lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20090315214244.GA12965@Krystal>
Date:	Sun, 15 Mar 2009 17:42:44 -0400
From:	Mathieu Desnoyers <mathieu.desnoyers@...ymtl.ca>
To:	Frederic Weisbecker <fweisbec@...il.com>
Cc:	Ingo Molnar <mingo@...e.hu>, akpm@...ux-foundation.org,
	Steven Rostedt <rostedt@...dmis.org>,
	LKML <linux-kernel@...r.kernel.org>,
	"Paul E. McKenney" <paulmck@...ux.vnet.ibm.com>
Subject: Re: [RFC patch 21/21] LTTng Kernel Trace Thread Flag API

* Frederic Weisbecker (fweisbec@...il.com) wrote:
> On Sun, Mar 15, 2009 at 03:04:01PM -0400, Mathieu Desnoyers wrote:
> > Add an API to set/clear the kernel wide tracing thread flags. Implemented in
> > kernel/sched.c. Updates thread flags *asynchronously* while holding the tasklist
> > lock.
> > 
> > Upon fork, the flag must be re-copied while the tasklist lock is held.
> > 
> > Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@...ymtl.ca>
> > ---
> >  include/linux/sched.h |    3 ++
> >  kernel/fork.c         |    9 ++++++++
> >  kernel/sched.c        |   55 ++++++++++++++++++++++++++++++++++++++++++++++++++
> >  3 files changed, 67 insertions(+)
> > 
> > Index: linux-2.6-lttng/include/linux/sched.h
> > ===================================================================
> > --- linux-2.6-lttng.orig/include/linux/sched.h	2009-01-09 18:15:54.000000000 -0500
> > +++ linux-2.6-lttng/include/linux/sched.h	2009-01-09 18:17:50.000000000 -0500
> > @@ -2300,6 +2300,9 @@ static inline void mm_init_owner(struct 
> >  
> >  #define TASK_STATE_TO_CHAR_STR "RSDTtZX"
> >  
> > +extern void clear_kernel_trace_flag_all_tasks(void);
> > +extern void set_kernel_trace_flag_all_tasks(void);
> > +
> >  #endif /* __KERNEL__ */
> >  
> >  #endif
> > Index: linux-2.6-lttng/kernel/fork.c
> > ===================================================================
> > --- linux-2.6-lttng.orig/kernel/fork.c	2009-01-09 18:17:38.000000000 -0500
> > +++ linux-2.6-lttng/kernel/fork.c	2009-01-09 18:17:50.000000000 -0500
> > @@ -1218,6 +1218,15 @@ static struct task_struct *copy_process(
> >  			!cpu_online(task_cpu(p))))
> >  		set_task_cpu(p, smp_processor_id());
> >  
> > +	/*
> > +	 * The state of the parent's TIF_KTRACE flag may have changed
> > +	 * since it was copied in dup_task_struct() so we re-copy it here.
> > +	 */
> > +	if (test_thread_flag(TIF_KERNEL_TRACE))
> > +		set_tsk_thread_flag(p, TIF_KERNEL_TRACE);
> > +	else
> > +		clear_tsk_thread_flag(p, TIF_KERNEL_TRACE);
> > +
> 
> 
> I thought about that too. But the thread info is already
> copied on a fork right? Which includes the thread flags.
> 
> But perhaps I'm wrong and miss one detail about this copy...
> 

If I remember correctly, the copy you are talking about is done before
taking the task list write lock, and therefore races with the task list
read lock we take below.


> 
> >  	/* CLONE_PARENT re-uses the old parent */
> >  	if (clone_flags & (CLONE_PARENT|CLONE_THREAD))
> >  		p->real_parent = current->real_parent;
> > Index: linux-2.6-lttng/kernel/sched.c
> > ===================================================================
> > --- linux-2.6-lttng.orig/kernel/sched.c	2009-01-09 18:17:38.000000000 -0500
> > +++ linux-2.6-lttng/kernel/sched.c	2009-01-09 18:17:50.000000000 -0500
> > @@ -9395,3 +9395,58 @@ struct cgroup_subsys cpuacct_subsys = {
> >  	.subsys_id = cpuacct_subsys_id,
> >  };
> >  #endif	/* CONFIG_CGROUP_CPUACCT */
> > +
> > +static DEFINE_MUTEX(kernel_trace_mutex);
> > +static int kernel_trace_refcount;
> > +
> > +/**
> > + * clear_kernel_trace_flag_all_tasks - clears all TIF_KERNEL_TRACE thread flags.
> > + *
> > + * This function iterates on all threads in the system to clear their
> > + * TIF_KERNEL_TRACE flag. Setting the TIF_KERNEL_TRACE flag with the
> > + * tasklist_lock held in copy_process() makes sure that once we finish clearing
> > + * the thread flags, all threads have their flags cleared.
> > + */
> > +void clear_kernel_trace_flag_all_tasks(void)
> > +{
> > +	struct task_struct *p;
> > +	struct task_struct *t;
> > +
> > +	mutex_lock(&kernel_trace_mutex);
> > +	if (--kernel_trace_refcount)
> > +		goto end;
> > +	read_lock(&tasklist_lock);
> > +	do_each_thread(p, t) {
> > +		clear_tsk_thread_flag(t, TIF_KERNEL_TRACE);
> > +	} while_each_thread(p, t);
> > +	read_unlock(&tasklist_lock);
> > +end:
> > +	mutex_unlock(&kernel_trace_mutex);
> > +}
> > +EXPORT_SYMBOL_GPL(clear_kernel_trace_flag_all_tasks);
> > +
> > +/**
> > + * set_kernel_trace_flag_all_tasks - sets all TIF_KERNEL_TRACE thread flags.
> > + *
> > + * This function iterates on all threads in the system to set their
> > + * TIF_KERNEL_TRACE flag. Setting the TIF_KERNEL_TRACE flag with the
> > + * tasklist_lock held in copy_process() makes sure that once we finish setting
> > + * the thread flags, all threads have their flags set.
> > + */
> > +void set_kernel_trace_flag_all_tasks(void)
> > +{
> > +	struct task_struct *p;
> > +	struct task_struct *t;
> > +
> > +	mutex_lock(&kernel_trace_mutex);
> > +	if (kernel_trace_refcount++)
> > +		goto end;
> > +	read_lock(&tasklist_lock);
> > +	do_each_thread(p, t) {
> > +		set_tsk_thread_flag(t, TIF_KERNEL_TRACE);
> > +	} while_each_thread(p, t);
> > +	read_unlock(&tasklist_lock);
> > +end:
> > +	mutex_unlock(&kernel_trace_mutex);
> > +}
> > +EXPORT_SYMBOL_GPL(set_kernel_trace_flag_all_tasks);
> 
> 
> Speaking about optimizations and latency. If you want to avoid
> the overhead of a whole tasklist iteration making any writer spinning on
> you, may be you can iterate it using rcu.
> 
> But it can be slightly racy in that you could miss a freshly inserted
> task while playing with a nano delayed version of the list. So this is not
> necessarily a good idea.

I have not found any good way to do it racelessly with RCU read lock
yet. We'd have to think carefully about it.

Mathieu



> 
> 
> > 
> > -- 
> > Mathieu Desnoyers
> > OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68
> 

-- 
Mathieu Desnoyers
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