2 * Copyright (c) 2014 Andrew Turner
3 * Copyright (c) 2015-2017 Ruslan Bukin <br@bsdpad.com>
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.
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).
14 * Redistribution and use in source and binary forms, with or without
15 * modification, are permitted provided that the following conditions
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.
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
36 #include "opt_platform.h"
38 #include <sys/cdefs.h>
39 __FBSDID("$FreeBSD$");
41 #include <sys/param.h>
42 #include <sys/systm.h>
48 #include <sys/imgact.h>
50 #include <sys/kernel.h>
51 #include <sys/limits.h>
52 #include <sys/linker.h>
53 #include <sys/msgbuf.h>
56 #include <sys/ptrace.h>
57 #include <sys/reboot.h>
58 #include <sys/rwlock.h>
59 #include <sys/sched.h>
60 #include <sys/signalvar.h>
61 #include <sys/syscallsubr.h>
62 #include <sys/sysent.h>
63 #include <sys/sysproto.h>
64 #include <sys/ucontext.h>
67 #include <vm/vm_kern.h>
68 #include <vm/vm_object.h>
69 #include <vm/vm_page.h>
71 #include <vm/vm_map.h>
72 #include <vm/vm_pager.h>
74 #include <machine/riscvreg.h>
75 #include <machine/cpu.h>
76 #include <machine/kdb.h>
77 #include <machine/machdep.h>
78 #include <machine/pcb.h>
79 #include <machine/reg.h>
80 #include <machine/trap.h>
81 #include <machine/vmparam.h>
82 #include <machine/intr.h>
83 #include <machine/sbi.h>
85 #include <machine/asm.h>
88 #include <machine/fpe.h>
92 #include <dev/fdt/fdt_common.h>
93 #include <dev/ofw/openfirm.h>
96 struct pcpu __pcpu[MAXCPU];
98 static struct trapframe proc0_tf;
100 vm_paddr_t phys_avail[PHYS_AVAIL_SIZE + 2];
101 vm_paddr_t dump_avail[PHYS_AVAIL_SIZE + 2];
108 #define DTB_SIZE_MAX (1024 * 1024)
110 #define PHYSMAP_SIZE (2 * (VM_PHYSSEG_MAX - 1))
111 vm_paddr_t physmap[PHYSMAP_SIZE];
114 struct kva_md_info kmi;
116 int64_t dcache_line_size; /* The minimum D cache line size */
117 int64_t icache_line_size; /* The minimum I cache line size */
118 int64_t idcache_line_size; /* The minimum cache line size */
121 extern int *initstack_end;
123 uintptr_t mcall_trap(uintptr_t mcause, uintptr_t* regs);
126 mcall_trap(uintptr_t mcause, uintptr_t* regs)
133 cpu_startup(void *dummy)
138 vm_ksubmap_init(&kmi);
140 vm_pager_bufferinit();
143 SYSINIT(cpu, SI_SUB_CPU, SI_ORDER_FIRST, cpu_startup, NULL);
146 cpu_idle_wakeup(int cpu)
153 fill_regs(struct thread *td, struct reg *regs)
155 struct trapframe *frame;
157 frame = td->td_frame;
158 regs->sepc = frame->tf_sepc;
159 regs->sstatus = frame->tf_sstatus;
160 regs->ra = frame->tf_ra;
161 regs->sp = frame->tf_sp;
162 regs->gp = frame->tf_gp;
163 regs->tp = frame->tf_tp;
165 memcpy(regs->t, frame->tf_t, sizeof(regs->t));
166 memcpy(regs->s, frame->tf_s, sizeof(regs->s));
167 memcpy(regs->a, frame->tf_a, sizeof(regs->a));
173 set_regs(struct thread *td, struct reg *regs)
175 struct trapframe *frame;
177 frame = td->td_frame;
178 frame->tf_sepc = regs->sepc;
179 frame->tf_ra = regs->ra;
180 frame->tf_sp = regs->sp;
181 frame->tf_gp = regs->gp;
182 frame->tf_tp = regs->tp;
184 memcpy(frame->tf_t, regs->t, sizeof(frame->tf_t));
185 memcpy(frame->tf_s, regs->s, sizeof(frame->tf_s));
186 memcpy(frame->tf_a, regs->a, sizeof(frame->tf_a));
192 fill_fpregs(struct thread *td, struct fpreg *regs)
199 if ((pcb->pcb_fpflags & PCB_FP_STARTED) != 0) {
201 * If we have just been running FPE instructions we will
202 * need to save the state to memcpy it below.
