xref: /openbmc/linux/arch/arm64/kernel/head.S (revision f2a89d3b)
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
211ENTRY(stext)
212	bl	preserve_boot_args
213	bl	el2_setup			// Drop to EL1, w20=cpu_boot_mode
214	adrp	x24, __PHYS_OFFSET
215	and	x23, x24, MIN_KIMG_ALIGN - 1	// KASLR offset, defaults to 0
216	bl	set_cpu_boot_mode_flag
217	bl	__create_page_tables		// x25=TTBR0, x26=TTBR1
218	/*
219	 * The following calls CPU setup code, see arch/arm64/mm/proc.S for
220	 * details.
221	 * On return, the CPU will be ready for the MMU to be turned on and
222	 * the TCR will have been set.
223	 */
224	bl	__cpu_setup			// initialise processor
225	adr_l	x27, __primary_switch		// address to jump to after
226						// MMU has been enabled
227	b	__enable_mmu
228ENDPROC(stext)
229
230/*
231 * Preserve the arguments passed by the bootloader in x0 .. x3
232 */
233preserve_boot_args:
234	mov	x21, x0				// x21=FDT
235
236	adr_l	x0, boot_args			// record the contents of
237	stp	x21, x1, [x0]			// x0 .. x3 at kernel entry
238	stp	x2, x3, [x0, #16]
239
240	dmb	sy				// needed before dc ivac with
241						// MMU off
242
243	add	x1, x0, #0x20			// 4 x 8 bytes
244	b	__inval_cache_range		// tail call
245ENDPROC(preserve_boot_args)
246
247/*
248 * Macro to create a table entry to the next page.
249 *
250 *	tbl:	page table address
251 *	virt:	virtual address
252 *	shift:	#imm page table shift
253 *	ptrs:	#imm pointers per table page
254 *
255 * Preserves:	virt
256 * Corrupts:	tmp1, tmp2
257 * Returns:	tbl -> next level table page address
258 */
259	.macro	create_table_entry, tbl, virt, shift, ptrs, tmp1, tmp2
260	lsr	\tmp1, \virt, #\shift
261	and	\tmp1, \tmp1, #\ptrs - 1	// table index
262	add	\tmp2, \tbl, #PAGE_SIZE
263	orr	\tmp2, \tmp2, #PMD_TYPE_TABLE	// address of next table and entry type
264	str	\tmp2, [\tbl, \tmp1, lsl #3]
265	add	\tbl, \tbl, #PAGE_SIZE		// next level table page
266	.endm
267
268/*
269 * Macro to populate the PGD (and possibily PUD) for the corresponding
270 * block entry in the next level (tbl) for the given virtual address.
271 *
272 * Preserves:	tbl, next, virt
273 * Corrupts:	tmp1, tmp2
274 */
275	.macro	create_pgd_entry, tbl, virt, tmp1, tmp2
276	create_table_entry \tbl, \virt, PGDIR_SHIFT, PTRS_PER_PGD, \tmp1, \tmp2
277#if SWAPPER_PGTABLE_LEVELS > 3
278	create_table_entry \tbl, \virt, PUD_SHIFT, PTRS_PER_PUD, \tmp1, \tmp2
279#endif
280#if SWAPPER_PGTABLE_LEVELS > 2
281	create_table_entry \tbl, \virt, SWAPPER_TABLE_SHIFT, PTRS_PER_PTE, \tmp1, \tmp2
282#endif
283	.endm
284
285/*
286 * Macro to populate block entries in the page table for the start..end
287 * virtual range (inclusive).
