.%d.%d.%d:%d",
+ uaddr[0], uaddr[1], uaddr[2], uaddr[3], ntohs (addr->sin_port));
+ return buf;
+}
+
+static void set_close_on_exec (int fd)
+{
+ int flags = fcntl (fd, F_GETFD);
+ if (flags == -1)
+ flags = 0;
+ flags |= FD_CLOEXEC;
+ fcntl (fd, F_SETFD, flags);
+}
+
+static void close_client (struct ev_tcp *client)
+{
+ close (client->io.fd);
+ if (client_chain == client)
+ client_chain = client->next;
+ if (client->next)
+ client->next->prev = client->prev;
+ if (client->prev)
+ client->prev->next = client->next;
+ free (client);
+}
+
+static void client_message (struct ev_tcp *io, unsigned int length, unsigned char *header)
+{
+ unsigned char ch;
+ uint32_t magic, type, mlen, seq;
+ int hver, mver;
+ int i;
+
+ if (AUDIT_RMW_IS_MAGIC (header, length)) {
+ AUDIT_RMW_UNPACK_HEADER (header, hver, mver, type, mlen, seq)
+
+ ch = header[length];
+ header[length] = 0;
+ if (length > 1 && header[length-1] == '\n')
+ header[length-1] = 0;
+ enqueue_formatted_event (header+AUDIT_RMW_HEADER_SIZE, io->io.fd, seq);
+ header[length] = ch;
+ } else {
+ header[length] = 0;
+ if (length > 1 && header[length-1] == '\n')
+ header[length-1] = 0;
+ enqueue_formatted_event (header, 0, 0);
+ }
+}
+
+static void auditd_tcp_client_handler( struct ev_loop *loop, struct ev_io *_io, int revents )
+{
+ struct ev_tcp *io = (struct ev_tcp *) _io;
+ int i, r;
+ int total_this_call = 0;
+
+ /* The socket is non-blocking, but we have a limited buffer
+ size. In the event that we get a packet that's bigger than
+ our buffer, we need to read it in multiple parts. Thus, we
+ keep reading/parsing/processing until we run out of ready
+ data. */
+read_more:
+ r = read (io->io.fd,
+ io->buffer + io->bufptr,
+ MAX_AUDIT_MESSAGE_LENGTH - io->bufptr);
+
+ if (r < 0 && errno == EAGAIN)
+ r = 0;
+
+ /* We need to keep track of the difference between "no data
+ * because it's closed" and "no data because we've read it
+ * all". */
+ if (r == 0 && total_this_call > 0) {
+ return;
+ }
+
+ /* If the connection is gracefully closed, the first read we
+ try will return zero. If the connection times out or
+ otherwise fails, the read will return -1. */
+ if (r <= 0) {
+ if (r < 0)
+ audit_msg (LOG_WARNING, "client %s socket closed unexpectedly",
+ sockaddr_to_ip (&io->addr));
+
+ /* There may have been a final message without a LF. */
+ if (io->bufptr) {
+ client_message (io, io->bufptr, io->buffer);
+
+ }
+
+ ev_io_stop (loop, _io);
+ close_client (io);
+ return;
+ }
+
+ total_this_call += r;
+
+more_messages:
+ if (AUDIT_RMW_IS_MAGIC (io->buffer, io->bufptr+r)) {
+ uint32_t type, len, seq;
+ int hver, mver;
+ unsigned char *header = (unsigned char *)io->buffer;
+
+ io->bufptr += r;
+
+ if (io->bufptr < AUDIT_RMW_HEADER_SIZE)
+ return;
+
+ AUDIT_RMW_UNPACK_HEADER (header, hver, mver, type, len, seq);
+
+ i = len;
+ i += AUDIT_RMW_HEADER_SIZE;
+
+ /* See if we have enough bytes to extract the whole message. */
+ if (io->bufptr < i)
+ return;
+
+ } else {
+ /* At this point, the buffer has IO->BUFPTR+R bytes in it.
