[apple/xnu.git] / bsd / kern / kern_core.c

/*
 * Copyright (c) 2000-2006 Apple Computer, Inc. All rights reserved.
 *
 * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
 * 
 * This file contains Original Code and/or Modifications of Original Code
 * as defined in and that are subject to the Apple Public Source License
 * Version 2.0 (the 'License'). You may not use this file except in
 * compliance with the License. The rights granted to you under the License
 * may not be used to create, or enable the creation or redistribution of,
 * unlawful or unlicensed copies of an Apple operating system, or to
 * circumvent, violate, or enable the circumvention or violation of, any
 * terms of an Apple operating system software license agreement.
 * 
 * Please obtain a copy of the License at
 * http://www.opensource.apple.com/apsl/ and read it before using this file.
 * 
 * The Original Code and all software distributed under the License are
 * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
 * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
 * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
 * Please see the License for the specific language governing rights and
 * limitations under the License.
 * 
 * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
 */
/* Copyright (c) 1991 NeXT Computer, Inc.  All rights reserved.
 *
 *	File:	bsd/kern/kern_core.c
 *
 *	This file contains machine independent code for performing core dumps.
 *
 */

#include <mach/vm_param.h>
#include <mach/thread_status.h>

#include <sys/param.h>
#include <sys/systm.h>
#include <sys/signalvar.h>
#include <sys/resourcevar.h>
#include <sys/namei.h>
#include <sys/vnode_internal.h>
#include <sys/proc_internal.h>
#include <sys/kauth.h>
#include <sys/timeb.h>
#include <sys/times.h>
#include <sys/acct.h>
#include <sys/file_internal.h>
#include <sys/uio.h>
#include <sys/kernel.h>
#include <sys/stat.h>

#include <mach-o/loader.h>
#include <mach/vm_region.h>
#include <mach/vm_statistics.h>

#include <vm/vm_kern.h>
#include <vm/vm_protos.h> /* last */
#include <vm/vm_map.h>		/* current_map() */
#include <mach/mach_vm.h>	/* mach_vm_region_recurse() */
#include <mach/task.h>		/* task_suspend() */
#include <kern/task.h>		/* get_task_numacts() */

#include <security/audit/audit.h>

typedef struct {
	int	flavor;			/* the number for this flavor */
	mach_msg_type_number_t	count;	/* count of ints in this flavor */
} mythread_state_flavor_t;

#if defined (__i386__) || defined (__x86_64__)
mythread_state_flavor_t thread_flavor_array [] = { 
		{x86_THREAD_STATE, x86_THREAD_STATE_COUNT},
		{x86_FLOAT_STATE, x86_FLOAT_STATE_COUNT},
		{x86_EXCEPTION_STATE, x86_EXCEPTION_STATE_COUNT},
		};
int mynum_flavors=3;
#else
#error architecture not supported
#endif


typedef struct {
	vm_offset_t header; 
	int  hoffset;
	mythread_state_flavor_t *flavors;
	int tstate_size;
	int flavor_count;
} tir_t;

/* XXX should be static */
void collectth_state(thread_t th_act, void *tirp);

extern int freespace_mb(vnode_t vp);

/* XXX not in a Mach header anywhere */
kern_return_t thread_getstatus(register thread_t act, int flavor,
	thread_state_t tstate, mach_msg_type_number_t *count);
void task_act_iterate_wth_args(task_t, void(*)(thread_t, void *), void *);
extern kern_return_t task_suspend_internal(task_t);

static cpu_type_t process_cpu_type(proc_t proc);
static cpu_type_t process_cpu_subtype(proc_t proc);

#ifdef SECURE_KERNEL
__XNU_PRIVATE_EXTERN int do_coredump = 0;	/* default: don't dump cores */
#else
__XNU_PRIVATE_EXTERN int do_coredump = 1;	/* default: dump cores */
#endif
__XNU_PRIVATE_EXTERN int sugid_coredump = 0; /* default: but not SGUID binaries */


/* cpu_type returns only the most generic indication of the current CPU. */
/* in a core we want to know the kind of process. */

static cpu_type_t
process_cpu_type(proc_t core_proc)
{
	cpu_type_t what_we_think;
#if defined (__i386__) || defined (__x86_64__)
    if (IS_64BIT_PROCESS(core_proc)) {
		what_we_think = CPU_TYPE_X86_64;
	} else {
		what_we_think = CPU_TYPE_I386;
	}
#endif
	return what_we_think;
}

static cpu_type_t
process_cpu_subtype(proc_t core_proc)
{
	cpu_type_t what_we_think;
#if defined (__i386__) || defined (__x86_64__)
    if (IS_64BIT_PROCESS(core_proc)) {
		what_we_think = CPU_SUBTYPE_X86_64_ALL;
	} else {
		what_we_think = CPU_SUBTYPE_I386_ALL;
	}
#endif
	return what_we_think;
}

void
collectth_state(thread_t th_act, void *tirp)
{
	vm_offset_t	header;
	int  hoffset, i ;
	mythread_state_flavor_t *flavors;
	struct thread_command	*tc;
	tir_t *t = (tir_t *)tirp;

