[PATCH 5.14 014/151] drm/nouveau/ga102-: support ttm buffer moves via copy engine

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

From: Ben Skeggs <bskeggs@redhat.com>

commit 49b2dfc081826874705b27f7970631319628ee7f upstream.

We don't currently have any kind of real acceleration on Ampere GPUs,
but the TTM memcpy() fallback paths aren't really designed to handle
copies between different devices, such as on Optimus systems, and
result in a kernel OOPS.

A few options were investigated to try and fix this, but didn't work
out, and likely would have resulted in a very unpleasant experience
for users anyway.

This commit adds just enough support for setting up a single channel
connected to a copy engine, which the kernel can use to accelerate
the buffer copies between devices.  Userspace has no access to this
incomplete channel support, but it's suitable for TTM's needs.

A more complete implementation of host(fifo) for Ampere GPUs is in
the works, but the required changes are far too invasive that they
would be unsuitable to backport to fix this issue on current kernels.

v2: fix GPFIFO length in RAMFC (reported by Karol)

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Cc: Lyude Paul <lyude@redhat.com>
Cc: Karol Herbst <kherbst@redhat.com>
Cc: <stable@vger.kernel.org> # v5.12+
Reviewed-by: Karol Herbst <kherbst@redhat.com>
Tested-by: Karol Herbst <kherbst@redhat.com>
Signed-off-by: Karol Herbst <kherbst@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20210916220406.666454-1-skeggsb@gmail.com
Signed-off-by: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
---
 drivers/gpu/drm/nouveau/include/nvif/class.h       |    2 
 drivers/gpu/drm/nouveau/include/nvkm/engine/fifo.h |    1 
 drivers/gpu/drm/nouveau/nouveau_bo.c               |    1 
 drivers/gpu/drm/nouveau/nouveau_chan.c             |    6 
 drivers/gpu/drm/nouveau/nouveau_drm.c              |    4 
 drivers/gpu/drm/nouveau/nv84_fence.c               |    2 
 drivers/gpu/drm/nouveau/nvkm/engine/device/base.c  |    3 
 drivers/gpu/drm/nouveau/nvkm/engine/fifo/Kbuild    |    1 
 drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga102.c   |  308 +++++++++++++++++++++
 drivers/gpu/drm/nouveau/nvkm/subdev/top/ga100.c    |    7 
 10 files changed, 329 insertions(+), 6 deletions(-)
 create mode 100644 drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga102.c

--- a/drivers/gpu/drm/nouveau/include/nvif/class.h
+++ b/drivers/gpu/drm/nouveau/include/nvif/class.h
@@ -71,6 +71,7 @@
 #define PASCAL_CHANNEL_GPFIFO_A                       /* cla06f.h */ 0x0000c06f
 #define VOLTA_CHANNEL_GPFIFO_A                        /* clc36f.h */ 0x0000c36f
 #define TURING_CHANNEL_GPFIFO_A                       /* clc36f.h */ 0x0000c46f
+#define AMPERE_CHANNEL_GPFIFO_B                       /* clc36f.h */ 0x0000c76f
 
 #define NV50_DISP                                     /* cl5070.h */ 0x00005070
 #define G82_DISP                                      /* cl5070.h */ 0x00008270
@@ -200,6 +201,7 @@
 #define PASCAL_DMA_COPY_B                                            0x0000c1b5
 #define VOLTA_DMA_COPY_A                                             0x0000c3b5
 #define TURING_DMA_COPY_A                                            0x0000c5b5
+#define AMPERE_DMA_COPY_B                                            0x0000c7b5
 