+ The first IO->BUFPTR bytes do not have a LF in them (we've
+ already checked), we must check the R new bytes. */
+
+ for (i = io->bufptr; i < io->bufptr + r; i ++)
+ if (io->buffer [i] == '\n')
+ break;
+
+ io->bufptr += r;
+
+ /* Check for a partial message, with no LF yet. */
+ if (i == io->bufptr)
+ return;
+
+ i ++;
+ }
+
+ /* We have an I-byte message in buffer. */
+ client_message (io, i, io->buffer);
+
+ /* Now copy any remaining bytes to the beginning of the
+ buffer. */
+ memmove (io->buffer, io->buffer + i, io->bufptr);
+ io->bufptr -= i;
+
+ /* See if this packet had more than one message in it. */
+ if (io->bufptr > 0) {
+ r = io->bufptr;
+ io->bufptr = 0;
+ goto more_messages;
+ }
+
+ /* Go back and see if there's more data to read. */
+ goto read_more;
+}
+
+static void auditd_tcp_listen_handler( struct ev_loop *loop, struct ev_io *_io, int revents )
+{
+ int one=1;
+ int afd;
+ socklen_t aaddrlen;
+ struct sockaddr_in aaddr;
+ struct linger li;
+ struct ev_tcp *client;
+ unsigned char *uaddr;
+
+ /* Accept the connection and see where it's coming from. */
+ aaddrlen = sizeof(aaddr);
+ afd = accept (listen_socket, (struct sockaddr *)&aaddr, &aaddrlen);
+ if (afd == -1) {
+ audit_msg(LOG_ERR, "Unable to accept TCP connection");
+ return;
+ }
+
+ uaddr = (unsigned char *)&aaddr.sin_addr;
+
+ /* Verify it's coming from an authorized port. We assume the firewall will
+ block attempts from unauthorized machines. */
+ if (min_port > ntohs (aaddr.sin_port) || ntohs (aaddr.sin_port) > max_port) {
+ audit_msg(LOG_ERR, "TCP connection from %s rejected", sockaddr_to_ip (&aaddr));
+ close (afd);
+ return;
+ }
+
+ setsockopt(afd, SOL_SOCKET, SO_REUSEADDR, (char *)&one, sizeof (int));
+ setsockopt(afd, SOL_SOCKET, SO_KEEPALIVE, (char *)&one, sizeof (int));
+ setsockopt(afd, IPPROTO_TCP, TCP_NODELAY, (char *)&one, sizeof (int));
+ fcntl(afd, F_SETFL, O_NONBLOCK | O_NDELAY);
+ set_close_on_exec (afd);
+
+ client = (struct ev_tcp *) malloc (sizeof (struct ev_tcp));
+ if (client == NULL) {
+ audit_msg(LOG_CRIT, "Unable to allocate TCP client data");
+ close (afd);
+ return;
+ }
+
+ memset (client, 0, sizeof (struct ev_tcp));
+
+ ev_io_init (&(client->io), auditd_tcp_client_handler, afd, EV_READ | EV_ERROR);
+ ev_io_start (loop, &(client->io));
+
+ memcpy (&client->addr, &aaddr, sizeof (struct sockaddr_in));
+
+ /* Keep a linked list of active clients. */
+ client->next = client_chain;
+ if (client->next)
+ client->next->prev = client;
+ client_chain = client;
+}
+
+int auditd_tcp_listen_init ( struct ev_loop *loop, struct daemon_conf *config )
+{
+ struct sockaddr_in address;
+ int one = 1;
+ int flags;
+
+ /* If the port is not set, that means we aren't going to
+ listen for connections. */
+ if (config->tcp_listen_port == 0)
+ return;
+
+ listen_socket = socket (AF_INET, SOCK_STREAM, 0);
+ if (listen_socket == 0) {
+ audit_msg(LOG_ERR, "Cannot create tcp listener socket");
+ return 1;
+ }
+
+ set_close_on_exec (listen_socket);
+ setsockopt(listen_socket, SOL_SOCKET, SO_REUSEADDR, (char *)&one, sizeof (int));
+
+ memset (&address, 0, sizeof(address));
+ address.sin_family = htons(AF_INET);
+ address.sin_port = htons(config->tcp_listen_port);
+ address.sin_addr.s_addr = htonl(INADDR_ANY);
+
+ /* This avoids problems if auditd needs to be restarted. */
+ setsockopt(listen_socket, SOL_SOCKET, SO_REUSEADDR, (char *)&one, sizeof (int));
+
+ if ( bind ( listen_socket, (struct sockaddr *)&address, sizeof(address)) ) {
+ audit_msg(LOG_ERR, "Cannot bind tcp listener socket to port %d",
+ config->tcp_listen_port);
+ close(listen_socket);
+ return 1;
+ }
+
+ listen(listen_socket, config->tcp_listen_queue);
+
+ audit_msg(LOG_DEBUG, "Listening on TCP port %d", config->tcp_listen_port);
+
+ ev_io_init (&tcp_listen_watcher, auditd_tcp_listen_handler, listen_socket, EV_READ);
+ ev_io_start (loop, &tcp_listen_watcher);
+
+ min_port = config->tcp_client_min_port;
+ max_port = config->tcp_client_max_port;
+
+ return 0;
+}
+
+void auditd_tcp_listen_uninit ( struct ev_loop *loop )
+{
+ ev_io_stop ( loop, &tcp_listen_watcher );
+ close ( listen_socket );
+
+ while (client_chain) {
+ close_client (client_chain);
+ }
+}
diff -N -U 3 -x .svn -r pristine/src/auditd-listen.h trunk/src/auditd-listen.h
