cpufunc_asm_armv8.S revision 1.8 1 1.8 ryo /* $NetBSD: cpufunc_asm_armv8.S,v 1.8 2021/02/11 08:35:11 ryo Exp $ */
2 1.1 ryo
3 1.1 ryo /*-
4 1.1 ryo * Copyright (c) 2014 Robin Randhawa
5 1.1 ryo * Copyright (c) 2015 The FreeBSD Foundation
6 1.1 ryo * All rights reserved.
7 1.1 ryo *
8 1.1 ryo * Portions of this software were developed by Andrew Turner
9 1.1 ryo * under sponsorship from the FreeBSD Foundation
10 1.1 ryo *
11 1.1 ryo * Redistribution and use in source and binary forms, with or without
12 1.1 ryo * modification, are permitted provided that the following conditions
13 1.1 ryo * are met:
14 1.1 ryo * 1. Redistributions of source code must retain the above copyright
15 1.1 ryo * notice, this list of conditions and the following disclaimer.
16 1.1 ryo * 2. Redistributions in binary form must reproduce the above copyright
17 1.1 ryo * notice, this list of conditions and the following disclaimer in the
18 1.1 ryo * documentation and/or other materials provided with the distribution.
19 1.1 ryo *
20 1.1 ryo * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
21 1.1 ryo * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 1.1 ryo * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 1.1 ryo * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
24 1.1 ryo * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 1.1 ryo * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 1.1 ryo * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 1.1 ryo * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 1.1 ryo * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 1.1 ryo * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 1.1 ryo * SUCH DAMAGE.
31 1.1 ryo *
32 1.1 ryo * $FreeBSD: head/sys/arm64/arm64/cpufunc_asm.S 313347 2017-02-06 17:50:09Z andrew $
33 1.1 ryo */
34 1.1 ryo
35 1.3 ryo #include "opt_cputypes.h"
36 1.8 ryo #include "opt_gprof.h"
37 1.1 ryo #include "opt_multiprocessor.h"
38 1.1 ryo #include <aarch64/asm.h>
39 1.1 ryo
40 1.1 ryo .text
41 1.1 ryo .align 2
42 1.1 ryo
43 1.1 ryo /*
44 1.1 ryo * Macro to handle the cache. This takes the start address in x0, length
45 1.5 ryo * in x1. It will corrupt x2-x5.
46 1.1 ryo */
47 1.7 ryo .macro cache_handle_range dcop = "", icop = ""
48 1.1 ryo mrs x3, ctr_el0
49 1.5 ryo mov x4, #4 /* size of word */
50 1.7 ryo .ifnb \dcop
51 1.1 ryo ubfx x2, x3, #16, #4 /* x2 = D cache shift */
52 1.5 ryo lsl x2, x4, x2 /* x2 = D cache line size */
53 1.5 ryo .endif
54 1.7 ryo .ifnb \icop
55 1.1 ryo and x3, x3, #15 /* x3 = I cache shift */
56 1.5 ryo lsl x3, x4, x3 /* x3 = I cache line size */
57 1.1 ryo .endif
58 1.7 ryo .ifnb \dcop
59 1.5 ryo sub x4, x2, #1 /* Get the address mask */
60 1.5 ryo and x4, x0, x4 /* Get the low bits of the address */
61 1.5 ryo add x5, x1, x4 /* Add these to the size */
62 1.5 ryo bic x4, x0, x4 /* Clear the low bit of the address */
63 1.1 ryo 1:
64 1.5 ryo dc \dcop, x4
65 1.5 ryo add x4, x4, x2 /* Move to the next line */
66 1.5 ryo subs x5, x5, x2 /* Reduce the size */
67 1.5 ryo b.hi 1b /* Check if we are done */
68 1.1 ryo dsb ish
69 1.1 ryo .endif
70 1.7 ryo .ifnb \icop
71 1.5 ryo sub x4, x3, #1 /* Get the address mask */
72 1.5 ryo and x4, x0, x4 /* Get the low bits of the address */
73 1.5 ryo add x5, x1, x4 /* Add these to the size */
74 1.5 ryo bic x4, x0, x4 /* Clear the low bit of the address */
75 1.5 ryo 1:
76 1.5 ryo ic \icop, x4
77 1.5 ryo add x4, x4, x3 /* Move to the next line */
78 1.5 ryo subs x5, x5, x3 /* Reduce the size */
79 1.1 ryo b.hi 1b /* Check if we are done */
80 1.5 ryo dsb ish
81 1.1 ryo isb
82 1.1 ryo .endif
83 1.1 ryo .endm
84 1.1 ryo
85 1.1 ryo
86 1.1 ryo ENTRY(aarch64_nullop)
87 1.1 ryo ret
88 1.1 ryo END(aarch64_nullop)
89 1.1 ryo
90 1.1 ryo ENTRY(aarch64_cpuid)
91 1.1 ryo mrs x0, midr_el1
92 1.1 ryo ret
93 1.1 ryo END(aarch64_cpuid)
94 1.1 ryo
95 1.1 ryo /*
96 1.1 ryo * void aarch64_dcache_wb_range(vaddr_t, vsize_t)
97 1.1 ryo */
98 1.1 ryo ENTRY(aarch64_dcache_wb_range)
99 1.1 ryo cache_handle_range dcop = cvac
100 1.5 ryo ret
101 1.1 ryo END(aarch64_dcache_wb_range)
102 1.1 ryo
103 1.1 ryo /*
104 1.1 ryo * void aarch64_dcache_wbinv_range(vaddr_t, vsize_t)
105 1.1 ryo */
106 1.1 ryo ENTRY(aarch64_dcache_wbinv_range)
107 1.1 ryo cache_handle_range dcop = civac
108 1.5 ryo ret
109 1.1 ryo END(aarch64_dcache_wbinv_range)
110 1.1 ryo
111 1.1 ryo /*
112 1.1 ryo * void aarch64_dcache_inv_range(vaddr_t, vsize_t)
113 1.1 ryo *
114 1.1 ryo * Note, we must not invalidate everything. If the range is too big we
115 1.1 ryo * must use wb-inv of the entire cache.
