Skip to contentSkip to conversation

audit 1.7.4 released

DISCUSSION Steve Grubb 4 participants 10 messages May 19, 2008 – May 27, 2008

lore
Share & tools
lore
#0 lore raw

Hi,

I've just released a new version of the audit daemon. It can be downloaded from http://people.redhat.com/sgrubb/audit It will also be in rawhide tomorrow. The Changelog is:

- Fix interpreting of keys in syscall records
- Interpret audit rule config change list fields
- Don't error on name=(null) PATH records in ausearch/report
- Add key report to aureport
- Fix --end today to be now
- Added python bindings for auparse_goto_record_num
- Update system-config-audit to 0.4.7 (Miloslav Trmac)
- Add support for the filetype field option in auditctl
- In audispd boost priority after starting children

This release is a mix of bug fixes and new features. The bug fixes are what is driving the release earlier than what I'd like. I was doing some testing and found that a lot of keys were not being interpreted correctly. I think many were coming back as (null) which looks pretty normal if you don't use the keys. Anyways, this is fixed.

I also found that ausearch/report were not processing some events correctly when the PATH record's name field was (null). The result of this was that the event was being discarded in search results.

With the new interest in keys, I added a key report to aureport. This presents a listing of what keys & quantities have been found in a given time frame. During testing of that, I found that "--end today" was not behaving as I expected. I really think that when you do aureport --start yesterday --end today, you should see events from yesterday at midnight until now.

I added an interpretation for the list in audit watch add/delete events. This will now print the list's name like exit,entry, user, etc.

This release also adds support for a new rule field in he 2.6.26 kernel. If you wanted to audit setting the execute bit via the chmod syscal, you would normally write a rule something like this:

-a always,exit -S chmod -F a1&0111

but the problem is that this will trigger on chmod 0755 of directories which is pretty common if you want the directory to be searchable. So we added a new option to let you specify what the object's type is, filetype. The new rule would look like this:

-a always,exit -S chmod -F a1&0111 -F filetype=file

filetype can be file, dir, socket, symlink, char, block, or fifo.

And last item I wanted to comment on was the change in priority boost for audispd. I moved the call to nice() until after the child processes were started. This is because audispd should not have to fight with its children for time slices at the higher priority. It has an internal queue that can be extended by admin configurable parameters. The children are now started with the priority inherited from auditd.

Please let me know if you run across any problems with this release.

-Steve

#1 lore raw

Steve,

I am testing 1.7.4 (with mls permissive policy):
audit-viewer-0.2-2.fc9.x86_64
audit-libs-python-1.7.4-1.fc9.x86_64
system-config-audit-0.4.7-1.fc9.x86_64
audit-1.7.4-1.fc9.x86_64
audit-libs-devel-1.7.4-1.fc9.x86_64
audit-debuginfo-1.7.3-1.fc9.x86_64
audit-libs-1.7.4-1.fc9.x86_64
audit-libs-1.7.4-1.fc9.i386

I moved all the old audit out of the way, so all records would be new, and see this after reboot:

[root@hugo ~]# aureport -h -i --summary

Host Summary Report
===========================
total host
===========================
223 ?
12 homeserver
8 127.0.0.1
6 0.0.0.0

The "?" entries are application audits - I am going to look, maybe they have an error on the way we are sending those in.

The ones I don't understand are the "0.0.0.0" entries. Here is an example of one of those:

[root@hugo ~]# ausearch -hn 0.0.0.0 -i --just-one
----
type=SOCKADDR msg=audit(05/27/2008 10:30:22.163:13193) : saddr=inet host:0.0.0.0 serv:711
type=SYSCALL msg=audit(05/27/2008 10:30:22.163:13193) : arch=x86_64 syscall=bind success=yes exit=0 a0=5 a1=7fff63dbb220 a2=10 a3=89ea70 items=0 ppid=1 pid=2647 auid=unset uid=root gid=root euid=root suid=root fsuid=root egid=root sgid=root fsgid=root tty=(none) ses=4294967295 comm=rpc.rquotad exe=/usr/sbin/rpc.rquotad subj=system_u:system_r:initrc_t:s0-s15:c0.c1023 key=(null) type=AVC msg=audit(05/27/2008 10:30:22.163:13193) : avc: denied { name_bind } for pid=2647 comm=rpc.rquotad src=711 scontext=system_u:system_r:initrc_t:s0-s15:c0.c1023 tcontext=system_u:object_r:hi_reserved_port_t:s0 tclass=udp_socket

