Merge pull request #8069 from grpc/revert-8068-revert-7279-grand-unified-closures

Revert "Revert "Grand unified closures""

27 files changed, 809 insertions, 146 deletions

diff --git a/src/core/lib/channel/channel_stack.c b/src/core/lib/channel/channel_stack.c
index 98f304f2da..0655b9353f 100644
--- a/src/core/lib/channel/channel_stack.c
+++ b/src/core/lib/channel/channel_stack.c
@@ -32,6 +32,7 @@
  */
 
 #include "src/core/lib/channel/channel_stack.h"
+#include <grpc/support/alloc.h>
 #include <grpc/support/log.h>
 
 #include <stdlib.h>
@@ -270,21 +271,27 @@ grpc_call_stack *grpc_call_stack_from_top_element(grpc_call_element *elem) {
       sizeof(grpc_call_stack)));
 }
 
+static void destroy_op(grpc_exec_ctx *exec_ctx, void *op, grpc_error *error) {
+  gpr_free(op);
+}
+
 void grpc_call_element_send_cancel(grpc_exec_ctx *exec_ctx,
                                    grpc_call_element *cur_elem) {
-  grpc_transport_stream_op op;
-  memset(&op, 0, sizeof(op));
-  op.cancel_error = GRPC_ERROR_CANCELLED;
-  grpc_call_next_op(exec_ctx, cur_elem, &op);
+  grpc_transport_stream_op *op = gpr_malloc(sizeof(*op));
+  memset(op, 0, sizeof(*op));
+  op->cancel_error = GRPC_ERROR_CANCELLED;
+  op->on_complete = grpc_closure_create(destroy_op, op);
+  grpc_call_next_op(exec_ctx, cur_elem, op);
 }
 
 void grpc_call_element_send_cancel_with_message(grpc_exec_ctx *exec_ctx,
                                                 grpc_call_element *cur_elem,
                                                 grpc_status_code status,
                                                 gpr_slice *optional_message) {
-  grpc_transport_stream_op op;
-  memset(&op, 0, sizeof(op));
-  grpc_transport_stream_op_add_cancellation_with_message(&op, status,
+  grpc_transport_stream_op *op = gpr_malloc(sizeof(*op));
+  memset(op, 0, sizeof(*op));
+  op->on_complete = grpc_closure_create(destroy_op, op);
+  grpc_transport_stream_op_add_cancellation_with_message(op, status,
                                                          optional_message);
-  grpc_call_next_op(exec_ctx, cur_elem, &op);
+  grpc_call_next_op(exec_ctx, cur_elem, op);
 }
diff --git a/src/core/lib/channel/compress_filter.c b/src/core/lib/channel/compress_filter.c
index 134180e619..0981d59f63 100644
--- a/src/core/lib/channel/compress_filter.c
+++ b/src/core/lib/channel/compress_filter.c
@@ -60,7 +60,7 @@ typedef struct call_data {
   /** If true, contents of \a compression_algorithm are authoritative */
   int has_compression_algorithm;
 
-  grpc_transport_stream_op send_op;
+  grpc_transport_stream_op *send_op;
   uint32_t send_length;
   uint32_t send_flags;
   gpr_slice incoming_slice;
@@ -199,11 +199,11 @@ static void finish_send_message(grpc_exec_ctx *exec_ctx,
 
   grpc_slice_buffer_stream_init(&calld->replacement_stream, &calld->slices,
                                 calld->send_flags);
-  calld->send_op.send_message = &calld->replacement_stream.base;
-  calld->post_send = calld->send_op.on_complete;
-  calld->send_op.on_complete = &calld->send_done;
+  calld->send_op->send_message = &calld->replacement_stream.base;
+  calld->post_send = calld->send_op->on_complete;
+  calld->send_op->on_complete = &calld->send_done;
 
-  grpc_call_next_op(exec_ctx, elem, &calld->send_op);
+  grpc_call_next_op(exec_ctx, elem, calld->send_op);
 }
 
 static void got_slice(grpc_exec_ctx *exec_ctx, void *elemp, grpc_error *error) {
@@ -220,7 +220,7 @@ static void got_slice(grpc_exec_ctx *exec_ctx, void *elemp, grpc_error *error) {
 static void continue_send_message(grpc_exec_ctx *exec_ctx,
                                   grpc_call_element *elem) {
   call_data *calld = elem->call_data;
-  while (grpc_byte_stream_next(exec_ctx, calld->send_op.send_message,
+  while (grpc_byte_stream_next(exec_ctx, calld->send_op->send_message,
                                &calld->incoming_slice, ~(size_t)0,
                                &calld->got_slice)) {
     gpr_slice_buffer_add(&calld->slices, calld->incoming_slice);
@@ -243,7 +243,7 @@ static void compress_start_transport_stream_op(grpc_exec_ctx *exec_ctx,
   }
   if (op->send_message != NULL && !skip_compression(elem) &&
       0 == (op->send_message->flags & GRPC_WRITE_NO_COMPRESS)) {
-    calld->send_op = *op;
+    calld->send_op = op;
     calld->send_length = op->send_message->length;
     calld->send_flags = op->send_message->flags;
     continue_send_message(exec_ctx, elem);
diff --git a/src/core/lib/iomgr/closure.c b/src/core/lib/iomgr/closure.c
index 0b6c3b2539..1ba0a5c141 100644
--- a/src/core/lib/iomgr/closure.c
+++ b/src/core/lib/iomgr/closure.c
@@ -41,6 +41,10 @@ void grpc_closure_init(grpc_closure *closure, grpc_iomgr_cb_func cb,
   closure->cb_arg = cb_arg;
 }
 
+void grpc_closure_list_init(grpc_closure_list *closure_list) {
+  closure_list->head = closure_list->tail = NULL;
+}
+
 void grpc_closure_list_append(grpc_closure_list *closure_list,
                               grpc_closure *closure, grpc_error *error) {
   if (closure == NULL) {
diff --git a/src/core/lib/iomgr/closure.h b/src/core/lib/iomgr/closure.h
index 08e59a168e..c1a22b6021 100644
--- a/src/core/lib/iomgr/closure.h
+++ b/src/core/lib/iomgr/closure.h
@@ -37,6 +37,7 @@
 #include <grpc/support/port_platform.h>
 #include <stdbool.h>
 #include "src/core/lib/iomgr/error.h"
+#include "src/core/lib/support/mpscq.h"
 
 struct grpc_closure;
 typedef struct grpc_closure grpc_closure;
@@ -60,6 +61,14 @@ typedef void (*grpc_iomgr_cb_func)(grpc_exec_ctx *exec_ctx, void *arg,
 
 /** A closure over a grpc_iomgr_cb_func. */
 struct grpc_closure {
+  /** Once queued, next indicates the next queued closure; before then, scratch
+   *  space */
+  union {
+    grpc_closure *next;
+    gpr_mpscq_node atm_next;
+    uintptr_t scratch;
+  } next_data;
+
   /** Bound callback. */
   grpc_iomgr_cb_func cb;
 
@@ -68,13 +77,6 @@ struct grpc_closure {
 
   /** Once queued, the result of the closure. Before then: scratch space */
   grpc_error *error;
-
-  /** Once queued, next indicates the next queued closure; before then, scratch
-   *  space */
-  union {
-    grpc_closure *next;
-    uintptr_t scratch;
-  } next_data;
 };
 
 /** Initializes \a closure with \a cb and \a cb_arg. */
@@ -87,6 +89,8 @@ grpc_closure *grpc_closure_create(grpc_iomgr_cb_func cb, void *cb_arg);
 #define GRPC_CLOSURE_LIST_INIT \
   { NULL, NULL }
 
