bsd/kern/kern_core.c

   1 /*
   2  * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
   3  *
   4  * @APPLE_LICENSE_HEADER_START@
   5  *
   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. Please obtain a copy of the License at
  10  * http://www.opensource.apple.com/apsl/ and read it before using this
  11  * file.
  12  *
  13  * The Original Code and all software distributed under the License are
  14  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  15  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
  16  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
  17  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  18  * Please see the License for the specific language governing rights and
  19  * limitations under the License.
  20  *
  21  * @APPLE_LICENSE_HEADER_END@
  22  */
  23 /* Copyright (c) 1991 NeXT Computer, Inc.  All rights reserved.
  24  *
  25  *      File:   bsd/kern/kern_core.c
  26  *
  27  *      This file contains machine independent code for performing core dumps.
  28  *
  29  */
  30
  31 #include <mach/vm_param.h>
  32 #include <mach/thread_status.h>
  33
  34 #include <sys/param.h>
  35 #include <sys/systm.h>
  36 #include <sys/signalvar.h>
  37 #include <sys/resourcevar.h>
  38 #include <sys/namei.h>
  39 #include <sys/vnode_internal.h>
  40 #include <sys/proc_internal.h>
  41 #include <sys/kauth.h>
  42 #include <sys/timeb.h>
  43 #include <sys/times.h>
  44 #include <sys/acct.h>
  45 #include <sys/file_internal.h>
  46 #include <sys/uio.h>
  47 #include <sys/kernel.h>
  48 #include <sys/stat.h>
  49
  50 #include <mach-o/loader.h>
  51 #include <mach/vm_region.h>
  52 #include <mach/vm_statistics.h>
  53
  54 #include <vm/vm_kern.h>
  55 #include <vm/vm_protos.h> /* last */
  56 #include <vm/vm_map.h>          /* current_map() */
  57 #include <mach/mach_vm.h>       /* mach_vm_region_recurse() */
  58 #include <mach/task.h>          /* task_suspend() */
  59 #include <kern/task.h>          /* get_task_numacts() */
  60
  61 typedef struct {
  62         int     flavor;                 /* the number for this flavor */
  63         int     count;                  /* count of ints in this flavor */
  64 } mythread_state_flavor_t;
  65
  66 #if defined (__ppc__)
  67
  68 mythread_state_flavor_t thread_flavor_array[]={
  69                 {PPC_THREAD_STATE , PPC_THREAD_STATE_COUNT},
  70                 {PPC_FLOAT_STATE, PPC_FLOAT_STATE_COUNT},
  71                 {PPC_EXCEPTION_STATE, PPC_EXCEPTION_STATE_COUNT},
  72                 {PPC_VECTOR_STATE, PPC_VECTOR_STATE_COUNT}
  73                 };
  74 int mynum_flavors=4;
  75 #elif defined (__i386__)
  76 mythread_state_flavor_t thread_flavor_array [] = {
  77                 {i386_THREAD_STATE, i386_THREAD_STATE_COUNT},
  78                 {i386_THREAD_FPSTATE, i386_THREAD_FPSTATE_COUNT},
  79                 {i386_THREAD_EXCEPTSTATE, i386_THREAD_EXCEPTSTATE_COUNT},
  80                 {i386_THREAD_CTHREADSTATE, i386_THREAD_CTHREADSTATE_COUNT},
  81                 {i386_NEW_THREAD_STATE, i386_NEW_THREAD_STATE_COUNT},
  82                 {i386_FLOAT_STATE, i386_FLOAT_STATE_COUNT},
  83                 {i386_ISA_PORT_MAP_STATE, i386_ISA_PORT_MAP_STATE_COUNT},
  84                 {i386_V86_ASSIST_STATE, i386_V86_ASSIST_STATE_COUNT},
  85                 {THREAD_SYSCALL_STATE, i386_THREAD_SYSCALL_STATE_COUNT}
  86                 };
  87 int mynum_flavors=9;
  88
  89 #else
  90 #error architecture not supported
  91 #endif
  92
  93
  94 typedef struct {
  95         vm_offset_t header;
  96         int  hoffset;
  97         mythread_state_flavor_t *flavors;
  98         int tstate_size;
  99 } tir_t;
 100
 101 /* XXX should be static */
 102 void collectth_state(thread_t th_act, void *tirp);
 103
 104 /* XXX not in a Mach header anywhere */
 105 kern_return_t thread_getstatus(register thread_t act, int flavor,
 106         thread_state_t tstate, mach_msg_type_number_t *count);
 107 void task_act_iterate_wth_args(task_t, void(*)(thread_t, void *), void *);
 108
 109
 110 __private_extern__ int do_coredump = 1; /* default: dump cores */
 111 __private_extern__ int sugid_coredump = 0; /* default: but not SGUID binaries */
 112
 113 void
 114 collectth_state(thread_t th_act, void *tirp)
 115 {
 116         vm_offset_t     header;
 117         int  hoffset, i ;
 118         mythread_state_flavor_t *flavors;
 119         struct thread_command   *tc;
 120         tir_t *t = (tir_t *)tirp;
 121
 122                 /*
 123                  *      Fill in thread command structure.
 124                  */
 125                 header = t->header;
 126                 hoffset = t->hoffset;
 127                 flavors = t->flavors;
 128
 129                 tc = (struct thread_command *) (header + hoffset);
 130                 tc->cmd = LC_THREAD;
 131                 tc->cmdsize = sizeof(struct thread_command)
 132                                 + t->tstate_size;
 133                 hoffset += sizeof(struct thread_command);
 134                 /*
 135                  * Follow with a struct thread_state_flavor and
 136                  * the appropriate thread state struct for each
 137                  * thread state flavor.
 138                  */
 139                 for (i = 0; i < mynum_flavors; i++) {
 140                         *(mythread_state_flavor_t *)(header+hoffset) =
 141                           flavors[i];
 142                         hoffset += sizeof(mythread_state_flavor_t);
 143                         thread_getstatus(th_act, flavors[i].flavor,
 144                                         (thread_state_t)(header+hoffset),
 145                                         &flavors[i].count);
 146                         hoffset += flavors[i].count*sizeof(int);
 147                 }
 148
 149                 t->hoffset = hoffset;
 150 }
 151
 152 /*
 153  * Create a core image on the file "core".
 154  */
 155 #define MAX_TSTATE_FLAVORS      10
 156 int
 157 coredump(struct proc *p)
 158 {
 159         int error=0;
 160         kauth_cred_t cred = kauth_cred_get();
 161         struct vnode_attr va;
 162         struct vfs_context context;
 163         vm_map_t        map;
 164         int             thread_count, segment_count;
 165         int             command_size, header_size, tstate_size;
 166         int             hoffset;
 167         off_t           foffset;
 168         vm_map_offset_t vmoffset;
 169         vm_offset_t     header;
 170         vm_map_size_t   vmsize;
 171         vm_prot_t       prot;
 172         vm_prot_t       maxprot;
 173         vm_inherit_t    inherit;
 174         int             error1;
 175         task_t          task;
 176         char            core_name[MAXCOMLEN+6];
 177         char            *name;
 178         mythread_state_flavor_t flavors[MAX_TSTATE_FLAVORS];
 179         vm_size_t       mapsize;
 180         int             i;
 181         int nesting_depth = 0;
 182         kern_return_t   kret;
 183         struct vm_region_submap_info_64 vbr;
 184         int vbrcount=0;
 185         tir_t tir1;
 186         struct vnode * vp;
 187         struct mach_header      *mh;
 188         struct mach_header_64   *mh64;
 189         int             is_64 = 0;
 190         size_t          mach_header_sz = sizeof(struct mach_header);
 191         size_t          segment_command_sz = sizeof(struct segment_command);
 192
 193         if (do_coredump == 0 ||         /* Not dumping at all */
 194             ( (sugid_coredump == 0) &&  /* Not dumping SUID/SGID binaries */
 195               ( (cred->cr_svuid != cred->cr_ruid) ||
 196                 (cred->cr_svgid != cred->cr_rgid)))) {
 197
 198                 return (EFAULT);
 199         }
 200
 201         if (IS_64BIT_PROCESS(p)) {
 202                 is_64 = 1;
 203                 mach_header_sz = sizeof(struct mach_header_64);
 204                 segment_command_sz = sizeof(struct segment_command_64);
 205         }
 206
 207         task = current_task();
 208         map = current_map();
 209         mapsize = get_vmmap_size(map);
 210
 211         if (mapsize >=  p->p_rlimit[RLIMIT_CORE].rlim_cur)
 212                 return (EFAULT);
 213         (void) task_suspend(task);
 214
 215         /* create name according to sysctl'able format string */
 216         name = proc_core_name(p->p_comm, kauth_cred_getuid(cred), p->p_pid);
 217
 218         /* if name creation fails, fall back to historical behaviour... */
 219         if (name == NULL) {
 220                 sprintf(core_name, "/cores/core.%d", p->p_pid);
 221                 name = core_name;
 222         }
 223         context.vc_proc = p;
 224         context.vc_ucred = cred;
 225
 226         if ((error = vnode_open(name, (O_CREAT | FWRITE | O_NOFOLLOW), S_IRUSR, VNODE_LOOKUP_NOFOLLOW, &vp, &context)))
 227                 return (error);
 228
 229         VATTR_INIT(&va);
 230         VATTR_WANTED(&va, va_nlink);
 231         /* Don't dump to non-regular files or files with links. */
 232         if (vp->v_type != VREG ||
 233             vnode_getattr(vp, &va, &context) || va.va_nlink != 1) {
 234                 error = EFAULT;
 235                 goto out;
 236         }
 237
 238         VATTR_INIT(&va);        /* better to do it here than waste more stack in vnode_setsize */
 239         VATTR_SET(&va, va_data_size, 0);
 240         vnode_setattr(vp, &va, &context);
 241         p->p_acflag |= ACORE;
 242
 243         /*
 244          *      If the task is modified while dumping the file
 245          *      (e.g., changes in threads or VM, the resulting
 246          *      file will not necessarily be correct.
 247          */
 248
 249         thread_count = get_task_numacts(task);
 250         segment_count = get_vmmap_entries(map); /* XXX */
 251         bcopy(thread_flavor_array,flavors,sizeof(thread_flavor_array));
 252         tstate_size = 0;
 253         for (i = 0; i < mynum_flavors; i++)
 254                 tstate_size += sizeof(mythread_state_flavor_t) +
 255                   (flavors[i].count * sizeof(int));
 256
 257         command_size = segment_count * segment_command_sz +
 258           thread_count*sizeof(struct thread_command) +
 259           tstate_size*thread_count;
 260
 261         header_size = command_size + mach_header_sz;
 262
 263         (void) kmem_alloc(kernel_map,
 264                                     (vm_offset_t *)&header,
 265                                     (vm_size_t)header_size);
 266
 267         /*
 268          *      Set up Mach-O header.
 269          */
 270         if (is_64) {
 271                 mh64 = (struct mach_header_64 *)header;
 272                 mh64->magic = MH_MAGIC_64;
 273                 mh64->cputype = cpu_type();
 274                 mh64->cpusubtype = cpu_subtype();
 275                 mh64->filetype = MH_CORE;
 276                 mh64->ncmds = segment_count + thread_count;
 277                 mh64->sizeofcmds = command_size;
 278                 mh64->reserved = 0;             /* 8 byte alignment */
 279         } else {
 280                 mh = (struct mach_header *)header;
 281                 mh->magic = MH_MAGIC;
 282                 mh->cputype = cpu_type();
 283                 mh->cpusubtype = cpu_subtype();
 284                 mh->filetype = MH_CORE;
 285                 mh->ncmds = segment_count + thread_count;
 286                 mh->sizeofcmds = command_size;
 287         }
 288
 289         hoffset = mach_header_sz;       /* offset into header */
 290         foffset = round_page(header_size);      /* offset into file */
 291         vmoffset = MACH_VM_MIN_ADDRESS;         /* offset into VM */
 292
 293         /*
 294          * We use to check for an error, here, now we try and get
 295          * as much as we can
 296          */
 297         while (segment_count > 0) {
 298                 struct segment_command          *sc;
 299                 struct segment_command_64       *sc64;
 300
 301                 /*
 302                  *      Get region information for next region.
 303                  */
 304
 305                 while (1) {
 306                         vbrcount = VM_REGION_SUBMAP_INFO_COUNT_64;
 307                         if((kret = mach_vm_region_recurse(map,
 308                                         &vmoffset, &vmsize, &nesting_depth,
 309                                         (vm_region_recurse_info_t)&vbr,
 310                                         &vbrcount)) != KERN_SUCCESS) {
 311                                 break;
 312                         }
 313                         /*
 314                          * If we get a valid mapping back, but we're dumping
 315                          * a 32 bit process,  and it's over the allowable
 316                          * address space of a 32 bit process, it's the same
 317                          * as if mach_vm_region_recurse() failed.
 318                          */
 319                         if (!(is_64) &&
 320                             (vmoffset + vmsize > VM_MAX_ADDRESS)) {
 321                                 kret = KERN_INVALID_ADDRESS;
 322                                 break;
 323                         }
 324                         if(vbr.is_submap) {
 325                                 nesting_depth++;
 326                                 continue;
 327                         } else {
 328                                 break;
 329                         }
 330                 }
 331                 if(kret != KERN_SUCCESS)
 332                         break;
 333
 334                 prot = vbr.protection;
 335                 maxprot = vbr.max_protection;
 336                 inherit = vbr.inheritance;
 337                 /*
 338                  *      Fill in segment command structure.
 339                  */
 340                 if (is_64) {
 341                         sc64 = (struct segment_command_64 *)(header + hoffset);
 342                         sc64->cmd = LC_SEGMENT_64;
 343                         sc64->cmdsize = sizeof(struct segment_command_64);
 344                         /* segment name is zeroed by kmem_alloc */
 345                         sc64->segname[0] = 0;
 346                         sc64->vmaddr = vmoffset;
 347                         sc64->vmsize = vmsize;
 348                         sc64->fileoff = foffset;
 349                         sc64->filesize = vmsize;
 350                         sc64->maxprot = maxprot;
 351                         sc64->initprot = prot;
 352                         sc64->nsects = 0;
 353                 } else  {
 354                         sc = (struct segment_command *) (header + hoffset);
 355                         sc->cmd = LC_SEGMENT;
 356                         sc->cmdsize = sizeof(struct segment_command);
 357                         /* segment name is zeroed by kmem_alloc */
 358                         sc->segname[0] = 0;
 359                         sc->vmaddr = CAST_DOWN(vm_offset_t,vmoffset);
 360                         sc->vmsize = CAST_DOWN(vm_size_t,vmsize);
 361                         sc->fileoff = CAST_DOWN(uint32_t,foffset);
 362                         sc->filesize = CAST_DOWN(uint32_t,vmsize);
 363                         sc->maxprot = maxprot;
 364                         sc->initprot = prot;
 365                         sc->nsects = 0;
 366                 }
 367
 368                 /*
 369                  *      Write segment out.  Try as hard as possible to
 370                  *      get read access to the data.
 371                  */
 372                 if ((prot & VM_PROT_READ) == 0) {
 373                         mach_vm_protect(map, vmoffset, vmsize, FALSE,
 374                                            prot|VM_PROT_READ);
 375                 }
 376                 /*
 377                  *      Only actually perform write if we can read.
 378                  *      Note: if we can't read, then we end up with
 379                  *      a hole in the file.
 380                  */
 381                 if ((maxprot & VM_PROT_READ) == VM_PROT_READ
 382                         && vbr.user_tag != VM_MEMORY_IOKIT
 383                         && coredumpok(map,vmoffset)) {
 384                         vm_map_size_t   tmp_vmsize = vmsize;
 385                         off_t           xfer_foffset = foffset;
 386
 387                         //LP64todo - works around vn_rdwr_64() 2G limit
 388                         while (tmp_vmsize > 0) {
 389                                 vm_map_size_t   xfer_vmsize = tmp_vmsize;
 390                                 if (xfer_vmsize > INT_MAX)
 391                                         xfer_vmsize = INT_MAX;
 392                                 error = vn_rdwr_64(UIO_WRITE, vp,
 393                                                 vmoffset, xfer_vmsize, xfer_foffset,
 394                                         (IS_64BIT_PROCESS(p) ? UIO_USERSPACE64 : UIO_USERSPACE32),
 395                                         IO_NODELOCKED|IO_UNIT, cred, (int *) 0, p);
 396                                 tmp_vmsize -= xfer_vmsize;
 397                                 xfer_foffset += xfer_vmsize;
 398                         }
 399                 }
 400
 401                 hoffset += segment_command_sz;
 402                 foffset += vmsize;
 403                 vmoffset += vmsize;
 404                 segment_count--;
 405         }
 406
 407         /*
 408          * If there are remaining segments which have not been written
 409          * out because break in the loop above, then they were not counted
 410          * because they exceed the real address space of the executable
 411          * type: remove them from the header's count.  This is OK, since
 412          * we are allowed to have a sparse area following the segments.
 413          */
 414         if (is_64) {
 415                 mh64->ncmds -= segment_count;
 416         } else {
 417                 mh->ncmds -= segment_count;
 418         }
 419
 420         tir1.header = header;
 421         tir1.hoffset = hoffset;
 422         tir1.flavors = flavors;
 423         tir1.tstate_size = tstate_size;
 424         task_act_iterate_wth_args(task, collectth_state,&tir1);
 425
 426         /*
 427          *      Write out the Mach header at the beginning of the
 428          *      file.  OK to use a 32 bit write for this.
 429          */
 430         error = vn_rdwr(UIO_WRITE, vp, (caddr_t)header, header_size, (off_t)0,
 431                         UIO_SYSSPACE32, IO_NODELOCKED|IO_UNIT, cred, (int *) 0, p);
 432         kmem_free(kernel_map, header, header_size);
 433 out:
 434         error1 = vnode_close(vp, FWRITE, &context);
 435         if (error == 0)
 436                 error = error1;
 437
 438         return (error);
 439 }