The Design and Implementation of the FreeBSD Operating System, Second Edition
Now available: The Design and Implementation of the FreeBSD Operating System (Second Edition)


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]

FreeBSD/Linux Kernel Cross Reference
sys/kern/sys_process.c

Version: -  FREEBSD  -  FREEBSD-12-STABLE  -  FREEBSD-12-0  -  FREEBSD-11-STABLE  -  FREEBSD-11-2  -  FREEBSD-11-1  -  FREEBSD-11-0  -  FREEBSD-10-STABLE  -  FREEBSD-10-4  -  FREEBSD-10-3  -  FREEBSD-10-2  -  FREEBSD-10-1  -  FREEBSD-10-0  -  FREEBSD-9-STABLE  -  FREEBSD-9-3  -  FREEBSD-9-2  -  FREEBSD-9-1  -  FREEBSD-9-0  -  FREEBSD-8-STABLE  -  FREEBSD-8-4  -  FREEBSD-8-3  -  FREEBSD-8-2  -  FREEBSD-8-1  -  FREEBSD-8-0  -  FREEBSD-7-STABLE  -  FREEBSD-7-4  -  FREEBSD-7-3  -  FREEBSD-7-2  -  FREEBSD-7-1  -  FREEBSD-7-0  -  FREEBSD-6-STABLE  -  FREEBSD-6-4  -  FREEBSD-6-3  -  FREEBSD-6-2  -  FREEBSD-6-1  -  FREEBSD-6-0  -  FREEBSD-5-STABLE  -  FREEBSD-5-5  -  FREEBSD-5-4  -  FREEBSD-5-3  -  FREEBSD-5-2  -  FREEBSD-5-1  -  FREEBSD-5-0  -  FREEBSD-4-STABLE  -  FREEBSD-3-STABLE  -  FREEBSD22  -  linux-2.6  -  linux-2.4.22  -  MK83  -  MK84  -  PLAN9  -  DFBSD  -  NETBSD  -  NETBSD5  -  NETBSD4  -  NETBSD3  -  NETBSD20  -  OPENBSD  -  xnu-517  -  xnu-792  -  xnu-792.6.70  -  xnu-1228  -  xnu-1456.1.26  -  xnu-1699.24.8  -  xnu-2050.18.24  -  OPENSOLARIS  -  minix-3-1-1 
SearchContext: -  none  -  3  -  10 

    1 /*-
    2  * Copyright (c) 1994, Sean Eric Fagan
    3  * All rights reserved.
    4  *
    5  * Redistribution and use in source and binary forms, with or without
    6  * modification, are permitted provided that the following conditions
    7  * are met:
    8  * 1. Redistributions of source code must retain the above copyright
    9  *    notice, this list of conditions and the following disclaimer.
   10  * 2. Redistributions in binary form must reproduce the above copyright
   11  *    notice, this list of conditions and the following disclaimer in the
   12  *    documentation and/or other materials provided with the distribution.
   13  * 3. All advertising materials mentioning features or use of this software
   14  *    must display the following acknowledgement:
   15  *      This product includes software developed by Sean Eric Fagan.
   16  * 4. The name of the author may not be used to endorse or promote products
   17  *    derived from this software without specific prior written permission.
   18  *
   19  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
   20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
   21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
   22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
   23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
   24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
   25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
   26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
   27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
   28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
   29  * SUCH DAMAGE.
   30  */
   31 
   32 #include <sys/cdefs.h>
   33 __FBSDID("$FreeBSD: releng/10.1/sys/kern/sys_process.c 270264 2014-08-21 10:46:19Z kib $");
   34 
   35 #include "opt_compat.h"
   36 
   37 #include <sys/param.h>
   38 #include <sys/systm.h>
   39 #include <sys/lock.h>
   40 #include <sys/mutex.h>
   41 #include <sys/syscallsubr.h>
   42 #include <sys/sysent.h>
   43 #include <sys/sysproto.h>
   44 #include <sys/priv.h>
   45 #include <sys/proc.h>
   46 #include <sys/procctl.h>
   47 #include <sys/vnode.h>
   48 #include <sys/ptrace.h>
   49 #include <sys/rwlock.h>
   50 #include <sys/sx.h>
   51 #include <sys/malloc.h>
   52 #include <sys/signalvar.h>
   53 
   54 #include <machine/reg.h>
   55 
   56 #include <security/audit/audit.h>
   57 
   58 #include <vm/vm.h>
   59 #include <vm/pmap.h>
   60 #include <vm/vm_extern.h>
   61 #include <vm/vm_map.h>
   62 #include <vm/vm_kern.h>
   63 #include <vm/vm_object.h>
   64 #include <vm/vm_page.h>
   65 #include <vm/vm_param.h>
   66 
   67 #ifdef COMPAT_FREEBSD32
   68 #include <sys/procfs.h>
   69 #include <compat/freebsd32/freebsd32_signal.h>
   70 
   71 struct ptrace_io_desc32 {
   72         int             piod_op;
   73         uint32_t        piod_offs;
   74         uint32_t        piod_addr;
   75         uint32_t        piod_len;
   76 };
   77 
   78 struct ptrace_vm_entry32 {
   79         int             pve_entry;
   80         int             pve_timestamp;
   81         uint32_t        pve_start;
   82         uint32_t        pve_end;
   83         uint32_t        pve_offset;
   84         u_int           pve_prot;
   85         u_int           pve_pathlen;
   86         int32_t         pve_fileid;
   87         u_int           pve_fsid;
   88         uint32_t        pve_path;
   89 };
   90 
   91 struct ptrace_lwpinfo32 {
   92         lwpid_t pl_lwpid;       /* LWP described. */
   93         int     pl_event;       /* Event that stopped the LWP. */
   94         int     pl_flags;       /* LWP flags. */
   95         sigset_t        pl_sigmask;     /* LWP signal mask */
   96         sigset_t        pl_siglist;     /* LWP pending signal */
   97         struct siginfo32 pl_siginfo;    /* siginfo for signal */
   98         char    pl_tdname[MAXCOMLEN + 1];       /* LWP name. */
   99         int     pl_child_pid;           /* New child pid */
  100 };
  101 
  102 #endif
  103 
  104 /*
  105  * Functions implemented using PROC_ACTION():
  106  *
  107  * proc_read_regs(proc, regs)
  108  *      Get the current user-visible register set from the process
  109  *      and copy it into the regs structure (<machine/reg.h>).
  110  *      The process is stopped at the time read_regs is called.
  111  *
  112  * proc_write_regs(proc, regs)
  113  *      Update the current register set from the passed in regs
  114  *      structure.  Take care to avoid clobbering special CPU
  115  *      registers or privileged bits in the PSL.
  116  *      Depending on the architecture this may have fix-up work to do,
  117  *      especially if the IAR or PCW are modified.
  118  *      The process is stopped at the time write_regs is called.
  119  *
  120  * proc_read_fpregs, proc_write_fpregs
  121  *      deal with the floating point register set, otherwise as above.
  122  *
  123  * proc_read_dbregs, proc_write_dbregs
  124  *      deal with the processor debug register set, otherwise as above.
