gpu: nvgpu: gv11b: gfxp wfi timeout

For gv11b, configured gfx preemption wfi timeout in usec.
Set timeout unit as usec in gr_gv11b_init_preemption_state.
Used default timeout as 1msec and this timeout value can
be modified through sysfs node:
/sys/devices/gpu.0/gfxp_wfi_timeout_count

For gp10b:
gfxp_wfi_timeout_count is in syclk cycles

For gv11b:
gfxp_wfi_timeout_count is in usec

Bug 2003668

Change-Id: I68d52ce996a83df90b8b3a8164debb07e5cb370f
Signed-off-by: seshendra Gadagottu <sgadagottu@nvidia.com>
Reviewed-on: https://git-master.nvidia.com/r/1599658
Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com>
Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>

12 files changed, 100 insertions, 6 deletions

diff --git a/drivers/gpu/nvgpu/common/linux/sysfs.c b/drivers/gpu/nvgpu/common/linux/sysfs.c
index b1e7d1ed..053c1a12 100644
--- a/drivers/gpu/nvgpu/common/linux/sysfs.c
+++ b/drivers/gpu/nvgpu/common/linux/sysfs.c
@@ -965,8 +965,10 @@ static ssize_t gfxp_wfi_timeout_count_store(struct device *dev,
         if (kstrtoul(buf, 10, &val) < 0)
                 return -EINVAL;
 
-        if (val >= 100*1000*1000) /* 100ms @ 1Ghz */
-                return -EINVAL;
+        if (g->ops.gr.get_max_gfxp_wfi_timeout_count) {
+                if (val >= g->ops.gr.get_max_gfxp_wfi_timeout_count(g))
+                        return -EINVAL;
+        }
 
         gr->gfxp_wfi_timeout_count = val;
 
diff --git a/drivers/gpu/nvgpu/common/linux/vgpu/gp10b/vgpu_hal_gp10b.c b/drivers/gpu/nvgpu/common/linux/vgpu/gp10b/vgpu_hal_gp10b.c
index b35cac90..acd0ad5d 100644
--- a/drivers/gpu/nvgpu/common/linux/vgpu/gp10b/vgpu_hal_gp10b.c
+++ b/drivers/gpu/nvgpu/common/linux/vgpu/gp10b/vgpu_hal_gp10b.c
@@ -209,6 +209,10 @@ static const struct gpu_ops vgpu_gp10b_ops = {
                 .set_ctxsw_preemption_mode =
                                         vgpu_gr_gp10b_set_ctxsw_preemption_mode,
                 .init_ctxsw_hdr_data = gr_gp10b_init_ctxsw_hdr_data,
+                .init_gfxp_wfi_timeout_count =
+                        gr_gp10b_init_gfxp_wfi_timeout_count,
+                .get_max_gfxp_wfi_timeout_count =
+                        gr_gp10b_get_max_gfxp_wfi_timeout_count,
         },
         .fb = {
                 .reset = fb_gk20a_reset,
diff --git a/drivers/gpu/nvgpu/common/linux/vgpu/gv11b/vgpu_hal_gv11b.c b/drivers/gpu/nvgpu/common/linux/vgpu/gv11b/vgpu_hal_gv11b.c
index a7491b20..a470377c 100644
--- a/drivers/gpu/nvgpu/common/linux/vgpu/gv11b/vgpu_hal_gv11b.c
+++ b/drivers/gpu/nvgpu/common/linux/vgpu/gv11b/vgpu_hal_gv11b.c
@@ -248,6 +248,10 @@ static const struct gpu_ops vgpu_gv11b_ops = {
                         gr_gv11b_handle_tpc_sm_ecc_exception,
                 .decode_egpc_addr = gv11b_gr_decode_egpc_addr,
                 .init_ctxsw_hdr_data = gr_gp10b_init_ctxsw_hdr_data,
+                .init_gfxp_wfi_timeout_count =
+                        gr_gv11b_init_gfxp_wfi_timeout_count,
+                .get_max_gfxp_wfi_timeout_count =
+                        gr_gv11b_get_max_gfxp_wfi_timeout_count,
         },
         .fb = {
                 .reset = gv11b_fb_reset,
diff --git a/drivers/gpu/nvgpu/gk20a/gk20a.h b/drivers/gpu/nvgpu/gk20a/gk20a.h
index 8f3900a1..a361648f 100644
--- a/drivers/gpu/nvgpu/gk20a/gk20a.h
+++ b/drivers/gpu/nvgpu/gk20a/gk20a.h
@@ -433,7 +433,9 @@ struct gpu_ops {
                 void (*disable_rd_coalesce)(struct gk20a *g);
                 void (*init_ctxsw_hdr_data)(struct gk20a *g,
                                         struct nvgpu_mem *mem);
-
+                void (*init_gfxp_wfi_timeout_count)(struct gk20a *g);
+                unsigned long (*get_max_gfxp_wfi_timeout_count)
+                                        (struct gk20a *g);
         } gr;
         struct {
                 void (*init_hw)(struct gk20a *g);
diff --git a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
index 64b54699..73b49927 100644
--- a/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
+++ b/drivers/gpu/nvgpu/gk20a/gr_gk20a.c
@@ -79,7 +79,6 @@
 #define CTXSW_MEM_SCRUBBING_TIMEOUT_DEFAULT 10
 #define FECS_ARB_CMD_TIMEOUT_MAX 40
 #define FECS_ARB_CMD_TIMEOUT_DEFAULT 2
-#define GFXP_WFI_TIMEOUT_COUNT_DEFAULT 100000
 
