xref: /linux/arch/arm64/include/asm/assembler.h (revision 32786fdc9506aeba98278c1844d4bfb766863832)
1 /*
2  * Based on arch/arm/include/asm/assembler.h, arch/arm/mm/proc-macros.S
3  *
4  * Copyright (C) 1996-2000 Russell King
5  * Copyright (C) 2012 ARM Ltd.
6  *
7  * This program is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License version 2 as
9  * published by the Free Software Foundation.
10  *
11  * This program is distributed in the hope that it will be useful,
12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
14  * GNU General Public License for more details.
15  *
16  * You should have received a copy of the GNU General Public License
17  * along with this program.  If not, see <http://www.gnu.org/licenses/>.
18  */
19 #ifndef __ASSEMBLY__
20 #error "Only include this from assembly code"
21 #endif
22 
23 #ifndef __ASM_ASSEMBLER_H
24 #define __ASM_ASSEMBLER_H
25 
26 #include <asm/asm-offsets.h>
27 #include <asm/cpufeature.h>
28 #include <asm/page.h>
29 #include <asm/pgtable-hwdef.h>
30 #include <asm/ptrace.h>
31 #include <asm/thread_info.h>
32 
33 /*
34  * Enable and disable interrupts.
35  */
36 	.macro	disable_irq
37 	msr	daifset, #2
38 	.endm
39 
40 	.macro	enable_irq
41 	msr	daifclr, #2
42 	.endm
43 
44 	.macro	save_and_disable_irq, flags
45 	mrs	\flags, daif
46 	msr	daifset, #2
47 	.endm
48 
49 	.macro	restore_irq, flags
50 	msr	daif, \flags
51 	.endm
52 
53 /*
54  * Enable and disable debug exceptions.
55  */
56 	.macro	disable_dbg
57 	msr	daifset, #8
58 	.endm
59 
60 	.macro	enable_dbg
61 	msr	daifclr, #8
62 	.endm
63 
64 	.macro	disable_step_tsk, flgs, tmp
65 	tbz	\flgs, #TIF_SINGLESTEP, 9990f
66 	mrs	\tmp, mdscr_el1
67 	bic	\tmp, \tmp, #1
68 	msr	mdscr_el1, \tmp
69 	isb	// Synchronise with enable_dbg
70 9990:
71 	.endm
72 
73 	.macro	enable_step_tsk, flgs, tmp
74 	tbz	\flgs, #TIF_SINGLESTEP, 9990f
75 	disable_dbg
76 	mrs	\tmp, mdscr_el1
77 	orr	\tmp, \tmp, #1
78 	msr	mdscr_el1, \tmp
79 9990:
80 	.endm
81 
82 /*
83  * Enable both debug exceptions and interrupts. This is likely to be
84  * faster than two daifclr operations, since writes to this register
85  * are self-synchronising.
86  */
87 	.macro	enable_dbg_and_irq
88 	msr	daifclr, #(8 | 2)
89 	.endm
90 
91 /*
92  * SMP data memory barrier
93  */
94 	.macro	smp_dmb, opt
95 	dmb	\opt
96 	.endm
97 
98 /*
99  * NOP sequence
100  */
101 	.macro	nops, num
102 	.rept	\num
103 	nop
104 	.endr
105 	.endm
106 
107 /*
108  * Emit an entry into the exception table
109  */
110 	.macro		_asm_extable, from, to
111 	.pushsection	__ex_table, "a"
112 	.align		3
113 	.long		(\from - .), (\to - .)
114 	.popsection
115 	.endm
116 
117 #define USER(l, x...)				\
118 9999:	x;					\
119 	_asm_extable	9999b, l
120 
121 /*
122  * Register aliases.
123  */
124 lr	.req	x30		// link register
125 
126 /*
127  * Vector entry
128  */
129 	 .macro	ventry	label
130 	.align	7
131 	b	\label
132 	.endm
133 
134 /*
135  * Select code when configured for BE.
