gpu: nvgpu: Allow disabling CDE functionality

CDE is a Tegra SoC specific feature. Add new config option
CONFIG_NVGPU_SUPPORT_CDE and #ifdef all CDE specific code with it.

JIRA NVGPU-4

Change-Id: I6f0b0047d6ba2b5c36c2eb9b8a1514776741f5b5
Signed-off-by: Terje Bergstrom <tbergstrom@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1648002
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

12 files changed, 80 insertions, 30 deletions

diff --git a/drivers/gpu/nvgpu/Kconfig b/drivers/gpu/nvgpu/Kconfig
index 9e75339f..13b27813 100644
--- a/drivers/gpu/nvgpu/Kconfig
+++ b/drivers/gpu/nvgpu/Kconfig
@@ -128,3 +128,10 @@ config GK20A_VIDMEM
           Enable support for using and allocating buffers in a distinct video
           memory aperture (in contrast to general system memory), available on
           GPUs that have their own banks. PCIe GPUs have this, for example.
+
+config NVGPU_SUPPORT_CDE
+        bool "Support extraction of comptags for CDE"
+        depends on GK20A
+        default y
+        help
+          Enable support for extraction of comptags for CDE.
diff --git a/drivers/gpu/nvgpu/Makefile b/drivers/gpu/nvgpu/Makefile
index c99a48dc..4169bc9b 100644
--- a/drivers/gpu/nvgpu/Makefile
+++ b/drivers/gpu/nvgpu/Makefile
@@ -38,12 +38,9 @@ nvgpu-y := \
         common/linux/vm.o \
         common/linux/intr.o \
         common/linux/sysfs.o \
-        common/linux/cde.o \
         common/linux/io.o \
         common/linux/io_usermode.o \
         common/linux/rwsem.o \
-        common/linux/cde_gm20b.o \
-        common/linux/cde_gp10b.o \
         common/linux/comptags.o \
         common/linux/dmabuf.o \
         common/linux/sched.o \
@@ -132,7 +129,6 @@ nvgpu-$(CONFIG_DEBUG_FS) += \
         common/linux/debug.o \
         common/linux/debug_gr.o \
         common/linux/debug_fifo.o \
-        common/linux/debug_cde.o \
         common/linux/debug_ce.o \
         common/linux/debug_pmu.o \
         common/linux/debug_sched.o \
@@ -301,3 +297,13 @@ nvgpu-$(CONFIG_TEGRA_GR_VIRTUALIZATION) += \
         common/linux/vgpu/gv11b/vgpu_fifo_gv11b.o \
         common/linux/vgpu/gv11b/vgpu_subctx_gv11b.o \
         common/linux/vgpu/gv11b/vgpu_tsg_gv11b.o
+
+nvgpu-$(CONFIG_NVGPU_SUPPORT_CDE) += \
+        common/linux/cde.o \
+        common/linux/cde_gm20b.o \
+        common/linux/cde_gp10b.o
+
+ifeq ($(CONFIG_DEBUG_FS),y)
+nvgpu-$(CONFIG_NVGPU_SUPPORT_CDE) += \
+        common/linux/debug_cde.o
+endif
diff --git a/drivers/gpu/nvgpu/common/linux/cde.c b/drivers/gpu/nvgpu/common/linux/cde.c
index 894e776d..b366acc4 100644
--- a/drivers/gpu/nvgpu/common/linux/cde.c
+++ b/drivers/gpu/nvgpu/common/linux/cde.c
@@ -44,6 +44,8 @@
 #include "os_linux.h"
 #include "dmabuf.h"
 #include "channel.h"
+#include "cde_gm20b.h"
+#include "cde_gp10b.h"
 
