drm/nve0/gr: initial fuc implementation, based on fermi's code

Currently identical except the available chipset register lists.  This will
*not* currently work and is disabled by default because of this.

May get merged again later, remains to be seen what further changes will be
required.

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

6 files changed, 3235 insertions, 69 deletions

diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc
new file mode 100644
index 00000000000..7b715fda276
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc
@@ -0,0 +1,451 @@
+/* fuc microcode for nve0 PGRAPH/GPC
+ *
+ * Copyright 2011 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.
+ *
+ * Authors: Ben Skeggs
+ */
+
+/* To build:
+ *    m4 nve0_grgpc.fuc | envyas -a -w -m fuc -V nva3 -o nve0_grgpc.fuc.h
+ */
+
+/* TODO
+ * - bracket certain functions with scratch writes, useful for debugging
+ * - watchdog timer around ctx operations
+ */
+
+.section #nve0_grgpc_data
+include(`nve0.fuc')
+gpc_id:			.b32 0
+gpc_mmio_list_head:	.b32 0
+gpc_mmio_list_tail:	.b32 0
+
+tpc_count:		.b32 0
+tpc_mask:		.b32 0
+tpc_mmio_list_head:	.b32 0
+tpc_mmio_list_tail:	.b32 0
+
+cmd_queue:		queue_init
+
+// chipset descriptions
+chipsets:
+.b8  0xe4 0 0 0
+.b16 #nve4_gpc_mmio_head
+.b16 #nve4_gpc_mmio_tail
+.b16 #nve4_tpc_mmio_head
+.b16 #nve4_tpc_mmio_tail
+.b8  0xe7 0 0 0
+.b16 #nve4_gpc_mmio_head
+.b16 #nve4_gpc_mmio_tail
+.b16 #nve4_tpc_mmio_head
+.b16 #nve4_tpc_mmio_tail
+.b8  0 0 0 0
+
+// GPC mmio lists
+nve4_gpc_mmio_head:
+mmctx_data(0x000380, 1)
+mmctx_data(0x000400, 2)
+mmctx_data(0x00040c, 3)
+mmctx_data(0x000450, 9)
+mmctx_data(0x000600, 1)
+mmctx_data(0x000684, 1)
+mmctx_data(0x000700, 5)
+mmctx_data(0x000800, 1)
+mmctx_data(0x000808, 3)
+mmctx_data(0x000828, 1)
+mmctx_data(0x000830, 1)
+mmctx_data(0x0008d8, 1)
+mmctx_data(0x0008e0, 1)
+mmctx_data(0x0008e8, 6)
+mmctx_data(0x00091c, 1)
+mmctx_data(0x000924, 3)
+mmctx_data(0x000b00, 1)
+mmctx_data(0x000b08, 6)
+mmctx_data(0x000bb8, 1)
+mmctx_data(0x000c08, 1)
+mmctx_data(0x000c10, 8)
+mmctx_data(0x000c40, 1)
+mmctx_data(0x000c6c, 1)
+mmctx_data(0x000c80, 1)
+mmctx_data(0x000c8c, 1)
+mmctx_data(0x001000, 3)
+mmctx_data(0x001014, 1)
+mmctx_data(0x003024, 1)
+mmctx_data(0x0030c0, 2)
+mmctx_data(0x0030e4, 1)
+mmctx_data(0x003100, 6)
+mmctx_data(0x0031d0, 1)
+mmctx_data(0x0031e0, 2)
+nve4_gpc_mmio_tail:
+
+// TPC mmio lists
+nve4_tpc_mmio_head:
+mmctx_data(0x000048, 1)
+mmctx_data(0x000064, 1)
+mmctx_data(0x000088, 1)
+mmctx_data(0x000200, 6)
+mmctx_data(0x00021c, 2)
+mmctx_data(0x000230, 1)
+mmctx_data(0x0002c4, 1)
+mmctx_data(0x000400, 3)
+mmctx_data(0x000420, 3)
+mmctx_data(0x0004e8, 1)
+mmctx_data(0x0004f4, 1)
+mmctx_data(0x000604, 4)
+mmctx_data(0x000644, 22)
+mmctx_data(0x0006ac, 2)
+mmctx_data(0x0006c8, 1)
+mmctx_data(0x000730, 8)
+mmctx_data(0x000758, 1)
+mmctx_data(0x000778, 1)
+nve4_tpc_mmio_tail:
+
+.section #nve0_grgpc_code
+bra #init
+define(`include_code')
+include(`nve0.fuc')
+
+// reports an exception to the host
+//
+// In: $r15 error code (see nve0.fuc)
+//
+error:
+	push $r14
+	mov $r14 -0x67ec 	// 0x9814
+	sethi $r14 0x400000
+	call #nv_wr32		// HUB_CTXCTL_CC_SCRATCH[5] = error code
+	add b32 $r14 0x41c
+	mov $r15 1
+	call #nv_wr32		// HUB_CTXCTL_INTR_UP_SET
+	pop $r14
+	ret
+
+// GPC fuc initialisation, executed by triggering ucode start, will
+// fall through to main loop after completion.
+//
+// Input:
+//   CC_SCRATCH[0]: chipset (PMC_BOOT_0 read returns 0x0bad0bad... sigh)
+//   CC_SCRATCH[1]: context base
+//
+// Output:
+//   CC_SCRATCH[0]:
+//	     31:31: set to signal completion
+//   CC_SCRATCH[1]:
+//	      31:0: GPC context size
+//
+init:
+	clear b32 $r0
+	mov $sp $r0
+
+	// enable fifo access
+	mov $r1 0x1200
+	mov $r2 2
+	iowr I[$r1 + 0x000] $r2		// FIFO_ENABLE
+
+	// setup i0 handler, and route all interrupts to it
+	mov $r1 #ih
+	mov $iv0 $r1
+	mov $r1 0x400
+	iowr I[$r1 + 0x300] $r0		// INTR_DISPATCH
+
+	// enable fifo interrupt
+	mov $r2 4
+	iowr I[$r1 + 0x000] $r2		// INTR_EN_SET
+
+	// enable interrupts
+	bset $flags ie0
+
+	// figure out which GPC we are, and how many TPCs we have
+	mov $r1 0x608
+	shl b32 $r1 6
+	iord $r2 I[$r1 + 0x000]		// UNITS
+	mov $r3 1
+	and $r2 0x1f
+	shl b32 $r3 $r2
+	sub b32 $r3 1
+	st b32 D[$r0 + #tpc_count] $r2
+	st b32 D[$r0 + #tpc_mask] $r3
+	add b32 $r1 0x400
+	iord $r2 I[$r1 + 0x000]		// MYINDEX
+	st b32 D[$r0 + #gpc_id] $r2
+
+	// find context data for this chipset
+	mov $r2 0x800
+	shl b32 $r2 6
+	iord $r2 I[$r2 + 0x000]		// CC_SCRATCH[0]
+	mov $r1 #chipsets - 12
+	init_find_chipset:
+		add b32 $r1 12
+		ld b32 $r3 D[$r1 + 0x00]
+		cmpu b32 $r3 $r2
+		bra e #init_context
+		cmpu b32 $r3 0
+		bra ne #init_find_chipset
+		// unknown chipset
+		ret
+
+	// initialise context base, and size tracking
+	init_context:
+	mov $r2 0x800
+	shl b32 $r2 6
+	iord $r2 I[$r2 + 0x100]	// CC_SCRATCH[1], initial base
+	clear b32 $r3		// track GPC context size here
+
+	// set mmctx base addresses now so we don't have to do it later,
+	// they don't currently ever change
+	mov $r4 0x700
+	shl b32 $r4 6
+	shr b32 $r5 $r2 8
+	iowr I[$r4 + 0x000] $r5		// MMCTX_SAVE_SWBASE
+	iowr I[$r4 + 0x100] $r5		// MMCTX_LOAD_SWBASE
+
+	// calculate GPC mmio context size, store the chipset-specific
+	// mmio list pointers somewhere we can get at them later without
+	// re-parsing the chipset list
+	clear b32 $r14
+	clear b32 $r15
+	ld b16 $r14 D[$r1 + 4]
+	ld b16 $r15 D[$r1 + 6]
+	st b16 D[$r0 + #gpc_mmio_list_head] $r14
+	st b16 D[$r0 + #gpc_mmio_list_tail] $r15
+	call #mmctx_size
+	add b32 $r2 $r15
+	add b32 $r3 $r15
+
+	// calculate per-TPC mmio context size, store the list pointers
+	ld b16 $r14 D[$r1 + 8]
+	ld b16 $r15 D[$r1 + 10]
+	st b16 D[$r0 + #tpc_mmio_list_head] $r14
+	st b16 D[$r0 + #tpc_mmio_list_tail] $r15
+	call #mmctx_size
+	ld b32 $r14 D[$r0 + #tpc_count]
+	mulu $r14 $r15
+	add b32 $r2 $r14
+	add b32 $r3 $r14
+
+	// round up base/size to 256 byte boundary (for strand SWBASE)
+	add b32 $r4 0x1300
+	shr b32 $r3 2
+	iowr I[$r4 + 0x000] $r3		// MMCTX_LOAD_COUNT, wtf for?!?
+	shr b32 $r2 8
+	shr b32 $r3 6
+	add b32 $r2 1
+	add b32 $r3 1
+	shl b32 $r2 8
+	shl b32 $r3 8
+
+	// calculate size of strand context data
+	mov b32 $r15 $r2
+	call #strand_ctx_init
+	add b32 $r3 $r15
+
+	// save context size, and tell HUB we're done
+	mov $r1 0x800
+	shl b32 $r1 6
+	iowr I[$r1 + 0x100] $r3		// CC_SCRATCH[1]  = context size
+	add b32 $r1 0x800
+	clear b32 $r2
+	bset $r2 31
+	iowr I[$r1 + 0x000] $r2		// CC_SCRATCH[0] |= 0x80000000
+
+// Main program loop, very simple, sleeps until woken up by the interrupt
+// handler, pulls a command from the queue and executes its handler
+//
+main:
+	bset $flags $p0
+	sleep $p0
+	mov $r13 #cmd_queue
+	call #queue_get
+	bra $p1 #main
+
+	// 0x0000-0x0003 are all context transfers
+	cmpu b32 $r14 0x04
+	bra nc #main_not_ctx_xfer
+		// fetch $flags and mask off $p1/$p2
+		mov $r1 $flags
+		mov $r2 0x0006
+		not b32 $r2
+		and $r1 $r2
+		// set $p1/$p2 according to transfer type
+		shl b32 $r14 1
+		or $r1 $r14
+		mov $flags $r1
+		// transfer context data
+		call #ctx_xfer
+		bra #main
+
+	main_not_ctx_xfer:
+	shl b32 $r15 $r14 16
+	or $r15 E_BAD_COMMAND
+	call #error
+	bra #main
+
+// interrupt handler
+ih:
+	push $r8
+	mov $r8 $flags
+	push $r8
+	push $r9
+	push $r10
+	push $r11
+	push $r13
+	push $r14
+	push $r15
+
+	// incoming fifo command?
+	iord $r10 I[$r0 + 0x200]	// INTR
+	and $r11 $r10 0x00000004
+	bra e #ih_no_fifo
+		// queue incoming fifo command for later processing
+		mov $r11 0x1900
+		mov $r13 #cmd_queue
+		iord $r14 I[$r11 + 0x100]	// FIFO_CMD
+		iord $r15 I[$r11 + 0x000]	// FIFO_DATA
+		call #queue_put
+		add b32 $r11 0x400
+		mov $r14 1
+		iowr I[$r11 + 0x000] $r14	// FIFO_ACK
+
+	// ack, and wake up main()
+	ih_no_fifo:
+	iowr I[$r0 + 0x100] $r10	// INTR_ACK
+
+	pop $r15
+	pop $r14
+	pop $r13
+	pop $r11
+	pop $r10
+	pop $r9
+	pop $r8
+	mov $flags $r8
+	pop $r8
+	bclr $flags $p0
+	iret
+
+// Set this GPC's bit in HUB_BAR, used to signal completion of various
+// activities to the HUB fuc
+//
+hub_barrier_done:
+	mov $r15 1
+	ld b32 $r14 D[$r0 + #gpc_id]
+	shl b32 $r15 $r14
+	mov $r14 -0x6be8 	// 0x409418 - HUB_BAR_SET
+	sethi $r14 0x400000
+	call #nv_wr32
+	ret
+
+// Disables various things, waits a bit, and re-enables them..