207 memcpy(regs->fp_x, pcb->pcb_x, sizeof(regs->fp_x));
208 regs->fp_fcsr = pcb->pcb_fcsr;
211 memset(regs, 0, sizeof(*regs));
217 set_fpregs(struct thread *td, struct fpreg *regs)
220 struct trapframe *frame;
223 frame = td->td_frame;
226 memcpy(pcb->pcb_x, regs->fp_x, sizeof(regs->fp_x));
227 pcb->pcb_fcsr = regs->fp_fcsr;
228 pcb->pcb_fpflags |= PCB_FP_STARTED;
229 frame->tf_sstatus &= ~SSTATUS_FS_MASK;
230 frame->tf_sstatus |= SSTATUS_FS_CLEAN;
237 fill_dbregs(struct thread *td, struct dbreg *regs)
240 panic("fill_dbregs");
244 set_dbregs(struct thread *td, struct dbreg *regs)
251 ptrace_set_pc(struct thread *td, u_long addr)
254 td->td_frame->tf_sepc = addr;
259 ptrace_single_step(struct thread *td)
267 ptrace_clear_single_step(struct thread *td)
275 exec_setregs(struct thread *td, struct image_params *imgp, u_long stack)
277 struct trapframe *tf;
283 memset(tf, 0, sizeof(struct trapframe));
286 tf->tf_sp = STACKALIGN(stack);
287 tf->tf_ra = imgp->entry_addr;
288 tf->tf_sepc = imgp->entry_addr;
290 pcb->pcb_fpflags &= ~PCB_FP_STARTED;
293 /* Sanity check these are the same size, they will be memcpy'd to and fro */
294 CTASSERT(sizeof(((struct trapframe *)0)->tf_a) ==
295 sizeof((struct gpregs *)0)->gp_a);
296 CTASSERT(sizeof(((struct trapframe *)0)->tf_s) ==
297 sizeof((struct gpregs *)0)->gp_s);
298 CTASSERT(sizeof(((struct trapframe *)0)->tf_t) ==
299 sizeof((struct gpregs *)0)->gp_t);
300 CTASSERT(sizeof(((struct trapframe *)0)->tf_a) ==
301 sizeof((struct reg *)0)->a);
302 CTASSERT(sizeof(((struct trapframe *)0)->tf_s) ==
303 sizeof((struct reg *)0)->s);
304 CTASSERT(sizeof(((struct trapframe *)0)->tf_t) ==
305 sizeof((struct reg *)0)->t);
307 /* Support for FDT configurations only. */
311 get_mcontext(struct thread *td, mcontext_t *mcp, int clear_ret)
313 struct trapframe *tf = td->td_frame;
315 memcpy(mcp->mc_gpregs.gp_t, tf->tf_t, sizeof(mcp->mc_gpregs.gp_t));
316 memcpy(mcp->mc_gpregs.gp_s, tf->tf_s, sizeof(mcp->mc_gpregs.gp_s));
317 memcpy(mcp->mc_gpregs.gp_a, tf->tf_a, sizeof(mcp->mc_gpregs.gp_a));
319 if (clear_ret & GET_MC_CLEAR_RET) {
320 mcp->mc_gpregs.gp_a[0] = 0;
321 mcp->mc_gpregs.gp_t[0] = 0; /* clear syscall error */
324 mcp->mc_gpregs.gp_ra = tf->tf_ra;
325 mcp->mc_gpregs.gp_sp = tf->tf_sp;
326 mcp->mc_gpregs.gp_gp = tf->tf_gp;
327 mcp->mc_gpregs.gp_tp = tf->tf_tp;
328 mcp->mc_gpregs.gp_sepc = tf->tf_sepc;
329 mcp->mc_gpregs.gp_sstatus = tf->tf_sstatus;
335 set_mcontext(struct thread *td, mcontext_t *mcp)
