xref: /openbmc/linux/arch/arm64/kernel/head.S (revision 8571e645)
1/*
2 * Low-level CPU initialisation
3 * Based on arch/arm/kernel/head.S
4 *
5 * Copyright (C) 1994-2002 Russell King
6 * Copyright (C) 2003-2012 ARM Ltd.
7 * Authors:	Catalin Marinas <catalin.marinas@arm.com>
8 *		Will Deacon <will.deacon@arm.com>
9 *
10 * This program is free software; you can redistribute it and/or modify
11 * it under the terms of the GNU General Public License version 2 as
12 * published by the Free Software Foundation.
13 *
14 * This program is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
17 * GNU General Public License for more details.
18 *
19 * You should have received a copy of the GNU General Public License
20 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
21 */
22
23#include <linux/linkage.h>
24#include <linux/init.h>
25#include <linux/irqchip/arm-gic-v3.h>
26
27#include <asm/assembler.h>
28#include <asm/ptrace.h>
29#include <asm/asm-offsets.h>
30#include <asm/cache.h>
31#include <asm/cputype.h>
32#include <asm/elf.h>
33#include <asm/kernel-pgtable.h>
34#include <asm/kvm_arm.h>
35#include <asm/memory.h>
36#include <asm/pgtable-hwdef.h>
37#include <asm/pgtable.h>
38#include <asm/page.h>
39#include <asm/smp.h>
40#include <asm/sysreg.h>
41#include <asm/thread_info.h>
42#include <asm/virt.h>
43
44#define __PHYS_OFFSET	(KERNEL_START - TEXT_OFFSET)
45
46#if (TEXT_OFFSET & 0xfff) != 0
47#error TEXT_OFFSET must be at least 4KB aligned
48#elif (PAGE_OFFSET & 0x1fffff) != 0
49#error PAGE_OFFSET must be at least 2MB aligned
50#elif TEXT_OFFSET > 0x1fffff
51#error TEXT_OFFSET must be less than 2MB
52#endif
53
54#define KERNEL_START	_text
55#define KERNEL_END	_end
56
57/*
58 * Kernel startup entry point.
59 * ---------------------------
60 *
61 * The requirements are:
62 *   MMU = off, D-cache = off, I-cache = on or off,
63 *   x0 = physical address to the FDT blob.
64 *
65 * This code is mostly position independent so you call this at
66 * __pa(PAGE_OFFSET + TEXT_OFFSET).
67 *
68 * Note that the callee-saved registers are used for storing variables
69 * that are useful before the MMU is enabled. The allocations are described
70 * in the entry routines.
71 */
72	__HEAD
73_head:
74	/*
75	 * DO NOT MODIFY. Image header expected by Linux boot-loaders.
76	 */
77#ifdef CONFIG_EFI
78	/*
79	 * This add instruction has no meaningful effect except that
80	 * its opcode forms the magic "MZ" signature required by UEFI.
81	 */
82	add	x13, x18, #0x16
83	b	stext
84#else
85	b	stext				// branch to kernel start, magic
86	.long	0				// reserved
87#endif
88	le64sym	_kernel_offset_le		// Image load offset from start of RAM, little-endian
89	le64sym	_kernel_size_le			// Effective size of kernel image, little-endian
90	le64sym	_kernel_flags_le		// Informative flags, little-endian
91	.quad	0				// reserved
92	.quad	0				// reserved
93	.quad	0				// reserved
94	.byte	0x41				// Magic number, "ARM\x64"
95	.byte	0x52
96	.byte	0x4d
97	.byte	0x64
98#ifdef CONFIG_EFI
99	.long	pe_header - _head		// Offset to the PE header.
100#else
101	.word	0				// reserved
102#endif
103
104#ifdef CONFIG_EFI
105	.globl	__efistub_stext_offset
106	.set	__efistub_stext_offset, stext - _head
107	.align 3
108pe_header:
109	.ascii	"PE"
110	.short 	0
111coff_header:
112	.short	0xaa64				// AArch64
113	.short	2				// nr_sections
114	.long	0 				// TimeDateStamp
115	.long	0				// PointerToSymbolTable
116	.long	1				// NumberOfSymbols
117	.short	section_table - optional_header	// SizeOfOptionalHeader
118	.short	0x206				// Characteristics.
