xref: /linux/arch/arm64/include/asm/assembler.h (revision eb2f932100288dbb881eadfed02e1459c6b9504c)
1 /* SPDX-License-Identifier: GPL-2.0-only */
2 /*
3  * Based on arch/arm/include/asm/assembler.h, arch/arm/mm/proc-macros.S
4  *
5  * Copyright (C) 1996-2000 Russell King
6  * Copyright (C) 2012 ARM Ltd.
7  */
8 #ifndef __ASSEMBLY__
9 #error "Only include this from assembly code"
10 #endif
11 
12 #ifndef __ASM_ASSEMBLER_H
13 #define __ASM_ASSEMBLER_H
14 
15 #include <linux/export.h>
16 
17 #include <asm/alternative.h>
18 #include <asm/asm-bug.h>
19 #include <asm/asm-extable.h>
20 #include <asm/asm-offsets.h>
21 #include <asm/cpufeature.h>
22 #include <asm/cputype.h>
23 #include <asm/debug-monitors.h>
24 #include <asm/page.h>
25 #include <asm/pgtable-hwdef.h>
26 #include <asm/ptrace.h>
27 #include <asm/thread_info.h>
28 
29 	/*
30 	 * Provide a wxN alias for each wN register so what we can paste a xN
31 	 * reference after a 'w' to obtain the 32-bit version.
32 	 */
33 	.irp	n,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28,29,30
34 	wx\n	.req	w\n
35 	.endr
36 
37 	.macro disable_daif
38 	msr	daifset, #0xf
39 	.endm
40 
41 	.macro enable_daif
42 	msr	daifclr, #0xf
43 	.endm
44 
45 /*
46  * Save/restore interrupts.
47  */
48 	.macro	save_and_disable_irq, flags
49 	mrs	\flags, daif
50 	msr	daifset, #3
51 	.endm
52 
53 	.macro	restore_irq, flags
54 	msr	daif, \flags
55 	.endm
56 
57 	.macro	enable_dbg
58 	msr	daifclr, #8
59 	.endm
60 
61 	.macro	disable_step_tsk, flgs, tmp
62 	tbz	\flgs, #TIF_SINGLESTEP, 9990f
63 	mrs	\tmp, mdscr_el1
64 	bic	\tmp, \tmp, #DBG_MDSCR_SS
65 	msr	mdscr_el1, \tmp
66 	isb	// Synchronise with enable_dbg
67 9990:
68 	.endm
69 
70 	/* call with daif masked */
71 	.macro	enable_step_tsk, flgs, tmp
72 	tbz	\flgs, #TIF_SINGLESTEP, 9990f
73 	mrs	\tmp, mdscr_el1
74 	orr	\tmp, \tmp, #DBG_MDSCR_SS
75 	msr	mdscr_el1, \tmp
76 9990:
77 	.endm
78 
79 /*
80  * RAS Error Synchronization barrier
81  */
82 	.macro  esb
83 #ifdef CONFIG_ARM64_RAS_EXTN
84 	hint    #16
85 #else
86 	nop
87 #endif
88 	.endm
89 
90 /*
91  * Value prediction barrier
92  */
93 	.macro	csdb
94 	hint	#20
95 	.endm
96 
97 /*
98  * Clear Branch History instruction
99  */
100 	.macro clearbhb
101 	hint	#22
102 	.endm
103 
104 /*
105  * Speculation barrier
106  */
107 	.macro	sb
108 alternative_if_not ARM64_HAS_SB
109 	dsb	nsh
110 	isb
111 alternative_else
112 	SB_BARRIER_INSN
113 	nop
114 alternative_endif
115 	.endm
116 
117 /*
118  * NOP sequence
119  */
120 	.macro	nops, num
121 	.rept	\num
122 	nop
123 	.endr
124 	.endm
125 
126 /*
127  * Register aliases.
128  */
129 lr	.req	x30		// link register
130 
131 /*
132  * Vector entry
133  */
134 	 .macro	ventry	label
135 	.align	7
136 	b	\label
137 	.endm
138 
139 /*
140  * Select code when configured for BE.
141  */
142 #ifdef CONFIG_CPU_BIG_ENDIAN
143 #define CPU_BE(code...) code
144 #else
145 #define CPU_BE(code...)
146 #endif
147 
148 /*
149  * Select code when configured for LE.
150  */
151 #ifdef CONFIG_CPU_BIG_ENDIAN
152 #define CPU_LE(code...)