 #include <nvgpu/hw/gk20a/hw_ccsr_gk20a.h>
 #include <nvgpu/hw/gk20a/hw_pbdma_gk20a.h>
@@ -1749,3 +1751,24 @@ int gk20a_mark_compressible_write(struct gk20a *g, u32 buffer_fd,
         dma_buf_put(dmabuf);
         return 0;
 }
+
+int nvgpu_cde_init_ops(struct nvgpu_os_linux *l)
+{
+        struct gk20a *g = &l->g;
+        u32 ver = g->params.gpu_arch + g->params.gpu_impl;
+
+        switch (ver) {
+        case GK20A_GPUID_GM20B:
+        case GK20A_GPUID_GM20B_B:
+                l->ops.cde = gm20b_cde_ops.cde;
+                break;
+        case NVGPU_GPUID_GP10B:
+                l->ops.cde = gp10b_cde_ops.cde;
+                break;
+        default:
+                /* CDE is optional, so today ignoring unknown chip is fine */
+                break;
+        }
+
+        return 0;
+}
diff --git a/drivers/gpu/nvgpu/common/linux/cde.h b/drivers/gpu/nvgpu/common/linux/cde.h
index 91ea9b88..115558dc 100644
--- a/drivers/gpu/nvgpu/common/linux/cde.h
+++ b/drivers/gpu/nvgpu/common/linux/cde.h
@@ -321,5 +321,6 @@ int gk20a_prepare_compressible_read(
 int gk20a_mark_compressible_write(
                 struct gk20a *g, u32 buffer_fd, u32 valid_compbits, u64 offset,
                 u32 zbc_color);
+int nvgpu_cde_init_ops(struct nvgpu_os_linux *l);
 
 #endif
diff --git a/drivers/gpu/nvgpu/common/linux/debug.c b/drivers/gpu/nvgpu/common/linux/debug.c
index af3b9964..80a1f7db 100644
--- a/drivers/gpu/nvgpu/common/linux/debug.c
+++ b/drivers/gpu/nvgpu/common/linux/debug.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (C) 2017 NVIDIA Corporation.  All rights reserved.
+ * Copyright (C) 2017-2018 NVIDIA Corporation.  All rights reserved.
  *
  * This software is licensed under the terms of the GNU General Public
  * License version 2, as published by the Free Software Foundation, and
@@ -377,7 +377,9 @@ void gk20a_debug_init(struct gk20a *g, const char *debugfs_symlink)
         gr_gk20a_debugfs_init(g);
         gk20a_pmu_debugfs_init(g);
         gk20a_railgating_debugfs_init(g);
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         gk20a_cde_debugfs_init(g);
+#endif
         gk20a_ce_debugfs_init(g);
         nvgpu_alloc_debugfs_init(g);
         nvgpu_hal_debugfs_init(g);
diff --git a/drivers/gpu/nvgpu/common/linux/driver_common.c b/drivers/gpu/nvgpu/common/linux/driver_common.c
index 66078a85..b3333e37 100644
--- a/drivers/gpu/nvgpu/common/linux/driver_common.c
+++ b/drivers/gpu/nvgpu/common/linux/driver_common.c
@@ -148,7 +148,9 @@ static void nvgpu_init_pm_vars(struct gk20a *g)
         g->aggressive_sync_destroy = platform->aggressive_sync_destroy;
         g->aggressive_sync_destroy_thresh = platform->aggressive_sync_destroy_thresh;
         g->has_syncpoints = platform->has_syncpoints;
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         g->has_cde = platform->has_cde;
+#endif
         g->ptimer_src_freq = platform->ptimer_src_freq;
         g->support_pmu = support_gk20a_pmu(dev_from_gk20a(g));
         g->can_railgate = platform->can_railgate_init;
diff --git a/drivers/gpu/nvgpu/common/linux/ioctl_ctrl.c b/drivers/gpu/nvgpu/common/linux/ioctl_ctrl.c
index 0b449692..71a9bee6 100644
--- a/drivers/gpu/nvgpu/common/linux/ioctl_ctrl.c
+++ b/drivers/gpu/nvgpu/common/linux/ioctl_ctrl.c
@@ -333,10 +333,12 @@ static int gk20a_ctrl_prepare_compressible_read(
                 struct gk20a *g,
                 struct nvgpu_gpu_prepare_compressible_read_args *args)
 {
+        int ret = -ENOSYS;
+
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
         struct nvgpu_fence fence;
         struct gk20a_fence *fence_out = NULL;
-        int ret = 0;
         int flags = args->submit_flags;
 