288 *
289 * Preserves:	tbl, flags
290 * Corrupts:	phys, start, end, pstate
291 */
292	.macro	create_block_map, tbl, flags, phys, start, end
293	lsr	\phys, \phys, #SWAPPER_BLOCK_SHIFT
294	lsr	\start, \start, #SWAPPER_BLOCK_SHIFT
295	and	\start, \start, #PTRS_PER_PTE - 1	// table index
296	orr	\phys, \flags, \phys, lsl #SWAPPER_BLOCK_SHIFT	// table entry
297	lsr	\end, \end, #SWAPPER_BLOCK_SHIFT
298	and	\end, \end, #PTRS_PER_PTE - 1		// table end index
2999999:	str	\phys, [\tbl, \start, lsl #3]		// store the entry
300	add	\start, \start, #1			// next entry
301	add	\phys, \phys, #SWAPPER_BLOCK_SIZE		// next block
302	cmp	\start, \end
303	b.ls	9999b
304	.endm
305
306/*
307 * Setup the initial page tables. We only setup the barest amount which is
308 * required to get the kernel running. The following sections are required:
309 *   - identity mapping to enable the MMU (low address, TTBR0)
310 *   - first few MB of the kernel linear mapping to jump to once the MMU has
311 *     been enabled
312 */
313__create_page_tables:
314	adrp	x25, idmap_pg_dir
315	adrp	x26, swapper_pg_dir
316	mov	x28, lr
317
318	/*
319	 * Invalidate the idmap and swapper page tables to avoid potential
320	 * dirty cache lines being evicted.
321	 */
322	mov	x0, x25
323	add	x1, x26, #SWAPPER_DIR_SIZE
324	bl	__inval_cache_range
325
326	/*
327	 * Clear the idmap and swapper page tables.
328	 */
329	mov	x0, x25
330	add	x6, x26, #SWAPPER_DIR_SIZE
3311:	stp	xzr, xzr, [x0], #16
332	stp	xzr, xzr, [x0], #16
333	stp	xzr, xzr, [x0], #16
334	stp	xzr, xzr, [x0], #16
335	cmp	x0, x6
336	b.lo	1b
337
338	mov	x7, SWAPPER_MM_MMUFLAGS
339
340	/*
341	 * Create the identity mapping.
342	 */
343	mov	x0, x25				// idmap_pg_dir
344	adrp	x3, __idmap_text_start		// __pa(__idmap_text_start)
345
346#ifndef CONFIG_ARM64_VA_BITS_48
347#define EXTRA_SHIFT	(PGDIR_SHIFT + PAGE_SHIFT - 3)
348#define EXTRA_PTRS	(1 << (48 - EXTRA_SHIFT))
349
350	/*
351	 * If VA_BITS < 48, it may be too small to allow for an ID mapping to be
352	 * created that covers system RAM if that is located sufficiently high
353	 * in the physical address space. So for the ID map, use an extended
354	 * virtual range in that case, by configuring an additional translation
355	 * level.
356	 * First, we have to verify our assumption that the current value of
357	 * VA_BITS was chosen such that all translation levels are fully
358	 * utilised, and that lowering T0SZ will always result in an additional
359	 * translation level to be configured.
360	 */
361#if VA_BITS != EXTRA_SHIFT
362#error "Mismatch between VA_BITS and page size/number of translation levels"
363#endif
364
365	/*
366	 * Calculate the maximum allowed value for TCR_EL1.T0SZ so that the
367	 * entire ID map region can be mapped. As T0SZ == (64 - #bits used),
368	 * this number conveniently equals the number of leading zeroes in
369	 * the physical address of __idmap_text_end.
370	 */
371	adrp	x5, __idmap_text_end
372	clz	x5, x5
373	cmp	x5, TCR_T0SZ(VA_BITS)	// default T0SZ small enough?
374	b.ge	1f			// .. then skip additional level
375
376	adr_l	x6, idmap_t0sz
377	str	x5, [x6]
378	dmb	sy
379	dc	ivac, x6		// Invalidate potentially stale cache line
380
381	create_table_entry x0, x3, EXTRA_SHIFT, EXTRA_PTRS, x5, x6
3821:
383#endif
384
385	create_pgd_entry x0, x3, x5, x6
386	mov	x5, x3				// __pa(__idmap_text_start)
387	adr_l	x6, __idmap_text_end		// __pa(__idmap_text_end)
388	create_block_map x0, x7, x3, x5, x6
389
390	/*
391	 * Map the kernel image (starting with PHYS_OFFSET).
