drm/nouveau/fb/gf100-: allocate mmu debug buffers

Later chipsets require setting this up both in FB and GR, so let's just
move the allocation to FB.

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>

drivers/gpu/drm/nouveau/include/nvkm/subdev/fb.h

@@ -55,6 +55,9 @@ struct nvkm_fb {
 		struct nvkm_fb_tile region[16];
 		int regions;
 	} tile;
+
+	struct nvkm_memory *mmu_rd;
+	struct nvkm_memory *mmu_wr;
 };
 
 bool nvkm_fb_memtype_valid(struct nvkm_fb *, u32 memtype);

drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.c

@@ -1616,30 +1616,10 @@ gf100_gr_oneinit(struct nvkm_gr *base)
 {
 	struct gf100_gr *gr = gf100_gr(base);
 	struct nvkm_device *device = gr->base.engine.subdev.device;
-	int ret, i, j;
+	int i, j;
 
 	nvkm_pmu_pgob(device->pmu, false);
 
-	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 256, false,
-			      &gr->unk4188b4);
-	if (ret)
-		return ret;
-
-	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 256, false,
-			      &gr->unk4188b8);
-	if (ret)
-		return ret;
-
-	nvkm_kmap(gr->unk4188b4);
-	for (i = 0; i < 0x1000; i += 4)
-		nvkm_wo32(gr->unk4188b4, i, 0x00000010);
-	nvkm_done(gr->unk4188b4);
-
-	nvkm_kmap(gr->unk4188b8);
-	for (i = 0; i < 0x1000; i += 4)
-		nvkm_wo32(gr->unk4188b8, i, 0x00000010);
-	nvkm_done(gr->unk4188b8);
-
 	gr->rop_nr = gr->func->rops(gr);
 	gr->gpc_nr = nvkm_rd32(device, 0x409604) & 0x0000001f;
 	for (i = 0; i < gr->gpc_nr; i++) {
@@ -1736,8 +1716,6 @@ gf100_gr_dtor(struct nvkm_gr *base)
 	gf100_gr_dtor_init(gr->fuc_sw_ctx);
 	gf100_gr_dtor_init(gr->fuc_sw_nonctx);
 
-	nvkm_memory_del(&gr->unk4188b8);
-	nvkm_memory_del(&gr->unk4188b4);
 	return gr;
 }
 
@@ -1822,6 +1800,7 @@ int
 gf100_gr_init(struct gf100_gr *gr)
 {
 	struct nvkm_device *device = gr->base.engine.subdev.device;
+	struct nvkm_fb *fb = device->fb;
 	const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total);
 	u32 data[TPC_MAX / 8] = {};
 	u8  tpcnr[GPC_MAX];
@@ -1834,8 +1813,8 @@ gf100_gr_init(struct gf100_gr *gr)
 	nvkm_wr32(device, GPC_BCAST(0x088c), 0x00000000);
 	nvkm_wr32(device, GPC_BCAST(0x0890), 0x00000000);
 	nvkm_wr32(device, GPC_BCAST(0x0894), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(gr->unk4188b4) >> 8);
-	nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(gr->unk4188b8) >> 8);
+	nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(fb->mmu_wr) >> 8);
+	nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(fb->mmu_rd) >> 8);
 
 	gf100_gr_mmio(gr, gr->func->mmio);
 

drivers/gpu/drm/nouveau/nvkm/engine/gr/gf100.h

@@ -101,9 +101,6 @@ struct gf100_gr {
 	u8 ppc_mask[GPC_MAX];
 	u8 ppc_tpc_nr[GPC_MAX][4];
 
-	struct nvkm_memory *unk4188b4;
-	struct nvkm_memory *unk4188b8;
-
 	struct gf100_gr_data mmio_data[4];
 	struct gf100_gr_mmio mmio_list[4096/8];
 	u32  size;

drivers/gpu/drm/nouveau/nvkm/engine/gr/gk104.c

@@ -24,6 +24,8 @@
 #include "gf100.h"
 #include "ctxgf100.h"
 
