1 /* $NetBSD: booke_machdep.c,v 1.35 2024/03/05 14:15:34 thorpej Exp $ */
2 /*-
3 * Copyright (c) 2010, 2011 The NetBSD Foundation, Inc.
4 * All rights reserved.
5 *
6 * This code is derived from software contributed to The NetBSD Foundation
7 * by Raytheon BBN Technologies Corp and Defense Advanced Research Projects
8 * Agency and which was developed by Matt Thomas of 3am Software Foundry.
9 *
10 * This material is based upon work supported by the Defense Advanced Research
11 * Projects Agency and Space and Naval Warfare Systems Center, Pacific, under
12 * Contract No. N66001-09-C-2073.
13 * Approved for Public Release, Distribution Unlimited
14 *
15 * Redistribution and use in source and binary forms, with or without
16 * modification, are permitted provided that the following conditions
17 * are met:
18 * 1. Redistributions of source code must retain the above copyright
19 * notice, this list of conditions and the following disclaimer.
20 * 2. Redistributions in binary form must reproduce the above copyright
21 * notice, this list of conditions and the following disclaimer in the
22 * documentation and/or other materials provided with the distribution.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
25 * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
26 * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
27 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
28 * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
35 */
36
37 #define __INTR_PRIVATE
38 #define _POWERPC_BUS_DMA_PRIVATE
39
40 #include <sys/cdefs.h>
41 __KERNEL_RCSID(0, "$NetBSD: booke_machdep.c,v 1.35 2024/03/05 14:15:34 thorpej Exp $");
42
43 #include "ksyms.h"
44
45 #ifdef _KERNEL_OPT
46 #include "opt_ddb.h"
47 #include "opt_modular.h"
48 #include "opt_multiprocessor.h"
49 #endif
50
51 #include <sys/param.h>
52 #include <sys/cpu.h>
53 #include <sys/device.h>
54 #include <sys/intr.h>
55 #include <sys/mount.h>
56 #include <sys/msgbuf.h>
57 #include <sys/kernel.h>
58 #include <sys/reboot.h>
59 #include <sys/bus.h>
60 #include <sys/cpu.h>
61
62 #include <uvm/uvm_extern.h>
63
64 #include <dev/cons.h>
65
66 #include <powerpc/pcb.h>
67 #include <powerpc/spr.h>
68 #include <powerpc/booke/spr.h>
69 #include <powerpc/booke/cpuvar.h>
70
71 /*
72 * Global variables used here and there
73 */
74 paddr_t msgbuf_paddr;
75 psize_t pmemsize;
76 struct vm_map *phys_map;
77
78 #ifdef MODULAR
79 register_t cpu_psluserset = PSL_USERSET;
80 register_t cpu_pslusermod = PSL_USERMOD;
81 register_t cpu_pslusermask = PSL_USERMASK;
82 #endif
83
84 static bus_addr_t booke_dma_phys_to_bus_mem(bus_dma_tag_t, bus_addr_t);
85 static bus_addr_t booke_dma_bus_mem_to_phys(bus_dma_tag_t, bus_addr_t);
86
87
88 struct powerpc_bus_dma_tag booke_bus_dma_tag = {
89 ._dmamap_create = _bus_dmamap_create,
90 ._dmamap_destroy = _bus_dmamap_destroy,
91 ._dmamap_load = _bus_dmamap_load,
92 ._dmamap_load_mbuf = _bus_dmamap_load_mbuf,
93 ._dmamap_load_uio = _bus_dmamap_load_uio,
94 ._dmamap_load_raw = _bus_dmamap_load_raw,
95 ._dmamap_unload = _bus_dmamap_unload,
96 /*
97 * The caches on BookE are coherent so we don't need to do any special
98 * cache synchronization.
