xref: /openbmc/linux/arch/arm64/kernel/head.S (revision 6c8c1406)
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	__INIT
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	 *  x20        primary_entry() .. __primary_switch()    CPU boot mode
81	 *  x21        primary_entry() .. start_kernel()        FDT pointer passed at boot in x0
82	 *  x22        create_idmap() .. start_kernel()         ID map VA of the DT blob
83	 *  x23        primary_entry() .. start_kernel()        physical misalignment/KASLR offset
84	 *  x24        __primary_switch()                       linear map KASLR seed
85	 *  x25        primary_entry() .. start_kernel()        supported VA size
86	 *  x28        create_idmap()                           callee preserved temp register
87	 */
88SYM_CODE_START(primary_entry)
89	bl	preserve_boot_args
90	bl	init_kernel_el			// w0=cpu_boot_mode
91	mov	x20, x0
92	bl	create_idmap
93
94	/*
95	 * The following calls CPU setup code, see arch/arm64/mm/proc.S for
96	 * details.
97	 * On return, the CPU will be ready for the MMU to be turned on and
98	 * the TCR will have been set.
99	 */
100#if VA_BITS > 48
101	mrs_s	x0, SYS_ID_AA64MMFR2_EL1
102	tst	x0, #0xf << ID_AA64MMFR2_EL1_VARange_SHIFT
103	mov	x0, #VA_BITS
104	mov	x25, #VA_BITS_MIN
105	csel	x25, x25, x0, eq
106	mov	x0, x25
107#endif
108	bl	__cpu_setup			// initialise processor
109	b	__primary_switch
110SYM_CODE_END(primary_entry)
111
112/*
113 * Preserve the arguments passed by the bootloader in x0 .. x3
114 */
115SYM_CODE_START_LOCAL(preserve_boot_args)
116	mov	x21, x0				// x21=FDT
117
118	adr_l	x0, boot_args			// record the contents of
119	stp	x21, x1, [x0]			// x0 .. x3 at kernel entry
120	stp	x2, x3, [x0, #16]
121
122	dmb	sy				// needed before dc ivac with
123						// MMU off
124
125	add	x1, x0, #0x20			// 4 x 8 bytes
126	b	dcache_inval_poc		// tail call
127SYM_CODE_END(preserve_boot_args)
128
129SYM_FUNC_START_LOCAL(clear_page_tables)
130	/*
131	 * Clear the init page tables.
132	 */
133	adrp	x0, init_pg_dir
134	adrp	x1, init_pg_end
135	sub	x2, x1, x0
136	mov	x1, xzr
137	b	__pi_memset			// tail call
138SYM_FUNC_END(clear_page_tables)
139
140/*
141 * Macro to populate page table entries, these entries can be pointers to the next level
142 * or last level entries pointing to physical memory.
143 *
144 *	tbl:	page table address
145 *	rtbl:	pointer to page table or physical memory
146 *	index:	start index to write
147 *	eindex:	end index to write - [index, eindex] written to
148 *	flags:	flags for pagetable entry to or in
149 *	inc:	increment to rtbl between each entry
150 *	tmp1:	temporary variable
151 *
152 * Preserves:	tbl, eindex, flags, inc
153 * Corrupts:	index, tmp1
154 * Returns:	rtbl
155 */
156	.macro populate_entries, tbl, rtbl, index, eindex, flags, inc, tmp1
157.Lpe\@:	phys_to_pte \tmp1, \rtbl
158	orr	\tmp1, \tmp1, \flags	// tmp1 = table entry
159	str	\tmp1, [\tbl, \index, lsl #3]
160	add	\rtbl, \rtbl, \inc	// rtbl = pa next level
161	add	\index, \index, #1
162	cmp	\index, \eindex
163	b.ls	.Lpe\@
164	.endm
165
166/*
167 * Compute indices of table entries from virtual address range. If multiple entries
168 * were needed in the previous page table level then the next page table level is assumed
169 * to be composed of multiple pages. (This effectively scales the end index).
