diff --git a/sys/kern/kern_descrip.c b/sys/kern/kern_descrip.c
index 585f2124eab1..36092c9acd42 100644
--- a/sys/kern/kern_descrip.c
+++ b/sys/kern/kern_descrip.c
@@ -1,5141 +1,5137 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1982, 1986, 1989, 1991, 1993
* The Regents of the University of California. All rights reserved.
* (c) UNIX System Laboratories, Inc.
* All or some portions of this file are derived from material licensed
* to the University of California by American Telephone and Telegraph
* Co. or Unix System Laboratories, Inc. and are reproduced herein with
* the permission of UNIX System Laboratories, Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)kern_descrip.c 8.6 (Berkeley) 4/19/94
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_capsicum.h"
#include "opt_ddb.h"
#include "opt_ktrace.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/capsicum.h>
#include <sys/conf.h>
#include <sys/fcntl.h>
#include <sys/file.h>
#include <sys/filedesc.h>
#include <sys/filio.h>
#include <sys/jail.h>
#include <sys/kernel.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/mutex.h>
#include <sys/namei.h>
#include <sys/selinfo.h>
#include <sys/poll.h>
#include <sys/priv.h>
#include <sys/proc.h>
#include <sys/protosw.h>
#include <sys/racct.h>
#include <sys/resourcevar.h>
#include <sys/sbuf.h>
#include <sys/signalvar.h>
#include <sys/kdb.h>
#include <sys/smr.h>
#include <sys/stat.h>
#include <sys/sx.h>
#include <sys/syscallsubr.h>
#include <sys/sysctl.h>
#include <sys/sysproto.h>
#include <sys/unistd.h>
#include <sys/user.h>
#include <sys/vnode.h>
-#ifdef KTRACE
#include <sys/ktrace.h>
-#endif
#include <net/vnet.h>
#include <security/audit/audit.h>
#include <vm/uma.h>
#include <vm/vm.h>
#include <ddb/ddb.h>
static MALLOC_DEFINE(M_FILEDESC, "filedesc", "Open file descriptor table");
static MALLOC_DEFINE(M_PWD, "pwd", "Descriptor table vnodes");
static MALLOC_DEFINE(M_PWDDESC, "pwddesc", "Pwd descriptors");
static MALLOC_DEFINE(M_FILEDESC_TO_LEADER, "filedesc_to_leader",
"file desc to leader structures");
static MALLOC_DEFINE(M_SIGIO, "sigio", "sigio structures");
MALLOC_DEFINE(M_FILECAPS, "filecaps", "descriptor capabilities");
MALLOC_DECLARE(M_FADVISE);
static __read_mostly uma_zone_t file_zone;
static __read_mostly uma_zone_t filedesc0_zone;
__read_mostly uma_zone_t pwd_zone;
VFS_SMR_DECLARE;
static int closefp(struct filedesc *fdp, int fd, struct file *fp,
struct thread *td, bool holdleaders, bool audit);
static int fd_first_free(struct filedesc *fdp, int low, int size);
static void fdgrowtable(struct filedesc *fdp, int nfd);
static void fdgrowtable_exp(struct filedesc *fdp, int nfd);
static void fdunused(struct filedesc *fdp, int fd);
static void fdused(struct filedesc *fdp, int fd);
static int getmaxfd(struct thread *td);
static u_long *filecaps_copy_prep(const struct filecaps *src);
static void filecaps_copy_finish(const struct filecaps *src,
struct filecaps *dst, u_long *ioctls);
static u_long *filecaps_free_prep(struct filecaps *fcaps);
static void filecaps_free_finish(u_long *ioctls);
static struct pwd *pwd_alloc(void);
/*
* Each process has:
*
* - An array of open file descriptors (fd_ofiles)
* - An array of file flags (fd_ofileflags)
* - A bitmap recording which descriptors are in use (fd_map)
*
* A process starts out with NDFILE descriptors. The value of NDFILE has
* been selected based the historical limit of 20 open files, and an
* assumption that the majority of processes, especially short-lived
* processes like shells, will never need more.
*
* If this initial allocation is exhausted, a larger descriptor table and
* map are allocated dynamically, and the pointers in the process's struct
* filedesc are updated to point to those. This is repeated every time
* the process runs out of file descriptors (provided it hasn't hit its
* resource limit).
*
* Since threads may hold references to individual descriptor table
* entries, the tables are never freed. Instead, they are placed on a
* linked list and freed only when the struct filedesc is released.
*/
#define NDFILE 20
#define NDSLOTSIZE sizeof(NDSLOTTYPE)
#define NDENTRIES (NDSLOTSIZE * __CHAR_BIT)
#define NDSLOT(x) ((x) / NDENTRIES)
#define NDBIT(x) ((NDSLOTTYPE)1 << ((x) % NDENTRIES))
#define NDSLOTS(x) (((x) + NDENTRIES - 1) / NDENTRIES)
/*
* SLIST entry used to keep track of ofiles which must be reclaimed when
* the process exits.
*/
struct freetable {
struct fdescenttbl *ft_table;
SLIST_ENTRY(freetable) ft_next;
};
/*
* Initial allocation: a filedesc structure + the head of SLIST used to
* keep track of old ofiles + enough space for NDFILE descriptors.
*/
struct fdescenttbl0 {
int fdt_nfiles;
struct filedescent fdt_ofiles[NDFILE];
};
struct filedesc0 {
struct filedesc fd_fd;
SLIST_HEAD(, freetable) fd_free;
struct fdescenttbl0 fd_dfiles;
NDSLOTTYPE fd_dmap[NDSLOTS(NDFILE)];
};
/*
* Descriptor management.
*/
static int __exclusive_cache_line openfiles; /* actual number of open files */
struct mtx sigio_lock; /* mtx to protect pointers to sigio */
void __read_mostly (*mq_fdclose)(struct thread *td, int fd, struct file *fp);
/*
* If low >= size, just return low. Otherwise find the first zero bit in the
* given bitmap, starting at low and not exceeding size - 1. Return size if
* not found.
*/
static int
fd_first_free(struct filedesc *fdp, int low, int size)
{
NDSLOTTYPE *map = fdp->fd_map;
NDSLOTTYPE mask;
int off, maxoff;
if (low >= size)
return (low);
off = NDSLOT(low);
if (low % NDENTRIES) {
mask = ~(~(NDSLOTTYPE)0 >> (NDENTRIES - (low % NDENTRIES)));
if ((mask &= ~map[off]) != 0UL)
return (off * NDENTRIES + ffsl(mask) - 1);
++off;
}
for (maxoff = NDSLOTS(size); off < maxoff; ++off)
if (map[off] != ~0UL)
return (off * NDENTRIES + ffsl(~map[off]) - 1);
return (size);
}
/*
* Find the last used fd.
*
* Call this variant if fdp can't be modified by anyone else (e.g, during exec).
* Otherwise use fdlastfile.
*/
int
fdlastfile_single(struct filedesc *fdp)
{
NDSLOTTYPE *map = fdp->fd_map;
int off, minoff;
off = NDSLOT(fdp->fd_nfiles - 1);
for (minoff = NDSLOT(0); off >= minoff; --off)
if (map[off] != 0)
return (off * NDENTRIES + flsl(map[off]) - 1);
return (-1);
}
int
fdlastfile(struct filedesc *fdp)
{
FILEDESC_LOCK_ASSERT(fdp);
return (fdlastfile_single(fdp));
}
static int
fdisused(struct filedesc *fdp, int fd)
{
KASSERT(fd >= 0 && fd < fdp->fd_nfiles,
("file descriptor %d out of range (0, %d)", fd, fdp->fd_nfiles));
return ((fdp->fd_map[NDSLOT(fd)] & NDBIT(fd)) != 0);
}
/*
* Mark a file descriptor as used.
*/
static void
fdused_init(struct filedesc *fdp, int fd)
{
KASSERT(!fdisused(fdp, fd), ("fd=%d is already used", fd));
fdp->fd_map[NDSLOT(fd)] |= NDBIT(fd);
}
static void
fdused(struct filedesc *fdp, int fd)
{
FILEDESC_XLOCK_ASSERT(fdp);
fdused_init(fdp, fd);
if (fd == fdp->fd_freefile)
fdp->fd_freefile++;
}
/*
* Mark a file descriptor as unused.
*/
static void
fdunused(struct filedesc *fdp, int fd)
{
FILEDESC_XLOCK_ASSERT(fdp);
KASSERT(fdisused(fdp, fd), ("fd=%d is already unused", fd));
KASSERT(fdp->fd_ofiles[fd].fde_file == NULL,
("fd=%d is still in use", fd));
fdp->fd_map[NDSLOT(fd)] &= ~NDBIT(fd);
if (fd < fdp->fd_freefile)
fdp->fd_freefile = fd;
}
/*
* Free a file descriptor.
*
* Avoid some work if fdp is about to be destroyed.
*/
static inline void
fdefree_last(struct filedescent *fde)
{
filecaps_free(&fde->fde_caps);
}
static inline void
fdfree(struct filedesc *fdp, int fd)
{
struct filedescent *fde;
FILEDESC_XLOCK_ASSERT(fdp);
fde = &fdp->fd_ofiles[fd];
#ifdef CAPABILITIES
seqc_write_begin(&fde->fde_seqc);
#endif
fde->fde_file = NULL;
#ifdef CAPABILITIES
seqc_write_end(&fde->fde_seqc);
#endif
fdefree_last(fde);
fdunused(fdp, fd);
}
/*
* System calls on descriptors.
*/
#ifndef _SYS_SYSPROTO_H_
struct getdtablesize_args {
int dummy;
};
#endif
/* ARGSUSED */
int
sys_getdtablesize(struct thread *td, struct getdtablesize_args *uap)
{
#ifdef RACCT
uint64_t lim;
#endif
td->td_retval[0] = getmaxfd(td);
#ifdef RACCT
PROC_LOCK(td->td_proc);
lim = racct_get_limit(td->td_proc, RACCT_NOFILE);
PROC_UNLOCK(td->td_proc);
if (lim < td->td_retval[0])
td->td_retval[0] = lim;
#endif
return (0);
}
/*
* Duplicate a file descriptor to a particular value.
*
* Note: keep in mind that a potential race condition exists when closing
* descriptors from a shared descriptor table (via rfork).
*/
#ifndef _SYS_SYSPROTO_H_
struct dup2_args {
u_int from;
u_int to;
};
#endif
/* ARGSUSED */
int
sys_dup2(struct thread *td, struct dup2_args *uap)
{
return (kern_dup(td, FDDUP_FIXED, 0, (int)uap->from, (int)uap->to));
}
/*
* Duplicate a file descriptor.
*/
#ifndef _SYS_SYSPROTO_H_
struct dup_args {
u_int fd;
};
#endif
/* ARGSUSED */
int
sys_dup(struct thread *td, struct dup_args *uap)
{
return (kern_dup(td, FDDUP_NORMAL, 0, (int)uap->fd, 0));
}
/*
* The file control system call.
*/
#ifndef _SYS_SYSPROTO_H_
struct fcntl_args {
int fd;
int cmd;
long arg;
};
#endif
/* ARGSUSED */
int
sys_fcntl(struct thread *td, struct fcntl_args *uap)
{
return (kern_fcntl_freebsd(td, uap->fd, uap->cmd, uap->arg));
}
int
kern_fcntl_freebsd(struct thread *td, int fd, int cmd, long arg)
{
struct flock fl;
struct __oflock ofl;
intptr_t arg1;
int error, newcmd;
error = 0;
newcmd = cmd;
switch (cmd) {
case F_OGETLK:
case F_OSETLK:
case F_OSETLKW:
/*
* Convert old flock structure to new.
*/
error = copyin((void *)(intptr_t)arg, &ofl, sizeof(ofl));
fl.l_start = ofl.l_start;
fl.l_len = ofl.l_len;
fl.l_pid = ofl.l_pid;
fl.l_type = ofl.l_type;
fl.l_whence = ofl.l_whence;
fl.l_sysid = 0;
switch (cmd) {
case F_OGETLK:
newcmd = F_GETLK;
break;
case F_OSETLK:
newcmd = F_SETLK;
break;
case F_OSETLKW:
newcmd = F_SETLKW;
break;
}
arg1 = (intptr_t)&fl;
break;
case F_GETLK:
case F_SETLK:
case F_SETLKW:
case F_SETLK_REMOTE:
error = copyin((void *)(intptr_t)arg, &fl, sizeof(fl));
arg1 = (intptr_t)&fl;
break;
default:
arg1 = arg;
break;
}
if (error)
return (error);
error = kern_fcntl(td, fd, newcmd, arg1);
if (error)
return (error);
if (cmd == F_OGETLK) {
ofl.l_start = fl.l_start;
ofl.l_len = fl.l_len;
ofl.l_pid = fl.l_pid;
ofl.l_type = fl.l_type;
ofl.l_whence = fl.l_whence;
error = copyout(&ofl, (void *)(intptr_t)arg, sizeof(ofl));
} else if (cmd == F_GETLK) {
error = copyout(&fl, (void *)(intptr_t)arg, sizeof(fl));
}
return (error);
}
int
kern_fcntl(struct thread *td, int fd, int cmd, intptr_t arg)
{
struct filedesc *fdp;
struct flock *flp;
struct file *fp, *fp2;
struct filedescent *fde;
struct proc *p;
struct vnode *vp;
struct mount *mp;
int error, flg, seals, tmp;
uint64_t bsize;
off_t foffset;
error = 0;
flg = F_POSIX;
p = td->td_proc;
fdp = p->p_fd;
AUDIT_ARG_FD(cmd);
AUDIT_ARG_CMD(cmd);
switch (cmd) {
case F_DUPFD:
tmp = arg;
error = kern_dup(td, FDDUP_FCNTL, 0, fd, tmp);
break;
case F_DUPFD_CLOEXEC:
tmp = arg;
error = kern_dup(td, FDDUP_FCNTL, FDDUP_FLAG_CLOEXEC, fd, tmp);
break;
case F_DUP2FD:
tmp = arg;
error = kern_dup(td, FDDUP_FIXED, 0, fd, tmp);
break;
case F_DUP2FD_CLOEXEC:
tmp = arg;
error = kern_dup(td, FDDUP_FIXED, FDDUP_FLAG_CLOEXEC, fd, tmp);
break;
case F_GETFD:
error = EBADF;
FILEDESC_SLOCK(fdp);
fde = fdeget_locked(fdp, fd);
if (fde != NULL) {
td->td_retval[0] =
(fde->fde_flags & UF_EXCLOSE) ? FD_CLOEXEC : 0;
error = 0;
}
FILEDESC_SUNLOCK(fdp);
break;
case F_SETFD:
error = EBADF;
FILEDESC_XLOCK(fdp);
fde = fdeget_locked(fdp, fd);
if (fde != NULL) {
fde->fde_flags = (fde->fde_flags & ~UF_EXCLOSE) |
(arg & FD_CLOEXEC ? UF_EXCLOSE : 0);
error = 0;
}
FILEDESC_XUNLOCK(fdp);
break;
case F_GETFL:
error = fget_fcntl(td, fd, &cap_fcntl_rights, F_GETFL, &fp);
if (error != 0)
break;
td->td_retval[0] = OFLAGS(fp->f_flag);
fdrop(fp, td);
break;
case F_SETFL:
error = fget_fcntl(td, fd, &cap_fcntl_rights, F_SETFL, &fp);
if (error != 0)
break;
if (fp->f_ops == &path_fileops) {
fdrop(fp, td);
error = EBADF;
break;
}
do {
tmp = flg = fp->f_flag;
tmp &= ~FCNTLFLAGS;
tmp |= FFLAGS(arg & ~O_ACCMODE) & FCNTLFLAGS;
} while (atomic_cmpset_int(&fp->f_flag, flg, tmp) == 0);
tmp = fp->f_flag & FNONBLOCK;
error = fo_ioctl(fp, FIONBIO, &tmp, td->td_ucred, td);
if (error != 0) {
fdrop(fp, td);
break;
}
tmp = fp->f_flag & FASYNC;
error = fo_ioctl(fp, FIOASYNC, &tmp, td->td_ucred, td);
if (error == 0) {
fdrop(fp, td);
break;
}
atomic_clear_int(&fp->f_flag, FNONBLOCK);
tmp = 0;
(void)fo_ioctl(fp, FIONBIO, &tmp, td->td_ucred, td);
fdrop(fp, td);
break;
case F_GETOWN:
error = fget_fcntl(td, fd, &cap_fcntl_rights, F_GETOWN, &fp);
if (error != 0)
break;
error = fo_ioctl(fp, FIOGETOWN, &tmp, td->td_ucred, td);
if (error == 0)
td->td_retval[0] = tmp;
fdrop(fp, td);
break;
case F_SETOWN:
error = fget_fcntl(td, fd, &cap_fcntl_rights, F_SETOWN, &fp);
if (error != 0)
break;
tmp = arg;
error = fo_ioctl(fp, FIOSETOWN, &tmp, td->td_ucred, td);
fdrop(fp, td);
break;
case F_SETLK_REMOTE:
error = priv_check(td, PRIV_NFS_LOCKD);
if (error != 0)
return (error);
flg = F_REMOTE;
goto do_setlk;
case F_SETLKW:
flg |= F_WAIT;
/* FALLTHROUGH F_SETLK */
case F_SETLK:
do_setlk:
flp = (struct flock *)arg;
if ((flg & F_REMOTE) != 0 && flp->l_sysid == 0) {
error = EINVAL;
break;
}
error = fget_unlocked(fdp, fd, &cap_flock_rights, &fp);
if (error != 0)
break;
if (fp->f_type != DTYPE_VNODE || fp->f_ops == &path_fileops) {
error = EBADF;
fdrop(fp, td);
break;
}
if (flp->l_whence == SEEK_CUR) {
foffset = foffset_get(fp);
if (foffset < 0 ||
(flp->l_start > 0 &&
foffset > OFF_MAX - flp->l_start)) {
error = EOVERFLOW;
fdrop(fp, td);
break;
}
flp->l_start += foffset;
}
vp = fp->f_vnode;
switch (flp->l_type) {
case F_RDLCK:
if ((fp->f_flag & FREAD) == 0) {
error = EBADF;
break;
}
if ((p->p_leader->p_flag & P_ADVLOCK) == 0) {
PROC_LOCK(p->p_leader);
p->p_leader->p_flag |= P_ADVLOCK;
PROC_UNLOCK(p->p_leader);
}
error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_SETLK,
flp, flg);
break;
case F_WRLCK:
if ((fp->f_flag & FWRITE) == 0) {
error = EBADF;
break;
}
if ((p->p_leader->p_flag & P_ADVLOCK) == 0) {
PROC_LOCK(p->p_leader);
p->p_leader->p_flag |= P_ADVLOCK;
PROC_UNLOCK(p->p_leader);
}
error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_SETLK,
flp, flg);
break;
case F_UNLCK:
error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_UNLCK,
flp, flg);
break;
case F_UNLCKSYS:
if (flg != F_REMOTE) {
error = EINVAL;
break;
}
error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader,
F_UNLCKSYS, flp, flg);
break;
default:
error = EINVAL;
break;
}
if (error != 0 || flp->l_type == F_UNLCK ||
flp->l_type == F_UNLCKSYS) {
fdrop(fp, td);
break;
}
/*
* Check for a race with close.
*
* The vnode is now advisory locked (or unlocked, but this case
* is not really important) as the caller requested.
* We had to drop the filedesc lock, so we need to recheck if
* the descriptor is still valid, because if it was closed
* in the meantime we need to remove advisory lock from the
* vnode - close on any descriptor leading to an advisory
* locked vnode, removes that lock.
* We will return 0 on purpose in that case, as the result of
* successful advisory lock might have been externally visible
* already. This is fine - effectively we pretend to the caller
* that the closing thread was a bit slower and that the
* advisory lock succeeded before the close.
*/
error = fget_unlocked(fdp, fd, &cap_no_rights, &fp2);
if (error != 0) {
fdrop(fp, td);
break;
}
if (fp != fp2) {
flp->l_whence = SEEK_SET;
flp->l_start = 0;
flp->l_len = 0;
flp->l_type = F_UNLCK;
(void) VOP_ADVLOCK(vp, (caddr_t)p->p_leader,
F_UNLCK, flp, F_POSIX);
}
fdrop(fp, td);
fdrop(fp2, td);
break;
case F_GETLK:
error = fget_unlocked(fdp, fd, &cap_flock_rights, &fp);
if (error != 0)
break;
if (fp->f_type != DTYPE_VNODE || fp->f_ops == &path_fileops) {
error = EBADF;
fdrop(fp, td);
break;
}
flp = (struct flock *)arg;
if (flp->l_type != F_RDLCK && flp->l_type != F_WRLCK &&
flp->l_type != F_UNLCK) {
error = EINVAL;
fdrop(fp, td);
break;
}
if (flp->l_whence == SEEK_CUR) {
foffset = foffset_get(fp);
if ((flp->l_start > 0 &&
foffset > OFF_MAX - flp->l_start) ||
(flp->l_start < 0 &&
foffset < OFF_MIN - flp->l_start)) {
error = EOVERFLOW;
fdrop(fp, td);
break;
}
flp->l_start += foffset;
}
vp = fp->f_vnode;
error = VOP_ADVLOCK(vp, (caddr_t)p->p_leader, F_GETLK, flp,
F_POSIX);
fdrop(fp, td);
break;
case F_ADD_SEALS:
error = fget_unlocked(fdp, fd, &cap_no_rights, &fp);
if (error != 0)
break;
error = fo_add_seals(fp, arg);
fdrop(fp, td);
break;
case F_GET_SEALS:
error = fget_unlocked(fdp, fd, &cap_no_rights, &fp);
if (error != 0)
break;
if (fo_get_seals(fp, &seals) == 0)
td->td_retval[0] = seals;
else
error = EINVAL;
fdrop(fp, td);
break;
case F_RDAHEAD:
arg = arg ? 128 * 1024: 0;
/* FALLTHROUGH */
case F_READAHEAD:
error = fget_unlocked(fdp, fd, &cap_no_rights, &fp);
if (error != 0)
break;
if (fp->f_type != DTYPE_VNODE || fp->f_ops == &path_fileops) {
fdrop(fp, td);
error = EBADF;
break;
}
vp = fp->f_vnode;
if (vp->v_type != VREG) {
fdrop(fp, td);
error = ENOTTY;
break;
}
/*
* Exclusive lock synchronizes against f_seqcount reads and
* writes in sequential_heuristic().
*/
error = vn_lock(vp, LK_EXCLUSIVE);
if (error != 0) {
fdrop(fp, td);
break;
}
if (arg >= 0) {
bsize = fp->f_vnode->v_mount->mnt_stat.f_iosize;
arg = MIN(arg, INT_MAX - bsize + 1);
fp->f_seqcount[UIO_READ] = MIN(IO_SEQMAX,
(arg + bsize - 1) / bsize);
atomic_set_int(&fp->f_flag, FRDAHEAD);
} else {
atomic_clear_int(&fp->f_flag, FRDAHEAD);
}
VOP_UNLOCK(vp);
fdrop(fp, td);
break;
case F_ISUNIONSTACK:
/*
* Check if the vnode is part of a union stack (either the
* "union" flag from mount(2) or unionfs).
*
* Prior to introduction of this op libc's readdir would call
* fstatfs(2), in effect unnecessarily copying kilobytes of
* data just to check fs name and a mount flag.
*
* Fixing the code to handle everything in the kernel instead
* is a non-trivial endeavor and has low priority, thus this
* horrible kludge facilitates the current behavior in a much
* cheaper manner until someone(tm) sorts this out.
*/
error = fget_unlocked(fdp, fd, &cap_no_rights, &fp);
if (error != 0)
break;
if (fp->f_type != DTYPE_VNODE) {
fdrop(fp, td);
error = EBADF;
break;
}
vp = fp->f_vnode;
/*
* Since we don't prevent dooming the vnode even non-null mp
* found can become immediately stale. This is tolerable since
* mount points are type-stable (providing safe memory access)
* and any vfs op on this vnode going forward will return an
* error (meaning return value in this case is meaningless).
*/
mp = atomic_load_ptr(&vp->v_mount);
if (__predict_false(mp == NULL)) {
fdrop(fp, td);
error = EBADF;
break;
}
td->td_retval[0] = 0;
if (mp->mnt_kern_flag & MNTK_UNIONFS ||
mp->mnt_flag & MNT_UNION)
td->td_retval[0] = 1;
fdrop(fp, td);
break;
default:
error = EINVAL;
break;
}
return (error);
}
static int
getmaxfd(struct thread *td)
{
return (min((int)lim_cur(td, RLIMIT_NOFILE), maxfilesperproc));
}
/*
* Common code for dup, dup2, fcntl(F_DUPFD) and fcntl(F_DUP2FD).
*/
int
kern_dup(struct thread *td, u_int mode, int flags, int old, int new)
{
struct filedesc *fdp;
struct filedescent *oldfde, *newfde;
struct proc *p;
struct file *delfp, *oldfp;
u_long *oioctls, *nioctls;
int error, maxfd;
p = td->td_proc;
fdp = p->p_fd;
oioctls = NULL;
MPASS((flags & ~(FDDUP_FLAG_CLOEXEC)) == 0);
MPASS(mode < FDDUP_LASTMODE);
AUDIT_ARG_FD(old);
/* XXXRW: if (flags & FDDUP_FIXED) AUDIT_ARG_FD2(new); */
/*
* Verify we have a valid descriptor to dup from and possibly to
* dup to. Unlike dup() and dup2(), fcntl()'s F_DUPFD should
* return EINVAL when the new descriptor is out of bounds.
*/
if (old < 0)
return (EBADF);
if (new < 0)
return (mode == FDDUP_FCNTL ? EINVAL : EBADF);
maxfd = getmaxfd(td);
if (new >= maxfd)
return (mode == FDDUP_FCNTL ? EINVAL : EBADF);
error = EBADF;
FILEDESC_XLOCK(fdp);
if (fget_locked(fdp, old) == NULL)
goto unlock;
if ((mode == FDDUP_FIXED || mode == FDDUP_MUSTREPLACE) && old == new) {
td->td_retval[0] = new;
if (flags & FDDUP_FLAG_CLOEXEC)
fdp->fd_ofiles[new].fde_flags |= UF_EXCLOSE;
error = 0;
goto unlock;
}
oldfde = &fdp->fd_ofiles[old];
oldfp = oldfde->fde_file;
if (!fhold(oldfp))
goto unlock;
/*
* If the caller specified a file descriptor, make sure the file
* table is large enough to hold it, and grab it. Otherwise, just
* allocate a new descriptor the usual way.
*/
switch (mode) {
case FDDUP_NORMAL:
case FDDUP_FCNTL:
if ((error = fdalloc(td, new, &new)) != 0) {
fdrop(oldfp, td);
goto unlock;
}
break;
case FDDUP_MUSTREPLACE:
/* Target file descriptor must exist. */
if (fget_locked(fdp, new) == NULL) {
fdrop(oldfp, td);
goto unlock;
}
break;
case FDDUP_FIXED:
if (new >= fdp->fd_nfiles) {
/*
* The resource limits are here instead of e.g.
* fdalloc(), because the file descriptor table may be
* shared between processes, so we can't really use
* racct_add()/racct_sub(). Instead of counting the
* number of actually allocated descriptors, just put
* the limit on the size of the file descriptor table.
*/
#ifdef RACCT
if (RACCT_ENABLED()) {
error = racct_set_unlocked(p, RACCT_NOFILE, new + 1);
if (error != 0) {
error = EMFILE;
fdrop(oldfp, td);
goto unlock;
}
}
#endif
fdgrowtable_exp(fdp, new + 1);
}
if (!fdisused(fdp, new))
fdused(fdp, new);
break;
default:
KASSERT(0, ("%s unsupported mode %d", __func__, mode));
}
KASSERT(old != new, ("new fd is same as old"));
/* Refetch oldfde because the table may have grown and old one freed. */
oldfde = &fdp->fd_ofiles[old];
KASSERT(oldfp == oldfde->fde_file,
("fdt_ofiles shift from growth observed at fd %d",
old));
newfde = &fdp->fd_ofiles[new];
delfp = newfde->fde_file;
nioctls = filecaps_copy_prep(&oldfde->fde_caps);
/*
* Duplicate the source descriptor.
*/
#ifdef CAPABILITIES
seqc_write_begin(&newfde->fde_seqc);
#endif
oioctls = filecaps_free_prep(&newfde->fde_caps);
memcpy(newfde, oldfde, fde_change_size);
filecaps_copy_finish(&oldfde->fde_caps, &newfde->fde_caps,
nioctls);
if ((flags & FDDUP_FLAG_CLOEXEC) != 0)
newfde->fde_flags = oldfde->fde_flags | UF_EXCLOSE;
else
newfde->fde_flags = oldfde->fde_flags & ~UF_EXCLOSE;
#ifdef CAPABILITIES
seqc_write_end(&newfde->fde_seqc);
#endif
td->td_retval[0] = new;
error = 0;
if (delfp != NULL) {
(void) closefp(fdp, new, delfp, td, true, false);
FILEDESC_UNLOCK_ASSERT(fdp);
} else {
unlock:
FILEDESC_XUNLOCK(fdp);
}
filecaps_free_finish(oioctls);
return (error);
}
static void
sigiofree(struct sigio *sigio)
{
crfree(sigio->sio_ucred);
free(sigio, M_SIGIO);
}
static struct sigio *
funsetown_locked(struct sigio *sigio)
{
struct proc *p;
struct pgrp *pg;
SIGIO_ASSERT_LOCKED();
if (sigio == NULL)
return (NULL);
*(sigio->sio_myref) = NULL;
if (sigio->sio_pgid < 0) {
pg = sigio->sio_pgrp;
PGRP_LOCK(pg);
SLIST_REMOVE(&sigio->sio_pgrp->pg_sigiolst, sigio,
sigio, sio_pgsigio);
PGRP_UNLOCK(pg);
} else {
p = sigio->sio_proc;
PROC_LOCK(p);
SLIST_REMOVE(&sigio->sio_proc->p_sigiolst, sigio,
sigio, sio_pgsigio);
PROC_UNLOCK(p);
}
return (sigio);
}
/*
* If sigio is on the list associated with a process or process group,
* disable signalling from the device, remove sigio from the list and
* free sigio.
*/
void
funsetown(struct sigio **sigiop)
{
struct sigio *sigio;
/* Racy check, consumers must provide synchronization. */
if (*sigiop == NULL)
return;
SIGIO_LOCK();
sigio = funsetown_locked(*sigiop);
SIGIO_UNLOCK();
if (sigio != NULL)
sigiofree(sigio);
}
/*
* Free a list of sigio structures. The caller must ensure that new sigio
* structures cannot be added after this point. For process groups this is
* guaranteed using the proctree lock; for processes, the P_WEXIT flag serves
* as an interlock.
*/
void
funsetownlst(struct sigiolst *sigiolst)
{
struct proc *p;
struct pgrp *pg;
struct sigio *sigio, *tmp;
/* Racy check. */
sigio = SLIST_FIRST(sigiolst);
if (sigio == NULL)
return;
p = NULL;
pg = NULL;
SIGIO_LOCK();
sigio = SLIST_FIRST(sigiolst);
if (sigio == NULL) {
SIGIO_UNLOCK();
return;
}
/*
* Every entry of the list should belong to a single proc or pgrp.
*/
if (sigio->sio_pgid < 0) {
pg = sigio->sio_pgrp;
sx_assert(&proctree_lock, SX_XLOCKED);
PGRP_LOCK(pg);
} else /* if (sigio->sio_pgid > 0) */ {
p = sigio->sio_proc;
PROC_LOCK(p);
KASSERT((p->p_flag & P_WEXIT) != 0,
("%s: process %p is not exiting", __func__, p));
}
SLIST_FOREACH(sigio, sigiolst, sio_pgsigio) {
*sigio->sio_myref = NULL;
if (pg != NULL) {
KASSERT(sigio->sio_pgid < 0,
("Proc sigio in pgrp sigio list"));
KASSERT(sigio->sio_pgrp == pg,
("Bogus pgrp in sigio list"));
} else /* if (p != NULL) */ {
KASSERT(sigio->sio_pgid > 0,
("Pgrp sigio in proc sigio list"));
KASSERT(sigio->sio_proc == p,
("Bogus proc in sigio list"));
}
}
if (pg != NULL)
PGRP_UNLOCK(pg);
else
PROC_UNLOCK(p);
SIGIO_UNLOCK();
SLIST_FOREACH_SAFE(sigio, sigiolst, sio_pgsigio, tmp)
sigiofree(sigio);
}
/*
* This is common code for FIOSETOWN ioctl called by fcntl(fd, F_SETOWN, arg).
*
* After permission checking, add a sigio structure to the sigio list for
* the process or process group.
*/
int
fsetown(pid_t pgid, struct sigio **sigiop)
{
struct proc *proc;
struct pgrp *pgrp;
struct sigio *osigio, *sigio;
int ret;
if (pgid == 0) {
funsetown(sigiop);
return (0);
}
ret = 0;
sigio = malloc(sizeof(struct sigio), M_SIGIO, M_WAITOK);
sigio->sio_pgid = pgid;
sigio->sio_ucred = crhold(curthread->td_ucred);
sigio->sio_myref = sigiop;
sx_slock(&proctree_lock);
SIGIO_LOCK();
osigio = funsetown_locked(*sigiop);
if (pgid > 0) {
proc = pfind(pgid);
if (proc == NULL) {
ret = ESRCH;
goto fail;
}
/*
* Policy - Don't allow a process to FSETOWN a process
* in another session.
*
* Remove this test to allow maximum flexibility or
* restrict FSETOWN to the current process or process
* group for maximum safety.
*/
if (proc->p_session != curthread->td_proc->p_session) {
PROC_UNLOCK(proc);
ret = EPERM;
goto fail;
}
sigio->sio_proc = proc;
SLIST_INSERT_HEAD(&proc->p_sigiolst, sigio, sio_pgsigio);
PROC_UNLOCK(proc);
} else /* if (pgid < 0) */ {
pgrp = pgfind(-pgid);
if (pgrp == NULL) {
ret = ESRCH;
goto fail;
}
/*
* Policy - Don't allow a process to FSETOWN a process
* in another session.
*
* Remove this test to allow maximum flexibility or
* restrict FSETOWN to the current process or process
* group for maximum safety.
*/
if (pgrp->pg_session != curthread->td_proc->p_session) {
PGRP_UNLOCK(pgrp);
ret = EPERM;
goto fail;
}
SLIST_INSERT_HEAD(&pgrp->pg_sigiolst, sigio, sio_pgsigio);
sigio->sio_pgrp = pgrp;
PGRP_UNLOCK(pgrp);
}
sx_sunlock(&proctree_lock);
*sigiop = sigio;
SIGIO_UNLOCK();
if (osigio != NULL)
sigiofree(osigio);
return (0);
fail:
SIGIO_UNLOCK();
sx_sunlock(&proctree_lock);
sigiofree(sigio);
if (osigio != NULL)
sigiofree(osigio);
return (ret);
}
/*
* This is common code for FIOGETOWN ioctl called by fcntl(fd, F_GETOWN, arg).
*/
pid_t
fgetown(struct sigio **sigiop)
{
pid_t pgid;
SIGIO_LOCK();
pgid = (*sigiop != NULL) ? (*sigiop)->sio_pgid : 0;
SIGIO_UNLOCK();
return (pgid);
}
static int
closefp_impl(struct filedesc *fdp, int fd, struct file *fp, struct thread *td,
bool audit)
{
int error;
FILEDESC_XLOCK_ASSERT(fdp);
/*
* We now hold the fp reference that used to be owned by the
* descriptor array. We have to unlock the FILEDESC *AFTER*
* knote_fdclose to prevent a race of the fd getting opened, a knote
* added, and deleteing a knote for the new fd.
*/
if (__predict_false(!TAILQ_EMPTY(&fdp->fd_kqlist)))
knote_fdclose(td, fd);
/*
* We need to notify mqueue if the object is of type mqueue.
*/
if (__predict_false(fp->f_type == DTYPE_MQUEUE))
mq_fdclose(td, fd, fp);
FILEDESC_XUNLOCK(fdp);
#ifdef AUDIT
if (AUDITING_TD(td) && audit)
audit_sysclose(td, fd, fp);
#endif
error = closef(fp, td);
/*
* All paths leading up to closefp() will have already removed or
* replaced the fd in the filedesc table, so a restart would not
* operate on the same file.
*/
if (error == ERESTART)
error = EINTR;
return (error);
}
static int
closefp_hl(struct filedesc *fdp, int fd, struct file *fp, struct thread *td,
bool holdleaders, bool audit)
{
int error;
FILEDESC_XLOCK_ASSERT(fdp);
if (holdleaders) {
if (td->td_proc->p_fdtol != NULL) {
/*
* Ask fdfree() to sleep to ensure that all relevant
* process leaders can be traversed in closef().
*/
fdp->fd_holdleaderscount++;
} else {
holdleaders = false;
}
}
error = closefp_impl(fdp, fd, fp, td, audit);
if (holdleaders) {
FILEDESC_XLOCK(fdp);
fdp->fd_holdleaderscount--;
if (fdp->fd_holdleaderscount == 0 &&
fdp->fd_holdleaderswakeup != 0) {
fdp->fd_holdleaderswakeup = 0;
wakeup(&fdp->fd_holdleaderscount);
}
FILEDESC_XUNLOCK(fdp);
}
return (error);
}
static int
closefp(struct filedesc *fdp, int fd, struct file *fp, struct thread *td,
bool holdleaders, bool audit)
{
FILEDESC_XLOCK_ASSERT(fdp);
if (__predict_false(td->td_proc->p_fdtol != NULL)) {
return (closefp_hl(fdp, fd, fp, td, holdleaders, audit));
} else {
return (closefp_impl(fdp, fd, fp, td, audit));
}
}
/*
* Close a file descriptor.
*/
#ifndef _SYS_SYSPROTO_H_
struct close_args {
int fd;
};
#endif
/* ARGSUSED */
int
sys_close(struct thread *td, struct close_args *uap)
{
return (kern_close(td, uap->fd));
}
int
kern_close(struct thread *td, int fd)
{
struct filedesc *fdp;
struct file *fp;
fdp = td->td_proc->p_fd;
FILEDESC_XLOCK(fdp);
if ((fp = fget_locked(fdp, fd)) == NULL) {
FILEDESC_XUNLOCK(fdp);
return (EBADF);
}
fdfree(fdp, fd);
/* closefp() drops the FILEDESC lock for us. */
return (closefp(fdp, fd, fp, td, true, true));
}
int
kern_close_range(struct thread *td, u_int lowfd, u_int highfd)
{
struct filedesc *fdp;
const struct fdescenttbl *fdt;
struct file *fp;
int fd;
/*
* Check this prior to clamping; closefrom(3) with only fd 0, 1, and 2
* open should not be a usage error. From a close_range() perspective,
* close_range(3, ~0U, 0) in the same scenario should also likely not
* be a usage error as all fd above 3 are in-fact already closed.
*/
if (highfd < lowfd) {
return (EINVAL);
}
fdp = td->td_proc->p_fd;
FILEDESC_XLOCK(fdp);
fdt = atomic_load_ptr(&fdp->fd_files);
highfd = MIN(highfd, fdt->fdt_nfiles - 1);
fd = lowfd;
if (__predict_false(fd > highfd)) {
goto out_locked;
}
for (;;) {
fp = fdt->fdt_ofiles[fd].fde_file;
if (fp == NULL) {
if (fd == highfd)
goto out_locked;
} else {
fdfree(fdp, fd);
(void) closefp(fdp, fd, fp, td, true, true);
if (fd == highfd)
goto out_unlocked;
FILEDESC_XLOCK(fdp);
fdt = atomic_load_ptr(&fdp->fd_files);
}
fd++;
}
out_locked:
FILEDESC_XUNLOCK(fdp);
out_unlocked:
return (0);
}
#ifndef _SYS_SYSPROTO_H_
struct close_range_args {
u_int lowfd;
u_int highfd;
int flags;
};
#endif
int
sys_close_range(struct thread *td, struct close_range_args *uap)
{
AUDIT_ARG_FD(uap->lowfd);
AUDIT_ARG_CMD(uap->highfd);
AUDIT_ARG_FFLAGS(uap->flags);
/* No flags currently defined */
if (uap->flags != 0)
return (EINVAL);
return (kern_close_range(td, uap->lowfd, uap->highfd));
}
#ifdef COMPAT_FREEBSD12
/*
* Close open file descriptors.
*/
#ifndef _SYS_SYSPROTO_H_
struct freebsd12_closefrom_args {
int lowfd;
};
#endif
/* ARGSUSED */
int
freebsd12_closefrom(struct thread *td, struct freebsd12_closefrom_args *uap)
{
u_int lowfd;
AUDIT_ARG_FD(uap->lowfd);
/*
* Treat negative starting file descriptor values identical to
* closefrom(0) which closes all files.
*/
lowfd = MAX(0, uap->lowfd);
return (kern_close_range(td, lowfd, ~0U));
}
#endif /* COMPAT_FREEBSD12 */
#if defined(COMPAT_43)
/*
* Return status information about a file descriptor.
*/
#ifndef _SYS_SYSPROTO_H_
struct ofstat_args {
int fd;
struct ostat *sb;
};
#endif
/* ARGSUSED */
int
ofstat(struct thread *td, struct ofstat_args *uap)
{
struct ostat oub;
struct stat ub;
int error;
error = kern_fstat(td, uap->fd, &ub);
if (error == 0) {
cvtstat(&ub, &oub);
error = copyout(&oub, uap->sb, sizeof(oub));
}
return (error);
}
#endif /* COMPAT_43 */
#if defined(COMPAT_FREEBSD11)
int
freebsd11_fstat(struct thread *td, struct freebsd11_fstat_args *uap)
{
struct stat sb;
struct freebsd11_stat osb;
int error;
error = kern_fstat(td, uap->fd, &sb);
if (error != 0)
return (error);
error = freebsd11_cvtstat(&sb, &osb);
if (error == 0)
error = copyout(&osb, uap->sb, sizeof(osb));
return (error);
}
#endif /* COMPAT_FREEBSD11 */
/*
* Return status information about a file descriptor.
*/
#ifndef _SYS_SYSPROTO_H_
struct fstat_args {
int fd;
struct stat *sb;
};
#endif
/* ARGSUSED */
int
sys_fstat(struct thread *td, struct fstat_args *uap)
{
struct stat ub;
int error;
error = kern_fstat(td, uap->fd, &ub);
if (error == 0)
error = copyout(&ub, uap->sb, sizeof(ub));
return (error);
}
int
kern_fstat(struct thread *td, int fd, struct stat *sbp)
{
struct file *fp;
int error;
AUDIT_ARG_FD(fd);
error = fget(td, fd, &cap_fstat_rights, &fp);
if (__predict_false(error != 0))
return (error);
AUDIT_ARG_FILE(td->td_proc, fp);
error = fo_stat(fp, sbp, td->td_ucred, td);
fdrop(fp, td);
#ifdef __STAT_TIME_T_EXT
sbp->st_atim_ext = 0;
sbp->st_mtim_ext = 0;
sbp->st_ctim_ext = 0;
sbp->st_btim_ext = 0;
#endif
#ifdef KTRACE
if (KTRPOINT(td, KTR_STRUCT))
ktrstat_error(sbp, error);
#endif
return (error);
}
#if defined(COMPAT_FREEBSD11)
/*
* Return status information about a file descriptor.
*/
#ifndef _SYS_SYSPROTO_H_
struct freebsd11_nfstat_args {
int fd;
struct nstat *sb;
};
#endif
/* ARGSUSED */
int
freebsd11_nfstat(struct thread *td, struct freebsd11_nfstat_args *uap)
{
struct nstat nub;
struct stat ub;
int error;
error = kern_fstat(td, uap->fd, &ub);
if (error == 0) {
freebsd11_cvtnstat(&ub, &nub);
error = copyout(&nub, uap->sb, sizeof(nub));
}
return (error);
}
#endif /* COMPAT_FREEBSD11 */
/*
* Return pathconf information about a file descriptor.
*/
#ifndef _SYS_SYSPROTO_H_
struct fpathconf_args {
int fd;
int name;
};
#endif
/* ARGSUSED */
int
sys_fpathconf(struct thread *td, struct fpathconf_args *uap)
{
long value;
int error;
error = kern_fpathconf(td, uap->fd, uap->name, &value);
if (error == 0)
td->td_retval[0] = value;
return (error);
}
int
kern_fpathconf(struct thread *td, int fd, int name, long *valuep)
{
struct file *fp;
struct vnode *vp;
int error;
error = fget(td, fd, &cap_fpathconf_rights, &fp);
if (error != 0)
return (error);
if (name == _PC_ASYNC_IO) {
*valuep = _POSIX_ASYNCHRONOUS_IO;
goto out;
}
vp = fp->f_vnode;
if (vp != NULL) {
vn_lock(vp, LK_SHARED | LK_RETRY);
error = VOP_PATHCONF(vp, name, valuep);
VOP_UNLOCK(vp);
} else if (fp->f_type == DTYPE_PIPE || fp->f_type == DTYPE_SOCKET) {
if (name != _PC_PIPE_BUF) {
error = EINVAL;
} else {
*valuep = PIPE_BUF;
error = 0;
}
} else {
error = EOPNOTSUPP;
}
out:
fdrop(fp, td);
return (error);
}
/*
* Copy filecaps structure allocating memory for ioctls array if needed.
*
* The last parameter indicates whether the fdtable is locked. If it is not and
* ioctls are encountered, copying fails and the caller must lock the table.
*
* Note that if the table was not locked, the caller has to check the relevant
* sequence counter to determine whether the operation was successful.
*/
bool
filecaps_copy(const struct filecaps *src, struct filecaps *dst, bool locked)
{
size_t size;
if (src->fc_ioctls != NULL && !locked)
return (false);
memcpy(dst, src, sizeof(*src));
if (src->fc_ioctls == NULL)
return (true);
KASSERT(src->fc_nioctls > 0,
("fc_ioctls != NULL, but fc_nioctls=%hd", src->fc_nioctls));
size = sizeof(src->fc_ioctls[0]) * src->fc_nioctls;
dst->fc_ioctls = malloc(size, M_FILECAPS, M_WAITOK);
memcpy(dst->fc_ioctls, src->fc_ioctls, size);
return (true);
}
static u_long *
filecaps_copy_prep(const struct filecaps *src)
{
u_long *ioctls;
size_t size;
if (__predict_true(src->fc_ioctls == NULL))
return (NULL);
KASSERT(src->fc_nioctls > 0,
("fc_ioctls != NULL, but fc_nioctls=%hd", src->fc_nioctls));
size = sizeof(src->fc_ioctls[0]) * src->fc_nioctls;
ioctls = malloc(size, M_FILECAPS, M_WAITOK);
return (ioctls);
}
static void
filecaps_copy_finish(const struct filecaps *src, struct filecaps *dst,
u_long *ioctls)
{
size_t size;
*dst = *src;
if (__predict_true(src->fc_ioctls == NULL)) {
MPASS(ioctls == NULL);
return;
}
size = sizeof(src->fc_ioctls[0]) * src->fc_nioctls;
dst->fc_ioctls = ioctls;
bcopy(src->fc_ioctls, dst->fc_ioctls, size);
}
/*
* Move filecaps structure to the new place and clear the old place.
*/
void
filecaps_move(struct filecaps *src, struct filecaps *dst)
{
*dst = *src;
bzero(src, sizeof(*src));
}
/*
* Fill the given filecaps structure with full rights.
*/
static void
filecaps_fill(struct filecaps *fcaps)
{
CAP_ALL(&fcaps->fc_rights);
fcaps->fc_ioctls = NULL;
fcaps->fc_nioctls = -1;
fcaps->fc_fcntls = CAP_FCNTL_ALL;
}
/*
* Free memory allocated within filecaps structure.
*/
void
filecaps_free(struct filecaps *fcaps)
{
free(fcaps->fc_ioctls, M_FILECAPS);
bzero(fcaps, sizeof(*fcaps));
}
static u_long *
filecaps_free_prep(struct filecaps *fcaps)
{
u_long *ioctls;
ioctls = fcaps->fc_ioctls;
bzero(fcaps, sizeof(*fcaps));
return (ioctls);
}
static void
filecaps_free_finish(u_long *ioctls)
{
free(ioctls, M_FILECAPS);
}
/*
* Validate the given filecaps structure.
*/
static void
filecaps_validate(const struct filecaps *fcaps, const char *func)
{
KASSERT(cap_rights_is_valid(&fcaps->fc_rights),
("%s: invalid rights", func));
KASSERT((fcaps->fc_fcntls & ~CAP_FCNTL_ALL) == 0,
("%s: invalid fcntls", func));
KASSERT(fcaps->fc_fcntls == 0 ||
cap_rights_is_set(&fcaps->fc_rights, CAP_FCNTL),
("%s: fcntls without CAP_FCNTL", func));
KASSERT(fcaps->fc_ioctls != NULL ? fcaps->fc_nioctls > 0 :
(fcaps->fc_nioctls == -1 || fcaps->fc_nioctls == 0),
("%s: invalid ioctls", func));
KASSERT(fcaps->fc_nioctls == 0 ||
cap_rights_is_set(&fcaps->fc_rights, CAP_IOCTL),
("%s: ioctls without CAP_IOCTL", func));
}
static void
fdgrowtable_exp(struct filedesc *fdp, int nfd)
{
int nfd1;
FILEDESC_XLOCK_ASSERT(fdp);
nfd1 = fdp->fd_nfiles * 2;
if (nfd1 < nfd)
nfd1 = nfd;
fdgrowtable(fdp, nfd1);
}
/*
* Grow the file table to accommodate (at least) nfd descriptors.
