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>

drivers/gpu/drm/nouveau/core/engine/graph/fuc/gpcnve0.fuc.h

+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,
+};

drivers/gpu/drm/nouveau/core/engine/graph/fuc/nve0.fuc

+/* 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
+')

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) {