• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
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 <asm-generic/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_daif, flags
49 	mrs	\flags, daif
50 	msr	daifset, #0xf
51 	.endm
52 
53 	.macro	save_and_disable_irq, flags
54 	mrs	\flags, daif
55 	msr	daifset, #3
56 	.endm
57 
58 	.macro	restore_irq, flags
59 	msr	daif, \flags
60 	.endm
61 
62 	.macro	enable_dbg
63 	msr	daifclr, #8
64 	.endm
65 
66 	.macro	disable_step_tsk, flgs, tmp
67 	tbz	\flgs, #TIF_SINGLESTEP, 9990f
68 	mrs	\tmp, mdscr_el1
69 	bic	\tmp, \tmp, #DBG_MDSCR_SS
70 	msr	mdscr_el1, \tmp
71 	isb	// Synchronise with enable_dbg
72 9990:
73 	.endm
74 
75 	/* call with daif masked */
76 	.macro	enable_step_tsk, flgs, tmp
77 	tbz	\flgs, #TIF_SINGLESTEP, 9990f
78 	mrs	\tmp, mdscr_el1
79 	orr	\tmp, \tmp, #DBG_MDSCR_SS
80 	msr	mdscr_el1, \tmp
81 9990:
82 	.endm
83 
84 /*
85  * RAS Error Synchronization barrier
86  */
87 	.macro  esb
88 #ifdef CONFIG_ARM64_RAS_EXTN
89 	hint    #16
90 #else
91 	nop
92 #endif
93 	.endm
94 
95 /*
96  * Value prediction barrier
97  */
98 	.macro	csdb
99 	hint	#20
100 	.endm
101 
102 /*
103  * Clear Branch History instruction
104  */
105 	.macro clearbhb
106 	hint	#22
107 	.endm
108 
109 /*
110  * Speculation barrier
111  */
112 	.macro	sb
113 alternative_if_not ARM64_HAS_SB
114 	dsb	nsh
115 	isb
116 alternative_else
117 	SB_BARRIER_INSN
118 	nop
119 alternative_endif
120 	.endm
121 
122 /*
123  * NOP sequence
124  */
125 	.macro	nops, num
126 	.rept	\num
127 	nop
128 	.endr
129 	.endm
130 
131 /*
132  * Register aliases.
133  */
134 lr	.req	x30		// link register
135 
136 /*
137  * Vector entry
138  */
139 	 .macro	ventry	label
140 	.align	7
141 	b	\label
142 	.endm
143 
144 /*
145  * Select code when configured for BE.
146  */
147 #ifdef CONFIG_CPU_BIG_ENDIAN
148 #define CPU_BE(code...) code
149 #else
150 #define CPU_BE(code...)
151 #endif
152 
153 /*
154  * Select code when configured for LE.
155  */
156 #ifdef CONFIG_CPU_BIG_ENDIAN
157 #define CPU_LE(code...)
158 #else
159 #define CPU_LE(code...) code
160 #endif
161 
162 /*
163  * Define a macro that constructs a 64-bit value by concatenating two
164  * 32-bit registers. Note that on big endian systems the order of the
165  * registers is swapped.
166  */
167 #ifndef CONFIG_CPU_BIG_ENDIAN
168 	.macro	regs_to_64, rd, lbits, hbits
169 #else
170 	.macro	regs_to_64, rd, hbits, lbits
171 #endif
172 	orr	\rd, \lbits, \hbits, lsl #32
173 	.endm
174 
175 /*
176  * Pseudo-ops for PC-relative adr/ldr/str <reg>, <symbol> where
177  * <symbol> is within the range +/- 4 GB of the PC.
178  */
179 	/*
180 	 * @dst: destination register (64 bit wide)
181 	 * @sym: name of the symbol
182 	 */
183 	.macro	adr_l, dst, sym
184 	adrp	\dst, \sym
185 	add	\dst, \dst, :lo12:\sym
186 	.endm
187 
188 	/*
189 	 * @dst: destination register (32 or 64 bit wide)
190 	 * @sym: name of the symbol
191 	 * @tmp: optional 64-bit scratch register to be used if <dst> is a
192 	 *       32-bit wide register, in which case it cannot be used to hold
193 	 *       the address
194 	 */
195 	.macro	ldr_l, dst, sym, tmp=
196 	.ifb	\tmp
197 	adrp	\dst, \sym
198 	ldr	\dst, [\dst, :lo12:\sym]
199 	.else
200 	adrp	\tmp, \sym
201 	ldr	\dst, [\tmp, :lo12:\sym]
202 	.endif
203 	.endm
204 
205 	/*
206 	 * @src: source register (32 or 64 bit wide)
207 	 * @sym: name of the symbol
208 	 * @tmp: mandatory 64-bit scratch register to calculate the address
209 	 *       while <src> needs to be preserved.
