grpc 1.68.1 → 1.74.1

data/src/core/handshaker/security/secure_endpoint.cc

@@ -30,19 +30,25 @@
 
 #include <algorithm>
 #include <atomic>
+#include <cstddef>
 #include <memory>
+#include <optional>
+#include <regex>
 #include <utility>
+#include <vector>
 
 #include "absl/base/thread_annotations.h"
 #include "absl/log/check.h"
 #include "absl/log/log.h"
 #include "absl/status/status.h"
 #include "absl/strings/string_view.h"
-#include "absl/types/optional.h"
+#include "absl/types/span.h"
 #include "src/core/lib/debug/trace.h"
+#include "src/core/lib/experiments/experiments.h"
 #include "src/core/lib/iomgr/closure.h"
 #include "src/core/lib/iomgr/endpoint.h"
 #include "src/core/lib/iomgr/error.h"
+#include "src/core/lib/iomgr/event_engine_shims/endpoint.h"
 #include "src/core/lib/iomgr/exec_ctx.h"
 #include "src/core/lib/iomgr/iomgr_fwd.h"
 #include "src/core/lib/resource_quota/api.h"
@@ -63,206 +69,137 @@
 static void on_read(void* user_data, grpc_error_handle error);
 static void on_write(void* user_data, grpc_error_handle error);
 
+namespace grpc_core {
 namespace {
-struct secure_endpoint : public grpc_endpoint {
-  secure_endpoint(const grpc_endpoint_vtable* vtbl,
-                  tsi_frame_protector* protector,
-                  tsi_zero_copy_grpc_protector* zero_copy_protector,
-                  grpc_core::OrphanablePtr<grpc_endpoint> endpoint,
-                  grpc_slice* leftover_slices,
-                  const grpc_channel_args* channel_args,
-                  size_t leftover_nslices)
-      : wrapped_ep(std::move(endpoint)),
-        protector(protector),
-        zero_copy_protector(zero_copy_protector) {
-    this->vtable = vtbl;
-    gpr_mu_init(&protector_mu);
-    GRPC_CLOSURE_INIT(&on_read, ::on_read, this, grpc_schedule_on_exec_ctx);
-    GRPC_CLOSURE_INIT(&on_write, ::on_write, this, grpc_schedule_on_exec_ctx);
-    grpc_slice_buffer_init(&source_buffer);
-    grpc_slice_buffer_init(&leftover_bytes);
-    for (size_t i = 0; i < leftover_nslices; i++) {
-      grpc_slice_buffer_add(&leftover_bytes,
-                            grpc_core::CSliceRef(leftover_slices[i]));
+class FrameProtector : public RefCounted<FrameProtector> {
+ public:
+  FrameProtector(tsi_frame_protector* protector,
+                 tsi_zero_copy_grpc_protector* zero_copy_protector,
+                 grpc_slice* leftover_slices, size_t leftover_nslices,
+                 const ChannelArgs& args)
+      : protector_(protector),
+        zero_copy_protector_(zero_copy_protector),
+        memory_owner_(args.GetObject<ResourceQuota>()
+                          ->memory_quota()
+                          ->CreateMemoryOwner()),
+        self_reservation_(memory_owner_.MakeReservation(sizeof(*this))) {
+    GRPC_TRACE_LOG(secure_endpoint, INFO)
+        << "FrameProtector: " << this << " protector: " << protector_
+        << " zero_copy_protector: " << zero_copy_protector_
+        << " leftover_nslices: " << leftover_nslices;
+    if (leftover_nslices > 0) {
+      leftover_bytes_ = std::make_unique<SliceBuffer>();
+      for (size_t i = 0; i < leftover_nslices; i++) {
+        leftover_bytes_->Append(Slice(CSliceRef(leftover_slices[i])));
+      }
     }
-    grpc_slice_buffer_init(&output_buffer);
-    memory_owner = grpc_core::ResourceQuotaFromChannelArgs(channel_args)
-                       ->memory_quota()
-                       ->CreateMemoryOwner();
-    self_reservation = memory_owner.MakeReservation(sizeof(*this));
-    if (zero_copy_protector) {
-      read_staging_buffer = grpc_empty_slice();
-      write_staging_buffer = grpc_empty_slice();
+    if (zero_copy_protector_ != nullptr) {
+      read_staging_buffer_ = grpc_empty_slice();
+      write_staging_buffer_ = grpc_empty_slice();
     } else {
-      read_staging_buffer =
-          memory_owner.MakeSlice(grpc_core::MemoryRequest(STAGING_BUFFER_SIZE));
-      write_staging_buffer =
-          memory_owner.MakeSlice(grpc_core::MemoryRequest(STAGING_BUFFER_SIZE));
+      read_staging_buffer_ =
+          memory_owner_.MakeSlice(MemoryRequest(STAGING_BUFFER_SIZE));
+      write_staging_buffer_ =
+          memory_owner_.MakeSlice(MemoryRequest(STAGING_BUFFER_SIZE));
     }
-    has_posted_reclaimer.store(false, std::memory_order_relaxed);
-    min_progress_size = 1;
-    grpc_slice_buffer_init(&protector_staging_buffer);
-    gpr_ref_init(&ref, 1);
   }
 
-  ~secure_endpoint() {
-    tsi_frame_protector_destroy(protector);
-    tsi_zero_copy_grpc_protector_destroy(zero_copy_protector);
-    grpc_slice_buffer_destroy(&source_buffer);
-    grpc_slice_buffer_destroy(&leftover_bytes);
-    grpc_core::CSliceUnref(read_staging_buffer);
-    grpc_core::CSliceUnref(write_staging_buffer);
-    grpc_slice_buffer_destroy(&output_buffer);
-    grpc_slice_buffer_destroy(&protector_staging_buffer);
-    gpr_mu_destroy(&protector_mu);
+  ~FrameProtector() override {
+    tsi_frame_protector_destroy(protector_);
+    tsi_zero_copy_grpc_protector_destroy(zero_copy_protector_);
+    CSliceUnref(read_staging_buffer_);
+    CSliceUnref(write_staging_buffer_);
   }
 
-  grpc_core::OrphanablePtr<grpc_endpoint> wrapped_ep;
-  struct tsi_frame_protector* protector;
-  struct tsi_zero_copy_grpc_protector* zero_copy_protector;
-  gpr_mu protector_mu;
-  grpc_core::Mutex read_mu;
-  grpc_core::Mutex write_mu;
-  // saved upper level callbacks and user_data.
-  grpc_closure* read_cb = nullptr;
-  grpc_closure* write_cb = nullptr;
-  grpc_closure on_read;
-  grpc_closure on_write;
-  grpc_slice_buffer* read_buffer = nullptr;
-  grpc_slice_buffer source_buffer;
-  // saved handshaker leftover data to unprotect.
-  grpc_slice_buffer leftover_bytes;
-  // buffers for read and write
-  grpc_slice read_staging_buffer ABSL_GUARDED_BY(read_mu);
-  grpc_slice write_staging_buffer ABSL_GUARDED_BY(write_mu);
-  grpc_slice_buffer output_buffer;
-  grpc_core::MemoryOwner memory_owner;
-  grpc_core::MemoryAllocator::Reservation self_reservation;
-  std::atomic<bool> has_posted_reclaimer;
-  int min_progress_size;
-  grpc_slice_buffer protector_staging_buffer;
-  gpr_refcount ref;
-};
-}  // namespace
-
-static void destroy(secure_endpoint* ep) { delete ep; }
+  Mutex* read_mu() ABSL_LOCK_RETURNED(read_mu_) { return &read_mu_; }
+  Mutex* write_mu() ABSL_LOCK_RETURNED(write_mu_) { return &write_mu_; }
 
