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 extern void trap(struct trapframe *frame);
128 extern void syscall(struct trapframe *frame);
130 static int trap_pfault(struct trapframe *, int, vm_offset_t);
131 static void trap_fatal(struct trapframe *, vm_offset_t);
132 void dblfault_handler(void);
134 extern inthand_t IDTVEC(lcall_syscall);
136 #define MAX_TRAP_MSG 30
137 static char *trap_msg[] = {
139 "privileged instruction fault", /* 1 T_PRIVINFLT */
141 "breakpoint instruction fault", /* 3 T_BPTFLT */
144 "arithmetic trap", /* 6 T_ARITHTRAP */
147 "general protection fault", /* 9 T_PROTFLT */
148 "trace trap", /* 10 T_TRCTRAP */
150 "page fault", /* 12 T_PAGEFLT */
152 "alignment fault", /* 14 T_ALIGNFLT */
156 "integer divide fault", /* 18 T_DIVIDE */
157 "non-maskable interrupt trap", /* 19 T_NMI */
158 "overflow trap", /* 20 T_OFLOW */
159 "FPU bounds check fault", /* 21 T_BOUND */
160 "FPU device not available", /* 22 T_DNA */
161 "double fault", /* 23 T_DOUBLEFLT */
162 "FPU operand fetch fault", /* 24 T_FPOPFLT */
163 "invalid TSS fault", /* 25 T_TSSFLT */
164 "segment not present fault", /* 26 T_SEGNPFLT */
165 "stack fault", /* 27 T_STKFLT */
166 "machine check trap", /* 28 T_MCHK */
167 "SIMD floating-point exception", /* 29 T_XMMFLT */
168 "reserved (unknown) fault", /* 30 T_RESERVED */
171 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
172 extern int has_f00f_bug;
176 static int kdb_on_nmi = 1;
177 SYSCTL_INT(_machdep, OID_AUTO, kdb_on_nmi, CTLFLAG_RW,
178 &kdb_on_nmi, 0, "Go to KDB on NMI");
180 static int panic_on_nmi = 1;
181 SYSCTL_INT(_machdep, OID_AUTO, panic_on_nmi, CTLFLAG_RW,
182 &panic_on_nmi, 0, "Panic on NMI");
183 static int prot_fault_translation = 0;
184 SYSCTL_INT(_machdep, OID_AUTO, prot_fault_translation, CTLFLAG_RW,
185 &prot_fault_translation, 0, "Select signal to deliver on protection fault");
188 * Exception, fault, and trap interface to the FreeBSD kernel.
189 * This common code is called from assembly language IDT gate entry
190 * routines that prepare a suitable stack frame, and restore this
191 * frame after the exception has been processed.
195 trap(struct trapframe *frame)
197 struct thread *td = curthread;
198 struct proc *p = td->td_proc;
199 int i = 0, ucode = 0, code;
205 static int lastalert = 0;
208 PCPU_INC(cnt.v_trap);
209 type = frame->tf_trapno;
212 /* Handler for NMI IPIs used for stopping CPUs. */
214 if (ipi_nmi_handler() == 0)
226 if (type == T_RESERVED) {
227 trap_fatal(frame, 0);
233 * CPU PMCs interrupt using an NMI so we check for that first.
234 * If the HWPMC module is active, 'pmc_hook' will point to
235 * the function to be called. A return value of '1' from the
236 * hook means that the NMI was handled by it and that we can
237 * return immediately.
239 if (type == T_NMI && pmc_intr &&
240 (*pmc_intr)(PCPU_GET(cpuid), frame))
244 if (type == T_MCHK) {
246 trap_fatal(frame, 0);
252 * A trap can occur while DTrace executes a probe. Before
253 * executing the probe, DTrace blocks re-scheduling and sets
254 * a flag in it's per-cpu flags to indicate that it doesn't
255 * want to fault. On returning from the the probe, the no-fault
256 * flag is cleared and finally re-scheduling is enabled.
258 * If the DTrace kernel module has registered a trap handler,
259 * call it and if it returns non-zero, assume that it has
260 * handled the trap and modified the trap frame so that this
261 * function can return normally.
