Delete stale non-e2e devices for users, take 2 (#14595)

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.

1 files changed, 30 insertions, 1 deletions

diff --git a/synapse/handlers/device.py b/synapse/handlers/device.py
index d4750a32e6..7674c187ef 100644
--- a/synapse/handlers/device.py
+++ b/synapse/handlers/device.py
@@ -52,6 +52,7 @@ from synapse.util import stringutils
 from synapse.util.async_helpers import Linearizer
 from synapse.util.caches.expiringcache import ExpiringCache
 from synapse.util.cancellation import cancellable
+from synapse.util.iterutils import batch_iter
 from synapse.util.metrics import measure_func
 from synapse.util.retryutils import NotRetryingDestination
 
@@ -421,6 +422,9 @@ class DeviceHandler(DeviceWorkerHandler):
 
         self._check_device_name_length(initial_device_display_name)
 
+        # Prune the user's device list if they already have a lot of devices.
+        await self._prune_too_many_devices(user_id)
+
         if device_id is not None:
             new_device = await self.store.store_device(
                 user_id=user_id,
@@ -452,6 +456,31 @@ class DeviceHandler(DeviceWorkerHandler):
 
         raise errors.StoreError(500, "Couldn't generate a device ID.")
 
+    async def _prune_too_many_devices(self, user_id: str) -> None:
+        """Delete any excess old devices this user may have."""
+        device_ids = await self.store.check_too_many_devices_for_user(user_id)
+        if not device_ids:
+            return
+
+        # We don't want to block and try and delete tonnes of devices at once,
+        # so we cap the number of devices we delete synchronously.
+        first_batch, remaining_device_ids = device_ids[:10], device_ids[10:]
+        await self.delete_devices(user_id, first_batch)
+
+        if not remaining_device_ids:
+            return
+
+        # Now spawn a background loop that deletes the rest.
+        async def _prune_too_many_devices_loop() -> None:
+            for batch in batch_iter(remaining_device_ids, 10):
+                await self.delete_devices(user_id, batch)
+
+                await self.clock.sleep(1)
+
+        run_as_background_process(
+            "_prune_too_many_devices_loop", _prune_too_many_devices_loop
+        )
+
     async def _delete_stale_devices(self) -> None:
         """Background task that deletes devices which haven't been accessed for more than
         a configured time period.
@@ -481,7 +510,7 @@ class DeviceHandler(DeviceWorkerHandler):
             device_ids = [d for d in device_ids if d != except_device_id]
         await self.delete_devices(user_id, device_ids)
 
-    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:
         """Delete several devices
 
         Args: