zookeeper 1.2.4-java → 1.2.5-java

data/CHANGELOG

@@ -1,3 +1,24 @@
+v1.2.5 cleanup locking in ZookeeperBase
+
+  * There were several situations where we would hold the lock before calling
+    a method on CZookeeper (inquisitors and #create in particular). This
+    exposed us to deadlocks in situations where an async event would be
+    delivered (probably a SESSION_EXPIRED event), but the callback block could
+    not be called because the dispatch thread would block on the mutex being
+    held by the caller of create. 
+
+    This version cleans up that usage, and ensures that the only time we hold
+    the mutex is during startup/shutdown (when the value of @czk may be changing),
+    and in all other cases we grab the mutex, dereference, and unlock then perform
+    whatever action on the reference. 
+
+  * Add a safety net to Continuation (which will soon be called 'Promise' or
+    'Future' at the request of @eric). If any operation takes more than 30s
+    an exception will be raised in the calling thread. The session timeout
+    setting makes it so that no operation should take more than 20s, so we know
+    if we haven't received a reply in *longer* than that, something has gone
+    awry. 
+
 v1.2.4 fix buffer overflow in CZookeeper client_id code
   
   * the 'passwd' part of the struct is a char[16], but isn't null terminated.

data/ext/c_zookeeper.rb

@@ -176,6 +176,7 @@ class CZookeeper
     return false unless wait_until_running(timeout)
 
     @mutex.synchronize do
+      # TODO: use deadline here
       @state_cond.wait(timeout) unless (@state == ZOO_CONNECTED_STATE)
     end
 
@@ -191,15 +192,12 @@ class CZookeeper
       end
 
       cnt = Continuation.new(meth, *args)
-      @reg.lock 
-      begin
+      @reg.synchronize do |r|
         if meth == :state
-          @reg.state_check << cnt
+          r.state_check << cnt
         else
-          @reg.pending << cnt
+          r.pending << cnt
         end
-      ensure
-        @reg.unlock rescue nil
       end
       wake_event_loop!
       cnt.value

data/ext/extconf.rb

@@ -42,6 +42,11 @@ $LDFLAGS = "#{$libraries} #{$LDFLAGS}"
 $LIBPATH = ["#{HERE}/lib"]
 $DEFLIBPATH = []
 
+# fix for rbenv?
+if $LIBRUBYARG == ''
+  $LIBRUBYARG = RbConfig::CONFIG['LIBRUBYARG']
+end
+
 def safe_sh(cmd)
   puts cmd
   system(cmd)
@@ -87,6 +92,8 @@ Dir.chdir("#{HERE}/lib") do
 end
 $LIBS << " -lzookeeper_st_gem"
 
+have_func('rb_thread_blocking_region')
+
 $CFLAGS << ' -Wall' if ZK_DEV
 create_makefile 'zookeeper_c'
 

data/ext/zkrb.c

@@ -824,7 +824,6 @@ static VALUE method_zkrb_iterate_event_loop(VALUE self) {
   return INT2FIX(rc);
 }
 
-
 static VALUE method_has_events(VALUE self) {
   VALUE rb_event;
   FETCH_DATA_PTR(self, zk);
@@ -882,16 +881,8 @@ static VALUE method_recv_timeout(VALUE self) {
 // returns a CZookeeper::ClientId object with the values set for session_id and passwd
 static VALUE method_client_id(VALUE self) {
   FETCH_DATA_PTR(self, zk);
-  char buf[32];
   const clientid_t *cid = zoo_client_id(zk->zh);
 
-/*  if (strlen(cid->passwd) != 16) { */
-/*    zkrb_debug("passwd is not null-terminated");*/
-/*  } else {*/
-/*    hexbufify(buf, cid->passwd, 16);*/
-/*    zkrb_debug("password in hex is: %s", buf);*/
-/*  }*/
-
   VALUE session_id = LL2NUM(cid->client_id);
   VALUE passwd = rb_str_new(cid->passwd, 16);
 

data/ext/zkrb_wrapper_compat.c

@@ -4,10 +4,10 @@
 
 VALUE zkrb_thread_blocking_region(zkrb_blocking_function_t *func, void *data1) {
 
-#ifdef ZKRB_RUBY_187
-  return func(data1);
-#else
+#ifdef HAVE_RB_THREAD_BLOCKING_REGION
   return rb_thread_blocking_region((rb_blocking_function_t *)func, data1, RUBY_UBF_IO, 0);
+#else
+  return func(data1);
 #endif
 
 }

data/ext/zookeeper_base.rb

@@ -31,17 +31,15 @@ class ZookeeperBase
   ZOO_LOG_LEVEL_DEBUG  = 4
 
 
-  def_delegators :@czk, :get_children, :exists, :delete, :get, :set,
+  def_delegators :czk, :get_children, :exists, :delete, :get, :set,
     :set_acl, :get_acl, :client_id, :sync, :wait_until_connected
 
