Re: [PATCH v2] cpuidle: change enter_s2idle() prototype

["Rafael J. Wysocki" <rafael@kernel.org>]

On Fri, Jul 24, 2020 at 1:50 PM Neal Liu <neal.liu@mediatek.com> wrote:
>
> On Fri, 2020-07-24 at 13:20 +0200, Rafael J. Wysocki wrote:
> > On Fri, Jul 24, 2020 at 12:24 PM Neal Liu <neal.liu@mediatek.com> wrote:
> > >
> > > On Fri, 2020-07-24 at 11:57 +0200, Rafael J. Wysocki wrote:
> > > > On Thu, Jul 23, 2020 at 9:07 PM Sami Tolvanen <samitolvanen@google.com> wrote:
> > > > >
> > > > > On Mon, Jul 20, 2020 at 04:21:34PM +0800, Neal Liu wrote:
> > > > > > Gentle ping on this patch.
> > > > > >
> > > > > >
> > > > > > On Fri, 2020-07-10 at 11:08 +0800, Neal Liu wrote:
> > > > > > > On Thu, 2020-07-09 at 14:18 +0200, Rafael J. Wysocki wrote:
> > > > > > > > On Mon, Jul 6, 2020 at 5:13 AM Neal Liu <neal.liu@mediatek.com> wrote:
> > > > > > > > >
> > > > > > > > > Control Flow Integrity(CFI) is a security mechanism that disallows
> > > > > > > > > changes to the original control flow graph of a compiled binary,
> > > > > > > > > making it significantly harder to perform such attacks.
> > > > > > > > >
> > > > > > > > > init_state_node() assign same function callback to different
> > > > > > > > > function pointer declarations.
> > > > > > > > >
> > > > > > > > > static int init_state_node(struct cpuidle_state *idle_state,
> > > > > > > > >                            const struct of_device_id *matches,
> > > > > > > > >                            struct device_node *state_node) { ...
> > > > > > > > >         idle_state->enter = match_id->data; ...
> > > > > > > > >         idle_state->enter_s2idle = match_id->data; }
> > > > > > > > >
> > > > > > > > > Function declarations:
> > > > > > > > >
> > > > > > > > > struct cpuidle_state { ...
> > > > > > > > >         int (*enter) (struct cpuidle_device *dev,
> > > > > > > > >                       struct cpuidle_driver *drv,
> > > > > > > > >                       int index);
> > > > > > > > >
> > > > > > > > >         void (*enter_s2idle) (struct cpuidle_device *dev,
> > > > > > > > >                               struct cpuidle_driver *drv,
> > > > > > > > >                               int index); };
> > > > > > > > >
> > > > > > > > > In this case, either enter() or enter_s2idle() would cause CFI check
> > > > > > > > > failed since they use same callee.
> > > > > > > >
> > > > > > > > Can you please explain this in a bit more detail?
> > > > > > > >
> > > > > > > > As it stands, I don't understand the problem statement enough to apply
> > > > > > > > the patch.
> > > > > > > >
> > > > > > >
> > > > > > > Okay, Let's me try to explain more details.
> > > > > > > Control Flow Integrity(CFI) is a security mechanism that disallows
> > > > > > > changes to the original control flow graph of a compiled binary, making
> > > > > > > it significantly harder to perform such attacks.
> > > > > > >
> > > > > > > There are multiple control flow instructions that could be manipulated
> > > > > > > by the attacker and subvert control flow. The target instructions that
> > > > > > > use data to determine the actual destination.
> > > > > > > - indirect jump
> > > > > > > - indirect call
> > > > > > > - return
> > > > > > >
> > > > > > > In this case, function prototype between caller and callee are mismatch.
> > > > > > > Caller: (type A)funcA
> > > > > > > Callee: (type A)funcB
> > > > > > > Callee: (type C)funcC
> > > > > > >
> > > > > > > funcA calls funcB -> no problem
> > > > > > > funcA calls funcC -> CFI check failed
> > > > > > >
> > > > > > > That's why we try to align function prototype.
> > > > > > > Please feel free to feedback if you have any questions.
> > > > >
> > > > > I think you should include a better explanation in the commit message.
> > > > > Perhaps something like this?
> > > > >
> > > > >   init_state_node assigns the same callback function to both enter and
> > > > >   enter_s2idle despite mismatching function types, which trips indirect
> > > > >   call checking with Control-Flow Integrity (CFI).
> > > > >
> > > > > > > > > Align function prototype of enter() since it needs return value for
> > > > > > > > > some use cases. The return value of enter_s2idle() is no
> > > > > > > > > need currently.
> > > > > > > >
> > > > > > > > So last time I requested you to document why ->enter_s2idle needs to
> > > > > > > > return an int in the code, which has not been done.  Please do that.
> > > > >
> > > > > Rafael, are you happy with the commit message documenting the reason,
> > > > > or would you prefer to also add a comment before enter_s2idle?
> > > >
> > > > As I said before, it would be good to have a comment in the code as
> > > > well or people will be wondering why it is necessary to return
> > > > anything from that callback, because its return value is never used.
> > > >
> > > > Thanks!
> > >
> > > Is it okay to add these comments before enter_s2idle?
> > >
> > > /*
> > >  * Align function type since init_state_node assigns the same callback
> >
> > init_state_node()
> >
> > >  * function to both enter and enter_s2idle despite mismatching function
> >
> > ->enter_s2idle
> >
> > >  * types, which trips indirect call checking with Control-Flow Integrity
> > >  * (CFI).
> > >  */
> > > int (*enter_s2idle)(struct cpuidle_device *dev,
> > >                     struct cpuidle_driver *drv,
> > >                     int index);
> >
> > But IMO it would be sufficient to add something like this to the
> > existing comment regarding ->enter_s2idle:
> >
> > "This callback may point to the same function as ->enter if all of the
> > above requirements are met by it."
> >
> > That would explain why the signature is the same sufficiently in my view.
> >
> > Thanks!
>
> For clarification, do you mean add this comment on enter_s2idle function
> pointer declaration is enough?

Yes, I do.

Thanks!