2 * Copyright 2014 Svatopluk Kraus <onwahe@gmail.com>
3 * Copyright 2014 Michal Meloun <meloun@miracle.cz>
6 * Redistribution and use in source and binary forms, with or without
7 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
11 * 2. Redistributions in binary form must reproduce the above copyright
12 * notice, this list of conditions and the following disclaimer in the
13 * documentation and/or other materials provided with the distribution.
15 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
16 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
17 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
18 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
19 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
20 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
21 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
22 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
23 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
24 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 #ifndef MACHINE_CPU_V6_H
30 #define MACHINE_CPU_V6_H
32 /* There are no user serviceable parts here, they may change without notice */
34 #error Only include this file in the kernel
37 #include <machine/acle-compat.h>
38 #include "machine/atomic.h"
39 #include "machine/cpufunc.h"
40 #include "machine/cpuinfo.h"
41 #include "machine/sysreg.h"
43 #define CPU_ASID_KERNEL 0
45 vm_offset_t dcache_wb_pou_checked(vm_offset_t, vm_size_t);
46 vm_offset_t icache_inv_pou_checked(vm_offset_t, vm_size_t);
49 * Macros to generate CP15 (system control processor) read/write functions.
53 #define _RF0(fname, aname...) \
54 static __inline register_t \
58 __asm __volatile("mrc\t" _FX(aname): "=r" (reg)); \
62 #define _R64F0(fname, aname) \
63 static __inline uint64_t \
67 __asm __volatile("mrrc\t" _FX(aname): "=r" (reg)); \
71 #define _WF0(fname, aname...) \
72 static __inline void \
75 __asm __volatile("mcr\t" _FX(aname)); \
78 #define _WF1(fname, aname...) \
79 static __inline void \
80 fname(register_t reg) \
82 __asm __volatile("mcr\t" _FX(aname):: "r" (reg)); \
85 #define _W64F1(fname, aname...) \
86 static __inline void \
89 __asm __volatile("mcrr\t" _FX(aname):: "r" (reg)); \
93 * Raw CP15 maintenance operations
94 * !!! not for external use !!!
99 _WF0(_CP15_TLBIALL, CP15_TLBIALL) /* Invalidate entire unified TLB */
100 #if __ARM_ARCH >= 7 && defined SMP
101 _WF0(_CP15_TLBIALLIS, CP15_TLBIALLIS) /* Invalidate entire unified TLB IS */
103 _WF1(_CP15_TLBIASID, CP15_TLBIASID(%0)) /* Invalidate unified TLB by ASID */
104 #if __ARM_ARCH >= 7 && defined SMP
105 _WF1(_CP15_TLBIASIDIS, CP15_TLBIASIDIS(%0)) /* Invalidate unified TLB by ASID IS */
107 _WF1(_CP15_TLBIMVAA, CP15_TLBIMVAA(%0)) /* Invalidate unified TLB by MVA, all ASID */
108 #if __ARM_ARCH >= 7 && defined SMP
109 _WF1(_CP15_TLBIMVAAIS, CP15_TLBIMVAAIS(%0)) /* Invalidate unified TLB by MVA, all ASID IS */
111 _WF1(_CP15_TLBIMVA, CP15_TLBIMVA(%0)) /* Invalidate unified TLB by MVA */
113 _WF1(_CP15_TTB_SET, CP15_TTBR0(%0))
115 /* Cache and Branch predictor */
117 _WF0(_CP15_BPIALL, CP15_BPIALL) /* Branch predictor invalidate all */
118 #if __ARM_ARCH >= 7 && defined SMP
119 _WF0(_CP15_BPIALLIS, CP15_BPIALLIS) /* Branch predictor invalidate all IS */
121 _WF1(_CP15_BPIMVA, CP15_BPIMVA(%0)) /* Branch predictor invalidate by MVA */
122 _WF1(_CP15_DCCIMVAC, CP15_DCCIMVAC(%0)) /* Data cache clean and invalidate by MVA PoC */
123 _WF1(_CP15_DCCISW, CP15_DCCISW(%0)) /* Data cache clean and invalidate by set/way */
124 _WF1(_CP15_DCCMVAC, CP15_DCCMVAC(%0)) /* Data cache clean by MVA PoC */
126 _WF1(_CP15_DCCMVAU, CP15_DCCMVAU(%0)) /* Data cache clean by MVA PoU */
128 _WF1(_CP15_DCCSW, CP15_DCCSW(%0)) /* Data cache clean by set/way */
129 _WF1(_CP15_DCIMVAC, CP15_DCIMVAC(%0)) /* Data cache invalidate by MVA PoC */
130 _WF1(_CP15_DCISW, CP15_DCISW(%0)) /* Data cache invalidate by set/way */
131 _WF0(_CP15_ICIALLU, CP15_ICIALLU) /* Instruction cache invalidate all PoU */
132 #if __ARM_ARCH >= 7 && defined SMP
133 _WF0(_CP15_ICIALLUIS, CP15_ICIALLUIS) /* Instruction cache invalidate all PoU IS */
135 _WF1(_CP15_ICIMVAU, CP15_ICIMVAU(%0)) /* Instruction cache invalidate */
138 * Publicly accessible functions
141 /* CP14 Debug Registers */
142 _RF0(cp14_dbgdidr_get, CP14_DBGDIDR(%0))
143 _RF0(cp14_dbgprsr_get, CP14_DBGPRSR(%0))
144 _RF0(cp14_dbgoslsr_get, CP14_DBGOSLSR(%0))
145 _RF0(cp14_dbgosdlr_get, CP14_DBGOSDLR(%0))
146 _RF0(cp14_dbgdscrint_get, CP14_DBGDSCRint(%0))
148 _WF1(cp14_dbgdscr_v6_set, CP14_DBGDSCRext_V6(%0))
149 _WF1(cp14_dbgdscr_v7_set, CP14_DBGDSCRext_V7(%0))
150 _WF1(cp14_dbgvcr_set, CP14_DBGVCR(%0))
151 _WF1(cp14_dbgoslar_set, CP14_DBGOSLAR(%0))
153 /* Various control registers */
155 _RF0(cp15_cpacr_get, CP15_CPACR(%0))
156 _WF1(cp15_cpacr_set, CP15_CPACR(%0))
157 _RF0(cp15_dfsr_get, CP15_DFSR(%0))
158 _RF0(cp15_ifsr_get, CP15_IFSR(%0))