210 	 */
211 	.macro	str_l, src, sym, tmp
212 	adrp	\tmp, \sym
213 	str	\src, [\tmp, :lo12:\sym]
214 	.endm
215 
216 	/*
217 	 * @dst: destination register
218 	 */
219 #if defined(__KVM_NVHE_HYPERVISOR__) || defined(__KVM_VHE_HYPERVISOR__)
220 	.macro	get_this_cpu_offset, dst
221 	mrs	\dst, tpidr_el2
222 	.endm
223 #else
224 	.macro	get_this_cpu_offset, dst
225 alternative_if_not ARM64_HAS_VIRT_HOST_EXTN
226 	mrs	\dst, tpidr_el1
227 alternative_else
228 	mrs	\dst, tpidr_el2
229 alternative_endif
230 	.endm
231 
232 	.macro	set_this_cpu_offset, src
233 alternative_if_not ARM64_HAS_VIRT_HOST_EXTN
234 	msr	tpidr_el1, \src
235 alternative_else
236 	msr	tpidr_el2, \src
237 alternative_endif
238 	.endm
239 #endif
240 
241 	/*
242 	 * @dst: Result of per_cpu(sym, smp_processor_id()) (can be SP)
243 	 * @sym: The name of the per-cpu variable
244 	 * @tmp: scratch register
245 	 */
246 	.macro adr_this_cpu, dst, sym, tmp
247 	adrp	\tmp, \sym
248 	add	\dst, \tmp, #:lo12:\sym
249 	get_this_cpu_offset \tmp
250 	add	\dst, \dst, \tmp
251 	.endm
252 
253 	/*
254 	 * @dst: Result of READ_ONCE(per_cpu(sym, smp_processor_id()))
255 	 * @sym: The name of the per-cpu variable
256 	 * @tmp: scratch register
257 	 */
258 	.macro ldr_this_cpu dst, sym, tmp
259 	adr_l	\dst, \sym
260 	get_this_cpu_offset \tmp
261 	ldr	\dst, [\dst, \tmp]
262 	.endm
263 
264 /*
265  * vma_vm_mm - get mm pointer from vma pointer (vma->vm_mm)
266  */
267 	.macro	vma_vm_mm, rd, rn
268 	ldr	\rd, [\rn, #VMA_VM_MM]
269 	.endm
270 
271 /*
272  * read_ctr - read CTR_EL0. If the system has mismatched register fields,
273  * provide the system wide safe value from arm64_ftr_reg_ctrel0.sys_val
274  */
275 	.macro	read_ctr, reg
276 #ifndef __KVM_NVHE_HYPERVISOR__
277 alternative_if_not ARM64_MISMATCHED_CACHE_TYPE
278 	mrs	\reg, ctr_el0			// read CTR
279 	nop
280 alternative_else
281 	ldr_l	\reg, arm64_ftr_reg_ctrel0 + ARM64_FTR_SYSVAL
282 alternative_endif
283 #else
284 alternative_if_not ARM64_KVM_PROTECTED_MODE
285 	ASM_BUG()
286 alternative_else_nop_endif
287 alternative_cb ARM64_ALWAYS_SYSTEM, kvm_compute_final_ctr_el0
288 	movz	\reg, #0
289 	movk	\reg, #0, lsl #16
290 	movk	\reg, #0, lsl #32
291 	movk	\reg, #0, lsl #48
292 alternative_cb_end
293 #endif
294 	.endm
295 
296 
297 /*
298  * raw_dcache_line_size - get the minimum D-cache line size on this CPU
299  * from the CTR register.