+//
+// Not sure how exactly this helps, perhaps "ENABLE" is not such a
+// good description for the bits we turn off?  Anyways, without this,
+// funny things happen.
+//
+ctx_redswitch:
+	mov $r14 0x614
+	shl b32 $r14 6
+	mov $r15 0x020
+	iowr I[$r14] $r15	// GPC_RED_SWITCH = POWER
+	mov $r15 8
+	ctx_redswitch_delay:
+		sub b32 $r15 1
+		bra ne #ctx_redswitch_delay
+	mov $r15 0xa20
+	iowr I[$r14] $r15	// GPC_RED_SWITCH = UNK11, ENABLE, POWER
+	ret
+
+// Transfer GPC context data between GPU and storage area
+//
+// In: $r15 context base address
+//     $p1 clear on save, set on load
+//     $p2 set if opposite direction done/will be done, so:
+//		on save it means: "a load will follow this save"
+//		on load it means: "a save preceeded this load"
+//
+ctx_xfer:
+	// set context base address
+	mov $r1 0xa04
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r15// MEM_BASE
+	bra not $p1 #ctx_xfer_not_load
+		call #ctx_redswitch
+	ctx_xfer_not_load:
+
+	// strands
+	mov $r1 0x4afc
+	sethi $r1 0x20000
+	mov $r2 0xc
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x0c
+	call #strand_wait
+	mov $r2 0x47fc
+	sethi $r2 0x20000
+	iowr I[$r2] $r0		// STRAND_FIRST_GENE(0x3f) = 0x00
+	xbit $r2 $flags $p1
+	add b32 $r2 3
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x03/0x04 (SAVE/LOAD)
+
+	// mmio context
+	xbit $r10 $flags $p1	// direction
+	or $r10 2		// first
+	mov $r11 0x0000
+	sethi $r11 0x500000
+	ld b32 $r12 D[$r0 + #gpc_id]
+	shl b32 $r12 15
+	add b32 $r11 $r12	// base = NV_PGRAPH_GPCn
+	ld b32 $r12 D[$r0 + #gpc_mmio_list_head]
+	ld b32 $r13 D[$r0 + #gpc_mmio_list_tail]
+	mov $r14 0		// not multi
+	call #mmctx_xfer
+
+	// per-TPC mmio context
+	xbit $r10 $flags $p1	// direction
+	or $r10 4		// last
+	mov $r11 0x4000
+	sethi $r11 0x500000	// base = NV_PGRAPH_GPC0_TPC0
+	ld b32 $r12 D[$r0 + #gpc_id]
+	shl b32 $r12 15
+	add b32 $r11 $r12	// base = NV_PGRAPH_GPCn_TPC0
+	ld b32 $r12 D[$r0 + #tpc_mmio_list_head]
+	ld b32 $r13 D[$r0 + #tpc_mmio_list_tail]
+	ld b32 $r15 D[$r0 + #tpc_mask]
+	mov $r14 0x800		// stride = 0x800
+	call #mmctx_xfer
+
+	// wait for strands to finish
+	call #strand_wait
+
+	// if load, or a save without a load following, do some
+	// unknown stuff that's done after finishing a block of
+	// strand commands
+	bra $p1 #ctx_xfer_post
+	bra not $p2 #ctx_xfer_done
+	ctx_xfer_post:
+		mov $r1 0x4afc
+		sethi $r1 0x20000
+		mov $r2 0xd
+		iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x0d
+		call #strand_wait
+
+	// mark completion in HUB's barrier
+	ctx_xfer_done:
+	call #hub_barrier_done
+	ret
+
+.align 256
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc.h b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc.h
new file mode 100644
index 00000000000..26c2165bad0
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc.h
@@ -0,0 +1,530 @@
+uint32_t nve0_grgpc_data[] = {
+/* 0x0000: gpc_id */
+	0x00000000,
+/* 0x0004: gpc_mmio_list_head */
+	0x00000000,
+/* 0x0008: gpc_mmio_list_tail */
+	0x00000000,
+/* 0x000c: tpc_count */
+	0x00000000,
+/* 0x0010: tpc_mask */
+	0x00000000,
+/* 0x0014: tpc_mmio_list_head */
+	0x00000000,
+/* 0x0018: tpc_mmio_list_tail */
+	0x00000000,
+/* 0x001c: cmd_queue */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0064: chipsets */
+	0x000000e4,
+	0x01040080,
+	0x014c0104,
+	0x000000e7,
+	0x01040080,
+	0x014c0104,
+	0x00000000,
+/* 0x0080: nve4_gpc_mmio_head */
+	0x00000380,
+	0x04000400,
+	0x0800040c,
+	0x20000450,
+	0x00000600,
+	0x00000684,
+	0x10000700,
+	0x00000800,
+	0x08000808,
+	0x00000828,
+	0x00000830,
+	0x000008d8,
+	0x000008e0,
+	0x140008e8,
+	0x0000091c,
+	0x08000924,
+	0x00000b00,
+	0x14000b08,
+	0x00000bb8,
+	0x00000c08,
+	0x1c000c10,
+	0x00000c40,
+	0x00000c6c,
+	0x00000c80,
+	0x00000c8c,
+	0x08001000,
+	0x00001014,
+	0x00003024,
+	0x040030c0,
+	0x000030e4,
+	0x14003100,
+	0x000031d0,
+	0x040031e0,
+/* 0x0104: nve4_gpc_mmio_tail */
+/* 0x0104: nve4_tpc_mmio_head */
+	0x00000048,
+	0x00000064,
+	0x00000088,
+	0x14000200,
+	0x0400021c,
+	0x00000230,
+	0x000002c4,
+	0x08000400,
+	0x08000420,
+	0x000004e8,
+	0x000004f4,
+	0x0c000604,
+	0x54000644,
+	0x040006ac,
+	0x000006c8,
+	0x1c000730,
+	0x00000758,
+	0x00000778,
+};
+
+uint32_t nve0_grgpc_code[] = {
+	0x03060ef5,
+/* 0x0004: queue_put */
+	0x9800d898,
+	0x86f001d9,
+	0x0489b808,
+	0xf00c1bf4,
+	0x21f502f7,
+	0x00f802ec,
+/* 0x001c: queue_put_next */
+	0xb60798c4,
+	0x8dbb0384,
+	0x0880b600,
+	0x80008e80,
+	0x90b6018f,
+	0x0f94f001,
+	0xf801d980,
+/* 0x0039: queue_get */
+	0x0131f400,
+	0x9800d898,
+	0x89b801d9,
+	0x210bf404,
+	0xb60789c4,
+	0x9dbb0394,
+	0x0890b600,
+	0x98009e98,
+	0x80b6019f,
+	0x0f84f001,
+	0xf400d880,
+/* 0x0066: queue_get_done */
+	0x00f80132,
+/* 0x0068: nv_rd32 */
+	0x0728b7f1,
+	0xb906b4b6,
+	0xc9f002ec,
+	0x00bcd01f,
+/* 0x0078: nv_rd32_wait */
+	0xc800bccf,
+	0x1bf41fcc,
+	0x06a7f0fa,
+	0x010321f5,
+	0xf840bfcf,
+/* 0x008d: nv_wr32 */
+	0x28b7f100,
+	0x06b4b607,
+	0xb980bfd0,
+	0xc9f002ec,
+	0x1ec9f01f,
+/* 0x00a3: nv_wr32_wait */
+	0xcf00bcd0,
+	0xccc800bc,
+	0xfa1bf41f,
+/* 0x00ae: watchdog_reset */
+	0x87f100f8,
+	0x84b60430,
+	0x1ff9f006,
+	0xf8008fd0,
+/* 0x00bd: watchdog_clear */
+	0x3087f100,
+	0x0684b604,
+	0xf80080d0,
+/* 0x00c9: wait_donez */
+	0x3c87f100,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d000,
+	0x081887f1,
+	0xd00684b6,
+/* 0x00e2: wait_done_wait_donez */
+	0x87f1008a,
+	0x84b60400,
+	0x0088cf06,
+	0xf4888aff,
+	0x87f1f31b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00099,
+/* 0x0103: wait_doneo */
+	0xf100f800,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00099f0,
+	0x87f10089,
+	0x84b60818,
+	0x008ad006,
+/* 0x011c: wait_done_wait_doneo */
+	0x040087f1,
+	0xcf0684b6,
+	0x8aff0088,
+	0xf30bf488,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0099f094,
+	0xf80089d0,
+/* 0x013d: mmctx_size */
+/* 0x013f: nv_mmctx_size_loop */
+	0x9894bd00,
+	0x85b600e8,
+	0x0180b61a,
+	0xbb0284b6,
+	0xe0b60098,
+	0x04efb804,
+	0xb9eb1bf4,
+	0x00f8029f,
+/* 0x015c: mmctx_xfer */
+	0x083c87f1,
+	0xbd0684b6,
+	0x0199f094,
+	0xf10089d0,
+	0xb6071087,
+	0x94bd0684,
+	0xf405bbfd,
+	0x8bd0090b,
+	0x0099f000,
+/* 0x0180: mmctx_base_disabled */
+	0xf405eefd,
+	0x8ed00c0b,
+	0xc08fd080,
+/* 0x018f: mmctx_multi_disabled */
+	0xb70199f0,
+	0xc8010080,
+	0xb4b600ab,
+	0x0cb9f010,
+	0xb601aec8,
+	0xbefd11e4,
+	0x008bd005,
+/* 0x01a8: mmctx_exec_loop */
+/* 0x01a8: mmctx_wait_free */
+	0xf0008ecf,
+	0x0bf41fe4,
+	0x00ce98fa,
+	0xd005e9fd,
+	0xc0b6c08e,
+	0x04cdb804,
+	0xc8e81bf4,
+	0x1bf402ab,
+/* 0x01c9: mmctx_fini_wait */
+	0x008bcf18,
+	0xb01fb4f0,
+	0x1bf410b4,
+	0x02a7f0f7,
+	0xf4c921f4,
+/* 0x01de: mmctx_stop */
+	0xabc81b0e,
+	0x10b4b600,
+	0xf00cb9f0,
+	0x8bd012b9,
+/* 0x01ed: mmctx_stop_wait */
+	0x008bcf00,
+	0xf412bbc8,
+/* 0x01f6: mmctx_done */
+	0x87f1fa1b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00199,
+/* 0x0207: strand_wait */
+	0xf900f800,
+	0x02a7f0a0,
+	0xfcc921f4,
+/* 0x0213: strand_pre */
+	0xf100f8a0,
+	0xf04afc87,
+	0x97f00283,
+	0x0089d00c,
+	0x020721f5,
+/* 0x0226: strand_post */
+	0x87f100f8,
+	0x83f04afc,
+	0x0d97f002,
+	0xf50089d0,
+	0xf8020721,
+/* 0x0239: strand_set */
+	0xfca7f100,
+	0x02a3f04f,
+	0x0500aba2,
+	0xd00fc7f0,
+	0xc7f000ac,
+	0x00bcd00b,
+	0x020721f5,
+	0xf000aed0,
+	0xbcd00ac7,
+	0x0721f500,
+/* 0x0263: strand_ctx_init */
+	0xf100f802,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x21f50089,
+	0xe7f00213,
+	0x3921f503,
+	0xfca7f102,
+	0x02a3f046,
+	0x0400aba0,
+	0xf040a0d0,
+	0xbcd001c7,
+	0x0721f500,
+	0x010c9202,
+	0xf000acd0,
+	0xbcd002c7,
+	0x0721f500,
+	0x2621f502,
+	0x8087f102,
+	0x0684b608,
+	0xb70089cf,
+	0x95220080,
+/* 0x02ba: ctx_init_strand_loop */
+	0x8ed008fe,
+	0x408ed000,
+	0xb6808acf,
+	0xa0b606a5,
+	0x00eabb01,
+	0xb60480b6,
+	0x1bf40192,
+	0x08e4b6e8,
+	0xf1f2efbc,
+	0xb6085c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x00f80089,
+/* 0x02ec: error */
+	0xe7f1e0f9,
+	0xe3f09814,
+	0x8d21f440,
+	0x041ce0b7,
+	0xf401f7f0,
+	0xe0fc8d21,
+/* 0x0306: init */
+	0x04bd00f8,
+	0xf10004fe,
+	0xf0120017,
+	0x12d00227,
+	0x3e17f100,
+	0x0010fe04,
+	0x040017f1,
+	0xf0c010d0,
+	0x12d00427,
+	0x1031f400,
+	0x060817f1,
+	0xcf0614b6,
+	0x37f00012,
+	0x1f24f001,
+	0xb60432bb,
+	0x02800132,
+	0x04038003,
+	0x040010b7,
+	0x800012cf,
+	0x27f10002,
+	0x24b60800,
+	0x0022cf06,
+/* 0x035f: init_find_chipset */