Is the host "0.0.0.0" field here a bug?
Once we aggregate these would be tough to separate. Also the localhost ones I guess:

[root@hugo ~]# ausearch -hn 127.0.0.1 -i --just-one
----
type=SOCKADDR msg=audit(05/27/2008 10:30:22.022:13190) : saddr=inet host:127.0.0.1 serv:750
type=SYSCALL msg=audit(05/27/2008 10:30:22.022:13190) : arch=x86_64 syscall=sendto success=yes exit=28 a0=6 a1=7f56310606e0 a2=1c a3=0 items=0 ppid=1 pid=2189 auid=unset uid=rpc gid=root euid=rpc suid=rpc fsuid=rpc egid=root sgid=root fsgid=root tty=(none) ses=4294967295 comm=rpcbind exe=/sbin/rpcbind subj=system_u:system_r:initrc_t:s0-s15:c0.c1023 key=(null) type=AVC msg=audit(05/27/2008 10:30:22.022:13190) : avc: denied { recvfrom } for pid=2189 comm=rpcbind saddr=127.0.0.1 src=111 daddr=127.0.0.1 dest=750 netif=lo scontext=system_u:system_r:nfsd_t:s0-s15:c0.c1023 tcontext=system_u:system_r:initrc_t:s0-s15:c0.c1023 tclass=association

Thx,
LCB.

Signature · 2 lines
LC (Lenny) Bruzenak
[email protected]
#2 lore raw
Quoted 47 lines · LC Bruzenak

Steve,

I am testing 1.7.4 (with mls permissive policy):
audit-viewer-0.2-2.fc9.x86_64
audit-libs-python-1.7.4-1.fc9.x86_64
system-config-audit-0.4.7-1.fc9.x86_64
audit-1.7.4-1.fc9.x86_64
audit-libs-devel-1.7.4-1.fc9.x86_64
audit-debuginfo-1.7.3-1.fc9.x86_64
audit-libs-1.7.4-1.fc9.x86_64
audit-libs-1.7.4-1.fc9.i386

I moved all the old audit out of the way, so all records would be new, and see this after reboot:

[root@hugo ~]# aureport -h -i --summary

Host Summary Report
===========================
total host
===========================
223 ?
12 homeserver
8 127.0.0.1
6 0.0.0.0

The "?" entries are application audits - I am going to look, maybe they have an error on the way we are sending those in.

The ones I don't understand are the "0.0.0.0" entries. Here is an example of one of those:

[root@hugo ~]# ausearch -hn 0.0.0.0 -i --just-one
----
type=SOCKADDR msg=audit(05/27/2008 10:30:22.163:13193) : saddr=inet host:0.0.0.0 serv:711
type=SYSCALL msg=audit(05/27/2008 10:30:22.163:13193) : arch=x86_64 syscall=bind success=yes exit=0 a0=5 a1=7fff63dbb220 a2=10 a3=89ea70 items=0 ppid=1 pid=2647 auid=unset uid=root gid=root euid=root suid=root fsuid=root egid=root sgid=root fsgid=root tty=(none) ses=4294967295 comm=rpc.rquotad exe=/usr/sbin/rpc.rquotad subj=system_u:system_r:initrc_t:s0-s15:c0.c1023 key=(null) type=AVC msg=audit(05/27/2008 10:30:22.163:13193) : avc: denied { name_bind } for pid=2647 comm=rpc.rquotad src=711 scontext=system_u:system_r:initrc_t:s0-s15:c0.c1023 tcontext=system_u:object_r:hi_reserved_port_t:s0 tclass=udp_socket

Quoted 1 line

Is the host "0.0.0.0" field here a bug?

Isn't this telling up that they are calling bind on any interface not a specific address?

the const struct sockaddr *addr part of the bind(2) call is IN_ADDRANY what whatever the semantics are...

-Eric

#3 lore raw

On Tue, 2008-05-27 at 11:59 -0400, Eric Paris wrote:
...