337 struct trapframe *tf;
341 memcpy(tf->tf_t, mcp->mc_gpregs.gp_t, sizeof(tf->tf_t));
342 memcpy(tf->tf_s, mcp->mc_gpregs.gp_s, sizeof(tf->tf_s));
343 memcpy(tf->tf_a, mcp->mc_gpregs.gp_a, sizeof(tf->tf_a));
345 tf->tf_ra = mcp->mc_gpregs.gp_ra;
346 tf->tf_sp = mcp->mc_gpregs.gp_sp;
347 tf->tf_gp = mcp->mc_gpregs.gp_gp;
348 tf->tf_sepc = mcp->mc_gpregs.gp_sepc;
349 tf->tf_sstatus = mcp->mc_gpregs.gp_sstatus;
355 get_fpcontext(struct thread *td, mcontext_t *mcp)
362 curpcb = curthread->td_pcb;
364 KASSERT(td->td_pcb == curpcb, ("Invalid fpe pcb"));
366 if ((curpcb->pcb_fpflags & PCB_FP_STARTED) != 0) {
368 * If we have just been running FPE instructions we will
369 * need to save the state to memcpy it below.
373 KASSERT((curpcb->pcb_fpflags & ~PCB_FP_USERMASK) == 0,
374 ("Non-userspace FPE flags set in get_fpcontext"));
375 memcpy(mcp->mc_fpregs.fp_x, curpcb->pcb_x,
376 sizeof(mcp->mc_fpregs));
377 mcp->mc_fpregs.fp_fcsr = curpcb->pcb_fcsr;
378 mcp->mc_fpregs.fp_flags = curpcb->pcb_fpflags;
379 mcp->mc_flags |= _MC_FP_VALID;
387 set_fpcontext(struct thread *td, mcontext_t *mcp)
394 if ((mcp->mc_flags & _MC_FP_VALID) != 0) {
395 curpcb = curthread->td_pcb;
396 /* FPE usage is enabled, override registers. */
397 memcpy(curpcb->pcb_x, mcp->mc_fpregs.fp_x,
398 sizeof(mcp->mc_fpregs));
399 curpcb->pcb_fcsr = mcp->mc_fpregs.fp_fcsr;
400 curpcb->pcb_fpflags = mcp->mc_fpregs.fp_flags & PCB_FP_USERMASK;
414 if (!sched_runnable())
429 __asm __volatile("wfi");
433 * Flush the D-cache for non-DMA I/O so that the I-cache can
434 * be made coherent later.
437 cpu_flush_dcache(void *ptr, size_t len)
443 /* Get current clock frequency for the given CPU ID. */
445 cpu_est_clockrate(int cpu_id, uint64_t *rate)
448 panic("cpu_est_clockrate");
452 cpu_pcpu_init(struct pcpu *pcpu, int cpuid, size_t size)
462 if (td->td_md.md_spinlock_count == 0) {
463 td->td_md.md_spinlock_count = 1;
464 td->td_md.md_saved_sstatus_ie = intr_disable();
466 td->td_md.md_spinlock_count++;
474 register_t sstatus_ie;
478 sstatus_ie = td->td_md.md_saved_sstatus_ie;
479 td->td_md.md_spinlock_count--;
480 if (td->td_md.md_spinlock_count == 0)
481 intr_restore(sstatus_ie);
484 #ifndef _SYS_SYSPROTO_H_
485 struct sigreturn_args {
491 sys_sigreturn(struct thread *td, struct sigreturn_args *uap)
499 if (copyin(uap->sigcntxp, &uc, sizeof(uc)))
503 * Make sure the processor mode has not been tampered with and
504 * interrupts have not been disabled.