+void grpc_closure_list_init(grpc_closure_list *list);
+
 /** add \a closure to the end of \a list
     and set \a closure's result to \a error */
 void grpc_closure_list_append(grpc_closure_list *list, grpc_closure *closure,
diff --git a/src/core/lib/iomgr/combiner.c b/src/core/lib/iomgr/combiner.c
new file mode 100644
index 0000000000..831bdb4aff
--- /dev/null
+++ b/src/core/lib/iomgr/combiner.c
@@ -0,0 +1,293 @@
+/*
+ *
+ * Copyright 2016, Google Inc.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *     * Neither the name of Google Inc. nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include "src/core/lib/iomgr/combiner.h"
+
+#include <string.h>
+
+#include <grpc/support/alloc.h>
+#include <grpc/support/log.h>
+
+#include "src/core/lib/iomgr/workqueue.h"
+#include "src/core/lib/profiling/timers.h"
+
+int grpc_combiner_trace = 0;
+
+#define GRPC_COMBINER_TRACE(fn) \
+  do {                          \
+    if (grpc_combiner_trace) {  \
+      fn;                       \
+    }                           \
+  } while (0)
+
+struct grpc_combiner {
+  grpc_workqueue *optional_workqueue;
+  gpr_mpscq queue;
+  // state is:
+  // lower bit - zero if orphaned
+  // other bits - number of items queued on the lock
+  gpr_atm state;
+  bool take_async_break_before_final_list;
+  grpc_closure_list final_list;
+  grpc_closure continue_finishing;
+};
+
+grpc_combiner *grpc_combiner_create(grpc_workqueue *optional_workqueue) {
+  grpc_combiner *lock = gpr_malloc(sizeof(*lock));
+  lock->optional_workqueue = optional_workqueue;
+  gpr_atm_no_barrier_store(&lock->state, 1);
+  gpr_mpscq_init(&lock->queue);
+  lock->take_async_break_before_final_list = false;
+  grpc_closure_list_init(&lock->final_list);
+  GRPC_COMBINER_TRACE(gpr_log(GPR_DEBUG, "C:%p create", lock));
+  return lock;
+}
+
+static void really_destroy(grpc_exec_ctx *exec_ctx, grpc_combiner *lock) {
+  GRPC_COMBINER_TRACE(gpr_log(GPR_DEBUG, "C:%p really_destroy", lock));
+  GPR_ASSERT(gpr_atm_no_barrier_load(&lock->state) == 0);
+  gpr_mpscq_destroy(&lock->queue);
+  GRPC_WORKQUEUE_UNREF(exec_ctx, lock->optional_workqueue, "combiner");
+  gpr_free(lock);
+}
+
+void grpc_combiner_destroy(grpc_exec_ctx *exec_ctx, grpc_combiner *lock) {
+  gpr_atm old_state = gpr_atm_full_fetch_add(&lock->state, -1);
+  GRPC_COMBINER_TRACE(gpr_log(
+      GPR_DEBUG, "C:%p really_destroy old_state=%" PRIdPTR, lock, old_state));
+  if (old_state == 1) {
+    really_destroy(exec_ctx, lock);
+  }
+}
+
+static bool maybe_finish_one(grpc_exec_ctx *exec_ctx, grpc_combiner *lock);
+static void finish(grpc_exec_ctx *exec_ctx, grpc_combiner *lock);
+
+static void continue_finishing_mainline(grpc_exec_ctx *exec_ctx, void *arg,
+                                        grpc_error *error) {
+  GPR_TIMER_BEGIN("combiner.continue_executing_mainline", 0);
+  grpc_combiner *lock = arg;
+  GRPC_COMBINER_TRACE(
+      gpr_log(GPR_DEBUG, "C:%p continue_finishing_mainline", lock));
+  GPR_ASSERT(exec_ctx->active_combiner == NULL);
+  exec_ctx->active_combiner = lock;
+  if (maybe_finish_one(exec_ctx, lock)) finish(exec_ctx, lock);
+  GPR_ASSERT(exec_ctx->active_combiner == lock);
+  exec_ctx->active_combiner = NULL;
+  GPR_TIMER_END("combiner.continue_executing_mainline", 0);
+}
+
+static void execute_final(grpc_exec_ctx *exec_ctx, grpc_combiner *lock) {
+  GPR_TIMER_BEGIN("combiner.execute_final", 0);
+  grpc_closure *c = lock->final_list.head;
+  GPR_ASSERT(c != NULL);
+  grpc_closure_list_init(&lock->final_list);
+  lock->take_async_break_before_final_list = false;
+  int loops = 0;
+  while (c != NULL) {
+    GRPC_COMBINER_TRACE(
+        gpr_log(GPR_DEBUG, "C:%p execute_final[%d] c=%p", lock, loops, c));
+    grpc_closure *next = c->next_data.next;
+    grpc_error *error = c->error;
+    c->cb(exec_ctx, c->cb_arg, error);
+    GRPC_ERROR_UNREF(error);
+    c = next;
+    loops++;
+  }
+  GPR_TIMER_END("combiner.execute_final", 0);
+}
+
+static void continue_executing_final(grpc_exec_ctx *exec_ctx, void *arg,
+                                     grpc_error *error) {
+  GPR_TIMER_BEGIN("combiner.continue_executing_final", 0);
+  grpc_combiner *lock = arg;
+  GRPC_COMBINER_TRACE(
+      gpr_log(GPR_DEBUG, "C:%p continue_executing_final", lock));
+  GPR_ASSERT(exec_ctx->active_combiner == NULL);
+  exec_ctx->active_combiner = lock;
+  // quick peek to see if new things have turned up on the queue: if so, go back
+  // to executing them before the final list
+  if ((gpr_atm_acq_load(&lock->state) >> 1) > 1) {
+    if (maybe_finish_one(exec_ctx, lock)) finish(exec_ctx, lock);
+  } else {
+    execute_final(exec_ctx, lock);
+    finish(exec_ctx, lock);
+  }
+  GPR_ASSERT(exec_ctx->active_combiner == lock);
+  exec_ctx->active_combiner = NULL;
+  GPR_TIMER_END("combiner.continue_executing_final", 0);
+}
+
+static bool start_execute_final(grpc_exec_ctx *exec_ctx, grpc_combiner *lock) {
+  GPR_TIMER_BEGIN("combiner.start_execute_final", 0);
+  GPR_ASSERT(exec_ctx->active_combiner == lock);
+  GRPC_COMBINER_TRACE(
+      gpr_log(GPR_DEBUG,
+              "C:%p start_execute_final take_async_break_before_final_list=%d",
+              lock, lock->take_async_break_before_final_list));
+  if (lock->take_async_break_before_final_list) {
+    grpc_closure_init(&lock->continue_finishing, continue_executing_final,
+                      lock);
+    grpc_exec_ctx_sched(exec_ctx, &lock->continue_finishing, GRPC_ERROR_NONE,
+                        GRPC_WORKQUEUE_REF(lock->optional_workqueue, "sched"));
+    GPR_TIMER_END("combiner.start_execute_final", 0);
+    return false;
+  } else {
+    execute_final(exec_ctx, lock);
+    GPR_TIMER_END("combiner.start_execute_final", 0);
+    return true;
+  }
+}
+
+static bool maybe_finish_one(grpc_exec_ctx *exec_ctx, grpc_combiner *lock) {
+  GPR_TIMER_BEGIN("combiner.maybe_finish_one", 0);
+  gpr_mpscq_node *n = gpr_mpscq_pop(&lock->queue);
+  GRPC_COMBINER_TRACE(
+      gpr_log(GPR_DEBUG, "C:%p maybe_finish_one n=%p", lock, n));
+  GPR_ASSERT(exec_ctx->active_combiner == lock);
+  if (n == NULL) {
+    // Queue is in an transiently inconsistent state: a new item is being queued
+    // but is not visible to this thread yet.
+    // Use this as a cue that we should go off and do something else for a while
+    // (and come back later)
+    grpc_closure_init(&lock->continue_finishing, continue_finishing_mainline,
+                      lock);
+    grpc_exec_ctx_sched(exec_ctx, &lock->continue_finishing, GRPC_ERROR_NONE,
+                        GRPC_WORKQUEUE_REF(lock->optional_workqueue, "sched"));
+    GPR_TIMER_END("combiner.maybe_finish_one", 0);
+    return false;
+  }
+  grpc_closure *cl = (grpc_closure *)n;
+  grpc_error *error = cl->error;
+  cl->cb(exec_ctx, cl->cb_arg, error);
+  GRPC_ERROR_UNREF(error);
+  GPR_TIMER_END("combiner.maybe_finish_one", 0);
+  return true;
+}
+
+static void finish(grpc_exec_ctx *exec_ctx, grpc_combiner *lock) {
+  bool (*executor)(grpc_exec_ctx * exec_ctx, grpc_combiner * lock);
+  GPR_TIMER_BEGIN("combiner.finish", 0);
+  int loops = 0;
+  do {
+    executor = maybe_finish_one;
+    gpr_atm old_state = gpr_atm_full_fetch_add(&lock->state, -2);
+    GRPC_COMBINER_TRACE(gpr_log(GPR_DEBUG,
+                                "C:%p finish[%d] old_state=%" PRIdPTR, lock,
+                                loops, old_state));
+    switch (old_state) {
+      default:
+        // we have multiple queued work items: just continue executing them
+        break;
+      case 5:  // we're down to one queued item: if it's the final list we
+      case 4:  // should do that
+        if (!grpc_closure_list_empty(lock->final_list)) {
+          executor = start_execute_final;
+        }
+        break;
+      case 3:  // had one count, one unorphaned --> unlocked unorphaned
+        GPR_TIMER_END("combiner.finish", 0);
+        return;
+      case 2:  // and one count, one orphaned --> unlocked and orphaned
+        really_destroy(exec_ctx, lock);
+        GPR_TIMER_END("combiner.finish", 0);
+        return;
+      case 1:
+      case 0:
+        // these values are illegal - representing an already unlocked or
+        // deleted lock
+        GPR_UNREACHABLE_CODE(return );
+    }
+    loops++;
+  } while (executor(exec_ctx, lock));
+  GPR_TIMER_END("combiner.finish", 0);
+}
+
+void grpc_combiner_execute(grpc_exec_ctx *exec_ctx, grpc_combiner *lock,
+                           grpc_closure *cl, grpc_error *error) {
+  GRPC_COMBINER_TRACE(
+      gpr_log(GPR_DEBUG, "C:%p grpc_combiner_execute c=%p", lock, cl));
+  GPR_TIMER_BEGIN("combiner.execute", 0);
+  gpr_atm last = gpr_atm_full_fetch_add(&lock->state, 2);
+  GPR_ASSERT(last & 1);  // ensure lock has not been destroyed
+  if (last == 1) {
+    exec_ctx->active_combiner = lock;
+    GPR_TIMER_BEGIN("combiner.execute_first_cb", 0);
+    cl->cb(exec_ctx, cl->cb_arg, error);
+    GPR_TIMER_END("combiner.execute_first_cb", 0);
+    GRPC_ERROR_UNREF(error);
+    finish(exec_ctx, lock);
+    GPR_ASSERT(exec_ctx->active_combiner == lock);
+    exec_ctx->active_combiner = NULL;
+  } else {
+    cl->error = error;
+    gpr_mpscq_push(&lock->queue, &cl->next_data.atm_next);
+  }
+  GPR_TIMER_END("combiner.execute", 0);
+}
+
+static void enqueue_finally(grpc_exec_ctx *exec_ctx, void *closure,
+                            grpc_error *error) {
+  grpc_combiner_execute_finally(exec_ctx, exec_ctx->active_combiner, closure,
+                                GRPC_ERROR_REF(error), false);
+}
+
+void grpc_combiner_execute_finally(grpc_exec_ctx *exec_ctx, grpc_combiner *lock,
+                                   grpc_closure *closure, grpc_error *error,
+                                   bool force_async_break) {
+  GRPC_COMBINER_TRACE(gpr_log(
+      GPR_DEBUG,
+      "C:%p grpc_combiner_execute_finally c=%p force_async_break=%d; ac=%p",
+      lock, closure, force_async_break, exec_ctx->active_combiner));
+  GPR_TIMER_BEGIN("combiner.execute_finally", 0);
+  if (exec_ctx->active_combiner != lock) {
+    GPR_TIMER_MARK("slowpath", 0);
+    grpc_combiner_execute(exec_ctx, lock,
+                          grpc_closure_create(enqueue_finally, closure), error);
+    GPR_TIMER_END("combiner.execute_finally", 0);
+    return;
+  }
+
+  if (force_async_break) {
+    lock->take_async_break_before_final_list = true;
+  }
+  if (grpc_closure_list_empty(lock->final_list)) {
+    gpr_atm_full_fetch_add(&lock->state, 2);
+  }
+  grpc_closure_list_append(&lock->final_list, closure, error);
+  GPR_TIMER_END("combiner.execute_finally", 0);
+}
+
+void grpc_combiner_force_async_finally(grpc_combiner *lock) {
+  lock->take_async_break_before_final_list = true;
+}
diff --git a/src/core/lib/iomgr/combiner.h b/src/core/lib/iomgr/combiner.h
new file mode 100644
index 0000000000..1409db24b9
--- /dev/null
+++ b/src/core/lib/iomgr/combiner.h
@@ -0,0 +1,71 @@
+/*
+ *
+ * Copyright 2016, Google Inc.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *     * Neither the name of Google Inc. nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#ifndef GRPC_CORE_LIB_IOMGR_COMBINER_H
+#define GRPC_CORE_LIB_IOMGR_COMBINER_H
+
+#include <stddef.h>
+
+#include <grpc/support/atm.h>
+#include "src/core/lib/iomgr/exec_ctx.h"
+#include "src/core/lib/support/mpscq.h"
+
+// Provides serialized access to some resource.
+// Each action queued on a combiner is executed serially in a borrowed thread.
+// The actual thread executing actions may change over time (but there will only
+// every be one at a time).
+
+// Initialize the lock, with an optional workqueue to shift load to when
+// necessary
+grpc_combiner *grpc_combiner_create(grpc_workqueue *optional_workqueue);
+// Destroy the lock
+void grpc_combiner_destroy(grpc_exec_ctx *exec_ctx, grpc_combiner *lock);
+// Execute \a action within the lock.
+void grpc_combiner_execute(grpc_exec_ctx *exec_ctx, grpc_combiner *lock,
+                           grpc_closure *closure, grpc_error *error);
+// Execute \a action within the lock just prior to unlocking.
+// if \a hint_async_break is true, the combiner tries to hand execution to
+// another thread before finishing the primary queue of combined closures and
+// executing the finally list.
+// Deprecation warning: \a hint_async_break will be removed in a future version
+// Takes a very slow and round-about path if not called from a
+// grpc_combiner_execute closure.
+void grpc_combiner_execute_finally(grpc_exec_ctx *exec_ctx, grpc_combiner *lock,
+                                   grpc_closure *closure, grpc_error *error,
+                                   bool hint_async_break);
+// Deprecated: force the finally list execution onto another thread
+void grpc_combiner_force_async_finally(grpc_combiner *lock);
+
+extern int grpc_combiner_trace;
+
+#endif /* GRPC_CORE_LIB_IOMGR_COMBINER_H */
diff --git a/src/core/lib/iomgr/error.c b/src/core/lib/iomgr/error.c
index 149c55663c..e366961936 100644
--- a/src/core/lib/iomgr/error.c
+++ b/src/core/lib/iomgr/error.c
@@ -332,7 +332,7 @@ grpc_error *grpc_error_add_child(grpc_error *src, grpc_error *child) {
   return new;
 }
 