*/
static void
fdgrowtable(struct filedesc *fdp, int nfd)
{
struct filedesc0 *fdp0;
struct freetable *ft;
struct fdescenttbl *ntable;
struct fdescenttbl *otable;
int nnfiles, onfiles;
NDSLOTTYPE *nmap, *omap;
KASSERT(fdp->fd_nfiles > 0, ("zero-length file table"));
/* save old values */
onfiles = fdp->fd_nfiles;
otable = fdp->fd_files;
omap = fdp->fd_map;
/* compute the size of the new table */
nnfiles = NDSLOTS(nfd) * NDENTRIES; /* round up */
if (nnfiles <= onfiles)
/* the table is already large enough */
return;
/*
* Allocate a new table. We need enough space for the number of
* entries, file entries themselves and the struct freetable we will use
* when we decommission the table and place it on the freelist.
* We place the struct freetable in the middle so we don't have
* to worry about padding.
*/
ntable = malloc(offsetof(struct fdescenttbl, fdt_ofiles) +
nnfiles * sizeof(ntable->fdt_ofiles[0]) +
sizeof(struct freetable),
M_FILEDESC, M_ZERO | M_WAITOK);
/* copy the old data */
ntable->fdt_nfiles = nnfiles;
memcpy(ntable->fdt_ofiles, otable->fdt_ofiles,
onfiles * sizeof(ntable->fdt_ofiles[0]));
/*
* Allocate a new map only if the old is not large enough. It will
* grow at a slower rate than the table as it can map more
* entries than the table can hold.
*/
if (NDSLOTS(nnfiles) > NDSLOTS(onfiles)) {
nmap = malloc(NDSLOTS(nnfiles) * NDSLOTSIZE, M_FILEDESC,
M_ZERO | M_WAITOK);
/* copy over the old data and update the pointer */
memcpy(nmap, omap, NDSLOTS(onfiles) * sizeof(*omap));
fdp->fd_map = nmap;
}
/*
* Make sure that ntable is correctly initialized before we replace
* fd_files poiner. Otherwise fget_unlocked() may see inconsistent
* data.
*/
atomic_store_rel_ptr((volatile void *)&fdp->fd_files, (uintptr_t)ntable);
/*
* Free the old file table when not shared by other threads or processes.
* The old file table is considered to be shared when either are true:
* - The process has more than one thread.
* - The file descriptor table has been shared via fdshare().
*
* When shared, the old file table will be placed on a freelist
* which will be processed when the struct filedesc is released.
*
* Note that if onfiles == NDFILE, we're dealing with the original
* static allocation contained within (struct filedesc0 *)fdp,
* which must not be freed.
*/
if (onfiles > NDFILE) {
/*
* Note we may be called here from fdinit while allocating a
* table for a new process in which case ->p_fd points
* elsewhere.
*/
if (curproc->p_fd != fdp || FILEDESC_IS_ONLY_USER(fdp)) {
free(otable, M_FILEDESC);
} else {
ft = (struct freetable *)&otable->fdt_ofiles[onfiles];
fdp0 = (struct filedesc0 *)fdp;
ft->ft_table = otable;
SLIST_INSERT_HEAD(&fdp0->fd_free, ft, ft_next);
}
}
/*
* The map does not have the same possibility of threads still
* holding references to it. So always free it as long as it
* does not reference the original static allocation.
*/
if (NDSLOTS(onfiles) > NDSLOTS(NDFILE))
free(omap, M_FILEDESC);
}
/*
* Allocate a file descriptor for the process.
*/
int
fdalloc(struct thread *td, int minfd, int *result)
{
struct proc *p = td->td_proc;
struct filedesc *fdp = p->p_fd;
int fd, maxfd, allocfd;
#ifdef RACCT
int error;
#endif
FILEDESC_XLOCK_ASSERT(fdp);
if (fdp->fd_freefile > minfd)
minfd = fdp->fd_freefile;
maxfd = getmaxfd(td);
/*
* Search the bitmap for a free descriptor starting at minfd.
* If none is found, grow the file table.
*/
fd = fd_first_free(fdp, minfd, fdp->fd_nfiles);
if (__predict_false(fd >= maxfd))
return (EMFILE);
if (__predict_false(fd >= fdp->fd_nfiles)) {
allocfd = min(fd * 2, maxfd);
#ifdef RACCT
if (RACCT_ENABLED()) {
error = racct_set_unlocked(p, RACCT_NOFILE, allocfd);
if (error != 0)
return (EMFILE);
}
#endif
/*
* fd is already equal to first free descriptor >= minfd, so
* we only need to grow the table and we are done.
*/
fdgrowtable_exp(fdp, allocfd);
}
/*
* Perform some sanity checks, then mark the file descriptor as
* used and return it to the caller.
*/
KASSERT(fd >= 0 && fd < min(maxfd, fdp->fd_nfiles),
("invalid descriptor %d", fd));
KASSERT(!fdisused(fdp, fd),
("fd_first_free() returned non-free descriptor"));
KASSERT(fdp->fd_ofiles[fd].fde_file == NULL,
("file descriptor isn't free"));
fdused(fdp, fd);
*result = fd;
return (0);
}
/*
* Allocate n file descriptors for the process.
*/
int
fdallocn(struct thread *td, int minfd, int *fds, int n)
{
struct proc *p = td->td_proc;
struct filedesc *fdp = p->p_fd;
int i;
FILEDESC_XLOCK_ASSERT(fdp);
for (i = 0; i < n; i++)
if (fdalloc(td, 0, &fds[i]) != 0)
break;
if (i < n) {
for (i--; i >= 0; i--)
fdunused(fdp, fds[i]);
return (EMFILE);
}
return (0);
}
/*
* Create a new open file structure and allocate a file descriptor for the
* process that refers to it. We add one reference to the file for the
* descriptor table and one reference for resultfp. This is to prevent us
* being preempted and the entry in the descriptor table closed after we
* release the FILEDESC lock.
*/
int
falloc_caps(struct thread *td, struct file **resultfp, int *resultfd, int flags,
struct filecaps *fcaps)
{
struct file *fp;
int error, fd;
MPASS(resultfp != NULL);
MPASS(resultfd != NULL);
error = _falloc_noinstall(td, &fp, 2);
if (__predict_false(error != 0)) {
return (error);
}
error = finstall_refed(td, fp, &fd, flags, fcaps);
if (__predict_false(error != 0)) {
falloc_abort(td, fp);
return (error);
}
*resultfp = fp;
*resultfd = fd;
return (0);
}
/*
* Create a new open file structure without allocating a file descriptor.
*/
int
_falloc_noinstall(struct thread *td, struct file **resultfp, u_int n)
{
struct file *fp;
int maxuserfiles = maxfiles - (maxfiles / 20);
int openfiles_new;
static struct timeval lastfail;
static int curfail;
KASSERT(resultfp != NULL, ("%s: resultfp == NULL", __func__));
MPASS(n > 0);
openfiles_new = atomic_fetchadd_int(&openfiles, 1) + 1;
if ((openfiles_new >= maxuserfiles &&
priv_check(td, PRIV_MAXFILES) != 0) ||
openfiles_new >= maxfiles) {
atomic_subtract_int(&openfiles, 1);
if (ppsratecheck(&lastfail, &curfail, 1)) {
printf("kern.maxfiles limit exceeded by uid %i, (%s) "
"please see tuning(7).\n", td->td_ucred->cr_ruid, td->td_proc->p_comm);
}
return (ENFILE);
}
fp = uma_zalloc(file_zone, M_WAITOK);
bzero(fp, sizeof(*fp));
refcount_init(&fp->f_count, n);
fp->f_cred = crhold(td->td_ucred);
fp->f_ops = &badfileops;
*resultfp = fp;
return (0);
}
void
falloc_abort(struct thread *td, struct file *fp)
{
/*
* For assertion purposes.
*/
refcount_init(&fp->f_count, 0);
_fdrop(fp, td);
}
/*
* Install a file in a file descriptor table.
*/
void
_finstall(struct filedesc *fdp, struct file *fp, int fd, int flags,
struct filecaps *fcaps)
{
struct filedescent *fde;
MPASS(fp != NULL);
if (fcaps != NULL)
filecaps_validate(fcaps, __func__);
FILEDESC_XLOCK_ASSERT(fdp);
fde = &fdp->fd_ofiles[fd];
#ifdef CAPABILITIES
seqc_write_begin(&fde->fde_seqc);
#endif
fde->fde_file = fp;
fde->fde_flags = (flags & O_CLOEXEC) != 0 ? UF_EXCLOSE : 0;
if (fcaps != NULL)
filecaps_move(fcaps, &fde->fde_caps);
else
filecaps_fill(&fde->fde_caps);
#ifdef CAPABILITIES
seqc_write_end(&fde->fde_seqc);
#endif
}
int
finstall_refed(struct thread *td, struct file *fp, int *fd, int flags,
struct filecaps *fcaps)
{
struct filedesc *fdp = td->td_proc->p_fd;
int error;
MPASS(fd != NULL);
FILEDESC_XLOCK(fdp);
error = fdalloc(td, 0, fd);
if (__predict_true(error == 0)) {
_finstall(fdp, fp, *fd, flags, fcaps);
}
FILEDESC_XUNLOCK(fdp);
return (error);
}
int
finstall(struct thread *td, struct file *fp, int *fd, int flags,
struct filecaps *fcaps)
{
int error;
MPASS(fd != NULL);
if (!fhold(fp))
return (EBADF);
error = finstall_refed(td, fp, fd, flags, fcaps);
if (__predict_false(error != 0)) {
fdrop(fp, td);
}
return (error);
}
/*
* Build a new filedesc structure from another.
*
* If fdp is not NULL, return with it shared locked.
*/
struct filedesc *
fdinit(struct filedesc *fdp, bool prepfiles, int *lastfile)
{
struct filedesc0 *newfdp0;
struct filedesc *newfdp;
if (prepfiles)
MPASS(lastfile != NULL);
else
MPASS(lastfile == NULL);
newfdp0 = uma_zalloc(filedesc0_zone, M_WAITOK | M_ZERO);
newfdp = &newfdp0->fd_fd;
/* Create the file descriptor table. */
FILEDESC_LOCK_INIT(newfdp);
refcount_init(&newfdp->fd_refcnt, 1);
refcount_init(&newfdp->fd_holdcnt, 1);
newfdp->fd_map = newfdp0->fd_dmap;
newfdp->fd_files = (struct fdescenttbl *)&newfdp0->fd_dfiles;
newfdp->fd_files->fdt_nfiles = NDFILE;
if (fdp == NULL)
return (newfdp);
FILEDESC_SLOCK(fdp);
if (!prepfiles) {
FILEDESC_SUNLOCK(fdp);
return (newfdp);
}
for (;;) {
*lastfile = fdlastfile(fdp);
if (*lastfile < newfdp->fd_nfiles)
break;
FILEDESC_SUNLOCK(fdp);
fdgrowtable(newfdp, *lastfile + 1);
FILEDESC_SLOCK(fdp);
}
return (newfdp);
}
/*
* Build a pwddesc structure from another.
* Copy the current, root, and jail root vnode references.
*
* If pdp is not NULL, return with it shared locked.
*/
struct pwddesc *
pdinit(struct pwddesc *pdp, bool keeplock)
{
struct pwddesc *newpdp;
struct pwd *newpwd;
newpdp = malloc(sizeof(*newpdp), M_PWDDESC, M_WAITOK | M_ZERO);
PWDDESC_LOCK_INIT(newpdp);
refcount_init(&newpdp->pd_refcount, 1);
newpdp->pd_cmask = CMASK;
if (pdp == NULL) {
newpwd = pwd_alloc();
smr_serialized_store(&newpdp->pd_pwd, newpwd, true);
return (newpdp);
}
PWDDESC_XLOCK(pdp);
newpwd = pwd_hold_pwddesc(pdp);
smr_serialized_store(&newpdp->pd_pwd, newpwd, true);
if (!keeplock)
PWDDESC_XUNLOCK(pdp);
return (newpdp);
}
/*
* Hold either filedesc or pwddesc of the passed process.
*
* The process lock is used to synchronize against the target exiting and
* freeing the data.
*
* Clearing can be ilustrated in 3 steps:
* 1. set the pointer to NULL. Either routine can race against it, hence
* atomic_load_ptr.
* 2. observe the process lock as not taken. Until then fdhold/pdhold can
* race to either still see the pointer or find NULL. It is still safe to
* grab a reference as clearing is stalled.
* 3. after the lock is observed as not taken, any fdhold/pdhold calls are
* guaranteed to see NULL, making it safe to finish clearing
*/
static struct filedesc *
fdhold(struct proc *p)
{
struct filedesc *fdp;
PROC_LOCK_ASSERT(p, MA_OWNED);
fdp = atomic_load_ptr(&p->p_fd);
if (fdp != NULL)
refcount_acquire(&fdp->fd_holdcnt);
return (fdp);
}
static struct pwddesc *
pdhold(struct proc *p)
{
struct pwddesc *pdp;
PROC_LOCK_ASSERT(p, MA_OWNED);
pdp = atomic_load_ptr(&p->p_pd);
if (pdp != NULL)
refcount_acquire(&pdp->pd_refcount);
return (pdp);
}
static void
fddrop(struct filedesc *fdp)
{
if (refcount_load(&fdp->fd_holdcnt) > 1) {
if (refcount_release(&fdp->fd_holdcnt) == 0)
return;
}
FILEDESC_LOCK_DESTROY(fdp);
uma_zfree(filedesc0_zone, fdp);
}
static void
pddrop(struct pwddesc *pdp)
{
struct pwd *pwd;
if (refcount_release_if_not_last(&pdp->pd_refcount))
return;
PWDDESC_XLOCK(pdp);
if (refcount_release(&pdp->pd_refcount) == 0) {
PWDDESC_XUNLOCK(pdp);
return;
}
pwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
pwd_set(pdp, NULL);
PWDDESC_XUNLOCK(pdp);
pwd_drop(pwd);
PWDDESC_LOCK_DESTROY(pdp);
free(pdp, M_PWDDESC);
}
/*
* Share a filedesc structure.
*/
struct filedesc *
fdshare(struct filedesc *fdp)
{
refcount_acquire(&fdp->fd_refcnt);
return (fdp);
}
/*
* Share a pwddesc structure.
*/
struct pwddesc *
pdshare(struct pwddesc *pdp)
{
refcount_acquire(&pdp->pd_refcount);
return (pdp);
}
/*
* Unshare a filedesc structure, if necessary by making a copy
*/
void
fdunshare(struct thread *td)
{
struct filedesc *tmp;
struct proc *p = td->td_proc;
if (refcount_load(&p->p_fd->fd_refcnt) == 1)
return;
tmp = fdcopy(p->p_fd);
fdescfree(td);
p->p_fd = tmp;
}
/*
* Unshare a pwddesc structure.
*/
void
pdunshare(struct thread *td)
{
struct pwddesc *pdp;
struct proc *p;
p = td->td_proc;
/* Not shared. */
if (p->p_pd->pd_refcount == 1)
return;
pdp = pdcopy(p->p_pd);
pdescfree(td);
p->p_pd = pdp;
}
void
fdinstall_remapped(struct thread *td, struct filedesc *fdp)
{
fdescfree(td);
td->td_proc->p_fd = fdp;
}
/*
* Copy a filedesc structure. A NULL pointer in returns a NULL reference,
* this is to ease callers, not catch errors.
*/
struct filedesc *
fdcopy(struct filedesc *fdp)
{
struct filedesc *newfdp;
struct filedescent *nfde, *ofde;
int i, lastfile;
MPASS(fdp != NULL);
newfdp = fdinit(fdp, true, &lastfile);
/* copy all passable descriptors (i.e. not kqueue) */
newfdp->fd_freefile = -1;
for (i = 0; i <= lastfile; ++i) {
ofde = &fdp->fd_ofiles[i];
if (ofde->fde_file == NULL ||
(ofde->fde_file->f_ops->fo_flags & DFLAG_PASSABLE) == 0 ||
!fhold(ofde->fde_file)) {
if (newfdp->fd_freefile == -1)
newfdp->fd_freefile = i;
continue;
}
nfde = &newfdp->fd_ofiles[i];
*nfde = *ofde;
filecaps_copy(&ofde->fde_caps, &nfde->fde_caps, true);
fdused_init(newfdp, i);
}
if (newfdp->fd_freefile == -1)
newfdp->fd_freefile = i;
FILEDESC_SUNLOCK(fdp);
return (newfdp);
}
/*
* Copy a pwddesc structure.
*/
struct pwddesc *
pdcopy(struct pwddesc *pdp)
{
struct pwddesc *newpdp;
MPASS(pdp != NULL);
newpdp = pdinit(pdp, true);
newpdp->pd_cmask = pdp->pd_cmask;
PWDDESC_XUNLOCK(pdp);
return (newpdp);
}
/*
* Copies a filedesc structure, while remapping all file descriptors
* stored inside using a translation table.
*
* File descriptors are copied over to the new file descriptor table,
* regardless of whether the close-on-exec flag is set.
*/
int
fdcopy_remapped(struct filedesc *fdp, const int *fds, size_t nfds,
struct filedesc **ret)
{
struct filedesc *newfdp;
struct filedescent *nfde, *ofde;
int error, i, lastfile;
MPASS(fdp != NULL);
newfdp = fdinit(fdp, true, &lastfile);
if (nfds > lastfile + 1) {
/* New table cannot be larger than the old one. */
error = E2BIG;
goto bad;
}
/* Copy all passable descriptors (i.e. not kqueue). */
newfdp->fd_freefile = nfds;
for (i = 0; i < nfds; ++i) {
if (fds[i] < 0 || fds[i] > lastfile) {
/* File descriptor out of bounds. */
error = EBADF;
goto bad;
}
ofde = &fdp->fd_ofiles[fds[i]];
if (ofde->fde_file == NULL) {
/* Unused file descriptor. */
error = EBADF;
goto bad;
}
if ((ofde->fde_file->f_ops->fo_flags & DFLAG_PASSABLE) == 0) {
/* File descriptor cannot be passed. */
error = EINVAL;
goto bad;
}
if (!fhold(ofde->fde_file)) {
error = EBADF;
goto bad;
}
nfde = &newfdp->fd_ofiles[i];
*nfde = *ofde;
filecaps_copy(&ofde->fde_caps, &nfde->fde_caps, true);
fdused_init(newfdp, i);
}
FILEDESC_SUNLOCK(fdp);
*ret = newfdp;
return (0);
bad:
FILEDESC_SUNLOCK(fdp);
fdescfree_remapped(newfdp);
return (error);
}
/*
* Clear POSIX style locks. This is only used when fdp looses a reference (i.e.
* one of processes using it exits) and the table used to be shared.
*/
static void
fdclearlocks(struct thread *td)
{
struct filedesc *fdp;
struct filedesc_to_leader *fdtol;
struct flock lf;
struct file *fp;
struct proc *p;
struct vnode *vp;
int i, lastfile;
p = td->td_proc;
fdp = p->p_fd;
fdtol = p->p_fdtol;
MPASS(fdtol != NULL);
FILEDESC_XLOCK(fdp);
KASSERT(fdtol->fdl_refcount > 0,
("filedesc_to_refcount botch: fdl_refcount=%d",
fdtol->fdl_refcount));
if (fdtol->fdl_refcount == 1 &&
(p->p_leader->p_flag & P_ADVLOCK) != 0) {
lastfile = fdlastfile(fdp);
for (i = 0; i <= lastfile; i++) {
fp = fdp->fd_ofiles[i].fde_file;
if (fp == NULL || fp->f_type != DTYPE_VNODE ||
!fhold(fp))
continue;
FILEDESC_XUNLOCK(fdp);
lf.l_whence = SEEK_SET;
lf.l_start = 0;
lf.l_len = 0;
lf.l_type = F_UNLCK;
vp = fp->f_vnode;
(void) VOP_ADVLOCK(vp,
(caddr_t)p->p_leader, F_UNLCK,
&lf, F_POSIX);
FILEDESC_XLOCK(fdp);
fdrop(fp, td);
}
}
retry:
if (fdtol->fdl_refcount == 1) {
if (fdp->fd_holdleaderscount > 0 &&
(p->p_leader->p_flag & P_ADVLOCK) != 0) {
/*
* close() or kern_dup() has cleared a reference
* in a shared file descriptor table.
*/
fdp->fd_holdleaderswakeup = 1;
sx_sleep(&fdp->fd_holdleaderscount,
FILEDESC_LOCK(fdp), PLOCK, "fdlhold", 0);
goto retry;
}
if (fdtol->fdl_holdcount > 0) {
/*
* Ensure that fdtol->fdl_leader remains
* valid in closef().
*/
fdtol->fdl_wakeup = 1;
sx_sleep(fdtol, FILEDESC_LOCK(fdp), PLOCK,
"fdlhold", 0);
goto retry;
}
}
fdtol->fdl_refcount--;
if (fdtol->fdl_refcount == 0 &&
fdtol->fdl_holdcount == 0) {
fdtol->fdl_next->fdl_prev = fdtol->fdl_prev;
fdtol->fdl_prev->fdl_next = fdtol->fdl_next;
} else
fdtol = NULL;
p->p_fdtol = NULL;
FILEDESC_XUNLOCK(fdp);
if (fdtol != NULL)
free(fdtol, M_FILEDESC_TO_LEADER);
}
/*
* Release a filedesc structure.
*/
static void
fdescfree_fds(struct thread *td, struct filedesc *fdp, bool needclose)
{
struct filedesc0 *fdp0;
struct freetable *ft, *tft;
struct filedescent *fde;
struct file *fp;
int i, lastfile;
KASSERT(refcount_load(&fdp->fd_refcnt) == 0,
("%s: fd table %p carries references", __func__, fdp));
/*
* Serialize with threads iterating over the table, if any.
*/
if (refcount_load(&fdp->fd_holdcnt) > 1) {
FILEDESC_XLOCK(fdp);
FILEDESC_XUNLOCK(fdp);
}
lastfile = fdlastfile_single(fdp);
for (i = 0; i <= lastfile; i++) {
fde = &fdp->fd_ofiles[i];
fp = fde->fde_file;
if (fp != NULL) {
fdefree_last(fde);
if (needclose)
(void) closef(fp, td);
else
fdrop(fp, td);
}
}
if (NDSLOTS(fdp->fd_nfiles) > NDSLOTS(NDFILE))
free(fdp->fd_map, M_FILEDESC);
if (fdp->fd_nfiles > NDFILE)
free(fdp->fd_files, M_FILEDESC);
fdp0 = (struct filedesc0 *)fdp;
SLIST_FOREACH_SAFE(ft, &fdp0->fd_free, ft_next, tft)
free(ft->ft_table, M_FILEDESC);
fddrop(fdp);
}
void
fdescfree(struct thread *td)
{
struct proc *p;
struct filedesc *fdp;
p = td->td_proc;
fdp = p->p_fd;
MPASS(fdp != NULL);
#ifdef RACCT
if (RACCT_ENABLED())
racct_set_unlocked(p, RACCT_NOFILE, 0);
#endif
if (p->p_fdtol != NULL)
fdclearlocks(td);
/*
* Check fdhold for an explanation.
*/
atomic_store_ptr(&p->p_fd, NULL);
atomic_thread_fence_seq_cst();
PROC_WAIT_UNLOCKED(p);
if (refcount_release(&fdp->fd_refcnt) == 0)
return;
fdescfree_fds(td, fdp, 1);
}
void
pdescfree(struct thread *td)
{
struct proc *p;
struct pwddesc *pdp;
p = td->td_proc;
pdp = p->p_pd;
MPASS(pdp != NULL);
/*
* Check pdhold for an explanation.
*/
atomic_store_ptr(&p->p_pd, NULL);
atomic_thread_fence_seq_cst();
PROC_WAIT_UNLOCKED(p);
pddrop(pdp);
}
void
fdescfree_remapped(struct filedesc *fdp)
{
#ifdef INVARIANTS
/* fdescfree_fds() asserts that fd_refcnt == 0. */
if (!refcount_release(&fdp->fd_refcnt))
panic("%s: fd table %p has extra references", __func__, fdp);
#endif
fdescfree_fds(curthread, fdp, 0);
}
/*
* For setugid programs, we don't want to people to use that setugidness
* to generate error messages which write to a file which otherwise would
* otherwise be off-limits to the process. We check for filesystems where
* the vnode can change out from under us after execve (like [lin]procfs).
*
* Since fdsetugidsafety calls this only for fd 0, 1 and 2, this check is
* sufficient. We also don't check for setugidness since we know we are.
*/
static bool
is_unsafe(struct file *fp)
{
struct vnode *vp;
if (fp->f_type != DTYPE_VNODE)
return (false);
vp = fp->f_vnode;
return ((vp->v_vflag & VV_PROCDEP) != 0);
}
/*
* Make this setguid thing safe, if at all possible.
*/
void
fdsetugidsafety(struct thread *td)
{
struct filedesc *fdp;
struct file *fp;
int i;
fdp = td->td_proc->p_fd;
KASSERT(refcount_load(&fdp->fd_refcnt) == 1,
("the fdtable should not be shared"));
MPASS(fdp->fd_nfiles >= 3);
for (i = 0; i <= 2; i++) {
fp = fdp->fd_ofiles[i].fde_file;
if (fp != NULL && is_unsafe(fp)) {
FILEDESC_XLOCK(fdp);
knote_fdclose(td, i);
/*
* NULL-out descriptor prior to close to avoid
* a race while close blocks.
*/
fdfree(fdp, i);
FILEDESC_XUNLOCK(fdp);
(void) closef(fp, td);
}
}
}
/*
* If a specific file object occupies a specific file descriptor, close the
* file descriptor entry and drop a reference on the file object. This is a
* convenience function to handle a subsequent error in a function that calls
* falloc() that handles the race that another thread might have closed the
* file descriptor out from under the thread creating the file object.
*/
void
fdclose(struct thread *td, struct file *fp, int idx)
{
struct filedesc *fdp = td->td_proc->p_fd;
FILEDESC_XLOCK(fdp);
if (fdp->fd_ofiles[idx].fde_file == fp) {
fdfree(fdp, idx);
FILEDESC_XUNLOCK(fdp);
fdrop(fp, td);
} else
FILEDESC_XUNLOCK(fdp);
}
/*
* Close any files on exec?
*/
void
fdcloseexec(struct thread *td)
{
struct filedesc *fdp;
struct filedescent *fde;
struct file *fp;
int i, lastfile;
fdp = td->td_proc->p_fd;
KASSERT(refcount_load(&fdp->fd_refcnt) == 1,
("the fdtable should not be shared"));
lastfile = fdlastfile_single(fdp);
for (i = 0; i <= lastfile; i++) {
fde = &fdp->fd_ofiles[i];
fp = fde->fde_file;
if (fp != NULL && (fp->f_type == DTYPE_MQUEUE ||
(fde->fde_flags & UF_EXCLOSE))) {
FILEDESC_XLOCK(fdp);
fdfree(fdp, i);
(void) closefp(fdp, i, fp, td, false, false);
FILEDESC_UNLOCK_ASSERT(fdp);
}
}
}
/*
* It is unsafe for set[ug]id processes to be started with file
* descriptors 0..2 closed, as these descriptors are given implicit
* significance in the Standard C library. fdcheckstd() will create a
* descriptor referencing /dev/null for each of stdin, stdout, and
* stderr that is not already open.
*/
int
fdcheckstd(struct thread *td)
{
struct filedesc *fdp;
register_t save;
int i, error, devnull;
fdp = td->td_proc->p_fd;
KASSERT(refcount_load(&fdp->fd_refcnt) == 1,
("the fdtable should not be shared"));
MPASS(fdp->fd_nfiles >= 3);
devnull = -1;
for (i = 0; i <= 2; i++) {
if (fdp->fd_ofiles[i].fde_file != NULL)
continue;
save = td->td_retval[0];
if (devnull != -1) {
error = kern_dup(td, FDDUP_FIXED, 0, devnull, i);
} else {
error = kern_openat(td, AT_FDCWD, "/dev/null",
UIO_SYSSPACE, O_RDWR, 0);
if (error == 0) {
devnull = td->td_retval[0];
KASSERT(devnull == i, ("we didn't get our fd"));
}
}
td->td_retval[0] = save;
if (error != 0)
return (error);
}
return (0);
}
/*
* Internal form of close. Decrement reference count on file structure.
* Note: td may be NULL when closing a file that was being passed in a
* message.
*/
int
closef(struct file *fp, struct thread *td)
{
struct vnode *vp;
struct flock lf;
struct filedesc_to_leader *fdtol;
struct filedesc *fdp;
MPASS(td != NULL);
/*
* POSIX record locking dictates that any close releases ALL
* locks owned by this process. This is handled by setting
* a flag in the unlock to free ONLY locks obeying POSIX
* semantics, and not to free BSD-style file locks.
* If the descriptor was in a message, POSIX-style locks
* aren't passed with the descriptor, and the thread pointer
* will be NULL. Callers should be careful only to pass a
* NULL thread pointer when there really is no owning
* context that might have locks, or the locks will be
* leaked.
*/
if (fp->f_type == DTYPE_VNODE) {
vp = fp->f_vnode;
if ((td->td_proc->p_leader->p_flag & P_ADVLOCK) != 0) {
lf.l_whence = SEEK_SET;
lf.l_start = 0;
lf.l_len = 0;
lf.l_type = F_UNLCK;
(void) VOP_ADVLOCK(vp, (caddr_t)td->td_proc->p_leader,
F_UNLCK, &lf, F_POSIX);
}
fdtol = td->td_proc->p_fdtol;
if (fdtol != NULL) {
/*
* Handle special case where file descriptor table is
* shared between multiple process leaders.
*/
fdp = td->td_proc->p_fd;
FILEDESC_XLOCK(fdp);
for (fdtol = fdtol->fdl_next;
fdtol != td->td_proc->p_fdtol;
fdtol = fdtol->fdl_next) {
if ((fdtol->fdl_leader->p_flag &
P_ADVLOCK) == 0)
continue;
fdtol->fdl_holdcount++;
FILEDESC_XUNLOCK(fdp);
lf.l_whence = SEEK_SET;
lf.l_start = 0;
lf.l_len = 0;
lf.l_type = F_UNLCK;
vp = fp->f_vnode;
(void) VOP_ADVLOCK(vp,
(caddr_t)fdtol->fdl_leader, F_UNLCK, &lf,
F_POSIX);
FILEDESC_XLOCK(fdp);
fdtol->fdl_holdcount--;
if (fdtol->fdl_holdcount == 0 &&
fdtol->fdl_wakeup != 0) {
fdtol->fdl_wakeup = 0;
wakeup(fdtol);
}
}
FILEDESC_XUNLOCK(fdp);
}
}
return (fdrop_close(fp, td));
}
/*
* Hack for file descriptor passing code.
*/
void
closef_nothread(struct file *fp)
{
fdrop(fp, NULL);
}
/*
* Initialize the file pointer with the specified properties.
*
* The ops are set with release semantics to be certain that the flags, type,
* and data are visible when ops is. This is to prevent ops methods from being
* called with bad data.
*/
void
finit(struct file *fp, u_int flag, short type, void *data, struct fileops *ops)
{
fp->f_data = data;
fp->f_flag = flag;
fp->f_type = type;
atomic_store_rel_ptr((volatile uintptr_t *)&fp->f_ops, (uintptr_t)ops);
}
void
finit_vnode(struct file *fp, u_int flag, void *data, struct fileops *ops)
{
fp->f_seqcount[UIO_READ] = 1;
fp->f_seqcount[UIO_WRITE] = 1;
finit(fp, (flag & FMASK) | (fp->f_flag & FHASLOCK), DTYPE_VNODE,
data, ops);
}
int
fget_cap_locked(struct filedesc *fdp, int fd, cap_rights_t *needrightsp,
struct file **fpp, struct filecaps *havecapsp)
{
struct filedescent *fde;
int error;
FILEDESC_LOCK_ASSERT(fdp);
fde = fdeget_locked(fdp, fd);
if (fde == NULL) {
error = EBADF;
goto out;
}
#ifdef CAPABILITIES
error = cap_check(cap_rights_fde_inline(fde), needrightsp);
if (error != 0)
goto out;
#endif
if (havecapsp != NULL)
filecaps_copy(&fde->fde_caps, havecapsp, true);
*fpp = fde->fde_file;
error = 0;
out:
return (error);
}
int
fget_cap(struct thread *td, int fd, cap_rights_t *needrightsp,
struct file **fpp, struct filecaps *havecapsp)
{
struct filedesc *fdp = td->td_proc->p_fd;
int error;
#ifndef CAPABILITIES
error = fget_unlocked(fdp, fd, needrightsp, fpp);
if (havecapsp != NULL && error == 0)
filecaps_fill(havecapsp);
#else
struct file *fp;
seqc_t seq;
*fpp = NULL;
for (;;) {
error = fget_unlocked_seq(fdp, fd, needrightsp, &fp, &seq);
if (error != 0)
return (error);
if (havecapsp != NULL) {
if (!filecaps_copy(&fdp->fd_ofiles[fd].fde_caps,
havecapsp, false)) {
fdrop(fp, td);
goto get_locked;
}
}
if (!fd_modified(fdp, fd, seq))
break;
fdrop(fp, td);
}
*fpp = fp;
return (0);
get_locked:
FILEDESC_SLOCK(fdp);
error = fget_cap_locked(fdp, fd, needrightsp, fpp, havecapsp);
if (error == 0 && !fhold(*fpp))
error = EBADF;
FILEDESC_SUNLOCK(fdp);
#endif
return (error);
}
#ifdef CAPABILITIES
int
fgetvp_lookup_smr(int fd, struct nameidata *ndp, struct vnode **vpp, bool *fsearch)
{
const struct filedescent *fde;
const struct fdescenttbl *fdt;
struct filedesc *fdp;
struct file *fp;
struct vnode *vp;
const cap_rights_t *haverights;
cap_rights_t rights;
seqc_t seq;
VFS_SMR_ASSERT_ENTERED();
rights = *ndp->ni_rightsneeded;
cap_rights_set_one(&rights, CAP_LOOKUP);
fdp = curproc->p_fd;
fdt = fdp->fd_files;
if (__predict_false((u_int)fd >= fdt->fdt_nfiles))
return (EBADF);
seq = seqc_read_notmodify(fd_seqc(fdt, fd));
fde = &fdt->fdt_ofiles[fd];
haverights = cap_rights_fde_inline(fde);
fp = fde->fde_file;
if (__predict_false(fp == NULL))
return (EAGAIN);
if (__predict_false(cap_check_inline_transient(haverights, &rights)))
return (EAGAIN);
*fsearch = ((fp->f_flag & FSEARCH) != 0);
vp = fp->f_vnode;
if (__predict_false(vp == NULL || vp->v_type != VDIR)) {
return (EAGAIN);
}
if (!filecaps_copy(&fde->fde_caps, &ndp->ni_filecaps, false)) {
return (EAGAIN);
}
/*
* Use an acquire barrier to force re-reading of fdt so it is
* refreshed for verification.
*/
atomic_thread_fence_acq();
fdt = fdp->fd_files;
if (__predict_false(!seqc_consistent_nomb(fd_seqc(fdt, fd), seq)))
return (EAGAIN);
/*
* If file descriptor doesn't have all rights,
* all lookups relative to it must also be
* strictly relative.
*
* Not yet supported by fast path.
*/
CAP_ALL(&rights);
if (!cap_rights_contains(&ndp->ni_filecaps.fc_rights, &rights) ||
ndp->ni_filecaps.fc_fcntls != CAP_FCNTL_ALL ||
ndp->ni_filecaps.fc_nioctls != -1) {
#ifdef notyet
ndp->ni_lcf |= NI_LCF_STRICTRELATIVE;
#else
return (EAGAIN);
#endif
}
*vpp = vp;
return (0);
}
#else
int
fgetvp_lookup_smr(int fd, struct nameidata *ndp, struct vnode **vpp, bool *fsearch)
{
const struct fdescenttbl *fdt;
struct filedesc *fdp;
struct file *fp;
struct vnode *vp;
VFS_SMR_ASSERT_ENTERED();
fdp = curproc->p_fd;
fdt = fdp->fd_files;
if (__predict_false((u_int)fd >= fdt->fdt_nfiles))
return (EBADF);
fp = fdt->fdt_ofiles[fd].fde_file;
if (__predict_false(fp == NULL))
return (EAGAIN);
*fsearch = ((fp->f_flag & FSEARCH) != 0);
vp = fp->f_vnode;
if (__predict_false(vp == NULL || vp->v_type != VDIR)) {
return (EAGAIN);
}
/*
* Use an acquire barrier to force re-reading of fdt so it is
* refreshed for verification.
*/
atomic_thread_fence_acq();
fdt = fdp->fd_files;
if (__predict_false(fp != fdt->fdt_ofiles[fd].fde_file))
return (EAGAIN);
filecaps_fill(&ndp->ni_filecaps);
*vpp = vp;
return (0);
}
#endif
int
fget_unlocked_seq(struct filedesc *fdp, int fd, cap_rights_t *needrightsp,
struct file **fpp, seqc_t *seqp)
{
#ifdef CAPABILITIES
const struct filedescent *fde;
#endif
const struct fdescenttbl *fdt;
struct file *fp;
#ifdef CAPABILITIES
seqc_t seq;
cap_rights_t haverights;
int error;
#endif
fdt = fdp->fd_files;
if (__predict_false((u_int)fd >= fdt->fdt_nfiles))
return (EBADF);
/*
* Fetch the descriptor locklessly. We avoid fdrop() races by
* never raising a refcount above 0. To accomplish this we have
* to use a cmpset loop rather than an atomic_add. The descriptor
* must be re-verified once we acquire a reference to be certain
* that the identity is still correct and we did not lose a race
* due to preemption.
*/
for (;;) {
#ifdef CAPABILITIES
seq = seqc_read_notmodify(fd_seqc(fdt, fd));
fde = &fdt->fdt_ofiles[fd];
haverights = *cap_rights_fde_inline(fde);
fp = fde->fde_file;
if (!seqc_consistent(fd_seqc(fdt, fd), seq))
continue;
#else
fp = fdt->fdt_ofiles[fd].fde_file;
#endif
if (fp == NULL)
return (EBADF);
#ifdef CAPABILITIES
error = cap_check_inline(&haverights, needrightsp);
if (error != 0)
return (error);
#endif
if (__predict_false(!refcount_acquire_if_not_zero(&fp->f_count))) {
/*
* Force a reload. Other thread could reallocate the
* table before this fd was closed, so it is possible
* that there is a stale fp pointer in cached version.
*/
fdt = atomic_load_ptr(&fdp->fd_files);
continue;
}
/*
* Use an acquire barrier to force re-reading of fdt so it is
* refreshed for verification.
*/
atomic_thread_fence_acq();
fdt = fdp->fd_files;
#ifdef CAPABILITIES
if (seqc_consistent_nomb(fd_seqc(fdt, fd), seq))
#else
if (fp == fdt->fdt_ofiles[fd].fde_file)
#endif
break;
fdrop(fp, curthread);
}
*fpp = fp;
if (seqp != NULL) {
#ifdef CAPABILITIES
*seqp = seq;
#endif
}
return (0);
}
/*
* See the comments in fget_unlocked_seq for an explanation of how this works.
*
* This is a simplified variant which bails out to the aforementioned routine
* if anything goes wrong. In practice this only happens when userspace is
* racing with itself.
*/
int
fget_unlocked(struct filedesc *fdp, int fd, cap_rights_t *needrightsp,
struct file **fpp)
{
#ifdef CAPABILITIES
const struct filedescent *fde;
#endif
const struct fdescenttbl *fdt;
struct file *fp;
#ifdef CAPABILITIES
seqc_t seq;
const cap_rights_t *haverights;
#endif
fdt = fdp->fd_files;
if (__predict_false((u_int)fd >= fdt->fdt_nfiles))
return (EBADF);
#ifdef CAPABILITIES
seq = seqc_read_notmodify(fd_seqc(fdt, fd));
fde = &fdt->fdt_ofiles[fd];
haverights = cap_rights_fde_inline(fde);
fp = fde->fde_file;
#else
fp = fdt->fdt_ofiles[fd].fde_file;
#endif
if (__predict_false(fp == NULL))
goto out_fallback;
#ifdef CAPABILITIES
if (__predict_false(cap_check_inline_transient(haverights, needrightsp)))
goto out_fallback;
#endif
if (__predict_false(!refcount_acquire_if_not_zero(&fp->f_count)))
goto out_fallback;
/*
* Use an acquire barrier to force re-reading of fdt so it is
* refreshed for verification.
*/
atomic_thread_fence_acq();
fdt = fdp->fd_files;
#ifdef CAPABILITIES
if (__predict_false(!seqc_consistent_nomb(fd_seqc(fdt, fd), seq)))
#else
if (__predict_false(fp != fdt->fdt_ofiles[fd].fde_file))
#endif
goto out_fdrop;
*fpp = fp;
return (0);
out_fdrop:
fdrop(fp, curthread);
out_fallback:
return (fget_unlocked_seq(fdp, fd, needrightsp, fpp, NULL));
}
/*
* Translate fd -> file when the caller guarantees the file descriptor table
* can't be changed by others.
*
* Note this does not mean the file object itself is only visible to the caller,
* merely that it wont disappear without having to be referenced.
*
* Must be paired with fput_only_user.
*/
#ifdef CAPABILITIES
int
fget_only_user(struct filedesc *fdp, int fd, cap_rights_t *needrightsp,
struct file **fpp)
{
const struct filedescent *fde;
const struct fdescenttbl *fdt;
const cap_rights_t *haverights;
struct file *fp;
int error;
MPASS(FILEDESC_IS_ONLY_USER(fdp));
if (__predict_false(fd >= fdp->fd_nfiles))
return (EBADF);
fdt = fdp->fd_files;
fde = &fdt->fdt_ofiles[fd];
fp = fde->fde_file;
if (__predict_false(fp == NULL))
return (EBADF);
MPASS(refcount_load(&fp->f_count) > 0);
haverights = cap_rights_fde_inline(fde);
error = cap_check_inline(haverights, needrightsp);
if (__predict_false(error != 0))
return (error);
*fpp = fp;
return (0);
}
#else
int
fget_only_user(struct filedesc *fdp, int fd, cap_rights_t *needrightsp,
struct file **fpp)
{
struct file *fp;
MPASS(FILEDESC_IS_ONLY_USER(fdp));
if (__predict_false(fd >= fdp->fd_nfiles))
return (EBADF);
fp = fdp->fd_ofiles[fd].fde_file;
if (__predict_false(fp == NULL))
return (EBADF);
MPASS(refcount_load(&fp->f_count) > 0);
*fpp = fp;
return (0);
}
#endif
/*
* Extract the file pointer associated with the specified descriptor for the
* current user process.
*
* If the descriptor doesn't exist or doesn't match 'flags', EBADF is
* returned.
*
* File's rights will be checked against the capability rights mask.
*
* If an error occurred the non-zero error is returned and *fpp is set to
* NULL. Otherwise *fpp is held and set and zero is returned. Caller is
* responsible for fdrop().
*/
static __inline int
_fget(struct thread *td, int fd, struct file **fpp, int flags,
cap_rights_t *needrightsp)
{
struct filedesc *fdp;
struct file *fp;
int error;
*fpp = NULL;
fdp = td->td_proc->p_fd;
error = fget_unlocked(fdp, fd, needrightsp, &fp);
if (__predict_false(error != 0))
return (error);
if (__predict_false(fp->f_ops == &badfileops)) {
fdrop(fp, td);
return (EBADF);
}
/*
* FREAD and FWRITE failure return EBADF as per POSIX.
*/
error = 0;
switch (flags) {
case FREAD:
case FWRITE:
if ((fp->f_flag & flags) == 0)
error = EBADF;
break;
case FEXEC:
if ((fp->f_flag & (FREAD | FEXEC)) == 0 ||
((fp->f_flag & FWRITE) != 0))
error = EBADF;
break;
case 0:
break;
default:
KASSERT(0, ("wrong flags"));
}
if (error != 0) {
fdrop(fp, td);
return (error);
}
*fpp = fp;
return (0);
}
int
fget(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
{
return (_fget(td, fd, fpp, 0, rightsp));
}
int
fget_mmap(struct thread *td, int fd, cap_rights_t *rightsp, vm_prot_t *maxprotp,
struct file **fpp)
{
int error;
#ifndef CAPABILITIES
error = _fget(td, fd, fpp, 0, rightsp);
if (maxprotp != NULL)
*maxprotp = VM_PROT_ALL;
return (error);
#else
cap_rights_t fdrights;
struct filedesc *fdp;
struct file *fp;
seqc_t seq;
*fpp = NULL;
fdp = td->td_proc->p_fd;
MPASS(cap_rights_is_set(rightsp, CAP_MMAP));
for (;;) {
error = fget_unlocked_seq(fdp, fd, rightsp, &fp, &seq);
if (__predict_false(error != 0))
return (error);
if (__predict_false(fp->f_ops == &badfileops)) {
fdrop(fp, td);
return (EBADF);
}
if (maxprotp != NULL)
fdrights = *cap_rights(fdp, fd);
if (!fd_modified(fdp, fd, seq))
break;
fdrop(fp, td);
}
/*
* If requested, convert capability rights to access flags.
*/
if (maxprotp != NULL)
*maxprotp = cap_rights_to_vmprot(&fdrights);
*fpp = fp;
return (0);
#endif
}
int
fget_read(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
{
return (_fget(td, fd, fpp, FREAD, rightsp));
}
int
fget_write(struct thread *td, int fd, cap_rights_t *rightsp, struct file **fpp)
{
return (_fget(td, fd, fpp, FWRITE, rightsp));
}
int
fget_fcntl(struct thread *td, int fd, cap_rights_t *rightsp, int needfcntl,
struct file **fpp)
{
struct filedesc *fdp = td->td_proc->p_fd;
#ifndef CAPABILITIES
return (fget_unlocked(fdp, fd, rightsp, fpp));
#else
struct file *fp;
int error;
seqc_t seq;
*fpp = NULL;
MPASS(cap_rights_is_set(rightsp, CAP_FCNTL));
for (;;) {
error = fget_unlocked_seq(fdp, fd, rightsp, &fp, &seq);
if (error != 0)
return (error);
error = cap_fcntl_check(fdp, fd, needfcntl);
if (!fd_modified(fdp, fd, seq))
break;
fdrop(fp, td);
}
if (error != 0) {
fdrop(fp, td);
return (error);
}
*fpp = fp;
return (0);
#endif
}
/*
* Like fget() but loads the underlying vnode, or returns an error if the
* descriptor does not represent a vnode. Note that pipes use vnodes but
* never have VM objects. The returned vnode will be vref()'d.
*
* XXX: what about the unused flags ?
*/
static __inline int
_fgetvp(struct thread *td, int fd, int flags, cap_rights_t *needrightsp,
struct vnode **vpp)
{
struct file *fp;
int error;
*vpp = NULL;
error = _fget(td, fd, &fp, flags, needrightsp);
if (error != 0)
return (error);
if (fp->f_vnode == NULL) {
error = EINVAL;
} else {
*vpp = fp->f_vnode;
vref(*vpp);
}
fdrop(fp, td);
return (error);
}
int
fgetvp(struct thread *td, int fd, cap_rights_t *rightsp, struct vnode **vpp)
{
return (_fgetvp(td, fd, 0, rightsp, vpp));
}
int
fgetvp_rights(struct thread *td, int fd, cap_rights_t *needrightsp,
struct filecaps *havecaps, struct vnode **vpp)
{
struct filecaps caps;
struct file *fp;
int error;
error = fget_cap(td, fd, needrightsp, &fp, &caps);
if (error != 0)
return (error);
if (fp->f_ops == &badfileops) {
error = EBADF;
goto out;
}
if (fp->f_vnode == NULL) {
error = EINVAL;
goto out;
}
*havecaps = caps;
*vpp = fp->f_vnode;
vref(*vpp);
fdrop(fp, td);
return (0);
out:
filecaps_free(&caps);
fdrop(fp, td);
return (error);
}
int
fgetvp_read(struct thread *td, int fd, cap_rights_t *rightsp, struct vnode **vpp)
{
return (_fgetvp(td, fd, FREAD, rightsp, vpp));
}
int
fgetvp_exec(struct thread *td, int fd, cap_rights_t *rightsp, struct vnode **vpp)
{
return (_fgetvp(td, fd, FEXEC, rightsp, vpp));
}
#ifdef notyet
int
fgetvp_write(struct thread *td, int fd, cap_rights_t *rightsp,
struct vnode **vpp)
{
return (_fgetvp(td, fd, FWRITE, rightsp, vpp));
}
#endif
/*
* Handle the last reference to a file being closed.
*
* Without the noinline attribute clang keeps inlining the func thorough this
* file when fdrop is used.
*/
int __noinline
_fdrop(struct file *fp, struct thread *td)
{
int error;
#ifdef INVARIANTS
int count;
count = refcount_load(&fp->f_count);
if (count != 0)
panic("fdrop: fp %p count %d", fp, count);
#endif
error = fo_close(fp, td);
atomic_subtract_int(&openfiles, 1);
crfree(fp->f_cred);
free(fp->f_advice, M_FADVISE);
uma_zfree(file_zone, fp);
return (error);
}
/*
* Apply an advisory lock on a file descriptor.
*
* Just attempt to get a record lock of the requested type on the entire file
* (l_whence = SEEK_SET, l_start = 0, l_len = 0).
*/
#ifndef _SYS_SYSPROTO_H_
struct flock_args {
int fd;
int how;
};
#endif
/* ARGSUSED */
int
sys_flock(struct thread *td, struct flock_args *uap)
{
struct file *fp;
struct vnode *vp;
struct flock lf;
int error;
error = fget(td, uap->fd, &cap_flock_rights, &fp);
if (error != 0)
return (error);
if (fp->f_type != DTYPE_VNODE || fp->f_ops == &path_fileops) {
fdrop(fp, td);
return (EOPNOTSUPP);
}
vp = fp->f_vnode;
lf.l_whence = SEEK_SET;
lf.l_start = 0;
lf.l_len = 0;
if (uap->how & LOCK_UN) {
lf.l_type = F_UNLCK;
atomic_clear_int(&fp->f_flag, FHASLOCK);
error = VOP_ADVLOCK(vp, (caddr_t)fp, F_UNLCK, &lf, F_FLOCK);
goto done2;
}
if (uap->how & LOCK_EX)
lf.l_type = F_WRLCK;
else if (uap->how & LOCK_SH)
lf.l_type = F_RDLCK;
else {
error = EBADF;
goto done2;
}
atomic_set_int(&fp->f_flag, FHASLOCK);
error = VOP_ADVLOCK(vp, (caddr_t)fp, F_SETLK, &lf,
(uap->how & LOCK_NB) ? F_FLOCK : F_FLOCK | F_WAIT);
done2:
fdrop(fp, td);
return (error);
}
/*
* Duplicate the specified descriptor to a free descriptor.