-#ifndef NDEBUG
-#define SECURE_ENDPOINT_UNREF(ep, reason) \
-  secure_endpoint_unref((ep), (reason), __FILE__, __LINE__)
-#define SECURE_ENDPOINT_REF(ep, reason) \
-  secure_endpoint_ref((ep), (reason), __FILE__, __LINE__)
-static void secure_endpoint_unref(secure_endpoint* ep, const char* reason,
-                                  const char* file, int line) {
-  if (GRPC_TRACE_FLAG_ENABLED(secure_endpoint)) {
-    gpr_atm val = gpr_atm_no_barrier_load(&ep->ref.count);
-    VLOG(2).AtLocation(file, line) << "SECENDP unref " << ep << " : " << reason
-                                   << " " << val << " -> " << val - 1;
-  }
-  if (gpr_unref(&ep->ref)) {
-    destroy(ep);
-  }
-}
-
-static void secure_endpoint_ref(secure_endpoint* ep, const char* reason,
-                                const char* file, int line) {
-  if (GRPC_TRACE_FLAG_ENABLED(secure_endpoint)) {
-    gpr_atm val = gpr_atm_no_barrier_load(&ep->ref.count);
-    VLOG(2).AtLocation(file, line) << "SECENDP   ref " << ep << " : " << reason
-                                   << " " << val << " -> " << val + 1;
-  }
-  gpr_ref(&ep->ref);
-}
-#else
-#define SECURE_ENDPOINT_UNREF(ep, reason) secure_endpoint_unref((ep))
-#define SECURE_ENDPOINT_REF(ep, reason) secure_endpoint_ref((ep))
-static void secure_endpoint_unref(secure_endpoint* ep) {
-  if (gpr_unref(&ep->ref)) {
-    destroy(ep);
+  void TraceOp(absl::string_view op, grpc_slice_buffer* slices) {
+    if (GRPC_TRACE_FLAG_ENABLED(secure_endpoint)) {
+      size_t i;
+      if (slices->length < 64) {
+        for (i = 0; i < slices->count; i++) {
+          char* data =
+              grpc_dump_slice(slices->slices[i], GPR_DUMP_HEX | GPR_DUMP_ASCII);
+          LOG(INFO) << op << " " << this << ": " << data;
+          gpr_free(data);
+        }
+      } else {
+        grpc_slice first = GRPC_SLICE_MALLOC(64);
+        grpc_slice_buffer_copy_first_into_buffer(slices, 64,
+                                                 GRPC_SLICE_START_PTR(first));
+        char* data = grpc_dump_slice(first, GPR_DUMP_HEX | GPR_DUMP_ASCII);
+        LOG(INFO) << op << " first:" << this << ": " << data;
+        gpr_free(data);
+        CSliceUnref(first);
+      }
+    }
   }
-}
 
-static void secure_endpoint_ref(secure_endpoint* ep) { gpr_ref(&ep->ref); }
-#endif
+  void MaybePostReclaimer() {
+    if (!has_posted_reclaimer_.exchange(true, std::memory_order_relaxed)) {
+      memory_owner_.PostReclaimer(
+          ReclamationPass::kBenign,
+          [self = Ref()](std::optional<ReclamationSweep> sweep) {
+            if (sweep.has_value()) {
+              GRPC_TRACE_LOG(resource_quota, INFO)
+                  << "secure endpoint: benign reclamation to free memory";
+              grpc_slice temp_read_slice;
+              grpc_slice temp_write_slice;
 
-static void maybe_post_reclaimer(secure_endpoint* ep) {
-  if (!ep->has_posted_reclaimer) {
-    SECURE_ENDPOINT_REF(ep, "benign_reclaimer");
-    ep->has_posted_reclaimer.exchange(true, std::memory_order_relaxed);
-    ep->memory_owner.PostReclaimer(
-        grpc_core::ReclamationPass::kBenign,
-        [ep](absl::optional<grpc_core::ReclamationSweep> sweep) {
-          if (sweep.has_value()) {
-            GRPC_TRACE_LOG(resource_quota, INFO)
-                << "secure endpoint: benign reclamation to free memory";
-            grpc_slice temp_read_slice;
-            grpc_slice temp_write_slice;
-
-            ep->read_mu.Lock();
-            temp_read_slice = ep->read_staging_buffer;
-            ep->read_staging_buffer = grpc_empty_slice();
-            ep->read_mu.Unlock();
-
-            ep->write_mu.Lock();
-            temp_write_slice = ep->write_staging_buffer;
-            ep->write_staging_buffer = grpc_empty_slice();
-            ep->write_mu.Unlock();
-
-            grpc_core::CSliceUnref(temp_read_slice);
-            grpc_core::CSliceUnref(temp_write_slice);
-            ep->has_posted_reclaimer.exchange(false, std::memory_order_relaxed);
-          }
-          SECURE_ENDPOINT_UNREF(ep, "benign_reclaimer");
-        });
-  }
-}
+              self->read_mu_.Lock();
+              temp_read_slice =
+                  std::exchange(self->read_staging_buffer_, grpc_empty_slice());
+              self->read_mu_.Unlock();
 
-static void flush_read_staging_buffer(secure_endpoint* ep, uint8_t** cur,
-                                      uint8_t** end)
-    ABSL_EXCLUSIVE_LOCKS_REQUIRED(ep->read_mu) {
-  grpc_slice_buffer_add_indexed(ep->read_buffer, ep->read_staging_buffer);
-  ep->read_staging_buffer =
-      ep->memory_owner.MakeSlice(grpc_core::MemoryRequest(STAGING_BUFFER_SIZE));
-  *cur = GRPC_SLICE_START_PTR(ep->read_staging_buffer);
-  *end = GRPC_SLICE_END_PTR(ep->read_staging_buffer);
-}
+              self->write_mu_.Lock();
+              temp_write_slice = std::exchange(self->write_staging_buffer_,
+                                               grpc_empty_slice());
+              self->write_mu_.Unlock();
 
-static void call_read_cb(secure_endpoint* ep, grpc_error_handle error) {
-  if (GRPC_TRACE_FLAG_ENABLED(secure_endpoint) && ABSL_VLOG_IS_ON(2)) {
-    size_t i;
-    for (i = 0; i < ep->read_buffer->count; i++) {
-      char* data = grpc_dump_slice(ep->read_buffer->slices[i],
-                                   GPR_DUMP_HEX | GPR_DUMP_ASCII);
-      VLOG(2) << "READ " << ep << ": " << data;
-      gpr_free(data);
+              CSliceUnref(temp_read_slice);
+              CSliceUnref(temp_write_slice);
+              self->has_posted_reclaimer_.store(false,
+                                                std::memory_order_relaxed);
+            }
+          });
     }
   }
-  ep->read_buffer = nullptr;
-  grpc_core::ExecCtx::Run(DEBUG_LOCATION, ep->read_cb, error);
-  SECURE_ENDPOINT_UNREF(ep, "read");
-}
 
-static void on_read(void* user_data, grpc_error_handle error) {
-  unsigned i;
-  uint8_t keep_looping = 0;
-  tsi_result result = TSI_OK;
-  secure_endpoint* ep = static_cast<secure_endpoint*>(user_data);
+  void FlushReadStagingBuffer(uint8_t** cur, uint8_t** end)
+      ABSL_EXCLUSIVE_LOCKS_REQUIRED(read_mu_) {
+    grpc_slice_buffer_add_indexed(read_buffer_, read_staging_buffer_);
+    read_staging_buffer_ =
+        memory_owner_.MakeSlice(MemoryRequest(STAGING_BUFFER_SIZE));
+    *cur = GRPC_SLICE_START_PTR(read_staging_buffer_);
+    *end = GRPC_SLICE_END_PTR(read_staging_buffer_);
+  }
 