99 */
100 //._dmamap_sync = _bus_dmamap_sync,
101 ._dmamem_alloc = _bus_dmamem_alloc,
102 ._dmamem_free = _bus_dmamem_free,
103 ._dmamem_map = _bus_dmamem_map,
104 ._dmamem_unmap = _bus_dmamem_unmap,
105 ._dmamem_mmap = _bus_dmamem_mmap,
106 ._dma_phys_to_bus_mem = booke_dma_phys_to_bus_mem,
107 ._dma_bus_mem_to_phys = booke_dma_bus_mem_to_phys,
108 };
109
110 static bus_addr_t
booke_dma_phys_to_bus_mem(bus_dma_tag_t t,bus_addr_t a)111 booke_dma_phys_to_bus_mem(bus_dma_tag_t t, bus_addr_t a)
112 {
113 return a;
114 }
115
116 static bus_addr_t
booke_dma_bus_mem_to_phys(bus_dma_tag_t t,bus_addr_t a)117 booke_dma_bus_mem_to_phys(bus_dma_tag_t t, bus_addr_t a)
118 {
119 return a;
120 }
121
122 struct cpu_md_ops cpu_md_ops;
123
124 struct cpu_softc cpu_softc[] = {
125 [0] = {
126 .cpu_ci = &cpu_info[0],
127 },
128 #ifdef MULTIPROCESSOR
129 [CPU_MAXNUM-1] = {
130 .cpu_ci = &cpu_info[CPU_MAXNUM-1],
131 },
132 #endif
133 };
134 struct cpu_info cpu_info[] = {
135 [0] = {
136 .ci_curlwp = &lwp0,
137 .ci_tlb_info = &pmap_tlb0_info,
138 .ci_softc = &cpu_softc[0],
139 .ci_cpl = IPL_HIGH,
140 .ci_idepth = -1,
141 .ci_pmap_kern_segtab = &pmap_kern_segtab,
142 },
143 #ifdef MULTIPROCESSOR
144 [CPU_MAXNUM-1] = {
145 .ci_curlwp = NULL,
146 .ci_tlb_info = &pmap_tlb0_info,
147 .ci_softc = &cpu_softc[CPU_MAXNUM-1],
148 .ci_cpl = IPL_HIGH,
149 .ci_idepth = -1,
150 .ci_pmap_kern_segtab = &pmap_kern_segtab,
151 },
152 #endif
153 };
154 __CTASSERT(__arraycount(cpu_info) == __arraycount(cpu_softc));
155
156 /*
157 * This should probably be in autoconf! XXX
158 */
159 char machine[] = MACHINE; /* from <machine/param.h> */
160 char machine_arch[] = MACHINE_ARCH; /* from <machine/param.h> */
161
162 char bootpath[256];
163
164 #if NKSYMS || defined(DDB) || defined(MODULAR)
165 void *startsym, *endsym;
166 #endif
167
168 #if defined(MULTIPROCESSOR)
169 volatile struct cpu_hatch_data cpu_hatch_data __cacheline_aligned;
170 #endif
171
172 int fake_mapiodev = 1;
173
174 void
booke_cpu_startup(const char * model)175 booke_cpu_startup(const char *model)
176 {
177 vaddr_t minaddr, maxaddr;
178 char pbuf[9];
179
180 cpu_setmodel("%s", model);
181
182 printf("%s%s", copyright, version);
183
184 format_bytes(pbuf, sizeof(pbuf), ctob((uint64_t)physmem));
185 printf("total memory = %s\n", pbuf);
186
187 minaddr = 0;
188 /*
189 * Allocate a submap for physio
190 */
191 phys_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
192 VM_PHYS_SIZE, 0, false, NULL);
193
194 /*
195 * No need to allocate an mbuf cluster submap. Mbuf clusters
196 * are allocated via the pool allocator, and we use direct-mapped
197 * pool pages.
198 */
199
200 format_bytes(pbuf, sizeof(pbuf), ptoa(uvm_availmem(false)));
201 printf("avail memory = %s\n", pbuf);
202
203 /*
204 * Register the tlb's evcnts
205 */
206 pmap_tlb_info_evcnt_attach(curcpu()->ci_tlb_info);
207
208 /*
209 * Set up the board properties database.
210 */
211 board_info_init();
212
213 /*
214 * Now that we have VM, malloc()s are OK in bus_space.
