/*
* Update disk usage, and take corrective action.
*/
int
lfs_chkdq1(struct inode *ip, int64_t change, kauth_cred_t cred, int flags)
{
struct dquot *dq;
int i;
int ncurblocks, error;
if ((error = lfs_getinoquota(ip)) != 0)
return error;
if (change == 0)
return (0);
if (change < 0) {
for (i = 0; i < ULFS_MAXQUOTAS; i++) {
if ((dq = ip->i_dquot[i]) == NODQUOT)
continue;
mutex_enter(&dq->dq_interlock);
ncurblocks = dq->dq_curblocks + change;
if (ncurblocks >= 0)
dq->dq_curblocks = ncurblocks;
else
dq->dq_curblocks = 0;
dq->dq_flags &= ~DQ_WARN(QL_BLOCK);
dq->dq_flags |= DQ_MOD;
mutex_exit(&dq->dq_interlock);
}
return (0);
}
for (i = 0; i < ULFS_MAXQUOTAS; i++) {
if ((dq = ip->i_dquot[i]) == NODQUOT)
continue;
if ((flags & FORCE) == 0 &&
kauth_authorize_system(cred, KAUTH_SYSTEM_FS_QUOTA,
KAUTH_REQ_SYSTEM_FS_QUOTA_NOLIMIT, KAUTH_ARG(i),
KAUTH_ARG(QL_BLOCK), NULL) != 0) {
mutex_enter(&dq->dq_interlock);
error = chkdqchg(ip, change, cred, i);
mutex_exit(&dq->dq_interlock);
if (error != 0)
return (error);
}
}
for (i = 0; i < ULFS_MAXQUOTAS; i++) {
if ((dq = ip->i_dquot[i]) == NODQUOT)
continue;
mutex_enter(&dq->dq_interlock);
dq->dq_curblocks += change;
dq->dq_flags |= DQ_MOD;
mutex_exit(&dq->dq_interlock);
}
return (0);
}
/* This function is used by clock_settime and settimeofday */
static int
settime1(struct proc *p, const struct timespec *ts, bool check_kauth)
{
struct timespec delta, now;
int s;
/* WHAT DO WE DO ABOUT PENDING REAL-TIME TIMEOUTS??? */
s = splclock();
nanotime(&now);
timespecsub(ts, &now, &delta);
if (check_kauth && kauth_authorize_system(kauth_cred_get(),
KAUTH_SYSTEM_TIME, KAUTH_REQ_SYSTEM_TIME_SYSTEM, __UNCONST(ts),
&delta, KAUTH_ARG(check_kauth ? false : true)) != 0) {
splx(s);
return (EPERM);
}
#ifdef notyet
if ((delta.tv_sec < 86400) && securelevel > 0) { /* XXX elad - notyet */
splx(s);
return (EPERM);
}
#endif
tc_setclock(ts);
timespecadd(&boottime, &delta, &boottime);
resettodr();
splx(s);
return (0);
}
/* XXX shouldn't all this be in kauth ? */
static int
quota_get_auth(struct mount *mp, struct lwp *l, uid_t id) {
/* The user can always query about his own quota. */
if (id == kauth_cred_getuid(l->l_cred))
return 0;
return kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FS_QUOTA,
KAUTH_REQ_SYSTEM_FS_QUOTA_GET, mp, KAUTH_ARG(id), NULL);
}
static int
kill1(struct lwp *l, pid_t pid, ksiginfo_t *ksi, register_t *retval)
{
int error;
struct proc *p;
if ((u_int)ksi->ksi_signo >= NSIG)
return EINVAL;
if (pid != l->l_proc->p_pid) {
if (ksi->ksi_pid != l->l_proc->p_pid)
return EPERM;
if (ksi->ksi_uid != kauth_cred_geteuid(l->l_cred))
return EPERM;
switch (ksi->ksi_code) {
case SI_USER:
case SI_QUEUE:
break;
default:
return EPERM;
}
}
if (pid > 0) {
/* kill single process */
mutex_enter(proc_lock);
p = proc_find_raw(pid);
if (p == NULL || (p->p_stat != SACTIVE && p->p_stat != SSTOP)) {
mutex_exit(proc_lock);
/* IEEE Std 1003.1-2001: return success for zombies */
return p ? 0 : ESRCH;
}
mutex_enter(p->p_lock);
error = kauth_authorize_process(l->l_cred,
KAUTH_PROCESS_SIGNAL, p, KAUTH_ARG(ksi->ksi_signo),
NULL, NULL);
if (!error && ksi->ksi_signo) {
kpsignal2(p, ksi);
}
mutex_exit(p->p_lock);
mutex_exit(proc_lock);
return error;
}
switch (pid) {
case -1: /* broadcast signal */
return killpg1(l, ksi, 0, 1);
case 0: /* signal own process group */
return killpg1(l, ksi, 0, 0);
default: /* negative explicit process group */
return killpg1(l, ksi, -pid, 0);
}
/* NOTREACHED */
}
int
ipcperm(kauth_cred_t cred, struct ipc_perm *perm, int mode)
{
int error;
error = kauth_authorize_system(cred, KAUTH_SYSTEM_SYSVIPC,
KAUTH_REQ_SYSTEM_SYSVIPC_BYPASS, perm, KAUTH_ARG(mode), NULL);
if (error == 0)
return (0);
/* Adjust EPERM and EACCES errors until there's a better way to do this. */
if (mode != IPC_M)
error = EACCES;
return error;
}
static int
linux_do_tkill(struct lwp *l, int tgid, int tid, int signum)
{
struct proc *p;
struct lwp *t;
ksiginfo_t ksi;
int error;
if (signum < 0 || signum >= LINUX__NSIG)
return EINVAL;
signum = linux_to_native_signo[signum];
if (tgid == -1) {
tgid = tid;
}
KSI_INIT(&ksi);
ksi.ksi_signo = signum;
ksi.ksi_code = SI_LWP;
ksi.ksi_pid = l->l_proc->p_pid;
ksi.ksi_uid = kauth_cred_geteuid(l->l_cred);
ksi.ksi_lid = tid;
mutex_enter(proc_lock);
p = proc_find(tgid);
if (p == NULL) {
mutex_exit(proc_lock);
return ESRCH;
}
mutex_enter(p->p_lock);
error = kauth_authorize_process(l->l_cred,
KAUTH_PROCESS_SIGNAL, p, KAUTH_ARG(signum), NULL, NULL);
if ((t = lwp_find(p, ksi.ksi_lid)) == NULL)
error = ESRCH;
else if (signum != 0)
kpsignal2(p, &ksi);
mutex_exit(p->p_lock);
mutex_exit(proc_lock);
return error;
}
static int
linux_clone_nptl(struct lwp *l, const struct linux_sys_clone_args *uap, register_t *retval)
{
/* {
syscallarg(int) flags;
syscallarg(void *) stack;
syscallarg(void *) parent_tidptr;
syscallarg(void *) tls;
syscallarg(void *) child_tidptr;
} */
struct proc *p;
struct lwp *l2;
struct linux_emuldata *led;
void *parent_tidptr, *tls, *child_tidptr;
struct schedstate_percpu *spc;
vaddr_t uaddr;
lwpid_t lid;
int flags, tnprocs, error;
p = l->l_proc;
flags = SCARG(uap, flags);
parent_tidptr = SCARG(uap, parent_tidptr);
tls = SCARG(uap, tls);
child_tidptr = SCARG(uap, child_tidptr);
tnprocs = atomic_inc_uint_nv(&nprocs);
if (__predict_false(tnprocs >= maxproc) ||
kauth_authorize_process(l->l_cred, KAUTH_PROCESS_FORK, p,
KAUTH_ARG(tnprocs), NULL, NULL) != 0) {
atomic_dec_uint(&nprocs);
return EAGAIN;
}
uaddr = uvm_uarea_alloc();
if (__predict_false(uaddr == 0)) {
atomic_dec_uint(&nprocs);
return ENOMEM;
}
error = lwp_create(l, p, uaddr, LWP_DETACHED | LWP_PIDLID,
SCARG(uap, stack), 0, child_return, NULL, &l2, l->l_class);
if (__predict_false(error)) {
DPRINTF(("%s: lwp_create error=%d\n", __func__, error));
atomic_dec_uint(&nprocs);
uvm_uarea_free(uaddr);
return error;
}
lid = l2->l_lid;
/* LINUX_CLONE_CHILD_CLEARTID: clear TID in child's memory on exit() */
if (flags & LINUX_CLONE_CHILD_CLEARTID) {
led = l2->l_emuldata;
led->led_clear_tid = child_tidptr;
}
/* LINUX_CLONE_PARENT_SETTID: store child's TID in parent's memory */
if (flags & LINUX_CLONE_PARENT_SETTID) {
if ((error = copyout(&lid, parent_tidptr, sizeof(lid))) != 0)
printf("%s: LINUX_CLONE_PARENT_SETTID "
"failed (parent_tidptr = %p tid = %d error=%d)\n",
__func__, parent_tidptr, lid, error);
}
/* LINUX_CLONE_CHILD_SETTID: store child's TID in child's memory */
if (flags & LINUX_CLONE_CHILD_SETTID) {
if ((error = copyout(&lid, child_tidptr, sizeof(lid))) != 0)
printf("%s: LINUX_CLONE_CHILD_SETTID "
"failed (child_tidptr = %p, tid = %d error=%d)\n",
__func__, child_tidptr, lid, error);
}
if (flags & LINUX_CLONE_SETTLS) {
error = LINUX_LWP_SETPRIVATE(l2, tls);
if (error) {
DPRINTF(("%s: LINUX_LWP_SETPRIVATE %d\n", __func__,
error));
lwp_exit(l2);
return error;
}
}
/*
* Set the new LWP running, unless the process is stopping,
* then the LWP is created stopped.
