]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/kern/kern_core.c
xnu-1228.12.14.tar.gz
[apple/xnu.git] / bsd / kern / kern_core.c
index 052a0d537a50537caf619d8fcaca5ded9b899e26..70fb531b359c4497125858701368bab1cb21fd7d 100644 (file)
@@ -1,16 +1,19 @@
 /*
- * Copyright (c) 2000-2002 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2006 Apple Computer, Inc. All rights reserved.
  *
- * @APPLE_LICENSE_HEADER_START@
- * 
- * Copyright (c) 1999-2003 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. Please obtain a copy of the License at
- * http://www.opensource.apple.com/apsl/ and read it before using this
- * file.
+ * 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
@@ -20,7 +23,7 @@
  * Please see the License for the specific language governing rights and
  * limitations under the License.
  * 
- * @APPLE_LICENSE_HEADER_END@
+ * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /* Copyright (c) 1991 NeXT Computer, Inc.  All rights reserved.
  *
 #include <sys/signalvar.h>
 #include <sys/resourcevar.h>
 #include <sys/namei.h>
-#include <sys/vnode.h>
-#include <sys/proc.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/buf.h>
 #include <sys/acct.h>
-#include <sys/file.h>
+#include <sys/file_internal.h>
 #include <sys/uio.h>
 #include <sys/kernel.h>
 #include <sys/stat.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() */
 
 typedef struct {
        int     flavor;                 /* the number for this flavor */
-       int     count;                  /* count of ints in this flavor */
+       mach_msg_type_number_t  count;  /* count of ints in this flavor */
 } mythread_state_flavor_t;
 
 #if defined (__ppc__)
+/* 64 bit */
+mythread_state_flavor_t thread_flavor_array64[]={
+               {PPC_THREAD_STATE64 , PPC_THREAD_STATE64_COUNT},
+               {PPC_FLOAT_STATE, PPC_FLOAT_STATE_COUNT}, 
+               {PPC_EXCEPTION_STATE64, PPC_EXCEPTION_STATE64_COUNT},
+               {PPC_VECTOR_STATE, PPC_VECTOR_STATE_COUNT}
+               };
 
+/* 32 bit */
 mythread_state_flavor_t thread_flavor_array[]={
                {PPC_THREAD_STATE , PPC_THREAD_STATE_COUNT},
                {PPC_FLOAT_STATE, PPC_FLOAT_STATE_COUNT}, 
                {PPC_EXCEPTION_STATE, PPC_EXCEPTION_STATE_COUNT},
                {PPC_VECTOR_STATE, PPC_VECTOR_STATE_COUNT}
                };
-int mynum_flavors=4;
+
 #elif defined (__i386__)
 mythread_state_flavor_t thread_flavor_array [] = { 
-               {i386_THREAD_STATE, i386_THREAD_STATE_COUNT},
-               {i386_THREAD_FPSTATE, i386_THREAD_FPSTATE_COUNT},
-               {i386_THREAD_EXCEPTSTATE, i386_THREAD_EXCEPTSTATE_COUNT},
-               {i386_THREAD_CTHREADSTATE, i386_THREAD_CTHREADSTATE_COUNT},
-               {i386_NEW_THREAD_STATE, i386_NEW_THREAD_STATE_COUNT},
-               {i386_FLOAT_STATE, i386_FLOAT_STATE_COUNT},
-               {i386_ISA_PORT_MAP_STATE, i386_ISA_PORT_MAP_STATE_COUNT},
-               {i386_V86_ASSIST_STATE, i386_V86_ASSIST_STATE_COUNT},
-               {THREAD_SYSCALL_STATE, i386_THREAD_SYSCALL_STATE_COUNT}
+               {x86_THREAD_STATE, x86_THREAD_STATE_COUNT},
+               {x86_FLOAT_STATE, x86_FLOAT_STATE_COUNT},
+               {x86_EXCEPTION_STATE, x86_EXCEPTION_STATE_COUNT},
                };
-int mynum_flavors=9;
-
+int mynum_flavors=3;
 #else
 #error architecture not supported
 #endif
@@ -93,15 +102,34 @@ typedef struct {
        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);
+
+/* 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 *);
+
+
+#ifdef SECURE_KERNEL
+__private_extern__ int do_coredump = 0;        /* default: don't dump cores */
+#else
+__private_extern__ int do_coredump = 1;        /* default: dump cores */
+#endif
+__private_extern__ int sugid_coredump = 0; /* default: but not SGUID binaries */
+
 void
-collectth_state(thread_act_t th_act, tir_t *t)
+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.
                 */
@@ -119,89 +147,124 @@ collectth_state(thread_act_t th_act, tir_t *t)
                 * the appropriate thread state struct for each
                 * thread state flavor.
                 */
-               for (i = 0; i < mynum_flavors; i++) {
+               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),
+                                       (thread_state_t)(header+hoffset),
                                        &flavors[i].count);
                        hoffset += flavors[i].count*sizeof(int);
                }
 
