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"
56 #include <sys/param.h>
58 #include <sys/systm.h>
60 #include <sys/pioctl.h>
61 #include <sys/ptrace.h>
63 #include <sys/kernel.h>
66 #include <sys/mutex.h>
67 #include <sys/resourcevar.h>
68 #include <sys/signalvar.h>
69 #include <sys/syscall.h>
70 #include <sys/sysctl.h>
71 #include <sys/sysent.h>
73 #include <sys/vmmeter.h>
75 #include <sys/pmckern.h>
77 #include <security/audit/audit.h>
80 #include <vm/vm_param.h>
82 #include <vm/vm_kern.h>
83 #include <vm/vm_map.h>
84 #include <vm/vm_page.h>
85 #include <vm/vm_extern.h>
87 #include <machine/cpu.h>
88 #include <machine/intr_machdep.h>
90 #include <machine/md_var.h>
91 #include <machine/pcb.h>
93 #include <machine/smp.h>
95 #include <machine/tss.h>
96 #include <machine/vm86.h>
99 #include <sys/syslog.h>
100 #include <machine/clock.h>
104 #include <sys/dtrace_bsd.h>
107 * This is a hook which is initialised by the dtrace module
108 * to handle traps which might occur during DTrace probe
111 dtrace_trap_func_t dtrace_trap_func;
113 dtrace_doubletrap_func_t dtrace_doubletrap_func;
116 * This is a hook which is initialised by the systrace module
117 * when it is loaded. This keeps the DTrace syscall provider
118 * implementation opaque.
120 systrace_probe_func_t systrace_probe_func;
123 * These hooks are necessary for the pid, usdt and fasttrap providers.
125 dtrace_fasttrap_probe_ptr_t dtrace_fasttrap_probe_ptr;
126 dtrace_pid_probe_ptr_t dtrace_pid_probe_ptr;
127 dtrace_return_probe_ptr_t dtrace_return_probe_ptr;
130 extern void trap(struct trapframe *frame);
131 extern void syscall(struct trapframe *frame);
133 static int trap_pfault(struct trapframe *, int, vm_offset_t);
134 static void trap_fatal(struct trapframe *, vm_offset_t);
135 void dblfault_handler(void);
137 extern inthand_t IDTVEC(lcall_syscall);
139 #define MAX_TRAP_MSG 33
140 static char *trap_msg[] = {
142 "privileged instruction fault", /* 1 T_PRIVINFLT */
144 "breakpoint instruction fault", /* 3 T_BPTFLT */
147 "arithmetic trap", /* 6 T_ARITHTRAP */
150 "general protection fault", /* 9 T_PROTFLT */
151 "trace trap", /* 10 T_TRCTRAP */
153 "page fault", /* 12 T_PAGEFLT */
155 "alignment fault", /* 14 T_ALIGNFLT */
159 "integer divide fault", /* 18 T_DIVIDE */
160 "non-maskable interrupt trap", /* 19 T_NMI */
161 "overflow trap", /* 20 T_OFLOW */
162 "FPU bounds check fault", /* 21 T_BOUND */
163 "FPU device not available", /* 22 T_DNA */
164 "double fault", /* 23 T_DOUBLEFLT */
165 "FPU operand fetch fault", /* 24 T_FPOPFLT */
166 "invalid TSS fault", /* 25 T_TSSFLT */
167 "segment not present fault", /* 26 T_SEGNPFLT */
168 "stack fault", /* 27 T_STKFLT */
169 "machine check trap", /* 28 T_MCHK */
170 "SIMD floating-point exception", /* 29 T_XMMFLT */
171 "reserved (unknown) fault", /* 30 T_RESERVED */
172 "", /* 31 unused (reserved) */
173 "DTrace pid return trap", /* 32 T_DTRACE_RET */
174 "DTrace fasttrap probe trap", /* 33 T_DTRACE_PROBE */
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");
186 TUNABLE_INT("machdep.kdb_on_nmi", &kdb_on_nmi);
188 static int panic_on_nmi = 1;
189 SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
190 &panic_on_nmi, 0, "Panic on NMI");
191 TUNABLE_INT("machdep.panic_on_nmi", &panic_on_nmi);
192 static int prot_fault_translation = 0;
193 SYSCTL_INT(_machdep, OID_AUTO, prot_fault_translation, CTLFLAG_RW,
194 &prot_fault_translation, 0, "Select signal to deliver on protection fault");
197 * Exception, fault, and trap interface to the FreeBSD kernel.