+#include <subdev/fb.h>
+
 #include <nvif/class.h>
 
 /*******************************************************************************
@@ -181,6 +183,7 @@ int
 gk104_gr_init(struct gf100_gr *gr)
 {
 	struct nvkm_device *device = gr->base.engine.subdev.device;
+	struct nvkm_fb *fb = device->fb;
 	const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total);
 	u32 data[TPC_MAX / 8] = {};
 	u8  tpcnr[GPC_MAX];
@@ -193,8 +196,8 @@ gk104_gr_init(struct gf100_gr *gr)
 	nvkm_wr32(device, GPC_BCAST(0x088c), 0x00000000);
 	nvkm_wr32(device, GPC_BCAST(0x0890), 0x00000000);
 	nvkm_wr32(device, GPC_BCAST(0x0894), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(gr->unk4188b4) >> 8);
-	nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(gr->unk4188b8) >> 8);
+	nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(fb->mmu_wr) >> 8);
+	nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(fb->mmu_rd) >> 8);
 
 	gf100_gr_mmio(gr, gr->func->mmio);
 

drivers/gpu/drm/nouveau/nvkm/engine/gr/gk20a.c

@@ -22,6 +22,7 @@
 #include "gf100.h"
 #include "ctxgf100.h"
 
+#include <subdev/fb.h>
 #include <subdev/timer.h>
 
 #include <nvif/class.h>
@@ -219,6 +220,7 @@ int
 gk20a_gr_init(struct gf100_gr *gr)
 {
 	struct nvkm_device *device = gr->base.engine.subdev.device;
+	struct nvkm_fb *fb = device->fb;
 	const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total);
 	u32 data[TPC_MAX / 8] = {};
 	u8  tpcnr[GPC_MAX];
@@ -239,8 +241,8 @@ gk20a_gr_init(struct gf100_gr *gr)
 		return ret;
 
 	/* MMU debug buffer */
-	nvkm_wr32(device, 0x100cc8, nvkm_memory_addr(gr->unk4188b4) >> 8);
-	nvkm_wr32(device, 0x100ccc, nvkm_memory_addr(gr->unk4188b8) >> 8);
+	nvkm_wr32(device, 0x100cc8, nvkm_memory_addr(fb->mmu_wr) >> 8);
+	nvkm_wr32(device, 0x100ccc, nvkm_memory_addr(fb->mmu_rd) >> 8);
 
 	if (gr->func->init_gpc_mmu)
 		gr->func->init_gpc_mmu(gr);

drivers/gpu/drm/nouveau/nvkm/engine/gr/gm107.c

@@ -26,6 +26,7 @@
 
 #include <subdev/bios.h>
 #include <subdev/bios/P0260.h>
+#include <subdev/fb.h>
 
 #include <nvif/class.h>
 
@@ -311,6 +312,7 @@ int
 gm107_gr_init(struct gf100_gr *gr)
 {
 	struct nvkm_device *device = gr->base.engine.subdev.device;
+	struct nvkm_fb *fb = device->fb;
 	const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total);
 	u32 data[TPC_MAX / 8] = {};
 	u8  tpcnr[GPC_MAX];
@@ -320,8 +322,8 @@ gm107_gr_init(struct gf100_gr *gr)
 	nvkm_wr32(device, GPC_BCAST(0x0880), 0x00000000);
 	nvkm_wr32(device, GPC_BCAST(0x0890), 0x00000000);
 	nvkm_wr32(device, GPC_BCAST(0x0894), 0x00000000);
-	nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(gr->unk4188b4) >> 8);
-	nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(gr->unk4188b8) >> 8);
+	nvkm_wr32(device, GPC_BCAST(0x08b4), nvkm_memory_addr(fb->mmu_wr) >> 8);
+	nvkm_wr32(device, GPC_BCAST(0x08b8), nvkm_memory_addr(fb->mmu_rd) >> 8);
 
 	gf100_gr_mmio(gr, gr->func->mmio);
 

drivers/gpu/drm/nouveau/nvkm/engine/gr/gm200.c

@@ -24,6 +24,7 @@
 #include "gf100.h"
 #include "ctxgf100.h"
 
+#include <subdev/fb.h>
 #include <subdev/secboot.h>
 
 #include <nvif/class.h>
@@ -56,6 +57,7 @@ int
 gm200_gr_init(struct gf100_gr *gr)
 {
 	struct nvkm_device *device = gr->base.engine.subdev.device;
+	struct nvkm_fb *fb = device->fb;
 	const u32 magicgpc918 = DIV_ROUND_UP(0x00800000, gr->tpc_total);
 	u32 data[TPC_MAX / 8] = {};
 	u8  tpcnr[GPC_MAX];
@@ -63,8 +65,8 @@ gm200_gr_init(struct gf100_gr *gr)
 	int i;
 