119						// IMAGE_FILE_DEBUG_STRIPPED |
120						// IMAGE_FILE_EXECUTABLE_IMAGE |
121						// IMAGE_FILE_LINE_NUMS_STRIPPED
122optional_header:
123	.short	0x20b				// PE32+ format
124	.byte	0x02				// MajorLinkerVersion
125	.byte	0x14				// MinorLinkerVersion
126	.long	_end - stext			// SizeOfCode
127	.long	0				// SizeOfInitializedData
128	.long	0				// SizeOfUninitializedData
129	.long	__efistub_entry - _head		// AddressOfEntryPoint
130	.long	__efistub_stext_offset		// BaseOfCode
131
132extra_header_fields:
133	.quad	0				// ImageBase
134	.long	0x1000				// SectionAlignment
135	.long	PECOFF_FILE_ALIGNMENT		// FileAlignment
136	.short	0				// MajorOperatingSystemVersion
137	.short	0				// MinorOperatingSystemVersion
138	.short	0				// MajorImageVersion
139	.short	0				// MinorImageVersion
140	.short	0				// MajorSubsystemVersion
141	.short	0				// MinorSubsystemVersion
142	.long	0				// Win32VersionValue
143
144	.long	_end - _head			// SizeOfImage
145
146	// Everything before the kernel image is considered part of the header
147	.long	__efistub_stext_offset		// SizeOfHeaders
148	.long	0				// CheckSum
149	.short	0xa				// Subsystem (EFI application)
150	.short	0				// DllCharacteristics
151	.quad	0				// SizeOfStackReserve
152	.quad	0				// SizeOfStackCommit
153	.quad	0				// SizeOfHeapReserve
154	.quad	0				// SizeOfHeapCommit
155	.long	0				// LoaderFlags
156	.long	0x6				// NumberOfRvaAndSizes
157
158	.quad	0				// ExportTable
159	.quad	0				// ImportTable
160	.quad	0				// ResourceTable
161	.quad	0				// ExceptionTable
162	.quad	0				// CertificationTable
163	.quad	0				// BaseRelocationTable
164
165	// Section table
166section_table:
167
168	/*
169	 * The EFI application loader requires a relocation section
170	 * because EFI applications must be relocatable.  This is a
171	 * dummy section as far as we are concerned.
172	 */
173	.ascii	".reloc"
174	.byte	0
175	.byte	0			// end of 0 padding of section name
176	.long	0
177	.long	0
178	.long	0			// SizeOfRawData
179	.long	0			// PointerToRawData
180	.long	0			// PointerToRelocations
181	.long	0			// PointerToLineNumbers
182	.short	0			// NumberOfRelocations
183	.short	0			// NumberOfLineNumbers
184	.long	0x42100040		// Characteristics (section flags)
185
186
187	.ascii	".text"
188	.byte	0
189	.byte	0
190	.byte	0        		// end of 0 padding of section name
191	.long	_end - stext		// VirtualSize
192	.long	__efistub_stext_offset	// VirtualAddress
193	.long	_edata - stext		// SizeOfRawData
194	.long	__efistub_stext_offset	// PointerToRawData
195
196	.long	0		// PointerToRelocations (0 for executables)
197	.long	0		// PointerToLineNumbers (0 for executables)
198	.short	0		// NumberOfRelocations  (0 for executables)
199	.short	0		// NumberOfLineNumbers  (0 for executables)
200	.long	0xe0500020	// Characteristics (section flags)
201
202	/*
203	 * EFI will load stext onwards at the 4k section alignment
204	 * described in the PE/COFF header. To ensure that instruction
205	 * sequences using an adrp and a :lo12: immediate will function
206	 * correctly at this alignment, we must ensure that stext is
207	 * placed at a 4k boundary in the Image to begin with.
208	 */
209	.align 12
210#endif
211
212ENTRY(stext)
213	bl	preserve_boot_args
214	bl	el2_setup			// Drop to EL1, w20=cpu_boot_mode
215	mov	x23, xzr			// KASLR offset, defaults to 0
216	adrp	x24, __PHYS_OFFSET
217	bl	set_cpu_boot_mode_flag
218	bl	__create_page_tables		// x25=TTBR0, x26=TTBR1
219	/*
220	 * The following calls CPU setup code, see arch/arm64/mm/proc.S for
221	 * details.
