Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   Khazhismel Kumykov <khazhy@chromium.org>
To:     Alexander Viro <viro@zeniv.linux.org.uk>,
        Andrew Morton <akpm@linux-foundation.org>,
        "Matthew Wilcox (Oracle)" <willy@infradead.org>,
        Jan Kara <jack@suse.cz>
Cc:     linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
        linux-mm@kvack.org, Khazhismel Kumykov <khazhy@google.com>
Subject: [PATCH v2] writeback: avoid use-after-free after removing device
Date:   Mon,  1 Aug 2022 08:50:34 -0700
Message-Id: <20220801155034.3772543-1-khazhy@google.com>
In-Reply-To: <20220729215123.1998585-1-khazhy@google.com>
References: <20220729215123.1998585-1-khazhy@google.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

When a disk is removed, bdi_unregister gets called to stop further
writeback and wait for associated delayed work to complete. However,
wb_inode_writeback_end() may schedule bandwidth estimation dwork after
this has completed, which can result in the timer attempting to access
the just freed bdi_writeback.

Fix this by checking if the bdi_writeback is alive, similar to when
scheduling writeback work.

Since this requires wb->work_lock, and wb_inode_writeback_end() may get
called from interrupt, switch wb->work_lock to an irqsafe lock.

Fixes: 45a2966fd641 ("writeback: fix bandwidth estimate for spiky workload")
Signed-off-by: Khazhismel Kumykov <khazhy@google.com>
---
 fs/fs-writeback.c   | 12 ++++++------
 mm/backing-dev.c    | 10 +++++-----
 mm/page-writeback.c |  6 +++++-
 3 files changed, 16 insertions(+), 12 deletions(-)

v2: made changelog a bit more verbose

diff --git a/fs/fs-writeback.c b/fs/fs-writeback.c
index 05221366a16d..08a1993ab7fd 100644
--- a/fs/fs-writeback.c
+++ b/fs/fs-writeback.c
@@ -134,10 +134,10 @@ static bool inode_io_list_move_locked(struct inode *inode,
 
 static void wb_wakeup(struct bdi_writeback *wb)
 {
-	spin_lock_bh(&wb->work_lock);
+	spin_lock_irq(&wb->work_lock);
 	if (test_bit(WB_registered, &wb->state))
 		mod_delayed_work(bdi_wq, &wb->dwork, 0);
-	spin_unlock_bh(&wb->work_lock);
+	spin_unlock_irq(&wb->work_lock);
 }
 
 static void finish_writeback_work(struct bdi_writeback *wb,
@@ -164,7 +164,7 @@ static void wb_queue_work(struct bdi_writeback *wb,
 	if (work->done)
 		atomic_inc(&work->done->cnt);
 
-	spin_lock_bh(&wb->work_lock);
+	spin_lock_irq(&wb->work_lock);
 
 	if (test_bit(WB_registered, &wb->state)) {
 		list_add_tail(&work->list, &wb->work_list);
@@ -172,7 +172,7 @@ static void wb_queue_work(struct bdi_writeback *wb,
 	} else
 		finish_writeback_work(wb, work);
 
-	spin_unlock_bh(&wb->work_lock);
+	spin_unlock_irq(&wb->work_lock);
 }
 
 /**
@@ -2082,13 +2082,13 @@ static struct wb_writeback_work *get_next_work_item(struct bdi_writeback *wb)
 {
 	struct wb_writeback_work *work = NULL;
 
-	spin_lock_bh(&wb->work_lock);
+	spin_lock_irq(&wb->work_lock);
 	if (!list_empty(&wb->work_list)) {
 		work = list_entry(wb->work_list.next,
 				  struct wb_writeback_work, list);
 		list_del_init(&work->list);
 	}
-	spin_unlock_bh(&wb->work_lock);
+	spin_unlock_irq(&wb->work_lock);
 	return work;
 }
 
diff --git a/mm/backing-dev.c b/mm/backing-dev.c
index 95550b8fa7fe..de65cb1e5f76 100644
--- a/mm/backing-dev.c
+++ b/mm/backing-dev.c
@@ -260,10 +260,10 @@ void wb_wakeup_delayed(struct bdi_writeback *wb)
 	unsigned long timeout;
 
 	timeout = msecs_to_jiffies(dirty_writeback_interval * 10);
-	spin_lock_bh(&wb->work_lock);
+	spin_lock_irq(&wb->work_lock);
 	if (test_bit(WB_registered, &wb->state))
 		queue_delayed_work(bdi_wq, &wb->dwork, timeout);
-	spin_unlock_bh(&wb->work_lock);
+	spin_unlock_irq(&wb->work_lock);
 }
 
 static void wb_update_bandwidth_workfn(struct work_struct *work)
@@ -334,12 +334,12 @@ static void cgwb_remove_from_bdi_list(struct bdi_writeback *wb);
 static void wb_shutdown(struct bdi_writeback *wb)
 {
 	/* Make sure nobody queues further work */
-	spin_lock_bh(&wb->work_lock);
+	spin_lock_irq(&wb->work_lock);
 	if (!test_and_clear_bit(WB_registered, &wb->state)) {
-		spin_unlock_bh(&wb->work_lock);
+		spin_unlock_irq(&wb->work_lock);
 		return;
 	}
-	spin_unlock_bh(&wb->work_lock);
+	spin_unlock_irq(&wb->work_lock);
 
 	cgwb_remove_from_bdi_list(wb);
 	/*
diff --git a/mm/page-writeback.c b/mm/page-writeback.c
index 55c2776ae699..3c34db15cf70 100644
--- a/mm/page-writeback.c
+++ b/mm/page-writeback.c
@@ -2867,6 +2867,7 @@ static void wb_inode_writeback_start(struct bdi_writeback *wb)
 
 static void wb_inode_writeback_end(struct bdi_writeback *wb)
 {
+	unsigned long flags;
 	atomic_dec(&wb->writeback_inodes);
 	/*
 	 * Make sure estimate of writeback throughput gets updated after
@@ -2875,7 +2876,10 @@ static void wb_inode_writeback_end(struct bdi_writeback *wb)
 	 * that if multiple inodes end writeback at a similar time, they get
 	 * batched into one bandwidth update.
 	 */
-	queue_delayed_work(bdi_wq, &wb->bw_dwork, BANDWIDTH_INTERVAL);
+	spin_lock_irqsave(&wb->work_lock, flags);
+	if (test_bit(WB_registered, &wb->state))
+		queue_delayed_work(bdi_wq, &wb->bw_dwork, BANDWIDTH_INTERVAL);
+	spin_unlock_irqrestore(&wb->work_lock, flags);
 }
 
 bool __folio_end_writeback(struct folio *folio)
-- 
2.37.1.455.g008518b4e5-goog