153 #else
154 #define CPU_LE(code...) code
155 #endif
156 
157 /*
158  * Define a macro that constructs a 64-bit value by concatenating two
159  * 32-bit registers. Note that on big endian systems the order of the
160  * registers is swapped.
161  */
162 #ifndef CONFIG_CPU_BIG_ENDIAN
163 	.macro	regs_to_64, rd, lbits, hbits
164 #else
165 	.macro	regs_to_64, rd, hbits, lbits
166 #endif
167 	orr	\rd, \lbits, \hbits, lsl #32
168 	.endm
169 
170 /*
171  * Pseudo-ops for PC-relative adr/ldr/str <reg>, <symbol> where
172  * <symbol> is within the range +/- 4 GB of the PC.
173  */
174 	/*
175 	 * @dst: destination register (64 bit wide)
176 	 * @sym: name of the symbol
177 	 */
178 	.macro	adr_l, dst, sym
179 	adrp	\dst, \sym
180 	add	\dst, \dst, :lo12:\sym
181 	.endm
182 
183 	/*
184 	 * @dst: destination register (32 or 64 bit wide)
185 	 * @sym: name of the symbol
186 	 * @tmp: optional 64-bit scratch register to be used if <dst> is a
187 	 *       32-bit wide register, in which case it cannot be used to hold
188 	 *       the address
189 	 */
190 	.macro	ldr_l, dst, sym, tmp=
191 	.ifb	\tmp
192 	adrp	\dst, \sym
193 	ldr	\dst, [\dst, :lo12:\sym]
194 	.else
195 	adrp	\tmp, \sym
196 	ldr	\dst, [\tmp, :lo12:\sym]
197 	.endif
198 	.endm
199 
200 	/*
201 	 * @src: source register (32 or 64 bit wide)
202 	 * @sym: name of the symbol
203 	 * @tmp: mandatory 64-bit scratch register to calculate the address
204 	 *       while <src> needs to be preserved.
205 	 */
206 	.macro	str_l, src, sym, tmp
207 	adrp	\tmp, \sym
208 	str	\src, [\tmp, :lo12:\sym]
209 	.endm
210 
211 	/*
212 	 * @dst: destination register
213 	 */
214 #if defined(__KVM_NVHE_HYPERVISOR__) || defined(__KVM_VHE_HYPERVISOR__)
215 	.macro	get_this_cpu_offset, dst
216 	mrs	\dst, tpidr_el2
217 	.endm
218 #else
219 	.macro	get_this_cpu_offset, dst
220 alternative_if_not ARM64_HAS_VIRT_HOST_EXTN
221 	mrs	\dst, tpidr_el1
222 alternative_else
223 	mrs	\dst, tpidr_el2
224 alternative_endif
225 	.endm
226 
227 	.macro	set_this_cpu_offset, src
228 alternative_if_not ARM64_HAS_VIRT_HOST_EXTN
229 	msr	tpidr_el1, \src
230 alternative_else
231 	msr	tpidr_el2, \src
232 alternative_endif
233 	.endm
234 #endif
235 
236 	/*
237 	 * @dst: Result of per_cpu(sym, smp_processor_id()) (can be SP)
238 	 * @sym: The name of the per-cpu variable
239 	 * @tmp: scratch register
240 	 */
241 	.macro adr_this_cpu, dst, sym, tmp
242 	adrp	\tmp, \sym
243 	add	\dst, \tmp, #:lo12:\sym
244 	get_this_cpu_offset \tmp
245 	add	\dst, \dst, \tmp
246 	.endm
247 
248 	/*
249 	 * @dst: Result of READ_ONCE(per_cpu(sym, smp_processor_id()))
250 	 * @sym: The name of the per-cpu variable
251 	 * @tmp: scratch register
252 	 */
253 	.macro ldr_this_cpu dst, sym, tmp
254 	adr_l	\dst, \sym
255 	get_this_cpu_offset \tmp
256 	ldr	\dst, [\dst, \tmp]
257 	.endm
258 
259 /*
260  * vma_vm_mm - get mm pointer from vma pointer (vma->vm_mm)
261  */
262 	.macro	vma_vm_mm, rd, rn
263 	ldr	\rd, [\rn, #VMA_VM_MM]
264 	.endm
265 
266 /*
267  * read_ctr - read CTR_EL0. If the system has mismatched register fields,
268  * provide the system wide safe value from arm64_ftr_reg_ctrel0.sys_val
269  */
270 	.macro	read_ctr, reg
271 #ifndef __KVM_NVHE_HYPERVISOR__
272 alternative_if_not ARM64_MISMATCHED_CACHE_TYPE
273 	mrs	\reg, ctr_el0			// read CTR
274 	nop
275 alternative_else
276 	ldr_l	\reg, arm64_ftr_reg_ctrel0 + ARM64_FTR_SYSVAL
277 alternative_endif
278 #else
279 alternative_if_not ARM64_KVM_PROTECTED_MODE
280 	ASM_BUG()
281 alternative_else_nop_endif
282 alternative_cb ARM64_ALWAYS_SYSTEM, kvm_compute_final_ctr_el0
283 	movz	\reg, #0
284 	movk	\reg, #0, lsl #16
285 	movk	\reg, #0, lsl #32
286 	movk	\reg, #0, lsl #48
287 alternative_cb_end
288 #endif
289 	.endm
290 
291 
292 /*
293  * raw_dcache_line_size - get the minimum D-cache line size on this CPU
294  * from the CTR register.