136  */
137 #ifdef CONFIG_CPU_BIG_ENDIAN
138 #define CPU_BE(code...) code
139 #else
140 #define CPU_BE(code...)
141 #endif
142 
143 /*
144  * Select code when configured for LE.
145  */
146 #ifdef CONFIG_CPU_BIG_ENDIAN
147 #define CPU_LE(code...)
148 #else
149 #define CPU_LE(code...) code
150 #endif
151 
152 /*
153  * Define a macro that constructs a 64-bit value by concatenating two
154  * 32-bit registers. Note that on big endian systems the order of the
155  * registers is swapped.
156  */
157 #ifndef CONFIG_CPU_BIG_ENDIAN
158 	.macro	regs_to_64, rd, lbits, hbits
159 #else
160 	.macro	regs_to_64, rd, hbits, lbits
161 #endif
162 	orr	\rd, \lbits, \hbits, lsl #32
163 	.endm
164 
165 /*
166  * Pseudo-ops for PC-relative adr/ldr/str <reg>, <symbol> where
167  * <symbol> is within the range +/- 4 GB of the PC.
168  */
169 	/*
170 	 * @dst: destination register (64 bit wide)
171 	 * @sym: name of the symbol
172 	 * @tmp: optional scratch register to be used if <dst> == sp, which
173 	 *       is not allowed in an adrp instruction
174 	 */
175 	.macro	adr_l, dst, sym, tmp=
176 	.ifb	\tmp
177 	adrp	\dst, \sym
178 	add	\dst, \dst, :lo12:\sym
179 	.else
180 	adrp	\tmp, \sym
181 	add	\dst, \tmp, :lo12:\sym
182 	.endif
183 	.endm
184 
185 	/*
186 	 * @dst: destination register (32 or 64 bit wide)
187 	 * @sym: name of the symbol
188 	 * @tmp: optional 64-bit scratch register to be used if <dst> is a
189 	 *       32-bit wide register, in which case it cannot be used to hold
190 	 *       the address
191 	 */
192 	.macro	ldr_l, dst, sym, tmp=
193 	.ifb	\tmp
194 	adrp	\dst, \sym
195 	ldr	\dst, [\dst, :lo12:\sym]
196 	.else
197 	adrp	\tmp, \sym
198 	ldr	\dst, [\tmp, :lo12:\sym]
199 	.endif
200 	.endm
201 
202 	/*
203 	 * @src: source register (32 or 64 bit wide)
204 	 * @sym: name of the symbol
205 	 * @tmp: mandatory 64-bit scratch register to calculate the address
206 	 *       while <src> needs to be preserved.
207 	 */
208 	.macro	str_l, src, sym, tmp
209 	adrp	\tmp, \sym
210 	str	\src, [\tmp, :lo12:\sym]
211 	.endm
212 
213 	/*
214 	 * @dst: Result of per_cpu(sym, smp_processor_id())
215 	 * @sym: The name of the per-cpu variable
216 	 * @tmp: scratch register
217 	 */
218 	.macro adr_this_cpu, dst, sym, tmp
219 	adr_l	\dst, \sym
220 	mrs	\tmp, tpidr_el1
221 	add	\dst, \dst, \tmp
222 	.endm
223 
224 	/*
225 	 * @dst: Result of READ_ONCE(per_cpu(sym, smp_processor_id()))
226 	 * @sym: The name of the per-cpu variable
227 	 * @tmp: scratch register
228 	 */
229 	.macro ldr_this_cpu dst, sym, tmp
230 	adr_l	\dst, \sym
231 	mrs	\tmp, tpidr_el1
232 	ldr	\dst, [\dst, \tmp]
233 	.endm
234 
235 /*
236  * vma_vm_mm - get mm pointer from vma pointer (vma->vm_mm)
237  */
238 	.macro	vma_vm_mm, rd, rn
239 	ldr	\rd, [\rn, #VMA_VM_MM]
240 	.endm
241 
242 /*
243  * mmid - get context id from mm pointer (mm->context.id)
244  */
245 	.macro	mmid, rd, rn
246 	ldr	\rd, [\rn, #MM_CONTEXT_ID]
247 	.endm
248 /*
249  * read_ctr - read CTR_EL0. If the system has mismatched
250  * cache line sizes, provide the system wide safe value
251  * from arm64_ftr_reg_ctrel0.sys_val
252  */
253 	.macro	read_ctr, reg
254 alternative_if_not ARM64_MISMATCHED_CACHE_LINE_SIZE
255 	mrs	\reg, ctr_el0			// read CTR
256 	nop
257 alternative_else
258 	ldr_l	\reg, arm64_ftr_reg_ctrel0 + ARM64_FTR_SYSVAL
259 alternative_endif
260 	.endm
261 
262 
263 /*
264  * raw_dcache_line_size - get the minimum D-cache line size on this CPU
265  * from the CTR register.