 static int gk20a_init_gr_bind_fecs_elpg(struct gk20a *g);
 
@@ -4840,7 +4839,8 @@ static int gk20a_init_gr_setup_sw(struct gk20a *g)
         if (g->ops.gr.init_czf_bypass)
                 g->ops.gr.init_czf_bypass(g);
 
-        gr->gfxp_wfi_timeout_count = GFXP_WFI_TIMEOUT_COUNT_DEFAULT;
+        if (g->ops.gr.init_gfxp_wfi_timeout_count)
+                g->ops.gr.init_gfxp_wfi_timeout_count(g);
 
         nvgpu_mutex_init(&gr->ctx_mutex);
         nvgpu_spinlock_init(&gr->ch_tlb_lock);
diff --git a/drivers/gpu/nvgpu/gp10b/gr_gp10b.c b/drivers/gpu/nvgpu/gp10b/gr_gp10b.c
index 27a42a45..c232c4ab 100644
--- a/drivers/gpu/nvgpu/gp10b/gr_gp10b.c
+++ b/drivers/gpu/nvgpu/gp10b/gr_gp10b.c
@@ -46,6 +46,8 @@
 #include <nvgpu/hw/gp10b/hw_mc_gp10b.h>
 #include <nvgpu/hw/gp10b/hw_fuse_gp10b.h>
 
+#define GFXP_WFI_TIMEOUT_COUNT_DEFAULT 100000
+
 bool gr_gp10b_is_valid_class(struct gk20a *g, u32 class_num)
 {
         bool valid = false;
@@ -2351,3 +2353,16 @@ void gr_gp10b_init_ctxsw_hdr_data(struct gk20a *g, struct nvgpu_mem *mem)
         nvgpu_mem_wr(g, mem,
                         ctxsw_prog_main_image_num_cilp_save_ops_o(), 0);
 }
+
+void gr_gp10b_init_gfxp_wfi_timeout_count(struct gk20a *g)
+{
+        struct gr_gk20a *gr = &g->gr;
+
+        gr->gfxp_wfi_timeout_count = GFXP_WFI_TIMEOUT_COUNT_DEFAULT;
+}
+
+unsigned long gr_gp10b_get_max_gfxp_wfi_timeout_count(struct gk20a *g)
+{
+        /* 100msec @ 1GHZ */
+        return (100 * 1000 * 1000UL);
+}
diff --git a/drivers/gpu/nvgpu/gp10b/gr_gp10b.h b/drivers/gpu/nvgpu/gp10b/gr_gp10b.h
index 55117022..3b0f0f2e 100644
--- a/drivers/gpu/nvgpu/gp10b/gr_gp10b.h
+++ b/drivers/gpu/nvgpu/gp10b/gr_gp10b.h
@@ -144,5 +144,7 @@ void gr_gp10b_set_preemption_buffer_va(struct gk20a *g,
 int gr_gp10b_set_czf_bypass(struct gk20a *g, struct channel_gk20a *ch);
 void gr_gp10b_init_czf_bypass(struct gk20a *g);
 void gr_gp10b_init_ctxsw_hdr_data(struct gk20a *g, struct nvgpu_mem *mem);
+void gr_gp10b_init_gfxp_wfi_timeout_count(struct gk20a *g);
+unsigned long gr_gp10b_get_max_gfxp_wfi_timeout_count(struct gk20a *g);
 
