Skip to contentSkip to conversation

ausearch on aggregation - syscall difference

DISCUSSION LC Bruzenak 3 participants 7 messages Oct 24, 2008

lore
Share & tools
lore
#0 lore raw

I have a test (virtual) machine running a 32-bit F9 OS. My aggregating machine is a 64-bit F9 box.

source (32-bit machine) :

[root@v1 ~]# ausearch -ts today -i -a 10038
----
node=v1 type=SYSCALL msg=audit(10/24/2008 11:11:59.162:10038) : arch=i386 syscall=socketcall(recv) success=yes exit=5 a0=a a1=bfcc1f80 a2=25b0c4 a3=0 items=0 ppid=1 pid=11761 auid=root uid=root gid=root euid=root suid=root fsuid=root egid=root sgid=root fsgid=root tty=(none) ses=1 comm=prelude-manager exe=/usr/bin/prelude-manager subj=system_u:system_r:prelude_t:s0-s15:c0.c1023 key=(null) node=v1 type=AVC msg=audit(10/24/2008 11:11:59.162:10038) : avc: denied { read } for pid=11761 comm=prelude-manager laddr=127.0.0.1 lport=4690 faddr=127.0.0.1 fport=36291 scontext=system_u:system_r:prelude_t:s0-s15:c0.c1023 tcontext=system_u:system_r:prelude_t:s15:c0.c1023 tclass=tcp_socket

aggregating machine (64-bit) :

[root@dell1 ~]# ausearch -ts today -i -a 10038
----
node=v1 type=SYSCALL msg=audit(10/24/2008 11:11:59.162:10038) : arch=i386 syscall=getuid success=yes exit=5 a0=a a1=bfcc1f80 a2=25b0c4 a3=0 items=0 ppid=1 pid=11761 auid=root uid=root gid=root euid=root suid=root fsuid=root egid=root sgid=root fsgid=root tty=(none) ses=1 comm=prelude-manager exe=/usr/bin/prelude-manager subj=system_u:system_r:prelude_t:s0-s15:c0.c1023 key=(null) node=v1 type=AVC msg=audit(10/24/2008 11:11:59.162:10038) : avc: denied { read } for pid=11761 comm=prelude-manager laddr=127.0.0.1 lport=4690 faddr=127.0.0.1 fport=36291 scontext=system_u:system_r:prelude_t:s0-s15:c0.c1023 tcontext=system_u:system_r:prelude_t:s15:c0.c1023 tclass=tcp_socket

Note that the syscall is listed differently.
This is using the 1.7.7 code (on F9), I have not yet moved over to 1.7.8 in case it may be fixed there.

Also, (at the latest) after F10 GA release I'll be migrating there.

Thx,
LCB.

Signature · 2 lines
LC (Lenny) Bruzenak
[email protected]
#1 lore raw
Quoted 1 line · On Friday 24 October 2008 13:08:41 LC Bruzenak wrote:

Note that the syscall is listed differently.

Interesting.

Quoted 2 lines

This is using the 1.7.7 code (on F9), I have not yet moved over to 1.7.8 in case it may be fixed there.

Nope...nothing was changed there to fix it. This is the first I'd heard of the problem..Can you show me the raw record?

ausearch -ts today -a 10038 --raw

Thanks,
-Steve

#2 lore raw
Quoted 9 lines · Steve Grubb
Quoted 1 line · On Friday 24 October 2008 13:08:41 LC Bruzenak wrote:

Note that the syscall is listed differently.

Interesting.

Quoted 2 lines

This is using the 1.7.7 code (on F9), I have not yet moved over to 1.7.8 in case it may be fixed there.

Quoted 7 lines

Nope...nothing was changed there to fix it. This is the first I'd heard of the problem..Can you show me the raw record?

ausearch -ts today -a 10038 --raw

Thanks,
-Steve

I noticed it because with the audit-viewer I cannot see the "msg=" part of TRUSTED_APP records. I submitted a bugtraq
(https://fedorahosted.org/audit-viewer/ticket/6) for that. So, that made me look at the ausearch results to get all the info.

Additionally, I believe there is a policy issue which caused this in the first place...

Quoted 1 line

From the aggregating machine:

[root@dell1 ~]# ausearch -ts today -a 10038 --raw
node=v1 type=AVC msg=audit(1224864719.162:10038): avc: denied { read } for pid=11761 comm="prelude-manager" laddr=127.0.0.1 lport=4690 faddr=127.0.0.1 fport=36291 scontext=system_u:system_r:prelude_t:s0-s15:c0.c1023 tcontext=system_u:system_r:prelude_t:s15:c0.c1023 tclass=tcp_socket node=v1 type=SYSCALL msg=audit(1224864719.162:10038): arch=40000003 syscall=102 success=yes exit=5 a0=a a1=bfcc1f80 a2=25b0c4 a3=0 items=0 ppid=1 pid=11761 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=(none) ses=1 comm="prelude-manager" exe="/usr/bin/prelude-manager" subj=system_u:system_r:prelude_t:s0-s15:c0.c1023 key=(null)

Quoted 1 line

From the originating machine:

[root@v1 ~]# ausearch -ts today -a 10038 --raw
node=v1 type=AVC msg=audit(1224864719.162:10038): avc: denied { read } for pid=11761 comm="prelude-manager" laddr=127.0.0.1 lport=4690 faddr=127.0.0.1 fport=36291 scontext=system_u:system_r:prelude_t:s0-s15:c0.c1023 tcontext=system_u:system_r:prelude_t:s15:c0.c1023 tclass=tcp_socket node=v1 type=SYSCALL msg=audit(1224864719.162:10038): arch=40000003 syscall=102 success=yes exit=5 a0=a a1=bfcc1f80 a2=25b0c4 a3=0 items=0 ppid=1 pid=11761 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=(none) ses=1 comm="prelude-manager" exe="/usr/bin/prelude-manager" subj=system_u:system_r:prelude_t:s0-s15:c0.c1023 key=(null)

So it looks like the architectures interpretation (-i) of the syscall is where it differs?