         fence.id = args->fence.syncpt_id;
@@ -377,18 +379,21 @@ static int gk20a_ctrl_prepare_compressible_read(
                 }
         }
         gk20a_fence_put(fence_out);
+#endif
 
-        return 0;
+        return ret;
 }
 
 static int gk20a_ctrl_mark_compressible_write(
                 struct gk20a *g,
                 struct nvgpu_gpu_mark_compressible_write_args *args)
 {
-        int ret;
+        int ret = -ENOSYS;
 
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         ret = gk20a_mark_compressible_write(g, args->handle,
                         args->valid_compbits, args->offset, args->zbc_color);
+#endif
 
         return ret;
 }
diff --git a/drivers/gpu/nvgpu/common/linux/module.c b/drivers/gpu/nvgpu/common/linux/module.c
index 67dfeff3..5dfd8ff7 100644
--- a/drivers/gpu/nvgpu/common/linux/module.c
+++ b/drivers/gpu/nvgpu/common/linux/module.c
@@ -48,17 +48,18 @@
 #include "module.h"
 #include "module_usermode.h"
 #include "intr.h"
-#include "cde.h"
 #include "ioctl.h"
 #include "sim.h"
 
 #include "os_linux.h"
-#include "cde_gm20b.h"
-#include "cde_gp10b.h"
 #include "ctxsw_trace.h"
 #include "driver_common.h"
 #include "channel.h"
 
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
+#include "cde.h"
+#endif
+
 #define CLASS_NAME "nvidia-gpu"
 /* TODO: Change to e.g. "nvidia-gpu%s" once we have symlinks in place. */
 
@@ -180,23 +181,13 @@ static int gk20a_restore_registers(struct gk20a *g)
 
 static int nvgpu_init_os_linux_ops(struct nvgpu_os_linux *l)
 {
-        struct gk20a *g = &l->g;
-        u32 ver = g->params.gpu_arch + g->params.gpu_impl;
-
-        switch (ver) {
-        case GK20A_GPUID_GM20B:
-        case GK20A_GPUID_GM20B_B:
-                l->ops.cde = gm20b_cde_ops.cde;
-                break;
-        case NVGPU_GPUID_GP10B:
-                l->ops.cde = gp10b_cde_ops.cde;
-                break;
-        default:
-                /* CDE is optional, so today ignoring unknown chip is fine */
-                break;
-        }
+        int err = 0;
 
-        return 0;
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
+        err = nvgpu_cde_init_ops(l);
+#endif
+
+        return err;
 }
 
 int nvgpu_finalize_poweron_linux(struct nvgpu_os_linux *l)
@@ -285,8 +276,10 @@ int gk20a_pm_finalize_poweron(struct device *dev)
 
         gk20a_scale_resume(dev_from_gk20a(g));
 
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         if (platform->has_cde)
                 gk20a_init_cde_support(l);
+#endif
 
         err = gk20a_sched_ctrl_init(g);
         if (err) {
@@ -325,7 +318,9 @@ static int gk20a_lockout_registers(struct gk20a *g)
 static int gk20a_pm_prepare_poweroff(struct device *dev)
 {
         struct gk20a *g = get_gk20a(dev);
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
+#endif
         int ret = 0;
         struct gk20a_platform *platform = gk20a_get_platform(dev);
         bool irqs_enabled;
@@ -348,7 +343,9 @@ static int gk20a_pm_prepare_poweroff(struct device *dev)
 
         gk20a_scale_suspend(dev);
 