116 1.1 ryo */
117 1.1 ryo ENTRY(aarch64_dcache_inv_range)
118 1.1 ryo cache_handle_range dcop = ivac
119 1.5 ryo ret
120 1.1 ryo END(aarch64_dcache_inv_range)
121 1.1 ryo
122 1.1 ryo /*
123 1.1 ryo * void aarch64_idcache_wbinv_range(vaddr_t, vsize_t)
124 1.1 ryo */
125 1.1 ryo ENTRY(aarch64_idcache_wbinv_range)
126 1.5 ryo cache_handle_range dcop = civac, icop = ivau
127 1.5 ryo ret
128 1.1 ryo END(aarch64_idcache_wbinv_range)
129 1.1 ryo
130 1.1 ryo /*
131 1.1 ryo * void aarch64_icache_sync_range(vaddr_t, vsize_t)
132 1.1 ryo */
133 1.1 ryo ENTRY(aarch64_icache_sync_range)
134 1.5 ryo cache_handle_range dcop = cvau, icop = ivau
135 1.5 ryo ret
136 1.1 ryo END(aarch64_icache_sync_range)
137 1.1 ryo
138 1.1 ryo /*
139 1.6 ryo * void aarch64_icache_inv_range(vaddr_t, vsize_t)
140 1.6 ryo */
141 1.6 ryo ENTRY(aarch64_icache_inv_range)
142 1.6 ryo cache_handle_range icop = ivau
143 1.6 ryo ret
144 1.6 ryo END(aarch64_icache_inv_range)
145 1.6 ryo
146 1.6 ryo /*
147 1.6 ryo * void aarch64_icache_barrier_range(vaddr_t, vsize_t)
148 1.6 ryo */
149 1.6 ryo ENTRY(aarch64_icache_barrier_range)
150 1.6 ryo dsb ishst
151 1.6 ryo isb
152 1.6 ryo ret
153 1.6 ryo END(aarch64_icache_barrier_range)
154 1.6 ryo
155 1.6 ryo /*
156 1.1 ryo * void aarch64_icache_inv_all(void)
157 1.1 ryo */
158 1.1 ryo ENTRY(aarch64_icache_inv_all)
159 1.1 ryo dsb ish
160 1.1 ryo #ifdef MULTIPROCESSOR
161 1.2 ryo ic ialluis
162 1.2 ryo #else
163 1.1 ryo ic iallu
164 1.1 ryo #endif
165 1.1 ryo dsb ish
166 1.1 ryo isb
167 1.1 ryo ret
168 1.2 ryo END(aarch64_icache_inv_all)
169 1.1 ryo
170 1.1 ryo
171 1.1 ryo
172 1.1 ryo ENTRY(aarch64_drain_writebuf)
173 1.1 ryo dsb sy
174 1.1 ryo ret
175 1.1 ryo END(aarch64_drain_writebuf)
176 1.1 ryo
177 1.1 ryo
178 1.1 ryo /*
179 1.1 ryo * TLB ops
180 1.1 ryo */
181 1.1 ryo
182 1.1 ryo /* void aarch64_set_ttbr0(uint64_t ttbr0) */
183 1.1 ryo ENTRY(aarch64_set_ttbr0)
184 1.1 ryo dsb ish
185 1.1 ryo msr ttbr0_el1, x0
186 1.1 ryo dsb ish
187 1.1 ryo isb
188 1.1 ryo ret
189 1.1 ryo END(aarch64_set_ttbr0)
190 1.1 ryo
191 1.3 ryo #ifdef CPU_THUNDERX
192 1.3 ryo /*
193 1.3 ryo * Cavium erratum 27456
194 1.3 ryo * void aarch64_set_ttbr0_thunderx(uint64_t ttbr0)
195 1.3 ryo */
196 1.3 ryo ENTRY(aarch64_set_ttbr0_thunderx)
197 1.3 ryo dsb ish
198 1.3 ryo msr ttbr0_el1, x0
199 1.3 ryo isb
200 1.3 ryo ic iallu
201 1.3 ryo dsb nsh
202 1.3 ryo isb
203 1.3 ryo ret
204 1.3 ryo END(aarch64_set_ttbr0_thunderx)
205 1.3 ryo #endif /* CPU_THUNDERX */
206 1.3 ryo
207 1.1 ryo /* void aarch64_tlbi_all(void) */
