diff --git a/src/intel/vulkan/anv_batch_chain.c b/src/intel/vulkan/anv_batch_chain.c
index 22975ddc0c8..8da9aa84682 100644
--- a/src/intel/vulkan/anv_batch_chain.c
+++ b/src/intel/vulkan/anv_batch_chain.c
@@ -1287,7 +1287,8 @@ anv_queue_exec_locked(struct anv_queue *queue,
                       uint32_t signal_count,
                       const struct vk_sync_signal *signals,
                       struct anv_query_pool *perf_query_pool,
-                      uint32_t perf_query_pass)
+                      uint32_t perf_query_pass,
+                      struct anv_utrace_submit *utrace_submit)
 {
    struct anv_device *device = queue->device;
    VkResult result = VK_SUCCESS;
@@ -1311,7 +1312,8 @@ anv_queue_exec_locked(struct anv_queue *queue,
          cmd_buffer_count, cmd_buffers,
          needs_companion_sync ? 0 : signal_count, signals,
          perf_query_pool,
-         perf_query_pass);
+         perf_query_pass,
+         utrace_submit);
    if (result != VK_SUCCESS)
       return result;
 
@@ -1328,7 +1330,8 @@ anv_queue_exec_locked(struct anv_queue *queue,
                                                 1, &companion_sync,
                                                 0, NULL,
                                                 signal_count, signals,
-                                                NULL, 0);
+                                                NULL, 0,
+                                                NULL);
    }
 
    return result;
@@ -1342,7 +1345,8 @@ can_chain_query_pools(struct anv_query_pool *p1, struct anv_query_pool *p2)
 
 static VkResult
 anv_queue_submit_locked(struct anv_queue *queue,
-                        struct vk_queue_submit *submit)
+                        struct vk_queue_submit *submit,
+                        struct anv_utrace_submit *utrace_submit)
 {
    VkResult result;
 
@@ -1363,7 +1367,8 @@ anv_queue_submit_locked(struct anv_queue *queue,
                                      NULL /* cmd_buffers */,
                                      submit->signal_count, submit->signals,
                                      NULL /* perf_query_pool */,
-                                     0 /* perf_query_pass */);
+                                     0 /* perf_query_pass */,
+                                     utrace_submit);
       if (result != VK_SUCCESS)
          return result;
    } else {
@@ -1401,7 +1406,8 @@ anv_queue_submit_locked(struct anv_queue *queue,
                                      next == end ? submit->signal_count : 0,
                                      next == end ? submit->signals : NULL,
                                      perf_query_pool,
-                                     submit->perf_pass_index);
+                                     submit->perf_pass_index,
+                                     next == end ? utrace_submit : NULL);
             if (result != VK_SUCCESS)
                return result;
             if (next < end) {
@@ -1456,10 +1462,22 @@ anv_queue_submit(struct vk_queue *vk_queue,
       return VK_SUCCESS;
    }
 
+   /* Flush the trace points first before taking the lock as the flushing
+    * might try to take that same lock.
+    */
+   struct anv_utrace_submit *utrace_submit = NULL;
+   result = anv_device_utrace_flush_cmd_buffers(
+      queue,
+      submit->command_buffer_count,
+      (struct anv_cmd_buffer **)submit->command_buffers,
+      &utrace_submit);
+   if (result != VK_SUCCESS)
+      return result;
+
    pthread_mutex_lock(&device->mutex);
 
    uint64_t start_ts = intel_ds_begin_submit(&queue->ds);
-   result = anv_queue_submit_locked(queue, submit);
+   result = anv_queue_submit_locked(queue, submit, utrace_submit);
    /* Take submission ID under lock */
    intel_ds_end_submit(&queue->ds, start_ts);
 