170 *
171 *	vstart:	virtual address of start of range
172 *	vend:	virtual address of end of range - we map [vstart, vend]
173 *	shift:	shift used to transform virtual address into index
174 *	order:  #imm 2log(number of entries in page table)
175 *	istart:	index in table corresponding to vstart
176 *	iend:	index in table corresponding to vend
177 *	count:	On entry: how many extra entries were required in previous level, scales
178 *			  our end index.
179 *		On exit: returns how many extra entries required for next page table level
180 *
181 * Preserves:	vstart, vend
182 * Returns:	istart, iend, count
183 */
184	.macro compute_indices, vstart, vend, shift, order, istart, iend, count
185	ubfx	\istart, \vstart, \shift, \order
186	ubfx	\iend, \vend, \shift, \order
187	add	\iend, \iend, \count, lsl \order
188	sub	\count, \iend, \istart
189	.endm
190
191/*
192 * Map memory for specified virtual address range. Each level of page table needed supports
193 * multiple entries. If a level requires n entries the next page table level is assumed to be
194 * formed from n pages.
195 *
196 *	tbl:	location of page table
197 *	rtbl:	address to be used for first level page table entry (typically tbl + PAGE_SIZE)
198 *	vstart:	virtual address of start of range
199 *	vend:	virtual address of end of range - we map [vstart, vend - 1]
200 *	flags:	flags to use to map last level entries
201 *	phys:	physical address corresponding to vstart - physical memory is contiguous
202 *	order:  #imm 2log(number of entries in PGD table)
203 *
204 * If extra_shift is set, an extra level will be populated if the end address does
205 * not fit in 'extra_shift' bits. This assumes vend is in the TTBR0 range.
206 *
207 * Temporaries:	istart, iend, tmp, count, sv - these need to be different registers
208 * Preserves:	vstart, flags
209 * Corrupts:	tbl, rtbl, vend, istart, iend, tmp, count, sv
210 */
211	.macro map_memory, tbl, rtbl, vstart, vend, flags, phys, order, istart, iend, tmp, count, sv, extra_shift
212	sub \vend, \vend, #1
213	add \rtbl, \tbl, #PAGE_SIZE
214	mov \count, #0
215
216	.ifnb	\extra_shift
217	tst	\vend, #~((1 << (\extra_shift)) - 1)
218	b.eq	.L_\@
219	compute_indices \vstart, \vend, #\extra_shift, #(PAGE_SHIFT - 3), \istart, \iend, \count
220	mov \sv, \rtbl
221	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
222	mov \tbl, \sv
223	.endif
224.L_\@:
225	compute_indices \vstart, \vend, #PGDIR_SHIFT, #\order, \istart, \iend, \count
226	mov \sv, \rtbl
227	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
228	mov \tbl, \sv
229
230#if SWAPPER_PGTABLE_LEVELS > 3
231	compute_indices \vstart, \vend, #PUD_SHIFT, #(PAGE_SHIFT - 3), \istart, \iend, \count
232	mov \sv, \rtbl
233	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
234	mov \tbl, \sv
235#endif
236
237#if SWAPPER_PGTABLE_LEVELS > 2
238	compute_indices \vstart, \vend, #SWAPPER_TABLE_SHIFT, #(PAGE_SHIFT - 3), \istart, \iend, \count
239	mov \sv, \rtbl
240	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
241	mov \tbl, \sv
242#endif
243
244	compute_indices \vstart, \vend, #SWAPPER_BLOCK_SHIFT, #(PAGE_SHIFT - 3), \istart, \iend, \count
245	bic \rtbl, \phys, #SWAPPER_BLOCK_SIZE - 1
246	populate_entries \tbl, \rtbl, \istart, \iend, \flags, #SWAPPER_BLOCK_SIZE, \tmp
247	.endm
248
249/*
250 * Remap a subregion created with the map_memory macro with modified attributes
251 * or output address. The entire remapped region must have been covered in the
252 * invocation of map_memory.