-  {
-    grpc_core::MutexLock l(&ep->read_mu);
+  void FinishRead(bool ok) {
+    TraceOp("FinishRead", read_buffer_);
+    // TODO(yangg) experiment with moving this block after read_cb to see if it
+    // helps latency
+    source_buffer_.Clear();
+    if (!ok) grpc_slice_buffer_reset_and_unref(read_buffer_);
+    read_buffer_ = nullptr;
+  }
 
-    // If we were shut down after this callback was scheduled with OK
-    // status but before it was invoked, we need to treat that as an error.
-    if (ep->wrapped_ep == nullptr && error.ok()) {
-      error = absl::CancelledError("secure endpoint shutdown");
+  absl::Status Unprotect(absl::Status read_status)
+      ABSL_EXCLUSIVE_LOCKS_REQUIRED(read_mu_) {
+    if (shutdown_) {
+      return absl::CancelledError("secure endpoint shutdown");
     }
 
-    uint8_t* cur = GRPC_SLICE_START_PTR(ep->read_staging_buffer);
-    uint8_t* end = GRPC_SLICE_END_PTR(ep->read_staging_buffer);
+    GRPC_LATENT_SEE_INNER_SCOPE("unprotect");
+    bool keep_looping = false;
+    tsi_result result = TSI_OK;
+
+    uint8_t* cur = GRPC_SLICE_START_PTR(read_staging_buffer_);
+    uint8_t* end = GRPC_SLICE_END_PTR(read_staging_buffer_);
 
-    if (!error.ok()) {
-      grpc_slice_buffer_reset_and_unref(ep->read_buffer);
-    } else if (ep->zero_copy_protector != nullptr) {
+    if (!read_status.ok()) {
+      grpc_slice_buffer_reset_and_unref(read_buffer_);
+    } else if (zero_copy_protector_ != nullptr) {
       // Use zero-copy grpc protector to unprotect.
       int min_progress_size = 1;
       // Get the size of the last frame which is not yet fully decrypted.
@@ -270,18 +207,18 @@ static void on_read(void* user_data, grpc_error_handle error) {
       // passed to the TCP layer to indicate the minimum number of
       // bytes that need to be read to make meaningful progress. This would
       // avoid reading of small slices from the network.
-      // TODO(vigneshbabu): Set min_progress_size in the regular (non-zero-copy)
-      // frame protector code path as well.
+      // TODO(vigneshbabu): Set min_progress_size in the regular
+      // (non-zero-copy) frame protector code path as well.
       result = tsi_zero_copy_grpc_protector_unprotect(
-          ep->zero_copy_protector, &ep->source_buffer, ep->read_buffer,
+          zero_copy_protector_, source_buffer_.c_slice_buffer(), read_buffer_,
           &min_progress_size);
       min_progress_size = std::max(1, min_progress_size);
-      ep->min_progress_size = result != TSI_OK ? 1 : min_progress_size;
+      min_progress_size_ = result != TSI_OK ? 1 : min_progress_size;
     } else {
       // Use frame protector to unprotect.
       // TODO(yangg) check error, maybe bail out early
-      for (i = 0; i < ep->source_buffer.count; i++) {
-        grpc_slice encrypted = ep->source_buffer.slices[i];
+      for (size_t i = 0; i < source_buffer_.Count(); i++) {
+        grpc_slice encrypted = source_buffer_.c_slice_buffer()->slices[i];
         uint8_t* message_bytes = GRPC_SLICE_START_PTR(encrypted);
         size_t message_size = GRPC_SLICE_LENGTH(encrypted);
 
@@ -289,11 +226,17 @@ static void on_read(void* user_data, grpc_error_handle error) {
           size_t unprotected_buffer_size_written =
               static_cast<size_t>(end - cur);
           size_t processed_message_size = message_size;
-          gpr_mu_lock(&ep->protector_mu);
-          result = tsi_frame_protector_unprotect(
-              ep->protector, message_bytes, &processed_message_size, cur,
-              &unprotected_buffer_size_written);
-          gpr_mu_unlock(&ep->protector_mu);
+          if (IsTsiFrameProtectorWithoutLocksEnabled()) {
+            result = tsi_frame_protector_unprotect(
+                protector_, message_bytes, &processed_message_size, cur,
+                &unprotected_buffer_size_written);
+          } else {
+            protector_mu_.Lock();
+            result = tsi_frame_protector_unprotect(
+                protector_, message_bytes, &processed_message_size, cur,
+                &unprotected_buffer_size_written);
+            protector_mu_.Unlock();
+          }
           if (result != TSI_OK) {
             LOG(ERROR) << "Decryption error: " << tsi_result_to_string(result);
             break;
@@ -303,152 +246,130 @@ static void on_read(void* user_data, grpc_error_handle error) {
           cur += unprotected_buffer_size_written;
 
           if (cur == end) {
-            flush_read_staging_buffer(ep, &cur, &end);
+            FlushReadStagingBuffer(&cur, &end);
             // Force to enter the loop again to extract buffered bytes in
-            // protector. The bytes could be buffered because of running out of
-            // staging_buffer. If this happens at the end of all slices, doing
-            // another unprotect avoids leaving data in the protector.
-            keep_looping = 1;
+            // protector. The bytes could be buffered because of running out
+            // of staging_buffer. If this happens at the end of all slices,
+            // doing another unprotect avoids leaving data in the protector.
+            keep_looping = true;
           } else if (unprotected_buffer_size_written > 0) {
-            keep_looping = 1;
+            keep_looping = true;
           } else {
-            keep_looping = 0;
+            keep_looping = false;
           }
         }
         if (result != TSI_OK) break;
       }
 
-      if (cur != GRPC_SLICE_START_PTR(ep->read_staging_buffer)) {
+      if (cur != GRPC_SLICE_START_PTR(read_staging_buffer_)) {
         grpc_slice_buffer_add(
-            ep->read_buffer,
+            read_buffer_,
             grpc_slice_split_head(
-                &ep->read_staging_buffer,
+                &read_staging_buffer_,
                 static_cast<size_t>(
-                    cur - GRPC_SLICE_START_PTR(ep->read_staging_buffer))));
+                    cur - GRPC_SLICE_START_PTR(read_staging_buffer_))));
       }
     }
-  }
 
-  if (!error.ok()) {
-    call_read_cb(
-        ep, GRPC_ERROR_CREATE_REFERENCING("Secure read failed", &error, 1));
-    return;
-  }
+    if (read_status.ok() && result != TSI_OK) {
+      read_status = GRPC_ERROR_CREATE(
+          absl::StrCat("Unwrap failed (", tsi_result_to_string(result), ")"));
+    }
 
-  // TODO(yangg) experiment with moving this block after read_cb to see if it
-  // helps latency
-  grpc_slice_buffer_reset_and_unref(&ep->source_buffer);
+    GRPC_TRACE_LOG(secure_endpoint, INFO)
+        << "Unprotect: " << this << " read_status: " << read_status;
 
-  if (result != TSI_OK) {
-    grpc_slice_buffer_reset_and_unref(ep->read_buffer);
-    call_read_cb(
-        ep, GRPC_ERROR_CREATE(absl::StrCat("Unwrap failed (",
-                                           tsi_result_to_string(result), ")")));
-    return;
+    return read_status;
   }
 
-  call_read_cb(ep, absl::OkStatus());
-}
-
-static void endpoint_read(grpc_endpoint* secure_ep, grpc_slice_buffer* slices,
-                          grpc_closure* cb, bool urgent,
-                          int /*min_progress_size*/) {
-  secure_endpoint* ep = reinterpret_cast<secure_endpoint*>(secure_ep);
-  ep->read_cb = cb;
-  ep->read_buffer = slices;
-  grpc_slice_buffer_reset_and_unref(ep->read_buffer);
+  void BeginRead(grpc_slice_buffer* slices) {
+    read_buffer_ = slices;
+    grpc_slice_buffer_reset_and_unref(read_buffer_);
+  }
 
