/*
* Destroy a process structure that resulted from a call to forkproc(), but
* which must be returned to the system because of a subsequent failure
* preventing it from becoming active.
*
* Parameters: p The incomplete process from forkproc()
*
* Returns: (void)
*
* Note: This function should only be used in an error handler following
* a call to forkproc().
*
* Operations occur in reverse order of those in forkproc().
*/
void
forkproc_free(proc_t p)
{
/* We held signal and a transition locks; drop them */
proc_signalend(p, 0);
proc_transend(p, 0);
/*
* If we have our own copy of the resource limits structure, we
* need to free it. If it's a shared copy, we need to drop our
* reference on it.
*/
proc_limitdrop(p, 0);
p->p_limit = NULL;
#if SYSV_SHM
/* Need to drop references to the shared memory segment(s), if any */
if (p->vm_shm) {
/*
* Use shmexec(): we have no address space, so no mappings
*
* XXX Yes, the routine is badly named.
*/
shmexec(p);
}
#endif
/* Need to undo the effects of the fdcopy(), if any */
fdfree(p);
/*
* Drop the reference on a text vnode pointer, if any
* XXX This code is broken in forkproc(); see <rdar://4256419>;
* XXX if anyone ever uses this field, we will be extremely unhappy.
*/
if (p->p_textvp) {
vnode_rele(p->p_textvp);
p->p_textvp = NULL;
}
/* Stop the profiling clock */
stopprofclock(p);
/* Release the credential reference */
kauth_cred_unref(&p->p_ucred);
proc_list_lock();
/* Decrement the count of processes in the system */
nprocs--;
proc_list_unlock();
thread_call_free(p->p_rcall);
/* Free allocated memory */
FREE_ZONE(p->p_sigacts, sizeof *p->p_sigacts, M_SIGACTS);
FREE_ZONE(p->p_stats, sizeof *p->p_stats, M_PSTATS);
proc_checkdeadrefs(p);
FREE_ZONE(p, sizeof *p, M_PROC);
}
/*
* Much like before, but we can afford to take faults here. If the
* update fails, we simply turn off profiling.
*/
void
addupc_task(struct proc *p, u_long pc, u_int nticks)
{
struct uprof *prof;
caddr_t addr;
u_int i;
u_short v;
/* Testing P_PROFIL may be unnecessary, but is certainly safe. */
if ((p->p_flag & P_PROFIL) == 0 || nticks == 0)
return;
prof = &p->p_p->ps_prof;
if (pc < prof->pr_off ||
(i = PC_TO_INDEX(pc, prof)) >= prof->pr_size)
return;
addr = prof->pr_base + i;
if (copyin(addr, (caddr_t)&v, sizeof(v)) == 0) {
v += nticks;
if (copyout((caddr_t)&v, addr, sizeof(v)) == 0)
return;
}
stopprofclock(p);
}
/* ARGSUSED */
int
sys_profil(struct proc *p, void *v, register_t *retval)
{
struct sys_profil_args /* {
syscallarg(caddr_t) samples;
syscallarg(size_t) size;
syscallarg(u_long) offset;
syscallarg(u_int) scale;
} */ *uap = v;
struct uprof *upp;
int s;
if (SCARG(uap, scale) > (1 << 16))
return (EINVAL);
if (SCARG(uap, scale) == 0) {
stopprofclock(p);
return (0);
}
upp = &p->p_p->ps_prof;
/* Block profile interrupts while changing state. */
s = splstatclock();
upp->pr_off = SCARG(uap, offset);
upp->pr_scale = SCARG(uap, scale);
upp->pr_base = (caddr_t)SCARG(uap, samples);
upp->pr_size = SCARG(uap, size);
startprofclock(p);
splx(s);
return (0);
}
/*
* sysctl helper routine for kern.profiling subtree. enables/disables
* kernel profiling and gives out copies of the profiling data.
*/
static int
sysctl_kern_profiling(SYSCTLFN_ARGS)
{
struct gmonparam *gp = &_gmonparam;
int error;
struct sysctlnode node;
node = *rnode;
switch (node.sysctl_num) {
case GPROF_STATE:
node.sysctl_data = &gp->state;
break;
case GPROF_COUNT:
node.sysctl_data = gp->kcount;
node.sysctl_size = gp->kcountsize;
break;
case GPROF_FROMS:
node.sysctl_data = gp->froms;
node.sysctl_size = gp->fromssize;
break;
case GPROF_TOS:
node.sysctl_data = gp->tos;
node.sysctl_size = gp->tossize;
break;
case GPROF_GMONPARAM:
node.sysctl_data = gp;
node.sysctl_size = sizeof(*gp);
break;
default:
return (EOPNOTSUPP);
}
error = sysctl_lookup(SYSCTLFN_CALL(&node));
if (error || newp == NULL)
return (error);
if (node.sysctl_num == GPROF_STATE) {
mutex_spin_enter(&proc0.p_stmutex);
if (gp->state == GMON_PROF_OFF)
stopprofclock(&proc0);
else
startprofclock(&proc0);
mutex_spin_exit(&proc0.p_stmutex);
}
return (0);
}
/*
* Return kernel profiling information.
*/
int
sysctl_doprof(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
size_t newlen)
{
struct gmonparam *gp = &_gmonparam;
int error;
/* all sysctl names at this level are terminal */
if (namelen != 1)
return (ENOTDIR); /* overloaded */
switch (name[0]) {
case GPROF_STATE:
error = sysctl_int(oldp, oldlenp, newp, newlen, &gp->state);
if (error)
return (error);
if (gp->state == GMON_PROF_OFF)
stopprofclock(&proc0);
else
startprofclock(&proc0);
return (0);
case GPROF_COUNT:
return (sysctl_struct(oldp, oldlenp, newp, newlen,
gp->kcount, gp->kcountsize));
case GPROF_FROMS:
return (sysctl_struct(oldp, oldlenp, newp, newlen,
gp->froms, gp->fromssize));
case GPROF_TOS:
return (sysctl_struct(oldp, oldlenp, newp, newlen,
gp->tos, gp->tossize));
case GPROF_GMONPARAM:
return (sysctl_rdstruct(oldp, oldlenp, newp, gp, sizeof *gp));
default:
return (EOPNOTSUPP);
}
/* NOTREACHED */
}
/* ARGSUSED */
int
sys_execve(struct proc *p, void *v, register_t *retval)
{
struct sys_execve_args /* {
syscallarg(const char *) path;
syscallarg(char *const *) argp;
syscallarg(char *const *) envp;
} */ *uap = v;
int error;
struct exec_package pack;
struct nameidata nid;
struct vattr attr;
struct ucred *cred = p->p_ucred;
char *argp;
char * const *cpp, *dp, *sp;
#ifdef KTRACE
char *env_start;
#endif
struct process *pr = p->p_p;
long argc, envc;
size_t len, sgap;
#ifdef MACHINE_STACK_GROWS_UP
size_t slen;
#endif
char *stack;
struct ps_strings arginfo;
struct vmspace *vm = pr->ps_vmspace;
char **tmpfap;
extern struct emul emul_native;
#if NSYSTRACE > 0
int wassugid = ISSET(pr->ps_flags, PS_SUGID | PS_SUGIDEXEC);
size_t pathbuflen;
#endif
char *pathbuf = NULL;
struct vnode *otvp;
/* get other threads to stop */
if ((error = single_thread_set(p, SINGLE_UNWIND, 1)))
return (error);
/*
* Cheap solution to complicated problems.