 #define FERMI_DECOMPRESS                                             0x000090b8
 
--- a/drivers/gpu/drm/nouveau/include/nvkm/engine/fifo.h
+++ b/drivers/gpu/drm/nouveau/include/nvkm/engine/fifo.h
@@ -77,4 +77,5 @@ int gp100_fifo_new(struct nvkm_device *,
 int gp10b_fifo_new(struct nvkm_device *, enum nvkm_subdev_type, int inst, struct nvkm_fifo **);
 int gv100_fifo_new(struct nvkm_device *, enum nvkm_subdev_type, int inst, struct nvkm_fifo **);
 int tu102_fifo_new(struct nvkm_device *, enum nvkm_subdev_type, int inst, struct nvkm_fifo **);
+int ga102_fifo_new(struct nvkm_device *, enum nvkm_subdev_type, int inst, struct nvkm_fifo **);
 #endif
--- a/drivers/gpu/drm/nouveau/nouveau_bo.c
+++ b/drivers/gpu/drm/nouveau/nouveau_bo.c
@@ -844,6 +844,7 @@ nouveau_bo_move_init(struct nouveau_drm
 			    struct ttm_resource *, struct ttm_resource *);
 		int (*init)(struct nouveau_channel *, u32 handle);
 	} _methods[] = {
+		{  "COPY", 4, 0xc7b5, nve0_bo_move_copy, nve0_bo_move_init },
 		{  "COPY", 4, 0xc5b5, nve0_bo_move_copy, nve0_bo_move_init },
 		{  "GRCE", 0, 0xc5b5, nve0_bo_move_copy, nvc0_bo_move_init },
 		{  "COPY", 4, 0xc3b5, nve0_bo_move_copy, nve0_bo_move_init },
--- a/drivers/gpu/drm/nouveau/nouveau_chan.c
+++ b/drivers/gpu/drm/nouveau/nouveau_chan.c
@@ -250,7 +250,8 @@ static int
 nouveau_channel_ind(struct nouveau_drm *drm, struct nvif_device *device,
 		    u64 runlist, bool priv, struct nouveau_channel **pchan)
 {
-	static const u16 oclasses[] = { TURING_CHANNEL_GPFIFO_A,
+	static const u16 oclasses[] = { AMPERE_CHANNEL_GPFIFO_B,
+					TURING_CHANNEL_GPFIFO_A,
 					VOLTA_CHANNEL_GPFIFO_A,
 					PASCAL_CHANNEL_GPFIFO_A,
 					MAXWELL_CHANNEL_GPFIFO_A,
@@ -386,7 +387,8 @@ nouveau_channel_init(struct nouveau_chan
 
 	nvif_object_map(&chan->user, NULL, 0);
 
-	if (chan->user.oclass >= FERMI_CHANNEL_GPFIFO) {
+	if (chan->user.oclass >= FERMI_CHANNEL_GPFIFO &&
+	    chan->user.oclass < AMPERE_CHANNEL_GPFIFO_B) {
 		ret = nvif_notify_ctor(&chan->user, "abi16ChanKilled",
 				       nouveau_channel_killed,
 				       true, NV906F_V0_NTFY_KILLED,
--- a/drivers/gpu/drm/nouveau/nouveau_drm.c
+++ b/drivers/gpu/drm/nouveau/nouveau_drm.c
@@ -345,6 +345,9 @@ nouveau_accel_gr_init(struct nouveau_drm
 	u32 arg0, arg1;
 	int ret;
 
+	if (device->info.family >= NV_DEVICE_INFO_V0_AMPERE)
+		return;
+
 	/* Allocate channel that has access to the graphics engine. */
 	if (device->info.family >= NV_DEVICE_INFO_V0_KEPLER) {
 		arg0 = nvif_fifo_runlist(device, NV_DEVICE_HOST_RUNLIST_ENGINES_GR);
@@ -469,6 +472,7 @@ nouveau_accel_init(struct nouveau_drm *d
 		case PASCAL_CHANNEL_GPFIFO_A:
 		case VOLTA_CHANNEL_GPFIFO_A:
 		case TURING_CHANNEL_GPFIFO_A:
+		case AMPERE_CHANNEL_GPFIFO_B:
 			ret = nvc0_fence_create(drm);
 			break;
 		default:
--- a/drivers/gpu/drm/nouveau/nv84_fence.c
+++ b/drivers/gpu/drm/nouveau/nv84_fence.c
@@ -204,7 +204,7 @@ nv84_fence_create(struct nouveau_drm *dr
 	priv->base.context_new = nv84_fence_context_new;
 	priv->base.context_del = nv84_fence_context_del;
 
-	priv->base.uevent = true;
+	priv->base.uevent = drm->client.device.info.family < NV_DEVICE_INFO_V0_AMPERE;
 
 	mutex_init(&priv->mutex);
 