222	 * On return, the CPU will be ready for the MMU to be turned on and
223	 * the TCR will have been set.
224	 */
225	ldr	x27, 0f				// address to jump to after
226						// MMU has been enabled
227	adr_l	lr, __enable_mmu		// return (PIC) address
228	b	__cpu_setup			// initialise processor
229ENDPROC(stext)
230	.align	3
2310:	.quad	__mmap_switched - (_head - TEXT_OFFSET) + KIMAGE_VADDR
232
233/*
234 * Preserve the arguments passed by the bootloader in x0 .. x3
235 */
236preserve_boot_args:
237	mov	x21, x0				// x21=FDT
238
239	adr_l	x0, boot_args			// record the contents of
240	stp	x21, x1, [x0]			// x0 .. x3 at kernel entry
241	stp	x2, x3, [x0, #16]
242
243	dmb	sy				// needed before dc ivac with
244						// MMU off
245
246	add	x1, x0, #0x20			// 4 x 8 bytes
247	b	__inval_cache_range		// tail call
248ENDPROC(preserve_boot_args)
249
250/*
251 * Macro to create a table entry to the next page.
252 *
253 *	tbl:	page table address
254 *	virt:	virtual address
255 *	shift:	#imm page table shift
256 *	ptrs:	#imm pointers per table page
257 *
258 * Preserves:	virt
259 * Corrupts:	tmp1, tmp2
260 * Returns:	tbl -> next level table page address
261 */
262	.macro	create_table_entry, tbl, virt, shift, ptrs, tmp1, tmp2
263	lsr	\tmp1, \virt, #\shift
264	and	\tmp1, \tmp1, #\ptrs - 1	// table index
265	add	\tmp2, \tbl, #PAGE_SIZE
266	orr	\tmp2, \tmp2, #PMD_TYPE_TABLE	// address of next table and entry type
267	str	\tmp2, [\tbl, \tmp1, lsl #3]
268	add	\tbl, \tbl, #PAGE_SIZE		// next level table page
269	.endm
270
271/*
272 * Macro to populate the PGD (and possibily PUD) for the corresponding
273 * block entry in the next level (tbl) for the given virtual address.
274 *
275 * Preserves:	tbl, next, virt
276 * Corrupts:	tmp1, tmp2
277 */
278	.macro	create_pgd_entry, tbl, virt, tmp1, tmp2
279	create_table_entry \tbl, \virt, PGDIR_SHIFT, PTRS_PER_PGD, \tmp1, \tmp2
280#if SWAPPER_PGTABLE_LEVELS > 3
281	create_table_entry \tbl, \virt, PUD_SHIFT, PTRS_PER_PUD, \tmp1, \tmp2
282#endif
283#if SWAPPER_PGTABLE_LEVELS > 2
284	create_table_entry \tbl, \virt, SWAPPER_TABLE_SHIFT, PTRS_PER_PTE, \tmp1, \tmp2
285#endif
286	.endm
287
288/*
289 * Macro to populate block entries in the page table for the start..end
290 * virtual range (inclusive).
291 *
292 * Preserves:	tbl, flags
293 * Corrupts:	phys, start, end, pstate
294 */
295	.macro	create_block_map, tbl, flags, phys, start, end
296	lsr	\phys, \phys, #SWAPPER_BLOCK_SHIFT
297	lsr	\start, \start, #SWAPPER_BLOCK_SHIFT
298	and	\start, \start, #PTRS_PER_PTE - 1	// table index
299	orr	\phys, \flags, \phys, lsl #SWAPPER_BLOCK_SHIFT	// table entry
300	lsr	\end, \end, #SWAPPER_BLOCK_SHIFT
301	and	\end, \end, #PTRS_PER_PTE - 1		// table end index
3029999:	str	\phys, [\tbl, \start, lsl #3]		// store the entry
303	add	\start, \start, #1			// next entry
304	add	\phys, \phys, #SWAPPER_BLOCK_SIZE		// next block
305	cmp	\start, \end
306	b.ls	9999b
307	.endm
308
309/*
310 * Setup the initial page tables. We only setup the barest amount which is
311 * required to get the kernel running. The following sections are required:
312 *   - identity mapping to enable the MMU (low address, TTBR0)
313 *   - first few MB of the kernel linear mapping to jump to once the MMU has
314 *     been enabled
315 */
316__create_page_tables:
317	adrp	x25, idmap_pg_dir
318	adrp	x26, swapper_pg_dir
319	mov	x28, lr
320
321	/*
322	 * Invalidate the idmap and swapper page tables to avoid potential
323	 * dirty cache lines being evicted.