215 */
216 bus_space_mallocok();
217 fake_mapiodev = 0;
218
219 #ifdef MULTIPROCESSOR
220 pmap_kernel()->pm_active = kcpuset_running;
221 pmap_kernel()->pm_onproc = kcpuset_running;
222
223 for (size_t i = 1; i < __arraycount(cpu_info); i++) {
224 struct cpu_info * const ci = &cpu_info[i];
225 struct cpu_softc * const cpu = &cpu_softc[i];
226 cpu->cpu_ci = ci;
227 cpu->cpu_bst = cpu_softc[0].cpu_bst;
228 cpu->cpu_le_bst = cpu_softc[0].cpu_le_bst;
229 cpu->cpu_bsh = cpu_softc[0].cpu_bsh;
230 cpu->cpu_highmem = cpu_softc[0].cpu_highmem;
231 ci->ci_softc = cpu;
232 ci->ci_tlb_info = &pmap_tlb0_info;
233 ci->ci_cpl = IPL_HIGH;
234 ci->ci_idepth = -1;
235 ci->ci_pmap_kern_segtab = curcpu()->ci_pmap_kern_segtab;
236 }
237
238 kcpuset_create(&cpuset_info.cpus_running, true);
239 kcpuset_create(&cpuset_info.cpus_hatched, true);
240 kcpuset_create(&cpuset_info.cpus_paused, true);
241 kcpuset_create(&cpuset_info.cpus_resumed, true);
242 kcpuset_create(&cpuset_info.cpus_halted, true);
243
244 kcpuset_set(cpuset_info.cpus_running, cpu_number());
245 #endif /* MULTIPROCESSOR */
246 }
247
248 static void
dumpsys(void)249 dumpsys(void)
250 {
251
252 printf("dumpsys: TBD\n");
253 }
254
255 /*
256 * Halt or reboot the machine after syncing/dumping according to howto.
257 */
258 void
cpu_reboot(int howto,char * what)259 cpu_reboot(int howto, char *what)
260 {
261 static int syncing;
262 static char str[256];
263 char *ap = str, *ap1 = ap;
264
265 boothowto = howto;
266 if (!cold && !(howto & RB_NOSYNC) && !syncing) {
267 syncing = 1;
268 vfs_shutdown(); /* sync */
269 }
270
271 splhigh();
272
273 if (!cold && (howto & RB_DUMP))
274 dumpsys();
275
276 doshutdownhooks();
277
278 pmf_system_shutdown(boothowto);
279
280 if ((howto & RB_POWERDOWN) == RB_POWERDOWN) {
281 /* Power off here if we know how...*/
282 }
283
284 if (howto & RB_HALT) {
285 printf("The operating system has halted.\n"
286 "Press any key to reboot.\n\n");
287 cnpollc(1); /* For proper keyboard command handling */
288 cngetc();
289 cnpollc(0);
290 }
291
292 printf("rebooting\n\n");
293 if (what && *what) {
294 if (strlen(what) > sizeof str - 5)
295 printf("boot string too large, ignored\n");
296 else {
297 strcpy(str, what);
298 ap1 = ap = str + strlen(str);
299 *ap++ = ' ';
300 }
301 }
302 *ap++ = '-';
303 if (howto & RB_SINGLE)
304 *ap++ = 's';
305 if (howto & RB_KDB)
306 *ap++ = 'd';
307 *ap++ = 0;
308 if (ap[-2] == '-')
309 *ap1 = 0;
310
311 /* flush cache for msgbuf */
312 dcache_wb(msgbuf_paddr, round_page(MSGBUFSIZE));
313
314 __asm volatile("msync; isync");
315 (*cpu_md_ops.md_cpu_reset)();
316
317 printf("%s: md_cpu_reset() failed!\n", __func__);
318 #ifdef DDB
319 for (;;)
320 Debugger();
321 #else
322 for (;;)
323 /* nothing */;
324 #endif
325 }
326
327 /*
328 * mapiodev:
329 *
330 * Allocate vm space and mapin the I/O address. Use reserved TLB
331 * mapping if one is found.
332 */
333 void *
mapiodev(paddr_t pa,psize_t len,bool prefetchable)334 mapiodev(paddr_t pa, psize_t len, bool prefetchable)
335 {
336 const vsize_t off = pa & PAGE_MASK;
337
338 /*
339 * See if we have reserved TLB entry for the pa. This needs to be
340 * true for console as we can't use uvm during early bootstrap.
