gpu: nvgpu: Enable IO coherency on GV100

This reverts commit 848af2ce6de6140323a6ffe3075bf8021e119434.

This is a revert of a revert, etc, etc. It re-enables IO coherence again.

JIRA EVLR-2333

Change-Id: Ibf97dce2f892e48a1200a06cd38a1c5d9603be04
Signed-off-by: Alex Waterman <alexw@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1669722
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

2 files changed, 11 insertions, 7 deletions

diff --git a/drivers/gpu/nvgpu/gm20b/bus_gm20b.c b/drivers/gpu/nvgpu/gm20b/bus_gm20b.c
index 34c8d4b7..cdd70d5b 100644
--- a/drivers/gpu/nvgpu/gm20b/bus_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/bus_gm20b.c
@@ -25,6 +25,7 @@
 #include <nvgpu/timers.h>
 #include <nvgpu/bus.h>
 #include <nvgpu/mm.h>
+#include <nvgpu/enabled.h>
 
 #include "bus_gm20b.h"
 #include "gk20a/gk20a.h"
@@ -43,8 +44,9 @@ int gm20b_bus_bar1_bind(struct gk20a *g, struct nvgpu_mem *bar1_inst)
 
         gk20a_writel(g, bus_bar1_block_r(),
                      nvgpu_aperture_mask(g, bar1_inst,
-                       bus_bar1_block_target_sys_mem_ncoh_f(),
-                       bus_bar1_block_target_vid_mem_f()) |
+                                         bus_bar1_block_target_sys_mem_ncoh_f(),
+                                         bus_bar1_block_target_sys_mem_coh_f(),
+                                         bus_bar1_block_target_vid_mem_f()) |
                      bus_bar1_block_mode_virtual_f() |
                      bus_bar1_block_ptr_f(ptr_v));
         nvgpu_timeout_init(g, &timeout, 1000, NVGPU_TIMER_RETRY_TIMER);
diff --git a/drivers/gpu/nvgpu/gm20b/fifo_gm20b.c b/drivers/gpu/nvgpu/gm20b/fifo_gm20b.c
index 0762e8bd..15612995 100644
--- a/drivers/gpu/nvgpu/gm20b/fifo_gm20b.c
+++ b/drivers/gpu/nvgpu/gm20b/fifo_gm20b.c
@@ -32,6 +32,7 @@
 #include <nvgpu/atomic.h>
 #include <nvgpu/barrier.h>
 #include <nvgpu/mm.h>
+#include <nvgpu/enabled.h>
 
 #include <nvgpu/hw/gm20b/hw_ccsr_gm20b.h>
 #include <nvgpu/hw/gm20b/hw_ram_gm20b.h>
@@ -51,11 +52,12 @@ void channel_gm20b_bind(struct channel_gk20a *c)
 
 
         gk20a_writel(g, ccsr_channel_inst_r(c->chid),
-                ccsr_channel_inst_ptr_f(inst_ptr) |
-                nvgpu_aperture_mask(g, &c->inst_block,
-                 ccsr_channel_inst_target_sys_mem_ncoh_f(),
-                 ccsr_channel_inst_target_vid_mem_f()) |
-                ccsr_channel_inst_bind_true_f());
+                     ccsr_channel_inst_ptr_f(inst_ptr) |
+                     nvgpu_aperture_mask(g, &c->inst_block,
+                                ccsr_channel_inst_target_sys_mem_ncoh_f(),
+                                ccsr_channel_inst_target_sys_mem_coh_f(),
+                                ccsr_channel_inst_target_vid_mem_f()) |
+                     ccsr_channel_inst_bind_true_f());
 
         gk20a_writel(g, ccsr_channel_r(c->chid),
                 (gk20a_readl(g, ccsr_channel_r(c->chid)) &