xref: /openbmc/linux/arch/sparc/kernel/head_64.S (revision 62e7ca52)
1/* head.S: Initial boot code for the Sparc64 port of Linux.
2 *
3 * Copyright (C) 1996, 1997, 2007 David S. Miller (davem@davemloft.net)
4 * Copyright (C) 1996 David Sitsky (David.Sitsky@anu.edu.au)
5 * Copyright (C) 1997, 1998 Jakub Jelinek (jj@sunsite.mff.cuni.cz)
6 * Copyright (C) 1997 Miguel de Icaza (miguel@nuclecu.unam.mx)
7 */
8
9#include <linux/version.h>
10#include <linux/errno.h>
11#include <linux/threads.h>
12#include <linux/init.h>
13#include <linux/linkage.h>
14#include <asm/thread_info.h>
15#include <asm/asi.h>
16#include <asm/pstate.h>
17#include <asm/ptrace.h>
18#include <asm/spitfire.h>
19#include <asm/page.h>
20#include <asm/pgtable.h>
21#include <asm/errno.h>
22#include <asm/signal.h>
23#include <asm/processor.h>
24#include <asm/lsu.h>
25#include <asm/dcr.h>
26#include <asm/dcu.h>
27#include <asm/head.h>
28#include <asm/ttable.h>
29#include <asm/mmu.h>
30#include <asm/cpudata.h>
31#include <asm/pil.h>
32#include <asm/estate.h>
33#include <asm/sfafsr.h>
34#include <asm/unistd.h>
35
36/* This section from from _start to sparc64_boot_end should fit into
37 * 0x0000000000404000 to 0x0000000000408000.
38 */
39	.text
40	.globl	start, _start, stext, _stext
41_start:
42start:
43_stext:
44stext:
45! 0x0000000000404000
46	b	sparc64_boot
47	 flushw					/* Flush register file.      */
48
49/* This stuff has to be in sync with SILO and other potential boot loaders
50 * Fields should be kept upward compatible and whenever any change is made,
51 * HdrS version should be incremented.
52 */
53        .global root_flags, ram_flags, root_dev
54        .global sparc_ramdisk_image, sparc_ramdisk_size
55	.global sparc_ramdisk_image64
56
57        .ascii  "HdrS"
58        .word   LINUX_VERSION_CODE
59
60	/* History:
61	 *
62	 * 0x0300 : Supports being located at other than 0x4000
63	 * 0x0202 : Supports kernel params string
64	 * 0x0201 : Supports reboot_command
65	 */
66	.half   0x0301          /* HdrS version */
67
68root_flags:
69        .half   1
70root_dev:
71        .half   0
72ram_flags:
73        .half   0
74sparc_ramdisk_image:
75        .word   0
76sparc_ramdisk_size:
77        .word   0
78        .xword  reboot_command
79	.xword	bootstr_info
80sparc_ramdisk_image64:
81	.xword	0
82	.word	_end
83
84	/* PROM cif handler code address is in %o4.  */
85sparc64_boot:
86	mov	%o4, %l7
87
88	/* We need to remap the kernel.  Use position independent
89	 * code to remap us to KERNBASE.
90	 *
91	 * SILO can invoke us with 32-bit address masking enabled,
92	 * so make sure that's clear.
93	 */
94	rdpr	%pstate, %g1
95	andn	%g1, PSTATE_AM, %g1
96	wrpr	%g1, 0x0, %pstate
97	ba,a,pt	%xcc, 1f
98
99	.globl	prom_finddev_name, prom_chosen_path, prom_root_node
100	.globl	prom_getprop_name, prom_mmu_name, prom_peer_name
101	.globl	prom_callmethod_name, prom_translate_name, prom_root_compatible
102	.globl	prom_map_name, prom_unmap_name, prom_mmu_ihandle_cache
103	.globl	prom_boot_mapped_pc, prom_boot_mapping_mode
104	.globl	prom_boot_mapping_phys_high, prom_boot_mapping_phys_low
