drm/nouveau/acr: use common falcon HS FW code for ACR FWs

Adds context binding and support for FWs with a bootloader to the code
that was added to load VPR scrubber HS binaries, and ports ACR over to
using all of it.

- gv100 split from gp108 to handle FW exit status differences

Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
Reviewed-by: Lyude Paul <lyude@redhat.com>

30 files changed, 648 insertions, 856 deletions

diff --git a/drivers/gpu/drm/nouveau/include/nvkm/core/falcon.h b/drivers/gpu/drm/nouveau/include/nvkm/core/falcon.h
index d8fd2cbd2872..4c8b668bacb4 100644
--- a/drivers/gpu/drm/nouveau/include/nvkm/core/falcon.h
+++ b/drivers/gpu/drm/nouveau/include/nvkm/core/falcon.h
@@ -6,6 +6,7 @@
 enum nvkm_falcon_mem {
 	IMEM,
 	DMEM,
+	EMEM,
 };
 
 static inline const char *
@@ -14,6 +15,7 @@ nvkm_falcon_mem(enum nvkm_falcon_mem mem)
 	switch (mem) {
 	case IMEM: return "imem";
 	case DMEM: return "dmem";
+	case EMEM: return "emem";
 	default:
 		WARN_ON(1);
 		return "?mem";
@@ -25,6 +27,8 @@ struct nvkm_falcon_func_pio {
 	int max;
 	void (*wr_init)(struct nvkm_falcon *, u8 port, bool sec, u32 mem_base);
 	void (*wr)(struct nvkm_falcon *, u8 port, const u8 *img, int len, u16 tag);
+	void (*rd_init)(struct nvkm_falcon *, u8 port, u32 mem_base);
+	void (*rd)(struct nvkm_falcon *, u8 port, const u8 *img, int len);
 };
 
 int nvkm_falcon_ctor(const struct nvkm_falcon_func *, struct nvkm_subdev *owner,
@@ -33,27 +37,25 @@ void nvkm_falcon_dtor(struct nvkm_falcon *);
 int nvkm_falcon_reset(struct nvkm_falcon *);
 int nvkm_falcon_pio_wr(struct nvkm_falcon *, const u8 *img, u32 img_base, u8 port,
 		       enum nvkm_falcon_mem mem_type, u32 mem_base, int len, u16 tag, bool sec);
+int nvkm_falcon_pio_rd(struct nvkm_falcon *, u8 port, enum nvkm_falcon_mem type, u32 mem_base,
+		       const u8 *img, u32 img_base, int len);
 
 int gm200_flcn_reset_wait_mem_scrubbing(struct nvkm_falcon *);
 int gm200_flcn_disable(struct nvkm_falcon *);
 int gm200_flcn_enable(struct nvkm_falcon *);
+void gm200_flcn_bind_inst(struct nvkm_falcon *, int, u64);
+int gm200_flcn_bind_stat(struct nvkm_falcon *, bool);
 extern const struct nvkm_falcon_func_pio gm200_flcn_imem_pio;
 extern const struct nvkm_falcon_func_pio gm200_flcn_dmem_pio;
 
 int gp102_flcn_reset_eng(struct nvkm_falcon *);
+extern const struct nvkm_falcon_func_pio gp102_flcn_emem_pio;
 
 void nvkm_falcon_v1_load_imem(struct nvkm_falcon *,
 			      void *, u32, u32, u16, u8, bool);
 void nvkm_falcon_v1_load_dmem(struct nvkm_falcon *, void *, u32, u32, u8);
-void nvkm_falcon_v1_read_dmem(struct nvkm_falcon *, u32, u32, u8, void *);
-void nvkm_falcon_v1_bind_context(struct nvkm_falcon *, struct nvkm_memory *);
-int nvkm_falcon_v1_wait_for_halt(struct nvkm_falcon *, u32);
-int nvkm_falcon_v1_clear_interrupt(struct nvkm_falcon *, u32);
-void nvkm_falcon_v1_set_start_addr(struct nvkm_falcon *, u32 start_addr);
 void nvkm_falcon_v1_start(struct nvkm_falcon *);
 
-void gp102_sec2_flcn_bind_context(struct nvkm_falcon *, struct nvkm_memory *);
-
 #define FLCN_PRINTK(f,l,p,fmt,a...) ({                                                          \
 	if ((f)->owner->name != (f)->name)                                                      \
 		nvkm_printk___((f)->owner, (f)->user, NV_DBG_##l, p, "%s:"fmt, (f)->name, ##a); \
@@ -70,7 +72,9 @@ struct nvkm_falcon_fw {
 	const struct nvkm_falcon_fw_func {
 		int (*signature)(struct nvkm_falcon_fw *, u32 *sig_base_src);
 		int (*reset)(struct nvkm_falcon_fw *);
+		int (*setup)(struct nvkm_falcon_fw *);
 		int (*load)(struct nvkm_falcon_fw *);
+		int (*load_bld)(struct nvkm_falcon_fw *);
 		int (*boot)(struct nvkm_falcon_fw *,
 			    u32 *mbox0, u32 *mbox1, u32 mbox0_ok, u32 irqsclr);
 	} *func;
@@ -96,11 +100,14 @@ struct nvkm_falcon_fw {
 	u32 dmem_size;
 	u32 dmem_sign;
 
+	u8 *boot;
+	u32 boot_size;
 	u32 boot_addr;
 
 	struct nvkm_falcon *falcon;
 	struct nvkm_memory *inst;
 	struct nvkm_vmm *vmm;
+	struct nvkm_vma *vma;
 };
 
 int nvkm_falcon_fw_ctor(const struct nvkm_falcon_fw_func *, const char *name, struct nvkm_device *,
diff --git a/drivers/gpu/drm/nouveau/include/nvkm/core/firmware.h b/drivers/gpu/drm/nouveau/include/nvkm/core/firmware.h
index 8453025891a5..d4e507e252b1 100644
--- a/drivers/gpu/drm/nouveau/include/nvkm/core/firmware.h
+++ b/drivers/gpu/drm/nouveau/include/nvkm/core/firmware.h
@@ -1,6 +1,7 @@
 /* SPDX-License-Identifier: MIT */
 #ifndef __NVKM_FIRMWARE_H__
 #define __NVKM_FIRMWARE_H__
+#include <core/memory.h>
 #include <core/option.h>
 #include <core/subdev.h>
 
@@ -8,6 +9,7 @@ struct nvkm_firmware {
 	const struct nvkm_firmware_func {
 		enum nvkm_firmware_type {
 			NVKM_FIRMWARE_IMG_RAM,
+			NVKM_FIRMWARE_IMG_DMA,
 		} type;
 	} *func;
 	const char *name;
@@ -15,6 +17,12 @@ struct nvkm_firmware {
 
 	int len;
 	u8 *img;
+	u64 phys;
+
+	struct nvkm_firmware_mem {
+		struct nvkm_memory memory;
+		struct scatterlist sgl;
+	} mem;
 };
 
 int nvkm_firmware_ctor(const struct nvkm_firmware_func *, const char *name, struct nvkm_device *,
diff --git a/drivers/gpu/drm/nouveau/include/nvkm/engine/falcon.h b/drivers/gpu/drm/nouveau/include/nvkm/engine/falcon.h
index ca751199379e..f576ca246d10 100644
--- a/drivers/gpu/drm/nouveau/include/nvkm/engine/falcon.h
+++ b/drivers/gpu/drm/nouveau/include/nvkm/engine/falcon.h
@@ -64,9 +64,22 @@ struct nvkm_falcon_func {
 	int (*reset_wait_mem_scrubbing)(struct nvkm_falcon *);
 
 	u32 debug;
+	void (*bind_inst)(struct nvkm_falcon *, int target, u64 addr);
+	int (*bind_stat)(struct nvkm_falcon *, bool intr);
+	bool bind_intr;
+
 	const struct nvkm_falcon_func_pio *imem_pio;
 	const struct nvkm_falcon_func_pio *dmem_pio;
 
+	u32 emem_addr;
+	const struct nvkm_falcon_func_pio *emem_pio;
+
+	struct {
+		u32 head;
+		u32 tail;
+		u32 stride;
+	} cmdq, msgq;
+
 	struct {
 		u32 *data;
 		u32  size;
@@ -78,24 +91,10 @@ struct nvkm_falcon_func {
 	void (*init)(struct nvkm_falcon *);
 	void (*intr)(struct nvkm_falcon *, struct nvkm_chan *);
 
-	u32 fbif;
-
 	void (*load_imem)(struct nvkm_falcon *, void *, u32, u32, u16, u8, bool);
 	void (*load_dmem)(struct nvkm_falcon *, void *, u32, u32, u8);
-	void (*read_dmem)(struct nvkm_falcon *, u32, u32, u8, void *);
-	u32 emem_addr;
-	void (*bind_context)(struct nvkm_falcon *, struct nvkm_memory *);
-	int (*wait_for_halt)(struct nvkm_falcon *, u32);
-	int (*clear_interrupt)(struct nvkm_falcon *, u32);
-	void (*set_start_addr)(struct nvkm_falcon *, u32 start_addr);
 	void (*start)(struct nvkm_falcon *);
 
-	struct {
-		u32 head;
-		u32 tail;
-		u32 stride;
-	} cmdq, msgq;
-
 	struct nvkm_sclass sclass[];
 };
 
@@ -122,10 +121,5 @@ nvkm_falcon_mask(struct nvkm_falcon *falcon, u32 addr, u32 mask, u32 val)
 void nvkm_falcon_load_imem(struct nvkm_falcon *, void *, u32, u32, u16, u8,
 			   bool);
 void nvkm_falcon_load_dmem(struct nvkm_falcon *, void *, u32, u32, u8);
-void nvkm_falcon_read_dmem(struct nvkm_falcon *, u32, u32, u8, void *);
-void nvkm_falcon_bind_context(struct nvkm_falcon *, struct nvkm_memory *);
-void nvkm_falcon_set_start_addr(struct nvkm_falcon *, u32);
 void nvkm_falcon_start(struct nvkm_falcon *);
-int nvkm_falcon_wait_for_halt(struct nvkm_falcon *, u32);
-int nvkm_falcon_clear_interrupt(struct nvkm_falcon *, u32);
 #endif
diff --git a/drivers/gpu/drm/nouveau/include/nvkm/subdev/acr.h b/drivers/gpu/drm/nouveau/include/nvkm/subdev/acr.h
index fe5024b6d0f3..12c0b8395262 100644
--- a/drivers/gpu/drm/nouveau/include/nvkm/subdev/acr.h
+++ b/drivers/gpu/drm/nouveau/include/nvkm/subdev/acr.h
@@ -36,7 +36,7 @@ struct nvkm_acr {
 	const struct nvkm_acr_func *func;
 	struct nvkm_subdev subdev;
 
-	struct list_head hsfw, hsf;
+	struct list_head hsfw;
 	struct list_head lsfw, lsf;
 
 	u64 managed_falcons;
@@ -65,6 +65,7 @@ int gm20b_acr_new(struct nvkm_device *, enum nvkm_subdev_type, int inst, struct
 int gp102_acr_new(struct nvkm_device *, enum nvkm_subdev_type, int inst, struct nvkm_acr **);
 int gp108_acr_new(struct nvkm_device *, enum nvkm_subdev_type, int inst, struct nvkm_acr **);
 int gp10b_acr_new(struct nvkm_device *, enum nvkm_subdev_type, int inst, struct nvkm_acr **);
+int gv100_acr_new(struct nvkm_device *, enum nvkm_subdev_type, int inst, struct nvkm_acr **);
 int tu102_acr_new(struct nvkm_device *, enum nvkm_subdev_type, int inst, struct nvkm_acr **);
 
 struct nvkm_acr_lsfw {
diff --git a/drivers/gpu/drm/nouveau/nvkm/core/firmware.c b/drivers/gpu/drm/nouveau/nvkm/core/firmware.c
index aa1bf6b5a8cd..fcf2a002f6cb 100644
--- a/drivers/gpu/drm/nouveau/nvkm/core/firmware.c
+++ b/drivers/gpu/drm/nouveau/nvkm/core/firmware.c
@@ -22,6 +22,9 @@
 #include <core/device.h>
 #include <core/firmware.h>
 
+#include <subdev/fb.h>
+#include <subdev/mmu.h>
+
 int
 nvkm_firmware_load_name(const struct nvkm_subdev *subdev, const char *base,
 			const char *name, int ver, const struct firmware **pfw)
@@ -108,9 +111,70 @@ nvkm_firmware_put(const struct firmware *fw)
 	release_firmware(fw);
 }
 
