xref: /openbmc/linux/arch/arm64/kernel/head.S (revision 2359ccdd)
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#include "efi-header.S"
46
47#define __PHYS_OFFSET	(KERNEL_START - TEXT_OFFSET)
48
49#if (TEXT_OFFSET & 0xfff) != 0
50#error TEXT_OFFSET must be at least 4KB aligned
51#elif (PAGE_OFFSET & 0x1fffff) != 0
52#error PAGE_OFFSET must be at least 2MB aligned
53#elif TEXT_OFFSET > 0x1fffff
54#error TEXT_OFFSET must be less than 2MB
55#endif
56
57/*
58 * Kernel startup entry point.
59 * ---------------------------
60 *
61 * The requirements are:
62 *   MMU = off, D-cache = off, I-cache = on or off,
63 *   x0 = physical address to the FDT blob.
64 *
65 * This code is mostly position independent so you call this at
66 * __pa(PAGE_OFFSET + TEXT_OFFSET).
67 *
68 * Note that the callee-saved registers are used for storing variables
69 * that are useful before the MMU is enabled. The allocations are described
70 * in the entry routines.
71 */
72	__HEAD
73_head:
74	/*
75	 * DO NOT MODIFY. Image header expected by Linux boot-loaders.
76	 */
77#ifdef CONFIG_EFI
78	/*
79	 * This add instruction has no meaningful effect except that
80	 * its opcode forms the magic "MZ" signature required by UEFI.
81	 */
82	add	x13, x18, #0x16
83	b	stext
84#else
85	b	stext				// branch to kernel start, magic
86	.long	0				// reserved
87#endif
88	le64sym	_kernel_offset_le		// Image load offset from start of RAM, little-endian
89	le64sym	_kernel_size_le			// Effective size of kernel image, little-endian
90	le64sym	_kernel_flags_le		// Informative flags, little-endian
91	.quad	0				// reserved
92	.quad	0				// reserved
93	.quad	0				// reserved
94	.ascii	"ARM\x64"			// Magic number
95#ifdef CONFIG_EFI
96	.long	pe_header - _head		// Offset to the PE header.
97
98pe_header:
99	__EFI_PE_HEADER
100#else
101	.long	0				// reserved
102#endif
103
104	__INIT
105
106	/*
107	 * The following callee saved general purpose registers are used on the
108	 * primary lowlevel boot path:
109	 *
110	 *  Register   Scope                      Purpose
111	 *  x21        stext() .. start_kernel()  FDT pointer passed at boot in x0
112	 *  x23        stext() .. start_kernel()  physical misalignment/KASLR offset
113	 *  x28        __create_page_tables()     callee preserved temp register
114	 *  x19/x20    __primary_switch()         callee preserved temp registers
115	 */
116ENTRY(stext)
117	bl	preserve_boot_args
118	bl	el2_setup			// Drop to EL1, w0=cpu_boot_mode
119	adrp	x23, __PHYS_OFFSET
120	and	x23, x23, MIN_KIMG_ALIGN - 1	// KASLR offset, defaults to 0
121	bl	set_cpu_boot_mode_flag
122	bl	__create_page_tables
123	/*
124	 * The following calls CPU setup code, see arch/arm64/mm/proc.S for
125	 * details.
126	 * On return, the CPU will be ready for the MMU to be turned on and
127	 * the TCR will have been set.
128	 */
129	bl	__cpu_setup			// initialise processor
130	b	__primary_switch
131ENDPROC(stext)
132
133/*
134 * Preserve the arguments passed by the bootloader in x0 .. x3
135 */
136preserve_boot_args:
137	mov	x21, x0				// x21=FDT
138
139	adr_l	x0, boot_args			// record the contents of
140	stp	x21, x1, [x0]			// x0 .. x3 at kernel entry
141	stp	x2, x3, [x0, #16]
142
143	dmb	sy				// needed before dc ivac with
144						// MMU off
145
146	mov	x1, #0x20			// 4 x 8 bytes
147	b	__inval_dcache_area		// tail call
148ENDPROC(preserve_boot_args)
149
150/*
151 * Macro to create a table entry to the next page.