392	 */
393	mov	x0, x26				// swapper_pg_dir
394	mov_q	x5, KIMAGE_VADDR + TEXT_OFFSET	// compile time __va(_text)
395	add	x5, x5, x23			// add KASLR displacement
396	create_pgd_entry x0, x5, x3, x6
397	adrp	x6, _end			// runtime __pa(_end)
398	adrp	x3, _text			// runtime __pa(_text)
399	sub	x6, x6, x3			// _end - _text
400	add	x6, x6, x5			// runtime __va(_end)
401	create_block_map x0, x7, x3, x5, x6
402
403	/*
404	 * Since the page tables have been populated with non-cacheable
405	 * accesses (MMU disabled), invalidate the idmap and swapper page
406	 * tables again to remove any speculatively loaded cache lines.
407	 */
408	mov	x0, x25
409	add	x1, x26, #SWAPPER_DIR_SIZE
410	dmb	sy
411	bl	__inval_cache_range
412
413	ret	x28
414ENDPROC(__create_page_tables)
415	.ltorg
416
417/*
418 * The following fragment of code is executed with the MMU enabled.
419 */
420	.set	initial_sp, init_thread_union + THREAD_START_SP
421__primary_switched:
422	mov	x28, lr				// preserve LR
423	adr_l	x8, vectors			// load VBAR_EL1 with virtual
424	msr	vbar_el1, x8			// vector table address
425	isb
426
427	// Clear BSS
428	adr_l	x0, __bss_start
429	mov	x1, xzr
430	adr_l	x2, __bss_stop
431	sub	x2, x2, x0
432	bl	__pi_memset
433	dsb	ishst				// Make zero page visible to PTW
434
435	adr_l	sp, initial_sp, x4
436	mov	x4, sp
437	and	x4, x4, #~(THREAD_SIZE - 1)
438	msr	sp_el0, x4			// Save thread_info
439	str_l	x21, __fdt_pointer, x5		// Save FDT pointer
440
441	ldr_l	x4, kimage_vaddr		// Save the offset between
442	sub	x4, x4, x24			// the kernel virtual and
443	str_l	x4, kimage_voffset, x5		// physical mappings
444
445	mov	x29, #0
446#ifdef CONFIG_KASAN
447	bl	kasan_early_init
448#endif
449#ifdef CONFIG_RANDOMIZE_BASE
450	tst	x23, ~(MIN_KIMG_ALIGN - 1)	// already running randomized?
451	b.ne	0f
452	mov	x0, x21				// pass FDT address in x0
453	mov	x1, x23				// pass modulo offset in x1
454	bl	kaslr_early_init		// parse FDT for KASLR options
455	cbz	x0, 0f				// KASLR disabled? just proceed
456	orr	x23, x23, x0			// record KASLR offset
457	ret	x28				// we must enable KASLR, return
458						// to __enable_mmu()
4590:
460#endif
461	b	start_kernel
462ENDPROC(__primary_switched)
463
464/*
465 * end early head section, begin head code that is also used for
466 * hotplug and needs to have the same protections as the text region
467 */
468	.section ".text","ax"
469
470ENTRY(kimage_vaddr)
471	.quad		_text - TEXT_OFFSET
472
473/*
474 * If we're fortunate enough to boot at EL2, ensure that the world is
475 * sane before dropping to EL1.
476 *
477 * Returns either BOOT_CPU_MODE_EL1 or BOOT_CPU_MODE_EL2 in x20 if
478 * booted in EL1 or EL2 respectively.
479 */
480ENTRY(el2_setup)
481	mrs	x0, CurrentEL
482	cmp	x0, #CurrentEL_EL2
483	b.ne	1f
484	mrs	x0, sctlr_el2
485CPU_BE(	orr	x0, x0, #(1 << 25)	)	// Set the EE bit for EL2
486CPU_LE(	bic	x0, x0, #(1 << 25)	)	// Clear the EE bit for EL2
487	msr	sctlr_el2, x0
488	b	2f
4891:	mrs	x0, sctlr_el1
490CPU_BE(	orr	x0, x0, #(3 << 24)	)	// Set the EE and E0E bits for EL1
491CPU_LE(	bic	x0, x0, #(3 << 24)	)	// Clear the EE and E0E bits for EL1
492	msr	sctlr_el1, x0
493	mov	w20, #BOOT_CPU_MODE_EL1		// This cpu booted in EL1
494	isb
495	ret
496
4972:
498#ifdef CONFIG_ARM64_VHE
499	/*
500	 * Check for VHE being present. For the rest of the EL2 setup,
501	 * x2 being non-zero indicates that we do have VHE, and that the
502	 * kernel is intended to run at EL2.
