As the potential failure of the kvmalloc_array(),
it should be better to check and restore the 'data'
if fails in order to avoid the dereference of the
NULL pointer.

Fixes: 6ae746711263 ("hv_netvsc: Add per-cpu ethtool stats for netvsc")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
---
 drivers/net/hyperv/netvsc_drv.c | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/drivers/net/hyperv/netvsc_drv.c b/drivers/net/hyperv/netvsc_drv.c
index 3646469433b1..018c4a5f6f44 100644
--- a/drivers/net/hyperv/netvsc_drv.c
+++ b/drivers/net/hyperv/netvsc_drv.c
@@ -1587,6 +1587,12 @@ static void netvsc_get_ethtool_stats(struct net_device *dev,
 	pcpu_sum = kvmalloc_array(num_possible_cpus(),
 				  sizeof(struct netvsc_ethtool_pcpu_stats),
 				  GFP_KERNEL);
+	if (!pcpu_sum) {
+		for (j = 0; j < i; j++)
+			data[j] = 0;
+		return;
+	}
+
 	netvsc_get_pcpu_stats(dev, pcpu_sum);
 	for_each_present_cpu(cpu) {
 		struct netvsc_ethtool_pcpu_stats *this_sum = &pcpu_sum[cpu];
-- 
2.25.1

On Fri, 11 Mar 2022 10:43:44 +0800
Jiasheng Jiang <jiasheng@iscas.ac.cn> wrote:

> +	if (!pcpu_sum) {
> +		for (j = 0; j < i; j++)
> +			data[j] = 0;
> +		return

Why is unrolled zero (memset) needed? The data area comes from
ethtool_get_stats and is already zeroed (vzalloc).


There does look like at TOCTOU error here with on the number of stats.
Code doesn't look hotplug safe.
Not sure, but that issue might have been raised during review.