152 *
153 *	tbl:	page table address
154 *	virt:	virtual address
155 *	shift:	#imm page table shift
156 *	ptrs:	#imm pointers per table page
157 *
158 * Preserves:	virt
159 * Corrupts:	ptrs, tmp1, tmp2
160 * Returns:	tbl -> next level table page address
161 */
162	.macro	create_table_entry, tbl, virt, shift, ptrs, tmp1, tmp2
163	add	\tmp1, \tbl, #PAGE_SIZE
164	phys_to_pte \tmp2, \tmp1
165	orr	\tmp2, \tmp2, #PMD_TYPE_TABLE	// address of next table and entry type
166	lsr	\tmp1, \virt, #\shift
167	sub	\ptrs, \ptrs, #1
168	and	\tmp1, \tmp1, \ptrs		// table index
169	str	\tmp2, [\tbl, \tmp1, lsl #3]
170	add	\tbl, \tbl, #PAGE_SIZE		// next level table page
171	.endm
172
173/*
174 * Macro to populate page table entries, these entries can be pointers to the next level
175 * or last level entries pointing to physical memory.
176 *
177 *	tbl:	page table address
178 *	rtbl:	pointer to page table or physical memory
179 *	index:	start index to write
180 *	eindex:	end index to write - [index, eindex] written to
181 *	flags:	flags for pagetable entry to or in
182 *	inc:	increment to rtbl between each entry
183 *	tmp1:	temporary variable
184 *
185 * Preserves:	tbl, eindex, flags, inc
186 * Corrupts:	index, tmp1
187 * Returns:	rtbl
188 */
189	.macro populate_entries, tbl, rtbl, index, eindex, flags, inc, tmp1
190.Lpe\@:	phys_to_pte \tmp1, \rtbl
191	orr	\tmp1, \tmp1, \flags	// tmp1 = table entry
192	str	\tmp1, [\tbl, \index, lsl #3]
193	add	\rtbl, \rtbl, \inc	// rtbl = pa next level
194	add	\index, \index, #1
195	cmp	\index, \eindex
196	b.ls	.Lpe\@
197	.endm
198
199/*
200 * Compute indices of table entries from virtual address range. If multiple entries
201 * were needed in the previous page table level then the next page table level is assumed
202 * to be composed of multiple pages. (This effectively scales the end index).
203 *
204 *	vstart:	virtual address of start of range
205 *	vend:	virtual address of end of range
206 *	shift:	shift used to transform virtual address into index
207 *	ptrs:	number of entries in page table
208 *	istart:	index in table corresponding to vstart
209 *	iend:	index in table corresponding to vend
210 *	count:	On entry: how many extra entries were required in previous level, scales
211 *			  our end index.
212 *		On exit: returns how many extra entries required for next page table level
213 *
214 * Preserves:	vstart, vend, shift, ptrs
215 * Returns:	istart, iend, count
216 */
217	.macro compute_indices, vstart, vend, shift, ptrs, istart, iend, count
218	lsr	\iend, \vend, \shift
219	mov	\istart, \ptrs
220	sub	\istart, \istart, #1
221	and	\iend, \iend, \istart	// iend = (vend >> shift) & (ptrs - 1)
222	mov	\istart, \ptrs
223	mul	\istart, \istart, \count
224	add	\iend, \iend, \istart	// iend += (count - 1) * ptrs
225					// our entries span multiple tables
226
227	lsr	\istart, \vstart, \shift
228	mov	\count, \ptrs
229	sub	\count, \count, #1
230	and	\istart, \istart, \count
231
232	sub	\count, \iend, \istart
233	.endm
234
235/*
236 * Map memory for specified virtual address range. Each level of page table needed supports
237 * multiple entries. If a level requires n entries the next page table level is assumed to be
238 * formed from n pages.