		/*
		 *	Fill in thread command structure.
		 */
		header = t->header;
		hoffset = t->hoffset;
		flavors = t->flavors;
	
		tc = (struct thread_command *) (header + hoffset);
		tc->cmd = LC_THREAD;
		tc->cmdsize = sizeof(struct thread_command)
				+ t->tstate_size;
		hoffset += sizeof(struct thread_command);
		/*
		 * Follow with a struct thread_state_flavor and
		 * the appropriate thread state struct for each
		 * thread state flavor.
		 */
		for (i = 0; i < t->flavor_count; i++) {
			*(mythread_state_flavor_t *)(header+hoffset) =
			  flavors[i];
			hoffset += sizeof(mythread_state_flavor_t);
			thread_getstatus(th_act, flavors[i].flavor,
					(thread_state_t)(header+hoffset),
					&flavors[i].count);
			hoffset += flavors[i].count*sizeof(int);
		}

		t->hoffset = hoffset;
}


/*
 * coredump
 *
 * Description:	Create a core image on the file "core" for the process
 *		indicated
 *
 * Parameters:	core_proc			Process to dump core [*]
 *				reserve_mb			If non-zero, leave filesystem with
 *									at least this much free space.
 *				ignore_ulimit		If set, ignore the process's core file ulimit.	
 *
 * Returns:	0				Success
 *		EFAULT				Failed
 *
 * IMPORTANT:	This function can only be called on the current process, due
 *		to assumptions below; see variable declaration section for
 *		details.
 */
#define	MAX_TSTATE_FLAVORS	10
int
coredump(proc_t core_proc, uint32_t reserve_mb, int ignore_ulimit)
{
/* Begin assumptions that limit us to only the current process */
	vfs_context_t ctx = vfs_context_current();
	vm_map_t	map = current_map();
	task_t		task = current_task();
/* End assumptions */
	kauth_cred_t cred = vfs_context_ucred(ctx);
	int error = 0;
	struct vnode_attr va;
	int		thread_count, segment_count;
	int		command_size, header_size, tstate_size;
	int		hoffset;
	off_t		foffset;
	mach_vm_offset_t vmoffset;
	vm_offset_t	header;
	mach_vm_size_t	vmsize;
	vm_prot_t	prot;
	vm_prot_t	maxprot;
	vm_inherit_t	inherit;
	int		error1 = 0;
	char		stack_name[MAXCOMLEN+6];
	char		*alloced_name = NULL;
	char		*name;
	mythread_state_flavor_t flavors[MAX_TSTATE_FLAVORS];
	vm_size_t	mapsize;
	int		i;
	uint32_t nesting_depth = 0;
	kern_return_t	kret;
	struct vm_region_submap_info_64 vbr;
	mach_msg_type_number_t vbrcount = 0;
	tir_t tir1;
	struct vnode * vp;
	struct mach_header	*mh = NULL;	/* protected by is_64 */
	struct mach_header_64	*mh64 = NULL;	/* protected by is_64 */
	int		is_64 = 0;
	size_t		mach_header_sz = sizeof(struct mach_header);
	size_t		segment_command_sz = sizeof(struct segment_command);

	if (do_coredump == 0 ||		/* Not dumping at all */
	    ( (sugid_coredump == 0) &&	/* Not dumping SUID/SGID binaries */
	      ( (kauth_cred_getsvuid(cred) != kauth_cred_getruid(cred)) ||
	        (kauth_cred_getsvgid(cred) != kauth_cred_getrgid(cred))))) {

#if CONFIG_AUDIT
		audit_proc_coredump(core_proc, NULL, EFAULT);
#endif
		return (EFAULT);
	}

	if (IS_64BIT_PROCESS(core_proc)) {
		is_64 = 1;
		mach_header_sz = sizeof(struct mach_header_64);
		segment_command_sz = sizeof(struct segment_command_64);
	}

	mapsize = get_vmmap_size(map);

	if ((mapsize >=  core_proc->p_rlimit[RLIMIT_CORE].rlim_cur) && (ignore_ulimit == 0))
		return (EFAULT);
	(void) task_suspend_internal(task);

	MALLOC(alloced_name, char *, MAXPATHLEN, M_TEMP, M_NOWAIT | M_ZERO);

	/* create name according to sysctl'able format string */
	/* if name creation fails, fall back to historical behaviour... */
	if (alloced_name == NULL ||
	    proc_core_name(core_proc->p_comm, kauth_cred_getuid(cred),
			   core_proc->p_pid, alloced_name, MAXPATHLEN)) {
		snprintf(stack_name, sizeof(stack_name),
			 "/cores/core.%d", core_proc->p_pid);
		name = stack_name;
	} else
		name = alloced_name;

	if ((error = vnode_open(name, (O_CREAT | FWRITE | O_NOFOLLOW), S_IRUSR, VNODE_LOOKUP_NOFOLLOW, &vp, ctx)))
		goto out2;

	VATTR_INIT(&va);
	VATTR_WANTED(&va, va_nlink);
	/* Don't dump to non-regular files or files with links. */
	if (vp->v_type != VREG ||
	    vnode_getattr(vp, &va, ctx) || va.va_nlink != 1) {
		error = EFAULT;
		goto out;
	}