*/
mutex_enter(p->p_lock);
lwp_lock(l2);
spc = &l2->l_cpu->ci_schedstate;
if ((l->l_flag & (LW_WREBOOT | LW_WSUSPEND | LW_WEXIT)) == 0) {
if (p->p_stat == SSTOP || (p->p_sflag & PS_STOPPING) != 0) {
KASSERT(l2->l_wchan == NULL);
l2->l_stat = LSSTOP;
p->p_nrlwps--;
lwp_unlock_to(l2, spc->spc_lwplock);
} else {
KASSERT(lwp_locked(l2, spc->spc_mutex));
l2->l_stat = LSRUN;
sched_enqueue(l2, false);
lwp_unlock(l2);
}
} else {
l2->l_stat = LSSUSPENDED;
p->p_nrlwps--;
lwp_unlock_to(l2, spc->spc_lwplock);
}
mutex_exit(p->p_lock);
retval[0] = lid;
retval[1] = 0;
return 0;
}
int
adosfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
{
struct lwp *l = curlwp;
struct vnode *devvp;
struct adosfs_args *args = data;
struct adosfsmount *amp;
int error;
mode_t accessmode;
if (*data_len < sizeof *args)
return EINVAL;
if (mp->mnt_flag & MNT_GETARGS) {
amp = VFSTOADOSFS(mp);
if (amp == NULL)
return EIO;
args->uid = amp->uid;
args->gid = amp->gid;
args->mask = amp->mask;
args->fspec = NULL;
*data_len = sizeof *args;
return 0;
}
if ((mp->mnt_flag & MNT_RDONLY) == 0)
return (EROFS);
if ((mp->mnt_flag & MNT_UPDATE) && args->fspec == NULL)
return EOPNOTSUPP;
/*
* Not an update, or updating the name: look up the name
* and verify that it refers to a sensible block device.
*/
error = namei_simple_user(args->fspec,
NSM_FOLLOW_NOEMULROOT, &devvp);
if (error != 0)
return (error);
if (devvp->v_type != VBLK) {
vrele(devvp);
return (ENOTBLK);
}
if (bdevsw_lookup(devvp->v_rdev) == NULL) {
vrele(devvp);
return (ENXIO);
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*/
accessmode = VREAD;
if ((mp->mnt_flag & MNT_RDONLY) == 0)
accessmode |= VWRITE;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(accessmode));
VOP_UNLOCK(devvp);
if (error) {
vrele(devvp);
return (error);
}
/* MNT_UPDATE? */
if ((error = adosfs_mountfs(devvp, mp, l)) != 0) {
vrele(devvp);
return (error);
}
amp = VFSTOADOSFS(mp);
amp->uid = args->uid;
amp->gid = args->gid;
amp->mask = args->mask;
return set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
mp->mnt_op->vfs_name, mp, l);
}
/*
* mount syscall vfsop.
*
* Returns 0 on success.
*/
static int
efs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
{
struct lwp *l = curlwp;
struct efs_args *args = data;
struct pathbuf *pb;
struct nameidata devnd;
struct efs_mount *emp;
struct vnode *devvp;
int err, mode;
if (args == NULL)
return EINVAL;
if (*data_len < sizeof *args)
return EINVAL;
if (mp->mnt_flag & MNT_GETARGS) {
if ((emp = VFSTOEFS(mp)) == NULL)
return (EIO);
args->fspec = NULL;
args->version = EFS_MNT_VERSION;
*data_len = sizeof *args;
return 0;
}
if (mp->mnt_flag & MNT_UPDATE)
return (EOPNOTSUPP); /* XXX read-only */
/* look up our device's vnode. it is returned locked */
err = pathbuf_copyin(args->fspec, &pb);
if (err) {
return err;
}
NDINIT(&devnd, LOOKUP, FOLLOW | LOCKLEAF, pb);
if ((err = namei(&devnd))) {
pathbuf_destroy(pb);
return (err);
}
devvp = devnd.ni_vp;
pathbuf_destroy(pb);
if (devvp->v_type != VBLK) {
vput(devvp);
return (ENOTBLK);
}
/* XXX - rdonly */
mode = FREAD;
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*/
err = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(VREAD));
if (err) {
vput(devvp);
return (err);
}
if ((err = VOP_OPEN(devvp, mode, l->l_cred))) {
vput(devvp);
return (err);
}
err = efs_mount_common(mp, path, devvp, args);
if (err) {
VOP_CLOSE(devvp, mode, l->l_cred);
vput(devvp);
return (err);
}
VOP_UNLOCK(devvp);
return (0);
}
static int
quota_handle_cmd_clear(struct mount *mp, struct lwp *l,
prop_dictionary_t cmddict, int type, prop_array_t datas)
{
prop_array_t replies;
prop_object_iterator_t iter;
prop_dictionary_t data;
uint32_t id;
struct ufsmount *ump = VFSTOUFS(mp);
int error, defaultq = 0;
const char *idstr;
if ((ump->um_flags & UFS_QUOTA2) == 0)
return EOPNOTSUPP;
replies = prop_array_create();
if (replies == NULL)
return ENOMEM;
iter = prop_array_iterator(datas);
if (iter == NULL) {
prop_object_release(replies);
return ENOMEM;
}
while ((data = prop_object_iterator_next(iter)) != NULL) {
if (!prop_dictionary_get_uint32(data, "id", &id)) {
if (!prop_dictionary_get_cstring_nocopy(data, "id",
&idstr))
continue;
if (strcmp(idstr, "default"))
continue;
id = 0;
defaultq = 1;
} else {
defaultq = 0;
}
error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_FS_QUOTA,
KAUTH_REQ_SYSTEM_FS_QUOTA_MANAGE, mp, KAUTH_ARG(id), NULL);
if (error != 0)
goto err;
#ifdef QUOTA2
if (ump->um_flags & UFS_QUOTA2) {
error = quota2_handle_cmd_clear(ump, type, id, defaultq,
data);
} else
#endif
panic("quota_handle_cmd_get: no support ?");
if (error && error != ENOENT)
goto err;
}
prop_object_iterator_release(iter);
if (!prop_dictionary_set_and_rel(cmddict, "data", replies)) {
error = ENOMEM;
} else {
error = 0;
}
return error;
err:
prop_object_iterator_release(iter);
prop_object_release(replies);
return error;
}
int
hfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
{
struct lwp *l = curlwp;
struct hfs_args *args = data;
struct vnode *devvp;
struct hfsmount *hmp;
int error = 0;
int update;
mode_t accessmode;
if (args == NULL)
return EINVAL;
if (*data_len < sizeof *args)
return EINVAL;
#ifdef HFS_DEBUG
printf("vfsop = hfs_mount()\n");
#endif /* HFS_DEBUG */
if (mp->mnt_flag & MNT_GETARGS) {
hmp = VFSTOHFS(mp);
if (hmp == NULL)
return EIO;
args->fspec = NULL;
*data_len = sizeof *args;
return 0;
}
if (data == NULL)
return EINVAL;
/* FIXME: For development ONLY - disallow remounting for now */
#if 0
update = mp->mnt_flag & MNT_UPDATE;
#else
update = 0;
#endif
/* Check arguments */
if (args->fspec != NULL) {
/*
* Look up the name and verify that it's sane.
