1 /*-
2 * Copyright (c) 2014 Andrew Turner
3 * Copyright (c) 2015-2017 Ruslan Bukin <br@bsdpad.com>
4 * All rights reserved.
5 *
6 * Portions of this software were developed by SRI International and the
7 * University of Cambridge Computer Laboratory under DARPA/AFRL contract
8 * FA8750-10-C-0237 ("CTSRD"), as part of the DARPA CRASH research programme.
9 *
10 * Portions of this software were developed by the University of Cambridge
11 * Computer Laboratory as part of the CTSRD Project, with support from the
12 * UK Higher Education Innovation Fund (HEIF).
13 *
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
16 * are met:
17 * 1. Redistributions of source code must retain the above copyright
18 * notice, this list of conditions and the following disclaimer.
19 * 2. Redistributions in binary form must reproduce the above copyright
20 * notice, this list of conditions and the following disclaimer in the
21 * documentation and/or other materials provided with the distribution.
22 *
23 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * SUCH DAMAGE.
34 */
35
36 #include "opt_ddb.h"
37 #include "opt_kstack_pages.h"
38 #include "opt_platform.h"
39
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/boot.h>
43 #include <sys/buf.h>
44 #include <sys/bus.h>
45 #include <sys/cons.h>
46 #include <sys/cpu.h>
47 #include <sys/devmap.h>
48 #include <sys/exec.h>
49 #include <sys/imgact.h>
50 #include <sys/kdb.h>
51 #include <sys/kernel.h>
52 #include <sys/ktr.h>
53 #include <sys/limits.h>
54 #include <sys/linker.h>
55 #include <sys/msgbuf.h>
56 #include <sys/pcpu.h>
57 #include <sys/physmem.h>
58 #include <sys/proc.h>
59 #include <sys/ptrace.h>
60 #include <sys/reboot.h>
61 #include <sys/reg.h>
62 #include <sys/rwlock.h>
63 #include <sys/sched.h>
64 #include <sys/signalvar.h>
65 #include <sys/syscallsubr.h>
66 #include <sys/sysent.h>
67 #include <sys/sysproto.h>
68 #include <sys/tslog.h>
69 #include <sys/ucontext.h>
70 #include <sys/vmmeter.h>
71
72 #include <vm/vm.h>
73 #include <vm/vm_param.h>
74 #include <vm/vm_kern.h>
75 #include <vm/vm_object.h>
76 #include <vm/vm_page.h>
77 #include <vm/vm_phys.h>
78 #include <vm/pmap.h>
79 #include <vm/vm_map.h>
80 #include <vm/vm_pager.h>
81
82 #include <machine/cpu.h>
83 #include <machine/fpe.h>
84 #include <machine/intr.h>
85 #include <machine/kdb.h>
86 #include <machine/machdep.h>
87 #include <machine/metadata.h>
88 #include <machine/pcb.h>
89 #include <machine/pte.h>
90 #include <machine/riscvreg.h>
91 #include <machine/sbi.h>
92 #include <machine/trap.h>
93 #include <machine/vmparam.h>
94
95 #ifdef DDB
96 #include <ddb/ddb.h>
97 #endif
98
99 #ifdef FDT
100 #include <contrib/libfdt/libfdt.h>
101 #include <dev/fdt/fdt_common.h>
102 #include <dev/ofw/openfirm.h>
103 #endif
104
105 struct pcpu __pcpu[MAXCPU];
106
107 static struct trapframe proc0_tf;
108
109 int early_boot = 1;
110 int cold = 1;
111
112 #define DTB_SIZE_MAX (1024 * 1024)
113
114 struct kva_md_info kmi;
115
116 #define BOOT_HART_INVALID 0xffffffff
117 uint32_t boot_hart = BOOT_HART_INVALID; /* The hart we booted on. */
118
119 cpuset_t all_harts;
120
121 extern int *end;
122
123 static char static_kenv[PAGE_SIZE];
124
125 static void
cpu_startup(void * dummy)126 cpu_startup(void *dummy)
127 {
128
129 sbi_print_version();
130 printcpuinfo(0);
131
132 printf("real memory = %ju (%ju MB)\n", ptoa((uintmax_t)realmem),
133 ptoa((uintmax_t)realmem) / (1024 * 1024));
134
135 /*
136 * Display any holes after the first chunk of extended memory.
