[PATCH] vsprintf: Do not break early boot with probing addresses

Petr Mladek pmladek at suse.com
Mon May 13 22:24:24 AEST 2019


On Fri 2019-05-10 12:40:58, Steven Rostedt wrote:
> On Fri, 10 May 2019 18:32:58 +0200
> Martin Schwidefsky <schwidefsky at de.ibm.com> wrote:
> 
> > On Fri, 10 May 2019 12:24:01 -0400
> > Steven Rostedt <rostedt at goodmis.org> wrote:
> > 
> > > On Fri, 10 May 2019 10:42:13 +0200
> > > Petr Mladek <pmladek at suse.com> wrote:
> > >   
> > > >  static const char *check_pointer_msg(const void *ptr)
> > > >  {
> > > > -	char byte;
> > > > -
> > > >  	if (!ptr)
> > > >  		return "(null)";
> > > >  
> > > > -	if (probe_kernel_address(ptr, byte))
> > > > +	if ((unsigned long)ptr < PAGE_SIZE || IS_ERR_VALUE(ptr))
> > > >  		return "(efault)";
> > > >      
> > > 
> > > 
> > > 	< PAGE_SIZE ?
> > > 
> > > do you mean: < TASK_SIZE ?  
> > 
> > The check with < TASK_SIZE would break on s390. The 'ptr' is
> > in the kernel address space, *not* in the user address space.
> > Remember s390 has two separate address spaces for kernel/user
> > the check < TASK_SIZE only makes sense with a __user pointer.
> > 
> 
> So we allow this to read user addresses? Can't that cause a fault?

I did some quick check and did not found anywhere a user pointer
being dereferenced via vsprintf().

In each case, %s did the check (ptr < PAGE_SIZE) even before this
patchset. The other checks are in %p format modifiers that are
used to print various kernel structures.

Finally, it accesses the pointer directly. I am not completely sure
but I think that it would not work that easily with an address
from the user address space.

Best Regards,
Petr


More information about the Linuxppc-dev mailing list