*/
error = namei_simple_user(args->fspec,
NSM_FOLLOW_NOEMULROOT, &devvp);
if (error != 0)
return error;
if (!update) {
/*
* Be sure this is a valid block device
*/
if (devvp->v_type != VBLK)
error = ENOTBLK;
else if (bdevsw_lookup(devvp->v_rdev) == NULL)
error = ENXIO;
} else {
/*
* Be sure we're still naming the same device
* used for our initial mount
*/
hmp = VFSTOHFS(mp);
if (devvp != hmp->hm_devvp)
error = EINVAL;
}
} else {
if (update) {
/* Use the extant mount */
hmp = VFSTOHFS(mp);
devvp = hmp->hm_devvp;
vref(devvp);
} else {
/* New mounts must have a filename for the device */
return EINVAL;
}
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*
* Permission to update a mount is checked higher, so here we presume
* updating the mount is okay (for example, as far as securelevel goes)
* which leaves us with the normal check.
*/
if (error == 0) {
accessmode = VREAD;
if (update ?
(mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
(mp->mnt_flag & MNT_RDONLY) == 0)
accessmode |= VWRITE;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
KAUTH_ARG(accessmode));
VOP_UNLOCK(devvp);
}
if (error != 0)
goto error;
if (update) {
printf("HFS: live remounting not yet supported!\n");
error = EINVAL;
goto error;
}
if ((error = hfs_mountfs(devvp, mp, l, args->fspec)) != 0)
goto error;
error = set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
mp->mnt_op->vfs_name, mp, l);
#ifdef HFS_DEBUG
if(!update) {
char* volname;
hmp = VFSTOHFS(mp);
volname = malloc(hmp->hm_vol.name.length + 1, M_TEMP, M_WAITOK);
if (volname == NULL)
printf("could not allocate volname; ignored\n");
else {
if (hfs_unicode_to_ascii(hmp->hm_vol.name.unicode,
hmp->hm_vol.name.length, volname) == NULL)
printf("could not convert volume name to ascii; ignored\n");
else
printf("mounted volume \"%s\"\n", volname);
free(volname, M_TEMP);
}
}
#endif /* HFS_DEBUG */
return error;
error:
vrele(devvp);
return error;
}
int
v7fs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
{
struct lwp *l = curlwp;
struct v7fs_args *args = data;
struct v7fs_mount *v7fsmount = (void *)mp->mnt_data;
struct vnode *devvp = NULL;
int error = 0;
bool update = mp->mnt_flag & MNT_UPDATE;
DPRINTF("mnt_flag=%x %s\n", mp->mnt_flag, update ? "update" : "");
if (*data_len < sizeof(*args))
return EINVAL;
if (mp->mnt_flag & MNT_GETARGS) {
if (!v7fsmount)
return EIO;
args->fspec = NULL;
args->endian = v7fsmount->core->endian;
*data_len = sizeof(*args);
return 0;
}
DPRINTF("args->fspec=%s endian=%d\n", args->fspec, args->endian);
if (args->fspec == NULL) {
/* nothing to do. */
return EINVAL;
}
if (args->fspec != NULL) {
/* Look up the name and verify that it's sane. */
error = namei_simple_user(args->fspec,
NSM_FOLLOW_NOEMULROOT, &devvp);
if (error != 0)
return (error);
DPRINTF("mount device=%lx\n", (long)devvp->v_rdev);
if (!update) {
/*
* Be sure this is a valid block device
*/
if (devvp->v_type != VBLK)
error = ENOTBLK;
else if (bdevsw_lookup(devvp->v_rdev) == NULL)
error = ENXIO;
} else {
KDASSERT(v7fsmount);
/*
* Be sure we're still naming the same device
* used for our initial mount
*/
if (devvp != v7fsmount->devvp) {
DPRINTF("devvp %p != %p rootvp=%p\n", devvp,
v7fsmount->devvp, rootvp);
if (rootvp == v7fsmount->devvp) {
vrele(devvp);
devvp = rootvp;
vref(devvp);
} else {
error = EINVAL;
}
}
}
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*
* Permission to update a mount is checked higher, so here we presume
* updating the mount is okay (for example, as far as securelevel goes)
* which leaves us with the normal check.
*/
if (error == 0) {
int accessmode = VREAD;
if (update ?
(mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
(mp->mnt_flag & MNT_RDONLY) == 0)
accessmode |= VWRITE;
error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
KAUTH_ARG(accessmode));
}
if (error) {
vrele(devvp);
return error;
}
if (!update) {
if ((error = v7fs_openfs(devvp, mp, l))) {
vrele(devvp);
return error;
}
if ((error = v7fs_mountfs(devvp, mp, args->endian))) {
v7fs_closefs(devvp, mp);
VOP_UNLOCK(devvp);
vrele(devvp);
return error;
}
VOP_UNLOCK(devvp);
} else if (mp->mnt_flag & MNT_RDONLY) {
/* XXX: r/w -> read only */
}
return set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
mp->mnt_op->vfs_name, mp, l);
}
/*
* VFS Operations.
*
* mount system call
*/
int
cd9660_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
{
struct lwp *l = curlwp;
struct vnode *devvp;
struct iso_args *args = data;
int error;
struct iso_mnt *imp = VFSTOISOFS(mp);
if (*data_len < sizeof *args)
return EINVAL;
if (mp->mnt_flag & MNT_GETARGS) {
if (imp == NULL)
return EIO;
args->fspec = NULL;
args->flags = imp->im_flags;
*data_len = sizeof (*args);
return 0;
}
if ((mp->mnt_flag & MNT_RDONLY) == 0)
return (EROFS);
if ((mp->mnt_flag & MNT_UPDATE) && args->fspec == NULL)
return EINVAL;
/*
* Not an update, or updating the name: look up the name
* and verify that it refers to a sensible block device.
*/
error = namei_simple_user(args->fspec,
NSM_FOLLOW_NOEMULROOT, &devvp);
if (error != 0)
return (error);
if (devvp->v_type != VBLK) {
vrele(devvp);
return ENOTBLK;
}
if (bdevsw_lookup(devvp->v_rdev) == NULL) {
vrele(devvp);
return ENXIO;
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*/
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(VREAD));
VOP_UNLOCK(devvp);
if (error) {
vrele(devvp);
return (error);
}
if ((mp->mnt_flag & MNT_UPDATE) == 0) {
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = VOP_OPEN(devvp, FREAD, FSCRED);
VOP_UNLOCK(devvp);
if (error)
goto fail;
error = iso_mountfs(devvp, mp, l, args);
if (error) {
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
(void)VOP_CLOSE(devvp, FREAD, NOCRED);
VOP_UNLOCK(devvp);
goto fail;
}
} else {
vrele(devvp);
if (devvp != imp->im_devvp &&
devvp->v_rdev != imp->im_devvp->v_rdev)
return (EINVAL); /* needs translation */
}
return set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
mp->mnt_op->vfs_name, mp, l);
fail:
vrele(devvp);
return (error);
}
/*
* Process debugging system call.
*/
int
sys_ptrace(struct lwp *l, const struct sys_ptrace_args *uap, register_t *retval)
{
/* {
syscallarg(int) req;
syscallarg(pid_t) pid;
syscallarg(void *) addr;
syscallarg(int) data;
} */
struct proc *p = l->l_proc;
struct lwp *lt;
struct proc *t; /* target process */
struct uio uio;
struct iovec iov;
struct ptrace_io_desc piod;
struct ptrace_lwpinfo pl;
struct vmspace *vm;
int error, write, tmp, req, pheld;
int signo;
ksiginfo_t ksi;
#ifdef COREDUMP
char *path;
#endif
error = 0;
req = SCARG(uap, req);
/*
* If attaching or detaching, we need to get a write hold on the
* proclist lock so that we can re-parent the target process.
*/
mutex_enter(proc_lock);
/* "A foolish consistency..." XXX */
if (req == PT_TRACE_ME) {
t = p;
mutex_enter(t->p_lock);
} else {
/* Find the process we're supposed to be operating on. */
if ((t = p_find(SCARG(uap, pid), PFIND_LOCKED)) == NULL) {
mutex_exit(proc_lock);
return (ESRCH);
}
/* XXX-elad */
mutex_enter(t->p_lock);
error = kauth_authorize_process(l->l_cred, KAUTH_PROCESS_CANSEE,
t, KAUTH_ARG(KAUTH_REQ_PROCESS_CANSEE_ENTRY), NULL, NULL);
if (error) {
mutex_exit(proc_lock);
mutex_exit(t->p_lock);
return (ESRCH);
}
}
/*
* Grab a reference on the process to prevent it from execing or
* exiting.
