On 2018-07-20 18:14, Paul Moore wrote:
> On Wed, Jun 6, 2018 at 1:01 PM Richard Guy Briggs <rgb@xxxxxxxxxx> wrote:
> > Standalone audit records have the timestamp and serial number generated
> > on the fly and as such are unique, making them standalone. This new
> > function audit_alloc_local() generates a local audit context that will
> > be used only for a standalone record and its auxiliary record(s). The
> > context is discarded immediately after the local associated records are
> > produced.
> >
> > Signed-off-by: Richard Guy Briggs <rgb@xxxxxxxxxx>
> > ---
> > include/linux/audit.h | 8 ++++++++
> > kernel/auditsc.c | 25 +++++++++++++++++++++++--
> > 2 files changed, 31 insertions(+), 2 deletions(-)
>
> ...
>
> > --- a/kernel/auditsc.c
> > +++ b/kernel/auditsc.c
> > @@ -916,8 +916,11 @@ static inline void audit_free_aux(struct audit_context *context)
> > static inline struct audit_context *audit_alloc_context(enum audit_state state)
> > {
> > struct audit_context *context;
> > + gfp_t gfpflags;
> >
> > - context = kzalloc(sizeof(*context), GFP_KERNEL);
> > + /* We can be called in atomic context via audit_tg() */
> > + gfpflags = (in_atomic() || irqs_disabled()) ? GFP_ATOMIC : GFP_KERNEL;
>
> Instead of trying to guess the proper gfp flags, and likely getting it
> wrong at some point (the in_atomic() comment in preempt.h don't give
> me the warm fuzzies), why not pass the gfp flags as an argument?
>
> Right now it looks like we would only have two callers: audit_alloc()
> and audit_audit_local(). The audit_alloc() invocation would simply
> pass GFP_KERNEL and we could allow the audit_alloc_local() callers to
> specify the gfp flags when calling audit_alloc_local() (although I
> suspect that will always be GFP_ATOMIC since we should only be calling
> audit_alloc_local() from interrupt-like context, in all other cases we
> should use the audit_context from the current task_struct.
Ok, I'll explicitly pass it in.
> > + context = kzalloc(sizeof(*context), gfpflags);
> > if (!context)
> > return NULL;
> > context->state = state;
> > @@ -993,8 +996,26 @@ struct audit_task_info init_struct_audit = {
> > .ctx = NULL,
> > };
> >
> > -static inline void audit_free_context(struct audit_context *context)
> > +struct audit_context *audit_alloc_local(void)
> > {
>
> Let's see where this goes, but we may want to rename this slightly to
> indicate that this should only be called from interrupt context when
> we can't rely on current's audit_context. Bonus points if we can find
> a way to enforce this with a WARN() assertion so we can better catch
> abuse.
I'll see what I can come up with.
> > + struct audit_context *context;
> > +
> > + if (!audit_ever_enabled)
> > + return NULL; /* Return if not auditing. */
> > +
> > + context = audit_alloc_context(AUDIT_RECORD_CONTEXT);
> > + if (!context)
> > + return NULL;
> > + context->serial = audit_serial();
> > + context->ctime = current_kernel_time64();
> > + context->in_syscall = 1;
>
> Setting in_syscall is both interesting and a bit troubling, if for no
> other reason than I expect most (all?) callers to be in an interrupt
> context when audit_alloc_local() is called. Setting in_syscall would
> appear to be conceptually in this case. Can you help explain why this
> is the right thing to do, or necessary to ensure things are handled
> correctly?
I'll admit this is cheating a bit, but seemed harmless. It is needed so
that auditsc_get_stamp() from audit_get_stamp() from audit_log_start()
doesn't bail on me without giving me its already assigned time and
serial values rather than generating a new one. I did look to see if
there were any other undesireable side effects and found none, so I'm
tmepted to rename the ->in_syscall to something a bit more helpful. I
could add a new audit_context structure member to make
auditsc_get_stamp() co-operative, but this seems wasteful and
unnecessary.
> Looking quickly at the audit code, it seems to only be used on record
> and/or syscall termination to end things properly as well as in some
> of the PATH record code paths to limit filename collection to actual
> syscalls. However, this was just a quick look so I could be missing
> some important points.
>
> > + return context;
> > +}
> > +
> > +void audit_free_context(struct audit_context *context)
> > +{
> > + if (!context)
> > + return;
> > audit_free_names(context);
> > unroll_tree_refs(context, NULL, 0);
> > free_tree_refs(context);
> > --
> > 1.8.3.1
> >
> > --
> > Linux-audit mailing list
> > Linux-audit@xxxxxxxxxx
> > https://www.redhat.com/mailman/listinfo/linux-audit
>
> --
> paul moore
> www.paul-moore.com
- RGB
--
Richard Guy Briggs <rgb@xxxxxxxxxx>
Sr. S/W Engineer, Kernel Security, Base Operating Systems
Remote, Ottawa, Red Hat Canada
IRC: rgb, SunRaycer
Voice: +1.647.777.2635, Internal: (81) 32635
--
To unsubscribe from this list: send the line "unsubscribe linux-api" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