505 * Supervisor interrupts in user mode are always enabled.
507 sstatus = uc.uc_mcontext.mc_gpregs.gp_sstatus;
508 if ((sstatus & SSTATUS_SPP) != 0)
511 error = set_mcontext(td, &uc.uc_mcontext);
515 set_fpcontext(td, &uc.uc_mcontext);
517 /* Restore signal mask. */
518 kern_sigprocmask(td, SIG_SETMASK, &uc.uc_sigmask, NULL, 0);
520 return (EJUSTRETURN);
524 * Construct a PCB from a trapframe. This is called from kdb_trap() where
525 * we want to start a backtrace from the function that caused us to enter
526 * the debugger. We have the context in the trapframe, but base the trace
527 * on the PCB. The PCB doesn't have to be perfect, as long as it contains
528 * enough for a backtrace.
531 makectx(struct trapframe *tf, struct pcb *pcb)
534 memcpy(pcb->pcb_t, tf->tf_t, sizeof(tf->tf_t));
535 memcpy(pcb->pcb_s, tf->tf_s, sizeof(tf->tf_s));
536 memcpy(pcb->pcb_a, tf->tf_a, sizeof(tf->tf_a));
538 pcb->pcb_ra = tf->tf_ra;
539 pcb->pcb_sp = tf->tf_sp;
540 pcb->pcb_gp = tf->tf_gp;
541 pcb->pcb_tp = tf->tf_tp;
542 pcb->pcb_sepc = tf->tf_sepc;
546 sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
548 struct sigframe *fp, frame;
549 struct sysentvec *sysent;
550 struct trapframe *tf;
559 PROC_LOCK_ASSERT(p, MA_OWNED);
561 sig = ksi->ksi_signo;
563 mtx_assert(&psp->ps_mtx, MA_OWNED);
566 onstack = sigonstack(tf->tf_sp);
568 CTR4(KTR_SIG, "sendsig: td=%p (%s) catcher=%p sig=%d", td, p->p_comm,
571 /* Allocate and validate space for the signal handler context. */
572 if ((td->td_pflags & TDP_ALTSTACK) != 0 && !onstack &&
573 SIGISMEMBER(psp->ps_sigonstack, sig)) {
574 fp = (struct sigframe *)((uintptr_t)td->td_sigstk.ss_sp +
575 td->td_sigstk.ss_size);
577 fp = (struct sigframe *)td->td_frame->tf_sp;
580 /* Make room, keeping the stack aligned */
582 fp = (struct sigframe *)STACKALIGN(fp);
584 /* Fill in the frame to copy out */
585 bzero(&frame, sizeof(frame));
586 get_mcontext(td, &frame.sf_uc.uc_mcontext, 0);
587 get_fpcontext(td, &frame.sf_uc.uc_mcontext);
588 frame.sf_si = ksi->ksi_info;
589 frame.sf_uc.uc_sigmask = *mask;
590 frame.sf_uc.uc_stack = td->td_sigstk;
591 frame.sf_uc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK) != 0 ?
