There is a small race between copy_process() and sched_fork()
where child->sched_task_group point to an already freed pointer.
parent doing fork() | someone moving the parent
to another cgroup
-------------------------------+-------------------------------
copy_process()
+ dup_task_struct()<1>
parent move to another cgroup,
and free the old cgroup. <2>
+ sched_fork()
+ __set_task_cpu()<3>
+ task_fork_fair()
+ sched_slice()<4>
In the worst case, this bug can lead to "use-after-free" and
cause panic as shown above,
(1)parent copy its sched_task_group to child at <1>;
(2)someone move the parent to another cgroup and free the old
cgroup at <2>;
(3)the sched_task_group and cfs_rq that belong to the old cgroup
will be accessed at <3> and <4>, which cause a panic:
[89249.732198] BUG: unable to handle kernel NULL pointer
dereference at 0000000000000000
[89249.732701] PGD 8000001fa0a86067 P4D 8000001fa0a86067 PUD
2029955067 PMD 0
[89249.733005] Oops: 0000 [#1] SMP PTI
[89249.733288] CPU: 7 PID: 648398 Comm: ebizzy Kdump: loaded
Tainted: G OE --------- - - 4.18.0.x86_64+ #1
[89249.734318] RIP: 0010:sched_slice+0x84/0xc0
....
[89249.737910] Call Trace:
[89249.738181] task_fork_fair+0x81/0x120
[89249.738457] sched_fork+0x132/0x240
[89249.738732] copy_process.part.5+0x675/0x20e0
[89249.739010] ? __handle_mm_fault+0x63f/0x690
[89249.739286] _do_fork+0xcd/0x3b0
[89249.739558] do_syscall_64+0x5d/0x1d0
[89249.739830] entry_SYSCALL_64_after_hwframe+0x65/0xca
[89249.740107] RIP: 0033:0x7f04418cd7e1
Between cgroup_can_fork() and cgroup_post_fork(), the cgroup
membership is fixed and thus sched_task_group can't change. So
call sched_fork() after cgroup_can_fork() and update the child's
sched_task_group before it is used.
Fixes: 8323f26ce342 ("sched: Fix race in task_group")
Signed-off-by: Zhang Qiao <[email protected]>
---
Changes since v1:
- revert changes of sched_post_fork()
- call sched_fork() after cgroup_can_fork()
---
include/linux/sched/task.h | 3 ++-
kernel/fork.c | 17 +++++++++--------
kernel/sched/core.c | 9 ++++++++-
3 files changed, 19 insertions(+), 10 deletions(-)
diff --git a/include/linux/sched/task.h b/include/linux/sched/task.h
index ef02be869cf2..57eb217f1a86 100644
--- a/include/linux/sched/task.h
+++ b/include/linux/sched/task.h
@@ -53,7 +53,8 @@ extern int lockdep_tasklist_lock_is_held(void);
extern asmlinkage void schedule_tail(struct task_struct *prev);
extern void init_idle(struct task_struct *idle, int cpu);
-extern int sched_fork(unsigned long clone_flags, struct task_struct *p);
+extern int sched_fork(unsigned long clone_flags, struct task_struct *p,
+ struct kernel_clone_args *kargs);
extern void sched_post_fork(struct task_struct *p);
extern void sched_dead(struct task_struct *p);
diff --git a/kernel/fork.c b/kernel/fork.c
index 38681ad44c76..1874fe754a56 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -2160,11 +2160,6 @@ static __latent_entropy struct task_struct *copy_process(
p->bpf_ctx = NULL;
#endif
- /* Perform scheduler related setup. Assign this task to a CPU. */
- retval = sched_fork(clone_flags, p);
- if (retval)
- goto bad_fork_cleanup_policy;
-
retval = perf_event_init_task(p, clone_flags);
if (retval)
goto bad_fork_cleanup_policy;
@@ -2295,6 +2290,11 @@ static __latent_entropy struct task_struct *copy_process(
if (retval)
goto bad_fork_put_pidfd;
+ /* Perform scheduler related setup. Assign this task to a CPU. */
+ retval = sched_fork(clone_flags, p, args);
+ if (retval)
+ goto bad_fork_cancel_cgroup;
+
/*
* From this point on we must avoid any synchronous user-space
* communication until we take the tasklist-lock. In particular, we do
@@ -2343,13 +2343,13 @@ static __latent_entropy struct task_struct *copy_process(
/* Don't start children in a dying pid namespace */
if (unlikely(!(ns_of_pid(pid)->pid_allocated & PIDNS_ADDING))) {
retval = -ENOMEM;
- goto bad_fork_cancel_cgroup;
+ goto bad_fork_clean_sched;
}
/* Let kill terminate clone/fork in the middle */
if (fatal_signal_pending(current)) {
retval = -EINTR;
- goto bad_fork_cancel_cgroup;
+ goto bad_fork_clean_sched;
}
/* past the last point of failure */
@@ -2416,10 +2416,11 @@ static __latent_entropy struct task_struct *copy_process(
return p;
-bad_fork_cancel_cgroup:
+bad_fork_clean_sched:
sched_core_free(p);
spin_unlock(¤t->sighand->siglock);
write_unlock_irq(&tasklist_lock);
+bad_fork_cancel_cgroup:
cgroup_cancel_fork(p, args);
bad_fork_put_pidfd:
if (clone_flags & CLONE_PIDFD) {
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index c4462c454ab9..acb30e418c4d 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -4326,10 +4326,17 @@ int sysctl_schedstats(struct ctl_table *table, int write, void *buffer,
/*
* fork()/clone()-time setup:
*/
-int sched_fork(unsigned long clone_flags, struct task_struct *p)
+int sched_fork(unsigned long clone_flags, struct task_struct *p,
+ struct kernel_clone_args *kargs)
{
unsigned long flags;
+#ifdef CONFIG_CGROUP_SCHED
+ struct task_group *tg;
+ tg = container_of(kargs->cset->subsys[cpu_cgrp_id],
+ struct task_group, css);
+ p->sched_task_group = autogroup_task_group(p, tg);
+#endif
__sched_fork(clone_flags, p);
/*
* We mark the process as NEW here. This guarantees that
--
2.25.1
Hello,
On Sat, Sep 11, 2021 at 03:50:54PM +0800, Zhang Qiao wrote:
> Between cgroup_can_fork() and cgroup_post_fork(), the cgroup
> membership is fixed and thus sched_task_group can't change. So
> call sched_fork() after cgroup_can_fork() and update the child's
> sched_task_group before it is used.
The part being fixed looks correct to me but it's difficult to for me to
assess whether the whole relocation of the sched_fork() hook doesn't change
anything else. Besides, even if we decide to relocate the sched_fork hook, I
think it'd be better to separate the two changes - one is a relatively safe
bug fix, the other is a code reorganization with possibly subtle side
effects. So, I think it'd be better to produce a patch which just fixes the
bug even if that ends up introducing another function in the flow.
Thanks.
--
tejun
?? 2021/9/14 2:01, Tejun Heo д??:
> Hello,
>
> On Sat, Sep 11, 2021 at 03:50:54PM +0800, Zhang Qiao wrote:
>> Between cgroup_can_fork() and cgroup_post_fork(), the cgroup
>> membership is fixed and thus sched_task_group can't change. So
>> call sched_fork() after cgroup_can_fork() and update the child's
>> sched_task_group before it is used.
>
> The part being fixed looks correct to me but it's difficult to for me to
> assess whether the whole relocation of the sched_fork() hook doesn't change
Indeed is this, although I checked and tested it many times.
> anything else. Besides, even if we decide to relocate the sched_fork hook, I
> think it'd be better to separate the two changes - one is a relatively safe
> bug fix, the other is a code reorganization with possibly subtle side
> effects. So, I think it'd be better to produce a patch which just fixes the
> bug even if that ends up introducing another function in the flow.
Perhaps we can fix this bug at sched_post_fork(), which is executed between cgroup_can_fork()
and cgroup_post_fork(). Relocate the fectching of task_group and access it at sched_post_fork().
thanks.
>
> Thanks.
>
On Tue, Sep 14, 2021 at 04:20:02PM +0800, Zhang Qiao wrote:
> Perhaps we can fix this bug at sched_post_fork(), which is executed between cgroup_can_fork()
> and cgroup_post_fork(). Relocate the fectching of task_group and access it at sched_post_fork().
Can't we just add a can_fork callback in cpu_cgrp_subsys?
--
tejun
?? 2021/9/15 0:23, Tejun Heo д??:
> On Tue, Sep 14, 2021 at 04:20:02PM +0800, Zhang Qiao wrote:
>> Perhaps we can fix this bug at sched_post_fork(), which is executed between cgroup_can_fork()
>> and cgroup_post_fork(). Relocate the fectching of task_group and access it at sched_post_fork().
>
> Can't we just add a can_fork callback in cpu_cgrp_subsys?
Because __set_task_cpu and task_fork callback need access the sched_task_group field, so the task_fork
callback must be called after update the child's sched_task_group, if perform these
operations in the newly added can_fork callback, it may lead to some dependencies(the can_fork will
not be executed when !CONFIG_CGROUP_SCHED).
__set_task_cpu and task_fork callback mainly do some field initialization and vruntime calculation,
these things just need to be done before waking up the new task. So i think it is no problem to defer
do this things at sched_post_fork().
thanks.
>