2 * Copyright (C) 1994, David Greenman
3 * Copyright (c) 1990, 1993
4 * The Regents of the University of California. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
7 * the University of Utah, and William Jolitz.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by the University of
20 * California, Berkeley and its contributors.
21 * 4. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * from: @(#)trap.c 7.4 (Berkeley) 5/13/91
40 #include <sys/cdefs.h>
41 __FBSDID("$FreeBSD$");
44 * 386 Trap and System call handling
47 #include "opt_clock.h"
49 #include "opt_hwpmc_hooks.h"
52 #include "opt_kdtrace.h"
53 #include "opt_ktrace.h"
57 #include <sys/param.h>
59 #include <sys/systm.h>
61 #include <sys/pioctl.h>
62 #include <sys/ptrace.h>
64 #include <sys/kernel.h>
67 #include <sys/mutex.h>
68 #include <sys/resourcevar.h>
69 #include <sys/signalvar.h>
70 #include <sys/syscall.h>
71 #include <sys/sysctl.h>
72 #include <sys/sysent.h>
74 #include <sys/vmmeter.h>
76 #include <sys/ktrace.h>
79 #include <sys/pmckern.h>
81 #include <security/audit/audit.h>
84 #include <vm/vm_param.h>
86 #include <vm/vm_kern.h>
87 #include <vm/vm_map.h>
88 #include <vm/vm_page.h>
89 #include <vm/vm_extern.h>
91 #include <machine/cpu.h>
92 #include <machine/intr_machdep.h>
93 #include <machine/mca.h>
94 #include <machine/md_var.h>
95 #include <machine/pcb.h>
97 #include <machine/smp.h>
99 #include <machine/tss.h>
100 #include <machine/vm86.h>
103 #include <sys/syslog.h>
104 #include <machine/clock.h>
108 #include <sys/dtrace_bsd.h>
111 * This is a hook which is initialised by the dtrace module
112 * to handle traps which might occur during DTrace probe
115 dtrace_trap_func_t dtrace_trap_func;
117 dtrace_doubletrap_func_t dtrace_doubletrap_func;
120 * This is a hook which is initialised by the systrace module
121 * when it is loaded. This keeps the DTrace syscall provider
122 * implementation opaque.
124 systrace_probe_func_t systrace_probe_func;
127 * These hooks are necessary for the pid, usdt and fasttrap providers.
129 dtrace_fasttrap_probe_ptr_t dtrace_fasttrap_probe_ptr;
130 dtrace_pid_probe_ptr_t dtrace_pid_probe_ptr;
131 dtrace_return_probe_ptr_t dtrace_return_probe_ptr;
134 extern void trap(struct trapframe *frame);
135 extern void syscall(struct trapframe *frame);
137 static int trap_pfault(struct trapframe *, int, vm_offset_t);
138 static void trap_fatal(struct trapframe *, vm_offset_t);
139 void dblfault_handler(void);
141 extern inthand_t IDTVEC(lcall_syscall);
143 #define MAX_TRAP_MSG 30
144 static char *trap_msg[] = {
146 "privileged instruction fault", /* 1 T_PRIVINFLT */
148 "breakpoint instruction fault", /* 3 T_BPTFLT */
151 "arithmetic trap", /* 6 T_ARITHTRAP */
154 "general protection fault", /* 9 T_PROTFLT */
155 "trace trap", /* 10 T_TRCTRAP */
157 "page fault", /* 12 T_PAGEFLT */
159 "alignment fault", /* 14 T_ALIGNFLT */
163 "integer divide fault", /* 18 T_DIVIDE */
164 "non-maskable interrupt trap", /* 19 T_NMI */
165 "overflow trap", /* 20 T_OFLOW */
166 "FPU bounds check fault", /* 21 T_BOUND */
167 "FPU device not available", /* 22 T_DNA */
168 "double fault", /* 23 T_DOUBLEFLT */
169 "FPU operand fetch fault", /* 24 T_FPOPFLT */
170 "invalid TSS fault", /* 25 T_TSSFLT */
171 "segment not present fault", /* 26 T_SEGNPFLT */
172 "stack fault", /* 27 T_STKFLT */
173 "machine check trap", /* 28 T_MCHK */
174 "SIMD floating-point exception", /* 29 T_XMMFLT */
175 "reserved (unknown) fault", /* 30 T_RESERVED */
178 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
179 extern int has_f00f_bug;
183 static int kdb_on_nmi = 1;
184 SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RW,
185 &kdb_on_nmi, 0, "Go to KDB on NMI");
187 static int panic_on_nmi = 1;
188 SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
189 &panic_on_nmi, 0, "Panic on NMI");
190 static int prot_fault_translation = 0;
191 SYSCTL_INT(_machdep, OID_AUTO, prot_fault_translation, CTLFLAG_RW,
192 &prot_fault_translation, 0, "Select signal to deliver on protection fault");
195 * Exception, fault, and trap interface to the FreeBSD kernel.