592 (onstack ? SS_ONSTACK : 0) : SS_DISABLE;
593 mtx_unlock(&psp->ps_mtx);
594 PROC_UNLOCK(td->td_proc);
596 /* Copy the sigframe out to the user's stack. */
597 if (copyout(&frame, fp, sizeof(*fp)) != 0) {
598 /* Process has trashed its stack. Kill it. */
599 CTR2(KTR_SIG, "sendsig: sigexit td=%p fp=%p", td, fp);
605 tf->tf_a[1] = (register_t)&fp->sf_si;
606 tf->tf_a[2] = (register_t)&fp->sf_uc;
608 tf->tf_sepc = (register_t)catcher;
609 tf->tf_sp = (register_t)fp;
611 sysent = p->p_sysent;
612 if (sysent->sv_sigcode_base != 0)
613 tf->tf_ra = (register_t)sysent->sv_sigcode_base;
615 tf->tf_ra = (register_t)(sysent->sv_psstrings -
616 *(sysent->sv_szsigcode));
618 CTR3(KTR_SIG, "sendsig: return td=%p pc=%#x sp=%#x", td, tf->tf_sepc,
622 mtx_lock(&psp->ps_mtx);
626 init_proc0(vm_offset_t kstack)
632 proc_linkup0(&proc0, &thread0);
633 thread0.td_kstack = kstack;
634 thread0.td_pcb = (struct pcb *)(thread0.td_kstack) - 1;
635 thread0.td_pcb->pcb_fpflags = 0;
636 thread0.td_frame = &proc0_tf;
637 pcpup->pc_curpcb = thread0.td_pcb;
641 add_physmap_entry(uint64_t base, uint64_t length, vm_paddr_t *physmap,
644 u_int i, insert_idx, _physmap_idx;
646 _physmap_idx = *physmap_idxp;
652 * Find insertion point while checking for overlap. Start off by
653 * assuming the new entry will be added to the end.
655 insert_idx = _physmap_idx;
656 for (i = 0; i <= _physmap_idx; i += 2) {
657 if (base < physmap[i + 1]) {
658 if (base + length <= physmap[i]) {
662 if (boothowto & RB_VERBOSE)
664 "Overlapping memory regions, ignoring second region\n");
669 /* See if we can prepend to the next entry. */
670 if (insert_idx <= _physmap_idx &&
671 base + length == physmap[insert_idx]) {
672 physmap[insert_idx] = base;
676 /* See if we can append to the previous entry. */
677 if (insert_idx > 0 && base == physmap[insert_idx - 1]) {
678 physmap[insert_idx - 1] += length;
683 *physmap_idxp = _physmap_idx;
684 if (_physmap_idx == PHYSMAP_SIZE) {
686 "Too many segments in the physical address map, giving up\n");
691 * Move the last 'N' entries down to make room for the new
694 for (i = _physmap_idx; i > insert_idx; i -= 2) {
695 physmap[i] = physmap[i - 2];
696 physmap[i + 1] = physmap[i - 1];
699 /* Insert the new entry. */
700 physmap[insert_idx] = base;
701 physmap[insert_idx + 1] = base + length;
703 printf("physmap[%d] = 0x%016lx\n", insert_idx, base);
704 printf("physmap[%d] = 0x%016lx\n", insert_idx + 1, base + length);
710 try_load_dtb(caddr_t kmdp, vm_offset_t dtbp)
713 #if defined(FDT_DTB_STATIC)
714 dtbp = (vm_offset_t)&fdt_static_dtb;
717 if (dtbp == (vm_offset_t)NULL) {
718 printf("ERROR loading DTB\n");
722 if (OF_install(OFW_FDT, 0) == FALSE)
723 panic("Cannot install FDT");
725 if (OF_init((void *)dtbp) != 0)
726 panic("OF_init failed with the found device tree");
738 * Fake up a boot descriptor table.
739 * RISCVTODO: This needs to be done via loader (when it's available).