-static const char *no_error_string = "null";
+static const char *no_error_string = "\"No Error\"";
 static const char *oom_error_string = "\"Out of memory\"";
 static const char *cancelled_error_string = "\"Cancelled\"";
 
diff --git a/src/core/lib/iomgr/ev_epoll_linux.c b/src/core/lib/iomgr/ev_epoll_linux.c
index eba347125e..740920d760 100644
--- a/src/core/lib/iomgr/ev_epoll_linux.c
+++ b/src/core/lib/iomgr/ev_epoll_linux.c
@@ -1531,6 +1531,8 @@ static grpc_error *pollset_work(grpc_exec_ctx *exec_ctx, grpc_pollset *pollset,
 
 static void pollset_add_fd(grpc_exec_ctx *exec_ctx, grpc_pollset *pollset,
                            grpc_fd *fd) {
+  GPR_TIMER_BEGIN("pollset_add_fd", 0);
+
   grpc_error *error = GRPC_ERROR_NONE;
 
   gpr_mu_lock(&pollset->mu);
@@ -1643,6 +1645,8 @@ retry:
   gpr_mu_unlock(&pollset->mu);
 
   GRPC_LOG_IF_ERROR("pollset_add_fd", error);
+
+  GPR_TIMER_END("pollset_add_fd", 0);
 }
 
 /*******************************************************************************
diff --git a/src/core/lib/iomgr/exec_ctx.h b/src/core/lib/iomgr/exec_ctx.h
index 917f332f03..1895ee6245 100644
--- a/src/core/lib/iomgr/exec_ctx.h
+++ b/src/core/lib/iomgr/exec_ctx.h
@@ -40,8 +40,8 @@
 
 /** A workqueue represents a list of work to be executed asynchronously.
     Forward declared here to avoid a circular dependency with workqueue.h. */
-struct grpc_workqueue;
 typedef struct grpc_workqueue grpc_workqueue;
+typedef struct grpc_combiner grpc_combiner;
 
 #ifndef GRPC_EXECUTION_CONTEXT_SANITIZER
 /** Execution context.
@@ -66,13 +66,15 @@ typedef struct grpc_workqueue grpc_workqueue;
  */
 struct grpc_exec_ctx {
   grpc_closure_list closure_list;
+  /** currently active combiner: updated only via combiner.c */
+  grpc_combiner *active_combiner;
   bool cached_ready_to_finish;
   void *check_ready_to_finish_arg;
   bool (*check_ready_to_finish)(grpc_exec_ctx *exec_ctx, void *arg);
 };
 
 #define GRPC_EXEC_CTX_INIT_WITH_FINISH_CHECK(finish_check, finish_check_arg) \
-  { GRPC_CLOSURE_LIST_INIT, false, finish_check_arg, finish_check }
+  { GRPC_CLOSURE_LIST_INIT, NULL, false, finish_check_arg, finish_check }
 #else
 struct grpc_exec_ctx {
   bool cached_ready_to_finish;
diff --git a/src/core/lib/iomgr/tcp_posix.c b/src/core/lib/iomgr/tcp_posix.c
index 974d5ae479..92767721d5 100644
--- a/src/core/lib/iomgr/tcp_posix.c
+++ b/src/core/lib/iomgr/tcp_posix.c
@@ -379,10 +379,19 @@ static void tcp_handle_write(grpc_exec_ctx *exec_ctx, void *arg /* grpc_tcp */,
   }
 
   if (!tcp_flush(tcp, &error)) {
+    if (grpc_tcp_trace) {
+      gpr_log(GPR_DEBUG, "write: delayed");
+    }
     grpc_fd_notify_on_write(exec_ctx, tcp->em_fd, &tcp->write_closure);
   } else {
     cb = tcp->write_cb;
     tcp->write_cb = NULL;
+    if (grpc_tcp_trace) {
+      const char *str = grpc_error_string(error);
+      gpr_log(GPR_DEBUG, "write: %s", str);
+      grpc_error_free_string(str);
+    }
+
     GPR_TIMER_BEGIN("tcp_handle_write.cb", 0);
     cb->cb(exec_ctx, cb->cb_arg, error);
     GPR_TIMER_END("tcp_handle_write.cb", 0);
@@ -425,8 +434,16 @@ static void tcp_write(grpc_exec_ctx *exec_ctx, grpc_endpoint *ep,
   if (!tcp_flush(tcp, &error)) {
     TCP_REF(tcp, "write");
     tcp->write_cb = cb;
+    if (grpc_tcp_trace) {
+      gpr_log(GPR_DEBUG, "write: delayed");
+    }
     grpc_fd_notify_on_write(exec_ctx, tcp->em_fd, &tcp->write_closure);
   } else {
+    if (grpc_tcp_trace) {
+      const char *str = grpc_error_string(error);
+      gpr_log(GPR_DEBUG, "write: %s", str);
+      grpc_error_free_string(str);
+    }
     grpc_exec_ctx_sched(exec_ctx, cb, error, NULL);
   }
 
diff --git a/src/core/lib/iomgr/workqueue.h b/src/core/lib/iomgr/workqueue.h
index 7156e490d7..b2805dc66c 100644
--- a/src/core/lib/iomgr/workqueue.h
+++ b/src/core/lib/iomgr/workqueue.h
@@ -50,10 +50,6 @@
 
 /* grpc_workqueue is forward declared in exec_ctx.h */
 
-/* Deprecated: do not use.
-   This has *already* been removed in a future commit. */
-void grpc_workqueue_flush(grpc_exec_ctx *exec_ctx, grpc_workqueue *workqueue);
-
 /* Reference counting functions. Use the macro's always
    (GRPC_WORKQUEUE_{REF,UNREF}).
 
