[RFC PATCH 3/4] audit: store the auditd PID as a pid struct instead of pid_t

[Paul Moore]

On Mon, Apr 10, 2017 at 12:30 AM, Richard Guy Briggs <rgb at redhat.com> wrote:
> On 2017-03-21 14:59, Paul Moore wrote:
>> From: Paul Moore <paul at paul-moore.com>
>>
>> This is arguably the right thing to do, and will make it easier when
>> we start supporting multiple audit daemons in different namespaces.
>
> I had tried this several years ago inspired by Eric Biederman's work for
> the same reasons:
>         https://www.redhat.com/archives/linux-audit/2014-February/msg00116.html
>
> A lot has changed since then...  A couple of comments in-line...
>
>> Signed-off-by: Paul Moore <paul at paul-moore.com>
>> ---
>>  kernel/audit.c |   84 ++++++++++++++++++++++++++++++++++++++------------------
>>  kernel/audit.h |    2 +
>>  2 files changed, 58 insertions(+), 28 deletions(-)
>>
>> diff --git a/kernel/audit.c b/kernel/audit.c
>> index 6cbf47a372e8..b718bf3a73f8 100644
>> --- a/kernel/audit.c
>> +++ b/kernel/audit.c
>> @@ -58,6 +58,7 @@
>>  #include <linux/rcupdate.h>
>>  #include <linux/mutex.h>
>>  #include <linux/gfp.h>
>> +#include <linux/pid.h>
>>
>>  #include <linux/audit.h>
>>
>> @@ -117,7 +118,7 @@ struct audit_net {
>>   * or the included spinlock for writing.
>>   */
>>  static struct auditd_connection {
>> -     int pid;
>> +     struct pid *pid;
>>       u32 portid;
>>       struct net *net;
>>       spinlock_t lock;
>> @@ -221,18 +222,41 @@ struct audit_reply {
>>   * Description:
>>   * Return 1 if the task is a registered audit daemon, 0 otherwise.
>>   */
>> -int auditd_test_task(const struct task_struct *task)
>> +int auditd_test_task(struct task_struct *task)
>
> Does the compiler complain if this is left as const?

Yep, it runs afoul with the task_tgid() call.

> Reviewed-by: Richard Guy Briggs <rgb at redhat.com>

Thanks.  Merged.

-- 
paul moore
www.paul-moore.com