Received-SPF: pass (google.com: domain of linux-kernel+bounces-200476-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
From: Hillf Danton <hdanton@sina.com>
To: Leon Romanovsky <leon@kernel.org>
Cc: Tejun Heo <tj@kernel.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Lai Jiangshan <jiangshanlai@gmail.com>,
	Zqiang <qiang.zhang1211@gmail.com>,
	linux-kernel@vger.kernel.org,
	Gal Pressman <gal@nvidia.com>,
	Tariq Toukan <tariqt@nvidia.com>,
	RDMA mailing list <linux-rdma@vger.kernel.org>
Subject: Re: [PATCH -rc] workqueue: Reimplement UAF fix to avoid lockdep worning
Date: Tue,  4 Jun 2024 18:54:56 +0800
Message-Id: <20240604105456.1668-1-hdanton@sina.com>
In-Reply-To: <20240604080958.GL3884@unreal>
References: <4c4f1fb769a609a61010cb6d884ab2841ef716d3.1716885172.git.leon@kernel.org> <ZljyqODpCD0_5-YD@slm.duckdns.org> <20240531034851.GF3884@unreal> <Zl4jPImmEeRuYQjz@slm.duckdns.org>
Precedence: bulk
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit

On Tue, 4 Jun 2024 11:09:58 +0300 Leon Romanovsky <leon@kernel.org>
> On Mon, Jun 03, 2024 at 10:10:36AM -1000, Tejun Heo wrote:
> > 
> > And KASAN is reporting use-after-free on a completely unrelated VFS object.
> > I can't tell for sure from the logs alone but lockdep_register_key()
> > iterates entries in the hashtable trying to find whether the key is a
> > duplicate and it could be that that walk is triggering the use-after-free
> > warning. If so, it doesn't really have much to do with workqueue. The
> > corruption happened elsewhere and workqueue just happens to traverse the
> > hashtable afterwards.
> 
> The problem is that revert of commit 643445531829
> ("workqueue: Fix UAF report by KASAN in pwq_release_workfn()")
> fixed these use-after-free reports.
> 
Given revert makes sense,

	if (alloc_and_link_pwqs(wq) < 0)
		goto err_unreg_lockdep;

err_unreg_lockdep:
	wq_unregister_lockdep(wq);
	wq_free_lockdep(wq);
err_free_wq:
	free_workqueue_attrs(wq->unbound_attrs);
	kfree(wq);	<-- freed
	return NULL;

the difference 643445531829 makes is double free.

	alloc_and_link_pwqs(struct workqueue_struct *wq)
	if (ret)
		kthread_flush_worker(pwq_release_worker);
		  pwq_release_workfn()
		  if (is_last) {
			wq_unregister_lockdep(wq);
			call_rcu(&wq->rcu, rcu_free_wq); <-- freed again
		  }