*/
int
dupfdopen(struct thread *td, struct filedesc *fdp, int dfd, int mode,
int openerror, int *indxp)
{
struct filedescent *newfde, *oldfde;
struct file *fp;
u_long *ioctls;
int error, indx;
KASSERT(openerror == ENODEV || openerror == ENXIO,
("unexpected error %d in %s", openerror, __func__));
/*
* If the to-be-dup'd fd number is greater than the allowed number
* of file descriptors, or the fd to be dup'd has already been
* closed, then reject.
*/
FILEDESC_XLOCK(fdp);
if ((fp = fget_locked(fdp, dfd)) == NULL) {
FILEDESC_XUNLOCK(fdp);
return (EBADF);
}
error = fdalloc(td, 0, &indx);
if (error != 0) {
FILEDESC_XUNLOCK(fdp);
return (error);
}
/*
* There are two cases of interest here.
*
* For ENODEV simply dup (dfd) to file descriptor (indx) and return.
*
* For ENXIO steal away the file structure from (dfd) and store it in
* (indx). (dfd) is effectively closed by this operation.
*/
switch (openerror) {
case ENODEV:
/*
* Check that the mode the file is being opened for is a
* subset of the mode of the existing descriptor.
*/
if (((mode & (FREAD|FWRITE)) | fp->f_flag) != fp->f_flag) {
fdunused(fdp, indx);
FILEDESC_XUNLOCK(fdp);
return (EACCES);
}
if (!fhold(fp)) {
fdunused(fdp, indx);
FILEDESC_XUNLOCK(fdp);
return (EBADF);
}
newfde = &fdp->fd_ofiles[indx];
oldfde = &fdp->fd_ofiles[dfd];
ioctls = filecaps_copy_prep(&oldfde->fde_caps);
#ifdef CAPABILITIES
seqc_write_begin(&newfde->fde_seqc);
#endif
memcpy(newfde, oldfde, fde_change_size);
filecaps_copy_finish(&oldfde->fde_caps, &newfde->fde_caps,
ioctls);
#ifdef CAPABILITIES
seqc_write_end(&newfde->fde_seqc);
#endif
break;
case ENXIO:
/*
* Steal away the file pointer from dfd and stuff it into indx.
*/
newfde = &fdp->fd_ofiles[indx];
oldfde = &fdp->fd_ofiles[dfd];
#ifdef CAPABILITIES
seqc_write_begin(&newfde->fde_seqc);
#endif
memcpy(newfde, oldfde, fde_change_size);
oldfde->fde_file = NULL;
fdunused(fdp, dfd);
#ifdef CAPABILITIES
seqc_write_end(&newfde->fde_seqc);
#endif
break;
}
FILEDESC_XUNLOCK(fdp);
*indxp = indx;
return (0);
}
/*
* This sysctl determines if we will allow a process to chroot(2) if it
* has a directory open:
* 0: disallowed for all processes.
* 1: allowed for processes that were not already chroot(2)'ed.
* 2: allowed for all processes.
*/
static int chroot_allow_open_directories = 1;
SYSCTL_INT(_kern, OID_AUTO, chroot_allow_open_directories, CTLFLAG_RW,
&chroot_allow_open_directories, 0,
"Allow a process to chroot(2) if it has a directory open");
/*
* Helper function for raised chroot(2) security function: Refuse if
* any filedescriptors are open directories.
*/
static int
chroot_refuse_vdir_fds(struct filedesc *fdp)
{
struct vnode *vp;
struct file *fp;
int fd, lastfile;
FILEDESC_LOCK_ASSERT(fdp);
lastfile = fdlastfile(fdp);
for (fd = 0; fd <= lastfile; fd++) {
fp = fget_locked(fdp, fd);
if (fp == NULL)
continue;
if (fp->f_type == DTYPE_VNODE) {
vp = fp->f_vnode;
if (vp->v_type == VDIR)
return (EPERM);
}
}
return (0);
}
static void
pwd_fill(struct pwd *oldpwd, struct pwd *newpwd)
{
if (newpwd->pwd_cdir == NULL && oldpwd->pwd_cdir != NULL) {
vrefact(oldpwd->pwd_cdir);
newpwd->pwd_cdir = oldpwd->pwd_cdir;
}
if (newpwd->pwd_rdir == NULL && oldpwd->pwd_rdir != NULL) {
vrefact(oldpwd->pwd_rdir);
newpwd->pwd_rdir = oldpwd->pwd_rdir;
}
if (newpwd->pwd_jdir == NULL && oldpwd->pwd_jdir != NULL) {
vrefact(oldpwd->pwd_jdir);
newpwd->pwd_jdir = oldpwd->pwd_jdir;
}
}
struct pwd *
pwd_hold_pwddesc(struct pwddesc *pdp)
{
struct pwd *pwd;
PWDDESC_ASSERT_XLOCKED(pdp);
pwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
if (pwd != NULL)
refcount_acquire(&pwd->pwd_refcount);
return (pwd);
}
bool
pwd_hold_smr(struct pwd *pwd)
{
MPASS(pwd != NULL);
if (__predict_true(refcount_acquire_if_not_zero(&pwd->pwd_refcount))) {
return (true);
}
return (false);
}
struct pwd *
pwd_hold(struct thread *td)
{
struct pwddesc *pdp;
struct pwd *pwd;
pdp = td->td_proc->p_pd;
vfs_smr_enter();
pwd = vfs_smr_entered_load(&pdp->pd_pwd);
if (pwd_hold_smr(pwd)) {
vfs_smr_exit();
return (pwd);
}
vfs_smr_exit();
PWDDESC_XLOCK(pdp);
pwd = pwd_hold_pwddesc(pdp);
MPASS(pwd != NULL);
PWDDESC_XUNLOCK(pdp);
return (pwd);
}
static struct pwd *
pwd_alloc(void)
{
struct pwd *pwd;
pwd = uma_zalloc_smr(pwd_zone, M_WAITOK);
bzero(pwd, sizeof(*pwd));
refcount_init(&pwd->pwd_refcount, 1);
return (pwd);
}
void
pwd_drop(struct pwd *pwd)
{
if (!refcount_release(&pwd->pwd_refcount))
return;
if (pwd->pwd_cdir != NULL)
vrele(pwd->pwd_cdir);
if (pwd->pwd_rdir != NULL)
vrele(pwd->pwd_rdir);
if (pwd->pwd_jdir != NULL)
vrele(pwd->pwd_jdir);
uma_zfree_smr(pwd_zone, pwd);
}
/*
* The caller is responsible for invoking priv_check() and
* mac_vnode_check_chroot() to authorize this operation.
*/
int
pwd_chroot(struct thread *td, struct vnode *vp)
{
struct pwddesc *pdp;
struct filedesc *fdp;
struct pwd *newpwd, *oldpwd;
int error;
fdp = td->td_proc->p_fd;
pdp = td->td_proc->p_pd;
newpwd = pwd_alloc();
FILEDESC_SLOCK(fdp);
PWDDESC_XLOCK(pdp);
oldpwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
if (chroot_allow_open_directories == 0 ||
(chroot_allow_open_directories == 1 &&
oldpwd->pwd_rdir != rootvnode)) {
error = chroot_refuse_vdir_fds(fdp);
FILEDESC_SUNLOCK(fdp);
if (error != 0) {
PWDDESC_XUNLOCK(pdp);
pwd_drop(newpwd);
return (error);
}
} else {
FILEDESC_SUNLOCK(fdp);
}
vrefact(vp);
newpwd->pwd_rdir = vp;
if (oldpwd->pwd_jdir == NULL) {
vrefact(vp);
newpwd->pwd_jdir = vp;
}
pwd_fill(oldpwd, newpwd);
pwd_set(pdp, newpwd);
PWDDESC_XUNLOCK(pdp);
pwd_drop(oldpwd);
return (0);
}
void
pwd_chdir(struct thread *td, struct vnode *vp)
{
struct pwddesc *pdp;
struct pwd *newpwd, *oldpwd;
VNPASS(vp->v_usecount > 0, vp);
newpwd = pwd_alloc();
pdp = td->td_proc->p_pd;
PWDDESC_XLOCK(pdp);
oldpwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
newpwd->pwd_cdir = vp;
pwd_fill(oldpwd, newpwd);
pwd_set(pdp, newpwd);
PWDDESC_XUNLOCK(pdp);
pwd_drop(oldpwd);
}
/*
* jail_attach(2) changes both root and working directories.
*/
int
pwd_chroot_chdir(struct thread *td, struct vnode *vp)
{
struct pwddesc *pdp;
struct filedesc *fdp;
struct pwd *newpwd, *oldpwd;
int error;
fdp = td->td_proc->p_fd;
pdp = td->td_proc->p_pd;
newpwd = pwd_alloc();
FILEDESC_SLOCK(fdp);
PWDDESC_XLOCK(pdp);
oldpwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
error = chroot_refuse_vdir_fds(fdp);
FILEDESC_SUNLOCK(fdp);
if (error != 0) {
PWDDESC_XUNLOCK(pdp);
pwd_drop(newpwd);
return (error);
}
vrefact(vp);
newpwd->pwd_rdir = vp;
vrefact(vp);
newpwd->pwd_cdir = vp;
if (oldpwd->pwd_jdir == NULL) {
vrefact(vp);
newpwd->pwd_jdir = vp;
}
pwd_fill(oldpwd, newpwd);
pwd_set(pdp, newpwd);
PWDDESC_XUNLOCK(pdp);
pwd_drop(oldpwd);
return (0);
}
void
pwd_ensure_dirs(void)
{
struct pwddesc *pdp;
struct pwd *oldpwd, *newpwd;
pdp = curproc->p_pd;
PWDDESC_XLOCK(pdp);
oldpwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
if (oldpwd->pwd_cdir != NULL && oldpwd->pwd_rdir != NULL) {
PWDDESC_XUNLOCK(pdp);
return;
}
PWDDESC_XUNLOCK(pdp);
newpwd = pwd_alloc();
PWDDESC_XLOCK(pdp);
oldpwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
pwd_fill(oldpwd, newpwd);
if (newpwd->pwd_cdir == NULL) {
vrefact(rootvnode);
newpwd->pwd_cdir = rootvnode;
}
if (newpwd->pwd_rdir == NULL) {
vrefact(rootvnode);
newpwd->pwd_rdir = rootvnode;
}
pwd_set(pdp, newpwd);
PWDDESC_XUNLOCK(pdp);
pwd_drop(oldpwd);
}
void
pwd_set_rootvnode(void)
{
struct pwddesc *pdp;
struct pwd *oldpwd, *newpwd;
pdp = curproc->p_pd;
newpwd = pwd_alloc();
PWDDESC_XLOCK(pdp);
oldpwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
vrefact(rootvnode);
newpwd->pwd_cdir = rootvnode;
vrefact(rootvnode);
newpwd->pwd_rdir = rootvnode;
pwd_fill(oldpwd, newpwd);
pwd_set(pdp, newpwd);
PWDDESC_XUNLOCK(pdp);
pwd_drop(oldpwd);
}
/*
* Scan all active processes and prisons to see if any of them have a current
* or root directory of `olddp'. If so, replace them with the new mount point.
*/
void
mountcheckdirs(struct vnode *olddp, struct vnode *newdp)
{
struct pwddesc *pdp;
struct pwd *newpwd, *oldpwd;
struct prison *pr;
struct proc *p;
int nrele;
if (vrefcnt(olddp) == 1)
return;
nrele = 0;
newpwd = pwd_alloc();
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
PROC_LOCK(p);
pdp = pdhold(p);
PROC_UNLOCK(p);
if (pdp == NULL)
continue;
PWDDESC_XLOCK(pdp);
oldpwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
if (oldpwd == NULL ||
(oldpwd->pwd_cdir != olddp &&
oldpwd->pwd_rdir != olddp &&
oldpwd->pwd_jdir != olddp)) {
PWDDESC_XUNLOCK(pdp);
pddrop(pdp);
continue;
}
if (oldpwd->pwd_cdir == olddp) {
vrefact(newdp);
newpwd->pwd_cdir = newdp;
}
if (oldpwd->pwd_rdir == olddp) {
vrefact(newdp);
newpwd->pwd_rdir = newdp;
}
if (oldpwd->pwd_jdir == olddp) {
vrefact(newdp);
newpwd->pwd_jdir = newdp;
}
pwd_fill(oldpwd, newpwd);
pwd_set(pdp, newpwd);
PWDDESC_XUNLOCK(pdp);
pwd_drop(oldpwd);
pddrop(pdp);
newpwd = pwd_alloc();
}
sx_sunlock(&allproc_lock);
pwd_drop(newpwd);
if (rootvnode == olddp) {
vrefact(newdp);
rootvnode = newdp;
nrele++;
}
mtx_lock(&prison0.pr_mtx);
if (prison0.pr_root == olddp) {
vrefact(newdp);
prison0.pr_root = newdp;
nrele++;
}
mtx_unlock(&prison0.pr_mtx);
sx_slock(&allprison_lock);
TAILQ_FOREACH(pr, &allprison, pr_list) {
mtx_lock(&pr->pr_mtx);
if (pr->pr_root == olddp) {
vrefact(newdp);
pr->pr_root = newdp;
nrele++;
}
mtx_unlock(&pr->pr_mtx);
}
sx_sunlock(&allprison_lock);
while (nrele--)
vrele(olddp);
}
struct filedesc_to_leader *
filedesc_to_leader_alloc(struct filedesc_to_leader *old, struct filedesc *fdp, struct proc *leader)
{
struct filedesc_to_leader *fdtol;
fdtol = malloc(sizeof(struct filedesc_to_leader),
M_FILEDESC_TO_LEADER, M_WAITOK);
fdtol->fdl_refcount = 1;
fdtol->fdl_holdcount = 0;
fdtol->fdl_wakeup = 0;
fdtol->fdl_leader = leader;
if (old != NULL) {
FILEDESC_XLOCK(fdp);
fdtol->fdl_next = old->fdl_next;
fdtol->fdl_prev = old;
old->fdl_next = fdtol;
fdtol->fdl_next->fdl_prev = fdtol;
FILEDESC_XUNLOCK(fdp);
} else {
fdtol->fdl_next = fdtol;
fdtol->fdl_prev = fdtol;
}
return (fdtol);
}
static int
sysctl_kern_proc_nfds(SYSCTL_HANDLER_ARGS)
{
NDSLOTTYPE *map;
struct filedesc *fdp;
int count, off, minoff;
if (*(int *)arg1 != 0)
return (EINVAL);
fdp = curproc->p_fd;
count = 0;
FILEDESC_SLOCK(fdp);
map = fdp->fd_map;
off = NDSLOT(fdp->fd_nfiles - 1);
for (minoff = NDSLOT(0); off >= minoff; --off)
count += bitcountl(map[off]);
FILEDESC_SUNLOCK(fdp);
return (SYSCTL_OUT(req, &count, sizeof(count)));
}
static SYSCTL_NODE(_kern_proc, KERN_PROC_NFDS, nfds,
CTLFLAG_RD|CTLFLAG_CAPRD|CTLFLAG_MPSAFE, sysctl_kern_proc_nfds,
"Number of open file descriptors");
/*
* Get file structures globally.
*/
static int
sysctl_kern_file(SYSCTL_HANDLER_ARGS)
{
struct xfile xf;
struct filedesc *fdp;
struct file *fp;
struct proc *p;
int error, n, lastfile;
error = sysctl_wire_old_buffer(req, 0);
if (error != 0)
return (error);
if (req->oldptr == NULL) {
n = 0;
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
PROC_LOCK(p);
if (p->p_state == PRS_NEW) {
PROC_UNLOCK(p);
continue;
}
fdp = fdhold(p);
PROC_UNLOCK(p);
if (fdp == NULL)
continue;
/* overestimates sparse tables. */
n += fdp->fd_nfiles;
fddrop(fdp);
}
sx_sunlock(&allproc_lock);
return (SYSCTL_OUT(req, 0, n * sizeof(xf)));
}
error = 0;
bzero(&xf, sizeof(xf));
xf.xf_size = sizeof(xf);
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
PROC_LOCK(p);
if (p->p_state == PRS_NEW) {
PROC_UNLOCK(p);
continue;
}
if (p_cansee(req->td, p) != 0) {
PROC_UNLOCK(p);
continue;
}
xf.xf_pid = p->p_pid;
xf.xf_uid = p->p_ucred->cr_uid;
fdp = fdhold(p);
PROC_UNLOCK(p);
if (fdp == NULL)
continue;
FILEDESC_SLOCK(fdp);
lastfile = fdlastfile(fdp);
for (n = 0; refcount_load(&fdp->fd_refcnt) > 0 && n <= lastfile;
n++) {
if ((fp = fdp->fd_ofiles[n].fde_file) == NULL)
continue;
xf.xf_fd = n;
xf.xf_file = (uintptr_t)fp;
xf.xf_data = (uintptr_t)fp->f_data;
xf.xf_vnode = (uintptr_t)fp->f_vnode;
xf.xf_type = (uintptr_t)fp->f_type;
xf.xf_count = refcount_load(&fp->f_count);
xf.xf_msgcount = 0;
xf.xf_offset = foffset_get(fp);
xf.xf_flag = fp->f_flag;
error = SYSCTL_OUT(req, &xf, sizeof(xf));
if (error)
break;
}
FILEDESC_SUNLOCK(fdp);
fddrop(fdp);
if (error)
break;
}
sx_sunlock(&allproc_lock);
return (error);
}
SYSCTL_PROC(_kern, KERN_FILE, file, CTLTYPE_OPAQUE|CTLFLAG_RD|CTLFLAG_MPSAFE,
0, 0, sysctl_kern_file, "S,xfile", "Entire file table");
#ifdef KINFO_FILE_SIZE
CTASSERT(sizeof(struct kinfo_file) == KINFO_FILE_SIZE);
#endif
static int
xlate_fflags(int fflags)
{
static const struct {
int fflag;
int kf_fflag;
} fflags_table[] = {
{ FAPPEND, KF_FLAG_APPEND },
{ FASYNC, KF_FLAG_ASYNC },
{ FFSYNC, KF_FLAG_FSYNC },
{ FHASLOCK, KF_FLAG_HASLOCK },
{ FNONBLOCK, KF_FLAG_NONBLOCK },
{ FREAD, KF_FLAG_READ },
{ FWRITE, KF_FLAG_WRITE },
{ O_CREAT, KF_FLAG_CREAT },
{ O_DIRECT, KF_FLAG_DIRECT },
{ O_EXCL, KF_FLAG_EXCL },
{ O_EXEC, KF_FLAG_EXEC },
{ O_EXLOCK, KF_FLAG_EXLOCK },
{ O_NOFOLLOW, KF_FLAG_NOFOLLOW },
{ O_SHLOCK, KF_FLAG_SHLOCK },
{ O_TRUNC, KF_FLAG_TRUNC }
};
unsigned int i;
int kflags;
kflags = 0;
for (i = 0; i < nitems(fflags_table); i++)
if (fflags & fflags_table[i].fflag)
kflags |= fflags_table[i].kf_fflag;
return (kflags);
}
/* Trim unused data from kf_path by truncating the structure size. */
void
pack_kinfo(struct kinfo_file *kif)
{
kif->kf_structsize = offsetof(struct kinfo_file, kf_path) +
strlen(kif->kf_path) + 1;
kif->kf_structsize = roundup(kif->kf_structsize, sizeof(uint64_t));
}
static void
export_file_to_kinfo(struct file *fp, int fd, cap_rights_t *rightsp,
struct kinfo_file *kif, struct filedesc *fdp, int flags)
{
int error;
bzero(kif, sizeof(*kif));
/* Set a default type to allow for empty fill_kinfo() methods. */
kif->kf_type = KF_TYPE_UNKNOWN;
kif->kf_flags = xlate_fflags(fp->f_flag);
if (rightsp != NULL)
kif->kf_cap_rights = *rightsp;
else
cap_rights_init_zero(&kif->kf_cap_rights);
kif->kf_fd = fd;
kif->kf_ref_count = refcount_load(&fp->f_count);
kif->kf_offset = foffset_get(fp);
/*
* This may drop the filedesc lock, so the 'fp' cannot be
* accessed after this call.
*/
error = fo_fill_kinfo(fp, kif, fdp);
if (error == 0)
kif->kf_status |= KF_ATTR_VALID;
if ((flags & KERN_FILEDESC_PACK_KINFO) != 0)
pack_kinfo(kif);
else
kif->kf_structsize = roundup2(sizeof(*kif), sizeof(uint64_t));
}
static void
export_vnode_to_kinfo(struct vnode *vp, int fd, int fflags,
struct kinfo_file *kif, int flags)
{
int error;
bzero(kif, sizeof(*kif));
kif->kf_type = KF_TYPE_VNODE;
error = vn_fill_kinfo_vnode(vp, kif);
if (error == 0)
kif->kf_status |= KF_ATTR_VALID;
kif->kf_flags = xlate_fflags(fflags);
cap_rights_init_zero(&kif->kf_cap_rights);
kif->kf_fd = fd;
kif->kf_ref_count = -1;
kif->kf_offset = -1;
if ((flags & KERN_FILEDESC_PACK_KINFO) != 0)
pack_kinfo(kif);
else
kif->kf_structsize = roundup2(sizeof(*kif), sizeof(uint64_t));
vrele(vp);
}
struct export_fd_buf {
struct filedesc *fdp;
struct pwddesc *pdp;
struct sbuf *sb;
ssize_t remainder;
struct kinfo_file kif;
int flags;
};
static int
export_kinfo_to_sb(struct export_fd_buf *efbuf)
{
struct kinfo_file *kif;
kif = &efbuf->kif;
if (efbuf->remainder != -1) {
if (efbuf->remainder < kif->kf_structsize) {
/* Terminate export. */
efbuf->remainder = 0;
return (0);
}
efbuf->remainder -= kif->kf_structsize;
}
return (sbuf_bcat(efbuf->sb, kif, kif->kf_structsize) == 0 ? 0 : ENOMEM);
}
static int
export_file_to_sb(struct file *fp, int fd, cap_rights_t *rightsp,
struct export_fd_buf *efbuf)
{
int error;
if (efbuf->remainder == 0)
return (0);
export_file_to_kinfo(fp, fd, rightsp, &efbuf->kif, efbuf->fdp,
efbuf->flags);
FILEDESC_SUNLOCK(efbuf->fdp);
error = export_kinfo_to_sb(efbuf);
FILEDESC_SLOCK(efbuf->fdp);
return (error);
}
static int
export_vnode_to_sb(struct vnode *vp, int fd, int fflags,
struct export_fd_buf *efbuf)
{
int error;
if (efbuf->remainder == 0)
return (0);
if (efbuf->pdp != NULL)
PWDDESC_XUNLOCK(efbuf->pdp);
export_vnode_to_kinfo(vp, fd, fflags, &efbuf->kif, efbuf->flags);
error = export_kinfo_to_sb(efbuf);
if (efbuf->pdp != NULL)
PWDDESC_XLOCK(efbuf->pdp);
return (error);
}
/*
* Store a process file descriptor information to sbuf.
*
* Takes a locked proc as argument, and returns with the proc unlocked.
*/
int
kern_proc_filedesc_out(struct proc *p, struct sbuf *sb, ssize_t maxlen,
int flags)
{
struct file *fp;
struct filedesc *fdp;
struct pwddesc *pdp;
struct export_fd_buf *efbuf;
struct vnode *cttyvp, *textvp, *tracevp;
struct pwd *pwd;
int error, i, lastfile;
cap_rights_t rights;
PROC_LOCK_ASSERT(p, MA_OWNED);
/* ktrace vnode */
- tracevp = p->p_tracevp;
- if (tracevp != NULL)
- vrefact(tracevp);
+ tracevp = ktr_get_tracevp(p, true);
/* text vnode */
textvp = p->p_textvp;
if (textvp != NULL)
vrefact(textvp);
/* Controlling tty. */
cttyvp = NULL;
if (p->p_pgrp != NULL && p->p_pgrp->pg_session != NULL) {
cttyvp = p->p_pgrp->pg_session->s_ttyvp;
if (cttyvp != NULL)
vrefact(cttyvp);
}
fdp = fdhold(p);
pdp = pdhold(p);
PROC_UNLOCK(p);
efbuf = malloc(sizeof(*efbuf), M_TEMP, M_WAITOK);
efbuf->fdp = NULL;
efbuf->pdp = NULL;
efbuf->sb = sb;
efbuf->remainder = maxlen;
efbuf->flags = flags;
if (tracevp != NULL)
export_vnode_to_sb(tracevp, KF_FD_TYPE_TRACE, FREAD | FWRITE,
efbuf);
if (textvp != NULL)
export_vnode_to_sb(textvp, KF_FD_TYPE_TEXT, FREAD, efbuf);
if (cttyvp != NULL)
export_vnode_to_sb(cttyvp, KF_FD_TYPE_CTTY, FREAD | FWRITE,
efbuf);
error = 0;
if (pdp == NULL || fdp == NULL)
goto fail;
efbuf->fdp = fdp;
efbuf->pdp = pdp;
PWDDESC_XLOCK(pdp);
pwd = pwd_hold_pwddesc(pdp);
if (pwd != NULL) {
/* working directory */
if (pwd->pwd_cdir != NULL) {
vrefact(pwd->pwd_cdir);
export_vnode_to_sb(pwd->pwd_cdir, KF_FD_TYPE_CWD,
FREAD, efbuf);
}
/* root directory */
if (pwd->pwd_rdir != NULL) {
vrefact(pwd->pwd_rdir);
export_vnode_to_sb(pwd->pwd_rdir, KF_FD_TYPE_ROOT,
FREAD, efbuf);
}
/* jail directory */
if (pwd->pwd_jdir != NULL) {
vrefact(pwd->pwd_jdir);
export_vnode_to_sb(pwd->pwd_jdir, KF_FD_TYPE_JAIL,
FREAD, efbuf);
}
}
PWDDESC_XUNLOCK(pdp);
if (pwd != NULL)
pwd_drop(pwd);
FILEDESC_SLOCK(fdp);
lastfile = fdlastfile(fdp);
for (i = 0; refcount_load(&fdp->fd_refcnt) > 0 && i <= lastfile; i++) {
if ((fp = fdp->fd_ofiles[i].fde_file) == NULL)
continue;
#ifdef CAPABILITIES
rights = *cap_rights(fdp, i);
#else /* !CAPABILITIES */
rights = cap_no_rights;
#endif
/*
* Create sysctl entry. It is OK to drop the filedesc
* lock inside of export_file_to_sb() as we will
* re-validate and re-evaluate its properties when the
* loop continues.
*/
error = export_file_to_sb(fp, i, &rights, efbuf);
if (error != 0 || efbuf->remainder == 0)
break;
}
FILEDESC_SUNLOCK(fdp);
fail:
if (fdp != NULL)
fddrop(fdp);
if (pdp != NULL)
pddrop(pdp);
free(efbuf, M_TEMP);
return (error);
}
#define FILEDESC_SBUF_SIZE (sizeof(struct kinfo_file) * 5)
/*
* Get per-process file descriptors for use by procstat(1), et al.
*/
static int
sysctl_kern_proc_filedesc(SYSCTL_HANDLER_ARGS)
{
struct sbuf sb;
struct proc *p;
ssize_t maxlen;
int error, error2, *name;
name = (int *)arg1;
sbuf_new_for_sysctl(&sb, NULL, FILEDESC_SBUF_SIZE, req);
sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
if (error != 0) {
sbuf_delete(&sb);
return (error);
}
maxlen = req->oldptr != NULL ? req->oldlen : -1;
error = kern_proc_filedesc_out(p, &sb, maxlen,
KERN_FILEDESC_PACK_KINFO);
error2 = sbuf_finish(&sb);
sbuf_delete(&sb);
return (error != 0 ? error : error2);
}
#ifdef COMPAT_FREEBSD7
#ifdef KINFO_OFILE_SIZE
CTASSERT(sizeof(struct kinfo_ofile) == KINFO_OFILE_SIZE);
#endif
static void
kinfo_to_okinfo(struct kinfo_file *kif, struct kinfo_ofile *okif)
{
okif->kf_structsize = sizeof(*okif);
okif->kf_type = kif->kf_type;
okif->kf_fd = kif->kf_fd;
okif->kf_ref_count = kif->kf_ref_count;
okif->kf_flags = kif->kf_flags & (KF_FLAG_READ | KF_FLAG_WRITE |
KF_FLAG_APPEND | KF_FLAG_ASYNC | KF_FLAG_FSYNC | KF_FLAG_NONBLOCK |
KF_FLAG_DIRECT | KF_FLAG_HASLOCK);
okif->kf_offset = kif->kf_offset;
if (kif->kf_type == KF_TYPE_VNODE)
okif->kf_vnode_type = kif->kf_un.kf_file.kf_file_type;
else
okif->kf_vnode_type = KF_VTYPE_VNON;
strlcpy(okif->kf_path, kif->kf_path, sizeof(okif->kf_path));
if (kif->kf_type == KF_TYPE_SOCKET) {
okif->kf_sock_domain = kif->kf_un.kf_sock.kf_sock_domain0;
okif->kf_sock_type = kif->kf_un.kf_sock.kf_sock_type0;
okif->kf_sock_protocol = kif->kf_un.kf_sock.kf_sock_protocol0;
okif->kf_sa_local = kif->kf_un.kf_sock.kf_sa_local;
okif->kf_sa_peer = kif->kf_un.kf_sock.kf_sa_peer;
} else {
okif->kf_sa_local.ss_family = AF_UNSPEC;
okif->kf_sa_peer.ss_family = AF_UNSPEC;
}
}
static int
export_vnode_for_osysctl(struct vnode *vp, int type, struct kinfo_file *kif,
struct kinfo_ofile *okif, struct pwddesc *pdp, struct sysctl_req *req)
{
int error;
vrefact(vp);
PWDDESC_XUNLOCK(pdp);
export_vnode_to_kinfo(vp, type, 0, kif, KERN_FILEDESC_PACK_KINFO);
kinfo_to_okinfo(kif, okif);
error = SYSCTL_OUT(req, okif, sizeof(*okif));
PWDDESC_XLOCK(pdp);
return (error);
}
/*
* Get per-process file descriptors for use by procstat(1), et al.
*/
static int
sysctl_kern_proc_ofiledesc(SYSCTL_HANDLER_ARGS)
{
struct kinfo_ofile *okif;
struct kinfo_file *kif;
struct filedesc *fdp;
struct pwddesc *pdp;
struct pwd *pwd;
int error, i, lastfile, *name;
struct file *fp;
struct proc *p;
name = (int *)arg1;
error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
if (error != 0)
return (error);
fdp = fdhold(p);
if (fdp != NULL)
pdp = pdhold(p);
PROC_UNLOCK(p);
if (fdp == NULL || pdp == NULL) {
if (fdp != NULL)
fddrop(fdp);
return (ENOENT);
}
kif = malloc(sizeof(*kif), M_TEMP, M_WAITOK);
okif = malloc(sizeof(*okif), M_TEMP, M_WAITOK);
PWDDESC_XLOCK(pdp);
pwd = pwd_hold_pwddesc(pdp);
if (pwd != NULL) {
if (pwd->pwd_cdir != NULL)
export_vnode_for_osysctl(pwd->pwd_cdir, KF_FD_TYPE_CWD, kif,
okif, pdp, req);
if (pwd->pwd_rdir != NULL)
export_vnode_for_osysctl(pwd->pwd_rdir, KF_FD_TYPE_ROOT, kif,
okif, pdp, req);
if (pwd->pwd_jdir != NULL)
export_vnode_for_osysctl(pwd->pwd_jdir, KF_FD_TYPE_JAIL, kif,
okif, pdp, req);
}
PWDDESC_XUNLOCK(pdp);
if (pwd != NULL)
pwd_drop(pwd);
FILEDESC_SLOCK(fdp);
lastfile = fdlastfile(fdp);
for (i = 0; refcount_load(&fdp->fd_refcnt) > 0 && i <= lastfile; i++) {
if ((fp = fdp->fd_ofiles[i].fde_file) == NULL)
continue;
export_file_to_kinfo(fp, i, NULL, kif, fdp,
KERN_FILEDESC_PACK_KINFO);
FILEDESC_SUNLOCK(fdp);
kinfo_to_okinfo(kif, okif);
error = SYSCTL_OUT(req, okif, sizeof(*okif));
FILEDESC_SLOCK(fdp);
if (error)
break;
}
FILEDESC_SUNLOCK(fdp);
fddrop(fdp);
pddrop(pdp);
free(kif, M_TEMP);
free(okif, M_TEMP);
return (0);
}
static SYSCTL_NODE(_kern_proc, KERN_PROC_OFILEDESC, ofiledesc,
CTLFLAG_RD|CTLFLAG_MPSAFE, sysctl_kern_proc_ofiledesc,
"Process ofiledesc entries");
#endif /* COMPAT_FREEBSD7 */
int
vntype_to_kinfo(int vtype)
{
struct {
int vtype;
int kf_vtype;
} vtypes_table[] = {
{ VBAD, KF_VTYPE_VBAD },
{ VBLK, KF_VTYPE_VBLK },
{ VCHR, KF_VTYPE_VCHR },
{ VDIR, KF_VTYPE_VDIR },
{ VFIFO, KF_VTYPE_VFIFO },
{ VLNK, KF_VTYPE_VLNK },
{ VNON, KF_VTYPE_VNON },
{ VREG, KF_VTYPE_VREG },
{ VSOCK, KF_VTYPE_VSOCK }
};
unsigned int i;
/*
* Perform vtype translation.
*/
for (i = 0; i < nitems(vtypes_table); i++)
if (vtypes_table[i].vtype == vtype)
return (vtypes_table[i].kf_vtype);
return (KF_VTYPE_UNKNOWN);
}
static SYSCTL_NODE(_kern_proc, KERN_PROC_FILEDESC, filedesc,
CTLFLAG_RD|CTLFLAG_MPSAFE, sysctl_kern_proc_filedesc,
"Process filedesc entries");
/*
* Store a process current working directory information to sbuf.
*
* Takes a locked proc as argument, and returns with the proc unlocked.
*/
int
kern_proc_cwd_out(struct proc *p, struct sbuf *sb, ssize_t maxlen)
{
struct pwddesc *pdp;
struct pwd *pwd;
struct export_fd_buf *efbuf;
struct vnode *cdir;
int error;
PROC_LOCK_ASSERT(p, MA_OWNED);
pdp = pdhold(p);
PROC_UNLOCK(p);
if (pdp == NULL)
return (EINVAL);
efbuf = malloc(sizeof(*efbuf), M_TEMP, M_WAITOK);
efbuf->pdp = pdp;
efbuf->sb = sb;
efbuf->remainder = maxlen;
PWDDESC_XLOCK(pdp);
pwd = PWDDESC_XLOCKED_LOAD_PWD(pdp);
cdir = pwd->pwd_cdir;
if (cdir == NULL) {
error = EINVAL;
} else {
vrefact(cdir);
error = export_vnode_to_sb(cdir, KF_FD_TYPE_CWD, FREAD, efbuf);
}
PWDDESC_XUNLOCK(pdp);
pddrop(pdp);
free(efbuf, M_TEMP);
return (error);
}
/*
* Get per-process current working directory.
*/
static int
sysctl_kern_proc_cwd(SYSCTL_HANDLER_ARGS)
{
struct sbuf sb;
struct proc *p;
ssize_t maxlen;
int error, error2, *name;
name = (int *)arg1;
sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_file), req);
sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
if (error != 0) {
sbuf_delete(&sb);
return (error);
}
maxlen = req->oldptr != NULL ? req->oldlen : -1;
error = kern_proc_cwd_out(p, &sb, maxlen);
error2 = sbuf_finish(&sb);
sbuf_delete(&sb);
return (error != 0 ? error : error2);
}
static SYSCTL_NODE(_kern_proc, KERN_PROC_CWD, cwd, CTLFLAG_RD|CTLFLAG_MPSAFE,
sysctl_kern_proc_cwd, "Process current working directory");
#ifdef DDB
/*
* For the purposes of debugging, generate a human-readable string for the
* file type.
*/
static const char *
file_type_to_name(short type)
{
switch (type) {
case 0:
return ("zero");
case DTYPE_VNODE:
return ("vnode");
case DTYPE_SOCKET:
return ("socket");
case DTYPE_PIPE:
return ("pipe");
case DTYPE_FIFO:
return ("fifo");
case DTYPE_KQUEUE:
return ("kqueue");
case DTYPE_CRYPTO:
return ("crypto");
case DTYPE_MQUEUE:
return ("mqueue");
case DTYPE_SHM:
return ("shm");
case DTYPE_SEM:
return ("ksem");
case DTYPE_PTS:
return ("pts");
case DTYPE_DEV:
return ("dev");
case DTYPE_PROCDESC:
return ("proc");
case DTYPE_EVENTFD:
return ("eventfd");
case DTYPE_LINUXTFD:
return ("ltimer");
default:
return ("unkn");
}
}
/*
* For the purposes of debugging, identify a process (if any, perhaps one of
* many) that references the passed file in its file descriptor array. Return
* NULL if none.
*/
static struct proc *
file_to_first_proc(struct file *fp)
{
struct filedesc *fdp;
struct proc *p;
int n;
FOREACH_PROC_IN_SYSTEM(p) {
if (p->p_state == PRS_NEW)
continue;
fdp = p->p_fd;
if (fdp == NULL)
continue;
for (n = 0; n < fdp->fd_nfiles; n++) {
if (fp == fdp->fd_ofiles[n].fde_file)
return (p);
}
}
return (NULL);
}
static void
db_print_file(struct file *fp, int header)
{
#define XPTRWIDTH ((int)howmany(sizeof(void *) * NBBY, 4))
struct proc *p;
if (header)
db_printf("%*s %6s %*s %8s %4s %5s %6s %*s %5s %s\n",
XPTRWIDTH, "File", "Type", XPTRWIDTH, "Data", "Flag",
"GCFl", "Count", "MCount", XPTRWIDTH, "Vnode", "FPID",
"FCmd");
p = file_to_first_proc(fp);
db_printf("%*p %6s %*p %08x %04x %5d %6d %*p %5d %s\n", XPTRWIDTH,
fp, file_type_to_name(fp->f_type), XPTRWIDTH, fp->f_data,
fp->f_flag, 0, refcount_load(&fp->f_count), 0, XPTRWIDTH, fp->f_vnode,
p != NULL ? p->p_pid : -1, p != NULL ? p->p_comm : "-");
#undef XPTRWIDTH
}
DB_SHOW_COMMAND(file, db_show_file)
{
struct file *fp;
if (!have_addr) {
db_printf("usage: show file <addr>\n");
return;
}
fp = (struct file *)addr;
db_print_file(fp, 1);
}
DB_SHOW_COMMAND(files, db_show_files)
{
struct filedesc *fdp;
struct file *fp;
struct proc *p;
int header;
int n;
header = 1;
FOREACH_PROC_IN_SYSTEM(p) {
if (p->p_state == PRS_NEW)
continue;
if ((fdp = p->p_fd) == NULL)
continue;
for (n = 0; n < fdp->fd_nfiles; ++n) {
if ((fp = fdp->fd_ofiles[n].fde_file) == NULL)
continue;
db_print_file(fp, header);
header = 0;
}
}
}
#endif
SYSCTL_INT(_kern, KERN_MAXFILESPERPROC, maxfilesperproc, CTLFLAG_RW,
&maxfilesperproc, 0, "Maximum files allowed open per process");
SYSCTL_INT(_kern, KERN_MAXFILES, maxfiles, CTLFLAG_RW,
&maxfiles, 0, "Maximum number of files");
SYSCTL_INT(_kern, OID_AUTO, openfiles, CTLFLAG_RD,
&openfiles, 0, "System-wide number of open files");
/* ARGSUSED*/
static void
filelistinit(void *dummy)
{
file_zone = uma_zcreate("Files", sizeof(struct file), NULL, NULL,
NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
filedesc0_zone = uma_zcreate("filedesc0", sizeof(struct filedesc0),
NULL, NULL, NULL, NULL, UMA_ALIGN_PTR, 0);
pwd_zone = uma_zcreate("PWD", sizeof(struct pwd), NULL, NULL,
NULL, NULL, UMA_ALIGN_PTR, UMA_ZONE_SMR);
/*
* XXXMJG this is a temporary hack due to boot ordering issues against
* the vnode zone.
*/
vfs_smr = uma_zone_get_smr(pwd_zone);
mtx_init(&sigio_lock, "sigio lock", NULL, MTX_DEF);
}
SYSINIT(select, SI_SUB_LOCK, SI_ORDER_FIRST, filelistinit, NULL);
/*-------------------------------------------------------------------*/
static int
badfo_readwrite(struct file *fp, struct uio *uio, struct ucred *active_cred,
int flags, struct thread *td)
{
return (EBADF);
}
static int
badfo_truncate(struct file *fp, off_t length, struct ucred *active_cred,
struct thread *td)
{
return (EINVAL);
}
static int
badfo_ioctl(struct file *fp, u_long com, void *data, struct ucred *active_cred,
struct thread *td)
{
return (EBADF);
}
static int
badfo_poll(struct file *fp, int events, struct ucred *active_cred,
struct thread *td)
{
return (0);
}
static int
badfo_kqfilter(struct file *fp, struct knote *kn)
{
return (EBADF);
}
static int
badfo_stat(struct file *fp, struct stat *sb, struct ucred *active_cred,
struct thread *td)
{
return (EBADF);
}
static int
badfo_close(struct file *fp, struct thread *td)
{
return (0);
}
static int
badfo_chmod(struct file *fp, mode_t mode, struct ucred *active_cred,
struct thread *td)
{
return (EBADF);
}
static int
badfo_chown(struct file *fp, uid_t uid, gid_t gid, struct ucred *active_cred,
struct thread *td)
{
return (EBADF);
}
static int
badfo_sendfile(struct file *fp, int sockfd, struct uio *hdr_uio,
struct uio *trl_uio, off_t offset, size_t nbytes, off_t *sent, int flags,
struct thread *td)
{
return (EBADF);
}
static int
badfo_fill_kinfo(struct file *fp, struct kinfo_file *kif, struct filedesc *fdp)
{
return (0);
}
struct fileops badfileops = {
.fo_read = badfo_readwrite,
.fo_write = badfo_readwrite,
.fo_truncate = badfo_truncate,
.fo_ioctl = badfo_ioctl,
.fo_poll = badfo_poll,
.fo_kqfilter = badfo_kqfilter,
.fo_stat = badfo_stat,
.fo_close = badfo_close,
.fo_chmod = badfo_chmod,
.fo_chown = badfo_chown,
.fo_sendfile = badfo_sendfile,
.fo_fill_kinfo = badfo_fill_kinfo,
};
static int
path_poll(struct file *fp, int events, struct ucred *active_cred,
struct thread *td)
{
return (POLLNVAL);
}
static int
path_close(struct file *fp, struct thread *td)
{
MPASS(fp->f_type == DTYPE_VNODE);
fp->f_ops = &badfileops;
vdrop(fp->f_vnode);
return (0);
}
struct fileops path_fileops = {
.fo_read = badfo_readwrite,
.fo_write = badfo_readwrite,
.fo_truncate = badfo_truncate,
.fo_ioctl = badfo_ioctl,
.fo_poll = path_poll,
.fo_kqfilter = vn_kqfilter_opath,
.fo_stat = vn_statfile,
.fo_close = path_close,
.fo_chmod = badfo_chmod,
.fo_chown = badfo_chown,
.fo_sendfile = badfo_sendfile,
.fo_fill_kinfo = vn_fill_kinfo,
.fo_flags = DFLAG_PASSABLE,
};
int
invfo_rdwr(struct file *fp, struct uio *uio, struct ucred *active_cred,
int flags, struct thread *td)
{
return (EOPNOTSUPP);
}
int
invfo_truncate(struct file *fp, off_t length, struct ucred *active_cred,
struct thread *td)
{
return (EINVAL);
}
int
invfo_ioctl(struct file *fp, u_long com, void *data,
struct ucred *active_cred, struct thread *td)
{
return (ENOTTY);
}
int
invfo_poll(struct file *fp, int events, struct ucred *active_cred,
struct thread *td)
{
return (poll_no_poll(events));
}
int
invfo_kqfilter(struct file *fp, struct knote *kn)
{
return (EINVAL);
}
int
invfo_chmod(struct file *fp, mode_t mode, struct ucred *active_cred,
struct thread *td)
{
return (EINVAL);
}
int
invfo_chown(struct file *fp, uid_t uid, gid_t gid, struct ucred *active_cred,
struct thread *td)
{
return (EINVAL);
}
int
invfo_sendfile(struct file *fp, int sockfd, struct uio *hdr_uio,
struct uio *trl_uio, off_t offset, size_t nbytes, off_t *sent, int flags,
struct thread *td)
{
return (EINVAL);
}
/*-------------------------------------------------------------------*/
/*
* File Descriptor pseudo-device driver (/dev/fd/).
*
* Opening minor device N dup()s the file (if any) connected to file
* descriptor N belonging to the calling process. Note that this driver
* consists of only the ``open()'' routine, because all subsequent
* references to this file will be direct to the other driver.
*
* XXX: we could give this one a cloning event handler if necessary.
*/
/* ARGSUSED */
static int
fdopen(struct cdev *dev, int mode, int type, struct thread *td)
{
/*
* XXX Kludge: set curthread->td_dupfd to contain the value of the
* the file descriptor being sought for duplication. The error
* return ensures that the vnode for this device will be released
* by vn_open. Open will detect this special error and take the
* actions in dupfdopen below. Other callers of vn_open or VOP_OPEN
* will simply report the error.