* Mark this process as "leave me alone, I'm execing".
*/
atomic_setbits_int(&pr->ps_flags, PS_INEXEC);
#if NSYSTRACE > 0
if (ISSET(p->p_flag, P_SYSTRACE)) {
systrace_execve0(p);
pathbuf = pool_get(&namei_pool, PR_WAITOK);
error = copyinstr(SCARG(uap, path), pathbuf, MAXPATHLEN,
&pathbuflen);
if (error != 0)
goto clrflag;
}
#endif
if (pathbuf != NULL) {
NDINIT(&nid, LOOKUP, NOFOLLOW, UIO_SYSSPACE, pathbuf, p);
} else {
NDINIT(&nid, LOOKUP, NOFOLLOW, UIO_USERSPACE,
SCARG(uap, path), p);
}
/*
* initialize the fields of the exec package.
*/
if (pathbuf != NULL)
pack.ep_name = pathbuf;
else
pack.ep_name = (char *)SCARG(uap, path);
pack.ep_hdr = malloc(exec_maxhdrsz, M_EXEC, M_WAITOK);
pack.ep_hdrlen = exec_maxhdrsz;
pack.ep_hdrvalid = 0;
pack.ep_ndp = &nid;
pack.ep_interp = NULL;
pack.ep_emul_arg = NULL;
VMCMDSET_INIT(&pack.ep_vmcmds);
pack.ep_vap = &attr;
pack.ep_emul = &emul_native;
pack.ep_flags = 0;
/* see if we can run it. */
if ((error = check_exec(p, &pack)) != 0) {
goto freehdr;
}
/* XXX -- THE FOLLOWING SECTION NEEDS MAJOR CLEANUP */
/* allocate an argument buffer */
argp = km_alloc(NCARGS, &kv_exec, &kp_pageable, &kd_waitok);
#ifdef DIAGNOSTIC
if (argp == NULL)
panic("execve: argp == NULL");
#endif
dp = argp;
argc = 0;
/* copy the fake args list, if there's one, freeing it as we go */
if (pack.ep_flags & EXEC_HASARGL) {
tmpfap = pack.ep_fa;
while (*tmpfap != NULL) {
char *cp;
cp = *tmpfap;
while (*cp)
*dp++ = *cp++;
*dp++ = '\0';
free(*tmpfap, M_EXEC, 0);
tmpfap++; argc++;
}
free(pack.ep_fa, M_EXEC, 0);
pack.ep_flags &= ~EXEC_HASARGL;
}
/* Now get argv & environment */
if (!(cpp = SCARG(uap, argp))) {
error = EFAULT;
goto bad;
}
if (pack.ep_flags & EXEC_SKIPARG)
cpp++;
while (1) {
len = argp + ARG_MAX - dp;
if ((error = copyin(cpp, &sp, sizeof(sp))) != 0)
goto bad;
if (!sp)
break;
if ((error = copyinstr(sp, dp, len, &len)) != 0) {
if (error == ENAMETOOLONG)
error = E2BIG;
goto bad;
}
dp += len;
cpp++;
argc++;
}
/* must have at least one argument */
if (argc == 0) {
error = EINVAL;
goto bad;
}
#ifdef KTRACE
if (KTRPOINT(p, KTR_EXECARGS))
ktrexec(p, KTR_EXECARGS, argp, dp - argp);
#endif
envc = 0;
/* environment does not need to be there */
if ((cpp = SCARG(uap, envp)) != NULL ) {
#ifdef KTRACE
env_start = dp;
#endif
while (1) {
len = argp + ARG_MAX - dp;
if ((error = copyin(cpp, &sp, sizeof(sp))) != 0)
goto bad;
if (!sp)
break;
if ((error = copyinstr(sp, dp, len, &len)) != 0) {
if (error == ENAMETOOLONG)
error = E2BIG;
goto bad;
}
dp += len;
cpp++;
envc++;
}
#ifdef KTRACE
if (KTRPOINT(p, KTR_EXECENV))
ktrexec(p, KTR_EXECENV, env_start, dp - env_start);
#endif
}
dp = (char *)(((long)dp + _STACKALIGNBYTES) & ~_STACKALIGNBYTES);
sgap = STACKGAPLEN;
/*
* If we have enabled random stackgap, the stack itself has already
* been moved from a random location, but is still aligned to a page
* boundary. Provide the lower bits of random placement now.
*/
if (stackgap_random != 0) {
sgap += arc4random() & PAGE_MASK;
sgap = (sgap + _STACKALIGNBYTES) & ~_STACKALIGNBYTES;
}
/* Now check if args & environ fit into new stack */
len = ((argc + envc + 2 + pack.ep_emul->e_arglen) * sizeof(char *) +
sizeof(long) + dp + sgap + sizeof(struct ps_strings)) - argp;
len = (len + _STACKALIGNBYTES) &~ _STACKALIGNBYTES;
if (len > pack.ep_ssize) { /* in effect, compare to initial limit */
error = ENOMEM;
goto bad;
}
/* adjust "active stack depth" for process VSZ */
pack.ep_ssize = len; /* maybe should go elsewhere, but... */
/*
* we're committed: any further errors will kill the process, so
* kill the other threads now.
*/
single_thread_set(p, SINGLE_EXIT, 0);
/*
* Prepare vmspace for remapping. Note that uvmspace_exec can replace
* pr_vmspace!