 #endif
diff --git a/drivers/gpu/nvgpu/gp10b/hal_gp10b.c b/drivers/gpu/nvgpu/gp10b/hal_gp10b.c
index f13c2735..5ae6b638 100644
--- a/drivers/gpu/nvgpu/gp10b/hal_gp10b.c
+++ b/drivers/gpu/nvgpu/gp10b/hal_gp10b.c
@@ -332,6 +332,10 @@ static const struct gpu_ops gp10b_ops = {
                 .create_gr_sysfs = gr_gp10b_create_sysfs,
                 .set_ctxsw_preemption_mode = gr_gp10b_set_ctxsw_preemption_mode,
                 .init_ctxsw_hdr_data = gr_gp10b_init_ctxsw_hdr_data,
+                .init_gfxp_wfi_timeout_count =
+                                gr_gp10b_init_gfxp_wfi_timeout_count,
+                .get_max_gfxp_wfi_timeout_count =
+                                gr_gp10b_get_max_gfxp_wfi_timeout_count,
         },
         .fb = {
                 .reset = fb_gk20a_reset,
diff --git a/drivers/gpu/nvgpu/gv11b/gr_gv11b.c b/drivers/gpu/nvgpu/gv11b/gr_gv11b.c
index f2397108..9d6d8d05 100644
--- a/drivers/gpu/nvgpu/gv11b/gr_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/gr_gv11b.c
@@ -54,6 +54,8 @@
 #include <nvgpu/hw/gv11b/hw_therm_gv11b.h>
 #include <nvgpu/hw/gv11b/hw_fb_gv11b.h>
 
+#define GFXP_WFI_TIMEOUT_COUNT_IN_USEC_DEFAULT 1000
+
 bool gr_gv11b_is_valid_class(struct gk20a *g, u32 class_num)
 {
         bool valid = false;
@@ -3625,3 +3627,35 @@ void gr_gv11b_init_gpc_mmu(struct gk20a *g)
         gk20a_writel(g, gr_gpcs_pri_mmu_debug_rd_r(),
                         gk20a_readl(g, fb_mmu_debug_rd_r()));
 }
+
+int gr_gv11b_init_preemption_state(struct gk20a *g)
+{
+        u32 debug_2;
+        struct gr_gk20a *gr = &g->gr;
+
+        nvgpu_log_fn(g, " ");
+
+        gk20a_writel(g, gr_fe_gfxp_wfi_timeout_r(),
+                        gr_fe_gfxp_wfi_timeout_count_f(
+                        gr->gfxp_wfi_timeout_count));
+
+        debug_2 = gk20a_readl(g, gr_debug_2_r());
+        debug_2 = set_field(debug_2,
+                        gr_debug_2_gfxp_wfi_timeout_unit_m(),
+                        gr_debug_2_gfxp_wfi_timeout_unit_usec_f());
+        gk20a_writel(g, gr_debug_2_r(), debug_2);
+
+        return 0;
+}
+void gr_gv11b_init_gfxp_wfi_timeout_count(struct gk20a *g)
+{
+        struct gr_gk20a *gr = &g->gr;
+
+        gr->gfxp_wfi_timeout_count = GFXP_WFI_TIMEOUT_COUNT_IN_USEC_DEFAULT;
+}
+
+unsigned long gr_gv11b_get_max_gfxp_wfi_timeout_count(struct gk20a *g)
+{
+        /* 100 msec in usec count */
+        return (100 * 1000UL);
+}
diff --git a/drivers/gpu/nvgpu/gv11b/gr_gv11b.h b/drivers/gpu/nvgpu/gv11b/gr_gv11b.h
index 1941f239..7c56f62d 100644
--- a/drivers/gpu/nvgpu/gv11b/gr_gv11b.h
+++ b/drivers/gpu/nvgpu/gv11b/gr_gv11b.h
@@ -213,5 +213,8 @@ void gv11b_gr_egpc_etpc_priv_addr_table(struct gk20a *g, u32 addr,
                  u32 gpc, u32 broadcast_flags, u32 *priv_addr_table, u32 *t);
 u32 gv11b_gr_get_egpc_base(struct gk20a *g);
 void gr_gv11b_init_gpc_mmu(struct gk20a *g);
+int gr_gv11b_init_preemption_state(struct gk20a *g);
+void gr_gv11b_init_gfxp_wfi_timeout_count(struct gk20a *g);
+unsigned long gr_gv11b_get_max_gfxp_wfi_timeout_count(struct gk20a *g);
 