*/
td->td_dupfd = dev2unit(dev);
return (ENODEV);
}
static struct cdevsw fildesc_cdevsw = {
.d_version = D_VERSION,
.d_open = fdopen,
.d_name = "FD",
};
static void
fildesc_drvinit(void *unused)
{
struct cdev *dev;
dev = make_dev_credf(MAKEDEV_ETERNAL, &fildesc_cdevsw, 0, NULL,
UID_ROOT, GID_WHEEL, 0666, "fd/0");
make_dev_alias(dev, "stdin");
dev = make_dev_credf(MAKEDEV_ETERNAL, &fildesc_cdevsw, 1, NULL,
UID_ROOT, GID_WHEEL, 0666, "fd/1");
make_dev_alias(dev, "stdout");
dev = make_dev_credf(MAKEDEV_ETERNAL, &fildesc_cdevsw, 2, NULL,
UID_ROOT, GID_WHEEL, 0666, "fd/2");
make_dev_alias(dev, "stderr");
}
SYSINIT(fildescdev, SI_SUB_DRIVERS, SI_ORDER_MIDDLE, fildesc_drvinit, NULL);
diff --git a/sys/kern/kern_exec.c b/sys/kern/kern_exec.c
index 3413a5d024d4..2936e246f706 100644
--- a/sys/kern/kern_exec.c
+++ b/sys/kern/kern_exec.c
@@ -1,1874 +1,1868 @@
/*-
* SPDX-License-Identifier: BSD-2-Clause-FreeBSD
*
* Copyright (c) 1993, David Greenman
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
*
* THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_capsicum.h"
#include "opt_hwpmc_hooks.h"
#include "opt_ktrace.h"
#include "opt_vm.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/acct.h>
#include <sys/capsicum.h>
#include <sys/eventhandler.h>
#include <sys/exec.h>
#include <sys/fcntl.h>
#include <sys/filedesc.h>
#include <sys/imgact.h>
#include <sys/imgact_elf.h>
#include <sys/kernel.h>
#include <sys/lock.h>
#include <sys/malloc.h>
#include <sys/mman.h>
#include <sys/mount.h>
#include <sys/mutex.h>
#include <sys/namei.h>
#include <sys/priv.h>
#include <sys/proc.h>
#include <sys/ptrace.h>
#include <sys/resourcevar.h>
#include <sys/rwlock.h>
#include <sys/sched.h>
#include <sys/sdt.h>
#include <sys/sf_buf.h>
#include <sys/shm.h>
#include <sys/signalvar.h>
#include <sys/smp.h>
#include <sys/stat.h>
#include <sys/syscallsubr.h>
#include <sys/sysctl.h>
#include <sys/sysent.h>
#include <sys/sysproto.h>
#include <sys/timers.h>
#include <sys/umtx.h>
#include <sys/vnode.h>
#include <sys/wait.h>
#ifdef KTRACE
#include <sys/ktrace.h>
#endif
#include <vm/vm.h>
#include <vm/vm_param.h>
#include <vm/pmap.h>
#include <vm/vm_page.h>
#include <vm/vm_map.h>
#include <vm/vm_kern.h>
#include <vm/vm_extern.h>
#include <vm/vm_object.h>
#include <vm/vm_pager.h>
#ifdef HWPMC_HOOKS
#include <sys/pmckern.h>
#endif
#include <machine/reg.h>
#include <security/audit/audit.h>
#include <security/mac/mac_framework.h>
#ifdef KDTRACE_HOOKS
#include <sys/dtrace_bsd.h>
dtrace_execexit_func_t dtrace_fasttrap_exec;
#endif
SDT_PROVIDER_DECLARE(proc);
SDT_PROBE_DEFINE1(proc, , , exec, "char *");
SDT_PROBE_DEFINE1(proc, , , exec__failure, "int");
SDT_PROBE_DEFINE1(proc, , , exec__success, "char *");
MALLOC_DEFINE(M_PARGS, "proc-args", "Process arguments");
int coredump_pack_fileinfo = 1;
SYSCTL_INT(_kern, OID_AUTO, coredump_pack_fileinfo, CTLFLAG_RWTUN,
&coredump_pack_fileinfo, 0,
"Enable file path packing in 'procstat -f' coredump notes");
int coredump_pack_vmmapinfo = 1;
SYSCTL_INT(_kern, OID_AUTO, coredump_pack_vmmapinfo, CTLFLAG_RWTUN,
&coredump_pack_vmmapinfo, 0,
"Enable file path packing in 'procstat -v' coredump notes");
static int sysctl_kern_ps_strings(SYSCTL_HANDLER_ARGS);
static int sysctl_kern_usrstack(SYSCTL_HANDLER_ARGS);
static int sysctl_kern_stackprot(SYSCTL_HANDLER_ARGS);
static int do_execve(struct thread *td, struct image_args *args,
struct mac *mac_p, struct vmspace *oldvmspace);
/* XXX This should be vm_size_t. */
SYSCTL_PROC(_kern, KERN_PS_STRINGS, ps_strings, CTLTYPE_ULONG|CTLFLAG_RD|
CTLFLAG_CAPRD|CTLFLAG_MPSAFE, NULL, 0, sysctl_kern_ps_strings, "LU",
"Location of process' ps_strings structure");
/* XXX This should be vm_size_t. */
SYSCTL_PROC(_kern, KERN_USRSTACK, usrstack, CTLTYPE_ULONG|CTLFLAG_RD|
CTLFLAG_CAPRD|CTLFLAG_MPSAFE, NULL, 0, sysctl_kern_usrstack, "LU",
"Top of process stack");
SYSCTL_PROC(_kern, OID_AUTO, stackprot, CTLTYPE_INT|CTLFLAG_RD|CTLFLAG_MPSAFE,
NULL, 0, sysctl_kern_stackprot, "I",
"Stack memory permissions");
u_long ps_arg_cache_limit = PAGE_SIZE / 16;
SYSCTL_ULONG(_kern, OID_AUTO, ps_arg_cache_limit, CTLFLAG_RW,
&ps_arg_cache_limit, 0,
"Process' command line characters cache limit");
static int disallow_high_osrel;
SYSCTL_INT(_kern, OID_AUTO, disallow_high_osrel, CTLFLAG_RW,
&disallow_high_osrel, 0,
"Disallow execution of binaries built for higher version of the world");
static int map_at_zero = 0;
SYSCTL_INT(_security_bsd, OID_AUTO, map_at_zero, CTLFLAG_RWTUN, &map_at_zero, 0,
"Permit processes to map an object at virtual address 0.");
static int
sysctl_kern_ps_strings(SYSCTL_HANDLER_ARGS)
{
struct proc *p;
int error;
p = curproc;
#ifdef SCTL_MASK32
if (req->flags & SCTL_MASK32) {
unsigned int val;
val = (unsigned int)p->p_sysent->sv_psstrings;
error = SYSCTL_OUT(req, &val, sizeof(val));
} else
#endif
error = SYSCTL_OUT(req, &p->p_sysent->sv_psstrings,
sizeof(p->p_sysent->sv_psstrings));
return error;
}
static int
sysctl_kern_usrstack(SYSCTL_HANDLER_ARGS)
{
struct proc *p;
int error;
p = curproc;
#ifdef SCTL_MASK32
if (req->flags & SCTL_MASK32) {
unsigned int val;
val = (unsigned int)p->p_sysent->sv_usrstack;
error = SYSCTL_OUT(req, &val, sizeof(val));
} else
#endif
error = SYSCTL_OUT(req, &p->p_sysent->sv_usrstack,
sizeof(p->p_sysent->sv_usrstack));
return error;
}
static int
sysctl_kern_stackprot(SYSCTL_HANDLER_ARGS)
{
struct proc *p;
p = curproc;
return (SYSCTL_OUT(req, &p->p_sysent->sv_stackprot,
sizeof(p->p_sysent->sv_stackprot)));
}
/*
* Each of the items is a pointer to a `const struct execsw', hence the
* double pointer here.
*/
static const struct execsw **execsw;
#ifndef _SYS_SYSPROTO_H_
struct execve_args {
char *fname;
char **argv;
char **envv;
};
#endif
int
sys_execve(struct thread *td, struct execve_args *uap)
{
struct image_args args;
struct vmspace *oldvmspace;
int error;
error = pre_execve(td, &oldvmspace);
if (error != 0)
return (error);
error = exec_copyin_args(&args, uap->fname, UIO_USERSPACE,
uap->argv, uap->envv);
if (error == 0)
error = kern_execve(td, &args, NULL, oldvmspace);
post_execve(td, error, oldvmspace);
AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td);
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct fexecve_args {
int fd;
char **argv;
char **envv;
};
#endif
int
sys_fexecve(struct thread *td, struct fexecve_args *uap)
{
struct image_args args;
struct vmspace *oldvmspace;
int error;
error = pre_execve(td, &oldvmspace);
if (error != 0)
return (error);
error = exec_copyin_args(&args, NULL, UIO_SYSSPACE,
uap->argv, uap->envv);
if (error == 0) {
args.fd = uap->fd;
error = kern_execve(td, &args, NULL, oldvmspace);
}
post_execve(td, error, oldvmspace);
AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td);
return (error);
}
#ifndef _SYS_SYSPROTO_H_
struct __mac_execve_args {
char *fname;
char **argv;
char **envv;
struct mac *mac_p;
};
#endif
int
sys___mac_execve(struct thread *td, struct __mac_execve_args *uap)
{
#ifdef MAC
struct image_args args;
struct vmspace *oldvmspace;
int error;
error = pre_execve(td, &oldvmspace);
if (error != 0)
return (error);
error = exec_copyin_args(&args, uap->fname, UIO_USERSPACE,
uap->argv, uap->envv);
if (error == 0)
error = kern_execve(td, &args, uap->mac_p, oldvmspace);
post_execve(td, error, oldvmspace);
AUDIT_SYSCALL_EXIT(error == EJUSTRETURN ? 0 : error, td);
return (error);
#else
return (ENOSYS);
#endif
}
int
pre_execve(struct thread *td, struct vmspace **oldvmspace)
{
struct proc *p;
int error;
KASSERT(td == curthread, ("non-current thread %p", td));
error = 0;
p = td->td_proc;
if ((p->p_flag & P_HADTHREADS) != 0) {
PROC_LOCK(p);
if (thread_single(p, SINGLE_BOUNDARY) != 0)
error = ERESTART;
PROC_UNLOCK(p);
}
KASSERT(error != 0 || (td->td_pflags & TDP_EXECVMSPC) == 0,
("nested execve"));
*oldvmspace = p->p_vmspace;
return (error);
}
void
post_execve(struct thread *td, int error, struct vmspace *oldvmspace)
{
struct proc *p;
KASSERT(td == curthread, ("non-current thread %p", td));
p = td->td_proc;
if ((p->p_flag & P_HADTHREADS) != 0) {
PROC_LOCK(p);
/*
* If success, we upgrade to SINGLE_EXIT state to
* force other threads to suicide.
*/
if (error == EJUSTRETURN)
thread_single(p, SINGLE_EXIT);
else
thread_single_end(p, SINGLE_BOUNDARY);
PROC_UNLOCK(p);
}
exec_cleanup(td, oldvmspace);
}
/*
* kern_execve() has the astonishing property of not always returning to
* the caller. If sufficiently bad things happen during the call to
* do_execve(), it can end up calling exit1(); as a result, callers must
* avoid doing anything which they might need to undo (e.g., allocating
* memory).
*/
int
kern_execve(struct thread *td, struct image_args *args, struct mac *mac_p,
struct vmspace *oldvmspace)
{
AUDIT_ARG_ARGV(args->begin_argv, args->argc,
exec_args_get_begin_envv(args) - args->begin_argv);
AUDIT_ARG_ENVV(exec_args_get_begin_envv(args), args->envc,
args->endp - exec_args_get_begin_envv(args));
return (do_execve(td, args, mac_p, oldvmspace));
}
static void
execve_nosetid(struct image_params *imgp)
{
imgp->credential_setid = false;
if (imgp->newcred != NULL) {
crfree(imgp->newcred);
imgp->newcred = NULL;
}
}
/*
* 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 vmspace *oldvmspace)
{
struct proc *p = td->td_proc;
struct nameidata nd;
struct ucred *oldcred;
struct uidinfo *euip = NULL;
uintptr_t stack_base;
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;
+ struct ktr_io_params *kiop;
#endif
struct vnode *oldtextvp = NULL, *newtextvp;
int credential_changing;
#ifdef MAC
struct label *interpvplabel = NULL;
int will_transition;
#endif
#ifdef HWPMC_HOOKS
struct pmckern_procexec pe;
#endif
int error, i, orig_osrel;
uint32_t orig_fctl0;
static const char fexecv_proc_title[] = "(fexecv)";
imgp = &image_params;
+ kiop = NULL;
/*
* 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;
orig_osrel = p->p_osrel;
orig_fctl0 = p->p_fctl0;
#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 | LOCKSHARED | 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_SHARED | LK_RETRY);
AUDIT_ARG_VNODE1(newtextvp);
imgp->vp = newtextvp;
}
/*
* Check file permissions. Also 'opens' file and sets its vnode to
* text mode.
*/
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);
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);
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_SHARED | 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);
imgp->newcred = crdup(oldcred);
vn_lock(imgp->vp, LK_SHARED | 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);
if (vn_fullpath(imgp->vp, &imgp->execpath, &imgp->freepath) != 0)
imgp->execpath = args->fname;
vn_lock(imgp->vp, LK_SHARED | 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)
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);
/*
* The text reference needs to be removed for scripts.
* There is a short period before we determine that
* something is a script where text reference is active.
* The vnode lock is held over this entire period
* so nothing should illegitimately be blocked.
*/
MPASS(imgp->textset);
VOP_UNSET_TEXT_CHECKED(newtextvp);
imgp->textset = false;
/* 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;
execve_nosetid(imgp);
imgp->execpath = NULL;
free(imgp->freepath, M_TEMP);
imgp->freepath = NULL;
/* set new name to that of the interpreter */
NDINIT(&nd, LOOKUP, ISOPEN | LOCKLEAF | LOCKSHARED | 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);
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.
*/
error = (*p->p_sysent->sv_copyout_strings)(imgp, &stack_base);
if (error != 0) {
vn_lock(imgp->vp, LK_SHARED | LK_RETRY);
goto exec_fail_dealloc;
}
/*
* 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);
pdunshare(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_flag2 & P2_STKGAP_DISABLE_EXEC) == 0)
p->p_flag2 &= ~P2_STKGAP_DISABLE;
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);
}
if (imgp->sysent->sv_setid_allowed != NULL &&
!(*imgp->sysent->sv_setid_allowed)(td, imgp))
execve_nosetid(imgp);
/*
* 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);
+ kiop = NULL;
#ifdef KTRACE
- if (p->p_tracecred != NULL &&
- priv_check_cred(p->p_tracecred, PRIV_DEBUG_DIFFCRED))
- ktrprocexec(p, &tracecred, &tracevp);
+ kiop = ktrprocexec(p);
#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);
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);
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, stack_base);
VOP_MMAPPED(imgp->vp);
SDT_PROBE1(proc, , , exec__success, args->fname);
exec_fail_dealloc:
if (error != 0) {
p->p_osrel = orig_osrel;
p->p_fctl0 = orig_fctl0;
}
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 (imgp->textset)
VOP_UNSET_TEXT_CHECKED(imgp->vp);
if (error != 0)
vput(imgp->vp);
else
VOP_UNLOCK(imgp->vp);
}
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);
}
} 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
+ ktr_io_params_free(kiop);
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 */
exec_cleanup(td, oldvmspace);
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);
}
void
exec_cleanup(struct thread *td, struct vmspace *oldvmspace)
{
if ((td->td_pflags & TDP_EXECVMSPC) != 0) {
KASSERT(td->td_proc->p_vmspace != oldvmspace,
("oldvmspace still used"));
vmspace_free(oldvmspace);
td->td_pflags &= ~TDP_EXECVMSPC;
}
}
int
exec_map_first_page(struct image_params *imgp)
{
vm_object_t object;
vm_page_t m;
int error;
if (imgp->firstpage != NULL)
exec_unmap_first_page(imgp);
object = imgp->vp->v_object;
if (object == NULL)
return (EACCES);
#if VM_NRESERVLEVEL > 0
if ((object->flags & OBJ_COLORED) == 0) {
VM_OBJECT_WLOCK(object);
vm_object_color(object, 0);
VM_OBJECT_WUNLOCK(object);
}
#endif
error = vm_page_grab_valid_unlocked(&m, object, 0,
VM_ALLOC_COUNT(VM_INITIAL_PAGEIN) |
VM_ALLOC_NORMAL | VM_ALLOC_NOBUSY | VM_ALLOC_WIRED);
if (error != VM_PAGER_OK)
return (EIO);
imgp->firstpage = sf_buf_alloc(m, 0);
imgp->image_header = (char *)sf_buf_kva(imgp->firstpage);
return (0);
}
void
exec_unmap_first_page(struct image_params *imgp)
{
vm_page_t m;
if (imgp->firstpage != NULL) {
m = sf_buf_page(imgp->firstpage);
sf_buf_free(imgp->firstpage);
imgp->firstpage = NULL;
vm_page_unwire(m, PQ_ACTIVE);
}
}
/*
* Destroy old address space, and allocate a new stack.
* The new stack is only sgrowsiz large because it is grown
* automatically on a page fault.
*/
int
exec_new_vmspace(struct image_params *imgp, struct sysentvec *sv)
{
int error;
struct proc *p = imgp->proc;
struct vmspace *vmspace = p->p_vmspace;
struct thread *td = curthread;
vm_object_t obj;
struct rlimit rlim_stack;
vm_offset_t sv_minuser, stack_addr;
vm_map_t map;
vm_prot_t stack_prot;
u_long ssiz;
imgp->vmspace_destroyed = 1;
imgp->sysent = sv;
sigfastblock_clear(td);
umtx_exec(p);
itimers_exec(p);
if (sv->sv_onexec != NULL)
sv->sv_onexec(p, imgp);
EVENTHANDLER_DIRECT_INVOKE(process_exec, p, imgp);
/*
* Blow away entire process VM, if address space not shared,
* otherwise, create a new VM space so that other threads are
* not disrupted
*/
map = &vmspace->vm_map;
if (map_at_zero)
sv_minuser = sv->sv_minuser;
else
sv_minuser = MAX(sv->sv_minuser, PAGE_SIZE);
if (refcount_load(&vmspace->vm_refcnt) == 1 &&
vm_map_min(map) == sv_minuser &&
vm_map_max(map) == sv->sv_maxuser &&
cpu_exec_vmspace_reuse(p, map)) {
shmexit(vmspace);
pmap_remove_pages(vmspace_pmap(vmspace));
vm_map_remove(map, vm_map_min(map), vm_map_max(map));
/*
* An exec terminates mlockall(MCL_FUTURE).
* ASLR and W^X states must be re-evaluated.
*/
vm_map_lock(map);
vm_map_modflags(map, 0, MAP_WIREFUTURE | MAP_ASLR |
MAP_ASLR_IGNSTART | MAP_WXORX);
vm_map_unlock(map);
} else {
error = vmspace_exec(p, sv_minuser, sv->sv_maxuser);
if (error)
return (error);
vmspace = p->p_vmspace;
map = &vmspace->vm_map;
}
map->flags |= imgp->map_flags;
/* Map a shared page */
obj = sv->sv_shared_page_obj;
if (obj != NULL) {
vm_object_reference(obj);
error = vm_map_fixed(map, obj, 0,
sv->sv_shared_page_base, sv->sv_shared_page_len,
VM_PROT_READ | VM_PROT_EXECUTE,
VM_PROT_READ | VM_PROT_EXECUTE,
MAP_INHERIT_SHARE | MAP_ACC_NO_CHARGE);
if (error != KERN_SUCCESS) {
vm_object_deallocate(obj);
return (vm_mmap_to_errno(error));
}
}
/* Allocate a new stack */
if (imgp->stack_sz != 0) {
ssiz = trunc_page(imgp->stack_sz);
PROC_LOCK(p);
lim_rlimit_proc(p, RLIMIT_STACK, &rlim_stack);
PROC_UNLOCK(p);
if (ssiz > rlim_stack.rlim_max)
ssiz = rlim_stack.rlim_max;
if (ssiz > rlim_stack.rlim_cur) {
rlim_stack.rlim_cur = ssiz;
kern_setrlimit(curthread, RLIMIT_STACK, &rlim_stack);
}
} else if (sv->sv_maxssiz != NULL) {
ssiz = *sv->sv_maxssiz;
} else {
ssiz = maxssiz;
}
imgp->eff_stack_sz = lim_cur(curthread, RLIMIT_STACK);
if (ssiz < imgp->eff_stack_sz)
imgp->eff_stack_sz = ssiz;
stack_addr = sv->sv_usrstack - ssiz;
stack_prot = obj != NULL && imgp->stack_prot != 0 ?
imgp->stack_prot : sv->sv_stackprot;
error = vm_map_stack(map, stack_addr, (vm_size_t)ssiz, stack_prot,
VM_PROT_ALL, MAP_STACK_GROWS_DOWN);
if (error != KERN_SUCCESS) {
uprintf("exec_new_vmspace: mapping stack size %#jx prot %#x "
"failed mach error %d errno %d\n", (uintmax_t)ssiz,
stack_prot, error, vm_mmap_to_errno(error));
return (vm_mmap_to_errno(error));
}
/*
* vm_ssize and vm_maxsaddr are somewhat antiquated concepts, but they
* are still used to enforce the stack rlimit on the process stack.
*/
vmspace->vm_ssize = sgrowsiz >> PAGE_SHIFT;
vmspace->vm_maxsaddr = (char *)stack_addr;
return (0);
}
/*
* Copy out argument and environment strings from the old process address
* space into the temporary string buffer.
*/
int
exec_copyin_args(struct image_args *args, const char *fname,
enum uio_seg segflg, char **argv, char **envv)
{
u_long arg, env;
int error;
bzero(args, sizeof(*args));
if (argv == NULL)
return (EFAULT);
/*
* Allocate demand-paged memory for the file name, argument, and
* environment strings.
*/
error = exec_alloc_args(args);
if (error != 0)
return (error);
/*
* Copy the file name.
*/
error = exec_args_add_fname(args, fname, segflg);
if (error != 0)
goto err_exit;
/*
* extract arguments first
*/
for (;;) {
error = fueword(argv++, &arg);
if (error == -1) {
error = EFAULT;
goto err_exit;
}
if (arg == 0)
break;
error = exec_args_add_arg(args, (char *)(uintptr_t)arg,
UIO_USERSPACE);
if (error != 0)
goto err_exit;
}
/*
* extract environment strings
*/
if (envv) {
for (;;) {
error = fueword(envv++, &env);
if (error == -1) {
error = EFAULT;
goto err_exit;
}
if (env == 0)
break;
error = exec_args_add_env(args,
(char *)(uintptr_t)env, UIO_USERSPACE);
if (error != 0)
goto err_exit;
}
}
return (0);
err_exit:
exec_free_args(args);
return (error);
}
int
exec_copyin_data_fds(struct thread *td, struct image_args *args,
const void *data, size_t datalen, const int *fds, size_t fdslen)
{
struct filedesc *ofdp;
const char *p;
int *kfds;
int error;
memset(args, '\0', sizeof(*args));
ofdp = td->td_proc->p_fd;
if (datalen >= ARG_MAX || fdslen >= ofdp->fd_nfiles)
return (E2BIG);
error = exec_alloc_args(args);
if (error != 0)
return (error);
args->begin_argv = args->buf;
args->stringspace = ARG_MAX;
if (datalen > 0) {
/*
* Argument buffer has been provided. Copy it into the
* kernel as a single string and add a terminating null
* byte.
*/
error = copyin(data, args->begin_argv, datalen);
if (error != 0)
goto err_exit;
args->begin_argv[datalen] = '\0';
args->endp = args->begin_argv + datalen + 1;
args->stringspace -= datalen + 1;
/*
* Traditional argument counting. Count the number of
* null bytes.
*/
for (p = args->begin_argv; p < args->endp; ++p)
if (*p == '\0')
++args->argc;
} else {
/* No argument buffer provided. */
args->endp = args->begin_argv;
}
/* Create new file descriptor table. */
kfds = malloc(fdslen * sizeof(int), M_TEMP, M_WAITOK);
error = copyin(fds, kfds, fdslen * sizeof(int));
if (error != 0) {
free(kfds, M_TEMP);
goto err_exit;
}
error = fdcopy_remapped(ofdp, kfds, fdslen, &args->fdp);
free(kfds, M_TEMP);
if (error != 0)
goto err_exit;
return (0);
err_exit:
exec_free_args(args);
return (error);
}
struct exec_args_kva {
vm_offset_t addr;
u_int gen;
SLIST_ENTRY(exec_args_kva) next;
};
DPCPU_DEFINE_STATIC(struct exec_args_kva *, exec_args_kva);
static SLIST_HEAD(, exec_args_kva) exec_args_kva_freelist;
static struct mtx exec_args_kva_mtx;
static u_int exec_args_gen;
static void
exec_prealloc_args_kva(void *arg __unused)
{
struct exec_args_kva *argkva;
u_int i;
SLIST_INIT(&exec_args_kva_freelist);
mtx_init(&exec_args_kva_mtx, "exec args kva", NULL, MTX_DEF);
for (i = 0; i < exec_map_entries; i++) {
argkva = malloc(sizeof(*argkva), M_PARGS, M_WAITOK);
argkva->addr = kmap_alloc_wait(exec_map, exec_map_entry_size);
argkva->gen = exec_args_gen;
SLIST_INSERT_HEAD(&exec_args_kva_freelist, argkva, next);
}
}
SYSINIT(exec_args_kva, SI_SUB_EXEC, SI_ORDER_ANY, exec_prealloc_args_kva, NULL);
static vm_offset_t
exec_alloc_args_kva(void **cookie)
{
struct exec_args_kva *argkva;
argkva = (void *)atomic_readandclear_ptr(
(uintptr_t *)DPCPU_PTR(exec_args_kva));
if (argkva == NULL) {
mtx_lock(&exec_args_kva_mtx);
while ((argkva = SLIST_FIRST(&exec_args_kva_freelist)) == NULL)
(void)mtx_sleep(&exec_args_kva_freelist,
&exec_args_kva_mtx, 0, "execkva", 0);
SLIST_REMOVE_HEAD(&exec_args_kva_freelist, next);
mtx_unlock(&exec_args_kva_mtx);
}
*(struct exec_args_kva **)cookie = argkva;
return (argkva->addr);
}
static void
exec_release_args_kva(struct exec_args_kva *argkva, u_int gen)
{
vm_offset_t base;
base = argkva->addr;
if (argkva->gen != gen) {
(void)vm_map_madvise(exec_map, base, base + exec_map_entry_size,
MADV_FREE);
argkva->gen = gen;
}
if (!atomic_cmpset_ptr((uintptr_t *)DPCPU_PTR(exec_args_kva),
(uintptr_t)NULL, (uintptr_t)argkva)) {
mtx_lock(&exec_args_kva_mtx);
SLIST_INSERT_HEAD(&exec_args_kva_freelist, argkva, next);
wakeup_one(&exec_args_kva_freelist);
mtx_unlock(&exec_args_kva_mtx);
}
}
static void
exec_free_args_kva(void *cookie)
{
exec_release_args_kva(cookie, exec_args_gen);
}
static void
exec_args_kva_lowmem(void *arg __unused)
{
SLIST_HEAD(, exec_args_kva) head;
struct exec_args_kva *argkva;
u_int gen;
int i;
gen = atomic_fetchadd_int(&exec_args_gen, 1) + 1;
/*
* Force an madvise of each KVA range. Any currently allocated ranges
* will have MADV_FREE applied once they are freed.
*/
SLIST_INIT(&head);
mtx_lock(&exec_args_kva_mtx);
SLIST_SWAP(&head, &exec_args_kva_freelist, exec_args_kva);
mtx_unlock(&exec_args_kva_mtx);
while ((argkva = SLIST_FIRST(&head)) != NULL) {
SLIST_REMOVE_HEAD(&head, next);
exec_release_args_kva(argkva, gen);
}
CPU_FOREACH(i) {
argkva = (void *)atomic_readandclear_ptr(
(uintptr_t *)DPCPU_ID_PTR(i, exec_args_kva));
if (argkva != NULL)
exec_release_args_kva(argkva, gen);
}
}
EVENTHANDLER_DEFINE(vm_lowmem, exec_args_kva_lowmem, NULL,
EVENTHANDLER_PRI_ANY);
/*
* Allocate temporary demand-paged, zero-filled memory for the file name,
* argument, and environment strings.
*/
int
exec_alloc_args(struct image_args *args)
{
args->buf = (char *)exec_alloc_args_kva(&args->bufkva);
return (0);
}
void
exec_free_args(struct image_args *args)
{
if (args->buf != NULL) {
exec_free_args_kva(args->bufkva);
args->buf = NULL;
}
if (args->fname_buf != NULL) {
free(args->fname_buf, M_TEMP);
args->fname_buf = NULL;
}
if (args->fdp != NULL)
fdescfree_remapped(args->fdp);
}
/*
* A set to functions to fill struct image args.
*
* NOTE: exec_args_add_fname() must be called (possibly with a NULL
* fname) before the other functions. All exec_args_add_arg() calls must
* be made before any exec_args_add_env() calls. exec_args_adjust_args()
* may be called any time after exec_args_add_fname().
*
* exec_args_add_fname() - install path to be executed
* exec_args_add_arg() - append an argument string
* exec_args_add_env() - append an env string
* exec_args_adjust_args() - adjust location of the argument list to
* allow new arguments to be prepended
*/
int
exec_args_add_fname(struct image_args *args, const char *fname,
enum uio_seg segflg)
{
int error;
size_t length;
KASSERT(args->fname == NULL, ("fname already appended"));
KASSERT(args->endp == NULL, ("already appending to args"));
if (fname != NULL) {
args->fname = args->buf;
error = segflg == UIO_SYSSPACE ?
copystr(fname, args->fname, PATH_MAX, &length) :
copyinstr(fname, args->fname, PATH_MAX, &length);
if (error != 0)
return (error == ENAMETOOLONG ? E2BIG : error);
} else
length = 0;
/* Set up for _arg_*()/_env_*() */
args->endp = args->buf + length;
/* begin_argv must be set and kept updated */
args->begin_argv = args->endp;
KASSERT(exec_map_entry_size - length >= ARG_MAX,
("too little space remaining for arguments %zu < %zu",
exec_map_entry_size - length, (size_t)ARG_MAX));
args->stringspace = ARG_MAX;
return (0);
}
static int
exec_args_add_str(struct image_args *args, const char *str,
enum uio_seg segflg, int *countp)
{
int error;
size_t length;
KASSERT(args->endp != NULL, ("endp not initialized"));
KASSERT(args->begin_argv != NULL, ("begin_argp not initialized"));
error = (segflg == UIO_SYSSPACE) ?
copystr(str, args->endp, args->stringspace, &length) :
copyinstr(str, args->endp, args->stringspace, &length);
if (error != 0)
return (error == ENAMETOOLONG ? E2BIG : error);
args->stringspace -= length;
args->endp += length;
(*countp)++;
return (0);
}
int
exec_args_add_arg(struct image_args *args, const char *argp,
enum uio_seg segflg)
{
KASSERT(args->envc == 0, ("appending args after env"));
return (exec_args_add_str(args, argp, segflg, &args->argc));
}
int
exec_args_add_env(struct image_args *args, const char *envp,
enum uio_seg segflg)
{
if (args->envc == 0)
args->begin_envv = args->endp;
return (exec_args_add_str(args, envp, segflg, &args->envc));
}
int
exec_args_adjust_args(struct image_args *args, size_t consume, ssize_t extend)
{
ssize_t offset;
KASSERT(args->endp != NULL, ("endp not initialized"));
KASSERT(args->begin_argv != NULL, ("begin_argp not initialized"));
offset = extend - consume;
if (args->stringspace < offset)
return (E2BIG);
memmove(args->begin_argv + extend, args->begin_argv + consume,
args->endp - args->begin_argv + consume);
if (args->envc > 0)
args->begin_envv += offset;
args->endp += offset;
args->stringspace -= offset;
return (0);
}
char *
exec_args_get_begin_envv(struct image_args *args)
{
KASSERT(args->endp != NULL, ("endp not initialized"));
if (args->envc > 0)
return (args->begin_envv);
return (args->endp);
}
void
exec_stackgap(struct image_params *imgp, uintptr_t *dp)
{
if (imgp->sysent->sv_stackgap == NULL ||
(imgp->proc->p_fctl0 & (NT_FREEBSD_FCTL_ASLR_DISABLE |
NT_FREEBSD_FCTL_ASG_DISABLE)) != 0 ||
(imgp->map_flags & MAP_ASLR) == 0)
return;
imgp->sysent->sv_stackgap(imgp, dp);
}
/*
* Copy strings out to the new process address space, constructing new arg
* and env vector tables. Return a pointer to the base so that it can be used
* as the initial stack pointer.
*/
int
exec_copyout_strings(struct image_params *imgp, uintptr_t *stack_base)
{
int argc, envc;
char **vectp;
char *stringp;
uintptr_t destp, ustringp;
struct ps_strings *arginfo;
struct proc *p;
size_t execpath_len;
int error, szsigcode, szps;
char canary[sizeof(long) * 8];
szps = sizeof(pagesizes[0]) * MAXPAGESIZES;
/*
* Calculate string base and vector table pointers.
* Also deal with signal trampoline code for this exec type.
*/
if (imgp->execpath != NULL && imgp->auxargs != NULL)
execpath_len = strlen(imgp->execpath) + 1;
else
execpath_len = 0;
p = imgp->proc;
szsigcode = 0;
arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
imgp->ps_strings = arginfo;
if (p->p_sysent->sv_sigcode_base == 0) {
if (p->p_sysent->sv_szsigcode != NULL)
szsigcode = *(p->p_sysent->sv_szsigcode);
}
destp = (uintptr_t)arginfo;
/*
* install sigcode
*/
if (szsigcode != 0) {
destp -= szsigcode;
destp = rounddown2(destp, sizeof(void *));
error = copyout(p->p_sysent->sv_sigcode, (void *)destp,
szsigcode);
if (error != 0)
return (error);
}
/*
* Copy the image path for the rtld.
*/
if (execpath_len != 0) {
destp -= execpath_len;
destp = rounddown2(destp, sizeof(void *));
imgp->execpathp = (void *)destp;
error = copyout(imgp->execpath, imgp->execpathp, execpath_len);
if (error != 0)
return (error);
}
/*
* Prepare the canary for SSP.
*/
arc4rand(canary, sizeof(canary), 0);
destp -= sizeof(canary);
imgp->canary = (void *)destp;
error = copyout(canary, imgp->canary, sizeof(canary));
if (error != 0)
return (error);
imgp->canarylen = sizeof(canary);
/*
* Prepare the pagesizes array.
*/
destp -= szps;
destp = rounddown2(destp, sizeof(void *));
imgp->pagesizes = (void *)destp;
error = copyout(pagesizes, imgp->pagesizes, szps);
if (error != 0)
return (error);
imgp->pagesizeslen = szps;
/*
* Allocate room for the argument and environment strings.
*/
destp -= ARG_MAX - imgp->args->stringspace;
destp = rounddown2(destp, sizeof(void *));
ustringp = destp;
exec_stackgap(imgp, &destp);
if (imgp->auxargs) {
/*
* Allocate room on the stack for the ELF auxargs
* array. It has up to AT_COUNT entries.
*/
destp -= AT_COUNT * sizeof(Elf_Auxinfo);
destp = rounddown2(destp, sizeof(void *));
}
vectp = (char **)destp;
/*
* Allocate room for the argv[] and env vectors including the
* terminating NULL pointers.
*/
vectp -= imgp->args->argc + 1 + imgp->args->envc + 1;
/*
* vectp also becomes our initial stack base
*/
*stack_base = (uintptr_t)vectp;
stringp = imgp->args->begin_argv;
argc = imgp->args->argc;
envc = imgp->args->envc;
/*
* Copy out strings - arguments and environment.
*/
error = copyout(stringp, (void *)ustringp,
ARG_MAX - imgp->args->stringspace);
if (error != 0)
return (error);
/*
* Fill in "ps_strings" struct for ps, w, etc.
*/
imgp->argv = vectp;
if (suword(&arginfo->ps_argvstr, (long)(intptr_t)vectp) != 0 ||
suword32(&arginfo->ps_nargvstr, argc) != 0)
return (EFAULT);
/*
* Fill in argument portion of vector table.
*/
for (; argc > 0; --argc) {
if (suword(vectp++, ustringp) != 0)
return (EFAULT);
while (*stringp++ != 0)
ustringp++;
ustringp++;
}
/* a null vector table pointer separates the argp's from the envp's */
if (suword(vectp++, 0) != 0)
return (EFAULT);
imgp->envv = vectp;
if (suword(&arginfo->ps_envstr, (long)(intptr_t)vectp) != 0 ||
suword32(&arginfo->ps_nenvstr, envc) != 0)
return (EFAULT);
/*
* Fill in environment portion of vector table.
*/
for (; envc > 0; --envc) {
if (suword(vectp++, ustringp) != 0)
return (EFAULT);
while (*stringp++ != 0)
ustringp++;
ustringp++;
}
/* end of vector table is a null pointer */
if (suword(vectp, 0) != 0)
return (EFAULT);
if (imgp->auxargs) {
vectp++;
error = imgp->sysent->sv_copyout_auxargs(imgp,
(uintptr_t)vectp);
if (error != 0)
return (error);
}
return (0);
}
/*
* Check permissions of file to execute.
* Called with imgp->vp locked.
* Return 0 for success or error code on failure.
*/
int
exec_check_permissions(struct image_params *imgp)
{
struct vnode *vp = imgp->vp;
struct vattr *attr = imgp->attr;
struct thread *td;
int error;
td = curthread;
/* Get file attributes */
error = VOP_GETATTR(vp, attr, td->td_ucred);
if (error)
return (error);
#ifdef MAC
error = mac_vnode_check_exec(td->td_ucred, imgp->vp, imgp);
if (error)
return (error);
#endif
/*
* 1) Check if file execution is disabled for the filesystem that
* this file resides on.
* 2) Ensure that at least one execute bit is on. Otherwise, a
* privileged user will always succeed, and we don't want this
* to happen unless the file really is executable.
* 3) Ensure that the file is a regular file.
*/
if ((vp->v_mount->mnt_flag & MNT_NOEXEC) ||
(attr->va_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0 ||
(attr->va_type != VREG))
return (EACCES);
/*
* Zero length files can't be exec'd
*/
if (attr->va_size == 0)
return (ENOEXEC);
/*
* Check for execute permission to file based on current credentials.
*/
error = VOP_ACCESS(vp, VEXEC, td->td_ucred, td);
if (error)
return (error);
/*
* Check number of open-for-writes on the file and deny execution
* if there are any.
*
* Add a text reference 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.
*/
error = VOP_SET_TEXT(vp);
if (error != 0)
return (error);
imgp->textset = true;
/*
* Call filesystem specific open routine (which does nothing in the
* general case).
*/
error = VOP_OPEN(vp, FREAD, td->td_ucred, td, NULL);
if (error == 0)
imgp->opened = 1;
return (error);
}
/*
* Exec handler registration
*/
int
exec_register(const struct execsw *execsw_arg)
{
const struct execsw **es, **xs, **newexecsw;
u_int count = 2; /* New slot and trailing NULL */
if (execsw)
for (es = execsw; *es; es++)
count++;
newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
xs = newexecsw;
if (execsw)
for (es = execsw; *es; es++)
*xs++ = *es;
*xs++ = execsw_arg;
*xs = NULL;
if (execsw)
free(execsw, M_TEMP);
execsw = newexecsw;
return (0);
}
int
exec_unregister(const struct execsw *execsw_arg)
{
const struct execsw **es, **xs, **newexecsw;
int count = 1;
if (execsw == NULL)
panic("unregister with no handlers left?\n");
for (es = execsw; *es; es++) {
if (*es == execsw_arg)
break;
}
if (*es == NULL)
return (ENOENT);
for (es = execsw; *es; es++)
if (*es != execsw_arg)
count++;
newexecsw = malloc(count * sizeof(*es), M_TEMP, M_WAITOK);
xs = newexecsw;
for (es = execsw; *es; es++)
if (*es != execsw_arg)
*xs++ = *es;
*xs = NULL;
if (execsw)
free(execsw, M_TEMP);
execsw = newexecsw;
return (0);
}
diff --git a/sys/kern/kern_ktrace.c b/sys/kern/kern_ktrace.c
index 26fba786e1e9..c923149ed129 100644
--- a/sys/kern/kern_ktrace.c
+++ b/sys/kern/kern_ktrace.c
@@ -1,1316 +1,1378 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1989, 1993
* The Regents of the University of California.
* Copyright (c) 2005 Robert N. M. Watson
* All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)kern_ktrace.c 8.2 (Berkeley) 9/23/93
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_ktrace.h"
#include <sys/param.h>
#include <sys/capsicum.h>
#include <sys/systm.h>
#include <sys/fcntl.h>
#include <sys/kernel.h>
#include <sys/kthread.h>
#include <sys/lock.h>
#include <sys/mutex.h>
#include <sys/malloc.h>
#include <sys/mount.h>
#include <sys/namei.h>
#include <sys/priv.h>
#include <sys/proc.h>
#include <sys/unistd.h>
#include <sys/vnode.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <sys/ktrace.h>
#include <sys/sx.h>
#include <sys/sysctl.h>
#include <sys/sysent.h>
#include <sys/syslog.h>
#include <sys/sysproto.h>
#include <security/mac/mac_framework.h>
/*
* The ktrace facility allows the tracing of certain key events in user space
* processes, such as system calls, signal delivery, context switches, and
* user generated events using utrace(2). It works by streaming event
* records and data to a vnode associated with the process using the
* ktrace(2) system call. In general, records can be written directly from
* the context that generates the event. One important exception to this is
* during a context switch, where sleeping is not permitted. To handle this
* case, trace events are generated using in-kernel ktr_request records, and
* then delivered to disk at a convenient moment -- either immediately, the
* next traceable event, at system call return, or at process exit.
*
* When dealing with multiple threads or processes writing to the same event
* log, ordering guarantees are weak: specifically, if an event has multiple
* records (i.e., system call enter and return), they may be interlaced with
* records from another event. Process and thread ID information is provided
* in the record, and user applications can de-interlace events if required.
*/
static MALLOC_DEFINE(M_KTRACE, "KTRACE", "KTRACE");
#ifdef KTRACE
FEATURE(ktrace, "Kernel support for system-call tracing");
#ifndef KTRACE_REQUEST_POOL
#define KTRACE_REQUEST_POOL 100
#endif
struct ktr_request {
struct ktr_header ktr_header;
void *ktr_buffer;
union {
struct ktr_proc_ctor ktr_proc_ctor;
struct ktr_cap_fail ktr_cap_fail;
struct ktr_syscall ktr_syscall;
struct ktr_sysret ktr_sysret;
struct ktr_genio ktr_genio;
struct ktr_psig ktr_psig;
struct ktr_csw ktr_csw;
struct ktr_fault ktr_fault;
struct ktr_faultend ktr_faultend;
struct ktr_struct_array ktr_struct_array;
} ktr_data;
STAILQ_ENTRY(ktr_request) ktr_list;
};
static int data_lengths[] = {
[KTR_SYSCALL] = offsetof(struct ktr_syscall, ktr_args),
[KTR_SYSRET] = sizeof(struct ktr_sysret),
[KTR_NAMEI] = 0,
[KTR_GENIO] = sizeof(struct ktr_genio),
[KTR_PSIG] = sizeof(struct ktr_psig),
[KTR_CSW] = sizeof(struct ktr_csw),
[KTR_USER] = 0,
[KTR_STRUCT] = 0,
[KTR_SYSCTL] = 0,
[KTR_PROCCTOR] = sizeof(struct ktr_proc_ctor),
[KTR_PROCDTOR] = 0,
[KTR_CAPFAIL] = sizeof(struct ktr_cap_fail),
[KTR_FAULT] = sizeof(struct ktr_fault),
[KTR_FAULTEND] = sizeof(struct ktr_faultend),
[KTR_STRUCT_ARRAY] = sizeof(struct ktr_struct_array),
};
static STAILQ_HEAD(, ktr_request) ktr_free;
static SYSCTL_NODE(_kern, OID_AUTO, ktrace, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
"KTRACE options");
static u_int ktr_requestpool = KTRACE_REQUEST_POOL;
TUNABLE_INT("kern.ktrace.request_pool", &ktr_requestpool);
u_int ktr_geniosize = PAGE_SIZE;
SYSCTL_UINT(_kern_ktrace, OID_AUTO, genio_size, CTLFLAG_RWTUN, &ktr_geniosize,
0, "Maximum size of genio event payload");
static int print_message = 1;
static struct mtx ktrace_mtx;
static struct sx ktrace_sx;
+struct ktr_io_params {
+ struct vnode *vp;
+ struct ucred *cr;
+ u_int refs;
+};
+
static void ktrace_init(void *dummy);
static int sysctl_kern_ktrace_request_pool(SYSCTL_HANDLER_ARGS);
static u_int ktrace_resize_pool(u_int oldsize, u_int newsize);
static struct ktr_request *ktr_getrequest_entered(struct thread *td, int type);
static struct ktr_request *ktr_getrequest(int type);
static void ktr_submitrequest(struct thread *td, struct ktr_request *req);
-static void ktr_freeproc(struct proc *p, struct ucred **uc,
- struct vnode **vp);
+static struct ktr_io_params *ktr_freeproc(struct proc *p);
static void ktr_freerequest(struct ktr_request *req);
static void ktr_freerequest_locked(struct ktr_request *req);
static void ktr_writerequest(struct thread *td, struct ktr_request *req);
static int ktrcanset(struct thread *,struct proc *);
-static int ktrsetchildren(struct thread *,struct proc *,int,int,struct vnode *);
-static int ktrops(struct thread *,struct proc *,int,int,struct vnode *);
+static int ktrsetchildren(struct thread *, struct proc *, int, int,
+ struct ktr_io_params *);
+static int ktrops(struct thread *, struct proc *, int, int,
+ struct ktr_io_params *);
static void ktrprocctor_entered(struct thread *, struct proc *);
/*
* ktrace itself generates events, such as context switches, which we do not
* wish to trace. Maintain a flag, TDP_INKTRACE, on each thread to determine
* whether or not it is in a region where tracing of events should be
* suppressed.
*/
static void
ktrace_enter(struct thread *td)
{
KASSERT(!(td->td_pflags & TDP_INKTRACE), ("ktrace_enter: flag set"));
td->td_pflags |= TDP_INKTRACE;
}
static void
ktrace_exit(struct thread *td)
{
KASSERT(td->td_pflags & TDP_INKTRACE, ("ktrace_exit: flag not set"));
td->td_pflags &= ~TDP_INKTRACE;
}
static void
ktrace_assert(struct thread *td)
{
KASSERT(td->td_pflags & TDP_INKTRACE, ("ktrace_assert: flag not set"));
}
static void
ktrace_init(void *dummy)
{
struct ktr_request *req;
int i;
mtx_init(&ktrace_mtx, "ktrace", NULL, MTX_DEF | MTX_QUIET);
sx_init(&ktrace_sx, "ktrace_sx");
STAILQ_INIT(&ktr_free);
for (i = 0; i < ktr_requestpool; i++) {
req = malloc(sizeof(struct ktr_request), M_KTRACE, M_WAITOK);
STAILQ_INSERT_HEAD(&ktr_free, req, ktr_list);
}
}
SYSINIT(ktrace_init, SI_SUB_KTRACE, SI_ORDER_ANY, ktrace_init, NULL);
static int
sysctl_kern_ktrace_request_pool(SYSCTL_HANDLER_ARGS)
{
struct thread *td;
u_int newsize, oldsize, wantsize;
int error;
/* Handle easy read-only case first to avoid warnings from GCC. */
if (!req->newptr) {
oldsize = ktr_requestpool;
return (SYSCTL_OUT(req, &oldsize, sizeof(u_int)));
}
error = SYSCTL_IN(req, &wantsize, sizeof(u_int));
if (error)
return (error);
td = curthread;
ktrace_enter(td);
oldsize = ktr_requestpool;
newsize = ktrace_resize_pool(oldsize, wantsize);
ktrace_exit(td);
error = SYSCTL_OUT(req, &oldsize, sizeof(u_int));
if (error)
return (error);
if (wantsize > oldsize && newsize < wantsize)
return (ENOSPC);
return (0);
}
SYSCTL_PROC(_kern_ktrace, OID_AUTO, request_pool,
CTLTYPE_UINT | CTLFLAG_RW | CTLFLAG_NEEDGIANT, &ktr_requestpool, 0,
sysctl_kern_ktrace_request_pool, "IU",
"Pool buffer size for ktrace(1)");
static u_int
ktrace_resize_pool(u_int oldsize, u_int newsize)
{
STAILQ_HEAD(, ktr_request) ktr_new;
struct ktr_request *req;
int bound;
print_message = 1;
bound = newsize - oldsize;
if (bound == 0)
return (ktr_requestpool);
if (bound < 0) {
mtx_lock(&ktrace_mtx);
/* Shrink pool down to newsize if possible. */
while (bound++ < 0) {
req = STAILQ_FIRST(&ktr_free);
if (req == NULL)
break;
STAILQ_REMOVE_HEAD(&ktr_free, ktr_list);
ktr_requestpool--;
free(req, M_KTRACE);
}
} else {
/* Grow pool up to newsize. */
STAILQ_INIT(&ktr_new);
while (bound-- > 0) {
req = malloc(sizeof(struct ktr_request), M_KTRACE,
M_WAITOK);
STAILQ_INSERT_HEAD(&ktr_new, req, ktr_list);
}
mtx_lock(&ktrace_mtx);
STAILQ_CONCAT(&ktr_free, &ktr_new);
ktr_requestpool += (newsize - oldsize);
}
mtx_unlock(&ktrace_mtx);
return (ktr_requestpool);
}
/* ktr_getrequest() assumes that ktr_comm[] is the same size as td_name[]. */
CTASSERT(sizeof(((struct ktr_header *)NULL)->ktr_comm) ==
(sizeof((struct thread *)NULL)->td_name));
static struct ktr_request *
ktr_getrequest_entered(struct thread *td, int type)
{
struct ktr_request *req;
struct proc *p = td->td_proc;
int pm;
mtx_lock(&ktrace_mtx);
if (!KTRCHECK(td, type)) {
mtx_unlock(&ktrace_mtx);
return (NULL);
}
req = STAILQ_FIRST(&ktr_free);
if (req != NULL) {
STAILQ_REMOVE_HEAD(&ktr_free, ktr_list);
req->ktr_header.ktr_type = type;
if (p->p_traceflag & KTRFAC_DROP) {
req->ktr_header.ktr_type |= KTR_DROP;
p->p_traceflag &= ~KTRFAC_DROP;
}
mtx_unlock(&ktrace_mtx);
microtime(&req->ktr_header.ktr_time);
req->ktr_header.ktr_pid = p->p_pid;
req->ktr_header.ktr_tid = td->td_tid;
bcopy(td->td_name, req->ktr_header.ktr_comm,
sizeof(req->ktr_header.ktr_comm));
req->ktr_buffer = NULL;
req->ktr_header.ktr_len = 0;
} else {
p->p_traceflag |= KTRFAC_DROP;
pm = print_message;
print_message = 0;
mtx_unlock(&ktrace_mtx);
if (pm)
printf("Out of ktrace request objects.\n");
}
return (req);
}
static struct ktr_request *
ktr_getrequest(int type)
{
struct thread *td = curthread;
struct ktr_request *req;
ktrace_enter(td);
req = ktr_getrequest_entered(td, type);
if (req == NULL)
ktrace_exit(td);
return (req);
}
/*
* Some trace generation environments don't permit direct access to VFS,
* such as during a context switch where sleeping is not allowed. Under these
* circumstances, queue a request to the thread to be written asynchronously
* later.
*/
static void
ktr_enqueuerequest(struct thread *td, struct ktr_request *req)
{
mtx_lock(&ktrace_mtx);
STAILQ_INSERT_TAIL(&td->td_proc->p_ktr, req, ktr_list);
mtx_unlock(&ktrace_mtx);
thread_lock(td);
td->td_flags |= TDF_ASTPENDING;
thread_unlock(td);
}
/*
* Drain any pending ktrace records from the per-thread queue to disk. This
* is used both internally before committing other records, and also on
* system call return. We drain all the ones we can find at the time when
* drain is requested, but don't keep draining after that as those events
* may be approximately "after" the current event.
*/
static void
ktr_drain(struct thread *td)
{
struct ktr_request *queued_req;
STAILQ_HEAD(, ktr_request) local_queue;
ktrace_assert(td);
sx_assert(&ktrace_sx, SX_XLOCKED);
STAILQ_INIT(&local_queue);
if (!STAILQ_EMPTY(&td->td_proc->p_ktr)) {
mtx_lock(&ktrace_mtx);
STAILQ_CONCAT(&local_queue, &td->td_proc->p_ktr);
mtx_unlock(&ktrace_mtx);
while ((queued_req = STAILQ_FIRST(&local_queue))) {
STAILQ_REMOVE_HEAD(&local_queue, ktr_list);
ktr_writerequest(td, queued_req);
ktr_freerequest(queued_req);
}
}
}
/*
* Submit a trace record for immediate commit to disk -- to be used only
* where entering VFS is OK. First drain any pending records that may have
* been cached in the thread.