295  */
296 	.macro	raw_dcache_line_size, reg, tmp
297 	mrs	\tmp, ctr_el0			// read CTR
298 	ubfm	\tmp, \tmp, #16, #19		// cache line size encoding
299 	mov	\reg, #4			// bytes per word
300 	lsl	\reg, \reg, \tmp		// actual cache line size
301 	.endm
302 
303 /*
304  * dcache_line_size - get the safe D-cache line size across all CPUs
305  */
306 	.macro	dcache_line_size, reg, tmp
307 	read_ctr	\tmp
308 	ubfm		\tmp, \tmp, #16, #19	// cache line size encoding
309 	mov		\reg, #4		// bytes per word
310 	lsl		\reg, \reg, \tmp	// actual cache line size
311 	.endm
312 
313 /*
314  * raw_icache_line_size - get the minimum I-cache line size on this CPU
315  * from the CTR register.
316  */
317 	.macro	raw_icache_line_size, reg, tmp
318 	mrs	\tmp, ctr_el0			// read CTR
319 	and	\tmp, \tmp, #0xf		// cache line size encoding
320 	mov	\reg, #4			// bytes per word
321 	lsl	\reg, \reg, \tmp		// actual cache line size
322 	.endm
323 
324 /*
325  * icache_line_size - get the safe I-cache line size across all CPUs
326  */
327 	.macro	icache_line_size, reg, tmp
328 	read_ctr	\tmp
329 	and		\tmp, \tmp, #0xf	// cache line size encoding
330 	mov		\reg, #4		// bytes per word
331 	lsl		\reg, \reg, \tmp	// actual cache line size
332 	.endm
333 
334 /*
335  * tcr_set_t0sz - update TCR.T0SZ so that we can load the ID map
336  */
337 	.macro	tcr_set_t0sz, valreg, t0sz
338 	bfi	\valreg, \t0sz, #TCR_T0SZ_OFFSET, #TCR_TxSZ_WIDTH
339 	.endm
340 
341 /*
342  * tcr_set_t1sz - update TCR.T1SZ
343  */
344 	.macro	tcr_set_t1sz, valreg, t1sz
345 	bfi	\valreg, \t1sz, #TCR_T1SZ_OFFSET, #TCR_TxSZ_WIDTH
346 	.endm
347 
348 /*
349  * idmap_get_t0sz - get the T0SZ value needed to cover the ID map
350  *
351  * Calculate the maximum allowed value for TCR_EL1.T0SZ so that the
352  * entire ID map region can be mapped. As T0SZ == (64 - #bits used),
353  * this number conveniently equals the number of leading zeroes in
354  * the physical address of _end.
