xref: /openbmc/linux/arch/arm64/kernel/head.S (revision a77e393c)
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/boot.h>
29#include <asm/ptrace.h>
30#include <asm/asm-offsets.h>
31#include <asm/cache.h>
32#include <asm/cputype.h>
33#include <asm/elf.h>
34#include <asm/kernel-pgtable.h>
35#include <asm/kvm_arm.h>
36#include <asm/memory.h>
37#include <asm/pgtable-hwdef.h>
38#include <asm/pgtable.h>
39#include <asm/page.h>
40#include <asm/smp.h>
41#include <asm/sysreg.h>
42#include <asm/thread_info.h>
43#include <asm/virt.h>
44
45#define __PHYS_OFFSET	(KERNEL_START - TEXT_OFFSET)
46
47#if (TEXT_OFFSET & 0xfff) != 0
48#error TEXT_OFFSET must be at least 4KB aligned
49#elif (PAGE_OFFSET & 0x1fffff) != 0
50#error PAGE_OFFSET must be at least 2MB aligned
51#elif TEXT_OFFSET > 0x1fffff
52#error TEXT_OFFSET must be less than 2MB
53#endif
54
55/*
56 * Kernel startup entry point.
57 * ---------------------------
58 *
59 * The requirements are:
60 *   MMU = off, D-cache = off, I-cache = on or off,
61 *   x0 = physical address to the FDT blob.
62 *
63 * This code is mostly position independent so you call this at
64 * __pa(PAGE_OFFSET + TEXT_OFFSET).
65 *
66 * Note that the callee-saved registers are used for storing variables
67 * that are useful before the MMU is enabled. The allocations are described
68 * in the entry routines.
69 */
70	__HEAD
71_head:
72	/*
73	 * DO NOT MODIFY. Image header expected by Linux boot-loaders.
74	 */
75#ifdef CONFIG_EFI
76	/*
77	 * This add instruction has no meaningful effect except that
78	 * its opcode forms the magic "MZ" signature required by UEFI.
79	 */
80	add	x13, x18, #0x16
81	b	stext
82#else
83	b	stext				// branch to kernel start, magic
84	.long	0				// reserved
85#endif
86	le64sym	_kernel_offset_le		// Image load offset from start of RAM, little-endian
87	le64sym	_kernel_size_le			// Effective size of kernel image, little-endian
88	le64sym	_kernel_flags_le		// Informative flags, little-endian
89	.quad	0				// reserved
90	.quad	0				// reserved
91	.quad	0				// reserved
92	.byte	0x41				// Magic number, "ARM\x64"
93	.byte	0x52
94	.byte	0x4d
95	.byte	0x64
96#ifdef CONFIG_EFI
97	.long	pe_header - _head		// Offset to the PE header.
98#else
99	.word	0				// reserved
100#endif
101
102#ifdef CONFIG_EFI
103	.align 3
104pe_header:
105	.ascii	"PE"
106	.short 	0
107coff_header:
108	.short	0xaa64				// AArch64
109	.short	2				// nr_sections
110	.long	0 				// TimeDateStamp
111	.long	0				// PointerToSymbolTable
112	.long	1				// NumberOfSymbols
113	.short	section_table - optional_header	// SizeOfOptionalHeader
114	.short	0x206				// Characteristics.