300  */
301 	.macro	raw_dcache_line_size, reg, tmp
302 	mrs	\tmp, ctr_el0			// read CTR
303 	ubfm	\tmp, \tmp, #16, #19		// cache line size encoding
304 	mov	\reg, #4			// bytes per word
305 	lsl	\reg, \reg, \tmp		// actual cache line size
306 	.endm
307 
308 /*
309  * dcache_line_size - get the safe D-cache line size across all CPUs
310  */
311 	.macro	dcache_line_size, reg, tmp
312 	read_ctr	\tmp
313 	ubfm		\tmp, \tmp, #16, #19	// cache line size encoding
314 	mov		\reg, #4		// bytes per word
315 	lsl		\reg, \reg, \tmp	// actual cache line size
316 	.endm
317 
318 /*
319  * raw_icache_line_size - get the minimum I-cache line size on this CPU
320  * from the CTR register.
321  */
322 	.macro	raw_icache_line_size, reg, tmp
323 	mrs	\tmp, ctr_el0			// read CTR
324 	and	\tmp, \tmp, #0xf		// cache line size encoding
325 	mov	\reg, #4			// bytes per word
326 	lsl	\reg, \reg, \tmp		// actual cache line size
327 	.endm
328 
329 /*
330  * icache_line_size - get the safe I-cache line size across all CPUs
331  */
332 	.macro	icache_line_size, reg, tmp
333 	read_ctr	\tmp
334 	and		\tmp, \tmp, #0xf	// cache line size encoding
335 	mov		\reg, #4		// bytes per word
336 	lsl		\reg, \reg, \tmp	// actual cache line size
337 	.endm
338 
339 /*
340  * tcr_set_t0sz - update TCR.T0SZ so that we can load the ID map
341  */
342 	.macro	tcr_set_t0sz, valreg, t0sz
343 	bfi	\valreg, \t0sz, #TCR_T0SZ_OFFSET, #TCR_TxSZ_WIDTH
344 	.endm
345 
346 /*
347  * tcr_set_t1sz - update TCR.T1SZ
348  */
349 	.macro	tcr_set_t1sz, valreg, t1sz
350 	bfi	\valreg, \t1sz, #TCR_T1SZ_OFFSET, #TCR_TxSZ_WIDTH
351 	.endm
352 
353 /*
354  * idmap_get_t0sz - get the T0SZ value needed to cover the ID map
355  *
356  * Calculate the maximum allowed value for TCR_EL1.T0SZ so that the
357  * entire ID map region can be mapped. As T0SZ == (64 - #bits used),
358  * this number conveniently equals the number of leading zeroes in
359  * the physical address of _end.
360  */
361 	.macro	idmap_get_t0sz, reg
362 	adrp	\reg, _end
363 	orr	\reg, \reg, #(1 << VA_BITS_MIN) - 1
364 	clz	\reg, \reg
365 	.endm
366 
367 /*
368  * tcr_compute_pa_size - set TCR.(I)PS to the highest supported
369  * ID_AA64MMFR0_EL1.PARange value
370  *
371  *	tcr:		register with the TCR_ELx value to be updated
372  *	pos:		IPS or PS bitfield position
373  *	tmp{0,1}:	temporary registers
374  */
375 	.macro	tcr_compute_pa_size, tcr, pos, tmp0, tmp1
376 	mrs	\tmp0, ID_AA64MMFR0_EL1
377 	// Narrow PARange to fit the PS field in TCR_ELx
378 	ubfx	\tmp0, \tmp0, #ID_AA64MMFR0_EL1_PARANGE_SHIFT, #3
379 	mov	\tmp1, #ID_AA64MMFR0_EL1_PARANGE_MAX
380 	cmp	\tmp0, \tmp1
381 	csel	\tmp0, \tmp1, \tmp0, hi
382 	bfi	\tcr, \tmp0, \pos, #3
383 	.endm
384 
385 	.macro __dcache_op_workaround_clean_cache, op, addr
386 alternative_if_not ARM64_WORKAROUND_CLEAN_CACHE
387 	dc	\op, \addr
388 alternative_else
389 	dc	civac, \addr
390 alternative_endif
391 	.endm
392 
393 /*
394  * Macro to perform a data cache maintenance for the interval
395  * [start, end) with dcache line size explicitly provided.