*/
if (!rw_tryenter(&t->p_reflock, RW_READER)) {
mutex_exit(proc_lock);
mutex_exit(t->p_lock);
return EBUSY;
}
/* Make sure we can operate on it. */
switch (req) {
case PT_TRACE_ME:
/* Saying that you're being traced is always legal. */
break;
case PT_ATTACH:
/*
* You can't attach to a process if:
* (1) it's the process that's doing the attaching,
*/
if (t->p_pid == p->p_pid) {
error = EINVAL;
break;
}
/*
* (2) it's a system process
*/
if (t->p_flag & PK_SYSTEM) {
error = EPERM;
break;
}
/*
* (3) it's already being traced, or
*/
if (ISSET(t->p_slflag, PSL_TRACED)) {
error = EBUSY;
break;
}
/*
* (4) the tracer is chrooted, and its root directory is
* not at or above the root directory of the tracee
*/
mutex_exit(t->p_lock); /* XXXSMP */
tmp = proc_isunder(t, l);
mutex_enter(t->p_lock); /* XXXSMP */
if (!tmp) {
error = EPERM;
break;
}
break;
case PT_READ_I:
case PT_READ_D:
case PT_WRITE_I:
case PT_WRITE_D:
case PT_IO:
#ifdef PT_GETREGS
case PT_GETREGS:
#endif
#ifdef PT_SETREGS
case PT_SETREGS:
#endif
#ifdef PT_GETFPREGS
case PT_GETFPREGS:
#endif
#ifdef PT_SETFPREGS
case PT_SETFPREGS:
#endif
#ifdef __HAVE_PTRACE_MACHDEP
PTRACE_MACHDEP_REQUEST_CASES
#endif
/*
* You can't read/write the memory or registers of a process
* if the tracer is chrooted, and its root directory is not at
* or above the root directory of the tracee.
*/
mutex_exit(t->p_lock); /* XXXSMP */
tmp = proc_isunder(t, l);
mutex_enter(t->p_lock); /* XXXSMP */
if (!tmp) {
error = EPERM;
break;
}
/*FALLTHROUGH*/
case PT_CONTINUE:
case PT_KILL:
case PT_DETACH:
case PT_LWPINFO:
case PT_SYSCALL:
#ifdef COREDUMP
case PT_DUMPCORE:
#endif
#ifdef PT_STEP
case PT_STEP:
#endif
/*
* You can't do what you want to the process if:
* (1) It's not being traced at all,
*/
if (!ISSET(t->p_slflag, PSL_TRACED)) {
error = EPERM;
break;
}
/*
* (2) it's being traced by procfs (which has
* different signal delivery semantics),
*/
if (ISSET(t->p_slflag, PSL_FSTRACE)) {
uprintf("file system traced\n");
error = EBUSY;
break;
}
/*
* (3) it's not being traced by _you_, or
*/
if (t->p_pptr != p) {
uprintf("parent %d != %d\n", t->p_pptr->p_pid, p->p_pid);
error = EBUSY;
break;
}
/*
* (4) it's not currently stopped.
*/
if (t->p_stat != SSTOP || !t->p_waited /* XXXSMP */) {
uprintf("stat %d flag %d\n", t->p_stat,
!t->p_waited);
error = EBUSY;
break;
}
break;
default: /* It was not a legal request. */
error = EINVAL;
break;
}
if (error == 0)
error = kauth_authorize_process(l->l_cred,
KAUTH_PROCESS_PTRACE, t, KAUTH_ARG(req),
NULL, NULL);
if (error != 0) {
mutex_exit(proc_lock);
mutex_exit(t->p_lock);
rw_exit(&t->p_reflock);
return error;
}
/* Do single-step fixup if needed. */
FIX_SSTEP(t);
/*
* XXX NJWLWP
*
* The entire ptrace interface needs work to be useful to a
* process with multiple LWPs. For the moment, we'll kluge
* this; memory access will be fine, but register access will
* be weird.
*/
lt = LIST_FIRST(&t->p_lwps);
KASSERT(lt != NULL);
lwp_addref(lt);
/*
* Which locks do we need held? XXX Ugly.
*/
switch (req) {
#ifdef PT_STEP
case PT_STEP:
#endif
case PT_CONTINUE:
case PT_DETACH:
case PT_KILL:
case PT_SYSCALL:
case PT_ATTACH:
case PT_TRACE_ME:
pheld = 1;
break;
default:
mutex_exit(proc_lock);
mutex_exit(t->p_lock);
pheld = 0;
break;
}
/* Now do the operation. */
write = 0;
*retval = 0;
tmp = 0;
switch (req) {
case PT_TRACE_ME:
/* Just set the trace flag. */
SET(t->p_slflag, PSL_TRACED);
t->p_opptr = t->p_pptr;
break;
case PT_WRITE_I: /* XXX no separate I and D spaces */
case PT_WRITE_D:
#if defined(__HAVE_RAS)
/*
* Can't write to a RAS
*/
if (ras_lookup(t, SCARG(uap, addr)) != (void *)-1) {
error = EACCES;
break;
}
#endif
write = 1;
tmp = SCARG(uap, data);
/* FALLTHROUGH */
case PT_READ_I: /* XXX no separate I and D spaces */
case PT_READ_D:
/* write = 0 done above. */
iov.iov_base = (void *)&tmp;
iov.iov_len = sizeof(tmp);
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = (off_t)(unsigned long)SCARG(uap, addr);
uio.uio_resid = sizeof(tmp);
uio.uio_rw = write ? UIO_WRITE : UIO_READ;
UIO_SETUP_SYSSPACE(&uio);
error = process_domem(l, lt, &uio);
if (!write)
*retval = tmp;
break;
case PT_IO:
error = copyin(SCARG(uap, addr), &piod, sizeof(piod));
if (error)
break;
switch (piod.piod_op) {
case PIOD_READ_D:
case PIOD_READ_I:
uio.uio_rw = UIO_READ;
break;
case PIOD_WRITE_D:
case PIOD_WRITE_I:
/*
* Can't write to a RAS
*/
if (ras_lookup(t, SCARG(uap, addr)) != (void *)-1) {
return (EACCES);
}
uio.uio_rw = UIO_WRITE;
break;
default:
error = EINVAL;
break;
}
if (error)
break;
error = proc_vmspace_getref(l->l_proc, &vm);
if (error)
break;
iov.iov_base = piod.piod_addr;
iov.iov_len = piod.piod_len;
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = (off_t)(unsigned long)piod.piod_offs;
uio.uio_resid = piod.piod_len;
uio.uio_vmspace = vm;
error = process_domem(l, lt, &uio);
piod.piod_len -= uio.uio_resid;
(void) copyout(&piod, SCARG(uap, addr), sizeof(piod));
uvmspace_free(vm);
break;
#ifdef COREDUMP
case PT_DUMPCORE:
if ((path = SCARG(uap, addr)) != NULL) {
char *dst;
int len = SCARG(uap, data);
if (len < 0 || len >= MAXPATHLEN) {
error = EINVAL;
break;
}
dst = malloc(len + 1, M_TEMP, M_WAITOK);
if ((error = copyin(path, dst, len)) != 0) {
free(dst, M_TEMP);
break;
}
path = dst;
path[len] = '\0';
}
error = coredump(lt, path);
if (path)
free(path, M_TEMP);
break;
#endif
#ifdef PT_STEP
case PT_STEP:
/*
* From the 4.4BSD PRM:
* "Execution continues as in request PT_CONTINUE; however
* as soon as possible after execution of at least one
* instruction, execution stops again. [ ... ]"
*/
#endif
case PT_CONTINUE:
case PT_SYSCALL:
case PT_DETACH:
if (req == PT_SYSCALL) {
if (!ISSET(t->p_slflag, PSL_SYSCALL)) {
SET(t->p_slflag, PSL_SYSCALL);
#ifdef __HAVE_SYSCALL_INTERN
(*t->p_emul->e_syscall_intern)(t);
#endif
}
} else {
if (ISSET(t->p_slflag, PSL_SYSCALL)) {
CLR(t->p_slflag, PSL_SYSCALL);
#ifdef __HAVE_SYSCALL_INTERN
(*t->p_emul->e_syscall_intern)(t);
#endif
}
}
p->p_trace_enabled = trace_is_enabled(p);
/*
* From the 4.4BSD PRM:
* "The data argument is taken as a signal number and the
* child's execution continues at location addr as if it
* incurred that signal. Normally the signal number will
* be either 0 to indicate that the signal that caused the
* stop should be ignored, or that value fetched out of
* the process's image indicating which signal caused
* the stop. If addr is (int *)1 then execution continues
* from where it stopped."