198 * This common code is called from assembly language IDT gate entry
199 * routines that prepare a suitable stack frame, and restore this
200 * frame after the exception has been processed.
204 trap(struct trapframe *frame)
206 struct thread *td = curthread;
207 struct proc *p = td->td_proc;
208 int i = 0, ucode = 0, code;
214 static int lastalert = 0;
217 PCPU_INC(cnt.v_trap);
218 type = frame->tf_trapno;
221 /* Handler for NMI IPIs used for stopping CPUs. */
223 if (ipi_nmi_handler() == 0)
235 if (type == T_RESERVED) {
236 trap_fatal(frame, 0);
242 * CPU PMCs interrupt using an NMI so we check for that first.
243 * If the HWPMC module is active, 'pmc_hook' will point to
244 * the function to be called. A return value of '1' from the
245 * hook means that the NMI was handled by it and that we can
246 * return immediately.
248 if (type == T_NMI && pmc_intr &&
249 (*pmc_intr)(PCPU_GET(cpuid), frame))
253 if (type == T_MCHK) {
255 trap_fatal(frame, 0);
261 * A trap can occur while DTrace executes a probe. Before
262 * executing the probe, DTrace blocks re-scheduling and sets
263 * a flag in it's per-cpu flags to indicate that it doesn't
264 * want to fault. On returning from the probe, the no-fault
265 * flag is cleared and finally re-scheduling is enabled.
267 * If the DTrace kernel module has registered a trap handler,
268 * call it and if it returns non-zero, assume that it has
269 * handled the trap and modified the trap frame so that this
270 * function can return normally.
272 if (type == T_DTRACE_PROBE || type == T_DTRACE_RET ||
276 fill_frame_regs(frame, ®s);
277 if (type == T_DTRACE_PROBE &&
278 dtrace_fasttrap_probe_ptr != NULL &&
279 dtrace_fasttrap_probe_ptr(®s) == 0)
281 if (type == T_BPTFLT &&
282 dtrace_pid_probe_ptr != NULL &&
283 dtrace_pid_probe_ptr(®s) == 0)
285 if (type == T_DTRACE_RET &&
286 dtrace_return_probe_ptr != NULL &&
287 dtrace_return_probe_ptr(®s) == 0)
290 if ((type == T_PROTFLT || type == T_PAGEFLT) &&
291 dtrace_trap_func != NULL && (*dtrace_trap_func)(frame, type))
295 if ((frame->tf_eflags & PSL_I) == 0) {
297 * Buggy application or kernel code has disabled
298 * interrupts and then trapped. Enabling interrupts
299 * now is wrong, but it is better than running with
300 * interrupts disabled until they are accidentally
303 if (ISPL(frame->tf_cs) == SEL_UPL || (frame->tf_eflags & PSL_VM))
305 "pid %ld (%s): trap %d with interrupts disabled\n",
306 (long)curproc->p_pid, curthread->td_name, type);
307 else if (type != T_BPTFLT && type != T_TRCTRAP &&
308 frame->tf_eip != (int)cpu_switch_load_gs) {
310 * XXX not quite right, since this may be for a
311 * multiple fault in user mode.
313 printf("kernel trap %d with interrupts disabled\n",
316 * Page faults need interrupts disabled until later,
317 * and we shouldn't enable interrupts while holding
318 * a spin lock or if servicing an NMI.
320 if (type != T_NMI && type != T_PAGEFLT &&
321 td->td_md.md_spinlock_count == 0)
326 code = frame->tf_err;
327 if (type == T_PAGEFLT) {
329 * For some Cyrix CPUs, %cr2 is clobbered by
330 * interrupts. This problem is worked around by using
331 * an interrupt gate for the pagefault handler. We
332 * are finally ready to read %cr2 and then must
333 * reenable interrupts.
335 * If we get a page fault while in a critical section, then
336 * it is most likely a fatal kernel page fault. The kernel
337 * is already going to panic trying to get a sleep lock to
338 * do the VM lookup, so just consider it a fatal trap so the
339 * kernel can print out a useful trap message and even get
342 * If we get a page fault while holding a non-sleepable
343 * lock, then it is most likely a fatal kernel page fault.
344 * If WITNESS is enabled, then it's going to whine about
345 * bogus LORs with various VM locks, so just skip to the
346 * fatal trap handling directly.