341 */
342 void * const p = tlb_mapiodev(pa, len, prefetchable);
343 if (p != NULL)
344 return p;
345
346 if (fake_mapiodev)
347 panic("mapiodev: no TLB entry reserved for %llx+%llx",
348 (long long)pa, (long long)len);
349
350 const paddr_t orig_pa = pa;
351 const psize_t orig_len = len;
352 vsize_t align = 0;
353 pa = trunc_page(pa);
354 len = round_page(off + len);
355 /*
356 * If we are allocating a large amount (>= 1MB) try to get an
357 * aligned VA region for it so try to do a large mapping for it.
358 */
359 if ((len & (len - 1)) == 0 && len >= 0x100000)
360 align = len;
361
362 vaddr_t va = uvm_km_alloc(kernel_map, len, align, UVM_KMF_VAONLY);
363
364 if (va == 0 && align > 0) {
365 /*
366 * Large aligned request failed. Let's just get anything.
367 */
368 align = 0;
369 va = uvm_km_alloc(kernel_map, len, align, UVM_KMF_VAONLY);
370 }
371 if (va == 0)
372 return NULL;
373
374 if (align) {
375 /*
376 * Now try to map that via one big TLB entry.
377 */
378 pt_entry_t pte = pte_make_kenter_pa(pa, NULL,
379 VM_PROT_READ|VM_PROT_WRITE,
380 prefetchable ? 0 : PMAP_NOCACHE);
381 if (!tlb_ioreserve(va, len, pte)) {
382 void * const p0 = tlb_mapiodev(orig_pa, orig_len,
383 prefetchable);
384 KASSERT(p0 != NULL);
385 return p0;
386 }
387 }
388
389 for (va += len, pa += len; len > 0; len -= PAGE_SIZE) {
390 va -= PAGE_SIZE;
391 pa -= PAGE_SIZE;
392 pmap_kenter_pa(va, pa, VM_PROT_READ|VM_PROT_WRITE,
393 prefetchable ? 0 : PMAP_NOCACHE);
394 }
395 pmap_update(pmap_kernel());
396 return (void *)(va + off);
397 }
398
399 void
unmapiodev(vaddr_t va,vsize_t len)400 unmapiodev(vaddr_t va, vsize_t len)
401 {
402 /* Nothing to do for reserved (ie. not uvm_km_alloc'd) mappings. */
403 if (va < VM_MIN_KERNEL_ADDRESS || va > VM_MAX_KERNEL_ADDRESS) {
404 tlb_unmapiodev(va, len);
405 return;
406 }
407
408 len = round_page((va & PAGE_MASK) + len);
409 va = trunc_page(va);
410
411 pmap_kremove(va, len);
412 uvm_km_free(kernel_map, va, len, UVM_KMF_VAONLY);
413 }
414
415 void
cpu_evcnt_attach(struct cpu_info * ci)416 cpu_evcnt_attach(struct cpu_info *ci)
417 {
418 struct cpu_softc * const cpu = ci->ci_softc;
419 const char * const xname = ci->ci_data.cpu_name;
420
421 evcnt_attach_dynamic_nozero(&ci->ci_ev_clock, EVCNT_TYPE_INTR,
422 NULL, xname, "clock");
423 evcnt_attach_dynamic_nozero(&cpu->cpu_ev_late_clock, EVCNT_TYPE_INTR,
424 NULL, xname, "late clock");
425 evcnt_attach_dynamic_nozero(&cpu->cpu_ev_exec_trap_sync, EVCNT_TYPE_TRAP,
426 NULL, xname, "exec pages synced (trap)");
427 evcnt_attach_dynamic_nozero(&ci->ci_ev_traps, EVCNT_TYPE_TRAP,
428 NULL, xname, "traps");
429 evcnt_attach_dynamic_nozero(&ci->ci_ev_kdsi, EVCNT_TYPE_TRAP,