diff --git a/src/core/lib/iomgr/workqueue_posix.c b/src/core/lib/iomgr/workqueue_posix.c
index e0d6dac230..ecfea68f56 100644
--- a/src/core/lib/iomgr/workqueue_posix.c
+++ b/src/core/lib/iomgr/workqueue_posix.c
@@ -44,6 +44,7 @@
 #include <grpc/support/useful.h>
 
 #include "src/core/lib/iomgr/ev_posix.h"
+#include "src/core/lib/profiling/timers.h"
 
 static void on_readable(grpc_exec_ctx *exec_ctx, void *arg, grpc_error *error);
 
@@ -52,8 +53,7 @@ grpc_error *grpc_workqueue_create(grpc_exec_ctx *exec_ctx,
   char name[32];
   *workqueue = gpr_malloc(sizeof(grpc_workqueue));
   gpr_ref_init(&(*workqueue)->refs, 1);
-  gpr_mu_init(&(*workqueue)->mu);
-  (*workqueue)->closure_list.head = (*workqueue)->closure_list.tail = NULL;
+  gpr_atm_no_barrier_store(&(*workqueue)->state, 1);
   grpc_error *err = grpc_wakeup_fd_init(&(*workqueue)->wakeup_fd);
   if (err != GRPC_ERROR_NONE) {
     gpr_free(*workqueue);
@@ -62,6 +62,7 @@ grpc_error *grpc_workqueue_create(grpc_exec_ctx *exec_ctx,
   sprintf(name, "workqueue:%p", (void *)(*workqueue));
   (*workqueue)->wakeup_read_fd = grpc_fd_create(
       GRPC_WAKEUP_FD_GET_READ_FD(&(*workqueue)->wakeup_fd), name);
+  gpr_mpscq_init(&(*workqueue)->queue);
   grpc_closure_init(&(*workqueue)->read_closure, on_readable, *workqueue);
   grpc_fd_notify_on_read(exec_ctx, (*workqueue)->wakeup_read_fd,
                          &(*workqueue)->read_closure);
@@ -70,57 +71,79 @@ grpc_error *grpc_workqueue_create(grpc_exec_ctx *exec_ctx,
 
 static void workqueue_destroy(grpc_exec_ctx *exec_ctx,
                               grpc_workqueue *workqueue) {
-  grpc_exec_ctx_enqueue_list(exec_ctx, &workqueue->closure_list, NULL);
   grpc_fd_shutdown(exec_ctx, workqueue->wakeup_read_fd);
 }
 
+static void workqueue_orphan(grpc_exec_ctx *exec_ctx,
+                             grpc_workqueue *workqueue) {
+  if (gpr_atm_full_fetch_add(&workqueue->state, -1) == 1) {
+    workqueue_destroy(exec_ctx, workqueue);
+  }
+}
+
 #ifdef GRPC_WORKQUEUE_REFCOUNT_DEBUG
 void grpc_workqueue_ref(grpc_workqueue *workqueue, const char *file, int line,
                         const char *reason) {
+  if (workqueue == NULL) return;
   gpr_log(file, line, GPR_LOG_SEVERITY_DEBUG, "WORKQUEUE:%p   ref %d -> %d %s",
           workqueue, (int)workqueue->refs.count, (int)workqueue->refs.count + 1,
           reason);
+  gpr_ref(&workqueue->refs);
+}
 #else
 void grpc_workqueue_ref(grpc_workqueue *workqueue) {
-#endif
+  if (workqueue == NULL) return;
   gpr_ref(&workqueue->refs);
 }
+#endif
 
 #ifdef GRPC_WORKQUEUE_REFCOUNT_DEBUG
 void grpc_workqueue_unref(grpc_exec_ctx *exec_ctx, grpc_workqueue *workqueue,
                           const char *file, int line, const char *reason) {
+  if (workqueue == NULL) return;
   gpr_log(file, line, GPR_LOG_SEVERITY_DEBUG, "WORKQUEUE:%p unref %d -> %d %s",
           workqueue, (int)workqueue->refs.count, (int)workqueue->refs.count - 1,
           reason);
+  if (gpr_unref(&workqueue->refs)) {
+    workqueue_orphan(exec_ctx, workqueue);
+  }
+}
 #else
 void grpc_workqueue_unref(grpc_exec_ctx *exec_ctx, grpc_workqueue *workqueue) {
-#endif
+  if (workqueue == NULL) return;
   if (gpr_unref(&workqueue->refs)) {
-    workqueue_destroy(exec_ctx, workqueue);
+    workqueue_orphan(exec_ctx, workqueue);
   }
 }
+#endif
+
+static void drain(grpc_exec_ctx *exec_ctx, grpc_workqueue *workqueue) {
+  abort();
+}
 
-void grpc_workqueue_flush(grpc_exec_ctx *exec_ctx, grpc_workqueue *workqueue) {
-  gpr_mu_lock(&workqueue->mu);
-  grpc_exec_ctx_enqueue_list(exec_ctx, &workqueue->closure_list, NULL);
-  gpr_mu_unlock(&workqueue->mu);
+static void wakeup(grpc_exec_ctx *exec_ctx, grpc_workqueue *workqueue) {
+  GPR_TIMER_MARK("workqueue.wakeup", 0);
+  grpc_error *err = grpc_wakeup_fd_wakeup(&workqueue->wakeup_fd);
+  if (!GRPC_LOG_IF_ERROR("wakeupfd_wakeup", err)) {
+    drain(exec_ctx, workqueue);
+  }
 }
 
 static void on_readable(grpc_exec_ctx *exec_ctx, void *arg, grpc_error *error) {
+  GPR_TIMER_BEGIN("workqueue.on_readable", 0);
+
   grpc_workqueue *workqueue = arg;
 
   if (error != GRPC_ERROR_NONE) {
-    gpr_mu_destroy(&workqueue->mu);
     /* HACK: let wakeup_fd code know that we stole the fd */
     workqueue->wakeup_fd.read_fd = 0;
     grpc_wakeup_fd_destroy(&workqueue->wakeup_fd);
     grpc_fd_orphan(exec_ctx, workqueue->wakeup_read_fd, NULL, NULL, "destroy");
+    GPR_ASSERT(gpr_atm_no_barrier_load(&workqueue->state) == 0);
     gpr_free(workqueue);
   } else {
-    gpr_mu_lock(&workqueue->mu);
-    grpc_exec_ctx_enqueue_list(exec_ctx, &workqueue->closure_list, NULL);
     error = grpc_wakeup_fd_consume_wakeup(&workqueue->wakeup_fd);
-    gpr_mu_unlock(&workqueue->mu);
+    gpr_mpscq_node *n = gpr_mpscq_pop(&workqueue->queue);
     if (error == GRPC_ERROR_NONE) {
       grpc_fd_notify_on_read(exec_ctx, workqueue->wakeup_read_fd,
                              &workqueue->read_closure);
@@ -128,24 +151,46 @@ static void on_readable(grpc_exec_ctx *exec_ctx, void *arg, grpc_error *error) {
       /* recurse to get error handling */
       on_readable(exec_ctx, arg, error);
     }
+    if (n == NULL) {
+      /* try again - queue in an inconsistant state */
+      wakeup(exec_ctx, workqueue);
+    } else {
+      switch (gpr_atm_full_fetch_add(&workqueue->state, -2)) {
+        case 3:  // had one count, one unorphaned --> done, unorphaned
+          break;
+        case 2:  // had one count, one orphaned --> done, orphaned
+          workqueue_destroy(exec_ctx, workqueue);
+          break;
+        case 1:
+        case 0:
+          // these values are illegal - representing an already done or
+          // deleted workqueue
+          GPR_UNREACHABLE_CODE(break);
+        default:
+          // schedule a wakeup since there's more to do
+          wakeup(exec_ctx, workqueue);
+      }
+      grpc_closure *cl = (grpc_closure *)n;
+      grpc_error *clerr = cl->error;
+      cl->cb(exec_ctx, cl->cb_arg, clerr);
+      GRPC_ERROR_UNREF(clerr);
+    }
   }
+
+  GPR_TIMER_END("workqueue.on_readable", 0);
 }
 
 void grpc_workqueue_enqueue(grpc_exec_ctx *exec_ctx, grpc_workqueue *workqueue,
                             grpc_closure *closure, grpc_error *error) {
-  grpc_error *push_error = GRPC_ERROR_NONE;
-  gpr_mu_lock(&workqueue->mu);
-  if (grpc_closure_list_empty(workqueue->closure_list)) {
-    push_error = grpc_wakeup_fd_wakeup(&workqueue->wakeup_fd);
-  }
-  grpc_closure_list_append(&workqueue->closure_list, closure, error);
-  if (push_error != GRPC_ERROR_NONE) {
-    const char *msg = grpc_error_string(push_error);
-    gpr_log(GPR_ERROR, "Failed to push to workqueue: %s", msg);
-    grpc_error_free_string(msg);
-    grpc_exec_ctx_enqueue_list(exec_ctx, &workqueue->closure_list, NULL);
+  GPR_TIMER_BEGIN("workqueue.enqueue", 0);
+  gpr_atm last = gpr_atm_full_fetch_add(&workqueue->state, 2);
+  GPR_ASSERT(last & 1);
+  closure->error = error;
+  gpr_mpscq_push(&workqueue->queue, &closure->next_data.atm_next);
+  if (last == 1) {
+    wakeup(exec_ctx, workqueue);
   }
-  gpr_mu_unlock(&workqueue->mu);
+  GPR_TIMER_END("workqueue.enqueue", 0);
 }
 
 #endif /* GPR_POSIX_SOCKET */
diff --git a/src/core/lib/iomgr/workqueue_posix.h b/src/core/lib/iomgr/workqueue_posix.h
index 0f26ba58e2..03ee21cef7 100644
--- a/src/core/lib/iomgr/workqueue_posix.h
+++ b/src/core/lib/iomgr/workqueue_posix.h
@@ -35,14 +35,17 @@
 #define GRPC_CORE_LIB_IOMGR_WORKQUEUE_POSIX_H
 
 #include "src/core/lib/iomgr/wakeup_fd_posix.h"
+#include "src/core/lib/support/mpscq.h"
 
 struct grpc_fd;
 
 struct grpc_workqueue {
   gpr_refcount refs;
-
-  gpr_mu mu;
-  grpc_closure_list closure_list;
+  gpr_mpscq queue;
+  // state is:
+  // lower bit - zero if orphaned
+  // other bits - number of items enqueued
+  gpr_atm state;
 
   grpc_wakeup_fd wakeup_fd;
   struct grpc_fd *wakeup_read_fd;
diff --git a/src/core/lib/iomgr/workqueue_windows.c b/src/core/lib/iomgr/workqueue_windows.c
index 23e2dea185..ee81dc248e 100644
--- a/src/core/lib/iomgr/workqueue_windows.c
+++ b/src/core/lib/iomgr/workqueue_windows.c
@@ -42,8 +42,6 @@
 // context, which is at least correct, if not performant or in the spirit of
 // workqueues.
 