105	.globl	prom_compatible_name, prom_cpu_path, prom_cpu_compatible
106	.globl	is_sun4v, sun4v_chip_type, prom_set_trap_table_name
107prom_peer_name:
108	.asciz	"peer"
109prom_compatible_name:
110	.asciz	"compatible"
111prom_finddev_name:
112	.asciz	"finddevice"
113prom_chosen_path:
114	.asciz	"/chosen"
115prom_cpu_path:
116	.asciz	"/cpu"
117prom_getprop_name:
118	.asciz	"getprop"
119prom_mmu_name:
120	.asciz	"mmu"
121prom_callmethod_name:
122	.asciz	"call-method"
123prom_translate_name:
124	.asciz	"translate"
125prom_map_name:
126	.asciz	"map"
127prom_unmap_name:
128	.asciz	"unmap"
129prom_set_trap_table_name:
130	.asciz	"SUNW,set-trap-table"
131prom_sun4v_name:
132	.asciz	"sun4v"
133prom_niagara_prefix:
134	.asciz	"SUNW,UltraSPARC-T"
135prom_sparc_prefix:
136	.asciz	"SPARC-"
137prom_sparc64x_prefix:
138	.asciz	"SPARC64-X"
139	.align	4
140prom_root_compatible:
141	.skip	64
142prom_cpu_compatible:
143	.skip	64
144prom_root_node:
145	.word	0
146prom_mmu_ihandle_cache:
147	.word	0
148prom_boot_mapped_pc:
149	.word	0
150prom_boot_mapping_mode:
151	.word	0
152	.align	8
153prom_boot_mapping_phys_high:
154	.xword	0
155prom_boot_mapping_phys_low:
156	.xword	0
157is_sun4v:
158	.word	0
159sun4v_chip_type:
160	.word	SUN4V_CHIP_INVALID
1611:
162	rd	%pc, %l0
163
164	mov	(1b - prom_peer_name), %l1
165	sub	%l0, %l1, %l1
166	mov	0, %l2
167
168	/* prom_root_node = prom_peer(0) */
169	stx	%l1, [%sp + 2047 + 128 + 0x00]	! service, "peer"
170	mov	1, %l3
171	stx	%l3, [%sp + 2047 + 128 + 0x08]	! num_args, 1
172	stx	%l3, [%sp + 2047 + 128 + 0x10]	! num_rets, 1
173	stx	%l2, [%sp + 2047 + 128 + 0x18]	! arg1, 0
174	stx	%g0, [%sp + 2047 + 128 + 0x20]	! ret1
175	call	%l7
176	 add	%sp, (2047 + 128), %o0		! argument array
177
178	ldx	[%sp + 2047 + 128 + 0x20], %l4	! prom root node
179	mov	(1b - prom_root_node), %l1
180	sub	%l0, %l1, %l1
181	stw	%l4, [%l1]
182
183	mov	(1b - prom_getprop_name), %l1
184	mov	(1b - prom_compatible_name), %l2
185	mov	(1b - prom_root_compatible), %l5
186	sub	%l0, %l1, %l1
187	sub	%l0, %l2, %l2
188	sub	%l0, %l5, %l5
189
190	/* prom_getproperty(prom_root_node, "compatible",
191	 *                  &prom_root_compatible, 64)
192	 */
193	stx	%l1, [%sp + 2047 + 128 + 0x00]	! service, "getprop"
194	mov	4, %l3
195	stx	%l3, [%sp + 2047 + 128 + 0x08]	! num_args, 4
196	mov	1, %l3
197	stx	%l3, [%sp + 2047 + 128 + 0x10]	! num_rets, 1
198	stx	%l4, [%sp + 2047 + 128 + 0x18]	! arg1, prom_root_node
199	stx	%l2, [%sp + 2047 + 128 + 0x20]	! arg2, "compatible"
200	stx	%l5, [%sp + 2047 + 128 + 0x28]	! arg3, &prom_root_compatible
201	mov	64, %l3
202	stx	%l3, [%sp + 2047 + 128 + 0x30]	! arg4, size
203	stx	%g0, [%sp + 2047 + 128 + 0x38]	! ret1
204	call	%l7
205	 add	%sp, (2047 + 128), %o0		! argument array
206
207	mov	(1b - prom_finddev_name), %l1
208	mov	(1b - prom_chosen_path), %l2
209	mov	(1b - prom_boot_mapped_pc), %l3
210	sub	%l0, %l1, %l1
211	sub	%l0, %l2, %l2
212	sub	%l0, %l3, %l3
213	stw	%l0, [%l3]
214	sub	%sp, (192 + 128), %sp
215
216	/* chosen_node = prom_finddevice("/chosen") */
217	stx	%l1, [%sp + 2047 + 128 + 0x00]	! service, "finddevice"
