xref: /linux/arch/arm64/kernel/head.S (revision 156010ed9c2ac1e9df6c11b1f688cf8a6e0152e6)
1/* SPDX-License-Identifier: GPL-2.0-only */
2/*
3 * Low-level CPU initialisation
4 * Based on arch/arm/kernel/head.S
5 *
6 * Copyright (C) 1994-2002 Russell King
7 * Copyright (C) 2003-2012 ARM Ltd.
8 * Authors:	Catalin Marinas <catalin.marinas@arm.com>
9 *		Will Deacon <will.deacon@arm.com>
10 */
11
12#include <linux/linkage.h>
13#include <linux/init.h>
14#include <linux/pgtable.h>
15
16#include <asm/asm_pointer_auth.h>
17#include <asm/assembler.h>
18#include <asm/boot.h>
19#include <asm/bug.h>
20#include <asm/ptrace.h>
21#include <asm/asm-offsets.h>
22#include <asm/cache.h>
23#include <asm/cputype.h>
24#include <asm/el2_setup.h>
25#include <asm/elf.h>
26#include <asm/image.h>
27#include <asm/kernel-pgtable.h>
28#include <asm/kvm_arm.h>
29#include <asm/memory.h>
30#include <asm/pgtable-hwdef.h>
31#include <asm/page.h>
32#include <asm/scs.h>
33#include <asm/smp.h>
34#include <asm/sysreg.h>
35#include <asm/thread_info.h>
36#include <asm/virt.h>
37
38#include "efi-header.S"
39
40#if (PAGE_OFFSET & 0x1fffff) != 0
41#error PAGE_OFFSET must be at least 2MB aligned
42#endif
43
44/*
45 * Kernel startup entry point.
46 * ---------------------------
47 *
48 * The requirements are:
49 *   MMU = off, D-cache = off, I-cache = on or off,
50 *   x0 = physical address to the FDT blob.
51 *
52 * Note that the callee-saved registers are used for storing variables
53 * that are useful before the MMU is enabled. The allocations are described
54 * in the entry routines.
55 */
56	__HEAD
57	/*
58	 * DO NOT MODIFY. Image header expected by Linux boot-loaders.
59	 */
60	efi_signature_nop			// special NOP to identity as PE/COFF executable
61	b	primary_entry			// branch to kernel start, magic
62	.quad	0				// Image load offset from start of RAM, little-endian
63	le64sym	_kernel_size_le			// Effective size of kernel image, little-endian
64	le64sym	_kernel_flags_le		// Informative flags, little-endian
65	.quad	0				// reserved
66	.quad	0				// reserved
67	.quad	0				// reserved
68	.ascii	ARM64_IMAGE_MAGIC		// Magic number
69	.long	.Lpe_header_offset		// Offset to the PE header.
70
71	__EFI_PE_HEADER
72
73	.section ".idmap.text","awx"
74
75	/*
76	 * The following callee saved general purpose registers are used on the
77	 * primary lowlevel boot path:
78	 *
79	 *  Register   Scope                      Purpose
80	 *  x19        primary_entry() .. start_kernel()        whether we entered with the MMU on
81	 *  x20        primary_entry() .. __primary_switch()    CPU boot mode
82	 *  x21        primary_entry() .. start_kernel()        FDT pointer passed at boot in x0
83	 *  x22        create_idmap() .. start_kernel()         ID map VA of the DT blob
84	 *  x23        primary_entry() .. start_kernel()        physical misalignment/KASLR offset
85	 *  x24        __primary_switch()                       linear map KASLR seed
86	 *  x25        primary_entry() .. start_kernel()        supported VA size
87	 *  x28        create_idmap()                           callee preserved temp register
88	 */
89SYM_CODE_START(primary_entry)
90	bl	record_mmu_state
91	bl	preserve_boot_args
92	bl	create_idmap
93
94	/*
95	 * If we entered with the MMU and caches on, clean the ID mapped part
96	 * of the primary boot code to the PoC so we can safely execute it with
97	 * the MMU off.
98	 */
99	cbz	x19, 0f
100	adrp	x0, __idmap_text_start
101	adr_l	x1, __idmap_text_end
102	bl	dcache_clean_poc
1030:	mov	x0, x19
104	bl	init_kernel_el			// w0=cpu_boot_mode
105	mov	x20, x0
106
107	/*
108	 * The following calls CPU setup code, see arch/arm64/mm/proc.S for
109	 * details.
110	 * On return, the CPU will be ready for the MMU to be turned on and
111	 * the TCR will have been set.