--- a/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c
@@ -2602,6 +2602,7 @@ nv172_chipset = {
 	.top      = { 0x00000001, ga100_top_new },
 	.disp     = { 0x00000001, ga102_disp_new },
 	.dma      = { 0x00000001, gv100_dma_new },
+	.fifo     = { 0x00000001, ga102_fifo_new },
 };
 
 static const struct nvkm_device_chip
@@ -2622,6 +2623,7 @@ nv174_chipset = {
 	.top      = { 0x00000001, ga100_top_new },
 	.disp     = { 0x00000001, ga102_disp_new },
 	.dma      = { 0x00000001, gv100_dma_new },
+	.fifo     = { 0x00000001, ga102_fifo_new },
 };
 
 static const struct nvkm_device_chip
@@ -2642,6 +2644,7 @@ nv177_chipset = {
 	.top      = { 0x00000001, ga100_top_new },
 	.disp     = { 0x00000001, ga102_disp_new },
 	.dma      = { 0x00000001, gv100_dma_new },
+	.fifo     = { 0x00000001, ga102_fifo_new },
 };
 
 static int
--- a/drivers/gpu/drm/nouveau/nvkm/engine/fifo/Kbuild
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/Kbuild
@@ -18,6 +18,7 @@ nvkm-y += nvkm/engine/fifo/gp100.o
 nvkm-y += nvkm/engine/fifo/gp10b.o
 nvkm-y += nvkm/engine/fifo/gv100.o
 nvkm-y += nvkm/engine/fifo/tu102.o
+nvkm-y += nvkm/engine/fifo/ga102.o
 