396  *
397  * 	op:		operation passed to dc instruction
398  * 	domain:		domain used in dsb instruciton
399  * 	start:          starting virtual address of the region
400  * 	end:            end virtual address of the region
401  *	linesz:		dcache line size
402  * 	fixup:		optional label to branch to on user fault
403  * 	Corrupts:       start, end, tmp
404  */
405 	.macro dcache_by_myline_op op, domain, start, end, linesz, tmp, fixup
406 	sub	\tmp, \linesz, #1
407 	bic	\start, \start, \tmp
408 .Ldcache_op\@:
409 	.ifc	\op, cvau
410 	__dcache_op_workaround_clean_cache \op, \start
411 	.else
412 	.ifc	\op, cvac
413 	__dcache_op_workaround_clean_cache \op, \start
414 	.else
415 	.ifc	\op, cvap
416 	sys	3, c7, c12, 1, \start	// dc cvap
417 	.else
418 	.ifc	\op, cvadp
419 	sys	3, c7, c13, 1, \start	// dc cvadp
420 	.else
421 	dc	\op, \start
422 	.endif
423 	.endif
424 	.endif
425 	.endif
426 	add	\start, \start, \linesz
427 	cmp	\start, \end
428 	b.lo	.Ldcache_op\@
429 	dsb	\domain
430 
431 	_cond_uaccess_extable .Ldcache_op\@, \fixup
432 	.endm
433 
434 /*
435  * Macro to perform a data cache maintenance for the interval
436  * [start, end)
437  *
438  * 	op:		operation passed to dc instruction
439  * 	domain:		domain used in dsb instruciton
440  * 	start:          starting virtual address of the region
441  * 	end:            end virtual address of the region
442  * 	fixup:		optional label to branch to on user fault
443  * 	Corrupts:       start, end, tmp1, tmp2
444  */
445 	.macro dcache_by_line_op op, domain, start, end, tmp1, tmp2, fixup
446 	dcache_line_size \tmp1, \tmp2
447 	dcache_by_myline_op \op, \domain, \start, \end, \tmp1, \tmp2, \fixup
448 	.endm
449 
450 /*
451  * Macro to perform an instruction cache maintenance for the interval
452  * [start, end)
453  *
454  * 	start, end:	virtual addresses describing the region
455  *	fixup:		optional label to branch to on user fault
456  * 	Corrupts:	tmp1, tmp2
457  */
458 	.macro invalidate_icache_by_line start, end, tmp1, tmp2, fixup
459 	icache_line_size \tmp1, \tmp2
460 	sub	\tmp2, \tmp1, #1
461 	bic	\tmp2, \start, \tmp2
462 .Licache_op\@:
463 	ic	ivau, \tmp2			// invalidate I line PoU
464 	add	\tmp2, \tmp2, \tmp1
465 	cmp	\tmp2, \end
466 	b.lo	.Licache_op\@
467 	dsb	ish
468 	isb
469 
470 	_cond_uaccess_extable .Licache_op\@, \fixup
471 	.endm
472 
473 /*
474  * load_ttbr1 - install @pgtbl as a TTBR1 page table
475  * pgtbl preserved
476  * tmp1/tmp2 clobbered, either may overlap with pgtbl
477  */
478 	.macro		load_ttbr1, pgtbl, tmp1, tmp2
479 	phys_to_ttbr	\tmp1, \pgtbl
480 	offset_ttbr1 	\tmp1, \tmp2
481 	msr		ttbr1_el1, \tmp1
482 	isb
483 	.endm
484 
485 /*
486  * To prevent the possibility of old and new partial table walks being visible
487  * in the tlb, switch the ttbr to a zero page when we invalidate the old
488  * records. D4.7.1 'General TLB maintenance requirements' in ARM DDI 0487A.i
489  * Even switching to our copied tables will cause a changed output address at
490  * each stage of the walk.