112	 */
113#if VA_BITS > 48
114	mrs_s	x0, SYS_ID_AA64MMFR2_EL1
115	tst	x0, #0xf << ID_AA64MMFR2_EL1_VARange_SHIFT
116	mov	x0, #VA_BITS
117	mov	x25, #VA_BITS_MIN
118	csel	x25, x25, x0, eq
119	mov	x0, x25
120#endif
121	bl	__cpu_setup			// initialise processor
122	b	__primary_switch
123SYM_CODE_END(primary_entry)
124
125	__INIT
126SYM_CODE_START_LOCAL(record_mmu_state)
127	mrs	x19, CurrentEL
128	cmp	x19, #CurrentEL_EL2
129	mrs	x19, sctlr_el1
130	b.ne	0f
131	mrs	x19, sctlr_el2
1320:
133CPU_LE( tbnz	x19, #SCTLR_ELx_EE_SHIFT, 1f	)
134CPU_BE( tbz	x19, #SCTLR_ELx_EE_SHIFT, 1f	)
135	tst	x19, #SCTLR_ELx_C		// Z := (C == 0)
136	and	x19, x19, #SCTLR_ELx_M		// isolate M bit
137	csel	x19, xzr, x19, eq		// clear x19 if Z
138	ret
139
140	/*
141	 * Set the correct endianness early so all memory accesses issued
142	 * before init_kernel_el() occur in the correct byte order. Note that
143	 * this means the MMU must be disabled, or the active ID map will end
144	 * up getting interpreted with the wrong byte order.
145	 */
1461:	eor	x19, x19, #SCTLR_ELx_EE
147	bic	x19, x19, #SCTLR_ELx_M
148	b.ne	2f
149	pre_disable_mmu_workaround
150	msr	sctlr_el2, x19
151	b	3f
152	pre_disable_mmu_workaround
1532:	msr	sctlr_el1, x19
1543:	isb
155	mov	x19, xzr
156	ret
157SYM_CODE_END(record_mmu_state)
158
159/*
160 * Preserve the arguments passed by the bootloader in x0 .. x3
161 */
162SYM_CODE_START_LOCAL(preserve_boot_args)
163	mov	x21, x0				// x21=FDT
164
165	adr_l	x0, boot_args			// record the contents of
166	stp	x21, x1, [x0]			// x0 .. x3 at kernel entry
167	stp	x2, x3, [x0, #16]
168
169	cbnz	x19, 0f				// skip cache invalidation if MMU is on
170	dmb	sy				// needed before dc ivac with
171						// MMU off
172
173	add	x1, x0, #0x20			// 4 x 8 bytes
174	b	dcache_inval_poc		// tail call
1750:	str_l   x19, mmu_enabled_at_boot, x0
176	ret
177SYM_CODE_END(preserve_boot_args)
178
179SYM_FUNC_START_LOCAL(clear_page_tables)
180	/*
181	 * Clear the init page tables.
182	 */
183	adrp	x0, init_pg_dir
184	adrp	x1, init_pg_end
185	sub	x2, x1, x0
186	mov	x1, xzr
187	b	__pi_memset			// tail call
188SYM_FUNC_END(clear_page_tables)
189
190/*
191 * Macro to populate page table entries, these entries can be pointers to the next level
192 * or last level entries pointing to physical memory.
193 *
194 *	tbl:	page table address
195 *	rtbl:	pointer to page table or physical memory
196 *	index:	start index to write
197 *	eindex:	end index to write - [index, eindex] written to
198 *	flags:	flags for pagetable entry to or in
199 *	inc:	increment to rtbl between each entry
200 *	tmp1:	temporary variable
201 *
202 * Preserves:	tbl, eindex, flags, inc
203 * Corrupts:	index, tmp1
204 * Returns:	rtbl
205 */
206	.macro populate_entries, tbl, rtbl, index, eindex, flags, inc, tmp1
207.Lpe\@:	phys_to_pte \tmp1, \rtbl
208	orr	\tmp1, \tmp1, \flags	// tmp1 = table entry
209	str	\tmp1, [\tbl, \index, lsl #3]
210	add	\rtbl, \rtbl, \inc	// rtbl = pa next level
211	add	\index, \index, #1
212	cmp	\index, \eindex
213	b.ls	.Lpe\@
214	.endm
215
216/*
217 * Compute indices of table entries from virtual address range. If multiple entries
218 * were needed in the previous page table level then the next page table level is assumed
219 * to be composed of multiple pages. (This effectively scales the end index).
220 *
221 *	vstart:	virtual address of start of range
222 *	vend:	virtual address of end of range - we map [vstart, vend]
223 *	shift:	shift used to transform virtual address into index
224 *	order:  #imm 2log(number of entries in page table)
225 *	istart:	index in table corresponding to vstart
226 *	iend:	index in table corresponding to vend
227 *	count:	On entry: how many extra entries were required in previous level, scales
228 *			  our end index.