430 &ci->ci_ev_traps, xname, "kernel DSI traps");
431 evcnt_attach_dynamic_nozero(&ci->ci_ev_udsi, EVCNT_TYPE_TRAP,
432 &ci->ci_ev_traps, xname, "user DSI traps");
433 evcnt_attach_dynamic_nozero(&ci->ci_ev_udsi_fatal, EVCNT_TYPE_TRAP,
434 &ci->ci_ev_udsi, xname, "user DSI failures");
435 evcnt_attach_dynamic_nozero(&ci->ci_ev_kisi, EVCNT_TYPE_TRAP,
436 &ci->ci_ev_traps, xname, "kernel ISI traps");
437 evcnt_attach_dynamic_nozero(&ci->ci_ev_isi, EVCNT_TYPE_TRAP,
438 &ci->ci_ev_traps, xname, "user ISI traps");
439 evcnt_attach_dynamic_nozero(&ci->ci_ev_isi_fatal, EVCNT_TYPE_TRAP,
440 &ci->ci_ev_isi, xname, "user ISI failures");
441 evcnt_attach_dynamic_nozero(&ci->ci_ev_scalls, EVCNT_TYPE_TRAP,
442 &ci->ci_ev_traps, xname, "system call traps");
443 evcnt_attach_dynamic_nozero(&ci->ci_ev_pgm, EVCNT_TYPE_TRAP,
444 &ci->ci_ev_traps, xname, "PGM traps");
445 evcnt_attach_dynamic_nozero(&ci->ci_ev_debug, EVCNT_TYPE_TRAP,
446 &ci->ci_ev_traps, xname, "debug traps");
447 evcnt_attach_dynamic_nozero(&ci->ci_ev_fpu, EVCNT_TYPE_TRAP,
448 &ci->ci_ev_traps, xname, "FPU unavailable traps");
449 evcnt_attach_dynamic_nozero(&ci->ci_ev_fpusw, EVCNT_TYPE_MISC,
450 &ci->ci_ev_fpu, xname, "FPU context switches");
451 evcnt_attach_dynamic_nozero(&ci->ci_ev_ali, EVCNT_TYPE_TRAP,
452 &ci->ci_ev_traps, xname, "user alignment traps");
453 evcnt_attach_dynamic_nozero(&ci->ci_ev_ali_fatal, EVCNT_TYPE_TRAP,
454 &ci->ci_ev_ali, xname, "user alignment traps");
455 evcnt_attach_dynamic_nozero(&ci->ci_ev_umchk, EVCNT_TYPE_TRAP,
456 &ci->ci_ev_umchk, xname, "user MCHK failures");
457 evcnt_attach_dynamic_nozero(&ci->ci_ev_vec, EVCNT_TYPE_TRAP,
458 &ci->ci_ev_traps, xname, "SPE unavailable");
459 evcnt_attach_dynamic_nozero(&ci->ci_ev_vecsw, EVCNT_TYPE_MISC,
460 &ci->ci_ev_vec, xname, "SPE context switches");
461 evcnt_attach_dynamic_nozero(&ci->ci_ev_ipi, EVCNT_TYPE_INTR,
462 NULL, xname, "IPIs");
463 evcnt_attach_dynamic_nozero(&ci->ci_ev_tlbmiss_soft, EVCNT_TYPE_TRAP,
464 &ci->ci_ev_traps, xname, "soft tlb misses");
465 evcnt_attach_dynamic_nozero(&ci->ci_ev_dtlbmiss_hard, EVCNT_TYPE_TRAP,
466 &ci->ci_ev_traps, xname, "data tlb misses");
467 evcnt_attach_dynamic_nozero(&ci->ci_ev_itlbmiss_hard, EVCNT_TYPE_TRAP,
468 &ci->ci_ev_traps, xname, "inst tlb misses");
469 }
470
471 #ifdef MULTIPROCESSOR
472 register_t
cpu_hatch(void)473 cpu_hatch(void)
474 {
475 struct cpuset_info * const csi = &cpuset_info;
476 const size_t id = cpu_number();
477
478 /*
479 * We've hatched so tell the spinup code.
480 */
481 kcpuset_set(csi->cpus_hatched, id);
482
483 /*
484 * Loop until running bit for this cpu is set.
