Re: [PATCH] audit: set context->dummy even when audit is off

[Chris Mason <clm@fb.com>]

On 1 Nov 2019, at 10:16, Steve Grubb wrote:

> On Friday, November 1, 2019 9:24:17 AM EDT Chris Mason wrote:
>> On 31 Oct 2019, at 19:27, Paul Moore wrote:
>>> On Thu, Oct 31, 2019 at 12:40 PM Chris Mason <clm@fb.com> wrote:
>>> [ ... ]
>>> Hi Chris,
>>>
>>> This is a rather hasty email as I'm at a conference right now, but I
>>> wanted to convey that I'm not opposed to making sure that the NTP
>>> records obey the audit configuration (that was the original intent
>>> after all), I think it is just that we are all a little confused as 
>>> to
>>> why you are seeing the NTP records *and*only* the NTP records.
>>
>> This part is harder to nail down because there's a window during boot
>> where journald has enabled audit but chef hasn't yet run in and 
>> turned
>> it off, so we get a lot of logs early and then mostly ntp after that.
>
> This is the root of the problem. Journald should never turn on audit 
> since it
> has no idea if auditd even has rules to load. What if the end user 
> does not
> want auditing? By blindly enabling audit without knowing if its 
> wanted, it
> causes a system performance hit even with no rules loaded. It would be 
> best
> if journald leaves audit alone. If it wants to listen on the multicast
> socket, so be it. It should just listen and not try to alter the 
> system.
>
> Back to ntp, it sounds like the ntp record needs to check for 
> audit_enabled
> rather than the dummy context.

I'm not against sprinkling more audit_enabled checks, but we'd have to 
change audit_inode() and a bunch of the other callers of 
audit_dummy_context() as well.

-chris