115						// IMAGE_FILE_DEBUG_STRIPPED |
116						// IMAGE_FILE_EXECUTABLE_IMAGE |
117						// IMAGE_FILE_LINE_NUMS_STRIPPED
118optional_header:
119	.short	0x20b				// PE32+ format
120	.byte	0x02				// MajorLinkerVersion
121	.byte	0x14				// MinorLinkerVersion
122	.long	_end - efi_header_end		// SizeOfCode
123	.long	0				// SizeOfInitializedData
124	.long	0				// SizeOfUninitializedData
125	.long	__efistub_entry - _head		// AddressOfEntryPoint
126	.long	efi_header_end - _head		// BaseOfCode
127
128extra_header_fields:
129	.quad	0				// ImageBase
130	.long	0x1000				// SectionAlignment
131	.long	PECOFF_FILE_ALIGNMENT		// FileAlignment
132	.short	0				// MajorOperatingSystemVersion
133	.short	0				// MinorOperatingSystemVersion
134	.short	0				// MajorImageVersion
135	.short	0				// MinorImageVersion
136	.short	0				// MajorSubsystemVersion
137	.short	0				// MinorSubsystemVersion
138	.long	0				// Win32VersionValue
139
140	.long	_end - _head			// SizeOfImage
141
142	// Everything before the kernel image is considered part of the header
143	.long	efi_header_end - _head		// SizeOfHeaders
144	.long	0				// CheckSum
145	.short	0xa				// Subsystem (EFI application)
146	.short	0				// DllCharacteristics
147	.quad	0				// SizeOfStackReserve
148	.quad	0				// SizeOfStackCommit
149	.quad	0				// SizeOfHeapReserve
150	.quad	0				// SizeOfHeapCommit
151	.long	0				// LoaderFlags
152	.long	0x6				// NumberOfRvaAndSizes
153
154	.quad	0				// ExportTable
155	.quad	0				// ImportTable
156	.quad	0				// ResourceTable
157	.quad	0				// ExceptionTable
158	.quad	0				// CertificationTable
159	.quad	0				// BaseRelocationTable
160
161	// Section table
162section_table:
163
164	/*
165	 * The EFI application loader requires a relocation section
166	 * because EFI applications must be relocatable.  This is a
167	 * dummy section as far as we are concerned.
168	 */
169	.ascii	".reloc"
170	.byte	0
171	.byte	0			// end of 0 padding of section name
172	.long	0
173	.long	0
174	.long	0			// SizeOfRawData
175	.long	0			// PointerToRawData
176	.long	0			// PointerToRelocations
177	.long	0			// PointerToLineNumbers
178	.short	0			// NumberOfRelocations
179	.short	0			// NumberOfLineNumbers
180	.long	0x42100040		// Characteristics (section flags)
181
182
183	.ascii	".text"
184	.byte	0
185	.byte	0
186	.byte	0        		// end of 0 padding of section name
187	.long	_end - efi_header_end	// VirtualSize
188	.long	efi_header_end - _head	// VirtualAddress
189	.long	_edata - efi_header_end	// SizeOfRawData
190	.long	efi_header_end - _head	// PointerToRawData
191
192	.long	0		// PointerToRelocations (0 for executables)
193	.long	0		// PointerToLineNumbers (0 for executables)
194	.short	0		// NumberOfRelocations  (0 for executables)
195	.short	0		// NumberOfLineNumbers  (0 for executables)
196	.long	0xe0500020	// Characteristics (section flags)
197
198	/*
199	 * EFI will load .text onwards at the 4k section alignment
200	 * described in the PE/COFF header. To ensure that instruction
201	 * sequences using an adrp and a :lo12: immediate will function
202	 * correctly at this alignment, we must ensure that .text is
203	 * placed at a 4k boundary in the Image to begin with.
204	 */
205	.align 12
206efi_header_end:
207#endif
208
209	__INIT
210
211	/*
212	 * The following callee saved general purpose registers are used on the
213	 * primary lowlevel boot path:
214	 *
215	 *  Register   Scope                      Purpose
216	 *  x21        stext() .. start_kernel()  FDT pointer passed at boot in x0
217	 *  x23        stext() .. start_kernel()  physical misalignment/KASLR offset
218	 *  x28        __create_page_tables()     callee preserved temp register
219	 *  x19/x20    __primary_switch()         callee preserved temp registers
220	 */
221ENTRY(stext)
222	bl	preserve_boot_args
223	bl	el2_setup			// Drop to EL1, w0=cpu_boot_mode
224	adrp	x23, __PHYS_OFFSET
225	and	x23, x23, MIN_KIMG_ALIGN - 1	// KASLR offset, defaults to 0
226	bl	set_cpu_boot_mode_flag
227	bl	__create_page_tables
228	/*
229	 * The following calls CPU setup code, see arch/arm64/mm/proc.S for
230	 * details.
231	 * On return, the CPU will be ready for the MMU to be turned on and
232	 * the TCR will have been set.
