diff mbox series

scsi: virtio_scsi: Fix a NULL pointer dereference in virtscsi_rescan_hotunplug()

Message ID 20211130171901.202229-1-zhou1615@umn.edu (mailing list archive)
State New, archived
Headers show
Series scsi: virtio_scsi: Fix a NULL pointer dereference in virtscsi_rescan_hotunplug() | expand

Commit Message

Zhou Qingyang Nov. 30, 2021, 5:19 p.m. UTC
In virtscsi_rescan_hotunplug(), kmalloc() is directly used in memset(),
which could lead to a NULL pointer dereference on failure of
kmalloc().

Fix this bug by adding a check of inq_result.

This bug was found by a static analyzer. The analysis employs
differential checking to identify inconsistent security operations
(e.g., checks or kfrees) between two code paths and confirms that the
inconsistent operations are not recovered in the current function or
the callers, so they constitute bugs.

Note that, as a bug found by static analysis, it can be a false
positive or hard to trigger. Multiple researchers have cross-reviewed
the bug.

Builds with CONFIG_SCSI_VIRTIO=m show no new warnings,
and our static analyzer no longer warns about this code.

Fixes: 5ff843721467 ("scsi: virtio_scsi: unplug LUNs when events missed")
Signed-off-by: Zhou Qingyang <zhou1615@umn.edu>
---
 drivers/scsi/virtio_scsi.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

Comments

Paolo Bonzini Nov. 30, 2021, 5:35 p.m. UTC | #1
On 11/30/21 18:19, Zhou Qingyang wrote:
> --- a/drivers/scsi/virtio_scsi.c
> +++ b/drivers/scsi/virtio_scsi.c
> @@ -337,7 +337,11 @@ static void virtscsi_rescan_hotunplug(struct virtio_scsi *vscsi)
>   	unsigned char scsi_cmd[MAX_COMMAND_SIZE];
>   	int result, inquiry_len, inq_result_len = 256;
>   	char *inq_result = kmalloc(inq_result_len, GFP_KERNEL);
> -
> +	if (!inq_result) {
> +		pr_err("%s:no enough memory for inq_result\n",
> +			__func__);
> +		return;
> +	}
>   	shost_for_each_device(sdev, shost) {
>   		inquiry_len = sdev->inquiry_len ? sdev->inquiry_len : 36;
>   

In practice this will never happen, since the kmalloc is very small, so 
I think it's easier to just return early without a printk.  On the other 
hand, if the out-of-memory really could happen, this should be a 
pr_err_ratelimited.

Paolo
diff mbox series

Patch

diff --git a/drivers/scsi/virtio_scsi.c b/drivers/scsi/virtio_scsi.c
index 28e1d98ae102..5309f2a3a4cb 100644
--- a/drivers/scsi/virtio_scsi.c
+++ b/drivers/scsi/virtio_scsi.c
@@ -337,7 +337,11 @@  static void virtscsi_rescan_hotunplug(struct virtio_scsi *vscsi)
 	unsigned char scsi_cmd[MAX_COMMAND_SIZE];
 	int result, inquiry_len, inq_result_len = 256;
 	char *inq_result = kmalloc(inq_result_len, GFP_KERNEL);
-
+	if (!inq_result) {
+		pr_err("%s:no enough memory for inq_result\n",
+			__func__);
+		return;
+	}
 	shost_for_each_device(sdev, shost) {
 		inquiry_len = sdev->inquiry_len ? sdev->inquiry_len : 36;