239 *
240 *	tbl:	location of page table
241 *	rtbl:	address to be used for first level page table entry (typically tbl + PAGE_SIZE)
242 *	vstart:	start address to map
243 *	vend:	end address to map - we map [vstart, vend]
244 *	flags:	flags to use to map last level entries
245 *	phys:	physical address corresponding to vstart - physical memory is contiguous
246 *	pgds:	the number of pgd entries
247 *
248 * Temporaries:	istart, iend, tmp, count, sv - these need to be different registers
249 * Preserves:	vstart, vend, flags
250 * Corrupts:	tbl, rtbl, istart, iend, tmp, count, sv
251 */
252	.macro map_memory, tbl, rtbl, vstart, vend, flags, phys, pgds, istart, iend, tmp, count, sv
253	add \rtbl, \tbl, #PAGE_SIZE
254	mov \sv, \rtbl
255	mov \count, #0
256	compute_indices \vstart, \vend, #PGDIR_SHIFT, \pgds, \istart, \iend, \count
257	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
258	mov \tbl, \sv
259	mov \sv, \rtbl
260
261#if SWAPPER_PGTABLE_LEVELS > 3
262	compute_indices \vstart, \vend, #PUD_SHIFT, #PTRS_PER_PUD, \istart, \iend, \count
263	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
264	mov \tbl, \sv
265	mov \sv, \rtbl
266#endif
267
268#if SWAPPER_PGTABLE_LEVELS > 2
269	compute_indices \vstart, \vend, #SWAPPER_TABLE_SHIFT, #PTRS_PER_PMD, \istart, \iend, \count
270	populate_entries \tbl, \rtbl, \istart, \iend, #PMD_TYPE_TABLE, #PAGE_SIZE, \tmp
271	mov \tbl, \sv
272#endif
273
274	compute_indices \vstart, \vend, #SWAPPER_BLOCK_SHIFT, #PTRS_PER_PTE, \istart, \iend, \count
275	bic \count, \phys, #SWAPPER_BLOCK_SIZE - 1
276	populate_entries \tbl, \count, \istart, \iend, \flags, #SWAPPER_BLOCK_SIZE, \tmp
277	.endm
278
279/*
280 * Setup the initial page tables. We only setup the barest amount which is
281 * required to get the kernel running. The following sections are required:
282 *   - identity mapping to enable the MMU (low address, TTBR0)
283 *   - first few MB of the kernel linear mapping to jump to once the MMU has
284 *     been enabled
285 */
286__create_page_tables:
287	mov	x28, lr
288
289	/*
290	 * Invalidate the idmap and swapper page tables to avoid potential
291	 * dirty cache lines being evicted.
292	 */
293	adrp	x0, idmap_pg_dir
294	adrp	x1, swapper_pg_end
295	sub	x1, x1, x0
296	bl	__inval_dcache_area
297
298	/*
299	 * Clear the idmap and swapper page tables.
300	 */
301	adrp	x0, idmap_pg_dir
302	adrp	x1, swapper_pg_end
303	sub	x1, x1, x0
3041:	stp	xzr, xzr, [x0], #16
305	stp	xzr, xzr, [x0], #16
306	stp	xzr, xzr, [x0], #16
307	stp	xzr, xzr, [x0], #16
308	subs	x1, x1, #64
309	b.ne	1b
310
311	mov	x7, SWAPPER_MM_MMUFLAGS
312
313	/*
314	 * Create the identity mapping.
315	 */
316	adrp	x0, idmap_pg_dir
317	adrp	x3, __idmap_text_start		// __pa(__idmap_text_start)
318
319	/*
320	 * VA_BITS may be too small to allow for an ID mapping to be created
321	 * that covers system RAM if that is located sufficiently high in the
322	 * physical address space. So for the ID map, use an extended virtual
323	 * range in that case, and configure an additional translation level
324	 * if needed.
325	 *
326	 * Calculate the maximum allowed value for TCR_EL1.T0SZ so that the
327	 * entire ID map region can be mapped. As T0SZ == (64 - #bits used),
328	 * this number conveniently equals the number of leading zeroes in
329	 * the physical address of __idmap_text_end.
330	 */
331	adrp	x5, __idmap_text_end
332	clz	x5, x5
333	cmp	x5, TCR_T0SZ(VA_BITS)	// default T0SZ small enough?
334	b.ge	1f			// .. then skip VA range extension
335
336	adr_l	x6, idmap_t0sz
337	str	x5, [x6]
338	dmb	sy
339	dc	ivac, x6		// Invalidate potentially stale cache line
340
341#if (VA_BITS < 48)
342#define EXTRA_SHIFT	(PGDIR_SHIFT + PAGE_SHIFT - 3)
343#define EXTRA_PTRS	(1 << (PHYS_MASK_SHIFT - EXTRA_SHIFT))
344
345	/*
346	 * If VA_BITS < 48, we have to configure an additional table level.