196 * This common code is called from assembly language IDT gate entry
197 * routines that prepare a suitable stack frame, and restore this
198 * frame after the exception has been processed.
202 trap(struct trapframe *frame)
204 struct thread *td = curthread;
205 struct proc *p = td->td_proc;
206 int i = 0, ucode = 0, code;
212 static int lastalert = 0;
215 PCPU_INC(cnt.v_trap);
216 type = frame->tf_trapno;
219 /* Handler for NMI IPIs used for stopping CPUs. */
221 if (ipi_nmi_handler() == 0)
233 if (type == T_RESERVED) {
234 trap_fatal(frame, 0);
240 * CPU PMCs interrupt using an NMI so we check for that first.
241 * If the HWPMC module is active, 'pmc_hook' will point to
242 * the function to be called. A return value of '1' from the
243 * hook means that the NMI was handled by it and that we can
244 * return immediately.
246 if (type == T_NMI && pmc_intr &&
247 (*pmc_intr)(PCPU_GET(cpuid), frame))
251 if (type == T_MCHK) {
253 trap_fatal(frame, 0);
259 * A trap can occur while DTrace executes a probe. Before
260 * executing the probe, DTrace blocks re-scheduling and sets
261 * a flag in it's per-cpu flags to indicate that it doesn't
262 * want to fault. On returning from the probe, the no-fault
263 * flag is cleared and finally re-scheduling is enabled.
265 * If the DTrace kernel module has registered a trap handler,
266 * call it and if it returns non-zero, assume that it has
267 * handled the trap and modified the trap frame so that this
268 * function can return normally.
270 if ((type == T_PROTFLT || type == T_PAGEFLT) &&
271 dtrace_trap_func != NULL)
272 if ((*dtrace_trap_func)(frame, type))
274 if (type == T_DTRACE_PROBE || type == T_DTRACE_RET ||
278 regs.r_fs = frame->tf_fs;
279 regs.r_es = frame->tf_es;
280 regs.r_ds = frame->tf_ds;
281 regs.r_edi = frame->tf_edi;
282 regs.r_esi = frame->tf_esi;
283 regs.r_ebp = frame->tf_ebp;
284 regs.r_ebx = frame->tf_ebx;
285 regs.r_edx = frame->tf_edx;
286 regs.r_ecx = frame->tf_ecx;
287 regs.r_eax = frame->tf_eax;
288 regs.r_eip = frame->tf_eip;
289 regs.r_cs = frame->tf_cs;
290 regs.r_eflags = frame->tf_eflags;
291 regs.r_esp = frame->tf_esp;
292 regs.r_ss = frame->tf_ss;
293 if (type == T_DTRACE_PROBE &&
294 dtrace_fasttrap_probe_ptr != NULL &&
295 dtrace_fasttrap_probe_ptr(®s) == 0)
297 if (type == T_BPTFLT &&
298 dtrace_pid_probe_ptr != NULL &&
299 dtrace_pid_probe_ptr(®s) == 0)
301 if (type == T_DTRACE_RET &&
302 dtrace_return_probe_ptr != NULL &&
303 dtrace_return_probe_ptr(®s) == 0)
308 if ((frame->tf_eflags & PSL_I) == 0) {
310 * Buggy application or kernel code has disabled
311 * interrupts and then trapped. Enabling interrupts
312 * now is wrong, but it is better than running with
313 * interrupts disabled until they are accidentally
316 if (ISPL(frame->tf_cs) == SEL_UPL || (frame->tf_eflags & PSL_VM))
318 "pid %ld (%s): trap %d with interrupts disabled\n",
319 (long)curproc->p_pid, curthread->td_name, type);
320 else if (type != T_BPTFLT && type != T_TRCTRAP &&
321 frame->tf_eip != (int)cpu_switch_load_gs) {
323 * XXX not quite right, since this may be for a
324 * multiple fault in user mode.