+	0xb65817f0,
+	0x13980c10,
+	0x0432b800,
+	0xb00b0bf4,
+	0x1bf40034,
+/* 0x0373: init_context */
+	0xf100f8f1,
+	0xb6080027,
+	0x22cf0624,
+	0xf134bd40,
+	0xb6070047,
+	0x25950644,
+	0x0045d008,
+	0xbd4045d0,
+	0x58f4bde4,
+	0x1f58021e,
+	0x020e4003,
+	0xf5040f40,
+	0xbb013d21,
+	0x3fbb002f,
+	0x041e5800,
+	0x40051f58,
+	0x0f400a0e,
+	0x3d21f50c,
+	0x030e9801,
+	0xbb00effd,
+	0x3ebb002e,
+	0x0040b700,
+	0x0235b613,
+	0xb60043d0,
+	0x35b60825,
+	0x0120b606,
+	0xb60130b6,
+	0x34b60824,
+	0x022fb908,
+	0x026321f5,
+	0xf1003fbb,
+	0xb6080017,
+	0x13d00614,
+	0x0010b740,
+	0xf024bd08,
+	0x12d01f29,
+/* 0x0401: main */
+	0x0031f400,
+	0xf00028f4,
+	0x21f41cd7,
+	0xf401f439,
+	0xf404e4b0,
+	0x81fe1e18,
+	0x0627f001,
+	0x12fd20bd,
+	0x01e4b604,
+	0xfe051efd,
+	0x21f50018,
+	0x0ef404c3,
+/* 0x0431: main_not_ctx_xfer */
+	0x10ef94d3,
+	0xf501f5f0,
+	0xf402ec21,
+/* 0x043e: ih */
+	0x80f9c60e,
+	0xf90188fe,
+	0xf990f980,
+	0xf9b0f9a0,
+	0xf9e0f9d0,
+	0x800acff0,
+	0xf404abc4,
+	0xb7f11d0b,
+	0xd7f01900,
+	0x40becf1c,
+	0xf400bfcf,
+	0xb0b70421,
+	0xe7f00400,
+	0x00bed001,
+/* 0x0474: ih_no_fifo */
+	0xfc400ad0,
+	0xfce0fcf0,
+	0xfcb0fcd0,
+	0xfc90fca0,
+	0x0088fe80,
+	0x32f480fc,
+/* 0x048f: hub_barrier_done */
+	0xf001f800,
+	0x0e9801f7,
+	0x04febb00,
+	0x9418e7f1,
+	0xf440e3f0,
+	0x00f88d21,
+/* 0x04a4: ctx_redswitch */
+	0x0614e7f1,
+	0xf006e4b6,
+	0xefd020f7,
+	0x08f7f000,
+/* 0x04b4: ctx_redswitch_delay */
+	0xf401f2b6,
+	0xf7f1fd1b,
+	0xefd00a20,
+/* 0x04c3: ctx_xfer */
+	0xf100f800,
+	0xb60a0417,
+	0x1fd00614,
+	0x0711f400,
+	0x04a421f5,
+/* 0x04d4: ctx_xfer_not_load */
+	0x4afc17f1,
+	0xf00213f0,
+	0x12d00c27,
+	0x0721f500,
+	0xfc27f102,
+	0x0223f047,
+	0xf00020d0,
+	0x20b6012c,
+	0x0012d003,
+	0xf001acf0,
+	0xb7f002a5,
+	0x50b3f000,
+	0xb6000c98,
+	0xbcbb0fc4,
+	0x010c9800,
+	0xf0020d98,
+	0x21f500e7,
+	0xacf0015c,
+	0x04a5f001,
+	0x4000b7f1,
+	0x9850b3f0,
+	0xc4b6000c,
+	0x00bcbb0f,
+	0x98050c98,
+	0x0f98060d,
+	0x00e7f104,
+	0x5c21f508,
+	0x0721f501,
+	0x0601f402,
+/* 0x054b: ctx_xfer_post */
+	0xf11412f4,
+	0xf04afc17,
+	0x27f00213,
+	0x0012d00d,
+	0x020721f5,
+/* 0x055c: ctx_xfer_done */
+	0x048f21f5,
+	0x000000f8,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc
new file mode 100644
index 00000000000..816ae57343f
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc
@@ -0,0 +1,806 @@
+/* fuc microcode for nve0 PGRAPH/HUB
+ *
+ * Copyright 2011 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.
+ *
+ * Authors: Ben Skeggs
+ */
+
+/* To build:
+ *    m4 nve0_grhub.fuc | envyas -a -w -m fuc -V nva3 -o nve0_grhub.fuc.h
+ */
+
+.section #nve0_grhub_data
+include(`nve0.fuc')
+gpc_count:		.b32 0
+rop_count:		.b32 0
+cmd_queue:		queue_init
+hub_mmio_list_head:	.b32 0
+hub_mmio_list_tail:	.b32 0
+
+ctx_current:		.b32 0
+
+chipsets:
+.b8  0xe4 0 0 0
+.b16 #nve4_hub_mmio_head
+.b16 #nve4_hub_mmio_tail
+.b8  0xe7 0 0 0
+.b16 #nve4_hub_mmio_head
+.b16 #nve4_hub_mmio_tail
+.b8  0 0 0 0
+
+nve4_hub_mmio_head:
+mmctx_data(0x17e91c, 2)
+mmctx_data(0x400204, 2)
+mmctx_data(0x404010, 7)
+mmctx_data(0x4040a8, 9)
+mmctx_data(0x4040d0, 7)
+mmctx_data(0x4040f8, 1)
+mmctx_data(0x404130, 3)
+mmctx_data(0x404150, 3)
+mmctx_data(0x404164, 1)
+mmctx_data(0x4041a0, 4)
+mmctx_data(0x404200, 4)
+mmctx_data(0x404404, 14)
+mmctx_data(0x404460, 4)
+mmctx_data(0x404480, 1)
+mmctx_data(0x404498, 1)
+mmctx_data(0x404604, 4)
+mmctx_data(0x404618, 4)
+mmctx_data(0x40462c, 2)
+mmctx_data(0x404640, 1)
+mmctx_data(0x404654, 1)
+mmctx_data(0x404660, 1)
+mmctx_data(0x404678, 19)
+mmctx_data(0x4046c8, 3)
+mmctx_data(0x404700, 3)
+mmctx_data(0x404718, 10)
+mmctx_data(0x404744, 2)
+mmctx_data(0x404754, 1)
+mmctx_data(0x405800, 1)
+mmctx_data(0x405830, 3)
+mmctx_data(0x405854, 1)
+mmctx_data(0x405870, 4)
+mmctx_data(0x405a00, 2)
+mmctx_data(0x405a18, 1)
+mmctx_data(0x405b00, 1)
+mmctx_data(0x405b10, 1)
+mmctx_data(0x406020, 1)
+mmctx_data(0x406028, 4)
+mmctx_data(0x4064a8, 2)
+mmctx_data(0x4064b4, 2)
+mmctx_data(0x4064c0, 12)
+mmctx_data(0x4064fc, 1)
+mmctx_data(0x407040, 1)
+mmctx_data(0x407804, 1)
+mmctx_data(0x40780c, 6)
+mmctx_data(0x4078bc, 1)
+mmctx_data(0x408000, 7)
+mmctx_data(0x408064, 1)
+mmctx_data(0x408800, 3)
+mmctx_data(0x408840, 1)
+mmctx_data(0x408900, 3)
+mmctx_data(0x408980, 1)
+nve4_hub_mmio_tail:
+
+.align 256
+chan_data:
+chan_mmio_count:	.b32 0
+chan_mmio_address:	.b32 0
+
+.align 256
+xfer_data: 		.b32 0
+
+.section #nve0_grhub_code
+bra #init
+define(`include_code')
+include(`nve0.fuc')
+
+// reports an exception to the host
+//
+// In: $r15 error code (see nve0.fuc)
+//
+error:
+	push $r14
+	mov $r14 0x814
+	shl b32 $r14 6
+	iowr I[$r14 + 0x000] $r15	// CC_SCRATCH[5] = error code
+	mov $r14 0xc1c
+	shl b32 $r14 6
+	mov $r15 1
+	iowr I[$r14 + 0x000] $r15	// INTR_UP_SET
+	pop $r14
+	ret
+
+// HUB fuc initialisation, executed by triggering ucode start, will
+// fall through to main loop after completion.
+//
+// Input:
+//   CC_SCRATCH[0]: chipset (PMC_BOOT_0 read returns 0x0bad0bad... sigh)
+//
+// Output:
+//   CC_SCRATCH[0]:
+//	     31:31: set to signal completion
+//   CC_SCRATCH[1]:
+//	      31:0: total PGRAPH context size
+//
+init:
+	clear b32 $r0
+	mov $sp $r0
+	mov $xdbase $r0
+
+	// enable fifo access
+	mov $r1 0x1200
+	mov $r2 2
+	iowr I[$r1 + 0x000] $r2	// FIFO_ENABLE
+
+	// setup i0 handler, and route all interrupts to it
+	mov $r1 #ih
+	mov $iv0 $r1
+	mov $r1 0x400
+	iowr I[$r1 + 0x300] $r0	// INTR_DISPATCH
+
+	// route HUB_CHANNEL_SWITCH to fuc interrupt 8
+	mov $r3 0x404
+	shl b32 $r3 6
+	mov $r2 0x2003		// { HUB_CHANNEL_SWITCH, ZERO } -> intr 8
+	iowr I[$r3 + 0x000] $r2
+
+	// not sure what these are, route them because NVIDIA does, and
+	// the IRQ handler will signal the host if we ever get one.. we
+	// may find out if/why we need to handle these if so..
+	//
+	mov $r2 0x2004
+	iowr I[$r3 + 0x004] $r2 // { 0x04, ZERO } -> intr 9
+	mov $r2 0x200b
+	iowr I[$r3 + 0x008] $r2 // { 0x0b, ZERO } -> intr 10
+	mov $r2 0x200c
+	iowr I[$r3 + 0x01c] $r2 // { 0x0c, ZERO } -> intr 15
+
+	// enable all INTR_UP interrupts
+	mov $r2 0xc24
+	shl b32 $r2 6
+	not b32 $r3 $r0
+	iowr I[$r2] $r3
+
+	// enable fifo, ctxsw, 9, 10, 15 interrupts
+	mov $r2 -0x78fc		// 0x8704
+	sethi $r2 0
+	iowr I[$r1 + 0x000] $r2	// INTR_EN_SET
+
+	// fifo level triggered, rest edge
+	sub b32 $r1 0x100
+	mov $r2 4
+	iowr I[$r1] $r2
+
+	// enable interrupts
+	bset $flags ie0
+
+	// fetch enabled GPC/ROP counts
+	mov $r14 -0x69fc	// 0x409604
+	sethi $r14 0x400000
+	call #nv_rd32
+	extr $r1 $r15 16:20
+	st b32 D[$r0 + #rop_count] $r1
+	and $r15 0x1f
+	st b32 D[$r0 + #gpc_count] $r15
+
+	// set BAR_REQMASK to GPC mask
+	mov $r1 1
+	shl b32 $r1 $r15
+	sub b32 $r1 1
+	mov $r2 0x40c
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r1
+	iowr I[$r2 + 0x100] $r1
+
+	// find context data for this chipset
+	mov $r2 0x800
+	shl b32 $r2 6
+	iord $r2 I[$r2 + 0x000]		// CC_SCRATCH[0]
+	mov $r15 #chipsets - 8
+	init_find_chipset:
+		add b32 $r15 8
+		ld b32 $r3 D[$r15 + 0x00]
+		cmpu b32 $r3 $r2
+		bra e #init_context
+		cmpu b32 $r3 0
+		bra ne #init_find_chipset
+		// unknown chipset
+		ret
+
+	// context size calculation, reserve first 256 bytes for use by fuc
+	init_context:
+	mov $r1 256
+
+	// calculate size of mmio context data
+	ld b16 $r14 D[$r15 + 4]
+	ld b16 $r15 D[$r15 + 6]
+	sethi $r14 0
+	st b32 D[$r0 + #hub_mmio_list_head] $r14
+	st b32 D[$r0 + #hub_mmio_list_tail] $r15
+	call #mmctx_size
+
+	// set mmctx base addresses now so we don't have to do it later,
+	// they don't (currently) ever change
+	mov $r3 0x700
+	shl b32 $r3 6
+	shr b32 $r4 $r1 8
+	iowr I[$r3 + 0x000] $r4		// MMCTX_SAVE_SWBASE
+	iowr I[$r3 + 0x100] $r4		// MMCTX_LOAD_SWBASE
+	add b32 $r3 0x1300
+	add b32 $r1 $r15
+	shr b32 $r15 2
+	iowr I[$r3 + 0x000] $r15	// MMCTX_LOAD_COUNT, wtf for?!?
+
+	// strands, base offset needs to be aligned to 256 bytes
+	shr b32 $r1 8
+	add b32 $r1 1
+	shl b32 $r1 8
+	mov b32 $r15 $r1
+	call #strand_ctx_init
+	add b32 $r1 $r15
+
+	// initialise each GPC in sequence by passing in the offset of its
+	// context data in GPCn_CC_SCRATCH[1], and starting its FUC (which
+	// has previously been uploaded by the host) running.
+	//
+	// the GPC fuc init sequence will set GPCn_CC_SCRATCH[0] bit 31
+	// when it has completed, and return the size of its context data
+	// in GPCn_CC_SCRATCH[1]