--- pristine/src/auditd-listen.h 1969-12-31 19:00:00.000000000 -0500
+++ trunk/src/auditd-listen.h 2008-08-08 17:01:37.000000000 -0400
@@ -0,0 +1,30 @@
+/* auditd-config.h --
+ * Copyright 2004-2007 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
+ *
+ * Authors:
+ * DJ Delorie <dj(a)redhat.com>
+ *
+ */
+
+#ifndef AUDITD_LISTEN_H
+#define AUDITD_LISTEN_H
+
+int auditd_tcp_listen_init ( struct ev_loop *loop, struct daemon_conf *config );
+void auditd_tcp_listen_uninit ( struct ev_loop *loop );
+
+#endif
diff -N -U 3 -x .svn -r pristine/src/auditd-reconfig.c trunk/src/auditd-reconfig.c
--- pristine/src/auditd-reconfig.c 2008-08-04 12:47:30.000000000 -0400
+++ trunk/src/auditd-reconfig.c 2008-08-12 18:54:20.000000000 -0400
@@ -28,9 +28,9 @@
#include <unistd.h>
#include <string.h>
#include <stdio.h>
-#include "auditd-config.h"
-#include "auditd-event.h"
#include "libaudit.h"
+#include "auditd-event.h"
+#include "auditd-config.h"
#include "private.h"
/* This is the configuration manager code */
diff -N -U 3 -x .svn -r pristine/src/auditd.c trunk/src/auditd.c
--- pristine/src/auditd.c 2008-08-08 14:36:19.000000000 -0400
+++ trunk/src/auditd.c 2008-08-13 20:51:35.000000000 -0400
@@ -40,8 +40,8 @@
#include <getopt.h>
#include "libaudit.h"
-#include "auditd-config.h"
#include "auditd-event.h"
+#include "auditd-config.h"
#include "auditd-dispatch.h"
#include "private.h"
@@ -670,7 +670,15 @@
ev_signal_init (&sigchld_watcher, child_handler, SIGCHLD);
ev_signal_start (loop, &sigchld_watcher);
- ev_loop (loop, 0);
+ if (auditd_tcp_listen_init (loop, &config)) {
+ tell_parent (FAILURE);
+ stop = 1;
+ }
+
+ if (!stop)
+ ev_loop (loop, 0);
+
+ auditd_tcp_listen_uninit (loop);
/* Write message to log that we are going down */
15 years, 2 months
[RFC] Drop support for old audit rule API
by Steve Grubb
Hi,
Another thought for the next major release is to drop support for the old
audit rules API in libaudit. A long time ago, the function prototypes for:
extern int audit_request_rules_list(int fd);
extern int audit_add_rule(int fd, struct audit_rule *rule,
int flags, int action);
extern int audit_delete_rule(int fd, struct audit_rule *rule,
int flags, int action);
extern int audit_rule_syscall(struct audit_rule *rule, int scall);
extern int audit_rule_syscallbyname(struct audit_rule *rule,
const char *scall);
extern int audit_rule_fieldpair(struct audit_rule *rule, const char *pair,
int flags);
extern void audit_rule_free(struct audit_rule *rule);
Were moved to a private internal header. This was to encourage any external
apps from using them, but if they had been previously compiled to know about
the API, they would still be able to link and run.
As far as I know, auditctl & autrace are the only apps that would be affected.
There very well may be some 3rd party apps I don't know about that this could
impact. Hopefully they are on this list or moved their function calls when
the prototype disappeared.
The proposal is to drop the above mentioned functions and bump the soname
number so that anything linking against libaudit has to be recompiled to make
sure they have no old function calls. At some point in the future, we can
also remove the kernel's support for the old rule format.
Let's discuss....
-Steve
15 years, 4 months
[RFC] Do away with entry filter
by Steve Grubb
Hello,
I will be forking the user space audit code soon to start the next major
series. I have a couple thoughts I'd like to share with people to see what
they think.
The first item is doing away with the entry filter for syscall auditing. You
normally run across this filter when you write rules such as:
-a always,entry -S open
The reason I think we can do away with it is that its purpose has changed. Way
back in the early days 2.6.6 -> 2.6.15 kernels, there was this notion that
the audit code could be made to have little impact on the performance of the
system if we give hints about what is needed by using "possible" actions.
The problem with "possible" was that people forgot to use it and had exit
filter rules that had no data to operate on. So, we changed the kernel to
always collect the data it needed in case an exit filter would trigger an
event. This was optimized and performance was pretty good. So, that kind of
left the entry filter without a purpose.