229 *		On exit: returns how many extra entries required for next page table level
230 *
231 * Preserves:	vstart, vend
232 * Returns:	istart, iend, count
233 */
234	.macro compute_indices, vstart, vend, shift, order, istart, iend, count
235	ubfx	\istart, \vstart, \shift, \order
236	ubfx	\iend, \vend, \shift, \order
237	add	\iend, \iend, \count, lsl \order
238	sub	\count, \iend, \istart
239	.endm
240
241/*
242 * Map memory for specified virtual address range. Each level of page table needed supports
243 * multiple entries. If a level requires n entries the next page table level is assumed to be
244 * formed from n pages.
245 *
246 *	tbl:	location of page table
247 *	rtbl:	address to be used for first level page table entry (typically tbl + PAGE_SIZE)
248 *	vstart:	virtual address of start of range
249 *	vend:	virtual address of end of range - we map [vstart, vend - 1]
250 *	flags:	flags to use to map last level entries
251 *	phys:	physical address corresponding to vstart - physical memory is contiguous
252 *	order:  #imm 2log(number of entries in PGD table)
253 *
254 * If extra_shift is set, an extra level will be populated if the end address does
255 * not fit in 'extra_shift' bits. This assumes vend is in the TTBR0 range.
256 *
257 * Temporaries:	istart, iend, tmp, count, sv - these need to be different registers
258 * Preserves:	vstart, flags
259 * Corrupts:	tbl, rtbl, vend, istart, iend, tmp, count, sv
260 */
261	.macro map_memory, tbl, rtbl, vstart, vend, flags, phys, order, istart, iend, tmp, count, sv, extra_shift
262	sub \vend, \vend, #1
263	add \rtbl, \tbl, #PAGE_SIZE
264	mov \count, #0
265
266	.ifnb	\extra_shift
267	tst	\vend, #~((1 << (\extra_shift)) - 1)
268	b.eq	.L_\@
269	compute_indices \vstart, \vend, #\extra_shift, #(PAGE_SHIFT - 3), \istart, \iend, \count
270	mov \sv, \rtbl
271	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
272	mov \tbl, \sv
273	.endif
274.L_\@:
275	compute_indices \vstart, \vend, #PGDIR_SHIFT, #\order, \istart, \iend, \count
276	mov \sv, \rtbl
277	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
278	mov \tbl, \sv
279
280#if SWAPPER_PGTABLE_LEVELS > 3
281	compute_indices \vstart, \vend, #PUD_SHIFT, #(PAGE_SHIFT - 3), \istart, \iend, \count
282	mov \sv, \rtbl
283	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
284	mov \tbl, \sv
285#endif
286
287#if SWAPPER_PGTABLE_LEVELS > 2
288	compute_indices \vstart, \vend, #SWAPPER_TABLE_SHIFT, #(PAGE_SHIFT - 3), \istart, \iend, \count
289	mov \sv, \rtbl
290	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
291	mov \tbl, \sv
292#endif
293
294	compute_indices \vstart, \vend, #SWAPPER_BLOCK_SHIFT, #(PAGE_SHIFT - 3), \istart, \iend, \count
295	bic \rtbl, \phys, #SWAPPER_BLOCK_SIZE - 1
296	populate_entries \tbl, \rtbl, \istart, \iend, \flags, #SWAPPER_BLOCK_SIZE, \tmp
297	.endm
298
299/*
300 * Remap a subregion created with the map_memory macro with modified attributes
301 * or output address. The entire remapped region must have been covered in the
302 * invocation of map_memory.
303 *
304 * x0: last level table address (returned in first argument to map_memory)
305 * x1: start VA of the existing mapping
306 * x2: start VA of the region to update
307 * x3: end VA of the region to update (exclusive)
308 * x4: start PA associated with the region to update
309 * x5: attributes to set on the updated region
310 * x6: order of the last level mappings
311 */
312SYM_FUNC_START_LOCAL(remap_region)
313	sub	x3, x3, #1		// make end inclusive
314
315	// Get the index offset for the start of the last level table
316	lsr	x1, x1, x6
317	bfi	x1, xzr, #0, #PAGE_SHIFT - 3
318
319	// Derive the start and end indexes into the last level table
320	// associated with the provided region
321	lsr	x2, x2, x6
322	lsr	x3, x3, x6
323	sub	x2, x2, x1
324	sub	x3, x3, x1
325
326	mov	x1, #1
327	lsl	x6, x1, x6		// block size at this level
328
329	populate_entries x0, x4, x2, x3, x5, x6, x7
330	ret
331SYM_FUNC_END(remap_region)
332
333SYM_FUNC_START_LOCAL(create_idmap)
334	mov	x28, lr
335	/*
336	 * The ID map carries a 1:1 mapping of the physical address range
337	 * covered by the loaded image, which could be anywhere in DRAM. This
338	 * means that the required size of the VA (== PA) space is decided at
339	 * boot time, and could be more than the configured size of the VA
340	 * space for ordinary kernel and user space mappings.