 #endif
diff --git a/drivers/gpu/nvgpu/gv11b/hal_gv11b.c b/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
index 5649d758..e6cf0e62 100644
--- a/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
@@ -361,7 +361,7 @@ static const struct gpu_ops gv11b_ops = {
                 .set_czf_bypass = NULL,
                 .pre_process_sm_exception = gr_gv11b_pre_process_sm_exception,
                 .set_preemption_buffer_va = gr_gv11b_set_preemption_buffer_va,
-                .init_preemption_state = NULL,
+                .init_preemption_state = gr_gv11b_init_preemption_state,
                 .update_boosted_ctx = gr_gp10b_update_boosted_ctx,
                 .set_bes_crop_debug3 = gr_gp10b_set_bes_crop_debug3,
                 .create_gr_sysfs = gr_gv11b_create_sysfs,
@@ -388,6 +388,10 @@ static const struct gpu_ops gv11b_ops = {
                         gr_gv11b_handle_tpc_sm_ecc_exception,
                 .decode_egpc_addr = gv11b_gr_decode_egpc_addr,
                 .init_ctxsw_hdr_data = gr_gp10b_init_ctxsw_hdr_data,
+                .init_gfxp_wfi_timeout_count =
+                                gr_gv11b_init_gfxp_wfi_timeout_count,
+                .get_max_gfxp_wfi_timeout_count =
+                        gr_gv11b_get_max_gfxp_wfi_timeout_count,
         },
         .fb = {
                 .reset = gv11b_fb_reset,
diff --git a/drivers/gpu/nvgpu/include/nvgpu/hw/gv11b/hw_gr_gv11b.h b/drivers/gpu/nvgpu/include/nvgpu/hw/gv11b/hw_gr_gv11b.h
index 692b7ba3..805d8b0e 100644
--- a/drivers/gpu/nvgpu/include/nvgpu/hw/gv11b/hw_gr_gv11b.h
+++ b/drivers/gpu/nvgpu/include/nvgpu/hw/gv11b/hw_gr_gv11b.h
@@ -4292,6 +4292,10 @@ static inline u32 gr_fe_gfxp_wfi_timeout_count_disabled_f(void)
 {
         return 0x0U;
 }
+static inline u32 gr_fe_gfxp_wfi_timeout_count_init_f(void)
+{
+        return 0x800U;
+}
 static inline u32 gr_gpcs_tpcs_sm_texio_control_r(void)
 {
         return 0x00419bd8U;
@@ -4936,4 +4940,20 @@ static inline u32 gr_fecs_falcon_ecc_uncorrected_err_count_unique_total_v(u32 r)
 {
         return (r >> 16U) & 0xffffU;
 }
+static inline u32 gr_debug_2_r(void)
+{
+        return 0x00400088U;
+}
+static inline u32 gr_debug_2_gfxp_wfi_timeout_unit_m(void)
+{
+        return 0x1U << 27U;
+}
+static inline u32 gr_debug_2_gfxp_wfi_timeout_unit_usec_f(void)
+{
+        return 0x0U;
+}
+static inline u32 gr_debug_2_gfxp_wfi_timeout_unit_sysclk_f(void)
+{
+        return 0x8000000U;
+}
 #endif