Thx,
LCB.

Signature · 2 lines
LC (Lenny) Bruzenak
[email protected]
#4 lore raw
In reply to Steve Grubb
Quoted 9 lines · Steve Grubb
Quoted 1 line · On Friday 24 October 2008 13:08:41 LC Bruzenak wrote:

Note that the syscall is listed differently.

Interesting.

Quoted 2 lines

This is using the 1.7.7 code (on F9), I have not yet moved over to 1.7.8 in case it may be fixed there.

Quoted 8 lines

Nope...nothing was changed there to fix it. This is the first I'd heard of the problem..Can you show me the raw record?

ausearch -ts today -a 10038 --raw

Thanks,
-Steve

[root@dell1 ~]# ausyscall 102
getuid
[root@dell1 ~]# ausyscall i386 102
socketcall

Signature · 2 lines
LC (Lenny) Bruzenak
[email protected]
#5 lore raw
In reply to LC Bruzenak
Quoted 12 lines · LC Bruzenak

I have a test (virtual) machine running a 32-bit F9 OS. My aggregating machine is a 64-bit F9 box.

source (32-bit machine) :

[root@v1 ~]# ausearch -ts today -i -a 10038
----
node=v1 type=SYSCALL msg=audit(10/24/2008 11:11:59.162:10038) : arch=i386 syscall=socketcall(recv) success=yes exit=5 a0=a a1=bfcc1f80 a2=25b0c4 a3=0 items=0 ppid=1 pid=11761 auid=root uid=root gid=root euid=root suid=root fsuid=root egid=root sgid=root fsgid=root tty=(none) ses=1 comm=prelude-manager exe=/usr/bin/prelude-manager subj=system_u:system_r:prelude_t:s0-s15:c0.c1023 key=(null) node=v1 type=AVC msg=audit(10/24/2008 11:11:59.162:10038) : avc: denied { read } for pid=11761 comm=prelude-manager laddr=127.0.0.1 lport=4690 faddr=127.0.0.1 fport=36291 scontext=system_u:system_r:prelude_t:s0-s15:c0.c1023 tcontext=system_u:system_r:prelude_t:s15:c0.c1023 tclass=tcp_socket

aggregating machine (64-bit) :

Quoted 9 lines

[root@dell1 ~]# ausearch -ts today -i -a 10038
----
node=v1 type=SYSCALL msg=audit(10/24/2008 11:11:59.162:10038) : arch=i386 syscall=getuid success=yes exit=5 a0=a a1=bfcc1f80 a2=25b0c4 a3=0 items=0 ppid=1 pid=11761 auid=root uid=root gid=root euid=root suid=root fsuid=root egid=root sgid=root fsgid=root tty=(none) ses=1 comm=prelude-manager exe=/usr/bin/prelude-manager subj=system_u:system_r:prelude_t:s0-s15:c0.c1023 key=(null) node=v1 type=AVC msg=audit(10/24/2008 11:11:59.162:10038) : avc: denied { read } for pid=11761 comm=prelude-manager laddr=127.0.0.1 lport=4690 faddr=127.0.0.1 fport=36291 scontext=system_u:system_r:prelude_t:s0-s15:c0.c1023 tcontext=system_u:system_r:prelude_t:s15:c0.c1023 tclass=tcp_socket

Note that the syscall is listed differently.
This is using the 1.7.7 code (on F9), I have not yet moved over to 1.7.8 in case it may be fixed there.

This problem occurs because ausearch naively assumes the log data it's parsing originated on the same machine it's running on. Instead of reading the arch from the audit record it calls audit_detect_machine() which calls uname(). It then uses the machine arch it found with uname() to interpret the syscall number. Auparse has the same problem.

Signature · 1 line
John Dennis <[email protected]>
#3 lore raw
In reply to LC Bruzenak
Quoted 2 lines · On Friday 24 October 2008 13:27:49 LC Bruzenak wrote:

So it looks like the architectures interpretation (-i) of the syscall is where it differs?

Yes, there was a collision between the unset value and the i386 value in the source code. This meant that it when it ran across I386 machines, it thought there was an error looking it up and reverted to looking up the uname machine value as a fallback. Svn commit 155 fixes this.

-Steve

#6 lore raw
In reply to John Dennis
Quoted 7 lines · John Dennis
Quoted 1 line

This problem occurs because ausearch naively assumes the log data it's parsing originated on the same machine it's running on. Instead of reading the arch from the audit record it calls audit_detect_machine() which calls uname(). It then uses the machine arch it found with uname() to interpret the syscall number. Auparse has the same problem.

The audit-viewer gets the right syscall for the event's arch.

LCB.

Signature · 2 lines
LC (Lenny) Bruzenak
[email protected]