Any entry rule can be written as an exit rule. But not every exit rule can be
written as an entry rule. So the logical choice is to consolidate on the exit
filter. The reason to do this is to improve performance. If we have an entry
rule that triggers, it marks the syscall excursion as auditable. When we get
to the exit filter, it iterates over the whole set of rules even though the
event is auditable. This is because there could be a never rule that would
suppress the output. Another problem introduced by having two filters is that
some fields are not available in the entry filter (exit for example), it adds
complexity in the auditctl program and the in-kernel rule parser to look for
these errors.
The way that we could make the change is for the audit package to silently
convert entry rules to exit in user space. It could output a warning that
entry rules are being converted and the admin should make the necessary
adjustments. Then after some time has elapsed so that distros have all
updated, drop support in the kernel for the entry filter.
Let's discuss...
Thanks,
-Steve
15 years, 4 months
exclude rule help
by LC Bruzenak
I'm trying to create an audit rule to exclude all rsync actions when run
by a process with a certain subject type (jcdx_fsbackup_t).
Our policy allows the action to happen, however as a result we generate
a ton of audit data on file access and recreation at the target point.
Anyone have a good idea of how to discard all these events? Ideally the
caller would send in a self-generated event such as "ryncing rick/src2/
to /temp-home" or similar. This is for a dedicated file backup
procedure.
Obviously I do not want to discard all rsync events, just when launched
by our trusted program. Nor would I really want all that program's
events discarded since I want it to be able to submit proactive events
which summarize its behavior.
Below are some event samples. Other ideas are welcome also if my
approach is myopic/flawed.
Thx,
LCB.
----
node=develop.local.austin.rr.com type=PATH msg=audit(06/22/2009 15:05:37.017:901990) : item=0 name=rick/path1/path2/path3/k4jtCX inode=203842 dev=fd:04 mode=file,600 ouid=root ogid=unknown(1106) rdev=00:00 obj=siterep_u:object_r:user_home_t:s0
node=develop.local.austin.rr.com type=CWD msg=audit(06/22/2009 15:05:37.017:901990) : cwd=/temp_home
node=develop.local.austin.rr.com type=SYSCALL msg=audit(06/22/2009 15:05:37.017:901990) : arch=x86_64 syscall=lchown success=yes exit=0 a0=7ffff2994e30 a1=1f7 a2=44c a3=1 items=1 ppid=5445 pid=5446 auid=rick uid=root gid=unknown(1106) euid=root suid=root fsuid=root egid=unknown(1106) sgid=unknown(1106) fsgid=unknown(1106) tty=pts2 ses=1 comm=rsync exe=/usr/bin/rsync subj=siterep_u:siterep_r:dx_fsbackup_t:s15:c0.c1023 key=perm_mod
...
----
node=develop.local.austin.rr.com type=PATH msg=audit(06/22/2009 15:05:37.017:901997) : item=3 name=rick/path1/path2/path3/f1 inode=203843 dev=fd:04 mode=file,444 ouid=unknown(503) ogid=unknown(1100) rdev=00:00 obj=system_u:object_r:user_home_t:s0
node=develop.local.austin.rr.com type=PATH msg=audit(06/22/2009 15:05:37.017:901997) : item=2 name=rick/path1/path2/path3/f2 inode=203843 dev=fd:04 mode=file,444 ouid=unknown(503) ogid=unknown(1100) rdev=00:00 obj=system_u:object_r:user_home_t:s0
node=develop.local.austin.rr.com type=PATH msg=audit(06/22/2009 15:05:37.017:901997) : item=1 name=rick/path1/path2/path3/d1 inode=203324 dev=fd:04 mode=dir,755 ouid=unknown(503) ogid=unknown(1100) rdev=00:00 obj=system_u:object_r:user_home_t:s0
node=develop.local.austin.rr.com type=PATH msg=audit(06/22/2009 15:05:37.017:901997) : item=0 name=rick/path1/path2/path3/d2 inode=203324 dev=fd:04 mode=dir,755 ouid=unknown(503) ogid=unknown(1100) rdev=00:00 obj=system_u:object_r:user_home_t:s0
node=develop.local.austin.rr.com type=CWD msg=audit(06/22/2009 15:05:37.017:901997) : cwd=/temp_home
node=develop.local.austin.rr.com type=SYSCALL msg=audit(06/22/2009 15:05:37.017:901997) : arch=x86_64 syscall=rename success=yes exit=0 a0=7ffff2994e30 a1=7ffff2996e30 a2=0 a3=0 items=4 ppid=5445 pid=5446 auid=rick uid=root gid=unknown(1106) euid=root suid=root fsuid=root egid=unknown(1106) sgid=unknown(1106) fsgid=unknown(1106) tty=pts2 ses=1 comm=rsync exe=/usr/bin/rsync subj=siterep_u:siterep_r:dx_fsbackup_t:s15:c0.c1023 key=delete
--
LC (Lenny) Bruzenak
lenny(a)magitekltd.com
15 years, 5 months
Gavin Appleton is out of the office.
by Gavin Appleton
I will be out of the office starting 22/06/2009 and will not return until
29/06/2009.