349 if (td->td_critnest != 0 ||
350 WITNESS_CHECK(WARN_SLEEPOK | WARN_GIANTOK, NULL,
351 "Kernel page fault") != 0)
352 trap_fatal(frame, eva);
357 if ((ISPL(frame->tf_cs) == SEL_UPL) ||
358 ((frame->tf_eflags & PSL_VM) &&
359 !(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL))) {
363 td->td_frame = frame;
364 addr = frame->tf_eip;
365 if (td->td_ucred != p->p_ucred)
366 cred_update_thread(td);
369 case T_PRIVINFLT: /* privileged instruction fault */
374 case T_BPTFLT: /* bpt instruction fault */
375 case T_TRCTRAP: /* trace trap */
377 frame->tf_eflags &= ~PSL_T;
379 ucode = (type == T_TRCTRAP ? TRAP_TRACE : TRAP_BRKPT);
382 case T_ARITHTRAP: /* arithmetic trap */
394 * The following two traps can happen in
395 * vm86 mode, and, if so, we want to handle
398 case T_PROTFLT: /* general protection fault */
399 case T_STKFLT: /* stack fault */
400 if (frame->tf_eflags & PSL_VM) {
401 i = vm86_emulate((struct vm86frame *)frame);
407 ucode = (type == T_PROTFLT) ? BUS_OBJERR : BUS_ADRERR;
409 case T_SEGNPFLT: /* segment not present fault */
413 case T_TSSFLT: /* invalid TSS fault */
417 case T_DOUBLEFLT: /* double fault */
423 case T_PAGEFLT: /* page fault */
425 i = trap_pfault(frame, TRUE, eva);
426 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
429 * The f00f hack workaround has triggered, so
430 * treat the fault as an illegal instruction
431 * (T_PRIVINFLT) instead of a page fault.
433 type = frame->tf_trapno = T_PRIVINFLT;
435 /* Proceed as in that case. */
449 if (prot_fault_translation == 0) {
452 * This check also covers the images
453 * without the ABI-tag ELF note.
455 if (SV_CURPROC_ABI() == SV_ABI_FREEBSD
456 && p->p_osrel >= P_OSREL_SIGSEGV) {
461 ucode = BUS_PAGE_FAULT;
463 } else if (prot_fault_translation == 1) {
465 * Always compat mode.
468 ucode = BUS_PAGE_FAULT;
471 * Always SIGSEGV mode.
480 case T_DIVIDE: /* integer divide fault */
489 # define TIMER_FREQ 1193182
491 if (time_second - lastalert > 10) {
492 log(LOG_WARNING, "NMI: power fail\n");
494 lastalert = time_second;
497 #else /* !POWERFAIL_NMI */
498 /* machine/parity/power fail/"kitchen sink" faults */
499 if (isa_nmi(code) == 0) {
502 * NMI can be hooked up to a pushbutton
506 printf ("NMI ... going to debugger\n");
507 kdb_trap(type, 0, frame);
511 } else if (panic_on_nmi)
512 panic("NMI indicates hardware failure");
514 #endif /* POWERFAIL_NMI */
517 case T_OFLOW: /* integer overflow fault */
522 case T_BOUND: /* bounds check fault */
529 KASSERT(PCB_USER_FPU(td->td_pcb),
530 ("kernel FPU ctx has leaked"));
531 /* transparent fault (due to context switch "late") */
535 uprintf("pid %d killed due to lack of floating point\n",
541 case T_FPOPFLT: /* FPU operand fetch fault */
546 case T_XMMFLT: /* SIMD floating-point exception */
554 KASSERT(cold || td->td_ucred != NULL,
555 ("kernel trap doesn't have ucred"));
557 case T_PAGEFLT: /* page fault */
558 (void) trap_pfault(frame, FALSE, eva);
563 KASSERT(!PCB_USER_FPU(td->td_pcb),
564 ("Unregistered use of FPU in kernel"));
570 case T_ARITHTRAP: /* arithmetic trap */
571 case T_XMMFLT: /* SIMD floating-point exception */
572 case T_FPOPFLT: /* FPU operand fetch fault */
574 * XXXKIB for now disable any FPU traps in kernel
575 * handler registration seems to be overkill
577 trap_fatal(frame, 0);
581 * The following two traps can happen in
582 * vm86 mode, and, if so, we want to handle
585 case T_PROTFLT: /* general protection fault */
586 case T_STKFLT: /* stack fault */
587 if (frame->tf_eflags & PSL_VM) {
588 i = vm86_emulate((struct vm86frame *)frame);
591 * returns to original process
593 vm86_trap((struct vm86frame *)frame);
596 if (type == T_STKFLT)
601 case T_SEGNPFLT: /* segment not present fault */
602 if (PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)
606 * Invalid %fs's and %gs's can be created using
607 * procfs or PT_SETREGS or by invalidating the
608 * underlying LDT entry. This causes a fault
609 * in kernel mode when the kernel attempts to
610 * switch contexts. Lose the bad context
611 * (XXX) so that we can continue, and generate
614 if (frame->tf_eip == (int)cpu_switch_load_gs) {
615 PCPU_GET(curpcb)->pcb_gs = 0;
618 kern_psignal(p, SIGBUS);
624 if (td->td_intr_nesting_level != 0)
628 * Invalid segment selectors and out of bounds
629 * %eip's and %esp's can be set up in user mode.
