Re: [PATCH v2 2/5] audit: account backlog waiting time in audit_receive()

[Paul Moore]

On Mon, May 22, 2023 at 12:22 AM Eiichi Tsukata
<eiichi.tsukata@xxxxxxxxxxx> wrote:
> > On May 20, 2023, at 5:54, Paul Moore <paul@xxxxxxxxxxxxxx> wrote:
> > On May 11, 2023 Eiichi Tsukata <eiichi.tsukata@xxxxxxxxxxx> wrote:
> >>
> >> Currently backlog waiting time in audit_receive() is not accounted as
> >> audit_backlog_wait_time_actual. Accounts it as well.
> >>
> >> Signed-off-by: Eiichi Tsukata <eiichi.tsukata@xxxxxxxxxxx>
> >> ---
> >> kernel/audit.c | 44 ++++++++++++++++++++++++++------------------
> >> 1 file changed, 26 insertions(+), 18 deletions(-)
> >
> > The audit_receive() wait is different from that in audit_log_start()
> > as processes calling into audit_receive() are performing an explicit
> > audit operation whereas those processes calling audit_log_start() are
> > likely doing something else, e.g. opening a file, that happens to
> > result in an audit record being generated.  The fact that the
> > audit_receive() accounting logic, as well as the timeout calculation
> > used, is different from audit_log_start() is intentional.
> >
>
> The intention still sounds a bit not clear to me as both cases wait using
> the same parameter “backlog_wait_time” and use the same wait
> queue.
>
> IMHO, it will be better to have some comprehensive code comments there.

A fair point.  I'll add that to the todo list.

-- 
paul-moore.com