1/*
2 * Copyright (C) 2007-2009 Michal Simek <monstr@monstr.eu>
3 * Copyright (C) 2007-2009 PetaLogix
4 * Copyright (C) 2006 Atmark Techno, Inc.
5 *
6 * MMU code derived from arch/ppc/kernel/head_4xx.S:
7 *    Copyright (c) 1995-1996 Gary Thomas <gdt@linuxppc.org>
8 *      Initial PowerPC version.
9 *    Copyright (c) 1996 Cort Dougan <cort@cs.nmt.edu>
10 *      Rewritten for PReP
11 *    Copyright (c) 1996 Paul Mackerras <paulus@cs.anu.edu.au>
12 *      Low-level exception handers, MMU support, and rewrite.
13 *    Copyright (c) 1997 Dan Malek <dmalek@jlc.net>
14 *      PowerPC 8xx modifications.
15 *    Copyright (c) 1998-1999 TiVo, Inc.
16 *      PowerPC 403GCX modifications.
17 *    Copyright (c) 1999 Grant Erickson <grant@lcse.umn.edu>
18 *      PowerPC 403GCX/405GP modifications.
19 *    Copyright 2000 MontaVista Software Inc.
20 *	PPC405 modifications
21 *      PowerPC 403GCX/405GP modifications.
22 * 	Author: MontaVista Software, Inc.
23 *         	frank_rowand@mvista.com or source@mvista.com
24 * 	   	debbie_chu@mvista.com
25 *
26 * This file is subject to the terms and conditions of the GNU General Public
27 * License. See the file "COPYING" in the main directory of this archive
28 * for more details.
29 */
30
31#include <linux/init.h>
32#include <linux/linkage.h>
33#include <asm/thread_info.h>
34#include <asm/page.h>
35#include <linux/of_fdt.h>		/* for OF_DT_HEADER */
36
37#include <asm/setup.h> /* COMMAND_LINE_SIZE */
38#include <asm/mmu.h>
39#include <asm/processor.h>
40
41.section .data
42.global empty_zero_page
43.align 12
44empty_zero_page:
45	.space	PAGE_SIZE
46.global swapper_pg_dir
47swapper_pg_dir:
48	.space	PAGE_SIZE
49
50.section .rodata
51.align 4
52endian_check:
53	.word	1
54
55	__HEAD
56ENTRY(_start)
57#if CONFIG_KERNEL_BASE_ADDR == 0
58	brai	TOPHYS(real_start)
59	.org	0x100
60real_start:
61#endif
62
63	mts	rmsr, r0
64/* Disable stack protection from bootloader */
65	mts	rslr, r0
66	addi	r8, r0, 0xFFFFFFFF
67	mts	rshr, r8
68/*
69 * According to Xilinx, msrclr instruction behaves like 'mfs rX,rpc'
70 * if the msrclr instruction is not enabled. We use this to detect
71 * if the opcode is available, by issuing msrclr and then testing the result.
72 * r8 == 0 - msr instructions are implemented
73 * r8 != 0 - msr instructions are not implemented
74 */
75	mfs	r1, rmsr
76	msrclr	r8, 0 /* clear nothing - just read msr for test */
77	cmpu	r8, r8, r1 /* r1 must contain msr reg content */
78
79/* r7 may point to an FDT, or there may be one linked in.
80   if it's in r7, we've got to save it away ASAP.