503	 */
504	mrs	x2, id_aa64mmfr1_el1
505	ubfx	x2, x2, #8, #4
506#else
507	mov	x2, xzr
508#endif
509
510	/* Hyp configuration. */
511	mov	x0, #HCR_RW			// 64-bit EL1
512	cbz	x2, set_hcr
513	orr	x0, x0, #HCR_TGE		// Enable Host Extensions
514	orr	x0, x0, #HCR_E2H
515set_hcr:
516	msr	hcr_el2, x0
517	isb
518
519	/* Generic timers. */
520	mrs	x0, cnthctl_el2
521	orr	x0, x0, #3			// Enable EL1 physical timers
522	msr	cnthctl_el2, x0
523	msr	cntvoff_el2, xzr		// Clear virtual offset
524
525#ifdef CONFIG_ARM_GIC_V3
526	/* GICv3 system register access */
527	mrs	x0, id_aa64pfr0_el1
528	ubfx	x0, x0, #24, #4
529	cmp	x0, #1
530	b.ne	3f
531
532	mrs_s	x0, ICC_SRE_EL2
533	orr	x0, x0, #ICC_SRE_EL2_SRE	// Set ICC_SRE_EL2.SRE==1
534	orr	x0, x0, #ICC_SRE_EL2_ENABLE	// Set ICC_SRE_EL2.Enable==1
535	msr_s	ICC_SRE_EL2, x0
536	isb					// Make sure SRE is now set
537	mrs_s	x0, ICC_SRE_EL2			// Read SRE back,
538	tbz	x0, #0, 3f			// and check that it sticks
539	msr_s	ICH_HCR_EL2, xzr		// Reset ICC_HCR_EL2 to defaults
540
5413:
542#endif
543
544	/* Populate ID registers. */
545	mrs	x0, midr_el1
546	mrs	x1, mpidr_el1
547	msr	vpidr_el2, x0
548	msr	vmpidr_el2, x1
549
550	/*
551	 * When VHE is not in use, early init of EL2 and EL1 needs to be
552	 * done here.
553	 * When VHE _is_ in use, EL1 will not be used in the host and
554	 * requires no configuration, and all non-hyp-specific EL2 setup
555	 * will be done via the _EL1 system register aliases in __cpu_setup.
556	 */
557	cbnz	x2, 1f
558
559	/* sctlr_el1 */
560	mov	x0, #0x0800			// Set/clear RES{1,0} bits
561CPU_BE(	movk	x0, #0x33d0, lsl #16	)	// Set EE and E0E on BE systems
562CPU_LE(	movk	x0, #0x30d0, lsl #16	)	// Clear EE and E0E on LE systems
563	msr	sctlr_el1, x0
564
565	/* Coprocessor traps. */
566	mov	x0, #0x33ff
567	msr	cptr_el2, x0			// Disable copro. traps to EL2
5681:
569
570#ifdef CONFIG_COMPAT
571	msr	hstr_el2, xzr			// Disable CP15 traps to EL2
572#endif
573
574	/* EL2 debug */
575	mrs	x0, id_aa64dfr0_el1		// Check ID_AA64DFR0_EL1 PMUVer
576	sbfx	x0, x0, #8, #4
577	cmp	x0, #1
578	b.lt	4f				// Skip if no PMU present
579	mrs	x0, pmcr_el0			// Disable debug access traps
580	ubfx	x0, x0, #11, #5			// to EL2 and allow access to
581	msr	mdcr_el2, x0			// all PMU counters from EL1
5824:
583
584	/* Stage-2 translation */
585	msr	vttbr_el2, xzr
586
587	cbz	x2, install_el2_stub
588
589	mov	w20, #BOOT_CPU_MODE_EL2		// This CPU booted in EL2
590	isb
591	ret
592
593install_el2_stub:
594	/* Hypervisor stub */
595	adrp	x0, __hyp_stub_vectors
596	add	x0, x0, #:lo12:__hyp_stub_vectors
597	msr	vbar_el2, x0
598
599	/* spsr */
600	mov	x0, #(PSR_F_BIT | PSR_I_BIT | PSR_A_BIT | PSR_D_BIT |\
601		      PSR_MODE_EL1h)
602	msr	spsr_el2, x0
603	msr	elr_el2, lr
604	mov	w20, #BOOT_CPU_MODE_EL2		// This CPU booted in EL2
605	eret
606ENDPROC(el2_setup)
607
608/*
609 * Sets the __boot_cpu_mode flag depending on the CPU boot mode passed
610 * in x20. See arch/arm64/include/asm/virt.h for more info.