324	 */
325	mov	x0, x25
326	add	x1, x26, #SWAPPER_DIR_SIZE
327	bl	__inval_cache_range
328
329	/*
330	 * Clear the idmap and swapper page tables.
331	 */
332	mov	x0, x25
333	add	x6, x26, #SWAPPER_DIR_SIZE
3341:	stp	xzr, xzr, [x0], #16
335	stp	xzr, xzr, [x0], #16
336	stp	xzr, xzr, [x0], #16
337	stp	xzr, xzr, [x0], #16
338	cmp	x0, x6
339	b.lo	1b
340
341	ldr	x7, =SWAPPER_MM_MMUFLAGS
342
343	/*
344	 * Create the identity mapping.
345	 */
346	mov	x0, x25				// idmap_pg_dir
347	adrp	x3, __idmap_text_start		// __pa(__idmap_text_start)
348
349#ifndef CONFIG_ARM64_VA_BITS_48
350#define EXTRA_SHIFT	(PGDIR_SHIFT + PAGE_SHIFT - 3)
351#define EXTRA_PTRS	(1 << (48 - EXTRA_SHIFT))
352
353	/*
354	 * If VA_BITS < 48, it may be too small to allow for an ID mapping to be
355	 * created that covers system RAM if that is located sufficiently high
356	 * in the physical address space. So for the ID map, use an extended
357	 * virtual range in that case, by configuring an additional translation
358	 * level.
359	 * First, we have to verify our assumption that the current value of
360	 * VA_BITS was chosen such that all translation levels are fully
361	 * utilised, and that lowering T0SZ will always result in an additional
362	 * translation level to be configured.
363	 */
364#if VA_BITS != EXTRA_SHIFT
365#error "Mismatch between VA_BITS and page size/number of translation levels"
366#endif
367
368	/*
369	 * Calculate the maximum allowed value for TCR_EL1.T0SZ so that the
370	 * entire ID map region can be mapped. As T0SZ == (64 - #bits used),
371	 * this number conveniently equals the number of leading zeroes in
372	 * the physical address of __idmap_text_end.
373	 */
374	adrp	x5, __idmap_text_end
375	clz	x5, x5
376	cmp	x5, TCR_T0SZ(VA_BITS)	// default T0SZ small enough?
377	b.ge	1f			// .. then skip additional level
378
379	adr_l	x6, idmap_t0sz
380	str	x5, [x6]
381	dmb	sy
382	dc	ivac, x6		// Invalidate potentially stale cache line
383
384	create_table_entry x0, x3, EXTRA_SHIFT, EXTRA_PTRS, x5, x6
3851:
386#endif
387
388	create_pgd_entry x0, x3, x5, x6
389	mov	x5, x3				// __pa(__idmap_text_start)
390	adr_l	x6, __idmap_text_end		// __pa(__idmap_text_end)
391	create_block_map x0, x7, x3, x5, x6
392
393	/*
394	 * Map the kernel image (starting with PHYS_OFFSET).
395	 */
396	mov	x0, x26				// swapper_pg_dir
397	ldr	x5, =KIMAGE_VADDR
398	add	x5, x5, x23			// add KASLR displacement
399	create_pgd_entry x0, x5, x3, x6
400	ldr	w6, kernel_img_size
401	add	x6, x6, x5
402	mov	x3, x24				// phys offset
403	create_block_map x0, x7, x3, x5, x6
404
405	/*
406	 * Since the page tables have been populated with non-cacheable
407	 * accesses (MMU disabled), invalidate the idmap and swapper page
408	 * tables again to remove any speculatively loaded cache lines.