81   We ensure r7 points to a valid FDT, just in case the bootloader
82   is broken or non-existent */
83	beqi	r7, no_fdt_arg			/* NULL pointer?  don't copy */
84/* Does r7 point to a valid FDT? Load HEADER magic number */
85	/* Run time Big/Little endian platform */
86	/* Save 1 as word and load byte - 0 - BIG, 1 - LITTLE */
87	lbui	r11, r0, TOPHYS(endian_check)
88	beqid	r11, big_endian /* DO NOT break delay stop dependency */
89	lw	r11, r0, r7 /* Big endian load in delay slot */
90	lwr	r11, r0, r7 /* Little endian load */
91big_endian:
92	rsubi	r11, r11, OF_DT_HEADER	/* Check FDT header */
93	beqi	r11, _prepare_copy_fdt
94	or	r7, r0, r0		/* clear R7 when not valid DTB */
95	bnei	r11, no_fdt_arg			/* No - get out of here */
96_prepare_copy_fdt:
97	or	r11, r0, r0 /* incremment */
98	ori	r4, r0, TOPHYS(_fdt_start)
99	ori	r3, r0, (0x10000 - 4)
100_copy_fdt:
101	lw	r12, r7, r11 /* r12 = r7 + r11 */
102	sw	r12, r4, r11 /* addr[r4 + r11] = r12 */
103	addik	r11, r11, 4 /* increment counting */
104	bgtid	r3, _copy_fdt /* loop for all entries */
105	addik	r3, r3, -4 /* descrement loop */
106no_fdt_arg:
107
108#ifndef CONFIG_CMDLINE_BOOL
109/*
110 * handling command line
111 * copy command line directly to cmd_line placed in data section.
112 */
113	beqid	r5, skip	/* Skip if NULL pointer */
114	or	r11, r0, r0		/* incremment */
115	ori	r4, r0, cmd_line	/* load address of command line */
116	tophys(r4,r4)			/* convert to phys address */
117	ori	r3, r0, COMMAND_LINE_SIZE - 1 /* number of loops */
118_copy_command_line:
119	/* r2=r5+r11 - r5 contain pointer to command line */
120	lbu	r2, r5, r11
121	beqid	r2, skip		/* Skip if no data */
122	sb	r2, r4, r11		/* addr[r4+r11]= r2 */
123	addik	r11, r11, 1		/* increment counting */
124	bgtid	r3, _copy_command_line	/* loop for all entries       */
125	addik	r3, r3, -1		/* decrement loop */
126	addik	r5, r4, 0		/* add new space for command line */
127	tovirt(r5,r5)
128skip:
129#endif /* CONFIG_CMDLINE_BOOL */
130
131#ifdef NOT_COMPILE
132/* save bram context */
133	or	r11, r0, r0				/* incremment */
134	ori	r4, r0, TOPHYS(_bram_load_start)	/* save bram context */
135	ori	r3, r0, (LMB_SIZE - 4)
136_copy_bram:
137	lw	r7, r0, r11		/* r7 = r0 + r11 */
138	sw	r7, r4, r11		/* addr[r4 + r11] = r7 */
139	addik	r11, r11, 4		/* increment counting */
140	bgtid	r3, _copy_bram		/* loop for all entries */
141	addik	r3, r3, -4		/* descrement loop */
142#endif
143	/* We have to turn on the MMU right away. */
144
145	/*
146	 * Set up the initial MMU state so we can do the first level of
147	 * kernel initialization.  This maps the first 16 MBytes of memory 1:1
148	 * virtual to physical.
149	 */
150	nop
151	addik	r3, r0, MICROBLAZE_TLB_SIZE -1	/* Invalidate all TLB entries */
152_invalidate:
153	mts	rtlbx, r3
154	mts	rtlbhi, r0			/* flush: ensure V is clear   */
155	mts	rtlblo, r0
156	bgtid	r3, _invalidate		/* loop for all entries       */
157	addik	r3, r3, -1
158	/* sync */
159
160	/* Setup the kernel PID */
161	mts	rpid,r0			/* Load the kernel PID */
162	nop
163	bri	4
164
165	/*
166	 * We should still be executing code at physical address area
167	 * RAM_BASEADDR at this point. However, kernel code is at
168	 * a virtual address. So, set up a TLB mapping to cover this once
169	 * translation is enabled.
