2021-05-15 22:38:47

by Khaled Romdhani

[permalink] [raw]
Subject: [PATCH-next] sched: Fix Null pointer derefrence

The 'curr' variable could be NULL and derefrenced by
pick_next_entity. Fix this by adding a check that prevent
the invocation of pick_next_entity with a NULL passed argument.

Addresses-Coverity: ("Explicit null dereferenced")
Signed-off-by: Khaled ROMDHANI <[email protected]>
---
kernel/sched/fair.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 161b92aa1c79..490948df94fe 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -7075,7 +7075,7 @@ static struct task_struct *pick_task_fair(struct rq *rq)
else
curr = NULL;

- if (unlikely(check_cfs_rq_runtime(cfs_rq)))
+ if (unlikely(check_cfs_rq_runtime(cfs_rq) || !curr))
goto again;
}

--
2.17.1



2021-05-16 09:25:41

by Peter Zijlstra

[permalink] [raw]
Subject: Re: [PATCH-next] sched: Fix Null pointer derefrence

On Sat, May 15, 2021 at 05:46:45PM +0100, Khaled ROMDHANI wrote:
> The 'curr' variable could be NULL and derefrenced by
> pick_next_entity. Fix this by adding a check that prevent
> the invocation of pick_next_entity with a NULL passed argument.

And why exactly is that a problem?


2021-05-16 16:38:57

by Khaled Romdhani

[permalink] [raw]
Subject: Re: [PATCH-next] sched: Fix Null pointer derefrence

On Sat, May 15, 2021 at 11:02:38PM +0200, Peter Zijlstra wrote:
> On Sat, May 15, 2021 at 05:46:45PM +0100, Khaled ROMDHANI wrote:
> > The 'curr' variable could be NULL and derefrenced by
> > pick_next_entity. Fix this by adding a check that prevent
> > the invocation of pick_next_entity with a NULL passed argument.
>
> And why exactly is that a problem?
>
Within pick_next_entity, we could have 'left = curr'.
Thus the function wakeup_preempt_entity which is invoked
from pick_next_entity, derefrence the the sched entity 'left'
that may be a NULL passed argument.

2021-05-16 16:55:06

by Peter Zijlstra

[permalink] [raw]
Subject: Re: [PATCH-next] sched: Fix Null pointer derefrence

On Sun, May 16, 2021 at 02:01:29PM +0100, Khaled Romdhani wrote:
> On Sat, May 15, 2021 at 11:02:38PM +0200, Peter Zijlstra wrote:
> > On Sat, May 15, 2021 at 05:46:45PM +0100, Khaled ROMDHANI wrote:
> > > The 'curr' variable could be NULL and derefrenced by
> > > pick_next_entity. Fix this by adding a check that prevent
> > > the invocation of pick_next_entity with a NULL passed argument.
> >
> > And why exactly is that a problem?
> >
> Within pick_next_entity, we could have 'left = curr'.
> Thus the function wakeup_preempt_entity which is invoked
> from pick_next_entity, derefrence the the sched entity 'left'
> that may be a NULL passed argument.

We call pick_next_entity() from pick_task_fair() (or
pick_next_task_fair(), same argument for both), we only call it when
cfs_rq->nr_running != 0. IOW we *know* there is at least one task.

Therefore, if curr == NULL, we *must* have left.

If you can't read code, stay away from Coverity, it's crap.