218	mov	1, %l3
219	stx	%l3, [%sp + 2047 + 128 + 0x08]	! num_args, 1
220	stx	%l3, [%sp + 2047 + 128 + 0x10]	! num_rets, 1
221	stx	%l2, [%sp + 2047 + 128 + 0x18]	! arg1, "/chosen"
222	stx	%g0, [%sp + 2047 + 128 + 0x20]	! ret1
223	call	%l7
224	 add	%sp, (2047 + 128), %o0		! argument array
225
226	ldx	[%sp + 2047 + 128 + 0x20], %l4	! chosen device node
227
228	mov	(1b - prom_getprop_name), %l1
229	mov	(1b - prom_mmu_name), %l2
230	mov	(1b - prom_mmu_ihandle_cache), %l5
231	sub	%l0, %l1, %l1
232	sub	%l0, %l2, %l2
233	sub	%l0, %l5, %l5
234
235	/* prom_mmu_ihandle_cache = prom_getint(chosen_node, "mmu") */
236	stx	%l1, [%sp + 2047 + 128 + 0x00]	! service, "getprop"
237	mov	4, %l3
238	stx	%l3, [%sp + 2047 + 128 + 0x08]	! num_args, 4
239	mov	1, %l3
240	stx	%l3, [%sp + 2047 + 128 + 0x10]	! num_rets, 1
241	stx	%l4, [%sp + 2047 + 128 + 0x18]	! arg1, chosen_node
242	stx	%l2, [%sp + 2047 + 128 + 0x20]	! arg2, "mmu"
243	stx	%l5, [%sp + 2047 + 128 + 0x28]	! arg3, &prom_mmu_ihandle_cache
244	mov	4, %l3
245	stx	%l3, [%sp + 2047 + 128 + 0x30]	! arg4, sizeof(arg3)
246	stx	%g0, [%sp + 2047 + 128 + 0x38]	! ret1
247	call	%l7
248	 add	%sp, (2047 + 128), %o0		! argument array
249
250	mov	(1b - prom_callmethod_name), %l1
251	mov	(1b - prom_translate_name), %l2
252	sub	%l0, %l1, %l1
253	sub	%l0, %l2, %l2
254	lduw	[%l5], %l5			! prom_mmu_ihandle_cache
255
256	stx	%l1, [%sp + 2047 + 128 + 0x00]	! service, "call-method"
257	mov	3, %l3
258	stx	%l3, [%sp + 2047 + 128 + 0x08]	! num_args, 3
259	mov	5, %l3
260	stx	%l3, [%sp + 2047 + 128 + 0x10]	! num_rets, 5
261	stx	%l2, [%sp + 2047 + 128 + 0x18]	! arg1: "translate"
262	stx	%l5, [%sp + 2047 + 128 + 0x20]	! arg2: prom_mmu_ihandle_cache
263	/* PAGE align */
264	srlx	%l0, 13, %l3
265	sllx	%l3, 13, %l3
266	stx	%l3, [%sp + 2047 + 128 + 0x28]	! arg3: vaddr, our PC
267	stx	%g0, [%sp + 2047 + 128 + 0x30]	! res1
268	stx	%g0, [%sp + 2047 + 128 + 0x38]	! res2
269	stx	%g0, [%sp + 2047 + 128 + 0x40]	! res3
270	stx	%g0, [%sp + 2047 + 128 + 0x48]	! res4
271	stx	%g0, [%sp + 2047 + 128 + 0x50]	! res5
272	call	%l7
273	 add	%sp, (2047 + 128), %o0		! argument array
274
275	ldx	[%sp + 2047 + 128 + 0x40], %l1	! translation mode
276	mov	(1b - prom_boot_mapping_mode), %l4
277	sub	%l0, %l4, %l4
278	stw	%l1, [%l4]
279	mov	(1b - prom_boot_mapping_phys_high), %l4
280	sub	%l0, %l4, %l4
281	ldx	[%sp + 2047 + 128 + 0x48], %l2	! physaddr high
282	stx	%l2, [%l4 + 0x0]
283	ldx	[%sp + 2047 + 128 + 0x50], %l3	! physaddr low
284	/* 4MB align */
285	srlx	%l3, ILOG2_4MB, %l3
286	sllx	%l3, ILOG2_4MB, %l3
287	stx	%l3, [%l4 + 0x8]
288
289	/* Leave service as-is, "call-method" */
290	mov	7, %l3
291	stx	%l3, [%sp + 2047 + 128 + 0x08]	! num_args, 7
292	mov	1, %l3
293	stx	%l3, [%sp + 2047 + 128 + 0x10]	! num_rets, 1
294	mov	(1b - prom_map_name), %l3
295	sub	%l0, %l3, %l3
296	stx	%l3, [%sp + 2047 + 128 + 0x18]	! arg1: "map"
297	/* Leave arg2 as-is, prom_mmu_ihandle_cache */
298	mov	-1, %l3
299	stx	%l3, [%sp + 2047 + 128 + 0x28]	! arg3: mode (-1 default)
300	/* 4MB align the kernel image size. */
301	set	(_end - KERNBASE), %l3