+#define nvkm_firmware_mem(p) container_of((p), struct nvkm_firmware, mem.memory)
+
+static int
+nvkm_firmware_mem_map(struct nvkm_memory *memory, u64 offset, struct nvkm_vmm *vmm,
+		      struct nvkm_vma *vma, void *argv, u32 argc)
+{
+	struct nvkm_firmware *fw = nvkm_firmware_mem(memory);
+	struct nvkm_vmm_map map = {
+		.memory = &fw->mem.memory,
+		.offset = offset,
+		.sgl = &fw->mem.sgl,
+	};
+
+	if (WARN_ON(fw->func->type != NVKM_FIRMWARE_IMG_DMA))
+		return -ENOSYS;
+
+	return nvkm_vmm_map(vmm, vma, argv, argc, &map);
+}
+
+static u64
+nvkm_firmware_mem_size(struct nvkm_memory *memory)
+{
+	return sg_dma_len(&nvkm_firmware_mem(memory)->mem.sgl);
+}
+
+static u64
+nvkm_firmware_mem_addr(struct nvkm_memory *memory)
+{
+	return nvkm_firmware_mem(memory)->phys;
+}
+
+static u8
+nvkm_firmware_mem_page(struct nvkm_memory *memory)
+{
+	return PAGE_SHIFT;
+}
+
+static enum nvkm_memory_target
+nvkm_firmware_mem_target(struct nvkm_memory *memory)
+{
+	return NVKM_MEM_TARGET_HOST;
+}
+
+static void *
+nvkm_firmware_mem_dtor(struct nvkm_memory *memory)
+{
+	return NULL;
+}
+
+static const struct nvkm_memory_func
+nvkm_firmware_mem = {
+	.dtor = nvkm_firmware_mem_dtor,
+	.target = nvkm_firmware_mem_target,
+	.page = nvkm_firmware_mem_page,
+	.addr = nvkm_firmware_mem_addr,
+	.size = nvkm_firmware_mem_size,
+	.map = nvkm_firmware_mem_map,
+};
+
 void
 nvkm_firmware_dtor(struct nvkm_firmware *fw)
 {
+	struct nvkm_memory *memory = &fw->mem.memory;
+
 	if (!fw->img)
 		return;
 
@@ -118,6 +182,10 @@ nvkm_firmware_dtor(struct nvkm_firmware *fw)
 	case NVKM_FIRMWARE_IMG_RAM:
 		kfree(fw->img);
 		break;
+	case NVKM_FIRMWARE_IMG_DMA:
+		nvkm_memory_unref(&memory);
+		dma_free_coherent(fw->device->dev, sg_dma_len(&fw->mem.sgl), fw->img, fw->phys);
+		break;
 	default:
 		WARN_ON(1);
 		break;
@@ -133,12 +201,28 @@ nvkm_firmware_ctor(const struct nvkm_firmware_func *func, const char *name,
 	fw->func = func;
 	fw->name = name;
 	fw->device = device;
+	fw->len = len;
 
 	switch (fw->func->type) {
 	case NVKM_FIRMWARE_IMG_RAM:
-		fw->len = len;
 		fw->img = kmemdup(src, fw->len, GFP_KERNEL);
 		break;
+	case NVKM_FIRMWARE_IMG_DMA: {
+		dma_addr_t addr;
+
+		len = ALIGN(fw->len, PAGE_SIZE);
+
+		fw->img = dma_alloc_coherent(fw->device->dev, len, &addr, GFP_KERNEL);
+		if (fw->img) {
+			memcpy(fw->img, src, fw->len);
+			fw->phys = addr;
+		}
+
+		sg_init_one(&fw->mem.sgl, fw->img, len);
+		sg_dma_address(&fw->mem.sgl) = fw->phys;
+		sg_dma_len(&fw->mem.sgl) = len;
+	}
+		break;
 	default:
 		WARN_ON(1);
 		return -EINVAL;
@@ -147,5 +231,6 @@ nvkm_firmware_ctor(const struct nvkm_firmware_func *func, const char *name,
 	if (!fw->img)
 		return -ENOMEM;
 
+	nvkm_memory_ctor(&nvkm_firmware_mem, &fw->mem.memory);
 	return 0;
 }
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c b/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c
index c94de7b45556..931a59581815 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/device/base.c
@@ -2364,7 +2364,7 @@ nv13b_chipset = {
 static const struct nvkm_device_chip
 nv140_chipset = {
 	.name = "GV100",
-	.acr      = { 0x00000001, gp108_acr_new },
+	.acr      = { 0x00000001, gv100_acr_new },
 	.bar      = { 0x00000001, gm107_bar_new },
 	.bios     = { 0x00000001, nvkm_bios_new },
 	.bus      = { 0x00000001, gf100_bus_new },
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp102.c b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp102.c
index ae910c9bdc0a..a5c987ce58ed 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp102.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/gp102.c
@@ -190,45 +190,6 @@ gp102_sec2_intr(struct nvkm_inth *inth)
 	return IRQ_HANDLED;
 }
 
-void
-gp102_sec2_flcn_bind_context(struct nvkm_falcon *falcon,
-			     struct nvkm_memory *ctx)
-{
-	struct nvkm_device *device = falcon->owner->device;
-
-	nvkm_falcon_v1_bind_context(falcon, ctx);
-	if (!ctx)
-		return;
-
-	/* Not sure if this is a WAR for a HW issue, or some additional
-	 * programming sequence that's needed to properly complete the
-	 * context switch we trigger above.
-	 *
-	 * Fixes unreliability of booting the SEC2 RTOS on Quadro P620,
-	 * particularly when resuming from suspend.
-	 *
-	 * Also removes the need for an odd workaround where we needed
-	 * to program SEC2's FALCON_CPUCTL_ALIAS_STARTCPU twice before
-	 * the SEC2 RTOS would begin executing.
-	 */
-	nvkm_msec(device, 10,
-		u32 irqstat = nvkm_falcon_rd32(falcon, 0x008);
-		u32 flcn0dc = nvkm_falcon_rd32(falcon, 0x0dc);
-		if ((irqstat & 0x00000008) &&
-		    (flcn0dc & 0x00007000) == 0x00005000)
-			break;
-	);
-
-	nvkm_falcon_mask(falcon, 0x004, 0x00000008, 0x00000008);
-	nvkm_falcon_mask(falcon, 0x058, 0x00000002, 0x00000002);
-
-	nvkm_msec(device, 10,
-		u32 flcn0dc = nvkm_falcon_rd32(falcon, 0x0dc);
-		if ((flcn0dc & 0x00007000) == 0x00000000)
-			break;
-	);
-}
-
 static const struct nvkm_falcon_func
 gp102_sec2_flcn = {
 	.disable = gm200_flcn_disable,
@@ -237,15 +198,13 @@ gp102_sec2_flcn = {
 	.reset_eng = gp102_flcn_reset_eng,
 	.reset_wait_mem_scrubbing = gm200_flcn_reset_wait_mem_scrubbing,
 	.debug = 0x408,
-	.fbif = 0x600,
-	.load_imem = nvkm_falcon_v1_load_imem,
-	.load_dmem = nvkm_falcon_v1_load_dmem,
-	.read_dmem = nvkm_falcon_v1_read_dmem,
+	.bind_inst = gm200_flcn_bind_inst,
+	.bind_stat = gm200_flcn_bind_stat,
+	.bind_intr = true,
+	.imem_pio = &gm200_flcn_imem_pio,
+	.dmem_pio = &gm200_flcn_dmem_pio,
 	.emem_addr = 0x01000000,
-	.bind_context = gp102_sec2_flcn_bind_context,
-	.wait_for_halt = nvkm_falcon_v1_wait_for_halt,
-	.clear_interrupt = nvkm_falcon_v1_clear_interrupt,
-	.set_start_addr = nvkm_falcon_v1_set_start_addr,
+	.emem_pio = &gp102_flcn_emem_pio,
 	.start = nvkm_falcon_v1_start,
 	.cmdq = { 0xa00, 0xa04, 8 },
 	.msgq = { 0xa30, 0xa34, 8 },
diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/tu102.c b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/tu102.c
index 0f859f323504..0afc4b2fa529 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/sec2/tu102.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/sec2/tu102.c
@@ -32,15 +32,13 @@ tu102_sec2_flcn = {
 	.reset_eng = gp102_flcn_reset_eng,
 	.reset_wait_mem_scrubbing = gm200_flcn_reset_wait_mem_scrubbing,
 	.debug = 0x408,
-	.fbif = 0x600,
-	.load_imem = nvkm_falcon_v1_load_imem,
-	.load_dmem = nvkm_falcon_v1_load_dmem,
-	.read_dmem = nvkm_falcon_v1_read_dmem,
+	.bind_inst = gm200_flcn_bind_inst,
+	.bind_stat = gm200_flcn_bind_stat,
+	.bind_intr = true,
+	.imem_pio = &gm200_flcn_imem_pio,
+	.dmem_pio = &gm200_flcn_dmem_pio,
 	.emem_addr = 0x01000000,
-	.bind_context = gp102_sec2_flcn_bind_context,
-	.wait_for_halt = nvkm_falcon_v1_wait_for_halt,
-	.clear_interrupt = nvkm_falcon_v1_clear_interrupt,
-	.set_start_addr = nvkm_falcon_v1_set_start_addr,
+	.emem_pio = &gp102_flcn_emem_pio,
 	.start = nvkm_falcon_v1_start,
 	.cmdq = { 0xc00, 0xc04, 8 },
 	.msgq = { 0xc80, 0xc84, 8 },
diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/base.c b/drivers/gpu/drm/nouveau/nvkm/falcon/base.c
index c15dda020c77..e4075aa441f3 100644
--- a/drivers/gpu/drm/nouveau/nvkm/falcon/base.c
+++ b/drivers/gpu/drm/nouveau/nvkm/falcon/base.c
@@ -22,6 +22,7 @@
 #include "priv.h"
 
 #include <subdev/mc.h>
+#include <subdev/timer.h>
 #include <subdev/top.h>
 
 static const struct nvkm_falcon_func_pio *
@@ -36,12 +37,49 @@ nvkm_falcon_pio(struct nvkm_falcon *falcon, enum nvkm_falcon_mem *mem_type, u32
 
 		*mem_base -= falcon->func->emem_addr;
 		fallthrough;
+	case EMEM:
+		return falcon->func->emem_pio;
 	default:
 		return NULL;
 	}
 }
 
 int
+nvkm_falcon_pio_rd(struct nvkm_falcon *falcon, u8 port, enum nvkm_falcon_mem mem_type, u32 mem_base,
+		   const u8 *img, u32 img_base, int len)
+{
+	const struct nvkm_falcon_func_pio *pio = nvkm_falcon_pio(falcon, &mem_type, &mem_base);
+	const char *type = nvkm_falcon_mem(mem_type);
+	int xfer_len;
+
+	if (WARN_ON(!pio || !pio->rd))
+		return -EINVAL;
+
+	FLCN_DBG(falcon, "%s %08x -> %08x bytes at %08x", type, mem_base, len, img_base);
+	if (WARN_ON(!len || (len & (pio->min - 1))))
+		return -EINVAL;
+
+	pio->rd_init(falcon, port, mem_base);
+	do {
+		xfer_len = min(len, pio->max);
+		pio->rd(falcon, port, img, xfer_len);
+
+		if (nvkm_printk_ok(falcon->owner, falcon->user, NV_DBG_TRACE)) {
+			for (img_base = 0; img_base < xfer_len; img_base += 4, mem_base += 4) {
+				if (((img_base / 4) % 8) == 0)
+					printk(KERN_INFO "%s %08x ->", type, mem_base);
+				printk(KERN_CONT " %08x", *(u32 *)(img + img_base));
+			}
+		}
+
+		img += xfer_len;
+		len -= xfer_len;
+	} while (len);
+
+	return 0;
+}
+
+int
 nvkm_falcon_pio_wr(struct nvkm_falcon *falcon, const u8 *img, u32 img_base, u8 port,
 		   enum nvkm_falcon_mem mem_type, u32 mem_base, int len, u16 tag, bool sec)
 {
@@ -104,35 +142,6 @@ nvkm_falcon_load_dmem(struct nvkm_falcon *falcon, void *data, u32 start,
 }
 
 void
-nvkm_falcon_read_dmem(struct nvkm_falcon *falcon, u32 start, u32 size, u8 port,
-		      void *data)
-{
-	mutex_lock(&falcon->dmem_mutex);
-
-	falcon->func->read_dmem(falcon, start, size, port, data);
-
-	mutex_unlock(&falcon->dmem_mutex);
-}
-
-void
-nvkm_falcon_bind_context(struct nvkm_falcon *falcon, struct nvkm_memory *inst)
-{
-	if (!falcon->func->bind_context) {
-		nvkm_error(falcon->user,
-			   "Context binding not supported on this falcon!\n");
-		return;
-	}
-
-	falcon->func->bind_context(falcon, inst);
-}
-
-void
-nvkm_falcon_set_start_addr(struct nvkm_falcon *falcon, u32 start_addr)
-{
-	falcon->func->set_start_addr(falcon, start_addr);
-}
-
-void
 nvkm_falcon_start(struct nvkm_falcon *falcon)
 {
 	falcon->func->start(falcon);
@@ -150,18 +159,6 @@ nvkm_falcon_reset(struct nvkm_falcon *falcon)
 	return nvkm_falcon_enable(falcon);
 }
 
