Re: [PATCH] Arm: mm: ftrace: Only set text back to ro after kernel has been marked ro

[Matthias Reichl <hias@horus.com>]

On Wed, Aug 23, 2017 at 03:03:51PM -0400, Steven Rostedt wrote:
> On Wed, 23 Aug 2017 11:48:13 -0700
> Kees Cook <keescook@chromium.org> wrote:
> 
> > > diff --git a/arch/arm/mm/init.c b/arch/arm/mm/init.c
> > > index ad80548..fd75f38 100644
> > > --- a/arch/arm/mm/init.c
> > > +++ b/arch/arm/mm/init.c
> > > @@ -745,19 +745,29 @@ static int __mark_rodata_ro(void *unused)
> > >         return 0;
> > >  }
> > >
> > > +static int kernel_set_to_readonly;  
> > 
> > Adding a comment here might be a good idea, something like:
> > 
> > /* Has system boot-up reached mark_rodata_ro() yet? */
> 
> I don't mind adding a comment, but the above is rather self explanatory
> (one can easily see that it is set in mark_rodata_ro() with a simple
> search).
> 
> If a comment is to be added, something a bit more descriptive of the
> functionality of the variable would be appropriate:
> 
> /*
>  * Ignore modifying kernel text permissions until the kernel core calls
>  * make_rodata_ro() at system start up.
>  */
> 
> I can resend with the comment, or whoever takes this could add it
> themselves.

Gentle ping: this patch doesn't seem to have landed in upstream
trees yet. Is any more work required?

It would be nice to have this fix added. Just tested next-20171205
on RPi B+, it oopses when the function tracer is enabled during boot.
next-20171205 plus this patch boots up fine.

so long,

Hias

> 
> -- Steve
> 
> 
> > 
> > Otherwise:
> > 
> > Acked-by: Kees Cook <keescook@chromium.org>
> > 
> > > +
> > >  void mark_rodata_ro(void)
> > >  {
> > > +       kernel_set_to_readonly = 1;
> > > +
> > >         stop_machine(__mark_rodata_ro, NULL, NULL);
> > >  }
> > >
> > >  void set_kernel_text_rw(void)
> > >  {
> > > +       if (!kernel_set_to_readonly)
> > > +               return;
> > > +
> > >         set_section_perms(ro_perms, ARRAY_SIZE(ro_perms), false,
> > >                                 current->active_mm);
> > >  }
> > >
> > >  void set_kernel_text_ro(void)
> > >  {
> > > +       if (!kernel_set_to_readonly)
> > > +               return;
> > > +
> > >         set_section_perms(ro_perms, ARRAY_SIZE(ro_perms), true,
> > >                                 current->active_mm);
> > >  }  
> > 
> > Does arm64 suffer from a similar condition? (It looks like no, as text
> > patching is done with a fixmap poke.)
> > 
> > -Kees
> > 
>