xnu-792.1.5.tar.gz

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

diff --git a/bsd/kern/kern_core.c b/bsd/kern/kern_core.c
index 8637468fb1853e632eb82e7bd070bf8cc337bd2a..d17444fd62c7b2c58d353586eeb41ffdbc38aef5 100644 (file)
--- a/bsd/kern/kern_core.c
+++ b/bsd/kern/kern_core.c
@@ -1,5 +1,5 @@
 /*
 /*

- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.

  *
  * @APPLE_LICENSE_HEADER_START@
  * 
  *
  * @APPLE_LICENSE_HEADER_START@
  * 


@@ -25,9 +25,6 @@
  *
  *     This file contains machine independent code for performing core dumps.
  *
  *
  *     This file contains machine independent code for performing core dumps.
  *

- * HISTORY
- * 16-Feb-91  Mike DeMoney (mike@next.com)
- *     Massaged into MI form from m68k/core.c.

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


@@ -38,21 +35,27 @@
 #include <sys/signalvar.h>
 #include <sys/resourcevar.h>
 #include <sys/namei.h>
 #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/timeb.h>
 #include <sys/times.h>

-#include <sys/buf.h>

 #include <sys/acct.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-o/loader.h>
 #include <mach/vm_region.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_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 */
 
 typedef struct {
        int     flavor;                 /* the number for this flavor */


@@ -64,9 +67,10 @@ typedef struct {
 mythread_state_flavor_t thread_flavor_array[]={
                {PPC_THREAD_STATE , PPC_THREAD_STATE_COUNT},
                {PPC_FLOAT_STATE, PPC_FLOAT_STATE_COUNT}, 
 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_EXCEPTION_STATE, PPC_EXCEPTION_STATE_COUNT},
+               {PPC_VECTOR_STATE, PPC_VECTOR_STATE_COUNT}

                };
                };

-int mynum_flavors=3;
+int mynum_flavors=4;

 #elif defined (__i386__)
 mythread_state_flavor_t thread_flavor_array [] = { 
                {i386_THREAD_STATE, i386_THREAD_STATE_COUNT},
 #elif defined (__i386__)
 mythread_state_flavor_t thread_flavor_array [] = { 
                {i386_THREAD_STATE, i386_THREAD_STATE_COUNT},


@@ -93,12 +97,27 @@ typedef struct {
        int tstate_size;
 } tir_t;
 
        int tstate_size;
 } tir_t;
 

-collectth_state(thread_act_t th_act, tir_t *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 *);
+
+
+__private_extern__ int do_coredump = 1;        /* default: dump cores */
+__private_extern__ int sugid_coredump = 0; /* default: but not SGUID binaries */
+
+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;
 {
        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.
                 */
                /*
                 *      Fill in thread command structure.
                 */


@@ -121,45 +140,42 @@ collectth_state(thread_act_t th_act, tir_t *t)
                          flavors[i];
                        hoffset += sizeof(mythread_state_flavor_t);
                        thread_getstatus(th_act, flavors[i].flavor,
                          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;
 }
                                        &flavors[i].count);
                        hoffset += flavors[i].count*sizeof(int);
                }
 
                t->hoffset = hoffset;
 }

+

 /*
  * Create a core image on the file "core".
  */
 #define        MAX_TSTATE_FLAVORS      10
 int
 /*
  * Create a core image on the file "core".
  */
 #define        MAX_TSTATE_FLAVORS      10
 int

-coredump(p)
-       register struct proc *p;
+coredump(struct proc *p)

 {
        int error=0;
 {
        int error=0;

-       register struct pcred *pcred = p->p_cred;
-       register struct ucred *cred = pcred->pc_ucred;
-       struct nameidata nd;
-       struct vattr    vattr;
+       kauth_cred_t cred = kauth_cred_get();
+       struct vnode_attr va;
+       struct vfs_context context;

        vm_map_t        map;
        int             thread_count, segment_count;
        int             command_size, header_size, tstate_size;
        vm_map_t        map;
        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;
        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_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;
        task_t          task;
        char            core_name[MAXCOMLEN+6];

+       char            *name;

        mythread_state_flavor_t flavors[MAX_TSTATE_FLAVORS];
        mythread_state_flavor_t flavors[MAX_TSTATE_FLAVORS];

-       vm_size_t       nflavors,mapsize;
+       vm_size_t       mapsize;

        int             i;
        int nesting_depth = 0;
        kern_return_t   kret;
        int             i;
        int nesting_depth = 0;
        kern_return_t   kret;


@@ -167,10 +183,25 @@ coredump(p)
        int vbrcount=0;
        tir_t tir1;
        struct vnode * vp;
        int vbrcount=0;
        tir_t tir1;
        struct vnode * vp;

+       struct mach_header      *mh;
+       struct mach_header_64   *mh64;
+       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);
                return (EFAULT);

+       }
+
+       if (IS_64BIT_PROCESS(p)) {
+               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();
 
        task = current_task();
        map = current_map();


@@ -180,31 +211,32 @@ coredump(p)
                return (EFAULT);
        (void) task_suspend(task);
 
                return (EFAULT);
        (void) task_suspend(task);
 

