i965: Use absolute addressing for constant buffer 0 on Kernel 4.16+.

By default, 3DSTATE_CONSTANT_* Constant Buffer 0 is relative to dynamic
state base address.  This makes it unusable for pushing UBOs.

There is a bit in the INSTPM register (or CS_DEBUG_MODE2 on Skylake)
which controls whether buffer 0 is relative to dynamic state base
address, or simply a normal pointer.  Setting that gives us full
flexibility.  This lets us push up to 4 UBO ranges.

We can't currently write this on Haswell and earlier, and will need
to update the kernel command parser, and then do the whole version
checking song and dance.  We also need a brand new kernel that supports
context isolation - on older kernels, newly created contexts inherit
register state from whatever happened to be running.  So, setting this
would have catastrophic impact on other drivers such as libva, Beignet,
or older Mesa.

See commit 8ec5a4e4a4 where we did this
once before, but had to revert it in commit 013d331220.

Reviewed-by: Francisco Jerez <currojerez@riseup.net>

src/mesa/drivers/dri/i965/brw_state_upload.c

@@ -49,6 +49,7 @@ static void
 brw_upload_initial_gpu_state(struct brw_context *brw)
 {
    const struct gen_device_info *devinfo = &brw->screen->devinfo;
+   const struct brw_compiler *compiler = brw->screen->compiler;
 
    /* On platforms with hardware contexts, we can set our initial GPU state
     * right away rather than doing it via state atoms.  This saves a small
@@ -115,6 +116,29 @@ brw_upload_initial_gpu_state(struct brw_context *brw)
       OUT_BATCH(0);
       ADVANCE_BATCH();
    }
+
+   /* Set the "CONSTANT_BUFFER Address Offset Disable" bit, so
+    * 3DSTATE_CONSTANT_XS buffer 0 is an absolute address.
+    *
+    * This is only safe on kernels with context isolation support.
+    */
+   if (!compiler->constant_buffer_0_is_relative) {
+      if (devinfo->gen >= 9) {
+         BEGIN_BATCH(3);
+         OUT_BATCH(MI_LOAD_REGISTER_IMM | (3 - 2));
+         OUT_BATCH(CS_DEBUG_MODE2);
+         OUT_BATCH(REG_MASK(CSDBG2_CONSTANT_BUFFER_ADDRESS_OFFSET_DISABLE) |
+                   CSDBG2_CONSTANT_BUFFER_ADDRESS_OFFSET_DISABLE);
+         ADVANCE_BATCH();
+      } else if (devinfo->gen == 8) {
+         BEGIN_BATCH(3);
+         OUT_BATCH(MI_LOAD_REGISTER_IMM | (3 - 2));
+         OUT_BATCH(INSTPM);
+         OUT_BATCH(REG_MASK(INSTPM_CONSTANT_BUFFER_ADDRESS_OFFSET_DISABLE) |
+                   INSTPM_CONSTANT_BUFFER_ADDRESS_OFFSET_DISABLE);
+         ADVANCE_BATCH();
+      }
+   }
 }
 
 static inline const struct brw_tracked_state *

src/mesa/drivers/dri/i965/intel_screen.c

@@ -2700,7 +2700,14 @@ __DRIconfig **intelInitScreen2(__DRIscreen *dri_screen)
    screen->compiler = brw_compiler_create(screen, devinfo);
    screen->compiler->shader_debug_log = shader_debug_log_mesa;
    screen->compiler->shader_perf_log = shader_perf_log_mesa;
-   screen->compiler->constant_buffer_0_is_relative = true;
+
+   /* Changing the meaning of constant buffer pointers from a dynamic state
+    * offset to an absolute address is only safe if the kernel isolates other
+    * contexts from our changes.
+    */
+   screen->compiler->constant_buffer_0_is_relative = devinfo->gen < 8 ||
+      !(screen->kernel_features & KERNEL_ALLOWS_CONTEXT_ISOLATION);
+
    screen->compiler->supports_pull_constants = true;
 
    screen->has_exec_fence =