302	set	((4 * 1024 * 1024) - 1), %l4
303	add	%l3, %l4, %l3
304	andn	%l3, %l4, %l3
305	stx	%l3, [%sp + 2047 + 128 + 0x30]	! arg4: roundup(ksize, 4MB)
306	sethi	%hi(KERNBASE), %l3
307	stx	%l3, [%sp + 2047 + 128 + 0x38]	! arg5: vaddr (KERNBASE)
308	stx	%g0, [%sp + 2047 + 128 + 0x40]	! arg6: empty
309	mov	(1b - prom_boot_mapping_phys_low), %l3
310	sub	%l0, %l3, %l3
311	ldx	[%l3], %l3
312	stx	%l3, [%sp + 2047 + 128 + 0x48]	! arg7: phys addr
313	call	%l7
314	 add	%sp, (2047 + 128), %o0		! argument array
315
316	add	%sp, (192 + 128), %sp
317
318	sethi	%hi(prom_root_compatible), %g1
319	or	%g1, %lo(prom_root_compatible), %g1
320	sethi	%hi(prom_sun4v_name), %g7
321	or	%g7, %lo(prom_sun4v_name), %g7
322	mov	5, %g3
32390:	ldub	[%g7], %g2
324	ldub	[%g1], %g4
325	cmp	%g2, %g4
326	bne,pn	%icc, 80f
327	 add	%g7, 1, %g7
328	subcc	%g3, 1, %g3
329	bne,pt	%xcc, 90b
330	 add	%g1, 1, %g1
331
332	sethi	%hi(is_sun4v), %g1
333	or	%g1, %lo(is_sun4v), %g1
334	mov	1, %g7
335	stw	%g7, [%g1]
336
337	/* cpu_node = prom_finddevice("/cpu") */
338	mov	(1b - prom_finddev_name), %l1
339	mov	(1b - prom_cpu_path), %l2
340	sub	%l0, %l1, %l1
341	sub	%l0, %l2, %l2
342	sub	%sp, (192 + 128), %sp
343
344	stx	%l1, [%sp + 2047 + 128 + 0x00]	! service, "finddevice"
345	mov	1, %l3
346	stx	%l3, [%sp + 2047 + 128 + 0x08]	! num_args, 1
347	stx	%l3, [%sp + 2047 + 128 + 0x10]	! num_rets, 1
348	stx	%l2, [%sp + 2047 + 128 + 0x18]	! arg1, "/cpu"
349	stx	%g0, [%sp + 2047 + 128 + 0x20]	! ret1
350	call	%l7
351	 add	%sp, (2047 + 128), %o0		! argument array
352
353	ldx	[%sp + 2047 + 128 + 0x20], %l4	! cpu device node
354
355	mov	(1b - prom_getprop_name), %l1
356	mov	(1b - prom_compatible_name), %l2
357	mov	(1b - prom_cpu_compatible), %l5
358	sub	%l0, %l1, %l1
359	sub	%l0, %l2, %l2
360	sub	%l0, %l5, %l5
361
362	/* prom_getproperty(cpu_node, "compatible",
363	 *                  &prom_cpu_compatible, 64)
364	 */
365	stx	%l1, [%sp + 2047 + 128 + 0x00]	! service, "getprop"
366	mov	4, %l3
367	stx	%l3, [%sp + 2047 + 128 + 0x08]	! num_args, 4
368	mov	1, %l3
369	stx	%l3, [%sp + 2047 + 128 + 0x10]	! num_rets, 1
370	stx	%l4, [%sp + 2047 + 128 + 0x18]	! arg1, cpu_node
371	stx	%l2, [%sp + 2047 + 128 + 0x20]	! arg2, "compatible"
372	stx	%l5, [%sp + 2047 + 128 + 0x28]	! arg3, &prom_cpu_compatible
373	mov	64, %l3
374	stx	%l3, [%sp + 2047 + 128 + 0x30]	! arg4, size
375	stx	%g0, [%sp + 2047 + 128 + 0x38]	! ret1
376	call	%l7
377	 add	%sp, (2047 + 128), %o0		! argument array
378
379	add	%sp, (192 + 128), %sp
380
381	sethi	%hi(prom_cpu_compatible), %g1
382	or	%g1, %lo(prom_cpu_compatible), %g1
383	sethi	%hi(prom_niagara_prefix), %g7
384	or	%g7, %lo(prom_niagara_prefix), %g7
385	mov	17, %g3
38690:	ldub	[%g7], %g2
387	ldub	[%g1], %g4
388	cmp	%g2, %g4
389	bne,pn	%icc, 89f
390	 add	%g7, 1, %g7
391	subcc	%g3, 1, %g3
392	bne,pt	%xcc, 90b
393	 add	%g1, 1, %g1
394	ba,pt	%xcc, 91f
395	 nop
396
39789:	sethi	%hi(prom_cpu_compatible), %g1
398	or	%g1, %lo(prom_cpu_compatible), %g1
399	sethi	%hi(prom_sparc_prefix), %g7
400	or	%g7, %lo(prom_sparc_prefix), %g7
401	mov	6, %g3
40290:	ldub	[%g7], %g2
403	ldub	[%g1], %g4
404	cmp	%g2, %g4
