[PATCH] fix a regression involving encrypted PVs

[David Lehman <dlehman redhat com>]

Hi,

I recently put in a patch that prevented us from marking as available
any encrypted PVs the user chose not to provide a passphrase for. The
unforeseen consequence, a result of the way we track information about
encrypted devices, was that if you give a passphrase for a preexisting
encrypted PV, delete it, then create a non-encrypted PV with the same
name (eg: sda2) the LVM button will reject you, saying you don't have
any available PVs. This is because the filtering I added for the first
case described is also erroneously catching this one. Following is a
patch that I have tested fairly thoroughly for the aforementioned
scenarios.

diff --git a/partitions.py b/partitions.py
index cfe770d..25af908 100644
--- a/partitions.py
+++ b/partitions.py
@@ -904,9 +904,14 @@ class Partitions:
             for part in partedUtils.get_lvm_partitions(disk):
                 partname = partedUtils.get_partition_name(part)
                 partrequest = self.getRequestByDeviceName(partname)
-                if partrequest.encryption is None and
cryptodev.isLuks("/dev/%s" % partname):
-                    # we don't want to treat encrypted an PV like a PV
if the
-                    # user chose not to provide a passphrase for this
device
+                if partrequest.encryption is None and \
+                   cryptodev.isLuks("/dev/%s" % partname) and \
+                   not self.encryptedDevices.get(partname):
+                    log.debug("ignoring PV %s since we cannot access
it's contents" % partname)
+                    # We don't want to treat encrypted an PV like a PV
if the
+                    # user chose not to provide a passphrase for this
device.
+                    # However, if the LUKS device belongs to a
just-deleted
+                    # request then we know it is available.
                     continue
                 used = 0
                 for volgroup in volgroups:


Dave