491  */
492 	.macro break_before_make_ttbr_switch zero_page, page_table, tmp, tmp2
493 	phys_to_ttbr \tmp, \zero_page
494 	msr	ttbr1_el1, \tmp
495 	isb
496 	tlbi	vmalle1
497 	dsb	nsh
498 	load_ttbr1 \page_table, \tmp, \tmp2
499 	.endm
500 
501 /*
502  * reset_pmuserenr_el0 - reset PMUSERENR_EL0 if PMUv3 present
503  */
504 	.macro	reset_pmuserenr_el0, tmpreg
505 	mrs	\tmpreg, id_aa64dfr0_el1
506 	sbfx	\tmpreg, \tmpreg, #ID_AA64DFR0_EL1_PMUVer_SHIFT, #4
507 	cmp	\tmpreg, #1			// Skip if no PMU present
508 	b.lt	9000f
509 	msr	pmuserenr_el0, xzr		// Disable PMU access from EL0
510 9000:
511 	.endm
512 
513 /*
514  * reset_amuserenr_el0 - reset AMUSERENR_EL0 if AMUv1 present
515  */
516 	.macro	reset_amuserenr_el0, tmpreg
517 	mrs	\tmpreg, id_aa64pfr0_el1	// Check ID_AA64PFR0_EL1
518 	ubfx	\tmpreg, \tmpreg, #ID_AA64PFR0_EL1_AMU_SHIFT, #4
519 	cbz	\tmpreg, .Lskip_\@		// Skip if no AMU present
520 	msr_s	SYS_AMUSERENR_EL0, xzr		// Disable AMU access from EL0
521 .Lskip_\@:
522 	.endm
523 /*
524  * copy_page - copy src to dest using temp registers t1-t8
525  */
526 	.macro copy_page dest:req src:req t1:req t2:req t3:req t4:req t5:req t6:req t7:req t8:req
527 9998:	ldp	\t1, \t2, [\src]
528 	ldp	\t3, \t4, [\src, #16]
529 	ldp	\t5, \t6, [\src, #32]
530 	ldp	\t7, \t8, [\src, #48]
531 	add	\src, \src, #64
532 	stnp	\t1, \t2, [\dest]
533 	stnp	\t3, \t4, [\dest, #16]
534 	stnp	\t5, \t6, [\dest, #32]
535 	stnp	\t7, \t8, [\dest, #48]
536 	add	\dest, \dest, #64
537 	tst	\src, #(PAGE_SIZE - 1)
538 	b.ne	9998b
539 	.endm
540 
541 /*
542  * Annotate a function as being unsuitable for kprobes.
543  */
544 #ifdef CONFIG_KPROBES
545 #define NOKPROBE(x)				\
546 	.pushsection "_kprobe_blacklist", "aw";	\
547 	.quad	x;				\
548 	.popsection;
549 #else
550 #define NOKPROBE(x)
551 #endif
552 
553 #if defined(CONFIG_KASAN_GENERIC) || defined(CONFIG_KASAN_SW_TAGS)
554 #define EXPORT_SYMBOL_NOKASAN(name)
555 #else
556 #define EXPORT_SYMBOL_NOKASAN(name)	EXPORT_SYMBOL(name)
557 #endif
558 
559 	/*
560 	 * Emit a 64-bit absolute little endian symbol reference in a way that
561 	 * ensures that it will be resolved at build time, even when building a
562 	 * PIE binary. This requires cooperation from the linker script, which
563 	 * must emit the lo32/hi32 halves individually.
564 	 */
565 	.macro	le64sym, sym
566 	.long	\sym\()_lo32
567 	.long	\sym\()_hi32
568 	.endm
569 
570 	/*
571 	 * mov_q - move an immediate constant into a 64-bit register using
572 	 *         between 2 and 4 movz/movk instructions (depending on the
573 	 *         magnitude and sign of the operand)
574 	 */
575 	.macro	mov_q, reg, val
576 	.if (((\val) >> 31) == 0 || ((\val) >> 31) == 0x1ffffffff)
577 	movz	\reg, :abs_g1_s:\val
578 	.else
579 	.if (((\val) >> 47) == 0 || ((\val) >> 47) == 0x1ffff)
580 	movz	\reg, :abs_g2_s:\val
581 	.else
582 	movz	\reg, :abs_g3:\val
583 	movk	\reg, :abs_g2_nc:\val
584 	.endif
585 	movk	\reg, :abs_g1_nc:\val
586 	.endif
587 	movk	\reg, :abs_g0_nc:\val
588 	.endm
589 
590 /*
591  * Return the current task_struct.
592  */
593 	.macro	get_current_task, rd
594 	mrs	\rd, sp_el0
595 	.endm
596 
597 /*
598  * Offset ttbr1 to allow for 48-bit kernel VAs set with 52-bit PTRS_PER_PGD.
599  * orr is used as it can cover the immediate value (and is idempotent).
600  * In future this may be nop'ed out when dealing with 52-bit kernel VAs.
601  * 	ttbr: Value of ttbr to set, modified.