137 */
138 if (bootverbose) {
139 int indx;
140
141 printf("Physical memory chunk(s):\n");
142 for (indx = 0; phys_avail[indx + 1] != 0; indx += 2) {
143 vm_paddr_t size;
144
145 size = phys_avail[indx + 1] - phys_avail[indx];
146 printf(
147 "0x%016jx - 0x%016jx, %ju bytes (%ju pages)\n",
148 (uintmax_t)phys_avail[indx],
149 (uintmax_t)phys_avail[indx + 1] - 1,
150 (uintmax_t)size, (uintmax_t)size / PAGE_SIZE);
151 }
152 }
153
154 vm_ksubmap_init(&kmi);
155
156 printf("avail memory = %ju (%ju MB)\n",
157 ptoa((uintmax_t)vm_free_count()),
158 ptoa((uintmax_t)vm_free_count()) / (1024 * 1024));
159 if (bootverbose)
160 devmap_print_table();
161
162 bufinit();
163 vm_pager_bufferinit();
164 }
165
166 SYSINIT(cpu, SI_SUB_CPU, SI_ORDER_FIRST, cpu_startup, NULL);
167
168 int
cpu_idle_wakeup(int cpu)169 cpu_idle_wakeup(int cpu)
170 {
171
172 return (0);
173 }
174
175 void
cpu_idle(int busy)176 cpu_idle(int busy)
177 {
178
179 spinlock_enter();
180 if (!busy)
181 cpu_idleclock();
182 if (!sched_runnable())
183 __asm __volatile(
184 "fence \n"
185 "wfi \n");
186 if (!busy)
187 cpu_activeclock();
188 spinlock_exit();
189 }
190
191 void
cpu_halt(void)192 cpu_halt(void)
193 {
194
195 /*
196 * Try to power down using the HSM SBI extension and fall back to a
197 * simple wfi loop.
198 */
199 intr_disable();
200 if (sbi_probe_extension(SBI_EXT_ID_HSM) != 0)
201 sbi_hsm_hart_stop();
202 for (;;)
203 __asm __volatile("wfi");
204 /* NOTREACHED */
205 }
206
207 /*
208 * Flush the D-cache for non-DMA I/O so that the I-cache can
209 * be made coherent later.
210 */
211 void
cpu_flush_dcache(void * ptr,size_t len)212 cpu_flush_dcache(void *ptr, size_t len)
213 {
214
215 /* TBD */
216 }
217
218 /* Get current clock frequency for the given CPU ID. */
219 int
cpu_est_clockrate(int cpu_id,uint64_t * rate)220 cpu_est_clockrate(int cpu_id, uint64_t *rate)
221 {
222 struct pcpu *pc;
223
224 pc = pcpu_find(cpu_id);
225 if (pc == NULL || rate == NULL)
226 return (EINVAL);
227
228 if (pc->pc_clock == 0)
229 return (EOPNOTSUPP);
230
231 *rate = pc->pc_clock;
232
233 return (0);
234 }
235
236 void
cpu_pcpu_init(struct pcpu * pcpu,int cpuid,size_t size)237 cpu_pcpu_init(struct pcpu *pcpu, int cpuid, size_t size)
238 {
239 }
240
241 void
spinlock_enter(void)242 spinlock_enter(void)
243 {
244 struct thread *td;
245 register_t reg;
246
247 td = curthread;
248 if (td->td_md.md_spinlock_count == 0) {
249 reg = intr_disable();
250 td->td_md.md_spinlock_count = 1;
251 td->td_md.md_saved_sstatus_ie = reg;
252 critical_enter();
253 } else
254 td->td_md.md_spinlock_count++;
255 }
256
257 void
spinlock_exit(void)258 spinlock_exit(void)
259 {
260 struct thread *td;
261 register_t sstatus_ie;
262
263 td = curthread;
264 sstatus_ie = td->td_md.md_saved_sstatus_ie;
265 td->td_md.md_spinlock_count--;
266 if (td->td_md.md_spinlock_count == 0) {
267 critical_exit();
268 intr_restore(sstatus_ie);
269 }
270 }
271
272 /*
273 * Construct a PCB from a trapframe. This is called from kdb_trap() where
274 * we want to start a backtrace from the function that caused us to enter
275 * the debugger. We have the context in the trapframe, but base the trace
276 * on the PCB. The PCB doesn't have to be perfect, as long as it contains
277 * enough for a backtrace.