326 printf("kernel trap %d with interrupts disabled\n",
329 * Page faults need interrupts disabled until later,
330 * and we shouldn't enable interrupts while holding
331 * a spin lock or if servicing an NMI.
333 if (type != T_NMI && type != T_PAGEFLT &&
334 td->td_md.md_spinlock_count == 0)
339 code = frame->tf_err;
340 if (type == T_PAGEFLT) {
342 * For some Cyrix CPUs, %cr2 is clobbered by
343 * interrupts. This problem is worked around by using
344 * an interrupt gate for the pagefault handler. We
345 * are finally ready to read %cr2 and then must
346 * reenable interrupts.
348 * If we get a page fault while in a critical section, then
349 * it is most likely a fatal kernel page fault. The kernel
350 * is already going to panic trying to get a sleep lock to
351 * do the VM lookup, so just consider it a fatal trap so the
352 * kernel can print out a useful trap message and even get
355 * If we get a page fault while holding a non-sleepable
356 * lock, then it is most likely a fatal kernel page fault.
357 * If WITNESS is enabled, then it's going to whine about
358 * bogus LORs with various VM locks, so just skip to the
359 * fatal trap handling directly.
362 if (td->td_critnest != 0 ||
363 WITNESS_CHECK(WARN_SLEEPOK | WARN_GIANTOK, NULL,
364 "Kernel page fault") != 0)
365 trap_fatal(frame, eva);
370 if ((ISPL(frame->tf_cs) == SEL_UPL) ||
371 ((frame->tf_eflags & PSL_VM) &&
372 !(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL))) {
376 td->td_frame = frame;
377 addr = frame->tf_eip;
378 if (td->td_ucred != p->p_ucred)
379 cred_update_thread(td);
382 case T_PRIVINFLT: /* privileged instruction fault */
387 case T_BPTFLT: /* bpt instruction fault */
388 case T_TRCTRAP: /* trace trap */
390 frame->tf_eflags &= ~PSL_T;
392 ucode = (type == T_TRCTRAP ? TRAP_TRACE : TRAP_BRKPT);
395 case T_ARITHTRAP: /* arithmetic trap */
407 * The following two traps can happen in
408 * vm86 mode, and, if so, we want to handle
411 case T_PROTFLT: /* general protection fault */
412 case T_STKFLT: /* stack fault */
413 if (frame->tf_eflags & PSL_VM) {
414 i = vm86_emulate((struct vm86frame *)frame);
420 ucode = (type == T_PROTFLT) ? BUS_OBJERR : BUS_ADRERR;
422 case T_SEGNPFLT: /* segment not present fault */
426 case T_TSSFLT: /* invalid TSS fault */
430 case T_DOUBLEFLT: /* double fault */
436 case T_PAGEFLT: /* page fault */
438 i = trap_pfault(frame, TRUE, eva);
439 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
442 * The f00f hack workaround has triggered, so
443 * treat the fault as an illegal instruction
444 * (T_PRIVINFLT) instead of a page fault.