602  */
603 	.macro	offset_ttbr1, ttbr, tmp
604 #ifdef CONFIG_ARM64_VA_BITS_52
605 	mrs_s	\tmp, SYS_ID_AA64MMFR2_EL1
606 	and	\tmp, \tmp, #(0xf << ID_AA64MMFR2_EL1_VARange_SHIFT)
607 	cbnz	\tmp, .Lskipoffs_\@
608 	orr	\ttbr, \ttbr, #TTBR1_BADDR_4852_OFFSET
609 .Lskipoffs_\@ :
610 #endif
611 	.endm
612 
613 /*
614  * Arrange a physical address in a TTBR register, taking care of 52-bit
615  * addresses.
616  *
617  * 	phys:	physical address, preserved
618  * 	ttbr:	returns the TTBR value
619  */
620 	.macro	phys_to_ttbr, ttbr, phys
621 #ifdef CONFIG_ARM64_PA_BITS_52
622 	orr	\ttbr, \phys, \phys, lsr #46
623 	and	\ttbr, \ttbr, #TTBR_BADDR_MASK_52
624 #else
625 	mov	\ttbr, \phys
626 #endif
627 	.endm
628 
629 	.macro	phys_to_pte, pte, phys
630 #ifdef CONFIG_ARM64_PA_BITS_52
631 	/*
632 	 * We assume \phys is 64K aligned and this is guaranteed by only
633 	 * supporting this configuration with 64K pages.
634 	 */
635 	orr	\pte, \phys, \phys, lsr #36
636 	and	\pte, \pte, #PTE_ADDR_MASK
637 #else
638 	mov	\pte, \phys
639 #endif
640 	.endm
641 
642 	.macro	pte_to_phys, phys, pte
643 	and	\phys, \pte, #PTE_ADDR_MASK
644 #ifdef CONFIG_ARM64_PA_BITS_52
645 	orr	\phys, \phys, \phys, lsl #PTE_ADDR_HIGH_SHIFT
646 	and	\phys, \phys, GENMASK_ULL(PHYS_MASK_SHIFT - 1, PAGE_SHIFT)
647 #endif
648 	.endm
649 
650 /*
651  * tcr_clear_errata_bits - Clear TCR bits that trigger an errata on this CPU.
652  */
653 	.macro	tcr_clear_errata_bits, tcr, tmp1, tmp2
654 #ifdef CONFIG_FUJITSU_ERRATUM_010001
655 	mrs	\tmp1, midr_el1
656 
657 	mov_q	\tmp2, MIDR_FUJITSU_ERRATUM_010001_MASK
658 	and	\tmp1, \tmp1, \tmp2
659 	mov_q	\tmp2, MIDR_FUJITSU_ERRATUM_010001
660 	cmp	\tmp1, \tmp2
661 	b.ne	10f
662 
663 	mov_q	\tmp2, TCR_CLEAR_FUJITSU_ERRATUM_010001
664 	bic	\tcr, \tcr, \tmp2
665 10:
666 #endif /* CONFIG_FUJITSU_ERRATUM_010001 */
667 	.endm
668 
669 /**
670  * Errata workaround prior to disable MMU. Insert an ISB immediately prior
671  * to executing the MSR that will change SCTLR_ELn[M] from a value of 1 to 0.
672  */
673 	.macro pre_disable_mmu_workaround
674 #ifdef CONFIG_QCOM_FALKOR_ERRATUM_E1041
675 	isb
676 #endif
677 	.endm
678 
679 	/*
680 	 * frame_push - Push @regcount callee saved registers to the stack,
681 	 *              starting at x19, as well as x29/x30, and set x29 to
682 	 *              the new value of sp. Add @extra bytes of stack space
683 	 *              for locals.
684 	 */
685 	.macro		frame_push, regcount:req, extra
686 	__frame		st, \regcount, \extra
687 	.endm
688 
689 	/*
690 	 * frame_pop  - Pop the callee saved registers from the stack that were
691 	 *              pushed in the most recent call to frame_push, as well
692 	 *              as x29/x30 and any extra stack space that may have been
693 	 *              allocated.
