| 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
| ||
|
Date: Thu, 26 Sep 2019 12:24:36 -0700
From: "Paul E. McKenney" <paulmck@...nel.org>
To: Joel Fernandes <joel@...lfernandes.org>
Cc: Dan Carpenter <dan.carpenter@...cle.com>, kbuild@...org,
kbuild-all@...org, LKML <linux-kernel@...r.kernel.org>
Subject: Re: [rcu:dev.2019.09.23a 62/77] kernel/rcu/tree.c:2882
kfree_call_rcu() warn: inconsistent returns 'irqsave:flags'.
On Thu, Sep 26, 2019 at 02:19:54PM -0400, Joel Fernandes wrote:
> On Tue, Sep 24, 2019 at 09:03:48AM -0700, Paul E. McKenney wrote:
> > On Tue, Sep 24, 2019 at 08:15:12AM -0400, Joel Fernandes wrote:
> > > On Tue, Sep 24, 2019 at 4:05 AM Dan Carpenter <dan.carpenter@...cle.com> wrote:
> > > >
> > > > tree: https://kernel.googlesource.com/pub/scm/linux/kernel/git/paulmck/linux-rcu.git dev.2019.09.23a
> > > > head: 97de53b94582c208ee239178b208b8e8b9472585
> > > > commit: 39676bb72323718a9e7bab1ba9c4a68319a96f62 [62/77] rcu: Add support for debug_objects debugging for kfree_rcu()
> > > >
> > > > If you fix the issue, kindly add following tag
> > > > Reported-by: kbuild test robot <lkp@...el.com>
> > > > Reported-by: Dan Carpenter <dan.carpenter@...cle.com>
> > > >
> > > > smatch warnings:
> > > > kernel/rcu/tree.c:2882 kfree_call_rcu() warn: inconsistent returns 'irqsave:flags'.
> > > > Locked on: line 2867
> > > > Unlocked on: line 2882
> > > >
> > > > git remote add rcu https://kernel.googlesource.com/pub/scm/linux/kernel/git/paulmck/linux-rcu.git
> > > > git remote update rcu
> > > > git checkout 39676bb72323718a9e7bab1ba9c4a68319a96f62
> > > > vim +2882 kernel/rcu/tree.c
> > > >
> > > > 5f5046cc15d514 Joel Fernandes (Google 2019-08-05 2850) void kfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
> > > > 5f5046cc15d514 Joel Fernandes (Google 2019-08-05 2851) {
> > > > 5f5046cc15d514 Joel Fernandes (Google 2019-08-05 2852) unsigned long flags;
> > > > 5f5046cc15d514 Joel Fernandes (Google 2019-08-05 2853) struct kfree_rcu_cpu *krcp;
> > > > 5f5046cc15d514 Joel Fernandes (Google 2019-08-05 2854)
> > > > 5f5046cc15d514 Joel Fernandes (Google 2019-08-05 2855) head->func = func;
> > > > 5f5046cc15d514 Joel Fernandes (Google 2019-08-05 2856)
> > > > 87970ccf5a9125 Paul E. McKenney 2019-09-19 2857 local_irq_save(flags); // For safely calling this_cpu_ptr().
> > > > ^^^^^^^^^^^^^^^^^^^^^
> > > > IRQs disabled.
> > > >
> > > > 5f5046cc15d514 Joel Fernandes (Google 2019-08-05 2858) krcp = this_cpu_ptr(&krc);
> > > > ff8db005e371cb Paul E. McKenney 2019-09-19 2859 if (krcp->initialized)
> > > > 5f5046cc15d514 Joel Fernandes (Google 2019-08-05 2860) spin_lock(&krcp->lock);
> > > > 5f5046cc15d514 Joel Fernandes (Google 2019-08-05 2861)
> > > > 87970ccf5a9125 Paul E. McKenney 2019-09-19 2862 // Queue the object but don't yet schedule the batch.
> > > > 39676bb7232371 Joel Fernandes (Google 2019-09-22 2863) if (debug_rcu_head_queue(head)) {
> > > > 39676bb7232371 Joel Fernandes (Google 2019-09-22 2864) // Probable double kfree_rcu(), just leak.
> > > > 39676bb7232371 Joel Fernandes (Google 2019-09-22 2865) WARN_ONCE(1, "%s(): Double-freed call. rcu_head %p\n",
> > > > 39676bb7232371 Joel Fernandes (Google 2019-09-22 2866) __func__, head);
> > > > 39676bb7232371 Joel Fernandes (Google 2019-09-22 2867) return;
> > > >
> > > > Need to re-enable before returning.
> > >
> > > Right. At this point the system has a bug anyway since it is a
> > > double-free. But yes, no reason to introduce more bugs. Will fix.
> > > Thank you, Dan!
> >
> > What Joel said!
> >
> > As long as I am doing several other reports anyway...
> >
> > How does the to-be-squashed patch below look?
>
> Hi Paul,
>
> Sorry for the late reply due to the conference..
>
> This patch looks Ok, the code in your tree though is slightly different mine
> I think because you reworked the kfree_rcu() code for working during early
> boot.
Yes, I did pull some of the changes earlier.
> After my talk tomorrow, I will go through your rcu/dev tree in detail to
> review the reworks you did in detail.
Sounds good, and best of everything with the talk!
> Let me know if you have the kfree_rcu() changes in a separate topic branch,
> as well. So it will make reviewing easier.
They will be, shortly after -rc1 comes out.
Thanx, Paul
> thanks,
>
> - Joel
>
>
> >
> > Thanx, Paul
> >
> > ------------------------------------------------------------------------
> >
> > commit 39af4c08038a8a6a7b1c9804fdd2c1921d583222
> > Author: Paul E. McKenney <paulmck@...nel.org>
> > Date: Tue Sep 24 09:02:00 2019 -0700
> >
> > squash! rcu: Add support for debug_objects debugging for kfree_rcu()
> >
> > [ paulmck: Fix IRQ per kbuild test robot <lkp@...el.com> feedback. ]
> > Signed-off-by: Paul E. McKenney <paulmck@...nel.org>
> >
> > diff --git a/kernel/rcu/tree.c b/kernel/rcu/tree.c
> > index 921daa7..edb7539 100644
> > --- a/kernel/rcu/tree.c
> > +++ b/kernel/rcu/tree.c
> > @@ -2870,7 +2870,7 @@ void kfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
> > // Probable double kfree_rcu(), just leak.
> > WARN_ONCE(1, "%s(): Double-freed call. rcu_head %p\n",
> > __func__, head);
> > - return;
> > + goto unlock_return;
> > }
> > head->func = func;
> > head->next = krcp->head;
> > @@ -2883,6 +2883,7 @@ void kfree_call_rcu(struct rcu_head *head, rcu_callback_t func)
> > schedule_delayed_work(&krcp->monitor_work, KFREE_DRAIN_JIFFIES);
> > }
> >
> > +unlock_return:
> > if (krcp->initialized)
> > spin_unlock(&krcp->lock);
> > local_irq_restore(flags);
Powered by blists - more mailing lists