-  SECURE_ENDPOINT_REF(ep, "read");
-  if (ep->leftover_bytes.count) {
-    grpc_slice_buffer_swap(&ep->leftover_bytes, &ep->source_buffer);
-    CHECK_EQ(ep->leftover_bytes.count, 0u);
-    on_read(ep, absl::OkStatus());
-    return;
+  bool MaybeCompleteReadImmediately() {
+    GRPC_TRACE_LOG(secure_endpoint, INFO)
+        << "MaybeCompleteReadImmediately: " << this
+        << " leftover_bytes_: " << leftover_bytes_.get();
+    if (leftover_bytes_ != nullptr) {
+      grpc_slice_buffer_swap(leftover_bytes_->c_slice_buffer(),
+                             source_buffer_.c_slice_buffer());
+      leftover_bytes_.reset();
+      return true;
+    }
+    return false;
   }
 
-  grpc_endpoint_read(ep->wrapped_ep.get(), &ep->source_buffer, &ep->on_read,
-                     urgent, /*min_progress_size=*/ep->min_progress_size);
-}
+  grpc_event_engine::experimental::SliceBuffer* source_buffer() {
+    return &source_buffer_;
+  }
 
-static void flush_write_staging_buffer(secure_endpoint* ep, uint8_t** cur,
-                                       uint8_t** end)
-    ABSL_EXCLUSIVE_LOCKS_REQUIRED(ep->write_mu) {
-  grpc_slice_buffer_add_indexed(&ep->output_buffer, ep->write_staging_buffer);
-  ep->write_staging_buffer =
-      ep->memory_owner.MakeSlice(grpc_core::MemoryRequest(STAGING_BUFFER_SIZE));
-  *cur = GRPC_SLICE_START_PTR(ep->write_staging_buffer);
-  *end = GRPC_SLICE_END_PTR(ep->write_staging_buffer);
-  maybe_post_reclaimer(ep);
-}
+  int min_progress_size() const { return min_progress_size_; }
 
-static void on_write(void* user_data, grpc_error_handle error) {
-  secure_endpoint* ep = static_cast<secure_endpoint*>(user_data);
-  grpc_closure* cb = ep->write_cb;
-  ep->write_cb = nullptr;
-  SECURE_ENDPOINT_UNREF(ep, "write");
-  grpc_core::EnsureRunInExecCtx([cb, error = std::move(error)]() {
-    grpc_core::Closure::Run(DEBUG_LOCATION, cb, error);
-  });
-}
+  void FlushWriteStagingBuffer(uint8_t** cur, uint8_t** end)
+      ABSL_EXCLUSIVE_LOCKS_REQUIRED(write_mu_) {
+    output_buffer_.AppendIndexed(
+        grpc_event_engine::experimental::Slice(write_staging_buffer_));
+    write_staging_buffer_ =
+        memory_owner_.MakeSlice(MemoryRequest(STAGING_BUFFER_SIZE));
+    *cur = GRPC_SLICE_START_PTR(write_staging_buffer_);
+    *end = GRPC_SLICE_END_PTR(write_staging_buffer_);
+    MaybePostReclaimer();
+  }
 
-static void endpoint_write(grpc_endpoint* secure_ep, grpc_slice_buffer* slices,
-                           grpc_closure* cb, void* arg, int max_frame_size) {
-  GRPC_LATENT_SEE_INNER_SCOPE("secure_endpoint write");
-  unsigned i;
-  tsi_result result = TSI_OK;
-  secure_endpoint* ep = reinterpret_cast<secure_endpoint*>(secure_ep);
+  tsi_result Protect(grpc_slice_buffer* slices, int max_frame_size)
+      ABSL_EXCLUSIVE_LOCKS_REQUIRED(write_mu_) {
+    if (shutdown_) return TSI_FAILED_PRECONDITION;
 
-  {
-    grpc_core::MutexLock l(&ep->write_mu);
-    uint8_t* cur = GRPC_SLICE_START_PTR(ep->write_staging_buffer);
-    uint8_t* end = GRPC_SLICE_END_PTR(ep->write_staging_buffer);
+    GRPC_LATENT_SEE_INNER_SCOPE("protect");
+    uint8_t* cur = GRPC_SLICE_START_PTR(write_staging_buffer_);
+    uint8_t* end = GRPC_SLICE_END_PTR(write_staging_buffer_);
 
-    grpc_slice_buffer_reset_and_unref(&ep->output_buffer);
+    output_buffer_.Clear();
 
-    if (GRPC_TRACE_FLAG_ENABLED(secure_endpoint) && ABSL_VLOG_IS_ON(2)) {
-      for (i = 0; i < slices->count; i++) {
-        char* data =
-            grpc_dump_slice(slices->slices[i], GPR_DUMP_HEX | GPR_DUMP_ASCII);
-        VLOG(2) << "WRITE " << ep << ": " << data;
-        gpr_free(data);
-      }
-    }
+    TraceOp("Protect", slices);
 
-    if (ep->zero_copy_protector != nullptr) {
+    tsi_result result = TSI_OK;
+    if (zero_copy_protector_ != nullptr) {
       // Use zero-copy grpc protector to protect.
-      result = TSI_OK;
       // Break the input slices into chunks of size = max_frame_size and call
       // tsi_zero_copy_grpc_protector_protect on each chunk. This ensures that
       // the protector cannot create frames larger than the specified
       // max_frame_size.
       while (slices->length > static_cast<size_t>(max_frame_size) &&
              result == TSI_OK) {
-        grpc_slice_buffer_move_first(slices,
-                                     static_cast<size_t>(max_frame_size),
-                                     &ep->protector_staging_buffer);
+        grpc_slice_buffer_move_first(
+            slices, static_cast<size_t>(max_frame_size),
+            protector_staging_buffer_.c_slice_buffer());
         result = tsi_zero_copy_grpc_protector_protect(
-            ep->zero_copy_protector, &ep->protector_staging_buffer,
-            &ep->output_buffer);
+            zero_copy_protector_, protector_staging_buffer_.c_slice_buffer(),
+            output_buffer_.c_slice_buffer());
       }
       if (result == TSI_OK && slices->length > 0) {
         result = tsi_zero_copy_grpc_protector_protect(
-            ep->zero_copy_protector, slices, &ep->output_buffer);
+            zero_copy_protector_, slices, output_buffer_.c_slice_buffer());
       }
-      grpc_slice_buffer_reset_and_unref(&ep->protector_staging_buffer);
+      protector_staging_buffer_.Clear();
     } else {
       // Use frame protector to protect.
-      for (i = 0; i < slices->count; i++) {
+      for (size_t i = 0; i < slices->count; i++) {
         grpc_slice plain = slices->slices[i];
         uint8_t* message_bytes = GRPC_SLICE_START_PTR(plain);
         size_t message_size = GRPC_SLICE_LENGTH(plain);
         while (message_size > 0) {
           size_t protected_buffer_size_to_send = static_cast<size_t>(end - cur);
           size_t processed_message_size = message_size;
-          gpr_mu_lock(&ep->protector_mu);
-          result = tsi_frame_protector_protect(ep->protector, message_bytes,
-                                               &processed_message_size, cur,
-                                               &protected_buffer_size_to_send);
-          gpr_mu_unlock(&ep->protector_mu);
+          if (IsTsiFrameProtectorWithoutLocksEnabled()) {
+            result = tsi_frame_protector_protect(
+                protector_, message_bytes, &processed_message_size, cur,
+                &protected_buffer_size_to_send);
+          } else {
+            protector_mu_.Lock();
+            result = tsi_frame_protector_protect(
+                protector_, message_bytes, &processed_message_size, cur,
+                &protected_buffer_size_to_send);
+            protector_mu_.Unlock();
+          }
           if (result != TSI_OK) {
             LOG(ERROR) << "Encryption error: " << tsi_result_to_string(result);
             break;
@@ -458,7 +379,7 @@ static void endpoint_write(grpc_endpoint* secure_ep, grpc_slice_buffer* slices,
           cur += protected_buffer_size_to_send;
 