Quoted 16 lines
Quoted 14 lines

----
type=SOCKADDR msg=audit(05/27/2008 10:30:22.163:13193) : saddr=inet host:0.0.0.0 serv:711
type=SYSCALL msg=audit(05/27/2008 10:30:22.163:13193) : arch=x86_64 syscall=bind success=yes exit=0 a0=5 a1=7fff63dbb220 a2=10 a3=89ea70 items=0 ppid=1 pid=2647 auid=unset uid=root gid=root euid=root suid=root fsuid=root egid=root sgid=root fsgid=root tty=(none) ses=4294967295 comm=rpc.rquotad exe=/usr/sbin/rpc.rquotad subj=system_u:system_r:initrc_t:s0-s15:c0.c1023 key=(null) type=AVC msg=audit(05/27/2008 10:30:22.163:13193) : avc: denied { name_bind } for pid=2647 comm=rpc.rquotad src=711 scontext=system_u:system_r:initrc_t:s0-s15:c0.c1023 tcontext=system_u:object_r:hi_reserved_port_t:s0 tclass=udp_socket

Quoted 1 line

Is the host "0.0.0.0" field here a bug?

Quoted 8 lines

Isn't this telling up that they are calling bind on any interface not a specific address?

the const struct sockaddr *addr part of the bind(2) call is IN_ADDRANY what whatever the semantics are...

-Eric

I understand; thanks.
Semantically this is probably the intended use of the "host" field.

When audit data is examined on one host this isn't a problem. But when aggregated with other host audit data, this record as standalone is indistinguishable from a similar one on a different host ... unless I'm missing something in the above record.

Or if we separate the aggregated data by filename/whatever which the search tools can use to differentiate audit hosts.

Thx,
LCB.

Signature · 2 lines
LC (Lenny) Bruzenak
[email protected]
#4 lore raw
In reply to LC Bruzenak
Quoted 1 line · On Tuesday 27 May 2008 11:50:31 LC Bruzenak wrote:

Is the host "0.0.0.0" field here a bug?

No, Eric's explanation sounds right.

Quoted 1 line

Once we aggregate these would be tough to separate.

That is why we added the node field. :) You should probably enable it with the name_format option.

-Steve

#5 lore raw

On Tue, 2008-05-27 at 12:10 -0400, Steve Grubb wrote:
...

Quoted 4 lines
Quoted 1 line

Once we aggregate these would be tough to separate.

That is why we added the node field. :) You should probably enable it with the name_format option.

I think I do have it:

[root@hugo audit]# grep name_format /etc/audit/auditd.conf name_format = hostname

LCB.

Signature · 2 lines
LC (Lenny) Bruzenak
[email protected]
#6 lore raw
Quoted 7 lines · On Tuesday 27 May 2008 12:16:18 LC Bruzenak wrote:
Quoted 2 lines

That is why we added the node field. :)  You should probably enable it with the name_format option.

I think I do have it:

[root@hugo audit]# grep name_format /etc/audit/auditd.conf name_format = hostname

You may also need to make the same change in
/etc/audisp/audispd.conf.

-Steve

#8 lore raw
In reply to LC Bruzenak
Quoted 11 lines · LC Bruzenak

On Tue, 2008-05-27 at 12:10 -0400, Steve Grubb wrote:
...

Quoted 4 lines
Quoted 1 line

Once we aggregate these would be tough to separate.

That is why we added the node field. :) You should probably enable it with the name_format option.

I think I do have it:

[root@hugo audit]# grep name_format /etc/audit/auditd.conf name_format = hostname

Isn't the audit dispatcher's role of adding the node name in the record? If so, only records going through the audispd would have this field.

-K

Signature · 2 lines
Klaus Heinrich Kiwi
Security Development - IBM Linux Technology Center
#9 lore raw
Quoted 9 lines · On Tuesday 27 May 2008 12:57:28 Klaus Heinrich Kiwi wrote:
Quoted 8 lines · LC Bruzenak

On Tue, 2008-05-27 at 12:10 -0400, Steve Grubb wrote:
...

Quoted 4 lines
Quoted 1 line

Once we aggregate these would be tough to separate.

That is why we added the node field. :) You should probably enable it with the name_format option.