208 1.1 ryo ENTRY(aarch64_tlbi_all)
209 1.1 ryo dsb ishst
210 1.1 ryo #ifdef MULTIPROCESSOR
211 1.1 ryo tlbi vmalle1is
212 1.1 ryo #else
213 1.1 ryo tlbi vmalle1
214 1.1 ryo #endif
215 1.1 ryo dsb ish
216 1.1 ryo isb
217 1.1 ryo ret
218 1.1 ryo END(aarch64_tlbi_all)
219 1.1 ryo
220 1.1 ryo /* void aarch64_tlbi_by_asid(int asid) */
221 1.1 ryo ENTRY(aarch64_tlbi_by_asid)
222 1.1 ryo /* x8 = bit 63[ASID]48, 47[RES0]0 */
223 1.1 ryo lsl x8, x0, #48
224 1.1 ryo dsb ishst
225 1.1 ryo #ifdef MULTIPROCESSOR
226 1.1 ryo tlbi aside1is, x8
227 1.1 ryo #else
228 1.1 ryo tlbi aside1, x8
229 1.1 ryo #endif
230 1.1 ryo dsb ish
231 1.1 ryo isb
232 1.1 ryo ret
233 1.1 ryo END(aarch64_tlbi_by_asid)
234 1.1 ryo
235 1.1 ryo /* aarch64_tlbi_by_va(vaddr_t va) */
236 1.1 ryo ENTRY(aarch64_tlbi_by_va)
237 1.1 ryo /* x8 = bit 63[RES0]44, 43[VA(55:12)]0 */
238 1.1 ryo ubfx x8, x0, #12, #44
239 1.1 ryo dsb ishst
240 1.1 ryo #ifdef MULTIPROCESSOR
241 1.1 ryo tlbi vaae1is, x8
242 1.1 ryo #else
243 1.1 ryo tlbi vaae1, x8
244 1.1 ryo #endif
245 1.1 ryo dsb ish
246 1.1 ryo isb
247 1.1 ryo ret
248 1.1 ryo END(aarch64_tlbi_by_va)
249 1.1 ryo
250 1.1 ryo /* aarch64_tlbi_by_va_ll(vaddr_t va) */
251 1.1 ryo ENTRY(aarch64_tlbi_by_va_ll)
252 1.1 ryo /* x8 = bit 63[RES0]44, 43[VA(55:12)]0 */
253 1.1 ryo ubfx x8, x0, #12, #44
254 1.1 ryo dsb ishst
255 1.1 ryo #ifdef MULTIPROCESSOR
256 1.1 ryo tlbi vaale1is, x8
257 1.1 ryo #else
258 1.1 ryo tlbi vaale1, x8
259 1.1 ryo #endif
260 1.1 ryo dsb ish
261 1.1 ryo isb
262 1.1 ryo ret
263 1.1 ryo END(aarch64_tlbi_by_va_ll)
264 1.1 ryo
265 1.1 ryo /* aarch64_tlbi_by_asid_va(int asid, vaddr_t va) */
266 1.1 ryo ENTRY(aarch64_tlbi_by_asid_va)
267 1.1 ryo /* x8 = bit 63[ASID]48, 47[RES0]44, 43[VA(55:12)]0 */
268 1.1 ryo lsl x8, x0, #48
269 1.1 ryo bfxil x8, x1, #12, #44
270 1.4 ryo dsb ishst
271 1.1 ryo #ifdef MULTIPROCESSOR
272 1.1 ryo tlbi vae1is, x8
273 1.1 ryo #else
274 1.1 ryo tlbi vae1, x8
275 1.1 ryo #endif
276 1.4 ryo dsb ish
277 1.4 ryo isb
278 1.1 ryo ret
279 1.1 ryo END(aarch64_tlbi_by_asid_va)
280 1.1 ryo
281 1.1 ryo /* aarch64_tlbi_by_asid_va_ll(int asid, vaddr_t va) */
282 1.1 ryo ENTRY(aarch64_tlbi_by_asid_va_ll)
283 1.1 ryo /* x8 = bit 63[ASID]48, 47[RES0]44, 43[VA(55:12)]0 */
284 1.1 ryo lsl x8, x0, #48
285 1.1 ryo bfxil x8, x1, #12, #44
286 1.4 ryo dsb ishst
287 1.1 ryo #ifdef MULTIPROCESSOR
288 1.1 ryo tlbi vale1is, x8
289 1.1 ryo #else
290 1.1 ryo tlbi vale1, x8
291 1.1 ryo #endif
292 1.4 ryo dsb ish
293 1.4 ryo isb
294 1.1 ryo ret
295 1.1 ryo END(aarch64_tlbi_by_asid_va_ll)
296