446 type = frame->tf_trapno = T_PRIVINFLT;
448 /* Proceed as in that case. */
462 if (prot_fault_translation == 0) {
465 * This check also covers the images
466 * without the ABI-tag ELF note.
468 if (SV_CURPROC_ABI() == SV_ABI_FREEBSD
469 && p->p_osrel >= P_OSREL_SIGSEGV) {
474 ucode = BUS_PAGE_FAULT;
476 } else if (prot_fault_translation == 1) {
478 * Always compat mode.
481 ucode = BUS_PAGE_FAULT;
484 * Always SIGSEGV mode.
493 case T_DIVIDE: /* integer divide fault */
502 # define TIMER_FREQ 1193182
504 if (time_second - lastalert > 10) {
505 log(LOG_WARNING, "NMI: power fail\n");
507 lastalert = time_second;
510 #else /* !POWERFAIL_NMI */
511 /* machine/parity/power fail/"kitchen sink" faults */
512 if (isa_nmi(code) == 0) {
515 * NMI can be hooked up to a pushbutton
519 printf ("NMI ... going to debugger\n");
520 kdb_trap(type, 0, frame);
524 } else if (panic_on_nmi)
525 panic("NMI indicates hardware failure");
527 #endif /* POWERFAIL_NMI */
530 case T_OFLOW: /* integer overflow fault */
535 case T_BOUND: /* bounds check fault */
542 KASSERT(PCB_USER_FPU(td->td_pcb),
543 ("kernel FPU ctx has leaked"));
544 /* transparent fault (due to context switch "late") */
548 uprintf("pid %d killed due to lack of floating point\n",
554 case T_FPOPFLT: /* FPU operand fetch fault */
559 case T_XMMFLT: /* SIMD floating-point exception */
567 KASSERT(cold || td->td_ucred != NULL,
568 ("kernel trap doesn't have ucred"));
570 case T_PAGEFLT: /* page fault */
571 (void) trap_pfault(frame, FALSE, eva);
576 KASSERT(!PCB_USER_FPU(td->td_pcb),
577 ("Unregistered use of FPU in kernel"));
583 case T_ARITHTRAP: /* arithmetic trap */
584 case T_XMMFLT: /* SIMD floating-point exception */
585 case T_FPOPFLT: /* FPU operand fetch fault */
587 * XXXKIB for now disable any FPU traps in kernel
588 * handler registration seems to be overkill
590 trap_fatal(frame, 0);
594 * The following two traps can happen in
595 * vm86 mode, and, if so, we want to handle
598 case T_PROTFLT: /* general protection fault */
599 case T_STKFLT: /* stack fault */
600 if (frame->tf_eflags & PSL_VM) {
601 i = vm86_emulate((struct vm86frame *)frame);
604 * returns to original process
606 vm86_trap((struct vm86frame *)frame);
609 if (type == T_STKFLT)
614 case T_SEGNPFLT: /* segment not present fault */
615 if (PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)
619 * Invalid %fs's and %gs's can be created using
620 * procfs or PT_SETREGS or by invalidating the
621 * underlying LDT entry. This causes a fault
622 * in kernel mode when the kernel attempts to
623 * switch contexts. Lose the bad context
624 * (XXX) so that we can continue, and generate
627 if (frame->tf_eip == (int)cpu_switch_load_gs) {
628 PCPU_GET(curpcb)->pcb_gs = 0;
637 if (td->td_intr_nesting_level != 0)
641 * Invalid segment selectors and out of bounds
642 * %eip's and %esp's can be set up in user mode.
