Messages in this thread | | | Date | Thu, 18 Jul 2013 16:51:37 +0200 | From | Oleg Nesterov <> | Subject | Re: [RFC PATCH V2] tracing: Check f_dentry before accessing event_file/call in inode->i_private |
| |
On 07/18, Masami Hiramatsu wrote: > > (2013/07/17 23:51), Oleg Nesterov wrote: > > Well, perhaps you are right... But this TRACE_EVENT_FL_REF_MASK code > > is new too, it is not that we only need a small fixlets to finish it. > > Would you mean that TRACE_EVENT_FL_REF_MASK may also have some problems?
It was you who initially pointed that it does have problems ;)
And, _afaics_ your patch which tries to fix this problem is not exactly correct.
It removes trace_array_get/put from tracing_open_generic_file() and tracing_release_generic_file(). This assumes that "call->flags++" is enough, but it is not.
Yes, the next patch adds the "flags & TRACE_EVENT_FL_REF_MASK" check into trace_remove_event_call() path. But this is still racy wrt instance_delete() unless I missed something.
IOW, I believe that either .open() should do trace_array_get(), or __trace_remove_event_dirs() needs another for-each-file loop which checks file->call->flags & TRACE_EVENT_FL_REF_MASK.
> > So I think that it makes sense to discuss the alternatives before we > > decide what exactly we should do. > > Your approach is also interesting for me, indeed. However, it is so > different from current one. I think you should clarify what bug you > would like to solve and how.
The same bugs which Steven's 1/4 tries to solve ;)
Oleg.
| |