	VATTR_INIT(&va);	/* better to do it here than waste more stack in vnode_setsize */
	VATTR_SET(&va, va_data_size, 0);
	vnode_setattr(vp, &va, ctx);
	core_proc->p_acflag |= ACORE;

	if ((reserve_mb > 0) &&
	    ((freespace_mb(vp) - (mapsize >> 20)) < reserve_mb)) {
		error = ENOSPC;
		goto out;
	}

	/*
	 *	If the task is modified while dumping the file
	 *	(e.g., changes in threads or VM, the resulting
	 *	file will not necessarily be correct.
	 */

	thread_count = get_task_numacts(task);
	segment_count = get_vmmap_entries(map);	/* XXX */
	tir1.flavor_count = sizeof(thread_flavor_array)/sizeof(mythread_state_flavor_t);
	bcopy(thread_flavor_array, flavors,sizeof(thread_flavor_array));
	tstate_size = 0;
	for (i = 0; i < tir1.flavor_count; i++)
		tstate_size += sizeof(mythread_state_flavor_t) +
		  (flavors[i].count * sizeof(int));
	command_size = segment_count * segment_command_sz +
	  thread_count*sizeof(struct thread_command) +
	  tstate_size*thread_count;

	header_size = command_size + mach_header_sz;

	if (kmem_alloc(kernel_map, &header, (vm_size_t)header_size) != KERN_SUCCESS) {
		error = ENOMEM;
		goto out;
	}

	/*
	 *	Set up Mach-O header.
	 */
	if (is_64) {
		mh64 = (struct mach_header_64 *)header;
		mh64->magic = MH_MAGIC_64;
		mh64->cputype = process_cpu_type(core_proc);
		mh64->cpusubtype = process_cpu_subtype(core_proc);
		mh64->filetype = MH_CORE;
		mh64->ncmds = segment_count + thread_count;
		mh64->sizeofcmds = command_size;
		mh64->reserved = 0;		/* 8 byte alignment */
	} else {
		mh = (struct mach_header *)header;
		mh->magic = MH_MAGIC;
		mh->cputype = process_cpu_type(core_proc);
		mh->cpusubtype = process_cpu_subtype(core_proc);
		mh->filetype = MH_CORE;
		mh->ncmds = segment_count + thread_count;
		mh->sizeofcmds = command_size;
	}

	hoffset = mach_header_sz;	/* offset into header */
	foffset = round_page(header_size);	/* offset into file */
	vmoffset = MACH_VM_MIN_ADDRESS;		/* offset into VM */

	/*
	 * We use to check for an error, here, now we try and get 
	 * as much as we can
	 */
	while (segment_count > 0) {
		struct segment_command		*sc;
		struct segment_command_64	*sc64;

		/*
		 *	Get region information for next region.
		 */
		
		while (1) {
			vbrcount = VM_REGION_SUBMAP_INFO_COUNT_64;
			if((kret = mach_vm_region_recurse(map, 
					&vmoffset, &vmsize, &nesting_depth, 
					(vm_region_recurse_info_t)&vbr,
					&vbrcount)) != KERN_SUCCESS) {
				break;
			}
			/*
			 * If we get a valid mapping back, but we're dumping
			 * a 32 bit process,  and it's over the allowable
			 * address space of a 32 bit process, it's the same
			 * as if mach_vm_region_recurse() failed.
			 */
			if (!(is_64) &&
			    (vmoffset + vmsize > VM_MAX_ADDRESS)) {
			    	kret = KERN_INVALID_ADDRESS;
				break;
			}
			if(vbr.is_submap) {
				nesting_depth++;
				continue;
			} else {
				break;
			}
		}
		if(kret != KERN_SUCCESS)
			break;

		prot = vbr.protection;
		maxprot = vbr.max_protection;
		inherit = vbr.inheritance;
		/*
		 *	Fill in segment command structure.
		 */
		if (is_64) {
			sc64 = (struct segment_command_64 *)(header + hoffset);
			sc64->cmd = LC_SEGMENT_64;
			sc64->cmdsize = sizeof(struct segment_command_64);
			/* segment name is zeroed by kmem_alloc */
			sc64->segname[0] = 0;
			sc64->vmaddr = vmoffset;
			sc64->vmsize = vmsize;
			sc64->fileoff = foffset;
			sc64->filesize = vmsize;
			sc64->maxprot = maxprot;
			sc64->initprot = prot;
			sc64->nsects = 0;
		} else  {
			sc = (struct segment_command *) (header + hoffset);
			sc->cmd = LC_SEGMENT;
			sc->cmdsize = sizeof(struct segment_command);
			/* segment name is zeroed by kmem_alloc */
			sc->segname[0] = 0;
			sc->vmaddr = CAST_DOWN_EXPLICIT(vm_offset_t,vmoffset);
			sc->vmsize = CAST_DOWN_EXPLICIT(vm_size_t,vmsize);
			sc->fileoff = CAST_DOWN_EXPLICIT(uint32_t,foffset); /* will never truncate */
			sc->filesize = CAST_DOWN_EXPLICIT(uint32_t,vmsize); /* will never truncate */
			sc->maxprot = maxprot;
			sc->initprot = prot;
			sc->nsects = 0;
		}