           if (cur == end) {
-            flush_write_staging_buffer(ep, &cur, &end);
+            FlushWriteStagingBuffer(&cur, &end);
           }
         }
         if (result != TSI_OK) break;
@@ -467,32 +388,210 @@ static void endpoint_write(grpc_endpoint* secure_ep, grpc_slice_buffer* slices,
         size_t still_pending_size;
         do {
           size_t protected_buffer_size_to_send = static_cast<size_t>(end - cur);
-          gpr_mu_lock(&ep->protector_mu);
-          result = tsi_frame_protector_protect_flush(
-              ep->protector, cur, &protected_buffer_size_to_send,
-              &still_pending_size);
-          gpr_mu_unlock(&ep->protector_mu);
+          if (IsTsiFrameProtectorWithoutLocksEnabled()) {
+            result = tsi_frame_protector_protect_flush(
+                protector_, cur, &protected_buffer_size_to_send,
+                &still_pending_size);
+          } else {
+            protector_mu_.Lock();
+            result = tsi_frame_protector_protect_flush(
+                protector_, cur, &protected_buffer_size_to_send,
+                &still_pending_size);
+            protector_mu_.Unlock();
+          }
           if (result != TSI_OK) break;
           cur += protected_buffer_size_to_send;
           if (cur == end) {
-            flush_write_staging_buffer(ep, &cur, &end);
+            FlushWriteStagingBuffer(&cur, &end);
           }
         } while (still_pending_size > 0);
-        if (cur != GRPC_SLICE_START_PTR(ep->write_staging_buffer)) {
-          grpc_slice_buffer_add(
-              &ep->output_buffer,
-              grpc_slice_split_head(
-                  &ep->write_staging_buffer,
+        if (cur != GRPC_SLICE_START_PTR(write_staging_buffer_)) {
+          output_buffer_.Append(
+              grpc_event_engine::experimental::Slice(grpc_slice_split_head(
+                  &write_staging_buffer_,
                   static_cast<size_t>(
-                      cur - GRPC_SLICE_START_PTR(ep->write_staging_buffer))));
+                      cur - GRPC_SLICE_START_PTR(write_staging_buffer_)))));
         }
       }
     }
+    // TODO(yangg) do different things according to the error type?
+    if (result != TSI_OK) output_buffer_.Clear();
+    return result;
+  }
+
+  grpc_event_engine::experimental::SliceBuffer* output_buffer() {
+    return &output_buffer_;
+  }
+
+  void Shutdown() {
+    shutdown_ = true;
+    memory_owner_.Reset();
+  }
+
+ private:
+  struct tsi_frame_protector* const protector_;
+  struct tsi_zero_copy_grpc_protector* const zero_copy_protector_;
+  Mutex mu_;
+  Mutex read_mu_;
+  Mutex write_mu_;
+  Mutex protector_mu_;
+  grpc_slice_buffer* read_buffer_ = nullptr;
+  grpc_event_engine::experimental::SliceBuffer source_buffer_;
+  // saved handshaker leftover data to unprotect.
+  std::unique_ptr<SliceBuffer> leftover_bytes_;
+  // buffers for read and write
+  grpc_slice read_staging_buffer_ ABSL_GUARDED_BY(read_mu_);
+  grpc_slice write_staging_buffer_ ABSL_GUARDED_BY(write_mu_);
+  grpc_event_engine::experimental::SliceBuffer output_buffer_;
+  MemoryOwner memory_owner_;
+  MemoryAllocator::Reservation self_reservation_;
+  std::atomic<bool> has_posted_reclaimer_{false};
+  int min_progress_size_ = 1;
+  SliceBuffer protector_staging_buffer_;
+  bool shutdown_ = false;
+};
+}  // namespace
+}  // namespace grpc_core
+
+namespace {
+struct secure_endpoint : public grpc_endpoint {
+  secure_endpoint(const grpc_endpoint_vtable* vtbl,
+                  tsi_frame_protector* protector,
+                  tsi_zero_copy_grpc_protector* zero_copy_protector,
+                  grpc_core::OrphanablePtr<grpc_endpoint> endpoint,
+                  grpc_slice* leftover_slices, size_t leftover_nslices,
+                  const grpc_core::ChannelArgs& args)
+      : wrapped_ep(std::move(endpoint)),
+        frame_protector(protector, zero_copy_protector, leftover_slices,
+                        leftover_nslices, args) {
+    this->vtable = vtbl;
+    GRPC_CLOSURE_INIT(&on_read, ::on_read, this, grpc_schedule_on_exec_ctx);
+    GRPC_CLOSURE_INIT(&on_write, ::on_write, this, grpc_schedule_on_exec_ctx);
+    gpr_ref_init(&ref, 1);
+  }
+
+  ~secure_endpoint() {}
+
+  grpc_core::OrphanablePtr<grpc_endpoint> wrapped_ep;
+  grpc_core::FrameProtector frame_protector;
+  // saved upper level callbacks and user_data.
+  grpc_closure* read_cb = nullptr;
+  grpc_closure* write_cb = nullptr;
+  grpc_closure on_read;
+  grpc_closure on_write;
+  gpr_refcount ref;
+};
+}  // namespace
+
+static void destroy(secure_endpoint* ep) { delete ep; }
+
+#ifndef NDEBUG
+#define SECURE_ENDPOINT_UNREF(ep, reason) \
+  secure_endpoint_unref((ep), (reason), __FILE__, __LINE__)
+#define SECURE_ENDPOINT_REF(ep, reason) \
+  secure_endpoint_ref((ep), (reason), __FILE__, __LINE__)
+static void secure_endpoint_unref(secure_endpoint* ep, const char* reason,
+                                  const char* file, int line) {
+  if (GRPC_TRACE_FLAG_ENABLED(secure_endpoint)) {
+    gpr_atm val = gpr_atm_no_barrier_load(&ep->ref.count);
+    VLOG(2).AtLocation(file, line) << "SECENDP unref " << ep << " : " << reason
+                                   << " " << val << " -> " << val - 1;
+  }
+  if (gpr_unref(&ep->ref)) {
+    destroy(ep);
+  }
+}
+
+static void secure_endpoint_ref(secure_endpoint* ep, const char* reason,
+                                const char* file, int line) {
+  if (GRPC_TRACE_FLAG_ENABLED(secure_endpoint)) {
+    gpr_atm val = gpr_atm_no_barrier_load(&ep->ref.count);
+    VLOG(2).AtLocation(file, line) << "SECENDP   ref " << ep << " : " << reason
+                                   << " " << val << " -> " << val + 1;
+  }
+  gpr_ref(&ep->ref);
+}
+#else
+#define SECURE_ENDPOINT_UNREF(ep, reason) secure_endpoint_unref((ep))
+#define SECURE_ENDPOINT_REF(ep, reason) secure_endpoint_ref((ep))
+static void secure_endpoint_unref(secure_endpoint* ep) {
+  if (gpr_unref(&ep->ref)) {
+    destroy(ep);
+  }
+}
+
+static void secure_endpoint_ref(secure_endpoint* ep) { gpr_ref(&ep->ref); }
+#endif
+
+static void call_read_cb(secure_endpoint* ep, grpc_error_handle error) {
+  ep->frame_protector.FinishRead(error.ok());
+  grpc_core::ExecCtx::Run(DEBUG_LOCATION, ep->read_cb, error);
+  SECURE_ENDPOINT_UNREF(ep, "read");
+}
+
+static void on_read(void* user_data, grpc_error_handle error) {
+  secure_endpoint* ep = reinterpret_cast<secure_endpoint*>(user_data);
+
+  {
+    grpc_core::MutexLock lock(ep->frame_protector.read_mu());
+    // If we were shut down after this callback was scheduled with OK
+    // status but before it was invoked, we need to treat that as an error.
+    if (ep->wrapped_ep == nullptr && error.ok()) {
+      error = absl::CancelledError("secure endpoint shutdown");
+    }
+    error = ep->frame_protector.Unprotect(std::move(error));
+  }
+
+  if (!error.ok()) {
+    call_read_cb(
+        ep, GRPC_ERROR_CREATE_REFERENCING("Secure read failed", &error, 1));
+    return;
+  }
+
+  call_read_cb(ep, absl::OkStatus());
+}
+
+static void endpoint_read(grpc_endpoint* secure_ep, grpc_slice_buffer* slices,
+                          grpc_closure* cb, bool urgent,
+                          int /*min_progress_size*/) {
+  secure_endpoint* ep = reinterpret_cast<secure_endpoint*>(secure_ep);
+  ep->read_cb = cb;
+  ep->frame_protector.BeginRead(slices);
+
+  SECURE_ENDPOINT_REF(ep, "read");
+  if (ep->frame_protector.MaybeCompleteReadImmediately()) {
+    on_read(ep, absl::OkStatus());
+    return;
+  }
+
+  grpc_endpoint_read(ep->wrapped_ep.get(),
+                     ep->frame_protector.source_buffer()->c_slice_buffer(),
+                     &ep->on_read, urgent,
+                     ep->frame_protector.min_progress_size());
+}
+
+static void on_write(void* user_data, grpc_error_handle error) {
+  secure_endpoint* ep = static_cast<secure_endpoint*>(user_data);
+  grpc_closure* cb = ep->write_cb;
+  ep->write_cb = nullptr;
+  SECURE_ENDPOINT_UNREF(ep, "write");
+  grpc_core::EnsureRunInExecCtx([cb, error = std::move(error)]() {
+    grpc_core::Closure::Run(DEBUG_LOCATION, cb, error);
+  });
+}
+
+static void endpoint_write(
+    grpc_endpoint* secure_ep, grpc_slice_buffer* slices, grpc_closure* cb,
+    grpc_event_engine::experimental::EventEngine::Endpoint::WriteArgs args) {
+  GRPC_LATENT_SEE_INNER_SCOPE("secure_endpoint write");
+  secure_endpoint* ep = reinterpret_cast<secure_endpoint*>(secure_ep);
+  tsi_result result;
+  {
+    grpc_core::MutexLock lock(ep->frame_protector.write_mu());
+    result = ep->frame_protector.Protect(slices, args.max_frame_size());
   }
 