630 * This causes a fault in kernel mode when the
631 * kernel tries to return to user mode. We want
632 * to get this fault so that we can fix the
633 * problem here and not have to check all the
634 * selectors and pointers when the user changes
637 if (frame->tf_eip == (int)doreti_iret) {
638 frame->tf_eip = (int)doreti_iret_fault;
641 if (frame->tf_eip == (int)doreti_popl_ds) {
642 frame->tf_eip = (int)doreti_popl_ds_fault;
645 if (frame->tf_eip == (int)doreti_popl_es) {
646 frame->tf_eip = (int)doreti_popl_es_fault;
649 if (frame->tf_eip == (int)doreti_popl_fs) {
650 frame->tf_eip = (int)doreti_popl_fs_fault;
653 if (PCPU_GET(curpcb)->pcb_onfault != NULL) {
655 (int)PCPU_GET(curpcb)->pcb_onfault;
662 * PSL_NT can be set in user mode and isn't cleared
663 * automatically when the kernel is entered. This
664 * causes a TSS fault when the kernel attempts to
665 * `iret' because the TSS link is uninitialized. We
666 * want to get this fault so that we can fix the
667 * problem here and not every time the kernel is
670 if (frame->tf_eflags & PSL_NT) {
671 frame->tf_eflags &= ~PSL_NT;
676 case T_TRCTRAP: /* trace trap */
677 if (frame->tf_eip == (int)IDTVEC(lcall_syscall)) {
679 * We've just entered system mode via the
680 * syscall lcall. Continue single stepping
681 * silently until the syscall handler has
686 if (frame->tf_eip == (int)IDTVEC(lcall_syscall) + 1) {
688 * The syscall handler has now saved the
689 * flags. Stop single stepping it.
691 frame->tf_eflags &= ~PSL_T;
695 * Ignore debug register trace traps due to
696 * accesses in the user's address space, which
697 * can happen under several conditions such as
698 * if a user sets a watchpoint on a buffer and
699 * then passes that buffer to a system call.
700 * We still want to get TRCTRAPS for addresses
701 * in kernel space because that is useful when
702 * debugging the kernel.
704 if (user_dbreg_trap() &&
705 !(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)) {
707 * Reset breakpoint bits because the
710 load_dr6(rdr6() & 0xfffffff0);
714 * FALLTHROUGH (TRCTRAP kernel mode, kernel address)
718 * If KDB is enabled, let it handle the debugger trap.
719 * Otherwise, debugger traps "can't happen".
722 if (kdb_trap(type, 0, frame))
730 if (time_second - lastalert > 10) {
731 log(LOG_WARNING, "NMI: power fail\n");
733 lastalert = time_second;
736 #else /* !POWERFAIL_NMI */
737 /* machine/parity/power fail/"kitchen sink" faults */
738 if (isa_nmi(code) == 0) {
741 * NMI can be hooked up to a pushbutton
745 printf ("NMI ... going to debugger\n");
746 kdb_trap(type, 0, frame);
750 } else if (panic_on_nmi == 0)
753 #endif /* POWERFAIL_NMI */
757 trap_fatal(frame, eva);
761 /* Translate fault for emulators (e.g. Linux) */
762 if (*p->p_sysent->sv_transtrap)
763 i = (*p->p_sysent->sv_transtrap)(i, type);
765 ksiginfo_init_trap(&ksi);
767 ksi.ksi_code = ucode;
768 ksi.ksi_addr = (void *)addr;
769 ksi.ksi_trapno = type;
770 trapsignal(td, &ksi);
773 if (type <= MAX_TRAP_MSG) {
774 uprintf("fatal process exception: %s",
776 if ((type == T_PAGEFLT) || (type == T_PROTFLT))
777 uprintf(", fault VA = 0x%lx", (u_long)eva);
784 mtx_assert(&Giant, MA_NOTOWNED);
785 KASSERT(PCB_USER_FPU(td->td_pcb),
786 ("Return from trap with kernel FPU ctx leaked"));
793 trap_pfault(frame, usermode, eva)
794 struct trapframe *frame;
799 struct vmspace *vm = NULL;
803 struct thread *td = curthread;
804 struct proc *p = td->td_proc;
806 va = trunc_page(eva);
807 if (va >= KERNBASE) {
809 * Don't allow user-mode faults in kernel address space.