347	 * First, we have to verify our assumption that the current value of
348	 * VA_BITS was chosen such that all translation levels are fully
349	 * utilised, and that lowering T0SZ will always result in an additional
350	 * translation level to be configured.
351	 */
352#if VA_BITS != EXTRA_SHIFT
353#error "Mismatch between VA_BITS and page size/number of translation levels"
354#endif
355
356	mov	x4, EXTRA_PTRS
357	create_table_entry x0, x3, EXTRA_SHIFT, x4, x5, x6
358#else
359	/*
360	 * If VA_BITS == 48, we don't have to configure an additional
361	 * translation level, but the top-level table has more entries.
362	 */
363	mov	x4, #1 << (PHYS_MASK_SHIFT - PGDIR_SHIFT)
364	str_l	x4, idmap_ptrs_per_pgd, x5
365#endif
3661:
367	ldr_l	x4, idmap_ptrs_per_pgd
368	mov	x5, x3				// __pa(__idmap_text_start)
369	adr_l	x6, __idmap_text_end		// __pa(__idmap_text_end)
370
371	map_memory x0, x1, x3, x6, x7, x3, x4, x10, x11, x12, x13, x14
372
373	/*
374	 * Map the kernel image (starting with PHYS_OFFSET).
375	 */
376	adrp	x0, swapper_pg_dir
377	mov_q	x5, KIMAGE_VADDR + TEXT_OFFSET	// compile time __va(_text)
378	add	x5, x5, x23			// add KASLR displacement
379	mov	x4, PTRS_PER_PGD
380	adrp	x6, _end			// runtime __pa(_end)
381	adrp	x3, _text			// runtime __pa(_text)
382	sub	x6, x6, x3			// _end - _text
383	add	x6, x6, x5			// runtime __va(_end)
384
385	map_memory x0, x1, x5, x6, x7, x3, x4, x10, x11, x12, x13, x14
386
387	/*
388	 * Since the page tables have been populated with non-cacheable
389	 * accesses (MMU disabled), invalidate the idmap and swapper page
390	 * tables again to remove any speculatively loaded cache lines.
391	 */
392	adrp	x0, idmap_pg_dir
393	adrp	x1, swapper_pg_end
394	sub	x1, x1, x0
395	dmb	sy
396	bl	__inval_dcache_area
397
398	ret	x28
399ENDPROC(__create_page_tables)
400	.ltorg
401
402/*
403 * The following fragment of code is executed with the MMU enabled.
404 *
405 *   x0 = __PHYS_OFFSET
406 */
407__primary_switched:
408	adrp	x4, init_thread_union
409	add	sp, x4, #THREAD_SIZE
410	adr_l	x5, init_task
411	msr	sp_el0, x5			// Save thread_info
412
413	adr_l	x8, vectors			// load VBAR_EL1 with virtual
414	msr	vbar_el1, x8			// vector table address
415	isb
416
417	stp	xzr, x30, [sp, #-16]!
418	mov	x29, sp
419
420	str_l	x21, __fdt_pointer, x5		// Save FDT pointer
421
422	ldr_l	x4, kimage_vaddr		// Save the offset between
423	sub	x4, x4, x0			// the kernel virtual and
424	str_l	x4, kimage_voffset, x5		// physical mappings
425
426	// Clear BSS
427	adr_l	x0, __bss_start
428	mov	x1, xzr
429	adr_l	x2, __bss_stop
430	sub	x2, x2, x0
431	bl	__pi_memset
432	dsb	ishst				// Make zero page visible to PTW
433
434#ifdef CONFIG_KASAN
435	bl	kasan_early_init
436#endif
437#ifdef CONFIG_RANDOMIZE_BASE
438	tst	x23, ~(MIN_KIMG_ALIGN - 1)	// already running randomized?