405	bne,pn	%icc, 4f
406	 add	%g7, 1, %g7
407	subcc	%g3, 1, %g3
408	bne,pt	%xcc, 90b
409	 add	%g1, 1, %g1
410
411	sethi	%hi(prom_cpu_compatible), %g1
412	or	%g1, %lo(prom_cpu_compatible), %g1
413	ldub	[%g1 + 6], %g2
414	cmp	%g2, 'T'
415	be,pt	%xcc, 70f
416	 cmp	%g2, 'M'
417	bne,pn	%xcc, 49f
418	 nop
419
42070:	ldub	[%g1 + 7], %g2
421	cmp	%g2, '3'
422	be,pt	%xcc, 5f
423	 mov	SUN4V_CHIP_NIAGARA3, %g4
424	cmp	%g2, '4'
425	be,pt	%xcc, 5f
426	 mov	SUN4V_CHIP_NIAGARA4, %g4
427	cmp	%g2, '5'
428	be,pt	%xcc, 5f
429	 mov	SUN4V_CHIP_NIAGARA5, %g4
430	ba,pt	%xcc, 49f
431	 nop
432
43391:	sethi	%hi(prom_cpu_compatible), %g1
434	or	%g1, %lo(prom_cpu_compatible), %g1
435	ldub	[%g1 + 17], %g2
436	cmp	%g2, '1'
437	be,pt	%xcc, 5f
438	 mov	SUN4V_CHIP_NIAGARA1, %g4
439	cmp	%g2, '2'
440	be,pt	%xcc, 5f
441	 mov	SUN4V_CHIP_NIAGARA2, %g4
442
4434:
444	/* Athena */
445	sethi	%hi(prom_cpu_compatible), %g1
446	or	%g1, %lo(prom_cpu_compatible), %g1
447	sethi	%hi(prom_sparc64x_prefix), %g7
448	or	%g7, %lo(prom_sparc64x_prefix), %g7
449	mov	9, %g3
45041:	ldub	[%g7], %g2
451	ldub	[%g1], %g4
452	cmp	%g2, %g4
453	bne,pn	%icc, 49f
454	add	%g7, 1, %g7
455	subcc	%g3, 1, %g3
456	bne,pt	%xcc, 41b
457	add	%g1, 1, %g1
458	mov	SUN4V_CHIP_SPARC64X, %g4
459	ba,pt	%xcc, 5f
460	nop
461
46249:
463	mov	SUN4V_CHIP_UNKNOWN, %g4
4645:	sethi	%hi(sun4v_chip_type), %g2
465	or	%g2, %lo(sun4v_chip_type), %g2
466	stw	%g4, [%g2]
467
46880:
469	BRANCH_IF_SUN4V(g1, jump_to_sun4u_init)
470	BRANCH_IF_CHEETAH_BASE(g1,g7,cheetah_boot)
471	BRANCH_IF_CHEETAH_PLUS_OR_FOLLOWON(g1,g7,cheetah_plus_boot)
472	ba,pt	%xcc, spitfire_boot
473	 nop
474
475cheetah_plus_boot:
476	/* Preserve OBP chosen DCU and DCR register settings.  */
477	ba,pt	%xcc, cheetah_generic_boot
478	 nop
479
480cheetah_boot:
481	mov	DCR_BPE | DCR_RPE | DCR_SI | DCR_IFPOE | DCR_MS, %g1
482	wr	%g1, %asr18
483
484	sethi	%uhi(DCU_ME|DCU_RE|DCU_HPE|DCU_SPE|DCU_SL|DCU_WE), %g7
485	or	%g7, %ulo(DCU_ME|DCU_RE|DCU_HPE|DCU_SPE|DCU_SL|DCU_WE), %g7
486	sllx	%g7, 32, %g7
487	or	%g7, DCU_DM | DCU_IM | DCU_DC | DCU_IC, %g7
488	stxa	%g7, [%g0] ASI_DCU_CONTROL_REG
489	membar	#Sync
490
491cheetah_generic_boot:
492	mov	TSB_EXTENSION_P, %g3
493	stxa	%g0, [%g3] ASI_DMMU
494	stxa	%g0, [%g3] ASI_IMMU
495	membar	#Sync
496
497	mov	TSB_EXTENSION_S, %g3
498	stxa	%g0, [%g3] ASI_DMMU
499	membar	#Sync
500
501	mov	TSB_EXTENSION_N, %g3
502	stxa	%g0, [%g3] ASI_DMMU
503	stxa	%g0, [%g3] ASI_IMMU
504	membar	#Sync
505
506	ba,a,pt	%xcc, jump_to_sun4u_init
507
508spitfire_boot:
509	/* Typically PROM has already enabled both MMU's and both on-chip
510	 * caches, but we do it here anyway just to be paranoid.
511	 */
512	mov	(LSU_CONTROL_IC|LSU_CONTROL_DC|LSU_CONTROL_IM|LSU_CONTROL_DM), %g1
513	stxa	%g1, [%g0] ASI_LSU_CONTROL
514	membar	#Sync
515
516jump_to_sun4u_init:
517	/*
518	 * Make sure we are in privileged mode, have address masking,
519         * using the ordinary globals and have enabled floating
520         * point.
521	 *
522	 * Again, typically PROM has left %pil at 13 or similar, and
523	 * (PSTATE_PRIV | PSTATE_PEF | PSTATE_IE) in %pstate.