*/
static void
ktr_submitrequest(struct thread *td, struct ktr_request *req)
{
ktrace_assert(td);
sx_xlock(&ktrace_sx);
ktr_drain(td);
ktr_writerequest(td, req);
ktr_freerequest(req);
sx_xunlock(&ktrace_sx);
ktrace_exit(td);
}
static void
ktr_freerequest(struct ktr_request *req)
{
mtx_lock(&ktrace_mtx);
ktr_freerequest_locked(req);
mtx_unlock(&ktrace_mtx);
}
static void
ktr_freerequest_locked(struct ktr_request *req)
{
mtx_assert(&ktrace_mtx, MA_OWNED);
if (req->ktr_buffer != NULL)
free(req->ktr_buffer, M_KTRACE);
STAILQ_INSERT_HEAD(&ktr_free, req, ktr_list);
}
+static void
+ktr_io_params_ref(struct ktr_io_params *kiop)
+{
+ mtx_assert(&ktrace_mtx, MA_OWNED);
+ kiop->refs++;
+}
+
+static struct ktr_io_params *
+ktr_io_params_rele(struct ktr_io_params *kiop)
+{
+ mtx_assert(&ktrace_mtx, MA_OWNED);
+ if (kiop == NULL)
+ return (NULL);
+ KASSERT(kiop->refs > 0, ("kiop ref == 0 %p", kiop));
+ return (--(kiop->refs) == 0 ? kiop : NULL);
+}
+
+void
+ktr_io_params_free(struct ktr_io_params *kiop)
+{
+ if (kiop == NULL)
+ return;
+
+ MPASS(kiop->refs == 0);
+ vn_close(kiop->vp, FWRITE, kiop->cr, curthread);
+ crfree(kiop->cr);
+ free(kiop, M_KTRACE);
+}
+
+static struct ktr_io_params *
+ktr_io_params_alloc(struct thread *td, struct vnode *vp)
+{
+ struct ktr_io_params *res;
+
+ res = malloc(sizeof(struct ktr_io_params), M_KTRACE, M_WAITOK);
+ res->vp = vp;
+ res->cr = crhold(td->td_ucred);
+ res->refs = 1;
+ return (res);
+}
+
/*
* Disable tracing for a process and release all associated resources.
* The caller is responsible for releasing a reference on the returned
* vnode and credentials.
*/
-static void
-ktr_freeproc(struct proc *p, struct ucred **uc, struct vnode **vp)
+static struct ktr_io_params *
+ktr_freeproc(struct proc *p)
{
+ struct ktr_io_params *kiop;
struct ktr_request *req;
PROC_LOCK_ASSERT(p, MA_OWNED);
mtx_assert(&ktrace_mtx, MA_OWNED);
- *uc = p->p_tracecred;
- p->p_tracecred = NULL;
- if (vp != NULL)
- *vp = p->p_tracevp;
- p->p_tracevp = NULL;
+ kiop = ktr_io_params_rele(p->p_ktrioparms);
+ p->p_ktrioparms = NULL;
p->p_traceflag = 0;
while ((req = STAILQ_FIRST(&p->p_ktr)) != NULL) {
STAILQ_REMOVE_HEAD(&p->p_ktr, ktr_list);
ktr_freerequest_locked(req);
}
+ return (kiop);
+}
+
+struct vnode *
+ktr_get_tracevp(struct proc *p, bool ref)
+{
+ struct vnode *vp;
+
+ PROC_LOCK_ASSERT(p, MA_OWNED);
+
+ if (p->p_ktrioparms != NULL) {
+ vp = p->p_ktrioparms->vp;
+ if (ref)
+ vrefact(vp);
+ } else {
+ vp = NULL;
+ }
+ return (vp);
}
void
ktrsyscall(int code, int narg, register_t args[])
{
struct ktr_request *req;
struct ktr_syscall *ktp;
size_t buflen;
char *buf = NULL;
if (__predict_false(curthread->td_pflags & TDP_INKTRACE))
return;
buflen = sizeof(register_t) * narg;
if (buflen > 0) {
buf = malloc(buflen, M_KTRACE, M_WAITOK);
bcopy(args, buf, buflen);
}
req = ktr_getrequest(KTR_SYSCALL);
if (req == NULL) {
if (buf != NULL)
free(buf, M_KTRACE);
return;
}
ktp = &req->ktr_data.ktr_syscall;
ktp->ktr_code = code;
ktp->ktr_narg = narg;
if (buflen > 0) {
req->ktr_header.ktr_len = buflen;
req->ktr_buffer = buf;
}
ktr_submitrequest(curthread, req);
}
void
ktrsysret(int code, int error, register_t retval)
{
struct ktr_request *req;
struct ktr_sysret *ktp;
if (__predict_false(curthread->td_pflags & TDP_INKTRACE))
return;
req = ktr_getrequest(KTR_SYSRET);
if (req == NULL)
return;
ktp = &req->ktr_data.ktr_sysret;
ktp->ktr_code = code;
ktp->ktr_error = error;
ktp->ktr_retval = ((error == 0) ? retval: 0); /* what about val2 ? */
ktr_submitrequest(curthread, req);
}
/*
* When a setuid process execs, disable tracing.
*
* XXX: We toss any pending asynchronous records.
*/
-void
-ktrprocexec(struct proc *p, struct ucred **uc, struct vnode **vp)
+struct ktr_io_params *
+ktrprocexec(struct proc *p)
{
+ struct ktr_io_params *kiop;
PROC_LOCK_ASSERT(p, MA_OWNED);
+
+ kiop = p->p_ktrioparms;
+ if (kiop == NULL || priv_check_cred(kiop->cr, PRIV_DEBUG_DIFFCRED))
+ return (NULL);
+
mtx_lock(&ktrace_mtx);
- ktr_freeproc(p, uc, vp);
+ kiop = ktr_freeproc(p);
mtx_unlock(&ktrace_mtx);
+ return (kiop);
}
/*
* When a process exits, drain per-process asynchronous trace records
* and disable tracing.
*/
void
ktrprocexit(struct thread *td)
{
struct ktr_request *req;
struct proc *p;
- struct ucred *cred;
- struct vnode *vp;
+ struct ktr_io_params *kiop;
p = td->td_proc;
if (p->p_traceflag == 0)
return;
ktrace_enter(td);
req = ktr_getrequest_entered(td, KTR_PROCDTOR);
if (req != NULL)
ktr_enqueuerequest(td, req);
sx_xlock(&ktrace_sx);
ktr_drain(td);
sx_xunlock(&ktrace_sx);
PROC_LOCK(p);
mtx_lock(&ktrace_mtx);
- ktr_freeproc(p, &cred, &vp);
+ kiop = ktr_freeproc(p);
mtx_unlock(&ktrace_mtx);
PROC_UNLOCK(p);
- if (vp != NULL)
- vrele(vp);
- if (cred != NULL)
- crfree(cred);
+ ktr_io_params_free(kiop);
ktrace_exit(td);
}
static void
ktrprocctor_entered(struct thread *td, struct proc *p)
{
struct ktr_proc_ctor *ktp;
struct ktr_request *req;
struct thread *td2;
ktrace_assert(td);
td2 = FIRST_THREAD_IN_PROC(p);
req = ktr_getrequest_entered(td2, KTR_PROCCTOR);
if (req == NULL)
return;
ktp = &req->ktr_data.ktr_proc_ctor;
ktp->sv_flags = p->p_sysent->sv_flags;
ktr_enqueuerequest(td2, req);
}
void
ktrprocctor(struct proc *p)
{
struct thread *td = curthread;
if ((p->p_traceflag & KTRFAC_MASK) == 0)
return;
ktrace_enter(td);
ktrprocctor_entered(td, p);
ktrace_exit(td);
}
/*
* When a process forks, enable tracing in the new process if needed.
*/
void
ktrprocfork(struct proc *p1, struct proc *p2)
{
- MPASS(p2->p_tracevp == NULL);
+ MPASS(p2->p_ktrioparms == NULL);
MPASS(p2->p_traceflag == 0);
if (p1->p_traceflag == 0)
return;
PROC_LOCK(p1);
mtx_lock(&ktrace_mtx);
if (p1->p_traceflag & KTRFAC_INHERIT) {
p2->p_traceflag = p1->p_traceflag;
- if ((p2->p_tracevp = p1->p_tracevp) != NULL) {
- VREF(p2->p_tracevp);
- KASSERT(p1->p_tracecred != NULL,
- ("ktrace vnode with no cred"));
- p2->p_tracecred = crhold(p1->p_tracecred);
- }
+ if ((p2->p_ktrioparms = p1->p_ktrioparms) != NULL)
+ p1->p_ktrioparms->refs++;
}
mtx_unlock(&ktrace_mtx);
PROC_UNLOCK(p1);
ktrprocctor(p2);
}
/*
* When a thread returns, drain any asynchronous records generated by the
* system call.
*/
void
ktruserret(struct thread *td)
{
ktrace_enter(td);
sx_xlock(&ktrace_sx);
ktr_drain(td);
sx_xunlock(&ktrace_sx);
ktrace_exit(td);
}
void
ktrnamei(path)
char *path;
{
struct ktr_request *req;
int namelen;
char *buf = NULL;
namelen = strlen(path);
if (namelen > 0) {
buf = malloc(namelen, M_KTRACE, M_WAITOK);
bcopy(path, buf, namelen);
}
req = ktr_getrequest(KTR_NAMEI);
if (req == NULL) {
if (buf != NULL)
free(buf, M_KTRACE);
return;
}
if (namelen > 0) {
req->ktr_header.ktr_len = namelen;
req->ktr_buffer = buf;
}
ktr_submitrequest(curthread, req);
}
void
ktrsysctl(int *name, u_int namelen)
{
struct ktr_request *req;
u_int mib[CTL_MAXNAME + 2];
char *mibname;
size_t mibnamelen;
int error;
/* Lookup name of mib. */
KASSERT(namelen <= CTL_MAXNAME, ("sysctl MIB too long"));
mib[0] = 0;
mib[1] = 1;
bcopy(name, mib + 2, namelen * sizeof(*name));
mibnamelen = 128;
mibname = malloc(mibnamelen, M_KTRACE, M_WAITOK);
error = kernel_sysctl(curthread, mib, namelen + 2, mibname, &mibnamelen,
NULL, 0, &mibnamelen, 0);
if (error) {
free(mibname, M_KTRACE);
return;
}
req = ktr_getrequest(KTR_SYSCTL);
if (req == NULL) {
free(mibname, M_KTRACE);
return;
}
req->ktr_header.ktr_len = mibnamelen;
req->ktr_buffer = mibname;
ktr_submitrequest(curthread, req);
}
void
ktrgenio(int fd, enum uio_rw rw, struct uio *uio, int error)
{
struct ktr_request *req;
struct ktr_genio *ktg;
int datalen;
char *buf;
if (error) {
free(uio, M_IOV);
return;
}
uio->uio_offset = 0;
uio->uio_rw = UIO_WRITE;
datalen = MIN(uio->uio_resid, ktr_geniosize);
buf = malloc(datalen, M_KTRACE, M_WAITOK);
error = uiomove(buf, datalen, uio);
free(uio, M_IOV);
if (error) {
free(buf, M_KTRACE);
return;
}
req = ktr_getrequest(KTR_GENIO);
if (req == NULL) {
free(buf, M_KTRACE);
return;
}
ktg = &req->ktr_data.ktr_genio;
ktg->ktr_fd = fd;
ktg->ktr_rw = rw;
req->ktr_header.ktr_len = datalen;
req->ktr_buffer = buf;
ktr_submitrequest(curthread, req);
}
void
ktrpsig(int sig, sig_t action, sigset_t *mask, int code)
{
struct thread *td = curthread;
struct ktr_request *req;
struct ktr_psig *kp;
req = ktr_getrequest(KTR_PSIG);
if (req == NULL)
return;
kp = &req->ktr_data.ktr_psig;
kp->signo = (char)sig;
kp->action = action;
kp->mask = *mask;
kp->code = code;
ktr_enqueuerequest(td, req);
ktrace_exit(td);
}
void
ktrcsw(int out, int user, const char *wmesg)
{
struct thread *td = curthread;
struct ktr_request *req;
struct ktr_csw *kc;
if (__predict_false(curthread->td_pflags & TDP_INKTRACE))
return;
req = ktr_getrequest(KTR_CSW);
if (req == NULL)
return;
kc = &req->ktr_data.ktr_csw;
kc->out = out;
kc->user = user;
if (wmesg != NULL)
strlcpy(kc->wmesg, wmesg, sizeof(kc->wmesg));
else
bzero(kc->wmesg, sizeof(kc->wmesg));
ktr_enqueuerequest(td, req);
ktrace_exit(td);
}
void
ktrstruct(const char *name, const void *data, size_t datalen)
{
struct ktr_request *req;
char *buf;
size_t buflen, namelen;
if (__predict_false(curthread->td_pflags & TDP_INKTRACE))
return;
if (data == NULL)
datalen = 0;
namelen = strlen(name) + 1;
buflen = namelen + datalen;
buf = malloc(buflen, M_KTRACE, M_WAITOK);
strcpy(buf, name);
bcopy(data, buf + namelen, datalen);
if ((req = ktr_getrequest(KTR_STRUCT)) == NULL) {
free(buf, M_KTRACE);
return;
}
req->ktr_buffer = buf;
req->ktr_header.ktr_len = buflen;
ktr_submitrequest(curthread, req);
}
void
ktrstruct_error(const char *name, const void *data, size_t datalen, int error)
{
if (error == 0)
ktrstruct(name, data, datalen);
}
void
ktrstructarray(const char *name, enum uio_seg seg, const void *data,
int num_items, size_t struct_size)
{
struct ktr_request *req;
struct ktr_struct_array *ksa;
char *buf;
size_t buflen, datalen, namelen;
int max_items;
if (__predict_false(curthread->td_pflags & TDP_INKTRACE))
return;
if (num_items < 0)
return;
/* Trim array length to genio size. */
max_items = ktr_geniosize / struct_size;
if (num_items > max_items) {
if (max_items == 0)
num_items = 1;
else
num_items = max_items;
}
datalen = num_items * struct_size;
if (data == NULL)
datalen = 0;
namelen = strlen(name) + 1;
buflen = namelen + datalen;
buf = malloc(buflen, M_KTRACE, M_WAITOK);
strcpy(buf, name);
if (seg == UIO_SYSSPACE)
bcopy(data, buf + namelen, datalen);
else {
if (copyin(data, buf + namelen, datalen) != 0) {
free(buf, M_KTRACE);
return;
}
}
if ((req = ktr_getrequest(KTR_STRUCT_ARRAY)) == NULL) {
free(buf, M_KTRACE);
return;
}
ksa = &req->ktr_data.ktr_struct_array;
ksa->struct_size = struct_size;
req->ktr_buffer = buf;
req->ktr_header.ktr_len = buflen;
ktr_submitrequest(curthread, req);
}
void
ktrcapfail(enum ktr_cap_fail_type type, const cap_rights_t *needed,
const cap_rights_t *held)
{
struct thread *td = curthread;
struct ktr_request *req;
struct ktr_cap_fail *kcf;
if (__predict_false(curthread->td_pflags & TDP_INKTRACE))
return;
req = ktr_getrequest(KTR_CAPFAIL);
if (req == NULL)
return;
kcf = &req->ktr_data.ktr_cap_fail;
kcf->cap_type = type;
if (needed != NULL)
kcf->cap_needed = *needed;
else
cap_rights_init(&kcf->cap_needed);
if (held != NULL)
kcf->cap_held = *held;
else
cap_rights_init(&kcf->cap_held);
ktr_enqueuerequest(td, req);
ktrace_exit(td);
}
void
ktrfault(vm_offset_t vaddr, int type)
{
struct thread *td = curthread;
struct ktr_request *req;
struct ktr_fault *kf;
if (__predict_false(curthread->td_pflags & TDP_INKTRACE))
return;
req = ktr_getrequest(KTR_FAULT);
if (req == NULL)
return;
kf = &req->ktr_data.ktr_fault;
kf->vaddr = vaddr;
kf->type = type;
ktr_enqueuerequest(td, req);
ktrace_exit(td);
}
void
ktrfaultend(int result)
{
struct thread *td = curthread;
struct ktr_request *req;
struct ktr_faultend *kf;
if (__predict_false(curthread->td_pflags & TDP_INKTRACE))
return;
req = ktr_getrequest(KTR_FAULTEND);
if (req == NULL)
return;
kf = &req->ktr_data.ktr_faultend;
kf->result = result;
ktr_enqueuerequest(td, req);
ktrace_exit(td);
}
#endif /* KTRACE */
/* Interface and common routines */
#ifndef _SYS_SYSPROTO_H_
struct ktrace_args {
char *fname;
int ops;
int facs;
int pid;
};
#endif
/* ARGSUSED */
int
sys_ktrace(struct thread *td, struct ktrace_args *uap)
{
#ifdef KTRACE
struct vnode *vp = NULL;
struct proc *p;
struct pgrp *pg;
int facs = uap->facs & ~KTRFAC_ROOT;
int ops = KTROP(uap->ops);
int descend = uap->ops & KTRFLAG_DESCEND;
int nfound, ret = 0;
int flags, error = 0;
struct nameidata nd;
- struct ucred *cred;
+ struct ktr_io_params *kiop, *old_kiop;
/*
* Need something to (un)trace.
*/
if (ops != KTROP_CLEARFILE && facs == 0)
return (EINVAL);
+ kiop = NULL;
ktrace_enter(td);
if (ops != KTROP_CLEAR) {
/*
* an operation which requires a file argument.
*/
NDINIT(&nd, LOOKUP, NOFOLLOW, UIO_USERSPACE, uap->fname, td);
flags = FREAD | FWRITE | O_NOFOLLOW;
error = vn_open(&nd, &flags, 0, NULL);
if (error) {
ktrace_exit(td);
return (error);
}
NDFREE(&nd, NDF_ONLY_PNBUF);
vp = nd.ni_vp;
VOP_UNLOCK(vp);
if (vp->v_type != VREG) {
(void) vn_close(vp, FREAD|FWRITE, td->td_ucred, td);
ktrace_exit(td);
return (EACCES);
}
+ kiop = ktr_io_params_alloc(td, vp);
}
/*
* Clear all uses of the tracefile.
*/
if (ops == KTROP_CLEARFILE) {
- int vrele_count;
-
- vrele_count = 0;
+restart:
sx_slock(&allproc_lock);
FOREACH_PROC_IN_SYSTEM(p) {
+ old_kiop = NULL;
PROC_LOCK(p);
- if (p->p_tracevp == vp) {
+ if (p->p_ktrioparms != NULL &&
+ p->p_ktrioparms->vp == vp) {
if (ktrcanset(td, p)) {
mtx_lock(&ktrace_mtx);
- ktr_freeproc(p, &cred, NULL);
+ old_kiop = ktr_freeproc(p);
mtx_unlock(&ktrace_mtx);
- vrele_count++;
- crfree(cred);
} else
error = EPERM;
}
PROC_UNLOCK(p);
+ if (old_kiop != NULL) {
+ sx_sunlock(&allproc_lock);
+ ktr_io_params_free(old_kiop);
+ goto restart;
+ }
}
sx_sunlock(&allproc_lock);
- if (vrele_count > 0) {
- while (vrele_count-- > 0)
- vrele(vp);
- }
goto done;
}
/*
* do it
*/
sx_slock(&proctree_lock);
if (uap->pid < 0) {
/*
* by process group
*/
pg = pgfind(-uap->pid);
if (pg == NULL) {
sx_sunlock(&proctree_lock);
error = ESRCH;
goto done;
}
/*
* ktrops() may call vrele(). Lock pg_members
* by the proctree_lock rather than pg_mtx.
*/
PGRP_UNLOCK(pg);
nfound = 0;
LIST_FOREACH(p, &pg->pg_members, p_pglist) {
PROC_LOCK(p);
if (p->p_state == PRS_NEW ||
p_cansee(td, p) != 0) {
PROC_UNLOCK(p);
continue;
}
nfound++;
if (descend)
- ret |= ktrsetchildren(td, p, ops, facs, vp);
+ ret |= ktrsetchildren(td, p, ops, facs, kiop);
else
- ret |= ktrops(td, p, ops, facs, vp);
+ ret |= ktrops(td, p, ops, facs, kiop);
}
if (nfound == 0) {
sx_sunlock(&proctree_lock);
error = ESRCH;
goto done;
}
} else {
/*
* by pid
*/
p = pfind(uap->pid);
if (p == NULL)
error = ESRCH;
else
error = p_cansee(td, p);
if (error) {
if (p != NULL)
PROC_UNLOCK(p);
sx_sunlock(&proctree_lock);
goto done;
}
if (descend)
- ret |= ktrsetchildren(td, p, ops, facs, vp);
+ ret |= ktrsetchildren(td, p, ops, facs, kiop);
else
- ret |= ktrops(td, p, ops, facs, vp);
+ ret |= ktrops(td, p, ops, facs, kiop);
}
sx_sunlock(&proctree_lock);
if (!ret)
error = EPERM;
done:
- if (vp != NULL)
- (void) vn_close(vp, FWRITE, td->td_ucred, td);
+ if (kiop != NULL) {
+ mtx_lock(&ktrace_mtx);
+ kiop = ktr_io_params_rele(kiop);
+ mtx_unlock(&ktrace_mtx);
+ ktr_io_params_free(kiop);
+ }
ktrace_exit(td);
return (error);
#else /* !KTRACE */
return (ENOSYS);
#endif /* KTRACE */
}
/* ARGSUSED */
int
sys_utrace(struct thread *td, struct utrace_args *uap)
{
#ifdef KTRACE
struct ktr_request *req;
void *cp;
int error;
if (!KTRPOINT(td, KTR_USER))
return (0);
if (uap->len > KTR_USER_MAXLEN)
return (EINVAL);
cp = malloc(uap->len, M_KTRACE, M_WAITOK);
error = copyin(uap->addr, cp, uap->len);
if (error) {
free(cp, M_KTRACE);
return (error);
}
req = ktr_getrequest(KTR_USER);
if (req == NULL) {
free(cp, M_KTRACE);
return (ENOMEM);
}
req->ktr_buffer = cp;
req->ktr_header.ktr_len = uap->len;
ktr_submitrequest(td, req);
return (0);
#else /* !KTRACE */
return (ENOSYS);
#endif /* KTRACE */
}
#ifdef KTRACE
static int
-ktrops(struct thread *td, struct proc *p, int ops, int facs, struct vnode *vp)
+ktrops(struct thread *td, struct proc *p, int ops, int facs,
+ struct ktr_io_params *new_kiop)
{
- struct vnode *tracevp = NULL;
- struct ucred *tracecred = NULL;
+ struct ktr_io_params *old_kiop;
PROC_LOCK_ASSERT(p, MA_OWNED);
if (!ktrcanset(td, p)) {
PROC_UNLOCK(p);
return (0);
}
if (p->p_flag & P_WEXIT) {
/* If the process is exiting, just ignore it. */
PROC_UNLOCK(p);
return (1);
}
+ old_kiop = NULL;
mtx_lock(&ktrace_mtx);
if (ops == KTROP_SET) {
- if (p->p_tracevp != vp) {
- /*
- * if trace file already in use, relinquish below
- */
- tracevp = p->p_tracevp;
- VREF(vp);
- p->p_tracevp = vp;
+ if (p->p_ktrioparms != NULL &&
+ p->p_ktrioparms->vp != new_kiop->vp) {
+ /* if trace file already in use, relinquish below */
+ old_kiop = ktr_io_params_rele(p->p_ktrioparms);
+ p->p_ktrioparms = NULL;
}
- if (p->p_tracecred != td->td_ucred) {
- tracecred = p->p_tracecred;
- p->p_tracecred = crhold(td->td_ucred);
+ if (p->p_ktrioparms == NULL) {
+ p->p_ktrioparms = new_kiop;
+ ktr_io_params_ref(new_kiop);
}
p->p_traceflag |= facs;
if (priv_check(td, PRIV_KTRACE) == 0)
p->p_traceflag |= KTRFAC_ROOT;
} else {
/* KTROP_CLEAR */
if (((p->p_traceflag &= ~facs) & KTRFAC_MASK) == 0)
/* no more tracing */
- ktr_freeproc(p, &tracecred, &tracevp);
+ old_kiop = ktr_freeproc(p);
}
mtx_unlock(&ktrace_mtx);
if ((p->p_traceflag & KTRFAC_MASK) != 0)
ktrprocctor_entered(td, p);
PROC_UNLOCK(p);
- if (tracevp != NULL)
- vrele(tracevp);
- if (tracecred != NULL)
- crfree(tracecred);
+ ktr_io_params_free(old_kiop);
return (1);
}
static int
ktrsetchildren(struct thread *td, struct proc *top, int ops, int facs,
- struct vnode *vp)
+ struct ktr_io_params *new_kiop)
{
struct proc *p;
int ret = 0;
p = top;
PROC_LOCK_ASSERT(p, MA_OWNED);
sx_assert(&proctree_lock, SX_LOCKED);
for (;;) {
- ret |= ktrops(td, p, ops, facs, vp);
+ ret |= ktrops(td, p, ops, facs, new_kiop);
/*
* If this process has children, descend to them next,
* otherwise do any siblings, and if done with this level,
* follow back up the tree (but not past top).
*/
if (!LIST_EMPTY(&p->p_children))
p = LIST_FIRST(&p->p_children);
else for (;;) {
if (p == top)
return (ret);
if (LIST_NEXT(p, p_sibling)) {
p = LIST_NEXT(p, p_sibling);
break;
}
p = p->p_pptr;
}
PROC_LOCK(p);
}
/*NOTREACHED*/
}
static void
ktr_writerequest(struct thread *td, struct ktr_request *req)
{
+ struct ktr_io_params *kiop;
struct ktr_header *kth;
struct vnode *vp;
struct proc *p;
struct ucred *cred;
struct uio auio;
struct iovec aiov[3];
struct mount *mp;
int datalen, buflen;
int error;
+ p = td->td_proc;
+
/*
* We hold the vnode and credential for use in I/O in case ktrace is
* disabled on the process as we write out the request.
*
* XXXRW: This is not ideal: we could end up performing a write after
* the vnode has been closed.
*/
mtx_lock(&ktrace_mtx);
- vp = td->td_proc->p_tracevp;
- cred = td->td_proc->p_tracecred;
+
+ kiop = p->p_ktrioparms;
/*
- * If vp is NULL, the vp has been cleared out from under this
- * request, so just drop it. Make sure the credential and vnode are
- * in sync: we should have both or neither.
+ * If kiop is NULL, it has been cleared out from under this
+ * request, so just drop it.
*/
- if (vp == NULL) {
- KASSERT(cred == NULL, ("ktr_writerequest: cred != NULL"));
+ if (kiop == NULL) {
mtx_unlock(&ktrace_mtx);
return;
}
- VREF(vp);
+
+ vp = kiop->vp;
+ cred = kiop->cr;
+
+ vrefact(vp);
KASSERT(cred != NULL, ("ktr_writerequest: cred == NULL"));
crhold(cred);
mtx_unlock(&ktrace_mtx);
kth = &req->ktr_header;
KASSERT(((u_short)kth->ktr_type & ~KTR_DROP) < nitems(data_lengths),
("data_lengths array overflow"));
datalen = data_lengths[(u_short)kth->ktr_type & ~KTR_DROP];
buflen = kth->ktr_len;
auio.uio_iov = &aiov[0];
auio.uio_offset = 0;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_rw = UIO_WRITE;
aiov[0].iov_base = (caddr_t)kth;
aiov[0].iov_len = sizeof(struct ktr_header);
auio.uio_resid = sizeof(struct ktr_header);
auio.uio_iovcnt = 1;
auio.uio_td = td;
if (datalen != 0) {
aiov[1].iov_base = (caddr_t)&req->ktr_data;
aiov[1].iov_len = datalen;
auio.uio_resid += datalen;
auio.uio_iovcnt++;
kth->ktr_len += datalen;
}
if (buflen != 0) {
KASSERT(req->ktr_buffer != NULL, ("ktrace: nothing to write"));
aiov[auio.uio_iovcnt].iov_base = req->ktr_buffer;
aiov[auio.uio_iovcnt].iov_len = buflen;
auio.uio_resid += buflen;
auio.uio_iovcnt++;
}
vn_start_write(vp, &mp, V_WAIT);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
#ifdef MAC
error = mac_vnode_check_write(cred, NOCRED, vp);
if (error == 0)
#endif
error = VOP_WRITE(vp, &auio, IO_UNIT | IO_APPEND, cred);
VOP_UNLOCK(vp);
vn_finished_write(mp);
crfree(cred);
- if (!error) {
+ if (error == 0) {
vrele(vp);
return;
}
/*
* If error encountered, give up tracing on this vnode on this
* process. Other processes might still be suitable for
* writes to this vnode.
*/
- p = td->td_proc;
log(LOG_NOTICE,
"ktrace write failed, errno %d, tracing stopped for pid %d\n",
error, p->p_pid);
- cred = NULL;
- sx_slock(&allproc_lock);
+
PROC_LOCK(p);
mtx_lock(&ktrace_mtx);
- if (p->p_tracevp == vp)
- ktr_freeproc(p, &cred, NULL);
+ if (p->p_ktrioparms != NULL && p->p_ktrioparms->vp == vp)
+ kiop = ktr_freeproc(p);
mtx_unlock(&ktrace_mtx);
PROC_UNLOCK(p);
- if (cred != NULL) {
- crfree(cred);
- cred = NULL;
- }
- sx_sunlock(&allproc_lock);
- vrele(vp);
+ ktr_io_params_free(kiop);
vrele(vp);
}
/*
* Return true if caller has permission to set the ktracing state
* of target. Essentially, the target can't possess any
* more permissions than the caller. KTRFAC_ROOT signifies that
* root previously set the tracing status on the target process, and
* so, only root may further change it.
*/
static int
ktrcanset(struct thread *td, struct proc *targetp)
{
PROC_LOCK_ASSERT(targetp, MA_OWNED);
if (targetp->p_traceflag & KTRFAC_ROOT &&
priv_check(td, PRIV_KTRACE))
return (0);
if (p_candebug(td, targetp) != 0)
return (0);
return (1);
}
#endif /* KTRACE */
diff --git a/sys/kern/kern_proc.c b/sys/kern/kern_proc.c
index 33f168836370..ec732e8db060 100644
--- a/sys/kern/kern_proc.c
+++ b/sys/kern/kern_proc.c
@@ -1,3374 +1,3377 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1982, 1986, 1989, 1991, 1993
* The Regents of the University of California. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)kern_proc.c 8.7 (Berkeley) 2/14/95
*/
#include <sys/cdefs.h>
__FBSDID("$FreeBSD$");
#include "opt_ddb.h"
#include "opt_ktrace.h"
#include "opt_kstack_pages.h"
#include "opt_stack.h"
#include <sys/param.h>
#include <sys/systm.h>
#include <sys/bitstring.h>
#include <sys/elf.h>
#include <sys/eventhandler.h>
#include <sys/exec.h>
#include <sys/jail.h>
#include <sys/kernel.h>
#include <sys/limits.h>
#include <sys/lock.h>
#include <sys/loginclass.h>
#include <sys/malloc.h>
#include <sys/mman.h>
#include <sys/mount.h>
#include <sys/mutex.h>
#include <sys/proc.h>
#include <sys/ptrace.h>
#include <sys/refcount.h>
#include <sys/resourcevar.h>
#include <sys/rwlock.h>
#include <sys/sbuf.h>
#include <sys/sysent.h>
#include <sys/sched.h>
#include <sys/smp.h>
#include <sys/stack.h>
#include <sys/stat.h>
#include <sys/dtrace_bsd.h>
#include <sys/sysctl.h>
#include <sys/filedesc.h>
#include <sys/tty.h>
#include <sys/signalvar.h>
#include <sys/sdt.h>
#include <sys/sx.h>
#include <sys/user.h>
#include <sys/vnode.h>
#include <sys/wait.h>
+#ifdef KTRACE
+#include <sys/ktrace.h>
+#endif
#ifdef DDB
#include <ddb/ddb.h>
#endif
#include <vm/vm.h>
#include <vm/vm_param.h>
#include <vm/vm_extern.h>
#include <vm/pmap.h>
#include <vm/vm_map.h>
#include <vm/vm_object.h>
#include <vm/vm_page.h>
#include <vm/uma.h>
#include <fs/devfs/devfs.h>
#ifdef COMPAT_FREEBSD32
#include <compat/freebsd32/freebsd32.h>
#include <compat/freebsd32/freebsd32_util.h>
#endif
SDT_PROVIDER_DEFINE(proc);
MALLOC_DEFINE(M_SESSION, "session", "session header");
static MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
static void doenterpgrp(struct proc *, struct pgrp *);
static void orphanpg(struct pgrp *pg);
static void fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp);
static void fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp);
static void fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp,
int preferthread);
static void pgdelete(struct pgrp *);
static int pgrp_init(void *mem, int size, int flags);
static int proc_ctor(void *mem, int size, void *arg, int flags);
static void proc_dtor(void *mem, int size, void *arg);
static int proc_init(void *mem, int size, int flags);
static void proc_fini(void *mem, int size);
static void pargs_free(struct pargs *pa);
/*
* Other process lists
*/
struct pidhashhead *pidhashtbl;
struct sx *pidhashtbl_lock;
u_long pidhash;
u_long pidhashlock;
struct pgrphashhead *pgrphashtbl;
u_long pgrphash;
struct proclist allproc;
struct sx __exclusive_cache_line allproc_lock;
struct sx __exclusive_cache_line proctree_lock;
struct mtx __exclusive_cache_line ppeers_lock;
struct mtx __exclusive_cache_line procid_lock;
uma_zone_t proc_zone;
uma_zone_t pgrp_zone;
/*
* The offset of various fields in struct proc and struct thread.
* These are used by kernel debuggers to enumerate kernel threads and
* processes.
*/
const int proc_off_p_pid = offsetof(struct proc, p_pid);
const int proc_off_p_comm = offsetof(struct proc, p_comm);
const int proc_off_p_list = offsetof(struct proc, p_list);
const int proc_off_p_hash = offsetof(struct proc, p_hash);
const int proc_off_p_threads = offsetof(struct proc, p_threads);
const int thread_off_td_tid = offsetof(struct thread, td_tid);
const int thread_off_td_name = offsetof(struct thread, td_name);
const int thread_off_td_oncpu = offsetof(struct thread, td_oncpu);
const int thread_off_td_pcb = offsetof(struct thread, td_pcb);
const int thread_off_td_plist = offsetof(struct thread, td_plist);
EVENTHANDLER_LIST_DEFINE(process_ctor);
EVENTHANDLER_LIST_DEFINE(process_dtor);
EVENTHANDLER_LIST_DEFINE(process_init);
EVENTHANDLER_LIST_DEFINE(process_fini);
EVENTHANDLER_LIST_DEFINE(process_exit);
EVENTHANDLER_LIST_DEFINE(process_fork);
EVENTHANDLER_LIST_DEFINE(process_exec);
int kstack_pages = KSTACK_PAGES;
SYSCTL_INT(_kern, OID_AUTO, kstack_pages, CTLFLAG_RD, &kstack_pages, 0,
"Kernel stack size in pages");
static int vmmap_skip_res_cnt = 0;
SYSCTL_INT(_kern, OID_AUTO, proc_vmmap_skip_resident_count, CTLFLAG_RW,
&vmmap_skip_res_cnt, 0,
"Skip calculation of the pages resident count in kern.proc.vmmap");
CTASSERT(sizeof(struct kinfo_proc) == KINFO_PROC_SIZE);
#ifdef COMPAT_FREEBSD32
CTASSERT(sizeof(struct kinfo_proc32) == KINFO_PROC32_SIZE);
#endif
/*
* Initialize global process hashing structures.
*/
void
procinit(void)
{
u_long i;
sx_init(&allproc_lock, "allproc");
sx_init(&proctree_lock, "proctree");
mtx_init(&ppeers_lock, "p_peers", NULL, MTX_DEF);
mtx_init(&procid_lock, "procid", NULL, MTX_DEF);
LIST_INIT(&allproc);
pidhashtbl = hashinit(maxproc / 4, M_PROC, &pidhash);
pidhashlock = (pidhash + 1) / 64;
if (pidhashlock > 0)
pidhashlock--;
pidhashtbl_lock = malloc(sizeof(*pidhashtbl_lock) * (pidhashlock + 1),
M_PROC, M_WAITOK | M_ZERO);
for (i = 0; i < pidhashlock + 1; i++)
sx_init_flags(&pidhashtbl_lock[i], "pidhash", SX_DUPOK);
pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash);
proc_zone = uma_zcreate("PROC", sched_sizeof_proc(),
proc_ctor, proc_dtor, proc_init, proc_fini,
UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
pgrp_zone = uma_zcreate("PGRP", sizeof(struct pgrp), NULL, NULL,
pgrp_init, NULL, UMA_ALIGN_PTR, UMA_ZONE_NOFREE);
uihashinit();
}
/*
* Prepare a proc for use.
*/
static int
proc_ctor(void *mem, int size, void *arg, int flags)
{
struct proc *p;
struct thread *td;
p = (struct proc *)mem;
#ifdef KDTRACE_HOOKS
kdtrace_proc_ctor(p);
#endif
EVENTHANDLER_DIRECT_INVOKE(process_ctor, p);
td = FIRST_THREAD_IN_PROC(p);
if (td != NULL) {
/* Make sure all thread constructors are executed */
EVENTHANDLER_DIRECT_INVOKE(thread_ctor, td);
}
return (0);
}
/*
* Reclaim a proc after use.
*/
static void
proc_dtor(void *mem, int size, void *arg)
{
struct proc *p;
struct thread *td;
/* INVARIANTS checks go here */
p = (struct proc *)mem;
td = FIRST_THREAD_IN_PROC(p);
if (td != NULL) {
#ifdef INVARIANTS
KASSERT((p->p_numthreads == 1),
("bad number of threads in exiting process"));
KASSERT(STAILQ_EMPTY(&p->p_ktr), ("proc_dtor: non-empty p_ktr"));
#endif
/* Free all OSD associated to this thread. */
osd_thread_exit(td);
td_softdep_cleanup(td);
MPASS(td->td_su == NULL);
/* Make sure all thread destructors are executed */
EVENTHANDLER_DIRECT_INVOKE(thread_dtor, td);
}
EVENTHANDLER_DIRECT_INVOKE(process_dtor, p);
#ifdef KDTRACE_HOOKS
kdtrace_proc_dtor(p);
#endif
if (p->p_ksi != NULL)
KASSERT(! KSI_ONQ(p->p_ksi), ("SIGCHLD queue"));
}
/*
* Initialize type-stable parts of a proc (when newly created).
*/
static int
proc_init(void *mem, int size, int flags)
{
struct proc *p;
p = (struct proc *)mem;
mtx_init(&p->p_mtx, "process lock", NULL, MTX_DEF | MTX_DUPOK | MTX_NEW);
mtx_init(&p->p_slock, "process slock", NULL, MTX_SPIN | MTX_NEW);
mtx_init(&p->p_statmtx, "pstatl", NULL, MTX_SPIN | MTX_NEW);
mtx_init(&p->p_itimmtx, "pitiml", NULL, MTX_SPIN | MTX_NEW);
mtx_init(&p->p_profmtx, "pprofl", NULL, MTX_SPIN | MTX_NEW);
cv_init(&p->p_pwait, "ppwait");
TAILQ_INIT(&p->p_threads); /* all threads in proc */
EVENTHANDLER_DIRECT_INVOKE(process_init, p);
p->p_stats = pstats_alloc();
p->p_pgrp = NULL;
return (0);
}
/*
* UMA should ensure that this function is never called.
* Freeing a proc structure would violate type stability.
*/
static void
proc_fini(void *mem, int size)
{
#ifdef notnow
struct proc *p;
p = (struct proc *)mem;
EVENTHANDLER_DIRECT_INVOKE(process_fini, p);
pstats_free(p->p_stats);
thread_free(FIRST_THREAD_IN_PROC(p));
mtx_destroy(&p->p_mtx);
if (p->p_ksi != NULL)
ksiginfo_free(p->p_ksi);
#else
panic("proc reclaimed");
#endif
}
static int
pgrp_init(void *mem, int size, int flags)
{
struct pgrp *pg;
pg = mem;
mtx_init(&pg->pg_mtx, "process group", NULL, MTX_DEF | MTX_DUPOK);
return (0);
}
/*
* PID space management.
*
* These bitmaps are used by fork_findpid.
*/
bitstr_t bit_decl(proc_id_pidmap, PID_MAX);
bitstr_t bit_decl(proc_id_grpidmap, PID_MAX);
bitstr_t bit_decl(proc_id_sessidmap, PID_MAX);
bitstr_t bit_decl(proc_id_reapmap, PID_MAX);
static bitstr_t *proc_id_array[] = {
proc_id_pidmap,
proc_id_grpidmap,
proc_id_sessidmap,
proc_id_reapmap,
};
void
proc_id_set(int type, pid_t id)
{
KASSERT(type >= 0 && type < nitems(proc_id_array),
("invalid type %d\n", type));
mtx_lock(&procid_lock);
KASSERT(bit_test(proc_id_array[type], id) == 0,
("bit %d already set in %d\n", id, type));
bit_set(proc_id_array[type], id);
mtx_unlock(&procid_lock);
}
void
proc_id_set_cond(int type, pid_t id)
{
KASSERT(type >= 0 && type < nitems(proc_id_array),
("invalid type %d\n", type));
if (bit_test(proc_id_array[type], id))
return;
mtx_lock(&procid_lock);
bit_set(proc_id_array[type], id);
mtx_unlock(&procid_lock);
}
void
proc_id_clear(int type, pid_t id)
{
KASSERT(type >= 0 && type < nitems(proc_id_array),
("invalid type %d\n", type));
mtx_lock(&procid_lock);
KASSERT(bit_test(proc_id_array[type], id) != 0,
("bit %d not set in %d\n", id, type));
bit_clear(proc_id_array[type], id);
mtx_unlock(&procid_lock);
}
/*
* Is p an inferior of the current process?
*/
int
inferior(struct proc *p)
{
sx_assert(&proctree_lock, SX_LOCKED);
PROC_LOCK_ASSERT(p, MA_OWNED);
for (; p != curproc; p = proc_realparent(p)) {
if (p->p_pid == 0)
return (0);
}
return (1);
}
/*
* Shared lock all the pid hash lists.
*/
void
pidhash_slockall(void)
{
u_long i;
for (i = 0; i < pidhashlock + 1; i++)
sx_slock(&pidhashtbl_lock[i]);
}
/*
* Shared unlock all the pid hash lists.
*/
void
pidhash_sunlockall(void)
{
u_long i;
for (i = 0; i < pidhashlock + 1; i++)
sx_sunlock(&pidhashtbl_lock[i]);
}
/*
* Similar to pfind_any(), this function finds zombies.
*/
struct proc *
pfind_any_locked(pid_t pid)
{
struct proc *p;
sx_assert(PIDHASHLOCK(pid), SX_LOCKED);
LIST_FOREACH(p, PIDHASH(pid), p_hash) {
if (p->p_pid == pid) {
PROC_LOCK(p);
if (p->p_state == PRS_NEW) {
PROC_UNLOCK(p);
p = NULL;
}
break;
}
}
return (p);
}
/*
* Locate a process by number.
*
* By not returning processes in the PRS_NEW state, we allow callers to avoid
* testing for that condition to avoid dereferencing p_ucred, et al.
*/
static __always_inline struct proc *
_pfind(pid_t pid, bool zombie)
{
struct proc *p;
p = curproc;
if (p->p_pid == pid) {
PROC_LOCK(p);
return (p);
}
sx_slock(PIDHASHLOCK(pid));
LIST_FOREACH(p, PIDHASH(pid), p_hash) {
if (p->p_pid == pid) {
PROC_LOCK(p);
if (p->p_state == PRS_NEW ||
(!zombie && p->p_state == PRS_ZOMBIE)) {
PROC_UNLOCK(p);
p = NULL;
}
break;
}
}
sx_sunlock(PIDHASHLOCK(pid));
return (p);
}
struct proc *
pfind(pid_t pid)
{
return (_pfind(pid, false));
}
/*
* Same as pfind but allow zombies.
*/
struct proc *
pfind_any(pid_t pid)
{
return (_pfind(pid, true));
}
/*
* Locate a process group by number.
* The caller must hold proctree_lock.
*/
struct pgrp *
pgfind(pid_t pgid)
{
struct pgrp *pgrp;
sx_assert(&proctree_lock, SX_LOCKED);
LIST_FOREACH(pgrp, PGRPHASH(pgid), pg_hash) {
if (pgrp->pg_id == pgid) {
PGRP_LOCK(pgrp);
return (pgrp);
}
}
return (NULL);
}
/*
* Locate process and do additional manipulations, depending on flags.
*/
int
pget(pid_t pid, int flags, struct proc **pp)
{
struct proc *p;
struct thread *td1;
int error;
p = curproc;
if (p->p_pid == pid) {
PROC_LOCK(p);
} else {
p = NULL;
if (pid <= PID_MAX) {
if ((flags & PGET_NOTWEXIT) == 0)
p = pfind_any(pid);
else
p = pfind(pid);
} else if ((flags & PGET_NOTID) == 0) {
td1 = tdfind(pid, -1);
if (td1 != NULL)
p = td1->td_proc;
}
if (p == NULL)
return (ESRCH);
if ((flags & PGET_CANSEE) != 0) {
error = p_cansee(curthread, p);
if (error != 0)
goto errout;
}
}
if ((flags & PGET_CANDEBUG) != 0) {
error = p_candebug(curthread, p);
if (error != 0)
goto errout;
}
if ((flags & PGET_ISCURRENT) != 0 && curproc != p) {
error = EPERM;
goto errout;
}
if ((flags & PGET_NOTWEXIT) != 0 && (p->p_flag & P_WEXIT) != 0) {
error = ESRCH;
goto errout;
}
if ((flags & PGET_NOTINEXEC) != 0 && (p->p_flag & P_INEXEC) != 0) {
/*
* XXXRW: Not clear ESRCH is the right error during proc
* execve().
*/
error = ESRCH;
goto errout;
}
if ((flags & PGET_HOLD) != 0) {
_PHOLD(p);
PROC_UNLOCK(p);
}
*pp = p;
return (0);
errout:
PROC_UNLOCK(p);
return (error);
}
/*
* Create a new process group.
* pgid must be equal to the pid of p.
* Begin a new session if required.
*/
int
enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp, struct session *sess)
{
sx_assert(&proctree_lock, SX_XLOCKED);
KASSERT(pgrp != NULL, ("enterpgrp: pgrp == NULL"));
KASSERT(p->p_pid == pgid,
("enterpgrp: new pgrp and pid != pgid"));
KASSERT(pgfind(pgid) == NULL,
("enterpgrp: pgrp with pgid exists"));
KASSERT(!SESS_LEADER(p),
("enterpgrp: session leader attempted setpgrp"));
if (sess != NULL) {
/*
* new session
*/
mtx_init(&sess->s_mtx, "session", NULL, MTX_DEF);
PROC_LOCK(p);
p->p_flag &= ~P_CONTROLT;
PROC_UNLOCK(p);
PGRP_LOCK(pgrp);
sess->s_leader = p;
sess->s_sid = p->p_pid;
proc_id_set(PROC_ID_SESSION, p->p_pid);
refcount_init(&sess->s_count, 1);
sess->s_ttyvp = NULL;
sess->s_ttydp = NULL;
sess->s_ttyp = NULL;
bcopy(p->p_session->s_login, sess->s_login,
sizeof(sess->s_login));
pgrp->pg_session = sess;
KASSERT(p == curproc,
("enterpgrp: mksession and p != curproc"));
} else {
pgrp->pg_session = p->p_session;
sess_hold(pgrp->pg_session);
PGRP_LOCK(pgrp);
}
pgrp->pg_id = pgid;
proc_id_set(PROC_ID_GROUP, p->p_pid);
LIST_INIT(&pgrp->pg_members);
pgrp->pg_flags = 0;
/*
* As we have an exclusive lock of proctree_lock,
* this should not deadlock.
*/
LIST_INSERT_HEAD(PGRPHASH(pgid), pgrp, pg_hash);
SLIST_INIT(&pgrp->pg_sigiolst);
PGRP_UNLOCK(pgrp);
doenterpgrp(p, pgrp);
return (0);
}
/*
* Move p to an existing process group
*/
int
enterthispgrp(struct proc *p, struct pgrp *pgrp)
{
sx_assert(&proctree_lock, SX_XLOCKED);
PROC_LOCK_ASSERT(p, MA_NOTOWNED);
PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
KASSERT(pgrp->pg_session == p->p_session,
("%s: pgrp's session %p, p->p_session %p proc %p\n",
__func__, pgrp->pg_session, p->p_session, p));
KASSERT(pgrp != p->p_pgrp,
("%s: p %p belongs to pgrp %p", __func__, p, pgrp));
doenterpgrp(p, pgrp);
return (0);
}
/*
* If true, any child of q which belongs to group pgrp, qualifies the
* process group pgrp as not orphaned.