439	b.ne	0f
440	mov	x0, x21				// pass FDT address in x0
441	bl	kaslr_early_init		// parse FDT for KASLR options
442	cbz	x0, 0f				// KASLR disabled? just proceed
443	orr	x23, x23, x0			// record KASLR offset
444	ldp	x29, x30, [sp], #16		// we must enable KASLR, return
445	ret					// to __primary_switch()
4460:
447#endif
448	add	sp, sp, #16
449	mov	x29, #0
450	mov	x30, #0
451	b	start_kernel
452ENDPROC(__primary_switched)
453
454/*
455 * end early head section, begin head code that is also used for
456 * hotplug and needs to have the same protections as the text region
457 */
458	.section ".idmap.text","awx"
459
460ENTRY(kimage_vaddr)
461	.quad		_text - TEXT_OFFSET
462
463/*
464 * If we're fortunate enough to boot at EL2, ensure that the world is
465 * sane before dropping to EL1.
466 *
467 * Returns either BOOT_CPU_MODE_EL1 or BOOT_CPU_MODE_EL2 in w0 if
468 * booted in EL1 or EL2 respectively.
469 */
470ENTRY(el2_setup)
471	msr	SPsel, #1			// We want to use SP_EL{1,2}
472	mrs	x0, CurrentEL
473	cmp	x0, #CurrentEL_EL2
474	b.eq	1f
475	mov_q	x0, (SCTLR_EL1_RES1 | ENDIAN_SET_EL1)
476	msr	sctlr_el1, x0
477	mov	w0, #BOOT_CPU_MODE_EL1		// This cpu booted in EL1
478	isb
479	ret
480
4811:	mov_q	x0, (SCTLR_EL2_RES1 | ENDIAN_SET_EL2)
482	msr	sctlr_el2, x0
483
484#ifdef CONFIG_ARM64_VHE
485	/*
486	 * Check for VHE being present. For the rest of the EL2 setup,
487	 * x2 being non-zero indicates that we do have VHE, and that the
488	 * kernel is intended to run at EL2.
489	 */
490	mrs	x2, id_aa64mmfr1_el1
491	ubfx	x2, x2, #8, #4
492#else
493	mov	x2, xzr
494#endif
495
496	/* Hyp configuration. */
497	mov	x0, #HCR_RW			// 64-bit EL1
498	cbz	x2, set_hcr
499	orr	x0, x0, #HCR_TGE		// Enable Host Extensions
500	orr	x0, x0, #HCR_E2H
501set_hcr:
502	msr	hcr_el2, x0
503	isb
504
505	/*
506	 * Allow Non-secure EL1 and EL0 to access physical timer and counter.
507	 * This is not necessary for VHE, since the host kernel runs in EL2,
508	 * and EL0 accesses are configured in the later stage of boot process.
509	 * Note that when HCR_EL2.E2H == 1, CNTHCTL_EL2 has the same bit layout
510	 * as CNTKCTL_EL1, and CNTKCTL_EL1 accessing instructions are redefined
511	 * to access CNTHCTL_EL2. This allows the kernel designed to run at EL1
512	 * to transparently mess with the EL0 bits via CNTKCTL_EL1 access in
513	 * EL2.
514	 */
515	cbnz	x2, 1f
516	mrs	x0, cnthctl_el2
517	orr	x0, x0, #3			// Enable EL1 physical timers
518	msr	cnthctl_el2, x0
5191:
520	msr	cntvoff_el2, xzr		// Clear virtual offset
521
522#ifdef CONFIG_ARM_GIC_V3
523	/* GICv3 system register access */
524	mrs	x0, id_aa64pfr0_el1
525	ubfx	x0, x0, #24, #4
526	cmp	x0, #1
527	b.ne	3f
528
529	mrs_s	x0, SYS_ICC_SRE_EL2
530	orr	x0, x0, #ICC_SRE_EL2_SRE	// Set ICC_SRE_EL2.SRE==1
531	orr	x0, x0, #ICC_SRE_EL2_ENABLE	// Set ICC_SRE_EL2.Enable==1
532	msr_s	SYS_ICC_SRE_EL2, x0
533	isb					// Make sure SRE is now set
534	mrs_s	x0, SYS_ICC_SRE_EL2		// Read SRE back,
535	tbz	x0, #0, 3f			// and check that it sticks
536	msr_s	SYS_ICH_HCR_EL2, xzr		// Reset ICC_HCR_EL2 to defaults
537
5383:
539#endif
540
541	/* Populate ID registers. */
542	mrs	x0, midr_el1
543	mrs	x1, mpidr_el1
544	msr	vpidr_el2, x0
545	msr	vmpidr_el2, x1
546
547#ifdef CONFIG_COMPAT
548	msr	hstr_el2, xzr			// Disable CP15 traps to EL2
549#endif
550
551	/* EL2 debug */
552	mrs	x1, id_aa64dfr0_el1		// Check ID_AA64DFR0_EL1 PMUVer
553	sbfx	x0, x1, #8, #4
554	cmp	x0, #1
555	b.lt	4f				// Skip if no PMU present
556	mrs	x0, pmcr_el0			// Disable debug access traps
557	ubfx	x0, x0, #11, #5			// to EL2 and allow access to
5584:
559	csel	x3, xzr, x0, lt			// all PMU counters from EL1
560
561	/* Statistical profiling */
562	ubfx	x0, x1, #32, #4			// Check ID_AA64DFR0_EL1 PMSVer
563	cbz	x0, 7f				// Skip if SPE not present
564	cbnz	x2, 6f				// VHE?
