Re: [PATCH v2 2/2] selinux: add basic filtering for audit trace events

[peter enderborg]

On 8/13/20 5:49 PM, Stephen Smalley wrote:
> On 8/13/20 11:35 AM, peter enderborg wrote:
>
>> On 8/13/20 5:05 PM, Casey Schaufler wrote:
>>> On 8/13/2020 7:48 AM, Thiébaud Weksteen wrote:
>>>> From: Peter Enderborg <peter.enderborg@xxxxxxxx>
>>>>
>>>> This patch adds further attributes to the event. These attributes are
>>>> helpful to understand the context of the message and can be used
>>>> to filter the events.
>>>>
>>>> There are three common items. Source context, target context and tclass.
>>>> There are also items from the outcome of operation performed.
>>>>
>>>> An event is similar to:
>>>>             <...>-1309  [002] ....  6346.691689: selinux_audited:
>>>>         requested=0x4000000 denied=0x4000000 audited=0x4000000
>>>>         result=-13 ssid=315 tsid=61
>>> It may not be my place to ask, but *please please please* don't
>>> externalize secids. I understand that it's easier to type "42"
>>> than "system_r:cupsd_t:s0-s0:c0.c1023", and that it's easier for
>>> your tools to parse and store the number. Once you start training
>>> people that system_r:cupsd_t:s0-s0:c0.c1023 is secid 42 you'll
>>> never be able to change it. The secid will start showing up in
>>> scripts. Bad  Things  Will  Happen.
>> Ok, it seems to mostly against having this performance options.
>> Yes, it is a kernel internal data. So is most of the kernel tracing.
>> I see it is a primary tool for kernel debugging but than can also be
>> used for user-space debugging tools.  Hiding data for debuggers
>> does not make any sense too me.
>
> To be clear, userspace tools can't use fixed secid values because secids are dynamically assigned by SELinux and thus secid 42 need not correspond to the same security context across different boots even with the same kernel and policy.  I wouldn't include them in the event unless it is common practice to include fields that can only be interpreted if you can debug the running kernel.  It would be akin to including kernel pointers in the event (albeit without the KASLR ramifications).
>
>
Just as a reference on my fedora system; out of 1808 events 244 as a pointer print. I don't see that there is any obfuscating aka "%pK" as there is for logs.