Re: [PATCH rcu 07/12] rcu: tiny: Record kvfree_call_rcu() call stack for KASAN

[Marco Elver]

On Tue, 21 Jun 2022 at 21:31, Paul E. McKenney <paulmck@xxxxxxxxxx> wrote:
>
> On Tue, Jun 21, 2022 at 12:01:29PM +0530, Neeraj Upadhyay wrote:
> >
> >
> > On 6/21/2022 3:50 AM, Paul E. McKenney wrote:
> > > From: Johannes Berg <johannes.berg@xxxxxxxxx>
> > >
> > > When running KASAN with Tiny RCU (e.g. under ARCH=um, where
> > > a working KASAN patch is now available), we don't get any
> > > information on the original kfree_rcu() (or similar) caller
> > > when a problem is reported, as Tiny RCU doesn't record this.
> > >
> > > Add the recording, which required pulling kvfree_call_rcu()
> > > out of line for the KASAN case since the recording function
> > > (kasan_record_aux_stack_noalloc) is neither exported, nor
> > > can we include kasan.h into rcutiny.h.
> > >
> > > without KASAN, the patch has no size impact (ARCH=um kernel):
> > >      text       data         bss         dec        hex    filename
> > >   6151515    4423154    33148520    43723189    29b29b5    linux
> > >   6151515    4423154    33148520    43723189    29b29b5    linux + patch
> > >
> > > with KASAN, the impact on my build was minimal:
> > >      text       data         bss         dec        hex    filename
> > > 13915539    7388050    33282304    54585893    340ea25    linux
> > > 13911266    7392114    33282304    54585684    340e954    linux + patch
> > >     -4273      +4064         +-0        -209
> > >
> > > Acked-by: Dmitry Vyukov <dvyukov@xxxxxxxxxx>
> > > Signed-off-by: Johannes Berg <johannes.berg@xxxxxxxxx>
> > > Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxx>
> > > ---
> > >   include/linux/rcutiny.h | 11 ++++++++++-
> > >   kernel/rcu/tiny.c       | 14 ++++++++++++++
> > >   2 files changed, 24 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/include/linux/rcutiny.h b/include/linux/rcutiny.h
> > > index 5fed476f977f6..d84e13f2c3848 100644
> > > --- a/include/linux/rcutiny.h
> > > +++ b/include/linux/rcutiny.h
> > > @@ -38,7 +38,7 @@ static inline void synchronize_rcu_expedited(void)
> > >    */
> > >   extern void kvfree(const void *addr);
> > > -static inline void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
> > > +static inline void __kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
> > >   {
> > >     if (head) {
> > >             call_rcu(head, func);
> > > @@ -51,6 +51,15 @@ static inline void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
> > >     kvfree((void *) func);
> > >   }
> > > +#ifdef CONFIG_KASAN_GENERIC
> > > +void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func);
> > > +#else
> > > +static inline void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
> > > +{
> > > +   __kvfree_call_rcu(head, func);
> > > +}
> > > +#endif
> > > +
> > >   void rcu_qs(void);
> > >   static inline void rcu_softirq_qs(void)
> > > diff --git a/kernel/rcu/tiny.c b/kernel/rcu/tiny.c
> > > index 340b3f8b090d4..58ff3721d975c 100644
> > > --- a/kernel/rcu/tiny.c
> > > +++ b/kernel/rcu/tiny.c
> > > @@ -217,6 +217,20 @@ bool poll_state_synchronize_rcu(unsigned long oldstate)
> > >   }
> > >   EXPORT_SYMBOL_GPL(poll_state_synchronize_rcu);
> > > +#ifdef CONFIG_KASAN_GENERIC
> > > +void kvfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
> > > +{
> > > +   if (head) {
> > > +           void *ptr = (void *) head - (unsigned long) func;
> > > +
> > > +           kasan_record_aux_stack_noalloc(ptr);
> > > +   }
> >
> > For the !head case; similar to Tree RCU's kvfree_call_rcu() implementation,
> > we do not need to record 'ptr' (which will be 'func')?
>
> My understanding is that we do not need to record in that case
> because __kvfree_call_rcu() will simply invoke the almost-zero-cost
> synchronize_rcu() and then invoke kfree().
>
> Johannes, Dmitry, Marco, anything that I am missing?

As-is looks sensible - doing kasan_record_aux_stack_noalloc() only
makes sense if the actual kfree() is not done with a callstack that
will point at the kvfree_call_rcu() caller. Otherwise we're doing
redundant work and just polluting the aux stack storage slots. So in
the case where kvfree_call_rcu() does synchronize_rcu() and kfree()
the kvfree_call_rcu() caller is in the callstack, and would be shown
on use-after-free bugs.

Thanks,
-- Marco