-int
-nvkm_falcon_wait_for_halt(struct nvkm_falcon *falcon, u32 ms)
-{
-	return falcon->func->wait_for_halt(falcon, ms);
-}
-
-int
-nvkm_falcon_clear_interrupt(struct nvkm_falcon *falcon, u32 mask)
-{
-	return falcon->func->clear_interrupt(falcon, mask);
-}
-
 static int
 nvkm_falcon_oneinit(struct nvkm_falcon *falcon)
 {
diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/cmdq.c b/drivers/gpu/drm/nouveau/nvkm/falcon/cmdq.c
index 44cf6a8862e1..211ebe7afac6 100644
--- a/drivers/gpu/drm/nouveau/nvkm/falcon/cmdq.c
+++ b/drivers/gpu/drm/nouveau/nvkm/falcon/cmdq.c
@@ -51,7 +51,7 @@ static void
 nvkm_falcon_cmdq_push(struct nvkm_falcon_cmdq *cmdq, void *data, u32 size)
 {
 	struct nvkm_falcon *falcon = cmdq->qmgr->falcon;
-	nvkm_falcon_load_dmem(falcon, data, cmdq->position, size, 0);
+	nvkm_falcon_pio_wr(falcon, data, 0, 0, DMEM, cmdq->position, size, 0, false);
 	cmdq->position += ALIGN(size, QUEUE_ALIGNMENT);
 }
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/fw.c b/drivers/gpu/drm/nouveau/nvkm/falcon/fw.c
index 12a899dbec35..13d52d7e4f60 100644
--- a/drivers/gpu/drm/nouveau/nvkm/falcon/fw.c
+++ b/drivers/gpu/drm/nouveau/nvkm/falcon/fw.c
@@ -93,6 +93,12 @@ nvkm_falcon_fw_boot(struct nvkm_falcon_fw *fw, struct nvkm_subdev *user,
 	fw->func->reset(fw);
 
 	FLCNFW_DBG(fw, "loading");
+	if (fw->func->setup) {
+		ret = fw->func->setup(fw);
+		if (ret)
+			goto done;
+	}
+
 	ret = fw->func->load(fw);
 	if (ret)
 		goto done;
@@ -114,15 +120,34 @@ int
 nvkm_falcon_fw_oneinit(struct nvkm_falcon_fw *fw, struct nvkm_falcon *falcon,
 		       struct nvkm_vmm *vmm, struct nvkm_memory *inst)
 {
+	int ret;
+
 	fw->falcon = falcon;
 	fw->vmm = nvkm_vmm_ref(vmm);
 	fw->inst = nvkm_memory_ref(inst);
+
+	if (fw->boot) {
+		FLCN_DBG(falcon, "mapping %s fw", fw->fw.name);
+		ret = nvkm_vmm_get(fw->vmm, 12, nvkm_memory_size(&fw->fw.mem.memory), &fw->vma);
+		if (ret) {
+			FLCN_ERR(falcon, "get %d", ret);
+			return ret;
+		}
+
+		ret = nvkm_memory_map(&fw->fw.mem.memory, 0, fw->vmm, fw->vma, NULL, 0);
+		if (ret) {
+			FLCN_ERR(falcon, "map %d", ret);
+			return ret;
+		}
+	}
+
 	return 0;
 }
 
 void
 nvkm_falcon_fw_dtor(struct nvkm_falcon_fw *fw)
 {
+	nvkm_vmm_put(fw->vmm, &fw->vma);
 	nvkm_vmm_unref(&fw->vmm);
 	nvkm_memory_unref(&fw->inst);
 	nvkm_falcon_fw_dtor_sigs(fw);
@@ -130,6 +155,11 @@ nvkm_falcon_fw_dtor(struct nvkm_falcon_fw *fw)
 }
 
 static const struct nvkm_firmware_func
+nvkm_falcon_fw_dma = {
+	.type = NVKM_FIRMWARE_IMG_DMA,
+};
+
+static const struct nvkm_firmware_func
 nvkm_falcon_fw = {
 	.type = NVKM_FIRMWARE_IMG_RAM,
 };
@@ -160,7 +190,7 @@ nvkm_falcon_fw_ctor(const struct nvkm_falcon_fw_func *func, const char *name,
 		    struct nvkm_device *device, bool dma, const void *src, u32 len,
 		    struct nvkm_falcon *falcon, struct nvkm_falcon_fw *fw)
 {
-	const struct nvkm_firmware_func *type = &nvkm_falcon_fw;
+	const struct nvkm_firmware_func *type = dma ? &nvkm_falcon_fw_dma : &nvkm_falcon_fw;
 	int ret;
 
 	fw->func = func;
@@ -181,6 +211,7 @@ nvkm_falcon_fw_ctor_hs(const struct nvkm_falcon_fw_func *func, const char *name,
 	const struct nvfw_bin_hdr *hdr;
 	const struct nvfw_hs_header *hshdr;
 	const struct nvfw_hs_load_header *lhdr;
+	const struct nvfw_bl_desc *desc;
 	u32 loc, sig;
 	int ret;
 
@@ -190,14 +221,31 @@ nvkm_falcon_fw_ctor_hs(const struct nvkm_falcon_fw_func *func, const char *name,
 
 	hdr = nvfw_bin_hdr(subdev, blob->data);
 	hshdr = nvfw_hs_header(subdev, blob->data + hdr->header_offset);
-	loc = *(u32 *)(blob->data + hshdr->patch_loc);
-	sig = *(u32 *)(blob->data + hshdr->patch_sig);
 
 	ret = nvkm_falcon_fw_ctor(func, name, subdev->device, bl != NULL,
 				  blob->data + hdr->data_offset, hdr->data_size, falcon, fw);
 	if (ret)
 		goto done;
 
+	/* Earlier FW releases by NVIDIA for Nouveau's use aren't in NVIDIA's
+	 * standard format, and don't have the indirection seen in the 0x10de
+	 * case.
+	 */
+	switch (hdr->bin_magic) {
+	case 0x000010de:
+		loc = *(u32 *)(blob->data + hshdr->patch_loc);
+		sig = *(u32 *)(blob->data + hshdr->patch_sig);
+		break;
+	case 0x3b1d14f0:
+		loc = hshdr->patch_loc;
+		sig = hshdr->patch_sig;
+		break;
+	default:
+		WARN_ON(1);
+		ret = -EINVAL;
+		goto done;
+	}
+
 	ret = nvkm_falcon_fw_sign(fw, loc, hshdr->sig_prod_size, blob->data,
 				  1, hshdr->sig_prod_offset + sig,
 				  1, hshdr->sig_dbg_offset + sig);
@@ -219,7 +267,26 @@ nvkm_falcon_fw_ctor_hs(const struct nvkm_falcon_fw_func *func, const char *name,
 	fw->dmem_size = lhdr->data_size;
 	fw->dmem_sign = loc - lhdr->data_dma_base;
 
-	fw->boot_addr = fw->nmem_base;
+	if (bl) {
+		nvkm_firmware_put(blob);
+
+		ret = nvkm_firmware_load_name(subdev, bl, "", ver, &blob);
+		if (ret)
+			return ret;
+
+		hdr = nvfw_bin_hdr(subdev, blob->data);
+		desc = nvfw_bl_desc(subdev, blob->data + hdr->header_offset);
+
+		fw->boot_addr = desc->start_tag << 8;
+		fw->boot_size = desc->code_size;
+		fw->boot = kmemdup(blob->data + hdr->data_offset + desc->code_off,
+				   fw->boot_size, GFP_KERNEL);
+		if (!fw->boot)
+			ret = -ENOMEM;
+	} else {
+		fw->boot_addr = fw->nmem_base;
+	}
+
 done:
 	if (ret)
 		nvkm_falcon_fw_dtor(fw);
diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c b/drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c
index 014ca38b8ff3..b61506776105 100644
--- a/drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c
+++ b/drivers/gpu/drm/nouveau/nvkm/falcon/gm200.c
@@ -21,10 +21,27 @@
  */
 #include "priv.h"
 
+#include <core/memory.h>
 #include <subdev/mc.h>
 #include <subdev/timer.h>
 
 static void
+gm200_flcn_pio_dmem_rd(struct nvkm_falcon *falcon, u8 port, const u8 *img, int len)
+{
+	while (len >= 4) {
+		*(u32 *)img = nvkm_falcon_rd32(falcon, 0x1c4 + (port * 8));
+		img += 4;
+		len -= 4;
+	}
+}
+
+static void
+gm200_flcn_pio_dmem_rd_init(struct nvkm_falcon *falcon, u8 port, u32 dmem_base)
+{
+	nvkm_falcon_wr32(falcon, 0x1c0 + (port * 8), BIT(25) | dmem_base);
+}
+
+static void
 gm200_flcn_pio_dmem_wr(struct nvkm_falcon *falcon, u8 port, const u8 *img, int len, u16 tag)
 {
 	while (len >= 4) {
@@ -46,6 +63,8 @@ gm200_flcn_dmem_pio = {
 	.max = 0x100,
 	.wr_init = gm200_flcn_pio_dmem_wr_init,
 	.wr = gm200_flcn_pio_dmem_wr,
+	.rd_init = gm200_flcn_pio_dmem_rd_init,
+	.rd = gm200_flcn_pio_dmem_rd,
 };
 
 static void
@@ -74,6 +93,24 @@ gm200_flcn_imem_pio = {
 };
 
 int
+gm200_flcn_bind_stat(struct nvkm_falcon *falcon, bool intr)
+{
+	if (intr && !(nvkm_falcon_rd32(falcon, 0x008) & 0x00000008))
+		return -1;
+
+	return (nvkm_falcon_rd32(falcon, 0x0dc) & 0x00007000) >> 12;
+}
+
+void
+gm200_flcn_bind_inst(struct nvkm_falcon *falcon, int target, u64 addr)
+{
+	nvkm_falcon_mask(falcon, 0x604, 0x00000007, 0x00000000); /* DMAIDX_VIRT */
+	nvkm_falcon_wr32(falcon, 0x054, (1 << 30) | (target << 28) | (addr >> 12));
+	nvkm_falcon_mask(falcon, 0x090, 0x00010000, 0x00010000);
+	nvkm_falcon_mask(falcon, 0x0a4, 0x00000008, 0x00000008);
+}
+
+int
 gm200_flcn_reset_wait_mem_scrubbing(struct nvkm_falcon *falcon)
 {
 	nvkm_falcon_mask(falcon, 0x040, 0x00000000, 0x00000000);
@@ -166,13 +203,60 @@ int
 gm200_flcn_fw_load(struct nvkm_falcon_fw *fw)
 {
 	struct nvkm_falcon *falcon = fw->falcon;
-	int ret;
+	int target, ret;
+
+	if (fw->inst) {
+		nvkm_falcon_mask(falcon, 0x048, 0x00000001, 0x00000001);
+
+		switch (nvkm_memory_target(fw->inst)) {
+		case NVKM_MEM_TARGET_VRAM: target = 0; break;
+		case NVKM_MEM_TARGET_HOST: target = 2; break;
+		case NVKM_MEM_TARGET_NCOH: target = 3; break;
+		default:
+			WARN_ON(1);
+			return -EINVAL;
+		}
+
+		falcon->func->bind_inst(falcon, target, nvkm_memory_addr(fw->inst));
+
+		if (nvkm_msec(falcon->owner->device, 10,
+			if (falcon->func->bind_stat(falcon, falcon->func->bind_intr) == 5)
+				break;
+		) < 0)
+			return -ETIMEDOUT;
 
-	if (1) {
+		nvkm_falcon_mask(falcon, 0x004, 0x00000008, 0x00000008);
+		nvkm_falcon_mask(falcon, 0x058, 0x00000002, 0x00000002);
+
+		if (nvkm_msec(falcon->owner->device, 10,
+			if (falcon->func->bind_stat(falcon, false) == 0)
+				break;
+		) < 0)
+			return -ETIMEDOUT;
+	} else {
 		nvkm_falcon_mask(falcon, 0x624, 0x00000080, 0x00000080);
 		nvkm_falcon_wr32(falcon, 0x10c, 0x00000000);
 	}
 
+	if (fw->boot) {
+		switch (nvkm_memory_target(&fw->fw.mem.memory)) {
+		case NVKM_MEM_TARGET_VRAM: target = 4; break;
+		case NVKM_MEM_TARGET_HOST: target = 5; break;
+		case NVKM_MEM_TARGET_NCOH: target = 6; break;
+		default:
+			WARN_ON(1);
+			return -EINVAL;
+		}
+
+		ret = nvkm_falcon_pio_wr(falcon, fw->boot, 0, 0,
+					 IMEM, falcon->code.limit - fw->boot_size, fw->boot_size,
+					 fw->boot_addr >> 8, false);
+		if (ret)
+			return ret;
+
+		return fw->func->load_bld(fw);
+	}
+
 	ret = nvkm_falcon_pio_wr(falcon, fw->fw.img + fw->nmem_base_img, fw->nmem_base_img, 0,
 				 IMEM, fw->nmem_base, fw->nmem_size, fw->nmem_base >> 8, false);
 	if (ret)
diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/gp102.c b/drivers/gpu/drm/nouveau/nvkm/falcon/gp102.c
index f49918530d0b..c70beacb8d30 100644
--- a/drivers/gpu/drm/nouveau/nvkm/falcon/gp102.c
+++ b/drivers/gpu/drm/nouveau/nvkm/falcon/gp102.c
@@ -21,6 +21,48 @@
  */
 #include "priv.h"
 