*/
uvmspace_exec(p, VM_MIN_ADDRESS, VM_MAXUSER_ADDRESS);
vm = pr->ps_vmspace;
/* Now map address space */
vm->vm_taddr = (char *)trunc_page(pack.ep_taddr);
vm->vm_tsize = atop(round_page(pack.ep_taddr + pack.ep_tsize) -
trunc_page(pack.ep_taddr));
vm->vm_daddr = (char *)trunc_page(pack.ep_daddr);
vm->vm_dsize = atop(round_page(pack.ep_daddr + pack.ep_dsize) -
trunc_page(pack.ep_daddr));
vm->vm_dused = 0;
vm->vm_ssize = atop(round_page(pack.ep_ssize));
vm->vm_maxsaddr = (char *)pack.ep_maxsaddr;
vm->vm_minsaddr = (char *)pack.ep_minsaddr;
/* create the new process's VM space by running the vmcmds */
#ifdef DIAGNOSTIC
if (pack.ep_vmcmds.evs_used == 0)
panic("execve: no vmcmds");
#endif
error = exec_process_vmcmds(p, &pack);
/* if an error happened, deallocate and punt */
if (error)
goto exec_abort;
/* old "stackgap" is gone now */
pr->ps_stackgap = 0;
#ifdef MACHINE_STACK_GROWS_UP
pr->ps_strings = (vaddr_t)vm->vm_maxsaddr + sgap;
if (uvm_map_protect(&vm->vm_map, (vaddr_t)vm->vm_maxsaddr,
trunc_page(pr->ps_strings), PROT_NONE, TRUE))
goto exec_abort;
#else
pr->ps_strings = (vaddr_t)vm->vm_minsaddr - sizeof(arginfo) - sgap;
if (uvm_map_protect(&vm->vm_map,
round_page(pr->ps_strings + sizeof(arginfo)),
(vaddr_t)vm->vm_minsaddr, PROT_NONE, TRUE))
goto exec_abort;
#endif
/* remember information about the process */
arginfo.ps_nargvstr = argc;
arginfo.ps_nenvstr = envc;
#ifdef MACHINE_STACK_GROWS_UP
stack = (char *)vm->vm_maxsaddr + sizeof(arginfo) + sgap;
slen = len - sizeof(arginfo) - sgap;
#else
stack = (char *)(vm->vm_minsaddr - len);
#endif
/* Now copy argc, args & environ to new stack */
if (!(*pack.ep_emul->e_copyargs)(&pack, &arginfo, stack, argp))
goto exec_abort;
/* copy out the process's ps_strings structure */
if (copyout(&arginfo, (char *)pr->ps_strings, sizeof(arginfo)))
goto exec_abort;
stopprofclock(pr); /* stop profiling */
fdcloseexec(p); /* handle close on exec */
execsigs(p); /* reset caught signals */
TCB_SET(p, NULL); /* reset the TCB address */
pr->ps_kbind_addr = 0; /* reset the kbind bits */
pr->ps_kbind_cookie = 0;
/* set command name & other accounting info */
memset(p->p_comm, 0, sizeof(p->p_comm));
len = min(nid.ni_cnd.cn_namelen, MAXCOMLEN);
memcpy(p->p_comm, nid.ni_cnd.cn_nameptr, len);
pr->ps_acflag &= ~AFORK;
/* record proc's vnode, for use by sysctl */
otvp = pr->ps_textvp;
vref(pack.ep_vp);
pr->ps_textvp = pack.ep_vp;
if (otvp)
vrele(otvp);
atomic_setbits_int(&pr->ps_flags, PS_EXEC);
if (pr->ps_flags & PS_PPWAIT) {
atomic_clearbits_int(&pr->ps_flags, PS_PPWAIT);
atomic_clearbits_int(&pr->ps_pptr->ps_flags, PS_ISPWAIT);
wakeup(pr->ps_pptr);
}
/*
* If process does execve() while it has a mismatched real,
* effective, or saved uid/gid, we set PS_SUGIDEXEC.
*/
if (cred->cr_uid != cred->cr_ruid ||
cred->cr_uid != cred->cr_svuid ||
cred->cr_gid != cred->cr_rgid ||
cred->cr_gid != cred->cr_svgid)
atomic_setbits_int(&pr->ps_flags, PS_SUGIDEXEC);
else
atomic_clearbits_int(&pr->ps_flags, PS_SUGIDEXEC);
atomic_clearbits_int(&pr->ps_flags, PS_TAMED);
tame_dropwpaths(pr);
/*
* deal with set[ug]id.
* MNT_NOEXEC has already been used to disable s[ug]id.
*/
if ((attr.va_mode & (VSUID | VSGID)) && proc_cansugid(p)) {
int i;
atomic_setbits_int(&pr->ps_flags, PS_SUGID|PS_SUGIDEXEC);
#ifdef KTRACE
/*
* If process is being ktraced, turn off - unless
* root set it.
*/
if (pr->ps_tracevp && !(pr->ps_traceflag & KTRFAC_ROOT))
ktrcleartrace(pr);
#endif
p->p_ucred = cred = crcopy(cred);
if (attr.va_mode & VSUID)
cred->cr_uid = attr.va_uid;
if (attr.va_mode & VSGID)
cred->cr_gid = attr.va_gid;
/*
* For set[ug]id processes, a few caveats apply to
* stdin, stdout, and stderr.
*/
error = 0;
fdplock(p->p_fd);
for (i = 0; i < 3; i++) {
struct file *fp = NULL;
/*
* NOTE - This will never return NULL because of
* immature fds. The file descriptor table is not
* shared because we're suid.
*/
fp = fd_getfile(p->p_fd, i);
/*
* Ensure that stdin, stdout, and stderr are already
* allocated. We do not want userland to accidentally
* allocate descriptors in this range which has implied
* meaning to libc.
*/
if (fp == NULL) {
short flags = FREAD | (i == 0 ? 0 : FWRITE);
struct vnode *vp;
int indx;
if ((error = falloc(p, &fp, &indx)) != 0)
break;
#ifdef DIAGNOSTIC
if (indx != i)
panic("sys_execve: falloc indx != i");
#endif
if ((error = cdevvp(getnulldev(), &vp)) != 0) {
fdremove(p->p_fd, indx);
closef(fp, p);
break;
}
if ((error = VOP_OPEN(vp, flags, cred, p)) != 0) {
fdremove(p->p_fd, indx);
closef(fp, p);
vrele(vp);
break;
}
if (flags & FWRITE)
vp->v_writecount++;
fp->f_flag = flags;
fp->f_type = DTYPE_VNODE;
fp->f_ops = &vnops;
fp->f_data = (caddr_t)vp;
FILE_SET_MATURE(fp, p);
}
}
fdpunlock(p->p_fd);
if (error)
goto exec_abort;
} else
atomic_clearbits_int(&pr->ps_flags, PS_SUGID);
/*
* Reset the saved ugids and update the process's copy of the
* creds if the creds have been changed
*/
if (cred->cr_uid != cred->cr_svuid ||
cred->cr_gid != cred->cr_svgid) {
/* make sure we have unshared ucreds */
p->p_ucred = cred = crcopy(cred);
cred->cr_svuid = cred->cr_uid;
cred->cr_svgid = cred->cr_gid;
}
if (pr->ps_ucred != cred) {
struct ucred *ocred;
ocred = pr->ps_ucred;
crhold(cred);
pr->ps_ucred = cred;
crfree(ocred);
}
if (pr->ps_flags & PS_SUGIDEXEC) {
int i, s = splclock();
timeout_del(&pr->ps_realit_to);
for (i = 0; i < nitems(pr->ps_timer); i++) {
timerclear(&pr->ps_timer[i].it_interval);
timerclear(&pr->ps_timer[i].it_value);
}
splx(s);
}
/* reset CPU time usage for the thread, but not the process */
timespecclear(&p->p_tu.tu_runtime);
p->p_tu.tu_uticks = p->p_tu.tu_sticks = p->p_tu.tu_iticks = 0;
km_free(argp, NCARGS, &kv_exec, &kp_pageable);
pool_put(&namei_pool, nid.ni_cnd.cn_pnbuf);
vn_close(pack.ep_vp, FREAD, cred, p);
/*
* notify others that we exec'd
*/
KNOTE(&pr->ps_klist, NOTE_EXEC);
/* setup new registers and do misc. setup. */
if (pack.ep_emul->e_fixup != NULL) {
if ((*pack.ep_emul->e_fixup)(p, &pack) != 0)
goto free_pack_abort;
}
#ifdef MACHINE_STACK_GROWS_UP
(*pack.ep_emul->e_setregs)(p, &pack, (u_long)stack + slen, retval);
#else
(*pack.ep_emul->e_setregs)(p, &pack, (u_long)stack, retval);
#endif
/* map the process's signal trampoline code */
if (exec_sigcode_map(pr, pack.ep_emul))
goto free_pack_abort;
#ifdef __HAVE_EXEC_MD_MAP
/* perform md specific mappings that process might need */
if (exec_md_map(p, &pack))
goto free_pack_abort;
#endif
if (pr->ps_flags & PS_TRACED)
psignal(p, SIGTRAP);
free(pack.ep_hdr, M_EXEC, pack.ep_hdrlen);
/*
* Call emulation specific exec hook. This can setup per-process
* p->p_emuldata or do any other per-process stuff an emulation needs.