409	 */
410	mov	x0, x25
411	add	x1, x26, #SWAPPER_DIR_SIZE
412	dmb	sy
413	bl	__inval_cache_range
414
415	ret	x28
416ENDPROC(__create_page_tables)
417
418kernel_img_size:
419	.long	_end - (_head - TEXT_OFFSET)
420	.ltorg
421
422/*
423 * The following fragment of code is executed with the MMU enabled.
424 */
425	.set	initial_sp, init_thread_union + THREAD_START_SP
426__mmap_switched:
427	mov	x28, lr				// preserve LR
428	adr_l	x8, vectors			// load VBAR_EL1 with virtual
429	msr	vbar_el1, x8			// vector table address
430	isb
431
432	// Clear BSS
433	adr_l	x0, __bss_start
434	mov	x1, xzr
435	adr_l	x2, __bss_stop
436	sub	x2, x2, x0
437	bl	__pi_memset
438	dsb	ishst				// Make zero page visible to PTW
439
440#ifdef CONFIG_RELOCATABLE
441
442	/*
443	 * Iterate over each entry in the relocation table, and apply the
444	 * relocations in place.
445	 */
446	adr_l	x8, __dynsym_start		// start of symbol table
447	adr_l	x9, __reloc_start		// start of reloc table
448	adr_l	x10, __reloc_end		// end of reloc table
449
4500:	cmp	x9, x10
451	b.hs	2f
452	ldp	x11, x12, [x9], #24
453	ldr	x13, [x9, #-8]
454	cmp	w12, #R_AARCH64_RELATIVE
455	b.ne	1f
456	add	x13, x13, x23			// relocate
457	str	x13, [x11, x23]
458	b	0b
459
4601:	cmp	w12, #R_AARCH64_ABS64
461	b.ne	0b
462	add	x12, x12, x12, lsl #1		// symtab offset: 24x top word
463	add	x12, x8, x12, lsr #(32 - 3)	// ... shifted into bottom word
464	ldrsh	w14, [x12, #6]			// Elf64_Sym::st_shndx
465	ldr	x15, [x12, #8]			// Elf64_Sym::st_value
466	cmp	w14, #-0xf			// SHN_ABS (0xfff1) ?
467	add	x14, x15, x23			// relocate
468	csel	x15, x14, x15, ne
469	add	x15, x13, x15
470	str	x15, [x11, x23]
471	b	0b
472
4732:	adr_l	x8, kimage_vaddr		// make relocated kimage_vaddr
474	dc	cvac, x8			// value visible to secondaries
475	dsb	sy				// with MMU off
476#endif
477
478	adr_l	sp, initial_sp, x4
479	mov	x4, sp
480	and	x4, x4, #~(THREAD_SIZE - 1)
481	msr	sp_el0, x4			// Save thread_info
482	str_l	x21, __fdt_pointer, x5		// Save FDT pointer
483
484	ldr_l	x4, kimage_vaddr		// Save the offset between
485	sub	x4, x4, x24			// the kernel virtual and
486	str_l	x4, kimage_voffset, x5		// physical mappings
487
488	mov	x29, #0
489#ifdef CONFIG_KASAN
490	bl	kasan_early_init
491#endif
492#ifdef CONFIG_RANDOMIZE_BASE
493	cbnz	x23, 0f				// already running randomized?
494	mov	x0, x21				// pass FDT address in x0
495	bl	kaslr_early_init		// parse FDT for KASLR options
496	cbz	x0, 0f				// KASLR disabled? just proceed
497	mov	x23, x0				// record KASLR offset
498	ret	x28				// we must enable KASLR, return
499						// to __enable_mmu()
5000:
501#endif
502	b	start_kernel
503ENDPROC(__mmap_switched)
504
505/*
506 * end early head section, begin head code that is also used for
507 * hotplug and needs to have the same protections as the text region
508 */
509	.section ".text","ax"
510
511ENTRY(kimage_vaddr)
512	.quad		_text - TEXT_OFFSET
513
514/*
515 * If we're fortunate enough to boot at EL2, ensure that the world is
516 * sane before dropping to EL1.