341	 *
342	 * There are three cases to consider here:
343	 * - 39 <= VA_BITS < 48, and the ID map needs up to 48 VA bits to cover
344	 *   the placement of the image. In this case, we configure one extra
345	 *   level of translation on the fly for the ID map only. (This case
346	 *   also covers 42-bit VA/52-bit PA on 64k pages).
347	 *
348	 * - VA_BITS == 48, and the ID map needs more than 48 VA bits. This can
349	 *   only happen when using 64k pages, in which case we need to extend
350	 *   the root level table rather than add a level. Note that we can
351	 *   treat this case as 'always extended' as long as we take care not
352	 *   to program an unsupported T0SZ value into the TCR register.
353	 *
354	 * - Combinations that would require two additional levels of
355	 *   translation are not supported, e.g., VA_BITS==36 on 16k pages, or
356	 *   VA_BITS==39/4k pages with 5-level paging, where the input address
357	 *   requires more than 47 or 48 bits, respectively.
358	 */
359#if (VA_BITS < 48)
360#define IDMAP_PGD_ORDER	(VA_BITS - PGDIR_SHIFT)
361#define EXTRA_SHIFT	(PGDIR_SHIFT + PAGE_SHIFT - 3)
362
363	/*
364	 * If VA_BITS < 48, we have to configure an additional table level.
365	 * First, we have to verify our assumption that the current value of
366	 * VA_BITS was chosen such that all translation levels are fully
367	 * utilised, and that lowering T0SZ will always result in an additional
368	 * translation level to be configured.
369	 */
370#if VA_BITS != EXTRA_SHIFT
371#error "Mismatch between VA_BITS and page size/number of translation levels"
372#endif
373#else
374#define IDMAP_PGD_ORDER	(PHYS_MASK_SHIFT - PGDIR_SHIFT)
375#define EXTRA_SHIFT
376	/*
377	 * If VA_BITS == 48, we don't have to configure an additional
378	 * translation level, but the top-level table has more entries.
379	 */
380#endif
381	adrp	x0, init_idmap_pg_dir
382	adrp	x3, _text
383	adrp	x6, _end + MAX_FDT_SIZE + SWAPPER_BLOCK_SIZE
384	mov	x7, SWAPPER_RX_MMUFLAGS
385
386	map_memory x0, x1, x3, x6, x7, x3, IDMAP_PGD_ORDER, x10, x11, x12, x13, x14, EXTRA_SHIFT
387
388	/* Remap the kernel page tables r/w in the ID map */
389	adrp	x1, _text
390	adrp	x2, init_pg_dir
391	adrp	x3, init_pg_end
392	bic	x4, x2, #SWAPPER_BLOCK_SIZE - 1
393	mov	x5, SWAPPER_RW_MMUFLAGS
394	mov	x6, #SWAPPER_BLOCK_SHIFT
395	bl	remap_region
396
397	/* Remap the FDT after the kernel image */
398	adrp	x1, _text
399	adrp	x22, _end + SWAPPER_BLOCK_SIZE
400	bic	x2, x22, #SWAPPER_BLOCK_SIZE - 1
401	bfi	x22, x21, #0, #SWAPPER_BLOCK_SHIFT		// remapped FDT address
402	add	x3, x2, #MAX_FDT_SIZE + SWAPPER_BLOCK_SIZE
403	bic	x4, x21, #SWAPPER_BLOCK_SIZE - 1
404	mov	x5, SWAPPER_RW_MMUFLAGS
405	mov	x6, #SWAPPER_BLOCK_SHIFT
406	bl	remap_region
407
408	/*
409	 * Since the page tables have been populated with non-cacheable
410	 * accesses (MMU disabled), invalidate those tables again to
411	 * remove any speculatively loaded cache lines.