170	 */
171
172	addik	r3,r0, CONFIG_KERNEL_START /* Load the kernel virtual address */
173	tophys(r4,r3)			/* Load the kernel physical address */
174
175	/* start to do TLB calculation */
176	addik	r12, r0, _end
177	rsub	r12, r3, r12
178	addik	r12, r12, CONFIG_LOWMEM_SIZE >> PTE_SHIFT /* that's the pad */
179
180	or r9, r0, r0 /* TLB0 = 0 */
181	or r10, r0, r0 /* TLB1 = 0 */
182
183	addik	r11, r12, -0x1000000
184	bgei	r11, GT16 /* size is greater than 16MB */
185	addik	r11, r12, -0x0800000
186	bgei	r11, GT8 /* size is greater than 8MB */
187	addik	r11, r12, -0x0400000
188	bgei	r11, GT4 /* size is greater than 4MB */
189	/* size is less than 4MB */
190	addik	r11, r12, -0x0200000
191	bgei	r11, GT2 /* size is greater than 2MB */
192	addik	r9, r0, 0x0100000 /* TLB0 must be 1MB */
193	addik	r11, r12, -0x0100000
194	bgei	r11, GT1 /* size is greater than 1MB */
195	/* TLB1 is 0 which is setup above */
196	bri tlb_end
197GT4: /* r11 contains the rest - will be either 1 or 4 */
198	ori r9, r0, 0x400000 /* TLB0 is 4MB */
199	bri TLB1
200GT16: /* TLB0 is 16MB */
201	addik	r9, r0, 0x1000000 /* means TLB0 is 16MB */
202TLB1:
203	/* must be used r2 because of subtract if failed */
204	addik	r2, r11, -0x0400000
205	bgei	r2, GT20 /* size is greater than 16MB */
206	/* size is >16MB and <20MB */
207	addik	r11, r11, -0x0100000
208	bgei	r11, GT17 /* size is greater than 17MB */
209	/* kernel is >16MB and < 17MB */
210GT1:
211	addik	r10, r0, 0x0100000 /* means TLB1 is 1MB */
212	bri tlb_end
213GT2: /* TLB0 is 0 and TLB1 will be 4MB */
214GT17: /* TLB1 is 4MB - kernel size <20MB */
215	addik	r10, r0, 0x0400000 /* means TLB1 is 4MB */
216	bri tlb_end
217GT8: /* TLB0 is still zero that's why I can use only TLB1 */
218GT20: /* TLB1 is 16MB - kernel size >20MB */
219	addik	r10, r0, 0x1000000 /* means TLB1 is 16MB */
220tlb_end:
221
222	/*
223	 * Configure and load two entries into TLB slots 0 and 1.
224	 * In case we are pinning TLBs, these are reserved in by the
225	 * other TLB functions.  If not reserving, then it doesn't
226	 * matter where they are loaded.
227	 */
228	andi	r4,r4,0xfffffc00	/* Mask off the real page number */
229	ori	r4,r4,(TLB_WR | TLB_EX)	/* Set the write and execute bits */
230
231	/*
232	 * TLB0 is always used - check if is not zero (r9 stores TLB0 value)
233	 * if is use TLB1 value and clear it (r10 stores TLB1 value)
234	 */
235	bnei	r9, tlb0_not_zero
236	add	r9, r10, r0
237	add	r10, r0, r0
238tlb0_not_zero:
239
240	/* look at the code below */
241	ori	r30, r0, 0x200
242	andi	r29, r9, 0x100000
243	bneid	r29, 1f
244	addik	r30, r30, 0x80
245	andi	r29, r9, 0x400000
246	bneid	r29, 1f
247	addik	r30, r30, 0x80
248	andi	r29, r9, 0x1000000
249	bneid	r29, 1f
250	addik	r30, r30, 0x80
2511:
252	andi	r3,r3,0xfffffc00	/* Mask off the effective page number */
253	ori	r3,r3,(TLB_VALID)
254	or	r3, r3, r30
255
256	/* Load tlb_skip size value which is index to first unused TLB entry */
257	lwi	r11, r0, TOPHYS(tlb_skip)
258	mts     rtlbx,r11		/* TLB slow 0 */
259