		/*
		 *	Write segment out.  Try as hard as possible to
		 *	get read access to the data.
		 */
		if ((prot & VM_PROT_READ) == 0) {
			mach_vm_protect(map, vmoffset, vmsize, FALSE,
					   prot|VM_PROT_READ);
		}
		/*
		 *	Only actually perform write if we can read.
		 *	Note: if we can't read, then we end up with
		 *	a hole in the file.
		 */
		if ((maxprot & VM_PROT_READ) == VM_PROT_READ
			&& vbr.user_tag != VM_MEMORY_IOKIT
			&& coredumpok(map,vmoffset)) {
			
			error = vn_rdwr_64(UIO_WRITE, vp, vmoffset, vmsize, foffset,
					(IS_64BIT_PROCESS(core_proc) ? UIO_USERSPACE64 : UIO_USERSPACE32), 
					IO_NOCACHE|IO_NODELOCKED|IO_UNIT, cred, (int64_t *) 0, core_proc);

		}

		hoffset += segment_command_sz;
		foffset += vmsize;
		vmoffset += vmsize;
		segment_count--;
	}

	/*
	 * If there are remaining segments which have not been written
	 * out because break in the loop above, then they were not counted
	 * because they exceed the real address space of the executable
	 * type: remove them from the header's count.  This is OK, since
	 * we are allowed to have a sparse area following the segments.
	 */
	if (is_64) {
		mh64->ncmds -= segment_count;
		mh64->sizeofcmds -= segment_count * segment_command_sz;
	} else {
		mh->ncmds -= segment_count;
		mh->sizeofcmds -= segment_count * segment_command_sz;
	}

	tir1.header = header;
	tir1.hoffset = hoffset;
	tir1.flavors = flavors;
	tir1.tstate_size = tstate_size;
	task_act_iterate_wth_args(task, collectth_state,&tir1);

	/*
	 *	Write out the Mach header at the beginning of the
	 *	file.  OK to use a 32 bit write for this.
	 */
	error = vn_rdwr(UIO_WRITE, vp, (caddr_t)header, header_size, (off_t)0,
			UIO_SYSSPACE, IO_NOCACHE|IO_NODELOCKED|IO_UNIT, cred, (int *) 0, core_proc);
	kmem_free(kernel_map, header, header_size);
out:
	error1 = vnode_close(vp, FWRITE, ctx);
out2:
#if CONFIG_AUDIT
	audit_proc_coredump(core_proc, name, error);
#endif
	if (alloced_name != NULL)
		FREE(alloced_name, M_TEMP);
	if (error == 0)
		error = error1;

	return (error);
}
Commit	Line	Data
1c79356b	1	/*
2d21ac55	2	* Copyright (c) 2000-2006 Apple Computer, Inc. All rights reserved.
5d5c5d0d	3	*
2d21ac55	4	* @APPLE_OSREFERENCE_LICENSE_HEADER_START@
1c79356b	5	*
2d21ac55 A	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.
8f6c56a5	14	*
2d21ac55 A	15	* Please obtain a copy of the License at
	16	* http://www.opensource.apple.com/apsl/ and read it before using this file.
	17	*
	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
8f6c56a5 A	20	* EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
8f6c56a5 A	21	* INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
2d21ac55 A	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.
8f6c56a5	25	*
2d21ac55	26	* @APPLE_OSREFERENCE_LICENSE_HEADER_END@
1c79356b A	27	*/
	28	/* Copyright (c) 1991 NeXT Computer, Inc. All rights reserved.
	29	*
	30	* File: bsd/kern/kern_core.c
	31	*
	32	* This file contains machine independent code for performing core dumps.
	33	*
1c79356b A	34	*/
	35
	36	#include <mach/vm_param.h>
	37	#include <mach/thread_status.h>
	38
	39	#include <sys/param.h>
	40	#include <sys/systm.h>
	41	#include <sys/signalvar.h>
	42	#include <sys/resourcevar.h>
	43	#include <sys/namei.h>
91447636 A	44	#include <sys/vnode_internal.h>
	45	#include <sys/proc_internal.h>
	46	#include <sys/kauth.h>
1c79356b A	47	#include <sys/timeb.h>
1c79356b A	48	#include <sys/times.h>
1c79356b	49	#include <sys/acct.h>
91447636	50	#include <sys/file_internal.h>
1c79356b A	51	#include <sys/uio.h>
	52	#include <sys/kernel.h>
	53	#include <sys/stat.h>
	54
	55	#include <mach-o/loader.h>
	56	#include <mach/vm_region.h>
765c9de3	57	#include <mach/vm_statistics.h>
1c79356b A	58
1c79356b A	59	#include <vm/vm_kern.h>
91447636 A	60	#include <vm/vm_protos.h> /* last */
	61	#include <vm/vm_map.h> /* current_map() */
	62	#include <mach/mach_vm.h> /* mach_vm_region_recurse() */
	63	#include <mach/task.h> /* task_suspend() */
	64	#include <kern/task.h> /* get_task_numacts() */
1c79356b	65
b0d623f7 A	66	#include <security/audit/audit.h>
b0d623f7 A	67
1c79356b A	68	typedef struct {
1c79356b A	69	int flavor; /* the number for this flavor */
2d21ac55	70	mach_msg_type_number_t count; /* count of ints in this flavor */
1c79356b A	71	} mythread_state_flavor_t;
1c79356b A	72
6d2010ae	73	#if defined (__i386__) \|\| defined (__x86_64__)
1c79356b	74	mythread_state_flavor_t thread_flavor_array [] = {
0c530ab8 A	75	{x86_THREAD_STATE, x86_THREAD_STATE_COUNT},
	76	{x86_FLOAT_STATE, x86_FLOAT_STATE_COUNT},
	77	{x86_EXCEPTION_STATE, x86_EXCEPTION_STATE_COUNT},
1c79356b	78	};
0c530ab8	79	int mynum_flavors=3;
1c79356b A	80	#else
	81	#error architecture not supported
	82	#endif
	83
	84
	85	typedef struct {
	86	vm_offset_t header;
	87	int hoffset;
	88	mythread_state_flavor_t *flavors;
	89	int tstate_size;
2d21ac55	90	int flavor_count;
1c79356b A	91	} tir_t;
1c79356b A	92
e5568f75	93	/* XXX should be static */
91447636	94	void collectth_state(thread_t th_act, void *tirp);
e5568f75	95
39236c6e A	96	extern int freespace_mb(vnode_t vp);
39236c6e A	97
e5568f75	98	/* XXX not in a Mach header anywhere */
91447636	99	kern_return_t thread_getstatus(register thread_t act, int flavor,
e5568f75	100	thread_state_t tstate, mach_msg_type_number_t *count);
91447636	101	void task_act_iterate_wth_args(task_t, void()(thread_t, void ), void *);
39236c6e	102	extern kern_return_t task_suspend_internal(task_t);
e5568f75	103
b0d623f7 A	104	static cpu_type_t process_cpu_type(proc_t proc);
b0d623f7 A	105	static cpu_type_t process_cpu_subtype(proc_t proc);
e5568f75	106
593a1d5f	107	#ifdef SECURE_KERNEL
39236c6e	108	__XNU_PRIVATE_EXTERN int do_coredump = 0; /* default: don't dump cores */
593a1d5f	109	#else
39236c6e	110	__XNU_PRIVATE_EXTERN int do_coredump = 1; /* default: dump cores */
593a1d5f	111	#endif
39236c6e	112	__XNU_PRIVATE_EXTERN int sugid_coredump = 0; /* default: but not SGUID binaries */
e5568f75	113
b0d623f7 A	114
	115	/* cpu_type returns only the most generic indication of the current CPU. */
	116	/* in a core we want to know the kind of process. */
	117
	118	static cpu_type_t
	119	process_cpu_type(proc_t core_proc)
	120	{
	121	cpu_type_t what_we_think;
	122	#if defined (__i386__) \|\| defined (__x86_64__)
	123	if (IS_64BIT_PROCESS(core_proc)) {
	124	what_we_think = CPU_TYPE_X86_64;
	125	} else {
	126	what_we_think = CPU_TYPE_I386;
	127	}
b0d623f7 A	128	#endif
	129	return what_we_think;
	130	}
	131
	132	static cpu_type_t
	133	process_cpu_subtype(proc_t core_proc)
	134	{
	135	cpu_type_t what_we_think;
	136	#if defined (__i386__) \|\| defined (__x86_64__)
	137	if (IS_64BIT_PROCESS(core_proc)) {
	138	what_we_think = CPU_SUBTYPE_X86_64_ALL;
	139	} else {
	140	what_we_think = CPU_SUBTYPE_I386_ALL;
	141	}
b0d623f7 A	142	#endif
	143	return what_we_think;
	144	}
	145
55e303ae	146	void
91447636	147	collectth_state(thread_t th_act, void *tirp)
1c79356b A	148	{
	149	vm_offset_t header;
	150	int hoffset, i ;
	151	mythread_state_flavor_t *flavors;
	152	struct thread_command *tc;
91447636 A	153	tir_t t = (tir_t )tirp;
91447636 A	154
1c79356b A	155	/*
	156	* Fill in thread command structure.
	157	*/
	158	header = t->header;
	159	hoffset = t->hoffset;
	160	flavors = t->flavors;
	161
	162	tc = (struct thread_command *) (header + hoffset);
	163	tc->cmd = LC_THREAD;
	164	tc->cmdsize = sizeof(struct thread_command)
	165	+ t->tstate_size;
	166	hoffset += sizeof(struct thread_command);
	167	/*
	168	* Follow with a struct thread_state_flavor and
	169	* the appropriate thread state struct for each
	170	* thread state flavor.
	171	*/
2d21ac55	172	for (i = 0; i < t->flavor_count; i++) {
1c79356b A	173	(mythread_state_flavor_t )(header+hoffset) =
	174	flavors[i];
	175	hoffset += sizeof(mythread_state_flavor_t);
	176	thread_getstatus(th_act, flavors[i].flavor,
e5568f75	177	(thread_state_t)(header+hoffset),
1c79356b A	178	&flavors[i].count);
	179	hoffset += flavors[i].count*sizeof(int);
	180	}
	181
	182	t->hoffset = hoffset;
	183	}
e5568f75	184
2d21ac55	185
1c79356b	186	/*
2d21ac55 A	187	* coredump
	188	*
	189	* Description: Create a core image on the file "core" for the process
	190	* indicated
	191	*
	192	* Parameters: core_proc Process to dump core [*]
39236c6e A	193	* reserve_mb If non-zero, leave filesystem with
	194	* at least this much free space.
	195	* ignore_ulimit If set, ignore the process's core file ulimit.
2d21ac55 A	196	*
	197	* Returns: 0 Success
	198	* EFAULT Failed
	199	*
	200	* IMPORTANT: This function can only be called on the current process, due
	201	* to assumptions below; see variable declaration section for
	202	* details.
1c79356b A	203	*/
	204	#define MAX_TSTATE_FLAVORS 10
	205	int
39236c6e	206	coredump(proc_t core_proc, uint32_t reserve_mb, int ignore_ulimit)
1c79356b	207	{
2d21ac55 A	208	/* Begin assumptions that limit us to only the current process */
	209	vfs_context_t ctx = vfs_context_current();
	210	vm_map_t map = current_map();
	211	task_t task = current_task();
	212	/* End assumptions */
	213	kauth_cred_t cred = vfs_context_ucred(ctx);
	214	int error = 0;
91447636	215	struct vnode_attr va;
1c79356b A	216	int thread_count, segment_count;
1c79356b A	217	int command_size, header_size, tstate_size;
91447636 A	218	int hoffset;
91447636 A	219	off_t foffset;
316670eb	220	mach_vm_offset_t vmoffset;
1c79356b	221	vm_offset_t header;
316670eb	222	mach_vm_size_t vmsize;
1c79356b A	223	vm_prot_t prot;
	224	vm_prot_t maxprot;
	225	vm_inherit_t inherit;
2d21ac55 A	226	int error1 = 0;
	227	char stack_name[MAXCOMLEN+6];
	228	char *alloced_name = NULL;
e5568f75	229	char *name;
1c79356b	230	mythread_state_flavor_t flavors[MAX_TSTATE_FLAVORS];
e5568f75	231	vm_size_t mapsize;
1c79356b	232	int i;
2d21ac55	233	uint32_t nesting_depth = 0;
1c79356b A	234	kern_return_t kret;
1c79356b A	235	struct vm_region_submap_info_64 vbr;
2d21ac55	236	mach_msg_type_number_t vbrcount = 0;
1c79356b A	237	tir_t tir1;
1c79356b A	238	struct vnode * vp;
2d21ac55 A	239	struct mach_header mh = NULL; / protected by is_64 */
2d21ac55 A	240	struct mach_header_64 mh64 = NULL; / protected by is_64 */
91447636 A	241	int is_64 = 0;
	242	size_t mach_header_sz = sizeof(struct mach_header);
	243	size_t segment_command_sz = sizeof(struct segment_command);
1c79356b	244
e5568f75 A	245	if (do_coredump == 0 \|\| /* Not dumping at all */
e5568f75 A	246	( (sugid_coredump == 0) && /* Not dumping SUID/SGID binaries */
6d2010ae A	247	( (kauth_cred_getsvuid(cred) != kauth_cred_getruid(cred)) \|\|
6d2010ae A	248	(kauth_cred_getsvgid(cred) != kauth_cred_getrgid(cred))))) {
b0d623f7 A	249
	250	#if CONFIG_AUDIT
	251	audit_proc_coredump(core_proc, NULL, EFAULT);
	252	#endif
1c79356b	253	return (EFAULT);
e5568f75	254	}
1c79356b	255
2d21ac55	256	if (IS_64BIT_PROCESS(core_proc)) {
91447636 A	257	is_64 = 1;
	258	mach_header_sz = sizeof(struct mach_header_64);
	259	segment_command_sz = sizeof(struct segment_command_64);
	260	}
	261
1c79356b A	262	mapsize = get_vmmap_size(map);
1c79356b A	263
39236c6e	264	if ((mapsize >= core_proc->p_rlimit[RLIMIT_CORE].rlim_cur) && (ignore_ulimit == 0))
1c79356b	265	return (EFAULT);
39236c6e	266	(void) task_suspend_internal(task);
1c79356b	267
2d21ac55	268	MALLOC(alloced_name, char *, MAXPATHLEN, M_TEMP, M_NOWAIT \| M_ZERO);
e5568f75	269
2d21ac55	270	/* create name according to sysctl'able format string */
e5568f75	271	/* if name creation fails, fall back to historical behaviour... */
b0d623f7 A	272	if (alloced_name == NULL \|\|
b0d623f7 A	273	proc_core_name(core_proc->p_comm, kauth_cred_getuid(cred),
2d21ac55 A	274	core_proc->p_pid, alloced_name, MAXPATHLEN)) {
	275	snprintf(stack_name, sizeof(stack_name),
	276	"/cores/core.%d", core_proc->p_pid);
	277	name = stack_name;
	278	} else
	279	name = alloced_name;
e5568f75	280
2d21ac55 A	281	if ((error = vnode_open(name, (O_CREAT \| FWRITE \| O_NOFOLLOW), S_IRUSR, VNODE_LOOKUP_NOFOLLOW, &vp, ctx)))
2d21ac55 A	282	goto out2;
91447636 A	283
	284	VATTR_INIT(&va);
	285	VATTR_WANTED(&va, va_nlink);
1c79356b A	286	/* Don't dump to non-regular files or files with links. */
1c79356b A	287	if (vp->v_type != VREG \|\|
2d21ac55	288	vnode_getattr(vp, &va, ctx) \|\| va.va_nlink != 1) {
1c79356b A	289	error = EFAULT;
	290	goto out;
	291	}
	292
91447636 A	293	VATTR_INIT(&va); /* better to do it here than waste more stack in vnode_setsize */
91447636 A	294	VATTR_SET(&va, va_data_size, 0);
2d21ac55 A	295	vnode_setattr(vp, &va, ctx);
2d21ac55 A	296	core_proc->p_acflag \|= ACORE;
1c79356b	297
39236c6e A	298	if ((reserve_mb > 0) &&
	299	((freespace_mb(vp) - (mapsize >> 20)) < reserve_mb)) {
	300	error = ENOSPC;
	301	goto out;
	302	}
	303
1c79356b A	304	/*
	305	* If the task is modified while dumping the file
	306	* (e.g., changes in threads or VM, the resulting
	307	* file will not necessarily be correct.
	308	*/
	309
	310	thread_count = get_task_numacts(task);
	311	segment_count = get_vmmap_entries(map); /* XXX */
6d2010ae A	312	tir1.flavor_count = sizeof(thread_flavor_array)/sizeof(mythread_state_flavor_t);
6d2010ae A	313	bcopy(thread_flavor_array, flavors,sizeof(thread_flavor_array));
1c79356b	314	tstate_size = 0;
2d21ac55	315	for (i = 0; i < tir1.flavor_count; i++)
1c79356b A	316	tstate_size += sizeof(mythread_state_flavor_t) +
1c79356b A	317	(flavors[i].count * sizeof(int));
91447636	318	command_size = segment_count * segment_command_sz +
1c79356b A	319	thread_count*sizeof(struct thread_command) +
	320	tstate_size*thread_count;
	321
91447636	322	header_size = command_size + mach_header_sz;
1c79356b	323
b0d623f7 A	324	if (kmem_alloc(kernel_map, &header, (vm_size_t)header_size) != KERN_SUCCESS) {
	325	error = ENOMEM;
	326	goto out;
	327	}
1c79356b A	328
	329	/*
	330	* Set up Mach-O header.
	331	*/
91447636 A	332	if (is_64) {
	333	mh64 = (struct mach_header_64 *)header;
	334	mh64->magic = MH_MAGIC_64;
b0d623f7 A	335	mh64->cputype = process_cpu_type(core_proc);
b0d623f7 A	336	mh64->cpusubtype = process_cpu_subtype(core_proc);
91447636 A	337	mh64->filetype = MH_CORE;
	338	mh64->ncmds = segment_count + thread_count;
	339	mh64->sizeofcmds = command_size;
	340	mh64->reserved = 0; /* 8 byte alignment */
	341	} else {
	342	mh = (struct mach_header *)header;
	343	mh->magic = MH_MAGIC;
b0d623f7 A	344	mh->cputype = process_cpu_type(core_proc);
b0d623f7 A	345	mh->cpusubtype = process_cpu_subtype(core_proc);
91447636 A	346	mh->filetype = MH_CORE;
	347	mh->ncmds = segment_count + thread_count;
	348	mh->sizeofcmds = command_size;
	349	}
	350
	351	hoffset = mach_header_sz; /* offset into header */
	352	foffset = round_page(header_size); /* offset into file */
	353	vmoffset = MACH_VM_MIN_ADDRESS; /* offset into VM */
	354
55e303ae A	355	/*
55e303ae A	356	* We use to check for an error, here, now we try and get
1c79356b A	357	* as much as we can
1c79356b A	358	*/
91447636 A	359	while (segment_count > 0) {
	360	struct segment_command *sc;
	361	struct segment_command_64 *sc64;
	362
1c79356b A	363	/*
	364	* Get region information for next region.
	365	*/
	366
	367	while (1) {
	368	vbrcount = VM_REGION_SUBMAP_INFO_COUNT_64;
91447636 A	369	if((kret = mach_vm_region_recurse(map,
	370	&vmoffset, &vmsize, &nesting_depth,
	371	(vm_region_recurse_info_t)&vbr,
	372	&vbrcount)) != KERN_SUCCESS) {