524         */
525	wrpr    %g0, (PSTATE_PRIV|PSTATE_PEF|PSTATE_IE), %pstate
526	wr	%g0, 0, %fprs
527
528	set	sun4u_init, %g2
529	jmpl    %g2 + %g0, %g0
530	 nop
531
532	__REF
533sun4u_init:
534	BRANCH_IF_SUN4V(g1, sun4v_init)
535
536	/* Set ctx 0 */
537	mov		PRIMARY_CONTEXT, %g7
538	stxa		%g0, [%g7] ASI_DMMU
539	membar		#Sync
540
541	mov		SECONDARY_CONTEXT, %g7
542	stxa		%g0, [%g7] ASI_DMMU
543	membar	#Sync
544
545	ba,pt		%xcc, sun4u_continue
546	 nop
547
548sun4v_init:
549	/* Set ctx 0 */
550	mov		PRIMARY_CONTEXT, %g7
551	stxa		%g0, [%g7] ASI_MMU
552	membar		#Sync
553
554	mov		SECONDARY_CONTEXT, %g7
555	stxa		%g0, [%g7] ASI_MMU
556	membar		#Sync
557	ba,pt		%xcc, niagara_tlb_fixup
558	 nop
559
560sun4u_continue:
561	BRANCH_IF_ANY_CHEETAH(g1, g7, cheetah_tlb_fixup)
562
563	ba,pt	%xcc, spitfire_tlb_fixup
564	 nop
565
566niagara_tlb_fixup:
567	mov	3, %g2		/* Set TLB type to hypervisor. */
568	sethi	%hi(tlb_type), %g1
569	stw	%g2, [%g1 + %lo(tlb_type)]
570
571	/* Patch copy/clear ops.  */
572	sethi	%hi(sun4v_chip_type), %g1
573	lduw	[%g1 + %lo(sun4v_chip_type)], %g1
574	cmp	%g1, SUN4V_CHIP_NIAGARA1
575	be,pt	%xcc, niagara_patch
576	 cmp	%g1, SUN4V_CHIP_NIAGARA2
577	be,pt	%xcc, niagara2_patch
578	 nop
579	cmp	%g1, SUN4V_CHIP_NIAGARA3
580	be,pt	%xcc, niagara2_patch
581	 nop
582	cmp	%g1, SUN4V_CHIP_NIAGARA4
583	be,pt	%xcc, niagara4_patch
584	 nop
585	cmp	%g1, SUN4V_CHIP_NIAGARA5
586	be,pt	%xcc, niagara4_patch
587	 nop
588
589	call	generic_patch_copyops
590	 nop
591	call	generic_patch_bzero
592	 nop
593	call	generic_patch_pageops
594	 nop
595
596	ba,a,pt	%xcc, 80f
597niagara4_patch:
598	call	niagara4_patch_copyops
599	 nop
600	call	niagara4_patch_bzero
601	 nop
602	call	niagara4_patch_pageops
603	 nop
604
605	ba,a,pt	%xcc, 80f
606
607niagara2_patch:
608	call	niagara2_patch_copyops
609	 nop
610	call	niagara_patch_bzero
611	 nop
612	call	niagara_patch_pageops
613	 nop
614
615	ba,a,pt	%xcc, 80f
616
617niagara_patch:
618	call	niagara_patch_copyops
619	 nop
620	call	niagara_patch_bzero
621	 nop
622	call	niagara_patch_pageops
623	 nop
624
62580:
626	/* Patch TLB/cache ops.  */
627	call	hypervisor_patch_cachetlbops
628	 nop
629
630	ba,pt	%xcc, tlb_fixup_done
631	 nop
632
633cheetah_tlb_fixup:
634	mov	2, %g2		/* Set TLB type to cheetah+. */
635	BRANCH_IF_CHEETAH_PLUS_OR_FOLLOWON(g1,g7,1f)
636
637	mov	1, %g2		/* Set TLB type to cheetah. */
638
6391:	sethi	%hi(tlb_type), %g1
640	stw	%g2, [%g1 + %lo(tlb_type)]
641
642	/* Patch copy/page operations to cheetah optimized versions. */
643	call	cheetah_patch_copyops
644	 nop
645	call	cheetah_patch_copy_page
646	 nop
647	call	cheetah_patch_cachetlbops
648	 nop
649
650	ba,pt	%xcc, tlb_fixup_done
651	 nop
652
653spitfire_tlb_fixup:
654	/* Set TLB type to spitfire. */
655	mov	0, %g2
656	sethi	%hi(tlb_type), %g1
657	stw	%g2, [%g1 + %lo(tlb_type)]
658
659tlb_fixup_done:
660	sethi	%hi(init_thread_union), %g6
661	or	%g6, %lo(init_thread_union), %g6
662	ldx	[%g6 + TI_TASK], %g4
663	mov	%sp, %l6
664
665	wr	%g0, ASI_P, %asi
666	mov	1, %g1
667	sllx	%g1, THREAD_SHIFT, %g1
668	sub	%g1, (STACKFRAME_SZ + STACK_BIAS), %g1
669	add	%g6, %g1, %sp
670	mov	0, %fp
671
672	/* Set per-cpu pointer initially to zero, this makes
673	 * the boot-cpu use the in-kernel-image per-cpu areas
674	 * before setup_per_cpu_area() is invoked.
675	 */
676	clr	%g5
677
678	wrpr	%g0, 0, %wstate
679	wrpr	%g0, 0x0, %tl
680
681	/* Clear the bss */
682	sethi	%hi(__bss_start), %o0
683	or	%o0, %lo(__bss_start), %o0
684	sethi	%hi(_end), %o1
685	or	%o1, %lo(_end), %o1
686	call	__bzero
687	 sub	%o1, %o0, %o1
688
689#ifdef CONFIG_LOCKDEP
690	/* We have this call this super early, as even prom_init can grab
691	 * spinlocks and thus call into the lockdep code.