643 * This causes a fault in kernel mode when the
644 * kernel tries to return to user mode. We want
645 * to get this fault so that we can fix the
646 * problem here and not have to check all the
647 * selectors and pointers when the user changes
650 if (frame->tf_eip == (int)doreti_iret) {
651 frame->tf_eip = (int)doreti_iret_fault;
654 if (frame->tf_eip == (int)doreti_popl_ds) {
655 frame->tf_eip = (int)doreti_popl_ds_fault;
658 if (frame->tf_eip == (int)doreti_popl_es) {
659 frame->tf_eip = (int)doreti_popl_es_fault;
662 if (frame->tf_eip == (int)doreti_popl_fs) {
663 frame->tf_eip = (int)doreti_popl_fs_fault;
666 if (PCPU_GET(curpcb)->pcb_onfault != NULL) {
668 (int)PCPU_GET(curpcb)->pcb_onfault;
675 * PSL_NT can be set in user mode and isn't cleared
676 * automatically when the kernel is entered. This
677 * causes a TSS fault when the kernel attempts to
678 * `iret' because the TSS link is uninitialized. We
679 * want to get this fault so that we can fix the
680 * problem here and not every time the kernel is
683 if (frame->tf_eflags & PSL_NT) {
684 frame->tf_eflags &= ~PSL_NT;
689 case T_TRCTRAP: /* trace trap */
690 if (frame->tf_eip == (int)IDTVEC(lcall_syscall)) {
692 * We've just entered system mode via the
693 * syscall lcall. Continue single stepping
694 * silently until the syscall handler has
699 if (frame->tf_eip == (int)IDTVEC(lcall_syscall) + 1) {
701 * The syscall handler has now saved the
702 * flags. Stop single stepping it.
704 frame->tf_eflags &= ~PSL_T;
708 * Ignore debug register trace traps due to
709 * accesses in the user's address space, which
710 * can happen under several conditions such as
711 * if a user sets a watchpoint on a buffer and
712 * then passes that buffer to a system call.
713 * We still want to get TRCTRAPS for addresses
714 * in kernel space because that is useful when
715 * debugging the kernel.
717 if (user_dbreg_trap() &&
718 !(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)) {
720 * Reset breakpoint bits because the
723 load_dr6(rdr6() & 0xfffffff0);
727 * FALLTHROUGH (TRCTRAP kernel mode, kernel address)
731 * If KDB is enabled, let it handle the debugger trap.
732 * Otherwise, debugger traps "can't happen".
735 if (kdb_trap(type, 0, frame))
743 if (time_second - lastalert > 10) {
744 log(LOG_WARNING, "NMI: power fail\n");
746 lastalert = time_second;
749 #else /* !POWERFAIL_NMI */
750 /* machine/parity/power fail/"kitchen sink" faults */
751 if (isa_nmi(code) == 0) {
754 * NMI can be hooked up to a pushbutton
758 printf ("NMI ... going to debugger\n");
759 kdb_trap(type, 0, frame);
763 } else if (panic_on_nmi == 0)
766 #endif /* POWERFAIL_NMI */
770 trap_fatal(frame, eva);
774 /* Translate fault for emulators (e.g. Linux) */
775 if (*p->p_sysent->sv_transtrap)
776 i = (*p->p_sysent->sv_transtrap)(i, type);
778 ksiginfo_init_trap(&ksi);
780 ksi.ksi_code = ucode;
781 ksi.ksi_addr = (void *)addr;
782 ksi.ksi_trapno = type;
783 trapsignal(td, &ksi);
786 if (type <= MAX_TRAP_MSG) {
787 uprintf("fatal process exception: %s",
789 if ((type == T_PAGEFLT) || (type == T_PROTFLT))
790 uprintf(", fault VA = 0x%lx", (u_long)eva);
797 mtx_assert(&Giant, MA_NOTOWNED);
798 KASSERT(PCB_USER_FPU(td->td_pcb),
799 ("Return from trap with kernel FPU ctx leaked"));
806 trap_pfault(frame, usermode, eva)
807 struct trapframe *frame;
812 struct vmspace *vm = NULL;
816 struct thread *td = curthread;
817 struct proc *p = td->td_proc;
819 va = trunc_page(eva);
820 if (va >= KERNBASE) {
822 * Don't allow user-mode faults in kernel address space.