517 *
518 * Returns either BOOT_CPU_MODE_EL1 or BOOT_CPU_MODE_EL2 in x20 if
519 * booted in EL1 or EL2 respectively.
520 */
521ENTRY(el2_setup)
522	mrs	x0, CurrentEL
523	cmp	x0, #CurrentEL_EL2
524	b.ne	1f
525	mrs	x0, sctlr_el2
526CPU_BE(	orr	x0, x0, #(1 << 25)	)	// Set the EE bit for EL2
527CPU_LE(	bic	x0, x0, #(1 << 25)	)	// Clear the EE bit for EL2
528	msr	sctlr_el2, x0
529	b	2f
5301:	mrs	x0, sctlr_el1
531CPU_BE(	orr	x0, x0, #(3 << 24)	)	// Set the EE and E0E bits for EL1
532CPU_LE(	bic	x0, x0, #(3 << 24)	)	// Clear the EE and E0E bits for EL1
533	msr	sctlr_el1, x0
534	mov	w20, #BOOT_CPU_MODE_EL1		// This cpu booted in EL1
535	isb
536	ret
537
5382:
539#ifdef CONFIG_ARM64_VHE
540	/*
541	 * Check for VHE being present. For the rest of the EL2 setup,
542	 * x2 being non-zero indicates that we do have VHE, and that the
543	 * kernel is intended to run at EL2.
544	 */
545	mrs	x2, id_aa64mmfr1_el1
546	ubfx	x2, x2, #8, #4
547#else
548	mov	x2, xzr
549#endif
550
551	/* Hyp configuration. */
552	mov	x0, #HCR_RW			// 64-bit EL1
553	cbz	x2, set_hcr
554	orr	x0, x0, #HCR_TGE		// Enable Host Extensions
555	orr	x0, x0, #HCR_E2H
556set_hcr:
557	msr	hcr_el2, x0
558	isb
559
560	/* Generic timers. */
561	mrs	x0, cnthctl_el2
562	orr	x0, x0, #3			// Enable EL1 physical timers
563	msr	cnthctl_el2, x0
564	msr	cntvoff_el2, xzr		// Clear virtual offset
565
566#ifdef CONFIG_ARM_GIC_V3
567	/* GICv3 system register access */
568	mrs	x0, id_aa64pfr0_el1
569	ubfx	x0, x0, #24, #4
570	cmp	x0, #1
571	b.ne	3f
572
573	mrs_s	x0, ICC_SRE_EL2
574	orr	x0, x0, #ICC_SRE_EL2_SRE	// Set ICC_SRE_EL2.SRE==1
575	orr	x0, x0, #ICC_SRE_EL2_ENABLE	// Set ICC_SRE_EL2.Enable==1
576	msr_s	ICC_SRE_EL2, x0
577	isb					// Make sure SRE is now set
578	mrs_s	x0, ICC_SRE_EL2			// Read SRE back,
579	tbz	x0, #0, 3f			// and check that it sticks
580	msr_s	ICH_HCR_EL2, xzr		// Reset ICC_HCR_EL2 to defaults
581
5823:
583#endif
584
585	/* Populate ID registers. */
586	mrs	x0, midr_el1
587	mrs	x1, mpidr_el1
588	msr	vpidr_el2, x0
589	msr	vmpidr_el2, x1
590
591	/*
592	 * When VHE is not in use, early init of EL2 and EL1 needs to be
593	 * done here.
594	 * When VHE _is_ in use, EL1 will not be used in the host and
595	 * requires no configuration, and all non-hyp-specific EL2 setup
596	 * will be done via the _EL1 system register aliases in __cpu_setup.