+	//
+	ld b32 $r3 D[$r0 + #gpc_count]
+	mov $r4 0x2000
+	sethi $r4 0x500000
+	init_gpc:
+		// setup, and start GPC ucode running
+		add b32 $r14 $r4 0x804
+		mov b32 $r15 $r1
+		call #nv_wr32			// CC_SCRATCH[1] = ctx offset
+		add b32 $r14 $r4 0x800
+		mov b32 $r15 $r2
+		call #nv_wr32			// CC_SCRATCH[0] = chipset
+		add b32 $r14 $r4 0x10c
+		clear b32 $r15
+		call #nv_wr32
+		add b32 $r14 $r4 0x104
+		call #nv_wr32			// ENTRY
+		add b32 $r14 $r4 0x100
+		mov $r15 2			// CTRL_START_TRIGGER
+		call #nv_wr32			// CTRL
+
+		// wait for it to complete, and adjust context size
+		add b32 $r14 $r4 0x800
+		init_gpc_wait:
+			call #nv_rd32
+			xbit $r15 $r15 31
+			bra e #init_gpc_wait
+		add b32 $r14 $r4 0x804
+		call #nv_rd32
+		add b32 $r1 $r15
+
+		// next!
+		add b32 $r4 0x8000
+		sub b32 $r3 1
+		bra ne #init_gpc
+
+	// save context size, and tell host we're ready
+	mov $r2 0x800
+	shl b32 $r2 6
+	iowr I[$r2 + 0x100] $r1		// CC_SCRATCH[1]  = context size
+	add b32 $r2 0x800
+	clear b32 $r1
+	bset $r1 31
+	iowr I[$r2 + 0x000] $r1		// CC_SCRATCH[0] |= 0x80000000
+
+// Main program loop, very simple, sleeps until woken up by the interrupt
+// handler, pulls a command from the queue and executes its handler
+//
+main:
+	// sleep until we have something to do
+	bset $flags $p0
+	sleep $p0
+	mov $r13 #cmd_queue
+	call #queue_get
+	bra $p1 #main
+
+	// context switch, requested by GPU?
+	cmpu b32 $r14 0x4001
+	bra ne #main_not_ctx_switch
+		trace_set(T_AUTO)
+		mov $r1 0xb00
+		shl b32 $r1 6
+		iord $r2 I[$r1 + 0x100]		// CHAN_NEXT
+		iord $r1 I[$r1 + 0x000]		// CHAN_CUR
+
+		xbit $r3 $r1 31
+		bra e #chsw_no_prev
+			xbit $r3 $r2 31
+			bra e #chsw_prev_no_next
+				push $r2
+				mov b32 $r2 $r1
+				trace_set(T_SAVE)
+				bclr $flags $p1
+				bset $flags $p2
+				call #ctx_xfer
+				trace_clr(T_SAVE);
+				pop $r2
+				trace_set(T_LOAD);
+				bset $flags $p1
+				call #ctx_xfer
+				trace_clr(T_LOAD);
+				bra #chsw_done
+			chsw_prev_no_next:
+				push $r2
+				mov b32 $r2 $r1
+				bclr $flags $p1
+				bclr $flags $p2
+				call #ctx_xfer
+				pop $r2
+				mov $r1 0xb00
+				shl b32 $r1 6
+				iowr I[$r1] $r2
+				bra #chsw_done
+		chsw_no_prev:
+			xbit $r3 $r2 31
+			bra e #chsw_done
+				bset $flags $p1
+				bclr $flags $p2
+				call #ctx_xfer
+
+		// ack the context switch request
+		chsw_done:
+		mov $r1 0xb0c
+		shl b32 $r1 6
+		mov $r2 1
+		iowr I[$r1 + 0x000] $r2		// 0x409b0c
+		trace_clr(T_AUTO)
+		bra #main
+
+	// request to set current channel? (*not* a context switch)
+	main_not_ctx_switch:
+	cmpu b32 $r14 0x0001
+	bra ne #main_not_ctx_chan
+		mov b32 $r2 $r15
+		call #ctx_chan
+		bra #main_done
+
+	// request to store current channel context?
+	main_not_ctx_chan:
+	cmpu b32 $r14 0x0002
+	bra ne #main_not_ctx_save
+		trace_set(T_SAVE)
+		bclr $flags $p1
+		bclr $flags $p2
+		call #ctx_xfer
+		trace_clr(T_SAVE)
+		bra #main_done
+
+	main_not_ctx_save:
+		shl b32 $r15 $r14 16
+		or $r15 E_BAD_COMMAND
+		call #error
+		bra #main
+
+	main_done:
+	mov $r1 0x820
+	shl b32 $r1 6
+	clear b32 $r2
+	bset $r2 31
+	iowr I[$r1 + 0x000] $r2		// CC_SCRATCH[0] |= 0x80000000
+	bra #main
+
+// interrupt handler
+ih:
+	push $r8
+	mov $r8 $flags
+	push $r8
+	push $r9
+	push $r10
+	push $r11
+	push $r13
+	push $r14
+	push $r15
+
+	// incoming fifo command?
+	iord $r10 I[$r0 + 0x200]	// INTR
+	and $r11 $r10 0x00000004
+	bra e #ih_no_fifo
+		// queue incoming fifo command for later processing
+		mov $r11 0x1900
+		mov $r13 #cmd_queue
+		iord $r14 I[$r11 + 0x100]	// FIFO_CMD
+		iord $r15 I[$r11 + 0x000]	// FIFO_DATA
+		call #queue_put
+		add b32 $r11 0x400
+		mov $r14 1
+		iowr I[$r11 + 0x000] $r14	// FIFO_ACK
+
+	// context switch request?
+	ih_no_fifo:
+	and $r11 $r10 0x00000100
+	bra e #ih_no_ctxsw
+		// enqueue a context switch for later processing
+		mov $r13 #cmd_queue
+		mov $r14 0x4001
+		call #queue_put
+
+	// anything we didn't handle, bring it to the host's attention
+	ih_no_ctxsw:
+	mov $r11 0x104
+	not b32 $r11
+	and $r11 $r10 $r11
+	bra e #ih_no_other
+		mov $r10 0xc1c
+		shl b32 $r10 6
+		iowr I[$r10] $r11	// INTR_UP_SET
+
+	// ack, and wake up main()
+	ih_no_other:
+	iowr I[$r0 + 0x100] $r10	// INTR_ACK
+
+	pop $r15
+	pop $r14
+	pop $r13
+	pop $r11
+	pop $r10
+	pop $r9
+	pop $r8
+	mov $flags $r8
+	pop $r8
+	bclr $flags $p0
+	iret
+
+// Not real sure, but, MEM_CMD 7 will hang forever if this isn't done
+ctx_4160s:
+	mov $r14 0x4160
+	sethi $r14 0x400000
+	mov $r15 1
+	call #nv_wr32
+	ctx_4160s_wait:
+		call #nv_rd32
+		xbit $r15 $r15 4
+		bra e #ctx_4160s_wait
+	ret
+
+// Without clearing again at end of xfer, some things cause PGRAPH
+// to hang with STATUS=0x00000007 until it's cleared.. fbcon can
+// still function with it set however...
+ctx_4160c:
+	mov $r14 0x4160
+	sethi $r14 0x400000
+	clear b32 $r15
+	call #nv_wr32
+	ret
+
+// Again, not real sure
+//
+// In: $r15 value to set 0x404170 to
+//
+ctx_4170s:
+	mov $r14 0x4170
+	sethi $r14 0x400000
+	or $r15 0x10
+	call #nv_wr32
+	ret
+
+// Waits for a ctx_4170s() call to complete
+//
+ctx_4170w:
+	mov $r14 0x4170
+	sethi $r14 0x400000
+	call #nv_rd32
+	and $r15 0x10
+	bra ne #ctx_4170w
+	ret
+
+// Disables various things, waits a bit, and re-enables them..
+//
+// Not sure how exactly this helps, perhaps "ENABLE" is not such a
+// good description for the bits we turn off?  Anyways, without this,
+// funny things happen.
+//
+ctx_redswitch:
+	mov $r14 0x614
+	shl b32 $r14 6
+	mov $r15 0x270
+	iowr I[$r14] $r15	// HUB_RED_SWITCH = ENABLE_GPC, POWER_ALL
+	mov $r15 8
+	ctx_redswitch_delay:
+		sub b32 $r15 1
+		bra ne #ctx_redswitch_delay
+	mov $r15 0x770
+	iowr I[$r14] $r15	// HUB_RED_SWITCH = ENABLE_ALL, POWER_ALL
+	ret
+
+// Not a clue what this is for, except that unless the value is 0x10, the
+// strand context is saved (and presumably restored) incorrectly..
+//
+// In: $r15 value to set to (0x00/0x10 are used)
+//
+ctx_86c:
+	mov $r14 0x86c
+	shl b32 $r14 6
+	iowr I[$r14] $r15	// HUB(0x86c) = val
+	mov $r14 -0x75ec
+	sethi $r14 0x400000
+	call #nv_wr32		// ROP(0xa14) = val
+	mov $r14 -0x5794
+	sethi $r14 0x410000
+	call #nv_wr32		// GPC(0x86c) = val
+	ret
+
+// ctx_load - load's a channel's ctxctl data, and selects its vm
+//
+// In: $r2 channel address
+//
+ctx_load:
+	trace_set(T_CHAN)
+
+	// switch to channel, somewhat magic in parts..
+	mov $r10 12		// DONE_UNK12
+	call #wait_donez
+	mov $r1 0xa24
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r0	// 0x409a24
+	mov $r3 0xb00
+	shl b32 $r3 6
+	iowr I[$r3 + 0x100] $r2	// CHAN_NEXT
+	mov $r1 0xa0c
+	shl b32 $r1 6
+	mov $r4 7
+	iowr I[$r1 + 0x000] $r2 // MEM_CHAN
+	iowr I[$r1 + 0x100] $r4	// MEM_CMD
+	ctx_chan_wait_0:
+		iord $r4 I[$r1 + 0x100]
+		and $r4 0x1f
+		bra ne #ctx_chan_wait_0
+	iowr I[$r3 + 0x000] $r2	// CHAN_CUR
+
+	// load channel header, fetch PGRAPH context pointer
+	mov $xtargets $r0
+	bclr $r2 31
+	shl b32 $r2 4
+	add b32 $r2 2
+
+	trace_set(T_LCHAN)
+	mov $r1 0xa04
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r2		// MEM_BASE
+	mov $r1 0xa20
+	shl b32 $r1 6
+	mov $r2 0x0002
+	sethi $r2 0x80000000
+	iowr I[$r1 + 0x000] $r2		// MEM_TARGET = vram
+	mov $r1 0x10			// chan + 0x0210
+	mov $r2 #xfer_data
+	sethi $r2 0x00020000		// 16 bytes
+	xdld $r1 $r2
+	xdwait
+	trace_clr(T_LCHAN)
+
+	// update current context
+	ld b32 $r1 D[$r0 + #xfer_data + 4]
+	shl b32 $r1 24
+	ld b32 $r2 D[$r0 + #xfer_data + 0]
+	shr b32 $r2 8
+	or $r1 $r2
+	st b32 D[$r0 + #ctx_current] $r1
+
+	// set transfer base to start of context, and fetch context header
+	trace_set(T_LCTXH)
+	mov $r2 0xa04
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r1		// MEM_BASE
+	mov $r2 1
+	mov $r1 0xa20
+	shl b32 $r1 6
+	iowr I[$r1 + 0x000] $r2		// MEM_TARGET = vm
+	mov $r1 #chan_data
+	sethi $r1 0x00060000		// 256 bytes
+	xdld $r0 $r1
+	xdwait
+	trace_clr(T_LCTXH)
+
+	trace_clr(T_CHAN)
+	ret
+
+// ctx_chan - handler for HUB_SET_CHAN command, will set a channel as
+//            the active channel for ctxctl, but not actually transfer
+//            any context data.  intended for use only during initial
+//            context construction.
+//
+// In: $r2 channel address
+//
+ctx_chan:
+	call #ctx_4160s
+	call #ctx_load
+	mov $r10 12			// DONE_UNK12
+	call #wait_donez
+	mov $r1 0xa10
+	shl b32 $r1 6
+	mov $r2 5
+	iowr I[$r1 + 0x000] $r2		// MEM_CMD = 5 (???)
