Skip to contentSkip to conversation

bitbake-worker: Guard against multiprocessing corruption of event data

PATCH New Richard Purdie 1 participant 1 message Sep 29, 2015

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 bitbake-worker: Guard against multiprocessing corruption of event data

In the forked child, we may use multiprocessing. There is only one event pipe to the worker controlling process and if we're unlucky, multiple processes can write to it at once corrupting the data by intermixing it.

We don't see this often but when we do, its quite puzzling. I suspect it only happens in tasks which use multiprocessng (do_rootfs, do_package) and is much more likely to happen when we have long messages, usually many times PAGE_SIZE since PAGE_SIZE writes are atomic. This makes it much more likely within do_roofs, when for example a subprocess lists the contents of a rootfs.

To fix this, we give each child a Lock() object and use this to serialise writes to the controlling worker.

bitbake/bin/bitbake-worker +9 -0
index 45a78ec..1dcd590 100755
@@ -10,6 +10,7 @@ import bb
 import select
 import errno
 import signal
+from multiprocessing import Lock
 
 # Users shouldn't be running this code directly
 if len(sys.argv) != 2 or not sys.argv[1].startswith("decafbad"):
@@ -44,6 +45,9 @@ except ImportError:
 
 worker_pipe = sys.stdout.fileno()
 bb.utils.nonblockingfd(worker_pipe)
+# Need to guard against multiprocessing being used in child processes
+# and multiple processes trying to write to the parent at the same time
+worker_pipe_lock = None
 
 handler = bb.event.LogHandler()
 logger.addHandler(handler)
@@ -85,10 +89,13 @@ def worker_flush():
 
 def worker_child_fire(event, d):
     global worker_pipe
+    global worker_pipe_lock
 
     data = "<event>" + pickle.dumps(event) + "</event>"
     try:
+        worker_pipe_lock.acquire()
         worker_pipe.write(data)
+        worker_pipe_lock.release()
     except IOError:
         sigterm_handler(None, None)
         raise
@@ -157,6 +164,7 @@ def fork_off_task(cfg, data, workerdata, fn, task, taskname, appends, taskdepdat
     if pid == 0:
         def child():
             global worker_pipe
+            global worker_pipe_lock
             pipein.close()
 
             signal.signal(signal.SIGTERM, sigterm_handler)
@@ -169,6 +177,7 @@ def fork_off_task(cfg, data, workerdata, fn, task, taskname, appends, taskdepdat
             bb.event.worker_pid = os.getpid()
             bb.event.worker_fire = worker_child_fire
             worker_pipe = pipeout
+            worker_pipe_lock = Lock()
 
             # Make the child the process group leader and ensure no
             # child process will be controlled by the current terminal