810 * An exception: if the faulting address is the invalid
811 * instruction entry in the IDT, then the Intel Pentium
812 * F00F bug workaround was triggered, and we need to
813 * treat it is as an illegal instruction, and not a page
816 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
817 if ((eva == (unsigned int)&idt[6]) && has_f00f_bug)
826 * This is a fault on non-kernel virtual memory.
827 * vm is initialized above to NULL. If curproc is NULL
828 * or curproc->p_vmspace is NULL the fault is fatal.
837 if (!usermode && (td->td_intr_nesting_level != 0 ||
838 PCPU_GET(curpcb)->pcb_onfault == NULL)) {
839 trap_fatal(frame, eva);
845 * PGEX_I is defined only if the execute disable bit capability is
846 * supported and enabled.
848 if (frame->tf_err & PGEX_W)
849 ftype = VM_PROT_WRITE;
851 else if ((frame->tf_err & PGEX_I) && pg_nx != 0)
852 ftype = VM_PROT_EXECUTE;
855 ftype = VM_PROT_READ;
857 if (map != kernel_map) {
859 * Keep swapout from messing with us during this
866 /* Fault in the user page: */
867 rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
874 * Don't have to worry about process locking or stacks in the
877 rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
879 if (rv == KERN_SUCCESS)
883 if (td->td_intr_nesting_level == 0 &&
884 PCPU_GET(curpcb)->pcb_onfault != NULL) {
885 frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault;
888 trap_fatal(frame, eva);
892 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV);
896 trap_fatal(frame, eva)
897 struct trapframe *frame;
902 struct soft_segment_descriptor softseg;
905 code = frame->tf_err;
906 type = frame->tf_trapno;
907 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg);
909 if (type <= MAX_TRAP_MSG)
910 msg = trap_msg[type];
913 printf("\n\nFatal trap %d: %s while in %s mode\n", type, msg,
914 frame->tf_eflags & PSL_VM ? "vm86" :
915 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
917 /* two separate prints in case of a trap on an unmapped page */
918 printf("cpuid = %d; ", PCPU_GET(cpuid));
919 printf("apic id = %02x\n", PCPU_GET(apic_id));
921 if (type == T_PAGEFLT) {
922 printf("fault virtual address = 0x%x\n", eva);
923 printf("fault code = %s %s, %s\n",
924 code & PGEX_U ? "user" : "supervisor",
925 code & PGEX_W ? "write" : "read",
926 code & PGEX_P ? "protection violation" : "page not present");
928 printf("instruction pointer = 0x%x:0x%x\n",
929 frame->tf_cs & 0xffff, frame->tf_eip);
930 if ((ISPL(frame->tf_cs) == SEL_UPL) || (frame->tf_eflags & PSL_VM)) {
931 ss = frame->tf_ss & 0xffff;
934 ss = GSEL(GDATA_SEL, SEL_KPL);
935 esp = (int)&frame->tf_esp;
937 printf("stack pointer = 0x%x:0x%x\n", ss, esp);
938 printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp);
939 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n",
940 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type);
941 printf(" = DPL %d, pres %d, def32 %d, gran %d\n",
942 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32,
944 printf("processor eflags = ");
945 if (frame->tf_eflags & PSL_T)
946 printf("trace trap, ");
947 if (frame->tf_eflags & PSL_I)
948 printf("interrupt enabled, ");
949 if (frame->tf_eflags & PSL_NT)
950 printf("nested task, ");
951 if (frame->tf_eflags & PSL_RF)
953 if (frame->tf_eflags & PSL_VM)
955 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12);
956 printf("current process = ");
958 printf("%lu (%s)\n", (u_long)curproc->p_pid, curthread->td_name);
964 if (debugger_on_panic || kdb_active) {
965 frame->tf_err = eva; /* smuggle fault address to ddb */
966 if (kdb_trap(type, 0, frame)) {
967 frame->tf_err = code; /* restore error code */
970 frame->tf_err = code; /* restore error code */
973 printf("trap number = %d\n", type);
974 if (type <= MAX_TRAP_MSG)
975 panic("%s", trap_msg[type]);
977 panic("unknown/reserved trap");
981 * Double fault handler. Called when a fault occurs while writing
982 * a frame for a trap/exception onto the stack. This usually occurs
983 * when the stack overflows (such is the case with infinite recursion,
986 * XXX Note that the current PTD gets replaced by IdlePTD when the
987 * task switch occurs. This means that the stack that was active at
988 * the time of the double fault is not available at <kstack> unless
989 * the machine was idle when the double fault occurred. The downside
990 * of this is that "trace <ebp>" in ddb won't work.