355  */
356 	.macro	idmap_get_t0sz, reg
357 	adrp	\reg, _end
358 	orr	\reg, \reg, #(1 << VA_BITS_MIN) - 1
359 	clz	\reg, \reg
360 	.endm
361 
362 /*
363  * tcr_compute_pa_size - set TCR.(I)PS to the highest supported
364  * ID_AA64MMFR0_EL1.PARange value
365  *
366  *	tcr:		register with the TCR_ELx value to be updated
367  *	pos:		IPS or PS bitfield position
368  *	tmp{0,1}:	temporary registers
369  */
370 	.macro	tcr_compute_pa_size, tcr, pos, tmp0, tmp1
371 	mrs	\tmp0, ID_AA64MMFR0_EL1
372 	// Narrow PARange to fit the PS field in TCR_ELx
373 	ubfx	\tmp0, \tmp0, #ID_AA64MMFR0_EL1_PARANGE_SHIFT, #3
374 	mov	\tmp1, #ID_AA64MMFR0_EL1_PARANGE_MAX
375 	cmp	\tmp0, \tmp1
376 	csel	\tmp0, \tmp1, \tmp0, hi
377 	bfi	\tcr, \tmp0, \pos, #3
378 	.endm
379 
380 	.macro __dcache_op_workaround_clean_cache, op, addr
381 alternative_if_not ARM64_WORKAROUND_CLEAN_CACHE
382 	dc	\op, \addr
383 alternative_else
384 	dc	civac, \addr
385 alternative_endif
386 	.endm
387 
388 /*
389  * Macro to perform a data cache maintenance for the interval
390  * [start, end) with dcache line size explicitly provided.
391  *
392  * 	op:		operation passed to dc instruction
393  * 	domain:		domain used in dsb instruciton
394  * 	start:          starting virtual address of the region
395  * 	end:            end virtual address of the region
396  *	linesz:		dcache line size
397  * 	fixup:		optional label to branch to on user fault
398  * 	Corrupts:       start, end, tmp
399  */
400 	.macro dcache_by_myline_op op, domain, start, end, linesz, tmp, fixup
401 	sub	\tmp, \linesz, #1
402 	bic	\start, \start, \tmp
403 .Ldcache_op\@:
404 	.ifc	\op, cvau
405 	__dcache_op_workaround_clean_cache \op, \start
406 	.else
407 	.ifc	\op, cvac
408 	__dcache_op_workaround_clean_cache \op, \start
409 	.else
410 	.ifc	\op, cvap
411 	sys	3, c7, c12, 1, \start	// dc cvap
412 	.else
413 	.ifc	\op, cvadp
414 	sys	3, c7, c13, 1, \start	// dc cvadp
415 	.else
416 	dc	\op, \start
417 	.endif
418 	.endif
419 	.endif
420 	.endif
421 	add	\start, \start, \linesz
422 	cmp	\start, \end
423 	b.lo	.Ldcache_op\@
424 	dsb	\domain
425 
426 	_cond_uaccess_extable .Ldcache_op\@, \fixup
427 	.endm
428 
429 /*
430  * Macro to perform a data cache maintenance for the interval
431  * [start, end)
432  *
433  * 	op:		operation passed to dc instruction
434  * 	domain:		domain used in dsb instruciton
435  * 	start:          starting virtual address of the region
436  * 	end:            end virtual address of the region
437  * 	fixup:		optional label to branch to on user fault
438  * 	Corrupts:       start, end, tmp1, tmp2
439  */
440 	.macro dcache_by_line_op op, domain, start, end, tmp1, tmp2, fixup
441 	dcache_line_size \tmp1, \tmp2
442 	dcache_by_myline_op \op, \domain, \start, \end, \tmp1, \tmp2, \fixup
443 	.endm
444 
445 /*
446  * Macro to perform an instruction cache maintenance for the interval
447  * [start, end)
448  *
449  * 	start, end:	virtual addresses describing the region
450  *	fixup:		optional label to branch to on user fault
451  * 	Corrupts:	tmp1, tmp2
452  */
453 	.macro invalidate_icache_by_line start, end, tmp1, tmp2, fixup
454 	icache_line_size \tmp1, \tmp2
455 	sub	\tmp2, \tmp1, #1
456 	bic	\tmp2, \start, \tmp2
457 .Licache_op\@:
458 	ic	ivau, \tmp2			// invalidate I line PoU
459 	add	\tmp2, \tmp2, \tmp1
460 	cmp	\tmp2, \end
461 	b.lo	.Licache_op\@
462 	dsb	ish
463 	isb
464 
465 	_cond_uaccess_extable .Licache_op\@, \fixup
466 	.endm
467 
468 /*
469  * load_ttbr1 - install @pgtbl as a TTBR1 page table
470  * pgtbl preserved
471  * tmp1/tmp2 clobbered, either may overlap with pgtbl
472  */
473 	.macro		load_ttbr1, pgtbl, tmp1, tmp2
474 	phys_to_ttbr	\tmp1, \pgtbl
475 	offset_ttbr1 	\tmp1, \tmp2
476 	msr		ttbr1_el1, \tmp1
477 	isb
478 	.endm
479 
480 /*
481  * To prevent the possibility of old and new partial table walks being visible
482  * in the tlb, switch the ttbr to a zero page when we invalidate the old
483  * records. D4.7.1 'General TLB maintenance requirements' in ARM DDI 0487A.i
484  * Even switching to our copied tables will cause a changed output address at
485  * each stage of the walk.