*
* If we are executing process of different emulation than the
* original forked process, call e_proc_exit() of the old emulation
* first, then e_proc_exec() of new emulation. If the emulation is
* same, the exec hook code should deallocate any old emulation
* resources held previously by this process.
*/
if (pr->ps_emul && pr->ps_emul->e_proc_exit &&
pr->ps_emul != pack.ep_emul)
(*pr->ps_emul->e_proc_exit)(p);
p->p_descfd = 255;
if ((pack.ep_flags & EXEC_HASFD) && pack.ep_fd < 255)
p->p_descfd = pack.ep_fd;
/*
* Call exec hook. Emulation code may NOT store reference to anything
* from &pack.
*/
if (pack.ep_emul->e_proc_exec)
(*pack.ep_emul->e_proc_exec)(p, &pack);
#if defined(KTRACE) && defined(COMPAT_LINUX)
/* update ps_emul, but don't ktrace it if native-execing-native */
if (pr->ps_emul != pack.ep_emul || pack.ep_emul != &emul_native) {
pr->ps_emul = pack.ep_emul;
if (KTRPOINT(p, KTR_EMUL))
ktremul(p);
}
#else
/* update ps_emul, the old value is no longer needed */
pr->ps_emul = pack.ep_emul;
#endif
atomic_clearbits_int(&pr->ps_flags, PS_INEXEC);
single_thread_clear(p, P_SUSPSIG);
#if NSYSTRACE > 0
if (ISSET(p->p_flag, P_SYSTRACE) &&
wassugid && !ISSET(pr->ps_flags, PS_SUGID | PS_SUGIDEXEC))
systrace_execve1(pathbuf, p);
#endif
if (pathbuf != NULL)
pool_put(&namei_pool, pathbuf);
return (0);
bad:
/* free the vmspace-creation commands, and release their references */
kill_vmcmds(&pack.ep_vmcmds);
/* kill any opened file descriptor, if necessary */
if (pack.ep_flags & EXEC_HASFD) {
pack.ep_flags &= ~EXEC_HASFD;
fdplock(p->p_fd);
(void) fdrelease(p, pack.ep_fd);
fdpunlock(p->p_fd);
}
if (pack.ep_interp != NULL)
pool_put(&namei_pool, pack.ep_interp);
if (pack.ep_emul_arg != NULL)
free(pack.ep_emul_arg, M_TEMP, pack.ep_emul_argsize);
/* close and put the exec'd file */
vn_close(pack.ep_vp, FREAD, cred, p);
pool_put(&namei_pool, nid.ni_cnd.cn_pnbuf);
km_free(argp, NCARGS, &kv_exec, &kp_pageable);
freehdr:
free(pack.ep_hdr, M_EXEC, pack.ep_hdrlen);
#if NSYSTRACE > 0
clrflag:
#endif
atomic_clearbits_int(&pr->ps_flags, PS_INEXEC);
single_thread_clear(p, P_SUSPSIG);
if (pathbuf != NULL)
pool_put(&namei_pool, pathbuf);
return (error);
exec_abort:
/*
* the old process doesn't exist anymore. exit gracefully.
* get rid of the (new) address space we have created, if any, get rid
* of our namei data and vnode, and exit noting failure
*/
uvm_deallocate(&vm->vm_map, VM_MIN_ADDRESS,
VM_MAXUSER_ADDRESS - VM_MIN_ADDRESS);
if (pack.ep_interp != NULL)
pool_put(&namei_pool, pack.ep_interp);
if (pack.ep_emul_arg != NULL)
free(pack.ep_emul_arg, M_TEMP, pack.ep_emul_argsize);
pool_put(&namei_pool, nid.ni_cnd.cn_pnbuf);
vn_close(pack.ep_vp, FREAD, cred, p);
km_free(argp, NCARGS, &kv_exec, &kp_pageable);
free_pack_abort:
free(pack.ep_hdr, M_EXEC, pack.ep_hdrlen);
if (pathbuf != NULL)
pool_put(&namei_pool, pathbuf);
exit1(p, W_EXITCODE(0, SIGABRT), EXIT_NORMAL);
/* NOTREACHED */
atomic_clearbits_int(&pr->ps_flags, PS_INEXEC);
return (0);
}
/*
* Exit: deallocate address space and other resources, change proc state to
* zombie, and unlink proc from allproc and parent's lists. Save exit status
* and rusage for wait(). Check for child processes and orphan them.
*/
void
exit1(struct thread *td, int rv)
{
struct proc *p, *nq, *q;
struct vnode *vtmp;
struct vnode *ttyvp = NULL;
struct plimit *plim;
mtx_assert(&Giant, MA_NOTOWNED);
p = td->td_proc;
/*
* XXX in case we're rebooting we just let init die in order to
* work around an unsolved stack overflow seen very late during
* shutdown on sparc64 when the gmirror worker process exists.
*/
if (p == initproc && rebooting == 0) {
printf("init died (signal %d, exit %d)\n",
WTERMSIG(rv), WEXITSTATUS(rv));
panic("Going nowhere without my init!");
}
/*
* MUST abort all other threads before proceeding past here.
*/
PROC_LOCK(p);
while (p->p_flag & P_HADTHREADS) {
/*
* First check if some other thread got here before us.
* If so, act appropriately: exit or suspend.
*/
thread_suspend_check(0);
/*
* Kill off the other threads. This requires
* some co-operation from other parts of the kernel
* so it may not be instantaneous. With this state set
* any thread entering the kernel from userspace will
* thread_exit() in trap(). Any thread attempting to
* sleep will return immediately with EINTR or EWOULDBLOCK
* which will hopefully force them to back out to userland
* freeing resources as they go. Any thread attempting
* to return to userland will thread_exit() from userret().
* thread_exit() will unsuspend us when the last of the
* other threads exits.
* If there is already a thread singler after resumption,
* calling thread_single will fail; in that case, we just
* re-check all suspension request, the thread should
* either be suspended there or exit.
*/
if (!thread_single(SINGLE_EXIT))
break;
/*
* All other activity in this process is now stopped.
* Threading support has been turned off.
*/
}
KASSERT(p->p_numthreads == 1,
("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
racct_sub(p, RACCT_NTHR, 1);
/*
* Wakeup anyone in procfs' PIOCWAIT. They should have a hold
* on our vmspace, so we should block below until they have
* released their reference to us. Note that if they have
* requested S_EXIT stops we will block here until they ack
* via PIOCCONT.
*/
_STOPEVENT(p, S_EXIT, rv);
/*
* Ignore any pending request to stop due to a stop signal.
* Once P_WEXIT is set, future requests will be ignored as
* well.
*/
p->p_flag &= ~P_STOPPED_SIG;
KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
/*
* Note that we are exiting and do another wakeup of anyone in
* PIOCWAIT in case they aren't listening for S_EXIT stops or
* decided to wait again after we told them we are exiting.
*/
p->p_flag |= P_WEXIT;
wakeup(&p->p_stype);
/*
* Wait for any processes that have a hold on our vmspace to
* release their reference.
