Re: [PATCH] pid: add null pointer check in pid_nr_ns()

[Alakesh Haloi]

On Wed, Dec 02, 2020 at 11:19:46AM -0600, Eric W. Biederman wrote:
> Alakesh Haloi <alakesh.haloi@xxxxxxxxx> writes:
> 
> > There has been at least one occurrence where a null pointer derefernce
> > panic was seen with following stack trace.
> >
> >  #0 [ffffff800bcd3800] machine_kexec at ffffff8008095fb4
> >  #1 [ffffff800bcd3860] __crash_kexec at ffffff8008122a30
> >  #2 [ffffff800bcd39f0] panic at ffffff80080aa054
> >  #3 [ffffff800bcd3ae0] die at ffffff800808aee8
> >  #4 [ffffff800bcd3b20] die_kernel_fault at ffffff8008099520
> >  #5 [ffffff800bcd3b50] __do_kernel_fault at ffffff8008098e50
> >  #6 [ffffff800bcd3b80] do_translation_fault at ffffff800809929c
> >  #7 [ffffff800bcd3b90] do_mem_abort at ffffff8008081204
> >  #8 [ffffff800bcd3d90] el1_ia at ffffff800808304c
> >      PC: ffffff80080c20ec  [pid_nr_ns+4]
> >      LR: ffffff80080c231c  [__task_pid_nr_ns+72]
> >      SP: ffffff800bcd3da0  PSTATE: 60000005
> >     X29: ffffff800bcd3da0  X28: ffffffc00691c380  X27: 0000000000000001
> >     X26: 00000000004ce8e8  X25: 00000000004ce8d0  X24: ffffffc00691c3e0
> >     X23: ffffffc004e8c000  X22: 0000000000000000  X21: ffffffc00b042ed2
> >     X20: ffffff800876a4f0  X19: 0000000000000000  X18: 0000000000000000
> >     X17: 0000000000000001  X16: 0000000000000000  X15: 0000000000000000
> >     X14: 0000000400000003  X13: 0000000000000008  X12: fefefefefefefeff
> >     X11: 0000000000000000  X10: 0000007fffffffff   X9: 00000000004ce8b0
> >      X8: 00000000004ce8b0   X7: 0000000000000000   X6: ffffffc00b042ed2
> >      X5: ffffffc00b042ed2   X4: 0000000000020008   X3: 53206e69616c702f
> >      X2: ffffff800876a4f0   X1: ffffff800876a4f0   X0: 53206e69616c702f
> >  #9 [ffffff800bcd3da0] pid_nr_ns at ffffff80080c20e8
> 
> I just skimmed through the callers of pid_nr_ns and now I am very
> puzzled. I don't see any of them where the namespace can be passed as
> NULL.
> 
> So I really suspect you have a larger but somewhere in the caller of
> pid_nr_ns.  Perhaps the memory was stomped and you were lucky it was
> NULL.
> 
> Without some more details I really don't think testing for a NULL
> namespace is useful or productive.  At best it will mask bugs in the
> callers
> 
> Eric
> 
Thanks Eric for your time. I do not have any evidence of memory
corruption yet, but I agree with your concerns about not needing a null
check here.

--Alakesh
> > Signed-off-by: Alakesh Haloi <alakesh.haloi@xxxxxxxxx>
> > Cc: stable@xxxxxxxxxxxxxxx
> > ---
> >  kernel/pid.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/kernel/pid.c b/kernel/pid.c
> > index a96bc4bf4f86..3767b9e1431d 100644
> > --- a/kernel/pid.c
> > +++ b/kernel/pid.c
> > @@ -474,7 +474,7 @@ pid_t pid_nr_ns(struct pid *pid, struct pid_namespace *ns)
> >  	struct upid *upid;
> >  	pid_t nr = 0;
> >  
> > -	if (pid && ns->level <= pid->level) {
> > +	if (pid && ns && ns->level <= pid->level) {
> >  		upid = &pid->numbers[ns->level];
> >  		if (upid->ns == ns)
> >  			nr = upid->nr;