Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
From:   Duoming Zhou <duoming@zju.edu.cn>
To:     krzk@kernel.org, linux-kernel@vger.kernel.org
Cc:     davem@davemloft.net, gregkh@linuxfoundation.org,
        alexander.deucher@amd.com, broonie@kernel.org,
        akpm@linux-foundation.org, netdev@vger.kernel.org,
        pabeni@redhat.com, Duoming Zhou <duoming@zju.edu.cn>
Subject: [PATCH V3 1/2] drivers: nfc: nfcmrvl: fix double free bugs caused by fw_dnld_over()
Date:   Tue, 12 Apr 2022 18:56:52 +0800
Message-Id: <a88510674c00ad2065698340e9fee89c93ad5d78.1649759498.git.duoming@zju.edu.cn>
In-Reply-To: <cover.1649759498.git.duoming@zju.edu.cn>
References: <cover.1649759498.git.duoming@zju.edu.cn>
In-Reply-To: <cover.1649759498.git.duoming@zju.edu.cn>
References: <cover.1649759498.git.duoming@zju.edu.cn>
Precedence: bulk

There are potential double free bugs in nfcmrvl usb driver among
fw_dnld_rx_work(), fw_dnld_timeout() and nfcmrvl_nci_unregister_dev().
All these three functions will call fw_dnld_over(). The fw_dnld_rx_work()
is a work item, the fw_dnld_timeout() is a timer handler and the
nfcmrvl_nci_unregister_dev() is called when nfcmrvl_nci device is
detaching. So these three functions could execute concurrently.

The race between fw_dnld_rx_work() and nfcmrvl_nci_unregister_dev()
can be shown as below:

   (Thread 1)               |      (Thread 2)
                            |       fw_dnld_rx_work
                            |        fw_dnld_over
                            |         release_firmware
                            |          kfree(fw); //(1)
nfcmrvl_disconnect          |
 nfcmrvl_nci_unregister_dev |
  nfcmrvl_fw_dnld_abort     |
   fw_dnld_over             |         ...
    if (priv->fw_dnld.fw)   |
    release_firmware        |
     kfree(fw); //(2)       |
     ...                    |         priv->fw_dnld.fw = NULL;

When the fw_dnld_rx_work work item is executing, we detach the device.
The release_firmware() will deallocate firmware in position (1),
but firmware will be deallocated again in position (2), which
leads to double free.

The race between fw_dnld_rx_work() and fw_dnld_timeout()
can be shown as below:

   (Thread 1)               |      (Thread 2)
                            |       fw_dnld_rx_work
                            |        fw_dnld_over
                            |         release_firmware
                            |          kfree(fw); //(1)
fw_dnld_timeout             |
 fw_dnld_over               |         ...
  if (priv->fw_dnld.fw)     |
   release_firmware         |
    kfree(fw); //(2)        |
     ...                    |         priv->fw_dnld.fw = NULL;

The release_firmware() will deallocate firmware in position (1),
but firmware will be deallocated again in position (2), which
leads to double free.

The race between fw_dnld_timeout() and nfcmrvl_nci_unregister_dev()
can be shown as below.

   (Thread 1)               |      (Thread 2)
                            |     nfcmrvl_disconnect
                            |      nfcmrvl_nci_unregister_dev
                            |       nfcmrvl_fw_dnld_abort
                            |        fw_dnld_over
                            |         release_firmware
                            |          kfree(fw); //(1)
fw_dnld_timeout             |
 fw_dnld_over               |         ...
  if (priv->fw_dnld.fw)     |
   release_firmware         |
    kfree(fw); //(2)        |
     ...                    |         priv->fw_dnld.fw = NULL;

The release_firmware() will deallocate firmware in position (1),
but firmware will be deallocated again in position (2), which
leads to double free.

This patch adds spin_lock_irq in fw_dnld_over() in order to synchronize
among different threads that call fw_dnld_over(). The priv->fw_dnld.fw will
be set to NULL after work item is finished and fw_dnld_over() called by
other threads will check whether priv->fw_dnld.fw is NULL. So the double
free bug could be prevented.

Fixes: 3194c6870158e3 ("NFC: nfcmrvl: add firmware download support")
Signed-off-by: Duoming Zhou <duoming@zju.edu.cn>
Reviewed-by: Lin Ma <linma@zju.edu.cn>
---
Changes in V3:
  - Make commit message more clearer.
  - Use spin_lock_irq to synchronize.

 drivers/nfc/nfcmrvl/fw_dnld.c | 3 +++
 drivers/nfc/nfcmrvl/fw_dnld.h | 2 ++
 2 files changed, 5 insertions(+)

diff --git a/drivers/nfc/nfcmrvl/fw_dnld.c b/drivers/nfc/nfcmrvl/fw_dnld.c
index e83f65596a8..c22a4556db5 100644
--- a/drivers/nfc/nfcmrvl/fw_dnld.c
+++ b/drivers/nfc/nfcmrvl/fw_dnld.c
@@ -92,12 +92,14 @@ static struct sk_buff *alloc_lc_skb(struct nfcmrvl_private *priv, uint8_t plen)
 
 static void fw_dnld_over(struct nfcmrvl_private *priv, u32 error)
 {
+	spin_lock_irq(&priv->fw_dnld.lock);
 	if (priv->fw_dnld.fw) {
 		release_firmware(priv->fw_dnld.fw);
 		priv->fw_dnld.fw = NULL;
 		priv->fw_dnld.header = NULL;
 		priv->fw_dnld.binary_config = NULL;
 	}
+	spin_unlock_irq(&priv->fw_dnld.lock);
 
 	atomic_set(&priv->ndev->cmd_cnt, 0);
 
@@ -451,6 +453,7 @@ int nfcmrvl_fw_dnld_init(struct nfcmrvl_private *priv)
 	if (!priv->fw_dnld.rx_wq)
 		return -ENOMEM;
 	skb_queue_head_init(&priv->fw_dnld.rx_q);
+	spin_lock_init(&priv->fw_dnld.lock);
 	return 0;
 }
 
diff --git a/drivers/nfc/nfcmrvl/fw_dnld.h b/drivers/nfc/nfcmrvl/fw_dnld.h
index 7c4d91b0191..e4fe0fb8aff 100644
--- a/drivers/nfc/nfcmrvl/fw_dnld.h
+++ b/drivers/nfc/nfcmrvl/fw_dnld.h
@@ -75,6 +75,8 @@ struct nfcmrvl_fw_dnld {
 	struct sk_buff_head rx_q;
 
 	struct timer_list timer;
+	/* To synchronize among different threads that call fw_dnld_over.*/
+	spinlock_t lock;
 };
 
 int nfcmrvl_fw_dnld_init(struct nfcmrvl_private *priv);
-- 
2.17.1