  125  *
  126  * proc_sstep(proc)
  127  *      Arrange for the process to trap after executing a single instruction.
  128  */
  129 
  130 #define PROC_ACTION(action) do {                                        \
  131         int error;                                                      \
  132                                                                         \
  133         PROC_LOCK_ASSERT(td->td_proc, MA_OWNED);                        \
  134         if ((td->td_proc->p_flag & P_INMEM) == 0)                       \
  135                 error = EIO;                                            \
  136         else                                                            \
  137                 error = (action);                                       \
  138         return (error);                                                 \
  139 } while(0)
  140 
  141 int
  142 proc_read_regs(struct thread *td, struct reg *regs)
  143 {
  144 
  145         PROC_ACTION(fill_regs(td, regs));
  146 }
  147 
  148 int
  149 proc_write_regs(struct thread *td, struct reg *regs)
  150 {
  151 
  152         PROC_ACTION(set_regs(td, regs));
  153 }
  154 
  155 int
  156 proc_read_dbregs(struct thread *td, struct dbreg *dbregs)
  157 {
  158 
  159         PROC_ACTION(fill_dbregs(td, dbregs));
  160 }
  161 
  162 int
  163 proc_write_dbregs(struct thread *td, struct dbreg *dbregs)
  164 {
  165 
  166         PROC_ACTION(set_dbregs(td, dbregs));
  167 }
  168 
  169 /*
  170  * Ptrace doesn't support fpregs at all, and there are no security holes
  171  * or translations for fpregs, so we can just copy them.
  172  */
  173 int
  174 proc_read_fpregs(struct thread *td, struct fpreg *fpregs)
  175 {
  176 
  177         PROC_ACTION(fill_fpregs(td, fpregs));
  178 }
  179 
  180 int
  181 proc_write_fpregs(struct thread *td, struct fpreg *fpregs)
  182 {
  183 
  184         PROC_ACTION(set_fpregs(td, fpregs));
  185 }
  186 
  187 #ifdef COMPAT_FREEBSD32
  188 /* For 32 bit binaries, we need to expose the 32 bit regs layouts. */
  189 int
  190 proc_read_regs32(struct thread *td, struct reg32 *regs32)
  191 {
  192 
  193         PROC_ACTION(fill_regs32(td, regs32));
  194 }
  195 
  196 int
  197 proc_write_regs32(struct thread *td, struct reg32 *regs32)
  198 {
  199 
  200         PROC_ACTION(set_regs32(td, regs32));
  201 }
  202 
  203 int
  204 proc_read_dbregs32(struct thread *td, struct dbreg32 *dbregs32)
  205 {
  206 
  207         PROC_ACTION(fill_dbregs32(td, dbregs32));
  208 }
  209 
  210 int
  211 proc_write_dbregs32(struct thread *td, struct dbreg32 *dbregs32)
  212 {
  213 
  214         PROC_ACTION(set_dbregs32(td, dbregs32));
  215 }
  216 
  217 int
  218 proc_read_fpregs32(struct thread *td, struct fpreg32 *fpregs32)
  219 {
  220 
  221         PROC_ACTION(fill_fpregs32(td, fpregs32));
  222 }
  223 
  224 int
  225 proc_write_fpregs32(struct thread *td, struct fpreg32 *fpregs32)
  226 {
  227 
  228         PROC_ACTION(set_fpregs32(td, fpregs32));
  229 }
  230 #endif
  231 
  232 int
  233 proc_sstep(struct thread *td)
  234 {
  235 
  236         PROC_ACTION(ptrace_single_step(td));
  237 }
  238 
  239 int
  240 proc_rwmem(struct proc *p, struct uio *uio)
  241 {
  242         vm_map_t map;
  243         vm_offset_t pageno;             /* page number */
  244         vm_prot_t reqprot;
  245         int error, fault_flags, page_offset, writing;
  246 
  247         /*
  248          * Assert that someone has locked this vmspace.  (Should be
  249          * curthread but we can't assert that.)  This keeps the process
  250          * from exiting out from under us until this operation completes.
  251          */
  252         KASSERT(p->p_lock >= 1, ("%s: process %p (pid %d) not held", __func__,
  253             p, p->p_pid));
  254 
  255         /*
  256          * The map we want...
  257          */
  258         map = &p->p_vmspace->vm_map;
  259 
  260         /*
  261          * If we are writing, then we request vm_fault() to create a private
  262          * copy of each page.  Since these copies will not be writeable by the
  263          * process, we must explicity request that they be dirtied.
  264          */
  265         writing = uio->uio_rw == UIO_WRITE;
  266         reqprot = writing ? VM_PROT_COPY | VM_PROT_READ : VM_PROT_READ;
  267         fault_flags = writing ? VM_FAULT_DIRTY : VM_FAULT_NORMAL;
  268 
  269         /*
  270          * Only map in one page at a time.  We don't have to, but it
  271          * makes things easier.  This way is trivial - right?
  272          */
  273         do {
  274                 vm_offset_t uva;
  275                 u_int len;
  276                 vm_page_t m;
  277 
  278                 uva = (vm_offset_t)uio->uio_offset;
  279 
  280                 /*
  281                  * Get the page number of this segment.
  282                  */
  283                 pageno = trunc_page(uva);
  284                 page_offset = uva - pageno;
  285 
  286                 /*
  287                  * How many bytes to copy
  288                  */
  289                 len = min(PAGE_SIZE - page_offset, uio->uio_resid);
  290 
  291                 /*
  292                  * Fault and hold the page on behalf of the process.
  293                  */
  294                 error = vm_fault_hold(map, pageno, reqprot, fault_flags, &m);
  295                 if (error != KERN_SUCCESS) {
  296                         if (error == KERN_RESOURCE_SHORTAGE)
  297                                 error = ENOMEM;
  298                         else
  299                                 error = EFAULT;
  300                         break;
  301                 }
  302 
  303                 /*
  304                  * Now do the i/o move.
  305                  */
  306                 error = uiomove_fromphys(&m, page_offset, len, uio);
  307 
  308                 /* Make the I-cache coherent for breakpoints. */
  309                 if (writing && error == 0) {
  310                         vm_map_lock_read(map);
  311                         if (vm_map_check_protection(map, pageno, pageno +
  312                             PAGE_SIZE, VM_PROT_EXECUTE))
  313                                 vm_sync_icache(map, uva, len);
  314                         vm_map_unlock_read(map);
  315                 }
  316 
  317                 /*
  318                  * Release the page.