253 *
254 * x0: last level table address (returned in first argument to map_memory)
255 * x1: start VA of the existing mapping
256 * x2: start VA of the region to update
257 * x3: end VA of the region to update (exclusive)
258 * x4: start PA associated with the region to update
259 * x5: attributes to set on the updated region
260 * x6: order of the last level mappings
261 */
262SYM_FUNC_START_LOCAL(remap_region)
263	sub	x3, x3, #1		// make end inclusive
264
265	// Get the index offset for the start of the last level table
266	lsr	x1, x1, x6
267	bfi	x1, xzr, #0, #PAGE_SHIFT - 3
268
269	// Derive the start and end indexes into the last level table
270	// associated with the provided region
271	lsr	x2, x2, x6
272	lsr	x3, x3, x6
273	sub	x2, x2, x1
274	sub	x3, x3, x1
275
276	mov	x1, #1
277	lsl	x6, x1, x6		// block size at this level
278
279	populate_entries x0, x4, x2, x3, x5, x6, x7
280	ret
281SYM_FUNC_END(remap_region)
282
283SYM_FUNC_START_LOCAL(create_idmap)
284	mov	x28, lr
285	/*
286	 * The ID map carries a 1:1 mapping of the physical address range
287	 * covered by the loaded image, which could be anywhere in DRAM. This
288	 * means that the required size of the VA (== PA) space is decided at
289	 * boot time, and could be more than the configured size of the VA
290	 * space for ordinary kernel and user space mappings.
291	 *
292	 * There are three cases to consider here:
293	 * - 39 <= VA_BITS < 48, and the ID map needs up to 48 VA bits to cover
294	 *   the placement of the image. In this case, we configure one extra
295	 *   level of translation on the fly for the ID map only. (This case
296	 *   also covers 42-bit VA/52-bit PA on 64k pages).
297	 *
298	 * - VA_BITS == 48, and the ID map needs more than 48 VA bits. This can
299	 *   only happen when using 64k pages, in which case we need to extend
300	 *   the root level table rather than add a level. Note that we can
301	 *   treat this case as 'always extended' as long as we take care not
302	 *   to program an unsupported T0SZ value into the TCR register.
303	 *
304	 * - Combinations that would require two additional levels of
305	 *   translation are not supported, e.g., VA_BITS==36 on 16k pages, or
306	 *   VA_BITS==39/4k pages with 5-level paging, where the input address
307	 *   requires more than 47 or 48 bits, respectively.
308	 */
309#if (VA_BITS < 48)
310#define IDMAP_PGD_ORDER	(VA_BITS - PGDIR_SHIFT)
311#define EXTRA_SHIFT	(PGDIR_SHIFT + PAGE_SHIFT - 3)
312
313	/*
314	 * If VA_BITS < 48, we have to configure an additional table level.
315	 * First, we have to verify our assumption that the current value of
316	 * VA_BITS was chosen such that all translation levels are fully
317	 * utilised, and that lowering T0SZ will always result in an additional
318	 * translation level to be configured.
319	 */
320#if VA_BITS != EXTRA_SHIFT
321#error "Mismatch between VA_BITS and page size/number of translation levels"
322#endif
323#else
324#define IDMAP_PGD_ORDER	(PHYS_MASK_SHIFT - PGDIR_SHIFT)
325#define EXTRA_SHIFT
326	/*
327	 * If VA_BITS == 48, we don't have to configure an additional
328	 * translation level, but the top-level table has more entries.