233	 */
234	bl	__cpu_setup			// initialise processor
235	b	__primary_switch
236ENDPROC(stext)
237
238/*
239 * Preserve the arguments passed by the bootloader in x0 .. x3
240 */
241preserve_boot_args:
242	mov	x21, x0				// x21=FDT
243
244	adr_l	x0, boot_args			// record the contents of
245	stp	x21, x1, [x0]			// x0 .. x3 at kernel entry
246	stp	x2, x3, [x0, #16]
247
248	dmb	sy				// needed before dc ivac with
249						// MMU off
250
251	add	x1, x0, #0x20			// 4 x 8 bytes
252	b	__inval_cache_range		// tail call
253ENDPROC(preserve_boot_args)
254
255/*
256 * Macro to create a table entry to the next page.
257 *
258 *	tbl:	page table address
259 *	virt:	virtual address
260 *	shift:	#imm page table shift
261 *	ptrs:	#imm pointers per table page
262 *
263 * Preserves:	virt
264 * Corrupts:	tmp1, tmp2
265 * Returns:	tbl -> next level table page address
266 */
267	.macro	create_table_entry, tbl, virt, shift, ptrs, tmp1, tmp2
268	lsr	\tmp1, \virt, #\shift
269	and	\tmp1, \tmp1, #\ptrs - 1	// table index
270	add	\tmp2, \tbl, #PAGE_SIZE
271	orr	\tmp2, \tmp2, #PMD_TYPE_TABLE	// address of next table and entry type
272	str	\tmp2, [\tbl, \tmp1, lsl #3]
273	add	\tbl, \tbl, #PAGE_SIZE		// next level table page
274	.endm
275
276/*
277 * Macro to populate the PGD (and possibily PUD) for the corresponding
278 * block entry in the next level (tbl) for the given virtual address.
279 *
280 * Preserves:	tbl, next, virt
281 * Corrupts:	tmp1, tmp2
282 */
283	.macro	create_pgd_entry, tbl, virt, tmp1, tmp2
284	create_table_entry \tbl, \virt, PGDIR_SHIFT, PTRS_PER_PGD, \tmp1, \tmp2
285#if SWAPPER_PGTABLE_LEVELS > 3
286	create_table_entry \tbl, \virt, PUD_SHIFT, PTRS_PER_PUD, \tmp1, \tmp2
287#endif
288#if SWAPPER_PGTABLE_LEVELS > 2
289	create_table_entry \tbl, \virt, SWAPPER_TABLE_SHIFT, PTRS_PER_PTE, \tmp1, \tmp2
290#endif
291	.endm
292
293/*
294 * Macro to populate block entries in the page table for the start..end
295 * virtual range (inclusive).
296 *
297 * Preserves:	tbl, flags
298 * Corrupts:	phys, start, end, pstate
299 */
300	.macro	create_block_map, tbl, flags, phys, start, end
301	lsr	\phys, \phys, #SWAPPER_BLOCK_SHIFT
302	lsr	\start, \start, #SWAPPER_BLOCK_SHIFT
303	and	\start, \start, #PTRS_PER_PTE - 1	// table index
304	orr	\phys, \flags, \phys, lsl #SWAPPER_BLOCK_SHIFT	// table entry
305	lsr	\end, \end, #SWAPPER_BLOCK_SHIFT
306	and	\end, \end, #PTRS_PER_PTE - 1		// table end index
3079999:	str	\phys, [\tbl, \start, lsl #3]		// store the entry
308	add	\start, \start, #1			// next entry
309	add	\phys, \phys, #SWAPPER_BLOCK_SIZE		// next block
310	cmp	\start, \end
311	b.ls	9999b
312	.endm
313
314/*
315 * Setup the initial page tables. We only setup the barest amount which is
316 * required to get the kernel running. The following sections are required:
317 *   - identity mapping to enable the MMU (low address, TTBR0)
318 *   - first few MB of the kernel linear mapping to jump to once the MMU has
319 *     been enabled
320 */
321__create_page_tables:
322	mov	x28, lr
323
324	/*
325	 * Invalidate the idmap and swapper page tables to avoid potential
326	 * dirty cache lines being evicted.