  319                  */
  320                 vm_page_lock(m);
  321                 vm_page_unhold(m);
  322                 vm_page_unlock(m);
  323 
  324         } while (error == 0 && uio->uio_resid > 0);
  325 
  326         return (error);
  327 }
  328 
  329 static int
  330 ptrace_vm_entry(struct thread *td, struct proc *p, struct ptrace_vm_entry *pve)
  331 {
  332         struct vattr vattr;
  333         vm_map_t map;
  334         vm_map_entry_t entry;
  335         vm_object_t obj, tobj, lobj;
  336         struct vmspace *vm;
  337         struct vnode *vp;
  338         char *freepath, *fullpath;
  339         u_int pathlen;
  340         int error, index;
  341 
  342         error = 0;
  343         obj = NULL;
  344 
  345         vm = vmspace_acquire_ref(p);
  346         map = &vm->vm_map;
  347         vm_map_lock_read(map);
  348 
  349         do {
  350                 entry = map->header.next;
  351                 index = 0;
  352                 while (index < pve->pve_entry && entry != &map->header) {
  353                         entry = entry->next;
  354                         index++;
  355                 }
  356                 if (index != pve->pve_entry) {
  357                         error = EINVAL;
  358                         break;
  359                 }
  360                 while (entry != &map->header &&
  361                     (entry->eflags & MAP_ENTRY_IS_SUB_MAP) != 0) {
  362                         entry = entry->next;
  363                         index++;
  364                 }
  365                 if (entry == &map->header) {
  366                         error = ENOENT;
  367                         break;
  368                 }
  369 
  370                 /* We got an entry. */
  371                 pve->pve_entry = index + 1;
  372                 pve->pve_timestamp = map->timestamp;
  373                 pve->pve_start = entry->start;
  374                 pve->pve_end = entry->end - 1;
  375                 pve->pve_offset = entry->offset;
  376                 pve->pve_prot = entry->protection;
  377 
  378                 /* Backing object's path needed? */
  379                 if (pve->pve_pathlen == 0)
  380                         break;
  381 
  382                 pathlen = pve->pve_pathlen;
  383                 pve->pve_pathlen = 0;
  384 
  385                 obj = entry->object.vm_object;
  386                 if (obj != NULL)
  387                         VM_OBJECT_RLOCK(obj);
  388         } while (0);
  389 
  390         vm_map_unlock_read(map);
  391         vmspace_free(vm);
  392 
  393         pve->pve_fsid = VNOVAL;
  394         pve->pve_fileid = VNOVAL;
  395 
  396         if (error == 0 && obj != NULL) {
  397                 lobj = obj;
  398                 for (tobj = obj; tobj != NULL; tobj = tobj->backing_object) {
  399                         if (tobj != obj)
  400                                 VM_OBJECT_RLOCK(tobj);
  401                         if (lobj != obj)
  402                                 VM_OBJECT_RUNLOCK(lobj);
  403                         lobj = tobj;
  404                         pve->pve_offset += tobj->backing_object_offset;
  405                 }
  406                 vp = (lobj->type == OBJT_VNODE) ? lobj->handle : NULL;
  407                 if (vp != NULL)
  408                         vref(vp);
  409                 if (lobj != obj)
  410                         VM_OBJECT_RUNLOCK(lobj);
  411                 VM_OBJECT_RUNLOCK(obj);
  412 
  413                 if (vp != NULL) {
  414                         freepath = NULL;
  415                         fullpath = NULL;
  416                         vn_fullpath(td, vp, &fullpath, &freepath);
  417                         vn_lock(vp, LK_SHARED | LK_RETRY);
  418                         if (VOP_GETATTR(vp, &vattr, td->td_ucred) == 0) {
  419                                 pve->pve_fileid = vattr.va_fileid;
  420                                 pve->pve_fsid = vattr.va_fsid;
  421                         }
  422                         vput(vp);
  423 
  424                         if (fullpath != NULL) {
  425                                 pve->pve_pathlen = strlen(fullpath) + 1;
  426                                 if (pve->pve_pathlen <= pathlen) {
  427                                         error = copyout(fullpath, pve->pve_path,
  428                                             pve->pve_pathlen);
  429                                 } else
  430                                         error = ENAMETOOLONG;
  431                         }
  432                         if (freepath != NULL)
  433                                 free(freepath, M_TEMP);
  434                 }
  435         }
  436 
  437         return (error);
  438 }
  439 
  440 #ifdef COMPAT_FREEBSD32
  441 static int      
  442 ptrace_vm_entry32(struct thread *td, struct proc *p,
  443     struct ptrace_vm_entry32 *pve32)
  444 {
  445         struct ptrace_vm_entry pve;
  446         int error;
  447 
  448         pve.pve_entry = pve32->pve_entry;
  449         pve.pve_pathlen = pve32->pve_pathlen;
  450         pve.pve_path = (void *)(uintptr_t)pve32->pve_path;
  451 
  452         error = ptrace_vm_entry(td, p, &pve);
  453         if (error == 0) {
  454                 pve32->pve_entry = pve.pve_entry;
  455                 pve32->pve_timestamp = pve.pve_timestamp;
  456                 pve32->pve_start = pve.pve_start;
  457                 pve32->pve_end = pve.pve_end;
  458                 pve32->pve_offset = pve.pve_offset;
  459                 pve32->pve_prot = pve.pve_prot;
  460                 pve32->pve_fileid = pve.pve_fileid;
  461                 pve32->pve_fsid = pve.pve_fsid;
  462         }
  463 
  464         pve32->pve_pathlen = pve.pve_pathlen;
  465         return (error);
  466 }
  467 
  468 static void
  469 ptrace_lwpinfo_to32(const struct ptrace_lwpinfo *pl,
  470     struct ptrace_lwpinfo32 *pl32)
  471 {
  472 
  473         pl32->pl_lwpid = pl->pl_lwpid;
  474         pl32->pl_event = pl->pl_event;
  475         pl32->pl_flags = pl->pl_flags;
  476         pl32->pl_sigmask = pl->pl_sigmask;
  477         pl32->pl_siglist = pl->pl_siglist;
  478         siginfo_to_siginfo32(&pl->pl_siginfo, &pl32->pl_siginfo);
  479         strcpy(pl32->pl_tdname, pl->pl_tdname);
  480         pl32->pl_child_pid = pl->pl_child_pid;
  481 }
  482 #endif /* COMPAT_FREEBSD32 */
  483 
  484 /*
  485  * Process debugging system call.
  486  */
  487 #ifndef _SYS_SYSPROTO_H_
  488 struct ptrace_args {
  489         int     req;
  490         pid_t   pid;
  491         caddr_t addr;
  492         int     data;
  493 };
  494 #endif
  495 
  496 #ifdef COMPAT_FREEBSD32
  497 /*
  498  * This CPP subterfuge is to try and reduce the number of ifdefs in
  499  * the body of the code.
  500  *   COPYIN(uap->addr, &r.reg, sizeof r.reg);
  501  * becomes either:
  502  *   copyin(uap->addr, &r.reg, sizeof r.reg);
  503  * or
  504  *   copyin(uap->addr, &r.reg32, sizeof r.reg32);
  505  * .. except this is done at runtime.
  506  */
  507 #define COPYIN(u, k, s)         wrap32 ? \
  508         copyin(u, k ## 32, s ## 32) : \
  509         copyin(u, k, s)
  510 #define COPYOUT(k, u, s)        wrap32 ? \
  511         copyout(k ## 32, u, s ## 32) : \
  512         copyout(k, u, s)
  513 #else
  514 #define COPYIN(u, k, s)         copyin(u, k, s)
  515 #define COPYOUT(k, u, s)        copyout(k, u, s)
  516 #endif
  517 int
  518 sys_ptrace(struct thread *td, struct ptrace_args *uap)
  519 {
  520         /*
  521          * XXX this obfuscation is to reduce stack usage, but the register
  522          * structs may be too large to put on the stack anyway.
  523          */
  524         union {
  525                 struct ptrace_io_desc piod;
  526                 struct ptrace_lwpinfo pl;
  527                 struct ptrace_vm_entry pve;
  528                 struct dbreg dbreg;
  529                 struct fpreg fpreg;
  530                 struct reg reg;
  531 #ifdef COMPAT_FREEBSD32
  532                 struct dbreg32 dbreg32;
  533                 struct fpreg32 fpreg32;
  534                 struct reg32 reg32;
  535                 struct ptrace_io_desc32 piod32;
  536                 struct ptrace_lwpinfo32 pl32;
  537                 struct ptrace_vm_entry32 pve32;
  538 #endif
  539         } r;
  540         void *addr;
  541         int error = 0;
  542 #ifdef COMPAT_FREEBSD32
  543         int wrap32 = 0;
  544 
  545         if (SV_CURPROC_FLAG(SV_ILP32))
  546                 wrap32 = 1;
  547 #endif
  548         AUDIT_ARG_PID(uap->pid);
  549         AUDIT_ARG_CMD(uap->req);
  550         AUDIT_ARG_VALUE(uap->data);
  551         addr = &r;
  552         switch (uap->req) {
  553         case PT_GETREGS:
  554         case PT_GETFPREGS:
  555         case PT_GETDBREGS:
  556         case PT_LWPINFO:
  557                 break;
  558         case PT_SETREGS:
  559                 error = COPYIN(uap->addr, &r.reg, sizeof r.reg);
  560                 break;
  561         case PT_SETFPREGS:
  562                 error = COPYIN(uap->addr, &r.fpreg, sizeof r.fpreg);
  563                 break;
  564         case PT_SETDBREGS:
  565                 error = COPYIN(uap->addr, &r.dbreg, sizeof r.dbreg);
  566                 break;
  567         case PT_IO:
  568                 error = COPYIN(uap->addr, &r.piod, sizeof r.piod);
  569                 break;
  570         case PT_VM_ENTRY:
  571                 error = COPYIN(uap->addr, &r.pve, sizeof r.pve);
  572                 break;
  573         default:
  574                 addr = uap->addr;
  575                 break;
  576         }
  577         if (error)
  578                 return (error);
  579 
  580         error = kern_ptrace(td, uap->req, uap->pid, addr, uap->data);
  581         if (error)
  582                 return (error);
  583 
  584         switch (uap->req) {
  585         case PT_VM_ENTRY:
  586                 error = COPYOUT(&r.pve, uap->addr, sizeof r.pve);
  587                 break;
  588         case PT_IO:
  589                 error = COPYOUT(&r.piod, uap->addr, sizeof r.piod);
  590                 break;
  591         case PT_GETREGS:
  592                 error = COPYOUT(&r.reg, uap->addr, sizeof r.reg);
  593                 break;
  594         case PT_GETFPREGS:
  595                 error = COPYOUT(&r.fpreg, uap->addr, sizeof r.fpreg);
  596                 break;
  597         case PT_GETDBREGS:
  598                 error = COPYOUT(&r.dbreg, uap->addr, sizeof r.dbreg);
  599                 break;
  600         case PT_LWPINFO:
  601                 error = copyout(&r.pl, uap->addr, uap->data);
  602                 break;
  603         }
  604 
  605         return (error);
  606 }
  607 #undef COPYIN
  608 #undef COPYOUT
  609 
  610 #ifdef COMPAT_FREEBSD32
  611 /*
  612  *   PROC_READ(regs, td2, addr);
  613  * becomes either:
  614  *   proc_read_regs(td2, addr);
  615  * or
  616  *   proc_read_regs32(td2, addr);
  617  * .. except this is done at runtime.  There is an additional
  618  * complication in that PROC_WRITE disallows 32 bit consumers
  619  * from writing to 64 bit address space targets.
  620  */
  621 #define PROC_READ(w, t, a)      wrap32 ? \
  622         proc_read_ ## w ## 32(t, a) : \
  623         proc_read_ ## w (t, a)
  624 #define PROC_WRITE(w, t, a)     wrap32 ? \
  625         (safe ? proc_write_ ## w ## 32(t, a) : EINVAL ) : \
  626         proc_write_ ## w (t, a)
  627 #else
  628 #define PROC_READ(w, t, a)      proc_read_ ## w (t, a)
  629 #define PROC_WRITE(w, t, a)     proc_write_ ## w (t, a)
  630 #endif
  631 
  632 int
  633 kern_ptrace(struct thread *td, int req, pid_t pid, void *addr, int data)
  634 {
  635         struct iovec iov;
  636         struct uio uio;
  637         struct proc *curp, *p, *pp;
  638         struct thread *td2 = NULL, *td3;
  639         struct ptrace_io_desc *piod = NULL;
  640         struct ptrace_lwpinfo *pl;
  641         int error, write, tmp, num;
  642         int proctree_locked = 0;
  643         lwpid_t tid = 0, *buf;
  644 #ifdef COMPAT_FREEBSD32
  645         int wrap32 = 0, safe = 0;
  646         struct ptrace_io_desc32 *piod32 = NULL;
  647         struct ptrace_lwpinfo32 *pl32 = NULL;
  648         struct ptrace_lwpinfo plr;
  649 #endif
  650 
  651         curp = td->td_proc;
  652 
  653         /* Lock proctree before locking the process. */
  654         switch (req) {
  655         case PT_TRACE_ME:
  656         case PT_ATTACH:
  657         case PT_STEP:
  658         case PT_CONTINUE:
  659         case PT_TO_SCE:
  660         case PT_TO_SCX:
  661         case PT_SYSCALL:
  662         case PT_FOLLOW_FORK:
  663         case PT_DETACH:
  664                 sx_xlock(&proctree_lock);
  665                 proctree_locked = 1;
  666                 break;
  667         default:
  668                 break;
  669         }
  670 
  671         write = 0;
  672         if (req == PT_TRACE_ME) {
  673                 p = td->td_proc;
  674                 PROC_LOCK(p);
  675         } else {
  676                 if (pid <= PID_MAX) {
  677                         if ((p = pfind(pid)) == NULL) {
  678                                 if (proctree_locked)
  679                                         sx_xunlock(&proctree_lock);
  680                                 return (ESRCH);
  681                         }
  682                 } else {
  683                         td2 = tdfind(pid, -1);
  684                         if (td2 == NULL) {
  685                                 if (proctree_locked)
  686                                         sx_xunlock(&proctree_lock);
  687                                 return (ESRCH);
  688                         }
  689                         p = td2->td_proc;
  690                         tid = pid;
  691                         pid = p->p_pid;
  692                 }
  693         }
  694         AUDIT_ARG_PROCESS(p);
  695 
  696         if ((p->p_flag & P_WEXIT) != 0) {
  697                 error = ESRCH;
  698                 goto fail;
  699         }
  700         if ((error = p_cansee(td, p)) != 0)
  701                 goto fail;
  702 
  703         if ((error = p_candebug(td, p)) != 0)
  704                 goto fail;
  705 
  706         /*
  707          * System processes can't be debugged.