-void grpc_workqueue_flush(grpc_exec_ctx *exec_ctx, grpc_workqueue *workqueue) {}
-
 #ifdef GRPC_WORKQUEUE_REFCOUNT_DEBUG
 void grpc_workqueue_ref(grpc_workqueue *workqueue, const char *file, int line,
                         const char *reason) {}
diff --git a/src/core/lib/security/transport/client_auth_filter.c b/src/core/lib/security/transport/client_auth_filter.c
index 2a1bf4d4e3..b366d1410f 100644
--- a/src/core/lib/security/transport/client_auth_filter.c
+++ b/src/core/lib/security/transport/client_auth_filter.c
@@ -40,6 +40,7 @@
 #include <grpc/support/string_util.h>
 
 #include "src/core/lib/channel/channel_stack.h"
+#include "src/core/lib/profiling/timers.h"
 #include "src/core/lib/security/context/security_context.h"
 #include "src/core/lib/security/credentials/credentials.h"
 #include "src/core/lib/security/transport/security_connector.h"
@@ -218,6 +219,8 @@ static void on_host_checked(grpc_exec_ctx *exec_ctx, void *user_data,
 static void auth_start_transport_op(grpc_exec_ctx *exec_ctx,
                                     grpc_call_element *elem,
                                     grpc_transport_stream_op *op) {
+  GPR_TIMER_BEGIN("auth_start_transport_op", 0);
+
   /* grab pointers to our data from the call element */
   call_data *calld = elem->call_data;
   channel_data *chand = elem->channel_data;
@@ -258,12 +261,14 @@ static void auth_start_transport_op(grpc_exec_ctx *exec_ctx,
       grpc_channel_security_connector_check_call_host(
           exec_ctx, chand->security_connector, call_host, chand->auth_context,
           on_host_checked, elem);
+      GPR_TIMER_END("auth_start_transport_op", 0);
       return; /* early exit */
     }
   }
 
   /* pass control down the stack */
   grpc_call_next_op(exec_ctx, elem, op);
+  GPR_TIMER_END("auth_start_transport_op", 0);
 }
 
 /* Constructor for call_data */
diff --git a/src/core/lib/security/transport/secure_endpoint.c b/src/core/lib/security/transport/secure_endpoint.c
index 0169ccd9ef..acb0113ea8 100644
--- a/src/core/lib/security/transport/secure_endpoint.c
+++ b/src/core/lib/security/transport/secure_endpoint.c
@@ -38,6 +38,7 @@
 #include <grpc/support/slice_buffer.h>
 #include <grpc/support/sync.h>
 #include "src/core/lib/debug/trace.h"
+#include "src/core/lib/profiling/timers.h"
 #include "src/core/lib/security/transport/tsi_error.h"
 #include "src/core/lib/support/string.h"
 #include "src/core/lib/tsi/transport_security_interface.h"
@@ -248,6 +249,8 @@ static void flush_write_staging_buffer(secure_endpoint *ep, uint8_t **cur,
 
 static void endpoint_write(grpc_exec_ctx *exec_ctx, grpc_endpoint *secure_ep,
                            gpr_slice_buffer *slices, grpc_closure *cb) {
+  GPR_TIMER_BEGIN("secure_endpoint.endpoint_write", 0);
+
   unsigned i;
   tsi_result result = TSI_OK;
   secure_endpoint *ep = (secure_endpoint *)secure_ep;
@@ -323,10 +326,12 @@ static void endpoint_write(grpc_exec_ctx *exec_ctx, grpc_endpoint *secure_ep,
         exec_ctx, cb,
         grpc_set_tsi_error_result(GRPC_ERROR_CREATE("Wrap failed"), result),
         NULL);
+    GPR_TIMER_END("secure_endpoint.endpoint_write", 0);
     return;
   }
 
   grpc_endpoint_write(exec_ctx, ep->wrapped_ep, &ep->output_buffer, cb);
+  GPR_TIMER_END("secure_endpoint.endpoint_write", 0);
 }
 
 static void endpoint_shutdown(grpc_exec_ctx *exec_ctx,
diff --git a/src/core/lib/security/transport/server_auth_filter.c b/src/core/lib/security/transport/server_auth_filter.c
index def16c8229..b2c6815af8 100644
--- a/src/core/lib/security/transport/server_auth_filter.c
+++ b/src/core/lib/security/transport/server_auth_filter.c
@@ -48,7 +48,7 @@ typedef struct call_data {
      up-call on transport_op, and remember to call our on_done_recv member after
      handling it. */
   grpc_closure auth_on_recv;
-  grpc_transport_stream_op transport_op;
+  grpc_transport_stream_op *transport_op;
   grpc_metadata_array md;
   const grpc_metadata *consumed_md;
   size_t num_consumed_md;
@@ -106,6 +106,10 @@ static grpc_mdelem *remove_consumed_md(void *user_data, grpc_mdelem *md) {
   return md;
 }
 
+static void destroy_op(grpc_exec_ctx *exec_ctx, void *arg, grpc_error *error) {
+  gpr_free(arg);
+}
+
 /* called from application code */
 static void on_md_processing_done(
     void *user_data, const grpc_metadata *consumed_md, size_t num_consumed_md,
@@ -131,21 +135,22 @@ static void on_md_processing_done(
     grpc_exec_ctx_sched(&exec_ctx, calld->on_done_recv, GRPC_ERROR_NONE, NULL);
   } else {
     gpr_slice message;
-    grpc_transport_stream_op close_op;
-    memset(&close_op, 0, sizeof(close_op));
+    grpc_transport_stream_op *close_op = gpr_malloc(sizeof(*close_op));
+    memset(close_op, 0, sizeof(*close_op));
     grpc_metadata_array_destroy(&calld->md);
     error_details = error_details != NULL
                         ? error_details
                         : "Authentication metadata processing failed.";
     message = gpr_slice_from_copied_string(error_details);
-    calld->transport_op.send_initial_metadata = NULL;
-    if (calld->transport_op.send_message != NULL) {
-      grpc_byte_stream_destroy(&exec_ctx, calld->transport_op.send_message);
-      calld->transport_op.send_message = NULL;
+    calld->transport_op->send_initial_metadata = NULL;
+    if (calld->transport_op->send_message != NULL) {
+      grpc_byte_stream_destroy(&exec_ctx, calld->transport_op->send_message);
+      calld->transport_op->send_message = NULL;
     }
-    calld->transport_op.send_trailing_metadata = NULL;
-    grpc_transport_stream_op_add_close(&close_op, status, &message);
-    grpc_call_next_op(&exec_ctx, elem, &close_op);
+    calld->transport_op->send_trailing_metadata = NULL;
+    close_op->on_complete = grpc_closure_create(destroy_op, close_op);
+    grpc_transport_stream_op_add_close(close_op, status, &message);
+    grpc_call_next_op(&exec_ctx, elem, close_op);
     grpc_exec_ctx_sched(&exec_ctx, calld->on_done_recv,
                         grpc_error_set_int(GRPC_ERROR_CREATE(error_details),
                                            GRPC_ERROR_INT_GRPC_STATUS, status),
@@ -182,7 +187,7 @@ static void set_recv_ops_md_callbacks(grpc_call_element *elem,
     calld->recv_initial_metadata = op->recv_initial_metadata;
     calld->on_done_recv = op->recv_initial_metadata_ready;
     op->recv_initial_metadata_ready = &calld->auth_on_recv;
-    calld->transport_op = *op;
+    calld->transport_op = op;
   }
 }
 