412	 */
413	cbnz	x19, 0f				// skip cache invalidation if MMU is on
414	dmb	sy
415
416	adrp	x0, init_idmap_pg_dir
417	adrp	x1, init_idmap_pg_end
418	bl	dcache_inval_poc
4190:	ret	x28
420SYM_FUNC_END(create_idmap)
421
422SYM_FUNC_START_LOCAL(create_kernel_mapping)
423	adrp	x0, init_pg_dir
424	mov_q	x5, KIMAGE_VADDR		// compile time __va(_text)
425#ifdef CONFIG_RELOCATABLE
426	add	x5, x5, x23			// add KASLR displacement
427#endif
428	adrp	x6, _end			// runtime __pa(_end)
429	adrp	x3, _text			// runtime __pa(_text)
430	sub	x6, x6, x3			// _end - _text
431	add	x6, x6, x5			// runtime __va(_end)
432	mov	x7, SWAPPER_RW_MMUFLAGS
433
434	map_memory x0, x1, x5, x6, x7, x3, (VA_BITS - PGDIR_SHIFT), x10, x11, x12, x13, x14
435
436	dsb	ishst				// sync with page table walker
437	ret
438SYM_FUNC_END(create_kernel_mapping)
439
440	/*
441	 * Initialize CPU registers with task-specific and cpu-specific context.
442	 *
443	 * Create a final frame record at task_pt_regs(current)->stackframe, so
444	 * that the unwinder can identify the final frame record of any task by
445	 * its location in the task stack. We reserve the entire pt_regs space
446	 * for consistency with user tasks and kthreads.
447	 */
448	.macro	init_cpu_task tsk, tmp1, tmp2
449	msr	sp_el0, \tsk
450
451	ldr	\tmp1, [\tsk, #TSK_STACK]
452	add	sp, \tmp1, #THREAD_SIZE
453	sub	sp, sp, #PT_REGS_SIZE
454
455	stp	xzr, xzr, [sp, #S_STACKFRAME]
456	add	x29, sp, #S_STACKFRAME
457
458	scs_load_current
459
460	adr_l	\tmp1, __per_cpu_offset
461	ldr	w\tmp2, [\tsk, #TSK_TI_CPU]
462	ldr	\tmp1, [\tmp1, \tmp2, lsl #3]
463	set_this_cpu_offset \tmp1
464	.endm
465
466/*
467 * The following fragment of code is executed with the MMU enabled.
468 *
469 *   x0 = __pa(KERNEL_START)
470 */
471SYM_FUNC_START_LOCAL(__primary_switched)
472	adr_l	x4, init_task
473	init_cpu_task x4, x5, x6
474
475	adr_l	x8, vectors			// load VBAR_EL1 with virtual
476	msr	vbar_el1, x8			// vector table address
477	isb
478
479	stp	x29, x30, [sp, #-16]!
480	mov	x29, sp
481
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, x0			// the kernel virtual and
486	str_l	x4, kimage_voffset, x5		// physical mappings
487
488	mov	x0, x20
489	bl	set_cpu_boot_mode_flag
490
491	// Clear BSS
492	adr_l	x0, __bss_start
493	mov	x1, xzr
494	adr_l	x2, __bss_stop
495	sub	x2, x2, x0
496	bl	__pi_memset
497	dsb	ishst				// Make zero page visible to PTW
498
499#if VA_BITS > 48
500	adr_l	x8, vabits_actual		// Set this early so KASAN early init
501	str	x25, [x8]			// ... observes the correct value
502	dc	civac, x8			// Make visible to booting secondaries
503#endif
504
505#ifdef CONFIG_RANDOMIZE_BASE
506	adrp	x5, memstart_offset_seed	// Save KASLR linear map seed
507	strh	w24, [x5, :lo12:memstart_offset_seed]
508#endif
509#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
510	bl	kasan_early_init
511#endif
512	mov	x0, x21				// pass FDT address in x0
513	bl	early_fdt_map			// Try mapping the FDT early
514	mov	x0, x20				// pass the full boot status
515	bl	init_feature_override		// Parse cpu feature overrides
516#ifdef CONFIG_UNWIND_PATCH_PAC_INTO_SCS
517	bl	scs_patch_vmlinux
518#endif
519	mov	x0, x20
520	bl	finalise_el2			// Prefer VHE if possible
521	ldp	x29, x30, [sp], #16
522	bl	start_kernel
523	ASM_BUG()
524SYM_FUNC_END(__primary_switched)
525
526/*
527 * end early head section, begin head code that is also used for
528 * hotplug and needs to have the same protections as the text region
529 */
530	.section ".idmap.text","awx"
531
532/*
533 * Starting from EL2 or EL1, configure the CPU to execute at the highest
534 * reachable EL supported by the kernel in a chosen default state. If dropping
535 * from EL2 to EL1, configure EL2 before configuring EL1.
536 *
537 * Since we cannot always rely on ERET synchronizing writes to sysregs (e.g. if
538 * SCTLR_ELx.EOS is clear), we place an ISB prior to ERET.