*/
/* Check that the data is a valid signal number or zero. */
if (SCARG(uap, data) < 0 || SCARG(uap, data) >= NSIG) {
error = EINVAL;
break;
}
uvm_lwp_hold(lt);
/* If the address parameter is not (int *)1, set the pc. */
if ((int *)SCARG(uap, addr) != (int *)1)
if ((error = process_set_pc(lt, SCARG(uap, addr))) != 0) {
uvm_lwp_rele(lt);
break;
}
#ifdef PT_STEP
/*
* Arrange for a single-step, if that's requested and possible.
*/
error = process_sstep(lt, req == PT_STEP);
if (error) {
uvm_lwp_rele(lt);
break;
}
#endif
uvm_lwp_rele(lt);
if (req == PT_DETACH) {
CLR(t->p_slflag, PSL_TRACED|PSL_FSTRACE|PSL_SYSCALL);
/* give process back to original parent or init */
if (t->p_opptr != t->p_pptr) {
struct proc *pp = t->p_opptr;
proc_reparent(t, pp ? pp : initproc);
}
/* not being traced any more */
t->p_opptr = NULL;
}
signo = SCARG(uap, data);
sendsig:
/* Finally, deliver the requested signal (or none). */
if (t->p_stat == SSTOP) {
/*
* Unstop the process. If it needs to take a
* signal, make all efforts to ensure that at
* an LWP runs to see it.
*/
t->p_xstat = signo;
proc_unstop(t);
} else if (signo != 0) {
KSI_INIT_EMPTY(&ksi);
ksi.ksi_signo = signo;
kpsignal2(t, &ksi);
}
break;
case PT_KILL:
/* just send the process a KILL signal. */
signo = SIGKILL;
goto sendsig; /* in PT_CONTINUE, above. */
case PT_ATTACH:
/*
* Go ahead and set the trace flag.
* Save the old parent (it's reset in
* _DETACH, and also in kern_exit.c:wait4()
* Reparent the process so that the tracing
* proc gets to see all the action.
* Stop the target.
*/
t->p_opptr = t->p_pptr;
if (t->p_pptr != p) {
struct proc *parent = t->p_pptr;
if (parent->p_lock < t->p_lock) {
if (!mutex_tryenter(parent->p_lock)) {
mutex_exit(t->p_lock);
mutex_enter(parent->p_lock);
}
} else if (parent->p_lock > t->p_lock) {
mutex_enter(parent->p_lock);
}
parent->p_slflag |= PSL_CHTRACED;
proc_reparent(t, p);
if (parent->p_lock != t->p_lock)
mutex_exit(parent->p_lock);
}
SET(t->p_slflag, PSL_TRACED);
signo = SIGSTOP;
goto sendsig;
case PT_LWPINFO:
if (SCARG(uap, data) != sizeof(pl)) {
error = EINVAL;
break;
}
error = copyin(SCARG(uap, addr), &pl, sizeof(pl));
if (error)
break;
tmp = pl.pl_lwpid;
lwp_delref(lt);
mutex_enter(t->p_lock);
if (tmp == 0)
lt = LIST_FIRST(&t->p_lwps);
else {
lt = lwp_find(t, tmp);
if (lt == NULL) {
mutex_exit(t->p_lock);
error = ESRCH;
break;
}
lt = LIST_NEXT(lt, l_sibling);
}
while (lt != NULL && lt->l_stat == LSZOMB)
lt = LIST_NEXT(lt, l_sibling);
pl.pl_lwpid = 0;
pl.pl_event = 0;
if (lt) {
lwp_addref(lt);
pl.pl_lwpid = lt->l_lid;
if (lt->l_lid == t->p_sigctx.ps_lwp)
pl.pl_event = PL_EVENT_SIGNAL;
}
mutex_exit(t->p_lock);
error = copyout(&pl, SCARG(uap, addr), sizeof(pl));
break;
#ifdef PT_SETREGS
case PT_SETREGS:
write = 1;
#endif
#ifdef PT_GETREGS
case PT_GETREGS:
/* write = 0 done above. */
#endif
#if defined(PT_SETREGS) || defined(PT_GETREGS)
tmp = SCARG(uap, data);
if (tmp != 0 && t->p_nlwps > 1) {
lwp_delref(lt);
mutex_enter(t->p_lock);
lt = lwp_find(t, tmp);
if (lt == NULL) {
mutex_exit(t->p_lock);
error = ESRCH;
break;
}
lwp_addref(lt);
mutex_exit(t->p_lock);
}
if (!process_validregs(lt))
error = EINVAL;
else {
error = proc_vmspace_getref(l->l_proc, &vm);
if (error)
break;
iov.iov_base = SCARG(uap, addr);
iov.iov_len = sizeof(struct reg);
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = 0;
uio.uio_resid = sizeof(struct reg);
uio.uio_rw = write ? UIO_WRITE : UIO_READ;
uio.uio_vmspace = vm;
error = process_doregs(l, lt, &uio);
uvmspace_free(vm);
}
break;
#endif
#ifdef PT_SETFPREGS
case PT_SETFPREGS:
write = 1;
#endif
#ifdef PT_GETFPREGS
case PT_GETFPREGS:
/* write = 0 done above. */
#endif
#if defined(PT_SETFPREGS) || defined(PT_GETFPREGS)
tmp = SCARG(uap, data);
if (tmp != 0 && t->p_nlwps > 1) {
lwp_delref(lt);
mutex_enter(t->p_lock);
lt = lwp_find(t, tmp);
if (lt == NULL) {
mutex_exit(t->p_lock);
error = ESRCH;
break;
}
lwp_addref(lt);
mutex_exit(t->p_lock);
}
if (!process_validfpregs(lt))
error = EINVAL;
else {
error = proc_vmspace_getref(l->l_proc, &vm);
if (error)
break;
iov.iov_base = SCARG(uap, addr);
iov.iov_len = sizeof(struct fpreg);
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = 0;
uio.uio_resid = sizeof(struct fpreg);
uio.uio_rw = write ? UIO_WRITE : UIO_READ;
uio.uio_vmspace = vm;
error = process_dofpregs(l, lt, &uio);
uvmspace_free(vm);
}
break;
#endif
#ifdef __HAVE_PTRACE_MACHDEP
PTRACE_MACHDEP_REQUEST_CASES
error = ptrace_machdep_dorequest(l, lt,
req, SCARG(uap, addr), SCARG(uap, data));
break;
#endif
}
if (pheld) {
mutex_exit(t->p_lock);
mutex_exit(proc_lock);
}
if (lt != NULL)
lwp_delref(lt);
rw_exit(&t->p_reflock);
return error;
}
int
msgctl1(struct lwp *l, int msqid, int cmd, struct msqid_ds *msqbuf)
{
kauth_cred_t cred = l->l_cred;
struct msqid_ds *msqptr;
kmsq_t *msq;
int error = 0, ix;
MSG_PRINTF(("call to msgctl1(%d, %d)\n", msqid, cmd));
ix = IPCID_TO_IX(msqid);
mutex_enter(&msgmutex);
if (ix < 0 || ix >= msginfo.msgmni) {
MSG_PRINTF(("msqid (%d) out of range (0<=msqid<%d)\n", ix,
msginfo.msgmni));
error = EINVAL;
goto unlock;
}
msq = &msqs[ix];
msqptr = &msq->msq_u;
if (msqptr->msg_qbytes == 0) {
MSG_PRINTF(("no such msqid\n"));
error = EINVAL;
goto unlock;
}
if (msqptr->msg_perm._seq != IPCID_TO_SEQ(msqid)) {
MSG_PRINTF(("wrong sequence number\n"));
error = EINVAL;
goto unlock;
}
switch (cmd) {
case IPC_RMID:
{
struct __msg *msghdr;
if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_M)) != 0)
break;
/* Free the message headers */
msghdr = msqptr->_msg_first;
while (msghdr != NULL) {
struct __msg *msghdr_tmp;
/* Free the segments of each message */
msqptr->_msg_cbytes -= msghdr->msg_ts;
msqptr->msg_qnum--;
msghdr_tmp = msghdr;
msghdr = msghdr->msg_next;
msg_freehdr(msghdr_tmp);
}
KASSERT(msqptr->_msg_cbytes == 0);
KASSERT(msqptr->msg_qnum == 0);
/* Mark it as free */
msqptr->msg_qbytes = 0;
cv_broadcast(&msq->msq_cv);
}
break;
case IPC_SET:
if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_M)))
break;
if (msqbuf->msg_qbytes > msqptr->msg_qbytes &&
kauth_authorize_system(cred, KAUTH_SYSTEM_SYSVIPC,
KAUTH_REQ_SYSTEM_SYSVIPC_MSGQ_OVERSIZE,
KAUTH_ARG(msqbuf->msg_qbytes),
KAUTH_ARG(msqptr->msg_qbytes), NULL) != 0) {
error = EPERM;
break;
}
if (msqbuf->msg_qbytes > msginfo.msgmnb) {
MSG_PRINTF(("can't increase msg_qbytes beyond %d "
"(truncating)\n", msginfo.msgmnb));
/* silently restrict qbytes to system limit */
msqbuf->msg_qbytes = msginfo.msgmnb;
}
if (msqbuf->msg_qbytes == 0) {
MSG_PRINTF(("can't reduce msg_qbytes to 0\n"));
error = EINVAL; /* XXX non-standard errno! */
break;
}
msqptr->msg_perm.uid = msqbuf->msg_perm.uid;
msqptr->msg_perm.gid = msqbuf->msg_perm.gid;
msqptr->msg_perm.mode = (msqptr->msg_perm.mode & ~0777) |
(msqbuf->msg_perm.mode & 0777);
msqptr->msg_qbytes = msqbuf->msg_qbytes;
msqptr->msg_ctime = time_second;
break;
case IPC_STAT:
if ((error = ipcperm(cred, &msqptr->msg_perm, IPC_R))) {
MSG_PRINTF(("requester doesn't have read access\n"));
break;
}
memcpy(msqbuf, msqptr, sizeof(struct msqid_ds));
break;
default:
MSG_PRINTF(("invalid command %d\n", cmd));
error = EINVAL;
break;
}
unlock:
mutex_exit(&msgmutex);
return (error);
}
/*
* mp - path - addr in user space of mount point (ie /usr or whatever)
* data - addr in user space of mount params including the name of the block
* special file to treat as a filesystem.