565	mrs_s	x4, SYS_PMBIDR_EL1		// If SPE available at EL2,
566	and	x4, x4, #(1 << SYS_PMBIDR_EL1_P_SHIFT)
567	cbnz	x4, 5f				// then permit sampling of physical
568	mov	x4, #(1 << SYS_PMSCR_EL2_PCT_SHIFT | \
569		      1 << SYS_PMSCR_EL2_PA_SHIFT)
570	msr_s	SYS_PMSCR_EL2, x4		// addresses and physical counter
5715:
572	mov	x1, #(MDCR_EL2_E2PB_MASK << MDCR_EL2_E2PB_SHIFT)
573	orr	x3, x3, x1			// If we don't have VHE, then
574	b	7f				// use EL1&0 translation.
5756:						// For VHE, use EL2 translation
576	orr	x3, x3, #MDCR_EL2_TPMS		// and disable access from EL1
5777:
578	msr	mdcr_el2, x3			// Configure debug traps
579
580	/* LORegions */
581	mrs	x1, id_aa64mmfr1_el1
582	ubfx	x0, x1, #ID_AA64MMFR1_LOR_SHIFT, 4
583	cbz	x0, 1f
584	msr_s	SYS_LORC_EL1, xzr
5851:
586
587	/* Stage-2 translation */
588	msr	vttbr_el2, xzr
589
590	cbz	x2, install_el2_stub
591
592	mov	w0, #BOOT_CPU_MODE_EL2		// This CPU booted in EL2
593	isb
594	ret
595
596install_el2_stub:
597	/*
598	 * When VHE is not in use, early init of EL2 and EL1 needs to be
599	 * done here.
600	 * When VHE _is_ in use, EL1 will not be used in the host and
601	 * requires no configuration, and all non-hyp-specific EL2 setup
602	 * will be done via the _EL1 system register aliases in __cpu_setup.
603	 */
604	mov_q	x0, (SCTLR_EL1_RES1 | ENDIAN_SET_EL1)
605	msr	sctlr_el1, x0
606
607	/* Coprocessor traps. */
608	mov	x0, #0x33ff
609	msr	cptr_el2, x0			// Disable copro. traps to EL2
610
611	/* SVE register access */
612	mrs	x1, id_aa64pfr0_el1
613	ubfx	x1, x1, #ID_AA64PFR0_SVE_SHIFT, #4
614	cbz	x1, 7f
615
616	bic	x0, x0, #CPTR_EL2_TZ		// Also disable SVE traps
617	msr	cptr_el2, x0			// Disable copro. traps to EL2
618	isb
619	mov	x1, #ZCR_ELx_LEN_MASK		// SVE: Enable full vector
620	msr_s	SYS_ZCR_EL2, x1			// length for EL1.
621
622	/* Hypervisor stub */
6237:	adr_l	x0, __hyp_stub_vectors
624	msr	vbar_el2, x0
625
626	/* spsr */
627	mov	x0, #(PSR_F_BIT | PSR_I_BIT | PSR_A_BIT | PSR_D_BIT |\
628		      PSR_MODE_EL1h)
629	msr	spsr_el2, x0
630	msr	elr_el2, lr
631	mov	w0, #BOOT_CPU_MODE_EL2		// This CPU booted in EL2
632	eret
633ENDPROC(el2_setup)
634
635/*
636 * Sets the __boot_cpu_mode flag depending on the CPU boot mode passed
637 * in w0. See arch/arm64/include/asm/virt.h for more info.