+	ctx_chan_wait:
+		iord $r2 I[$r1 + 0x000]
+		or $r2 $r2
+		bra ne #ctx_chan_wait
+	call #ctx_4160c
+	ret
+
+// Execute per-context state overrides list
+//
+// Only executed on the first load of a channel.  Might want to look into
+// removing this and having the host directly modify the channel's context
+// to change this state...  The nouveau DRM already builds this list as
+// it's definitely needed for NVIDIA's, so we may as well use it for now
+//
+// Input: $r1 mmio list length
+//
+ctx_mmio_exec:
+	// set transfer base to be the mmio list
+	ld b32 $r3 D[$r0 + #chan_mmio_address]
+	mov $r2 0xa04
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r3		// MEM_BASE
+
+	clear b32 $r3
+	ctx_mmio_loop:
+		// fetch next 256 bytes of mmio list if necessary
+		and $r4 $r3 0xff
+		bra ne #ctx_mmio_pull
+			mov $r5 #xfer_data
+			sethi $r5 0x00060000	// 256 bytes
+			xdld $r3 $r5
+			xdwait
+
+		// execute a single list entry
+		ctx_mmio_pull:
+		ld b32 $r14 D[$r4 + #xfer_data + 0x00]
+		ld b32 $r15 D[$r4 + #xfer_data + 0x04]
+		call #nv_wr32
+
+		// next!
+		add b32 $r3 8
+		sub b32 $r1 1
+		bra ne #ctx_mmio_loop
+
+	// set transfer base back to the current context
+	ctx_mmio_done:
+	ld b32 $r3 D[$r0 + #ctx_current]
+	iowr I[$r2 + 0x000] $r3		// MEM_BASE
+
+	// disable the mmio list now, we don't need/want to execute it again
+	st b32 D[$r0 + #chan_mmio_count] $r0
+	mov $r1 #chan_data
+	sethi $r1 0x00060000		// 256 bytes
+	xdst $r0 $r1
+	xdwait
+	ret
+
+// Transfer HUB context data between GPU and storage area
+//
+// In: $r2 channel address
+//     $p1 clear on save, set on load
+//     $p2 set if opposite direction done/will be done, so:
+//		on save it means: "a load will follow this save"
+//		on load it means: "a save preceeded this load"
+//
+ctx_xfer:
+	bra not $p1 #ctx_xfer_pre
+	bra $p2 #ctx_xfer_pre_load
+	ctx_xfer_pre:
+		mov $r15 0x10
+		call #ctx_86c
+		call #ctx_4160s
+		bra not $p1 #ctx_xfer_exec
+
+	ctx_xfer_pre_load:
+		mov $r15 2
+		call #ctx_4170s
+		call #ctx_4170w
+		call #ctx_redswitch
+		clear b32 $r15
+		call #ctx_4170s
+		call #ctx_load
+
+	// fetch context pointer, and initiate xfer on all GPCs
+	ctx_xfer_exec:
+	ld b32 $r1 D[$r0 + #ctx_current]
+	mov $r2 0x414
+	shl b32 $r2 6
+	iowr I[$r2 + 0x000] $r0	// BAR_STATUS = reset
+	mov $r14 -0x5b00
+	sethi $r14 0x410000
+	mov b32 $r15 $r1
+	call #nv_wr32		// GPC_BCAST_WRCMD_DATA = ctx pointer
+	add b32 $r14 4
+	xbit $r15 $flags $p1
+	xbit $r2 $flags $p2
+	shl b32 $r2 1
+	or $r15 $r2
+	call #nv_wr32		// GPC_BCAST_WRCMD_CMD = GPC_XFER(type)
+
+	// strands
+	mov $r1 0x4afc
+	sethi $r1 0x20000
+	mov $r2 0xc
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x0c
+	call #strand_wait
+	mov $r2 0x47fc
+	sethi $r2 0x20000
+	iowr I[$r2] $r0		// STRAND_FIRST_GENE(0x3f) = 0x00
+	xbit $r2 $flags $p1
+	add b32 $r2 3
+	iowr I[$r1] $r2		// STRAND_CMD(0x3f) = 0x03/0x04 (SAVE/LOAD)
+
+	// mmio context
+	xbit $r10 $flags $p1	// direction
+	or $r10 6		// first, last
+	mov $r11 0		// base = 0
+	ld b32 $r12 D[$r0 + #hub_mmio_list_head]
+	ld b32 $r13 D[$r0 + #hub_mmio_list_tail]
+	mov $r14 0		// not multi
+	call #mmctx_xfer
+
+	// wait for GPCs to all complete
+	mov $r10 8		// DONE_BAR
+	call #wait_doneo
+
+	// wait for strand xfer to complete
+	call #strand_wait
+
+	// post-op
+	bra $p1 #ctx_xfer_post
+		mov $r10 12		// DONE_UNK12
+		call #wait_donez
+		mov $r1 0xa10
+		shl b32 $r1 6
+		mov $r2 5
+		iowr I[$r1] $r2		// MEM_CMD
+		ctx_xfer_post_save_wait:
+			iord $r2 I[$r1]
+			or $r2 $r2
+			bra ne #ctx_xfer_post_save_wait
+
+	bra $p2 #ctx_xfer_done
+	ctx_xfer_post:
+		mov $r15 2
+		call #ctx_4170s
+		clear b32 $r15
+		call #ctx_86c
+		call #strand_post
+		call #ctx_4170w
+		clear b32 $r15
+		call #ctx_4170s
+
+		bra not $p1 #ctx_xfer_no_post_mmio
+		ld b32 $r1 D[$r0 + #chan_mmio_count]
+		or $r1 $r1
+		bra e #ctx_xfer_no_post_mmio
+			call #ctx_mmio_exec
+
+		ctx_xfer_no_post_mmio:
+		call #ctx_4160c
+
+	ctx_xfer_done:
+	ret
+
+.align 256
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc.h b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc.h
new file mode 100644
index 00000000000..05c3995639a
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/hubnve0.fuc.h
@@ -0,0 +1,924 @@
+uint32_t nve0_grhub_data[] = {
+/* 0x0000: gpc_count */
+	0x00000000,
+/* 0x0004: rop_count */
+	0x00000000,
+/* 0x0008: cmd_queue */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0050: hub_mmio_list_head */
+	0x00000000,
+/* 0x0054: hub_mmio_list_tail */
+	0x00000000,
+/* 0x0058: ctx_current */
+	0x00000000,
+/* 0x005c: chipsets */
+	0x000000e4,
+	0x013c0070,
+	0x000000e7,
+	0x013c0070,
+	0x00000000,
+/* 0x0070: nve4_hub_mmio_head */
+	0x0417e91c,
+	0x04400204,
+	0x18404010,
+	0x204040a8,
+	0x184040d0,
+	0x004040f8,
+	0x08404130,
+	0x08404150,
+	0x00404164,
+	0x0c4041a0,
+	0x0c404200,
+	0x34404404,
+	0x0c404460,
+	0x00404480,
+	0x00404498,
+	0x0c404604,
+	0x0c404618,
+	0x0440462c,
+	0x00404640,
+	0x00404654,
+	0x00404660,
+	0x48404678,
+	0x084046c8,
+	0x08404700,
+	0x24404718,
+	0x04404744,
+	0x00404754,
+	0x00405800,
+	0x08405830,
+	0x00405854,
+	0x0c405870,
+	0x04405a00,
+	0x00405a18,
+	0x00405b00,
+	0x00405b10,
+	0x00406020,
+	0x0c406028,
+	0x044064a8,
+	0x044064b4,
+	0x2c4064c0,
+	0x004064fc,
+	0x00407040,
+	0x00407804,
+	0x1440780c,
+	0x004078bc,
+	0x18408000,
+	0x00408064,
+	0x08408800,
+	0x00408840,
+	0x08408900,
+	0x00408980,
+/* 0x013c: nve4_hub_mmio_tail */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0200: chan_data */
+/* 0x0200: chan_mmio_count */
+	0x00000000,
+/* 0x0204: chan_mmio_address */
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+/* 0x0300: xfer_data */
+	0x00000000,
+};
+
+uint32_t nve0_grhub_code[] = {
+	0x03090ef5,
+/* 0x0004: queue_put */
+	0x9800d898,
+	0x86f001d9,
+	0x0489b808,
+	0xf00c1bf4,
+	0x21f502f7,
+	0x00f802ec,
+/* 0x001c: queue_put_next */
+	0xb60798c4,
+	0x8dbb0384,
+	0x0880b600,
+	0x80008e80,
+	0x90b6018f,
+	0x0f94f001,
+	0xf801d980,
+/* 0x0039: queue_get */
+	0x0131f400,
+	0x9800d898,
+	0x89b801d9,
+	0x210bf404,
+	0xb60789c4,
+	0x9dbb0394,
+	0x0890b600,
+	0x98009e98,
+	0x80b6019f,
+	0x0f84f001,
+	0xf400d880,
+/* 0x0066: queue_get_done */
+	0x00f80132,
+/* 0x0068: nv_rd32 */
+	0x0728b7f1,
+	0xb906b4b6,
+	0xc9f002ec,
+	0x00bcd01f,
+/* 0x0078: nv_rd32_wait */
+	0xc800bccf,
+	0x1bf41fcc,
+	0x06a7f0fa,
+	0x010321f5,
+	0xf840bfcf,
+/* 0x008d: nv_wr32 */
+	0x28b7f100,
+	0x06b4b607,
+	0xb980bfd0,
+	0xc9f002ec,
+	0x1ec9f01f,
+/* 0x00a3: nv_wr32_wait */
+	0xcf00bcd0,
+	0xccc800bc,
+	0xfa1bf41f,
+/* 0x00ae: watchdog_reset */
+	0x87f100f8,
+	0x84b60430,
+	0x1ff9f006,
+	0xf8008fd0,
+/* 0x00bd: watchdog_clear */
+	0x3087f100,
+	0x0684b604,
+	0xf80080d0,
+/* 0x00c9: wait_donez */
+	0x3c87f100,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d000,
+	0x081887f1,
+	0xd00684b6,
+/* 0x00e2: wait_done_wait_donez */
+	0x87f1008a,
+	0x84b60400,
+	0x0088cf06,
+	0xf4888aff,
+	0x87f1f31b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00099,
+/* 0x0103: wait_doneo */
+	0xf100f800,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00099f0,
+	0x87f10089,
+	0x84b60818,
+	0x008ad006,
+/* 0x011c: wait_done_wait_doneo */
+	0x040087f1,
+	0xcf0684b6,
+	0x8aff0088,
+	0xf30bf488,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0099f094,
+	0xf80089d0,
+/* 0x013d: mmctx_size */
+/* 0x013f: nv_mmctx_size_loop */
+	0x9894bd00,
+	0x85b600e8,
+	0x0180b61a,
+	0xbb0284b6,
+	0xe0b60098,
+	0x04efb804,
+	0xb9eb1bf4,
+	0x00f8029f,
+/* 0x015c: mmctx_xfer */
+	0x083c87f1,
+	0xbd0684b6,
+	0x0199f094,
+	0xf10089d0,
+	0xb6071087,
+	0x94bd0684,
+	0xf405bbfd,
+	0x8bd0090b,
+	0x0099f000,
+/* 0x0180: mmctx_base_disabled */
+	0xf405eefd,
+	0x8ed00c0b,
+	0xc08fd080,
+/* 0x018f: mmctx_multi_disabled */
+	0xb70199f0,
+	0xc8010080,
+	0xb4b600ab,
+	0x0cb9f010,
+	0xb601aec8,
+	0xbefd11e4,
+	0x008bd005,
+/* 0x01a8: mmctx_exec_loop */
+/* 0x01a8: mmctx_wait_free */
+	0xf0008ecf,
+	0x0bf41fe4,
+	0x00ce98fa,
+	0xd005e9fd,
+	0xc0b6c08e,
+	0x04cdb804,
+	0xc8e81bf4,
+	0x1bf402ab,
+/* 0x01c9: mmctx_fini_wait */
+	0x008bcf18,
+	0xb01fb4f0,
+	0x1bf410b4,
+	0x02a7f0f7,
+	0xf4c921f4,
+/* 0x01de: mmctx_stop */
+	0xabc81b0e,
+	0x10b4b600,
+	0xf00cb9f0,
+	0x8bd012b9,
+/* 0x01ed: mmctx_stop_wait */
+	0x008bcf00,
+	0xf412bbc8,
+/* 0x01f6: mmctx_done */
+	0x87f1fa1b,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00199,