*/
int
msdosfs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
{
struct lwp *l = curlwp;
struct vnode *devvp; /* vnode for blk device to mount */
struct msdosfs_args *args = data; /* holds data from mount request */
/* msdosfs specific mount control block */
struct msdosfsmount *pmp = NULL;
int error, flags;
mode_t accessmode;
if (*data_len < sizeof *args)
return EINVAL;
if (mp->mnt_flag & MNT_GETARGS) {
pmp = VFSTOMSDOSFS(mp);
if (pmp == NULL)
return EIO;
args->fspec = NULL;
args->uid = pmp->pm_uid;
args->gid = pmp->pm_gid;
args->mask = pmp->pm_mask;
args->flags = pmp->pm_flags;
args->version = MSDOSFSMNT_VERSION;
args->dirmask = pmp->pm_dirmask;
args->gmtoff = pmp->pm_gmtoff;
*data_len = sizeof *args;
return 0;
}
/*
* If not versioned (i.e. using old mount_msdos(8)), fill in
* the additional structure items with suitable defaults.
*/
if ((args->flags & MSDOSFSMNT_VERSIONED) == 0) {
args->version = 1;
args->dirmask = args->mask;
}
/*
* Reset GMT offset for pre-v3 mount structure args.
*/
if (args->version < 3)
args->gmtoff = 0;
/*
* If updating, check whether changing from read-only to
* read/write; if there is no device name, that's all we do.
*/
if (mp->mnt_flag & MNT_UPDATE) {
pmp = VFSTOMSDOSFS(mp);
error = 0;
if (!(pmp->pm_flags & MSDOSFSMNT_RONLY) &&
(mp->mnt_flag & MNT_RDONLY)) {
flags = WRITECLOSE;
if (mp->mnt_flag & MNT_FORCE)
flags |= FORCECLOSE;
error = vflush(mp, NULLVP, flags);
}
if (!error && (mp->mnt_flag & MNT_RELOAD))
/* not yet implemented */
error = EOPNOTSUPP;
if (error) {
DPRINTF(("vflush %d\n", error));
return (error);
}
if ((pmp->pm_flags & MSDOSFSMNT_RONLY) &&
(mp->mnt_iflag & IMNT_WANTRDWR)) {
/*
* If upgrade to read-write by non-root, then verify
* that user has necessary permissions on the device.
*
* Permission to update a mount is checked higher, so
* here we presume updating the mount is okay (for
* example, as far as securelevel goes) which leaves us
* with the normal check.
*/
devvp = pmp->pm_devvp;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = kauth_authorize_system(l->l_cred,
KAUTH_SYSTEM_MOUNT, KAUTH_REQ_SYSTEM_MOUNT_DEVICE,
mp, devvp, KAUTH_ARG(VREAD | VWRITE));
VOP_UNLOCK(devvp);
DPRINTF(("KAUTH_REQ_SYSTEM_MOUNT_DEVICE %d\n", error));
if (error)
return (error);
pmp->pm_flags &= ~MSDOSFSMNT_RONLY;
}
if (args->fspec == NULL) {
DPRINTF(("missing fspec\n"));
return EINVAL;
}
}
/*
* Not an update, or updating the name: look up the name
* and verify that it refers to a sensible block device.
*/
error = namei_simple_user(args->fspec,
NSM_FOLLOW_NOEMULROOT, &devvp);
if (error != 0) {
DPRINTF(("namei %d\n", error));
return (error);
}
if (devvp->v_type != VBLK) {
DPRINTF(("not block\n"));
vrele(devvp);
return (ENOTBLK);
}
if (bdevsw_lookup(devvp->v_rdev) == NULL) {
DPRINTF(("no block switch\n"));
vrele(devvp);
return (ENXIO);
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*/
accessmode = VREAD;
if ((mp->mnt_flag & MNT_RDONLY) == 0)
accessmode |= VWRITE;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp, KAUTH_ARG(accessmode));
VOP_UNLOCK(devvp);
if (error) {
DPRINTF(("KAUTH_REQ_SYSTEM_MOUNT_DEVICE %d\n", error));
vrele(devvp);
return (error);
}
if ((mp->mnt_flag & MNT_UPDATE) == 0) {
int xflags;
if (mp->mnt_flag & MNT_RDONLY)
xflags = FREAD;
else
xflags = FREAD|FWRITE;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = VOP_OPEN(devvp, xflags, FSCRED);
VOP_UNLOCK(devvp);
if (error) {
DPRINTF(("VOP_OPEN %d\n", error));
goto fail;
}
error = msdosfs_mountfs(devvp, mp, l, args);
if (error) {
DPRINTF(("msdosfs_mountfs %d\n", error));
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
(void) VOP_CLOSE(devvp, xflags, NOCRED);
VOP_UNLOCK(devvp);
goto fail;
}
#ifdef MSDOSFS_DEBUG /* only needed for the printf below */
pmp = VFSTOMSDOSFS(mp);
#endif
} else {
vrele(devvp);
if (devvp != pmp->pm_devvp) {
DPRINTF(("devvp %p pmp %p\n",
devvp, pmp->pm_devvp));
return (EINVAL); /* needs translation */
}
}
if ((error = update_mp(mp, args)) != 0) {
msdosfs_unmount(mp, MNT_FORCE);
DPRINTF(("update_mp %d\n", error));
return error;
}
#ifdef MSDOSFS_DEBUG
printf("msdosfs_mount(): mp %p, pmp %p, inusemap %p\n", mp, pmp, pmp->pm_inusemap);
#endif
return set_statvfs_info(path, UIO_USERSPACE, args->fspec, UIO_USERSPACE,
mp->mnt_op->vfs_name, mp, l);
fail:
vrele(devvp);
return (error);
}
/*
* VFS Operations.
*
* mount system call
*/
int
ext2fs_mount(struct mount *mp, const char *path, void *data, size_t *data_len)
{
struct lwp *l = curlwp;
struct vnode *devvp;
struct ufs_args *args = data;
struct ufsmount *ump = NULL;
struct m_ext2fs *fs;
int error = 0, flags, update;
mode_t accessmode;
if (args == NULL)
return EINVAL;
if (*data_len < sizeof *args)
return EINVAL;
if (mp->mnt_flag & MNT_GETARGS) {
ump = VFSTOUFS(mp);
if (ump == NULL)
return EIO;
memset(args, 0, sizeof *args);
args->fspec = NULL;
*data_len = sizeof *args;
return 0;
}
update = mp->mnt_flag & MNT_UPDATE;
/* Check arguments */
if (args->fspec != NULL) {
/*
* Look up the name and verify that it's sane.