327	 */
328	adrp	x0, idmap_pg_dir
329	adrp	x1, swapper_pg_dir + SWAPPER_DIR_SIZE
330	bl	__inval_cache_range
331
332	/*
333	 * Clear the idmap and swapper page tables.
334	 */
335	adrp	x0, idmap_pg_dir
336	adrp	x6, swapper_pg_dir + SWAPPER_DIR_SIZE
3371:	stp	xzr, xzr, [x0], #16
338	stp	xzr, xzr, [x0], #16
339	stp	xzr, xzr, [x0], #16
340	stp	xzr, xzr, [x0], #16
341	cmp	x0, x6
342	b.lo	1b
343
344	mov	x7, SWAPPER_MM_MMUFLAGS
345
346	/*
347	 * Create the identity mapping.
348	 */
349	adrp	x0, idmap_pg_dir
350	adrp	x3, __idmap_text_start		// __pa(__idmap_text_start)
351
352#ifndef CONFIG_ARM64_VA_BITS_48
353#define EXTRA_SHIFT	(PGDIR_SHIFT + PAGE_SHIFT - 3)
354#define EXTRA_PTRS	(1 << (48 - EXTRA_SHIFT))
355
356	/*
357	 * If VA_BITS < 48, it may be too small to allow for an ID mapping to be
358	 * created that covers system RAM if that is located sufficiently high
359	 * in the physical address space. So for the ID map, use an extended
360	 * virtual range in that case, by configuring an additional translation
361	 * level.
362	 * First, we have to verify our assumption that the current value of
363	 * VA_BITS was chosen such that all translation levels are fully
364	 * utilised, and that lowering T0SZ will always result in an additional
365	 * translation level to be configured.
366	 */
367#if VA_BITS != EXTRA_SHIFT
368#error "Mismatch between VA_BITS and page size/number of translation levels"
369#endif
370
371	/*
372	 * Calculate the maximum allowed value for TCR_EL1.T0SZ so that the
373	 * entire ID map region can be mapped. As T0SZ == (64 - #bits used),
374	 * this number conveniently equals the number of leading zeroes in
375	 * the physical address of __idmap_text_end.
376	 */
377	adrp	x5, __idmap_text_end
378	clz	x5, x5
379	cmp	x5, TCR_T0SZ(VA_BITS)	// default T0SZ small enough?
380	b.ge	1f			// .. then skip additional level
381
382	adr_l	x6, idmap_t0sz
383	str	x5, [x6]
384	dmb	sy
385	dc	ivac, x6		// Invalidate potentially stale cache line
386
387	create_table_entry x0, x3, EXTRA_SHIFT, EXTRA_PTRS, x5, x6
3881:
389#endif
390
391	create_pgd_entry x0, x3, x5, x6
392	mov	x5, x3				// __pa(__idmap_text_start)
393	adr_l	x6, __idmap_text_end		// __pa(__idmap_text_end)
394	create_block_map x0, x7, x3, x5, x6
395
396	/*
397	 * Map the kernel image (starting with PHYS_OFFSET).
398	 */
399	adrp	x0, swapper_pg_dir
400	mov_q	x5, KIMAGE_VADDR + TEXT_OFFSET	// compile time __va(_text)
401	add	x5, x5, x23			// add KASLR displacement
402	create_pgd_entry x0, x5, x3, x6
403	adrp	x6, _end			// runtime __pa(_end)
404	adrp	x3, _text			// runtime __pa(_text)
405	sub	x6, x6, x3			// _end - _text
406	add	x6, x6, x5			// runtime __va(_end)
407	create_block_map x0, x7, x3, x5, x6
408
409	/*
410	 * Since the page tables have been populated with non-cacheable
411	 * accesses (MMU disabled), invalidate the idmap and swapper page
412	 * tables again to remove any speculatively loaded cache lines.