 nvkm-y += nvkm/engine/fifo/chan.o
 nvkm-y += nvkm/engine/fifo/channv50.o
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/fifo/ga102.c
@@ -0,0 +1,308 @@
+/*
+ * Copyright 2021 Red Hat Inc.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+#define ga102_fifo(p) container_of((p), struct ga102_fifo, base.engine)
+#define ga102_chan(p) container_of((p), struct ga102_chan, object)
+#include <engine/fifo.h>
+#include "user.h"
+
+#include <core/memory.h>
+#include <subdev/mmu.h>
+#include <subdev/timer.h>
+#include <subdev/top.h>
+
+#include <nvif/cl0080.h>
+#include <nvif/clc36f.h>
+#include <nvif/class.h>
+
+struct ga102_fifo {
+	struct nvkm_fifo base;
+};
+
+struct ga102_chan {
+	struct nvkm_object object;
+
+	struct {
+		u32 runl;
+		u32 chan;
+	} ctrl;
+
+	struct nvkm_memory *mthd;
+	struct nvkm_memory *inst;
+	struct nvkm_memory *user;
+	struct nvkm_memory *runl;
+
+	struct nvkm_vmm *vmm;
+};
+
+static int
+ga102_chan_sclass(struct nvkm_object *object, int index, struct nvkm_oclass *oclass)
+{
+	if (index == 0) {
+		oclass->ctor = nvkm_object_new;
+		oclass->base = (struct nvkm_sclass) { -1, -1, AMPERE_DMA_COPY_B };
+		return 0;
+	}
+
+	return -EINVAL;
+}
+
+static int
+ga102_chan_map(struct nvkm_object *object, void *argv, u32 argc,
+	       enum nvkm_object_map *type, u64 *addr, u64 *size)
+{
+	struct ga102_chan *chan = ga102_chan(object);
+	struct nvkm_device *device = chan->object.engine->subdev.device;
+	u64 bar2 = nvkm_memory_bar2(chan->user);
+
+	if (bar2 == ~0ULL)
+		return -EFAULT;
+
+	*type = NVKM_OBJECT_MAP_IO;
+	*addr = device->func->resource_addr(device, 3) + bar2;
+	*size = 0x1000;
+	return 0;
+}
+
+static int
+ga102_chan_fini(struct nvkm_object *object, bool suspend)
+{
+	struct ga102_chan *chan = ga102_chan(object);
+	struct nvkm_device *device = chan->object.engine->subdev.device;
+
+	nvkm_wr32(device, chan->ctrl.chan, 0x00000003);
+
+	nvkm_wr32(device, chan->ctrl.runl + 0x098, 0x01000000);
+	nvkm_msec(device, 2000,
+		if (!(nvkm_rd32(device, chan->ctrl.runl + 0x098) & 0x00100000))
+			break;
+	);
+
+	nvkm_wr32(device, chan->ctrl.runl + 0x088, 0);
+
+	nvkm_wr32(device, chan->ctrl.chan, 0xffffffff);
+	return 0;
+}
+
+static int
+ga102_chan_init(struct nvkm_object *object)
+{
+	struct ga102_chan *chan = ga102_chan(object);
+	struct nvkm_device *device = chan->object.engine->subdev.device;
+
+	nvkm_mask(device, chan->ctrl.runl + 0x300, 0x80000000, 0x80000000);
+
+	nvkm_wr32(device, chan->ctrl.runl + 0x080, lower_32_bits(nvkm_memory_addr(chan->runl)));
+	nvkm_wr32(device, chan->ctrl.runl + 0x084, upper_32_bits(nvkm_memory_addr(chan->runl)));
+	nvkm_wr32(device, chan->ctrl.runl + 0x088, 2);
+
+	nvkm_wr32(device, chan->ctrl.chan, 0x00000002);
+	nvkm_wr32(device, chan->ctrl.runl + 0x0090, 0);
+	return 0;
+}
+
+static void *
+ga102_chan_dtor(struct nvkm_object *object)
+{
+	struct ga102_chan *chan = ga102_chan(object);
+
+	if (chan->vmm) {
+		nvkm_vmm_part(chan->vmm, chan->inst);
+		nvkm_vmm_unref(&chan->vmm);
+	}
+
+	nvkm_memory_unref(&chan->runl);
+	nvkm_memory_unref(&chan->user);
+	nvkm_memory_unref(&chan->inst);
+	nvkm_memory_unref(&chan->mthd);
+	return chan;
+}
+
+static const struct nvkm_object_func
+ga102_chan = {
+	.dtor = ga102_chan_dtor,
+	.init = ga102_chan_init,
+	.fini = ga102_chan_fini,
+	.map = ga102_chan_map,
+	.sclass = ga102_chan_sclass,
+};
+
+static int
+ga102_chan_new(struct nvkm_device *device,
+	       const struct nvkm_oclass *oclass, void *argv, u32 argc, struct nvkm_object **pobject)
+{
+	struct volta_channel_gpfifo_a_v0 *args = argv;
+	struct nvkm_top_device *tdev;
+	struct nvkm_vmm *vmm;
+	struct ga102_chan *chan;
+	int ret;
+
+	if (argc != sizeof(*args))
+		return -ENOSYS;
+
+	vmm = nvkm_uvmm_search(oclass->client, args->vmm);
+	if (IS_ERR(vmm))
+		return PTR_ERR(vmm);
+
+	if (!(chan = kzalloc(sizeof(*chan), GFP_KERNEL)))
+		return -ENOMEM;
+
+	nvkm_object_ctor(&ga102_chan, oclass, &chan->object);
+	*pobject = &chan->object;
+
+	list_for_each_entry(tdev, &device->top->device, head) {
+		if (tdev->type == NVKM_ENGINE_CE) {
+			chan->ctrl.runl = tdev->runlist;
+			break;
+		}
+	}
+
+	if (!chan->ctrl.runl)
+		return -ENODEV;
+
+	chan->ctrl.chan = nvkm_rd32(device, chan->ctrl.runl + 0x004) & 0xfffffff0;
+	args->token = nvkm_rd32(device, chan->ctrl.runl + 0x008) & 0xffff0000;
+
+	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 0x1000, true, &chan->mthd);