*/
error = namei_simple_user(args->fspec,
NSM_FOLLOW_NOEMULROOT, &devvp);
if (error != 0)
return error;
if (!update) {
/*
* Be sure this is a valid block device
*/
if (devvp->v_type != VBLK)
error = ENOTBLK;
else if (bdevsw_lookup(devvp->v_rdev) == NULL)
error = ENXIO;
} else {
/*
* Be sure we're still naming the same device
* used for our initial mount
*/
ump = VFSTOUFS(mp);
if (devvp != ump->um_devvp) {
if (devvp->v_rdev != ump->um_devvp->v_rdev)
error = EINVAL;
else {
vrele(devvp);
devvp = ump->um_devvp;
vref(devvp);
}
}
}
} else {
if (!update) {
/* New mounts must have a filename for the device */
return EINVAL;
} else {
ump = VFSTOUFS(mp);
devvp = ump->um_devvp;
vref(devvp);
}
}
/*
* If mount by non-root, then verify that user has necessary
* permissions on the device.
*
* Permission to update a mount is checked higher, so here we presume
* updating the mount is okay (for example, as far as securelevel goes)
* which leaves us with the normal check.
*/
if (error == 0) {
accessmode = VREAD;
if (update ?
(mp->mnt_iflag & IMNT_WANTRDWR) != 0 :
(mp->mnt_flag & MNT_RDONLY) == 0)
accessmode |= VWRITE;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = kauth_authorize_system(l->l_cred, KAUTH_SYSTEM_MOUNT,
KAUTH_REQ_SYSTEM_MOUNT_DEVICE, mp, devvp,
KAUTH_ARG(accessmode));
VOP_UNLOCK(devvp);
}
if (error) {
vrele(devvp);
return error;
}
if (!update) {
int xflags;
if (mp->mnt_flag & MNT_RDONLY)
xflags = FREAD;
else
xflags = FREAD|FWRITE;
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
error = VOP_OPEN(devvp, xflags, FSCRED);
VOP_UNLOCK(devvp);
if (error)
goto fail;
error = ext2fs_mountfs(devvp, mp);
if (error) {
vn_lock(devvp, LK_EXCLUSIVE | LK_RETRY);
(void)VOP_CLOSE(devvp, xflags, NOCRED);
VOP_UNLOCK(devvp);
goto fail;
}
ump = VFSTOUFS(mp);
fs = ump->um_e2fs;
} else {
/*
* Update the mount.
*/
/*
* The initial mount got a reference on this
* device, so drop the one obtained via
* namei(), above.
*/
vrele(devvp);
ump = VFSTOUFS(mp);
fs = ump->um_e2fs;
if (fs->e2fs_ronly == 0 && (mp->mnt_flag & MNT_RDONLY)) {
/*
* Changing from r/w to r/o
*/
flags = WRITECLOSE;
if (mp->mnt_flag & MNT_FORCE)
flags |= FORCECLOSE;
error = ext2fs_flushfiles(mp, flags);
if (error == 0 &&
ext2fs_cgupdate(ump, MNT_WAIT) == 0 &&
(fs->e2fs.e2fs_state & E2FS_ERRORS) == 0) {
fs->e2fs.e2fs_state = E2FS_ISCLEAN;
(void) ext2fs_sbupdate(ump, MNT_WAIT);
}
if (error)
return error;
fs->e2fs_ronly = 1;
}
if (mp->mnt_flag & MNT_RELOAD) {
error = ext2fs_reload(mp, l->l_cred, l);
if (error)
return error;
}
if (fs->e2fs_ronly && (mp->mnt_iflag & IMNT_WANTRDWR)) {
/*
* Changing from read-only to read/write
*/
fs->e2fs_ronly = 0;
if (fs->e2fs.e2fs_state == E2FS_ISCLEAN)
fs->e2fs.e2fs_state = 0;
else
fs->e2fs.e2fs_state = E2FS_ERRORS;
fs->e2fs_fmod = 1;
}
if (args->fspec == NULL)
return 0;
}
error = set_statvfs_info(path, UIO_USERSPACE, args->fspec,
UIO_USERSPACE, mp->mnt_op->vfs_name, mp, l);
if (error == 0)
ext2fs_sb_setmountinfo(fs, mp);
if (fs->e2fs_fmod != 0) { /* XXX */
fs->e2fs_fmod = 0;
if (fs->e2fs.e2fs_state == 0)
fs->e2fs.e2fs_wtime = time_second;
else
printf("%s: file system not clean; please fsck(8)\n",
mp->mnt_stat.f_mntfromname);
(void) ext2fs_cgupdate(ump, MNT_WAIT);
}
return error;
fail:
vrele(devvp);
return error;
}
/*
* General fork call. Note that another LWP in the process may call exec()
* or exit() while we are forking. It's safe to continue here, because
* neither operation will complete until all LWPs have exited the process.
*/
int
fork1(struct lwp *l1, int flags, int exitsig, void *stack, size_t stacksize,
void (*func)(void *), void *arg, register_t *retval,
struct proc **rnewprocp)
{
struct proc *p1, *p2, *parent;
struct plimit *p1_lim;
uid_t uid;
struct lwp *l2;
int count;
vaddr_t uaddr;
int tnprocs;
int tracefork;
int error = 0;
p1 = l1->l_proc;
uid = kauth_cred_getuid(l1->l_cred);
tnprocs = atomic_inc_uint_nv(&nprocs);
/*
* Although process entries are dynamically created, we still keep
* a global limit on the maximum number we will create.
*/
if (__predict_false(tnprocs >= maxproc))
error = -1;
else
error = kauth_authorize_process(l1->l_cred,
KAUTH_PROCESS_FORK, p1, KAUTH_ARG(tnprocs), NULL, NULL);
if (error) {
static struct timeval lasttfm;
atomic_dec_uint(&nprocs);
if (ratecheck(&lasttfm, &fork_tfmrate))
tablefull("proc", "increase kern.maxproc or NPROC");
if (forkfsleep)
kpause("forkmx", false, forkfsleep, NULL);
return EAGAIN;
}
/*
* Enforce limits.
*/
count = chgproccnt(uid, 1);
if (__predict_false(count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur)) {
if (kauth_authorize_process(l1->l_cred, KAUTH_PROCESS_RLIMIT,
p1, KAUTH_ARG(KAUTH_REQ_PROCESS_RLIMIT_BYPASS),
&p1->p_rlimit[RLIMIT_NPROC], KAUTH_ARG(RLIMIT_NPROC)) != 0) {
(void)chgproccnt(uid, -1);
atomic_dec_uint(&nprocs);
if (forkfsleep)
kpause("forkulim", false, forkfsleep, NULL);
return EAGAIN;
}
}
/*
* Allocate virtual address space for the U-area now, while it
* is still easy to abort the fork operation if we're out of
* kernel virtual address space.
*/
uaddr = uvm_uarea_alloc();
if (__predict_false(uaddr == 0)) {
(void)chgproccnt(uid, -1);
atomic_dec_uint(&nprocs);
return ENOMEM;
}
/*
* We are now committed to the fork. From here on, we may
* block on resources, but resource allocation may NOT fail.
*/
/* Allocate new proc. */
p2 = proc_alloc();
/*
* Make a proc table entry for the new process.
* Start by zeroing the section of proc that is zero-initialized,
* then copy the section that is copied directly from the parent.
*/
memset(&p2->p_startzero, 0,
(unsigned) ((char *)&p2->p_endzero - (char *)&p2->p_startzero));
memcpy(&p2->p_startcopy, &p1->p_startcopy,
(unsigned) ((char *)&p2->p_endcopy - (char *)&p2->p_startcopy));
TAILQ_INIT(&p2->p_sigpend.sp_info);
LIST_INIT(&p2->p_lwps);
LIST_INIT(&p2->p_sigwaiters);
/*
* Duplicate sub-structures as needed.
* Increase reference counts on shared objects.
* Inherit flags we want to keep. The flags related to SIGCHLD
* handling are important in order to keep a consistent behaviour
* for the child after the fork. If we are a 32-bit process, the
* child will be too.
*/
p2->p_flag =
p1->p_flag & (PK_SUGID | PK_NOCLDWAIT | PK_CLDSIGIGN | PK_32);
p2->p_emul = p1->p_emul;
p2->p_execsw = p1->p_execsw;
if (flags & FORK_SYSTEM) {
/*
* Mark it as a system process. Set P_NOCLDWAIT so that
* children are reparented to init(8) when they exit.
* init(8) can easily wait them out for us.
*/
p2->p_flag |= (PK_SYSTEM | PK_NOCLDWAIT);
}
mutex_init(&p2->p_stmutex, MUTEX_DEFAULT, IPL_HIGH);
mutex_init(&p2->p_auxlock, MUTEX_DEFAULT, IPL_NONE);
rw_init(&p2->p_reflock);
cv_init(&p2->p_waitcv, "wait");
cv_init(&p2->p_lwpcv, "lwpwait");
/*
* Share a lock between the processes if they are to share signal
* state: we must synchronize access to it.