	373	break;
	374	}
	375	/*
	376	* If we get a valid mapping back, but we're dumping
	377	* a 32 bit process, and it's over the allowable
	378	* address space of a 32 bit process, it's the same
	379	* as if mach_vm_region_recurse() failed.
	380	*/
	381	if (!(is_64) &&
	382	(vmoffset + vmsize > VM_MAX_ADDRESS)) {
	383	kret = KERN_INVALID_ADDRESS;
1c79356b A	384	break;
	385	}
	386	if(vbr.is_submap) {
	387	nesting_depth++;
	388	continue;
	389	} else {
	390	break;
	391	}
	392	}
	393	if(kret != KERN_SUCCESS)
	394	break;
	395
	396	prot = vbr.protection;
	397	maxprot = vbr.max_protection;
	398	inherit = vbr.inheritance;
	399	/*
	400	* Fill in segment command structure.
	401	*/
91447636 A	402	if (is_64) {
	403	sc64 = (struct segment_command_64 *)(header + hoffset);
	404	sc64->cmd = LC_SEGMENT_64;
	405	sc64->cmdsize = sizeof(struct segment_command_64);
	406	/* segment name is zeroed by kmem_alloc */
	407	sc64->segname[0] = 0;
	408	sc64->vmaddr = vmoffset;
	409	sc64->vmsize = vmsize;
	410	sc64->fileoff = foffset;
	411	sc64->filesize = vmsize;
	412	sc64->maxprot = maxprot;
	413	sc64->initprot = prot;
	414	sc64->nsects = 0;
	415	} else {
	416	sc = (struct segment_command *) (header + hoffset);
	417	sc->cmd = LC_SEGMENT;
	418	sc->cmdsize = sizeof(struct segment_command);
	419	/* segment name is zeroed by kmem_alloc */
	420	sc->segname[0] = 0;
b0d623f7 A	421	sc->vmaddr = CAST_DOWN_EXPLICIT(vm_offset_t,vmoffset);
	422	sc->vmsize = CAST_DOWN_EXPLICIT(vm_size_t,vmsize);
	423	sc->fileoff = CAST_DOWN_EXPLICIT(uint32_t,foffset); /* will never truncate */
	424	sc->filesize = CAST_DOWN_EXPLICIT(uint32_t,vmsize); /* will never truncate */
91447636 A	425	sc->maxprot = maxprot;
	426	sc->initprot = prot;
	427	sc->nsects = 0;
	428	}
1c79356b A	429
	430	/*
	431	* Write segment out. Try as hard as possible to
	432	* get read access to the data.
	433	*/
	434	if ((prot & VM_PROT_READ) == 0) {
91447636 A	435	mach_vm_protect(map, vmoffset, vmsize, FALSE,
91447636 A	436	prot\|VM_PROT_READ);
1c79356b A	437	}
	438	/*
	439	* Only actually perform write if we can read.
	440	* Note: if we can't read, then we end up with
	441	* a hole in the file.
	442	*/
55e303ae A	443	if ((maxprot & VM_PROT_READ) == VM_PROT_READ
	444	&& vbr.user_tag != VM_MEMORY_IOKIT
	445	&& coredumpok(map,vmoffset)) {
b0d623f7 A	446
b0d623f7 A	447	error = vn_rdwr_64(UIO_WRITE, vp, vmoffset, vmsize, foffset,
2d21ac55	448	(IS_64BIT_PROCESS(core_proc) ? UIO_USERSPACE64 : UIO_USERSPACE32),
b0d623f7 A	449	IO_NOCACHE\|IO_NODELOCKED\|IO_UNIT, cred, (int64_t *) 0, core_proc);
b0d623f7 A	450
1c79356b A	451	}
1c79356b A	452
91447636 A	453	hoffset += segment_command_sz;
	454	foffset += vmsize;
	455	vmoffset += vmsize;
1c79356b A	456	segment_count--;
	457	}
	458
91447636 A	459	/*
	460	* If there are remaining segments which have not been written
	461	* out because break in the loop above, then they were not counted
	462	* because they exceed the real address space of the executable
	463	* type: remove them from the header's count. This is OK, since
	464	* we are allowed to have a sparse area following the segments.
	465	*/
	466	if (is_64) {
	467	mh64->ncmds -= segment_count;
2d21ac55	468	mh64->sizeofcmds -= segment_count * segment_command_sz;
91447636 A	469	} else {
91447636 A	470	mh->ncmds -= segment_count;
2d21ac55	471	mh->sizeofcmds -= segment_count * segment_command_sz;
91447636 A	472	}
91447636 A	473
1c79356b A	474	tir1.header = header;
	475	tir1.hoffset = hoffset;
	476	tir1.flavors = flavors;
	477	tir1.tstate_size = tstate_size;
	478	task_act_iterate_wth_args(task, collectth_state,&tir1);
	479
1c79356b A	480	/*
1c79356b A	481	* Write out the Mach header at the beginning of the
91447636	482	* file. OK to use a 32 bit write for this.
1c79356b A	483	*/
1c79356b A	484	error = vn_rdwr(UIO_WRITE, vp, (caddr_t)header, header_size, (off_t)0,
b0d623f7	485	UIO_SYSSPACE, IO_NOCACHE\|IO_NODELOCKED\|IO_UNIT, cred, (int *) 0, core_proc);
1c79356b A	486	kmem_free(kernel_map, header, header_size);
1c79356b A	487	out:
2d21ac55 A	488	error1 = vnode_close(vp, FWRITE, ctx);
2d21ac55 A	489	out2:
b0d623f7 A	490	#if CONFIG_AUDIT
	491	audit_proc_coredump(core_proc, name, error);
	492	#endif
2d21ac55 A	493	if (alloced_name != NULL)
2d21ac55 A	494	FREE(alloced_name, M_TEMP);
1c79356b A	495	if (error == 0)
1c79356b A	496	error = error1;
91447636	497
55e303ae	498	return (error);
1c79356b	499	}