*/
static bool
isjobproc(struct proc *q, struct pgrp *pgrp)
{
sx_assert(&proctree_lock, SX_LOCKED);
return (q->p_pgrp != pgrp &&
q->p_pgrp->pg_session == pgrp->pg_session);
}
static struct proc *
jobc_reaper(struct proc *p)
{
struct proc *pp;
sx_assert(&proctree_lock, SA_LOCKED);
for (pp = p;;) {
pp = pp->p_reaper;
if (pp->p_reaper == pp ||
(pp->p_treeflag & P_TREE_GRPEXITED) == 0)
return (pp);
}
}
static struct proc *
jobc_parent(struct proc *p, struct proc *p_exiting)
{
struct proc *pp;
sx_assert(&proctree_lock, SA_LOCKED);
pp = proc_realparent(p);
if (pp->p_pptr == NULL || pp == p_exiting ||
(pp->p_treeflag & P_TREE_GRPEXITED) == 0)
return (pp);
return (jobc_reaper(pp));
}
static int
pgrp_calc_jobc(struct pgrp *pgrp)
{
struct proc *q;
int cnt;
#ifdef INVARIANTS
if (!mtx_owned(&pgrp->pg_mtx))
sx_assert(&proctree_lock, SA_LOCKED);
#endif
cnt = 0;
LIST_FOREACH(q, &pgrp->pg_members, p_pglist) {
if ((q->p_treeflag & P_TREE_GRPEXITED) != 0 ||
q->p_pptr == NULL)
continue;
if (isjobproc(jobc_parent(q, NULL), pgrp))
cnt++;
}
return (cnt);
}
/*
* Move p to a process group
*/
static void
doenterpgrp(struct proc *p, struct pgrp *pgrp)
{
struct pgrp *savepgrp;
struct proc *pp;
sx_assert(&proctree_lock, SX_XLOCKED);
PROC_LOCK_ASSERT(p, MA_NOTOWNED);
PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
PGRP_LOCK_ASSERT(p->p_pgrp, MA_NOTOWNED);
SESS_LOCK_ASSERT(p->p_session, MA_NOTOWNED);
savepgrp = p->p_pgrp;
pp = jobc_parent(p, NULL);
PGRP_LOCK(pgrp);
PGRP_LOCK(savepgrp);
if (isjobproc(pp, savepgrp) && pgrp_calc_jobc(savepgrp) == 1)
orphanpg(savepgrp);
PROC_LOCK(p);
LIST_REMOVE(p, p_pglist);
p->p_pgrp = pgrp;
PROC_UNLOCK(p);
LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
if (isjobproc(pp, pgrp))
pgrp->pg_flags &= ~PGRP_ORPHANED;
PGRP_UNLOCK(savepgrp);
PGRP_UNLOCK(pgrp);
if (LIST_EMPTY(&savepgrp->pg_members))
pgdelete(savepgrp);
}
/*
* remove process from process group
*/
int
leavepgrp(struct proc *p)
{
struct pgrp *savepgrp;
sx_assert(&proctree_lock, SX_XLOCKED);
savepgrp = p->p_pgrp;
PGRP_LOCK(savepgrp);
PROC_LOCK(p);
LIST_REMOVE(p, p_pglist);
p->p_pgrp = NULL;
PROC_UNLOCK(p);
PGRP_UNLOCK(savepgrp);
if (LIST_EMPTY(&savepgrp->pg_members))
pgdelete(savepgrp);
return (0);
}
/*
* delete a process group
*/
static void
pgdelete(struct pgrp *pgrp)
{
struct session *savesess;
struct tty *tp;
sx_assert(&proctree_lock, SX_XLOCKED);
PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
/*
* Reset any sigio structures pointing to us as a result of
* F_SETOWN with our pgid. The proctree lock ensures that
* new sigio structures will not be added after this point.
*/
funsetownlst(&pgrp->pg_sigiolst);
PGRP_LOCK(pgrp);
tp = pgrp->pg_session->s_ttyp;
LIST_REMOVE(pgrp, pg_hash);
savesess = pgrp->pg_session;
PGRP_UNLOCK(pgrp);
/* Remove the reference to the pgrp before deallocating it. */
if (tp != NULL) {
tty_lock(tp);
tty_rel_pgrp(tp, pgrp);
}
proc_id_clear(PROC_ID_GROUP, pgrp->pg_id);
uma_zfree(pgrp_zone, pgrp);
sess_release(savesess);
}
static void
fixjobc_kill(struct proc *p)
{
struct proc *q;
struct pgrp *pgrp;
sx_assert(&proctree_lock, SX_LOCKED);
PROC_LOCK_ASSERT(p, MA_NOTOWNED);
pgrp = p->p_pgrp;
PGRP_LOCK_ASSERT(pgrp, MA_NOTOWNED);
SESS_LOCK_ASSERT(pgrp->pg_session, MA_NOTOWNED);
/*
* p no longer affects process group orphanage for children.
* It is marked by the flag because p is only physically
* removed from its process group on wait(2).
*/
MPASS((p->p_treeflag & P_TREE_GRPEXITED) == 0);
p->p_treeflag |= P_TREE_GRPEXITED;
/*
* Check if exiting p orphans its own group.
*/
pgrp = p->p_pgrp;
if (isjobproc(jobc_parent(p, NULL), pgrp)) {
PGRP_LOCK(pgrp);
if (pgrp_calc_jobc(pgrp) == 0)
orphanpg(pgrp);
PGRP_UNLOCK(pgrp);
}
/*
* Check this process' children to see whether they qualify
* their process groups after reparenting to reaper.
*/
LIST_FOREACH(q, &p->p_children, p_sibling) {
pgrp = q->p_pgrp;
PGRP_LOCK(pgrp);
if (pgrp_calc_jobc(pgrp) == 0) {
/*
* We want to handle exactly the children that
* has p as realparent. Then, when calculating
* jobc_parent for children, we should ignore
* P_TREE_GRPEXITED flag already set on p.
*/
if (jobc_parent(q, p) == p && isjobproc(p, pgrp))
orphanpg(pgrp);
} else
pgrp->pg_flags &= ~PGRP_ORPHANED;
PGRP_UNLOCK(pgrp);
}
LIST_FOREACH(q, &p->p_orphans, p_orphan) {
pgrp = q->p_pgrp;
PGRP_LOCK(pgrp);
if (pgrp_calc_jobc(pgrp) == 0) {
if (isjobproc(p, pgrp))
orphanpg(pgrp);
} else
pgrp->pg_flags &= ~PGRP_ORPHANED;
PGRP_UNLOCK(pgrp);
}
}
void
killjobc(void)
{
struct session *sp;
struct tty *tp;
struct proc *p;
struct vnode *ttyvp;
p = curproc;
MPASS(p->p_flag & P_WEXIT);
sx_assert(&proctree_lock, SX_LOCKED);
if (SESS_LEADER(p)) {
sp = p->p_session;
/*
* 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);
}
devfs_ctty_unref(ttyvp);
sx_xlock(&proctree_lock);
}
}
fixjobc_kill(p);
}
/*
* A process group has become orphaned, mark it as such for signal
* delivery code. If there are any stopped processes in the group,
* hang-up all process in that group.
*/
static void
orphanpg(struct pgrp *pg)
{
struct proc *p;
PGRP_LOCK_ASSERT(pg, MA_OWNED);
pg->pg_flags |= PGRP_ORPHANED;
LIST_FOREACH(p, &pg->pg_members, p_pglist) {
PROC_LOCK(p);
if (P_SHOULDSTOP(p) == P_STOPPED_SIG) {
PROC_UNLOCK(p);
LIST_FOREACH(p, &pg->pg_members, p_pglist) {
PROC_LOCK(p);
kern_psignal(p, SIGHUP);
kern_psignal(p, SIGCONT);
PROC_UNLOCK(p);
}
return;
}
PROC_UNLOCK(p);
}
}
void
sess_hold(struct session *s)
{
refcount_acquire(&s->s_count);
}
void
sess_release(struct session *s)
{
if (refcount_release(&s->s_count)) {
if (s->s_ttyp != NULL) {
tty_lock(s->s_ttyp);
tty_rel_sess(s->s_ttyp, s);
}
proc_id_clear(PROC_ID_SESSION, s->s_sid);
mtx_destroy(&s->s_mtx);
free(s, M_SESSION);
}
}
#ifdef DDB
static void
db_print_pgrp_one(struct pgrp *pgrp, struct proc *p)
{
db_printf(
" pid %d at %p pr %d pgrp %p e %d jc %d\n",
p->p_pid, p, p->p_pptr == NULL ? -1 : p->p_pptr->p_pid,
p->p_pgrp, (p->p_treeflag & P_TREE_GRPEXITED) != 0,
p->p_pptr == NULL ? 0 : isjobproc(p->p_pptr, pgrp));
}
DB_SHOW_COMMAND(pgrpdump, pgrpdump)
{
struct pgrp *pgrp;
struct proc *p;
int i;
for (i = 0; i <= pgrphash; i++) {
if (!LIST_EMPTY(&pgrphashtbl[i])) {
db_printf("indx %d\n", i);
LIST_FOREACH(pgrp, &pgrphashtbl[i], pg_hash) {
db_printf(
" pgrp %p, pgid %d, sess %p, sesscnt %d, mem %p\n",
pgrp, (int)pgrp->pg_id, pgrp->pg_session,
pgrp->pg_session->s_count,
LIST_FIRST(&pgrp->pg_members));
LIST_FOREACH(p, &pgrp->pg_members, p_pglist)
db_print_pgrp_one(pgrp, p);
}
}
}
}
#endif /* DDB */
/*
* Calculate the kinfo_proc members which contain process-wide
* informations.
* Must be called with the target process locked.
*/
static void
fill_kinfo_aggregate(struct proc *p, struct kinfo_proc *kp)
{
struct thread *td;
PROC_LOCK_ASSERT(p, MA_OWNED);
kp->ki_estcpu = 0;
kp->ki_pctcpu = 0;
FOREACH_THREAD_IN_PROC(p, td) {
thread_lock(td);
kp->ki_pctcpu += sched_pctcpu(td);
kp->ki_estcpu += sched_estcpu(td);
thread_unlock(td);
}
}
/*
* Fill in any information that is common to all threads in the process.
* Must be called with the target process locked.
*/
static void
fill_kinfo_proc_only(struct proc *p, struct kinfo_proc *kp)
{
struct thread *td0;
struct ucred *cred;
struct sigacts *ps;
struct timeval boottime;
PROC_LOCK_ASSERT(p, MA_OWNED);
kp->ki_structsize = sizeof(*kp);
kp->ki_paddr = p;
kp->ki_addr =/* p->p_addr; */0; /* XXX */
kp->ki_args = p->p_args;
kp->ki_textvp = p->p_textvp;
#ifdef KTRACE
- kp->ki_tracep = p->p_tracevp;
+ kp->ki_tracep = ktr_get_tracevp(p, false);
kp->ki_traceflag = p->p_traceflag;
#endif
kp->ki_fd = p->p_fd;
kp->ki_pd = p->p_pd;
kp->ki_vmspace = p->p_vmspace;
kp->ki_flag = p->p_flag;
kp->ki_flag2 = p->p_flag2;
cred = p->p_ucred;
if (cred) {
kp->ki_uid = cred->cr_uid;
kp->ki_ruid = cred->cr_ruid;
kp->ki_svuid = cred->cr_svuid;
kp->ki_cr_flags = 0;
if (cred->cr_flags & CRED_FLAG_CAPMODE)
kp->ki_cr_flags |= KI_CRF_CAPABILITY_MODE;
/* XXX bde doesn't like KI_NGROUPS */
if (cred->cr_ngroups > KI_NGROUPS) {
kp->ki_ngroups = KI_NGROUPS;
kp->ki_cr_flags |= KI_CRF_GRP_OVERFLOW;
} else
kp->ki_ngroups = cred->cr_ngroups;
bcopy(cred->cr_groups, kp->ki_groups,
kp->ki_ngroups * sizeof(gid_t));
kp->ki_rgid = cred->cr_rgid;
kp->ki_svgid = cred->cr_svgid;
/* If jailed(cred), emulate the old P_JAILED flag. */
if (jailed(cred)) {
kp->ki_flag |= P_JAILED;
/* If inside the jail, use 0 as a jail ID. */
if (cred->cr_prison != curthread->td_ucred->cr_prison)
kp->ki_jid = cred->cr_prison->pr_id;
}
strlcpy(kp->ki_loginclass, cred->cr_loginclass->lc_name,
sizeof(kp->ki_loginclass));
}
ps = p->p_sigacts;
if (ps) {
mtx_lock(&ps->ps_mtx);
kp->ki_sigignore = ps->ps_sigignore;
kp->ki_sigcatch = ps->ps_sigcatch;
mtx_unlock(&ps->ps_mtx);
}
if (p->p_state != PRS_NEW &&
p->p_state != PRS_ZOMBIE &&
p->p_vmspace != NULL) {
struct vmspace *vm = p->p_vmspace;
kp->ki_size = vm->vm_map.size;
kp->ki_rssize = vmspace_resident_count(vm); /*XXX*/
FOREACH_THREAD_IN_PROC(p, td0) {
if (!TD_IS_SWAPPED(td0))
kp->ki_rssize += td0->td_kstack_pages;
}
kp->ki_swrss = vm->vm_swrss;
kp->ki_tsize = vm->vm_tsize;
kp->ki_dsize = vm->vm_dsize;
kp->ki_ssize = vm->vm_ssize;
} else if (p->p_state == PRS_ZOMBIE)
kp->ki_stat = SZOMB;
if (kp->ki_flag & P_INMEM)
kp->ki_sflag = PS_INMEM;
else
kp->ki_sflag = 0;
/* Calculate legacy swtime as seconds since 'swtick'. */
kp->ki_swtime = (ticks - p->p_swtick) / hz;
kp->ki_pid = p->p_pid;
kp->ki_nice = p->p_nice;
kp->ki_fibnum = p->p_fibnum;
kp->ki_start = p->p_stats->p_start;
getboottime(&boottime);
timevaladd(&kp->ki_start, &boottime);
PROC_STATLOCK(p);
rufetch(p, &kp->ki_rusage);
kp->ki_runtime = cputick2usec(p->p_rux.rux_runtime);
calcru(p, &kp->ki_rusage.ru_utime, &kp->ki_rusage.ru_stime);
PROC_STATUNLOCK(p);
calccru(p, &kp->ki_childutime, &kp->ki_childstime);
/* Some callers want child times in a single value. */
kp->ki_childtime = kp->ki_childstime;
timevaladd(&kp->ki_childtime, &kp->ki_childutime);
FOREACH_THREAD_IN_PROC(p, td0)
kp->ki_cow += td0->td_cow;
if (p->p_comm[0] != '\0')
strlcpy(kp->ki_comm, p->p_comm, sizeof(kp->ki_comm));
if (p->p_sysent && p->p_sysent->sv_name != NULL &&
p->p_sysent->sv_name[0] != '\0')
strlcpy(kp->ki_emul, p->p_sysent->sv_name, sizeof(kp->ki_emul));
kp->ki_siglist = p->p_siglist;
kp->ki_xstat = KW_EXITCODE(p->p_xexit, p->p_xsig);
kp->ki_acflag = p->p_acflag;
kp->ki_lock = p->p_lock;
if (p->p_pptr) {
kp->ki_ppid = p->p_oppid;
if (p->p_flag & P_TRACED)
kp->ki_tracer = p->p_pptr->p_pid;
}
}
/*
* Fill job-related process information.
*/
static void
fill_kinfo_proc_pgrp(struct proc *p, struct kinfo_proc *kp)
{
struct tty *tp;
struct session *sp;
struct pgrp *pgrp;
sx_assert(&proctree_lock, SA_LOCKED);
PROC_LOCK_ASSERT(p, MA_OWNED);
pgrp = p->p_pgrp;
if (pgrp == NULL)
return;
kp->ki_pgid = pgrp->pg_id;
kp->ki_jobc = pgrp_calc_jobc(pgrp);
sp = pgrp->pg_session;
tp = NULL;
if (sp != NULL) {
kp->ki_sid = sp->s_sid;
SESS_LOCK(sp);
strlcpy(kp->ki_login, sp->s_login, sizeof(kp->ki_login));
if (sp->s_ttyvp)
kp->ki_kiflag |= KI_CTTY;
if (SESS_LEADER(p))
kp->ki_kiflag |= KI_SLEADER;
tp = sp->s_ttyp;
SESS_UNLOCK(sp);
}
if ((p->p_flag & P_CONTROLT) && tp != NULL) {
kp->ki_tdev = tty_udev(tp);
kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
kp->ki_tpgid = tp->t_pgrp ? tp->t_pgrp->pg_id : NO_PID;
if (tp->t_session)
kp->ki_tsid = tp->t_session->s_sid;
} else {
kp->ki_tdev = NODEV;
kp->ki_tdev_freebsd11 = kp->ki_tdev; /* truncate */
}
}
/*
* Fill in information that is thread specific. Must be called with
* target process locked. If 'preferthread' is set, overwrite certain
* process-related fields that are maintained for both threads and
* processes.
*/
static void
fill_kinfo_thread(struct thread *td, struct kinfo_proc *kp, int preferthread)
{
struct proc *p;
p = td->td_proc;
kp->ki_tdaddr = td;
PROC_LOCK_ASSERT(p, MA_OWNED);
if (preferthread)
PROC_STATLOCK(p);
thread_lock(td);
if (td->td_wmesg != NULL)
strlcpy(kp->ki_wmesg, td->td_wmesg, sizeof(kp->ki_wmesg));
else
bzero(kp->ki_wmesg, sizeof(kp->ki_wmesg));
if (strlcpy(kp->ki_tdname, td->td_name, sizeof(kp->ki_tdname)) >=
sizeof(kp->ki_tdname)) {
strlcpy(kp->ki_moretdname,
td->td_name + sizeof(kp->ki_tdname) - 1,
sizeof(kp->ki_moretdname));
} else {
bzero(kp->ki_moretdname, sizeof(kp->ki_moretdname));
}
if (TD_ON_LOCK(td)) {
kp->ki_kiflag |= KI_LOCKBLOCK;
strlcpy(kp->ki_lockname, td->td_lockname,
sizeof(kp->ki_lockname));
} else {
kp->ki_kiflag &= ~KI_LOCKBLOCK;
bzero(kp->ki_lockname, sizeof(kp->ki_lockname));
}
if (p->p_state == PRS_NORMAL) { /* approximate. */
if (TD_ON_RUNQ(td) ||
TD_CAN_RUN(td) ||
TD_IS_RUNNING(td)) {
kp->ki_stat = SRUN;
} else if (P_SHOULDSTOP(p)) {
kp->ki_stat = SSTOP;
} else if (TD_IS_SLEEPING(td)) {
kp->ki_stat = SSLEEP;
} else if (TD_ON_LOCK(td)) {
kp->ki_stat = SLOCK;
} else {
kp->ki_stat = SWAIT;
}
} else if (p->p_state == PRS_ZOMBIE) {
kp->ki_stat = SZOMB;
} else {
kp->ki_stat = SIDL;
}
/* Things in the thread */
kp->ki_wchan = td->td_wchan;
kp->ki_pri.pri_level = td->td_priority;
kp->ki_pri.pri_native = td->td_base_pri;
/*
* Note: legacy fields; clamp at the old NOCPU value and/or
* the maximum u_char CPU value.
*/
if (td->td_lastcpu == NOCPU)
kp->ki_lastcpu_old = NOCPU_OLD;
else if (td->td_lastcpu > MAXCPU_OLD)
kp->ki_lastcpu_old = MAXCPU_OLD;
else
kp->ki_lastcpu_old = td->td_lastcpu;
if (td->td_oncpu == NOCPU)
kp->ki_oncpu_old = NOCPU_OLD;
else if (td->td_oncpu > MAXCPU_OLD)
kp->ki_oncpu_old = MAXCPU_OLD;
else
kp->ki_oncpu_old = td->td_oncpu;
kp->ki_lastcpu = td->td_lastcpu;
kp->ki_oncpu = td->td_oncpu;
kp->ki_tdflags = td->td_flags;
kp->ki_tid = td->td_tid;
kp->ki_numthreads = p->p_numthreads;
kp->ki_pcb = td->td_pcb;
kp->ki_kstack = (void *)td->td_kstack;
kp->ki_slptime = (ticks - td->td_slptick) / hz;
kp->ki_pri.pri_class = td->td_pri_class;
kp->ki_pri.pri_user = td->td_user_pri;
if (preferthread) {
rufetchtd(td, &kp->ki_rusage);
kp->ki_runtime = cputick2usec(td->td_rux.rux_runtime);
kp->ki_pctcpu = sched_pctcpu(td);
kp->ki_estcpu = sched_estcpu(td);
kp->ki_cow = td->td_cow;
}
/* We can't get this anymore but ps etc never used it anyway. */
kp->ki_rqindex = 0;
if (preferthread)
kp->ki_siglist = td->td_siglist;
kp->ki_sigmask = td->td_sigmask;
thread_unlock(td);
if (preferthread)
PROC_STATUNLOCK(p);
}
/*
* Fill in a kinfo_proc structure for the specified process.
* Must be called with the target process locked.
*/
void
fill_kinfo_proc(struct proc *p, struct kinfo_proc *kp)
{
MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
bzero(kp, sizeof(*kp));
fill_kinfo_proc_pgrp(p,kp);
fill_kinfo_proc_only(p, kp);
fill_kinfo_thread(FIRST_THREAD_IN_PROC(p), kp, 0);
fill_kinfo_aggregate(p, kp);
}
struct pstats *
pstats_alloc(void)
{
return (malloc(sizeof(struct pstats), M_SUBPROC, M_ZERO|M_WAITOK));
}
/*
* Copy parts of p_stats; zero the rest of p_stats (statistics).
*/
void
pstats_fork(struct pstats *src, struct pstats *dst)
{
bzero(&dst->pstat_startzero,
__rangeof(struct pstats, pstat_startzero, pstat_endzero));
bcopy(&src->pstat_startcopy, &dst->pstat_startcopy,
__rangeof(struct pstats, pstat_startcopy, pstat_endcopy));
}
void
pstats_free(struct pstats *ps)
{
free(ps, M_SUBPROC);
}
#ifdef COMPAT_FREEBSD32
/*
* This function is typically used to copy out the kernel address, so
* it can be replaced by assignment of zero.
*/
static inline uint32_t
ptr32_trim(const void *ptr)
{
uintptr_t uptr;
uptr = (uintptr_t)ptr;
return ((uptr > UINT_MAX) ? 0 : uptr);
}
#define PTRTRIM_CP(src,dst,fld) \
do { (dst).fld = ptr32_trim((src).fld); } while (0)
static void
freebsd32_kinfo_proc_out(const struct kinfo_proc *ki, struct kinfo_proc32 *ki32)
{
int i;
bzero(ki32, sizeof(struct kinfo_proc32));
ki32->ki_structsize = sizeof(struct kinfo_proc32);
CP(*ki, *ki32, ki_layout);
PTRTRIM_CP(*ki, *ki32, ki_args);
PTRTRIM_CP(*ki, *ki32, ki_paddr);
PTRTRIM_CP(*ki, *ki32, ki_addr);
PTRTRIM_CP(*ki, *ki32, ki_tracep);
PTRTRIM_CP(*ki, *ki32, ki_textvp);
PTRTRIM_CP(*ki, *ki32, ki_fd);
PTRTRIM_CP(*ki, *ki32, ki_vmspace);
PTRTRIM_CP(*ki, *ki32, ki_wchan);
CP(*ki, *ki32, ki_pid);
CP(*ki, *ki32, ki_ppid);
CP(*ki, *ki32, ki_pgid);
CP(*ki, *ki32, ki_tpgid);
CP(*ki, *ki32, ki_sid);
CP(*ki, *ki32, ki_tsid);
CP(*ki, *ki32, ki_jobc);
CP(*ki, *ki32, ki_tdev);
CP(*ki, *ki32, ki_tdev_freebsd11);
CP(*ki, *ki32, ki_siglist);
CP(*ki, *ki32, ki_sigmask);
CP(*ki, *ki32, ki_sigignore);
CP(*ki, *ki32, ki_sigcatch);
CP(*ki, *ki32, ki_uid);
CP(*ki, *ki32, ki_ruid);
CP(*ki, *ki32, ki_svuid);
CP(*ki, *ki32, ki_rgid);
CP(*ki, *ki32, ki_svgid);
CP(*ki, *ki32, ki_ngroups);
for (i = 0; i < KI_NGROUPS; i++)
CP(*ki, *ki32, ki_groups[i]);
CP(*ki, *ki32, ki_size);
CP(*ki, *ki32, ki_rssize);
CP(*ki, *ki32, ki_swrss);
CP(*ki, *ki32, ki_tsize);
CP(*ki, *ki32, ki_dsize);
CP(*ki, *ki32, ki_ssize);
CP(*ki, *ki32, ki_xstat);
CP(*ki, *ki32, ki_acflag);
CP(*ki, *ki32, ki_pctcpu);
CP(*ki, *ki32, ki_estcpu);
CP(*ki, *ki32, ki_slptime);
CP(*ki, *ki32, ki_swtime);
CP(*ki, *ki32, ki_cow);
CP(*ki, *ki32, ki_runtime);
TV_CP(*ki, *ki32, ki_start);
TV_CP(*ki, *ki32, ki_childtime);
CP(*ki, *ki32, ki_flag);
CP(*ki, *ki32, ki_kiflag);
CP(*ki, *ki32, ki_traceflag);
CP(*ki, *ki32, ki_stat);
CP(*ki, *ki32, ki_nice);
CP(*ki, *ki32, ki_lock);
CP(*ki, *ki32, ki_rqindex);
CP(*ki, *ki32, ki_oncpu);
CP(*ki, *ki32, ki_lastcpu);
/* XXX TODO: wrap cpu value as appropriate */
CP(*ki, *ki32, ki_oncpu_old);
CP(*ki, *ki32, ki_lastcpu_old);
bcopy(ki->ki_tdname, ki32->ki_tdname, TDNAMLEN + 1);
bcopy(ki->ki_wmesg, ki32->ki_wmesg, WMESGLEN + 1);
bcopy(ki->ki_login, ki32->ki_login, LOGNAMELEN + 1);
bcopy(ki->ki_lockname, ki32->ki_lockname, LOCKNAMELEN + 1);
bcopy(ki->ki_comm, ki32->ki_comm, COMMLEN + 1);
bcopy(ki->ki_emul, ki32->ki_emul, KI_EMULNAMELEN + 1);
bcopy(ki->ki_loginclass, ki32->ki_loginclass, LOGINCLASSLEN + 1);
bcopy(ki->ki_moretdname, ki32->ki_moretdname, MAXCOMLEN - TDNAMLEN + 1);
CP(*ki, *ki32, ki_tracer);
CP(*ki, *ki32, ki_flag2);
CP(*ki, *ki32, ki_fibnum);
CP(*ki, *ki32, ki_cr_flags);
CP(*ki, *ki32, ki_jid);
CP(*ki, *ki32, ki_numthreads);
CP(*ki, *ki32, ki_tid);
CP(*ki, *ki32, ki_pri);
freebsd32_rusage_out(&ki->ki_rusage, &ki32->ki_rusage);
freebsd32_rusage_out(&ki->ki_rusage_ch, &ki32->ki_rusage_ch);
PTRTRIM_CP(*ki, *ki32, ki_pcb);
PTRTRIM_CP(*ki, *ki32, ki_kstack);
PTRTRIM_CP(*ki, *ki32, ki_udata);
PTRTRIM_CP(*ki, *ki32, ki_tdaddr);
CP(*ki, *ki32, ki_sflag);
CP(*ki, *ki32, ki_tdflags);
}
#endif
static ssize_t
kern_proc_out_size(struct proc *p, int flags)
{
ssize_t size = 0;
PROC_LOCK_ASSERT(p, MA_OWNED);
if ((flags & KERN_PROC_NOTHREADS) != 0) {
#ifdef COMPAT_FREEBSD32
if ((flags & KERN_PROC_MASK32) != 0) {
size += sizeof(struct kinfo_proc32);
} else
#endif
size += sizeof(struct kinfo_proc);
} else {
#ifdef COMPAT_FREEBSD32
if ((flags & KERN_PROC_MASK32) != 0)
size += sizeof(struct kinfo_proc32) * p->p_numthreads;
else
#endif
size += sizeof(struct kinfo_proc) * p->p_numthreads;
}
PROC_UNLOCK(p);
return (size);
}
int
kern_proc_out(struct proc *p, struct sbuf *sb, int flags)
{
struct thread *td;
struct kinfo_proc ki;
#ifdef COMPAT_FREEBSD32
struct kinfo_proc32 ki32;
#endif
int error;
PROC_LOCK_ASSERT(p, MA_OWNED);
MPASS(FIRST_THREAD_IN_PROC(p) != NULL);
error = 0;
fill_kinfo_proc(p, &ki);
if ((flags & KERN_PROC_NOTHREADS) != 0) {
#ifdef COMPAT_FREEBSD32
if ((flags & KERN_PROC_MASK32) != 0) {
freebsd32_kinfo_proc_out(&ki, &ki32);
if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
error = ENOMEM;
} else
#endif
if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
error = ENOMEM;
} else {
FOREACH_THREAD_IN_PROC(p, td) {
fill_kinfo_thread(td, &ki, 1);
#ifdef COMPAT_FREEBSD32
if ((flags & KERN_PROC_MASK32) != 0) {
freebsd32_kinfo_proc_out(&ki, &ki32);
if (sbuf_bcat(sb, &ki32, sizeof(ki32)) != 0)
error = ENOMEM;
} else
#endif
if (sbuf_bcat(sb, &ki, sizeof(ki)) != 0)
error = ENOMEM;
if (error != 0)
break;
}
}
PROC_UNLOCK(p);
return (error);
}
static int
sysctl_out_proc(struct proc *p, struct sysctl_req *req, int flags)
{
struct sbuf sb;
struct kinfo_proc ki;
int error, error2;
if (req->oldptr == NULL)
return (SYSCTL_OUT(req, 0, kern_proc_out_size(p, flags)));
sbuf_new_for_sysctl(&sb, (char *)&ki, sizeof(ki), req);
sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
error = kern_proc_out(p, &sb, flags);
error2 = sbuf_finish(&sb);
sbuf_delete(&sb);
if (error != 0)
return (error);
else if (error2 != 0)
return (error2);
return (0);
}
int
proc_iterate(int (*cb)(struct proc *, void *), void *cbarg)
{
struct proc *p;
int error, i, j;
for (i = 0; i < pidhashlock + 1; i++) {
sx_slock(&proctree_lock);
sx_slock(&pidhashtbl_lock[i]);
for (j = i; j <= pidhash; j += pidhashlock + 1) {
LIST_FOREACH(p, &pidhashtbl[j], p_hash) {
if (p->p_state == PRS_NEW)
continue;
error = cb(p, cbarg);
PROC_LOCK_ASSERT(p, MA_NOTOWNED);
if (error != 0) {
sx_sunlock(&pidhashtbl_lock[i]);
sx_sunlock(&proctree_lock);
return (error);
}
}
}
sx_sunlock(&pidhashtbl_lock[i]);
sx_sunlock(&proctree_lock);
}
return (0);
}
struct kern_proc_out_args {
struct sysctl_req *req;
int flags;
int oid_number;
int *name;
};
static int
sysctl_kern_proc_iterate(struct proc *p, void *origarg)
{
struct kern_proc_out_args *arg = origarg;
int *name = arg->name;
int oid_number = arg->oid_number;
int flags = arg->flags;
struct sysctl_req *req = arg->req;
int error = 0;
PROC_LOCK(p);
KASSERT(p->p_ucred != NULL,
("process credential is NULL for non-NEW proc"));
/*
* Show a user only appropriate processes.
*/
if (p_cansee(curthread, p))
goto skip;
/*
* TODO - make more efficient (see notes below).
* do by session.
*/
switch (oid_number) {
case KERN_PROC_GID:
if (p->p_ucred->cr_gid != (gid_t)name[0])
goto skip;
break;
case KERN_PROC_PGRP:
/* could do this by traversing pgrp */
if (p->p_pgrp == NULL ||
p->p_pgrp->pg_id != (pid_t)name[0])
goto skip;
break;
case KERN_PROC_RGID:
if (p->p_ucred->cr_rgid != (gid_t)name[0])
goto skip;
break;
case KERN_PROC_SESSION:
if (p->p_session == NULL ||
p->p_session->s_sid != (pid_t)name[0])
goto skip;
break;
case KERN_PROC_TTY:
if ((p->p_flag & P_CONTROLT) == 0 ||
p->p_session == NULL)
goto skip;
/* XXX proctree_lock */
SESS_LOCK(p->p_session);
if (p->p_session->s_ttyp == NULL ||
tty_udev(p->p_session->s_ttyp) !=
(dev_t)name[0]) {
SESS_UNLOCK(p->p_session);
goto skip;
}
SESS_UNLOCK(p->p_session);
break;
case KERN_PROC_UID:
if (p->p_ucred->cr_uid != (uid_t)name[0])
goto skip;
break;
case KERN_PROC_RUID:
if (p->p_ucred->cr_ruid != (uid_t)name[0])
goto skip;
break;
case KERN_PROC_PROC:
break;
default:
break;
}
error = sysctl_out_proc(p, req, flags);
PROC_LOCK_ASSERT(p, MA_NOTOWNED);
return (error);
skip:
PROC_UNLOCK(p);
return (0);
}
static int
sysctl_kern_proc(SYSCTL_HANDLER_ARGS)
{
struct kern_proc_out_args iterarg;
int *name = (int *)arg1;
u_int namelen = arg2;
struct proc *p;
int flags, oid_number;
int error = 0;
oid_number = oidp->oid_number;
if (oid_number != KERN_PROC_ALL &&
(oid_number & KERN_PROC_INC_THREAD) == 0)
flags = KERN_PROC_NOTHREADS;
else {
flags = 0;
oid_number &= ~KERN_PROC_INC_THREAD;
}
#ifdef COMPAT_FREEBSD32
if (req->flags & SCTL_MASK32)
flags |= KERN_PROC_MASK32;
#endif
if (oid_number == KERN_PROC_PID) {
if (namelen != 1)
return (EINVAL);
error = sysctl_wire_old_buffer(req, 0);
if (error)
return (error);
sx_slock(&proctree_lock);
error = pget((pid_t)name[0], PGET_CANSEE, &p);
if (error == 0)
error = sysctl_out_proc(p, req, flags);
sx_sunlock(&proctree_lock);
return (error);
}
switch (oid_number) {
case KERN_PROC_ALL:
if (namelen != 0)
return (EINVAL);
break;
case KERN_PROC_PROC:
if (namelen != 0 && namelen != 1)
return (EINVAL);
break;
default:
if (namelen != 1)
return (EINVAL);
break;
}
if (req->oldptr == NULL) {
/* overestimate by 5 procs */
error = SYSCTL_OUT(req, 0, sizeof (struct kinfo_proc) * 5);
if (error)
return (error);
} else {
error = sysctl_wire_old_buffer(req, 0);
if (error != 0)
return (error);
}
iterarg.flags = flags;
iterarg.oid_number = oid_number;
iterarg.req = req;
iterarg.name = name;
error = proc_iterate(sysctl_kern_proc_iterate, &iterarg);
return (error);
}
struct pargs *
pargs_alloc(int len)
{
struct pargs *pa;
pa = malloc(sizeof(struct pargs) + len, M_PARGS,
M_WAITOK);
refcount_init(&pa->ar_ref, 1);
pa->ar_length = len;
return (pa);
}
static void
pargs_free(struct pargs *pa)
{
free(pa, M_PARGS);
}
void
pargs_hold(struct pargs *pa)
{
if (pa == NULL)
return;
refcount_acquire(&pa->ar_ref);
}
void
pargs_drop(struct pargs *pa)
{
if (pa == NULL)
return;
if (refcount_release(&pa->ar_ref))
pargs_free(pa);
}
static int
proc_read_string(struct thread *td, struct proc *p, const char *sptr, char *buf,
size_t len)
{
ssize_t n;
/*
* This may return a short read if the string is shorter than the chunk
* and is aligned at the end of the page, and the following page is not
* mapped.
*/
n = proc_readmem(td, p, (vm_offset_t)sptr, buf, len);
if (n <= 0)
return (ENOMEM);
return (0);
}
#define PROC_AUXV_MAX 256 /* Safety limit on auxv size. */
enum proc_vector_type {
PROC_ARG,
PROC_ENV,
PROC_AUX,
};
#ifdef COMPAT_FREEBSD32
static int
get_proc_vector32(struct thread *td, struct proc *p, char ***proc_vectorp,
size_t *vsizep, enum proc_vector_type type)
{
struct freebsd32_ps_strings pss;
Elf32_Auxinfo aux;
vm_offset_t vptr, ptr;
uint32_t *proc_vector32;
char **proc_vector;
size_t vsize, size;
int i, error;
error = 0;
if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
sizeof(pss)) != sizeof(pss))
return (ENOMEM);
switch (type) {
case PROC_ARG:
vptr = (vm_offset_t)PTRIN(pss.ps_argvstr);
vsize = pss.ps_nargvstr;
if (vsize > ARG_MAX)
return (ENOEXEC);
size = vsize * sizeof(int32_t);
break;
case PROC_ENV:
vptr = (vm_offset_t)PTRIN(pss.ps_envstr);
vsize = pss.ps_nenvstr;
if (vsize > ARG_MAX)
return (ENOEXEC);
size = vsize * sizeof(int32_t);
break;
case PROC_AUX:
vptr = (vm_offset_t)PTRIN(pss.ps_envstr) +
(pss.ps_nenvstr + 1) * sizeof(int32_t);
if (vptr % 4 != 0)
return (ENOEXEC);
for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
sizeof(aux))
return (ENOMEM);
if (aux.a_type == AT_NULL)
break;
ptr += sizeof(aux);
}
if (aux.a_type != AT_NULL)
return (ENOEXEC);
vsize = i + 1;
size = vsize * sizeof(aux);
break;
default:
KASSERT(0, ("Wrong proc vector type: %d", type));
return (EINVAL);
}
proc_vector32 = malloc(size, M_TEMP, M_WAITOK);
if (proc_readmem(td, p, vptr, proc_vector32, size) != size) {
error = ENOMEM;
goto done;
}
if (type == PROC_AUX) {
*proc_vectorp = (char **)proc_vector32;
*vsizep = vsize;
return (0);
}
proc_vector = malloc(vsize * sizeof(char *), M_TEMP, M_WAITOK);
for (i = 0; i < (int)vsize; i++)
proc_vector[i] = PTRIN(proc_vector32[i]);
*proc_vectorp = proc_vector;
*vsizep = vsize;
done:
free(proc_vector32, M_TEMP);
return (error);
}
#endif
static int
get_proc_vector(struct thread *td, struct proc *p, char ***proc_vectorp,
size_t *vsizep, enum proc_vector_type type)
{
struct ps_strings pss;
Elf_Auxinfo aux;
vm_offset_t vptr, ptr;
char **proc_vector;
size_t vsize, size;
int i;
#ifdef COMPAT_FREEBSD32
if (SV_PROC_FLAG(p, SV_ILP32) != 0)
return (get_proc_vector32(td, p, proc_vectorp, vsizep, type));
#endif
if (proc_readmem(td, p, (vm_offset_t)p->p_sysent->sv_psstrings, &pss,
sizeof(pss)) != sizeof(pss))
return (ENOMEM);
switch (type) {
case PROC_ARG:
vptr = (vm_offset_t)pss.ps_argvstr;
vsize = pss.ps_nargvstr;
if (vsize > ARG_MAX)
return (ENOEXEC);
size = vsize * sizeof(char *);
break;
case PROC_ENV:
vptr = (vm_offset_t)pss.ps_envstr;
vsize = pss.ps_nenvstr;
if (vsize > ARG_MAX)
return (ENOEXEC);
size = vsize * sizeof(char *);
break;
case PROC_AUX:
/*
* The aux array is just above env array on the stack. Check
* that the address is naturally aligned.
*/
vptr = (vm_offset_t)pss.ps_envstr + (pss.ps_nenvstr + 1)
* sizeof(char *);
#if __ELF_WORD_SIZE == 64
if (vptr % sizeof(uint64_t) != 0)
#else
if (vptr % sizeof(uint32_t) != 0)
#endif
return (ENOEXEC);
/*
* We count the array size reading the aux vectors from the
* stack until AT_NULL vector is returned. So (to keep the code
* simple) we read the process stack twice: the first time here
* to find the size and the second time when copying the vectors
* to the allocated proc_vector.
*/
for (ptr = vptr, i = 0; i < PROC_AUXV_MAX; i++) {
if (proc_readmem(td, p, ptr, &aux, sizeof(aux)) !=
sizeof(aux))
return (ENOMEM);
if (aux.a_type == AT_NULL)
break;
ptr += sizeof(aux);
}
/*
* If the PROC_AUXV_MAX entries are iterated over, and we have
* not reached AT_NULL, it is most likely we are reading wrong
* data: either the process doesn't have auxv array or data has
* been modified. Return the error in this case.
*/
if (aux.a_type != AT_NULL)
return (ENOEXEC);
vsize = i + 1;
size = vsize * sizeof(aux);
break;
default:
KASSERT(0, ("Wrong proc vector type: %d", type));
return (EINVAL); /* In case we are built without INVARIANTS. */
}
proc_vector = malloc(size, M_TEMP, M_WAITOK);
if (proc_readmem(td, p, vptr, proc_vector, size) != size) {
free(proc_vector, M_TEMP);
return (ENOMEM);
}
*proc_vectorp = proc_vector;
*vsizep = vsize;
return (0);
}
#define GET_PS_STRINGS_CHUNK_SZ 256 /* Chunk size (bytes) for ps_strings operations. */
static int
get_ps_strings(struct thread *td, struct proc *p, struct sbuf *sb,
enum proc_vector_type type)
{
size_t done, len, nchr, vsize;
int error, i;
char **proc_vector, *sptr;
char pss_string[GET_PS_STRINGS_CHUNK_SZ];
PROC_ASSERT_HELD(p);
/*
* We are not going to read more than 2 * (PATH_MAX + ARG_MAX) bytes.
*/
nchr = 2 * (PATH_MAX + ARG_MAX);
error = get_proc_vector(td, p, &proc_vector, &vsize, type);
if (error != 0)
return (error);
for (done = 0, i = 0; i < (int)vsize && done < nchr; i++) {
/*
* The program may have scribbled into its argv array, e.g. to
* remove some arguments. If that has happened, break out
* before trying to read from NULL.
*/
if (proc_vector[i] == NULL)
break;
for (sptr = proc_vector[i]; ; sptr += GET_PS_STRINGS_CHUNK_SZ) {
error = proc_read_string(td, p, sptr, pss_string,
sizeof(pss_string));
if (error != 0)
goto done;
len = strnlen(pss_string, GET_PS_STRINGS_CHUNK_SZ);
if (done + len >= nchr)
len = nchr - done - 1;
sbuf_bcat(sb, pss_string, len);
if (len != GET_PS_STRINGS_CHUNK_SZ)
break;
done += GET_PS_STRINGS_CHUNK_SZ;
}
sbuf_bcat(sb, "", 1);
done += len + 1;
}
done:
free(proc_vector, M_TEMP);
return (error);
}
int
proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb)
{
return (get_ps_strings(curthread, p, sb, PROC_ARG));
}
int
proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb)
{
return (get_ps_strings(curthread, p, sb, PROC_ENV));
}
int
proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb)
{
size_t vsize, size;
char **auxv;
int error;
error = get_proc_vector(td, p, &auxv, &vsize, PROC_AUX);
if (error == 0) {
#ifdef COMPAT_FREEBSD32
if (SV_PROC_FLAG(p, SV_ILP32) != 0)
size = vsize * sizeof(Elf32_Auxinfo);
else
#endif
size = vsize * sizeof(Elf_Auxinfo);
if (sbuf_bcat(sb, auxv, size) != 0)
error = ENOMEM;
free(auxv, M_TEMP);
}
return (error);
}
/*
* This sysctl allows a process to retrieve the argument list or process
* title for another process without groping around in the address space
* of the other process. It also allow a process to set its own "process
* title to a string of its own choice.
*/
static int
sysctl_kern_proc_args(SYSCTL_HANDLER_ARGS)
{
int *name = (int *)arg1;
u_int namelen = arg2;
struct pargs *newpa, *pa;
struct proc *p;
struct sbuf sb;
int flags, error = 0, error2;
pid_t pid;
if (namelen != 1)
return (EINVAL);
p = curproc;
pid = (pid_t)name[0];
if (pid == -1) {
pid = p->p_pid;
}
/*
* If the query is for this process and it is single-threaded, there
* is nobody to modify pargs, thus we can just read.
*/
if (pid == p->p_pid && p->p_numthreads == 1 && req->newptr == NULL &&
(pa = p->p_args) != NULL)
return (SYSCTL_OUT(req, pa->ar_args, pa->ar_length));
flags = PGET_CANSEE;
if (req->newptr != NULL)
flags |= PGET_ISCURRENT;
error = pget(pid, flags, &p);
if (error)
return (error);
pa = p->p_args;
if (pa != NULL) {
pargs_hold(pa);
PROC_UNLOCK(p);
error = SYSCTL_OUT(req, pa->ar_args, pa->ar_length);
pargs_drop(pa);
} else if ((p->p_flag & (P_WEXIT | P_SYSTEM)) == 0) {
_PHOLD(p);
PROC_UNLOCK(p);
sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
error = proc_getargv(curthread, p, &sb);
error2 = sbuf_finish(&sb);
PRELE(p);
sbuf_delete(&sb);
if (error == 0 && error2 != 0)
error = error2;
} else {
PROC_UNLOCK(p);
}
if (error != 0 || req->newptr == NULL)
return (error);
if (req->newlen > ps_arg_cache_limit - sizeof(struct pargs))
return (ENOMEM);
if (req->newlen == 0) {
/*
* Clear the argument pointer, so that we'll fetch arguments
* with proc_getargv() until further notice.
*/
newpa = NULL;
} else {
newpa = pargs_alloc(req->newlen);
error = SYSCTL_IN(req, newpa->ar_args, req->newlen);
if (error != 0) {
pargs_free(newpa);
return (error);
}
}
PROC_LOCK(p);
pa = p->p_args;
p->p_args = newpa;
PROC_UNLOCK(p);
pargs_drop(pa);
return (0);
}
/*
* This sysctl allows a process to retrieve environment of another process.
*/
static int
sysctl_kern_proc_env(SYSCTL_HANDLER_ARGS)
{
int *name = (int *)arg1;
u_int namelen = arg2;
struct proc *p;
struct sbuf sb;
int error, error2;
if (namelen != 1)
return (EINVAL);
error = pget((pid_t)name[0], PGET_WANTREAD, &p);
if (error != 0)
return (error);
if ((p->p_flag & P_SYSTEM) != 0) {
PRELE(p);
return (0);
}
sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
error = proc_getenvv(curthread, p, &sb);
error2 = sbuf_finish(&sb);
PRELE(p);
sbuf_delete(&sb);
return (error != 0 ? error : error2);
}
/*
* This sysctl allows a process to retrieve ELF auxiliary vector of
* another process.
*/
static int
sysctl_kern_proc_auxv(SYSCTL_HANDLER_ARGS)
{
int *name = (int *)arg1;
u_int namelen = arg2;
struct proc *p;
struct sbuf sb;
int error, error2;
if (namelen != 1)
return (EINVAL);
error = pget((pid_t)name[0], PGET_WANTREAD, &p);
if (error != 0)
return (error);
if ((p->p_flag & P_SYSTEM) != 0) {
PRELE(p);
return (0);
}
sbuf_new_for_sysctl(&sb, NULL, GET_PS_STRINGS_CHUNK_SZ, req);
sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
error = proc_getauxv(curthread, p, &sb);
error2 = sbuf_finish(&sb);
PRELE(p);
sbuf_delete(&sb);
return (error != 0 ? error : error2);
}
/*
* This sysctl allows a process to retrieve the path of the executable for
* itself or another process.
