POC delete stale non-e2e devices for users (#14038)

This should help reduce the number of devices e.g. simple bots the repeatedly login rack up.

We only delete non-e2e devices as they should be safe to delete, whereas if we delete e2e devices for a user we may accidentally break their ability to receive e2e keys for a message.

Co-authored-by: Patrick Cloke <clokep@users.noreply.github.com>
Co-authored-by: Sean Quah <8349537+squahtx@users.noreply.github.com>

synapse/storage/databases/main/devices.py

@@ -1533,6 +1533,70 @@ class DeviceBackgroundUpdateStore(SQLBaseStore):
 
         return rows
 
+    async def check_too_many_devices_for_user(self, user_id: str) -> Collection[str]:
+        """Check if the user has a lot of devices, and if so return the set of
+        devices we can prune.
+
+        This does *not* return hidden devices or devices with E2E keys.
+        """
+
+        num_devices = await self.db_pool.simple_select_one_onecol(
+            table="devices",
+            keyvalues={"user_id": user_id, "hidden": False},
+            retcol="COALESCE(COUNT(*), 0)",
+            desc="count_devices",
+        )
+
+        # We let users have up to ten devices without pruning.
+        if num_devices <= 10:
+            return ()
+
+        # We prune everything older than N days.
+        max_last_seen = self._clock.time_msec() - 14 * 24 * 60 * 60 * 1000
+
+        if num_devices > 50:
+            # If the user has more than 50 devices, then we chose a last seen
+            # that ensures we keep at most 50 devices.
+            sql = """
+                SELECT last_seen FROM devices
+                WHERE
+                    user_id = ?
+                    AND NOT hidden
+                    AND last_seen IS NOT NULL
+                    AND key_json IS NULL
+                ORDER BY last_seen DESC
+                LIMIT 1
+                OFFSET 50
+            """
+
+            rows = await self.db_pool.execute(
+                "check_too_many_devices_for_user_last_seen", None, sql, (user_id,)
+            )
+            if rows:
+                max_last_seen = max(rows[0][0], max_last_seen)
+
+        # Now fetch the devices to delete.
+        sql = """
+            SELECT DISTINCT device_id FROM devices
+            LEFT JOIN e2e_device_keys_json USING (user_id, device_id)
+            WHERE
+                user_id = ?
+                AND NOT hidden
+                AND last_seen < ?
+                AND key_json IS NULL
+        """
+
+        def check_too_many_devices_for_user_txn(
+            txn: LoggingTransaction,
+        ) -> Collection[str]:
+            txn.execute(sql, (user_id, max_last_seen))
+            return {device_id for device_id, in txn}
+
+        return await self.db_pool.runInteraction(
+            "check_too_many_devices_for_user",
+            check_too_many_devices_for_user_txn,
+        )
+
 
 class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore):
     # Because we have write access, this will be a StreamIdGenerator
@@ -1591,6 +1655,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore):
                 values={},
                 insertion_values={
                     "display_name": initial_device_display_name,
+                    "last_seen": self._clock.time_msec(),
                     "hidden": False,
                 },
                 desc="store_device",
@@ -1636,7 +1701,7 @@ class DeviceStore(DeviceWorkerStore, DeviceBackgroundUpdateStore):
             )
             raise StoreError(500, "Problem storing device.")
 
-    async def delete_devices(self, user_id: str, device_ids: List[str]) -> None:
+    async def delete_devices(self, user_id: str, device_ids: Collection[str]) -> None:
         """Deletes several devices.
 
         Args: