2 * Copyright (c) 2000-2005 Apple Computer, Inc. All rights reserved.
4 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
6 * This file contains Original Code and/or Modifications of Original Code
7 * as defined in and that are subject to the Apple Public Source License
8 * Version 2.0 (the 'License'). You may not use this file except in
9 * compliance with the License. The rights granted to you under the License
10 * may not be used to create, or enable the creation or redistribution of,
11 * unlawful or unlicensed copies of an Apple operating system, or to
12 * circumvent, violate, or enable the circumvention or violation of, any
13 * terms of an Apple operating system software license agreement.
15 * Please obtain a copy of the License at
16 * http://www.opensource.apple.com/apsl/ and read it before using this file.
18 * The Original Code and all software distributed under the License are
19 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
20 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
21 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
22 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
23 * Please see the License for the specific language governing rights and
24 * limitations under the License.
26 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
30 #include <mach_debug.h>
31 #include <mach_ldebug.h>
33 #include <mach/kern_return.h>
34 #include <mach/mach_traps.h>
35 #include <mach/thread_status.h>
36 #include <mach/vm_param.h>
38 #include <kern/counters.h>
39 #include <kern/cpu_data.h>
40 #include <kern/mach_param.h>
41 #include <kern/task.h>
42 #include <kern/thread.h>
43 #include <kern/sched_prim.h>
44 #include <kern/misc_protos.h>
45 #include <kern/assert.h>
47 #include <kern/syscall_sw.h>
48 #include <ipc/ipc_port.h>
49 #include <vm/vm_kern.h>
52 #include <i386/cpu_data.h>
53 #include <i386/cpu_number.h>
54 #include <i386/thread.h>
55 #include <i386/eflags.h>
56 #include <i386/proc_reg.h>
59 #include <i386/user_ldt.h>
61 #include <i386/iopb_entries.h>
62 #include <i386/machdep_call.h>
63 #include <i386/misc_protos.h>
64 #include <i386/cpu_data.h>
65 #include <i386/cpu_number.h>
66 #include <i386/mp_desc.h>
67 #include <i386/vmparam.h>
68 #include <i386/trap.h>
69 #include <mach/i386/syscall_sw.h>
70 #include <sys/syscall.h>
71 #include <sys/kdebug.h>
72 #include <sys/ktrace.h>
73 #include <sys/errno.h>
74 #include <../bsd/sys/sysent.h>
76 extern struct proc
*current_proc(void);
77 extern struct proc
* kernproc
;
98 void * find_user_regs(thread_t
);
100 unsigned int get_msr_exportmask(void);
102 unsigned int get_msr_nbits(void);
104 unsigned int get_msr_rbits(void);
107 thread_compose_cthread_desc(unsigned int addr
, pcb_t pcb
);
114 * Return the user stack pointer from the machine
115 * dependent thread state info.
119 __unused thread_t thread
,
121 thread_state_t tstate
,
122 __unused
unsigned int count
,
123 user_addr_t
*user_stack
,
131 case OLD_i386_THREAD_STATE
:
132 case x86_THREAD_STATE32
:
134 x86_thread_state32_t
*state25
;
136 state25
= (x86_thread_state32_t
*) tstate
;
139 *user_stack
= state25
->esp
;
141 *user_stack
= VM_USRSTACK32
;
142 if (customstack
&& state25
->esp
)
149 case x86_THREAD_STATE64
:
151 x86_thread_state64_t
*state25
;
153 state25
= (x86_thread_state64_t
*) tstate
;
156 *user_stack
= state25
->rsp
;
158 *user_stack
= VM_USRSTACK64
;
159 if (customstack
&& state25
->rsp
)
167 return (KERN_INVALID_ARGUMENT
);
170 return (KERN_SUCCESS
);
176 __unused thread_t thread
,
178 thread_state_t tstate
,
179 __unused
unsigned int count
,
180 mach_vm_offset_t
*entry_point
186 if (*entry_point
== 0)
187 *entry_point
= VM_MIN_ADDRESS
;
190 case OLD_i386_THREAD_STATE
:
191 case x86_THREAD_STATE32
:
193 x86_thread_state32_t
*state25
;
195 state25
= (x86_thread_state32_t
*) tstate
;
196 *entry_point
= state25
->eip
? state25
->eip
: VM_MIN_ADDRESS
;
200 case x86_THREAD_STATE64
:
202 x86_thread_state64_t
*state25
;
204 state25
= (x86_thread_state64_t
*) tstate
;
205 *entry_point
= state25
->rip
? state25
->rip
: VM_MIN_ADDRESS64
;
209 return (KERN_SUCCESS
);
214 * Duplicate parent state in child
227 if ((child_pcb
= child
->machine
.pcb
) == NULL
||
228 (parent_pcb
= parent
->machine
.pcb
) == NULL
)
229 return (KERN_FAILURE
);
231 * Copy over the i386_saved_state registers
233 if (cpu_mode_is64bit()) {
234 if (thread_is_64bit(parent
))
235 bcopy(USER_REGS64(parent
), USER_REGS64(child
), sizeof(x86_saved_state64_t
));
237 bcopy(USER_REGS32(parent
), USER_REGS32(child
), sizeof(x86_saved_state_compat32_t
));
239 bcopy(USER_REGS32(parent
), USER_REGS32(child
), sizeof(x86_saved_state32_t
));
242 * Check to see if parent is using floating point
243 * and if so, copy the registers to the child
245 fpu_dup_fxstate(parent
, child
);
249 * Copy the parent's cthread id and USER_CTHREAD descriptor, if 32-bit.
251 child_pcb
->cthread_self
= parent_pcb
->cthread_self
;
252 if (!thread_is_64bit(parent
))
253 child_pcb
->cthread_desc
= parent_pcb
->cthread_desc
;
256 * FIXME - should a user specified LDT, TSS and V86 info
257 * be duplicated as well?? - probably not.
259 // duplicate any use LDT entry that was set I think this is appropriate.
260 if (parent_pcb
->uldt_selector
!= 0) {
261 child_pcb
->uldt_selector
= parent_pcb
->uldt_selector
;
262 child_pcb
->uldt_desc
= parent_pcb
->uldt_desc
;
266 return (KERN_SUCCESS
);
270 * FIXME - thread_set_child
273 void thread_set_child(thread_t child
, int pid
);
275 thread_set_child(thread_t child
, int pid
)
278 if (thread_is_64bit(child
)) {
279 x86_saved_state64_t
*iss64
;
281 iss64
= USER_REGS64(child
);
285 iss64
->isf
.rflags
&= ~EFL_CF
;
287 x86_saved_state32_t
*iss32
;
289 iss32
= USER_REGS32(child
);
293 iss32
->efl
&= ~EFL_CF
;
298 void thread_set_parent(thread_t parent
, int pid
);
300 thread_set_parent(thread_t parent
, int pid
)
303 if (thread_is_64bit(parent
)) {
304 x86_saved_state64_t
*iss64
;
306 iss64
= USER_REGS64(parent
);
310 iss64
->isf
.rflags
&= ~EFL_CF
;
312 x86_saved_state32_t
*iss32
;
314 iss32
= USER_REGS32(parent
);
318 iss32
->efl
&= ~EFL_CF
;
325 * System Call handling code
328 extern struct proc
* i386_current_proc(void);
330 extern long fuword(vm_offset_t
);
333 /* following implemented in bsd/dev/i386/unix_signal.c */
334 int __pthread_cset(struct sysent
*);
336 void __pthread_creset(struct sysent
*);
340 machdep_syscall(x86_saved_state_t
*state
)
342 int args
[machdep_call_count
];
345 machdep_call_t
*entry
;
346 x86_saved_state32_t
*regs
;
348 assert(is_saved_state32(state
));
349 regs
= saved_state32(state
);
353 kprintf("machdep_syscall(0x%08x) code=%d\n", regs
, trapno
);
356 if (trapno
< 0 || trapno
>= machdep_call_count
) {
357 regs
->eax
= (unsigned int)kern_invalid(NULL
);
359 thread_exception_return();
362 entry
= &machdep_call_table
[trapno
];
363 nargs
= entry
->nargs
;
366 if (copyin((user_addr_t
) regs
->uesp
+ sizeof (int),
367 (char *) args
, (nargs
* sizeof (int)))) {
368 regs
->eax
= KERN_INVALID_ADDRESS
;
370 thread_exception_return();
376 regs
->eax
= (*entry
->routine
.args_0
)();
379 regs
->eax
= (*entry
->routine
.args_1
)(args
[0]);
382 regs
->eax
= (*entry
->routine
.args_2
)(args
[0], args
[1]);
385 if (!entry
->bsd_style
)
386 regs
->eax
= (*entry
->routine
.args_3
)(args
[0], args
[1], args
[2]);
391 error
= (*entry
->routine
.args_bsd_3
)(&rval
, args
[0], args
[1], args
[2]);
394 regs
->efl
|= EFL_CF
; /* carry bit */
397 regs
->efl
&= ~EFL_CF
;
402 regs
->eax
= (*entry
->routine
.args_4
)(args
[0], args
[1], args
[2], args
[3]);
406 panic("machdep_syscall: too many args");
408 if (current_thread()->funnel_lock
)
409 (void) thread_funnel_set(current_thread()->funnel_lock
, FALSE
);
411 thread_exception_return();
417 machdep_syscall64(x86_saved_state_t
*state
)
420 machdep_call_t
*entry
;
421 x86_saved_state64_t
*regs
;
423 assert(is_saved_state64(state
));
424 regs
= saved_state64(state
);
426 trapno
= regs
->rax
& SYSCALL_NUMBER_MASK
;
428 if (trapno
< 0 || trapno
>= machdep_call_count
) {
429 regs
->rax
= (unsigned int)kern_invalid(NULL
);
431 thread_exception_return();
434 entry
= &machdep_call_table64
[trapno
];
436 switch (entry
->nargs
) {
438 regs
->rax
= (*entry
->routine
.args_0
)();
441 regs
->rax
= (*entry
->routine
.args64_1
)(regs
->rdi
);
444 panic("machdep_syscall64: too many args");
446 if (current_thread()->funnel_lock
)
447 (void) thread_funnel_set(current_thread()->funnel_lock
, FALSE
);
449 thread_exception_return();
455 thread_compose_cthread_desc(unsigned int addr
, pcb_t pcb
)
457 struct real_descriptor desc
;
459 mp_disable_preemption();
463 desc
.base_low
= addr
& 0xffff;
464 desc
.base_med
= (addr
>> 16) & 0xff;
465 desc
.base_high
= (addr
>> 24) & 0xff;
466 desc
.access
= ACC_P
|ACC_PL_U
|ACC_DATA_W
;
467 desc
.granularity
= SZ_32
|SZ_G
;
468 pcb
->cthread_desc
= desc
;
469 *ldt_desc_p(USER_CTHREAD
) = desc
;
471 mp_enable_preemption();
473 return(KERN_SUCCESS
);
477 thread_set_cthread_self(uint32_t self
)
479 current_thread()->machine
.pcb
->cthread_self
= (uint64_t) self
;
481 return (KERN_SUCCESS
);
485 thread_get_cthread_self(void)
487 return ((kern_return_t
)current_thread()->machine
.pcb
->cthread_self
);
491 thread_fast_set_cthread_self(uint32_t self
)
494 x86_saved_state32_t
*iss
;
496 pcb
= (pcb_t
)current_thread()->machine
.pcb
;
497 thread_compose_cthread_desc(self
, pcb
);
498 pcb
->cthread_self
= (uint64_t) self
; /* preserve old func too */
499 iss
= saved_state32(pcb
->iss
);
500 iss
->gs
= USER_CTHREAD
;
502 return (USER_CTHREAD
);
506 thread_fast_set_cthread_self64(uint64_t self
)
509 x86_saved_state64_t
*iss
;
511 pcb
= current_thread()->machine
.pcb
;
513 /* check for canonical address, set 0 otherwise */
514 if (!IS_USERADDR64_CANONICAL(self
))
516 pcb
->cthread_self
= self
;
517 current_cpu_datap()->cpu_uber
.cu_user_gs_base
= self
;
519 /* XXX for 64-in-32 */
520 iss
= saved_state64(pcb
->iss
);
521 iss
->gs
= USER_CTHREAD
;
522 thread_compose_cthread_desc((uint32_t) self
, pcb
);
524 return (USER_CTHREAD
);
528 * thread_set_user_ldt routine is the interface for the user level
529 * settable ldt entry feature. allowing a user to create arbitrary
530 * ldt entries seems to be too large of a security hole, so instead
531 * this mechanism is in place to allow user level processes to have
532 * an ldt entry that can be used in conjunction with the FS register.
534 * Swapping occurs inside the pcb.c file along with initialization
535 * when a thread is created. The basic functioning theory is that the
536 * pcb->uldt_selector variable will contain either 0 meaning the
537 * process has not set up any entry, or the selector to be used in
538 * the FS register. pcb->uldt_desc contains the actual descriptor the
539 * user has set up stored in machine usable ldt format.
541 * Currently one entry is shared by all threads (USER_SETTABLE), but
542 * this could be changed in the future by changing how this routine
543 * allocates the selector. There seems to be no real reason at this
544 * time to have this added feature, but in the future it might be
547 * address is the linear address of the start of the data area size
548 * is the size in bytes of the area flags should always be set to 0
549 * for now. in the future it could be used to set R/W permisions or
550 * other functions. Currently the segment is created as a data segment
551 * up to 1 megabyte in size with full read/write permisions only.
553 * this call returns the segment selector or -1 if any error occurs
556 thread_set_user_ldt(uint32_t address
, uint32_t size
, uint32_t flags
)
559 struct fake_descriptor temp
;
563 return -1; // flags not supported
565 return -1; // size too big, 1 meg is the limit
567 mp_disable_preemption();
568 mycpu
= cpu_number();
570 // create a "fake" descriptor so we can use fix_desc()
571 // to build a real one...
572 // 32 bit default operation size
573 // standard read/write perms for a data segment
574 pcb
= (pcb_t
)current_thread()->machine
.pcb
;
575 temp
.offset
= address
;
576 temp
.lim_or_seg
= size
;
577 temp
.size_or_wdct
= SZ_32
;
578 temp
.access
= ACC_P
|ACC_PL_U
|ACC_DATA_W
;
580 // turn this into a real descriptor
583 // set up our data in the pcb
584 pcb
->uldt_desc
= *(struct real_descriptor
*)&temp
;
585 pcb
->uldt_selector
= USER_SETTABLE
; // set the selector value
587 // now set it up in the current table...
588 *ldt_desc_p(USER_SETTABLE
) = *(struct real_descriptor
*)&temp
;
590 mp_enable_preemption();
592 return USER_SETTABLE
;
595 #endif /* MACH_BSD */
598 typedef kern_return_t (*mach_call_t
)(void *);
600 struct mach_call_args
{
614 mach_call_arg_munger32(uint32_t sp
, int nargs
, int call_number
, struct mach_call_args
*args
);
618 mach_call_arg_munger32(uint32_t sp
, int nargs
, int call_number
, struct mach_call_args
*args
)
620 unsigned int args32
[9];
622 if (copyin((user_addr_t
)(sp
+ sizeof(int)), (char *)args32
, nargs
* sizeof (int)))
623 return KERN_INVALID_ARGUMENT
;
626 case 9: args
->arg9
= args32
[8];
627 case 8: args
->arg8
= args32
[7];
628 case 7: args
->arg7
= args32
[6];
629 case 6: args
->arg6
= args32
[5];
630 case 5: args
->arg5
= args32
[4];
631 case 4: args
->arg4
= args32
[3];
632 case 3: args
->arg3
= args32
[2];
633 case 2: args
->arg2
= args32
[1];
634 case 1: args
->arg1
= args32
[0];
636 if (call_number
== 90) {
637 /* munge_l for mach_wait_until_trap() */
638 args
->arg1
= (((uint64_t)(args32
[0])) | ((((uint64_t)(args32
[1]))<<32)));
640 if (call_number
== 93) {
641 /* munge_wl for mk_timer_arm_trap() */
642 args
->arg2
= (((uint64_t)(args32
[1])) | ((((uint64_t)(args32
[2]))<<32)));
649 __private_extern__
void
650 mach_call_munger(x86_saved_state_t
*state
);
655 mach_call_munger(x86_saved_state_t
*state
)
659 mach_call_t mach_call
;
660 kern_return_t retval
;
661 struct mach_call_args args
= { 0, 0, 0, 0, 0, 0, 0, 0, 0 };
662 x86_saved_state32_t
*regs
;
664 assert(is_saved_state32(state
));
665 regs
= saved_state32(state
);
667 call_number
= -(regs
->eax
);
669 kprintf("mach_call_munger(0x%08x) code=%d\n", regs
, call_number
);
672 if (call_number
< 0 || call_number
>= mach_trap_count
) {
673 i386_exception(EXC_SYSCALL
, call_number
, 1);
676 mach_call
= (mach_call_t
)mach_trap_table
[call_number
].mach_trap_function
;
678 if (mach_call
== (mach_call_t
)kern_invalid
) {
679 i386_exception(EXC_SYSCALL
, call_number
, 1);
682 argc
= mach_trap_table
[call_number
].mach_trap_arg_count
;
685 retval
= mach_call_arg_munger32(regs
->uesp
, argc
, call_number
, &args
);
687 if (retval
!= KERN_SUCCESS
) {
690 thread_exception_return();
694 KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_EXCP_SC
, (call_number
)) | DBG_FUNC_START
,
695 (int) args
.arg1
, (int) args
.arg2
, (int) args
.arg3
, (int) args
.arg4
, 0);
697 retval
= mach_call(&args
);
699 KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_EXCP_SC
,(call_number
)) | DBG_FUNC_END
,
703 thread_exception_return();
709 __private_extern__
void
710 mach_call_munger64(x86_saved_state_t
*state
);
715 mach_call_munger64(x86_saved_state_t
*state
)
719 mach_call_t mach_call
;
720 x86_saved_state64_t
*regs
;
722 assert(is_saved_state64(state
));
723 regs
= saved_state64(state
);
725 call_number
= regs
->rax
& SYSCALL_NUMBER_MASK
;
727 KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_EXCP_SC
, (call_number
)) | DBG_FUNC_START
,
728 (int) regs
->rdi
, (int) regs
->rsi
, (int) regs
->rdx
, (int) regs
->r10
, 0);
730 if (call_number
< 0 || call_number
>= mach_trap_count
) {
731 i386_exception(EXC_SYSCALL
, regs
->rax
, 1);
734 mach_call
= (mach_call_t
)mach_trap_table
[call_number
].mach_trap_function
;
736 if (mach_call
== (mach_call_t
)kern_invalid
) {
737 i386_exception(EXC_SYSCALL
, regs
->rax
, 1);
740 argc
= mach_trap_table
[call_number
].mach_trap_arg_count
;
745 copyin_count
= (argc
- 6) * sizeof(uint64_t);
747 if (copyin((user_addr_t
)(regs
->isf
.rsp
+ sizeof(user_addr_t
)), (char *)®s
->v_arg6
, copyin_count
)) {
748 regs
->rax
= KERN_INVALID_ARGUMENT
;
750 thread_exception_return();
754 regs
->rax
= (uint64_t)mach_call((void *)(®s
->rdi
));
756 KERNEL_DEBUG_CONSTANT(MACHDBG_CODE(DBG_MACH_EXCP_SC
,(call_number
)) | DBG_FUNC_END
,
757 (int)regs
->rax
, 0, 0, 0, 0);
759 thread_exception_return();
766 * thread_setuserstack:
768 * Sets the user stack pointer into the machine
769 * dependent thread state info.
774 mach_vm_address_t user_stack
)
776 if (thread_is_64bit(thread
)) {
777 x86_saved_state64_t
*iss64
;
779 iss64
= USER_REGS64(thread
);
781 iss64
->isf
.rsp
= (uint64_t)user_stack
;
783 x86_saved_state32_t
*iss32
;
785 iss32
= USER_REGS32(thread
);
787 iss32
->uesp
= CAST_DOWN(unsigned int, user_stack
);
792 * thread_adjuserstack:
794 * Returns the adjusted user stack pointer from the machine
795 * dependent thread state info. Used for small (<2G) deltas.
802 if (thread_is_64bit(thread
)) {
803 x86_saved_state64_t
*iss64
;
805 iss64
= USER_REGS64(thread
);
807 iss64
->isf
.rsp
+= adjust
;
809 return iss64
->isf
.rsp
;
811 x86_saved_state32_t
*iss32
;
813 iss32
= USER_REGS32(thread
);
815 iss32
->uesp
+= adjust
;
817 return CAST_USER_ADDR_T(iss32
->uesp
);
822 * thread_setentrypoint:
824 * Sets the user PC into the machine
825 * dependent thread state info.
828 thread_setentrypoint(thread_t thread
, mach_vm_address_t entry
)
830 if (thread_is_64bit(thread
)) {
831 x86_saved_state64_t
*iss64
;
833 iss64
= USER_REGS64(thread
);
835 iss64
->isf
.rip
= (uint64_t)entry
;
837 x86_saved_state32_t
*iss32
;
839 iss32
= USER_REGS32(thread
);
841 iss32
->eip
= CAST_DOWN(unsigned int, entry
);
847 thread_setsinglestep(thread_t thread
, int on
)
849 if (thread_is_64bit(thread
)) {
850 x86_saved_state64_t
*iss64
;
852 iss64
= USER_REGS64(thread
);
855 iss64
->isf
.rflags
|= EFL_TF
;
857 iss64
->isf
.rflags
&= ~EFL_TF
;
859 x86_saved_state32_t
*iss32
;
861 iss32
= USER_REGS32(thread
);
864 iss32
->efl
|= EFL_TF
;
866 iss32
->efl
&= ~EFL_TF
;
872 /* XXX this should be a struct savearea so that CHUD will work better on x86 */
877 return USER_STATE(thread
);