413	 */
414	adrp	x0, idmap_pg_dir
415	adrp	x1, swapper_pg_dir + SWAPPER_DIR_SIZE
416	dmb	sy
417	bl	__inval_cache_range
418
419	ret	x28
420ENDPROC(__create_page_tables)
421	.ltorg
422
423/*
424 * The following fragment of code is executed with the MMU enabled.
425 *
426 *   x0 = __PHYS_OFFSET
427 */
428__primary_switched:
429	adrp	x4, init_thread_union
430	add	sp, x4, #THREAD_SIZE
431	msr	sp_el0, x4			// Save thread_info
432
433	adr_l	x8, vectors			// load VBAR_EL1 with virtual
434	msr	vbar_el1, x8			// vector table address
435	isb
436
437	stp	xzr, x30, [sp, #-16]!
438	mov	x29, sp
439
440	str_l	x21, __fdt_pointer, x5		// Save FDT pointer
441
442	ldr_l	x4, kimage_vaddr		// Save the offset between
443	sub	x4, x4, x0			// the kernel virtual and
444	str_l	x4, kimage_voffset, x5		// physical mappings
445
446	// Clear BSS
447	adr_l	x0, __bss_start
448	mov	x1, xzr
449	adr_l	x2, __bss_stop
450	sub	x2, x2, x0
451	bl	__pi_memset
452	dsb	ishst				// Make zero page visible to PTW
453
454#ifdef CONFIG_KASAN
455	bl	kasan_early_init
456#endif
457#ifdef CONFIG_RANDOMIZE_BASE
458	tst	x23, ~(MIN_KIMG_ALIGN - 1)	// already running randomized?
459	b.ne	0f
460	mov	x0, x21				// pass FDT address in x0
461	mov	x1, x23				// pass modulo offset in x1
462	bl	kaslr_early_init		// parse FDT for KASLR options
463	cbz	x0, 0f				// KASLR disabled? just proceed
464	orr	x23, x23, x0			// record KASLR offset
465	ldp	x29, x30, [sp], #16		// we must enable KASLR, return
466	ret					// to __primary_switch()
4670:
468#endif
469	b	start_kernel
470ENDPROC(__primary_switched)
471
472/*
473 * end early head section, begin head code that is also used for
474 * hotplug and needs to have the same protections as the text region
475 */
476	.section ".idmap.text","ax"
477
478ENTRY(kimage_vaddr)
479	.quad		_text - TEXT_OFFSET
480
481/*
482 * If we're fortunate enough to boot at EL2, ensure that the world is
483 * sane before dropping to EL1.
484 *
485 * Returns either BOOT_CPU_MODE_EL1 or BOOT_CPU_MODE_EL2 in x20 if
486 * booted in EL1 or EL2 respectively.
487 */
488ENTRY(el2_setup)
489	mrs	x0, CurrentEL
490	cmp	x0, #CurrentEL_EL2
491	b.ne	1f
492	mrs	x0, sctlr_el2
493CPU_BE(	orr	x0, x0, #(1 << 25)	)	// Set the EE bit for EL2
494CPU_LE(	bic	x0, x0, #(1 << 25)	)	// Clear the EE bit for EL2
495	msr	sctlr_el2, x0
496	b	2f
4971:	mrs	x0, sctlr_el1
498CPU_BE(	orr	x0, x0, #(3 << 24)	)	// Set the EE and E0E bits for EL1
499CPU_LE(	bic	x0, x0, #(3 << 24)	)	// Clear the EE and E0E bits for EL1
500	msr	sctlr_el1, x0
501	mov	w0, #BOOT_CPU_MODE_EL1		// This cpu booted in EL1
502	isb
503	ret
504
5052:
506#ifdef CONFIG_ARM64_VHE
507	/*
508	 * Check for VHE being present. For the rest of the EL2 setup,
509	 * x2 being non-zero indicates that we do have VHE, and that the
510	 * kernel is intended to run at EL2.