  708          */
  709         if ((p->p_flag & P_SYSTEM) != 0) {
  710                 error = EINVAL;
  711                 goto fail;
  712         }
  713 
  714         if (tid == 0) {
  715                 if ((p->p_flag & P_STOPPED_TRACE) != 0) {
  716                         KASSERT(p->p_xthread != NULL, ("NULL p_xthread"));
  717                         td2 = p->p_xthread;
  718                 } else {
  719                         td2 = FIRST_THREAD_IN_PROC(p);
  720                 }
  721                 tid = td2->td_tid;
  722         }
  723 
  724 #ifdef COMPAT_FREEBSD32
  725         /*
  726          * Test if we're a 32 bit client and what the target is.
  727          * Set the wrap controls accordingly.
  728          */
  729         if (SV_CURPROC_FLAG(SV_ILP32)) {
  730                 if (SV_PROC_FLAG(td2->td_proc, SV_ILP32))
  731                         safe = 1;
  732                 wrap32 = 1;
  733         }
  734 #endif
  735         /*
  736          * Permissions check
  737          */
  738         switch (req) {
  739         case PT_TRACE_ME:
  740                 /* Always legal. */
  741                 break;
  742 
  743         case PT_ATTACH:
  744                 /* Self */
  745                 if (p->p_pid == td->td_proc->p_pid) {
  746                         error = EINVAL;
  747                         goto fail;
  748                 }
  749 
  750                 /* Already traced */
  751                 if (p->p_flag & P_TRACED) {
  752                         error = EBUSY;
  753                         goto fail;
  754                 }
  755 
  756                 /* Can't trace an ancestor if you're being traced. */
  757                 if (curp->p_flag & P_TRACED) {
  758                         for (pp = curp->p_pptr; pp != NULL; pp = pp->p_pptr) {
  759                                 if (pp == p) {
  760                                         error = EINVAL;
  761                                         goto fail;
  762                                 }
  763                         }
  764                 }
  765 
  766 
  767                 /* OK */
  768                 break;
  769 
  770         case PT_CLEARSTEP:
  771                 /* Allow thread to clear single step for itself */
  772                 if (td->td_tid == tid)
  773                         break;
  774 
  775                 /* FALLTHROUGH */
  776         default:
  777                 /* not being traced... */
  778                 if ((p->p_flag & P_TRACED) == 0) {
  779                         error = EPERM;
  780                         goto fail;
  781                 }
  782 
  783                 /* not being traced by YOU */
  784                 if (p->p_pptr != td->td_proc) {
  785                         error = EBUSY;
  786                         goto fail;
  787                 }
  788 
  789                 /* not currently stopped */
  790                 if ((p->p_flag & (P_STOPPED_SIG | P_STOPPED_TRACE)) == 0 ||
  791                     p->p_suspcount != p->p_numthreads  ||
  792                     (p->p_flag & P_WAITED) == 0) {
  793                         error = EBUSY;
  794                         goto fail;
  795                 }
  796 
  797                 if ((p->p_flag & P_STOPPED_TRACE) == 0) {
  798                         static int count = 0;
  799                         if (count++ == 0)
  800                                 printf("P_STOPPED_TRACE not set.\n");
  801                 }
  802 
  803                 /* OK */
  804                 break;
  805         }
  806 
  807         /* Keep this process around until we finish this request. */
  808         _PHOLD(p);
  809 
  810 #ifdef FIX_SSTEP
  811         /*
  812          * Single step fixup ala procfs
  813          */
  814         FIX_SSTEP(td2);
  815 #endif
  816 
  817         /*
  818          * Actually do the requests
  819          */
  820 
  821         td->td_retval[0] = 0;
  822 
  823         switch (req) {
  824         case PT_TRACE_ME:
  825                 /* set my trace flag and "owner" so it can read/write me */
  826                 p->p_flag |= P_TRACED;
  827                 if (p->p_flag & P_PPWAIT)
  828                         p->p_flag |= P_PPTRACE;
  829                 p->p_oppid = p->p_pptr->p_pid;
  830                 break;
  831 
  832         case PT_ATTACH:
  833                 /* security check done above */
  834                 /*
  835                  * It would be nice if the tracing relationship was separate
  836                  * from the parent relationship but that would require
  837                  * another set of links in the proc struct or for "wait"
  838                  * to scan the entire proc table.  To make life easier,
  839                  * we just re-parent the process we're trying to trace.
  840                  * The old parent is remembered so we can put things back
  841                  * on a "detach".
  842                  */
  843                 p->p_flag |= P_TRACED;
  844                 p->p_oppid = p->p_pptr->p_pid;
  845                 if (p->p_pptr != td->td_proc) {
  846                         proc_reparent(p, td->td_proc);
  847                 }
  848                 data = SIGSTOP;
  849                 goto sendsig;   /* in PT_CONTINUE below */
  850 
  851         case PT_CLEARSTEP:
  852                 error = ptrace_clear_single_step(td2);
  853                 break;
  854 
  855         case PT_SETSTEP:
  856                 error = ptrace_single_step(td2);
  857                 break;
  858 
  859         case PT_SUSPEND:
  860                 td2->td_dbgflags |= TDB_SUSPEND;
  861                 thread_lock(td2);
  862                 td2->td_flags |= TDF_NEEDSUSPCHK;
  863                 thread_unlock(td2);
  864                 break;
  865 
  866         case PT_RESUME:
  867                 td2->td_dbgflags &= ~TDB_SUSPEND;
  868                 break;
  869 
  870         case PT_FOLLOW_FORK:
  871                 if (data)
  872                         p->p_flag |= P_FOLLOWFORK;
  873                 else
  874                         p->p_flag &= ~P_FOLLOWFORK;
  875                 break;
  876 
  877         case PT_STEP:
  878         case PT_CONTINUE:
  879         case PT_TO_SCE:
  880         case PT_TO_SCX:
  881         case PT_SYSCALL:
  882         case PT_DETACH:
  883                 /* Zero means do not send any signal */
  884                 if (data < 0 || data > _SIG_MAXSIG) {
  885                         error = EINVAL;
  886                         break;
  887                 }
  888 
  889                 switch (req) {
  890                 case PT_STEP:
  891                         error = ptrace_single_step(td2);
  892                         if (error)
  893                                 goto out;
  894                         break;
  895                 case PT_CONTINUE:
  896                 case PT_TO_SCE:
  897                 case PT_TO_SCX:
  898                 case PT_SYSCALL:
  899                         if (addr != (void *)1) {
  900                                 error = ptrace_set_pc(td2,
  901                                     (u_long)(uintfptr_t)addr);
  902                                 if (error)
  903                                         goto out;
  904                         }
  905                         switch (req) {
  906                         case PT_TO_SCE:
  907                                 p->p_stops |= S_PT_SCE;
  908                                 break;
  909                         case PT_TO_SCX:
  910                                 p->p_stops |= S_PT_SCX;
  911                                 break;
  912                         case PT_SYSCALL:
  913                                 p->p_stops |= S_PT_SCE | S_PT_SCX;
  914                                 break;
  915                         }
  916                         break;
  917                 case PT_DETACH:
  918                         /* reset process parent */
  919                         if (p->p_oppid != p->p_pptr->p_pid) {
  920                                 PROC_LOCK(p->p_pptr);
  921                                 sigqueue_take(p->p_ksi);
  922                                 PROC_UNLOCK(p->p_pptr);
  923 
  924                                 pp = proc_realparent(p);
  925                                 proc_reparent(p, pp);
  926                                 if (pp == initproc)
  927                                         p->p_sigparent = SIGCHLD;
  928                         }
  929                         p->p_oppid = 0;
  930                         p->p_flag &= ~(P_TRACED | P_WAITED | P_FOLLOWFORK);
  931 
  932                         /* should we send SIGCHLD? */
  933                         /* childproc_continued(p); */
  934                         break;
  935                 }
  936 
  937         sendsig:
  938                 if (proctree_locked) {
  939                         sx_xunlock(&proctree_lock);
  940                         proctree_locked = 0;
  941                 }
  942                 p->p_xstat = data;
  943                 p->p_xthread = NULL;
  944                 if ((p->p_flag & (P_STOPPED_SIG | P_STOPPED_TRACE)) != 0) {
  945                         /* deliver or queue signal */
  946                         td2->td_dbgflags &= ~TDB_XSIG;
  947                         td2->td_xsig = data;
  948 
  949                         if (req == PT_DETACH) {
  950                                 FOREACH_THREAD_IN_PROC(p, td3)
  951                                         td3->td_dbgflags &= ~TDB_SUSPEND; 
  952                         }
  953                         /*
  954                          * unsuspend all threads, to not let a thread run,
  955                          * you should use PT_SUSPEND to suspend it before
  956                          * continuing process.