266  */
267 	.macro	raw_dcache_line_size, reg, tmp
268 	mrs	\tmp, ctr_el0			// read CTR
269 	ubfm	\tmp, \tmp, #16, #19		// cache line size encoding
270 	mov	\reg, #4			// bytes per word
271 	lsl	\reg, \reg, \tmp		// actual cache line size
272 	.endm
273 
274 /*
275  * dcache_line_size - get the safe D-cache line size across all CPUs
276  */
277 	.macro	dcache_line_size, reg, tmp
278 	read_ctr	\tmp
279 	ubfm		\tmp, \tmp, #16, #19	// cache line size encoding
280 	mov		\reg, #4		// bytes per word
281 	lsl		\reg, \reg, \tmp	// actual cache line size
282 	.endm
283 
284 /*
285  * raw_icache_line_size - get the minimum I-cache line size on this CPU
286  * from the CTR register.
287  */
288 	.macro	raw_icache_line_size, reg, tmp
289 	mrs	\tmp, ctr_el0			// read CTR
290 	and	\tmp, \tmp, #0xf		// cache line size encoding
291 	mov	\reg, #4			// bytes per word
292 	lsl	\reg, \reg, \tmp		// actual cache line size
293 	.endm
294 
295 /*
296  * icache_line_size - get the safe I-cache line size across all CPUs
297  */
298 	.macro	icache_line_size, reg, tmp
299 	read_ctr	\tmp
300 	and		\tmp, \tmp, #0xf	// cache line size encoding
301 	mov		\reg, #4		// bytes per word
302 	lsl		\reg, \reg, \tmp	// actual cache line size
303 	.endm
304 
305 /*
306  * tcr_set_idmap_t0sz - update TCR.T0SZ so that we can load the ID map
307  */
308 	.macro	tcr_set_idmap_t0sz, valreg, tmpreg
309 #ifndef CONFIG_ARM64_VA_BITS_48
310 	ldr_l	\tmpreg, idmap_t0sz
311 	bfi	\valreg, \tmpreg, #TCR_T0SZ_OFFSET, #TCR_TxSZ_WIDTH
312 #endif
313 	.endm
314 
315 /*
316  * Macro to perform a data cache maintenance for the interval
317  * [kaddr, kaddr + size)
318  *
319  * 	op:		operation passed to dc instruction
320  * 	domain:		domain used in dsb instruciton
321  * 	kaddr:		starting virtual address of the region
322  * 	size:		size of the region
323  * 	Corrupts:	kaddr, size, tmp1, tmp2
324  */
325 	.macro dcache_by_line_op op, domain, kaddr, size, tmp1, tmp2
326 	dcache_line_size \tmp1, \tmp2
327 	add	\size, \kaddr, \size
328 	sub	\tmp2, \tmp1, #1
329 	bic	\kaddr, \kaddr, \tmp2
330 9998:
331 	.if	(\op == cvau || \op == cvac)
332 alternative_if_not ARM64_WORKAROUND_CLEAN_CACHE
333 	dc	\op, \kaddr
334 alternative_else
335 	dc	civac, \kaddr
336 alternative_endif
337 	.else
338 	dc	\op, \kaddr
339 	.endif
340 	add	\kaddr, \kaddr, \tmp1
341 	cmp	\kaddr, \size
342 	b.lo	9998b
343 	dsb	\domain
344 	.endm
345 
346 /*
347  * reset_pmuserenr_el0 - reset PMUSERENR_EL0 if PMUv3 present
348  */
349 	.macro	reset_pmuserenr_el0, tmpreg