159 _WF1(cp15_prrr_set, CP15_PRRR(%0))
160 _WF1(cp15_nmrr_set, CP15_NMRR(%0))
161 _RF0(cp15_ttbr_get, CP15_TTBR0(%0))
162 _RF0(cp15_dfar_get, CP15_DFAR(%0))
164 _RF0(cp15_ifar_get, CP15_IFAR(%0))
165 _RF0(cp15_l2ctlr_get, CP15_L2CTLR(%0))
167 /* ARMv6+ and XScale */
168 _RF0(cp15_actlr_get, CP15_ACTLR(%0))
169 _WF1(cp15_actlr_set, CP15_ACTLR(%0))
171 _WF1(cp15_ats1cpr_set, CP15_ATS1CPR(%0))
172 _WF1(cp15_ats1cpw_set, CP15_ATS1CPW(%0))
173 _RF0(cp15_par_get, CP15_PAR(%0))
174 _RF0(cp15_sctlr_get, CP15_SCTLR(%0))
177 /*CPU id registers */
178 _RF0(cp15_midr_get, CP15_MIDR(%0))
179 _RF0(cp15_ctr_get, CP15_CTR(%0))
180 _RF0(cp15_tcmtr_get, CP15_TCMTR(%0))
181 _RF0(cp15_tlbtr_get, CP15_TLBTR(%0))
182 _RF0(cp15_mpidr_get, CP15_MPIDR(%0))
183 _RF0(cp15_revidr_get, CP15_REVIDR(%0))
184 _RF0(cp15_ccsidr_get, CP15_CCSIDR(%0))
185 _RF0(cp15_clidr_get, CP15_CLIDR(%0))
186 _RF0(cp15_aidr_get, CP15_AIDR(%0))
187 _WF1(cp15_csselr_set, CP15_CSSELR(%0))
188 _RF0(cp15_id_pfr0_get, CP15_ID_PFR0(%0))
189 _RF0(cp15_id_pfr1_get, CP15_ID_PFR1(%0))
190 _RF0(cp15_id_dfr0_get, CP15_ID_DFR0(%0))
191 _RF0(cp15_id_afr0_get, CP15_ID_AFR0(%0))
192 _RF0(cp15_id_mmfr0_get, CP15_ID_MMFR0(%0))
193 _RF0(cp15_id_mmfr1_get, CP15_ID_MMFR1(%0))
194 _RF0(cp15_id_mmfr2_get, CP15_ID_MMFR2(%0))
195 _RF0(cp15_id_mmfr3_get, CP15_ID_MMFR3(%0))
196 _RF0(cp15_id_isar0_get, CP15_ID_ISAR0(%0))
197 _RF0(cp15_id_isar1_get, CP15_ID_ISAR1(%0))
198 _RF0(cp15_id_isar2_get, CP15_ID_ISAR2(%0))
199 _RF0(cp15_id_isar3_get, CP15_ID_ISAR3(%0))
200 _RF0(cp15_id_isar4_get, CP15_ID_ISAR4(%0))
201 _RF0(cp15_id_isar5_get, CP15_ID_ISAR5(%0))
202 _RF0(cp15_cbar_get, CP15_CBAR(%0))
204 /* Performance Monitor registers */
206 #if __ARM_ARCH == 6 && defined(CPU_ARM1176)
207 _RF0(cp15_pmuserenr_get, CP15_PMUSERENR(%0))
208 _WF1(cp15_pmuserenr_set, CP15_PMUSERENR(%0))
209 _RF0(cp15_pmcr_get, CP15_PMCR(%0))
210 _WF1(cp15_pmcr_set, CP15_PMCR(%0))
211 _RF0(cp15_pmccntr_get, CP15_PMCCNTR(%0))
212 _WF1(cp15_pmccntr_set, CP15_PMCCNTR(%0))
214 _RF0(cp15_pmcr_get, CP15_PMCR(%0))
215 _WF1(cp15_pmcr_set, CP15_PMCR(%0))
216 _RF0(cp15_pmcnten_get, CP15_PMCNTENSET(%0))
217 _WF1(cp15_pmcnten_set, CP15_PMCNTENSET(%0))
218 _WF1(cp15_pmcnten_clr, CP15_PMCNTENCLR(%0))
219 _RF0(cp15_pmovsr_get, CP15_PMOVSR(%0))
220 _WF1(cp15_pmovsr_set, CP15_PMOVSR(%0))
221 _WF1(cp15_pmswinc_set, CP15_PMSWINC(%0))
222 _RF0(cp15_pmselr_get, CP15_PMSELR(%0))
223 _WF1(cp15_pmselr_set, CP15_PMSELR(%0))