329	 */
330#endif
331	adrp	x0, init_idmap_pg_dir
332	adrp	x3, _text
333	adrp	x6, _end + MAX_FDT_SIZE + SWAPPER_BLOCK_SIZE
334	mov	x7, SWAPPER_RX_MMUFLAGS
335
336	map_memory x0, x1, x3, x6, x7, x3, IDMAP_PGD_ORDER, x10, x11, x12, x13, x14, EXTRA_SHIFT
337
338	/* Remap the kernel page tables r/w in the ID map */
339	adrp	x1, _text
340	adrp	x2, init_pg_dir
341	adrp	x3, init_pg_end
342	bic	x4, x2, #SWAPPER_BLOCK_SIZE - 1
343	mov	x5, SWAPPER_RW_MMUFLAGS
344	mov	x6, #SWAPPER_BLOCK_SHIFT
345	bl	remap_region
346
347	/* Remap the FDT after the kernel image */
348	adrp	x1, _text
349	adrp	x22, _end + SWAPPER_BLOCK_SIZE
350	bic	x2, x22, #SWAPPER_BLOCK_SIZE - 1
351	bfi	x22, x21, #0, #SWAPPER_BLOCK_SHIFT		// remapped FDT address
352	add	x3, x2, #MAX_FDT_SIZE + SWAPPER_BLOCK_SIZE
353	bic	x4, x21, #SWAPPER_BLOCK_SIZE - 1
354	mov	x5, SWAPPER_RW_MMUFLAGS
355	mov	x6, #SWAPPER_BLOCK_SHIFT
356	bl	remap_region
357
358	/*
359	 * Since the page tables have been populated with non-cacheable
360	 * accesses (MMU disabled), invalidate those tables again to
361	 * remove any speculatively loaded cache lines.
362	 */
363	dmb	sy
364
365	adrp	x0, init_idmap_pg_dir
366	adrp	x1, init_idmap_pg_end
367	bl	dcache_inval_poc
368	ret	x28
369SYM_FUNC_END(create_idmap)
370
371SYM_FUNC_START_LOCAL(create_kernel_mapping)
372	adrp	x0, init_pg_dir
373	mov_q	x5, KIMAGE_VADDR		// compile time __va(_text)
374#ifdef CONFIG_RELOCATABLE
375	add	x5, x5, x23			// add KASLR displacement
376#endif
377	adrp	x6, _end			// runtime __pa(_end)
378	adrp	x3, _text			// runtime __pa(_text)
379	sub	x6, x6, x3			// _end - _text
380	add	x6, x6, x5			// runtime __va(_end)
381	mov	x7, SWAPPER_RW_MMUFLAGS
382
383	map_memory x0, x1, x5, x6, x7, x3, (VA_BITS - PGDIR_SHIFT), x10, x11, x12, x13, x14
384
385	dsb	ishst				// sync with page table walker
386	ret
387SYM_FUNC_END(create_kernel_mapping)
388
389	/*
390	 * Initialize CPU registers with task-specific and cpu-specific context.
391	 *
392	 * Create a final frame record at task_pt_regs(current)->stackframe, so
393	 * that the unwinder can identify the final frame record of any task by
394	 * its location in the task stack. We reserve the entire pt_regs space
395	 * for consistency with user tasks and kthreads.
396	 */
397	.macro	init_cpu_task tsk, tmp1, tmp2
398	msr	sp_el0, \tsk
399
400	ldr	\tmp1, [\tsk, #TSK_STACK]
401	add	sp, \tmp1, #THREAD_SIZE
402	sub	sp, sp, #PT_REGS_SIZE
403
404	stp	xzr, xzr, [sp, #S_STACKFRAME]
405	add	x29, sp, #S_STACKFRAME
406
407	scs_load \tsk
408
409	adr_l	\tmp1, __per_cpu_offset
410	ldr	w\tmp2, [\tsk, #TSK_TI_CPU]
411	ldr	\tmp1, [\tmp1, \tmp2, lsl #3]
412	set_this_cpu_offset \tmp1
413	.endm
414
415/*
416 * The following fragment of code is executed with the MMU enabled.