278 */
279 void
makectx(struct trapframe * tf,struct pcb * pcb)280 makectx(struct trapframe *tf, struct pcb *pcb)
281 {
282
283 memcpy(pcb->pcb_s, tf->tf_s, sizeof(tf->tf_s));
284
285 pcb->pcb_ra = tf->tf_sepc;
286 pcb->pcb_sp = tf->tf_sp;
287 pcb->pcb_gp = tf->tf_gp;
288 pcb->pcb_tp = tf->tf_tp;
289 }
290
291 static void
init_proc0(vm_offset_t kstack)292 init_proc0(vm_offset_t kstack)
293 {
294 struct pcpu *pcpup;
295
296 pcpup = &__pcpu[0];
297
298 proc_linkup0(&proc0, &thread0);
299 thread0.td_kstack = kstack;
300 thread0.td_kstack_pages = KSTACK_PAGES;
301 thread0.td_pcb = (struct pcb *)(thread0.td_kstack +
302 thread0.td_kstack_pages * PAGE_SIZE) - 1;
303 thread0.td_pcb->pcb_fpflags = 0;
304 thread0.td_frame = &proc0_tf;
305 pcpup->pc_curpcb = thread0.td_pcb;
306 }
307
308 #ifdef FDT
309 static void
try_load_dtb(void)310 try_load_dtb(void)
311 {
312 vm_offset_t dtbp;
313
314 dtbp = MD_FETCH(preload_kmdp, MODINFOMD_DTBP, vm_offset_t);
315
316 #if defined(FDT_DTB_STATIC)
317 /*
318 * In case the device tree blob was not retrieved (from metadata) try
319 * to use the statically embedded one.
320 */
321 if (dtbp == (vm_offset_t)NULL)
322 dtbp = (vm_offset_t)&fdt_static_dtb;
323 #endif
324
325 if (dtbp == (vm_offset_t)NULL) {
326 printf("ERROR loading DTB\n");
327 return;
328 }
329
330 if (!OF_install(OFW_FDT, 0))
331 panic("Cannot install FDT");
332
333 if (OF_init((void *)dtbp) != 0)
334 panic("OF_init failed with the found device tree");
335 }
336 #endif
337
338 /*
339 * Fake up a boot descriptor table.
340 */
341 static void
fake_preload_metadata(struct riscv_bootparams * rvbp)342 fake_preload_metadata(struct riscv_bootparams *rvbp)
343 {
344 static uint32_t fake_preload[48];
345 vm_offset_t lastaddr;
346 size_t fake_size, dtb_size;
347
348 #define PRELOAD_PUSH_VALUE(type, value) do { \
349 *(type *)((char *)fake_preload + fake_size) = (value); \
350 fake_size += sizeof(type); \
351 } while (0)
352
353 #define PRELOAD_PUSH_STRING(str) do { \
354 uint32_t ssize; \
355 ssize = strlen(str) + 1; \
356 PRELOAD_PUSH_VALUE(uint32_t, ssize); \
357 strcpy(((char *)fake_preload + fake_size), str); \
358 fake_size += ssize; \
359 fake_size = roundup(fake_size, sizeof(u_long)); \
360 } while (0)
361
362 fake_size = 0;
363 lastaddr = (vm_offset_t)&end;
364
365 PRELOAD_PUSH_VALUE(uint32_t, MODINFO_NAME);
366 PRELOAD_PUSH_STRING("kernel");
367 PRELOAD_PUSH_VALUE(uint32_t, MODINFO_TYPE);
368 PRELOAD_PUSH_STRING(preload_kerntype);
369
370 PRELOAD_PUSH_VALUE(uint32_t, MODINFO_ADDR);
371 PRELOAD_PUSH_VALUE(uint32_t, sizeof(vm_offset_t));
372 PRELOAD_PUSH_VALUE(uint64_t, KERNBASE);
373
374 PRELOAD_PUSH_VALUE(uint32_t, MODINFO_SIZE);
375 PRELOAD_PUSH_VALUE(uint32_t, sizeof(size_t));
376 PRELOAD_PUSH_VALUE(uint64_t, (size_t)((vm_offset_t)&end - KERNBASE));
377
378 /*
379 * Copy the DTB to KVA space. We are able to dereference the physical
380 * address due to the identity map created in locore.