224 _RF0(cp15_pmccntr_get, CP15_PMCCNTR(%0))
225 _WF1(cp15_pmccntr_set, CP15_PMCCNTR(%0))
226 _RF0(cp15_pmxevtyper_get, CP15_PMXEVTYPER(%0))
227 _WF1(cp15_pmxevtyper_set, CP15_PMXEVTYPER(%0))
228 _RF0(cp15_pmxevcntr_get, CP15_PMXEVCNTRR(%0))
229 _WF1(cp15_pmxevcntr_set, CP15_PMXEVCNTRR(%0))
230 _RF0(cp15_pmuserenr_get, CP15_PMUSERENR(%0))
231 _WF1(cp15_pmuserenr_set, CP15_PMUSERENR(%0))
232 _RF0(cp15_pminten_get, CP15_PMINTENSET(%0))
233 _WF1(cp15_pminten_set, CP15_PMINTENSET(%0))
234 _WF1(cp15_pminten_clr, CP15_PMINTENCLR(%0))
237 _RF0(cp15_tpidrurw_get, CP15_TPIDRURW(%0))
238 _WF1(cp15_tpidrurw_set, CP15_TPIDRURW(%0))
239 _RF0(cp15_tpidruro_get, CP15_TPIDRURO(%0))
240 _WF1(cp15_tpidruro_set, CP15_TPIDRURO(%0))
241 _RF0(cp15_tpidrpwr_get, CP15_TPIDRPRW(%0))
242 _WF1(cp15_tpidrpwr_set, CP15_TPIDRPRW(%0))
244 /* Generic Timer registers - only use when you know the hardware is available */
245 _RF0(cp15_cntfrq_get, CP15_CNTFRQ(%0))
246 _WF1(cp15_cntfrq_set, CP15_CNTFRQ(%0))
247 _RF0(cp15_cntkctl_get, CP15_CNTKCTL(%0))
248 _WF1(cp15_cntkctl_set, CP15_CNTKCTL(%0))
249 _RF0(cp15_cntp_tval_get, CP15_CNTP_TVAL(%0))
250 _WF1(cp15_cntp_tval_set, CP15_CNTP_TVAL(%0))
251 _RF0(cp15_cntp_ctl_get, CP15_CNTP_CTL(%0))
252 _WF1(cp15_cntp_ctl_set, CP15_CNTP_CTL(%0))
253 _RF0(cp15_cntv_tval_get, CP15_CNTV_TVAL(%0))
254 _WF1(cp15_cntv_tval_set, CP15_CNTV_TVAL(%0))
255 _RF0(cp15_cntv_ctl_get, CP15_CNTV_CTL(%0))
256 _WF1(cp15_cntv_ctl_set, CP15_CNTV_CTL(%0))
257 _RF0(cp15_cnthctl_get, CP15_CNTHCTL(%0))
258 _WF1(cp15_cnthctl_set, CP15_CNTHCTL(%0))
259 _RF0(cp15_cnthp_tval_get, CP15_CNTHP_TVAL(%0))
260 _WF1(cp15_cnthp_tval_set, CP15_CNTHP_TVAL(%0))
261 _RF0(cp15_cnthp_ctl_get, CP15_CNTHP_CTL(%0))
262 _WF1(cp15_cnthp_ctl_set, CP15_CNTHP_CTL(%0))
264 _R64F0(cp15_cntpct_get, CP15_CNTPCT(%Q0, %R0))
265 _R64F0(cp15_cntvct_get, CP15_CNTVCT(%Q0, %R0))
266 _R64F0(cp15_cntp_cval_get, CP15_CNTP_CVAL(%Q0, %R0))
267 _W64F1(cp15_cntp_cval_set, CP15_CNTP_CVAL(%Q0, %R0))
268 _R64F0(cp15_cntv_cval_get, CP15_CNTV_CVAL(%Q0, %R0))
269 _W64F1(cp15_cntv_cval_set, CP15_CNTV_CVAL(%Q0, %R0))
270 _R64F0(cp15_cntvoff_get, CP15_CNTVOFF(%Q0, %R0))
271 _W64F1(cp15_cntvoff_set, CP15_CNTVOFF(%Q0, %R0))
272 _R64F0(cp15_cnthp_cval_get, CP15_CNTHP_CVAL(%Q0, %R0))
273 _W64F1(cp15_cnthp_cval_set, CP15_CNTHP_CVAL(%Q0, %R0))
282 * Cache and TLB maintenance operations for armv6+ code. The #else block
283 * provides armv4/v5 implementations for a few of these used in common code.