417 *
418 *   x0 = __pa(KERNEL_START)
419 */
420SYM_FUNC_START_LOCAL(__primary_switched)
421	adr_l	x4, init_task
422	init_cpu_task x4, x5, x6
423
424	adr_l	x8, vectors			// load VBAR_EL1 with virtual
425	msr	vbar_el1, x8			// vector table address
426	isb
427
428	stp	x29, x30, [sp, #-16]!
429	mov	x29, sp
430
431	str_l	x21, __fdt_pointer, x5		// Save FDT pointer
432
433	ldr_l	x4, kimage_vaddr		// Save the offset between
434	sub	x4, x4, x0			// the kernel virtual and
435	str_l	x4, kimage_voffset, x5		// physical mappings
436
437	mov	x0, x20
438	bl	set_cpu_boot_mode_flag
439
440	// Clear BSS
441	adr_l	x0, __bss_start
442	mov	x1, xzr
443	adr_l	x2, __bss_stop
444	sub	x2, x2, x0
445	bl	__pi_memset
446	dsb	ishst				// Make zero page visible to PTW
447
448#if VA_BITS > 48
449	adr_l	x8, vabits_actual		// Set this early so KASAN early init
450	str	x25, [x8]			// ... observes the correct value
451	dc	civac, x8			// Make visible to booting secondaries
452#endif
453
454#ifdef CONFIG_RANDOMIZE_BASE
455	adrp	x5, memstart_offset_seed	// Save KASLR linear map seed
456	strh	w24, [x5, :lo12:memstart_offset_seed]
457#endif
458#if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
459	bl	kasan_early_init
460#endif
461	mov	x0, x21				// pass FDT address in x0
462	bl	early_fdt_map			// Try mapping the FDT early
463	mov	x0, x20				// pass the full boot status
464	bl	init_feature_override		// Parse cpu feature overrides
465	mov	x0, x20
466	bl	finalise_el2			// Prefer VHE if possible
467	ldp	x29, x30, [sp], #16
468	bl	start_kernel
469	ASM_BUG()
470SYM_FUNC_END(__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","awx"
477
478/*
479 * Starting from EL2 or EL1, configure the CPU to execute at the highest
480 * reachable EL supported by the kernel in a chosen default state. If dropping
481 * from EL2 to EL1, configure EL2 before configuring EL1.
482 *
483 * Since we cannot always rely on ERET synchronizing writes to sysregs (e.g. if
484 * SCTLR_ELx.EOS is clear), we place an ISB prior to ERET.
485 *
486 * Returns either BOOT_CPU_MODE_EL1 or BOOT_CPU_MODE_EL2 in x0 if
487 * booted in EL1 or EL2 respectively, with the top 32 bits containing
488 * potential context flags. These flags are *not* stored in __boot_cpu_mode.
489 */
490SYM_FUNC_START(init_kernel_el)
491	mrs	x0, CurrentEL
492	cmp	x0, #CurrentEL_EL2
493	b.eq	init_el2
494
495SYM_INNER_LABEL(init_el1, SYM_L_LOCAL)
496	mov_q	x0, INIT_SCTLR_EL1_MMU_OFF
497	msr	sctlr_el1, x0
498	isb
499	mov_q	x0, INIT_PSTATE_EL1
500	msr	spsr_el1, x0
501	msr	elr_el1, lr
502	mov	w0, #BOOT_CPU_MODE_EL1
503	eret
504
505SYM_INNER_LABEL(init_el2, SYM_L_LOCAL)
506	mov_q	x0, HCR_HOST_NVHE_FLAGS
507	msr	hcr_el2, x0
508	isb
509
510	init_el2_state
511
512	/* Hypervisor stub */
513	adr_l	x0, __hyp_stub_vectors
514	msr	vbar_el2, x0
515	isb
516
517	mov_q	x1, INIT_SCTLR_EL1_MMU_OFF
518
519	/*
520	 * Fruity CPUs seem to have HCR_EL2.E2H set to RES1,
521	 * making it impossible to start in nVHE mode. Is that
522	 * compliant with the architecture? Absolutely not!