*/
while (p->p_lock > 0)
msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
p->p_xstat = rv; /* Let event handler change exit status */
PROC_UNLOCK(p);
/* Drain the limit callout while we don't have the proc locked */
callout_drain(&p->p_limco);
#ifdef AUDIT
/*
* The Sun BSM exit token contains two components: an exit status as
* passed to exit(), and a return value to indicate what sort of exit
* it was. The exit status is WEXITSTATUS(rv), but it's not clear
* what the return value is.
*/
AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0);
AUDIT_SYSCALL_EXIT(0, td);
#endif
/* Are we a task leader? */
if (p == p->p_leader) {
mtx_lock(&ppeers_lock);
q = p->p_peers;
while (q != NULL) {
PROC_LOCK(q);
kern_psignal(q, SIGKILL);
PROC_UNLOCK(q);
q = q->p_peers;
}
while (p->p_peers != NULL)
msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
mtx_unlock(&ppeers_lock);
}
/*
* Check if any loadable modules need anything done at process exit.
* E.g. SYSV IPC stuff
* XXX what if one of these generates an error?
*/
EVENTHANDLER_INVOKE(process_exit, p);
/*
* If parent is waiting for us to exit or exec,
* P_PPWAIT is set; we will wakeup the parent below.
*/
PROC_LOCK(p);
rv = p->p_xstat; /* Event handler could change exit status */
stopprofclock(p);
p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
/*
* Stop the real interval timer. If the handler is currently
* executing, prevent it from rearming itself and let it finish.
*/
if (timevalisset(&p->p_realtimer.it_value) &&
callout_stop(&p->p_itcallout) == 0) {
timevalclear(&p->p_realtimer.it_interval);
msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
KASSERT(!timevalisset(&p->p_realtimer.it_value),
("realtime timer is still armed"));
}
PROC_UNLOCK(p);
/*
* Reset any sigio structures pointing to us as a result of
* F_SETOWN with our pid.
*/
funsetownlst(&p->p_sigiolst);
/*
* If this process has an nlminfo data area (for lockd), release it
*/
if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
(*nlminfo_release_p)(p);
/*
* Close open files and release open-file table.
* This may block!
*/
fdescfree(td);
/*
* If this thread tickled GEOM, we need to wait for the giggling to
* stop before we return to userland
*/
if (td->td_pflags & TDP_GEOM)
g_waitidle();
/*
* Remove ourself from our leader's peer list and wake our leader.
*/
mtx_lock(&ppeers_lock);
if (p->p_leader->p_peers) {
q = p->p_leader;
while (q->p_peers != p)
q = q->p_peers;
q->p_peers = p->p_peers;
wakeup(p->p_leader);
}
mtx_unlock(&ppeers_lock);
vmspace_exit(td);
sx_xlock(&proctree_lock);
if (SESS_LEADER(p)) {
struct session *sp = p->p_session;
struct tty *tp;
/*
* s_ttyp is not zero'd; we use this to indicate that
* the session once had a controlling terminal. (for
* logging and informational purposes)
*/
SESS_LOCK(sp);
ttyvp = sp->s_ttyvp;
tp = sp->s_ttyp;
sp->s_ttyvp = NULL;
sp->s_ttydp = NULL;
sp->s_leader = NULL;
SESS_UNLOCK(sp);
/*
* Signal foreground pgrp and revoke access to
* controlling terminal if it has not been revoked
* already.
*
* Because the TTY may have been revoked in the mean
* time and could already have a new session associated
* with it, make sure we don't send a SIGHUP to a
* foreground process group that does not belong to this
* session.
*/
if (tp != NULL) {
tty_lock(tp);
if (tp->t_session == sp)
tty_signal_pgrp(tp, SIGHUP);
tty_unlock(tp);
}
if (ttyvp != NULL) {
sx_xunlock(&proctree_lock);
if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
VOP_REVOKE(ttyvp, REVOKEALL);
VOP_UNLOCK(ttyvp, 0);
}
sx_xlock(&proctree_lock);
}
}
fixjobc(p, p->p_pgrp, 0);
sx_xunlock(&proctree_lock);
(void)acct_process(td);
/* Release the TTY now we've unlocked everything. */
if (ttyvp != NULL)
vrele(ttyvp);
#ifdef KTRACE
ktrprocexit(td);
#endif
/*
* Release reference to text vnode
*/
if ((vtmp = p->p_textvp) != NULL) {
p->p_textvp = NULL;
vrele(vtmp);
}
/*
* Release our limits structure.
*/
plim = p->p_limit;
p->p_limit = NULL;
lim_free(plim);
tidhash_remove(td);
/*
* Remove proc from allproc queue and pidhash chain.
* Place onto zombproc. Unlink from parent's child list.
*/
sx_xlock(&allproc_lock);
LIST_REMOVE(p, p_list);
LIST_INSERT_HEAD(&zombproc, p, p_list);
LIST_REMOVE(p, p_hash);
sx_xunlock(&allproc_lock);
/*
* Call machine-dependent code to release any
* machine-dependent resources other than the address space.
* The address space is released by "vmspace_exitfree(p)" in
* vm_waitproc().
*/
cpu_exit(td);
WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
/*
* Reparent all of our children to init.
*/
sx_xlock(&proctree_lock);
q = LIST_FIRST(&p->p_children);
if (q != NULL) /* only need this if any child is S_ZOMB */
wakeup(initproc);
for (; q != NULL; q = nq) {
nq = LIST_NEXT(q, p_sibling);
PROC_LOCK(q);
proc_reparent(q, initproc);
q->p_sigparent = SIGCHLD;
/*
* Traced processes are killed
* since their existence means someone is screwing up.
*/
if (q->p_flag & P_TRACED) {
struct thread *temp;
/*
* Since q was found on our children list, the
* proc_reparent() call moved q to the orphan
* list due to present P_TRACED flag. Clear
* orphan link for q now while q is locked.
*/
clear_orphan(q);
q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
FOREACH_THREAD_IN_PROC(q, temp)
temp->td_dbgflags &= ~TDB_SUSPEND;
kern_psignal(q, SIGKILL);
}
PROC_UNLOCK(q);
}
/*
* Also get rid of our orphans.
*/
while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
PROC_LOCK(q);
clear_orphan(q);
PROC_UNLOCK(q);
}
/* Save exit status. */
PROC_LOCK(p);
p->p_xthread = td;
/* Tell the prison that we are gone. */
prison_proc_free(p->p_ucred->cr_prison);
#ifdef KDTRACE_HOOKS
/*
* Tell the DTrace fasttrap provider about the exit if it
* has declared an interest.
*/
if (dtrace_fasttrap_exit)
dtrace_fasttrap_exit(p);
#endif
/*
* Notify interested parties of our demise.
*/
KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
#ifdef KDTRACE_HOOKS
int reason = CLD_EXITED;
if (WCOREDUMP(rv))
reason = CLD_DUMPED;
else if (WIFSIGNALED(rv))
reason = CLD_KILLED;
SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
#endif
/*
* Just delete all entries in the p_klist. At this point we won't
* report any more events, and there are nasty race conditions that
* can beat us if we don't.
*/
knlist_clear(&p->p_klist, 1);
/*
* If this is a process with a descriptor, we may not need to deliver
* a signal to the parent. proctree_lock is held over
* procdesc_exit() to serialize concurrent calls to close() and
* exit().