611 */
612set_cpu_boot_mode_flag:
613	adr_l	x1, __boot_cpu_mode
614	cmp	w20, #BOOT_CPU_MODE_EL2
615	b.ne	1f
616	add	x1, x1, #4
6171:	str	w20, [x1]			// This CPU has booted in EL1
618	dmb	sy
619	dc	ivac, x1			// Invalidate potentially stale cache line
620	ret
621ENDPROC(set_cpu_boot_mode_flag)
622
623/*
624 * We need to find out the CPU boot mode long after boot, so we need to
625 * store it in a writable variable.
626 *
627 * This is not in .bss, because we set it sufficiently early that the boot-time
628 * zeroing of .bss would clobber it.
629 */
630	.pushsection	.data..cacheline_aligned
631	.align	L1_CACHE_SHIFT
632ENTRY(__boot_cpu_mode)
633	.long	BOOT_CPU_MODE_EL2
634	.long	BOOT_CPU_MODE_EL1
635	.popsection
636
637	/*
638	 * This provides a "holding pen" for platforms to hold all secondary
639	 * cores are held until we're ready for them to initialise.
640	 */
641ENTRY(secondary_holding_pen)
642	bl	el2_setup			// Drop to EL1, w20=cpu_boot_mode
643	bl	set_cpu_boot_mode_flag
644	mrs	x0, mpidr_el1
645	mov_q	x1, MPIDR_HWID_BITMASK
646	and	x0, x0, x1
647	adr_l	x3, secondary_holding_pen_release
648pen:	ldr	x4, [x3]
649	cmp	x4, x0
650	b.eq	secondary_startup
651	wfe
652	b	pen
653ENDPROC(secondary_holding_pen)
654
655	/*
656	 * Secondary entry point that jumps straight into the kernel. Only to
657	 * be used where CPUs are brought online dynamically by the kernel.
658	 */
659ENTRY(secondary_entry)
660	bl	el2_setup			// Drop to EL1
661	bl	set_cpu_boot_mode_flag
662	b	secondary_startup
663ENDPROC(secondary_entry)
664
665secondary_startup:
666	/*
667	 * Common entry point for secondary CPUs.
668	 */
669	adrp	x25, idmap_pg_dir
670	adrp	x26, swapper_pg_dir
671	bl	__cpu_setup			// initialise processor
672
673	adr_l	x27, __secondary_switch		// address to jump to after enabling the MMU
674	b	__enable_mmu
675ENDPROC(secondary_startup)
676
677__secondary_switched:
678	adr_l	x5, vectors
679	msr	vbar_el1, x5
680	isb
681
682	adr_l	x0, secondary_data
683	ldr	x0, [x0, #CPU_BOOT_STACK]	// get secondary_data.stack
684	mov	sp, x0
685	and	x0, x0, #~(THREAD_SIZE - 1)
686	msr	sp_el0, x0			// save thread_info
687	mov	x29, #0
688	b	secondary_start_kernel
689ENDPROC(__secondary_switched)
690
691/*
692 * The booting CPU updates the failed status @__early_cpu_boot_status,
693 * with MMU turned off.
694 *
695 * update_early_cpu_boot_status tmp, status
696 *  - Corrupts tmp1, tmp2
697 *  - Writes 'status' to __early_cpu_boot_status and makes sure
698 *    it is committed to memory.