486  */
487 	.macro break_before_make_ttbr_switch zero_page, page_table, tmp, tmp2
488 	phys_to_ttbr \tmp, \zero_page
489 	msr	ttbr1_el1, \tmp
490 	isb
491 	tlbi	vmalle1
492 	dsb	nsh
493 	load_ttbr1 \page_table, \tmp, \tmp2
494 	.endm
495 
496 /*
497  * reset_pmuserenr_el0 - reset PMUSERENR_EL0 if PMUv3 present
498  */
499 	.macro	reset_pmuserenr_el0, tmpreg
500 	mrs	\tmpreg, id_aa64dfr0_el1
501 	sbfx	\tmpreg, \tmpreg, #ID_AA64DFR0_EL1_PMUVer_SHIFT, #4
502 	cmp	\tmpreg, #1			// Skip if no PMU present
503 	b.lt	9000f
504 	msr	pmuserenr_el0, xzr		// Disable PMU access from EL0
505 9000:
506 	.endm
507 
508 /*
509  * reset_amuserenr_el0 - reset AMUSERENR_EL0 if AMUv1 present
510  */
511 	.macro	reset_amuserenr_el0, tmpreg
512 	mrs	\tmpreg, id_aa64pfr0_el1	// Check ID_AA64PFR0_EL1
513 	ubfx	\tmpreg, \tmpreg, #ID_AA64PFR0_EL1_AMU_SHIFT, #4
514 	cbz	\tmpreg, .Lskip_\@		// Skip if no AMU present
515 	msr_s	SYS_AMUSERENR_EL0, xzr		// Disable AMU access from EL0
516 .Lskip_\@:
517 	.endm
518 /*
519  * copy_page - copy src to dest using temp registers t1-t8
520  */
521 	.macro copy_page dest:req src:req t1:req t2:req t3:req t4:req t5:req t6:req t7:req t8:req
522 9998:	ldp	\t1, \t2, [\src]
523 	ldp	\t3, \t4, [\src, #16]
524 	ldp	\t5, \t6, [\src, #32]
525 	ldp	\t7, \t8, [\src, #48]
526 	add	\src, \src, #64
527 	stnp	\t1, \t2, [\dest]
528 	stnp	\t3, \t4, [\dest, #16]
529 	stnp	\t5, \t6, [\dest, #32]
530 	stnp	\t7, \t8, [\dest, #48]
531 	add	\dest, \dest, #64
532 	tst	\src, #(PAGE_SIZE - 1)
533 	b.ne	9998b
534 	.endm
535 
536 /*
537  * Annotate a function as being unsuitable for kprobes.
538  */
539 #ifdef CONFIG_KPROBES
540 #define NOKPROBE(x)				\
541 	.pushsection "_kprobe_blacklist", "aw";	\
542 	.quad	x;				\
543 	.popsection;
544 #else
545 #define NOKPROBE(x)
546 #endif
547 
548 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
549 #define EXPORT_SYMBOL_NOKASAN(name)
550 #else
551 #define EXPORT_SYMBOL_NOKASAN(name)	EXPORT_SYMBOL(name)
552 #endif
553 
554 	/*
555 	 * Emit a 64-bit absolute little endian symbol reference in a way that
556 	 * ensures that it will be resolved at build time, even when building a
557 	 * PIE binary. This requires cooperation from the linker script, which
558 	 * must emit the lo32/hi32 halves individually.
559 	 */
560 	.macro	le64sym, sym
561 	.long	\sym\()_lo32
562 	.long	\sym\()_hi32
563 	.endm
564 
565 	/*
566 	 * mov_q - move an immediate constant into a 64-bit register using
567 	 *         between 2 and 4 movz/movk instructions (depending on the
568 	 *         magnitude and sign of the operand)
569 	 */
570 	.macro	mov_q, reg, val
571 	.if (((\val) >> 31) == 0 || ((\val) >> 31) == 0x1ffffffff)
572 	movz	\reg, :abs_g1_s:\val
573 	.else
574 	.if (((\val) >> 47) == 0 || ((\val) >> 47) == 0x1ffff)
575 	movz	\reg, :abs_g2_s:\val
576 	.else
577 	movz	\reg, :abs_g3:\val
578 	movk	\reg, :abs_g2_nc:\val
579 	.endif
580 	movk	\reg, :abs_g1_nc:\val
581 	.endif
582 	movk	\reg, :abs_g0_nc:\val
583 	.endm
584 
585 /*
586  * Return the current task_struct.