823 * An exception: if the faulting address is the invalid
824 * instruction entry in the IDT, then the Intel Pentium
825 * F00F bug workaround was triggered, and we need to
826 * treat it is as an illegal instruction, and not a page
829 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
830 if ((eva == (unsigned int)&idt[6]) && has_f00f_bug)
839 * This is a fault on non-kernel virtual memory.
840 * vm is initialized above to NULL. If curproc is NULL
841 * or curproc->p_vmspace is NULL the fault is fatal.
853 * PGEX_I is defined only if the execute disable bit capability is
854 * supported and enabled.
856 if (frame->tf_err & PGEX_W)
857 ftype = VM_PROT_WRITE;
859 else if ((frame->tf_err & PGEX_I) && pg_nx != 0)
860 ftype = VM_PROT_EXECUTE;
863 ftype = VM_PROT_READ;
865 if (map != kernel_map) {
867 * Keep swapout from messing with us during this
874 /* Fault in the user page: */
875 rv = vm_fault(map, va, ftype,
876 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY
884 * Don't have to worry about process locking or stacks in the
887 rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
889 if (rv == KERN_SUCCESS)
893 if (td->td_intr_nesting_level == 0 &&
894 PCPU_GET(curpcb)->pcb_onfault != NULL) {
895 frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault;
898 trap_fatal(frame, eva);
902 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV);
906 trap_fatal(frame, eva)
907 struct trapframe *frame;
912 struct soft_segment_descriptor softseg;
915 code = frame->tf_err;
916 type = frame->tf_trapno;
917 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg);
919 if (type <= MAX_TRAP_MSG)
920 msg = trap_msg[type];
923 printf("\n\nFatal trap %d: %s while in %s mode\n", type, msg,
924 frame->tf_eflags & PSL_VM ? "vm86" :
925 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
927 /* two separate prints in case of a trap on an unmapped page */
928 printf("cpuid = %d; ", PCPU_GET(cpuid));
929 printf("apic id = %02x\n", PCPU_GET(apic_id));
931 if (type == T_PAGEFLT) {
932 printf("fault virtual address = 0x%x\n", eva);
933 printf("fault code = %s %s, %s\n",
934 code & PGEX_U ? "user" : "supervisor",
935 code & PGEX_W ? "write" : "read",
936 code & PGEX_P ? "protection violation" : "page not present");
938 printf("instruction pointer = 0x%x:0x%x\n",
939 frame->tf_cs & 0xffff, frame->tf_eip);
940 if ((ISPL(frame->tf_cs) == SEL_UPL) || (frame->tf_eflags & PSL_VM)) {
941 ss = frame->tf_ss & 0xffff;
944 ss = GSEL(GDATA_SEL, SEL_KPL);
945 esp = (int)&frame->tf_esp;
947 printf("stack pointer = 0x%x:0x%x\n", ss, esp);
948 printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp);
949 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n",
950 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type);
951 printf(" = DPL %d, pres %d, def32 %d, gran %d\n",
952 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32,
954 printf("processor eflags = ");
955 if (frame->tf_eflags & PSL_T)
956 printf("trace trap, ");
957 if (frame->tf_eflags & PSL_I)
958 printf("interrupt enabled, ");
959 if (frame->tf_eflags & PSL_NT)
960 printf("nested task, ");
961 if (frame->tf_eflags & PSL_RF)
963 if (frame->tf_eflags & PSL_VM)
965 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12);
966 printf("current process = ");
968 printf("%lu (%s)\n", (u_long)curproc->p_pid, curthread->td_name);
974 if (debugger_on_panic || kdb_active) {
975 frame->tf_err = eva; /* smuggle fault address to ddb */
976 if (kdb_trap(type, 0, frame)) {
977 frame->tf_err = code; /* restore error code */
980 frame->tf_err = code; /* restore error code */
983 printf("trap number = %d\n", type);
984 if (type <= MAX_TRAP_MSG)
985 panic("%s", trap_msg[type]);
987 panic("unknown/reserved trap");
991 * Double fault handler. Called when a fault occurs while writing
992 * a frame for a trap/exception onto the stack. This usually occurs
993 * when the stack overflows (such is the case with infinite recursion,
996 * XXX Note that the current PTD gets replaced by IdlePTD when the
997 * task switch occurs. This means that the stack that was active at
998 * the time of the double fault is not available at <kstack> unless
999 * the machine was idle when the double fault occurred. The downside
1000 * of this is that "trace <ebp>" in ddb won't work.
