Linux-audit Archive on
 help / color / Atom feed
From: Steve Grubb <>
To: Vishnu Srinivasa Ramaprasad <>
Subject: Re: What value gets stored in a3 when -S execve?
Date: Wed, 27 May 2020 16:22:45 -0400
Message-ID: <4000710.EZmUHuUpmd@x2> (raw)
In-Reply-To: <>


On Wednesday, May 27, 2020 3:44:13 PM EDT Vishnu Srinivasa Ramaprasad wrote:
> I'm Vish and I am a newbie to auditd. My requirement is to log only
> shell/bash commands and custom commands executed by administrator users.
> I have created these rules in /etc/audit/rules.d/audit.rules, to ensure
> SYSCALL, EXECVE are being added to audit.log for administrator users with
> auid greater than 1000:
> -a exit,always -F arch=b64 -S execve -F auid>=1000 -F auid!=-1 -k log_cmd
> -a exit,always -F arch=b32 -S execve -F auid>=1000 -F auid!=-1 -k log_cmd
> After restarting auditd service, I had executed strace on a simple command
> which will display version of my project's toolkit:
> ~# strace -e trace=execve toolkit-version-show
> execve("/var/tmp/toolkit-version-show", ["toolkit-version-show"],
> 0x7ffef1fa38b0 /* 30 vars */) = 0
> Toolkit Version: 1.01
> +++ exited with 0 +++
> Later, I executed the ausearch command to check the log entry:
> ~# ausearch -i --start recent
> ----
> type=EXECVE msg=audit(05/27/2020 19:01:26.605:12725) : argc=2
> a0=/usr/bin/perl a1=/var/tmp/toolkit-version-show
> type=SYSCALL msg=audit(05/27/2020 19:01:26.605:12725) : arch=x86_64
> syscall=execve success=yes exit=0 a0=0x7ffef1fa2450 a1=0x7ffef1fa38a0
> a2=0x7ffef1fa38b0 a3=0x7f47f8669740 items=3 ppid=3641 pid=3643
> auid=administrator uid=root gid=root euid=root suid=root fsuid=root
> egid=root sgid=root fsgid=root tty=pts0 ses=8936 comm=toolkit-version-sho
> exe=/usr/bin/perl key=log_cmd
> ----
> type=EXECVE msg=audit(05/27/2020 19:01:26.601:12724) : argc=4 a0=strace
> a1=-e a2=trace=execve a3=toolkit-version-show
> type=SYSCALL msg=audit(05/27/2020 19:01:26.601:12724) : arch=x86_64
> syscall=execve success=yes exit=0 a0=0x55a2d44c9010 a1=0x55a2d449fe80
> a2=0x55a2d4389490 a3=0x8 items=2 ppid=3099 pid=3641 auid=administrator
> uid=root gid=root euid=root suid=root fsuid=root egid=root sgid=root
> fsgid=root tty=pts0 ses=8936 comm=strace exe=/usr/bin/strace key=log_cmd
> ----
> My understanding of a0 - a2: Please refer the syntax of execve() :
> int execve(const char*filename, char *const argv[], char *const envp[]);
> Based on syntax of execve() and output from strace, I could understand that
> in audit.log event entry:
> a0=0x7ffef1fa2450 represents filename argument of execve

It a pointer to the string.

> a1=0x7ffef1fa38a0 represents argv[] argument of execve
> a2=0x7ffef1fa38b0 represents envp[] argument of execve
> Question  1:  What does the a3=0x7f47f8669740  value represent?

whatever is on the stack in that position.

> As execve() has only 3 arguments (represented by a0,a1,a2), what value gets
> stored in a3?
> I have noticed a3 values to be varying from:
> a3=0x0
> a3=0x7
> a3=0x55a2d4389490
> a3=0x56a2d44adc00
> a3=0x8
> Question 2: a3=0x8 seems to be the value assigned for a majority of
> execve() syscalls.
> Is this a standard value set in case of main/primary system call, such as
> toolkit-version-show?

Nope. It depends entirely on what the previous syscalls or system activity 
were and where on the stack it currently is.

> Could I use this in a rule filter *-F a3=8* to log only primary (custom
> script) command executed by user and not internal commands executed by
> custom script like ls, cat, grep.. etc.:
> -a exit,always -F arch=b64 -S execve *-F a3=8* -F auid>=1000 -F auid!=-1 -k
> log_cmd
> -a exit,always -F arch=b32 -S execve *-F a3=8* -F auid>=1000 -F auid!=-1 -k
> log_cmd
> Would I miss logging a few primary syscalls, if a3 is not 0x80 in some
> cases?

Nope. you can't count on anything in a3 since it is not valid for that 

> Question 3: If a3=0x8 is not a standard value, Is it possible to identify
> primary custom command and log only that command, and not internal commands
> with a3=0x0 or a3=0x55a2d4389490?

You can always place a watch on the custom commands.
-a exit,always -F path=/path-to/custom-command -F perm=x -F key=command

> Question 4: Is it possible to filter out and not log syscalls with
> tty=(none)?

No. But there is a loose correlation with auid or sessionid being -1. That 
means it is a daemon. And if you only want commands run by people, then you 
want -F auid>=1000 -F auid!=-1.


Linux-audit mailing list

  reply index

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-05-27 19:44 Vishnu Srinivasa Ramaprasad
2020-05-27 20:22 ` Steve Grubb [this message]
2020-05-28 15:34   ` Vishnu Srinivasa Ramaprasad

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=4000710.EZmUHuUpmd@x2 \ \ \ \

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

Linux-audit Archive on

Archives are clonable:
	git clone --mirror linux-audit/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-audit linux-audit/ \
	public-inbox-index linux-audit

Example config snippet for mirrors

Newsgroup available over NNTP:

AGPL code for this site: git clone