587  */
588 	.macro	get_current_task, rd
589 	mrs	\rd, sp_el0
590 	.endm
591 
592 /*
593  * Offset ttbr1 to allow for 48-bit kernel VAs set with 52-bit PTRS_PER_PGD.
594  * orr is used as it can cover the immediate value (and is idempotent).
595  * In future this may be nop'ed out when dealing with 52-bit kernel VAs.
596  * 	ttbr: Value of ttbr to set, modified.
597  */
598 	.macro	offset_ttbr1, ttbr, tmp
599 #ifdef CONFIG_ARM64_VA_BITS_52
600 	mrs_s	\tmp, SYS_ID_AA64MMFR2_EL1
601 	and	\tmp, \tmp, #(0xf << ID_AA64MMFR2_EL1_VARange_SHIFT)
602 	cbnz	\tmp, .Lskipoffs_\@
603 	orr	\ttbr, \ttbr, #TTBR1_BADDR_4852_OFFSET
604 .Lskipoffs_\@ :
605 #endif
606 	.endm
607 
608 /*
609  * Arrange a physical address in a TTBR register, taking care of 52-bit
610  * addresses.
611  *
612  * 	phys:	physical address, preserved
613  * 	ttbr:	returns the TTBR value
614  */
615 	.macro	phys_to_ttbr, ttbr, phys
616 #ifdef CONFIG_ARM64_PA_BITS_52
617 	orr	\ttbr, \phys, \phys, lsr #46
618 	and	\ttbr, \ttbr, #TTBR_BADDR_MASK_52
619 #else
620 	mov	\ttbr, \phys
621 #endif
622 	.endm
623 
624 	.macro	phys_to_pte, pte, phys
625 #ifdef CONFIG_ARM64_PA_BITS_52
626 	/*
627 	 * We assume \phys is 64K aligned and this is guaranteed by only
628 	 * supporting this configuration with 64K pages.
629 	 */
630 	orr	\pte, \phys, \phys, lsr #36
631 	and	\pte, \pte, #PTE_ADDR_MASK
632 #else
633 	mov	\pte, \phys
634 #endif
635 	.endm
636 
637 	.macro	pte_to_phys, phys, pte
638 	and	\phys, \pte, #PTE_ADDR_MASK
639 #ifdef CONFIG_ARM64_PA_BITS_52
640 	orr	\phys, \phys, \phys, lsl #PTE_ADDR_HIGH_SHIFT
641 	and	\phys, \phys, GENMASK_ULL(PHYS_MASK_SHIFT - 1, PAGE_SHIFT)
642 #endif
643 	.endm
644 
645 /*
646  * tcr_clear_errata_bits - Clear TCR bits that trigger an errata on this CPU.
647  */
648 	.macro	tcr_clear_errata_bits, tcr, tmp1, tmp2
649 #ifdef CONFIG_FUJITSU_ERRATUM_010001
650 	mrs	\tmp1, midr_el1
651 
652 	mov_q	\tmp2, MIDR_FUJITSU_ERRATUM_010001_MASK
653 	and	\tmp1, \tmp1, \tmp2
654 	mov_q	\tmp2, MIDR_FUJITSU_ERRATUM_010001
655 	cmp	\tmp1, \tmp2
656 	b.ne	10f
657 
658 	mov_q	\tmp2, TCR_CLEAR_FUJITSU_ERRATUM_010001
659 	bic	\tcr, \tcr, \tmp2
660 10:
661 #endif /* CONFIG_FUJITSU_ERRATUM_010001 */
662 	.endm
663 
664 /**
665  * Errata workaround prior to disable MMU. Insert an ISB immediately prior
666  * to executing the MSR that will change SCTLR_ELn[M] from a value of 1 to 0.
667  */
668 	.macro pre_disable_mmu_workaround
669 #ifdef CONFIG_QCOM_FALKOR_ERRATUM_E1041
670 	isb
671 #endif
672 	.endm
673 
674 	/*
675 	 * frame_push - Push @regcount callee saved registers to the stack,
676 	 *              starting at x19, as well as x29/x30, and set x29 to
677 	 *              the new value of sp. Add @extra bytes of stack space
678 	 *              for locals.