+static void
+gp102_flcn_pio_emem_rd(struct nvkm_falcon *falcon, u8 port, const u8 *img, int len)
+{
+	while (len >= 4) {
+		*(u32 *)img = nvkm_falcon_rd32(falcon, 0xac4 + (port * 8));
+		img += 4;
+		len -= 4;
+	}
+}
+
+static void
+gp102_flcn_pio_emem_rd_init(struct nvkm_falcon *falcon, u8 port, u32 dmem_base)
+{
+	nvkm_falcon_wr32(falcon, 0xac0 + (port * 8), BIT(25) | dmem_base);
+}
+
+static void
+gp102_flcn_pio_emem_wr(struct nvkm_falcon *falcon, u8 port, const u8 *img, int len, u16 tag)
+{
+	while (len >= 4) {
+		nvkm_falcon_wr32(falcon, 0xac4 + (port * 8), *(u32 *)img);
+		img += 4;
+		len -= 4;
+	}
+}
+
+static void
+gp102_flcn_pio_emem_wr_init(struct nvkm_falcon *falcon, u8 port, bool sec, u32 emem_base)
+{
+	nvkm_falcon_wr32(falcon, 0xac0 + (port * 8), BIT(24) | emem_base);
+}
+
+const struct nvkm_falcon_func_pio
+gp102_flcn_emem_pio = {
+	.min = 4,
+	.max = 0x100,
+	.wr_init = gp102_flcn_pio_emem_wr_init,
+	.wr = gp102_flcn_pio_emem_wr,
+	.rd_init = gp102_flcn_pio_emem_rd_init,
+	.rd = gp102_flcn_pio_emem_rd,
+};
+
 int
 gp102_flcn_reset_eng(struct nvkm_falcon *falcon)
 {
diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/msgq.c b/drivers/gpu/drm/nouveau/nvkm/falcon/msgq.c
index 04f853151648..16b246fda666 100644
--- a/drivers/gpu/drm/nouveau/nvkm/falcon/msgq.c
+++ b/drivers/gpu/drm/nouveau/nvkm/falcon/msgq.c
@@ -68,7 +68,7 @@ nvkm_falcon_msgq_pop(struct nvkm_falcon_msgq *msgq, void *data, u32 size)
 		return -EINVAL;
 	}
 
-	nvkm_falcon_read_dmem(falcon, tail, size, 0, data);
+	nvkm_falcon_pio_rd(falcon, 0, DMEM, tail, data, 0, size);
 	msgq->position += ALIGN(size, QUEUE_ALIGNMENT);
 	return 0;
 }
diff --git a/drivers/gpu/drm/nouveau/nvkm/falcon/v1.c b/drivers/gpu/drm/nouveau/nvkm/falcon/v1.c
index 9a9e1e6f70a6..dd2ddc54ac60 100644
--- a/drivers/gpu/drm/nouveau/nvkm/falcon/v1.c
+++ b/drivers/gpu/drm/nouveau/nvkm/falcon/v1.c
@@ -64,44 +64,13 @@ nvkm_falcon_v1_load_imem(struct nvkm_falcon *falcon, void *data, u32 start,
 		nvkm_falcon_wr32(falcon, 0x184 + (port * 16), 0);
 }
 
-static void
-nvkm_falcon_v1_load_emem(struct nvkm_falcon *falcon, void *data, u32 start,
-			 u32 size, u8 port)
-{
-	u8 rem = size % 4;
-	int i;
-
-	size -= rem;
-
-	nvkm_falcon_wr32(falcon, 0xac0 + (port * 8), start | (0x1 << 24));
-	for (i = 0; i < size / 4; i++)
-		nvkm_falcon_wr32(falcon, 0xac4 + (port * 8), ((u32 *)data)[i]);
-
-	/*
-	 * If size is not a multiple of 4, mask the last word to ensure garbage
-	 * does not get written
-	 */
-	if (rem) {
-		u32 extra = ((u32 *)data)[i];
-
-		nvkm_falcon_wr32(falcon, 0xac4 + (port * 8),
-				 extra & (BIT(rem * 8) - 1));
-	}
-}
-
 void
 nvkm_falcon_v1_load_dmem(struct nvkm_falcon *falcon, void *data, u32 start,
 			 u32 size, u8 port)
 {
-	const struct nvkm_falcon_func *func = falcon->func;
 	u8 rem = size % 4;
 	int i;
 
-	if (func->emem_addr && start >= func->emem_addr)
-		return nvkm_falcon_v1_load_emem(falcon, data,
-						start - func->emem_addr, size,
-						port);
-
 	size -= rem;
 
 	nvkm_falcon_wr32(falcon, 0x1c0 + (port * 8), start | (0x1 << 24));
@@ -120,113 +89,6 @@ nvkm_falcon_v1_load_dmem(struct nvkm_falcon *falcon, void *data, u32 start,
 	}
 }
 
-static void
-nvkm_falcon_v1_read_emem(struct nvkm_falcon *falcon, u32 start, u32 size,
-			 u8 port, void *data)
-{
-	u8 rem = size % 4;
-	int i;
-
-	size -= rem;
-
-	nvkm_falcon_wr32(falcon, 0xac0 + (port * 8), start | (0x1 << 25));
-	for (i = 0; i < size / 4; i++)
-		((u32 *)data)[i] = nvkm_falcon_rd32(falcon, 0xac4 + (port * 8));
-
-	/*
-	 * If size is not a multiple of 4, mask the last word to ensure garbage
-	 * does not get read
-	 */
-	if (rem) {
-		u32 extra = nvkm_falcon_rd32(falcon, 0xac4 + (port * 8));
-
-		for (i = size; i < size + rem; i++) {
-			((u8 *)data)[i] = (u8)(extra & 0xff);
-			extra >>= 8;
-		}
-	}
-}
-
-void
-nvkm_falcon_v1_read_dmem(struct nvkm_falcon *falcon, u32 start, u32 size,
-			 u8 port, void *data)
-{
-	const struct nvkm_falcon_func *func = falcon->func;
-	u8 rem = size % 4;
-	int i;
-
-	if (func->emem_addr && start >= func->emem_addr)
-		return nvkm_falcon_v1_read_emem(falcon, start - func->emem_addr,
-						size, port, data);
-
-	size -= rem;
-
-	nvkm_falcon_wr32(falcon, 0x1c0 + (port * 8), start | (0x1 << 25));
-	for (i = 0; i < size / 4; i++)
-		((u32 *)data)[i] = nvkm_falcon_rd32(falcon, 0x1c4 + (port * 8));
-
-	/*
-	 * If size is not a multiple of 4, mask the last word to ensure garbage
-	 * does not get read
-	 */
-	if (rem) {
-		u32 extra = nvkm_falcon_rd32(falcon, 0x1c4 + (port * 8));
-
-		for (i = size; i < size + rem; i++) {
-			((u8 *)data)[i] = (u8)(extra & 0xff);
-			extra >>= 8;
-		}
-	}
-}
-
-void
-nvkm_falcon_v1_bind_context(struct nvkm_falcon *falcon, struct nvkm_memory *ctx)
-{
-	const u32 fbif = falcon->func->fbif;
-	u32 inst_loc;
-
-	/* disable instance block binding */
-	if (ctx == NULL) {
-		nvkm_falcon_wr32(falcon, 0x10c, 0x0);
-		return;
-	}
-
-	nvkm_falcon_wr32(falcon, 0x10c, 0x1);
-
-	/* setup apertures - virtual */
-	nvkm_falcon_wr32(falcon, fbif + 4 * FALCON_DMAIDX_UCODE, 0x4);
-	nvkm_falcon_wr32(falcon, fbif + 4 * FALCON_DMAIDX_VIRT, 0x0);
-	/* setup apertures - physical */
-	nvkm_falcon_wr32(falcon, fbif + 4 * FALCON_DMAIDX_PHYS_VID, 0x4);
-	nvkm_falcon_wr32(falcon, fbif + 4 * FALCON_DMAIDX_PHYS_SYS_COH, 0x5);
-	nvkm_falcon_wr32(falcon, fbif + 4 * FALCON_DMAIDX_PHYS_SYS_NCOH, 0x6);
-
-	/* Set context */
-	switch (nvkm_memory_target(ctx)) {
-	case NVKM_MEM_TARGET_VRAM: inst_loc = 0; break;
-	case NVKM_MEM_TARGET_HOST: inst_loc = 2; break;
-	case NVKM_MEM_TARGET_NCOH: inst_loc = 3; break;
-	default:
-		WARN_ON(1);
-		return;
-	}
-
-	/* Enable context */
-	nvkm_falcon_mask(falcon, 0x048, 0x1, 0x1);
-	nvkm_falcon_wr32(falcon, 0x054,
-			 ((nvkm_memory_addr(ctx) >> 12) & 0xfffffff) |
-			 (inst_loc << 28) | (1 << 30));
-
-	nvkm_falcon_mask(falcon, 0x090, 0x10000, 0x10000);
-	nvkm_falcon_mask(falcon, 0x0a4, 0x8, 0x8);
-}
-
-void
-nvkm_falcon_v1_set_start_addr(struct nvkm_falcon *falcon, u32 start_addr)
-{
-	nvkm_falcon_wr32(falcon, 0x104, start_addr);
-}
-
 void
 nvkm_falcon_v1_start(struct nvkm_falcon *falcon)
 {
@@ -237,32 +99,3 @@ nvkm_falcon_v1_start(struct nvkm_falcon *falcon)
 	else
 		nvkm_falcon_wr32(falcon, 0x100, 0x2);
 }
-
-int
-nvkm_falcon_v1_wait_for_halt(struct nvkm_falcon *falcon, u32 ms)
-{
-	struct nvkm_device *device = falcon->owner->device;
-	int ret;
-
-	ret = nvkm_wait_msec(device, ms, falcon->addr + 0x100, 0x10, 0x10);
-	if (ret < 0)
-		return ret;
-
-	return 0;
-}
-
-int
-nvkm_falcon_v1_clear_interrupt(struct nvkm_falcon *falcon, u32 mask)
-{
-	struct nvkm_device *device = falcon->owner->device;
-	int ret;
-
-	/* clear interrupt(s) */
-	nvkm_falcon_mask(falcon, 0x004, mask, mask);
-	/* wait until interrupts are cleared */
-	ret = nvkm_wait_msec(device, 10, falcon->addr + 0x008, mask, 0x0);
-	if (ret < 0)
-		return ret;
-
-	return 0;
-}
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/Kbuild b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/Kbuild
index 5b9f64a8957f..ce6ece75f07e 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/Kbuild
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/Kbuild
@@ -1,10 +1,10 @@
 # SPDX-License-Identifier: MIT
 nvkm-y += nvkm/subdev/acr/base.o
-nvkm-y += nvkm/subdev/acr/hsfw.o
 nvkm-y += nvkm/subdev/acr/lsfw.o
 nvkm-y += nvkm/subdev/acr/gm200.o
 nvkm-y += nvkm/subdev/acr/gm20b.o
 nvkm-y += nvkm/subdev/acr/gp102.o
 nvkm-y += nvkm/subdev/acr/gp108.o
+nvkm-y += nvkm/subdev/acr/gv100.o
 nvkm-y += nvkm/subdev/acr/gp10b.o
 nvkm-y += nvkm/subdev/acr/tu102.o
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/base.c b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/base.c
index 882d8d30bd64..795f3a649b12 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/base.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/base.c
@@ -24,43 +24,36 @@
 #include <core/firmware.h>
 #include <core/memory.h>
 #include <subdev/mmu.h>
+#include <subdev/gsp.h>
+#include <subdev/pmu.h>
+#include <engine/sec2.h>
+#include <engine/nvdec.h>
 
-static struct nvkm_acr_hsf *
-nvkm_acr_hsf_find(struct nvkm_acr *acr, const char *name)
+static struct nvkm_acr_hsfw *
+nvkm_acr_hsfw_find(struct nvkm_acr *acr, const char *name)
 {
-	struct nvkm_acr_hsf *hsf;
-	list_for_each_entry(hsf, &acr->hsf, head) {
-		if (!strcmp(hsf->name, name))
-			return hsf;
+	struct nvkm_acr_hsfw *hsfw;
+
+	list_for_each_entry(hsfw, &acr->hsfw, head) {
+		if (!strcmp(hsfw->fw.fw.name, name))
+			return hsfw;
 	}
+
 	return NULL;
 }
 
 int
-nvkm_acr_hsf_boot(struct nvkm_acr *acr, const char *name)
+nvkm_acr_hsfw_boot(struct nvkm_acr *acr, const char *name)
 {
 	struct nvkm_subdev *subdev = &acr->subdev;
-	struct nvkm_acr_hsf *hsf;
-	int ret;
+	struct nvkm_acr_hsfw *hsfw;
 
-	hsf = nvkm_acr_hsf_find(acr, name);
-	if (!hsf)
+	hsfw = nvkm_acr_hsfw_find(acr, name);
+	if (!hsfw)
 		return -EINVAL;
 