699 */
700
701	.macro	update_early_cpu_boot_status status, tmp1, tmp2
702	mov	\tmp2, #\status
703	adr_l	\tmp1, __early_cpu_boot_status
704	str	\tmp2, [\tmp1]
705	dmb	sy
706	dc	ivac, \tmp1			// Invalidate potentially stale cache line
707	.endm
708
709	.pushsection	.data..cacheline_aligned
710	.align	L1_CACHE_SHIFT
711ENTRY(__early_cpu_boot_status)
712	.long 	0
713	.popsection
714
715/*
716 * Enable the MMU.
717 *
718 *  x0  = SCTLR_EL1 value for turning on the MMU.
719 *  x27 = *virtual* address to jump to upon completion
720 *
721 * Other registers depend on the function called upon completion.
722 *
723 * Checks if the selected granule size is supported by the CPU.
724 * If it isn't, park the CPU
725 */
726	.section	".idmap.text", "ax"
727ENTRY(__enable_mmu)
728	mrs	x22, sctlr_el1			// preserve old SCTLR_EL1 value
729	mrs	x1, ID_AA64MMFR0_EL1
730	ubfx	x2, x1, #ID_AA64MMFR0_TGRAN_SHIFT, 4
731	cmp	x2, #ID_AA64MMFR0_TGRAN_SUPPORTED
732	b.ne	__no_granule_support
733	update_early_cpu_boot_status 0, x1, x2
734	msr	ttbr0_el1, x25			// load TTBR0
735	msr	ttbr1_el1, x26			// load TTBR1
736	isb
737	msr	sctlr_el1, x0
738	isb
739	/*
740	 * Invalidate the local I-cache so that any instructions fetched
741	 * speculatively from the PoC are discarded, since they may have
742	 * been dynamically patched at the PoU.
743	 */
744	ic	iallu
745	dsb	nsh
746	isb
747#ifdef CONFIG_RANDOMIZE_BASE
748	mov	x19, x0				// preserve new SCTLR_EL1 value
749	blr	x27
750
751	/*
752	 * If we return here, we have a KASLR displacement in x23 which we need
753	 * to take into account by discarding the current kernel mapping and
754	 * creating a new one.
755	 */
756	msr	sctlr_el1, x22			// disable the MMU
757	isb
758	bl	__create_page_tables		// recreate kernel mapping
759
760	msr	sctlr_el1, x19			// re-enable the MMU
761	isb
762	ic	iallu				// flush instructions fetched
763	dsb	nsh				// via old mapping
764	isb
765#endif
766	br	x27
767ENDPROC(__enable_mmu)
768
769__no_granule_support:
770	/* Indicate that this CPU can't boot and is stuck in the kernel */
771	update_early_cpu_boot_status CPU_STUCK_IN_KERNEL, x1, x2
7721:
773	wfe
774	wfi
775	b 1b
776ENDPROC(__no_granule_support)
777
778__primary_switch:
779#ifdef CONFIG_RELOCATABLE
780	/*
781	 * Iterate over each entry in the relocation table, and apply the
782	 * relocations in place.
783	 */
784	ldr	w9, =__rela_offset		// offset to reloc table
785	ldr	w10, =__rela_size		// size of reloc table
786
787	mov_q	x11, KIMAGE_VADDR		// default virtual offset
788	add	x11, x11, x23			// actual virtual offset
789	add	x9, x9, x11			// __va(.rela)
790	add	x10, x9, x10			// __va(.rela) + sizeof(.rela)
791
7920:	cmp	x9, x10
793	b.hs	1f
794	ldp	x11, x12, [x9], #24
795	ldr	x13, [x9, #-8]
796	cmp	w12, #R_AARCH64_RELATIVE
797	b.ne	0b
798	add	x13, x13, x23			// relocate
799	str	x13, [x11, x23]
800	b	0b
801
8021:
803#endif
804	ldr	x8, =__primary_switched
805	br	x8
806ENDPROC(__primary_switch)
807
808__secondary_switch:
809	ldr	x8, =__secondary_switched
810	br	x8
811ENDPROC(__secondary_switch)
812