diff --git a/src/core/lib/support/mpscq.c b/src/core/lib/support/mpscq.c
new file mode 100644
index 0000000000..5b9323275a
--- /dev/null
+++ b/src/core/lib/support/mpscq.c
@@ -0,0 +1,83 @@
+/*
+ *
+ * Copyright 2016, Google Inc.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *     * Neither the name of Google Inc. nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#include "src/core/lib/support/mpscq.h"
+
+#include <grpc/support/log.h>
+
+void gpr_mpscq_init(gpr_mpscq *q) {
+  gpr_atm_no_barrier_store(&q->head, (gpr_atm)&q->stub);
+  q->tail = &q->stub;
+  gpr_atm_no_barrier_store(&q->stub.next, (gpr_atm)NULL);
+}
+
+void gpr_mpscq_destroy(gpr_mpscq *q) {
+  GPR_ASSERT(gpr_atm_no_barrier_load(&q->head) == (gpr_atm)&q->stub);
+  GPR_ASSERT(q->tail == &q->stub);
+}
+
+void gpr_mpscq_push(gpr_mpscq *q, gpr_mpscq_node *n) {
+  gpr_atm_no_barrier_store(&n->next, (gpr_atm)NULL);
+  gpr_mpscq_node *prev =
+      (gpr_mpscq_node *)gpr_atm_full_xchg(&q->head, (gpr_atm)n);
+  gpr_atm_rel_store(&prev->next, (gpr_atm)n);
+}
+
+gpr_mpscq_node *gpr_mpscq_pop(gpr_mpscq *q) {
+  gpr_mpscq_node *tail = q->tail;
+  gpr_mpscq_node *next = (gpr_mpscq_node *)gpr_atm_acq_load(&tail->next);
+  if (tail == &q->stub) {
+    // indicates the list is actually (ephemerally) empty
+    if (next == NULL) return NULL;
+    q->tail = next;
+    tail = next;
+    next = (gpr_mpscq_node *)gpr_atm_acq_load(&tail->next);
+  }
+  if (next != NULL) {
+    q->tail = next;
+    return tail;
+  }
+  gpr_mpscq_node *head = (gpr_mpscq_node *)gpr_atm_acq_load(&q->head);
+  if (tail != head) {
+    // indicates a retry is in order: we're still adding
+    return NULL;
+  }
+  gpr_mpscq_push(q, &q->stub);
+  next = (gpr_mpscq_node *)gpr_atm_acq_load(&tail->next);
+  if (next != NULL) {
+    q->tail = next;
+    return tail;
+  }
+  // indicates a retry is in order: we're still adding
+  return NULL;
+}
diff --git a/src/core/lib/support/mpscq.h b/src/core/lib/support/mpscq.h
new file mode 100644
index 0000000000..977a117952
--- /dev/null
+++ b/src/core/lib/support/mpscq.h
@@ -0,0 +1,65 @@
+/*
+ *
+ * Copyright 2016, Google Inc.
+ * All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are
+ * met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ * notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above
+ * copyright notice, this list of conditions and the following disclaimer
+ * in the documentation and/or other materials provided with the
+ * distribution.
+ *     * Neither the name of Google Inc. nor the names of its
+ * contributors may be used to endorse or promote products derived from
+ * this software without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ *
+ */
+
+#ifndef GRPC_CORE_LIB_SUPPORT_MPSCQ_H
+#define GRPC_CORE_LIB_SUPPORT_MPSCQ_H
+
+#include <grpc/support/atm.h>
+#include <stddef.h>
+
+// Multiple-producer single-consumer lock free queue, based upon the
+// implementation from Dmitry Vyukov here:
+// http://www.1024cores.net/home/lock-free-algorithms/queues/intrusive-mpsc-node-based-queue
+
+// List node (include this in a data structure at the top, and add application
+// fields after it - to simulate inheritance)
+typedef struct gpr_mpscq_node { gpr_atm next; } gpr_mpscq_node;
+
+// Actual queue type
+typedef struct gpr_mpscq {
+  gpr_atm head;
+  // make sure head & tail don't share a cacheline
+  char padding[GPR_CACHELINE_SIZE];
+  gpr_mpscq_node *tail;
+  gpr_mpscq_node stub;
+} gpr_mpscq;
+
+void gpr_mpscq_init(gpr_mpscq *q);
+void gpr_mpscq_destroy(gpr_mpscq *q);
+// Push a node
+void gpr_mpscq_push(gpr_mpscq *q, gpr_mpscq_node *n);
+// Pop a node (returns NULL if no node is ready - which doesn't indicate that
+// the queue is empty!!)
+gpr_mpscq_node *gpr_mpscq_pop(gpr_mpscq *q);
+
+#endif /* GRPC_CORE_LIB_SUPPORT_MPSCQ_H */
diff --git a/src/core/lib/surface/call.c b/src/core/lib/surface/call.c
index 772681109a..119f5e82ab 100644
--- a/src/core/lib/surface/call.c
+++ b/src/core/lib/surface/call.c
@@ -109,6 +109,10 @@ typedef struct batch_control {
   uint8_t recv_message;
   uint8_t recv_final_op;
   uint8_t is_notify_tag_closure;
+
+  /* TODO(ctiller): now that this is inlined, figure out how much of the above
+                    state can be eliminated */
+  grpc_transport_stream_op op;
 } batch_control;
 
 struct grpc_call {
@@ -778,6 +782,7 @@ typedef struct termination_closure {
   grpc_error *error;
   grpc_closure *op_closure;
   enum { TC_CANCEL, TC_CLOSE } type;
+  grpc_transport_stream_op op;
 } termination_closure;
 
 static void done_termination(grpc_exec_ctx *exec_ctx, void *tcp,
@@ -797,26 +802,24 @@ static void done_termination(grpc_exec_ctx *exec_ctx, void *tcp,
 }
 
 static void send_cancel(grpc_exec_ctx *exec_ctx, void *tcp, grpc_error *error) {
-  grpc_transport_stream_op op;
   termination_closure *tc = tcp;
-  memset(&op, 0, sizeof(op));
-  op.cancel_error = tc->error;
+  memset(&tc->op, 0, sizeof(tc->op));
+  tc->op.cancel_error = tc->error;
   /* reuse closure to catch completion */
   grpc_closure_init(&tc->closure, done_termination, tc);
-  op.on_complete = &tc->closure;
-  execute_op(exec_ctx, tc->call, &op);
+  tc->op.on_complete = &tc->closure;
+  execute_op(exec_ctx, tc->call, &tc->op);
 }
 
 static void send_close(grpc_exec_ctx *exec_ctx, void *tcp, grpc_error *error) {
-  grpc_transport_stream_op op;
   termination_closure *tc = tcp;
-  memset(&op, 0, sizeof(op));
-  op.close_error = tc->error;
+  memset(&tc->op, 0, sizeof(tc->op));
+  tc->op.close_error = tc->error;
   /* reuse closure to catch completion */
   grpc_closure_init(&tc->closure, done_termination, tc);
-  tc->op_closure = op.on_complete;
-  op.on_complete = &tc->closure;
-  execute_op(exec_ctx, tc->call, &op);
+  tc->op_closure = tc->op.on_complete;
+  tc->op.on_complete = &tc->closure;
+  execute_op(exec_ctx, tc->call, &tc->op);
 }
 
 static grpc_call_error terminate_with_status(grpc_exec_ctx *exec_ctx,
@@ -1370,7 +1373,6 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
                                         grpc_call *call, const grpc_op *ops,
                                         size_t nops, void *notify_tag,
                                         int is_notify_tag_closure) {
-  grpc_transport_stream_op stream_op;
   size_t i;
   const grpc_op *op;
   batch_control *bctl;
@@ -1384,8 +1386,6 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
 
   GRPC_CALL_LOG_BATCH(GPR_INFO, call, ops, nops, notify_tag);
 
-  memset(&stream_op, 0, sizeof(stream_op));
-
   /* TODO(ctiller): this feels like it could be made lock-free */
   gpr_mu_lock(&call->mu);
   bctl = allocate_batch_control(call);
@@ -1394,6 +1394,9 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
   bctl->notify_tag = notify_tag;
   bctl->is_notify_tag_closure = (uint8_t)(is_notify_tag_closure != 0);
 