287 * TLB maintenance operations.
290 /* Local (i.e. not broadcasting ) operations. */
292 /* Flush all TLB entries (even global). */
294 tlb_flush_all_local(void)
302 /* Flush all not global TLB entries. */
304 tlb_flush_all_ng_local(void)
308 _CP15_TLBIASID(CPU_ASID_KERNEL);
312 /* Flush single TLB entry (even global). */
314 tlb_flush_local(vm_offset_t va)
317 KASSERT((va & PAGE_MASK) == 0, ("%s: va %#x not aligned", __func__, va));
320 _CP15_TLBIMVA(va | CPU_ASID_KERNEL);
324 /* Flush range of TLB entries (even global). */
326 tlb_flush_range_local(vm_offset_t va, vm_size_t size)
328 vm_offset_t eva = va + size;
330 KASSERT((va & PAGE_MASK) == 0, ("%s: va %#x not aligned", __func__, va));
331 KASSERT((size & PAGE_MASK) == 0, ("%s: size %#x not aligned", __func__,
335 for (; va < eva; va += PAGE_SIZE)
336 _CP15_TLBIMVA(va | CPU_ASID_KERNEL);
340 /* Broadcasting operations. */
341 #if __ARM_ARCH >= 7 && defined SMP
353 tlb_flush_all_ng(void)
357 _CP15_TLBIASIDIS(CPU_ASID_KERNEL);
362 tlb_flush(vm_offset_t va)
365 KASSERT((va & PAGE_MASK) == 0, ("%s: va %#x not aligned", __func__, va));
368 _CP15_TLBIMVAAIS(va);
373 tlb_flush_range(vm_offset_t va, vm_size_t size)
375 vm_offset_t eva = va + size;
377 KASSERT((va & PAGE_MASK) == 0, ("%s: va %#x not aligned", __func__, va));
378 KASSERT((size & PAGE_MASK) == 0, ("%s: size %#x not aligned", __func__,
382 for (; va < eva; va += PAGE_SIZE)
383 _CP15_TLBIMVAAIS(va);
388 #define tlb_flush_all() tlb_flush_all_local()
389 #define tlb_flush_all_ng() tlb_flush_all_ng_local()
390 #define tlb_flush(va) tlb_flush_local(va)
391 #define tlb_flush_range(va, size) tlb_flush_range_local(va, size)
396 * Cache maintenance operations.
399 /* Sync I and D caches to PoU */
401 icache_sync(vm_offset_t va, vm_size_t size)
403 vm_offset_t eva = va + size;
406 va &= ~cpuinfo.dcache_line_mask;
407 for ( ; va < eva; va += cpuinfo.dcache_line_size) {
408 #if __ARM_ARCH >= 7 && defined SMP
415 #if __ARM_ARCH >= 7 && defined SMP
424 /* Invalidate I cache */
428 #if __ARM_ARCH >= 7 && defined SMP
437 /* Invalidate branch predictor buffer */
441 #if __ARM_ARCH >= 7 && defined SMP
450 /* Write back D-cache to PoU */
452 dcache_wb_pou(vm_offset_t va, vm_size_t size)
454 vm_offset_t eva = va + size;
457 va &= ~cpuinfo.dcache_line_mask;
458 for ( ; va < eva; va += cpuinfo.dcache_line_size) {
459 #if __ARM_ARCH >= 7 && defined SMP
469 * Invalidate D-cache to PoC
471 * Caches are invalidated from outermost to innermost as fresh cachelines
472 * flow in this direction. In given range, if there was no dirty cacheline
473 * in any cache before, no stale cacheline should remain in them after this
474 * operation finishes.