597	 */
598	cbnz	x2, 1f
599
600	/* sctlr_el1 */
601	mov	x0, #0x0800			// Set/clear RES{1,0} bits
602CPU_BE(	movk	x0, #0x33d0, lsl #16	)	// Set EE and E0E on BE systems
603CPU_LE(	movk	x0, #0x30d0, lsl #16	)	// Clear EE and E0E on LE systems
604	msr	sctlr_el1, x0
605
606	/* Coprocessor traps. */
607	mov	x0, #0x33ff
608	msr	cptr_el2, x0			// Disable copro. traps to EL2
6091:
610
611#ifdef CONFIG_COMPAT
612	msr	hstr_el2, xzr			// Disable CP15 traps to EL2
613#endif
614
615	/* EL2 debug */
616	mrs	x0, id_aa64dfr0_el1		// Check ID_AA64DFR0_EL1 PMUVer
617	sbfx	x0, x0, #8, #4
618	cmp	x0, #1
619	b.lt	4f				// Skip if no PMU present
620	mrs	x0, pmcr_el0			// Disable debug access traps
621	ubfx	x0, x0, #11, #5			// to EL2 and allow access to
622	msr	mdcr_el2, x0			// all PMU counters from EL1
6234:
624
625	/* Stage-2 translation */
626	msr	vttbr_el2, xzr
627
628	cbz	x2, install_el2_stub
629
630	mov	w20, #BOOT_CPU_MODE_EL2		// This CPU booted in EL2
631	isb
632	ret
633
634install_el2_stub:
635	/* Hypervisor stub */
636	adrp	x0, __hyp_stub_vectors
637	add	x0, x0, #:lo12:__hyp_stub_vectors
638	msr	vbar_el2, x0
639
640	/* spsr */
641	mov	x0, #(PSR_F_BIT | PSR_I_BIT | PSR_A_BIT | PSR_D_BIT |\
642		      PSR_MODE_EL1h)
643	msr	spsr_el2, x0
644	msr	elr_el2, lr
645	mov	w20, #BOOT_CPU_MODE_EL2		// This CPU booted in EL2
646	eret
647ENDPROC(el2_setup)
648
649/*
650 * Sets the __boot_cpu_mode flag depending on the CPU boot mode passed
651 * in x20. See arch/arm64/include/asm/virt.h for more info.
652 */
653ENTRY(set_cpu_boot_mode_flag)
654	adr_l	x1, __boot_cpu_mode
655	cmp	w20, #BOOT_CPU_MODE_EL2
656	b.ne	1f
657	add	x1, x1, #4
6581:	str	w20, [x1]			// This CPU has booted in EL1
659	dmb	sy
660	dc	ivac, x1			// Invalidate potentially stale cache line
661	ret
662ENDPROC(set_cpu_boot_mode_flag)
663
664/*
665 * We need to find out the CPU boot mode long after boot, so we need to
666 * store it in a writable variable.
667 *
668 * This is not in .bss, because we set it sufficiently early that the boot-time
669 * zeroing of .bss would clobber it.
670 */
671	.pushsection	.data..cacheline_aligned
672	.align	L1_CACHE_SHIFT
673ENTRY(__boot_cpu_mode)
674	.long	BOOT_CPU_MODE_EL2
675	.long	BOOT_CPU_MODE_EL1
676	.popsection
677
678	/*
679	 * This provides a "holding pen" for platforms to hold all secondary
680	 * cores are held until we're ready for them to initialise.
681	 */
682ENTRY(secondary_holding_pen)
683	bl	el2_setup			// Drop to EL1, w20=cpu_boot_mode
684	bl	set_cpu_boot_mode_flag
685	mrs	x0, mpidr_el1
686	ldr     x1, =MPIDR_HWID_BITMASK
687	and	x0, x0, x1
688	adr_l	x3, secondary_holding_pen_release
689pen:	ldr	x4, [x3]
690	cmp	x4, x0
691	b.eq	secondary_startup
692	wfe
693	b	pen
694ENDPROC(secondary_holding_pen)
695
696	/*
697	 * Secondary entry point that jumps straight into the kernel. Only to
698	 * be used where CPUs are brought online dynamically by the kernel.
699	 */
700ENTRY(secondary_entry)
701	bl	el2_setup			// Drop to EL1
702	bl	set_cpu_boot_mode_flag
703	b	secondary_startup
704ENDPROC(secondary_entry)
705
706ENTRY(secondary_startup)
707	/*
708	 * Common entry point for secondary CPUs.