523	 */
524	mrs	x0, hcr_el2
525	and	x0, x0, #HCR_E2H
526	cbz	x0, 1f
527
528	/* Set a sane SCTLR_EL1, the VHE way */
529	msr_s	SYS_SCTLR_EL12, x1
530	mov	x2, #BOOT_CPU_FLAG_E2H
531	b	2f
532
5331:
534	msr	sctlr_el1, x1
535	mov	x2, xzr
5362:
537	msr	elr_el2, lr
538	mov	w0, #BOOT_CPU_MODE_EL2
539	orr	x0, x0, x2
540	eret
541SYM_FUNC_END(init_kernel_el)
542
543/*
544 * Sets the __boot_cpu_mode flag depending on the CPU boot mode passed
545 * in w0. See arch/arm64/include/asm/virt.h for more info.
546 */
547SYM_FUNC_START_LOCAL(set_cpu_boot_mode_flag)
548	adr_l	x1, __boot_cpu_mode
549	cmp	w0, #BOOT_CPU_MODE_EL2
550	b.ne	1f
551	add	x1, x1, #4
5521:	str	w0, [x1]			// Save CPU boot mode
553	ret
554SYM_FUNC_END(set_cpu_boot_mode_flag)
555
556	/*
557	 * This provides a "holding pen" for platforms to hold all secondary
558	 * cores are held until we're ready for them to initialise.
559	 */
560SYM_FUNC_START(secondary_holding_pen)
561	bl	init_kernel_el			// w0=cpu_boot_mode
562	mrs	x2, mpidr_el1
563	mov_q	x1, MPIDR_HWID_BITMASK
564	and	x2, x2, x1
565	adr_l	x3, secondary_holding_pen_release
566pen:	ldr	x4, [x3]
567	cmp	x4, x2
568	b.eq	secondary_startup
569	wfe
570	b	pen
571SYM_FUNC_END(secondary_holding_pen)
572
573	/*
574	 * Secondary entry point that jumps straight into the kernel. Only to
575	 * be used where CPUs are brought online dynamically by the kernel.
576	 */
577SYM_FUNC_START(secondary_entry)
578	bl	init_kernel_el			// w0=cpu_boot_mode
579	b	secondary_startup
580SYM_FUNC_END(secondary_entry)
581
582SYM_FUNC_START_LOCAL(secondary_startup)
583	/*
584	 * Common entry point for secondary CPUs.
585	 */
586	mov	x20, x0				// preserve boot mode
587	bl	finalise_el2
588	bl	__cpu_secondary_check52bitva
589#if VA_BITS > 48
590	ldr_l	x0, vabits_actual
591#endif
592	bl	__cpu_setup			// initialise processor
593	adrp	x1, swapper_pg_dir
594	adrp	x2, idmap_pg_dir
595	bl	__enable_mmu
596	ldr	x8, =__secondary_switched
597	br	x8
598SYM_FUNC_END(secondary_startup)
599
600SYM_FUNC_START_LOCAL(__secondary_switched)
601	mov	x0, x20
602	bl	set_cpu_boot_mode_flag
603	str_l	xzr, __early_cpu_boot_status, x3
604	adr_l	x5, vectors
605	msr	vbar_el1, x5
606	isb
607
608	adr_l	x0, secondary_data
609	ldr	x2, [x0, #CPU_BOOT_TASK]
610	cbz	x2, __secondary_too_slow
611
612	init_cpu_task x2, x1, x3
613
614#ifdef CONFIG_ARM64_PTR_AUTH
615	ptrauth_keys_init_cpu x2, x3, x4, x5
616#endif
617
618	bl	secondary_start_kernel
619	ASM_BUG()
620SYM_FUNC_END(__secondary_switched)
621
622SYM_FUNC_START_LOCAL(__secondary_too_slow)
623	wfe
624	wfi
625	b	__secondary_too_slow
626SYM_FUNC_END(__secondary_too_slow)
627
628/*
629 * The booting CPU updates the failed status @__early_cpu_boot_status,
630 * with MMU turned off.
