Name only physical volumes in a restricted LVM scan and keep its own reads from counting as activity

- An explicit --devices list makes LVM open every device in it, while a plain scan opens only known volumes; the list now names only LVM physical volumes, read from udev by lsblk, so no other disk is touched
- RAID, multipath and encrypted volumes follow the disks they are built on, instead of turning the whole check off
- After each pvs, lvs or vgs run, the disks it read take their counters as the new baseline, so an unused disk can become idle

AppImage/scripts/lvm_device_scope.py

@@ -11,11 +11,56 @@
 # Description:
 # Applies the automatic disk-read policy to LVM metadata scans in both
 # Monitor readers. Excluded and idle disks receive no power-mode probe;
-# their partitions are excluded too. No eligible devices means no scan.
+# every device built on them is excluded too, and the scan's own reads
+# are not taken as activity. No eligible devices means no scan.
 # ==========================================================
 
+import json
 import re
 import subprocess
+import threading
+
+_NAME = re.compile(r"[A-Za-z0-9._+-]+")
+_scan_lock = threading.Lock()
+_scanned_disks = set()
+
+
+def _topology():
+    """Every block device by kernel name: its type, its parents and whether
+    it is an LVM physical volume. lsblk reads this from udev, not the disks.
+    A RAID or multipath device appears once per member."""
+    result = subprocess.run(["lsblk", "-J", "-l", "-o", "KNAME,TYPE,PKNAME,FSTYPE"],
+                            capture_output=True, text=True, timeout=5)
+    if result.returncode:
+        return None
+    types, parents, volumes = {}, {}, set()
+    for device in json.loads(result.stdout).get("blockdevices") or []:
+        name, parent = device.get("kname"), device.get("pkname")
+        if not isinstance(name, str) or not _NAME.fullmatch(name):
+            return None
+        types[name] = device.get("type")
+        parents.setdefault(name, set())
+        if parent is not None:
+            if not isinstance(parent, str) or not _NAME.fullmatch(parent):
+                return None
+            parents[name].add(parent)
+        if device.get("fstype") == "LVM2_member" and device.get("type") != "lvm":
+            volumes.add(name)
+    return types, parents, volumes
+
+
+def _disks_under(name, parents, seen=None):
+    """The disks a device is built on: itself, or its parents' disks."""
+    seen = seen if seen is not None else set()
+    if name in seen:
+        return set()
+    seen.add(name)
+    if not parents.get(name):
+        return {name}
+    found = set()
+    for parent in parents[name]:
+        found |= _disks_under(parent, parents, seen)
+    return found
 
 
 def lvm_device_args(standby_check):
@@ -27,22 +72,12 @@ def lvm_device_args(standby_check):
     """
     try:
         from disk_temperature_history import disk_read_policy, excluded_disk_names
-        result = subprocess.run(["lsblk", "-rno", "NAME,TYPE,PKNAME"],
-                                capture_output=True, text=True, timeout=5)
-        if result.returncode:
+        topology = _topology()
+        if topology is None:
             return None
-        nodes = []
-        for line in result.stdout.splitlines():
-            parts = line.split()
-            if len(parts) < 2 or parts[1] not in ("disk", "part"):
-                continue
-            name = parts[0]
-            base = name if parts[1] == "disk" else parts[2] if len(parts) > 2 else ""
-            if not re.fullmatch(r"[A-Za-z0-9._+-]+", name) or not re.fullmatch(r"[A-Za-z0-9._+-]+", base):
-                return None
-            nodes.append((name, base))
-        disks = {name for name, base in nodes if name == base}
-        if not disks or any(base not in disks for _, base in nodes):
+        types, parents, volumes = topology
+        disks = {name for name, kind in types.items() if kind == "disk"}
+        if not disks:
             return None
         excluded = excluded_disk_names()
         blocked = set()
@@ -51,13 +86,31 @@ def lvm_device_args(standby_check):
             # excluded/idle disk. Some USB bridges do not honour -n standby.
             if disk_read_policy(disk, excluded) != "read" or standby_check(disk):
                 blocked.add(disk)
+        allowed = sorted(f"/dev/{name}" for name in volumes if not (_disks_under(name, parents) & blocked))
+        with _scan_lock:
+            _scanned_disks.clear()
+            _scanned_disks.update(disks - blocked)
         if not blocked:
             return []
-        devices = [f"/dev/{name}" for name, base in nodes if base not in blocked]
-        if not devices:
+        # An explicit list makes LVM open every device in it, so only the
+        # physical volumes are named: a RAID, multipath or encrypted volume
+        # built on a blocked disk would read that disk too and stays out.
+        if not allowed:
             return None
-        return ["--devices", ",".join(dict.fromkeys(devices))]
+        return ["--devices", ",".join(allowed)]
     except Exception:
         # Unavailable exclusions, topology or policy are not permission to
         # open devices that may have been deliberately excluded.
         return None
+
+
+def settle_scan_reads():
+    """The LVM scan reads the disks it is allowed to. Those reads are its
+    own, so they must not make an unused disk look busy at the next check."""
+    try:
+        from disk_temperature_history import settle_own_reads
+        with _scan_lock:
+            disks = set(_scanned_disks)
+        settle_own_reads(disks)
+    except Exception:
+        pass