+	if (ret)
+		return ret;
+
+	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 0x1000, true, &chan->inst);
+	if (ret)
+		return ret;
+
+	nvkm_kmap(chan->inst);
+	nvkm_wo32(chan->inst, 0x010, 0x0000face);
+	nvkm_wo32(chan->inst, 0x030, 0x7ffff902);
+	nvkm_wo32(chan->inst, 0x048, lower_32_bits(args->ioffset));
+	nvkm_wo32(chan->inst, 0x04c, upper_32_bits(args->ioffset) |
+				     (order_base_2(args->ilength / 8) << 16));
+	nvkm_wo32(chan->inst, 0x084, 0x20400000);
+	nvkm_wo32(chan->inst, 0x094, 0x30000001);
+	nvkm_wo32(chan->inst, 0x0ac, 0x00020000);
+	nvkm_wo32(chan->inst, 0x0e4, 0x00000000);
+	nvkm_wo32(chan->inst, 0x0e8, 0);
+	nvkm_wo32(chan->inst, 0x0f4, 0x00001000);
+	nvkm_wo32(chan->inst, 0x0f8, 0x10003080);
+	nvkm_mo32(chan->inst, 0x218, 0x00000000, 0x00000000);
+	nvkm_wo32(chan->inst, 0x220, lower_32_bits(nvkm_memory_bar2(chan->mthd)));
+	nvkm_wo32(chan->inst, 0x224, upper_32_bits(nvkm_memory_bar2(chan->mthd)));
+	nvkm_done(chan->inst);
+
+	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 0x1000, true, &chan->user);
+	if (ret)
+		return ret;
+
+	ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 0x1000, true, &chan->runl);
+	if (ret)
+		return ret;
+
+	nvkm_kmap(chan->runl);
+	nvkm_wo32(chan->runl, 0x00, 0x80030001);
+	nvkm_wo32(chan->runl, 0x04, 1);
+	nvkm_wo32(chan->runl, 0x08, 0);
+	nvkm_wo32(chan->runl, 0x0c, 0x00000000);
+	nvkm_wo32(chan->runl, 0x10, lower_32_bits(nvkm_memory_addr(chan->user)));
+	nvkm_wo32(chan->runl, 0x14, upper_32_bits(nvkm_memory_addr(chan->user)));
+	nvkm_wo32(chan->runl, 0x18, lower_32_bits(nvkm_memory_addr(chan->inst)));
+	nvkm_wo32(chan->runl, 0x1c, upper_32_bits(nvkm_memory_addr(chan->inst)));
+	nvkm_done(chan->runl);
+
+	ret = nvkm_vmm_join(vmm, chan->inst);
+	if (ret)
+		return ret;
+
+	chan->vmm = nvkm_vmm_ref(vmm);
+	return 0;
+}
+
+static const struct nvkm_device_oclass
+ga102_chan_oclass = {
+	.ctor = ga102_chan_new,
+};
+
+static int
+ga102_user_new(struct nvkm_device *device,
+	       const struct nvkm_oclass *oclass, void *argv, u32 argc, struct nvkm_object **pobject)
+{
+	return tu102_fifo_user_new(oclass, argv, argc, pobject);
+}
+
+static const struct nvkm_device_oclass
+ga102_user_oclass = {
+	.ctor = ga102_user_new,
+};
+
+static int
+ga102_fifo_sclass(struct nvkm_oclass *oclass, int index, const struct nvkm_device_oclass **class)
+{
+	if (index == 0) {
+		oclass->base = (struct nvkm_sclass) { -1, -1, VOLTA_USERMODE_A };
+		*class = &ga102_user_oclass;
+		return 0;
+	} else
+	if (index == 1) {
+		oclass->base = (struct nvkm_sclass) { 0, 0, AMPERE_CHANNEL_GPFIFO_B };
+		*class = &ga102_chan_oclass;
+		return 0;
+	}
+
+	return 2;
+}
+
+static int
+ga102_fifo_info(struct nvkm_engine *engine, u64 mthd, u64 *data)
+{
+	switch (mthd) {
+	case NV_DEVICE_HOST_CHANNELS: *data = 1; return 0;
+	default:
+		break;
+	}
+
+	return -ENOSYS;
+}
+
+static void *
+ga102_fifo_dtor(struct nvkm_engine *engine)
+{
+	return ga102_fifo(engine);
+}
+
+static const struct nvkm_engine_func
+ga102_fifo = {
+	.dtor = ga102_fifo_dtor,
+	.info = ga102_fifo_info,
+	.base.sclass = ga102_fifo_sclass,
+};
+
+int
+ga102_fifo_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst,
+	       struct nvkm_fifo **pfifo)
+{
+	struct ga102_fifo *fifo;
+
+	if (!(fifo = kzalloc(sizeof(*fifo), GFP_KERNEL)))
+		return -ENOMEM;
+
+	nvkm_engine_ctor(&ga102_fifo, device, type, inst, true, &fifo->base.engine);
+	*pfifo = &fifo->base;
+	return 0;
+}
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/top/ga100.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/top/ga100.c
@@ -54,7 +54,7 @@ ga100_top_oneinit(struct nvkm_top *top)
 			info->reset   = (data & 0x0000001f);
 			break;
 		case 2:
-			info->runlist = (data & 0x0000fc00) >> 10;
+			info->runlist = (data & 0x00fffc00);
 			info->engine  = (data & 0x00000003);
 			break;
 		default:
@@ -85,9 +85,10 @@ ga100_top_oneinit(struct nvkm_top *top)
 		}
 
 		nvkm_debug(subdev, "%02x.%d (%8s): addr %06x fault %2d "
-				   "runlist %2d engine %2d reset %2d\n", type, inst,
+				   "runlist %6x engine %2d reset %2d\n", type, inst,
 			   info->type == NVKM_SUBDEV_NR ? "????????" : nvkm_subdev_type[info->type],
-			   info->addr, info->fault, info->runlist, info->engine, info->reset);
+			   info->addr, info->fault, info->runlist < 0 ? 0 : info->runlist,
+			   info->engine, info->reset);
 		info = NULL;
 	}