485 */
486 while (!kcpuset_isset(csi->cpus_running, id)) {
487 continue;
488 }
489
490 /*
491 * Now that we are active, start the clocks.
492 */
493 cpu_initclocks();
494
495 /*
496 * Return sp of the idlelwp. Which we should be already using but ...
497 */
498 return curcpu()->ci_curpcb->pcb_sp;
499 }
500
501 void
cpu_boot_secondary_processors(void)502 cpu_boot_secondary_processors(void)
503 {
504 volatile struct cpuset_info * const csi = &cpuset_info;
505 CPU_INFO_ITERATOR cii;
506 struct cpu_info *ci;
507 kcpuset_t *running;
508
509 kcpuset_create(&running, true);
510
511 for (CPU_INFO_FOREACH(cii, ci)) {
512 /*
513 * Skip this CPU if it didn't successfully hatch.
514 */
515 if (!kcpuset_isset(csi->cpus_hatched, cpu_index(ci)))
516 continue;
517
518 KASSERT(!CPU_IS_PRIMARY(ci));
519 KASSERT(ci->ci_data.cpu_idlelwp);
520
521 kcpuset_set(running, cpu_index(ci));
522 }
523 KASSERT(kcpuset_match(csi->cpus_hatched, running));
524 if (!kcpuset_iszero(running)) {
525 kcpuset_merge(csi->cpus_running, running);
526 }
527 kcpuset_destroy(running);
528 }
529 #endif
530
531 uint32_t
cpu_read_4(bus_addr_t a)532 cpu_read_4(bus_addr_t a)
533 {
534 struct cpu_softc * const cpu = curcpu()->ci_softc;
535 // printf(" %s(%p, %x, %x)", __func__, cpu->cpu_bst, cpu->cpu_bsh, a);
536 return bus_space_read_4(cpu->cpu_bst, cpu->cpu_bsh, a);
537 }
538
539 uint8_t
cpu_read_1(bus_addr_t a)540 cpu_read_1(bus_addr_t a)
541 {
542 struct cpu_softc * const cpu = curcpu()->ci_softc;
543 // printf(" %s(%p, %x, %x)", __func__, cpu->cpu_bst, cpu->cpu_bsh, a);
544 return bus_space_read_1(cpu->cpu_bst, cpu->cpu_bsh, a);
545 }
546
547 void
cpu_write_4(bus_addr_t a,uint32_t v)548 cpu_write_4(bus_addr_t a, uint32_t v)
549 {
550 struct cpu_softc * const cpu = curcpu()->ci_softc;
551 bus_space_write_4(cpu->cpu_bst, cpu->cpu_bsh, a, v);
552 }
553
554 void
cpu_write_1(bus_addr_t a,uint8_t v)555 cpu_write_1(bus_addr_t a, uint8_t v)
556 {
557 struct cpu_softc * const cpu = curcpu()->ci_softc;
558 bus_space_write_1(cpu->cpu_bst, cpu->cpu_bsh, a, v);
559 }
560
561 void
booke_sstep(struct trapframe * tf)562 booke_sstep(struct trapframe *tf)
563 {
564 uint32_t insn;
565
566 KASSERT(tf->tf_srr1 & PSL_DE);
567 if (ufetch_32((const void *)tf->tf_srr0, &insn) != 0)
568 return;
569
570 register_t dbcr0 = DBCR0_IAC1 | DBCR0_IDM;
571 register_t dbcr1 = DBCR1_IAC1US_USER | DBCR1_IAC1ER_DS1;
572 if ((insn >> 28) == 4) {
573 uint32_t iac2 = 0;
574 if ((insn >> 26) == 0x12) {
575 const int32_t off = (((int32_t)insn << 6) >> 6) & ~3;
576 iac2 = ((insn & 2) ? 0 : tf->tf_srr0) + off;
577 dbcr0 |= DBCR0_IAC2;
578 } else if ((insn >> 26) == 0x10) {
579 const int16_t off = insn & ~3;
580 iac2 = ((insn & 2) ? 0 : tf->tf_srr0) + off;
581 dbcr0 |= DBCR0_IAC2;