679 	 */
680 	.macro		frame_push, regcount:req, extra
681 	__frame		st, \regcount, \extra
682 	.endm
683 
684 	/*
685 	 * frame_pop  - Pop the callee saved registers from the stack that were
686 	 *              pushed in the most recent call to frame_push, as well
687 	 *              as x29/x30 and any extra stack space that may have been
688 	 *              allocated.
689 	 */
690 	.macro		frame_pop
691 	__frame		ld
692 	.endm
693 
694 	.macro		__frame_regs, reg1, reg2, op, num
695 	.if		.Lframe_regcount == \num
696 	\op\()r		\reg1, [sp, #(\num + 1) * 8]
697 	.elseif		.Lframe_regcount > \num
698 	\op\()p		\reg1, \reg2, [sp, #(\num + 1) * 8]
699 	.endif
700 	.endm
701 
702 	.macro		__frame, op, regcount, extra=0
703 	.ifc		\op, st
704 	.if		(\regcount) < 0 || (\regcount) > 10
705 	.error		"regcount should be in the range [0 ... 10]"
706 	.endif
707 	.if		((\extra) % 16) != 0
708 	.error		"extra should be a multiple of 16 bytes"
709 	.endif
710 	.ifdef		.Lframe_regcount
711 	.if		.Lframe_regcount != -1
712 	.error		"frame_push/frame_pop may not be nested"
713 	.endif
714 	.endif
715 	.set		.Lframe_regcount, \regcount
716 	.set		.Lframe_extra, \extra
717 	.set		.Lframe_local_offset, ((\regcount + 3) / 2) * 16
718 	stp		x29, x30, [sp, #-.Lframe_local_offset - .Lframe_extra]!
719 	mov		x29, sp
720 	.endif
721 
722 	__frame_regs	x19, x20, \op, 1
723 	__frame_regs	x21, x22, \op, 3
724 	__frame_regs	x23, x24, \op, 5
725 	__frame_regs	x25, x26, \op, 7
726 	__frame_regs	x27, x28, \op, 9
727 
728 	.ifc		\op, ld
729 	.if		.Lframe_regcount == -1
730 	.error		"frame_push/frame_pop may not be nested"
731 	.endif
732 	ldp		x29, x30, [sp], #.Lframe_local_offset + .Lframe_extra
733 	.set		.Lframe_regcount, -1
734 	.endif
735 	.endm
736 
737 /*
738  * Set SCTLR_ELx to the @reg value, and invalidate the local icache
739  * in the process. This is called when setting the MMU on.
740  */
741 .macro set_sctlr, sreg, reg
742 	msr	\sreg, \reg
743 	isb
744 	/*
745 	 * Invalidate the local I-cache so that any instructions fetched
746 	 * speculatively from the PoC are discarded, since they may have
747 	 * been dynamically patched at the PoU.
748 	 */
749 	ic	iallu
750 	dsb	nsh
751 	isb
752 .endm
753 
754 .macro set_sctlr_el1, reg
755 	set_sctlr sctlr_el1, \reg
756 .endm
757 
758 .macro set_sctlr_el2, reg
759 	set_sctlr sctlr_el2, \reg
760 .endm
761 
762 	/*
763 	 * Check whether asm code should yield as soon as it is able. This is
764 	 * the case if we are currently running in task context, and the
765 	 * TIF_NEED_RESCHED flag is set. (Note that the TIF_NEED_RESCHED flag
766 	 * is stored negated in the top word of the thread_info::preempt_count
767 	 * field)
768 	 */
769 	.macro		cond_yield, lbl:req, tmp:req, tmp2
770 #ifdef CONFIG_PREEMPT_VOLUNTARY
771 	get_current_task \tmp
772 	ldr		\tmp, [\tmp, #TSK_TI_PREEMPT]
773 	/*
774 	 * If we are serving a softirq, there is no point in yielding: the
775 	 * softirq will not be preempted no matter what we do, so we should
776 	 * run to completion as quickly as we can. The preempt_count field will
777 	 * have BIT(SOFTIRQ_SHIFT) set in this case, so the zero check will
778 	 * catch this case too.
779 	 */
780 	cbz		\tmp, \lbl
781 #endif
782 	.endm
783 
784 /*
785  * Branch Target Identifier (BTI)
786  */
787 	.macro  bti, targets
788 	.equ	.L__bti_targets_c, 34
789 	.equ	.L__bti_targets_j, 36
790 	.equ	.L__bti_targets_jc,38
791 	hint	#.L__bti_targets_\targets
792 	.endm
793 
794 /*
795  * This macro emits a program property note section identifying
796  * architecture features which require special handling, mainly for
797  * use in assembly files included in the VDSO.