-	nvkm_debug(subdev, "executing %s binary\n", hsf->name);
-	ret = nvkm_falcon_get(hsf->falcon, subdev);
-	if (ret)
-		return ret;
-
-	ret = hsf->func->boot(acr, hsf);
-	nvkm_falcon_put(hsf->falcon, subdev);
-	if (ret) {
-		nvkm_error(subdev, "%s binary failed\n", hsf->name);
-		return ret;
-	}
-
-	nvkm_debug(subdev, "%s binary completed successfully\n", hsf->name);
-	return 0;
+	return nvkm_falcon_fw_boot(&hsfw->fw, subdev, true, NULL, NULL,
+				   hsfw->boot_mbox0, hsfw->intr_clear);
 }
 
 static struct nvkm_acr_lsf *
@@ -87,7 +80,7 @@ nvkm_acr_unload(struct nvkm_acr *acr)
 			acr->rtos = NULL;
 		}
 
-		nvkm_acr_hsf_boot(acr, "unload");
+		nvkm_acr_hsfw_boot(acr, "unload");
 		acr->done = false;
 	}
 }
@@ -213,7 +206,7 @@ static void
 nvkm_acr_cleanup(struct nvkm_acr *acr)
 {
 	nvkm_acr_lsfw_del_all(acr);
-	nvkm_acr_hsfw_del_all(acr);
+
 	nvkm_firmware_put(acr->wpr_fw);
 	acr->wpr_fw = NULL;
 }
@@ -226,6 +219,7 @@ nvkm_acr_oneinit(struct nvkm_subdev *subdev)
 	struct nvkm_acr_hsfw *hsfw;
 	struct nvkm_acr_lsfw *lsfw, *lsft;
 	struct nvkm_acr_lsf *lsf, *rtos;
+	struct nvkm_falcon *falcon;
 	u32 wpr_size = 0;
 	u64 falcons;
 	int ret, i;
@@ -343,8 +337,16 @@ nvkm_acr_oneinit(struct nvkm_subdev *subdev)
 
 	/* Load HS firmware blobs into ACR VMM. */
 	list_for_each_entry(hsfw, &acr->hsfw, head) {
-		nvkm_debug(subdev, "loading %s fw\n", hsfw->name);
-		ret = hsfw->func->load(acr, hsfw);
+		switch (hsfw->falcon_id) {
+		case NVKM_ACR_HSF_PMU : falcon = &device->pmu->falcon; break;
+		case NVKM_ACR_HSF_SEC2: falcon = &device->sec2->falcon; break;
+		case NVKM_ACR_HSF_GSP : falcon = &device->gsp->falcon; break;
+		default:
+			WARN_ON(1);
+			return -EINVAL;
+		}
+
+		ret = nvkm_falcon_fw_oneinit(&hsfw->fw, falcon, acr->vmm, acr->inst);
 		if (ret)
 			return ret;
 	}
@@ -358,15 +360,13 @@ static void *
 nvkm_acr_dtor(struct nvkm_subdev *subdev)
 {
 	struct nvkm_acr *acr = nvkm_acr(subdev);
-	struct nvkm_acr_hsf *hsf, *hst;
+	struct nvkm_acr_hsfw *hsfw, *hsft;
 	struct nvkm_acr_lsf *lsf, *lst;
 
-	list_for_each_entry_safe(hsf, hst, &acr->hsf, head) {
-		nvkm_vmm_put(acr->vmm, &hsf->vma);
-		nvkm_memory_unref(&hsf->ucode);
-		kfree(hsf->imem);
-		list_del(&hsf->head);
-		kfree(hsf);
+	list_for_each_entry_safe(hsfw, hsft, &acr->hsfw, head) {
+		nvkm_falcon_fw_dtor(&hsfw->fw);
+		list_del(&hsfw->head);
+		kfree(hsfw);
 	}
 
 	nvkm_vmm_part(acr->vmm, acr->inst);
@@ -427,7 +427,6 @@ nvkm_acr_new_(const struct nvkm_acr_fwif *fwif, struct nvkm_device *device,
 	nvkm_subdev_ctor(&nvkm_acr, device, type, inst, &acr->subdev);
 	INIT_LIST_HEAD(&acr->hsfw);
 	INIT_LIST_HEAD(&acr->lsfw);
-	INIT_LIST_HEAD(&acr->hsf);
 	INIT_LIST_HEAD(&acr->lsf);
 
 	fwif = nvkm_firmware_load(&acr->subdev, fwif, "Acr", acr);
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gm200.c b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gm200.c
index 82b4c8e1457c..7a11151af3bd 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gm200.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gm200.c
@@ -46,7 +46,7 @@ gm200_acr_nofw(struct nvkm_acr *acr, int ver, const struct nvkm_acr_fwif *fwif)
 int
 gm200_acr_init(struct nvkm_acr *acr)
 {
-	return nvkm_acr_hsf_boot(acr, "load");
+	return nvkm_acr_hsfw_boot(acr, "load");
 }
 
 void
@@ -219,162 +219,50 @@ gm200_acr_wpr_parse(struct nvkm_acr *acr)
 	return 0;
 }
 
-void
-gm200_acr_hsfw_bld(struct nvkm_acr *acr, struct nvkm_acr_hsf *hsf)
+int
+gm200_acr_hsfw_load_bld(struct nvkm_falcon_fw *fw)
 {
 	struct flcn_bl_dmem_desc_v1 hsdesc = {
 		.ctx_dma = FALCON_DMAIDX_VIRT,
-		.code_dma_base = hsf->vma->addr,
-		.non_sec_code_off = hsf->non_sec_addr,
-		.non_sec_code_size = hsf->non_sec_size,
-		.sec_code_off = hsf->sec_addr,
-		.sec_code_size = hsf->sec_size,
+		.code_dma_base = fw->vma->addr,
+		.non_sec_code_off = fw->nmem_base,
+		.non_sec_code_size = fw->nmem_size,
+		.sec_code_off = fw->imem_base,
+		.sec_code_size = fw->imem_size,
 		.code_entry_point = 0,
-		.data_dma_base = hsf->vma->addr + hsf->data_addr,
-		.data_size = hsf->data_size,
+		.data_dma_base = fw->vma->addr + fw->dmem_base_img,
+		.data_size = fw->dmem_size,
 	};
 
-	flcn_bl_dmem_desc_v1_dump(&acr->subdev, &hsdesc);
-
-	nvkm_falcon_load_dmem(hsf->falcon, &hsdesc, 0, sizeof(hsdesc), 0);
-}
-
-int
-gm200_acr_hsfw_boot(struct nvkm_acr *acr, struct nvkm_acr_hsf *hsf,
-		    u32 intr_clear, u32 mbox0_ok)
-{
-	struct nvkm_subdev *subdev = &acr->subdev;
-	struct nvkm_device *device = subdev->device;
-	struct nvkm_falcon *falcon = hsf->falcon;
-	u32 mbox0, mbox1;
-	int ret;
-
-	/* Reset falcon. */
-	nvkm_falcon_reset(falcon);
-	nvkm_falcon_bind_context(falcon, acr->inst);
-
-	/* Load bootloader into IMEM. */
-	nvkm_falcon_load_imem(falcon, hsf->imem,
-				      falcon->code.limit - hsf->imem_size,
-				      hsf->imem_size,
-				      hsf->imem_tag,
-				      0, false);
-
-	/* Load bootloader data into DMEM. */
-	hsf->func->bld(acr, hsf);
-
-	/* Boot the falcon. */
-	nvkm_mc_intr_mask(device, falcon->owner->type, falcon->owner->inst, false);
+	flcn_bl_dmem_desc_v1_dump(fw->falcon->user, &hsdesc);
 
-	nvkm_falcon_wr32(falcon, 0x040, 0xdeada5a5);
-	nvkm_falcon_set_start_addr(falcon, hsf->imem_tag << 8);
-	nvkm_falcon_start(falcon);
-	ret = nvkm_falcon_wait_for_halt(falcon, 100);
-	if (ret)
-		return ret;
-
-	/* Check for successful completion. */
-	mbox0 = nvkm_falcon_rd32(falcon, 0x040);
-	mbox1 = nvkm_falcon_rd32(falcon, 0x044);
-	nvkm_debug(subdev, "mailbox %08x %08x\n", mbox0, mbox1);
-	if (mbox0 && mbox0 != mbox0_ok)
-		return -EIO;
-
-	nvkm_falcon_clear_interrupt(falcon, intr_clear);
-	nvkm_mc_intr_mask(device, falcon->owner->type, falcon->owner->inst, true);
-	return ret;
+	return nvkm_falcon_pio_wr(fw->falcon, (u8 *)&hsdesc, 0, 0, DMEM, 0, sizeof(hsdesc), 0, 0);
 }
 
 int
-gm200_acr_hsfw_load(struct nvkm_acr *acr, struct nvkm_acr_hsfw *hsfw,
-		    struct nvkm_falcon *falcon)
+gm200_acr_hsfw_ctor(struct nvkm_acr *acr, const char *bl, const char *fw, const char *name, int ver,
+		    const struct nvkm_acr_hsf_fwif *fwif)
 {
-	struct nvkm_subdev *subdev = &acr->subdev;
-	struct nvkm_acr_hsf *hsf;
-	int ret;
-
-	/* Patch the appropriate signature (production/debug) into the FW
-	 * image, as determined by the mode the falcon is in.
-	 */
-	ret = nvkm_falcon_get(falcon, subdev);
-	if (ret)
-		return ret;
-
-	if (hsfw->sig.patch_loc) {
-		if (!falcon->debug) {
-			nvkm_debug(subdev, "patching production signature\n");
-			memcpy(hsfw->image + hsfw->sig.patch_loc,
-			       hsfw->sig.prod.data,
-			       hsfw->sig.prod.size);
-		} else {
-			nvkm_debug(subdev, "patching debug signature\n");
-			memcpy(hsfw->image + hsfw->sig.patch_loc,
-			       hsfw->sig.dbg.data,
-			       hsfw->sig.dbg.size);
-		}
-	}
-
-	nvkm_falcon_put(falcon, subdev);
+	struct nvkm_acr_hsfw *hsfw;
 
-	if (!(hsf = kzalloc(sizeof(*hsf), GFP_KERNEL)))
+	if (!(hsfw = kzalloc(sizeof(*hsfw), GFP_KERNEL)))
 		return -ENOMEM;
-	hsf->func = hsfw->func;
-	hsf->name = hsfw->name;
-	list_add_tail(&hsf->head, &acr->hsf);
-
-	hsf->imem_size = hsfw->imem_size;
-	hsf->imem_tag = hsfw->imem_tag;
-	hsf->imem = kmemdup(hsfw->imem, hsfw->imem_size, GFP_KERNEL);
-	if (!hsf->imem)
-		return -ENOMEM;
-
-	hsf->non_sec_addr = hsfw->non_sec_addr;
-	hsf->non_sec_size = hsfw->non_sec_size;
-	hsf->sec_addr = hsfw->sec_addr;
-	hsf->sec_size = hsfw->sec_size;
-	hsf->data_addr = hsfw->data_addr;
-	hsf->data_size = hsfw->data_size;
-
-	/* Make the FW image accessible to the HS bootloader. */
-	ret = nvkm_memory_new(subdev->device, NVKM_MEM_TARGET_INST,
-			      hsfw->image_size, 0x1000, false, &hsf->ucode);
-	if (ret)
-		return ret;
-
-	nvkm_kmap(hsf->ucode);
-	nvkm_wobj(hsf->ucode, 0, hsfw->image, hsfw->image_size);
-	nvkm_done(hsf->ucode);
-
-	ret = nvkm_vmm_get(acr->vmm, 12, nvkm_memory_size(hsf->ucode),
-			   &hsf->vma);
-	if (ret)
-		return ret;
-
-	ret = nvkm_memory_map(hsf->ucode, 0, acr->vmm, hsf->vma, NULL, 0);
-	if (ret)
-		return ret;
-
-	hsf->falcon = falcon;
-	return 0;
-}
 
-int
-gm200_acr_unload_boot(struct nvkm_acr *acr, struct nvkm_acr_hsf *hsf)
-{
-	return gm200_acr_hsfw_boot(acr, hsf, 0, 0x1d);
-}
+	hsfw->falcon_id = fwif->falcon_id;
+	hsfw->boot_mbox0 = fwif->boot_mbox0;
+	hsfw->intr_clear = fwif->intr_clear;
+	list_add_tail(&hsfw->head, &acr->hsfw);
 
-int
-gm200_acr_unload_load(struct nvkm_acr *acr, struct nvkm_acr_hsfw *hsfw)
-{
-	return gm200_acr_hsfw_load(acr, hsfw, &acr->subdev.device->pmu->falcon);
+	return nvkm_falcon_fw_ctor_hs(fwif->func, name, &acr->subdev, bl, fw, ver, NULL, &hsfw->fw);
 }
 
-const struct nvkm_acr_hsf_func
+const struct nvkm_falcon_fw_func
 gm200_acr_unload_0 = {
-	.load = gm200_acr_unload_load,
-	.boot = gm200_acr_unload_boot,
-	.bld = gm200_acr_hsfw_bld,
+	.signature = gm200_flcn_fw_signature,
+	.reset = gm200_flcn_fw_reset,
+	.load = gm200_flcn_fw_load,
+	.load_bld = gm200_acr_hsfw_load_bld,
+	.boot = gm200_flcn_fw_boot,
 };
 