539 *
540 * Returns either BOOT_CPU_MODE_EL1 or BOOT_CPU_MODE_EL2 in x0 if
541 * booted in EL1 or EL2 respectively, with the top 32 bits containing
542 * potential context flags. These flags are *not* stored in __boot_cpu_mode.
543 *
544 * x0: whether we are being called from the primary boot path with the MMU on
545 */
546SYM_FUNC_START(init_kernel_el)
547	mrs	x1, CurrentEL
548	cmp	x1, #CurrentEL_EL2
549	b.eq	init_el2
550
551SYM_INNER_LABEL(init_el1, SYM_L_LOCAL)
552	mov_q	x0, INIT_SCTLR_EL1_MMU_OFF
553	pre_disable_mmu_workaround
554	msr	sctlr_el1, x0
555	isb
556	mov_q	x0, INIT_PSTATE_EL1
557	msr	spsr_el1, x0
558	msr	elr_el1, lr
559	mov	w0, #BOOT_CPU_MODE_EL1
560	eret
561
562SYM_INNER_LABEL(init_el2, SYM_L_LOCAL)
563	msr	elr_el2, lr
564
565	// clean all HYP code to the PoC if we booted at EL2 with the MMU on
566	cbz	x0, 0f
567	adrp	x0, __hyp_idmap_text_start
568	adr_l	x1, __hyp_text_end
569	bl	dcache_clean_poc
5700:
571	mov_q	x0, HCR_HOST_NVHE_FLAGS
572	msr	hcr_el2, x0
573	isb
574
575	init_el2_state
576
577	/* Hypervisor stub */
578	adr_l	x0, __hyp_stub_vectors
579	msr	vbar_el2, x0
580	isb
581
582	mov_q	x1, INIT_SCTLR_EL1_MMU_OFF
583
584	/*
585	 * Fruity CPUs seem to have HCR_EL2.E2H set to RES1,
586	 * making it impossible to start in nVHE mode. Is that
587	 * compliant with the architecture? Absolutely not!
588	 */
589	mrs	x0, hcr_el2
590	and	x0, x0, #HCR_E2H
591	cbz	x0, 1f
592
593	/* Set a sane SCTLR_EL1, the VHE way */
594	pre_disable_mmu_workaround
595	msr_s	SYS_SCTLR_EL12, x1
596	mov	x2, #BOOT_CPU_FLAG_E2H
597	b	2f
598
5991:
600	pre_disable_mmu_workaround
601	msr	sctlr_el1, x1
602	mov	x2, xzr
6032:
604	mov	w0, #BOOT_CPU_MODE_EL2
605	orr	x0, x0, x2
606	eret
607SYM_FUNC_END(init_kernel_el)
608
609	/*
610	 * This provides a "holding pen" for platforms to hold all secondary
611	 * cores are held until we're ready for them to initialise.
612	 */
613SYM_FUNC_START(secondary_holding_pen)
614	mov	x0, xzr
615	bl	init_kernel_el			// w0=cpu_boot_mode
616	mrs	x2, mpidr_el1
617	mov_q	x1, MPIDR_HWID_BITMASK
618	and	x2, x2, x1
619	adr_l	x3, secondary_holding_pen_release
620pen:	ldr	x4, [x3]
621	cmp	x4, x2
622	b.eq	secondary_startup
623	wfe
624	b	pen
625SYM_FUNC_END(secondary_holding_pen)
626
627	/*
628	 * Secondary entry point that jumps straight into the kernel. Only to
629	 * be used where CPUs are brought online dynamically by the kernel.
630	 */
631SYM_FUNC_START(secondary_entry)
632	mov	x0, xzr
633	bl	init_kernel_el			// w0=cpu_boot_mode
634	b	secondary_startup
635SYM_FUNC_END(secondary_entry)
636
637SYM_FUNC_START_LOCAL(secondary_startup)
638	/*
639	 * Common entry point for secondary CPUs.
