vk: Implement workaround for occlusion queries

We have an issue with occlusion queries (PIPE_CONTROL depth writes)
after using the pipeline with the VS disabled. We work around it by
using a depth cache flush PIPE_CONTROL before doing a depth write.

Fixes dEQP-VK.query_pool.*

src/vulkan/anv_cmd_buffer.c

@@ -123,6 +123,7 @@ anv_cmd_state_init(struct anv_cmd_state *state)
    state->pipeline = NULL;
    state->restart_index = UINT32_MAX;
    state->dynamic = default_dynamic_state;
+   state->need_query_wa = true;
 
    state->gen7.index_buffer = NULL;
 }

src/vulkan/anv_meta.c

@@ -144,6 +144,11 @@ anv_meta_restore(const struct anv_meta_saved_state *state,
    anv_dynamic_state_copy(&cmd_buffer->state.dynamic, &state->dynamic,
                           state->dynamic_mask);
    cmd_buffer->state.dirty |= state->dynamic_mask;
+
+   /* Since we've used the pipeline with the VS disabled, set
+    * need_query_wa. See CmdBeginQuery.
+    */
+   cmd_buffer->state.need_query_wa = true;
 }
 
 VkImageViewType

src/vulkan/anv_private.h

@@ -1082,6 +1082,7 @@ struct anv_cmd_state {
    struct anv_state                             binding_tables[MESA_SHADER_STAGES];
    struct anv_state                             samplers[MESA_SHADER_STAGES];
    struct anv_dynamic_state                     dynamic;
+   bool                                         need_query_wa;
 
    struct {
       struct anv_buffer *                       index_buffer;

src/vulkan/gen8_cmd_buffer.c

@@ -896,6 +896,19 @@ void genX(CmdBeginQuery)(
    ANV_FROM_HANDLE(anv_cmd_buffer, cmd_buffer, commandBuffer);
    ANV_FROM_HANDLE(anv_query_pool, pool, queryPool);
 
+   /* Workaround: When meta uses the pipeline with the VS disabled, it seems
+    * that the pipelining of the depth write breaks. What we see is that
+    * samples from the render pass clear leaks into the first query
+    * immediately after the clear. Doing a pipecontrol with a post-sync
+    * operation and DepthStallEnable seems to work around the issue.
+    */
+   if (cmd_buffer->state.need_query_wa) {
+      cmd_buffer->state.need_query_wa = false;
+      anv_batch_emit(&cmd_buffer->batch, GENX(PIPE_CONTROL),
+                     .DepthCacheFlushEnable = true,
+                     .DepthStallEnable = true);
+   }
+
    switch (pool->type) {
    case VK_QUERY_TYPE_OCCLUSION:
       emit_ps_depth_count(&cmd_buffer->batch, &pool->bo,