*/
static int
sysctl_kern_proc_pathname(SYSCTL_HANDLER_ARGS)
{
pid_t *pidp = (pid_t *)arg1;
unsigned int arglen = arg2;
struct proc *p;
struct vnode *vp;
char *retbuf, *freebuf;
int error;
if (arglen != 1)
return (EINVAL);
if (*pidp == -1) { /* -1 means this process */
p = req->td->td_proc;
} else {
error = pget(*pidp, PGET_CANSEE, &p);
if (error != 0)
return (error);
}
vp = p->p_textvp;
if (vp == NULL) {
if (*pidp != -1)
PROC_UNLOCK(p);
return (0);
}
vref(vp);
if (*pidp != -1)
PROC_UNLOCK(p);
error = vn_fullpath(vp, &retbuf, &freebuf);
vrele(vp);
if (error)
return (error);
error = SYSCTL_OUT(req, retbuf, strlen(retbuf) + 1);
free(freebuf, M_TEMP);
return (error);
}
static int
sysctl_kern_proc_sv_name(SYSCTL_HANDLER_ARGS)
{
struct proc *p;
char *sv_name;
int *name;
int namelen;
int error;
namelen = arg2;
if (namelen != 1)
return (EINVAL);
name = (int *)arg1;
error = pget((pid_t)name[0], PGET_CANSEE, &p);
if (error != 0)
return (error);
sv_name = p->p_sysent->sv_name;
PROC_UNLOCK(p);
return (sysctl_handle_string(oidp, sv_name, 0, req));
}
#ifdef KINFO_OVMENTRY_SIZE
CTASSERT(sizeof(struct kinfo_ovmentry) == KINFO_OVMENTRY_SIZE);
#endif
#ifdef COMPAT_FREEBSD7
static int
sysctl_kern_proc_ovmmap(SYSCTL_HANDLER_ARGS)
{
vm_map_entry_t entry, tmp_entry;
unsigned int last_timestamp;
char *fullpath, *freepath;
struct kinfo_ovmentry *kve;
struct vattr va;
struct ucred *cred;
int error, *name;
struct vnode *vp;
struct proc *p;
vm_map_t map;
struct vmspace *vm;
name = (int *)arg1;
error = pget((pid_t)name[0], PGET_WANTREAD, &p);
if (error != 0)
return (error);
vm = vmspace_acquire_ref(p);
if (vm == NULL) {
PRELE(p);
return (ESRCH);
}
kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK);
map = &vm->vm_map;
vm_map_lock_read(map);
VM_MAP_ENTRY_FOREACH(entry, map) {
vm_object_t obj, tobj, lobj;
vm_offset_t addr;
if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
continue;
bzero(kve, sizeof(*kve));
kve->kve_structsize = sizeof(*kve);
kve->kve_private_resident = 0;
obj = entry->object.vm_object;
if (obj != NULL) {
VM_OBJECT_RLOCK(obj);
if (obj->shadow_count == 1)
kve->kve_private_resident =
obj->resident_page_count;
}
kve->kve_resident = 0;
addr = entry->start;
while (addr < entry->end) {
if (pmap_extract(map->pmap, addr))
kve->kve_resident++;
addr += PAGE_SIZE;
}
for (lobj = tobj = obj; tobj; tobj = tobj->backing_object) {
if (tobj != obj) {
VM_OBJECT_RLOCK(tobj);
kve->kve_offset += tobj->backing_object_offset;
}
if (lobj != obj)
VM_OBJECT_RUNLOCK(lobj);
lobj = tobj;
}
kve->kve_start = (void*)entry->start;
kve->kve_end = (void*)entry->end;
kve->kve_offset += (off_t)entry->offset;
if (entry->protection & VM_PROT_READ)
kve->kve_protection |= KVME_PROT_READ;
if (entry->protection & VM_PROT_WRITE)
kve->kve_protection |= KVME_PROT_WRITE;
if (entry->protection & VM_PROT_EXECUTE)
kve->kve_protection |= KVME_PROT_EXEC;
if (entry->eflags & MAP_ENTRY_COW)
kve->kve_flags |= KVME_FLAG_COW;
if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
last_timestamp = map->timestamp;
vm_map_unlock_read(map);
kve->kve_fileid = 0;
kve->kve_fsid = 0;
freepath = NULL;
fullpath = "";
if (lobj) {
kve->kve_type = vm_object_kvme_type(lobj, &vp);
if (kve->kve_type == KVME_TYPE_MGTDEVICE)
kve->kve_type = KVME_TYPE_UNKNOWN;
if (vp != NULL)
vref(vp);
if (lobj != obj)
VM_OBJECT_RUNLOCK(lobj);
kve->kve_ref_count = obj->ref_count;
kve->kve_shadow_count = obj->shadow_count;
VM_OBJECT_RUNLOCK(obj);
if (vp != NULL) {
vn_fullpath(vp, &fullpath, &freepath);
cred = curthread->td_ucred;
vn_lock(vp, LK_SHARED | LK_RETRY);
if (VOP_GETATTR(vp, &va, cred) == 0) {
kve->kve_fileid = va.va_fileid;
/* truncate */
kve->kve_fsid = va.va_fsid;
}
vput(vp);
}
} else {
kve->kve_type = KVME_TYPE_NONE;
kve->kve_ref_count = 0;
kve->kve_shadow_count = 0;
}
strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
if (freepath != NULL)
free(freepath, M_TEMP);
error = SYSCTL_OUT(req, kve, sizeof(*kve));
vm_map_lock_read(map);
if (error)
break;
if (last_timestamp != map->timestamp) {
vm_map_lookup_entry(map, addr - 1, &tmp_entry);
entry = tmp_entry;
}
}
vm_map_unlock_read(map);
vmspace_free(vm);
PRELE(p);
free(kve, M_TEMP);
return (error);
}
#endif /* COMPAT_FREEBSD7 */
#ifdef KINFO_VMENTRY_SIZE
CTASSERT(sizeof(struct kinfo_vmentry) == KINFO_VMENTRY_SIZE);
#endif
void
kern_proc_vmmap_resident(vm_map_t map, vm_map_entry_t entry,
int *resident_count, bool *super)
{
vm_object_t obj, tobj;
vm_page_t m, m_adv;
vm_offset_t addr;
vm_paddr_t pa;
vm_pindex_t pi, pi_adv, pindex;
*super = false;
*resident_count = 0;
if (vmmap_skip_res_cnt)
return;
pa = 0;
obj = entry->object.vm_object;
addr = entry->start;
m_adv = NULL;
pi = OFF_TO_IDX(entry->offset);
for (; addr < entry->end; addr += IDX_TO_OFF(pi_adv), pi += pi_adv) {
if (m_adv != NULL) {
m = m_adv;
} else {
pi_adv = atop(entry->end - addr);
pindex = pi;
for (tobj = obj;; tobj = tobj->backing_object) {
m = vm_page_find_least(tobj, pindex);
if (m != NULL) {
if (m->pindex == pindex)
break;
if (pi_adv > m->pindex - pindex) {
pi_adv = m->pindex - pindex;
m_adv = m;
}
}
if (tobj->backing_object == NULL)
goto next;
pindex += OFF_TO_IDX(tobj->
backing_object_offset);
}
}
m_adv = NULL;
if (m->psind != 0 && addr + pagesizes[1] <= entry->end &&
(addr & (pagesizes[1] - 1)) == 0 &&
(pmap_mincore(map->pmap, addr, &pa) & MINCORE_SUPER) != 0) {
*super = true;
pi_adv = atop(pagesizes[1]);
} else {
/*
* We do not test the found page on validity.
* Either the page is busy and being paged in,
* or it was invalidated. The first case
* should be counted as resident, the second
* is not so clear; we do account both.
*/
pi_adv = 1;
}
*resident_count += pi_adv;
next:;
}
}
/*
* Must be called with the process locked and will return unlocked.
*/
int
kern_proc_vmmap_out(struct proc *p, struct sbuf *sb, ssize_t maxlen, int flags)
{
vm_map_entry_t entry, tmp_entry;
struct vattr va;
vm_map_t map;
vm_object_t lobj, nobj, obj, tobj;
char *fullpath, *freepath;
struct kinfo_vmentry *kve;
struct ucred *cred;
struct vnode *vp;
struct vmspace *vm;
vm_offset_t addr;
unsigned int last_timestamp;
int error;
bool guard, super;
PROC_LOCK_ASSERT(p, MA_OWNED);
_PHOLD(p);
PROC_UNLOCK(p);
vm = vmspace_acquire_ref(p);
if (vm == NULL) {
PRELE(p);
return (ESRCH);
}
kve = malloc(sizeof(*kve), M_TEMP, M_WAITOK | M_ZERO);
error = 0;
map = &vm->vm_map;
vm_map_lock_read(map);
VM_MAP_ENTRY_FOREACH(entry, map) {
if (entry->eflags & MAP_ENTRY_IS_SUB_MAP)
continue;
addr = entry->end;
bzero(kve, sizeof(*kve));
obj = entry->object.vm_object;
if (obj != NULL) {
if ((obj->flags & OBJ_ANON) != 0)
kve->kve_obj = (uintptr_t)obj;
for (tobj = obj; tobj != NULL;
tobj = tobj->backing_object) {
VM_OBJECT_RLOCK(tobj);
kve->kve_offset += tobj->backing_object_offset;
lobj = tobj;
}
if (obj->backing_object == NULL)
kve->kve_private_resident =
obj->resident_page_count;
kern_proc_vmmap_resident(map, entry,
&kve->kve_resident, &super);
if (super)
kve->kve_flags |= KVME_FLAG_SUPER;
for (tobj = obj; tobj != NULL; tobj = nobj) {
nobj = tobj->backing_object;
if (tobj != obj && tobj != lobj)
VM_OBJECT_RUNLOCK(tobj);
}
} else {
lobj = NULL;
}
kve->kve_start = entry->start;
kve->kve_end = entry->end;
kve->kve_offset += entry->offset;
if (entry->protection & VM_PROT_READ)
kve->kve_protection |= KVME_PROT_READ;
if (entry->protection & VM_PROT_WRITE)
kve->kve_protection |= KVME_PROT_WRITE;
if (entry->protection & VM_PROT_EXECUTE)
kve->kve_protection |= KVME_PROT_EXEC;
if (entry->eflags & MAP_ENTRY_COW)
kve->kve_flags |= KVME_FLAG_COW;
if (entry->eflags & MAP_ENTRY_NEEDS_COPY)
kve->kve_flags |= KVME_FLAG_NEEDS_COPY;
if (entry->eflags & MAP_ENTRY_NOCOREDUMP)
kve->kve_flags |= KVME_FLAG_NOCOREDUMP;
if (entry->eflags & MAP_ENTRY_GROWS_UP)
kve->kve_flags |= KVME_FLAG_GROWS_UP;
if (entry->eflags & MAP_ENTRY_GROWS_DOWN)
kve->kve_flags |= KVME_FLAG_GROWS_DOWN;
if (entry->eflags & MAP_ENTRY_USER_WIRED)
kve->kve_flags |= KVME_FLAG_USER_WIRED;
guard = (entry->eflags & MAP_ENTRY_GUARD) != 0;
last_timestamp = map->timestamp;
vm_map_unlock_read(map);
freepath = NULL;
fullpath = "";
if (lobj != NULL) {
kve->kve_type = vm_object_kvme_type(lobj, &vp);
if (vp != NULL)
vref(vp);
if (lobj != obj)
VM_OBJECT_RUNLOCK(lobj);
kve->kve_ref_count = obj->ref_count;
kve->kve_shadow_count = obj->shadow_count;
VM_OBJECT_RUNLOCK(obj);
if (vp != NULL) {
vn_fullpath(vp, &fullpath, &freepath);
kve->kve_vn_type = vntype_to_kinfo(vp->v_type);
cred = curthread->td_ucred;
vn_lock(vp, LK_SHARED | LK_RETRY);
if (VOP_GETATTR(vp, &va, cred) == 0) {
kve->kve_vn_fileid = va.va_fileid;
kve->kve_vn_fsid = va.va_fsid;
kve->kve_vn_fsid_freebsd11 =
kve->kve_vn_fsid; /* truncate */
kve->kve_vn_mode =
MAKEIMODE(va.va_type, va.va_mode);
kve->kve_vn_size = va.va_size;
kve->kve_vn_rdev = va.va_rdev;
kve->kve_vn_rdev_freebsd11 =
kve->kve_vn_rdev; /* truncate */
kve->kve_status = KF_ATTR_VALID;
}
vput(vp);
}
} else {
kve->kve_type = guard ? KVME_TYPE_GUARD :
KVME_TYPE_NONE;
kve->kve_ref_count = 0;
kve->kve_shadow_count = 0;
}
strlcpy(kve->kve_path, fullpath, sizeof(kve->kve_path));
if (freepath != NULL)
free(freepath, M_TEMP);
/* Pack record size down */
if ((flags & KERN_VMMAP_PACK_KINFO) != 0)
kve->kve_structsize =
offsetof(struct kinfo_vmentry, kve_path) +
strlen(kve->kve_path) + 1;
else
kve->kve_structsize = sizeof(*kve);
kve->kve_structsize = roundup(kve->kve_structsize,
sizeof(uint64_t));
/* Halt filling and truncate rather than exceeding maxlen */
if (maxlen != -1 && maxlen < kve->kve_structsize) {
error = 0;
vm_map_lock_read(map);
break;
} else if (maxlen != -1)
maxlen -= kve->kve_structsize;
if (sbuf_bcat(sb, kve, kve->kve_structsize) != 0)
error = ENOMEM;
vm_map_lock_read(map);
if (error != 0)
break;
if (last_timestamp != map->timestamp) {
vm_map_lookup_entry(map, addr - 1, &tmp_entry);
entry = tmp_entry;
}
}
vm_map_unlock_read(map);
vmspace_free(vm);
PRELE(p);
free(kve, M_TEMP);
return (error);
}
static int
sysctl_kern_proc_vmmap(SYSCTL_HANDLER_ARGS)
{
struct proc *p;
struct sbuf sb;
int error, error2, *name;
name = (int *)arg1;
sbuf_new_for_sysctl(&sb, NULL, sizeof(struct kinfo_vmentry), req);
sbuf_clear_flags(&sb, SBUF_INCLUDENUL);
error = pget((pid_t)name[0], PGET_CANDEBUG | PGET_NOTWEXIT, &p);
if (error != 0) {
sbuf_delete(&sb);
return (error);
}
error = kern_proc_vmmap_out(p, &sb, -1, KERN_VMMAP_PACK_KINFO);
error2 = sbuf_finish(&sb);
sbuf_delete(&sb);
return (error != 0 ? error : error2);
}
#if defined(STACK) || defined(DDB)
static int
sysctl_kern_proc_kstack(SYSCTL_HANDLER_ARGS)
{
struct kinfo_kstack *kkstp;
int error, i, *name, numthreads;
lwpid_t *lwpidarray;
struct thread *td;
struct stack *st;
struct sbuf sb;
struct proc *p;
name = (int *)arg1;
error = pget((pid_t)name[0], PGET_NOTINEXEC | PGET_WANTREAD, &p);
if (error != 0)
return (error);
kkstp = malloc(sizeof(*kkstp), M_TEMP, M_WAITOK);
st = stack_create(M_WAITOK);
lwpidarray = NULL;
PROC_LOCK(p);
do {
if (lwpidarray != NULL) {
free(lwpidarray, M_TEMP);
lwpidarray = NULL;
}
numthreads = p->p_numthreads;
PROC_UNLOCK(p);
lwpidarray = malloc(sizeof(*lwpidarray) * numthreads, M_TEMP,
M_WAITOK | M_ZERO);
PROC_LOCK(p);
} while (numthreads < p->p_numthreads);
/*
* XXXRW: During the below loop, execve(2) and countless other sorts
* of changes could have taken place. Should we check to see if the
* vmspace has been replaced, or the like, in order to prevent
* giving a snapshot that spans, say, execve(2), with some threads
* before and some after? Among other things, the credentials could
* have changed, in which case the right to extract debug info might
* no longer be assured.
*/
i = 0;
FOREACH_THREAD_IN_PROC(p, td) {
KASSERT(i < numthreads,
("sysctl_kern_proc_kstack: numthreads"));
lwpidarray[i] = td->td_tid;
i++;
}
PROC_UNLOCK(p);
numthreads = i;
for (i = 0; i < numthreads; i++) {
td = tdfind(lwpidarray[i], p->p_pid);
if (td == NULL) {
continue;
}
bzero(kkstp, sizeof(*kkstp));
(void)sbuf_new(&sb, kkstp->kkst_trace,
sizeof(kkstp->kkst_trace), SBUF_FIXEDLEN);
thread_lock(td);
kkstp->kkst_tid = td->td_tid;
if (TD_IS_SWAPPED(td))
kkstp->kkst_state = KKST_STATE_SWAPPED;
else if (stack_save_td(st, td) == 0)
kkstp->kkst_state = KKST_STATE_STACKOK;
else
kkstp->kkst_state = KKST_STATE_RUNNING;
thread_unlock(td);
PROC_UNLOCK(p);
stack_sbuf_print(&sb, st);
sbuf_finish(&sb);
sbuf_delete(&sb);
error = SYSCTL_OUT(req, kkstp, sizeof(*kkstp));
if (error)
break;
}
PRELE(p);
if (lwpidarray != NULL)
free(lwpidarray, M_TEMP);
stack_destroy(st);
free(kkstp, M_TEMP);
return (error);
}
#endif
/*
* This sysctl allows a process to retrieve the full list of groups from
* itself or another process.
*/
static int
sysctl_kern_proc_groups(SYSCTL_HANDLER_ARGS)
{
pid_t *pidp = (pid_t *)arg1;
unsigned int arglen = arg2;
struct proc *p;
struct ucred *cred;
int error;
if (arglen != 1)
return (EINVAL);
if (*pidp == -1) { /* -1 means this process */
p = req->td->td_proc;
PROC_LOCK(p);
} else {
error = pget(*pidp, PGET_CANSEE, &p);
if (error != 0)
return (error);
}
cred = crhold(p->p_ucred);
PROC_UNLOCK(p);
error = SYSCTL_OUT(req, cred->cr_groups,
cred->cr_ngroups * sizeof(gid_t));
crfree(cred);
return (error);
}
/*
* This sysctl allows a process to retrieve or/and set the resource limit for
* another process.
*/
static int
sysctl_kern_proc_rlimit(SYSCTL_HANDLER_ARGS)
{
int *name = (int *)arg1;
u_int namelen = arg2;
struct rlimit rlim;
struct proc *p;
u_int which;
int flags, error;
if (namelen != 2)
return (EINVAL);
which = (u_int)name[1];
if (which >= RLIM_NLIMITS)
return (EINVAL);
if (req->newptr != NULL && req->newlen != sizeof(rlim))
return (EINVAL);
flags = PGET_HOLD | PGET_NOTWEXIT;
if (req->newptr != NULL)
flags |= PGET_CANDEBUG;
else
flags |= PGET_CANSEE;
error = pget((pid_t)name[0], flags, &p);
if (error != 0)
return (error);
/*
* Retrieve limit.
*/
if (req->oldptr != NULL) {
PROC_LOCK(p);
lim_rlimit_proc(p, which, &rlim);
PROC_UNLOCK(p);
}
error = SYSCTL_OUT(req, &rlim, sizeof(rlim));
if (error != 0)
goto errout;
/*
* Set limit.
*/
if (req->newptr != NULL) {
error = SYSCTL_IN(req, &rlim, sizeof(rlim));
if (error == 0)
error = kern_proc_setrlimit(curthread, p, which, &rlim);
}
errout:
PRELE(p);
return (error);
}
/*
* This sysctl allows a process to retrieve ps_strings structure location of
* another process.
*/
static int
sysctl_kern_proc_ps_strings(SYSCTL_HANDLER_ARGS)
{
int *name = (int *)arg1;
u_int namelen = arg2;
struct proc *p;
vm_offset_t ps_strings;
int error;
#ifdef COMPAT_FREEBSD32
uint32_t ps_strings32;
#endif
if (namelen != 1)
return (EINVAL);
error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
if (error != 0)
return (error);
#ifdef COMPAT_FREEBSD32
if ((req->flags & SCTL_MASK32) != 0) {
/*
* We return 0 if the 32 bit emulation request is for a 64 bit
* process.
*/
ps_strings32 = SV_PROC_FLAG(p, SV_ILP32) != 0 ?
PTROUT(p->p_sysent->sv_psstrings) : 0;
PROC_UNLOCK(p);
error = SYSCTL_OUT(req, &ps_strings32, sizeof(ps_strings32));
return (error);
}
#endif
ps_strings = p->p_sysent->sv_psstrings;
PROC_UNLOCK(p);
error = SYSCTL_OUT(req, &ps_strings, sizeof(ps_strings));
return (error);
}
/*
* This sysctl allows a process to retrieve umask of another process.
*/
static int
sysctl_kern_proc_umask(SYSCTL_HANDLER_ARGS)
{
int *name = (int *)arg1;
u_int namelen = arg2;
struct proc *p;
int error;
u_short cmask;
pid_t pid;
if (namelen != 1)
return (EINVAL);
pid = (pid_t)name[0];
p = curproc;
if (pid == p->p_pid || pid == 0) {
cmask = p->p_pd->pd_cmask;
goto out;
}
error = pget(pid, PGET_WANTREAD, &p);
if (error != 0)
return (error);
cmask = p->p_pd->pd_cmask;
PRELE(p);
out:
error = SYSCTL_OUT(req, &cmask, sizeof(cmask));
return (error);
}
/*
* This sysctl allows a process to set and retrieve binary osreldate of
* another process.
*/
static int
sysctl_kern_proc_osrel(SYSCTL_HANDLER_ARGS)
{
int *name = (int *)arg1;
u_int namelen = arg2;
struct proc *p;
int flags, error, osrel;
if (namelen != 1)
return (EINVAL);
if (req->newptr != NULL && req->newlen != sizeof(osrel))
return (EINVAL);
flags = PGET_HOLD | PGET_NOTWEXIT;
if (req->newptr != NULL)
flags |= PGET_CANDEBUG;
else
flags |= PGET_CANSEE;
error = pget((pid_t)name[0], flags, &p);
if (error != 0)
return (error);
error = SYSCTL_OUT(req, &p->p_osrel, sizeof(p->p_osrel));
if (error != 0)
goto errout;
if (req->newptr != NULL) {
error = SYSCTL_IN(req, &osrel, sizeof(osrel));
if (error != 0)
goto errout;
if (osrel < 0) {
error = EINVAL;
goto errout;
}
p->p_osrel = osrel;
}
errout:
PRELE(p);
return (error);
}
static int
sysctl_kern_proc_sigtramp(SYSCTL_HANDLER_ARGS)
{
int *name = (int *)arg1;
u_int namelen = arg2;
struct proc *p;
struct kinfo_sigtramp kst;
const struct sysentvec *sv;
int error;
#ifdef COMPAT_FREEBSD32
struct kinfo_sigtramp32 kst32;
#endif
if (namelen != 1)
return (EINVAL);
error = pget((pid_t)name[0], PGET_CANDEBUG, &p);
if (error != 0)
return (error);
sv = p->p_sysent;
#ifdef COMPAT_FREEBSD32
if ((req->flags & SCTL_MASK32) != 0) {
bzero(&kst32, sizeof(kst32));
if (SV_PROC_FLAG(p, SV_ILP32)) {
if (sv->sv_sigcode_base != 0) {
kst32.ksigtramp_start = sv->sv_sigcode_base;
kst32.ksigtramp_end = sv->sv_sigcode_base +
*sv->sv_szsigcode;
} else {
kst32.ksigtramp_start = sv->sv_psstrings -
*sv->sv_szsigcode;
kst32.ksigtramp_end = sv->sv_psstrings;
}
}
PROC_UNLOCK(p);
error = SYSCTL_OUT(req, &kst32, sizeof(kst32));
return (error);
}
#endif
bzero(&kst, sizeof(kst));
if (sv->sv_sigcode_base != 0) {
kst.ksigtramp_start = (char *)sv->sv_sigcode_base;
kst.ksigtramp_end = (char *)sv->sv_sigcode_base +
*sv->sv_szsigcode;
} else {
kst.ksigtramp_start = (char *)sv->sv_psstrings -
*sv->sv_szsigcode;
kst.ksigtramp_end = (char *)sv->sv_psstrings;
}
PROC_UNLOCK(p);
error = SYSCTL_OUT(req, &kst, sizeof(kst));
return (error);
}
static int
sysctl_kern_proc_sigfastblk(SYSCTL_HANDLER_ARGS)
{
int *name = (int *)arg1;
u_int namelen = arg2;
pid_t pid;
struct proc *p;
struct thread *td1;
uintptr_t addr;
#ifdef COMPAT_FREEBSD32
uint32_t addr32;
#endif
int error;
if (namelen != 1 || req->newptr != NULL)
return (EINVAL);
pid = (pid_t)name[0];
error = pget(pid, PGET_HOLD | PGET_NOTWEXIT | PGET_CANDEBUG, &p);
if (error != 0)
return (error);
PROC_LOCK(p);
#ifdef COMPAT_FREEBSD32
if (SV_CURPROC_FLAG(SV_ILP32)) {
if (!SV_PROC_FLAG(p, SV_ILP32)) {
error = EINVAL;
goto errlocked;
}
}
#endif
if (pid <= PID_MAX) {
td1 = FIRST_THREAD_IN_PROC(p);
} else {
FOREACH_THREAD_IN_PROC(p, td1) {
if (td1->td_tid == pid)
break;
}
}
if (td1 == NULL) {
error = ESRCH;
goto errlocked;
}
/*
* The access to the private thread flags. It is fine as far
* as no out-of-thin-air values are read from td_pflags, and
* usermode read of the td_sigblock_ptr is racy inherently,
* since target process might have already changed it
* meantime.
*/
if ((td1->td_pflags & TDP_SIGFASTBLOCK) != 0)
addr = (uintptr_t)td1->td_sigblock_ptr;
else
error = ENOTTY;
errlocked:
_PRELE(p);
PROC_UNLOCK(p);
if (error != 0)
return (error);
#ifdef COMPAT_FREEBSD32
if (SV_CURPROC_FLAG(SV_ILP32)) {
addr32 = addr;
error = SYSCTL_OUT(req, &addr32, sizeof(addr32));
} else
#endif
error = SYSCTL_OUT(req, &addr, sizeof(addr));
return (error);
}
SYSCTL_NODE(_kern, KERN_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE, 0,
"Process table");
SYSCTL_PROC(_kern_proc, KERN_PROC_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT|
CTLFLAG_MPSAFE, 0, 0, sysctl_kern_proc, "S,proc",
"Return entire process table");
static SYSCTL_NODE(_kern_proc, KERN_PROC_GID, gid, CTLFLAG_RD | CTLFLAG_MPSAFE,
sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, KERN_PROC_PGRP, pgrp, CTLFLAG_RD | CTLFLAG_MPSAFE,
sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, KERN_PROC_RGID, rgid, CTLFLAG_RD | CTLFLAG_MPSAFE,
sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, KERN_PROC_SESSION, sid, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, KERN_PROC_TTY, tty, CTLFLAG_RD | CTLFLAG_MPSAFE,
sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, KERN_PROC_UID, uid, CTLFLAG_RD | CTLFLAG_MPSAFE,
sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, KERN_PROC_RUID, ruid, CTLFLAG_RD | CTLFLAG_MPSAFE,
sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, KERN_PROC_PID, pid, CTLFLAG_RD | CTLFLAG_MPSAFE,
sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, KERN_PROC_PROC, proc, CTLFLAG_RD | CTLFLAG_MPSAFE,
sysctl_kern_proc, "Return process table, no threads");
static SYSCTL_NODE(_kern_proc, KERN_PROC_ARGS, args,
CTLFLAG_RW | CTLFLAG_CAPWR | CTLFLAG_ANYBODY | CTLFLAG_MPSAFE,
sysctl_kern_proc_args, "Process argument list");
static SYSCTL_NODE(_kern_proc, KERN_PROC_ENV, env, CTLFLAG_RD | CTLFLAG_MPSAFE,
sysctl_kern_proc_env, "Process environment");
static SYSCTL_NODE(_kern_proc, KERN_PROC_AUXV, auxv, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc_auxv, "Process ELF auxiliary vector");
static SYSCTL_NODE(_kern_proc, KERN_PROC_PATHNAME, pathname, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc_pathname, "Process executable path");
static SYSCTL_NODE(_kern_proc, KERN_PROC_SV_NAME, sv_name, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc_sv_name,
"Process syscall vector name (ABI type)");
static SYSCTL_NODE(_kern_proc, (KERN_PROC_GID | KERN_PROC_INC_THREAD), gid_td,
CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, (KERN_PROC_PGRP | KERN_PROC_INC_THREAD), pgrp_td,
CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, (KERN_PROC_RGID | KERN_PROC_INC_THREAD), rgid_td,
CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, (KERN_PROC_SESSION | KERN_PROC_INC_THREAD),
sid_td, CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, (KERN_PROC_TTY | KERN_PROC_INC_THREAD), tty_td,
CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, (KERN_PROC_UID | KERN_PROC_INC_THREAD), uid_td,
CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, (KERN_PROC_RUID | KERN_PROC_INC_THREAD), ruid_td,
CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, (KERN_PROC_PID | KERN_PROC_INC_THREAD), pid_td,
CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc, "Process table");
static SYSCTL_NODE(_kern_proc, (KERN_PROC_PROC | KERN_PROC_INC_THREAD), proc_td,
CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_kern_proc,
"Return process table, including threads");
#ifdef COMPAT_FREEBSD7
static SYSCTL_NODE(_kern_proc, KERN_PROC_OVMMAP, ovmmap, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc_ovmmap, "Old Process vm map entries");
#endif
static SYSCTL_NODE(_kern_proc, KERN_PROC_VMMAP, vmmap, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc_vmmap, "Process vm map entries");
#if defined(STACK) || defined(DDB)
static SYSCTL_NODE(_kern_proc, KERN_PROC_KSTACK, kstack, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc_kstack, "Process kernel stacks");
#endif
static SYSCTL_NODE(_kern_proc, KERN_PROC_GROUPS, groups, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc_groups, "Process groups");
static SYSCTL_NODE(_kern_proc, KERN_PROC_RLIMIT, rlimit, CTLFLAG_RW |
CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_rlimit,
"Process resource limits");
static SYSCTL_NODE(_kern_proc, KERN_PROC_PS_STRINGS, ps_strings, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc_ps_strings,
"Process ps_strings location");
static SYSCTL_NODE(_kern_proc, KERN_PROC_UMASK, umask, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc_umask, "Process umask");
static SYSCTL_NODE(_kern_proc, KERN_PROC_OSREL, osrel, CTLFLAG_RW |
CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_osrel,
"Process binary osreldate");
static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGTRAMP, sigtramp, CTLFLAG_RD |
CTLFLAG_MPSAFE, sysctl_kern_proc_sigtramp,
"Process signal trampoline location");
static SYSCTL_NODE(_kern_proc, KERN_PROC_SIGFASTBLK, sigfastblk, CTLFLAG_RD |
CTLFLAG_ANYBODY | CTLFLAG_MPSAFE, sysctl_kern_proc_sigfastblk,
"Thread sigfastblock address");
int allproc_gen;
/*
* stop_all_proc() purpose is to stop all process which have usermode,
* except current process for obvious reasons. This makes it somewhat
* unreliable when invoked from multithreaded process. The service
* must not be user-callable anyway.
*/
void
stop_all_proc(void)
{
struct proc *cp, *p;
int r, gen;
bool restart, seen_stopped, seen_exiting, stopped_some;
cp = curproc;
allproc_loop:
sx_xlock(&allproc_lock);
gen = allproc_gen;
seen_exiting = seen_stopped = stopped_some = restart = false;
LIST_REMOVE(cp, p_list);
LIST_INSERT_HEAD(&allproc, cp, p_list);
for (;;) {
p = LIST_NEXT(cp, p_list);
if (p == NULL)
break;
LIST_REMOVE(cp, p_list);
LIST_INSERT_AFTER(p, cp, p_list);
PROC_LOCK(p);
if ((p->p_flag & (P_KPROC | P_SYSTEM | P_TOTAL_STOP)) != 0) {
PROC_UNLOCK(p);
continue;
}
if ((p->p_flag & P_WEXIT) != 0) {
seen_exiting = true;
PROC_UNLOCK(p);
continue;
}
if (P_SHOULDSTOP(p) == P_STOPPED_SINGLE) {
/*
* Stopped processes are tolerated when there
* are no other processes which might continue
* them. P_STOPPED_SINGLE but not
* P_TOTAL_STOP process still has at least one
* thread running.
*/
seen_stopped = true;
PROC_UNLOCK(p);
continue;
}
sx_xunlock(&allproc_lock);
_PHOLD(p);
r = thread_single(p, SINGLE_ALLPROC);
if (r != 0)
restart = true;
else
stopped_some = true;
_PRELE(p);
PROC_UNLOCK(p);
sx_xlock(&allproc_lock);
}
/* Catch forked children we did not see in iteration. */
if (gen != allproc_gen)
restart = true;
sx_xunlock(&allproc_lock);
if (restart || stopped_some || seen_exiting || seen_stopped) {
kern_yield(PRI_USER);
goto allproc_loop;
}
}
void
resume_all_proc(void)
{
struct proc *cp, *p;
cp = curproc;
sx_xlock(&allproc_lock);
again:
LIST_REMOVE(cp, p_list);
LIST_INSERT_HEAD(&allproc, cp, p_list);
for (;;) {
p = LIST_NEXT(cp, p_list);
if (p == NULL)
break;
LIST_REMOVE(cp, p_list);
LIST_INSERT_AFTER(p, cp, p_list);
PROC_LOCK(p);
if ((p->p_flag & P_TOTAL_STOP) != 0) {
sx_xunlock(&allproc_lock);
_PHOLD(p);
thread_single_end(p, SINGLE_ALLPROC);
_PRELE(p);
PROC_UNLOCK(p);
sx_xlock(&allproc_lock);
} else {
PROC_UNLOCK(p);
}
}
/* Did the loop above missed any stopped process ? */
FOREACH_PROC_IN_SYSTEM(p) {
/* No need for proc lock. */
if ((p->p_flag & P_TOTAL_STOP) != 0)
goto again;
}
sx_xunlock(&allproc_lock);
}
/* #define TOTAL_STOP_DEBUG 1 */
#ifdef TOTAL_STOP_DEBUG
volatile static int ap_resume;
#include <sys/mount.h>
static int
sysctl_debug_stop_all_proc(SYSCTL_HANDLER_ARGS)
{
int error, val;
val = 0;
ap_resume = 0;
error = sysctl_handle_int(oidp, &val, 0, req);
if (error != 0 || req->newptr == NULL)
return (error);
if (val != 0) {
stop_all_proc();
syncer_suspend();
while (ap_resume == 0)
;
syncer_resume();
resume_all_proc();
}
return (0);
}
SYSCTL_PROC(_debug, OID_AUTO, stop_all_proc, CTLTYPE_INT | CTLFLAG_RW |
CTLFLAG_MPSAFE, __DEVOLATILE(int *, &ap_resume), 0,
sysctl_debug_stop_all_proc, "I",
"");
#endif
diff --git a/sys/sys/ktrace.h b/sys/sys/ktrace.h
index a0b02f7d3ac5..c4ab985722c0 100644
--- a/sys/sys/ktrace.h
+++ b/sys/sys/ktrace.h
@@ -1,309 +1,313 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1988, 1993
* The Regents of the University of California. All rights reserved.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)ktrace.h 8.1 (Berkeley) 6/2/93
* $FreeBSD$
*/
#ifndef _SYS_KTRACE_H_
#define _SYS_KTRACE_H_
#include <sys/caprights.h>
/*
* operations to ktrace system call (KTROP(op))
*/
#define KTROP_SET 0 /* set trace points */
#define KTROP_CLEAR 1 /* clear trace points */
#define KTROP_CLEARFILE 2 /* stop all tracing to file */
#define KTROP(o) ((o)&3) /* macro to extract operation */
/*
* flags (ORed in with operation)
*/
#define KTRFLAG_DESCEND 4 /* perform op on all children too */
/*
* ktrace record header
*/
struct ktr_header {
int ktr_len; /* length of buf */
short ktr_type; /* trace record type */
pid_t ktr_pid; /* process id */
char ktr_comm[MAXCOMLEN + 1];/* command name */
struct timeval ktr_time; /* timestamp */
intptr_t ktr_tid; /* was ktr_buffer */
};
/*
* Test for kernel trace point (MP SAFE).
*
* KTRCHECK() just checks that the type is enabled and is only for
* internal use in the ktrace subsystem. KTRPOINT() checks against
* ktrace recursion as well as checking that the type is enabled and
* is the public interface.
*/
#define KTRCHECK(td, type) ((td)->td_proc->p_traceflag & (1 << type))
#define KTRPOINT(td, type) (__predict_false(KTRCHECK((td), (type))))
#define KTRCHECKDRAIN(td) (!(STAILQ_EMPTY(&(td)->td_proc->p_ktr)))
#define KTRUSERRET(td) do { \
if (__predict_false(KTRCHECKDRAIN(td))) \
ktruserret(td); \
} while (0)
/*
* ktrace record types
*/
/*
* KTR_SYSCALL - system call record
*/
#define KTR_SYSCALL 1
struct ktr_syscall {
short ktr_code; /* syscall number */
short ktr_narg; /* number of arguments */
/*
* followed by ktr_narg register_t
*/
register_t ktr_args[1];
};
/*
* KTR_SYSRET - return from system call record
*/
#define KTR_SYSRET 2
struct ktr_sysret {
short ktr_code;
short ktr_eosys;
int ktr_error;
register_t ktr_retval;
};
/*
* KTR_NAMEI - namei record
*/
#define KTR_NAMEI 3
/* record contains pathname */
/*
* KTR_GENIO - trace generic process i/o
*/
#define KTR_GENIO 4
struct ktr_genio {
int ktr_fd;
enum uio_rw ktr_rw;
/*
* followed by data successfully read/written
*/
};
/*
* KTR_PSIG - trace processed signal
*/
#define KTR_PSIG 5
struct ktr_psig {
int signo;
sig_t action;
int code;
sigset_t mask;
};
/*
* KTR_CSW - trace context switches
*/
#define KTR_CSW 6
struct ktr_csw_old {
int out; /* 1 if switch out, 0 if switch in */
int user; /* 1 if usermode (ivcsw), 0 if kernel (vcsw) */
};
struct ktr_csw {
int out; /* 1 if switch out, 0 if switch in */
int user; /* 1 if usermode (ivcsw), 0 if kernel (vcsw) */
char wmesg[8];
};
/*
* KTR_USER - data coming from userland
*/
#define KTR_USER_MAXLEN 2048 /* maximum length of passed data */
#define KTR_USER 7
/*
* KTR_STRUCT - misc. structs
*/
#define KTR_STRUCT 8
/*
* record contains null-terminated struct name followed by
* struct contents
*/
struct sockaddr;
struct stat;
struct sysentvec;
/*
* KTR_SYSCTL - name of a sysctl MIB
*/
#define KTR_SYSCTL 9
/* record contains null-terminated MIB name */
/*
* KTR_PROCCTOR - trace process creation (multiple ABI support)
*/
#define KTR_PROCCTOR 10
struct ktr_proc_ctor {
u_int sv_flags; /* struct sysentvec sv_flags copy */
};
/*
* KTR_PROCDTOR - trace process destruction (multiple ABI support)
*/
#define KTR_PROCDTOR 11
/*
* KTR_CAPFAIL - trace capability check failures
*/
#define KTR_CAPFAIL 12
enum ktr_cap_fail_type {
CAPFAIL_NOTCAPABLE, /* insufficient capabilities in cap_check() */
CAPFAIL_INCREASE, /* attempt to increase capabilities */
CAPFAIL_SYSCALL, /* disallowed system call */
CAPFAIL_LOOKUP, /* disallowed VFS lookup */
};
struct ktr_cap_fail {
enum ktr_cap_fail_type cap_type;
cap_rights_t cap_needed;
cap_rights_t cap_held;
};
/*
* KTR_FAULT - page fault record
*/
#define KTR_FAULT 13
struct ktr_fault {
vm_offset_t vaddr;
int type;
};
/*
* KTR_FAULTEND - end of page fault record
*/
#define KTR_FAULTEND 14
struct ktr_faultend {
int result;
};
/*
* KTR_STRUCT_ARRAY - array of misc. structs
*/
#define KTR_STRUCT_ARRAY 15
struct ktr_struct_array {
size_t struct_size;
/*
* Followed by null-terminated structure name and then payload
* contents.
*/
};
/*
* KTR_DROP - If this bit is set in ktr_type, then at least one event
* between the previous record and this record was dropped.
*/
#define KTR_DROP 0x8000
/*
* kernel trace points (in p_traceflag)
*/
#define KTRFAC_MASK 0x00ffffff
#define KTRFAC_SYSCALL (1<<KTR_SYSCALL)
#define KTRFAC_SYSRET (1<<KTR_SYSRET)
#define KTRFAC_NAMEI (1<<KTR_NAMEI)
#define KTRFAC_GENIO (1<<KTR_GENIO)
#define KTRFAC_PSIG (1<<KTR_PSIG)
#define KTRFAC_CSW (1<<KTR_CSW)
#define KTRFAC_USER (1<<KTR_USER)
#define KTRFAC_STRUCT (1<<KTR_STRUCT)
#define KTRFAC_SYSCTL (1<<KTR_SYSCTL)
#define KTRFAC_PROCCTOR (1<<KTR_PROCCTOR)
#define KTRFAC_PROCDTOR (1<<KTR_PROCDTOR)
#define KTRFAC_CAPFAIL (1<<KTR_CAPFAIL)
#define KTRFAC_FAULT (1<<KTR_FAULT)
#define KTRFAC_FAULTEND (1<<KTR_FAULTEND)
#define KTRFAC_STRUCT_ARRAY (1<<KTR_STRUCT_ARRAY)
/*
* trace flags (also in p_traceflags)
*/
#define KTRFAC_ROOT 0x80000000 /* root set this trace */
#define KTRFAC_INHERIT 0x40000000 /* pass trace flags to children */
#define KTRFAC_DROP 0x20000000 /* last event was dropped */
#ifdef _KERNEL
+struct ktr_io_params;
+
+struct vnode *ktr_get_tracevp(struct proc *, bool);
+void ktr_io_params_free(struct ktr_io_params *);
void ktrnamei(char *);
void ktrcsw(int, int, const char *);
void ktrpsig(int, sig_t, sigset_t *, int);
void ktrfault(vm_offset_t, int);
void ktrfaultend(int);
void ktrgenio(int, enum uio_rw, struct uio *, int);
void ktrsyscall(int, int narg, register_t args[]);
void ktrsysctl(int *name, u_int namelen);
void ktrsysret(int, int, register_t);
void ktrprocctor(struct proc *);
-void ktrprocexec(struct proc *, struct ucred **, struct vnode **);
+struct ktr_io_params *ktrprocexec(struct proc *);
void ktrprocexit(struct thread *);
void ktrprocfork(struct proc *, struct proc *);
void ktruserret(struct thread *);
void ktrstruct(const char *, const void *, size_t);
void ktrstruct_error(const char *, const void *, size_t, int);
void ktrstructarray(const char *, enum uio_seg, const void *, int, size_t);
void ktrcapfail(enum ktr_cap_fail_type, const cap_rights_t *,
const cap_rights_t *);
#define ktrcaprights(s) \
ktrstruct("caprights", (s), sizeof(cap_rights_t))
#define ktritimerval(s) \
ktrstruct("itimerval", (s), sizeof(struct itimerval))
#define ktrsockaddr(s) \
ktrstruct("sockaddr", (s), ((struct sockaddr *)(s))->sa_len)
#define ktrstat(s) \
ktrstruct("stat", (s), sizeof(struct stat))
#define ktrstat_error(s, error) \
ktrstruct_error("stat", (s), sizeof(struct stat), error)
extern u_int ktr_geniosize;
#else
#include <sys/cdefs.h>
__BEGIN_DECLS
int ktrace(const char *, int, int, pid_t);
int utrace(const void *, size_t);
__END_DECLS
#endif
#endif
diff --git a/sys/sys/proc.h b/sys/sys/proc.h
index b47af58c34be..fb5714818163 100644
--- a/sys/sys/proc.h
+++ b/sys/sys/proc.h
@@ -1,1285 +1,1285 @@
/*-
* SPDX-License-Identifier: BSD-3-Clause
*
* Copyright (c) 1986, 1989, 1991, 1993
* The Regents of the University of California. All rights reserved.
* (c) UNIX System Laboratories, Inc.
* All or some portions of this file are derived from material licensed
* to the University of California by American Telephone and Telegraph
* Co. or Unix System Laboratories, Inc. and are reproduced herein with
* the permission of UNIX System Laboratories, Inc.
*
* Redistribution and use in source and binary forms, with or without
* modification, are permitted provided that the following conditions
* are met:
* 1. Redistributions of source code must retain the above copyright
* notice, this list of conditions and the following disclaimer.
* 2. Redistributions in binary form must reproduce the above copyright
* notice, this list of conditions and the following disclaimer in the
* documentation and/or other materials provided with the distribution.
* 3. Neither the name of the University nor the names of its contributors
* may be used to endorse or promote products derived from this software
* without specific prior written permission.
*
* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
* SUCH DAMAGE.
*
* @(#)proc.h 8.15 (Berkeley) 5/19/95
* $FreeBSD$
*/
#ifndef _SYS_PROC_H_
#define _SYS_PROC_H_
#include <sys/callout.h> /* For struct callout. */
#include <sys/event.h> /* For struct klist. */
#ifdef _KERNEL
#include <sys/_eventhandler.h>
#endif
#include <sys/condvar.h>
#ifndef _KERNEL
#include <sys/filedesc.h>
#endif
#include <sys/queue.h>
#include <sys/_lock.h>
#include <sys/lock_profile.h>
#include <sys/_mutex.h>
#include <sys/osd.h>
#include <sys/priority.h>
#include <sys/rtprio.h> /* XXX. */
#include <sys/runq.h>
#include <sys/resource.h>
#include <sys/sigio.h>
#include <sys/signal.h>
#include <sys/signalvar.h>
#ifndef _KERNEL
#include <sys/time.h> /* For structs itimerval, timeval. */
#else
#include <sys/pcpu.h>
#include <sys/systm.h>
#endif
#include <sys/ucontext.h>
#include <sys/ucred.h>
#include <sys/types.h>
#include <sys/_domainset.h>
#include <machine/proc.h> /* Machine-dependent proc substruct. */
#ifdef _KERNEL
#include <machine/cpu.h>
#endif
/*
* One structure allocated per session.
*
* List of locks
* (m) locked by s_mtx mtx
* (e) locked by proctree_lock sx
* (c) const until freeing
*/
struct session {
u_int s_count; /* Ref cnt; pgrps in session - atomic. */
struct proc *s_leader; /* (m + e) Session leader. */
struct vnode *s_ttyvp; /* (m) Vnode of controlling tty. */
struct cdev_priv *s_ttydp; /* (m) Device of controlling tty. */
struct tty *s_ttyp; /* (e) Controlling tty. */
pid_t s_sid; /* (c) Session ID. */
/* (m) Setlogin() name: */
char s_login[roundup(MAXLOGNAME, sizeof(long))];
struct mtx s_mtx; /* Mutex to protect members. */
};
/*
* One structure allocated per process group.
*
* List of locks
* (m) locked by pg_mtx mtx
* (e) locked by proctree_lock sx
* (c) const until freeing
*/
struct pgrp {
LIST_ENTRY(pgrp) pg_hash; /* (e) Hash chain. */
LIST_HEAD(, proc) pg_members; /* (m + e) Pointer to pgrp members. */
struct session *pg_session; /* (c) Pointer to session. */
struct sigiolst pg_sigiolst; /* (m) List of sigio sources. */
pid_t pg_id; /* (c) Process group id. */
struct mtx pg_mtx; /* Mutex to protect members */
int pg_flags; /* (m) PGRP_ flags */
};
#define PGRP_ORPHANED 0x00000001 /* Group is orphaned */
/*
* pargs, used to hold a copy of the command line, if it had a sane length.
*/
struct pargs {
u_int ar_ref; /* Reference count. */
u_int ar_length; /* Length. */
u_char ar_args[1]; /* Arguments. */
};
/*-
* Description of a process.
*
* This structure contains the information needed to manage a thread of
* control, known in UN*X as a process; it has references to substructures
* containing descriptions of things that the process uses, but may share
* with related processes. The process structure and the substructures
* are always addressable except for those marked "(CPU)" below,
* which might be addressable only on a processor on which the process
* is running.
*
* Below is a key of locks used to protect each member of struct proc. The
* lock is indicated by a reference to a specific character in parens in the
* associated comment.
* * - not yet protected
* a - only touched by curproc or parent during fork/wait
* b - created at fork, never changes
* (exception aiods switch vmspaces, but they are also
* marked 'P_SYSTEM' so hopefully it will be left alone)
* c - locked by proc mtx
* d - locked by allproc_lock lock
* e - locked by proctree_lock lock
* f - session mtx
* g - process group mtx
* h - callout_lock mtx
* i - by curproc or the master session mtx
* j - locked by proc slock
* k - only accessed by curthread
* k*- only accessed by curthread and from an interrupt
* kx- only accessed by curthread and by debugger
* l - the attaching proc or attaching proc parent
* m - Giant
* n - not locked, lazy
* o - ktrace lock
* q - td_contested lock
* r - p_peers lock
* s - see sleepq_switch(), sleeping_on_old_rtc(), and sleep(9)
* t - thread lock
* u - process stat lock
* w - process timer lock
* x - created at fork, only changes during single threading in exec
* y - created at first aio, doesn't change until exit or exec at which
* point we are single-threaded and only curthread changes it
* z - zombie threads lock
*
* If the locking key specifies two identifiers (for example, p_pptr) then
* either lock is sufficient for read access, but both locks must be held
* for write access.
*/
struct cpuset;
struct filecaps;
struct filemon;
struct kaioinfo;
struct kaudit_record;
struct kcov_info;
struct kdtrace_proc;
struct kdtrace_thread;
struct kq_timer_cb_data;
struct mqueue_notifier;
struct p_sched;
struct proc;
struct procdesc;
struct racct;
struct sbuf;
struct sleepqueue;
struct socket;
struct syscall_args;
struct td_sched;
struct thread;
struct trapframe;
struct turnstile;
struct vm_map;
struct vm_map_entry;
struct epoch_tracker;
/*
* XXX: Does this belong in resource.h or resourcevar.h instead?
* Resource usage extension. The times in rusage structs in the kernel are
* never up to date. The actual times are kept as runtimes and tick counts
* (with control info in the "previous" times), and are converted when
* userland asks for rusage info. Backwards compatibility prevents putting
* this directly in the user-visible rusage struct.
*
* Locking for p_rux: (cu) means (u) for p_rux and (c) for p_crux.
* Locking for td_rux: (t) for all fields.
*/
struct rusage_ext {
uint64_t rux_runtime; /* (cu) Real time. */
uint64_t rux_uticks; /* (cu) Statclock hits in user mode. */
uint64_t rux_sticks; /* (cu) Statclock hits in sys mode. */
uint64_t rux_iticks; /* (cu) Statclock hits in intr mode. */
uint64_t rux_uu; /* (c) Previous user time in usec. */
uint64_t rux_su; /* (c) Previous sys time in usec. */
uint64_t rux_tu; /* (c) Previous total time in usec. */
};
/*
* Kernel runnable context (thread).
* This is what is put to sleep and reactivated.
* Thread context. Processes may have multiple threads.