+/* 0x0207: strand_wait */
+	0xf900f800,
+	0x02a7f0a0,
+	0xfcc921f4,
+/* 0x0213: strand_pre */
+	0xf100f8a0,
+	0xf04afc87,
+	0x97f00283,
+	0x0089d00c,
+	0x020721f5,
+/* 0x0226: strand_post */
+	0x87f100f8,
+	0x83f04afc,
+	0x0d97f002,
+	0xf50089d0,
+	0xf8020721,
+/* 0x0239: strand_set */
+	0xfca7f100,
+	0x02a3f04f,
+	0x0500aba2,
+	0xd00fc7f0,
+	0xc7f000ac,
+	0x00bcd00b,
+	0x020721f5,
+	0xf000aed0,
+	0xbcd00ac7,
+	0x0721f500,
+/* 0x0263: strand_ctx_init */
+	0xf100f802,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x21f50089,
+	0xe7f00213,
+	0x3921f503,
+	0xfca7f102,
+	0x02a3f046,
+	0x0400aba0,
+	0xf040a0d0,
+	0xbcd001c7,
+	0x0721f500,
+	0x010c9202,
+	0xf000acd0,
+	0xbcd002c7,
+	0x0721f500,
+	0x2621f502,
+	0x8087f102,
+	0x0684b608,
+	0xb70089cf,
+	0x95220080,
+/* 0x02ba: ctx_init_strand_loop */
+	0x8ed008fe,
+	0x408ed000,
+	0xb6808acf,
+	0xa0b606a5,
+	0x00eabb01,
+	0xb60480b6,
+	0x1bf40192,
+	0x08e4b6e8,
+	0xf1f2efbc,
+	0xb6085c87,
+	0x94bd0684,
+	0xd00399f0,
+	0x00f80089,
+/* 0x02ec: error */
+	0xe7f1e0f9,
+	0xe4b60814,
+	0x00efd006,
+	0x0c1ce7f1,
+	0xf006e4b6,
+	0xefd001f7,
+	0xf8e0fc00,
+/* 0x0309: init */
+	0xfe04bd00,
+	0x07fe0004,
+	0x0017f100,
+	0x0227f012,
+	0xf10012d0,
+	0xfe05b917,
+	0x17f10010,
+	0x10d00400,
+	0x0437f1c0,
+	0x0634b604,
+	0x200327f1,
+	0xf10032d0,
+	0xd0200427,
+	0x27f10132,
+	0x32d0200b,
+	0x0c27f102,
+	0x0732d020,
+	0x0c2427f1,
+	0xb90624b6,
+	0x23d00003,
+	0x0427f100,
+	0x0023f087,
+	0xb70012d0,
+	0xf0010012,
+	0x12d00427,
+	0x1031f400,
+	0x9604e7f1,
+	0xf440e3f0,
+	0xf1c76821,
+	0x01018090,
+	0x801ff4f0,
+	0x17f0000f,
+	0x041fbb01,
+	0xf10112b6,
+	0xb6040c27,
+	0x21d00624,
+	0x4021d000,
+	0x080027f1,
+	0xcf0624b6,
+	0xf7f00022,
+/* 0x03a9: init_find_chipset */
+	0x08f0b654,
+	0xb800f398,
+	0x0bf40432,
+	0x0034b00b,
+	0xf8f11bf4,
+/* 0x03bd: init_context */
+	0x0017f100,
+	0x02fe5801,
+	0xf003ff58,
+	0x0e8000e3,
+	0x150f8014,
+	0x013d21f5,
+	0x070037f1,
+	0x950634b6,
+	0x34d00814,
+	0x4034d000,
+	0x130030b7,
+	0xb6001fbb,
+	0x3fd002f5,
+	0x0815b600,
+	0xb60110b6,
+	0x1fb90814,
+	0x6321f502,
+	0x001fbb02,
+	0xf1000398,
+	0xf0200047,
+/* 0x040e: init_gpc */
+	0x4ea05043,
+	0x1fb90804,
+	0x8d21f402,
+	0x08004ea0,
+	0xf4022fb9,
+	0x4ea08d21,
+	0xf4bd010c,
+	0xa08d21f4,
+	0xf401044e,
+	0x4ea08d21,
+	0xf7f00100,
+	0x8d21f402,
+	0x08004ea0,
+/* 0x0440: init_gpc_wait */
+	0xc86821f4,
+	0x0bf41fff,
+	0x044ea0fa,
+	0x6821f408,
+	0xb7001fbb,
+	0xb6800040,
+	0x1bf40132,
+	0x0027f1b4,
+	0x0624b608,
+	0xb74021d0,
+	0xbd080020,
+	0x1f19f014,
+/* 0x0473: main */
+	0xf40021d0,
+	0x28f40031,
+	0x08d7f000,
+	0xf43921f4,
+	0xe4b1f401,
+	0x1bf54001,
+	0x87f100d1,
+	0x84b6083c,
+	0xf094bd06,
+	0x89d00499,
+	0x0017f100,
+	0x0614b60b,
+	0xcf4012cf,
+	0x13c80011,
+	0x7e0bf41f,
+	0xf41f23c8,
+	0x20f95a0b,
+	0xf10212b9,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00799f0,
+	0x32f40089,
+	0x0231f401,
+	0x082921f5,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0799f094,
+	0xfc0089d0,
+	0x3c87f120,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d006,
+	0xf50131f4,
+	0xf1082921,
+	0xb6085c87,
+	0x94bd0684,
+	0xd00699f0,
+	0x0ef40089,
+/* 0x0509: chsw_prev_no_next */
+	0xb920f931,
+	0x32f40212,
+	0x0232f401,
+	0x082921f5,
+	0x17f120fc,
+	0x14b60b00,
+	0x0012d006,
+/* 0x0527: chsw_no_prev */
+	0xc8130ef4,
+	0x0bf41f23,
+	0x0131f40d,
+	0xf50232f4,
+/* 0x0537: chsw_done */
+	0xf1082921,
+	0xb60b0c17,
+	0x27f00614,
+	0x0012d001,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0499f094,
+	0xf50089d0,
+/* 0x0557: main_not_ctx_switch */
+	0xb0ff200e,
+	0x1bf401e4,
+	0x02f2b90d,
+	0x07b521f5,
+/* 0x0567: main_not_ctx_chan */
+	0xb0420ef4,
+	0x1bf402e4,
+	0x3c87f12e,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d007,
+	0xf40132f4,
+	0x21f50232,
+	0x87f10829,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00799,
+	0x110ef400,
+/* 0x0598: main_not_ctx_save */
+	0xf010ef94,
+	0x21f501f5,
+	0x0ef502ec,
+/* 0x05a6: main_done */
+	0x17f1fed1,
+	0x14b60820,
+	0xf024bd06,
+	0x12d01f29,
+	0xbe0ef500,
+/* 0x05b9: ih */
+	0xfe80f9fe,
+	0x80f90188,
+	0xa0f990f9,
+	0xd0f9b0f9,
+	0xf0f9e0f9,
+	0xc4800acf,
+	0x0bf404ab,
+	0x00b7f11d,
+	0x08d7f019,
+	0xcf40becf,
+	0x21f400bf,
+	0x00b0b704,
+	0x01e7f004,
+/* 0x05ef: ih_no_fifo */
+	0xe400bed0,
+	0xf40100ab,
+	0xd7f00d0b,
+	0x01e7f108,
+	0x0421f440,
+/* 0x0600: ih_no_ctxsw */
+	0x0104b7f1,
+	0xabffb0bd,
+	0x0d0bf4b4,
+	0x0c1ca7f1,
+	0xd006a4b6,
+/* 0x0616: ih_no_other */
+	0x0ad000ab,
+	0xfcf0fc40,
+	0xfcd0fce0,
+	0xfca0fcb0,
+	0xfe80fc90,
+	0x80fc0088,
+	0xf80032f4,
+/* 0x0631: ctx_4160s */
+	0x60e7f101,
+	0x40e3f041,
+	0xf401f7f0,
+/* 0x063e: ctx_4160s_wait */
+	0x21f48d21,
+	0x04ffc868,
+	0xf8fa0bf4,
+/* 0x0649: ctx_4160c */
+	0x60e7f100,
+	0x40e3f041,
+	0x21f4f4bd,
+/* 0x0657: ctx_4170s */
+	0xf100f88d,
+	0xf04170e7,
+	0xf5f040e3,
+	0x8d21f410,
+/* 0x0666: ctx_4170w */
+	0xe7f100f8,
+	0xe3f04170,
+	0x6821f440,
+	0xf410f4f0,
+	0x00f8f31b,
+/* 0x0678: ctx_redswitch */
+	0x0614e7f1,
+	0xf106e4b6,
+	0xd00270f7,
+	0xf7f000ef,
+/* 0x0689: ctx_redswitch_delay */
+	0x01f2b608,
+	0xf1fd1bf4,
+	0xd00770f7,
+	0x00f800ef,
+/* 0x0698: ctx_86c */
+	0x086ce7f1,
+	0xd006e4b6,
+	0xe7f100ef,
+	0xe3f08a14,
+	0x8d21f440,
+	0xa86ce7f1,
+	0xf441e3f0,
+	0x00f88d21,
+/* 0x06b8: ctx_load */
+	0x083c87f1,
+	0xbd0684b6,
+	0x0599f094,
+	0xf00089d0,
+	0x21f40ca7,
+	0x2417f1c9,
+	0x0614b60a,
+	0xf10010d0,
+	0xb60b0037,
+	0x32d00634,
+	0x0c17f140,
+	0x0614b60a,
+	0xd00747f0,
+	0x14d00012,
+/* 0x06f1: ctx_chan_wait_0 */
+	0x4014cf40,
+	0xf41f44f0,
+	0x32d0fa1b,
+	0x000bfe00,
+	0xb61f2af0,
+	0x20b60424,
+	0x3c87f102,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d008,
+	0x0a0417f1,
+	0xd00614b6,
+	0x17f10012,
+	0x14b60a20,
+	0x0227f006,
+	0x800023f1,
+	0xf00012d0,
+	0x27f11017,
+	0x23f00300,
+	0x0512fa02,
+	0x87f103f8,
+	0x84b6085c,
+	0xf094bd06,
+	0x89d00899,
+	0xc1019800,
+	0x981814b6,
+	0x25b6c002,
+	0x0512fd08,
+	0xf1160180,
+	0xb6083c87,
+	0x94bd0684,
+	0xd00999f0,
+	0x27f10089,
+	0x24b60a04,
+	0x0021d006,
+	0xf10127f0,
+	0xb60a2017,
+	0x12d00614,
+	0x0017f100,
+	0x0613f002,
+	0xf80501fa,
+	0x5c87f103,
+	0x0684b608,
+	0x99f094bd,
+	0x0089d009,
+	0x085c87f1,
+	0xbd0684b6,
+	0x0599f094,
+	0xf80089d0,
+/* 0x07b5: ctx_chan */
+	0x3121f500,
+	0xb821f506,
+	0x0ca7f006,
+	0xf1c921f4,
+	0xb60a1017,
+	0x27f00614,
+	0x0012d005,
+/* 0x07d0: ctx_chan_wait */
+	0xfd0012cf,
+	0x1bf40522,
+	0x4921f5fa,
+/* 0x07df: ctx_mmio_exec */
+	0x9800f806,
+	0x27f18103,
+	0x24b60a04,
+	0x0023d006,
+/* 0x07ee: ctx_mmio_loop */
+	0x34c434bd,
+	0x0f1bf4ff,
+	0x030057f1,
+	0xfa0653f0,
+	0x03f80535,
+/* 0x0800: ctx_mmio_pull */
+	0x98c04e98,
+	0x21f4c14f,
+	0x0830b68d,
+	0xf40112b6,
+/* 0x0812: ctx_mmio_done */
+	0x0398df1b,
+	0x0023d016,
+	0xf1800080,
+	0xf0020017,
+	0x01fa0613,
+	0xf803f806,
+/* 0x0829: ctx_xfer */
+	0x0611f400,
+/* 0x082f: ctx_xfer_pre */
+	0xf01102f4,
+	0x21f510f7,
+	0x21f50698,
+	0x11f40631,
+/* 0x083d: ctx_xfer_pre_load */
+	0x02f7f01c,
+	0x065721f5,
+	0x066621f5,
+	0x067821f5,
+	0x21f5f4bd,
+	0x21f50657,
+/* 0x0856: ctx_xfer_exec */
+	0x019806b8,
+	0x1427f116,
+	0x0624b604,
+	0xf10020d0,
+	0xf0a500e7,
+	0x1fb941e3,
+	0x8d21f402,
+	0xf004e0b6,
+	0x2cf001fc,
+	0x0124b602,
+	0xf405f2fd,
+	0x17f18d21,
+	0x13f04afc,
+	0x0c27f002,
+	0xf50012d0,
+	0xf1020721,
+	0xf047fc27,
+	0x20d00223,
+	0x012cf000,
+	0xd00320b6,
+	0xacf00012,
+	0x06a5f001,
+	0x9800b7f0,
+	0x0d98140c,
+	0x00e7f015,
+	0x015c21f5,
+	0xf508a7f0,
+	0xf5010321,
+	0xf4020721,
+	0xa7f02201,
+	0xc921f40c,
+	0x0a1017f1,
+	0xf00614b6,
+	0x12d00527,
+/* 0x08dd: ctx_xfer_post_save_wait */
+	0x0012cf00,
+	0xf40522fd,
+	0x02f4fa1b,
+/* 0x08e9: ctx_xfer_post */
+	0x02f7f032,
+	0x065721f5,
+	0x21f5f4bd,
+	0x21f50698,
+	0x21f50226,
+	0xf4bd0666,
+	0x065721f5,
+	0x981011f4,
+	0x11fd8001,
+	0x070bf405,
+	0x07df21f5,
+/* 0x0914: ctx_xfer_no_post_mmio */
+	0x064921f5,
+/* 0x0918: ctx_xfer_done */
+	0x000000f8,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+	0x00000000,
+};
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/fuc/nve0.fuc b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/nve0.fuc
new file mode 100644
index 00000000000..f16a5d53319
--- /dev/null
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/fuc/nve0.fuc
@@ -0,0 +1,400 @@
+/* fuc microcode util functions for nve0 PGRAPH
+ *
+ * Copyright 2011 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.