582 } else if ((insn & 0xfc00fffe) == 0x4c000420) {
583 iac2 = tf->tf_ctr;
584 dbcr0 |= DBCR0_IAC2;
585 } else if ((insn & 0xfc00fffe) == 0x4c000020) {
586 iac2 = tf->tf_lr;
587 dbcr0 |= DBCR0_IAC2;
588 }
589 if (dbcr0 & DBCR0_IAC2) {
590 dbcr1 |= DBCR1_IAC2US_USER | DBCR1_IAC2ER_DS1;
591 mtspr(SPR_IAC2, iac2);
592 }
593 }
594 mtspr(SPR_IAC1, tf->tf_srr0 + 4);
595 mtspr(SPR_DBCR1, dbcr1);
596 mtspr(SPR_DBCR0, dbcr0);
597 }
598
599 #ifdef DIAGNOSTIC
600 static inline void
swap_data(uint64_t * data,size_t a,size_t b)601 swap_data(uint64_t *data, size_t a, size_t b)
602 {
603 uint64_t swap = data[a];
604 data[a] = data[b];
605 data[b] = swap;
606 }
607
608 static void
sort_data(uint64_t * data,size_t count)609 sort_data(uint64_t *data, size_t count)
610 {
611 #if 0
612 /*
613 * Mostly classic bubble sort
614 */
615 do {
616 size_t new_count = 0;
617 for (size_t i = 1; i < count; i++) {
618 if (tbs[i - 1] > tbs[i]) {
619 swap_tbs(tbs, i - 1, i);
620 new_count = i;
621 }
622 }
623 count = new_count;
624 } while (count > 0);
625 #else
626 /*
627 * Comb sort
628 */
629 size_t gap = count;
630 bool swapped = false;
631 while (gap > 1 || swapped) {
632 if (gap > 1) {
633 /*
634 * phi = (1 + sqrt(5)) / 2 [golden ratio]
635 * N = 1 / (1 - e^-phi)) = 1.247330950103979
636 *
637 * We want to but can't use floating point to calculate
638 * gap = (size_t)((double)gap / N)
639 *
640 * So we will use the multiplicative inverse of N
641 * (module 65536) to achieve the division.
642 *
643 * iN = 2^16 / 1.24733... = 52540
644 * x / N == (x * iN) / 65536
645 */
646 gap = (gap * 52540) / 65536;
647 }
648
649 swapped = false;
650
651 for (size_t i = 0; gap + i < count; i++) {
652 if (data[i] > data[i + gap]) {
653 swap_data(data, i, i + gap);
654 swapped = true;
655 }
656 }
657 }
658 #endif
659 }
660 #endif
661
662 void
dump_splhist(struct cpu_info * ci,void (* pr)(const char *,...))663 dump_splhist(struct cpu_info *ci, void (*pr)(const char *, ...))
664 {
665 #ifdef DIAGNOSTIC
666 struct cpu_softc * const cpu = ci->ci_softc;
667 uint64_t tbs[NIPL*NIPL];
668 size_t ntbs = 0;
669 for (size_t to = 0; to < NIPL; to++) {
670 for (size_t from = 0; from < NIPL; from++) {
671 uint64_t tb = cpu->cpu_spl_tb[to][from];
672 if (tb == 0)
673 continue;
674 tbs[ntbs++] = (tb << 8) | (to << 4) | from;
675 }
676 }
677 sort_data(tbs, ntbs);
678
679 if (pr == NULL)
680 pr = printf;
681 uint64_t last_tb = 0;
682 for (size_t i = 0; i < ntbs; i++) {
683 uint64_t tb = tbs[i];
684 size_t from = tb & 15;
685 size_t to = (tb >> 4) & 15;
686 tb >>= 8;
687 (*pr)("%s(%zu) from %zu at %"PRId64"",
688 from < to ? "splraise" : "splx",
689 to, from, tb);
690 if (last_tb && from != IPL_NONE)
691 (*pr)(" (+%"PRId64")", tb - last_tb);
692 (*pr)("\n");
693 last_tb = tb;
694 }
695 #endif
696 }
697