+  grpc_transport_stream_op *stream_op = &bctl->op;
+  memset(stream_op, 0, sizeof(*stream_op));
+
   if (nops == 0) {
     GRPC_CALL_INTERNAL_REF(call, "completion");
     bctl->error = GRPC_ERROR_NONE;
@@ -1471,9 +1474,9 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
         }
         /* TODO(ctiller): just make these the same variable? */
         call->metadata_batch[0][0].deadline = call->send_deadline;
-        stream_op.send_initial_metadata =
+        stream_op->send_initial_metadata =
             &call->metadata_batch[0 /* is_receiving */][0 /* is_trailing */];
-        stream_op.send_initial_metadata_flags = op->flags;
+        stream_op->send_initial_metadata_flags = op->flags;
         break;
       case GRPC_OP_SEND_MESSAGE:
         if (!are_write_flags_valid(op->flags)) {
@@ -1493,7 +1496,7 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
         grpc_slice_buffer_stream_init(
             &call->sending_stream,
             &op->data.send_message->data.raw.slice_buffer, op->flags);
-        stream_op.send_message = &call->sending_stream.base;
+        stream_op->send_message = &call->sending_stream.base;
         break;
       case GRPC_OP_SEND_CLOSE_FROM_CLIENT:
         /* Flag validation: currently allow no flags */
@@ -1511,7 +1514,7 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
         }
         bctl->send_final_op = 1;
         call->sent_final_op = 1;
-        stream_op.send_trailing_metadata =
+        stream_op->send_trailing_metadata =
             &call->metadata_batch[0 /* is_receiving */][1 /* is_trailing */];
         break;
       case GRPC_OP_SEND_STATUS_FROM_SERVER:
@@ -1558,7 +1561,7 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
           error = GRPC_CALL_ERROR_INVALID_METADATA;
           goto done_with_error;
         }
-        stream_op.send_trailing_metadata =
+        stream_op->send_trailing_metadata =
             &call->metadata_batch[0 /* is_receiving */][1 /* is_trailing */];
         break;
       case GRPC_OP_RECV_INITIAL_METADATA:
@@ -1576,9 +1579,9 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
         grpc_closure_init(&call->receiving_initial_metadata_ready,
                           receiving_initial_metadata_ready, bctl);
         bctl->recv_initial_metadata = 1;
-        stream_op.recv_initial_metadata =
+        stream_op->recv_initial_metadata =
             &call->metadata_batch[1 /* is_receiving */][0 /* is_trailing */];
-        stream_op.recv_initial_metadata_ready =
+        stream_op->recv_initial_metadata_ready =
             &call->receiving_initial_metadata_ready;
         num_completion_callbacks_needed++;
         break;
@@ -1595,10 +1598,10 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
         call->receiving_message = 1;
         bctl->recv_message = 1;
         call->receiving_buffer = op->data.recv_message;
-        stream_op.recv_message = &call->receiving_stream;
+        stream_op->recv_message = &call->receiving_stream;
         grpc_closure_init(&call->receiving_stream_ready, receiving_stream_ready,
                           bctl);
-        stream_op.recv_message_ready = &call->receiving_stream_ready;
+        stream_op->recv_message_ready = &call->receiving_stream_ready;
         num_completion_callbacks_needed++;
         break;
       case GRPC_OP_RECV_STATUS_ON_CLIENT:
@@ -1624,9 +1627,9 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
         call->final_op.client.status_details_capacity =
             op->data.recv_status_on_client.status_details_capacity;
         bctl->recv_final_op = 1;
-        stream_op.recv_trailing_metadata =
+        stream_op->recv_trailing_metadata =
             &call->metadata_batch[1 /* is_receiving */][1 /* is_trailing */];
-        stream_op.collect_stats =
+        stream_op->collect_stats =
             &call->final_info.stats.transport_stream_stats;
         break;
       case GRPC_OP_RECV_CLOSE_ON_SERVER:
@@ -1647,9 +1650,9 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
         call->final_op.server.cancelled =
             op->data.recv_close_on_server.cancelled;
         bctl->recv_final_op = 1;
-        stream_op.recv_trailing_metadata =
+        stream_op->recv_trailing_metadata =
             &call->metadata_batch[1 /* is_receiving */][1 /* is_trailing */];
-        stream_op.collect_stats =
+        stream_op->collect_stats =
             &call->final_info.stats.transport_stream_stats;
         break;
     }
@@ -1661,12 +1664,12 @@ static grpc_call_error call_start_batch(grpc_exec_ctx *exec_ctx,
   }
   gpr_ref_init(&bctl->steps_to_complete, num_completion_callbacks_needed);
 
-  stream_op.context = call->context;
+  stream_op->context = call->context;
   grpc_closure_init(&bctl->finish_batch, finish_batch, bctl);
-  stream_op.on_complete = &bctl->finish_batch;
+  stream_op->on_complete = &bctl->finish_batch;
   gpr_mu_unlock(&call->mu);
 
-  execute_op(exec_ctx, call, &stream_op);
+  execute_op(exec_ctx, call, stream_op);
 
 done:
   GPR_TIMER_END("grpc_call_start_batch", 0);
diff --git a/src/core/lib/surface/channel.c b/src/core/lib/surface/channel.c
index 6d2b1c4935..52e78567bd 100644
--- a/src/core/lib/surface/channel.c
+++ b/src/core/lib/surface/channel.c
@@ -334,14 +334,13 @@ static void destroy_channel(grpc_exec_ctx *exec_ctx, void *arg,
 }
 
 void grpc_channel_destroy(grpc_channel *channel) {
-  grpc_transport_op op;
+  grpc_transport_op *op = grpc_make_transport_op(NULL);
   grpc_channel_element *elem;
   grpc_exec_ctx exec_ctx = GRPC_EXEC_CTX_INIT;
   GRPC_API_TRACE("grpc_channel_destroy(channel=%p)", 1, (channel));
-  memset(&op, 0, sizeof(op));
-  op.disconnect_with_error = GRPC_ERROR_CREATE("Channel Destroyed");
+  op->disconnect_with_error = GRPC_ERROR_CREATE("Channel Destroyed");
   elem = grpc_channel_stack_element(CHANNEL_STACK_FROM_CHANNEL(channel), 0);
-  elem->filter->start_transport_op(&exec_ctx, elem, &op);
+  elem->filter->start_transport_op(&exec_ctx, elem, op);
 
   GRPC_CHANNEL_INTERNAL_UNREF(&exec_ctx, channel, "channel");
 
diff --git a/src/core/lib/surface/channel_ping.c b/src/core/lib/surface/channel_ping.c
index 9818f9d2f2..0d2f01a649 100644
--- a/src/core/lib/surface/channel_ping.c
+++ b/src/core/lib/surface/channel_ping.c
@@ -61,19 +61,20 @@ static void ping_done(grpc_exec_ctx *exec_ctx, void *arg, grpc_error *error) {
 
 void grpc_channel_ping(grpc_channel *channel, grpc_completion_queue *cq,
                        void *tag, void *reserved) {
-  grpc_transport_op op;
+  GRPC_API_TRACE("grpc_channel_ping(channel=%p, cq=%p, tag=%p, reserved=%p)", 4,
+                 (channel, cq, tag, reserved));
+  grpc_transport_op *op = grpc_make_transport_op(NULL);
   ping_result *pr = gpr_malloc(sizeof(*pr));
   grpc_channel_element *top_elem =
       grpc_channel_stack_element(grpc_channel_get_channel_stack(channel), 0);
   grpc_exec_ctx exec_ctx = GRPC_EXEC_CTX_INIT;
   GPR_ASSERT(reserved == NULL);
-  memset(&op, 0, sizeof(op));
   pr->tag = tag;
   pr->cq = cq;
   grpc_closure_init(&pr->closure, ping_done, pr);
-  op.send_ping = &pr->closure;
-  op.bind_pollset = grpc_cq_pollset(cq);
+  op->send_ping = &pr->closure;
+  op->bind_pollset = grpc_cq_pollset(cq);
   grpc_cq_begin_op(cq, tag);
-  top_elem->filter->start_transport_op(&exec_ctx, top_elem, &op);
+  top_elem->filter->start_transport_op(&exec_ctx, top_elem, op);
   grpc_exec_ctx_finish(&exec_ctx);
 }
diff --git a/src/core/lib/surface/init.c b/src/core/lib/surface/init.c
index 5397913a21..edda0c85fa 100644
--- a/src/core/lib/surface/init.c
+++ b/src/core/lib/surface/init.c
@@ -47,6 +47,7 @@
 #include "src/core/lib/channel/http_server_filter.h"
 #include "src/core/lib/debug/trace.h"
 #include "src/core/lib/http/parser.h"
+#include "src/core/lib/iomgr/combiner.h"
 #include "src/core/lib/iomgr/executor.h"
 #include "src/core/lib/iomgr/iomgr.h"
 #include "src/core/lib/profiling/timers.h"
@@ -165,6 +166,7 @@ void grpc_init(void) {
     grpc_register_tracer("http1", &grpc_http1_trace);
     grpc_register_tracer("compression", &grpc_compression_trace);
     grpc_register_tracer("queue_pluck", &grpc_cq_pluck_trace);
+    grpc_register_tracer("combiner", &grpc_combiner_trace);
     // Default pluck trace to 1
     grpc_cq_pluck_trace = 1;
     grpc_register_tracer("queue_timeout", &grpc_cq_event_timeout_trace);
diff --git a/src/core/lib/surface/lame_client.c b/src/core/lib/surface/lame_client.c
index 19b78369dd..d32c884e8e 100644
--- a/src/core/lib/surface/lame_client.c
+++ b/src/core/lib/surface/lame_client.c
@@ -97,14 +97,14 @@ static void lame_start_transport_op(grpc_exec_ctx *exec_ctx,
     grpc_exec_ctx_sched(exec_ctx, op->on_connectivity_state_change,
                         GRPC_ERROR_NONE, NULL);
   }
-  if (op->on_consumed != NULL) {
-    grpc_exec_ctx_sched(exec_ctx, op->on_consumed, GRPC_ERROR_NONE, NULL);
-  }
   if (op->send_ping != NULL) {
     grpc_exec_ctx_sched(exec_ctx, op->send_ping,
                         GRPC_ERROR_CREATE("lame client channel"), NULL);
   }
   GRPC_ERROR_UNREF(op->disconnect_with_error);
+  if (op->on_consumed != NULL) {
+    grpc_exec_ctx_sched(exec_ctx, op->on_consumed, GRPC_ERROR_NONE, NULL);
+  }
 }
 
 static grpc_error *init_call_elem(grpc_exec_ctx *exec_ctx,
diff --git a/src/core/lib/surface/server.c b/src/core/lib/surface/server.c
index 55e6d99057..56fb80e92e 100644
--- a/src/core/lib/surface/server.c
+++ b/src/core/lib/surface/server.c
@@ -273,23 +273,20 @@ static void shutdown_cleanup(grpc_exec_ctx *exec_ctx, void *arg,
 }
 
 static void send_shutdown(grpc_exec_ctx *exec_ctx, grpc_channel *channel,
-                          bool send_goaway, grpc_error *send_disconnect) {
-  grpc_transport_op op;
-  struct shutdown_cleanup_args *sc;
+                          int send_goaway, grpc_error *send_disconnect) {
+  struct shutdown_cleanup_args *sc = gpr_malloc(sizeof(*sc));
+  grpc_closure_init(&sc->closure, shutdown_cleanup, sc);
+  grpc_transport_op *op = grpc_make_transport_op(&sc->closure);
   grpc_channel_element *elem;
 