+ *
+ * Authors: Ben Skeggs
+ */
+
+define(`mmctx_data', `.b32 eval((($2 - 1) << 26) | $1)')
+define(`queue_init', `.skip eval((2 * 4) + ((8 * 4) * 2))')
+
+ifdef(`include_code', `
+// Error codes
+define(`E_BAD_COMMAND', 0x01)
+define(`E_CMD_OVERFLOW', 0x02)
+
+// Util macros to help with debugging ucode hangs etc
+define(`T_WAIT', 0)
+define(`T_MMCTX', 1)
+define(`T_STRWAIT', 2)
+define(`T_STRINIT', 3)
+define(`T_AUTO', 4)
+define(`T_CHAN', 5)
+define(`T_LOAD', 6)
+define(`T_SAVE', 7)
+define(`T_LCHAN', 8)
+define(`T_LCTXH', 9)
+
+define(`trace_set', `
+	mov $r8 0x83c
+	shl b32 $r8 6
+	clear b32 $r9
+	bset $r9 $1
+	iowr I[$r8 + 0x000] $r9		// CC_SCRATCH[7]
+')
+
+define(`trace_clr', `
+	mov $r8 0x85c
+	shl b32 $r8 6
+	clear b32 $r9
+	bset $r9 $1
+	iowr I[$r8 + 0x000] $r9		// CC_SCRATCH[7]
+')
+
+// queue_put - add request to queue
+//
+// In : $r13 queue pointer
+//	$r14 command
+//	$r15 data
+//
+queue_put:
+	// make sure we have space..
+	ld b32 $r8 D[$r13 + 0x0]	// GET
+	ld b32 $r9 D[$r13 + 0x4]	// PUT
+	xor $r8 8
+	cmpu b32 $r8 $r9
+	bra ne #queue_put_next
+		mov $r15 E_CMD_OVERFLOW
+		call #error
+		ret
+
+	// store cmd/data on queue
+	queue_put_next:
+	and $r8 $r9 7
+	shl b32 $r8 3
+	add b32 $r8 $r13
+	add b32 $r8 8
+	st b32 D[$r8 + 0x0] $r14
+	st b32 D[$r8 + 0x4] $r15
+
+	// update PUT
+	add b32 $r9 1
+	and $r9 0xf
+	st b32 D[$r13 + 0x4] $r9
+	ret
+
+// queue_get - fetch request from queue
+//
+// In : $r13 queue pointer
+//
+// Out:	$p1  clear on success (data available)
+//	$r14 command
+// 	$r15 data
+//
+queue_get:
+	bset $flags $p1
+	ld b32 $r8 D[$r13 + 0x0]	// GET
+	ld b32 $r9 D[$r13 + 0x4]	// PUT
+	cmpu b32 $r8 $r9
+	bra e #queue_get_done
+		// fetch first cmd/data pair
+		and $r9 $r8 7
+		shl b32 $r9 3
+		add b32 $r9 $r13
+		add b32 $r9 8
+		ld b32 $r14 D[$r9 + 0x0]
+		ld b32 $r15 D[$r9 + 0x4]
+
+		// update GET
+		add b32 $r8 1
+		and $r8 0xf
+		st b32 D[$r13 + 0x0] $r8
+		bclr $flags $p1
+queue_get_done:
+	ret
+
+// nv_rd32 - read 32-bit value from nv register
+//
+// In : $r14 register
+// Out: $r15 value
+//
+nv_rd32:
+	mov $r11 0x728
+	shl b32 $r11 6
+	mov b32 $r12 $r14
+	bset $r12 31			// MMIO_CTRL_PENDING
+	iowr I[$r11 + 0x000] $r12	// MMIO_CTRL
+	nv_rd32_wait:
+		iord $r12 I[$r11 + 0x000]
+		xbit $r12 $r12 31
+		bra ne #nv_rd32_wait
+	mov $r10 6			// DONE_MMIO_RD
+	call #wait_doneo
+	iord $r15 I[$r11 + 0x100]	// MMIO_RDVAL
+	ret
+
+// nv_wr32 - write 32-bit value to nv register
+//
+// In : $r14 register
+//      $r15 value
+//
+nv_wr32:
+	mov $r11 0x728
+	shl b32 $r11 6
+	iowr I[$r11 + 0x200] $r15	// MMIO_WRVAL
+	mov b32 $r12 $r14
+	bset $r12 31			// MMIO_CTRL_PENDING
+	bset $r12 30			// MMIO_CTRL_WRITE
+	iowr I[$r11 + 0x000] $r12	// MMIO_CTRL
+	nv_wr32_wait:
+		iord $r12 I[$r11 + 0x000]
+		xbit $r12 $r12 31
+		bra ne #nv_wr32_wait
+	ret
+
+// (re)set watchdog timer
+//
+// In : $r15 timeout
+//
+watchdog_reset:
+	mov $r8 0x430
+	shl b32 $r8 6
+	bset $r15 31
+	iowr I[$r8 + 0x000] $r15
+	ret
+
+// clear watchdog timer
+watchdog_clear:
+	mov $r8 0x430
+	shl b32 $r8 6
+	iowr I[$r8 + 0x000] $r0
+	ret
+
+// wait_done{z,o} - wait on FUC_DONE bit to become clear/set
+//
+// In : $r10 bit to wait on
+//
+define(`wait_done', `
+$1:
+	trace_set(T_WAIT);
+	mov $r8 0x818
+	shl b32 $r8 6
+	iowr I[$r8 + 0x000] $r10	// CC_SCRATCH[6] = wait bit
+	wait_done_$1:
+		mov $r8 0x400
+		shl b32 $r8 6
+		iord $r8 I[$r8 + 0x000]	// DONE
+		xbit $r8 $r8 $r10
+		bra $2 #wait_done_$1
+	trace_clr(T_WAIT)
+	ret
+')
+wait_done(wait_donez, ne)
+wait_done(wait_doneo, e)
+
+// mmctx_size - determine size of a mmio list transfer
+//
+// In : $r14 mmio list head
+//      $r15 mmio list tail
+// Out: $r15 transfer size (in bytes)
+//
+mmctx_size:
+	clear b32 $r9
+	nv_mmctx_size_loop:
+		ld b32 $r8 D[$r14]
+		shr b32 $r8 26
+		add b32 $r8 1
+		shl b32 $r8 2
+		add b32 $r9 $r8
+		add b32 $r14 4
+		cmpu b32 $r14 $r15
+		bra ne #nv_mmctx_size_loop
+	mov b32 $r15 $r9
+	ret
+
+// mmctx_xfer - execute a list of mmio transfers
+//
+// In : $r10 flags
+//		bit 0: direction (0 = save, 1 = load)
+//		bit 1: set if first transfer
+//		bit 2: set if last transfer
+//	$r11 base
+//	$r12 mmio list head
+//	$r13 mmio list tail
+//	$r14 multi_stride
+//	$r15 multi_mask
+//
+mmctx_xfer:
+	trace_set(T_MMCTX)
+	mov $r8 0x710
+	shl b32 $r8 6
+	clear b32 $r9
+	or $r11 $r11
+	bra e #mmctx_base_disabled
+		iowr I[$r8 + 0x000] $r11	// MMCTX_BASE
+		bset $r9 0			// BASE_EN
+	mmctx_base_disabled:
+	or $r14 $r14
+	bra e #mmctx_multi_disabled
+		iowr I[$r8 + 0x200] $r14 	// MMCTX_MULTI_STRIDE
+		iowr I[$r8 + 0x300] $r15 	// MMCTX_MULTI_MASK
+		bset $r9 1			// MULTI_EN
+	mmctx_multi_disabled:
+	add b32 $r8 0x100
+
+	xbit $r11 $r10 0
+	shl b32 $r11 16			// DIR
+	bset $r11 12			// QLIMIT = 0x10
+	xbit $r14 $r10 1
+	shl b32 $r14 17
+	or $r11 $r14			// START_TRIGGER
+	iowr I[$r8 + 0x000] $r11	// MMCTX_CTRL
+
+	// loop over the mmio list, and send requests to the hw
+	mmctx_exec_loop:
+		// wait for space in mmctx queue
+		mmctx_wait_free:
+			iord $r14 I[$r8 + 0x000] // MMCTX_CTRL
+			and $r14 0x1f
+			bra e #mmctx_wait_free
+
+		// queue up an entry
+		ld b32 $r14 D[$r12]
+		or $r14 $r9
+		iowr I[$r8 + 0x300] $r14
+		add b32 $r12 4
+		cmpu b32 $r12 $r13
+		bra ne #mmctx_exec_loop
+
+	xbit $r11 $r10 2
+	bra ne #mmctx_stop
+		// wait for queue to empty
+		mmctx_fini_wait:
+			iord $r11 I[$r8 + 0x000]	// MMCTX_CTRL
+			and $r11 0x1f
+			cmpu b32 $r11 0x10
+			bra ne #mmctx_fini_wait
+		mov $r10 2				// DONE_MMCTX
+		call #wait_donez
+		bra #mmctx_done
+	mmctx_stop:
+		xbit $r11 $r10 0
+		shl b32 $r11 16			// DIR
+		bset $r11 12			// QLIMIT = 0x10
+		bset $r11 18			// STOP_TRIGGER
+		iowr I[$r8 + 0x000] $r11	// MMCTX_CTRL
+		mmctx_stop_wait:
+			// wait for STOP_TRIGGER to clear
+			iord $r11 I[$r8 + 0x000] // MMCTX_CTRL
+			xbit $r11 $r11 18
+			bra ne #mmctx_stop_wait
+	mmctx_done:
+	trace_clr(T_MMCTX)
+	ret
+
+// Wait for DONE_STRAND
+//
+strand_wait:
+	push $r10
+	mov $r10 2
+	call #wait_donez
+	pop $r10
+	ret
+
+// unknown - call before issuing strand commands
+//
+strand_pre:
+	mov $r8 0x4afc
+	sethi $r8 0x20000
+	mov $r9 0xc
+	iowr I[$r8] $r9
+	call #strand_wait
+	ret
+
+// unknown - call after issuing strand commands
+//
+strand_post:
+	mov $r8 0x4afc
+	sethi $r8 0x20000
+	mov $r9 0xd
+	iowr I[$r8] $r9
+	call #strand_wait
+	ret
+
+// Selects strand set?!
+//
+// In: $r14 id
+//
+strand_set:
+	mov $r10 0x4ffc
+	sethi $r10 0x20000
+	sub b32 $r11 $r10 0x500
+	mov $r12 0xf
+	iowr I[$r10 + 0x000] $r12		// 0x93c = 0xf
+	mov $r12 0xb
+	iowr I[$r11 + 0x000] $r12		// 0x928 = 0xb
+	call #strand_wait
+	iowr I[$r10 + 0x000] $r14		// 0x93c = <id>
+	mov $r12 0xa
+	iowr I[$r11 + 0x000] $r12		// 0x928 = 0xa
+	call #strand_wait
+	ret
+
+// Initialise strand context data
+//
+// In : $r15 context base
+// Out: $r15 context size (in bytes)
+//
+// Strandset(?) 3 hardcoded currently
+//
+strand_ctx_init:
+	trace_set(T_STRINIT)
+	call #strand_pre
+	mov $r14 3
+	call #strand_set
+	mov $r10 0x46fc
+	sethi $r10 0x20000
+	add b32 $r11 $r10 0x400
+	iowr I[$r10 + 0x100] $r0	// STRAND_FIRST_GENE = 0
+	mov $r12 1
+	iowr I[$r11 + 0x000] $r12	// STRAND_CMD = LATCH_FIRST_GENE
+	call #strand_wait
+	sub b32 $r12 $r0 1
+	iowr I[$r10 + 0x000] $r12	// STRAND_GENE_CNT = 0xffffffff
+	mov $r12 2
+	iowr I[$r11 + 0x000] $r12	// STRAND_CMD = LATCH_GENE_CNT
+	call #strand_wait
+	call #strand_post
+
+	// read the size of each strand, poke the context offset of
+	// each into STRAND_{SAVE,LOAD}_SWBASE now, no need to worry
+	// about it later then.
+	mov $r8 0x880
+	shl b32 $r8 6
+	iord $r9 I[$r8 + 0x000]		// STRANDS
+	add b32 $r8 0x2200
+	shr b32 $r14 $r15 8
+	ctx_init_strand_loop:
+		iowr I[$r8 + 0x000] $r14	// STRAND_SAVE_SWBASE
+		iowr I[$r8 + 0x100] $r14	// STRAND_LOAD_SWBASE
+		iord $r10 I[$r8 + 0x200]	// STRAND_SIZE
+		shr b32 $r10 6
+		add b32 $r10 1
+		add b32 $r14 $r10
+		add b32 $r8 4
+		sub b32 $r9 1
+		bra ne #ctx_init_strand_loop
+
+	shl b32 $r14 8
+	sub b32 $r15 $r14 $r15
+	trace_clr(T_STRINIT)
+	ret
+')
diff --git a/drivers/gpu/drm/nouveau/core/engine/graph/nve0.c b/drivers/gpu/drm/nouveau/core/engine/graph/nve0.c
index 2ba125bb5f3..7d3811201f7 100644
--- a/drivers/gpu/drm/nouveau/core/engine/graph/nve0.c
+++ b/drivers/gpu/drm/nouveau/core/engine/graph/nve0.c
@@ -23,6 +23,8 @@
  */
 