                t->hoffset = hoffset;
 }
+
+
 /*
- * Create a core image on the file "core".
+ * coredump
+ *
+ * Description:        Create a core image on the file "core" for the process
+ *             indicated
+ *
+ * Parameters: core_proc                       Process to dump core [*]
+ *
+ * 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(p)
-       register struct proc *p;
+coredump(proc_t core_proc)
 {
-       int error=0;
-       register struct pcred *pcred = p->p_cred;
-       register struct ucred *cred = pcred->pc_ucred;
-       struct nameidata nd;
-       struct vattr    vattr;
-       vm_map_t        map;
+/* 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, foffset, vmoffset;
+       int             hoffset;
+       off_t           foffset;
+       vm_map_offset_t vmoffset;
        vm_offset_t     header;
-       struct machine_slot     *ms;
-       struct mach_header      *mh;
-       struct segment_command  *sc;
-       struct thread_command   *tc;
-       vm_size_t       size;
+       vm_map_size_t   vmsize;
        vm_prot_t       prot;
        vm_prot_t       maxprot;
        vm_inherit_t    inherit;
-       vm_offset_t     offset;
-       int             error1;
-       task_t          task;
-       char            core_name[MAXCOMLEN+6];
+       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       nflavors,mapsize;
+       vm_size_t       mapsize;
        int             i;
-       int nesting_depth = 0;
+       uint32_t nesting_depth = 0;
        kern_return_t   kret;
        struct vm_region_submap_info_64 vbr;
-       int vbrcount=0;
+       mach_msg_type_number_t vbrcount = 0;
        tir_t tir1;
        struct vnode * vp;
-       extern boolean_t coredumpok(vm_map_t map, vm_offset_t va);  /* temp fix */
-       extern task_t current_task();   /* XXX */
+       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 (pcred->p_svuid != pcred->p_ruid || pcred->p_svgid != pcred->p_rgid)
+       if (do_coredump == 0 ||         /* Not dumping at all */
+           ( (sugid_coredump == 0) &&  /* Not dumping SUID/SGID binaries */
+             ( (cred->cr_svuid != cred->cr_ruid) ||
+               (cred->cr_svgid != cred->cr_rgid)))) {
+           
                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);
+       }
 
-       task = current_task();
-       map = current_map();
        mapsize = get_vmmap_size(map);
 
-       if (mapsize >=  p->p_rlimit[RLIMIT_CORE].rlim_cur)
+       if (mapsize >=  core_proc->p_rlimit[RLIMIT_CORE].rlim_cur)
                return (EFAULT);
        (void) task_suspend(task);
 
-       sprintf(core_name, "/cores/core.%d", p->p_pid);
-       NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_SYSSPACE, core_name, p);
-       if(error = vn_open(&nd, O_CREAT | FWRITE | O_NOFOLLOW, S_IRUSR ))
-               return (error);
-       vp = nd.ni_vp;
-       
+       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 (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 ||
-           VOP_GETATTR(vp, &vattr, cred, p) || vattr.va_nlink != 1) {
+           vnode_getattr(vp, &va, ctx) || va.va_nlink != 1) {
                error = EFAULT;
                goto out;
        }
 
-       VATTR_NULL(&vattr);
-       vattr.va_size = 0;
-       VOP_LEASE(vp, p, cred, LEASE_WRITE);
-       VOP_SETATTR(vp, &vattr, cred, p);
-       p->p_acflag |= ACORE;
+       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 the task is modified while dumping the file
@@ -211,56 +274,86 @@ coredump(p)
 
        thread_count = get_task_numacts(task);
        segment_count = get_vmmap_entries(map); /* XXX */
-       /*
-        * nflavors here is really the number of ints in flavors
-        * to meet the thread_getstatus() calling convention
-        */
-       nflavors = mynum_flavors;
-       bcopy(thread_flavor_array,flavors,sizeof(thread_flavor_array));
+#if defined (__ppc__)
+       if (is_64) {
+               tir1.flavor_count = sizeof(thread_flavor_array64)/sizeof(mythread_state_flavor_t);
+               bcopy(thread_flavor_array64, flavors,sizeof(thread_flavor_array64));
+       } else {
+#endif /* __ppc __ */
+               tir1.flavor_count = sizeof(thread_flavor_array)/sizeof(mythread_state_flavor_t);
+               bcopy(thread_flavor_array, flavors,sizeof(thread_flavor_array));
+#if defined (__ppc__)
+       }
+#endif /* __ppc __ */
        tstate_size = 0;
-       for (i = 0; i < nflavors; i++)
+       for (i = 0; i < tir1.flavor_count; i++)
                tstate_size += sizeof(mythread_state_flavor_t) +
                  (flavors[i].count * sizeof(int));
-
-       command_size = segment_count*sizeof(struct segment_command) +
+       command_size = segment_count * segment_command_sz +
          thread_count*sizeof(struct thread_command) +
          tstate_size*thread_count;
 