477 dcache_inv_poc(vm_offset_t va, vm_paddr_t pa, vm_size_t size)
479 vm_offset_t eva = va + size;
482 /* invalidate L2 first */
483 cpu_l2cache_inv_range(pa, size);
486 va &= ~cpuinfo.dcache_line_mask;
487 for ( ; va < eva; va += cpuinfo.dcache_line_size) {
494 * Discard D-cache lines to PoC, prior to overwrite by DMA engine.
496 * Normal invalidation does L2 then L1 to ensure that stale data from L2 doesn't
497 * flow into L1 while invalidating. This routine is intended to be used only
498 * when invalidating a buffer before a DMA operation loads new data into memory.
499 * The concern in this case is that dirty lines are not evicted to main memory,
500 * overwriting the DMA data. For that reason, the L1 is done first to ensure
501 * that an evicted L1 line doesn't flow to L2 after the L2 has been cleaned.
504 dcache_inv_poc_dma(vm_offset_t va, vm_paddr_t pa, vm_size_t size)
506 vm_offset_t eva = va + size;
508 /* invalidate L1 first */
510 va &= ~cpuinfo.dcache_line_mask;
511 for ( ; va < eva; va += cpuinfo.dcache_line_size) {
517 cpu_l2cache_inv_range(pa, size);
521 * Write back D-cache to PoC
523 * Caches are written back from innermost to outermost as dirty cachelines
524 * flow in this direction. In given range, no dirty cacheline should remain
525 * in any cache after this operation finishes.
528 dcache_wb_poc(vm_offset_t va, vm_paddr_t pa, vm_size_t size)
530 vm_offset_t eva = va + size;
533 va &= ~cpuinfo.dcache_line_mask;
534 for ( ; va < eva; va += cpuinfo.dcache_line_size) {
539 cpu_l2cache_wb_range(pa, size);
542 /* Write back and invalidate D-cache to PoC */
544 dcache_wbinv_poc(vm_offset_t sva, vm_paddr_t pa, vm_size_t size)
547 vm_offset_t eva = sva + size;
550 /* write back L1 first */
551 va = sva & ~cpuinfo.dcache_line_mask;
552 for ( ; va < eva; va += cpuinfo.dcache_line_size) {
557 /* then write back and invalidate L2 */
558 cpu_l2cache_wbinv_range(pa, size);
560 /* then invalidate L1 */
561 va = sva & ~cpuinfo.dcache_line_mask;
562 for ( ; va < eva; va += cpuinfo.dcache_line_size) {
568 /* Set TTB0 register */
570 cp15_ttbr_set(uint32_t reg)
578 tlb_flush_all_ng_local();
581 #else /* ! __ARM_ARCH >= 6 */
584 * armv4/5 compatibility shims.
586 * These functions provide armv4 cache maintenance using the new armv6 names.
587 * Included here are just the functions actually used now in common code; it may
588 * be necessary to add things here over time.
590 * The callers of the dcache functions expect these routines to handle address
591 * and size values which are not aligned to cacheline boundaries; the armv4 and
592 * armv5 asm code handles that.
596 dcache_inv_poc(vm_offset_t va, vm_paddr_t pa, vm_size_t size)
599 cpu_dcache_inv_range(va, size);
600 cpu_l2cache_inv_range(va, size);
604 dcache_inv_poc_dma(vm_offset_t va, vm_paddr_t pa, vm_size_t size)
607 /* See armv6 code, above, for why we do L2 before L1 in this case. */
608 cpu_l2cache_inv_range(va, size);
609 cpu_dcache_inv_range(va, size);
613 dcache_wb_poc(vm_offset_t va, vm_paddr_t pa, vm_size_t size)
616 cpu_dcache_wb_range(va, size);
617 cpu_l2cache_wb_range(va, size);
620 #endif /* __ARM_ARCH >= 6 */
624 #endif /* !MACHINE_CPU_V6_H */