631 *
632 * update_early_cpu_boot_status tmp, status
633 *  - Corrupts tmp1, tmp2
634 *  - Writes 'status' to __early_cpu_boot_status and makes sure
635 *    it is committed to memory.
636 */
637
638	.macro	update_early_cpu_boot_status status, tmp1, tmp2
639	mov	\tmp2, #\status
640	adr_l	\tmp1, __early_cpu_boot_status
641	str	\tmp2, [\tmp1]
642	dmb	sy
643	dc	ivac, \tmp1			// Invalidate potentially stale cache line
644	.endm
645
646/*
647 * Enable the MMU.
648 *
649 *  x0  = SCTLR_EL1 value for turning on the MMU.
650 *  x1  = TTBR1_EL1 value
651 *  x2  = ID map root table address
652 *
653 * Returns to the caller via x30/lr. This requires the caller to be covered
654 * by the .idmap.text section.
655 *
656 * Checks if the selected granule size is supported by the CPU.
657 * If it isn't, park the CPU
658 */
659SYM_FUNC_START(__enable_mmu)
660	mrs	x3, ID_AA64MMFR0_EL1
661	ubfx	x3, x3, #ID_AA64MMFR0_EL1_TGRAN_SHIFT, 4
662	cmp     x3, #ID_AA64MMFR0_EL1_TGRAN_SUPPORTED_MIN
663	b.lt    __no_granule_support
664	cmp     x3, #ID_AA64MMFR0_EL1_TGRAN_SUPPORTED_MAX
665	b.gt    __no_granule_support
666	phys_to_ttbr x2, x2
667	msr	ttbr0_el1, x2			// load TTBR0
668	load_ttbr1 x1, x1, x3
669
670	set_sctlr_el1	x0
671
672	ret
673SYM_FUNC_END(__enable_mmu)
674
675SYM_FUNC_START(__cpu_secondary_check52bitva)
676#if VA_BITS > 48
677	ldr_l	x0, vabits_actual
678	cmp	x0, #52
679	b.ne	2f
680
681	mrs_s	x0, SYS_ID_AA64MMFR2_EL1
682	and	x0, x0, #(0xf << ID_AA64MMFR2_EL1_VARange_SHIFT)
683	cbnz	x0, 2f
684
685	update_early_cpu_boot_status \
686		CPU_STUCK_IN_KERNEL | CPU_STUCK_REASON_52_BIT_VA, x0, x1
6871:	wfe
688	wfi
689	b	1b
690
691#endif
6922:	ret
693SYM_FUNC_END(__cpu_secondary_check52bitva)
694
695SYM_FUNC_START_LOCAL(__no_granule_support)
696	/* Indicate that this CPU can't boot and is stuck in the kernel */
697	update_early_cpu_boot_status \
698		CPU_STUCK_IN_KERNEL | CPU_STUCK_REASON_NO_GRAN, x1, x2
6991:
700	wfe
701	wfi
702	b	1b
703SYM_FUNC_END(__no_granule_support)
704
705#ifdef CONFIG_RELOCATABLE
706SYM_FUNC_START_LOCAL(__relocate_kernel)
707	/*
708	 * Iterate over each entry in the relocation table, and apply the
709	 * relocations in place.
710	 */
711	adr_l	x9, __rela_start
712	adr_l	x10, __rela_end
713	mov_q	x11, KIMAGE_VADDR		// default virtual offset
714	add	x11, x11, x23			// actual virtual offset
715
7160:	cmp	x9, x10
717	b.hs	1f
718	ldp	x12, x13, [x9], #24
719	ldr	x14, [x9, #-8]
720	cmp	w13, #R_AARCH64_RELATIVE
721	b.ne	0b
722	add	x14, x14, x23			// relocate
723	str	x14, [x12, x23]
724	b	0b
725
7261:
727#ifdef CONFIG_RELR
728	/*
729	 * Apply RELR relocations.