*/
if (p->p_procdesc == NULL || procdesc_exit(p)) {
/*
* Notify parent that we're gone. If parent has the
* PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
* notify process 1 instead (and hope it will handle this
* situation).
*/
PROC_LOCK(p->p_pptr);
mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
if (p->p_pptr->p_sigacts->ps_flag &
(PS_NOCLDWAIT | PS_CLDSIGIGN)) {
struct proc *pp;
mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
pp = p->p_pptr;
PROC_UNLOCK(pp);
proc_reparent(p, initproc);
p->p_sigparent = SIGCHLD;
PROC_LOCK(p->p_pptr);
/*
* Notify parent, so in case he was wait(2)ing or
* executing waitpid(2) with our pid, he will
* continue.
*/
wakeup(pp);
} else
mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
if (p->p_pptr == initproc)
kern_psignal(p->p_pptr, SIGCHLD);
else if (p->p_sigparent != 0) {
if (p->p_sigparent == SIGCHLD)
childproc_exited(p);
else /* LINUX thread */
kern_psignal(p->p_pptr, p->p_sigparent);
}
} else
PROC_LOCK(p->p_pptr);
sx_xunlock(&proctree_lock);
/*
* The state PRS_ZOMBIE prevents other proesses from sending
* signal to the process, to avoid memory leak, we free memory
* for signal queue at the time when the state is set.
*/
sigqueue_flush(&p->p_sigqueue);
sigqueue_flush(&td->td_sigqueue);
/*
* We have to wait until after acquiring all locks before
* changing p_state. We need to avoid all possible context
* switches (including ones from blocking on a mutex) while
* marked as a zombie. We also have to set the zombie state
* before we release the parent process' proc lock to avoid
* a lost wakeup. So, we first call wakeup, then we grab the
* sched lock, update the state, and release the parent process'
* proc lock.
*/
wakeup(p->p_pptr);
cv_broadcast(&p->p_pwait);
sched_exit(p->p_pptr, td);
PROC_SLOCK(p);
p->p_state = PRS_ZOMBIE;
PROC_UNLOCK(p->p_pptr);
/*
* Hopefully no one will try to deliver a signal to the process this
* late in the game.
*/
knlist_destroy(&p->p_klist);
/*
* Save our children's rusage information in our exit rusage.
*/
ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
/*
* Make sure the scheduler takes this thread out of its tables etc.
* This will also release this thread's reference to the ucred.
* Other thread parts to release include pcb bits and such.
*/
thread_exit();
}
/*
* In-kernel implementation of execve(). All arguments are assumed to be
* userspace pointers from the passed thread.
*/
static int
do_execve(struct thread *td, struct image_args *args, struct mac *mac_p)
{
struct proc *p = td->td_proc;
struct nameidata nd;
struct ucred *oldcred;
struct uidinfo *euip = NULL;
register_t *stack_base;
int error, i;
struct image_params image_params, *imgp;
struct vattr attr;
int (*img_first)(struct image_params *);
struct pargs *oldargs = NULL, *newargs = NULL;
struct sigacts *oldsigacts = NULL, *newsigacts = NULL;
#ifdef KTRACE
struct vnode *tracevp = NULL;
struct ucred *tracecred = NULL;
#endif
struct vnode *oldtextvp = NULL, *newtextvp;
int credential_changing;
int textset;
#ifdef MAC
struct label *interpvplabel = NULL;
int will_transition;
#endif
#ifdef HWPMC_HOOKS
struct pmckern_procexec pe;
#endif
static const char fexecv_proc_title[] = "(fexecv)";
imgp = &image_params;
/*
* Lock the process and set the P_INEXEC flag to indicate that
* it should be left alone until we're done here. This is
* necessary to avoid race conditions - e.g. in ptrace() -
* that might allow a local user to illicitly obtain elevated
* privileges.
*/
PROC_LOCK(p);
KASSERT((p->p_flag & P_INEXEC) == 0,
("%s(): process already has P_INEXEC flag", __func__));
p->p_flag |= P_INEXEC;
PROC_UNLOCK(p);
/*
* Initialize part of the common data
*/
bzero(imgp, sizeof(*imgp));
imgp->proc = p;
imgp->attr = &attr;
imgp->args = args;
oldcred = p->p_ucred;
#ifdef MAC
error = mac_execve_enter(imgp, mac_p);
if (error)
goto exec_fail;
#endif
/*
* Translate the file name. namei() returns a vnode pointer
* in ni_vp among other things.
*
* XXXAUDIT: It would be desirable to also audit the name of the
* interpreter if this is an interpreted binary.
*/
if (args->fname != NULL) {
NDINIT(&nd, LOOKUP, ISOPEN | LOCKLEAF | FOLLOW | SAVENAME
| AUDITVNODE1, UIO_SYSSPACE, args->fname, td);
}
SDT_PROBE1(proc, , , exec, args->fname);
interpret:
if (args->fname != NULL) {
#ifdef CAPABILITY_MODE
/*
* While capability mode can't reach this point via direct
* path arguments to execve(), we also don't allow
* interpreters to be used in capability mode (for now).
* Catch indirect lookups and return a permissions error.
*/
if (IN_CAPABILITY_MODE(td)) {
error = ECAPMODE;
goto exec_fail;
}
#endif
error = namei(&nd);
if (error)
goto exec_fail;
newtextvp = nd.ni_vp;
imgp->vp = newtextvp;
} else {
AUDIT_ARG_FD(args->fd);
/*
* Descriptors opened only with O_EXEC or O_RDONLY are allowed.
*/
error = fgetvp_exec(td, args->fd, &cap_fexecve_rights, &newtextvp);
if (error)
goto exec_fail;
vn_lock(newtextvp, LK_EXCLUSIVE | LK_RETRY);
AUDIT_ARG_VNODE1(newtextvp);
imgp->vp = newtextvp;
}
/*
* Check file permissions (also 'opens' file)
*/
error = exec_check_permissions(imgp);
if (error)
goto exec_fail_dealloc;
imgp->object = imgp->vp->v_object;
if (imgp->object != NULL)
vm_object_reference(imgp->object);
/*
* Set VV_TEXT now so no one can write to the executable while we're
* activating it.
*
* Remember if this was set before and unset it in case this is not
* actually an executable image.
*/
textset = VOP_IS_TEXT(imgp->vp);
VOP_SET_TEXT(imgp->vp);
error = exec_map_first_page(imgp);
if (error)
goto exec_fail_dealloc;
imgp->proc->p_osrel = 0;
imgp->proc->p_fctl0 = 0;
/*
* Implement image setuid/setgid.
*
* Determine new credentials before attempting image activators
* so that it can be used by process_exec handlers to determine
* credential/setid changes.
*
* Don't honor setuid/setgid if the filesystem prohibits it or if
* the process is being traced.
*
* We disable setuid/setgid/etc in capability mode on the basis
* that most setugid applications are not written with that
* environment in mind, and will therefore almost certainly operate
* incorrectly. In principle there's no reason that setugid
* applications might not be useful in capability mode, so we may want
* to reconsider this conservative design choice in the future.
*
* XXXMAC: For the time being, use NOSUID to also prohibit
* transitions on the file system.