-  memset(&op, 0, sizeof(op));
-  op.send_goaway = send_goaway;
-  sc = gpr_malloc(sizeof(*sc));
+  op->send_goaway = send_goaway;
   sc->slice = gpr_slice_from_copied_string("Server shutdown");
-  op.goaway_message = &sc->slice;
-  op.goaway_status = GRPC_STATUS_OK;
-  op.disconnect_with_error = send_disconnect;
-  grpc_closure_init(&sc->closure, shutdown_cleanup, sc);
-  op.on_consumed = &sc->closure;
+  op->goaway_message = &sc->slice;
+  op->goaway_status = GRPC_STATUS_OK;
+  op->disconnect_with_error = send_disconnect;
 
   elem = grpc_channel_stack_element(grpc_channel_get_channel_stack(channel), 0);
-  elem->filter->start_transport_op(exec_ctx, elem, &op);
+  elem->filter->start_transport_op(exec_ctx, elem, op);
 }
 
 static void channel_broadcaster_shutdown(grpc_exec_ctx *exec_ctx,
@@ -432,7 +429,8 @@ static void finish_destroy_channel(grpc_exec_ctx *exec_ctx, void *cd,
   server_unref(exec_ctx, server);
 }
 
-static void destroy_channel(grpc_exec_ctx *exec_ctx, channel_data *chand) {
+static void destroy_channel(grpc_exec_ctx *exec_ctx, channel_data *chand,
+                            grpc_error *error) {
   if (is_channel_orphaned(chand)) return;
   GPR_ASSERT(chand->server != NULL);
   orphan_channel(chand);
@@ -441,14 +439,20 @@ static void destroy_channel(grpc_exec_ctx *exec_ctx, channel_data *chand) {
   chand->finish_destroy_channel_closure.cb = finish_destroy_channel;
   chand->finish_destroy_channel_closure.cb_arg = chand;
 
-  grpc_transport_op op;
-  memset(&op, 0, sizeof(op));
-  op.set_accept_stream = true;
-  op.on_consumed = &chand->finish_destroy_channel_closure;
+  grpc_transport_op *op =
+      grpc_make_transport_op(&chand->finish_destroy_channel_closure);
+  op->set_accept_stream = true;
   grpc_channel_next_op(exec_ctx,
                        grpc_channel_stack_element(
                            grpc_channel_get_channel_stack(chand->channel), 0),
-                       &op);
+                       op);
+
+  if (error != GRPC_ERROR_NONE) {
+    const char *msg = grpc_error_string(error);
+    gpr_log(GPR_INFO, "Disconnected client: %s", msg);
+    grpc_error_free_string(msg);
+  }
+  GRPC_ERROR_UNREF(error);
 }
 
 static void cpstr(char **dest, size_t *capacity, grpc_mdstr *value) {
@@ -845,17 +849,16 @@ static void channel_connectivity_changed(grpc_exec_ctx *exec_ctx, void *cd,
   channel_data *chand = cd;
   grpc_server *server = chand->server;
   if (chand->connectivity_state != GRPC_CHANNEL_SHUTDOWN) {
-    grpc_transport_op op;
-    memset(&op, 0, sizeof(op));
-    op.on_connectivity_state_change = &chand->channel_connectivity_changed,
-    op.connectivity_state = &chand->connectivity_state;
+    grpc_transport_op *op = grpc_make_transport_op(NULL);
+    op->on_connectivity_state_change = &chand->channel_connectivity_changed,
+    op->connectivity_state = &chand->connectivity_state;
     grpc_channel_next_op(exec_ctx,
                          grpc_channel_stack_element(
                              grpc_channel_get_channel_stack(chand->channel), 0),
-                         &op);
+                         op);
   } else {
     gpr_mu_lock(&server->mu_global);
-    destroy_channel(exec_ctx, chand);
+    destroy_channel(exec_ctx, chand, GRPC_ERROR_REF(error));
     gpr_mu_unlock(&server->mu_global);
     GRPC_CHANNEL_INTERNAL_UNREF(exec_ctx, chand->channel, "connectivity");
   }
@@ -1119,7 +1122,7 @@ void grpc_server_setup_transport(grpc_exec_ctx *exec_ctx, grpc_server *s,
   size_t slots;
   uint32_t probes;
   uint32_t max_probes = 0;
-  grpc_transport_op op;
+  grpc_transport_op *op = NULL;
 
   channel =
       grpc_channel_create(exec_ctx, NULL, args, GRPC_SERVER_CHANNEL, transport);
@@ -1179,16 +1182,16 @@ void grpc_server_setup_transport(grpc_exec_ctx *exec_ctx, grpc_server *s,
   gpr_mu_unlock(&s->mu_global);
 
   GRPC_CHANNEL_INTERNAL_REF(channel, "connectivity");
-  memset(&op, 0, sizeof(op));
-  op.set_accept_stream = true;
-  op.set_accept_stream_fn = accept_stream;
-  op.set_accept_stream_user_data = chand;
-  op.on_connectivity_state_change = &chand->channel_connectivity_changed;
-  op.connectivity_state = &chand->connectivity_state;
+  op = grpc_make_transport_op(NULL);
+  op->set_accept_stream = true;
+  op->set_accept_stream_fn = accept_stream;
+  op->set_accept_stream_user_data = chand;
+  op->on_connectivity_state_change = &chand->channel_connectivity_changed;
+  op->connectivity_state = &chand->connectivity_state;
   if (gpr_atm_acq_load(&s->shutdown_flag) != 0) {
-    op.disconnect_with_error = GRPC_ERROR_CREATE("Server shutdown");
+    op->disconnect_with_error = GRPC_ERROR_CREATE("Server shutdown");
   }
-  grpc_transport_perform_op(exec_ctx, transport, &op);
+  grpc_transport_perform_op(exec_ctx, transport, op);
 }
 
 void done_published_shutdown(grpc_exec_ctx *exec_ctx, void *done_arg,
diff --git a/src/core/lib/transport/transport.c b/src/core/lib/transport/transport.c
index 857c3909d2..08f9d7e8d9 100644
--- a/src/core/lib/transport/transport.c
+++ b/src/core/lib/transport/transport.c
@@ -32,10 +32,14 @@
  */
 
 #include "src/core/lib/transport/transport.h"
+
+#include <string.h>
+
 #include <grpc/support/alloc.h>
 #include <grpc/support/atm.h>
 #include <grpc/support/log.h>
 #include <grpc/support/sync.h>
+
 #include "src/core/lib/support/string.h"
 #include "src/core/lib/transport/transport_impl.h"
 
@@ -247,3 +251,26 @@ void grpc_transport_stream_op_add_close(grpc_transport_stream_op *op,
   error = grpc_error_set_int(error, GRPC_ERROR_INT_GRPC_STATUS, status);
   add_error(op, &op->close_error, error);
 }
+
+typedef struct {
+  grpc_closure outer_on_complete;
+  grpc_closure *inner_on_complete;
+  grpc_transport_op op;
+} made_transport_op;
+
+static void destroy_made_transport_op(grpc_exec_ctx *exec_ctx, void *arg,
+                                      grpc_error *error) {
+  made_transport_op *op = arg;
+  grpc_exec_ctx_sched(exec_ctx, op->inner_on_complete, GRPC_ERROR_REF(error),
+                      NULL);
+  gpr_free(op);
+}
+
+grpc_transport_op *grpc_make_transport_op(grpc_closure *on_complete) {
+  made_transport_op *op = gpr_malloc(sizeof(*op));
+  grpc_closure_init(&op->outer_on_complete, destroy_made_transport_op, op);
+  op->inner_on_complete = on_complete;
+  memset(&op->op, 0, sizeof(op->op));
+  op->op.on_consumed = &op->outer_on_complete;
+  return &op->op;
+}
diff --git a/src/core/lib/transport/transport.h b/src/core/lib/transport/transport.h
index 26ed6cb839..8dc393fd61 100644
--- a/src/core/lib/transport/transport.h
+++ b/src/core/lib/transport/transport.h
@@ -100,6 +100,11 @@ void grpc_transport_move_one_way_stats(grpc_transport_one_way_stats *from,
 void grpc_transport_move_stats(grpc_transport_stream_stats *from,
                                grpc_transport_stream_stats *to);
 
+typedef struct {
+  grpc_closure closure;
+  void *args[2];
+} grpc_transport_private_op_data;
+
 /* Transport stream op: a set of operations to perform on a transport
    against a single stream */
 typedef struct grpc_transport_stream_op {
@@ -149,6 +154,12 @@ typedef struct grpc_transport_stream_op {
 
   /* Indexes correspond to grpc_context_index enum values */
   grpc_call_context_element *context;
+
+  /***************************************************************************
+   * remaining fields are initialized and used at the discretion of the
+   * transport implementation */
+
+  grpc_transport_private_op_data transport_private;
 } grpc_transport_stream_op;
 
 /** Transport op: a set of operations to perform on a transport as a whole */
@@ -182,6 +193,12 @@ typedef struct grpc_transport_op {
   grpc_pollset_set *bind_pollset_set;
   /** send a ping, call this back if not NULL */
   grpc_closure *send_ping;
+
+  /***************************************************************************
+   * remaining fields are initialized and used at the discretion of the
+   * transport implementation */
+
+  grpc_transport_private_op_data transport_private;
 } grpc_transport_op;
 
 /* Returns the amount of memory required to store a grpc_stream for this
@@ -273,6 +290,10 @@ void grpc_transport_destroy(grpc_exec_ctx *exec_ctx, grpc_transport *transport);
 char *grpc_transport_get_peer(grpc_exec_ctx *exec_ctx,
                               grpc_transport *transport);
 
+/* Allocate a grpc_transport_op, and preconfigure the on_consumed closure to
+   \a on_consumed and then delete the returned transport op */
+grpc_transport_op *grpc_make_transport_op(grpc_closure *on_consumed);
+
 #ifdef __cplusplus
 }
 #endif