 MODULE_FIRMWARE("nvidia/gm200/acr/ucode_unload.bin");
@@ -384,20 +272,15 @@ MODULE_FIRMWARE("nvidia/gp100/acr/ucode_unload.bin");
 
 static const struct nvkm_acr_hsf_fwif
 gm200_acr_unload_fwif[] = {
-	{ 0, nvkm_acr_hsfw_load, &gm200_acr_unload_0 },
+	{ 0, gm200_acr_hsfw_ctor, &gm200_acr_unload_0, NVKM_ACR_HSF_PMU, 0, 0x00000010 },
 	{}
 };
 
-int
-gm200_acr_load_boot(struct nvkm_acr *acr, struct nvkm_acr_hsf *hsf)
-{
-	return gm200_acr_hsfw_boot(acr, hsf, 0x10, 0);
-}
-
 static int
-gm200_acr_load_load(struct nvkm_acr *acr, struct nvkm_acr_hsfw *hsfw)
+gm200_acr_load_setup(struct nvkm_falcon_fw *fw)
 {
-	struct flcn_acr_desc *desc = (void *)&hsfw->image[hsfw->data_addr];
+	struct flcn_acr_desc *desc = (void *)&fw->fw.img[fw->dmem_base_img];
+	struct nvkm_acr *acr = fw->falcon->owner->device->acr;
 
 	desc->wpr_region_id = 1;
 	desc->regions.no_regions = 2;
@@ -408,15 +291,17 @@ gm200_acr_load_load(struct nvkm_acr *acr, struct nvkm_acr_hsfw *hsfw)
 	desc->regions.region_props[0].write_mask = 0xc;
 	desc->regions.region_props[0].client_mask = 0x2;
 	flcn_acr_desc_dump(&acr->subdev, desc);
-
-	return gm200_acr_hsfw_load(acr, hsfw, &acr->subdev.device->pmu->falcon);
+	return 0;
 }
 
-static const struct nvkm_acr_hsf_func
+static const struct nvkm_falcon_fw_func
 gm200_acr_load_0 = {
-	.load = gm200_acr_load_load,
-	.boot = gm200_acr_load_boot,
-	.bld = gm200_acr_hsfw_bld,
+	.signature = gm200_flcn_fw_signature,
+	.reset = gm200_flcn_fw_reset,
+	.setup = gm200_acr_load_setup,
+	.load = gm200_flcn_fw_load,
+	.load_bld = gm200_acr_hsfw_load_bld,
+	.boot = gm200_flcn_fw_boot,
 };
 
 MODULE_FIRMWARE("nvidia/gm200/acr/bl.bin");
@@ -433,7 +318,7 @@ MODULE_FIRMWARE("nvidia/gp100/acr/ucode_load.bin");
 
 static const struct nvkm_acr_hsf_fwif
 gm200_acr_load_fwif[] = {
-	{ 0, nvkm_acr_hsfw_load, &gm200_acr_load_0 },
+	{ 0, gm200_acr_hsfw_ctor, &gm200_acr_load_0, NVKM_ACR_HSF_PMU, 0, 0x00000010 },
 	{}
 };
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gm20b.c b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gm20b.c
index 54e996f2f630..ef5fb79128b1 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gm20b.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gm20b.c
@@ -45,43 +45,47 @@ gm20b_acr_wpr_alloc(struct nvkm_acr *acr, u32 wpr_size)
 			       wpr_size, 0, true, &acr->wpr);
 }
 
-static void
-gm20b_acr_load_bld(struct nvkm_acr *acr, struct nvkm_acr_hsf *hsf)
+static int
+gm20b_acr_hsfw_load_bld(struct nvkm_falcon_fw *fw)
 {
 	struct flcn_bl_dmem_desc hsdesc = {
 		.ctx_dma = FALCON_DMAIDX_VIRT,
-		.code_dma_base = hsf->vma->addr >> 8,
-		.non_sec_code_off = hsf->non_sec_addr,
-		.non_sec_code_size = hsf->non_sec_size,
-		.sec_code_off = hsf->sec_addr,
-		.sec_code_size = hsf->sec_size,
+		.code_dma_base = fw->vma->addr >> 8,
+		.non_sec_code_off = fw->nmem_base,
+		.non_sec_code_size = fw->nmem_size,
+		.sec_code_off = fw->imem_base,
+		.sec_code_size = fw->imem_size,
 		.code_entry_point = 0,
-		.data_dma_base = (hsf->vma->addr + hsf->data_addr) >> 8,
-		.data_size = hsf->data_size,
+		.data_dma_base = (fw->vma->addr + fw->dmem_base_img) >> 8,
+		.data_size = fw->dmem_size,
 	};
 
-	flcn_bl_dmem_desc_dump(&acr->subdev, &hsdesc);
+	flcn_bl_dmem_desc_dump(fw->falcon->user, &hsdesc);
 
-	nvkm_falcon_load_dmem(hsf->falcon, &hsdesc, 0, sizeof(hsdesc), 0);
+	return nvkm_falcon_pio_wr(fw->falcon, (u8 *)&hsdesc, 0, 0, DMEM, 0, sizeof(hsdesc), 0, 0);
 }
 
+
 static int
-gm20b_acr_load_load(struct nvkm_acr *acr, struct nvkm_acr_hsfw *hsfw)
+gm20b_acr_load_setup(struct nvkm_falcon_fw *fw)
 {
-	struct flcn_acr_desc *desc = (void *)&hsfw->image[hsfw->data_addr];
+	struct flcn_acr_desc *desc = (void *)&fw->fw.img[fw->dmem_base_img];
+	struct nvkm_acr *acr = fw->falcon->owner->device->acr;
 
 	desc->ucode_blob_base = nvkm_memory_addr(acr->wpr);
 	desc->ucode_blob_size = nvkm_memory_size(acr->wpr);
 	flcn_acr_desc_dump(&acr->subdev, desc);
-
-	return gm200_acr_hsfw_load(acr, hsfw, &acr->subdev.device->pmu->falcon);
+	return 0;
 }
 
-const struct nvkm_acr_hsf_func
+const struct nvkm_falcon_fw_func
 gm20b_acr_load_0 = {
-	.load = gm20b_acr_load_load,
-	.boot = gm200_acr_load_boot,
-	.bld = gm20b_acr_load_bld,
+	.signature = gm200_flcn_fw_signature,
+	.reset = gm200_flcn_fw_reset,
+	.setup = gm20b_acr_load_setup,
+	.load = gm200_flcn_fw_load,
+	.load_bld = gm20b_acr_hsfw_load_bld,
+	.boot = gm200_flcn_fw_boot,
 };
 
 #if IS_ENABLED(CONFIG_ARCH_TEGRA_210_SOC)
@@ -91,7 +95,7 @@ MODULE_FIRMWARE("nvidia/gm20b/acr/ucode_load.bin");
 
 static const struct nvkm_acr_hsf_fwif
 gm20b_acr_load_fwif[] = {
-	{ 0, nvkm_acr_hsfw_load, &gm20b_acr_load_0 },
+	{ 0, gm200_acr_hsfw_ctor, &gm20b_acr_load_0, NVKM_ACR_HSF_PMU, 0, 0x10 },
 	{}
 };
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp102.c b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp102.c
index fd97a935a380..f4c2d3729feb 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp102.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp102.c
@@ -187,14 +187,15 @@ MODULE_FIRMWARE("nvidia/gp107/acr/ucode_unload.bin");
 
 static const struct nvkm_acr_hsf_fwif
 gp102_acr_unload_fwif[] = {
-	{ 0, nvkm_acr_hsfw_load, &gm200_acr_unload_0 },
+	{ 0, gm200_acr_hsfw_ctor, &gm200_acr_unload_0, NVKM_ACR_HSF_PMU, 0x1d, 0x00000010 },
 	{}
 };
 
 int
-gp102_acr_load_load(struct nvkm_acr *acr, struct nvkm_acr_hsfw *hsfw)
+gp102_acr_load_setup(struct nvkm_falcon_fw *fw)
 {
-	struct flcn_acr_desc_v1 *desc = (void *)&hsfw->image[hsfw->data_addr];
+	struct flcn_acr_desc_v1 *desc = (void *)&fw->fw.img[fw->dmem_base_img];
+	struct nvkm_acr *acr = fw->falcon->owner->device->acr;
 
 	desc->wpr_region_id = 1;
 	desc->regions.no_regions = 2;
@@ -204,19 +205,19 @@ gp102_acr_load_load(struct nvkm_acr *acr, struct nvkm_acr_hsfw *hsfw)
 	desc->regions.region_props[0].read_mask = 0xf;
 	desc->regions.region_props[0].write_mask = 0xc;
 	desc->regions.region_props[0].client_mask = 0x2;
-	desc->regions.region_props[0].shadow_mem_start_addr =
-		acr->shadow_start >> 8;
+	desc->regions.region_props[0].shadow_mem_start_addr = acr->shadow_start >> 8;
 	flcn_acr_desc_v1_dump(&acr->subdev, desc);
-
-	return gm200_acr_hsfw_load(acr, hsfw,
-				  &acr->subdev.device->sec2->falcon);
+	return 0;
 }
 
-static const struct nvkm_acr_hsf_func
+static const struct nvkm_falcon_fw_func
 gp102_acr_load_0 = {
-	.load = gp102_acr_load_load,
-	.boot = gm200_acr_load_boot,
-	.bld = gm200_acr_hsfw_bld,
+	.signature = gm200_flcn_fw_signature,
+	.reset = gm200_flcn_fw_reset,
+	.setup = gp102_acr_load_setup,
+	.load = gm200_flcn_fw_load,
+	.load_bld = gm200_acr_hsfw_load_bld,
+	.boot = gm200_flcn_fw_boot,
 };
 
 MODULE_FIRMWARE("nvidia/gp102/acr/bl.bin");
@@ -233,7 +234,7 @@ MODULE_FIRMWARE("nvidia/gp107/acr/ucode_load.bin");
 
 static const struct nvkm_acr_hsf_fwif
 gp102_acr_load_fwif[] = {
-	{ 0, nvkm_acr_hsfw_load, &gp102_acr_load_0 },
+	{ 0, gm200_acr_hsfw_ctor, &gp102_acr_load_0, NVKM_ACR_HSF_SEC2, 0, 0x00000010 },
 	{}
 };
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp108.c b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp108.c
index 373d638a2177..6ab9d4959c17 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp108.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp108.c
@@ -25,63 +25,62 @@
 
 #include <nvfw/flcn.h>
 
-void
-gp108_acr_hsfw_bld(struct nvkm_acr *acr, struct nvkm_acr_hsf *hsf)
+int
+gp108_acr_hsfw_load_bld(struct nvkm_falcon_fw *fw)
 {
 	struct flcn_bl_dmem_desc_v2 hsdesc = {
 		.ctx_dma = FALCON_DMAIDX_VIRT,
-		.code_dma_base = hsf->vma->addr,
-		.non_sec_code_off = hsf->non_sec_addr,
-		.non_sec_code_size = hsf->non_sec_size,
-		.sec_code_off = hsf->sec_addr,
-		.sec_code_size = hsf->sec_size,
+		.code_dma_base = fw->vma->addr,
+		.non_sec_code_off = fw->nmem_base,
+		.non_sec_code_size = fw->nmem_size,
+		.sec_code_off = fw->imem_base,
+		.sec_code_size = fw->imem_size,
 		.code_entry_point = 0,
-		.data_dma_base = hsf->vma->addr + hsf->data_addr,
-		.data_size = hsf->data_size,
+		.data_dma_base = fw->vma->addr + fw->dmem_base_img,
+		.data_size = fw->dmem_size,
 		.argc = 0,
 		.argv = 0,
 	};
 
-	flcn_bl_dmem_desc_v2_dump(&acr->subdev, &hsdesc);
+	flcn_bl_dmem_desc_v2_dump(fw->falcon->user, &hsdesc);
 
-	nvkm_falcon_load_dmem(hsf->falcon, &hsdesc, 0, sizeof(hsdesc), 0);
+	return nvkm_falcon_pio_wr(fw->falcon, (u8 *)&hsdesc, 0, 0, DMEM, 0, sizeof(hsdesc), 0, 0);
 }
 
-const struct nvkm_acr_hsf_func
-gp108_acr_unload_0 = {
-	.load = gm200_acr_unload_load,
-	.boot = gm200_acr_unload_boot,
-	.bld = gp108_acr_hsfw_bld,
+const struct nvkm_falcon_fw_func
+gp108_acr_hsfw_0 = {
+	.signature = gm200_flcn_fw_signature,
+	.reset = gm200_flcn_fw_reset,
+	.load = gm200_flcn_fw_load,
+	.load_bld = gp108_acr_hsfw_load_bld,
+	.boot = gm200_flcn_fw_boot,
 };
 
 MODULE_FIRMWARE("nvidia/gp108/acr/unload_bl.bin");
 MODULE_FIRMWARE("nvidia/gp108/acr/ucode_unload.bin");
 