-       /*
-        *      Make sure all registers, etc. are in pcb so they get
-        *      into core file.
-        */
-#if defined (__ppc__)
-       fpu_save(current_act());
-       vec_save(current_act());
-#endif
-       sprintf(core_name, "/cores/core.%d", p->p_pid);
-       NDINIT(&nd, LOOKUP, FOLLOW, UIO_SYSSPACE, core_name, p);
-       if(error = vn_open(&nd, O_CREAT | FWRITE, S_IRUSR ))
-               return (error);
-       vp = nd.ni_vp;
-       
+       /* create name according to sysctl'able format string */
+       name = proc_core_name(p->p_comm, kauth_cred_getuid(cred), p->p_pid);
+
+       /* if name creation fails, fall back to historical behaviour... */
+       if (name == NULL) {
+               sprintf(core_name, "/cores/core.%d", p->p_pid);
+               name = core_name;
+       }
+       context.vc_proc = p;
+       context.vc_ucred = cred;
+
+       if ((error = vnode_open(name, (O_CREAT | FWRITE | O_NOFOLLOW), S_IRUSR, 0, &vp, &context)))
+               return (error);
+
+       VATTR_INIT(&va);
+       VATTR_WANTED(&va, va_nlink);

        /* Don't dump to non-regular files or files with links. */
        if (vp->v_type != VREG ||
        /* 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, &context) || va.va_nlink != 1) {

                error = EFAULT;
                goto out;
        }
 
                error = EFAULT;
                goto out;
        }
 

-       VATTR_NULL(&vattr);
-       vattr.va_size = 0;
-       VOP_LEASE(vp, p, cred, LEASE_WRITE);
-       VOP_SETATTR(vp, &vattr, cred, p);
+       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, &context);

        p->p_acflag |= ACORE;
 
        /*
        p->p_acflag |= ACORE;
 
        /*


@@ -215,65 +247,77 @@ coredump(p)
 
        thread_count = get_task_numacts(task);
        segment_count = get_vmmap_entries(map); /* XXX */
 
        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
-        */
-#if 0
-       nflavors = sizeof(flavors)/sizeof(int);
-       if (thread_getstatus(current_thread(), THREAD_STATE_FLAVOR_LIST,
-                               (thread_state_t)(flavors),
-                                &nflavors) != KERN_SUCCESS)
-           panic("core flavor list");
-       /* now convert to number of flavors */
-       nflavors /= sizeof(mythread_state_flavor_t)/sizeof(int);
-#else
-       nflavors = mynum_flavors;

        bcopy(thread_flavor_array,flavors,sizeof(thread_flavor_array));
        bcopy(thread_flavor_array,flavors,sizeof(thread_flavor_array));

-#endif

        tstate_size = 0;
        tstate_size = 0;

-       for (i = 0; i < nflavors; i++)
+       for (i = 0; i < mynum_flavors; i++)

                tstate_size += sizeof(mythread_state_flavor_t) +
                  (flavors[i].count * sizeof(int));
 
                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;
 
          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.
         */
                                    (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 */
+       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 */
        foffset = round_page(header_size);      /* offset into file */

-       vmoffset = VM_MIN_ADDRESS;              /* offset into VM */
-       /* We use to check for an error, here, now we try and get 
+       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
         */
         * 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;
                /*
                 *      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) {
                                break;
                        }
                        if(vbr.is_submap) {


@@ -292,90 +336,103 @@ coredump(p)
                /*
                 *      Fill in segment command structure.
                 */
                /*
                 *      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->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) {
 
                /*
                 *      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.
                 *      Note: if we can't read, then we end up with
                 *      a hole in the file.
                 */
                }
                /*
                 *      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) {
-                       error = vn_rdwr(UIO_WRITE, vp, (caddr_t)vmoffset, size, foffset,
-                               UIO_USERSPACE, IO_NODELOCKED|IO_UNIT, cred, (int *) 0, p);
+               if ((maxprot & VM_PROT_READ) == VM_PROT_READ
+                       && vbr.user_tag != VM_MEMORY_IOKIT
+                       && coredumpok(map,vmoffset)) {
+                       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(p) ? UIO_USERSPACE64 : UIO_USERSPACE32), 
+                                       IO_NODELOCKED|IO_UNIT, cred, (int *) 0, p);
+                               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--;
        }
 
                segment_count--;
        }
 

-#if 0 /* [ */
-       task_lock(task);
-       thread = (thread_t) queue_first(&task->thread_list);
-       while (thread_count > 0) {
-               /*
-                *      Fill in thread command structure.
-                */
-               tc = (struct thread_command *) (header + hoffset);
-               tc->cmd = LC_THREAD;
-               tc->cmdsize = sizeof(struct thread_command)
-                               + 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 < nflavors; i++) {
-                       *(mythread_state_flavor_t *)(header+hoffset) =
-                         flavors[i];
-                       hoffset += sizeof(mythread_state_flavor_t);
-                       thread_getstatus(thread, flavors[i].flavor,
-                                       (thread_state_t *)(header+hoffset),
-                                       &flavors[i].count);
-                       hoffset += flavors[i].count*sizeof(int);
-               }
-               thread = (thread_t) queue_next(&thread->thread_list);
-               thread_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;
+       } else {
+               mh->ncmds -= segment_count;

        }
        }

-       task_unlock(task);
-#else /* /* 0 ][ */
+

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

-#endif /* 0 ] */

        /*
         *      Write out the Mach header at the beginning of the
        /*
         *      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,
         */
        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, p);

        kmem_free(kernel_map, header, header_size);
 out:
        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, &context);

        if (error == 0)
                error = error1;
        if (error == 0)
                error = error1;

+
+       return (error);

 }
 }