640	 */
641	mov	x20, x0				// preserve boot mode
642	bl	__cpu_secondary_check52bitva
643#if VA_BITS > 48
644	ldr_l	x0, vabits_actual
645#endif
646	bl	__cpu_setup			// initialise processor
647	adrp	x1, swapper_pg_dir
648	adrp	x2, idmap_pg_dir
649	bl	__enable_mmu
650	ldr	x8, =__secondary_switched
651	br	x8
652SYM_FUNC_END(secondary_startup)
653
654	.text
655SYM_FUNC_START_LOCAL(__secondary_switched)
656	mov	x0, x20
657	bl	set_cpu_boot_mode_flag
658
659	mov	x0, x20
660	bl	finalise_el2
661
662	str_l	xzr, __early_cpu_boot_status, x3
663	adr_l	x5, vectors
664	msr	vbar_el1, x5
665	isb
666
667	adr_l	x0, secondary_data
668	ldr	x2, [x0, #CPU_BOOT_TASK]
669	cbz	x2, __secondary_too_slow
670
671	init_cpu_task x2, x1, x3
672
673#ifdef CONFIG_ARM64_PTR_AUTH
674	ptrauth_keys_init_cpu x2, x3, x4, x5
675#endif
676
677	bl	secondary_start_kernel
678	ASM_BUG()
679SYM_FUNC_END(__secondary_switched)
680
681SYM_FUNC_START_LOCAL(__secondary_too_slow)
682	wfe
683	wfi
684	b	__secondary_too_slow
685SYM_FUNC_END(__secondary_too_slow)
686
687/*
688 * Sets the __boot_cpu_mode flag depending on the CPU boot mode passed
689 * in w0. See arch/arm64/include/asm/virt.h for more info.
690 */
691SYM_FUNC_START_LOCAL(set_cpu_boot_mode_flag)
692	adr_l	x1, __boot_cpu_mode
693	cmp	w0, #BOOT_CPU_MODE_EL2
694	b.ne	1f
695	add	x1, x1, #4
6961:	str	w0, [x1]			// Save CPU boot mode
697	ret
698SYM_FUNC_END(set_cpu_boot_mode_flag)
699
700/*
701 * The booting CPU updates the failed status @__early_cpu_boot_status,
702 * with MMU turned off.
703 *
704 * update_early_cpu_boot_status tmp, status
705 *  - Corrupts tmp1, tmp2
706 *  - Writes 'status' to __early_cpu_boot_status and makes sure
707 *    it is committed to memory.
708 */
709
710	.macro	update_early_cpu_boot_status status, tmp1, tmp2
711	mov	\tmp2, #\status
712	adr_l	\tmp1, __early_cpu_boot_status
713	str	\tmp2, [\tmp1]
714	dmb	sy
715	dc	ivac, \tmp1			// Invalidate potentially stale cache line
716	.endm
717
718/*
719 * Enable the MMU.
720 *
721 *  x0  = SCTLR_EL1 value for turning on the MMU.
722 *  x1  = TTBR1_EL1 value
723 *  x2  = ID map root table address
724 *
725 * Returns to the caller via x30/lr. This requires the caller to be covered
726 * by the .idmap.text section.
727 *
728 * Checks if the selected granule size is supported by the CPU.
729 * If it isn't, park the CPU
730 */
731	.section ".idmap.text","awx"
732SYM_FUNC_START(__enable_mmu)
733	mrs	x3, ID_AA64MMFR0_EL1
734	ubfx	x3, x3, #ID_AA64MMFR0_EL1_TGRAN_SHIFT, 4
735	cmp     x3, #ID_AA64MMFR0_EL1_TGRAN_SUPPORTED_MIN
736	b.lt    __no_granule_support
737	cmp     x3, #ID_AA64MMFR0_EL1_TGRAN_SUPPORTED_MAX
738	b.gt    __no_granule_support
739	phys_to_ttbr x2, x2
740	msr	ttbr0_el1, x2			// load TTBR0
741	load_ttbr1 x1, x1, x3
742
743	set_sctlr_el1	x0
744
745	ret
746SYM_FUNC_END(__enable_mmu)
747
748SYM_FUNC_START(__cpu_secondary_check52bitva)
749#if VA_BITS > 48
750	ldr_l	x0, vabits_actual
751	cmp	x0, #52
752	b.ne	2f
753
754	mrs_s	x0, SYS_ID_AA64MMFR2_EL1
755	and	x0, x0, #(0xf << ID_AA64MMFR2_EL1_VARange_SHIFT)
756	cbnz	x0, 2f
757
758	update_early_cpu_boot_status \
759		CPU_STUCK_IN_KERNEL | CPU_STUCK_REASON_52_BIT_VA, x0, x1
7601:	wfe
761	wfi
762	b	1b
763
764#endif
7652:	ret
766SYM_FUNC_END(__cpu_secondary_check52bitva)
767
768SYM_FUNC_START_LOCAL(__no_granule_support)
769	/* Indicate that this CPU can't boot and is stuck in the kernel */
770	update_early_cpu_boot_status \
771		CPU_STUCK_IN_KERNEL | CPU_STUCK_REASON_NO_GRAN, x1, x2
7721:
773	wfe
774	wfi
775	b	1b
776SYM_FUNC_END(__no_granule_support)
777
778#ifdef CONFIG_RELOCATABLE
779SYM_FUNC_START_LOCAL(__relocate_kernel)
780	/*
781	 * Iterate over each entry in the relocation table, and apply the
782	 * relocations in place.