-MODULE_FIRMWARE("nvidia/gv100/acr/unload_bl.bin");
-MODULE_FIRMWARE("nvidia/gv100/acr/ucode_unload.bin");
-
 static const struct nvkm_acr_hsf_fwif
 gp108_acr_unload_fwif[] = {
-	{ 0, nvkm_acr_hsfw_load, &gp108_acr_unload_0 },
+	{ 0, gm200_acr_hsfw_ctor, &gp108_acr_hsfw_0, NVKM_ACR_HSF_PMU, 0x1d, 0x00000010 },
 	{}
 };
 
-static const struct nvkm_acr_hsf_func
+const struct nvkm_falcon_fw_func
 gp108_acr_load_0 = {
-	.load = gp102_acr_load_load,
-	.boot = gm200_acr_load_boot,
-	.bld = gp108_acr_hsfw_bld,
+	.signature = gm200_flcn_fw_signature,
+	.reset = gm200_flcn_fw_reset,
+	.setup = gp102_acr_load_setup,
+	.load = gm200_flcn_fw_load,
+	.load_bld = gp108_acr_hsfw_load_bld,
+	.boot = gm200_flcn_fw_boot,
 };
 
 MODULE_FIRMWARE("nvidia/gp108/acr/bl.bin");
 MODULE_FIRMWARE("nvidia/gp108/acr/ucode_load.bin");
 
-MODULE_FIRMWARE("nvidia/gv100/acr/bl.bin");
-MODULE_FIRMWARE("nvidia/gv100/acr/ucode_load.bin");
-
 static const struct nvkm_acr_hsf_fwif
 gp108_acr_load_fwif[] = {
-	{ 0, nvkm_acr_hsfw_load, &gp108_acr_load_0 },
+	{ 0, gm200_acr_hsfw_ctor, &gp108_acr_load_0, NVKM_ACR_HSF_SEC2, 0, 0x00000010 },
 	{}
 };
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp10b.c b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp10b.c
index f03ba028867b..a3422ab6deab 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp10b.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gp10b.c
@@ -28,7 +28,7 @@ MODULE_FIRMWARE("nvidia/gp10b/acr/ucode_load.bin");
 
 static const struct nvkm_acr_hsf_fwif
 gp10b_acr_load_fwif[] = {
-	{ 0, nvkm_acr_hsfw_load, &gm20b_acr_load_0 },
+	{ 0, gm200_acr_hsfw_ctor, &gm20b_acr_load_0, NVKM_ACR_HSF_PMU, 0, 0x00000010 },
 	{}
 };
 
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gv100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gv100.c
new file mode 100644
index 000000000000..4c5ca6b40027
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/gv100.c
@@ -0,0 +1,67 @@
+/*
+ * Copyright 2022 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.
+ */
+#include "priv.h"
+
+MODULE_FIRMWARE("nvidia/gv100/acr/unload_bl.bin");
+MODULE_FIRMWARE("nvidia/gv100/acr/ucode_unload.bin");
+
+static const struct nvkm_acr_hsf_fwif
+gv100_acr_unload_fwif[] = {
+	{ 0, gm200_acr_hsfw_ctor, &gp108_acr_hsfw_0, NVKM_ACR_HSF_PMU, 0, 0x00000000 },
+	{}
+};
+
+MODULE_FIRMWARE("nvidia/gv100/acr/bl.bin");
+MODULE_FIRMWARE("nvidia/gv100/acr/ucode_load.bin");
+
+static const struct nvkm_acr_hsf_fwif
+gv100_acr_load_fwif[] = {
+	{ 0, gm200_acr_hsfw_ctor, &gp108_acr_load_0, NVKM_ACR_HSF_SEC2, 0, 0x00000010 },
+	{}
+};
+
+static const struct nvkm_acr_func
+gv100_acr = {
+	.load = gv100_acr_load_fwif,
+	.unload = gv100_acr_unload_fwif,
+	.wpr_parse = gp102_acr_wpr_parse,
+	.wpr_layout = gp102_acr_wpr_layout,
+	.wpr_alloc = gp102_acr_wpr_alloc,
+	.wpr_build = gp102_acr_wpr_build,
+	.wpr_patch = gp102_acr_wpr_patch,
+	.wpr_check = gm200_acr_wpr_check,
+	.init = gm200_acr_init,
+};
+
+static const struct nvkm_acr_fwif
+gv100_acr_fwif[] = {
+	{  0, gp102_acr_load, &gv100_acr },
+	{ -1, gm200_acr_nofw, &gm200_acr },
+	{}
+};
+
+int
+gv100_acr_new(struct nvkm_device *device, enum nvkm_subdev_type type, int inst,
+	      struct nvkm_acr **pacr)
+{
+	return nvkm_acr_new_(gv100_acr_fwif, device, type, inst, pacr);
+}
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/hsfw.c b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/hsfw.c
deleted file mode 100644
index a6ea89a5d51a..000000000000
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/hsfw.c
+++ /dev/null
@@ -1,177 +0,0 @@
-/*
- * Copyright 2019 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.
- */
-#include "priv.h"
-
-#include <core/firmware.h>
-
-#include <nvfw/fw.h>
-#include <nvfw/hs.h>
-
-static void
-nvkm_acr_hsfw_del(struct nvkm_acr_hsfw *hsfw)
-{
-	list_del(&hsfw->head);
-	kfree(hsfw->imem);
-	kfree(hsfw->image);
-	kfree(hsfw->sig.prod.data);
-	kfree(hsfw->sig.dbg.data);
-	kfree(hsfw);
-}
-
-void
-nvkm_acr_hsfw_del_all(struct nvkm_acr *acr)
-{
-	struct nvkm_acr_hsfw *hsfw, *hsft;
-	list_for_each_entry_safe(hsfw, hsft, &acr->hsfw, head) {
-		nvkm_acr_hsfw_del(hsfw);
-	}
-}
-
-static int
-nvkm_acr_hsfw_load_image(struct nvkm_acr *acr, const char *name, int ver,
-			 struct nvkm_acr_hsfw *hsfw)
-{
-	struct nvkm_subdev *subdev = &acr->subdev;
-	const struct firmware *fw;
-	const struct nvfw_bin_hdr *hdr;
-	const struct nvfw_hs_header *fwhdr;
-	const struct nvfw_hs_load_header *lhdr;
-	u32 loc, sig;
-	int ret;
-
-	ret = nvkm_firmware_get(subdev, name, ver, &fw);
-	if (ret < 0)
-		return ret;
-
-	hdr = nvfw_bin_hdr(subdev, fw->data);
-	fwhdr = nvfw_hs_header(subdev, fw->data + hdr->header_offset);
-
-	/* Earlier FW releases by NVIDIA for Nouveau's use aren't in NVIDIA's
-	 * standard format, and don't have the indirection seen in the 0x10de
-	 * case.
-	 */
-	switch (hdr->bin_magic) {
-	case 0x000010de:
-		loc = *(u32 *)(fw->data + fwhdr->patch_loc);
-		sig = *(u32 *)(fw->data + fwhdr->patch_sig);
-		break;
-	case 0x3b1d14f0:
-		loc = fwhdr->patch_loc;
-		sig = fwhdr->patch_sig;
-		break;
-	default:
-		ret = -EINVAL;
-		goto done;
-	}
-
-	lhdr = nvfw_hs_load_header(subdev, fw->data + fwhdr->hdr_offset);
-
-	if (!(hsfw->image = kmalloc(hdr->data_size, GFP_KERNEL))) {
-		ret = -ENOMEM;
-		goto done;
-	}
-
-	memcpy(hsfw->image, fw->data + hdr->data_offset, hdr->data_size);
-	hsfw->image_size = hdr->data_size;
-	hsfw->non_sec_addr = lhdr->non_sec_code_off;
-	hsfw->non_sec_size = lhdr->non_sec_code_size;
-	hsfw->sec_addr = lhdr->apps[0];
-	hsfw->sec_size = lhdr->apps[lhdr->num_apps];
-	hsfw->data_addr = lhdr->data_dma_base;
-	hsfw->data_size = lhdr->data_size;
-
-	hsfw->sig.prod.size = fwhdr->sig_prod_size;
-	hsfw->sig.prod.data = kmemdup(fw->data + fwhdr->sig_prod_offset + sig,
-				      hsfw->sig.prod.size, GFP_KERNEL);
-	if (!hsfw->sig.prod.data) {
-		ret = -ENOMEM;
-		goto done;
-	}
-
-	hsfw->sig.dbg.size = fwhdr->sig_dbg_size;
-	hsfw->sig.dbg.data = kmemdup(fw->data + fwhdr->sig_dbg_offset + sig,
-				     hsfw->sig.dbg.size, GFP_KERNEL);
-	if (!hsfw->sig.dbg.data) {
-		ret = -ENOMEM;
-		goto done;
-	}
-
-	hsfw->sig.patch_loc = loc;
-done:
-	nvkm_firmware_put(fw);
-	return ret;
-}
-
-static int
-nvkm_acr_hsfw_load_bl(struct nvkm_acr *acr, const char *name, int ver,
-		      struct nvkm_acr_hsfw *hsfw)
-{
-	struct nvkm_subdev *subdev = &acr->subdev;
-	const struct nvfw_bin_hdr *hdr;
-	const struct nvfw_bl_desc *desc;
-	const struct firmware *fw;
-	u8 *data;
-	int ret;
-
-	ret = nvkm_firmware_get(subdev, name, ver, &fw);
-	if (ret)
-		return ret;
-
-	hdr = nvfw_bin_hdr(subdev, fw->data);
-	desc = nvfw_bl_desc(subdev, fw->data + hdr->header_offset);
-	data = (void *)fw->data + hdr->data_offset;
-
-	hsfw->imem_size = desc->code_size;
-	hsfw->imem_tag = desc->start_tag;
-	hsfw->imem = kmemdup(data + desc->code_off, desc->code_size, GFP_KERNEL);
-	nvkm_firmware_put(fw);
-	if (!hsfw->imem)
-		return -ENOMEM;
-	else
-		return 0;
-}
-
-int
-nvkm_acr_hsfw_load(struct nvkm_acr *acr, const char *bl, const char *fw,
-		   const char *name, int version,
-		   const struct nvkm_acr_hsf_fwif *fwif)
-{
-	struct nvkm_acr_hsfw *hsfw;
-	int ret;
-
-	if (!(hsfw = kzalloc(sizeof(*hsfw), GFP_KERNEL)))
-		return -ENOMEM;
-
-	hsfw->func = fwif->func;
-	hsfw->name = name;
-	list_add_tail(&hsfw->head, &acr->hsfw);
-
-	ret = nvkm_acr_hsfw_load_bl(acr, bl, version, hsfw);
-	if (ret)
-		goto done;
-
-	ret = nvkm_acr_hsfw_load_image(acr, fw, version, hsfw);
-done:
-	if (ret)
-		nvkm_acr_hsfw_del(hsfw);
-	return ret;
-}
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/priv.h b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/priv.h
index c30b841c9d35..66a764e24bc6 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/priv.h
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/priv.h
@@ -51,93 +51,50 @@ int gp102_acr_wpr_build_lsb(struct nvkm_acr *, struct nvkm_acr_lsfw *);
 void gp102_acr_wpr_patch(struct nvkm_acr *, s64);
 
 struct nvkm_acr_hsfw {
-	const struct nvkm_acr_hsf_func *func;
-	const char *name;
-	struct list_head head;
+	struct nvkm_falcon_fw fw;
+
+	enum nvkm_acr_hsf_id {
+		NVKM_ACR_HSF_PMU,
+		NVKM_ACR_HSF_SEC2,
+		NVKM_ACR_HSF_GSP,
+	} falcon_id;
+	u32 boot_mbox0;
+	u32 intr_clear;
 
-	u32 imem_size;
-	u32 imem_tag;
-	u32 *imem;
-
-	u8 *image;
-	u32 image_size;
-	u32 non_sec_addr;
-	u32 non_sec_size;
-	u32 sec_addr;
-	u32 sec_size;
-	u32 data_addr;
-	u32 data_size;
-
-	struct {
-		struct {
-			void *data;
-			u32 size;
-		} prod, dbg;
-		u32 patch_loc;
-	} sig;
+	struct list_head head;
 };
 
+int nvkm_acr_hsfw_boot(struct nvkm_acr *, const char *name);
+
 struct nvkm_acr_hsf_fwif {
 	int version;
 	int (*load)(struct nvkm_acr *, const char *bl, const char *fw,
 		    const char *name, int version,
 		    const struct nvkm_acr_hsf_fwif *);
-	const struct nvkm_acr_hsf_func *func;
-};
-
-int nvkm_acr_hsfw_load(struct nvkm_acr *, const char *, const char *,
-		       const char *, int, const struct nvkm_acr_hsf_fwif *);
-void nvkm_acr_hsfw_del_all(struct nvkm_acr *);
-
-struct nvkm_acr_hsf {
-	const struct nvkm_acr_hsf_func *func;
-	const char *name;
-	struct list_head head;
-
-	u32 imem_size;
-	u32 imem_tag;
-	u32 *imem;
-
-	u32 non_sec_addr;
-	u32 non_sec_size;
-	u32 sec_addr;
-	u32 sec_size;
-	u32 data_addr;
-	u32 data_size;
-
-	struct nvkm_memory *ucode;
-	struct nvkm_vma *vma;
-	struct nvkm_falcon *falcon;
-};
+	const struct nvkm_falcon_fw_func *func;
 