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         gk20a_cde_suspend(l);
+#endif
 
         ret = gk20a_prepare_poweroff(g);
         if (ret)
@@ -1239,7 +1236,9 @@ return_err:
 int nvgpu_remove(struct device *dev, struct class *class)
 {
         struct gk20a *g = get_gk20a(dev);
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         struct nvgpu_os_linux *l = nvgpu_os_linux_from_gk20a(g);
+#endif
         struct gk20a_platform *platform = gk20a_get_platform(dev);
         int err;
 
@@ -1251,8 +1250,10 @@ int nvgpu_remove(struct device *dev, struct class *class)
         if (nvgpu_mem_is_valid(&g->syncpt_mem))
                 nvgpu_dma_free(g, &g->syncpt_mem);
 
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         if (platform->has_cde)
                 gk20a_cde_destroy(l);
+#endif
 
 #ifdef CONFIG_GK20A_CTXSW_TRACE
         gk20a_ctxsw_trace_cleanup(g);
diff --git a/drivers/gpu/nvgpu/common/linux/platform_gk20a.h b/drivers/gpu/nvgpu/common/linux/platform_gk20a.h
index ba4880af..927a2cee 100644
--- a/drivers/gpu/nvgpu/common/linux/platform_gk20a.h
+++ b/drivers/gpu/nvgpu/common/linux/platform_gk20a.h
@@ -217,7 +217,9 @@ struct gk20a_platform {
         /* source frequency for ptimer in hz */
         u32 ptimer_src_freq;
 
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         bool has_cde;
+#endif
 
         /* soc name for finding firmware files */
         const char *soc_name;
diff --git a/drivers/gpu/nvgpu/common/linux/platform_gk20a_tegra.c b/drivers/gpu/nvgpu/common/linux/platform_gk20a_tegra.c
index 219dcd40..952a7ed1 100644
--- a/drivers/gpu/nvgpu/common/linux/platform_gk20a_tegra.c
+++ b/drivers/gpu/nvgpu/common/linux/platform_gk20a_tegra.c
@@ -948,7 +948,9 @@ struct gk20a_platform gm20b_tegra_platform = {
 
         .dump_platform_dependencies = gk20a_tegra_debug_dump,
 
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         .has_cde = true,
+#endif
 
         .soc_name = "tegra21x",
 
diff --git a/drivers/gpu/nvgpu/common/linux/platform_gp10b_tegra.c b/drivers/gpu/nvgpu/common/linux/platform_gp10b_tegra.c
index fd3c3e2c..c90277e7 100644
--- a/drivers/gpu/nvgpu/common/linux/platform_gp10b_tegra.c
+++ b/drivers/gpu/nvgpu/common/linux/platform_gp10b_tegra.c
@@ -409,7 +409,9 @@ struct gk20a_platform gp10b_tegra_platform = {
 
         .dump_platform_dependencies = gk20a_tegra_debug_dump,
 
+#ifdef CONFIG_NVGPU_SUPPORT_CDE
         .has_cde = true,
+#endif
 
         .clk_round_rate = gp10b_round_clk_rate,
         .get_clk_freqs = gp10b_clk_get_freqs,
diff --git a/drivers/gpu/nvgpu/common/linux/platform_gv11b_tegra.c b/drivers/gpu/nvgpu/common/linux/platform_gv11b_tegra.c
index ad56167a..407cd27f 100644
--- a/drivers/gpu/nvgpu/common/linux/platform_gv11b_tegra.c
+++ b/drivers/gpu/nvgpu/common/linux/platform_gv11b_tegra.c
@@ -211,9 +211,6 @@ static int gv11b_tegra_suspend(struct device *dev)
 struct gk20a_platform gv11b_tegra_platform = {
         .has_syncpoints = true,
 
-        /* no cde. use sysmem compression */
-        .has_cde = false,
-
         /* ptimer src frequency in hz*/
         .ptimer_src_freq        = 31250000,