Re: [PATCH] sched: fix another race when reading /proc/sched_debug

[Ingo Molnar <mingo@elte.hu>]

* Li Zefan <lizf@cn.fujitsu.com> wrote:

> I fixed an oops with the following commit:
> 
> | commit 24eb089950ce44603b30a3145a2c8520e2b55bb1
> | Author: Li Zefan <lizf@cn.fujitsu.com>
> | Date:   Thu Nov 6 12:53:32 2008 -0800
> |
> |    cgroups: fix invalid cgrp->dentry before cgroup has been completely removed
> |
> |    This fixes an oops when reading /proc/sched_debug.
> 
> The above commit fixed a race that reading /proc/sched_debug may access
> NULL cgrp->dentry if a cgroup is being removed.
> 
> But I found there's another different race, in that reading sched_debug
> may access a cgroup which hasn't been completed created, and thus
> dereference NULL cgrp->dentry!
> 
> cgroup_create()
>   cpu_cgroup_create()
>     register_fair_sched_group()
>       list_add_rcu(...)
> 				print_cfs_stats()
> 				  for_each_leaf_cfs_rq()
> 				    print_cfs_rq()
> 				      cgroup_path()
>   cgroup->dentry = dentry;
> 
> task_group is added to the global list before the cgroup has been created
> completely, if at this time print_cfs_stats() is called, it will access
> the half-created cgroup.
> 
> This patch fixes the bug by holding cgroup_lock() to wait the cgroup to
> be created completely before calling cgroup_path().
> 
> Signed-off-by: Li Zefan <lizf@cn.fujitsu.com>

applied to tip/sched/urgent, thanks!

> The patch is based on linus's git tree, and should go into 2.6.28,
> but it conflicts with the cleanup patch in sched/core:
> 	0a0db8f5c9d4bbb9bbfcc2b6cb6bce2d0ef4d73d

i merged it up in tip/master, could you please check whether it's ok?

	Ingo