511	 */
512	mrs	x2, id_aa64mmfr1_el1
513	ubfx	x2, x2, #8, #4
514#else
515	mov	x2, xzr
516#endif
517
518	/* Hyp configuration. */
519	mov	x0, #HCR_RW			// 64-bit EL1
520	cbz	x2, set_hcr
521	orr	x0, x0, #HCR_TGE		// Enable Host Extensions
522	orr	x0, x0, #HCR_E2H
523set_hcr:
524	msr	hcr_el2, x0
525	isb
526
527	/* Generic timers. */
528	mrs	x0, cnthctl_el2
529	orr	x0, x0, #3			// Enable EL1 physical timers
530	msr	cnthctl_el2, x0
531	msr	cntvoff_el2, xzr		// Clear virtual offset
532
533#ifdef CONFIG_ARM_GIC_V3
534	/* GICv3 system register access */
535	mrs	x0, id_aa64pfr0_el1
536	ubfx	x0, x0, #24, #4
537	cmp	x0, #1
538	b.ne	3f
539
540	mrs_s	x0, ICC_SRE_EL2
541	orr	x0, x0, #ICC_SRE_EL2_SRE	// Set ICC_SRE_EL2.SRE==1
542	orr	x0, x0, #ICC_SRE_EL2_ENABLE	// Set ICC_SRE_EL2.Enable==1
543	msr_s	ICC_SRE_EL2, x0
544	isb					// Make sure SRE is now set
545	mrs_s	x0, ICC_SRE_EL2			// Read SRE back,
546	tbz	x0, #0, 3f			// and check that it sticks
547	msr_s	ICH_HCR_EL2, xzr		// Reset ICC_HCR_EL2 to defaults
548
5493:
550#endif
551
552	/* Populate ID registers. */
553	mrs	x0, midr_el1
554	mrs	x1, mpidr_el1
555	msr	vpidr_el2, x0
556	msr	vmpidr_el2, x1
557
558	/*
559	 * When VHE is not in use, early init of EL2 and EL1 needs to be
560	 * done here.
561	 * When VHE _is_ in use, EL1 will not be used in the host and
562	 * requires no configuration, and all non-hyp-specific EL2 setup
563	 * will be done via the _EL1 system register aliases in __cpu_setup.
564	 */
565	cbnz	x2, 1f
566
567	/* sctlr_el1 */
568	mov	x0, #0x0800			// Set/clear RES{1,0} bits
569CPU_BE(	movk	x0, #0x33d0, lsl #16	)	// Set EE and E0E on BE systems
570CPU_LE(	movk	x0, #0x30d0, lsl #16	)	// Clear EE and E0E on LE systems
571	msr	sctlr_el1, x0
572
573	/* Coprocessor traps. */
574	mov	x0, #0x33ff
575	msr	cptr_el2, x0			// Disable copro. traps to EL2
5761:
577
578#ifdef CONFIG_COMPAT
579	msr	hstr_el2, xzr			// Disable CP15 traps to EL2
580#endif
581
582	/* EL2 debug */
583	mrs	x0, id_aa64dfr0_el1		// Check ID_AA64DFR0_EL1 PMUVer
584	sbfx	x0, x0, #8, #4
585	cmp	x0, #1
586	b.lt	4f				// Skip if no PMU present
587	mrs	x0, pmcr_el0			// Disable debug access traps
588	ubfx	x0, x0, #11, #5			// to EL2 and allow access to
589	msr	mdcr_el2, x0			// all PMU counters from EL1
5904:
591
592	/* Stage-2 translation */
593	msr	vttbr_el2, xzr
594
595	cbz	x2, install_el2_stub
596
597	mov	w0, #BOOT_CPU_MODE_EL2		// This CPU booted in EL2
598	isb
599	ret
600
601install_el2_stub:
602	/* Hypervisor stub */
603	adrp	x0, __hyp_stub_vectors
604	add	x0, x0, #:lo12:__hyp_stub_vectors
605	msr	vbar_el2, x0
606
607	/* spsr */
608	mov	x0, #(PSR_F_BIT | PSR_I_BIT | PSR_A_BIT | PSR_D_BIT |\
609		      PSR_MODE_EL1h)
610	msr	spsr_el2, x0
611	msr	elr_el2, lr
612	mov	w0, #BOOT_CPU_MODE_EL2		// This CPU booted in EL2
613	eret
614ENDPROC(el2_setup)
615
616/*
617 * Sets the __boot_cpu_mode flag depending on the CPU boot mode passed
618 * in x20. See arch/arm64/include/asm/virt.h for more info.