  957                          */
  958                         PROC_SLOCK(p);
  959                         p->p_flag &= ~(P_STOPPED_TRACE|P_STOPPED_SIG|P_WAITED);
  960                         thread_unsuspend(p);
  961                         PROC_SUNLOCK(p);
  962                         if (req == PT_ATTACH)
  963                                 kern_psignal(p, data);
  964                 } else {
  965                         if (data)
  966                                 kern_psignal(p, data);
  967                 }
  968                 break;
  969 
  970         case PT_WRITE_I:
  971         case PT_WRITE_D:
  972                 td2->td_dbgflags |= TDB_USERWR;
  973                 write = 1;
  974                 /* FALLTHROUGH */
  975         case PT_READ_I:
  976         case PT_READ_D:
  977                 PROC_UNLOCK(p);
  978                 tmp = 0;
  979                 /* write = 0 set above */
  980                 iov.iov_base = write ? (caddr_t)&data : (caddr_t)&tmp;
  981                 iov.iov_len = sizeof(int);
  982                 uio.uio_iov = &iov;
  983                 uio.uio_iovcnt = 1;
  984                 uio.uio_offset = (off_t)(uintptr_t)addr;
  985                 uio.uio_resid = sizeof(int);
  986                 uio.uio_segflg = UIO_SYSSPACE;  /* i.e.: the uap */
  987                 uio.uio_rw = write ? UIO_WRITE : UIO_READ;
  988                 uio.uio_td = td;
  989                 error = proc_rwmem(p, &uio);
  990                 if (uio.uio_resid != 0) {
  991                         /*
  992                          * XXX proc_rwmem() doesn't currently return ENOSPC,
  993                          * so I think write() can bogusly return 0.
  994                          * XXX what happens for short writes?  We don't want
  995                          * to write partial data.
  996                          * XXX proc_rwmem() returns EPERM for other invalid
  997                          * addresses.  Convert this to EINVAL.  Does this
  998                          * clobber returns of EPERM for other reasons?
  999                          */
 1000                         if (error == 0 || error == ENOSPC || error == EPERM)
 1001                                 error = EINVAL; /* EOF */
 1002                 }
 1003                 if (!write)
 1004                         td->td_retval[0] = tmp;
 1005                 PROC_LOCK(p);
 1006                 break;
 1007 
 1008         case PT_IO:
 1009 #ifdef COMPAT_FREEBSD32
 1010                 if (wrap32) {
 1011                         piod32 = addr;
 1012                         iov.iov_base = (void *)(uintptr_t)piod32->piod_addr;
 1013                         iov.iov_len = piod32->piod_len;
 1014                         uio.uio_offset = (off_t)(uintptr_t)piod32->piod_offs;
 1015                         uio.uio_resid = piod32->piod_len;
 1016                 } else
 1017 #endif
 1018                 {
 1019                         piod = addr;
 1020                         iov.iov_base = piod->piod_addr;
 1021                         iov.iov_len = piod->piod_len;
 1022                         uio.uio_offset = (off_t)(uintptr_t)piod->piod_offs;
 1023                         uio.uio_resid = piod->piod_len;
 1024                 }
 1025                 uio.uio_iov = &iov;
 1026                 uio.uio_iovcnt = 1;
 1027                 uio.uio_segflg = UIO_USERSPACE;
 1028                 uio.uio_td = td;
 1029 #ifdef COMPAT_FREEBSD32
 1030                 tmp = wrap32 ? piod32->piod_op : piod->piod_op;
 1031 #else
 1032                 tmp = piod->piod_op;
 1033 #endif
 1034                 switch (tmp) {
 1035                 case PIOD_READ_D:
 1036                 case PIOD_READ_I:
 1037                         uio.uio_rw = UIO_READ;
 1038                         break;
 1039                 case PIOD_WRITE_D:
 1040                 case PIOD_WRITE_I:
 1041                         td2->td_dbgflags |= TDB_USERWR;
 1042                         uio.uio_rw = UIO_WRITE;
 1043                         break;
 1044                 default:
 1045                         error = EINVAL;
 1046                         goto out;
 1047                 }
 1048                 PROC_UNLOCK(p);
 1049                 error = proc_rwmem(p, &uio);
 1050 #ifdef COMPAT_FREEBSD32
 1051                 if (wrap32)
 1052                         piod32->piod_len -= uio.uio_resid;
 1053                 else
 1054 #endif
 1055                         piod->piod_len -= uio.uio_resid;
 1056                 PROC_LOCK(p);
 1057                 break;
 1058 
 1059         case PT_KILL:
 1060                 data = SIGKILL;
 1061                 goto sendsig;   /* in PT_CONTINUE above */
 1062 
 1063         case PT_SETREGS:
 1064                 td2->td_dbgflags |= TDB_USERWR;
 1065                 error = PROC_WRITE(regs, td2, addr);
 1066                 break;
 1067 
 1068         case PT_GETREGS:
 1069                 error = PROC_READ(regs, td2, addr);
 1070                 break;
 1071 
 1072         case PT_SETFPREGS:
 1073                 td2->td_dbgflags |= TDB_USERWR;
 1074                 error = PROC_WRITE(fpregs, td2, addr);
 1075                 break;
 1076 
 1077         case PT_GETFPREGS:
 1078                 error = PROC_READ(fpregs, td2, addr);
 1079                 break;
 1080 
 1081         case PT_SETDBREGS:
 1082                 td2->td_dbgflags |= TDB_USERWR;
 1083                 error = PROC_WRITE(dbregs, td2, addr);
 1084                 break;
 1085 