   if (result != TSI_OK) {
-    // TODO(yangg) do different things according to the error type?
-    grpc_slice_buffer_reset_and_unref(&ep->output_buffer);
     grpc_core::ExecCtx::Run(
         DEBUG_LOCATION, cb,
         GRPC_ERROR_CREATE(
@@ -504,16 +603,17 @@ static void endpoint_write(grpc_endpoint* secure_ep, grpc_slice_buffer* slices,
   // output_buffer at any time until the write completes.
   SECURE_ENDPOINT_REF(ep, "write");
   ep->write_cb = cb;
-  grpc_endpoint_write(ep->wrapped_ep.get(), &ep->output_buffer, &ep->on_write,
-                      arg, max_frame_size);
+  grpc_endpoint_write(ep->wrapped_ep.get(),
+                      ep->frame_protector.output_buffer()->c_slice_buffer(),
+                      &ep->on_write, std::move(args));
 }
 
 static void endpoint_destroy(grpc_endpoint* secure_ep) {
   secure_endpoint* ep = reinterpret_cast<secure_endpoint*>(secure_ep);
-  ep->read_mu.Lock();
+  ep->frame_protector.read_mu()->Lock();
   ep->wrapped_ep.reset();
-  ep->memory_owner.Reset();
-  ep->read_mu.Unlock();
+  ep->frame_protector.Shutdown();
+  ep->frame_protector.read_mu()->Unlock();
   SECURE_ENDPOINT_UNREF(ep, "destroy");
 }
 
@@ -566,13 +666,415 @@ static const grpc_endpoint_vtable vtable = {endpoint_read,
                                             endpoint_get_fd,
                                             endpoint_can_track_err};
 