I will respond to your message when I return.
15 years, 5 months
[PATCH] audit: inode watches depend on CONFIG_AUDIT not CONFIG_AUDIT_SYSCALL
by Eric Paris
Even though one cannot make use of the audit watch code without
CONFIG_AUDIT_SYSCALL the spaghetti nature of the audit code means that
the audit rule filtering requires that it at least be compiled.
Thus build the audit_watch code when we build auditfilter like it was
before cfcad62c74abfef83762dc05a556d21bdf3980a2
Clearly this is a point of potential future cleanup..
Reported-by: Frans Pop <elendil(a)planet.nl>
Signed-off-by: Eric Paris <eparis(a)redhat.com>
---
kernel/Makefile | 4 ++--
1 files changed, 2 insertions(+), 2 deletions(-)
diff --git a/kernel/Makefile b/kernel/Makefile
index da75001..780c8dc 100644
--- a/kernel/Makefile
+++ b/kernel/Makefile
@@ -69,8 +69,8 @@ obj-$(CONFIG_IKCONFIG) += configs.o
obj-$(CONFIG_RESOURCE_COUNTERS) += res_counter.o
obj-$(CONFIG_STOP_MACHINE) += stop_machine.o
obj-$(CONFIG_KPROBES_SANITY_TEST) += test_kprobes.o
-obj-$(CONFIG_AUDIT) += audit.o auditfilter.o
-obj-$(CONFIG_AUDITSYSCALL) += auditsc.o audit_watch.o
+obj-$(CONFIG_AUDIT) += audit.o auditfilter.o audit_watch.o
+obj-$(CONFIG_AUDITSYSCALL) += auditsc.o
obj-$(CONFIG_GCOV_KERNEL) += gcov/
obj-$(CONFIG_AUDIT_TREE) += audit_tree.o
obj-$(CONFIG_KPROBES) += kprobes.o
15 years, 5 months
[PATCH 1/7] audit: convert audit watches to use fsnotify instead of inotify
by Eric Paris
Audit currently uses inotify to pin inodes in core and to detect when
watched inodes are deleted or unmounted. This patch uses fsnotify instead
of inotify.
Signed-off-by: Eric Paris <eparis(a)redhat.com>
---
include/linux/fsnotify_backend.h | 5 +
kernel/audit_watch.c | 209 +++++++++++++++++++++++++++-----------
2 files changed, 153 insertions(+), 61 deletions(-)
diff --git a/include/linux/fsnotify_backend.h b/include/linux/fsnotify_backend.h
index 44848aa..90e32cc 100644
--- a/include/linux/fsnotify_backend.h
+++ b/include/linux/fsnotify_backend.h
@@ -58,9 +58,12 @@
FS_MOVED_FROM | FS_MOVED_TO | FS_CREATE |\
FS_DELETE)
+#define FS_MOVE (FS_MOVED_FROM | FS_MOVED_TO)
+
/* listeners that hard code group numbers near the top */
#define DNOTIFY_GROUP_NUM UINT_MAX
-#define INOTIFY_GROUP_NUM (DNOTIFY_GROUP_NUM-1)
+#define AUDIT_WATCH_GROUP_NUM (DNOTIFY_GROUP_NUM-1)
+#define INOTIFY_GROUP_NUM (AUDIT_WATCH_GROUP_NUM-1)
struct fsnotify_group;
struct fsnotify_event;
diff --git a/kernel/audit_watch.c b/kernel/audit_watch.c
index 0c179fb..8d7f264 100644
--- a/kernel/audit_watch.c
+++ b/kernel/audit_watch.c
@@ -24,17 +24,17 @@
#include <linux/kthread.h>
#include <linux/mutex.h>
#include <linux/fs.h>
+#include <linux/fsnotify_backend.h>
#include <linux/namei.h>
#include <linux/netlink.h>
#include <linux/sched.h>
-#include <linux/inotify.h>
#include <linux/security.h>
#include "audit.h"
/*
* Reference counting:
*
- * audit_parent: lifetime is from audit_init_parent() to receipt of an IN_IGNORED
+ * audit_parent: lifetime is from audit_init_parent() to receipt of an FS_IGNORED
* event. Each audit_watch holds a reference to its associated parent.
*
* audit_watch: if added to lists, lifetime is from audit_init_watch() to
@@ -56,29 +56,30 @@ struct audit_watch {
struct audit_parent {
struct list_head ilist; /* tmp list used to free parents */
struct list_head watches; /* anchor for audit_watch->wlist */
- struct inotify_watch wdata; /* inotify watch data */
+ struct fsnotify_mark_entry mark; /* fsnotify mark on the inode */
unsigned flags; /* status flags */
};
static struct kmem_cache *audit_watch_cache __read_mostly;
static struct kmem_cache *audit_parent_cache __read_mostly;
-/* Inotify handle. */
-struct inotify_handle *audit_ih;
+/* fsnotify handle. */
+struct fsnotify_group *audit_watch_group;
/*
* audit_parent status flags:
*
* AUDIT_PARENT_INVALID - set anytime rules/watches are auto-removed due to
* a filesystem event to ensure we're adding audit watches to a valid parent.