Quoted 7 lines
Quoted 4 lines

I think I do have it:

[root@hugo audit]# grep name_format /etc/audit/auditd.conf name_format = hostname

Isn't the audit dispatcher's role of adding the node name in the record? If so, only records going through the audispd would have this field.

People may want the node name on disk as well as associated with events in the real time stream. So, there are separate enablers.

-Steve

#7 lore raw
In reply to Steve Grubb
Quoted 6 lines · Steve Grubb
Quoted 5 lines · On Tuesday 27 May 2008 12:16:18 LC Bruzenak wrote:
Quoted 2 lines

That is why we added the node field. :) You should probably enable it with the name_format option.

I think I do have it:

Quoted 7 lines
Quoted 2 lines

[root@hugo audit]# grep name_format /etc/audit/auditd.conf name_format = hostname

You may also need to make the same change in
/etc/audisp/audispd.conf.

-Steve

I thought this would solve it since for some reason I had "HOSTNAME" vice "hostname". I thought I had copied your instructions in the HOWTO, but when I checked, it is lower case. I rotated the logs, removed the old one, restarted audit, rebooted and still see nothing unique:

[root@hugo ~]# grep name_format /etc/audi*/*.conf
/etc/audisp/audispd.conf:name_format = hostname
/etc/audit/auditd.conf:name_format = hostname

[root@hugo ~]# ausearch -hn 0.0.0.0 -i --just-one
----
type=SOCKADDR msg=audit(05/27/2008 12:05:12.483:1299) : saddr=inet host:0.0.0.0 serv:708
type=SYSCALL msg=audit(05/27/2008 12:05:12.483:1299) : arch=x86_64 syscall=bind success=no exit=-98(Address already in use) a0=5 a1=7fff0d16b6c0 a2=10 a3=89ea70 items=0 ppid=1 pid=2645 auid=unset uid=root gid=root euid=root suid=root fsuid=root egid=root sgid=root fsgid=root tty=(none) ses=4294967295 comm=rpc.rquotad exe=/usr/sbin/rpc.rquotad subj=system_u:system_r:initrc_t:s0-s15:c0.c1023 key=(null) type=AVC msg=audit(05/27/2008 12:05:12.483:1299) : avc: denied { name_bind } for pid=2645 comm=rpc.rquotad src=708 scontext=system_u:system_r:initrc_t:s0-s15:c0.c1023 tcontext=system_u:object_r:hi_reserved_port_t:s0 tclass=udp_socket

[root@hugo ~]# ausearch -hn 127.0.0.1 -i --just-one
----
type=SOCKADDR msg=audit(05/27/2008 12:05:12.359:1296) : saddr=inet host:127.0.0.1 serv:748
type=SYSCALL msg=audit(05/27/2008 12:05:12.359:1296) : arch=x86_64 syscall=sendto success=yes exit=28 a0=6 a1=7f16f44936e0 a2=1c a3=0 items=0 ppid=1 pid=2187 auid=unset uid=rpc gid=root euid=rpc suid=rpc fsuid=rpc egid=root sgid=root fsgid=root tty=(none) ses=4294967295 comm=rpcbind exe=/sbin/rpcbind subj=system_u:system_r:initrc_t:s0-s15:c0.c1023 key=(null) type=AVC msg=audit(05/27/2008 12:05:12.359:1296) : avc: denied { recvfrom } for pid=2187 comm=rpcbind saddr=127.0.0.1 src=111 daddr=127.0.0.1 dest=748 netif=lo scontext=system_u:system_r:nfsd_t:s0-s15:c0.c1023 tcontext=system_u:system_r:initrc_t:s0-s15:c0.c1023 tclass=association

[root@hugo ~]# ausearch -hn 127.0.0.1 -i --just-one | grep host type=SOCKADDR msg=audit(05/27/2008 12:05:12.359:1296) : saddr=inet host:127.0.0.1 serv:748
[root@hugo ~]# ausearch -hn 0.0.0.0 -i --just-one | grep host type=SOCKADDR msg=audit(05/27/2008 12:05:12.483:1299) : saddr=inet host:0.0.0.0 serv:708

Appreciate the help; got more?
:)

Thx,
LCB.

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