Overwrite with zero instead of random when shredding

Writing zeroes is considerably faster than obtaining random
data for writing. On memory case, we should not worry so much
about possibility someone restoring old state after overwrites.

docs/install.md

@@ -629,7 +629,7 @@ the following custom parameters in a storage class:
 
 |key|meaning|optional|values|
 |---|-------|--------|-------------|
-|`eraseAfter`|Clear all data after use and before<br> deleting the volume|Yes|`true` (default),<br> `false`|
+|`eraseAfter`|Clear all data by overwriting with zeroes after use and before<br> deleting the volume|Yes|`true` (default),<br> `false`|
 |`kataContainers`|Prepare volume for use with DAX in Kata Containers.|Yes|`false/0/f/FALSE` (default),<br> `true/1/t/TRUE`|
 
 
@@ -705,7 +705,7 @@ ephemeral volumes. The volume request could use below fields as
 |key|meaning|optional|values|
 |---|-------|--------|-------------|
 |`size`|Size of the requested ephemeral volume as [Kubernetes memory string](https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/#meaning-of-memory) ("1Mi" = 1024*1024 bytes, "1e3K = 1000000 bytes)|No||
-|`eraseAfter`|Clear all data after use and before<br> deleting the volume|Yes|`true` (default),<br> `false`|
+|`eraseAfter`|Clear all data by overwriting with zeroes after use and before<br> deleting the volume|Yes|`true` (default),<br> `false`|
 |`kataContainers`|Prepare volume for use in Kata Containers.|Yes|`false/0/f/FALSE` (default),<br> `true/1/t/TRUE`|
 
 Try out ephemeral volume usage with the provided [example

pkg/pmem-device-manager/pmd-util.go

@@ -50,8 +50,11 @@ func clearDevice(dev *PmemDeviceInfo, flush bool) error {
 
 	if blocks == 0 {
 		klog.V(5).Infof("Wiping entire device: %s", dev.Path)
-		// use one iteration instead of shred's default=3 for speed
-		if _, err := pmemexec.RunCommand("shred", "-n", "1", dev.Path); err != nil {
+		// shred would write n times using random data, followed by optional write of zeroes.
+		// For faster operation, and because we consider zeroing enough for
+		// reasonable clearing in case of a memory device, we force zero iterations
+		// with random data, followed by one pass writing zeroes.
+		if _, err := pmemexec.RunCommand("shred", "-n", "0", "-z", dev.Path); err != nil {
 			return fmt.Errorf("device shred failure: %v", err.Error())
 		}
 	} else {