- * Technically not needed for IN_DELETE_SELF or IN_UNMOUNT events, as we cannot
- * receive them while we have nameidata, but must be used for IN_MOVE_SELF which
+ * Technically not needed for FS_DELETE_SELF or FS_UNMOUNT events, as we cannot
+ * receive them while we have nameidata, but must be used for FS_MOVE_SELF which
* we can receive while holding nameidata.
*/
#define AUDIT_PARENT_INVALID 0x001
-/* Inotify events we care about. */
-#define AUDIT_IN_WATCH IN_MOVE|IN_CREATE|IN_DELETE|IN_DELETE_SELF|IN_MOVE_SELF
+/* fsnotify events we care about. */
+#define AUDIT_FS_WATCH (FS_MOVE | FS_CREATE | FS_DELETE | FS_DELETE_SELF |\
+ FS_MOVE_SELF | FS_EVENT_ON_CHILD)
static void audit_free_parent(struct audit_parent *parent)
{
@@ -86,14 +87,45 @@ static void audit_free_parent(struct audit_parent *parent)
kmem_cache_free(audit_parent_cache, parent);
}
-static void audit_destroy_watch(struct inotify_watch *i_watch)
+static void audit_watch_free_mark(struct fsnotify_mark_entry *entry)
{
struct audit_parent *parent;
- parent = container_of(i_watch, struct audit_parent, wdata);
+ parent = container_of(entry, struct audit_parent, mark);
audit_free_parent(parent);
}
+static void audit_get_parent(struct audit_parent *parent)
+{
+ if (likely(parent))
+ fsnotify_get_mark(&parent->mark);
+}
+
+static void audit_put_parent(struct audit_parent *parent)
+{
+ if (likely(parent))
+ fsnotify_put_mark(&parent->mark);
+}
+
+/*
+ * Find and return the audit_parent on the given inode. If found a reference
+ * is taken on this parent.
+ */
+static inline struct audit_parent *audit_find_parent(struct inode *inode)
+{
+ struct audit_parent *parent = NULL;
+ struct fsnotify_mark_entry *entry;
+
+ spin_lock(&inode->i_lock);
+ entry = fsnotify_find_mark_entry(audit_watch_group, inode);
+ spin_unlock(&inode->i_lock);
+
+ if (entry)
+ parent = container_of(entry, struct audit_parent, mark);
+
+ return parent;
+}
+
void audit_get_watch(struct audit_watch *watch)
{
atomic_inc(&watch->count);
@@ -112,7 +144,7 @@ void audit_put_watch(struct audit_watch *watch)
void audit_remove_watch(struct audit_watch *watch)
{
list_del(&watch->wlist);
- put_inotify_watch(&watch->parent->wdata);
+ audit_put_parent(watch->parent);
watch->parent = NULL;
audit_put_watch(watch); /* match initial get */
}
@@ -132,8 +164,9 @@ int audit_watch_compare(struct audit_watch *watch, unsigned long ino, dev_t dev)
/* Initialize a parent watch entry. */
static struct audit_parent *audit_init_parent(struct path *parent_path)
{
+ struct inode *inode = parent_path->dentry->d_inode;
struct audit_parent *parent;
- s32 wd;
+ int ret;
parent = kmem_cache_zalloc(audit_parent_cache, GFP_KERNEL);
if (unlikely(!parent))
@@ -142,14 +175,14 @@ static struct audit_parent *audit_init_parent(struct path *parent_path)
INIT_LIST_HEAD(&parent->watches);
parent->flags = 0;
- inotify_init_watch(&parent->wdata);
- /* grab a ref so inotify watch hangs around until we take audit_filter_mutex */
- get_inotify_watch(&parent->wdata);
- wd = inotify_add_watch(audit_ih, &parent->wdata,
- parent_path->dentry->d_inode, AUDIT_IN_WATCH);
- if (wd < 0) {
+ fsnotify_init_mark(&parent->mark, audit_watch_free_mark);
+ parent->mark.mask = AUDIT_FS_WATCH;
+ /* grab a ref so fsnotify mark hangs around until we take audit_filter_mutex */
+ audit_get_parent(parent);
+ ret = fsnotify_add_mark(&parent->mark, audit_watch_group, inode);
+ if (ret < 0) {
audit_free_parent(parent);
- return ERR_PTR(wd);
+ return ERR_PTR(ret);
}
return parent;
@@ -178,7 +211,7 @@ int audit_to_watch(struct audit_krule *krule, char *path, int len, u32 op)
{
struct audit_watch *watch;
- if (!audit_ih)
+ if (!audit_watch_group)
return -EOPNOTSUPP;
if (path[0] != '/' || path[len-1] == '/' ||
@@ -216,7 +249,7 @@ static struct audit_watch *audit_dupe_watch(struct audit_watch *old)
new->dev = old->dev;
new->ino = old->ino;
- get_inotify_watch(&old->parent->wdata);
+ audit_get_parent(old->parent);
new->parent = old->parent;
out:
@@ -337,19 +370,21 @@ static void audit_remove_parent_watches(struct audit_parent *parent)
audit_remove_watch(w);
}
mutex_unlock(&audit_filter_mutex);
+
+ fsnotify_destroy_mark_by_entry(&parent->mark);
}
/* Unregister inotify watches for parents on in_list.