709	 */
710	adrp	x25, idmap_pg_dir
711	adrp	x26, swapper_pg_dir
712	bl	__cpu_setup			// initialise processor
713
714	ldr	x8, kimage_vaddr
715	ldr	w9, 0f
716	sub	x27, x8, w9, sxtw		// address to jump to after enabling the MMU
717	b	__enable_mmu
718ENDPROC(secondary_startup)
7190:	.long	(_text - TEXT_OFFSET) - __secondary_switched
720
721ENTRY(__secondary_switched)
722	adr_l	x5, vectors
723	msr	vbar_el1, x5
724	isb
725
726	adr_l	x0, secondary_data
727	ldr	x0, [x0, #CPU_BOOT_STACK]	// get secondary_data.stack
728	mov	sp, x0
729	and	x0, x0, #~(THREAD_SIZE - 1)
730	msr	sp_el0, x0			// save thread_info
731	mov	x29, #0
732	b	secondary_start_kernel
733ENDPROC(__secondary_switched)
734
735/*
736 * The booting CPU updates the failed status @__early_cpu_boot_status,
737 * with MMU turned off.
738 *
739 * update_early_cpu_boot_status tmp, status
740 *  - Corrupts tmp1, tmp2
741 *  - Writes 'status' to __early_cpu_boot_status and makes sure
742 *    it is committed to memory.
743 */
744
745	.macro	update_early_cpu_boot_status status, tmp1, tmp2
746	mov	\tmp2, #\status
747	adr_l	\tmp1, __early_cpu_boot_status
748	str	\tmp2, [\tmp1]
749	dmb	sy
750	dc	ivac, \tmp1			// Invalidate potentially stale cache line
751	.endm
752
753	.pushsection	.data..cacheline_aligned
754	.align	L1_CACHE_SHIFT
755ENTRY(__early_cpu_boot_status)
756	.long 	0
757	.popsection
758
759/*
760 * Enable the MMU.
761 *
762 *  x0  = SCTLR_EL1 value for turning on the MMU.
763 *  x27 = *virtual* address to jump to upon completion
764 *
765 * Other registers depend on the function called upon completion.
766 *
767 * Checks if the selected granule size is supported by the CPU.
768 * If it isn't, park the CPU
769 */
770	.section	".idmap.text", "ax"
771__enable_mmu:
772	mrs	x22, sctlr_el1			// preserve old SCTLR_EL1 value
773	mrs	x1, ID_AA64MMFR0_EL1
774	ubfx	x2, x1, #ID_AA64MMFR0_TGRAN_SHIFT, 4
775	cmp	x2, #ID_AA64MMFR0_TGRAN_SUPPORTED
776	b.ne	__no_granule_support
777	update_early_cpu_boot_status 0, x1, x2
778	msr	ttbr0_el1, x25			// load TTBR0
779	msr	ttbr1_el1, x26			// load TTBR1
780	isb
781	msr	sctlr_el1, x0
782	isb
783	/*
784	 * Invalidate the local I-cache so that any instructions fetched
785	 * speculatively from the PoC are discarded, since they may have
786	 * been dynamically patched at the PoU.
787	 */
788	ic	iallu
789	dsb	nsh
790	isb
791#ifdef CONFIG_RANDOMIZE_BASE
792	mov	x19, x0				// preserve new SCTLR_EL1 value
793	blr	x27
794
795	/*
796	 * If we return here, we have a KASLR displacement in x23 which we need
797	 * to take into account by discarding the current kernel mapping and
798	 * creating a new one.
799	 */
800	msr	sctlr_el1, x22			// disable the MMU
801	isb
802	bl	__create_page_tables		// recreate kernel mapping
803
804	msr	sctlr_el1, x19			// re-enable the MMU
805	isb
806	ic	iallu				// flush instructions fetched
807	dsb	nsh				// via old mapping
808	isb
809	add	x27, x27, x23			// relocated __mmap_switched
810#endif
811	br	x27
812ENDPROC(__enable_mmu)
813
814__no_granule_support:
815	/* Indicate that this CPU can't boot and is stuck in the kernel */
816	update_early_cpu_boot_status CPU_STUCK_IN_KERNEL, x1, x2
8171:
818	wfe
819	wfi
820	b 1b
821ENDPROC(__no_granule_support)
822