619 */
620set_cpu_boot_mode_flag:
621	adr_l	x1, __boot_cpu_mode
622	cmp	w0, #BOOT_CPU_MODE_EL2
623	b.ne	1f
624	add	x1, x1, #4
6251:	str	w0, [x1]			// This CPU has booted in EL1
626	dmb	sy
627	dc	ivac, x1			// Invalidate potentially stale cache line
628	ret
629ENDPROC(set_cpu_boot_mode_flag)
630
631/*
632 * These values are written with the MMU off, but read with the MMU on.
633 * Writers will invalidate the corresponding address, discarding up to a
634 * 'Cache Writeback Granule' (CWG) worth of data. The linker script ensures
635 * sufficient alignment that the CWG doesn't overlap another section.
636 */
637	.pushsection ".mmuoff.data.write", "aw"
638/*
639 * We need to find out the CPU boot mode long after boot, so we need to
640 * store it in a writable variable.
641 *
642 * This is not in .bss, because we set it sufficiently early that the boot-time
643 * zeroing of .bss would clobber it.
644 */
645ENTRY(__boot_cpu_mode)
646	.long	BOOT_CPU_MODE_EL2
647	.long	BOOT_CPU_MODE_EL1
648/*
649 * The booting CPU updates the failed status @__early_cpu_boot_status,
650 * with MMU turned off.
651 */
652ENTRY(__early_cpu_boot_status)
653	.long 	0
654
655	.popsection
656
657	/*
658	 * This provides a "holding pen" for platforms to hold all secondary
659	 * cores are held until we're ready for them to initialise.
660	 */
661ENTRY(secondary_holding_pen)
662	bl	el2_setup			// Drop to EL1, w0=cpu_boot_mode
663	bl	set_cpu_boot_mode_flag
664	mrs	x0, mpidr_el1
665	mov_q	x1, MPIDR_HWID_BITMASK
666	and	x0, x0, x1
667	adr_l	x3, secondary_holding_pen_release
668pen:	ldr	x4, [x3]
669	cmp	x4, x0
670	b.eq	secondary_startup
671	wfe
672	b	pen
673ENDPROC(secondary_holding_pen)
674
675	/*
676	 * Secondary entry point that jumps straight into the kernel. Only to
677	 * be used where CPUs are brought online dynamically by the kernel.
678	 */
679ENTRY(secondary_entry)
680	bl	el2_setup			// Drop to EL1
681	bl	set_cpu_boot_mode_flag
682	b	secondary_startup
683ENDPROC(secondary_entry)
684
685secondary_startup:
686	/*
687	 * Common entry point for secondary CPUs.
688	 */
689	bl	__cpu_setup			// initialise processor
690	bl	__enable_mmu
691	ldr	x8, =__secondary_switched
692	br	x8
693ENDPROC(secondary_startup)
694
695__secondary_switched:
696	adr_l	x5, vectors
697	msr	vbar_el1, x5
698	isb
699
700	adr_l	x0, secondary_data
701	ldr	x0, [x0, #CPU_BOOT_STACK]	// get secondary_data.stack
702	mov	sp, x0
703	and	x0, x0, #~(THREAD_SIZE - 1)
704	msr	sp_el0, x0			// save thread_info
705	mov	x29, #0
706	b	secondary_start_kernel
707ENDPROC(__secondary_switched)
708
709/*
710 * The booting CPU updates the failed status @__early_cpu_boot_status,
711 * with MMU turned off.
712 *
713 * update_early_cpu_boot_status tmp, status
714 *  - Corrupts tmp1, tmp2
715 *  - Writes 'status' to __early_cpu_boot_status and makes sure
716 *    it is committed to memory.
717 */
718
719	.macro	update_early_cpu_boot_status status, tmp1, tmp2
720	mov	\tmp2, #\status
721	adr_l	\tmp1, __early_cpu_boot_status
722	str	\tmp2, [\tmp1]
723	dmb	sy
724	dc	ivac, \tmp1			// Invalidate potentially stale cache line
725	.endm
726
727/*
728 * Enable the MMU.