263 if ((type == T_PROTFLT || type == T_PAGEFLT) &&
264 dtrace_trap_func != NULL)
265 if ((*dtrace_trap_func)(frame, type))
269 if ((frame->tf_eflags & PSL_I) == 0) {
271 * Buggy application or kernel code has disabled
272 * interrupts and then trapped. Enabling interrupts
273 * now is wrong, but it is better than running with
274 * interrupts disabled until they are accidentally
277 if (ISPL(frame->tf_cs) == SEL_UPL || (frame->tf_eflags & PSL_VM))
279 "pid %ld (%s): trap %d with interrupts disabled\n",
280 (long)curproc->p_pid, curthread->td_name, type);
281 else if (type != T_BPTFLT && type != T_TRCTRAP &&
282 frame->tf_eip != (int)cpu_switch_load_gs) {
284 * XXX not quite right, since this may be for a
285 * multiple fault in user mode.
287 printf("kernel trap %d with interrupts disabled\n",
290 * Page faults need interrupts disabled until later,
291 * and we shouldn't enable interrupts while holding
292 * a spin lock or if servicing an NMI.
294 if (type != T_NMI && type != T_PAGEFLT &&
295 td->td_md.md_spinlock_count == 0)
300 code = frame->tf_err;
301 if (type == T_PAGEFLT) {
303 * For some Cyrix CPUs, %cr2 is clobbered by
304 * interrupts. This problem is worked around by using
305 * an interrupt gate for the pagefault handler. We
306 * are finally ready to read %cr2 and then must
307 * reenable interrupts.
309 * If we get a page fault while in a critical section, then
310 * it is most likely a fatal kernel page fault. The kernel
311 * is already going to panic trying to get a sleep lock to
312 * do the VM lookup, so just consider it a fatal trap so the
313 * kernel can print out a useful trap message and even get
316 * If we get a page fault while holding a non-sleepable
317 * lock, then it is most likely a fatal kernel page fault.
318 * If WITNESS is enabled, then it's going to whine about
319 * bogus LORs with various VM locks, so just skip to the
320 * fatal trap handling directly.
323 if (td->td_critnest != 0 ||
324 WITNESS_CHECK(WARN_SLEEPOK | WARN_GIANTOK, NULL,
325 "Kernel page fault") != 0)
326 trap_fatal(frame, eva);
331 if ((ISPL(frame->tf_cs) == SEL_UPL) ||
332 ((frame->tf_eflags & PSL_VM) &&
333 !(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL))) {
337 td->td_frame = frame;
338 addr = frame->tf_eip;
339 if (td->td_ucred != p->p_ucred)
340 cred_update_thread(td);
343 case T_PRIVINFLT: /* privileged instruction fault */
348 case T_BPTFLT: /* bpt instruction fault */
349 case T_TRCTRAP: /* trace trap */
351 frame->tf_eflags &= ~PSL_T;
353 ucode = (type == T_TRCTRAP ? TRAP_TRACE : TRAP_BRKPT);
356 case T_ARITHTRAP: /* arithmetic trap */
368 * The following two traps can happen in
369 * vm86 mode, and, if so, we want to handle
372 case T_PROTFLT: /* general protection fault */
373 case T_STKFLT: /* stack fault */
374 if (frame->tf_eflags & PSL_VM) {
375 i = vm86_emulate((struct vm86frame *)frame);
381 ucode = (type == T_PROTFLT) ? BUS_OBJERR : BUS_ADRERR;
383 case T_SEGNPFLT: /* segment not present fault */
387 case T_TSSFLT: /* invalid TSS fault */
391 case T_DOUBLEFLT: /* double fault */
397 case T_PAGEFLT: /* page fault */
399 i = trap_pfault(frame, TRUE, eva);
400 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
403 * The f00f hack workaround has triggered, so
404 * treat the fault as an illegal instruction
405 * (T_PRIVINFLT) instead of a page fault.
407 type = frame->tf_trapno = T_PRIVINFLT;
409 /* Proceed as in that case. */
423 if (prot_fault_translation == 0) {
426 * This check also covers the images
427 * without the ABI-tag ELF note.
429 if (SV_CURPROC_ABI() == SV_ABI_FREEBSD
430 && p->p_osrel >= P_OSREL_SIGSEGV) {
435 ucode = BUS_PAGE_FAULT;
437 } else if (prot_fault_translation == 1) {
439 * Always compat mode.
442 ucode = BUS_PAGE_FAULT;
445 * Always SIGSEGV mode.
454 case T_DIVIDE: /* integer divide fault */
463 # define TIMER_FREQ 1193182
465 if (time_second - lastalert > 10) {
466 log(LOG_WARNING, "NMI: power fail\n");
468 lastalert = time_second;
471 #else /* !POWERFAIL_NMI */
472 /* machine/parity/power fail/"kitchen sink" faults */
473 if (isa_nmi(code) == 0) {
476 * NMI can be hooked up to a pushbutton
480 printf ("NMI ... going to debugger\n");
481 kdb_trap(type, 0, frame);
485 } else if (panic_on_nmi)
486 panic("NMI indicates hardware failure");
488 #endif /* POWERFAIL_NMI */
491 case T_OFLOW: /* integer overflow fault */
496 case T_BOUND: /* bounds check fault */
503 KASSERT(PCB_USER_FPU(td->td_pcb),
504 ("kernel FPU ctx has leaked"));
505 /* transparent fault (due to context switch "late") */
509 uprintf("pid %d killed due to lack of floating point\n",
515 case T_FPOPFLT: /* FPU operand fetch fault */
520 case T_XMMFLT: /* SIMD floating-point exception */
528 KASSERT(cold || td->td_ucred != NULL,
529 ("kernel trap doesn't have ucred"));
531 case T_PAGEFLT: /* page fault */
532 (void) trap_pfault(frame, FALSE, eva);
537 KASSERT(!PCB_USER_FPU(td->td_pcb),
538 ("Unregistered use of FPU in kernel"));
544 case T_ARITHTRAP: /* arithmetic trap */
545 case T_XMMFLT: /* SIMD floating-point exception */
546 case T_FPOPFLT: /* FPU operand fetch fault */
548 * XXXKIB for now disable any FPU traps in kernel
549 * handler registration seems to be overkill
551 trap_fatal(frame, 0);
555 * The following two traps can happen in
556 * vm86 mode, and, if so, we want to handle
559 case T_PROTFLT: /* general protection fault */
560 case T_STKFLT: /* stack fault */
561 if (frame->tf_eflags & PSL_VM) {
562 i = vm86_emulate((struct vm86frame *)frame);
565 * returns to original process
567 vm86_trap((struct vm86frame *)frame);
570 if (type == T_STKFLT)
575 case T_SEGNPFLT: /* segment not present fault */
576 if (PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)
580 * Invalid %fs's and %gs's can be created using
581 * procfs or PT_SETREGS or by invalidating the
582 * underlying LDT entry. This causes a fault
583 * in kernel mode when the kernel attempts to
584 * switch contexts. Lose the bad context
585 * (XXX) so that we can continue, and generate
588 if (frame->tf_eip == (int)cpu_switch_load_gs) {
589 PCPU_GET(curpcb)->pcb_gs = 0;
598 if (td->td_intr_nesting_level != 0)
602 * Invalid segment selectors and out of bounds
603 * %eip's and %esp's can be set up in user mode.
604 * This causes a fault in kernel mode when the
605 * kernel tries to return to user mode. We want
606 * to get this fault so that we can fix the
607 * problem here and not have to check all the
608 * selectors and pointers when the user changes
611 if (frame->tf_eip == (int)doreti_iret) {
612 frame->tf_eip = (int)doreti_iret_fault;
615 if (frame->tf_eip == (int)doreti_popl_ds) {
616 frame->tf_eip = (int)doreti_popl_ds_fault;
619 if (frame->tf_eip == (int)doreti_popl_es) {
620 frame->tf_eip = (int)doreti_popl_es_fault;
623 if (frame->tf_eip == (int)doreti_popl_fs) {
624 frame->tf_eip = (int)doreti_popl_fs_fault;
627 if (PCPU_GET(curpcb)->pcb_onfault != NULL) {
629 (int)PCPU_GET(curpcb)->pcb_onfault;
636 * PSL_NT can be set in user mode and isn't cleared
637 * automatically when the kernel is entered. This
638 * causes a TSS fault when the kernel attempts to
639 * `iret' because the TSS link is uninitialized. We
640 * want to get this fault so that we can fix the
641 * problem here and not every time the kernel is
644 if (frame->tf_eflags & PSL_NT) {
645 frame->tf_eflags &= ~PSL_NT;
650 case T_TRCTRAP: /* trace trap */
651 if (frame->tf_eip == (int)IDTVEC(lcall_syscall)) {
653 * We've just entered system mode via the
654 * syscall lcall. Continue single stepping
655 * silently until the syscall handler has
660 if (frame->tf_eip == (int)IDTVEC(lcall_syscall) + 1) {
662 * The syscall handler has now saved the
663 * flags. Stop single stepping it.
665 frame->tf_eflags &= ~PSL_T;
669 * Ignore debug register trace traps due to
670 * accesses in the user's address space, which
671 * can happen under several conditions such as
672 * if a user sets a watchpoint on a buffer and
673 * then passes that buffer to a system call.
674 * We still want to get TRCTRAPS for addresses
675 * in kernel space because that is useful when
676 * debugging the kernel.
678 if (user_dbreg_trap() &&
679 !(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)) {
681 * Reset breakpoint bits because the
684 load_dr6(rdr6() & 0xfffffff0);
688 * FALLTHROUGH (TRCTRAP kernel mode, kernel address)
692 * If KDB is enabled, let it handle the debugger trap.
693 * Otherwise, debugger traps "can't happen".
696 if (kdb_trap(type, 0, frame))
704 if (time_second - lastalert > 10) {
705 log(LOG_WARNING, "NMI: power fail\n");
707 lastalert = time_second;
710 #else /* !POWERFAIL_NMI */
711 /* machine/parity/power fail/"kitchen sink" faults */
712 if (isa_nmi(code) == 0) {
715 * NMI can be hooked up to a pushbutton
719 printf ("NMI ... going to debugger\n");
720 kdb_trap(type, 0, frame);
724 } else if (panic_on_nmi == 0)
727 #endif /* POWERFAIL_NMI */
731 trap_fatal(frame, eva);
735 /* Translate fault for emulators (e.g. Linux) */
736 if (*p->p_sysent->sv_transtrap)
737 i = (*p->p_sysent->sv_transtrap)(i, type);
739 ksiginfo_init_trap(&ksi);
741 ksi.ksi_code = ucode;
742 ksi.ksi_addr = (void *)addr;
743 ksi.ksi_trapno = type;
744 trapsignal(td, &ksi);
747 if (type <= MAX_TRAP_MSG) {
748 uprintf("fatal process exception: %s",
750 if ((type == T_PAGEFLT) || (type == T_PROTFLT))
751 uprintf(", fault VA = 0x%lx", (u_long)eva);
758 mtx_assert(&Giant, MA_NOTOWNED);
759 KASSERT(PCB_USER_FPU(td->td_pcb),
760 ("Return from trap with kernel FPU ctx leaked"));
767 trap_pfault(frame, usermode, eva)
768 struct trapframe *frame;
773 struct vmspace *vm = NULL;
777 struct thread *td = curthread;
778 struct proc *p = td->td_proc;
780 va = trunc_page(eva);
781 if (va >= KERNBASE) {
783 * Don't allow user-mode faults in kernel address space.
784 * An exception: if the faulting address is the invalid
785 * instruction entry in the IDT, then the Intel Pentium
786 * F00F bug workaround was triggered, and we need to
787 * treat it is as an illegal instruction, and not a page
790 #if defined(I586_CPU) && !defined(NO_F00F_HACK)
791 if ((eva == (unsigned int)&idt[6]) && has_f00f_bug)
800 * This is a fault on non-kernel virtual memory.
801 * vm is initialized above to NULL. If curproc is NULL
802 * or curproc->p_vmspace is NULL the fault is fatal.
814 * PGEX_I is defined only if the execute disable bit capability is
815 * supported and enabled.
817 if (frame->tf_err & PGEX_W)
818 ftype = VM_PROT_WRITE;
820 else if ((frame->tf_err & PGEX_I) && pg_nx != 0)
821 ftype = VM_PROT_EXECUTE;
824 ftype = VM_PROT_READ;
826 if (map != kernel_map) {
828 * Keep swapout from messing with us during this
835 /* Fault in the user page: */
836 rv = vm_fault(map, va, ftype,
837 (ftype & VM_PROT_WRITE) ? VM_FAULT_DIRTY
845 * Don't have to worry about process locking or stacks in the
848 rv = vm_fault(map, va, ftype, VM_FAULT_NORMAL);
850 if (rv == KERN_SUCCESS)
854 if (td->td_intr_nesting_level == 0 &&
855 PCPU_GET(curpcb)->pcb_onfault != NULL) {
856 frame->tf_eip = (int)PCPU_GET(curpcb)->pcb_onfault;
859 trap_fatal(frame, eva);
863 return((rv == KERN_PROTECTION_FAILURE) ? SIGBUS : SIGSEGV);
867 trap_fatal(frame, eva)
868 struct trapframe *frame;
873 struct soft_segment_descriptor softseg;
876 code = frame->tf_err;
877 type = frame->tf_trapno;
878 sdtossd(&gdt[IDXSEL(frame->tf_cs & 0xffff)].sd, &softseg);
880 if (type <= MAX_TRAP_MSG)
881 msg = trap_msg[type];
884 printf("\n\nFatal trap %d: %s while in %s mode\n", type, msg,
885 frame->tf_eflags & PSL_VM ? "vm86" :
886 ISPL(frame->tf_cs) == SEL_UPL ? "user" : "kernel");
888 /* two separate prints in case of a trap on an unmapped page */
889 printf("cpuid = %d; ", PCPU_GET(cpuid));
890 printf("apic id = %02x\n", PCPU_GET(apic_id));
892 if (type == T_PAGEFLT) {
893 printf("fault virtual address = 0x%x\n", eva);
894 printf("fault code = %s %s, %s\n",
895 code & PGEX_U ? "user" : "supervisor",
896 code & PGEX_W ? "write" : "read",
897 code & PGEX_P ? "protection violation" : "page not present");
899 printf("instruction pointer = 0x%x:0x%x\n",
900 frame->tf_cs & 0xffff, frame->tf_eip);
901 if ((ISPL(frame->tf_cs) == SEL_UPL) || (frame->tf_eflags & PSL_VM)) {
902 ss = frame->tf_ss & 0xffff;
905 ss = GSEL(GDATA_SEL, SEL_KPL);
906 esp = (int)&frame->tf_esp;
908 printf("stack pointer = 0x%x:0x%x\n", ss, esp);
909 printf("frame pointer = 0x%x:0x%x\n", ss, frame->tf_ebp);
910 printf("code segment = base 0x%x, limit 0x%x, type 0x%x\n",
911 softseg.ssd_base, softseg.ssd_limit, softseg.ssd_type);
912 printf(" = DPL %d, pres %d, def32 %d, gran %d\n",
913 softseg.ssd_dpl, softseg.ssd_p, softseg.ssd_def32,
915 printf("processor eflags = ");
916 if (frame->tf_eflags & PSL_T)
917 printf("trace trap, ");
918 if (frame->tf_eflags & PSL_I)
919 printf("interrupt enabled, ");
920 if (frame->tf_eflags & PSL_NT)
921 printf("nested task, ");
922 if (frame->tf_eflags & PSL_RF)
924 if (frame->tf_eflags & PSL_VM)
926 printf("IOPL = %d\n", (frame->tf_eflags & PSL_IOPL) >> 12);
927 printf("current process = ");
929 printf("%lu (%s)\n", (u_long)curproc->p_pid, curthread->td_name);
935 if (debugger_on_panic || kdb_active) {
936 frame->tf_err = eva; /* smuggle fault address to ddb */
937 if (kdb_trap(type, 0, frame)) {
938 frame->tf_err = code; /* restore error code */
941 frame->tf_err = code; /* restore error code */
944 printf("trap number = %d\n", type);
945 if (type <= MAX_TRAP_MSG)
946 panic("%s", trap_msg[type]);
948 panic("unknown/reserved trap");
952 * Double fault handler. Called when a fault occurs while writing
953 * a frame for a trap/exception onto the stack. This usually occurs
954 * when the stack overflows (such is the case with infinite recursion,
957 * XXX Note that the current PTD gets replaced by IdlePTD when the
958 * task switch occurs. This means that the stack that was active at
959 * the time of the double fault is not available at <kstack> unless
960 * the machine was idle when the double fault occurred. The downside
961 * of this is that "trace <ebp>" in ddb won't work.