*/
credential_changing = 0;
credential_changing |= (attr.va_mode & S_ISUID) &&
oldcred->cr_uid != attr.va_uid;
credential_changing |= (attr.va_mode & S_ISGID) &&
oldcred->cr_gid != attr.va_gid;
#ifdef MAC
will_transition = mac_vnode_execve_will_transition(oldcred, imgp->vp,
interpvplabel, imgp);
credential_changing |= will_transition;
#endif
/* Don't inherit PROC_PDEATHSIG_CTL value if setuid/setgid. */
if (credential_changing)
imgp->proc->p_pdeathsig = 0;
if (credential_changing &&
#ifdef CAPABILITY_MODE
((oldcred->cr_flags & CRED_FLAG_CAPMODE) == 0) &&
#endif
(imgp->vp->v_mount->mnt_flag & MNT_NOSUID) == 0 &&
(p->p_flag & P_TRACED) == 0) {
imgp->credential_setid = true;
VOP_UNLOCK(imgp->vp, 0);
imgp->newcred = crdup(oldcred);
if (attr.va_mode & S_ISUID) {
euip = uifind(attr.va_uid);
change_euid(imgp->newcred, euip);
}
vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
if (attr.va_mode & S_ISGID)
change_egid(imgp->newcred, attr.va_gid);
/*
* Implement correct POSIX saved-id behavior.
*
* XXXMAC: Note that the current logic will save the
* uid and gid if a MAC domain transition occurs, even
* though maybe it shouldn't.
*/
change_svuid(imgp->newcred, imgp->newcred->cr_uid);
change_svgid(imgp->newcred, imgp->newcred->cr_gid);
} else {
/*
* Implement correct POSIX saved-id behavior.
*
* XXX: It's not clear that the existing behavior is
* POSIX-compliant. A number of sources indicate that the
* saved uid/gid should only be updated if the new ruid is
* not equal to the old ruid, or the new euid is not equal
* to the old euid and the new euid is not equal to the old
* ruid. The FreeBSD code always updates the saved uid/gid.
* Also, this code uses the new (replaced) euid and egid as
* the source, which may or may not be the right ones to use.
*/
if (oldcred->cr_svuid != oldcred->cr_uid ||
oldcred->cr_svgid != oldcred->cr_gid) {
VOP_UNLOCK(imgp->vp, 0);
imgp->newcred = crdup(oldcred);
vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
change_svuid(imgp->newcred, imgp->newcred->cr_uid);
change_svgid(imgp->newcred, imgp->newcred->cr_gid);
}
}
/* The new credentials are installed into the process later. */
/*
* Do the best to calculate the full path to the image file.
*/
if (args->fname != NULL && args->fname[0] == '/')
imgp->execpath = args->fname;
else {
VOP_UNLOCK(imgp->vp, 0);
if (vn_fullpath(td, imgp->vp, &imgp->execpath,
&imgp->freepath) != 0)
imgp->execpath = args->fname;
vn_lock(imgp->vp, LK_EXCLUSIVE | LK_RETRY);
}
/*
* If the current process has a special image activator it
* wants to try first, call it. For example, emulating shell
* scripts differently.
*/
error = -1;
if ((img_first = imgp->proc->p_sysent->sv_imgact_try) != NULL)
error = img_first(imgp);
/*
* Loop through the list of image activators, calling each one.
* An activator returns -1 if there is no match, 0 on success,
* and an error otherwise.
*/
for (i = 0; error == -1 && execsw[i]; ++i) {
if (execsw[i]->ex_imgact == NULL ||
execsw[i]->ex_imgact == img_first) {
continue;
}
error = (*execsw[i]->ex_imgact)(imgp);
}
if (error) {
if (error == -1) {
if (textset == 0)
VOP_UNSET_TEXT(imgp->vp);
error = ENOEXEC;
}
goto exec_fail_dealloc;
}
/*
* Special interpreter operation, cleanup and loop up to try to
* activate the interpreter.
*/
if (imgp->interpreted) {
exec_unmap_first_page(imgp);
/*
* VV_TEXT needs to be unset for scripts. There is a short
* period before we determine that something is a script where
* VV_TEXT will be set. The vnode lock is held over this
* entire period so nothing should illegitimately be blocked.
*/
VOP_UNSET_TEXT(imgp->vp);
/* free name buffer and old vnode */
if (args->fname != NULL)
NDFREE(&nd, NDF_ONLY_PNBUF);
#ifdef MAC
mac_execve_interpreter_enter(newtextvp, &interpvplabel);
#endif
if (imgp->opened) {
VOP_CLOSE(newtextvp, FREAD, td->td_ucred, td);
imgp->opened = 0;
}
vput(newtextvp);
vm_object_deallocate(imgp->object);
imgp->object = NULL;
imgp->credential_setid = false;
if (imgp->newcred != NULL) {
crfree(imgp->newcred);
imgp->newcred = NULL;
}
imgp->execpath = NULL;
free(imgp->freepath, M_TEMP);
imgp->freepath = NULL;
/* set new name to that of the interpreter */
NDINIT(&nd, LOOKUP, LOCKLEAF | FOLLOW | SAVENAME,
UIO_SYSSPACE, imgp->interpreter_name, td);
args->fname = imgp->interpreter_name;
goto interpret;
}
/*
* NB: We unlock the vnode here because it is believed that none
* of the sv_copyout_strings/sv_fixup operations require the vnode.
*/
VOP_UNLOCK(imgp->vp, 0);
if (disallow_high_osrel &&
P_OSREL_MAJOR(p->p_osrel) > P_OSREL_MAJOR(__FreeBSD_version)) {
error = ENOEXEC;
uprintf("Osrel %d for image %s too high\n", p->p_osrel,
imgp->execpath != NULL ? imgp->execpath : "<unresolved>");
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
goto exec_fail_dealloc;
}
/* ABI enforces the use of Capsicum. Switch into capabilities mode. */
if (SV_PROC_FLAG(p, SV_CAPSICUM))
sys_cap_enter(td, NULL);
/*
* Copy out strings (args and env) and initialize stack base.
*/
stack_base = (*p->p_sysent->sv_copyout_strings)(imgp);
/*
* Stack setup.
*/
error = (*p->p_sysent->sv_fixup)(&stack_base, imgp);
if (error != 0) {
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
goto exec_fail_dealloc;
}
if (args->fdp != NULL) {
/* Install a brand new file descriptor table. */
fdinstall_remapped(td, args->fdp);
args->fdp = NULL;
} else {
/*
* Keep on using the existing file descriptor table. For
* security and other reasons, the file descriptor table
* cannot be shared after an exec.
*/
fdunshare(td);
/* close files on exec */
fdcloseexec(td);
}
/*
* Malloc things before we need locks.
*/
i = exec_args_get_begin_envv(imgp->args) - imgp->args->begin_argv;
/* Cache arguments if they fit inside our allowance */
if (ps_arg_cache_limit >= i + sizeof(struct pargs)) {
newargs = pargs_alloc(i);
bcopy(imgp->args->begin_argv, newargs->ar_args, i);
}
/*
* For security and other reasons, signal handlers cannot
* be shared after an exec. The new process gets a copy of the old
* handlers. In execsigs(), the new process will have its signals
* reset.