638 */
639set_cpu_boot_mode_flag:
640	adr_l	x1, __boot_cpu_mode
641	cmp	w0, #BOOT_CPU_MODE_EL2
642	b.ne	1f
643	add	x1, x1, #4
6441:	str	w0, [x1]			// This CPU has booted in EL1
645	dmb	sy
646	dc	ivac, x1			// Invalidate potentially stale cache line
647	ret
648ENDPROC(set_cpu_boot_mode_flag)
649
650/*
651 * These values are written with the MMU off, but read with the MMU on.
652 * Writers will invalidate the corresponding address, discarding up to a
653 * 'Cache Writeback Granule' (CWG) worth of data. The linker script ensures
654 * sufficient alignment that the CWG doesn't overlap another section.
655 */
656	.pushsection ".mmuoff.data.write", "aw"
657/*
658 * We need to find out the CPU boot mode long after boot, so we need to
659 * store it in a writable variable.
660 *
661 * This is not in .bss, because we set it sufficiently early that the boot-time
662 * zeroing of .bss would clobber it.
663 */
664ENTRY(__boot_cpu_mode)
665	.long	BOOT_CPU_MODE_EL2
666	.long	BOOT_CPU_MODE_EL1
667/*
668 * The booting CPU updates the failed status @__early_cpu_boot_status,
669 * with MMU turned off.
670 */
671ENTRY(__early_cpu_boot_status)
672	.long 	0
673
674	.popsection
675
676	/*
677	 * This provides a "holding pen" for platforms to hold all secondary
678	 * cores are held until we're ready for them to initialise.
679	 */
680ENTRY(secondary_holding_pen)
681	bl	el2_setup			// Drop to EL1, w0=cpu_boot_mode
682	bl	set_cpu_boot_mode_flag
683	mrs	x0, mpidr_el1
684	mov_q	x1, MPIDR_HWID_BITMASK
685	and	x0, x0, x1
686	adr_l	x3, secondary_holding_pen_release
687pen:	ldr	x4, [x3]
688	cmp	x4, x0
689	b.eq	secondary_startup
690	wfe
691	b	pen
692ENDPROC(secondary_holding_pen)
693
694	/*
695	 * Secondary entry point that jumps straight into the kernel. Only to
696	 * be used where CPUs are brought online dynamically by the kernel.
697	 */
698ENTRY(secondary_entry)
699	bl	el2_setup			// Drop to EL1
700	bl	set_cpu_boot_mode_flag
701	b	secondary_startup
702ENDPROC(secondary_entry)
703
704secondary_startup:
705	/*
706	 * Common entry point for secondary CPUs.
707	 */
708	bl	__cpu_setup			// initialise processor
709	bl	__enable_mmu
710	ldr	x8, =__secondary_switched
711	br	x8
712ENDPROC(secondary_startup)
713
714__secondary_switched:
715	adr_l	x5, vectors
716	msr	vbar_el1, x5
717	isb
718
719	adr_l	x0, secondary_data
720	ldr	x1, [x0, #CPU_BOOT_STACK]	// get secondary_data.stack
721	mov	sp, x1
722	ldr	x2, [x0, #CPU_BOOT_TASK]
723	msr	sp_el0, x2
724	mov	x29, #0
725	mov	x30, #0
726	b	secondary_start_kernel
727ENDPROC(__secondary_switched)
728
729/*
730 * The booting CPU updates the failed status @__early_cpu_boot_status,
731 * with MMU turned off.
732 *
733 * update_early_cpu_boot_status tmp, status
734 *  - Corrupts tmp1, tmp2
735 *  - Writes 'status' to __early_cpu_boot_status and makes sure
736 *    it is committed to memory.
737 */
738
739	.macro	update_early_cpu_boot_status status, tmp1, tmp2
740	mov	\tmp2, #\status
741	adr_l	\tmp1, __early_cpu_boot_status
742	str	\tmp2, [\tmp1]
743	dmb	sy
744	dc	ivac, \tmp1			// Invalidate potentially stale cache line
745	.endm
746
747/*
748 * Enable the MMU.