967 if (dtrace_doubletrap_func != NULL)
968 (*dtrace_doubletrap_func)();
970 printf("\nFatal double fault:\n");
971 printf("eip = 0x%x\n", PCPU_GET(common_tss.tss_eip));
972 printf("esp = 0x%x\n", PCPU_GET(common_tss.tss_esp));
973 printf("ebp = 0x%x\n", PCPU_GET(common_tss.tss_ebp));
975 /* two separate prints in case of a trap on an unmapped page */
976 printf("cpuid = %d; ", PCPU_GET(cpuid));
977 printf("apic id = %02x\n", PCPU_GET(apic_id));
979 panic("double fault");
983 cpu_fetch_syscall_args(struct thread *td, struct syscall_args *sa)
986 struct trapframe *frame;
991 frame = td->td_frame;
993 params = (caddr_t)frame->tf_esp + sizeof(int);
994 sa->code = frame->tf_eax;
997 * Need to check if this is a 32 bit or 64 bit syscall.
999 if (sa->code == SYS_syscall) {
1001 * Code is first argument, followed by actual args.
1003 sa->code = fuword(params);
1004 params += sizeof(int);
1005 } else if (sa->code == SYS___syscall) {
1007 * Like syscall, but code is a quad, so as to maintain
1008 * quad alignment for the rest of the arguments.
1010 sa->code = fuword(params);
1011 params += sizeof(quad_t);
1014 if (p->p_sysent->sv_mask)
1015 sa->code &= p->p_sysent->sv_mask;
1016 if (sa->code >= p->p_sysent->sv_size)
1017 sa->callp = &p->p_sysent->sv_table[0];
1019 sa->callp = &p->p_sysent->sv_table[sa->code];
1020 sa->narg = sa->callp->sy_narg;
1022 if (params != NULL && sa->narg != 0)
1023 error = copyin(params, (caddr_t)sa->args,
1024 (u_int)(sa->narg * sizeof(int)));
1029 td->td_retval[0] = 0;
1030 td->td_retval[1] = frame->tf_edx;
1037 * syscall - system call request C handler
1039 * A system call is essentially treated as a trap.
1042 syscall(struct trapframe *frame)
1045 struct syscall_args sa;
1046 register_t orig_tf_eflags;
1051 if (ISPL(frame->tf_cs) != SEL_UPL) {
1056 orig_tf_eflags = frame->tf_eflags;
1059 td->td_frame = frame;
1061 error = syscallenter(td, &sa);
1066 if ((orig_tf_eflags & PSL_T) && !(orig_tf_eflags & PSL_VM)) {
1067 frame->tf_eflags &= ~PSL_T;
1068 ksiginfo_init_trap(&ksi);
1069 ksi.ksi_signo = SIGTRAP;
1070 ksi.ksi_code = TRAP_TRACE;
1071 ksi.ksi_addr = (void *)frame->tf_eip;
1072 trapsignal(td, &ksi);
1075 KASSERT(PCB_USER_FPU(td->td_pcb),
1076 ("System call %s returning with kernel FPU ctx leaked",
1077 syscallname(td->td_proc, sa.code)));
1078 KASSERT(td->td_pcb->pcb_save == &td->td_pcb->pcb_user_save,
1079 ("System call %s returning with mangled pcb_save",
1080 syscallname(td->td_proc, sa.code)));
1082 syscallret(td, error, &sa);