742 fake_preload_metadata(struct riscv_bootparams *rvbp __unused)
744 static uint32_t fake_preload[35];
746 vm_offset_t zstart = 0, zend = 0;
748 vm_offset_t lastaddr;
753 fake_preload[i++] = MODINFO_NAME;
754 fake_preload[i++] = strlen("kernel") + 1;
755 strcpy((char*)&fake_preload[i++], "kernel");
757 fake_preload[i++] = MODINFO_TYPE;
758 fake_preload[i++] = strlen("elf64 kernel") + 1;
759 strcpy((char*)&fake_preload[i++], "elf64 kernel");
761 fake_preload[i++] = MODINFO_ADDR;
762 fake_preload[i++] = sizeof(vm_offset_t);
763 *(vm_offset_t *)&fake_preload[i++] =
764 (vm_offset_t)(KERNBASE + KERNENTRY);
766 fake_preload[i++] = MODINFO_SIZE;
767 fake_preload[i++] = sizeof(vm_offset_t);
768 fake_preload[i++] = (vm_offset_t)&end -
769 (vm_offset_t)(KERNBASE + KERNENTRY);
774 if (*(uint32_t *)KERNVIRTADDR == MAGIC_TRAMP_NUMBER) {
775 fake_preload[i++] = MODINFO_METADATA|MODINFOMD_SSYM;
776 fake_preload[i++] = sizeof(vm_offset_t);
777 fake_preload[i++] = *(uint32_t *)(KERNVIRTADDR + 4);
778 fake_preload[i++] = MODINFO_METADATA|MODINFOMD_ESYM;
779 fake_preload[i++] = sizeof(vm_offset_t);
780 fake_preload[i++] = *(uint32_t *)(KERNVIRTADDR + 8);
781 lastaddr = *(uint32_t *)(KERNVIRTADDR + 8);
783 zstart = *(uint32_t *)(KERNVIRTADDR + 4);
784 db_fetch_ksymtab(zstart, zend);
788 lastaddr = (vm_offset_t)&end;
789 fake_preload[i++] = 0;
791 preload_metadata = (void *)fake_preload;
797 initriscv(struct riscv_bootparams *rvbp)
799 struct mem_region mem_regions[FDT_MEM_REGIONS];
801 vm_offset_t rstart, rend;
804 vm_offset_t lastaddr;
809 /* Set the pcpu data, this is needed by pmap_bootstrap */
811 pcpu_init(pcpup, 0, sizeof(struct pcpu));
813 /* Set the pcpu pointer */
814 __asm __volatile("mv gp, %0" :: "r"(pcpup));
816 PCPU_SET(curthread, &thread0);
818 /* Set the module data location */
819 lastaddr = fake_preload_metadata(rvbp);
821 /* Find the kernel address */
822 kmdp = preload_search_by_type("elf kernel");
824 kmdp = preload_search_by_type("elf64 kernel");
826 boothowto = RB_VERBOSE | RB_SINGLE;
827 boothowto = RB_VERBOSE;
832 try_load_dtb(kmdp, rvbp->dtbp_virt);
835 /* Load the physical memory ranges */
839 /* Grab physical memory regions information from device tree. */
840 if (fdt_get_mem_regions(mem_regions, &mem_regions_sz, NULL) != 0)
841 panic("Cannot get physical memory regions");
844 e = s + DTB_SIZE_MAX;
846 for (i = 0; i < mem_regions_sz; i++) {
847 rstart = mem_regions[i].mr_start;
848 rend = (mem_regions[i].mr_start + mem_regions[i].mr_size);
850 if ((rstart < s) && (rend > e)) {
851 /* Exclude DTB region. */
852 add_physmap_entry(rstart, (s - rstart), physmap, &physmap_idx);
853 add_physmap_entry(e, (rend - e), physmap, &physmap_idx);
855 add_physmap_entry(mem_regions[i].mr_start,
856 mem_regions[i].mr_size, physmap, &physmap_idx);
861 /* Do basic tuning, hz etc */
866 /* Bootstrap enough of pmap to enter the kernel proper */
867 kernlen = (lastaddr - KERNBASE);
868 pmap_bootstrap(rvbp->kern_l1pt, mem_regions[0].mr_start, kernlen);
872 init_proc0(rvbp->kern_stack);
874 msgbufinit(msgbufp, msgbufsize);
876 init_param2(physmem);
884 bzero(void *buf, size_t len)