Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20190307123254.348-1-omosnace@redhat.com> <20190307123254.348-3-omosnace@redhat.com>
 <alpine.DEB.2.21.1903280039530.1789@nanos.tec.linutronix.de>
In-Reply-To: <alpine.DEB.2.21.1903280039530.1789@nanos.tec.linutronix.de>
From:   Ondrej Mosnacek <omosnace@redhat.com>
Date:   Mon, 1 Apr 2019 11:13:23 +0200
Message-ID: <CAFqZXNvQioPGUiU0q5R4nQLW7pQDsViNH-DR=SF4Fec+-CwksA@mail.gmail.com>
Subject: Re: [RFC PATCH ghak10 v6 2/2] ntp: Audit NTP parameters adjustment
To:     Thomas Gleixner <tglx@linutronix.de>
Cc:     Linux-Audit Mailing List <linux-audit@redhat.com>,
        Paul Moore <paul@paul-moore.com>,
        Richard Guy Briggs <rgb@redhat.com>,
        Steve Grubb <sgrubb@redhat.com>,
        Miroslav Lichvar <mlichvar@redhat.com>,
        John Stultz <john.stultz@linaro.org>,
        Stephen Boyd <sboyd@kernel.org>,
        Linux kernel mailing list <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: 8BIT
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu, Mar 28, 2019 at 1:02 AM Thomas Gleixner <tglx@linutronix.de> wrote:
> On Thu, 7 Mar 2019, Ondrej Mosnacek wrote:
>
> > Emit an audit record every time selected NTP parameters are modified
> > from userspace (via adjtimex(2) or clock_adjtime(2)).
> >
> > Such events will now generate records of type AUDIT_TIME_ADJNTPVAL
> > containing the following fields:
> >   - op -- which value was adjusted:
> >     - offset -- corresponding to the time_offset variable
> >     - freq   -- corresponding to the time_freq variable
> >     - status -- corresponding to the time_status variable
> >     - adjust -- corresponding to the time_adjust variable
> >     - tick   -- corresponding to the tick_usec variable
> >     - tai    -- corresponding to the timekeeping's TAI offset
> >   - old -- the old value
> >   - new -- the new value
> >
> > For reference, running the following commands:
> >
> >     auditctl -D
> >     auditctl -a exit,always -F arch=b64 -S adjtimex
> >     chronyd -q
> >
> > produces audit records like this:
> >
> > type=SYSCALL msg=audit(1530616044.507:5): arch=c000003e syscall=159 success=yes exit=5 a0=7fff57e78c00 a1=0 a2=4 a3=7f754ae28c0a items=0 ppid=626 pid=629 auid=0 uid=0 gid=0 euid=0 suid=0 fsuid=0 egid=0 sgid=0 fsgid=0 tty=(none) ses=1 comm="chronyd" exe="/usr/sbin/chronyd" subj=system_u:system_r:kernel_t:s0 key=(null)
>
> <SNIP gazillions of lines of unparseable garbage>
>
> Is it really necessary to put this into the changelog?

Yeah, sorry, I went a bit overboard with the record examples... I'll
try to provide simpler and less verbose examples in the next version.