381 */
382 lastaddr = roundup(lastaddr, sizeof(int));
383 PRELOAD_PUSH_VALUE(uint32_t, MODINFO_METADATA | MODINFOMD_DTBP);
384 PRELOAD_PUSH_VALUE(uint32_t, sizeof(vm_offset_t));
385 PRELOAD_PUSH_VALUE(vm_offset_t, lastaddr);
386 dtb_size = fdt_totalsize(rvbp->dtbp_phys);
387 memmove((void *)lastaddr, (const void *)rvbp->dtbp_phys, dtb_size);
388 lastaddr = roundup(lastaddr + dtb_size, sizeof(int));
389
390 PRELOAD_PUSH_VALUE(uint32_t, MODINFO_METADATA | MODINFOMD_KERNEND);
391 PRELOAD_PUSH_VALUE(uint32_t, sizeof(vm_offset_t));
392 PRELOAD_PUSH_VALUE(vm_offset_t, lastaddr);
393
394 PRELOAD_PUSH_VALUE(uint32_t, MODINFO_METADATA | MODINFOMD_HOWTO);
395 PRELOAD_PUSH_VALUE(uint32_t, sizeof(int));
396 PRELOAD_PUSH_VALUE(int, RB_VERBOSE);
397
398 /* End marker */
399 PRELOAD_PUSH_VALUE(uint32_t, 0);
400 PRELOAD_PUSH_VALUE(uint32_t, 0);
401 preload_metadata = (caddr_t)fake_preload;
402
403 /* Check if bootloader clobbered part of the kernel with the DTB. */
404 KASSERT(rvbp->dtbp_phys + dtb_size <= rvbp->kern_phys ||
405 rvbp->dtbp_phys >= rvbp->kern_phys + (lastaddr - KERNBASE),
406 ("FDT (%lx-%lx) and kernel (%lx-%lx) overlap", rvbp->dtbp_phys,
407 rvbp->dtbp_phys + dtb_size, rvbp->kern_phys,
408 rvbp->kern_phys + (lastaddr - KERNBASE)));
409 KASSERT(fake_size < sizeof(fake_preload),
410 ("Too many fake_preload items"));
411
412 if (boothowto & RB_VERBOSE)
413 printf("FDT phys (%lx-%lx), kernel phys (%lx-%lx)\n",
414 rvbp->dtbp_phys, rvbp->dtbp_phys + dtb_size,
415 rvbp->kern_phys, rvbp->kern_phys + (lastaddr - KERNBASE));
416 }
417
418 /* Support for FDT configurations only. */
419 CTASSERT(FDT);
420
421 #ifdef FDT
422 static void
parse_fdt_bootargs(void)423 parse_fdt_bootargs(void)
424 {
425 char bootargs[512];
426
427 bootargs[sizeof(bootargs) - 1] = '\0';
428 if (fdt_get_chosen_bootargs(bootargs, sizeof(bootargs) - 1) == 0) {
429 boothowto |= boot_parse_cmdline(bootargs);
430 }
431 }
432 #endif
433
434 static vm_offset_t
parse_metadata(void)435 parse_metadata(void)
436 {
437 vm_offset_t lastaddr;
438 #ifdef DDB
439 vm_offset_t ksym_start, ksym_end;
440 #endif
441 char *kern_envp;
442
443 /* Initialize preload_kmdp */
444 preload_initkmdp(true);
445
446 /* Read the boot metadata */
447 boothowto = MD_FETCH(preload_kmdp, MODINFOMD_HOWTO, int);
448 lastaddr = MD_FETCH(preload_kmdp, MODINFOMD_KERNEND, vm_offset_t);
449 kern_envp = MD_FETCH(preload_kmdp, MODINFOMD_ENVP, char *);
450 if (kern_envp != NULL)
451 init_static_kenv(kern_envp, 0);
452 else
453 init_static_kenv(static_kenv, sizeof(static_kenv));
454 #ifdef DDB
455 ksym_start = MD_FETCH(preload_kmdp, MODINFOMD_SSYM, uintptr_t);
456 ksym_end = MD_FETCH(preload_kmdp, MODINFOMD_ESYM, uintptr_t);
457 db_fetch_ksymtab(ksym_start, ksym_end, 0);
458 #endif
459 #ifdef FDT
460 try_load_dtb();
461 if (kern_envp == NULL)
462 parse_fdt_bootargs();
463 #endif
464 return (lastaddr);
465 }
466
467 void
initriscv(struct riscv_bootparams * rvbp)468 initriscv(struct riscv_bootparams *rvbp)
469 {
470 struct mem_region mem_regions[FDT_MEM_REGIONS];
471 struct pcpu *pcpup;
472 int mem_regions_sz;
473 vm_offset_t lastaddr;
474 vm_size_t kernlen;
475 #ifdef FDT
476 phandle_t chosen;
477 uint32_t hart;
478 #endif
479 char *env;
480
481 TSRAW(&thread0, TS_ENTER, __func__, NULL);
482
483 /* Set the pcpu data, this is needed by pmap_bootstrap */
484 pcpup = &__pcpu[0];
485 pcpu_init(pcpup, 0, sizeof(struct pcpu));
486
487 /* Set the pcpu pointer */
488 __asm __volatile("mv tp, %0" :: "r"(pcpup));
489
490 PCPU_SET(curthread, &thread0);
491
492 /* Initialize SBI interface. */
493 sbi_init();
494
495 /* Parse the boot metadata. */
496 if (rvbp->modulep != 0) {
497 preload_metadata = (caddr_t)rvbp->modulep;
498 } else {
499 fake_preload_metadata(rvbp);
500 }
501 lastaddr = parse_metadata();
502
503 #ifdef FDT
504 /*
505 * Look for the boot hart ID. This was either passed in directly from
506 * the SBI firmware and handled by locore, or was stored in the device
507 * tree by an earlier boot stage.