 1086         case PT_GETDBREGS:
 1087                 error = PROC_READ(dbregs, td2, addr);
 1088                 break;
 1089 
 1090         case PT_LWPINFO:
 1091                 if (data <= 0 ||
 1092 #ifdef COMPAT_FREEBSD32
 1093                     (!wrap32 && data > sizeof(*pl)) ||
 1094                     (wrap32 && data > sizeof(*pl32))) {
 1095 #else
 1096                     data > sizeof(*pl)) {
 1097 #endif
 1098                         error = EINVAL;
 1099                         break;
 1100                 }
 1101 #ifdef COMPAT_FREEBSD32
 1102                 if (wrap32) {
 1103                         pl = &plr;
 1104                         pl32 = addr;
 1105                 } else
 1106 #endif
 1107                 pl = addr;
 1108                 pl->pl_lwpid = td2->td_tid;
 1109                 pl->pl_event = PL_EVENT_NONE;
 1110                 pl->pl_flags = 0;
 1111                 if (td2->td_dbgflags & TDB_XSIG) {
 1112                         pl->pl_event = PL_EVENT_SIGNAL;
 1113                         if (td2->td_dbgksi.ksi_signo != 0 &&
 1114 #ifdef COMPAT_FREEBSD32
 1115                             ((!wrap32 && data >= offsetof(struct ptrace_lwpinfo,
 1116                             pl_siginfo) + sizeof(pl->pl_siginfo)) ||
 1117                             (wrap32 && data >= offsetof(struct ptrace_lwpinfo32,
 1118                             pl_siginfo) + sizeof(struct siginfo32)))
 1119 #else
 1120                             data >= offsetof(struct ptrace_lwpinfo, pl_siginfo)
 1121                             + sizeof(pl->pl_siginfo)
 1122 #endif
 1123                         ){
 1124                                 pl->pl_flags |= PL_FLAG_SI;
 1125                                 pl->pl_siginfo = td2->td_dbgksi.ksi_info;
 1126                         }
 1127                 }
 1128                 if ((pl->pl_flags & PL_FLAG_SI) == 0)
 1129                         bzero(&pl->pl_siginfo, sizeof(pl->pl_siginfo));
 1130                 if (td2->td_dbgflags & TDB_SCE)
 1131                         pl->pl_flags |= PL_FLAG_SCE;
 1132                 else if (td2->td_dbgflags & TDB_SCX)
 1133                         pl->pl_flags |= PL_FLAG_SCX;
 1134                 if (td2->td_dbgflags & TDB_EXEC)
 1135                         pl->pl_flags |= PL_FLAG_EXEC;
 1136                 if (td2->td_dbgflags & TDB_FORK) {
 1137                         pl->pl_flags |= PL_FLAG_FORKED;
 1138                         pl->pl_child_pid = td2->td_dbg_forked;
 1139                 }
 1140                 if (td2->td_dbgflags & TDB_CHILD)
 1141                         pl->pl_flags |= PL_FLAG_CHILD;
 1142                 pl->pl_sigmask = td2->td_sigmask;
 1143                 pl->pl_siglist = td2->td_siglist;
 1144                 strcpy(pl->pl_tdname, td2->td_name);
 1145 #ifdef COMPAT_FREEBSD32
 1146                 if (wrap32)
 1147                         ptrace_lwpinfo_to32(pl, pl32);
 1148 #endif
 1149                 break;
 1150 
 1151         case PT_GETNUMLWPS:
 1152                 td->td_retval[0] = p->p_numthreads;
 1153                 break;
 1154 
 1155         case PT_GETLWPLIST:
 1156                 if (data <= 0) {
 1157                         error = EINVAL;
 1158                         break;
 1159                 }
 1160                 num = imin(p->p_numthreads, data);
 1161                 PROC_UNLOCK(p);
 1162                 buf = malloc(num * sizeof(lwpid_t), M_TEMP, M_WAITOK);
 1163                 tmp = 0;
 1164                 PROC_LOCK(p);
 1165                 FOREACH_THREAD_IN_PROC(p, td2) {
 1166                         if (tmp >= num)
 1167                                 break;
 1168                         buf[tmp++] = td2->td_tid;
 1169                 }
 1170                 PROC_UNLOCK(p);
 1171                 error = copyout(buf, addr, tmp * sizeof(lwpid_t));
 1172                 free(buf, M_TEMP);
 1173                 if (!error)
 1174                         td->td_retval[0] = tmp;
 1175                 PROC_LOCK(p);
 1176                 break;
 1177 
 1178         case PT_VM_TIMESTAMP:
 1179                 td->td_retval[0] = p->p_vmspace->vm_map.timestamp;
 1180                 break;
 1181 
 1182         case PT_VM_ENTRY:
 1183                 PROC_UNLOCK(p);
 1184 #ifdef COMPAT_FREEBSD32
 1185                 if (wrap32)
 1186                         error = ptrace_vm_entry32(td, p, addr);
 1187                 else
 1188 #endif
 1189                 error = ptrace_vm_entry(td, p, addr);
 1190                 PROC_LOCK(p);
 1191                 break;
 1192 
 1193         default:
 1194 #ifdef __HAVE_PTRACE_MACHDEP
 1195                 if (req >= PT_FIRSTMACH) {
 1196                         PROC_UNLOCK(p);
 1197                         error = cpu_ptrace(td2, req, addr, data);
 1198                         PROC_LOCK(p);
 1199                 } else
 1200 #endif
 1201                         /* Unknown request. */
 1202                         error = EINVAL;
 1203                 break;
 1204         }
 1205 
 1206 out:
 1207         /* Drop our hold on this process now that the request has completed. */
 1208         _PRELE(p);
 1209 fail:
 1210         PROC_UNLOCK(p);
 1211         if (proctree_locked)
 1212                 sx_xunlock(&proctree_lock);
 1213         return (error);
 1214 }
 1215 #undef PROC_READ
 1216 #undef PROC_WRITE
 1217 
 1218 /*
 1219  * Stop a process because of a debugging event;
 1220  * stay stopped until p->p_step is cleared
 1221  * (cleared by PIOCCONT in procfs).