350 	mrs	\tmpreg, id_aa64dfr0_el1	// Check ID_AA64DFR0_EL1 PMUVer
351 	sbfx	\tmpreg, \tmpreg, #8, #4
352 	cmp	\tmpreg, #1			// Skip if no PMU present
353 	b.lt	9000f
354 	msr	pmuserenr_el0, xzr		// Disable PMU access from EL0
355 9000:
356 	.endm
357 
358 /*
359  * copy_page - copy src to dest using temp registers t1-t8
360  */
361 	.macro copy_page dest:req src:req t1:req t2:req t3:req t4:req t5:req t6:req t7:req t8:req
362 9998:	ldp	\t1, \t2, [\src]
363 	ldp	\t3, \t4, [\src, #16]
364 	ldp	\t5, \t6, [\src, #32]
365 	ldp	\t7, \t8, [\src, #48]
366 	add	\src, \src, #64
367 	stnp	\t1, \t2, [\dest]
368 	stnp	\t3, \t4, [\dest, #16]
369 	stnp	\t5, \t6, [\dest, #32]
370 	stnp	\t7, \t8, [\dest, #48]
371 	add	\dest, \dest, #64
372 	tst	\src, #(PAGE_SIZE - 1)
373 	b.ne	9998b
374 	.endm
375 
376 /*
377  * Annotate a function as position independent, i.e., safe to be called before
378  * the kernel virtual mapping is activated.
379  */
380 #define ENDPIPROC(x)			\
381 	.globl	__pi_##x;		\
382 	.type 	__pi_##x, %function;	\
383 	.set	__pi_##x, x;		\
384 	.size	__pi_##x, . - x;	\
385 	ENDPROC(x)
386 
387 	/*
388 	 * Emit a 64-bit absolute little endian symbol reference in a way that
389 	 * ensures that it will be resolved at build time, even when building a
390 	 * PIE binary. This requires cooperation from the linker script, which
391 	 * must emit the lo32/hi32 halves individually.
392 	 */
393 	.macro	le64sym, sym
394 	.long	\sym\()_lo32
395 	.long	\sym\()_hi32
396 	.endm
397 
398 	/*
399 	 * mov_q - move an immediate constant into a 64-bit register using
400 	 *         between 2 and 4 movz/movk instructions (depending on the
401 	 *         magnitude and sign of the operand)
402 	 */
403 	.macro	mov_q, reg, val
404 	.if (((\val) >> 31) == 0 || ((\val) >> 31) == 0x1ffffffff)
405 	movz	\reg, :abs_g1_s:\val
406 	.else
407 	.if (((\val) >> 47) == 0 || ((\val) >> 47) == 0x1ffff)
408 	movz	\reg, :abs_g2_s:\val
409 	.else
410 	movz	\reg, :abs_g3:\val
411 	movk	\reg, :abs_g2_nc:\val
412 	.endif
413 	movk	\reg, :abs_g1_nc:\val
414 	.endif
415 	movk	\reg, :abs_g0_nc:\val
416 	.endm
417 
418 /*
419  * Return the current thread_info.
420  */
421 	.macro	get_thread_info, rd
422 	mrs	\rd, sp_el0
423 	.endm
424 
425 /*
426  * Errata workaround post TTBR0_EL1 update.
427  */
428 	.macro	post_ttbr0_update_workaround
429 #ifdef CONFIG_CAVIUM_ERRATUM_27456
430 alternative_if ARM64_WORKAROUND_CAVIUM_27456
431 	ic	iallu
432 	dsb	nsh
433 	isb
434 alternative_else_nop_endif
435 #endif
436 	.endm
437 
438 #endif	/* __ASM_ASSEMBLER_H */
439