508 */
509 chosen = OF_finddevice("/chosen");
510 if (OF_getencprop(chosen, "boot-hartid", &hart, sizeof(hart)) != -1) {
511 boot_hart = hart;
512 }
513 #endif
514 if (boot_hart == BOOT_HART_INVALID) {
515 panic("Boot hart ID was not properly set");
516 }
517 pcpup->pc_hart = boot_hart;
518
519 #ifdef FDT
520 /*
521 * Exclude reserved memory specified by the device tree. Typically,
522 * this contains an entry for memory used by the runtime SBI firmware.
523 */
524 if (fdt_get_reserved_mem(mem_regions, &mem_regions_sz) == 0) {
525 physmem_exclude_regions(mem_regions, mem_regions_sz,
526 EXFLAG_NODUMP | EXFLAG_NOALLOC);
527 }
528
529 /* Grab physical memory regions information from device tree. */
530 if (fdt_get_mem_regions(mem_regions, &mem_regions_sz, NULL) != 0) {
531 panic("Cannot get physical memory regions");
532 }
533 physmem_hardware_regions(mem_regions, mem_regions_sz);
534 #endif
535
536 /*
537 * Identify CPU/ISA features.
538 */
539 identify_cpu(0);
540
541 /* Do basic tuning, hz etc */
542 init_param1();
543
544 #ifdef FDT
545 /*
546 * XXX: Unconditionally exclude the lowest 2MB of physical memory, as
547 * this area is assumed to contain the SBI firmware. This is a little
548 * fragile, but it is consistent with the platforms we support so far.
549 *
550 * TODO: remove this when the all regular booting methods properly
551 * report their reserved memory in the device tree.
552 */
553 physmem_exclude_region(mem_regions[0].mr_start, L2_SIZE,
554 EXFLAG_NODUMP | EXFLAG_NOALLOC);
555 #endif
556
557 /* Bootstrap enough of pmap to enter the kernel proper */
558 kernlen = (lastaddr - KERNBASE);
559 pmap_bootstrap(rvbp->kern_phys, kernlen);
560
561 physmem_init_kernel_globals();
562
563 /* Establish static device mappings */
564 devmap_bootstrap();
565
566 cninit();
567
568 /*
569 * Dump the boot metadata. We have to wait for cninit() since console
570 * output is required. If it's grossly incorrect the kernel will never
571 * make it this far.
572 */
573 if (getenv_is_true("debug.dump_modinfo_at_boot"))
574 preload_dump();
575
576 init_proc0(rvbp->kern_stack);
577
578 msgbufinit(msgbufp, msgbufsize);
579 mutex_init();
580 init_param2(physmem);
581 kdb_init();
582 #ifdef KDB
583 if ((boothowto & RB_KDB) != 0)
584 kdb_enter(KDB_WHY_BOOTFLAGS, "Boot flags requested debugger");
585 #endif
586
587 env = kern_getenv("kernelname");
588 if (env != NULL)
589 strlcpy(kernelname, env, sizeof(kernelname));
590
591 if (boothowto & RB_VERBOSE)
592 physmem_print_tables();
593
594 early_boot = 0;
595
596 if (bootverbose && kstack_pages != KSTACK_PAGES)
597 printf("kern.kstack_pages = %d ignored for thread0\n",
598 kstack_pages);
599
600 TSEXIT();
601 }
602