692	 */
693	call	lockdep_init
694	 nop
695#endif
696
697	mov	%l6, %o1			! OpenPROM stack
698	call	prom_init
699	 mov	%l7, %o0			! OpenPROM cif handler
700
701	/* Initialize current_thread_info()->cpu as early as possible.
702	 * In order to do that accurately we have to patch up the get_cpuid()
703	 * assembler sequences.  And that, in turn, requires that we know
704	 * if we are on a Starfire box or not.  While we're here, patch up
705	 * the sun4v sequences as well.
706	 */
707	call	check_if_starfire
708	 nop
709	call	per_cpu_patch
710	 nop
711	call	sun4v_patch
712	 nop
713
714#ifdef CONFIG_SMP
715	call	hard_smp_processor_id
716	 nop
717	cmp	%o0, NR_CPUS
718	blu,pt	%xcc, 1f
719	 nop
720	call	boot_cpu_id_too_large
721	 nop
722	/* Not reached... */
723
7241:
725#else
726	mov	0, %o0
727#endif
728	sth	%o0, [%g6 + TI_CPU]
729
730	call	prom_init_report
731	 nop
732
733	/* Off we go.... */
734	call	start_kernel
735	 nop
736	/* Not reached... */
737
738	.previous
739
740	/* This is meant to allow the sharing of this code between
741	 * boot processor invocation (via setup_tba() below) and
742	 * secondary processor startup (via trampoline.S).  The
743	 * former does use this code, the latter does not yet due
744	 * to some complexities.  That should be fixed up at some
745	 * point.
746	 *
747	 * There used to be enormous complexity wrt. transferring
748	 * over from the firmware's trap table to the Linux kernel's.
749	 * For example, there was a chicken & egg problem wrt. building
750	 * the OBP page tables, yet needing to be on the Linux kernel
751	 * trap table (to translate PAGE_OFFSET addresses) in order to
752	 * do that.
753	 *
754	 * We now handle OBP tlb misses differently, via linear lookups
755	 * into the prom_trans[] array.  So that specific problem no
756	 * longer exists.  Yet, unfortunately there are still some issues
757	 * preventing trampoline.S from using this code... ho hum.
758	 */
759	.globl	setup_trap_table
760setup_trap_table:
761	save	%sp, -192, %sp
762
763	/* Force interrupts to be disabled. */
764	rdpr	%pstate, %l0
765	andn	%l0, PSTATE_IE, %o1
766	wrpr	%o1, 0x0, %pstate
767	rdpr	%pil, %l1
768	wrpr	%g0, PIL_NORMAL_MAX, %pil
769
770	/* Make the firmware call to jump over to the Linux trap table.  */
771	sethi	%hi(is_sun4v), %o0
772	lduw	[%o0 + %lo(is_sun4v)], %o0
773	brz,pt	%o0, 1f
774	 nop
775
776	TRAP_LOAD_TRAP_BLOCK(%g2, %g3)
777	add	%g2, TRAP_PER_CPU_FAULT_INFO, %g2
778	stxa	%g2, [%g0] ASI_SCRATCHPAD
779
780	/* Compute physical address:
781	 *
782	 * paddr = kern_base + (mmfsa_vaddr - KERNBASE)
783	 */
784	sethi	%hi(KERNBASE), %g3
785	sub	%g2, %g3, %g2
786	sethi	%hi(kern_base), %g3
787	ldx	[%g3 + %lo(kern_base)], %g3
788	add	%g2, %g3, %o1
789	sethi	%hi(sparc64_ttable_tl0), %o0
790
791	set	prom_set_trap_table_name, %g2
792	stx	%g2, [%sp + 2047 + 128 + 0x00]
793	mov	2, %g2
794	stx	%g2, [%sp + 2047 + 128 + 0x08]
795	mov	0, %g2
796	stx	%g2, [%sp + 2047 + 128 + 0x10]
797	stx	%o0, [%sp + 2047 + 128 + 0x18]
798	stx	%o1, [%sp + 2047 + 128 + 0x20]
799	sethi	%hi(p1275buf), %g2
800	or	%g2, %lo(p1275buf), %g2
801	ldx	[%g2 + 0x08], %o1
802	call	%o1
803	 add	%sp, (2047 + 128), %o0
804
805	ba,pt	%xcc, 2f
806	 nop
807
8081:	sethi	%hi(sparc64_ttable_tl0), %o0
809	set	prom_set_trap_table_name, %g2
810	stx	%g2, [%sp + 2047 + 128 + 0x00]
811	mov	1, %g2
812	stx	%g2, [%sp + 2047 + 128 + 0x08]
813	mov	0, %g2
814	stx	%g2, [%sp + 2047 + 128 + 0x10]
815	stx	%o0, [%sp + 2047 + 128 + 0x18]
816	sethi	%hi(p1275buf), %g2
817	or	%g2, %lo(p1275buf), %g2
818	ldx	[%g2 + 0x08], %o1
819	call	%o1
820	 add	%sp, (2047 + 128), %o0
821
822	/* Start using proper page size encodings in ctx register.  */
8232:	sethi	%hi(sparc64_kern_pri_context), %g3