1005 #ifdef KDTRACE_HOOKS
1006 if (dtrace_doubletrap_func != NULL)
1007 (*dtrace_doubletrap_func)();
1009 printf("\nFatal double fault:\n");
1010 printf("eip = 0x%x\n", PCPU_GET(common_tss.tss_eip));
1011 printf("esp = 0x%x\n", PCPU_GET(common_tss.tss_esp));
1012 printf("ebp = 0x%x\n", PCPU_GET(common_tss.tss_ebp));
1014 /* two separate prints in case of a trap on an unmapped page */
1015 printf("cpuid = %d; ", PCPU_GET(cpuid));
1016 printf("apic id = %02x\n", PCPU_GET(apic_id));
1018 panic("double fault");
1022 cpu_fetch_syscall_args(struct thread *td, struct syscall_args *sa)
1025 struct trapframe *frame;
1030 frame = td->td_frame;
1032 params = (caddr_t)frame->tf_esp + sizeof(int);
1033 sa->code = frame->tf_eax;
1036 * Need to check if this is a 32 bit or 64 bit syscall.
1038 if (sa->code == SYS_syscall) {
1040 * Code is first argument, followed by actual args.
1042 sa->code = fuword(params);
1043 params += sizeof(int);
1044 } else if (sa->code == SYS___syscall) {
1046 * Like syscall, but code is a quad, so as to maintain
1047 * quad alignment for the rest of the arguments.
1049 sa->code = fuword(params);
1050 params += sizeof(quad_t);
1053 if (p->p_sysent->sv_mask)
1054 sa->code &= p->p_sysent->sv_mask;
1055 if (sa->code >= p->p_sysent->sv_size)
1056 sa->callp = &p->p_sysent->sv_table[0];
1058 sa->callp = &p->p_sysent->sv_table[sa->code];
1059 sa->narg = sa->callp->sy_narg;
1061 if (params != NULL && sa->narg != 0)
1062 error = copyin(params, (caddr_t)sa->args,
1063 (u_int)(sa->narg * sizeof(int)));
1068 td->td_retval[0] = 0;
1069 td->td_retval[1] = frame->tf_edx;
1076 * syscall - system call request C handler
1078 * A system call is essentially treated as a trap.
1081 syscall(struct trapframe *frame)
1084 struct syscall_args sa;
1085 register_t orig_tf_eflags;
1090 if (ISPL(frame->tf_cs) != SEL_UPL) {
1095 orig_tf_eflags = frame->tf_eflags;
1098 td->td_frame = frame;
1100 error = syscallenter(td, &sa);
1105 if ((orig_tf_eflags & PSL_T) && !(orig_tf_eflags & PSL_VM)) {
1106 frame->tf_eflags &= ~PSL_T;
1107 ksiginfo_init_trap(&ksi);
1108 ksi.ksi_signo = SIGTRAP;
1109 ksi.ksi_code = TRAP_TRACE;
1110 ksi.ksi_addr = (void *)frame->tf_eip;
1111 trapsignal(td, &ksi);
1114 KASSERT(PCB_USER_FPU(td->td_pcb),
1115 ("System call %s returning with kernel FPU ctx leaked",
1116 syscallname(td->td_proc, sa.code)));
1117 KASSERT(td->td_pcb->pcb_save == &td->td_pcb->pcb_user_save,
1118 ("System call %s returning with mangled pcb_save",
1119 syscallname(td->td_proc, sa.code)));
1121 syscallret(td, error, &sa);