-  # some state methods need to be more paranoid about locking to ensure the correct
-  # state is returned
-  # 
   def self.threadsafe_inquisitor(*syms)
     syms.each do |sym|
       class_eval(<<-EOM, __FILE__, __LINE__+1)
         def #{sym}
-          false|@mutex.synchronize { @czk and @czk.#{sym} }
+          c = @mutex.synchronize { @czk }
+          false|(c && c.#{sym})
         end
       EOM
     end
@@ -117,7 +115,7 @@ class ZookeeperBase
   # if either of these happen, the user will need to renegotiate a connection via reopen
   def assert_open
     @mutex.synchronize do
-      raise Exceptions::NotConnected if closed?
+      raise Exceptions::NotConnected if !@czk or @czk.closed?
       if forked?
         raise InheritedConnectionError, <<-EOS.gsub(/(?:^|\n)\s*/, ' ').strip
           You tried to use a connection inherited from another process 
@@ -152,7 +150,7 @@ class ZookeeperBase
   # is pretty damn annoying. this is used to clean things up.
   def create(*args)
     # since we don't care about the inputs, just glob args
-    rc, new_path = @mutex.synchronize { @czk.create(*args) }
+    rc, new_path = czk.create(*args)
     [rc, strip_chroot_from(new_path)]
   end
 
@@ -173,7 +171,7 @@ class ZookeeperBase
 
   def state
     return ZOO_CLOSED_STATE if closed?
-    @mutex.synchronize { @czk.state }
+    czk.state
   end
 
   def session_id
@@ -190,7 +188,9 @@ class ZookeeperBase
 
   # we are closed if there is no @czk instance or @czk.closed?
   def closed?
-    @mutex.synchronize { !@czk or @czk.closed? } 
+    czk.closed?
+  rescue Exceptions::NotConnected
+    true
   end
 
   def pause_before_fork_in_parent
@@ -243,6 +243,12 @@ protected
     super(req_id, meth_name, call_opts)
   end
 
+  def czk
+    rval = @mutex.synchronize { @czk }
+    raise Exceptions::NotConnected unless rval
+    rval
+  end
+
   # if we're chrooted, this method will strip the chroot prefix from +path+
   def strip_chroot_from(path)
     return path unless (chrooted? and path and path.start_with?(chroot_path))

data/lib/zookeeper/continuation.rb

@@ -6,6 +6,8 @@ module Zookeeper
     include Constants
     include Logger
 
+    OPERATION_TIMEOUT = 30 # seconds
+
     # for keeping track of which continuations are pending, and which ones have
     # been submitted and are awaiting a repsonse
     # 
@@ -79,8 +81,8 @@ module Zookeeper
     def initialize(meth, *args)
       @meth   = meth
       @args   = args
-      @mutex  = Mutex.new
-      @cond   = ConditionVariable.new
+      @mutex  = Monitor.new
+      @cond   = @mutex.new_cond
       @rval   = nil
 
       # make this error reporting more robust if necessary, right now, just set to state
@@ -88,9 +90,28 @@ module Zookeeper
     end
 
     # the caller calls this method and receives the response from the async loop
+    # this method has a hard-coded 30 second timeout as a safety feature. No
+    # call should take more than 20s (as the session timeout is set to 20s)
+    # so if any call takes longer than that, something has gone horribly wrong.
+    #
+    # @raise [ContinuationTimeoutError] if a response is not received within 30s
+    #
     def value
+      time_to_stop = Time.now + OPERATION_TIMEOUT
+      now = nil
+
       @mutex.synchronize do
-        @cond.wait(@mutex) until @rval or @error
+        while true
+          now = Time.now
+          break if @rval or @error or (now > time_to_stop)
+
+          deadline = time_to_stop.to_f - now.to_f
+          @cond.wait(deadline)
+        end
+
+        if now > time_to_stop
+          raise ContinuationTimeoutError, "response for meth: #{meth.inspect}, args: #{args.inspect}, not received within #{OPERATION_TIMEOUT} seconds"
+        end
 
         case @error
         when nil

data/lib/zookeeper/exceptions.rb

@@ -84,6 +84,10 @@ stacktrace:
   # maybe use this for continuation
   class InterruptedException  < ZookeeperException ; end
 
+  # raised when a continuation operation takes more time than is reasonable and
+  # the thread should be awoken. (i.e. prevents a call that never returns)
+  class ContinuationTimeoutError < ZookeeperException; end
+
   # raised when the user tries to use a connection after a fork()
   # without calling reopen() in the C client
   #

data/lib/zookeeper/version.rb

@@ -1,4 +1,4 @@
 module Zookeeper
-  VERSION = '1.2.4'
+  VERSION = '1.2.5'
   DRIVER_VERSION = '3.3.5'
 end

metadata

@@ -1,13 +1,13 @@
 --- !ruby/object:Gem::Specification 
 name: zookeeper
 version: !ruby/object:Gem::Version 
-  hash: 23
+  hash: 21
   prerelease: 
   segments: 
   - 1
   - 2
-  - 4
-  version: 1.2.4
+  - 5
+  version: 1.2.5
 platform: java
 authors: 
 - Phillip Pearson
@@ -20,7 +20,7 @@ autorequire:
 bindir: bin
 cert_chain: []
 
-date: 2012-05-24 00:00:00 Z
+date: 2012-06-03 00:00:00 Z
 dependencies: 
 - !ruby/object:Gem::Dependency 
   name: backports