- * Generates an IN_IGNORED event. */
+ * Generates an FS_IGNORED event. */
void audit_watch_inotify_unregister(struct list_head *in_list)
{
struct audit_parent *p, *n;
list_for_each_entry_safe(p, n, in_list, ilist) {
list_del(&p->ilist);
- inotify_rm_watch(audit_ih, &p->wdata);
- /* the unpin matching the pin in audit_remove_watch_rule() */
- unpin_inotify_watch(&p->wdata);
+ fsnotify_destroy_mark_by_entry(&p->mark);
+ /* matches the get in audit_remove_watch_rule() */
+ audit_put_parent(p);
}
}
@@ -383,7 +418,7 @@ static void audit_put_path(struct path *parent_path, struct path *watch_path)
path_put(watch_path);
}
-/* Associate the given rule with an existing parent inotify_watch.
+/* Associate the given rule with an existing parent.
* Caller must hold audit_filter_mutex. */
static void audit_add_to_parent(struct audit_krule *krule,
struct audit_parent *parent)
@@ -391,6 +426,8 @@ static void audit_add_to_parent(struct audit_krule *krule,
struct audit_watch *w, *watch = krule->watch;
int watch_found = 0;
+ BUG_ON(!mutex_is_locked(&audit_filter_mutex));
+
list_for_each_entry(w, &parent->watches, wlist) {
if (strcmp(watch->path, w->path))
continue;
@@ -407,7 +444,7 @@ static void audit_add_to_parent(struct audit_krule *krule,
}
if (!watch_found) {
- get_inotify_watch(&parent->wdata);
+ audit_get_parent(parent);
watch->parent = parent;
list_add(&watch->wlist, &parent->watches);
@@ -420,7 +457,6 @@ static void audit_add_to_parent(struct audit_krule *krule,
int audit_add_watch(struct audit_krule *krule, struct list_head **list)
{
struct audit_watch *watch = krule->watch;
- struct inotify_watch *i_watch;
struct audit_parent *parent;
struct path parent_path, watch_path;
int h, ret = 0;
@@ -446,8 +482,8 @@ int audit_add_watch(struct audit_krule *krule, struct list_head **list)
* inotify watch is found, inotify_find_watch() grabs a reference before
* returning.
*/
- if (inotify_find_watch(audit_ih, parent_path.dentry->d_inode,
- &i_watch) < 0) {
+ parent = audit_find_parent(parent_path.dentry->d_inode);
+ if (!parent) {
parent = audit_init_parent(&parent_path);
if (IS_ERR(parent)) {
/* caller expects mutex locked */
@@ -455,8 +491,7 @@ int audit_add_watch(struct audit_krule *krule, struct list_head **list)
ret = PTR_ERR(parent);
goto error;
}
- } else
- parent = container_of(i_watch, struct audit_parent, wdata);
+ }
mutex_lock(&audit_filter_mutex);
@@ -466,8 +501,8 @@ int audit_add_watch(struct audit_krule *krule, struct list_head **list)
else
audit_add_to_parent(krule, parent);
- /* match get in audit_init_parent or inotify_find_watch */
- put_inotify_watch(&parent->wdata);
+ /* match get in audit_find_parent or audit_init_parent */
+ audit_put_parent(parent);
h = audit_hash_ino((u32)watch->ino);
*list = &audit_inode_hash[h];
@@ -488,54 +523,108 @@ void audit_remove_watch_rule(struct audit_krule *krule, struct list_head *list)
audit_remove_watch(watch);
if (list_empty(&parent->watches)) {
- /* Put parent on the inotify un-registration
- * list. Grab a reference before releasing
+ /* Put parent on the un-registration list.
+ * Grab a reference before releasing
* audit_filter_mutex, to be released in
- * audit_inotify_unregister().
+ * audit_watch_inotify_unregister().