694 	 */
695 	.macro		frame_pop
696 	__frame		ld
697 	.endm
698 
699 	.macro		__frame_regs, reg1, reg2, op, num
700 	.if		.Lframe_regcount == \num
701 	\op\()r		\reg1, [sp, #(\num + 1) * 8]
702 	.elseif		.Lframe_regcount > \num
703 	\op\()p		\reg1, \reg2, [sp, #(\num + 1) * 8]
704 	.endif
705 	.endm
706 
707 	.macro		__frame, op, regcount, extra=0
708 	.ifc		\op, st
709 	.if		(\regcount) < 0 || (\regcount) > 10
710 	.error		"regcount should be in the range [0 ... 10]"
711 	.endif
712 	.if		((\extra) % 16) != 0
713 	.error		"extra should be a multiple of 16 bytes"
714 	.endif
715 	.ifdef		.Lframe_regcount
716 	.if		.Lframe_regcount != -1
717 	.error		"frame_push/frame_pop may not be nested"
718 	.endif
719 	.endif
720 	.set		.Lframe_regcount, \regcount
721 	.set		.Lframe_extra, \extra
722 	.set		.Lframe_local_offset, ((\regcount + 3) / 2) * 16
723 	stp		x29, x30, [sp, #-.Lframe_local_offset - .Lframe_extra]!
724 	mov		x29, sp
725 	.endif
726 
727 	__frame_regs	x19, x20, \op, 1
728 	__frame_regs	x21, x22, \op, 3
729 	__frame_regs	x23, x24, \op, 5
730 	__frame_regs	x25, x26, \op, 7
731 	__frame_regs	x27, x28, \op, 9
732 
733 	.ifc		\op, ld
734 	.if		.Lframe_regcount == -1
735 	.error		"frame_push/frame_pop may not be nested"
736 	.endif
737 	ldp		x29, x30, [sp], #.Lframe_local_offset + .Lframe_extra
738 	.set		.Lframe_regcount, -1
739 	.endif
740 	.endm
741 
742 /*
743  * Set SCTLR_ELx to the @reg value, and invalidate the local icache
744  * in the process. This is called when setting the MMU on.
745  */
746 .macro set_sctlr, sreg, reg
747 	msr	\sreg, \reg
748 	isb
749 	/*
750 	 * Invalidate the local I-cache so that any instructions fetched
751 	 * speculatively from the PoC are discarded, since they may have
752 	 * been dynamically patched at the PoU.
753 	 */
754 	ic	iallu
755 	dsb	nsh
756 	isb
757 .endm
758 
759 .macro set_sctlr_el1, reg
760 	set_sctlr sctlr_el1, \reg
761 .endm
762 
763 .macro set_sctlr_el2, reg
764 	set_sctlr sctlr_el2, \reg
765 .endm
766 
767 	/*
768 	 * Check whether preempt/bh-disabled asm code should yield as soon as
769 	 * it is able. This is the case if we are currently running in task
770 	 * context, and either a softirq is pending, or the TIF_NEED_RESCHED
771 	 * flag is set and re-enabling preemption a single time would result in
772 	 * a preempt count of zero. (Note that the TIF_NEED_RESCHED flag is
773 	 * stored negated in the top word of the thread_info::preempt_count
774 	 * field)
775 	 */
776 	.macro		cond_yield, lbl:req, tmp:req, tmp2:req
777 	get_current_task \tmp
778 	ldr		\tmp, [\tmp, #TSK_TI_PREEMPT]
779 	/*
780 	 * If we are serving a softirq, there is no point in yielding: the
781 	 * softirq will not be preempted no matter what we do, so we should
782 	 * run to completion as quickly as we can.
783 	 */
784 	tbnz		\tmp, #SOFTIRQ_SHIFT, .Lnoyield_\@
785 #ifdef CONFIG_PREEMPTION
786 	sub		\tmp, \tmp, #PREEMPT_DISABLE_OFFSET
787 	cbz		\tmp, \lbl
788 #endif
789 	adr_l		\tmp, irq_stat + IRQ_CPUSTAT_SOFTIRQ_PENDING
790 	get_this_cpu_offset	\tmp2
791 	ldr		w\tmp, [\tmp, \tmp2]
792 	cbnz		w\tmp, \lbl	// yield on pending softirq in task context
793 .Lnoyield_\@:
794 	.endm
795 
796 /*
797  * Branch Target Identifier (BTI)
798  */
799 	.macro  bti, targets
800 	.equ	.L__bti_targets_c, 34
801 	.equ	.L__bti_targets_j, 36
802 	.equ	.L__bti_targets_jc,38
803 	hint	#.L__bti_targets_\targets
804 	.endm
805 
806 /*
807  * This macro emits a program property note section identifying
808  * architecture features which require special handling, mainly for
809  * use in assembly files included in the VDSO.