+namespace grpc_event_engine::experimental {
+namespace {
+
+class SecureEndpoint final : public EventEngine::Endpoint {
+ public:
+  SecureEndpoint(
+      std::unique_ptr<grpc_event_engine::experimental::EventEngine::Endpoint>
+          wrapped_ep,
+      struct tsi_frame_protector* protector,
+      struct tsi_zero_copy_grpc_protector* zero_copy_protector,
+      grpc_slice* leftover_slices, size_t leftover_nslices,
+      const grpc_core::ChannelArgs& channel_args)
+      : impl_(grpc_core::MakeRefCounted<Impl>(
+            std::move(wrapped_ep), protector, zero_copy_protector,
+            leftover_slices, leftover_nslices, channel_args)) {}
+
+  ~SecureEndpoint() override { impl_->Shutdown(); }
+
+  bool Read(absl::AnyInvocable<void(absl::Status)> on_read, SliceBuffer* buffer,
+            ReadArgs in_args) override {
+    return impl_->Read(std::move(on_read), buffer, std::move(in_args));
+  }
+
+  bool Write(absl::AnyInvocable<void(absl::Status)> on_writable,
+             SliceBuffer* data, WriteArgs args) override {
+    return impl_->Write(std::move(on_writable), data, std::move(args));
+  }
+
+  const EventEngine::ResolvedAddress& GetPeerAddress() const override {
+    return impl_->GetPeerAddress();
+  }
+
+  const EventEngine::ResolvedAddress& GetLocalAddress() const override {
+    return impl_->GetLocalAddress();
+  }
+
+  void* QueryExtension(absl::string_view id) override {
+    return impl_->QueryExtension(id);
+  }
+
+  std::shared_ptr<TelemetryInfo> GetTelemetryInfo() const override {
+    return std::make_shared<Impl::TelemetryInfo>(impl_->GetTelemetryInfo());
+  }
+
+ private:
+  class Impl : public grpc_core::RefCounted<Impl> {
+   public:
+    class TelemetryInfo : public EventEngine::Endpoint::TelemetryInfo {
+     public:
+      explicit TelemetryInfo(
+          std::shared_ptr<EventEngine::Endpoint::TelemetryInfo>
+              wrapped_telemetry_info)
+          : wrapped_telemetry_info_(std::move(wrapped_telemetry_info)) {}
+
+      std::vector<size_t> AllWriteMetrics() const override {
+        return wrapped_telemetry_info_
+                   ? wrapped_telemetry_info_->AllWriteMetrics()
+                   : std::vector<size_t>{};
+      }
+
+      std::optional<absl::string_view> GetMetricName(
+          size_t key) const override {
+        return wrapped_telemetry_info_
+                   ? wrapped_telemetry_info_->GetMetricName(key)
+                   : std::nullopt;
+      }
+
+      std::optional<size_t> GetMetricKey(
+          absl::string_view name) const override {
+        return wrapped_telemetry_info_
+                   ? wrapped_telemetry_info_->GetMetricKey(name)
+                   : std::nullopt;
+      }
+
+      std::shared_ptr<EventEngine::Endpoint::MetricsSet> GetMetricsSet(
+          absl::Span<const size_t> keys) const override {
+        return wrapped_telemetry_info_
+                   ? wrapped_telemetry_info_->GetMetricsSet(keys)
+                   : nullptr;
+      }
+
+      std::shared_ptr<EventEngine::Endpoint::MetricsSet> GetFullMetricsSet()
+          const override {
+        return wrapped_telemetry_info_
+                   ? wrapped_telemetry_info_->GetFullMetricsSet()
+                   : nullptr;
+      }
+
+     private:
+      std::shared_ptr<EventEngine::Endpoint::TelemetryInfo>
+          wrapped_telemetry_info_;
+    };
+
+    Impl(std::unique_ptr<grpc_event_engine::experimental::EventEngine::Endpoint>
+             wrapped_ep,
+         struct tsi_frame_protector* protector,
+         struct tsi_zero_copy_grpc_protector* zero_copy_protector,
+         grpc_slice* leftover_slices, size_t leftover_nslices,
+         const grpc_core::ChannelArgs& channel_args)
+        : frame_protector_(protector, zero_copy_protector, leftover_slices,
+                           leftover_nslices, channel_args),
+          wrapped_ep_(std::move(wrapped_ep)),
+          event_engine_(channel_args.GetObjectRef<
+                        grpc_event_engine::experimental::EventEngine>()),
+          large_read_threshold_(std::max(
+              1, channel_args.GetInt(GRPC_ARG_DECRYPTION_OFFLOAD_THRESHOLD)
+                     .value_or(32 * 1024))),
+          large_write_threshold_(std::max(
+              1, channel_args.GetInt(GRPC_ARG_ENCRYPTION_OFFLOAD_THRESHOLD)
+                     .value_or(32 * 1024))),
+          max_buffered_writes_(std::max(
+              0, channel_args
+                     .GetInt(GRPC_ARG_ENCRYPTION_OFFLOAD_MAX_BUFFERED_WRITES)
+                     .value_or(1024 * 1024))) {}
+
+    bool Read(absl::AnyInvocable<void(absl::Status)> on_read,
+              SliceBuffer* buffer, ReadArgs args) {
+      on_read_ = std::move(on_read);
+      frame_protector_.BeginRead(buffer->c_slice_buffer());
+      if (frame_protector_.MaybeCompleteReadImmediately()) {
+        return MaybeFinishReadImmediately();
+      }
+      args.set_read_hint_bytes(frame_protector_.min_progress_size());
+      bool read_completed_immediately = wrapped_ep_->Read(
+          [impl = Ref()](absl::Status status) mutable {
+            FinishAsyncRead(std::move(impl), std::move(status));
+          },
+          frame_protector_.source_buffer(), std::move(args));
+      if (read_completed_immediately) return MaybeFinishReadImmediately();
+      return false;
+    }
+
+    bool Write(absl::AnyInvocable<void(absl::Status)> on_writable,
+               SliceBuffer* data, WriteArgs args) {
+      GRPC_LATENT_SEE_INNER_SCOPE("secure_endpoint write");
+      tsi_result result;
+      frame_protector_.TraceOp("Write", data->c_slice_buffer());
+      if (grpc_core::IsSecureEndpointOffloadLargeWritesEnabled()) {
+        // If we get a zero length frame, just complete without looking at
+        // anything further
+        if (data->Length() == 0) return true;
+        grpc_core::MutexLock lock(&write_queue_mu_);
+        // If there's been a failure observed asynchronously, then fail out with
+        // that error.
+        if (!writing_.ok()) {
+          event_engine_->Run(
+              [on_write = std::move(on_writable),
+               status = writing_.status()]() mutable { on_write(status); });
+          return false;
+        }
+        // If we're already writing (== encrypting on another thread) we need to
+        // queue the writes up to do after that completes.
+        // OR if we're not already writing but this write is large, we push it
+        // onto the event engine to encrypt later.
+        if (*writing_ || data->Length() > large_write_threshold_) {
+          // Since we don't call on_write until we've collected pending_writes
+          // in the FinishAsyncWrites path, and EventEngine insists that one
+          // write finishes before a second begins, we should never see a Write
+          // call here with a non-null pending_writes_.
+          CHECK(pending_writes_ == nullptr);
+          pending_writes_ = std::make_unique<SliceBuffer>(std::move(*data));
+          frame_protector_.TraceOp("Pending",
+                                   pending_writes_->c_slice_buffer());
+          // Wait for the previous write to finish before considering this one.
+          // Note that since EventEngine::Endpoint allows only one outstanding
+          // write, this will pause sending until the callback is invoked.
+          last_write_args_ = std::move(args);
+          on_write_ = std::move(on_writable);
+          if (!*writing_) {
+            writing_ = true;
+            event_engine_->Run([impl = Ref()]() mutable {
+              FinishAsyncWrite(std::move(impl));
+            });
+          }
+          return false;
+        }
+      }
+      // A small write: encrypt inline and write to the socket.
+      {
+        grpc_core::MutexLock lock(frame_protector_.write_mu());
+        result = frame_protector_.Protect(data->c_slice_buffer(),
+                                          args.max_frame_size());
+      }
+      if (result != TSI_OK) {
+        event_engine_->Run(
+            [on_writable = std::move(on_writable), result]() mutable {
+              on_writable(GRPC_ERROR_CREATE(absl::StrCat(
+                  "Wrap failed (", tsi_result_to_string(result), ")")));
+            });
+        return false;
+      }
+      on_write_ = std::move(on_writable);
+      frame_protector_.TraceOp(
+          "Write", frame_protector_.output_buffer()->c_slice_buffer());
+      return wrapped_ep_->Write(
+          [impl = Ref()](absl::Status status) mutable {
+            auto on_write = std::move(impl->on_write_);
+            impl.reset();
+            on_write(status);
+          },
+          frame_protector_.output_buffer(), std::move(args));
+    }
+
+    const EventEngine::ResolvedAddress& GetPeerAddress() const {
+      return wrapped_ep_->GetPeerAddress();
+    }
+
+    const EventEngine::ResolvedAddress& GetLocalAddress() const {
+      return wrapped_ep_->GetLocalAddress();
+    }
+
+    void* QueryExtension(absl::string_view id) {