 	/*XXX: belongs in fb */
-	nvkm_wr32(device, 0x100cc8, nvkm_memory_addr(gr->unk4188b4) >> 8);
-	nvkm_wr32(device, 0x100ccc, nvkm_memory_addr(gr->unk4188b8) >> 8);
+	nvkm_wr32(device, 0x100cc8, nvkm_memory_addr(fb->mmu_wr) >> 8);
+	nvkm_wr32(device, 0x100ccc, nvkm_memory_addr(fb->mmu_rd) >> 8);
 	nvkm_mask(device, 0x100cc4, 0x00040000, 0x00040000);
 	gr->func->init_gpc_mmu(gr);
 

drivers/gpu/drm/nouveau/nvkm/subdev/fb/base.c

@@ -24,6 +24,7 @@
 #include "priv.h"
 #include "ram.h"
 
+#include <core/memory.h>
 #include <subdev/bios.h>
 #include <subdev/bios/M0203.h>
 #include <engine/gr.h>
@@ -142,6 +143,9 @@ nvkm_fb_dtor(struct nvkm_subdev *subdev)
 	struct nvkm_fb *fb = nvkm_fb(subdev);
 	int i;
 
+	nvkm_memory_del(&fb->mmu_wr);
+	nvkm_memory_del(&fb->mmu_rd);
+
 	for (i = 0; i < fb->tile.regions; i++)
 		fb->func->tile.fini(fb, i, &fb->tile.region[i]);
 

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gf100.c

@@ -24,6 +24,9 @@
 #include "gf100.h"
 #include "ram.h"
 
+#include <core/memory.h>
+#include <core/option.h>
+
 extern const u8 gf100_pte_storage_type_map[256];
 
 bool
@@ -46,6 +49,28 @@ gf100_fb_intr(struct nvkm_fb *base)
 		nvkm_debug(subdev, "PBFB intr\n");
 }
 
+int
+gf100_fb_oneinit(struct nvkm_fb *fb)
+{
+	struct nvkm_device *device = fb->subdev.device;
+	int ret, size = 0x1000;
+
+	size = nvkm_longopt(device->cfgopt, "MmuDebugBufferSize", size);
+	size = min(size, 0x1000);
+
+	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, size, 0x1000,
+			      false, &fb->mmu_rd);
+	if (ret)
+		return ret;
+
+	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, size, 0x1000,
+			      false, &fb->mmu_wr);
+	if (ret)
+		return ret;
+
+	return 0;
+}
+
 void
 gf100_fb_init(struct nvkm_fb *base)
 {
@@ -98,6 +123,7 @@ gf100_fb_new_(const struct nvkm_fb_func *func, struct nvkm_device *device,
 static const struct nvkm_fb_func
 gf100_fb = {
 	.dtor = gf100_fb_dtor,
+	.oneinit = gf100_fb_oneinit,
 	.init = gf100_fb_init,
 	.intr = gf100_fb_intr,
 	.ram_new = gf100_ram_new,

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gk104.c

@@ -27,6 +27,7 @@
 static const struct nvkm_fb_func
 gk104_fb = {
 	.dtor = gf100_fb_dtor,
+	.oneinit = gf100_fb_oneinit,
 	.init = gf100_fb_init,
 	.intr = gf100_fb_intr,
 	.ram_new = gk104_ram_new,

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gk20a.c

@@ -30,6 +30,7 @@ gk20a_fb_init(struct nvkm_fb *fb)
 
 static const struct nvkm_fb_func
 gk20a_fb = {
+	.oneinit = gf100_fb_oneinit,
 	.init = gk20a_fb_init,
 	.memtype_valid = gf100_fb_memtype_valid,
 };

drivers/gpu/drm/nouveau/nvkm/subdev/fb/gm107.c

@@ -27,6 +27,7 @@
 static const struct nvkm_fb_func
 gm107_fb = {
 	.dtor = gf100_fb_dtor,
+	.oneinit = gf100_fb_oneinit,
 	.init = gf100_fb_init,
 	.intr = gf100_fb_intr,
 	.ram_new = gm107_ram_new,

drivers/gpu/drm/nouveau/nvkm/subdev/fb/priv.h

@@ -59,5 +59,6 @@ void nv44_fb_tile_prog(struct nvkm_fb *, int, struct nvkm_fb_tile *);
 void nv46_fb_tile_init(struct nvkm_fb *, int i, u32 addr, u32 size,
 		       u32 pitch, u32 flags, struct nvkm_fb_tile *);
 
+int gf100_fb_oneinit(struct nvkm_fb *);
 bool gf100_fb_memtype_valid(struct nvkm_fb *, u32);
 #endif