 1222  */
 1223 void
 1224 stopevent(struct proc *p, unsigned int event, unsigned int val)
 1225 {
 1226 
 1227         PROC_LOCK_ASSERT(p, MA_OWNED);
 1228         p->p_step = 1;
 1229         do {
 1230                 p->p_xstat = val;
 1231                 p->p_xthread = NULL;
 1232                 p->p_stype = event;     /* Which event caused the stop? */
 1233                 wakeup(&p->p_stype);    /* Wake up any PIOCWAIT'ing procs */
 1234                 msleep(&p->p_step, &p->p_mtx, PWAIT, "stopevent", 0);
 1235         } while (p->p_step);
 1236 }
 1237 
 1238 static int
 1239 protect_setchild(struct thread *td, struct proc *p, int flags)
 1240 {
 1241 
 1242         PROC_LOCK_ASSERT(p, MA_OWNED);
 1243         if (p->p_flag & P_SYSTEM || p_cansee(td, p) != 0)
 1244                 return (0);
 1245         if (flags & PPROT_SET) {
 1246                 p->p_flag |= P_PROTECTED;
 1247                 if (flags & PPROT_INHERIT)
 1248                         p->p_flag2 |= P2_INHERIT_PROTECTED;
 1249         } else {
 1250                 p->p_flag &= ~P_PROTECTED;
 1251                 p->p_flag2 &= ~P2_INHERIT_PROTECTED;
 1252         }
 1253         return (1);
 1254 }
 1255 
 1256 static int
 1257 protect_setchildren(struct thread *td, struct proc *top, int flags)
 1258 {
 1259         struct proc *p;
 1260         int ret;
 1261 
 1262         p = top;
 1263         ret = 0;
 1264         sx_assert(&proctree_lock, SX_LOCKED);
 1265         for (;;) {
 1266                 ret |= protect_setchild(td, p, flags);
 1267                 PROC_UNLOCK(p);
 1268                 /*
 1269                  * If this process has children, descend to them next,
 1270                  * otherwise do any siblings, and if done with this level,
 1271                  * follow back up the tree (but not past top).
 1272                  */
 1273                 if (!LIST_EMPTY(&p->p_children))
 1274                         p = LIST_FIRST(&p->p_children);
 1275                 else for (;;) {
 1276                         if (p == top) {
 1277                                 PROC_LOCK(p);
 1278                                 return (ret);
 1279                         }
 1280                         if (LIST_NEXT(p, p_sibling)) {
 1281                                 p = LIST_NEXT(p, p_sibling);
 1282                                 break;
 1283                         }
 1284                         p = p->p_pptr;
 1285                 }
 1286                 PROC_LOCK(p);
 1287         }
 1288 }
 1289 
 1290 static int
 1291 protect_set(struct thread *td, struct proc *p, int flags)
 1292 {
 1293         int error, ret;
 1294 
 1295         switch (PPROT_OP(flags)) {
 1296         case PPROT_SET:
 1297         case PPROT_CLEAR:
 1298                 break;
 1299         default:
 1300                 return (EINVAL);
 1301         }
 1302 
 1303         if ((PPROT_FLAGS(flags) & ~(PPROT_DESCEND | PPROT_INHERIT)) != 0)
 1304                 return (EINVAL);
 1305 
 1306         error = priv_check(td, PRIV_VM_MADV_PROTECT);
 1307         if (error)
 1308                 return (error);
 1309 
 1310         if (flags & PPROT_DESCEND)
 1311                 ret = protect_setchildren(td, p, flags);
 1312         else
 1313                 ret = protect_setchild(td, p, flags);
 1314         if (ret == 0)
 1315                 return (EPERM);
 1316         return (0);
 1317 }
 1318 
 1319 #ifndef _SYS_SYSPROTO_H_
 1320 struct procctl_args {
 1321         idtype_t idtype;
 1322         id_t    id;
 1323         int     com;
 1324         void    *data;
 1325 };
 1326 #endif
 1327 /* ARGSUSED */
 1328 int
 1329 sys_procctl(struct thread *td, struct procctl_args *uap)
 1330 {
 1331         int error, flags;
 1332         void *data;
 1333 
 1334         switch (uap->com) {
 1335         case PROC_SPROTECT:
 1336                 error = copyin(uap->data, &flags, sizeof(flags));
 1337                 if (error)
 1338                         return (error);
 1339                 data = &flags;
 1340                 break;
 1341         default:
 1342                 return (EINVAL);
 1343         }
 1344 
 1345         return (kern_procctl(td, uap->idtype, uap->id, uap->com, data));
 1346 }
 1347 
 1348 static int
 1349 kern_procctl_single(struct thread *td, struct proc *p, int com, void *data)
 1350 {
 1351 
 1352         PROC_LOCK_ASSERT(p, MA_OWNED);
 1353         switch (com) {
 1354         case PROC_SPROTECT:
 1355                 return (protect_set(td, p, *(int *)data));
 1356         default:
 1357                 return (EINVAL);
 1358         }
 1359 }
 1360 
 1361 int
 1362 kern_procctl(struct thread *td, idtype_t idtype, id_t id, int com, void *data)
 1363 {
 1364         struct pgrp *pg;
 1365         struct proc *p;
 1366         int error, first_error, ok;
 1367 
 1368         sx_slock(&proctree_lock);
 1369         switch (idtype) {
 1370         case P_PID:
 1371                 p = pfind(id);
 1372                 if (p == NULL) {
 1373                         error = ESRCH;
 1374                         break;
 1375                 }
 1376                 if (p->p_state == PRS_NEW)
 1377                         error = ESRCH;
 1378                 else
 1379                         error = p_cansee(td, p);
 1380                 if (error == 0)
 1381                         error = kern_procctl_single(td, p, com, data);
 1382                 PROC_UNLOCK(p);
 1383                 break;
 1384         case P_PGID:
 1385                 /*
 1386                  * Attempt to apply the operation to all members of the
 1387                  * group.  Ignore processes in the group that can't be
 1388                  * seen.  Ignore errors so long as at least one process is
 1389                  * able to complete the request successfully.
 1390                  */
 1391                 pg = pgfind(id);
 1392                 if (pg == NULL) {
 1393                         error = ESRCH;
 1394                         break;
 1395                 }
 1396                 PGRP_UNLOCK(pg);
 1397                 ok = 0;
 1398                 first_error = 0;
 1399                 LIST_FOREACH(p, &pg->pg_members, p_pglist) {
 1400                         PROC_LOCK(p);
 1401                         if (p->p_state == PRS_NEW || p_cansee(td, p) != 0) {
 1402                                 PROC_UNLOCK(p);
 1403                                 continue;
 1404                         }
 1405                         error = kern_procctl_single(td, p, com, data);
 1406                         PROC_UNLOCK(p);
 1407                         if (error == 0)
 1408                                 ok = 1;
 1409                         else if (first_error == 0)
 1410                                 first_error = error;
 1411                 }
 1412                 if (ok)
 1413                         error = 0;
 1414                 else if (first_error != 0)
 1415                         error = first_error;
 1416                 else
 1417                         /*
 1418                          * Was not able to see any processes in the
 1419                          * process group.
 1420                          */
 1421                         error = ESRCH;
 1422                 break;
 1423         default:
 1424                 error = EINVAL;
 1425                 break;
 1426         }
 1427         sx_sunlock(&proctree_lock);
 1428         return (error);
 1429 }

Cache object: 4f8e0e7199a183427f973b46b8cdd7ec


[ source navigation ] [ diff markup ] [ identifier search ] [ freetext search ] [ file search ] [ list types ] [ track identifier ]


This page is part of the FreeBSD/Linux Linux Kernel Cross-Reference, and was automatically generated using a modified version of the LXR engine.