*/
if (flags & FORK_SHARESIGS) {
p2->p_lock = p1->p_lock;
mutex_obj_hold(p1->p_lock);
} else
p2->p_lock = mutex_obj_alloc(MUTEX_DEFAULT, IPL_NONE);
kauth_proc_fork(p1, p2);
p2->p_raslist = NULL;
#if defined(__HAVE_RAS)
ras_fork(p1, p2);
#endif
/* bump references to the text vnode (for procfs) */
p2->p_textvp = p1->p_textvp;
if (p2->p_textvp)
vref(p2->p_textvp);
if (flags & FORK_SHAREFILES)
fd_share(p2);
else if (flags & FORK_CLEANFILES)
p2->p_fd = fd_init(NULL);
else
p2->p_fd = fd_copy();
/* XXX racy */
p2->p_mqueue_cnt = p1->p_mqueue_cnt;
if (flags & FORK_SHARECWD)
cwdshare(p2);
else
p2->p_cwdi = cwdinit();
/*
* Note: p_limit (rlimit stuff) is copy-on-write, so normally
* we just need increase pl_refcnt.
*/
p1_lim = p1->p_limit;
if (!p1_lim->pl_writeable) {
lim_addref(p1_lim);
p2->p_limit = p1_lim;
} else {
p2->p_limit = lim_copy(p1_lim);
}
if (flags & FORK_PPWAIT) {
/* Mark ourselves as waiting for a child. */
l1->l_pflag |= LP_VFORKWAIT;
p2->p_lflag = PL_PPWAIT;
p2->p_vforklwp = l1;
} else {
p2->p_lflag = 0;
}
p2->p_sflag = 0;
p2->p_slflag = 0;
parent = (flags & FORK_NOWAIT) ? initproc : p1;
p2->p_pptr = parent;
p2->p_ppid = parent->p_pid;
LIST_INIT(&p2->p_children);
p2->p_aio = NULL;
#ifdef KTRACE
/*
* Copy traceflag and tracefile if enabled.
* If not inherited, these were zeroed above.
*/
if (p1->p_traceflag & KTRFAC_INHERIT) {
mutex_enter(&ktrace_lock);
p2->p_traceflag = p1->p_traceflag;
if ((p2->p_tracep = p1->p_tracep) != NULL)
ktradref(p2);
mutex_exit(&ktrace_lock);
}
#endif
/*
* Create signal actions for the child process.
*/
p2->p_sigacts = sigactsinit(p1, flags & FORK_SHARESIGS);
mutex_enter(p1->p_lock);
p2->p_sflag |=
(p1->p_sflag & (PS_STOPFORK | PS_STOPEXEC | PS_NOCLDSTOP));
sched_proc_fork(p1, p2);
mutex_exit(p1->p_lock);
p2->p_stflag = p1->p_stflag;
/*
* p_stats.
* Copy parts of p_stats, and zero out the rest.
*/
p2->p_stats = pstatscopy(p1->p_stats);
/*
* Set up the new process address space.
*/
uvm_proc_fork(p1, p2, (flags & FORK_SHAREVM) ? true : false);
/*
* Finish creating the child process.
* It will return through a different path later.
*/
lwp_create(l1, p2, uaddr, (flags & FORK_PPWAIT) ? LWP_VFORK : 0,
stack, stacksize, (func != NULL) ? func : child_return, arg, &l2,
l1->l_class);
/*
* Inherit l_private from the parent.
* Note that we cannot use lwp_setprivate() here since that
* also sets the CPU TLS register, which is incorrect if the
* process has changed that without letting the kernel know.
*/
l2->l_private = l1->l_private;
/*
* If emulation has a process fork hook, call it now.
*/
if (p2->p_emul->e_proc_fork)
(*p2->p_emul->e_proc_fork)(p2, l1, flags);
/*
* ...and finally, any other random fork hooks that subsystems
* might have registered.
*/
doforkhooks(p2, p1);
SDT_PROBE(proc,,,create, p2, p1, flags, 0, 0);
/*
* It's now safe for the scheduler and other processes to see the
* child process.
*/
mutex_enter(proc_lock);
if (p1->p_session->s_ttyvp != NULL && p1->p_lflag & PL_CONTROLT)
p2->p_lflag |= PL_CONTROLT;
LIST_INSERT_HEAD(&parent->p_children, p2, p_sibling);
p2->p_exitsig = exitsig; /* signal for parent on exit */
/*
* We don't want to tracefork vfork()ed processes because they
* will not receive the SIGTRAP until it is too late.
*/
tracefork = (p1->p_slflag & (PSL_TRACEFORK|PSL_TRACED)) ==
(PSL_TRACEFORK|PSL_TRACED) && (flags && FORK_PPWAIT) == 0;
if (tracefork) {
p2->p_slflag |= PSL_TRACED;
p2->p_opptr = p2->p_pptr;
if (p2->p_pptr != p1->p_pptr) {
struct proc *parent1 = p2->p_pptr;
if (parent1->p_lock < p2->p_lock) {
if (!mutex_tryenter(parent1->p_lock)) {
mutex_exit(p2->p_lock);
mutex_enter(parent1->p_lock);
}
} else if (parent1->p_lock > p2->p_lock) {
mutex_enter(parent1->p_lock);
}
parent1->p_slflag |= PSL_CHTRACED;
proc_reparent(p2, p1->p_pptr);
if (parent1->p_lock != p2->p_lock)
mutex_exit(parent1->p_lock);
}
/*
* Set ptrace status.
*/
p1->p_fpid = p2->p_pid;
p2->p_fpid = p1->p_pid;
}
LIST_INSERT_AFTER(p1, p2, p_pglist);
LIST_INSERT_HEAD(&allproc, p2, p_list);
p2->p_trace_enabled = trace_is_enabled(p2);
#ifdef __HAVE_SYSCALL_INTERN
(*p2->p_emul->e_syscall_intern)(p2);
#endif
/*
* Update stats now that we know the fork was successful.
*/
uvmexp.forks++;
if (flags & FORK_PPWAIT)
uvmexp.forks_ppwait++;
if (flags & FORK_SHAREVM)
uvmexp.forks_sharevm++;
/*
* Pass a pointer to the new process to the caller.
*/
if (rnewprocp != NULL)
*rnewprocp = p2;
if (ktrpoint(KTR_EMUL))
p2->p_traceflag |= KTRFAC_TRC_EMUL;
/*
* Notify any interested parties about the new process.
*/
if (!SLIST_EMPTY(&p1->p_klist)) {
mutex_exit(proc_lock);
KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
mutex_enter(proc_lock);
}
/*
* Make child runnable, set start time, and add to run queue except
* if the parent requested the child to start in SSTOP state.
*/
mutex_enter(p2->p_lock);
/*
* Start profiling.
*/
if ((p2->p_stflag & PST_PROFIL) != 0) {
mutex_spin_enter(&p2->p_stmutex);
startprofclock(p2);
mutex_spin_exit(&p2->p_stmutex);
}
getmicrotime(&p2->p_stats->p_start);
p2->p_acflag = AFORK;
lwp_lock(l2);
KASSERT(p2->p_nrlwps == 1);
if (p2->p_sflag & PS_STOPFORK) {
struct schedstate_percpu *spc = &l2->l_cpu->ci_schedstate;
p2->p_nrlwps = 0;
p2->p_stat = SSTOP;
p2->p_waited = 0;
p1->p_nstopchild++;
l2->l_stat = LSSTOP;
KASSERT(l2->l_wchan == NULL);
lwp_unlock_to(l2, spc->spc_lwplock);
} else {
p2->p_nrlwps = 1;
p2->p_stat = SACTIVE;
l2->l_stat = LSRUN;
sched_enqueue(l2, false);
lwp_unlock(l2);
}
/*
* Return child pid to parent process,
* marking us as parent via retval[1].
*/
if (retval != NULL) {
retval[0] = p2->p_pid;
retval[1] = 0;
}
mutex_exit(p2->p_lock);
/*
* Preserve synchronization semantics of vfork. If waiting for
* child to exec or exit, sleep until it clears LP_VFORKWAIT.
*/
#if 0
while (l1->l_pflag & LP_VFORKWAIT) {
cv_wait(&l1->l_waitcv, proc_lock);
}
#else
while (p2->p_lflag & PL_PPWAIT)
cv_wait(&p1->p_waitcv, proc_lock);
#endif
/*
* Let the parent know that we are tracing its child.
*/
if (tracefork) {
ksiginfo_t ksi;
KSI_INIT_EMPTY(&ksi);
ksi.ksi_signo = SIGTRAP;
ksi.ksi_lid = l1->l_lid;
kpsignal(p1, &ksi, NULL);
}
mutex_exit(proc_lock);
return 0;
}