 #include "nvc0.h"
+#include "fuc/hubnve0.fuc.h"
+#include "fuc/gpcnve0.fuc.h"
 
 /*******************************************************************************
  * Graphics object classes
@@ -197,6 +199,7 @@ nve0_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
 	       struct nouveau_oclass *oclass, void *data, u32 size,
 	       struct nouveau_object **pobject)
 {
+	struct nouveau_device *device = nv_device(parent);
 	struct nvc0_graph_priv *priv;
 	int ret, i;
 
@@ -210,13 +213,15 @@ nve0_graph_ctor(struct nouveau_object *parent, struct nouveau_object *engine,
 	nv_engine(priv)->cclass = &nve0_graph_cclass;
 	nv_engine(priv)->sclass = nve0_graph_sclass;
 
-	nv_info(priv, "using external firmware\n");
-	if (nvc0_graph_ctor_fw(priv, "fuc409c", &priv->fuc409c) ||
-	    nvc0_graph_ctor_fw(priv, "fuc409d", &priv->fuc409d) ||
-	    nvc0_graph_ctor_fw(priv, "fuc41ac", &priv->fuc41ac) ||
-	    nvc0_graph_ctor_fw(priv, "fuc41ad", &priv->fuc41ad))
-		return -EINVAL;
-	priv->firmware = true;
+	if (nouveau_boolopt(device->cfgopt, "NvGrUseFW", true)) {
+		nv_info(priv, "using external firmware\n");
+		if (nvc0_graph_ctor_fw(priv, "fuc409c", &priv->fuc409c) ||
+		    nvc0_graph_ctor_fw(priv, "fuc409d", &priv->fuc409d) ||
+		    nvc0_graph_ctor_fw(priv, "fuc41ac", &priv->fuc41ac) ||
+		    nvc0_graph_ctor_fw(priv, "fuc41ad", &priv->fuc41ad))
+			return -EINVAL;
+		priv->firmware = true;
+	}
 
 	ret = nouveau_gpuobj_new(parent, NULL, 0x1000, 256, 0, &priv->unk4188b4);
 	if (ret)
@@ -386,81 +391,131 @@ static int
 nve0_graph_init_ctxctl(struct nvc0_graph_priv *priv)
 {
 	u32 r000260;
+	int i;
 
-	/* load fuc microcode */
-	r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
-	nvc0_graph_init_fw(priv, 0x409000, &priv->fuc409c, &priv->fuc409d);
-	nvc0_graph_init_fw(priv, 0x41a000, &priv->fuc41ac, &priv->fuc41ad);
-	nv_wr32(priv, 0x000260, r000260);
+	if (priv->firmware) {
+		/* load fuc microcode */
+		r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
+		nvc0_graph_init_fw(priv, 0x409000, &priv->fuc409c, &priv->fuc409d);
+		nvc0_graph_init_fw(priv, 0x41a000, &priv->fuc41ac, &priv->fuc41ad);
+		nv_wr32(priv, 0x000260, r000260);
+
+		/* start both of them running */
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x41a10c, 0x00000000);
+		nv_wr32(priv, 0x40910c, 0x00000000);
+		nv_wr32(priv, 0x41a100, 0x00000002);
+		nv_wr32(priv, 0x409100, 0x00000002);
+		if (!nv_wait(priv, 0x409800, 0x00000001, 0x00000001))
+			nv_error(priv, "0x409800 wait failed\n");
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x7fffffff);
+		nv_wr32(priv, 0x409504, 0x00000021);
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x00000000);
+		nv_wr32(priv, 0x409504, 0x00000010);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x10 timeout\n");
+			return -EBUSY;
+		}
+		priv->size = nv_rd32(priv, 0x409800);
+
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x00000000);
+		nv_wr32(priv, 0x409504, 0x00000016);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x16 timeout\n");
+			return -EBUSY;
+		}
 
-	/* start both of them running */
-	nv_wr32(priv, 0x409840, 0xffffffff);
-	nv_wr32(priv, 0x41a10c, 0x00000000);
-	nv_wr32(priv, 0x40910c, 0x00000000);
-	nv_wr32(priv, 0x41a100, 0x00000002);
-	nv_wr32(priv, 0x409100, 0x00000002);
-	if (!nv_wait(priv, 0x409800, 0x00000001, 0x00000001))
-		nv_error(priv, "0x409800 wait failed\n");
-
-	nv_wr32(priv, 0x409840, 0xffffffff);
-	nv_wr32(priv, 0x409500, 0x7fffffff);
-	nv_wr32(priv, 0x409504, 0x00000021);
-
-	nv_wr32(priv, 0x409840, 0xffffffff);
-	nv_wr32(priv, 0x409500, 0x00000000);
-	nv_wr32(priv, 0x409504, 0x00000010);
-	if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
-		nv_error(priv, "fuc09 req 0x10 timeout\n");
-		return -EBUSY;
-	}
-	priv->size = nv_rd32(priv, 0x409800);
+		nv_wr32(priv, 0x409840, 0xffffffff);
+		nv_wr32(priv, 0x409500, 0x00000000);
+		nv_wr32(priv, 0x409504, 0x00000025);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x25 timeout\n");
+			return -EBUSY;
+		}
 
-	nv_wr32(priv, 0x409840, 0xffffffff);
-	nv_wr32(priv, 0x409500, 0x00000000);
-	nv_wr32(priv, 0x409504, 0x00000016);
-	if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
-		nv_error(priv, "fuc09 req 0x16 timeout\n");
-		return -EBUSY;
-	}
+		nv_wr32(priv, 0x409800, 0x00000000);
+		nv_wr32(priv, 0x409500, 0x00000001);
+		nv_wr32(priv, 0x409504, 0x00000030);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x30 timeout\n");
+			return -EBUSY;
+		}
 
-	nv_wr32(priv, 0x409840, 0xffffffff);
-	nv_wr32(priv, 0x409500, 0x00000000);
-	nv_wr32(priv, 0x409504, 0x00000025);
-	if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
-		nv_error(priv, "fuc09 req 0x25 timeout\n");
-		return -EBUSY;
+		nv_wr32(priv, 0x409810, 0xb00095c8);
+		nv_wr32(priv, 0x409800, 0x00000000);
+		nv_wr32(priv, 0x409500, 0x00000001);
+		nv_wr32(priv, 0x409504, 0x00000031);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x31 timeout\n");
+			return -EBUSY;
+		}
+
+		nv_wr32(priv, 0x409810, 0x00080420);
+		nv_wr32(priv, 0x409800, 0x00000000);
+		nv_wr32(priv, 0x409500, 0x00000001);
+		nv_wr32(priv, 0x409504, 0x00000032);
+		if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
+			nv_error(priv, "fuc09 req 0x32 timeout\n");
+			return -EBUSY;
+		}
+
+		nv_wr32(priv, 0x409614, 0x00000070);
+		nv_wr32(priv, 0x409614, 0x00000770);
+		nv_wr32(priv, 0x40802c, 0x00000001);
+
+		if (priv->data == NULL) {
+			int ret = nve0_grctx_generate(priv);
+			if (ret) {
+				nv_error(priv, "failed to construct context\n");
+				return ret;
+			}
+		}
+
+		return 0;
 	}
 
-	nv_wr32(priv, 0x409800, 0x00000000);
-	nv_wr32(priv, 0x409500, 0x00000001);
-	nv_wr32(priv, 0x409504, 0x00000030);
-	if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
-		nv_error(priv, "fuc09 req 0x30 timeout\n");
-		return -EBUSY;
+	/* load HUB microcode */
+	r000260 = nv_mask(priv, 0x000260, 0x00000001, 0x00000000);
+	nv_wr32(priv, 0x4091c0, 0x01000000);
+	for (i = 0; i < sizeof(nve0_grhub_data) / 4; i++)
+		nv_wr32(priv, 0x4091c4, nve0_grhub_data[i]);
+
+	nv_wr32(priv, 0x409180, 0x01000000);
+	for (i = 0; i < sizeof(nve0_grhub_code) / 4; i++) {
+		if ((i & 0x3f) == 0)
+			nv_wr32(priv, 0x409188, i >> 6);
+		nv_wr32(priv, 0x409184, nve0_grhub_code[i]);
 	}
 
-	nv_wr32(priv, 0x409810, 0xb00095c8);
-	nv_wr32(priv, 0x409800, 0x00000000);
-	nv_wr32(priv, 0x409500, 0x00000001);
-	nv_wr32(priv, 0x409504, 0x00000031);
-	if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
-		nv_error(priv, "fuc09 req 0x31 timeout\n");
-		return -EBUSY;
+	/* load GPC microcode */
+	nv_wr32(priv, 0x41a1c0, 0x01000000);
+	for (i = 0; i < sizeof(nve0_grgpc_data) / 4; i++)
+		nv_wr32(priv, 0x41a1c4, nve0_grgpc_data[i]);
+
+	nv_wr32(priv, 0x41a180, 0x01000000);
+	for (i = 0; i < sizeof(nve0_grgpc_code) / 4; i++) {
+		if ((i & 0x3f) == 0)
+			nv_wr32(priv, 0x41a188, i >> 6);
+		nv_wr32(priv, 0x41a184, nve0_grgpc_code[i]);
 	}
+	nv_wr32(priv, 0x000260, r000260);
 
-	nv_wr32(priv, 0x409810, 0x00080420);
-	nv_wr32(priv, 0x409800, 0x00000000);
-	nv_wr32(priv, 0x409500, 0x00000001);
-	nv_wr32(priv, 0x409504, 0x00000032);
-	if (!nv_wait_ne(priv, 0x409800, 0xffffffff, 0x00000000)) {
-		nv_error(priv, "fuc09 req 0x32 timeout\n");
+	/* start HUB ucode running, it'll init the GPCs */
+	nv_wr32(priv, 0x409800, nv_device(priv)->chipset);
+	nv_wr32(priv, 0x40910c, 0x00000000);
+	nv_wr32(priv, 0x409100, 0x00000002);
+	if (!nv_wait(priv, 0x409800, 0x80000000, 0x80000000)) {
+		nv_error(priv, "HUB_INIT timed out\n");
+		nvc0_graph_ctxctl_debug(priv);
 		return -EBUSY;
 	}
 
-	nv_wr32(priv, 0x409614, 0x00000070);
-	nv_wr32(priv, 0x409614, 0x00000770);
-	nv_wr32(priv, 0x40802c, 0x00000001);
-
+	priv->size = nv_rd32(priv, 0x409804);
 	if (priv->data == NULL) {
 		int ret = nve0_grctx_generate(priv);
 		if (ret) {