+      return wrapped_ep_->QueryExtension(id);
+    }
+
+    void Shutdown() {
+      std::unique_ptr<EventEngine::Endpoint> wrapped_ep;
+      grpc_core::MutexLock write_lock(frame_protector_.write_mu());
+      grpc_core::MutexLock read_lock(frame_protector_.read_mu());
+      wrapped_ep = std::move(wrapped_ep_);
+      frame_protector_.Shutdown();
+    }
+
+    std::shared_ptr<TelemetryInfo> GetTelemetryInfo() const {
+      return std::make_shared<Impl::TelemetryInfo>(
+          wrapped_ep_->GetTelemetryInfo());
+    }
+
+   private:
+    bool MaybeFinishReadImmediately() {
+      GRPC_LATENT_SEE_INNER_SCOPE("secure_endpoint maybe finish read");
+      grpc_core::MutexLock lock(frame_protector_.read_mu());
+      // If the read is large, since we got the bytes whilst still calling read,
+      // offload the decryption to event engine.
+      // That way we can do the decryption off this thread (which is usually
+      // under some mutual exclusion device) and publish the bytes using the
+      // async callback path.
+      // (remember there's at most one outstanding read and write on an
+      // EventEngine::Endpoint allowed, so this doesn't risk ordering issues.)
+      if (grpc_core::IsSecureEndpointOffloadLargeReadsEnabled() &&
+          frame_protector_.source_buffer()->Length() > large_read_threshold_) {
+        event_engine_->Run([impl = Ref()]() mutable {
+          FinishAsyncRead(std::move(impl), absl::OkStatus());
+        });
+        return false;
+      }
+      frame_protector_.TraceOp(
+          "Read(Imm)", frame_protector_.source_buffer()->c_slice_buffer());
+      auto status = frame_protector_.Unprotect(absl::OkStatus());
+      frame_protector_.FinishRead(status.ok());
+      if (status.ok()) return true;
+      event_engine_->Run([impl = Ref(), status = std::move(status)]() mutable {
+        auto on_read = std::move(impl->on_read_);
+        impl.reset();
+        on_read(status);
+      });
+      return false;
+    }
+
+    static void FinishAsyncRead(grpc_core::RefCountedPtr<Impl> impl,
+                                absl::Status status) {
+      GRPC_LATENT_SEE_PARENT_SCOPE("secure endpoint finish async read");
+      {
+        grpc_core::MutexLock lock(impl->frame_protector_.read_mu());
+        if (status.ok() && impl->wrapped_ep_ == nullptr) {
+          status = absl::CancelledError("secure endpoint shutdown");
+        }
+        status = impl->frame_protector_.Unprotect(std::move(status));
+      }
+      if (status.ok()) {
+        impl->frame_protector_.TraceOp(
+            "Read", impl->frame_protector_.source_buffer()->c_slice_buffer());
+      }
+      auto on_read = std::move(impl->on_read_);
+      impl->frame_protector_.FinishRead(status.ok());
+      impl.reset();
+      on_read(status);
+    }
+
+    std::string WritingString() ABSL_EXCLUSIVE_LOCKS_REQUIRED(write_queue_mu_) {
+      if (!writing_.ok()) return writing_.status().ToString();
+      return *writing_ ? "true" : "false";
+    }
+
+    static void FailWrites(grpc_core::RefCountedPtr<Impl> impl,
+                           absl::Status status)
+        ABSL_LOCKS_EXCLUDED(frame_protector_.write_mu(), write_queue_mu_) {
+      impl->write_queue_mu_.Lock();
+      impl->writing_ = status;
+      auto on_write = std::move(impl->on_write_);
+      impl->write_queue_mu_.Unlock();
+      impl.reset();
+      if (on_write != nullptr) on_write(status);
+    };
+
+    static void FinishAsyncWrite(grpc_core::RefCountedPtr<Impl> impl) {
+      GRPC_LATENT_SEE_PARENT_SCOPE("secure endpoint finish async write");
+      tsi_result result;
+      std::unique_ptr<SliceBuffer> data;
+      WriteArgs args;
+      // If writes complete immediately we'll loop back to here.
+      while (true) {
+        {
+          // Check to see if we've written all the bytes.
+          grpc_core::ReleasableMutexLock lock(&impl->write_queue_mu_);
+          if (impl->pending_writes_ == nullptr) {
+            impl->writing_ = false;
+            DCHECK(impl->on_write_ == nullptr);
+            lock.Release();
+            return;
+          }
+          // There's more data - grab it under the queue lock.
+          data = std::move(impl->pending_writes_);
+          impl->frame_protector_.TraceOp("data", data->c_slice_buffer());
+          args = std::move(impl->last_write_args_);
+          DCHECK(impl->on_write_ != nullptr);
+        }
+        impl->event_engine_->Run(
+            [on_write = std::move(impl->on_write_)]() mutable {
+              on_write(absl::OkStatus());
+            });
+        // Now grab the frame protector write mutex - this is held for some
+        // time (we do the encryption inside of it) - so it's a different
+        // mutex to the queue mutex above.
+        grpc_core::ReleasableMutexLock lock(impl->frame_protector_.write_mu());
+        // If the endpoint closed whilst waiting for this callback, then fail
+        // out the write and we're done.
+        if (impl->wrapped_ep_ == nullptr) {
+          lock.Release();
+          FailWrites(std::move(impl),
+                     absl::CancelledError("secure endpoint shutdown"));
+          return;
+        }
+        result = impl->frame_protector_.Protect(data->c_slice_buffer(),
+                                                args.max_frame_size());
+        if (result != TSI_OK) {
+          lock.Release();
+          // Protection failed... fail the write and we're done.
+          FailWrites(std::move(impl),
+                     GRPC_ERROR_CREATE(absl::StrCat(
+                         "Wrap failed (", tsi_result_to_string(result), ")")));
+          return;
+        }
+        // Write out the protected bytes - returns true if it finishes
+        // immediately, in which case we'll loop.
+        const bool write_finished_immediately = impl->wrapped_ep_->Write(
+            [impl](absl::Status status) mutable {
+              // Async completion path: if we completed successfully then loop
+              // back into FinishAsyncWrite to see if there's more writing to
+              // do.
+              if (status.ok()) {
+                FinishAsyncWrite(std::move(impl));
+                return;
+              }
+              // Write failed: push the failure up via the callback if it's
+              // there.
+              FailWrites(std::move(impl), status);
+            },
+            impl->frame_protector_.output_buffer(), std::move(args));
+        if (!write_finished_immediately) break;
+      }
+    }
+
+    grpc_core::Mutex write_queue_mu_;
+    absl::StatusOr<bool> writing_ ABSL_GUARDED_BY(write_queue_mu_) = false;
+    WriteArgs last_write_args_ ABSL_GUARDED_BY(write_queue_mu_);
+    std::unique_ptr<SliceBuffer> pending_writes_
+        ABSL_GUARDED_BY(write_queue_mu_);
+    grpc_core::FrameProtector frame_protector_;
+    absl::AnyInvocable<void(absl::Status)> on_read_;
+    absl::AnyInvocable<void(absl::Status)> on_write_;
+    std::unique_ptr<EventEngine::Endpoint> wrapped_ep_;
+    std::shared_ptr<EventEngine> event_engine_;
+    const size_t large_read_threshold_;
+    const size_t large_write_threshold_;
+    const size_t max_buffered_writes_;
+  };
+
+  grpc_core::RefCountedPtr<Impl> impl_;
+};
+
+}  // namespace
+}  // namespace grpc_event_engine::experimental
+
 grpc_core::OrphanablePtr<grpc_endpoint> grpc_secure_endpoint_create(
     struct tsi_frame_protector* protector,
     struct tsi_zero_copy_grpc_protector* zero_copy_protector,
     grpc_core::OrphanablePtr<grpc_endpoint> to_wrap,
-    grpc_slice* leftover_slices, const grpc_channel_args* channel_args,
-    size_t leftover_nslices) {
+    grpc_slice* leftover_slices, size_t leftover_nslices,
+    const grpc_core::ChannelArgs& channel_args) {
+  if (!grpc_core::IsEventEngineSecureEndpointEnabled()) {
+    return grpc_legacy_secure_endpoint_create(
+        protector, zero_copy_protector, std::move(to_wrap), leftover_slices,
+        channel_args.ToC().get(), leftover_nslices);
+  }
+  if (grpc_event_engine::experimental::grpc_get_wrapped_event_engine_endpoint(
+          to_wrap.get()) != nullptr) {
+    std::unique_ptr<grpc_event_engine::experimental::EventEngine::Endpoint>
+        event_engine_endpoint = grpc_event_engine::experimental::
+            grpc_take_wrapped_event_engine_endpoint(to_wrap.release());
+    CHECK(event_engine_endpoint != nullptr);
+    return grpc_core::OrphanablePtr<grpc_endpoint>(
+        grpc_event_engine::experimental::grpc_event_engine_endpoint_create(
+            std::make_unique<grpc_event_engine::experimental::SecureEndpoint>(
+                std::move(event_engine_endpoint), protector,
+                zero_copy_protector, leftover_slices, leftover_nslices,
+                channel_args)));
+  }
   return grpc_core::MakeOrphanable<secure_endpoint>(
       &vtable, protector, zero_copy_protector, std::move(to_wrap),
-      leftover_slices, channel_args, leftover_nslices);
+      leftover_slices, leftover_nslices, channel_args);
 }