* If filesystem is going away, just leave
* the sucker alone, eviction will take
* care of it. */
- if (pin_inotify_watch(&parent->wdata))
- list_add(&parent->ilist, list);
+ audit_get_parent(parent);
+ list_add(&parent->ilist, list);
}
}
}
-/* Update watch data in audit rules based on inotify events. */
-static void audit_handle_ievent(struct inotify_watch *i_watch, u32 wd, u32 mask,
- u32 cookie, const char *dname, struct inode *inode)
+static bool audit_watch_should_send_event(struct fsnotify_group *group, struct inode *inode, __u32 mask)
+{
+ struct fsnotify_mark_entry *entry;
+ bool send;
+
+ spin_lock(&inode->i_lock);
+ entry = fsnotify_find_mark_entry(group, inode);
+ spin_unlock(&inode->i_lock);
+ if (!entry)
+ return false;
+
+ mask = (mask & ~FS_EVENT_ON_CHILD);
+ send = (entry->mask & mask);
+
+ /* find took a reference */
+ fsnotify_put_mark(entry);
+
+ return send;
+}
+
+/* Update watch data in audit rules based on fsnotify events. */
+static int audit_watch_handle_event(struct fsnotify_group *group, struct fsnotify_event *event)
{
+ struct inode *inode;
+ __u32 mask = event->mask;
+ const char *dname = event->file_name;
struct audit_parent *parent;
- parent = container_of(i_watch, struct audit_parent, wdata);
+ BUG_ON(group != audit_watch_group);
+
+ parent = audit_find_parent(event->to_tell);
+ if (unlikely(!parent))
+ return 0;
- if (mask & (IN_CREATE|IN_MOVED_TO) && inode)
+ switch (event->data_type) {
+ case (FSNOTIFY_EVENT_PATH):
+ inode = event->path.dentry->d_inode;
+ break;
+ case (FSNOTIFY_EVENT_INODE):
+ inode = event->inode;
+ break;
+ default:
+ BUG();
+ inode = NULL;
+ break;
+ };
+
+ if (mask & (FS_CREATE|FS_MOVED_TO) && inode)
audit_update_watch(parent, dname, inode->i_sb->s_dev, inode->i_ino, 0);
- else if (mask & (IN_DELETE|IN_MOVED_FROM))
+ else if (mask & (FS_DELETE|FS_MOVED_FROM))
audit_update_watch(parent, dname, (dev_t)-1, (unsigned long)-1, 1);
- /* inotify automatically removes the watch and sends IN_IGNORED */
- else if (mask & (IN_DELETE_SELF|IN_UNMOUNT))
- audit_remove_parent_watches(parent);
- /* inotify does not remove the watch, so remove it manually */
- else if(mask & IN_MOVE_SELF) {
+ else if (mask & (FS_DELETE_SELF|FS_UNMOUNT|FS_MOVE_SELF))
audit_remove_parent_watches(parent);
- inotify_remove_watch_locked(audit_ih, i_watch);
- } else if (mask & IN_IGNORED)
- put_inotify_watch(i_watch);
+ /* moved put_inotify_watch to freeing mark */
+
+ /* matched the ref taken by audit_find_parent */
+ audit_put_parent(parent);
+
+ return 0;
+}
+
+static void audit_watch_freeing_mark(struct fsnotify_mark_entry *entry, struct fsnotify_group *group)
+{
+ struct audit_parent *parent;
+
+ parent = container_of(entry, struct audit_parent, mark);
+ /* taken from audit_handle_ievent & FS_IGNORED please figure out what I match... */
+ audit_put_parent(parent);
}
-static const struct inotify_operations audit_inotify_ops = {
- .handle_event = audit_handle_ievent,
- .destroy_watch = audit_destroy_watch,
+static const struct fsnotify_ops audit_watch_fsnotify_ops = {
+ .should_send_event = audit_watch_should_send_event,
+ .handle_event = audit_watch_handle_event,
+ .free_group_priv = NULL,
+ .freeing_mark = audit_watch_freeing_mark,
+ .free_event_priv = NULL,
};
static int __init audit_watch_init(void)
{
audit_watch_cache = KMEM_CACHE(audit_watch, SLAB_PANIC);
audit_parent_cache = KMEM_CACHE(audit_parent, SLAB_PANIC);
- audit_ih = inotify_init(&audit_inotify_ops);
- if (IS_ERR(audit_ih))
- audit_panic("cannot initialize inotify handle");
+
+ audit_watch_group = fsnotify_obtain_group(AUDIT_WATCH_GROUP_NUM, AUDIT_FS_WATCH,
+ &audit_watch_fsnotify_ops);
+ if (IS_ERR(audit_watch_group)) {
+ audit_watch_group = NULL;
+ audit_panic("cannot create audit fsnotify group");
+ }
return 0;
}
subsys_initcall(audit_watch_init);
15 years, 5 months