*/
if (sigacts_shared(p->p_sigacts)) {
oldsigacts = p->p_sigacts;
newsigacts = sigacts_alloc();
sigacts_copy(newsigacts, oldsigacts);
}
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
PROC_LOCK(p);
if (oldsigacts)
p->p_sigacts = newsigacts;
/* Stop profiling */
stopprofclock(p);
/* reset caught signals */
execsigs(p);
/* name this process - nameiexec(p, ndp) */
bzero(p->p_comm, sizeof(p->p_comm));
if (args->fname)
bcopy(nd.ni_cnd.cn_nameptr, p->p_comm,
min(nd.ni_cnd.cn_namelen, MAXCOMLEN));
else if (vn_commname(newtextvp, p->p_comm, sizeof(p->p_comm)) != 0)
bcopy(fexecv_proc_title, p->p_comm, sizeof(fexecv_proc_title));
bcopy(p->p_comm, td->td_name, sizeof(td->td_name));
#ifdef KTR
sched_clear_tdname(td);
#endif
/*
* mark as execed, wakeup the process that vforked (if any) and tell
* it that it now has its own resources back
*/
p->p_flag |= P_EXEC;
if ((p->p_flag2 & P2_NOTRACE_EXEC) == 0)
p->p_flag2 &= ~P2_NOTRACE;
if (p->p_flag & P_PPWAIT) {
p->p_flag &= ~(P_PPWAIT | P_PPTRACE);
cv_broadcast(&p->p_pwait);
/* STOPs are no longer ignored, arrange for AST */
signotify(td);
}
/*
* Implement image setuid/setgid installation.
*/
if (imgp->credential_setid) {
/*
* Turn off syscall tracing for set-id programs, except for
* root. Record any set-id flags first to make sure that
* we do not regain any tracing during a possible block.
*/
setsugid(p);
#ifdef KTRACE
if (p->p_tracecred != NULL &&
priv_check_cred(p->p_tracecred, PRIV_DEBUG_DIFFCRED))
ktrprocexec(p, &tracecred, &tracevp);
#endif
/*
* Close any file descriptors 0..2 that reference procfs,
* then make sure file descriptors 0..2 are in use.
*
* Both fdsetugidsafety() and fdcheckstd() may call functions
* taking sleepable locks, so temporarily drop our locks.
*/
PROC_UNLOCK(p);
VOP_UNLOCK(imgp->vp, 0);
fdsetugidsafety(td);
error = fdcheckstd(td);
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
if (error != 0)
goto exec_fail_dealloc;
PROC_LOCK(p);
#ifdef MAC
if (will_transition) {
mac_vnode_execve_transition(oldcred, imgp->newcred,
imgp->vp, interpvplabel, imgp);
}
#endif
} else {
if (oldcred->cr_uid == oldcred->cr_ruid &&
oldcred->cr_gid == oldcred->cr_rgid)
p->p_flag &= ~P_SUGID;
}
/*
* Set the new credentials.
*/
if (imgp->newcred != NULL) {
proc_set_cred(p, imgp->newcred);
crfree(oldcred);
oldcred = NULL;
}
/*
* Store the vp for use in procfs. This vnode was referenced by namei
* or fgetvp_exec.
*/
oldtextvp = p->p_textvp;
p->p_textvp = newtextvp;
#ifdef KDTRACE_HOOKS
/*
* Tell the DTrace fasttrap provider about the exec if it
* has declared an interest.
*/
if (dtrace_fasttrap_exec)
dtrace_fasttrap_exec(p);
#endif
/*
* Notify others that we exec'd, and clear the P_INEXEC flag
* as we're now a bona fide freshly-execed process.
*/
KNOTE_LOCKED(p->p_klist, NOTE_EXEC);
p->p_flag &= ~P_INEXEC;
/* clear "fork but no exec" flag, as we _are_ execing */
p->p_acflag &= ~AFORK;
/*
* Free any previous argument cache and replace it with
* the new argument cache, if any.
*/
oldargs = p->p_args;
p->p_args = newargs;
newargs = NULL;
PROC_UNLOCK(p);
#ifdef HWPMC_HOOKS
/*
* Check if system-wide sampling is in effect or if the
* current process is using PMCs. If so, do exec() time
* processing. This processing needs to happen AFTER the
* P_INEXEC flag is cleared.
*/
if (PMC_SYSTEM_SAMPLING_ACTIVE() || PMC_PROC_IS_USING_PMCS(p)) {
VOP_UNLOCK(imgp->vp, 0);
pe.pm_credentialschanged = credential_changing;
pe.pm_entryaddr = imgp->entry_addr;
PMC_CALL_HOOK_X(td, PMC_FN_PROCESS_EXEC, (void *) &pe);
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
}
#endif
/* Set values passed into the program in registers. */
(*p->p_sysent->sv_setregs)(td, imgp, (u_long)(uintptr_t)stack_base);
vfs_mark_atime(imgp->vp, td->td_ucred);
SDT_PROBE1(proc, , , exec__success, args->fname);
exec_fail_dealloc:
if (imgp->firstpage != NULL)
exec_unmap_first_page(imgp);
if (imgp->vp != NULL) {
if (args->fname)
NDFREE(&nd, NDF_ONLY_PNBUF);
if (imgp->opened)
VOP_CLOSE(imgp->vp, FREAD, td->td_ucred, td);
if (error != 0)
vput(imgp->vp);
else
VOP_UNLOCK(imgp->vp, 0);
}
if (imgp->object != NULL)
vm_object_deallocate(imgp->object);
free(imgp->freepath, M_TEMP);
if (error == 0) {
if (p->p_ptevents & PTRACE_EXEC) {
PROC_LOCK(p);
if (p->p_ptevents & PTRACE_EXEC)
td->td_dbgflags |= TDB_EXEC;
PROC_UNLOCK(p);
}
/*
* Stop the process here if its stop event mask has
* the S_EXEC bit set.
*/
STOPEVENT(p, S_EXEC, 0);
} else {
exec_fail:
/* we're done here, clear P_INEXEC */
PROC_LOCK(p);
p->p_flag &= ~P_INEXEC;
PROC_UNLOCK(p);
SDT_PROBE1(proc, , , exec__failure, error);
}
if (imgp->newcred != NULL && oldcred != NULL)
crfree(imgp->newcred);
#ifdef MAC
mac_execve_exit(imgp);
mac_execve_interpreter_exit(interpvplabel);
#endif
exec_free_args(args);
/*
* Handle deferred decrement of ref counts.
*/
if (oldtextvp != NULL)
vrele(oldtextvp);
#ifdef KTRACE
if (tracevp != NULL)
vrele(tracevp);
if (tracecred != NULL)
crfree(tracecred);
#endif
pargs_drop(oldargs);
pargs_drop(newargs);
if (oldsigacts != NULL)
sigacts_free(oldsigacts);
if (euip != NULL)
uifree(euip);
if (error && imgp->vmspace_destroyed) {
/* sorry, no more process anymore. exit gracefully */
exit1(td, 0, SIGABRT);
/* NOT REACHED */
}
#ifdef KTRACE
if (error == 0)
ktrprocctor(p);
#endif
/*
* We don't want cpu_set_syscall_retval() to overwrite any of
* the register values put in place by exec_setregs().
* Implementations of cpu_set_syscall_retval() will leave
* registers unmodified when returning EJUSTRETURN.
*/
return (error == 0 ? EJUSTRETURN : error);
}