749 *
750 *  x0  = SCTLR_EL1 value for turning on the MMU.
751 *
752 * Returns to the caller via x30/lr. This requires the caller to be covered
753 * by the .idmap.text section.
754 *
755 * Checks if the selected granule size is supported by the CPU.
756 * If it isn't, park the CPU
757 */
758ENTRY(__enable_mmu)
759	mrs	x1, ID_AA64MMFR0_EL1
760	ubfx	x2, x1, #ID_AA64MMFR0_TGRAN_SHIFT, 4
761	cmp	x2, #ID_AA64MMFR0_TGRAN_SUPPORTED
762	b.ne	__no_granule_support
763	update_early_cpu_boot_status 0, x1, x2
764	adrp	x1, idmap_pg_dir
765	adrp	x2, swapper_pg_dir
766	phys_to_ttbr x3, x1
767	phys_to_ttbr x4, x2
768	msr	ttbr0_el1, x3			// load TTBR0
769	msr	ttbr1_el1, x4			// load TTBR1
770	isb
771	msr	sctlr_el1, x0
772	isb
773	/*
774	 * Invalidate the local I-cache so that any instructions fetched
775	 * speculatively from the PoC are discarded, since they may have
776	 * been dynamically patched at the PoU.
777	 */
778	ic	iallu
779	dsb	nsh
780	isb
781	ret
782ENDPROC(__enable_mmu)
783
784__no_granule_support:
785	/* Indicate that this CPU can't boot and is stuck in the kernel */
786	update_early_cpu_boot_status CPU_STUCK_IN_KERNEL, x1, x2
7871:
788	wfe
789	wfi
790	b	1b
791ENDPROC(__no_granule_support)
792
793#ifdef CONFIG_RELOCATABLE
794__relocate_kernel:
795	/*
796	 * Iterate over each entry in the relocation table, and apply the
797	 * relocations in place.
798	 */
799	ldr	w9, =__rela_offset		// offset to reloc table
800	ldr	w10, =__rela_size		// size of reloc table
801
802	mov_q	x11, KIMAGE_VADDR		// default virtual offset
803	add	x11, x11, x23			// actual virtual offset
804	add	x9, x9, x11			// __va(.rela)
805	add	x10, x9, x10			// __va(.rela) + sizeof(.rela)
806
8070:	cmp	x9, x10
808	b.hs	1f
809	ldp	x11, x12, [x9], #24
810	ldr	x13, [x9, #-8]
811	cmp	w12, #R_AARCH64_RELATIVE
812	b.ne	0b
813	add	x13, x13, x23			// relocate
814	str	x13, [x11, x23]
815	b	0b
8161:	ret
817ENDPROC(__relocate_kernel)
818#endif
819
820__primary_switch:
821#ifdef CONFIG_RANDOMIZE_BASE
822	mov	x19, x0				// preserve new SCTLR_EL1 value
823	mrs	x20, sctlr_el1			// preserve old SCTLR_EL1 value
824#endif
825
826	bl	__enable_mmu
827#ifdef CONFIG_RELOCATABLE
828	bl	__relocate_kernel
829#ifdef CONFIG_RANDOMIZE_BASE
830	ldr	x8, =__primary_switched
831	adrp	x0, __PHYS_OFFSET
832	blr	x8
833
834	/*
835	 * If we return here, we have a KASLR displacement in x23 which we need
836	 * to take into account by discarding the current kernel mapping and
837	 * creating a new one.
838	 */
839	pre_disable_mmu_workaround
840	msr	sctlr_el1, x20			// disable the MMU
841	isb
842	bl	__create_page_tables		// recreate kernel mapping
843
844	tlbi	vmalle1				// Remove any stale TLB entries
845	dsb	nsh
846
847	msr	sctlr_el1, x19			// re-enable the MMU
848	isb
849	ic	iallu				// flush instructions fetched
850	dsb	nsh				// via old mapping
851	isb
852
853	bl	__relocate_kernel
854#endif
855#endif
856	ldr	x8, =__primary_switched
857	adrp	x0, __PHYS_OFFSET
858	br	x8
859ENDPROC(__primary_switch)
860