diff --git a/src/intel/vulkan/anv_gem_stubs.c b/src/intel/vulkan/anv_gem_stubs.c
index 9e6ab8c4847..a22491012c4 100644
--- a/src/intel/vulkan/anv_gem_stubs.c
+++ b/src/intel/vulkan/anv_gem_stubs.c
@@ -74,7 +74,8 @@ stub_queue_exec_locked(struct anv_queue *queue,
                        uint32_t signal_count,
                        const struct vk_sync_signal *signals,
                        struct anv_query_pool *perf_query_pool,
-                       uint32_t perf_query_pass)
+                       uint32_t perf_query_pass,
+                       struct anv_utrace_submit *utrace_submit)
 {
    return VK_ERROR_UNKNOWN;
 }
diff --git a/src/intel/vulkan/anv_kmd_backend.h b/src/intel/vulkan/anv_kmd_backend.h
index bf3e093162f..ff1e969a222 100644
--- a/src/intel/vulkan/anv_kmd_backend.h
+++ b/src/intel/vulkan/anv_kmd_backend.h
@@ -86,7 +86,8 @@ struct anv_kmd_backend {
                                  uint32_t signal_count,
                                  const struct vk_sync_signal *signals,
                                  struct anv_query_pool *perf_query_pool,
-                                 uint32_t perf_query_pass);
+                                 uint32_t perf_query_pass,
+                                 struct anv_utrace_submit *utrace_submit);
    VkResult (*queue_exec_trace)(struct anv_queue *queue,
                                 struct anv_utrace_submit *submit);
    uint32_t (*bo_alloc_flags_to_bo_flags)(struct anv_device *device,
diff --git a/src/intel/vulkan/i915/anv_batch_chain.c b/src/intel/vulkan/i915/anv_batch_chain.c
index ed132d4b5ad..030a3b5d013 100644
--- a/src/intel/vulkan/i915/anv_batch_chain.c
+++ b/src/intel/vulkan/i915/anv_batch_chain.c
@@ -730,24 +730,16 @@ i915_queue_exec_locked(struct anv_queue *queue,
                        uint32_t signal_count,
                        const struct vk_sync_signal *signals,
                        struct anv_query_pool *perf_query_pool,
-                       uint32_t perf_query_pass)
+                       uint32_t perf_query_pass,
+                       struct anv_utrace_submit *utrace_submit)
 {
    struct anv_device *device = queue->device;
-   struct anv_utrace_submit *utrace_submit = NULL;
    struct anv_execbuf execbuf = {
       .alloc = &queue->device->vk.alloc,
       .alloc_scope = VK_SYSTEM_ALLOCATION_SCOPE_DEVICE,
       .perf_query_pass = perf_query_pass,
    };
-
-   /* Flush the trace points first, they need to be moved */
-   VkResult result =
-      anv_device_utrace_flush_cmd_buffers(queue,
-                                          cmd_buffer_count,
-                                          cmd_buffers,
-                                          &utrace_submit);
-   if (result != VK_SUCCESS)
-      goto error;
+   VkResult result;
 
    if (utrace_submit && !utrace_submit->batch_bo) {
       result = anv_execbuf_add_sync(device, &execbuf,
diff --git a/src/intel/vulkan/i915/anv_batch_chain.h b/src/intel/vulkan/i915/anv_batch_chain.h
index aa3dbe32437..f46f19c90bc 100644
--- a/src/intel/vulkan/i915/anv_batch_chain.h
+++ b/src/intel/vulkan/i915/anv_batch_chain.h
@@ -50,4 +50,5 @@ i915_queue_exec_locked(struct anv_queue *queue,
                        uint32_t signal_count,
                        const struct vk_sync_signal *signals,
                        struct anv_query_pool *perf_query_pool,
-                       uint32_t perf_query_pass);
+                       uint32_t perf_query_pass,
+                       struct anv_utrace_submit *utrace_submit);
diff --git a/src/intel/vulkan/xe/anv_batch_chain.c b/src/intel/vulkan/xe/anv_batch_chain.c
index 76a8a4fc2ec..28588aedfe2 100644
--- a/src/intel/vulkan/xe/anv_batch_chain.c
+++ b/src/intel/vulkan/xe/anv_batch_chain.c
@@ -266,17 +266,12 @@ xe_queue_exec_locked(struct anv_queue *queue,
                      uint32_t signal_count,
                      const struct vk_sync_signal *signals,
                      struct anv_query_pool *perf_query_pool,
-                     uint32_t perf_query_pass)
+                     uint32_t perf_query_pass,
+                     struct anv_utrace_submit *utrace_submit)
 {
    struct anv_device *device = queue->device;
-   struct anv_utrace_submit *utrace_submit = NULL;
    VkResult result;
 
-   result = anv_device_utrace_flush_cmd_buffers(queue, cmd_buffer_count,
-                                                cmd_buffers, &utrace_submit);
-   if (result != VK_SUCCESS)
-      return result;
-
    struct drm_xe_sync *xe_syncs = NULL;
    uint32_t xe_syncs_count = 0;
    result = xe_exec_process_syncs(queue, wait_count, waits,
diff --git a/src/intel/vulkan/xe/anv_batch_chain.h b/src/intel/vulkan/xe/anv_batch_chain.h
index a3a59f03d70..f664f9673dd 100644
--- a/src/intel/vulkan/xe/anv_batch_chain.h
+++ b/src/intel/vulkan/xe/anv_batch_chain.h
@@ -46,7 +46,8 @@ xe_queue_exec_locked(struct anv_queue *queue,
                      uint32_t signal_count,
                      const struct vk_sync_signal *signals,
                      struct anv_query_pool *perf_query_pool,
-                     uint32_t perf_query_pass);
+                     uint32_t perf_query_pass,
+                     struct anv_utrace_submit *utrace_submit);
 
 VkResult
 xe_queue_exec_utrace_locked(struct anv_queue *queue,