996 if (dtrace_doubletrap_func != NULL)
997 (*dtrace_doubletrap_func)();
999 printf("\nFatal double fault:\n");
1000 printf("eip = 0x%x\n", PCPU_GET(common_tss.tss_eip));
1001 printf("esp = 0x%x\n", PCPU_GET(common_tss.tss_esp));
1002 printf("ebp = 0x%x\n", PCPU_GET(common_tss.tss_ebp));
1004 /* two separate prints in case of a trap on an unmapped page */
1005 printf("cpuid = %d; ", PCPU_GET(cpuid));
1006 printf("apic id = %02x\n", PCPU_GET(apic_id));
1008 panic("double fault");
1012 cpu_fetch_syscall_args(struct thread *td, struct syscall_args *sa)
1015 struct trapframe *frame;
1020 frame = td->td_frame;
1022 params = (caddr_t)frame->tf_esp + sizeof(int);
1023 sa->code = frame->tf_eax;
1026 * Need to check if this is a 32 bit or 64 bit syscall.
1028 if (sa->code == SYS_syscall) {
1030 * Code is first argument, followed by actual args.
1032 sa->code = fuword(params);
1033 params += sizeof(int);
1034 } else if (sa->code == SYS___syscall) {
1036 * Like syscall, but code is a quad, so as to maintain
1037 * quad alignment for the rest of the arguments.
1039 sa->code = fuword(params);
1040 params += sizeof(quad_t);
1043 if (p->p_sysent->sv_mask)
1044 sa->code &= p->p_sysent->sv_mask;
1045 if (sa->code >= p->p_sysent->sv_size)
1046 sa->callp = &p->p_sysent->sv_table[0];
1048 sa->callp = &p->p_sysent->sv_table[sa->code];
1049 sa->narg = sa->callp->sy_narg;
1051 if (params != NULL && sa->narg != 0)
1052 error = copyin(params, (caddr_t)sa->args,
1053 (u_int)(sa->narg * sizeof(int)));
1058 td->td_retval[0] = 0;
1059 td->td_retval[1] = frame->tf_edx;
1065 #include "../../kern/subr_syscall.c"
1068 * syscall - system call request C handler
1070 * A system call is essentially treated as a trap.
1073 syscall(struct trapframe *frame)
1076 struct syscall_args sa;
1077 register_t orig_tf_eflags;
1082 if (ISPL(frame->tf_cs) != SEL_UPL) {
1087 orig_tf_eflags = frame->tf_eflags;
1090 td->td_frame = frame;
1092 error = syscallenter(td, &sa);
1097 if ((orig_tf_eflags & PSL_T) && !(orig_tf_eflags & PSL_VM)) {
1098 frame->tf_eflags &= ~PSL_T;
1099 ksiginfo_init_trap(&ksi);
1100 ksi.ksi_signo = SIGTRAP;
1101 ksi.ksi_code = TRAP_TRACE;
1102 ksi.ksi_addr = (void *)frame->tf_eip;
1103 trapsignal(td, &ksi);
1106 KASSERT(PCB_USER_FPU(td->td_pcb),
1107 ("System call %s returning with kernel FPU ctx leaked",
1108 syscallname(td->td_proc, sa.code)));
1109 KASSERT(td->td_pcb->pcb_save == &td->td_pcb->pcb_user_save,
1110 ("System call %s returning with mangled pcb_save",
1111 syscallname(td->td_proc, sa.code)));
1113 syscallret(td, error, &sa);