730	 *
731	 * RELR is a compressed format for storing relative relocations. The
732	 * encoded sequence of entries looks like:
733	 * [ AAAAAAAA BBBBBBB1 BBBBBBB1 ... AAAAAAAA BBBBBB1 ... ]
734	 *
735	 * i.e. start with an address, followed by any number of bitmaps. The
736	 * address entry encodes 1 relocation. The subsequent bitmap entries
737	 * encode up to 63 relocations each, at subsequent offsets following
738	 * the last address entry.
739	 *
740	 * The bitmap entries must have 1 in the least significant bit. The
741	 * assumption here is that an address cannot have 1 in lsb. Odd
742	 * addresses are not supported. Any odd addresses are stored in the RELA
743	 * section, which is handled above.
744	 *
745	 * Excluding the least significant bit in the bitmap, each non-zero
746	 * bit in the bitmap represents a relocation to be applied to
747	 * a corresponding machine word that follows the base address
748	 * word. The second least significant bit represents the machine
749	 * word immediately following the initial address, and each bit
750	 * that follows represents the next word, in linear order. As such,
751	 * a single bitmap can encode up to 63 relocations in a 64-bit object.
752	 *
753	 * In this implementation we store the address of the next RELR table
754	 * entry in x9, the address being relocated by the current address or
755	 * bitmap entry in x13 and the address being relocated by the current
756	 * bit in x14.
757	 */
758	adr_l	x9, __relr_start
759	adr_l	x10, __relr_end
760
7612:	cmp	x9, x10
762	b.hs	7f
763	ldr	x11, [x9], #8
764	tbnz	x11, #0, 3f			// branch to handle bitmaps
765	add	x13, x11, x23
766	ldr	x12, [x13]			// relocate address entry
767	add	x12, x12, x23
768	str	x12, [x13], #8			// adjust to start of bitmap
769	b	2b
770
7713:	mov	x14, x13
7724:	lsr	x11, x11, #1
773	cbz	x11, 6f
774	tbz	x11, #0, 5f			// skip bit if not set
775	ldr	x12, [x14]			// relocate bit
776	add	x12, x12, x23
777	str	x12, [x14]
778
7795:	add	x14, x14, #8			// move to next bit's address
780	b	4b
781
7826:	/*
783	 * Move to the next bitmap's address. 8 is the word size, and 63 is the
784	 * number of significant bits in a bitmap entry.
785	 */
786	add	x13, x13, #(8 * 63)
787	b	2b
788
7897:
790#endif
791	ret
792
793SYM_FUNC_END(__relocate_kernel)
794#endif
795
796SYM_FUNC_START_LOCAL(__primary_switch)
797	adrp	x1, reserved_pg_dir
798	adrp	x2, init_idmap_pg_dir
799	bl	__enable_mmu
800#ifdef CONFIG_RELOCATABLE
801	adrp	x23, KERNEL_START
802	and	x23, x23, MIN_KIMG_ALIGN - 1
803#ifdef CONFIG_RANDOMIZE_BASE
804	mov	x0, x22
805	adrp	x1, init_pg_end
806	mov	sp, x1
807	mov	x29, xzr
808	bl	__pi_kaslr_early_init
809	and	x24, x0, #SZ_2M - 1		// capture memstart offset seed
810	bic	x0, x0, #SZ_2M - 1
811	orr	x23, x23, x0			// record kernel offset
812#endif
813#endif
814	bl	clear_page_tables
815	bl	create_kernel_mapping
816
817	adrp	x1, init_pg_dir
818	load_ttbr1 x1, x1, x2
819#ifdef CONFIG_RELOCATABLE
820	bl	__relocate_kernel
821#endif
822	ldr	x8, =__primary_switched
823	adrp	x0, KERNEL_START		// __pa(KERNEL_START)
824	br	x8
825SYM_FUNC_END(__primary_switch)
826