linux-kernel - Re: [PATCH RT v3 4/5] rcu: Disable use_softirq on PREEMPT

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] [thread-next>] [day] [month] [year] [list]

Message-ID: <20190912221941.GD150506@google.com>
Date:   Thu, 12 Sep 2019 18:19:41 -0400
From:   Joel Fernandes <joel@...lfernandes.org>
To:     Scott Wood <swood@...hat.com>
Cc:     Sebastian Andrzej Siewior <bigeasy@...utronix.de>,
        linux-rt-users@...r.kernel.org, linux-kernel@...r.kernel.org,
        "Paul E . McKenney" <paulmck@...ux.ibm.com>,
        Thomas Gleixner <tglx@...utronix.de>,
        Steven Rostedt <rostedt@...dmis.org>,
        Peter Zijlstra <peterz@...radead.org>,
        Juri Lelli <juri.lelli@...hat.com>,
        Clark Williams <williams@...hat.com>, rcu@...r.kernel.org
Subject: Re: [PATCH RT v3 4/5] rcu: Disable use_softirq on PREEMPT_RT

On Thu, Sep 12, 2019 at 05:38:43PM -0400, Joel Fernandes wrote:
> Hi Scott,
> 
> Would you mind CC'ing rcu@...r.kernel.org on RCU related patches? I added it
> for this time.
> 
> On Wed, Sep 11, 2019 at 05:57:28PM +0100, Scott Wood wrote:
> > Besides restoring behavior that used to be default on RT, this avoids
> > a deadlock on scheduler locks:
[snip]
> > [  136.995194] 039:  May be due to missing lock nesting notation
> > 
> > [  137.001115] 039: 3 locks held by rcu_torture_rea/13474:
> > [  137.006341] 039:  #0:
> > [  137.008707] 039: 000000005f25146d
> > [  137.012024] 039:  (
> > [  137.014131] 039: &p->pi_lock
> > [  137.017015] 039: ){-...}
> > [  137.019558] 039: , at: try_to_wake_up+0x39/0x920
> > [  137.024175] 039:  #1:
> > [  137.026540] 039: 0000000011c8e51d
> > [  137.029859] 039:  (
> > [  137.031966] 039: &rq->lock
> > [  137.034679] 039: ){-...}
> > [  137.037217] 039: , at: try_to_wake_up+0x241/0x920
> > [  137.041924] 039:  #2:
> > [  137.044291] 039: 00000000098649b9
> > [  137.047610] 039:  (
> > [  137.049714] 039: rcu_read_lock
> > [  137.052774] 039: ){....}
> > [  137.055314] 039: , at: cpuacct_charge+0x33/0x1e0
> > [  137.059934] 039:
> > stack backtrace:
> > [  137.063425] 039: CPU: 39 PID: 13474 Comm: rcu_torture_rea Kdump: loaded Tainted: G            E     5.2.9-rt3.dbg+ #174
> > [  137.074197] 039: Hardware name: Intel Corporation S2600BT/S2600BT, BIOS SE5C620.86B.01.00.0763.022420181017 02/24/2018
> > [  137.084886] 039: Call Trace:
> > [  137.087773] 039:  <IRQ>
> > [  137.090226] 039:  dump_stack+0x5e/0x8b
> > [  137.093997] 039:  __lock_acquire+0x725/0x1100
> > [  137.098358] 039:  lock_acquire+0xc0/0x240
> > [  137.102374] 039:  ? try_to_wake_up+0x39/0x920
> > [  137.106737] 039:  _raw_spin_lock_irqsave+0x47/0x90
> > [  137.111534] 039:  ? try_to_wake_up+0x39/0x920
> > [  137.115910] 039:  try_to_wake_up+0x39/0x920
> > [  137.120098] 039:  rcu_read_unlock_special+0x65/0xb0
> > [  137.124977] 039:  __rcu_read_unlock+0x5d/0x70
> > [  137.129337] 039:  cpuacct_charge+0xd9/0x1e0
> > [  137.133522] 039:  ? cpuacct_charge+0x33/0x1e0
> > [  137.137880] 039:  update_curr+0x14b/0x420
> > [  137.141894] 039:  enqueue_entity+0x42/0x370
> > [  137.146080] 039:  enqueue_task_fair+0xa9/0x490
> > [  137.150528] 039:  activate_task+0x5a/0xf0
> > [  137.154539] 039:  ttwu_do_activate+0x4e/0x90
> > [  137.158813] 039:  try_to_wake_up+0x277/0x920
> > [  137.163086] 039:  irq_exit+0xb6/0xf0
[snip]
> > Signed-off-by: Scott Wood <swood@...hat.com>
> > ---
> > The prohibition on use_softirq should be able to be dropped once RT gets
> > the latest RCU code, but the question of what use_softirq should default
> > to on PREEMPT_RT remains.
> > 
> > v3: Use IS_ENABLED
> 
> Out of curiosity, does PREEMPT_RT use the NOCB callback offloading? If no,
> should it use it? IIUC, that does make the work the softirq have to do less
> work since the callbacks are executed in threaded context.
> 
> If yes, can RT tolerate use_softirq=false and what could a realistic softirq

s/use_softirq=false/use_softirq=true/

thanks,

 - Joel