729 *
730 *  x0  = SCTLR_EL1 value for turning on the MMU.
731 *
732 * Returns to the caller via x30/lr. This requires the caller to be covered
733 * by the .idmap.text section.
734 *
735 * Checks if the selected granule size is supported by the CPU.
736 * If it isn't, park the CPU
737 */
738ENTRY(__enable_mmu)
739	mrs	x1, ID_AA64MMFR0_EL1
740	ubfx	x2, x1, #ID_AA64MMFR0_TGRAN_SHIFT, 4
741	cmp	x2, #ID_AA64MMFR0_TGRAN_SUPPORTED
742	b.ne	__no_granule_support
743	update_early_cpu_boot_status 0, x1, x2
744	adrp	x1, idmap_pg_dir
745	adrp	x2, swapper_pg_dir
746	msr	ttbr0_el1, x1			// load TTBR0
747	msr	ttbr1_el1, x2			// load TTBR1
748	isb
749	msr	sctlr_el1, x0
750	isb
751	/*
752	 * Invalidate the local I-cache so that any instructions fetched
753	 * speculatively from the PoC are discarded, since they may have
754	 * been dynamically patched at the PoU.
755	 */
756	ic	iallu
757	dsb	nsh
758	isb
759	ret
760ENDPROC(__enable_mmu)
761
762__no_granule_support:
763	/* Indicate that this CPU can't boot and is stuck in the kernel */
764	update_early_cpu_boot_status CPU_STUCK_IN_KERNEL, x1, x2
7651:
766	wfe
767	wfi
768	b	1b
769ENDPROC(__no_granule_support)
770
771#ifdef CONFIG_RELOCATABLE
772__relocate_kernel:
773	/*
774	 * Iterate over each entry in the relocation table, and apply the
775	 * relocations in place.
776	 */
777	ldr	w9, =__rela_offset		// offset to reloc table
778	ldr	w10, =__rela_size		// size of reloc table
779
780	mov_q	x11, KIMAGE_VADDR		// default virtual offset
781	add	x11, x11, x23			// actual virtual offset
782	add	x9, x9, x11			// __va(.rela)
783	add	x10, x9, x10			// __va(.rela) + sizeof(.rela)
784
7850:	cmp	x9, x10
786	b.hs	1f
787	ldp	x11, x12, [x9], #24
788	ldr	x13, [x9, #-8]
789	cmp	w12, #R_AARCH64_RELATIVE
790	b.ne	0b
791	add	x13, x13, x23			// relocate
792	str	x13, [x11, x23]
793	b	0b
7941:	ret
795ENDPROC(__relocate_kernel)
796#endif
797
798__primary_switch:
799#ifdef CONFIG_RANDOMIZE_BASE
800	mov	x19, x0				// preserve new SCTLR_EL1 value
801	mrs	x20, sctlr_el1			// preserve old SCTLR_EL1 value
802#endif
803
804	bl	__enable_mmu
805#ifdef CONFIG_RELOCATABLE
806	bl	__relocate_kernel
807#ifdef CONFIG_RANDOMIZE_BASE
808	ldr	x8, =__primary_switched
809	adrp	x0, __PHYS_OFFSET
810	blr	x8
811
812	/*
813	 * If we return here, we have a KASLR displacement in x23 which we need
814	 * to take into account by discarding the current kernel mapping and
815	 * creating a new one.
816	 */
817	msr	sctlr_el1, x20			// disable the MMU
818	isb
819	bl	__create_page_tables		// recreate kernel mapping
820
821	tlbi	vmalle1				// Remove any stale TLB entries
822	dsb	nsh
823
824	msr	sctlr_el1, x19			// re-enable the MMU
825	isb
826	ic	iallu				// flush instructions fetched
827	dsb	nsh				// via old mapping
828	isb
829
830	bl	__relocate_kernel
831#endif
832#endif
833	ldr	x8, =__primary_switched
834	adrp	x0, __PHYS_OFFSET
835	br	x8
836ENDPROC(__primary_switch)
837