798  */
799 
800 #define NT_GNU_PROPERTY_TYPE_0  5
801 #define GNU_PROPERTY_AARCH64_FEATURE_1_AND      0xc0000000
802 
803 #define GNU_PROPERTY_AARCH64_FEATURE_1_BTI      (1U << 0)
804 #define GNU_PROPERTY_AARCH64_FEATURE_1_PAC      (1U << 1)
805 
806 #ifdef CONFIG_ARM64_BTI_KERNEL
807 #define GNU_PROPERTY_AARCH64_FEATURE_1_DEFAULT		\
808 		((GNU_PROPERTY_AARCH64_FEATURE_1_BTI |	\
809 		  GNU_PROPERTY_AARCH64_FEATURE_1_PAC))
810 #endif
811 
812 #ifdef GNU_PROPERTY_AARCH64_FEATURE_1_DEFAULT
813 .macro emit_aarch64_feature_1_and, feat=GNU_PROPERTY_AARCH64_FEATURE_1_DEFAULT
814 	.pushsection .note.gnu.property, "a"
815 	.align  3
816 	.long   2f - 1f
817 	.long   6f - 3f
818 	.long   NT_GNU_PROPERTY_TYPE_0
819 1:      .string "GNU"
820 2:
821 	.align  3
822 3:      .long   GNU_PROPERTY_AARCH64_FEATURE_1_AND
823 	.long   5f - 4f
824 4:
825 	/*
826 	 * This is described with an array of char in the Linux API
827 	 * spec but the text and all other usage (including binutils,
828 	 * clang and GCC) treat this as a 32 bit value so no swizzling
829 	 * is required for big endian.
830 	 */
831 	.long   \feat
832 5:
833 	.align  3
834 6:
835 	.popsection
836 .endm
837 
838 #else
839 .macro emit_aarch64_feature_1_and, feat=0
840 .endm
841 
842 #endif /* GNU_PROPERTY_AARCH64_FEATURE_1_DEFAULT */
843 
844 	.macro __mitigate_spectre_bhb_loop      tmp
845 #ifdef CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY
846 alternative_cb ARM64_ALWAYS_SYSTEM, spectre_bhb_patch_loop_iter
847 	mov	\tmp, #32		// Patched to correct the immediate
848 alternative_cb_end
849 .Lspectre_bhb_loop\@:
850 	b	. + 4
851 	subs	\tmp, \tmp, #1
852 	b.ne	.Lspectre_bhb_loop\@
853 	sb
854 #endif /* CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY */
855 	.endm
856 
857 	.macro mitigate_spectre_bhb_loop	tmp
858 #ifdef CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY
859 alternative_cb ARM64_ALWAYS_SYSTEM, spectre_bhb_patch_loop_mitigation_enable
860 	b	.L_spectre_bhb_loop_done\@	// Patched to NOP
861 alternative_cb_end
862 	__mitigate_spectre_bhb_loop	\tmp
863 .L_spectre_bhb_loop_done\@:
864 #endif /* CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY */
865 	.endm
866 
867 	/* Save/restores x0-x3 to the stack */
868 	.macro __mitigate_spectre_bhb_fw
869 #ifdef CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY
870 	stp	x0, x1, [sp, #-16]!
871 	stp	x2, x3, [sp, #-16]!
872 	mov	w0, #ARM_SMCCC_ARCH_WORKAROUND_3
873 alternative_cb ARM64_ALWAYS_SYSTEM, smccc_patch_fw_mitigation_conduit
874 	nop					// Patched to SMC/HVC #0
875 alternative_cb_end
876 	ldp	x2, x3, [sp], #16
877 	ldp	x0, x1, [sp], #16
878 #endif /* CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY */
879 	.endm
880 
881 	.macro mitigate_spectre_bhb_clear_insn
882 #ifdef CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY
883 alternative_cb ARM64_ALWAYS_SYSTEM, spectre_bhb_patch_clearbhb
884 	/* Patched to NOP when not supported */
885 	clearbhb
886 	isb
887 alternative_cb_end
888 #endif /* CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY */
889 	.endm
890 #endif	/* __ASM_ASSEMBLER_H */
891