>
> >
> > +void __audit_ntp_adjust(const char *type, s64 oldval, s64 newval)
> > +{
> > +     audit_log(audit_context(), GFP_ATOMIC, AUDIT_TIME_ADJNTPVAL,
>
> No.
>
> > +               "op=%s old=%lli new=%lli", type,
> > +               (long long)oldval, (long long)newval);
> > +}
> > +
> >  static void audit_log_task(struct audit_buffer *ab)
> >  {
> >       kuid_t auid, uid;
> > diff --git a/kernel/time/ntp.c b/kernel/time/ntp.c
> > index 36a2bef00125..5f456a84151a 100644
> > --- a/kernel/time/ntp.c
> > +++ b/kernel/time/ntp.c
> > @@ -17,6 +17,7 @@
> >  #include <linux/mm.h>
> >  #include <linux/module.h>
> >  #include <linux/rtc.h>
> > +#include <linux/audit.h>
> >
> >  #include "ntp_internal.h"
> >  #include "timekeeping_internal.h"
> > @@ -293,6 +294,8 @@ static inline s64 ntp_update_offset_fll(s64 offset64, long secs)
> >
> >  static void ntp_update_offset(long offset)
> >  {
> > +     s64 old_offset = time_offset;
> > +     s64 old_freq = time_freq;
> >       s64 freq_adj;
> >       s64 offset64;
> >       long secs;
> > @@ -341,6 +344,9 @@ static void ntp_update_offset(long offset)
> >       time_freq   = max(freq_adj, -MAXFREQ_SCALED);
> >
> >       time_offset = div_s64(offset64 << NTP_SCALE_SHIFT, NTP_INTERVAL_FREQ);
> > +
> > +     audit_ntp_adjust("offset", old_offset, time_offset);
> > +     audit_ntp_adjust("freq", old_freq, time_freq);
> >  }
> >
> >  /**
> > @@ -658,21 +664,31 @@ static inline void process_adj_status(const struct timex *txc)
> >
> >  static inline void process_adjtimex_modes(const struct timex *txc, s32 *time_tai)
> >  {
> > -     if (txc->modes & ADJ_STATUS)
> > -             process_adj_status(txc);
> > +     if (txc->modes & (ADJ_STATUS | ADJ_NANO | ADJ_MICRO)) {
> > +             int old_status = time_status;
> > +
> > +             if (txc->modes & ADJ_STATUS)
> > +                     process_adj_status(txc);
> > -     if (txc->modes & ADJ_NANO)
> > -             time_status |= STA_NANO;
> > +             if (txc->modes & ADJ_NANO)
> > +                     time_status |= STA_NANO;
> >
> > -     if (txc->modes & ADJ_MICRO)
> > -             time_status &= ~STA_NANO;
> > +             if (txc->modes & ADJ_MICRO)
> > +                     time_status &= ~STA_NANO;
> > +
> > +             audit_ntp_adjust("status", old_status, time_status);
> > +     }
> >
> >       if (txc->modes & ADJ_FREQUENCY) {
> > +             s64 old_freq = time_freq;
> > +
> >               time_freq = txc->freq * PPM_SCALE;
> >               time_freq = min(time_freq, MAXFREQ_SCALED);
> >               time_freq = max(time_freq, -MAXFREQ_SCALED);
> >               /* update pps_freq */
> >               pps_set_freq(time_freq);
> > +
> > +             audit_ntp_adjust("freq", old_freq, time_freq);
> >       }
> >
> >       if (txc->modes & ADJ_MAXERROR)
> > @@ -689,14 +705,18 @@ static inline void process_adjtimex_modes(const struct timex *txc, s32 *time_tai
> >               time_constant = max(time_constant, 0l);
> >       }
> >
> > -     if (txc->modes & ADJ_TAI && txc->constant > 0)
> > +     if (txc->modes & ADJ_TAI && txc->constant > 0) {
> > +             audit_ntp_adjust("tai", *time_tai, txc->constant);
> >               *time_tai = txc->constant;
> > +     }
> >
> >       if (txc->modes & ADJ_OFFSET)
> >               ntp_update_offset(txc->offset);
> >
> > -     if (txc->modes & ADJ_TICK)
> > +     if (txc->modes & ADJ_TICK) {
> > +             audit_ntp_adjust("tick", tick_usec, txc->tick);
> >               tick_usec = txc->tick;
> > +     }
> >
> >       if (txc->modes & (ADJ_TICK|ADJ_FREQUENCY|ADJ_OFFSET))
> >               ntp_update_frequency();
> > @@ -718,6 +738,8 @@ int __do_adjtimex(struct timex *txc, const struct timespec64 *ts, s32 *time_tai)
> >                       /* adjtime() is independent from ntp_adjtime() */
> >                       time_adjust = txc->offset;
> >                       ntp_update_frequency();
> > +
> > +                     audit_ntp_adjust("adjust", save_adjust, txc->offset);
> >               }
> >               txc->offset = save_adjust;
> >       } else {
>
> Not going to happen. We are not reshuffling all that code just to
> accomodate random audit log invocations in a critical section plus having a
> gazillion of GFP_ATOMIC allocation in the critical section just because.

OK, seems I underestimated the consequences of putting the logging
calls directly in there. While I was offline over the weekend I
already came up with a cleaner version that collects the changes in a
structure and does the logging outside of the critical section. I
currently does a few unnecessary writes into memory under
CONFIG_AUDIT=n, but if that is an issue I can boost the abstraction or
just add some #ifdefs to avoid that.

>
> The whole information can be reconstructed after the fact:
>
>    1) Copy the user space supplied struct timex to a buffer
>
>    2) Retrieve the current timex information _before_ invoking
>       do_adjtimex().
>
>    3) Look at the ret value and the resulting struct timex which is going
>       to be copied back to user space and figure out with the help of #1
>       and #2 what you need to log.
>
> That does not even need a single line of change in the NTP code and almost
> everything happens in fully preemptible context.

I worry that extracting everything from the timex structures might get
a little too complicated... Hopefully you'll be OK with the solution I
am preparing for v8 - it still adds a bit of code to ntp.c, but it's
much less intrusive.

Thanks a lot for the review!


--
Ondrej Mosnacek <omosnace at redhat dot com>
Software Engineer, Security Technologies
Red Hat, Inc.