-struct nvkm_acr_hsf_func {
-	int (*load)(struct nvkm_acr *, struct nvkm_acr_hsfw *);
-	int (*boot)(struct nvkm_acr *, struct nvkm_acr_hsf *);
-	void (*bld)(struct nvkm_acr *, struct nvkm_acr_hsf *);
+	enum nvkm_acr_hsf_id falcon_id;
+	u32 boot_mbox0;
+	u32 intr_clear;
 };
 
-int gm200_acr_hsfw_load(struct nvkm_acr *, struct nvkm_acr_hsfw *,
-			struct nvkm_falcon *);
-int gm200_acr_hsfw_boot(struct nvkm_acr *, struct nvkm_acr_hsf *,
-			u32 clear_intr, u32 mbox0_ok);
 
-int gm200_acr_load_boot(struct nvkm_acr *, struct nvkm_acr_hsf *);
+int gm200_acr_hsfw_ctor(struct nvkm_acr *, const char *, const char *, const char *, int,
+			const struct nvkm_acr_hsf_fwif *);
+int gm200_acr_hsfw_load_bld(struct nvkm_falcon_fw *);
+extern const struct nvkm_falcon_fw_func gm200_acr_unload_0;
 
-extern const struct nvkm_acr_hsf_func gm200_acr_unload_0;
-int gm200_acr_unload_load(struct nvkm_acr *, struct nvkm_acr_hsfw *);
-int gm200_acr_unload_boot(struct nvkm_acr *, struct nvkm_acr_hsf *);
-void gm200_acr_hsfw_bld(struct nvkm_acr *, struct nvkm_acr_hsf *);
+extern const struct nvkm_falcon_fw_func gm20b_acr_load_0;
 
-extern const struct nvkm_acr_hsf_func gm20b_acr_load_0;
+int gp102_acr_load_setup(struct nvkm_falcon_fw *);
 
-int gp102_acr_load_load(struct nvkm_acr *, struct nvkm_acr_hsfw *);
+extern const struct nvkm_falcon_fw_func gp108_acr_load_0;
 
-extern const struct nvkm_acr_hsf_func gp108_acr_unload_0;
-void gp108_acr_hsfw_bld(struct nvkm_acr *, struct nvkm_acr_hsf *);
+extern const struct nvkm_falcon_fw_func gp108_acr_hsfw_0;
+int gp108_acr_hsfw_load_bld(struct nvkm_falcon_fw *);
 
 int nvkm_acr_new_(const struct nvkm_acr_fwif *, struct nvkm_device *, enum nvkm_subdev_type,
 		  int inst, struct nvkm_acr **);
-int nvkm_acr_hsf_boot(struct nvkm_acr *, const char *name);
 
 struct nvkm_acr_lsf {
 	const struct nvkm_acr_lsf_func *func;
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/tu102.c b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/tu102.c
index 05a87e77525f..ad45f5cb452b 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/acr/tu102.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/acr/tu102.c
@@ -32,11 +32,11 @@
 static int
 tu102_acr_init(struct nvkm_acr *acr)
 {
-	int ret = nvkm_acr_hsf_boot(acr, "AHESASC");
+	int ret = nvkm_acr_hsfw_boot(acr, "AHESASC");
 	if (ret)
 		return ret;
 
-	return nvkm_acr_hsf_boot(acr, "ASB");
+	return nvkm_acr_hsfw_boot(acr, "ASB");
 }
 
 static int
@@ -85,12 +85,6 @@ tu102_acr_wpr_build(struct nvkm_acr *acr, struct nvkm_acr_lsf *rtos)
 }
 
 static int
-tu102_acr_hsfw_boot(struct nvkm_acr *acr, struct nvkm_acr_hsf *hsf)
-{
-	return gm200_acr_hsfw_boot(acr, hsf, 0, 0);
-}
-
-static int
 tu102_acr_hsfw_nofw(struct nvkm_acr *acr, const char *bl, const char *fw,
 		    const char *name, int version,
 		    const struct nvkm_acr_hsf_fwif *fwif)
@@ -115,24 +109,11 @@ MODULE_FIRMWARE("nvidia/tu117/acr/ucode_unload.bin");
 
 static const struct nvkm_acr_hsf_fwif
 tu102_acr_unload_fwif[] = {
-	{  0, nvkm_acr_hsfw_load, &gp108_acr_unload_0 },
+	{  0, gm200_acr_hsfw_ctor, &gp108_acr_hsfw_0, NVKM_ACR_HSF_PMU, 0, 0x00000000 },
 	{ -1, tu102_acr_hsfw_nofw },
 	{}
 };
 
-static int
-tu102_acr_asb_load(struct nvkm_acr *acr, struct nvkm_acr_hsfw *hsfw)
-{
-	return gm200_acr_hsfw_load(acr, hsfw, &acr->subdev.device->gsp->falcon);
-}
-
-static const struct nvkm_acr_hsf_func
-tu102_acr_asb_0 = {
-	.load = tu102_acr_asb_load,
-	.boot = tu102_acr_hsfw_boot,
-	.bld = gp108_acr_hsfw_bld,
-};
-
 MODULE_FIRMWARE("nvidia/tu102/acr/ucode_asb.bin");
 MODULE_FIRMWARE("nvidia/tu104/acr/ucode_asb.bin");
 MODULE_FIRMWARE("nvidia/tu106/acr/ucode_asb.bin");
@@ -141,18 +122,11 @@ MODULE_FIRMWARE("nvidia/tu117/acr/ucode_asb.bin");
 
 static const struct nvkm_acr_hsf_fwif
 tu102_acr_asb_fwif[] = {
-	{  0, nvkm_acr_hsfw_load, &tu102_acr_asb_0 },
+	{  0, gm200_acr_hsfw_ctor, &gp108_acr_hsfw_0, NVKM_ACR_HSF_GSP, 0, 0x00000000 },
 	{ -1, tu102_acr_hsfw_nofw },
 	{}
 };
 
-static const struct nvkm_acr_hsf_func
-tu102_acr_ahesasc_0 = {
-	.load = gp102_acr_load_load,
-	.boot = tu102_acr_hsfw_boot,
-	.bld = gp108_acr_hsfw_bld,
-};
-
 MODULE_FIRMWARE("nvidia/tu102/acr/bl.bin");
 MODULE_FIRMWARE("nvidia/tu102/acr/ucode_ahesasc.bin");
 
@@ -170,7 +144,7 @@ MODULE_FIRMWARE("nvidia/tu117/acr/ucode_ahesasc.bin");
 
 static const struct nvkm_acr_hsf_fwif
 tu102_acr_ahesasc_fwif[] = {
-	{  0, nvkm_acr_hsfw_load, &tu102_acr_ahesasc_0 },
+	{  0, gm200_acr_hsfw_ctor, &gp108_acr_load_0, NVKM_ACR_HSF_SEC2, 0, 0x00000000 },
 	{ -1, tu102_acr_hsfw_nofw },
 	{}
 };
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/gv100.c b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/gv100.c
index a247e57f7d9f..da6a809cd317 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/gv100.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/gsp/gv100.c
@@ -27,15 +27,11 @@ gv100_gsp_flcn = {
 	.enable = gm200_flcn_enable,
 	.reset_eng = gp102_flcn_reset_eng,
 	.reset_wait_mem_scrubbing = gm200_flcn_reset_wait_mem_scrubbing,
-	.fbif = 0x600,
-	.load_imem = nvkm_falcon_v1_load_imem,
-	.load_dmem = nvkm_falcon_v1_load_dmem,
-	.read_dmem = nvkm_falcon_v1_read_dmem,
-	.bind_context = gp102_sec2_flcn_bind_context,
-	.wait_for_halt = nvkm_falcon_v1_wait_for_halt,
-	.clear_interrupt = nvkm_falcon_v1_clear_interrupt,
-	.set_start_addr = nvkm_falcon_v1_set_start_addr,
-	.start = nvkm_falcon_v1_start,
+	.bind_inst = gm200_flcn_bind_inst,
+	.bind_stat = gm200_flcn_bind_stat,
+	.bind_intr = true,
+	.imem_pio = &gm200_flcn_imem_pio,
+	.dmem_pio = &gm200_flcn_dmem_pio,
 };
 
 static const struct nvkm_gsp_func
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm200.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm200.c
index 0bd854092da9..7359991f94c2 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm200.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gm200.c
@@ -23,6 +23,25 @@
  */
 #include "priv.h"
 
+static int
+gm200_pmu_flcn_bind_stat(struct nvkm_falcon *falcon, bool intr)
+{
+	nvkm_falcon_wr32(falcon, 0x200, 0x0000030e);
+	return (nvkm_falcon_rd32(falcon, 0x20c) & 0x00007000) >> 12;
+}
+
+void
+gm200_pmu_flcn_bind_inst(struct nvkm_falcon *falcon, int target, u64 addr)
+{
+	nvkm_falcon_wr32(falcon, 0xe00, 4); /* DMAIDX_UCODE */
+	nvkm_falcon_wr32(falcon, 0xe04, 0); /* DMAIDX_VIRT */
+	nvkm_falcon_wr32(falcon, 0xe08, 4); /* DMAIDX_PHYS_VID */
+	nvkm_falcon_wr32(falcon, 0xe0c, 5); /* DMAIDX_PHYS_SYS_COH */
+	nvkm_falcon_wr32(falcon, 0xe10, 6); /* DMAIDX_PHYS_SYS_NCOH */
+	nvkm_falcon_mask(falcon, 0x090, 0x00010000, 0x00010000);
+	nvkm_falcon_wr32(falcon, 0x480, (1 << 30) | (target << 28) | (addr >> 12));
+}
+
 const struct nvkm_falcon_func
 gm200_pmu_flcn = {
 	.disable = gm200_flcn_disable,
@@ -30,14 +49,10 @@ gm200_pmu_flcn = {
 	.reset_pmc = true,
 	.reset_wait_mem_scrubbing = gm200_flcn_reset_wait_mem_scrubbing,
 	.debug = 0xc08,
-	.fbif = 0xe00,
-	.load_imem = nvkm_falcon_v1_load_imem,
-	.load_dmem = nvkm_falcon_v1_load_dmem,
-	.read_dmem = nvkm_falcon_v1_read_dmem,
-	.bind_context = nvkm_falcon_v1_bind_context,
-	.wait_for_halt = nvkm_falcon_v1_wait_for_halt,
-	.clear_interrupt = nvkm_falcon_v1_clear_interrupt,
-	.set_start_addr = nvkm_falcon_v1_set_start_addr,
+	.bind_inst = gm200_pmu_flcn_bind_inst,
+	.bind_stat = gm200_pmu_flcn_bind_stat,
+	.imem_pio = &gm200_flcn_imem_pio,
+	.dmem_pio = &gm200_flcn_dmem_pio,
 	.start = nvkm_falcon_v1_start,
 	.cmdq = { 0x4a0, 0x4b0, 4 },
 	.msgq = { 0x4c8, 0x4cc, 0 },
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gp102.c b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gp102.c
index 47c7412f86e8..cd3148360996 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gp102.c
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/gp102.c
@@ -30,14 +30,10 @@ gp102_pmu_flcn = {
 	.reset_eng = gp102_flcn_reset_eng,
 	.reset_wait_mem_scrubbing = gm200_flcn_reset_wait_mem_scrubbing,
 	.debug = 0xc08,
-	.fbif = 0xe00,
-	.load_imem = nvkm_falcon_v1_load_imem,
-	.load_dmem = nvkm_falcon_v1_load_dmem,
-	.read_dmem = nvkm_falcon_v1_read_dmem,
-	.bind_context = nvkm_falcon_v1_bind_context,
-	.wait_for_halt = nvkm_falcon_v1_wait_for_halt,
-	.clear_interrupt = nvkm_falcon_v1_clear_interrupt,
-	.set_start_addr = nvkm_falcon_v1_set_start_addr,
+	.bind_inst = gm200_pmu_flcn_bind_inst,
+	.bind_stat = gm200_flcn_bind_stat,
+	.imem_pio = &gm200_flcn_imem_pio,
+	.dmem_pio = &gm200_flcn_dmem_pio,
 	.start = nvkm_falcon_v1_start,
 	.cmdq = { 0x4a0, 0x4b0, 4 },
 	.msgq = { 0x4c8, 0x4cc, 0 },
diff --git a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/priv.h b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/priv.h
index 75ede265a52c..2d0a8fa6f196 100644
--- a/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/priv.h
+++ b/drivers/gpu/drm/nouveau/nvkm/subdev/pmu/priv.h
@@ -46,6 +46,7 @@ void gp102_pmu_reset(struct nvkm_pmu *pmu);
 void gk110_pmu_pgob(struct nvkm_pmu *, bool);
 
 extern const struct nvkm_falcon_func gm200_pmu_flcn;
+void gm200_pmu_flcn_bind_inst(struct nvkm_falcon *, int, u64);
 
 extern const struct nvkm_pmu_func gm20b_pmu;
 void gm20b_pmu_acr_bld_patch(struct nvkm_acr *, u32, s64);