824	ldx	[%g3 + %lo(sparc64_kern_pri_context)], %g2
825
826	mov		PRIMARY_CONTEXT, %g1
827
828661:	stxa		%g2, [%g1] ASI_DMMU
829	.section	.sun4v_1insn_patch, "ax"
830	.word		661b
831	stxa		%g2, [%g1] ASI_MMU
832	.previous
833
834	membar	#Sync
835
836	BRANCH_IF_SUN4V(o2, 1f)
837
838	/* Kill PROM timer */
839	sethi	%hi(0x80000000), %o2
840	sllx	%o2, 32, %o2
841	wr	%o2, 0, %tick_cmpr
842
843	BRANCH_IF_ANY_CHEETAH(o2, o3, 1f)
844
845	ba,pt	%xcc, 2f
846	 nop
847
848	/* Disable STICK_INT interrupts. */
8491:
850	sethi	%hi(0x80000000), %o2
851	sllx	%o2, 32, %o2
852	wr	%o2, %asr25
853
8542:
855	wrpr	%g0, %g0, %wstate
856
857	call	init_irqwork_curcpu
858	 nop
859
860	/* Now we can restore interrupt state. */
861	wrpr	%l0, 0, %pstate
862	wrpr	%l1, 0x0, %pil
863
864	ret
865	 restore
866
867	.globl	setup_tba
868setup_tba:
869	save	%sp, -192, %sp
870
871	/* The boot processor is the only cpu which invokes this
872	 * routine, the other cpus set things up via trampoline.S.
873	 * So save the OBP trap table address here.
874	 */
875	rdpr	%tba, %g7
876	sethi	%hi(prom_tba), %o1
877	or	%o1, %lo(prom_tba), %o1
878	stx	%g7, [%o1]
879
880	call	setup_trap_table
881	 nop
882
883	ret
884	 restore
885sparc64_boot_end:
886
887#include "etrap_64.S"
888#include "rtrap_64.S"
889#include "winfixup.S"
890#include "fpu_traps.S"
891#include "ivec.S"
892#include "getsetcc.S"
893#include "utrap.S"
894#include "spiterrs.S"
895#include "cherrs.S"
896#include "misctrap.S"
897#include "syscalls.S"
898#include "helpers.S"
899#include "hvcalls.S"
900#include "sun4v_tlb_miss.S"
901#include "sun4v_ivec.S"
902#include "ktlb.S"
903#include "tsb.S"
904
905/*
906 * The following skip makes sure the trap table in ttable.S is aligned
907 * on a 32K boundary as required by the v9 specs for TBA register.
908 *
909 * We align to a 32K boundary, then we have the 32K kernel TSB,
910 * the 64K kernel 4MB TSB, and then the 32K aligned trap table.
911 */
9121:
913	.skip	0x4000 + _start - 1b
914
915! 0x0000000000408000
916
917	.globl	swapper_tsb
918swapper_tsb:
919	.skip	(32 * 1024)
920
921	.globl	swapper_4m_tsb
922swapper_4m_tsb:
923	.skip	(64 * 1024)
924
925! 0x0000000000420000
926
927	/* Some care needs to be exercised if you try to move the
928	 * location of the trap table relative to other things.  For
929	 * one thing there are br* instructions in some of the
930	 * trap table entires which branch back to code in ktlb.S
931	 * Those instructions can only handle a signed 16-bit
932	 * displacement.
933	 *
934	 * There is a binutils bug (bugzilla #4558) which causes
935	 * the relocation overflow checks for such instructions to
936	 * not be done correctly.  So bintuils will not notice the
937	 * error and will instead write junk into the relocation and
938	 * you'll have an unbootable kernel.
939	 */
940#include "ttable_64.S"
941
942! 0x0000000000428000
943
944#include "systbls_64.S"
945
946	.data
947	.align	8
948	.globl	prom_tba, tlb_type
949prom_tba:	.xword	0
950tlb_type:	.word	0	/* Must NOT end up in BSS */
951	.section	".fixup",#alloc,#execinstr
952
953	.globl	__ret_efault, __retl_efault, __ret_one, __retl_one
954ENTRY(__ret_efault)
955	ret
956	 restore %g0, -EFAULT, %o0
957ENDPROC(__ret_efault)
958
959ENTRY(__retl_efault)
960	retl
961	 mov	-EFAULT, %o0
962ENDPROC(__retl_efault)
963
964ENTRY(__retl_one)
965	retl
966	 mov	1, %o0
967ENDPROC(__retl_one)
968
969ENTRY(__ret_one_asi)
970	wr	%g0, ASI_AIUS, %asi
971	ret
972	 restore %g0, 1, %o0
973ENDPROC(__ret_one_asi)
974
975ENTRY(__retl_one_asi)
976	wr	%g0, ASI_AIUS, %asi
977	retl
978	 mov	1, %o0
979ENDPROC(__retl_one_asi)
980
981ENTRY(__retl_o1)
982	retl
983	 mov	%o1, %o0
984ENDPROC(__retl_o1)
985