783	 */
784	adr_l	x9, __rela_start
785	adr_l	x10, __rela_end
786	mov_q	x11, KIMAGE_VADDR		// default virtual offset
787	add	x11, x11, x23			// actual virtual offset
788
7890:	cmp	x9, x10
790	b.hs	1f
791	ldp	x12, x13, [x9], #24
792	ldr	x14, [x9, #-8]
793	cmp	w13, #R_AARCH64_RELATIVE
794	b.ne	0b
795	add	x14, x14, x23			// relocate
796	str	x14, [x12, x23]
797	b	0b
798
7991:
800#ifdef CONFIG_RELR
801	/*
802	 * Apply RELR relocations.
803	 *
804	 * RELR is a compressed format for storing relative relocations. The
805	 * encoded sequence of entries looks like:
806	 * [ AAAAAAAA BBBBBBB1 BBBBBBB1 ... AAAAAAAA BBBBBB1 ... ]
807	 *
808	 * i.e. start with an address, followed by any number of bitmaps. The
809	 * address entry encodes 1 relocation. The subsequent bitmap entries
810	 * encode up to 63 relocations each, at subsequent offsets following
811	 * the last address entry.
812	 *
813	 * The bitmap entries must have 1 in the least significant bit. The
814	 * assumption here is that an address cannot have 1 in lsb. Odd
815	 * addresses are not supported. Any odd addresses are stored in the RELA
816	 * section, which is handled above.
817	 *
818	 * Excluding the least significant bit in the bitmap, each non-zero
819	 * bit in the bitmap represents a relocation to be applied to
820	 * a corresponding machine word that follows the base address
821	 * word. The second least significant bit represents the machine
822	 * word immediately following the initial address, and each bit
823	 * that follows represents the next word, in linear order. As such,
824	 * a single bitmap can encode up to 63 relocations in a 64-bit object.
825	 *
826	 * In this implementation we store the address of the next RELR table
827	 * entry in x9, the address being relocated by the current address or
828	 * bitmap entry in x13 and the address being relocated by the current
829	 * bit in x14.
830	 */
831	adr_l	x9, __relr_start
832	adr_l	x10, __relr_end
833
8342:	cmp	x9, x10
835	b.hs	7f
836	ldr	x11, [x9], #8
837	tbnz	x11, #0, 3f			// branch to handle bitmaps
838	add	x13, x11, x23
839	ldr	x12, [x13]			// relocate address entry
840	add	x12, x12, x23
841	str	x12, [x13], #8			// adjust to start of bitmap
842	b	2b
843
8443:	mov	x14, x13
8454:	lsr	x11, x11, #1
846	cbz	x11, 6f
847	tbz	x11, #0, 5f			// skip bit if not set
848	ldr	x12, [x14]			// relocate bit
849	add	x12, x12, x23
850	str	x12, [x14]
851
8525:	add	x14, x14, #8			// move to next bit's address
853	b	4b
854
8556:	/*
856	 * Move to the next bitmap's address. 8 is the word size, and 63 is the
857	 * number of significant bits in a bitmap entry.
858	 */
859	add	x13, x13, #(8 * 63)
860	b	2b
861
8627:
863#endif
864	ret
865
866SYM_FUNC_END(__relocate_kernel)
867#endif
868
869SYM_FUNC_START_LOCAL(__primary_switch)
870	adrp	x1, reserved_pg_dir
871	adrp	x2, init_idmap_pg_dir
872	bl	__enable_mmu
873#ifdef CONFIG_RELOCATABLE
874	adrp	x23, KERNEL_START
875	and	x23, x23, MIN_KIMG_ALIGN - 1
876#ifdef CONFIG_RANDOMIZE_BASE
877	mov	x0, x22
878	adrp	x1, init_pg_end
879	mov	sp, x1
880	mov	x29, xzr
881	bl	__pi_kaslr_early_init
882	and	x24, x0, #SZ_2M - 1		// capture memstart offset seed
883	bic	x0, x0, #SZ_2M - 1
884	orr	x23, x23, x0			// record kernel offset
885#endif
886#endif
887	bl	clear_page_tables
888	bl	create_kernel_mapping
889
890	adrp	x1, init_pg_dir
891	load_ttbr1 x1, x1, x2
892#ifdef CONFIG_RELOCATABLE
893	bl	__relocate_kernel
894#endif
895	ldr	x8, =__primary_switched
896	adrp	x0, KERNEL_START		// __pa(KERNEL_START)
897	br	x8
898SYM_FUNC_END(__primary_switch)
899