-       header_size = command_size + sizeof(struct mach_header);
+       header_size = command_size + mach_header_sz;
 
-       (void) kmem_alloc_wired(kernel_map,
+       (void) kmem_alloc(kernel_map,
                                    (vm_offset_t *)&header,
                                    (vm_size_t)header_size);
 
        /*
         *      Set up Mach-O header.
         */
-       mh = (struct mach_header *) header;
-       ms = &machine_slot[cpu_number()];
-       mh->magic = MH_MAGIC;
-       mh->cputype = ms->cpu_type;
-       mh->cpusubtype = ms->cpu_subtype;
-       mh->filetype = MH_CORE;
-       mh->ncmds = segment_count + thread_count;
-       mh->sizeofcmds = command_size;
-
-       hoffset = sizeof(struct mach_header);   /* offset into header */
-       foffset = round_page_32(header_size);   /* offset into file */
-       vmoffset = VM_MIN_ADDRESS;              /* offset into VM */
+       if (is_64) {
+               mh64 = (struct mach_header_64 *)header;
+               mh64->magic = MH_MAGIC_64;
+               mh64->cputype = cpu_type();
+               mh64->cpusubtype = cpu_subtype();
+               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 = cpu_type();
+               mh->cpusubtype = cpu_subtype();
+               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){
+       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 = vm_region_recurse_64(map, 
-                                       &vmoffset, &size, &nesting_depth, 
-                                       &vbr, &vbrcount)) != KERN_SUCCESS) {
+                       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) {
@@ -279,26 +372,41 @@ coredump(p)
                /*
                 *      Fill in segment command structure.
                 */
-               sc = (struct segment_command *) (header + hoffset);
-               sc->cmd = LC_SEGMENT;
-               sc->cmdsize = sizeof(struct segment_command);
-               /* segment name is zerod by kmem_alloc */
-               sc->segname[0] = 0;
-               sc->vmaddr = vmoffset;
-               sc->vmsize = size;
-               sc->fileoff = foffset;
-               sc->filesize = size;
-               sc->maxprot = maxprot;
-               sc->initprot = prot;
-               sc->nsects = 0;
+               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(vm_offset_t,vmoffset);
+                       sc->vmsize = CAST_DOWN(vm_size_t,vmsize);
+                       sc->fileoff = CAST_DOWN(uint32_t,foffset);
+                       sc->filesize = CAST_DOWN(uint32_t,vmsize);
+                       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) {
-                       vm_protect(map, vmoffset, size, FALSE,
-                                  prot|VM_PROT_READ);
+                       mach_vm_protect(map, vmoffset, vmsize, FALSE,
+                                          prot|VM_PROT_READ);
                }
                /*
                 *      Only actually perform write if we can read.
@@ -308,16 +416,44 @@ coredump(p)
                if ((maxprot & VM_PROT_READ) == VM_PROT_READ
                        && vbr.user_tag != VM_MEMORY_IOKIT
                        && coredumpok(map,vmoffset)) {
-                       error = vn_rdwr(UIO_WRITE, vp, (caddr_t)vmoffset, size, foffset,
-                               UIO_USERSPACE, IO_NODELOCKED|IO_UNIT, cred, (int *) 0, p);
+                       vm_map_size_t   tmp_vmsize = vmsize;
+                       off_t           xfer_foffset = foffset;
+
+                       //LP64todo - works around vn_rdwr_64() 2G limit
+                       while (tmp_vmsize > 0) {
+                               vm_map_size_t   xfer_vmsize = tmp_vmsize;
+                               if (xfer_vmsize > INT_MAX)
+                                       xfer_vmsize = INT_MAX;
+                               error = vn_rdwr_64(UIO_WRITE, vp,
+                                               vmoffset, xfer_vmsize, xfer_foffset,
+                                       (IS_64BIT_PROCESS(core_proc) ? UIO_USERSPACE64 : UIO_USERSPACE32), 
+                                       IO_NODELOCKED|IO_UNIT, cred, (int *) 0, core_proc);
+                               tmp_vmsize -= xfer_vmsize;
+                               xfer_foffset += xfer_vmsize;
+                       }
                }
 
-               hoffset += sizeof(struct segment_command);
-               foffset += size;
-               vmoffset += size;
+               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;
@@ -326,15 +462,18 @@ coredump(p)
 
        /*
         *      Write out the Mach header at the beginning of the
-        *      file.
+        *      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_NODELOCKED|IO_UNIT, cred, (int *) 0, p);
+                       UIO_SYSSPACE32, IO_NODELOCKED|IO_UNIT, cred, (int *) 0, core_proc);
        kmem_free(kernel_map, header, header_size);
 out:
-       VOP_UNLOCK(vp, 0, p);
-       error1 = vn_close(vp, FWRITE, cred, p);
+       error1 = vnode_close(vp, FWRITE, ctx);
+out2:
+       if (alloced_name != NULL)
+               FREE(alloced_name, M_TEMP);
        if (error == 0)
                error = error1;
+
        return (error);
 }