260	mts	rtlblo,r4		/* Load the data portion of the entry */
261	mts	rtlbhi,r3		/* Load the tag portion of the entry */
262
263	/* Increase tlb_skip size */
264	addik	r11, r11, 1
265	swi	r11, r0, TOPHYS(tlb_skip)
266
267	/* TLB1 can be zeroes that's why we not setup it */
268	beqi	r10, jump_over2
269
270	/* look at the code below */
271	ori	r30, r0, 0x200
272	andi	r29, r10, 0x100000
273	bneid	r29, 1f
274	addik	r30, r30, 0x80
275	andi	r29, r10, 0x400000
276	bneid	r29, 1f
277	addik	r30, r30, 0x80
278	andi	r29, r10, 0x1000000
279	bneid	r29, 1f
280	addik	r30, r30, 0x80
2811:
282	addk	r4, r4, r9	/* previous addr + TLB0 size */
283	addk	r3, r3, r9
284
285	andi	r3,r3,0xfffffc00	/* Mask off the effective page number */
286	ori	r3,r3,(TLB_VALID)
287	or	r3, r3, r30
288
289	lwi	r11, r0, TOPHYS(tlb_skip)
290	mts     rtlbx, r11		/* r11 is used from TLB0 */
291
292	mts	rtlblo,r4		/* Load the data portion of the entry */
293	mts	rtlbhi,r3		/* Load the tag portion of the entry */
294
295	/* Increase tlb_skip size */
296	addik	r11, r11, 1
297	swi	r11, r0, TOPHYS(tlb_skip)
298
299jump_over2:
300	/*
301	 * Load a TLB entry for LMB, since we need access to
302	 * the exception vectors, using a 4k real==virtual mapping.
303	 */
304	/* Use temporary TLB_ID for LMB - clear this temporary mapping later */
305	ori	r11, r0, MICROBLAZE_LMB_TLB_ID
306	mts     rtlbx,r11
307
308	ori	r4,r0,(TLB_WR | TLB_EX)
309	ori	r3,r0,(TLB_VALID | TLB_PAGESZ(PAGESZ_4K))
310
311	mts	rtlblo,r4		/* Load the data portion of the entry */
312	mts	rtlbhi,r3		/* Load the tag portion of the entry */
313
314	/*
315	 * We now have the lower 16 Meg of RAM mapped into TLB entries, and the
316	 * caches ready to work.
317	 */
318turn_on_mmu:
319	ori	r15,r0,start_here
320	ori	r4,r0,MSR_KERNEL_VMS
321	mts	rmsr,r4
322	nop
323	rted	r15,0			/* enables MMU */
324	nop
325
326start_here:
327
328	/* Initialize small data anchors */
329	addik	r13, r0, _KERNEL_SDA_BASE_
330	addik	r2, r0, _KERNEL_SDA2_BASE_
331
332	/* Initialize stack pointer */
333	addik	r1, r0, init_thread_union + THREAD_SIZE - 4
334
335	/* Initialize r31 with current task address */
336	addik	r31, r0, init_task
337
338	addik	r11, r0, machine_early_init
339	brald	r15, r11
340	nop
341
342	/*
343	 * Initialize the MMU.
344	 */
345	bralid	r15, mmu_init
346	nop
347
348	/* Go back to running unmapped so we can load up new values
349	 * and change to using our exception vectors.
350	 * On the MicroBlaze, all we invalidate the used TLB entries to clear
351	 * the old 16M byte TLB mappings.
352	 */
353	ori	r15,r0,TOPHYS(kernel_load_context)
354	ori	r4,r0,MSR_KERNEL
355	mts	rmsr,r4
356	nop
357	bri	4
358	rted	r15,0
359	nop
360
361	/* Load up the kernel context */
362kernel_load_context:
363	ori	r5, r0, MICROBLAZE_LMB_TLB_ID
364	mts     rtlbx,r5
365	nop
366	mts	rtlbhi,r0
367	nop
368	addi	r15, r0, machine_halt
369	ori	r17, r0, start_kernel
370	ori	r4, r0, MSR_KERNEL_VMS
371	mts	rmsr, r4
372	nop
373	rted	r17, 0		/* enable MMU and jump to start_kernel */
374	nop
375