Re: ntp audit spew.

[Paul Moore]

On Tue, Sep 24, 2019 at 9:19 AM Steve Grubb  wrote:
> On Monday, September 23, 2019 12:14:14 PM EDT Paul Moore wrote:
> > On Mon, Sep 23, 2019 at 11:50 AM Dave Jones 
> wrote:
> > > I have some hosts that are constantly spewing audit messages like so:
> > >
> > > [46897.591182] audit: type=1333 audit(1569250288.663:220): op=offset
> > > old=2543677901372 new=2980866217213 [46897.591184] audit: type=1333
> > > audit(1569250288.663:221): op=freq old=-2443166611284 new=-2436281764244
> > > [48850.604005] audit: type=1333 audit(1569252241.675:222): op=offset
> > > old=1850302393317 new=3190241577926 [48850.604008] audit: type=1333
> > > audit(1569252241.675:223): op=freq old=-2436281764244 new=-2413071187316
> > > [49926.567270] audit: type=1333 audit(1569253317.638:224): op=offset
> > > old=2453141035832 new=2372389610455 [49926.567273] audit: type=1333
> > > audit(1569253317.638:225): op=freq old=-2413071187316 new=-2403561671476
> > >
> > > This gets emitted every time ntp makes an adjustment, which is apparently
> > > very frequent on some hosts.
> > >
> > >
> > > Audit isn't even enabled on these machines.
> > >
> > > # auditctl -l
> > > No rules
> >
> > [NOTE: added linux-audit to the CC line]
> >
> > There is an audit mailing list, please CC it when you have audit
> > concerns/questions/etc.
> >
> > What happens when you run 'auditctl -a never,task'?
>
> Actually, "-e 0" should turn it off. There is a general problem where systemd
> turns on auditing just because it can. The above rule just makes audit
> processes inauditable, but does not affect the kernel originating events.

The 'auditctl -s' output was lost when I trimmed/replied to DaveJ's
original email (sorry), but it appears that audit_enabled is already
'0':

> # auditctl -s
> enabled 0
> failure 1
> pid 0
> rate_limit 0
> backlog_limit 64
> lost 0
> backlog 0
> loginuid_immutable 0 unlocked

Original post from DaveJ:

* https://lore.kernel.org/lkml/20190923155041.ga14...@codemonkey.org.uk

-- 
paul moore
www.paul-moore.com

ntp audit spew.

[Dave Jones]

I have some hosts that are constantly spewing audit messages like so:

[46897.591182] audit: type=1333 audit(1569250288.663:220): op=offset 
old=2543677901372 new=2980866217213
[46897.591184] audit: type=1333 audit(1569250288.663:221): op=freq 
old=-2443166611284 new=-2436281764244
[48850.604005] audit: type=1333 audit(1569252241.675:222): op=offset 
old=1850302393317 new=3190241577926
[48850.604008] audit: type=1333 audit(1569252241.675:223): op=freq 
old=-2436281764244 new=-2413071187316
[49926.567270] audit: type=1333 audit(1569253317.638:224): op=offset 
old=2453141035832 new=2372389610455
[49926.567273] audit: type=1333 audit(1569253317.638:225): op=freq 
old=-2413071187316 new=-2403561671476

This gets emitted every time ntp makes an adjustment, which is apparently very 
frequent on some hosts.


Audit isn't even enabled on these machines.

# auditctl -l
No rules

# auditctl -s
enabled 0
failure 1
pid 0
rate_limit 0
backlog_limit 64
lost 0
backlog 0
loginuid_immutable 0 unlocked

Asides from the log spew, why is this code doing _anything_ when audit
isn't enabled ?

Something like this:


diff --git a/kernel/audit.c b/kernel/audit.c
index da8dc0db5bd3..1291d826c024 100644
--- a/kernel/audit.c
+++ b/kernel/audit.c
@@ -2340,6 +2340,9 @@ void audit_log(struct audit_context *ctx, gfp_t gfp_mask, 
int type,
struct audit_buffer *ab;
va_list args;
 
+   if (audit_initialized != AUDIT_INITIALIZED)
+   return;
+
ab = audit_log_start(ctx, gfp_mask, type);
if (ab) {
va_start(args, fmt);


Might silence the spew, but I'm concerned that the amount of work that
audit is doing on an unconfigured machine might warrant further
investigation.

("turn off CONFIG_AUDIT" isn't an option unfortunately, as this is a
one-size-fits-all kernel that runs on some other hosts that /do/ have
audit configured)

Dave