810  */
811 
812 #define NT_GNU_PROPERTY_TYPE_0  5
813 #define GNU_PROPERTY_AARCH64_FEATURE_1_AND      0xc0000000
814 
815 #define GNU_PROPERTY_AARCH64_FEATURE_1_BTI      (1U << 0)
816 #define GNU_PROPERTY_AARCH64_FEATURE_1_PAC      (1U << 1)
817 
818 #ifdef CONFIG_ARM64_BTI_KERNEL
819 #define GNU_PROPERTY_AARCH64_FEATURE_1_DEFAULT		\
820 		((GNU_PROPERTY_AARCH64_FEATURE_1_BTI |	\
821 		  GNU_PROPERTY_AARCH64_FEATURE_1_PAC))
822 #endif
823 
824 #ifdef GNU_PROPERTY_AARCH64_FEATURE_1_DEFAULT
825 .macro emit_aarch64_feature_1_and, feat=GNU_PROPERTY_AARCH64_FEATURE_1_DEFAULT
826 	.pushsection .note.gnu.property, "a"
827 	.align  3
828 	.long   2f - 1f
829 	.long   6f - 3f
830 	.long   NT_GNU_PROPERTY_TYPE_0
831 1:      .string "GNU"
832 2:
833 	.align  3
834 3:      .long   GNU_PROPERTY_AARCH64_FEATURE_1_AND
835 	.long   5f - 4f
836 4:
837 	/*
838 	 * This is described with an array of char in the Linux API
839 	 * spec but the text and all other usage (including binutils,
840 	 * clang and GCC) treat this as a 32 bit value so no swizzling
841 	 * is required for big endian.
842 	 */
843 	.long   \feat
844 5:
845 	.align  3
846 6:
847 	.popsection
848 .endm
849 
850 #else
851 .macro emit_aarch64_feature_1_and, feat=0
852 .endm
853 
854 #endif /* GNU_PROPERTY_AARCH64_FEATURE_1_DEFAULT */
855 
856 	.macro __mitigate_spectre_bhb_loop      tmp
857 #ifdef CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY
858 alternative_cb ARM64_ALWAYS_SYSTEM, spectre_bhb_patch_loop_iter
859 	mov	\tmp, #32		// Patched to correct the immediate
860 alternative_cb_end
861 .Lspectre_bhb_loop\@:
862 	b	. + 4
863 	subs	\tmp, \tmp, #1
864 	b.ne	.Lspectre_bhb_loop\@
865 	sb
866 #endif /* CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY */
867 	.endm
868 
869 	.macro mitigate_spectre_bhb_loop	tmp
870 #ifdef CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY
871 alternative_cb ARM64_ALWAYS_SYSTEM, spectre_bhb_patch_loop_mitigation_enable
872 	b	.L_spectre_bhb_loop_done\@	// Patched to NOP
873 alternative_cb_end
874 	__mitigate_spectre_bhb_loop	\tmp
875 .L_spectre_bhb_loop_done\@:
876 #endif /* CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY */
877 	.endm
878 
879 	/* Save/restores x0-x3 to the stack */
880 	.macro __mitigate_spectre_bhb_fw
881 #ifdef CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY
882 	stp	x0, x1, [sp, #-16]!
883 	stp	x2, x3, [sp, #-16]!
884 	mov	w0, #ARM_SMCCC_ARCH_WORKAROUND_3
885 alternative_cb ARM64_ALWAYS_SYSTEM, smccc_patch_fw_mitigation_conduit
886 	nop					// Patched to SMC/HVC #0
887 alternative_cb_end
888 	ldp	x2, x3, [sp], #16
889 	ldp	x0, x1, [sp], #16
890 #endif /* CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY */
891 	.endm
892 
893 	.macro mitigate_spectre_bhb_clear_insn
894 #ifdef CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY
895 alternative_cb ARM64_ALWAYS_SYSTEM, spectre_bhb_patch_clearbhb
896 	/* Patched to NOP when not supported */
897 	clearbhb
898 	isb
899 alternative_cb_end
900 #endif /* CONFIG_MITIGATE_SPECTRE_BRANCH_HISTORY */
901 	.endm
902 #endif	/* __ASM_ASSEMBLER_H */
903