*/
struct thread {
struct mtx *volatile td_lock; /* replaces sched lock */
struct proc *td_proc; /* (*) Associated process. */
TAILQ_ENTRY(thread) td_plist; /* (*) All threads in this proc. */
TAILQ_ENTRY(thread) td_runq; /* (t) Run queue. */
union {
TAILQ_ENTRY(thread) td_slpq; /* (t) Sleep queue. */
struct thread *td_zombie; /* Zombie list linkage */
};
TAILQ_ENTRY(thread) td_lockq; /* (t) Lock queue. */
LIST_ENTRY(thread) td_hash; /* (d) Hash chain. */
struct cpuset *td_cpuset; /* (t) CPU affinity mask. */
struct domainset_ref td_domain; /* (a) NUMA policy */
struct seltd *td_sel; /* Select queue/channel. */
struct sleepqueue *td_sleepqueue; /* (k) Associated sleep queue. */
struct turnstile *td_turnstile; /* (k) Associated turnstile. */
struct rl_q_entry *td_rlqe; /* (k) Associated range lock entry. */
struct umtx_q *td_umtxq; /* (c?) Link for when we're blocked. */
lwpid_t td_tid; /* (b) Thread ID. */
sigqueue_t td_sigqueue; /* (c) Sigs arrived, not delivered. */
#define td_siglist td_sigqueue.sq_signals
u_char td_lend_user_pri; /* (t) Lend user pri. */
u_char td_allocdomain; /* (b) NUMA domain backing this struct thread. */
/* Cleared during fork1() */
#define td_startzero td_flags
int td_flags; /* (t) TDF_* flags. */
int td_inhibitors; /* (t) Why can not run. */
int td_pflags; /* (k) Private thread (TDP_*) flags. */
int td_pflags2; /* (k) Private thread (TDP2_*) flags. */
int td_dupfd; /* (k) Ret value from fdopen. XXX */
int td_sqqueue; /* (t) Sleepqueue queue blocked on. */
const void *td_wchan; /* (t) Sleep address. */
const char *td_wmesg; /* (t) Reason for sleep. */
volatile u_char td_owepreempt; /* (k*) Preempt on last critical_exit */
u_char td_tsqueue; /* (t) Turnstile queue blocked on. */
short td_locks; /* (k) Debug: count of non-spin locks */
short td_rw_rlocks; /* (k) Count of rwlock read locks. */
short td_sx_slocks; /* (k) Count of sx shared locks. */
short td_lk_slocks; /* (k) Count of lockmgr shared locks. */
short td_stopsched; /* (k) Scheduler stopped. */
struct turnstile *td_blocked; /* (t) Lock thread is blocked on. */
const char *td_lockname; /* (t) Name of lock blocked on. */
LIST_HEAD(, turnstile) td_contested; /* (q) Contested locks. */
struct lock_list_entry *td_sleeplocks; /* (k) Held sleep locks. */
int td_intr_nesting_level; /* (k) Interrupt recursion. */
int td_pinned; /* (k) Temporary cpu pin count. */
struct ucred *td_realucred; /* (k) Reference to credentials. */
struct ucred *td_ucred; /* (k) Used credentials, temporarily switchable. */
struct plimit *td_limit; /* (k) Resource limits. */
int td_slptick; /* (t) Time at sleep. */
int td_blktick; /* (t) Time spent blocked. */
int td_swvoltick; /* (t) Time at last SW_VOL switch. */
int td_swinvoltick; /* (t) Time at last SW_INVOL switch. */
u_int td_cow; /* (*) Number of copy-on-write faults */
struct rusage td_ru; /* (t) rusage information. */
struct rusage_ext td_rux; /* (t) Internal rusage information. */
uint64_t td_incruntime; /* (t) Cpu ticks to transfer to proc. */
uint64_t td_runtime; /* (t) How many cpu ticks we've run. */
u_int td_pticks; /* (t) Statclock hits for profiling */
u_int td_sticks; /* (t) Statclock hits in system mode. */
u_int td_iticks; /* (t) Statclock hits in intr mode. */
u_int td_uticks; /* (t) Statclock hits in user mode. */
int td_intrval; /* (t) Return value for sleepq. */
sigset_t td_oldsigmask; /* (k) Saved mask from pre sigpause. */
volatile u_int td_generation; /* (k) For detection of preemption */
stack_t td_sigstk; /* (k) Stack ptr and on-stack flag. */
int td_xsig; /* (c) Signal for ptrace */
u_long td_profil_addr; /* (k) Temporary addr until AST. */
u_int td_profil_ticks; /* (k) Temporary ticks until AST. */
char td_name[MAXCOMLEN + 1]; /* (*) Thread name. */
struct file *td_fpop; /* (k) file referencing cdev under op */
int td_dbgflags; /* (c) Userland debugger flags */
siginfo_t td_si; /* (c) For debugger or core file */
int td_ng_outbound; /* (k) Thread entered ng from above. */
struct osd td_osd; /* (k) Object specific data. */
struct vm_map_entry *td_map_def_user; /* (k) Deferred entries. */
pid_t td_dbg_forked; /* (c) Child pid for debugger. */
struct vnode *td_vp_reserved;/* (k) Prealloated vnode. */
u_int td_no_sleeping; /* (k) Sleeping disabled count. */
void *td_su; /* (k) FFS SU private */
sbintime_t td_sleeptimo; /* (t) Sleep timeout. */
int td_rtcgen; /* (s) rtc_generation of abs. sleep */
int td_errno; /* (k) Error from last syscall. */
size_t td_vslock_sz; /* (k) amount of vslock-ed space */
struct kcov_info *td_kcov_info; /* (*) Kernel code coverage data */
u_int td_ucredref; /* (k) references on td_realucred */
#define td_endzero td_sigmask
/* Copied during fork1() or create_thread(). */
#define td_startcopy td_endzero
sigset_t td_sigmask; /* (c) Current signal mask. */
u_char td_rqindex; /* (t) Run queue index. */
u_char td_base_pri; /* (t) Thread base kernel priority. */
u_char td_priority; /* (t) Thread active priority. */
u_char td_pri_class; /* (t) Scheduling class. */
u_char td_user_pri; /* (t) User pri from estcpu and nice. */
u_char td_base_user_pri; /* (t) Base user pri */
u_char td_unused_0; /* no longer used field */
uintptr_t td_rb_list; /* (k) Robust list head. */
uintptr_t td_rbp_list; /* (k) Robust priv list head. */
uintptr_t td_rb_inact; /* (k) Current in-action mutex loc. */
struct syscall_args td_sa; /* (kx) Syscall parameters. Copied on
fork for child tracing. */
void *td_sigblock_ptr; /* (k) uptr for fast sigblock. */
uint32_t td_sigblock_val; /* (k) fast sigblock value read at
td_sigblock_ptr on kern entry */
#define td_endcopy td_pcb
/*
* Fields that must be manually set in fork1() or create_thread()
* or already have been set in the allocator, constructor, etc.
*/
struct pcb *td_pcb; /* (k) Kernel VA of pcb and kstack. */
enum td_states {
TDS_INACTIVE = 0x0,
TDS_INHIBITED,
TDS_CAN_RUN,
TDS_RUNQ,
TDS_RUNNING
} td_state; /* (t) thread state */
union {
register_t tdu_retval[2];
off_t tdu_off;
} td_uretoff; /* (k) Syscall aux returns. */
#define td_retval td_uretoff.tdu_retval
u_int td_cowgen; /* (k) Generation of COW pointers. */
/* LP64 hole */
struct callout td_slpcallout; /* (h) Callout for sleep. */
struct trapframe *td_frame; /* (k) */
vm_offset_t td_kstack; /* (a) Kernel VA of kstack. */
int td_kstack_pages; /* (a) Size of the kstack. */
volatile u_int td_critnest; /* (k*) Critical section nest level. */
struct mdthread td_md; /* (k) Any machine-dependent fields. */
struct kaudit_record *td_ar; /* (k) Active audit record, if any. */
struct lpohead td_lprof[2]; /* (a) lock profiling objects. */
struct kdtrace_thread *td_dtrace; /* (*) DTrace-specific data. */
struct vnet *td_vnet; /* (k) Effective vnet. */
const char *td_vnet_lpush; /* (k) Debugging vnet push / pop. */
struct trapframe *td_intr_frame;/* (k) Frame of the current irq */
struct proc *td_rfppwait_p; /* (k) The vforked child */
struct vm_page **td_ma; /* (k) uio pages held */
int td_ma_cnt; /* (k) size of *td_ma */
/* LP64 hole */
void *td_emuldata; /* Emulator state data */
int td_lastcpu; /* (t) Last cpu we were on. */
int td_oncpu; /* (t) Which cpu we are on. */
void *td_lkpi_task; /* LinuxKPI task struct pointer */
int td_pmcpend;
void *td_coredump; /* (c) coredump request. */
#ifdef EPOCH_TRACE
SLIST_HEAD(, epoch_tracker) td_epochs;
#endif
};
struct thread0_storage {
struct thread t0st_thread;
uint64_t t0st_sched[10];
};
struct mtx *thread_lock_block(struct thread *);
void thread_lock_block_wait(struct thread *);
void thread_lock_set(struct thread *, struct mtx *);
void thread_lock_unblock(struct thread *, struct mtx *);
#define THREAD_LOCK_ASSERT(td, type) \
mtx_assert((td)->td_lock, (type))
#define THREAD_LOCK_BLOCKED_ASSERT(td, type) \
do { \
struct mtx *__m = (td)->td_lock; \
if (__m != &blocked_lock) \
mtx_assert(__m, (type)); \
} while (0)
#ifdef INVARIANTS
#define THREAD_LOCKPTR_ASSERT(td, lock) \
do { \
struct mtx *__m; \
__m = (td)->td_lock; \
KASSERT(__m == (lock), \
("Thread %p lock %p does not match %p", td, __m, (lock))); \
} while (0)
#define THREAD_LOCKPTR_BLOCKED_ASSERT(td, lock) \
do { \
struct mtx *__m; \
__m = (td)->td_lock; \
KASSERT(__m == (lock) || __m == &blocked_lock, \
("Thread %p lock %p does not match %p", td, __m, (lock))); \
} while (0)
#define TD_LOCKS_INC(td) ((td)->td_locks++)
#define TD_LOCKS_DEC(td) do { \
KASSERT(SCHEDULER_STOPPED_TD(td) || (td)->td_locks > 0, \
("thread %p owns no locks", (td))); \
(td)->td_locks--; \
} while (0)
#else
#define THREAD_LOCKPTR_ASSERT(td, lock)
#define THREAD_LOCKPTR_BLOCKED_ASSERT(td, lock)
#define TD_LOCKS_INC(td)
#define TD_LOCKS_DEC(td)
#endif
/*
* Flags kept in td_flags:
* To change these you MUST have the scheduler lock.
*/
#define TDF_BORROWING 0x00000001 /* Thread is borrowing pri from another. */
#define TDF_INPANIC 0x00000002 /* Caused a panic, let it drive crashdump. */
#define TDF_INMEM 0x00000004 /* Thread's stack is in memory. */
#define TDF_SINTR 0x00000008 /* Sleep is interruptible. */
#define TDF_TIMEOUT 0x00000010 /* Timing out during sleep. */
#define TDF_IDLETD 0x00000020 /* This is a per-CPU idle thread. */
#define TDF_CANSWAP 0x00000040 /* Thread can be swapped. */
#define TDF_UNUSED80 0x00000080 /* unused. */
#define TDF_KTH_SUSP 0x00000100 /* kthread is suspended */
#define TDF_ALLPROCSUSP 0x00000200 /* suspended by SINGLE_ALLPROC */
#define TDF_BOUNDARY 0x00000400 /* Thread suspended at user boundary */
#define TDF_ASTPENDING 0x00000800 /* Thread has some asynchronous events. */
#define TDF_UNUSED12 0x00001000 /* --available-- */
#define TDF_SBDRY 0x00002000 /* Stop only on usermode boundary. */
#define TDF_UPIBLOCKED 0x00004000 /* Thread blocked on user PI mutex. */
#define TDF_NEEDSUSPCHK 0x00008000 /* Thread may need to suspend. */
#define TDF_NEEDRESCHED 0x00010000 /* Thread needs to yield. */
#define TDF_NEEDSIGCHK 0x00020000 /* Thread may need signal delivery. */
#define TDF_NOLOAD 0x00040000 /* Ignore during load avg calculations. */
#define TDF_SERESTART 0x00080000 /* ERESTART on stop attempts. */
#define TDF_THRWAKEUP 0x00100000 /* Libthr thread must not suspend itself. */
#define TDF_SEINTR 0x00200000 /* EINTR on stop attempts. */
#define TDF_SWAPINREQ 0x00400000 /* Swapin request due to wakeup. */
#define TDF_UNUSED23 0x00800000 /* --available-- */
#define TDF_SCHED0 0x01000000 /* Reserved for scheduler private use */
#define TDF_SCHED1 0x02000000 /* Reserved for scheduler private use */
#define TDF_SCHED2 0x04000000 /* Reserved for scheduler private use */
#define TDF_SCHED3 0x08000000 /* Reserved for scheduler private use */
#define TDF_ALRMPEND 0x10000000 /* Pending SIGVTALRM needs to be posted. */
#define TDF_PROFPEND 0x20000000 /* Pending SIGPROF needs to be posted. */
#define TDF_MACPEND 0x40000000 /* AST-based MAC event pending. */
/* Userland debug flags */
#define TDB_SUSPEND 0x00000001 /* Thread is suspended by debugger */
#define TDB_XSIG 0x00000002 /* Thread is exchanging signal under trace */
#define TDB_USERWR 0x00000004 /* Debugger modified memory or registers */
#define TDB_SCE 0x00000008 /* Thread performs syscall enter */
#define TDB_SCX 0x00000010 /* Thread performs syscall exit */
#define TDB_EXEC 0x00000020 /* TDB_SCX from exec(2) family */
#define TDB_FORK 0x00000040 /* TDB_SCX from fork(2) that created new
process */
#define TDB_STOPATFORK 0x00000080 /* Stop at the return from fork (child
only) */
#define TDB_CHILD 0x00000100 /* New child indicator for ptrace() */
#define TDB_BORN 0x00000200 /* New LWP indicator for ptrace() */
#define TDB_EXIT 0x00000400 /* Exiting LWP indicator for ptrace() */
#define TDB_VFORK 0x00000800 /* vfork indicator for ptrace() */
#define TDB_FSTP 0x00001000 /* The thread is PT_ATTACH leader */
#define TDB_STEP 0x00002000 /* (x86) PSL_T set for PT_STEP */
#define TDB_SSWITCH 0x00004000 /* Suspended in ptracestop */
#define TDB_COREDUMPRQ 0x00008000 /* Coredump request */
/*
* "Private" flags kept in td_pflags:
* These are only written by curthread and thus need no locking.
*/
#define TDP_OLDMASK 0x00000001 /* Need to restore mask after suspend. */
#define TDP_INKTR 0x00000002 /* Thread is currently in KTR code. */
#define TDP_INKTRACE 0x00000004 /* Thread is currently in KTRACE code. */
#define TDP_BUFNEED 0x00000008 /* Do not recurse into the buf flush */
#define TDP_COWINPROGRESS 0x00000010 /* Snapshot copy-on-write in progress. */
#define TDP_ALTSTACK 0x00000020 /* Have alternate signal stack. */
#define TDP_DEADLKTREAT 0x00000040 /* Lock acquisition - deadlock treatment. */
#define TDP_NOFAULTING 0x00000080 /* Do not handle page faults. */
#define TDP_SIGFASTBLOCK 0x00000100 /* Fast sigblock active */
#define TDP_OWEUPC 0x00000200 /* Call addupc() at next AST. */
#define TDP_ITHREAD 0x00000400 /* Thread is an interrupt thread. */
#define TDP_SYNCIO 0x00000800 /* Local override, disable async i/o. */
#define TDP_SCHED1 0x00001000 /* Reserved for scheduler private use */
#define TDP_SCHED2 0x00002000 /* Reserved for scheduler private use */
#define TDP_SCHED3 0x00004000 /* Reserved for scheduler private use */
#define TDP_SCHED4 0x00008000 /* Reserved for scheduler private use */
#define TDP_GEOM 0x00010000 /* Settle GEOM before finishing syscall */
#define TDP_SOFTDEP 0x00020000 /* Stuck processing softdep worklist */
#define TDP_NORUNNINGBUF 0x00040000 /* Ignore runningbufspace check */
#define TDP_WAKEUP 0x00080000 /* Don't sleep in umtx cond_wait */
#define TDP_INBDFLUSH 0x00100000 /* Already in BO_BDFLUSH, do not recurse */
#define TDP_KTHREAD 0x00200000 /* This is an official kernel thread */
#define TDP_CALLCHAIN 0x00400000 /* Capture thread's callchain */
#define TDP_IGNSUSP 0x00800000 /* Permission to ignore the MNTK_SUSPEND* */
#define TDP_AUDITREC 0x01000000 /* Audit record pending on thread */
#define TDP_RFPPWAIT 0x02000000 /* Handle RFPPWAIT on syscall exit */
#define TDP_RESETSPUR 0x04000000 /* Reset spurious page fault history. */
#define TDP_NERRNO 0x08000000 /* Last errno is already in td_errno */
#define TDP_UIOHELD 0x10000000 /* Current uio has pages held in td_ma */
#define TDP_FORKING 0x20000000 /* Thread is being created through fork() */
#define TDP_EXECVMSPC 0x40000000 /* Execve destroyed old vmspace */
#define TDP_SIGFASTPENDING 0x80000000 /* Pending signal due to sigfastblock */
#define TDP2_SBPAGES 0x00000001 /* Owns sbusy on some pages */
#define TDP2_COMPAT32RB 0x00000002 /* compat32 ABI for robust lists */
#define TDP2_ACCT 0x00000004 /* Doing accounting */
/*
* Reasons that the current thread can not be run yet.
* More than one may apply.
*/
#define TDI_SUSPENDED 0x0001 /* On suspension queue. */
#define TDI_SLEEPING 0x0002 /* Actually asleep! (tricky). */
#define TDI_SWAPPED 0x0004 /* Stack not in mem. Bad juju if run. */
#define TDI_LOCK 0x0008 /* Stopped on a lock. */
#define TDI_IWAIT 0x0010 /* Awaiting interrupt. */
#define TD_IS_SLEEPING(td) ((td)->td_inhibitors & TDI_SLEEPING)
#define TD_ON_SLEEPQ(td) ((td)->td_wchan != NULL)
#define TD_IS_SUSPENDED(td) ((td)->td_inhibitors & TDI_SUSPENDED)
#define TD_IS_SWAPPED(td) ((td)->td_inhibitors & TDI_SWAPPED)
#define TD_ON_LOCK(td) ((td)->td_inhibitors & TDI_LOCK)
#define TD_AWAITING_INTR(td) ((td)->td_inhibitors & TDI_IWAIT)
#define TD_IS_RUNNING(td) ((td)->td_state == TDS_RUNNING)
#define TD_ON_RUNQ(td) ((td)->td_state == TDS_RUNQ)
#define TD_CAN_RUN(td) ((td)->td_state == TDS_CAN_RUN)
#define TD_IS_INHIBITED(td) ((td)->td_state == TDS_INHIBITED)
#define TD_ON_UPILOCK(td) ((td)->td_flags & TDF_UPIBLOCKED)
#define TD_IS_IDLETHREAD(td) ((td)->td_flags & TDF_IDLETD)
#define TD_CAN_ABORT(td) (TD_ON_SLEEPQ((td)) && \
((td)->td_flags & TDF_SINTR) != 0)
#define KTDSTATE(td) \
(((td)->td_inhibitors & TDI_SLEEPING) != 0 ? "sleep" : \
((td)->td_inhibitors & TDI_SUSPENDED) != 0 ? "suspended" : \
((td)->td_inhibitors & TDI_SWAPPED) != 0 ? "swapped" : \
((td)->td_inhibitors & TDI_LOCK) != 0 ? "blocked" : \
((td)->td_inhibitors & TDI_IWAIT) != 0 ? "iwait" : "yielding")
#define TD_SET_INHIB(td, inhib) do { \
(td)->td_state = TDS_INHIBITED; \
(td)->td_inhibitors |= (inhib); \
} while (0)
#define TD_CLR_INHIB(td, inhib) do { \
if (((td)->td_inhibitors & (inhib)) && \
(((td)->td_inhibitors &= ~(inhib)) == 0)) \
(td)->td_state = TDS_CAN_RUN; \
} while (0)
#define TD_SET_SLEEPING(td) TD_SET_INHIB((td), TDI_SLEEPING)
#define TD_SET_SWAPPED(td) TD_SET_INHIB((td), TDI_SWAPPED)
#define TD_SET_LOCK(td) TD_SET_INHIB((td), TDI_LOCK)
#define TD_SET_SUSPENDED(td) TD_SET_INHIB((td), TDI_SUSPENDED)
#define TD_SET_IWAIT(td) TD_SET_INHIB((td), TDI_IWAIT)
#define TD_SET_EXITING(td) TD_SET_INHIB((td), TDI_EXITING)
#define TD_CLR_SLEEPING(td) TD_CLR_INHIB((td), TDI_SLEEPING)
#define TD_CLR_SWAPPED(td) TD_CLR_INHIB((td), TDI_SWAPPED)
#define TD_CLR_LOCK(td) TD_CLR_INHIB((td), TDI_LOCK)
#define TD_CLR_SUSPENDED(td) TD_CLR_INHIB((td), TDI_SUSPENDED)
#define TD_CLR_IWAIT(td) TD_CLR_INHIB((td), TDI_IWAIT)
#define TD_SET_RUNNING(td) (td)->td_state = TDS_RUNNING
#define TD_SET_RUNQ(td) (td)->td_state = TDS_RUNQ
#define TD_SET_CAN_RUN(td) (td)->td_state = TDS_CAN_RUN
#define TD_SBDRY_INTR(td) \
(((td)->td_flags & (TDF_SEINTR | TDF_SERESTART)) != 0)
#define TD_SBDRY_ERRNO(td) \
(((td)->td_flags & TDF_SEINTR) != 0 ? EINTR : ERESTART)
/*
* Process structure.
*/
struct proc {
LIST_ENTRY(proc) p_list; /* (d) List of all processes. */
TAILQ_HEAD(, thread) p_threads; /* (c) all threads. */
struct mtx p_slock; /* process spin lock */
struct ucred *p_ucred; /* (c) Process owner's identity. */
struct filedesc *p_fd; /* (b) Open files. */
struct filedesc_to_leader *p_fdtol; /* (b) Tracking node */
struct pwddesc *p_pd; /* (b) Cwd, chroot, jail, umask */
struct pstats *p_stats; /* (b) Accounting/statistics (CPU). */
struct plimit *p_limit; /* (c) Resource limits. */
struct callout p_limco; /* (c) Limit callout handle */
struct sigacts *p_sigacts; /* (x) Signal actions, state (CPU). */
int p_flag; /* (c) P_* flags. */
int p_flag2; /* (c) P2_* flags. */
enum p_states {
PRS_NEW = 0, /* In creation */
PRS_NORMAL, /* threads can be run. */
PRS_ZOMBIE
} p_state; /* (j/c) Process status. */
pid_t p_pid; /* (b) Process identifier. */
LIST_ENTRY(proc) p_hash; /* (d) Hash chain. */
LIST_ENTRY(proc) p_pglist; /* (g + e) List of processes in pgrp. */
struct proc *p_pptr; /* (c + e) Pointer to parent process. */
LIST_ENTRY(proc) p_sibling; /* (e) List of sibling processes. */
LIST_HEAD(, proc) p_children; /* (e) Pointer to list of children. */
struct proc *p_reaper; /* (e) My reaper. */
LIST_HEAD(, proc) p_reaplist; /* (e) List of my descendants
(if I am reaper). */
LIST_ENTRY(proc) p_reapsibling; /* (e) List of siblings - descendants of
the same reaper. */
struct mtx p_mtx; /* (n) Lock for this struct. */
struct mtx p_statmtx; /* Lock for the stats */
struct mtx p_itimmtx; /* Lock for the virt/prof timers */
struct mtx p_profmtx; /* Lock for the profiling */
struct ksiginfo *p_ksi; /* Locked by parent proc lock */
sigqueue_t p_sigqueue; /* (c) Sigs not delivered to a td. */
#define p_siglist p_sigqueue.sq_signals
pid_t p_oppid; /* (c + e) Real parent pid. */
/* The following fields are all zeroed upon creation in fork. */
#define p_startzero p_vmspace
struct vmspace *p_vmspace; /* (b) Address space. */
u_int p_swtick; /* (c) Tick when swapped in or out. */
u_int p_cowgen; /* (c) Generation of COW pointers. */
struct itimerval p_realtimer; /* (c) Alarm timer. */
struct rusage p_ru; /* (a) Exit information. */
struct rusage_ext p_rux; /* (cu) Internal resource usage. */
struct rusage_ext p_crux; /* (c) Internal child resource usage. */
int p_profthreads; /* (c) Num threads in addupc_task. */
volatile int p_exitthreads; /* (j) Number of threads exiting */
int p_traceflag; /* (o) Kernel trace points. */
- struct vnode *p_tracevp; /* (c + o) Trace to vnode. */
- struct ucred *p_tracecred; /* (o) Credentials to trace with. */
+ struct ktr_io_params *p_ktrioparms; /* (c + o) Params for ktrace. */
+ void *p_pad0;
struct vnode *p_textvp; /* (b) Vnode of executable. */
u_int p_lock; /* (c) Proclock (prevent swap) count. */
struct sigiolst p_sigiolst; /* (c) List of sigio sources. */
int p_sigparent; /* (c) Signal to parent on exit. */
int p_sig; /* (n) For core dump/debugger XXX. */
u_int p_ptevents; /* (c + e) ptrace() event mask. */
struct kaioinfo *p_aioinfo; /* (y) ASYNC I/O info. */
struct thread *p_singlethread;/* (c + j) If single threading this is it */
int p_suspcount; /* (j) Num threads in suspended mode. */
struct thread *p_xthread; /* (c) Trap thread */
int p_boundary_count;/* (j) Num threads at user boundary */
int p_pendingcnt; /* how many signals are pending */
struct itimers *p_itimers; /* (c) POSIX interval timers. */
struct procdesc *p_procdesc; /* (e) Process descriptor, if any. */
u_int p_treeflag; /* (e) P_TREE flags */
int p_pendingexits; /* (c) Count of pending thread exits. */
struct filemon *p_filemon; /* (c) filemon-specific data. */
int p_pdeathsig; /* (c) Signal from parent on exit. */
/* End area that is zeroed on creation. */
#define p_endzero p_magic
/* The following fields are all copied upon creation in fork. */
#define p_startcopy p_endzero
u_int p_magic; /* (b) Magic number. */
int p_osrel; /* (x) osreldate for the
binary (from ELF note, if any) */
uint32_t p_fctl0; /* (x) ABI feature control, ELF note */
char p_comm[MAXCOMLEN + 1]; /* (x) Process name. */
struct sysentvec *p_sysent; /* (b) Syscall dispatch info. */
struct pargs *p_args; /* (c) Process arguments. */
rlim_t p_cpulimit; /* (c) Current CPU limit in seconds. */
signed char p_nice; /* (c) Process "nice" value. */
int p_fibnum; /* in this routing domain XXX MRT */
pid_t p_reapsubtree; /* (e) Pid of the direct child of the
reaper which spawned
our subtree. */
uint16_t p_elf_machine; /* (x) ELF machine type */
uint64_t p_elf_flags; /* (x) ELF flags */
/* End area that is copied on creation. */
#define p_endcopy p_xexit
u_int p_xexit; /* (c) Exit code. */
u_int p_xsig; /* (c) Stop/kill sig. */
struct pgrp *p_pgrp; /* (c + e) Pointer to process group. */
struct knlist *p_klist; /* (c) Knotes attached to this proc. */
int p_numthreads; /* (c) Number of threads. */
struct mdproc p_md; /* Any machine-dependent fields. */
struct callout p_itcallout; /* (h + c) Interval timer callout. */
u_short p_acflag; /* (c) Accounting flags. */
struct proc *p_peers; /* (r) */
struct proc *p_leader; /* (b) */
void *p_emuldata; /* (c) Emulator state data. */
struct label *p_label; /* (*) Proc (not subject) MAC label. */
STAILQ_HEAD(, ktr_request) p_ktr; /* (o) KTR event queue. */
LIST_HEAD(, mqueue_notifier) p_mqnotifier; /* (c) mqueue notifiers.*/
struct kdtrace_proc *p_dtrace; /* (*) DTrace-specific data. */
struct cv p_pwait; /* (*) wait cv for exit/exec. */
uint64_t p_prev_runtime; /* (c) Resource usage accounting. */
struct racct *p_racct; /* (b) Resource accounting. */
int p_throttled; /* (c) Flag for racct pcpu throttling */
/*
* An orphan is the child that has been re-parented to the
* debugger as a result of attaching to it. Need to keep
* track of them for parent to be able to collect the exit
* status of what used to be children.
*/
LIST_ENTRY(proc) p_orphan; /* (e) List of orphan processes. */
LIST_HEAD(, proc) p_orphans; /* (e) Pointer to list of orphans. */
TAILQ_HEAD(, kq_timer_cb_data) p_kqtim_stop; /* (c) */
};
#define p_session p_pgrp->pg_session
#define p_pgid p_pgrp->pg_id
#define NOCPU (-1) /* For when we aren't on a CPU. */
#define NOCPU_OLD (255)
#define MAXCPU_OLD (254)
#define PROC_SLOCK(p) mtx_lock_spin(&(p)->p_slock)
#define PROC_SUNLOCK(p) mtx_unlock_spin(&(p)->p_slock)
#define PROC_SLOCK_ASSERT(p, type) mtx_assert(&(p)->p_slock, (type))
#define PROC_STATLOCK(p) mtx_lock_spin(&(p)->p_statmtx)
#define PROC_STATUNLOCK(p) mtx_unlock_spin(&(p)->p_statmtx)
#define PROC_STATLOCK_ASSERT(p, type) mtx_assert(&(p)->p_statmtx, (type))
#define PROC_ITIMLOCK(p) mtx_lock_spin(&(p)->p_itimmtx)
#define PROC_ITIMUNLOCK(p) mtx_unlock_spin(&(p)->p_itimmtx)
#define PROC_ITIMLOCK_ASSERT(p, type) mtx_assert(&(p)->p_itimmtx, (type))
#define PROC_PROFLOCK(p) mtx_lock_spin(&(p)->p_profmtx)
#define PROC_PROFUNLOCK(p) mtx_unlock_spin(&(p)->p_profmtx)
#define PROC_PROFLOCK_ASSERT(p, type) mtx_assert(&(p)->p_profmtx, (type))
/* These flags are kept in p_flag. */
#define P_ADVLOCK 0x00000001 /* Process may hold a POSIX advisory
lock. */
#define P_CONTROLT 0x00000002 /* Has a controlling terminal. */
#define P_KPROC 0x00000004 /* Kernel process. */
#define P_UNUSED3 0x00000008 /* --available-- */
#define P_PPWAIT 0x00000010 /* Parent is waiting for child to
exec/exit. */
#define P_PROFIL 0x00000020 /* Has started profiling. */
#define P_STOPPROF 0x00000040 /* Has thread requesting to stop
profiling. */
#define P_HADTHREADS 0x00000080 /* Has had threads (no cleanup
shortcuts) */
#define P_SUGID 0x00000100 /* Had set id privileges since last
exec. */
#define P_SYSTEM 0x00000200 /* System proc: no sigs, stats or
swapping. */
#define P_SINGLE_EXIT 0x00000400 /* Threads suspending should exit,
not wait. */
#define P_TRACED 0x00000800 /* Debugged process being traced. */
#define P_WAITED 0x00001000 /* Someone is waiting for us. */
#define P_WEXIT 0x00002000 /* Working on exiting. */
#define P_EXEC 0x00004000 /* Process called exec. */
#define P_WKILLED 0x00008000 /* Killed, go to kernel/user boundary
ASAP. */
#define P_CONTINUED 0x00010000 /* Proc has continued from a stopped
state. */
#define P_STOPPED_SIG 0x00020000 /* Stopped due to SIGSTOP/SIGTSTP. */
#define P_STOPPED_TRACE 0x00040000 /* Stopped because of tracing. */
#define P_STOPPED_SINGLE 0x00080000 /* Only 1 thread can continue (not to
user). */
#define P_PROTECTED 0x00100000 /* Do not kill on memory overcommit. */
#define P_SIGEVENT 0x00200000 /* Process pending signals changed. */
#define P_SINGLE_BOUNDARY 0x00400000 /* Threads should suspend at user
boundary. */
#define P_HWPMC 0x00800000 /* Process is using HWPMCs */
#define P_JAILED 0x01000000 /* Process is in jail. */
#define P_TOTAL_STOP 0x02000000 /* Stopped in stop_all_proc. */
#define P_INEXEC 0x04000000 /* Process is in execve(). */
#define P_STATCHILD 0x08000000 /* Child process stopped or exited. */
#define P_INMEM 0x10000000 /* Loaded into memory. */
#define P_SWAPPINGOUT 0x20000000 /* Process is being swapped out. */
#define P_SWAPPINGIN 0x40000000 /* Process is being swapped in. */
#define P_PPTRACE 0x80000000 /* PT_TRACEME by vforked child. */
#define P_STOPPED (P_STOPPED_SIG|P_STOPPED_SINGLE|P_STOPPED_TRACE)
#define P_SHOULDSTOP(p) ((p)->p_flag & P_STOPPED)
#define P_KILLED(p) ((p)->p_flag & P_WKILLED)
/* These flags are kept in p_flag2. */
#define P2_INHERIT_PROTECTED 0x00000001 /* New children get
P_PROTECTED. */
#define P2_NOTRACE 0x00000002 /* No ptrace(2) attach or
coredumps. */
#define P2_NOTRACE_EXEC 0x00000004 /* Keep P2_NOPTRACE on
exec(2). */
#define P2_AST_SU 0x00000008 /* Handles SU ast for
kthreads. */
#define P2_PTRACE_FSTP 0x00000010 /* SIGSTOP from PT_ATTACH not
yet handled. */
#define P2_TRAPCAP 0x00000020 /* SIGTRAP on ENOTCAPABLE */
#define P2_ASLR_ENABLE 0x00000040 /* Force enable ASLR. */
#define P2_ASLR_DISABLE 0x00000080 /* Force disable ASLR. */
#define P2_ASLR_IGNSTART 0x00000100 /* Enable ASLR to consume sbrk
area. */
#define P2_PROTMAX_ENABLE 0x00000200 /* Force enable implied
PROT_MAX. */
#define P2_PROTMAX_DISABLE 0x00000400 /* Force disable implied
PROT_MAX. */
#define P2_STKGAP_DISABLE 0x00000800 /* Disable stack gap for
MAP_STACK */
#define P2_STKGAP_DISABLE_EXEC 0x00001000 /* Stack gap disabled
after exec */
#define P2_ITSTOPPED 0x00002000
#define P2_PTRACEREQ 0x00004000 /* Active ptrace req */
/* Flags protected by proctree_lock, kept in p_treeflags. */
#define P_TREE_ORPHANED 0x00000001 /* Reparented, on orphan list */
#define P_TREE_FIRST_ORPHAN 0x00000002 /* First element of orphan
list */
#define P_TREE_REAPER 0x00000004 /* Reaper of subtree */
#define P_TREE_GRPEXITED 0x00000008 /* exit1() done with job ctl */
/*
* These were process status values (p_stat), now they are only used in
* legacy conversion code.
*/
#define SIDL 1 /* Process being created by fork. */
#define SRUN 2 /* Currently runnable. */
#define SSLEEP 3 /* Sleeping on an address. */
#define SSTOP 4 /* Process debugging or suspension. */
#define SZOMB 5 /* Awaiting collection by parent. */
#define SWAIT 6 /* Waiting for interrupt. */
#define SLOCK 7 /* Blocked on a lock. */
#define P_MAGIC 0xbeefface
#ifdef _KERNEL
/* Types and flags for mi_switch(). */
#define SW_TYPE_MASK 0xff /* First 8 bits are switch type */
#define SWT_NONE 0 /* Unspecified switch. */
#define SWT_PREEMPT 1 /* Switching due to preemption. */
#define SWT_OWEPREEMPT 2 /* Switching due to owepreempt. */
#define SWT_TURNSTILE 3 /* Turnstile contention. */
#define SWT_SLEEPQ 4 /* Sleepq wait. */
#define SWT_SLEEPQTIMO 5 /* Sleepq timeout wait. */
#define SWT_RELINQUISH 6 /* yield call. */
#define SWT_NEEDRESCHED 7 /* NEEDRESCHED was set. */
#define SWT_IDLE 8 /* Switching from the idle thread. */
#define SWT_IWAIT 9 /* Waiting for interrupts. */
#define SWT_SUSPEND 10 /* Thread suspended. */
#define SWT_REMOTEPREEMPT 11 /* Remote processor preempted. */
#define SWT_REMOTEWAKEIDLE 12 /* Remote processor preempted idle. */
#define SWT_COUNT 13 /* Number of switch types. */
/* Flags */
#define SW_VOL 0x0100 /* Voluntary switch. */
#define SW_INVOL 0x0200 /* Involuntary switch. */
#define SW_PREEMPT 0x0400 /* The invol switch is a preemption */
/* How values for thread_single(). */
#define SINGLE_NO_EXIT 0
#define SINGLE_EXIT 1
#define SINGLE_BOUNDARY 2
#define SINGLE_ALLPROC 3
#ifdef MALLOC_DECLARE
MALLOC_DECLARE(M_PARGS);
MALLOC_DECLARE(M_SESSION);
MALLOC_DECLARE(M_SUBPROC);
#endif
#define FOREACH_PROC_IN_SYSTEM(p) \
LIST_FOREACH((p), &allproc, p_list)
#define FOREACH_THREAD_IN_PROC(p, td) \
TAILQ_FOREACH((td), &(p)->p_threads, td_plist)
#define FIRST_THREAD_IN_PROC(p) TAILQ_FIRST(&(p)->p_threads)
/*
* We use process IDs <= pid_max <= PID_MAX; PID_MAX + 1 must also fit
* in a pid_t, as it is used to represent "no process group".
*/
#define PID_MAX 99999
#define NO_PID 100000
#define THREAD0_TID NO_PID
extern pid_t pid_max;
#define SESS_LEADER(p) ((p)->p_session->s_leader == (p))
/* Lock and unlock a process. */
#define PROC_LOCK(p) mtx_lock(&(p)->p_mtx)
#define PROC_TRYLOCK(p) mtx_trylock(&(p)->p_mtx)
#define PROC_UNLOCK(p) mtx_unlock(&(p)->p_mtx)
#define PROC_LOCKED(p) mtx_owned(&(p)->p_mtx)
#define PROC_WAIT_UNLOCKED(p) mtx_wait_unlocked(&(p)->p_mtx)
#define PROC_LOCK_ASSERT(p, type) mtx_assert(&(p)->p_mtx, (type))
/* Lock and unlock a process group. */
#define PGRP_LOCK(pg) mtx_lock(&(pg)->pg_mtx)
#define PGRP_UNLOCK(pg) mtx_unlock(&(pg)->pg_mtx)
#define PGRP_LOCKED(pg) mtx_owned(&(pg)->pg_mtx)
#define PGRP_LOCK_ASSERT(pg, type) mtx_assert(&(pg)->pg_mtx, (type))
#define PGRP_LOCK_PGSIGNAL(pg) do { \
if ((pg) != NULL) \
PGRP_LOCK(pg); \
} while (0)
#define PGRP_UNLOCK_PGSIGNAL(pg) do { \
if ((pg) != NULL) \
PGRP_UNLOCK(pg); \
} while (0)
/* Lock and unlock a session. */
#define SESS_LOCK(s) mtx_lock(&(s)->s_mtx)
#define SESS_UNLOCK(s) mtx_unlock(&(s)->s_mtx)
#define SESS_LOCKED(s) mtx_owned(&(s)->s_mtx)
#define SESS_LOCK_ASSERT(s, type) mtx_assert(&(s)->s_mtx, (type))
/*
* Non-zero p_lock ensures that:
* - exit1() is not performed until p_lock reaches zero;
* - the process' threads stack are not swapped out if they are currently
* not (P_INMEM).
*
* PHOLD() asserts that the process (except the current process) is
* not exiting, increments p_lock and swaps threads stacks into memory,
* if needed.
* _PHOLD() is same as PHOLD(), it takes the process locked.
* _PHOLD_LITE() also takes the process locked, but comparing with
* _PHOLD(), it only guarantees that exit1() is not executed,
* faultin() is not called.
*/
#define PHOLD(p) do { \
PROC_LOCK(p); \
_PHOLD(p); \
PROC_UNLOCK(p); \
} while (0)
#define _PHOLD(p) do { \
PROC_LOCK_ASSERT((p), MA_OWNED); \
KASSERT(!((p)->p_flag & P_WEXIT) || (p) == curproc, \
("PHOLD of exiting process %p", p)); \
(p)->p_lock++; \
if (((p)->p_flag & P_INMEM) == 0) \
faultin((p)); \
} while (0)
#define _PHOLD_LITE(p) do { \
PROC_LOCK_ASSERT((p), MA_OWNED); \
KASSERT(!((p)->p_flag & P_WEXIT) || (p) == curproc, \
("PHOLD of exiting process %p", p)); \
(p)->p_lock++; \
} while (0)
#define PROC_ASSERT_HELD(p) do { \
KASSERT((p)->p_lock > 0, ("process %p not held", p)); \
} while (0)
#define PRELE(p) do { \
PROC_LOCK((p)); \
_PRELE((p)); \
PROC_UNLOCK((p)); \
} while (0)
#define _PRELE(p) do { \
PROC_LOCK_ASSERT((p), MA_OWNED); \
PROC_ASSERT_HELD(p); \
(--(p)->p_lock); \
if (((p)->p_flag & P_WEXIT) && (p)->p_lock == 0) \
wakeup(&(p)->p_lock); \
} while (0)
#define PROC_ASSERT_NOT_HELD(p) do { \
KASSERT((p)->p_lock == 0, ("process %p held", p)); \
} while (0)
#define PROC_UPDATE_COW(p) do { \
PROC_LOCK_ASSERT((p), MA_OWNED); \
(p)->p_cowgen++; \
} while (0)
/* Check whether a thread is safe to be swapped out. */
#define thread_safetoswapout(td) ((td)->td_flags & TDF_CANSWAP)
/* Control whether or not it is safe for curthread to sleep. */
#define THREAD_NO_SLEEPING() do { \
curthread->td_no_sleeping++; \
MPASS(curthread->td_no_sleeping > 0); \
} while (0)
#define THREAD_SLEEPING_OK() do { \
MPASS(curthread->td_no_sleeping > 0); \
curthread->td_no_sleeping--; \
} while (0)
#define THREAD_CAN_SLEEP() ((curthread)->td_no_sleeping == 0)
#define PIDHASH(pid) (&pidhashtbl[(pid) & pidhash])
#define PIDHASHLOCK(pid) (&pidhashtbl_lock[((pid) & pidhashlock)])
extern LIST_HEAD(pidhashhead, proc) *pidhashtbl;
extern struct sx *pidhashtbl_lock;
extern u_long pidhash;
extern u_long pidhashlock;
#define PGRPHASH(pgid) (&pgrphashtbl[(pgid) & pgrphash])
extern LIST_HEAD(pgrphashhead, pgrp) *pgrphashtbl;
extern u_long pgrphash;
extern struct sx allproc_lock;
extern int allproc_gen;
extern struct sx proctree_lock;
extern struct mtx ppeers_lock;
extern struct mtx procid_lock;
extern struct proc proc0; /* Process slot for swapper. */
extern struct thread0_storage thread0_st; /* Primary thread in proc0. */
#define thread0 (thread0_st.t0st_thread)
extern struct vmspace vmspace0; /* VM space for proc0. */
extern int hogticks; /* Limit on kernel cpu hogs. */
extern int lastpid;
extern int nprocs, maxproc; /* Current and max number of procs. */
extern int maxprocperuid; /* Max procs per uid. */
extern u_long ps_arg_cache_limit;
LIST_HEAD(proclist, proc);
TAILQ_HEAD(procqueue, proc);
TAILQ_HEAD(threadqueue, thread);
extern struct proclist allproc; /* List of all processes. */
extern struct proc *initproc, *pageproc; /* Process slots for init, pager. */
extern struct uma_zone *proc_zone;
extern struct uma_zone *pgrp_zone;
struct proc *pfind(pid_t); /* Find process by id. */
struct proc *pfind_any(pid_t); /* Find (zombie) process by id. */
struct proc *pfind_any_locked(pid_t pid); /* Find process by id, locked. */
struct pgrp *pgfind(pid_t); /* Find process group by id. */
void pidhash_slockall(void); /* Shared lock all pid hash lists. */
void pidhash_sunlockall(void); /* Shared unlock all pid hash lists. */
struct fork_req {
int fr_flags;
int fr_pages;
int *fr_pidp;
struct proc **fr_procp;
int *fr_pd_fd;
int fr_pd_flags;
struct filecaps *fr_pd_fcaps;
int fr_flags2;
#define FR2_DROPSIG_CAUGHT 0x00000001 /* Drop caught non-DFL signals */
#define FR2_SHARE_PATHS 0x00000002 /* Invert sense of RFFDG for paths */
#define FR2_KPROC 0x00000004 /* Create a kernel process */
};
/*
* pget() flags.
*/
#define PGET_HOLD 0x00001 /* Hold the process. */
#define PGET_CANSEE 0x00002 /* Check against p_cansee(). */
#define PGET_CANDEBUG 0x00004 /* Check against p_candebug(). */
#define PGET_ISCURRENT 0x00008 /* Check that the found process is current. */
#define PGET_NOTWEXIT 0x00010 /* Check that the process is not in P_WEXIT. */
#define PGET_NOTINEXEC 0x00020 /* Check that the process is not in P_INEXEC. */
#define PGET_NOTID 0x00040 /* Do not assume tid if pid > PID_MAX. */
#define PGET_WANTREAD (PGET_HOLD | PGET_CANDEBUG | PGET_NOTWEXIT)
int pget(pid_t pid, int flags, struct proc **pp);
void ast(struct trapframe *framep);
struct thread *choosethread(void);
int cr_cansee(struct ucred *u1, struct ucred *u2);
int cr_canseesocket(struct ucred *cred, struct socket *so);
int cr_canseeothergids(struct ucred *u1, struct ucred *u2);
int cr_canseeotheruids(struct ucred *u1, struct ucred *u2);
int cr_canseejailproc(struct ucred *u1, struct ucred *u2);
int cr_cansignal(struct ucred *cred, struct proc *proc, int signum);
int enterpgrp(struct proc *p, pid_t pgid, struct pgrp *pgrp,
struct session *sess);
int enterthispgrp(struct proc *p, struct pgrp *pgrp);
void faultin(struct proc *p);
int fork1(struct thread *, struct fork_req *);
void fork_rfppwait(struct thread *);
void fork_exit(void (*)(void *, struct trapframe *), void *,
struct trapframe *);
void fork_return(struct thread *, struct trapframe *);
int inferior(struct proc *p);
void itimer_proc_continue(struct proc *p);
void kqtimer_proc_continue(struct proc *p);
void kern_proc_vmmap_resident(struct vm_map *map, struct vm_map_entry *entry,
int *resident_count, bool *super);
void kern_yield(int);
void kick_proc0(void);
void killjobc(void);
int leavepgrp(struct proc *p);
int maybe_preempt(struct thread *td);
void maybe_yield(void);
void mi_switch(int flags);
int p_candebug(struct thread *td, struct proc *p);
int p_cansee(struct thread *td, struct proc *p);
int p_cansched(struct thread *td, struct proc *p);
int p_cansignal(struct thread *td, struct proc *p, int signum);
int p_canwait(struct thread *td, struct proc *p);
struct pargs *pargs_alloc(int len);
void pargs_drop(struct pargs *pa);
void pargs_hold(struct pargs *pa);
int proc_getargv(struct thread *td, struct proc *p, struct sbuf *sb);
int proc_getauxv(struct thread *td, struct proc *p, struct sbuf *sb);
int proc_getenvv(struct thread *td, struct proc *p, struct sbuf *sb);
void procinit(void);
int proc_iterate(int (*cb)(struct proc *, void *), void *cbarg);
void proc_linkup0(struct proc *p, struct thread *td);
void proc_linkup(struct proc *p, struct thread *td);
struct proc *proc_realparent(struct proc *child);
void proc_reap(struct thread *td, struct proc *p, int *status, int options);
void proc_reparent(struct proc *child, struct proc *newparent, bool set_oppid);
void proc_add_orphan(struct proc *child, struct proc *parent);
void proc_set_traced(struct proc *p, bool stop);
void proc_wkilled(struct proc *p);
struct pstats *pstats_alloc(void);
void pstats_fork(struct pstats *src, struct pstats *dst);
void pstats_free(struct pstats *ps);
void proc_clear_orphan(struct proc *p);
void reaper_abandon_children(struct proc *p, bool exiting);
int securelevel_ge(struct ucred *cr, int level);
int securelevel_gt(struct ucred *cr, int level);
void sess_hold(struct session *);
void sess_release(struct session *);
int setrunnable(struct thread *, int);
void setsugid(struct proc *p);
int should_yield(void);
int sigonstack(size_t sp);
void stopevent(struct proc *, u_int, u_int);
struct thread *tdfind(lwpid_t, pid_t);
void threadinit(void);
void tidhash_add(struct thread *);
void tidhash_remove(struct thread *);
void cpu_idle(int);
int cpu_idle_wakeup(int);
extern void (*cpu_idle_hook)(sbintime_t); /* Hook to machdep CPU idler. */
void cpu_switch(struct thread *, struct thread *, struct mtx *);
void cpu_throw(struct thread *, struct thread *) __dead2;
void unsleep(struct thread *);
void userret(struct thread *, struct trapframe *);
void cpu_exit(struct thread *);
void exit1(struct thread *, int, int) __dead2;
void cpu_copy_thread(struct thread *td, struct thread *td0);
bool cpu_exec_vmspace_reuse(struct proc *p, struct vm_map *map);
int cpu_fetch_syscall_args(struct thread *td);
void cpu_fork(struct thread *, struct proc *, struct thread *, int);
void cpu_fork_kthread_handler(struct thread *, void (*)(void *), void *);
int cpu_procctl(struct thread *td, int idtype, id_t id, int com,
void *data);
void cpu_set_syscall_retval(struct thread *, int);
void cpu_set_upcall(struct thread *, void (*)(void *), void *,
stack_t *);
int cpu_set_user_tls(struct thread *, void *tls_base);
void cpu_thread_alloc(struct thread *);
void cpu_thread_clean(struct thread *);
void cpu_thread_exit(struct thread *);
void cpu_thread_free(struct thread *);
void cpu_thread_swapin(struct thread *);
void cpu_thread_swapout(struct thread *);
struct thread *thread_alloc(int pages);
int thread_alloc_stack(struct thread *, int pages);
int thread_check_susp(struct thread *td, bool sleep);
void thread_cow_get_proc(struct thread *newtd, struct proc *p);
void thread_cow_get(struct thread *newtd, struct thread *td);
void thread_cow_free(struct thread *td);
void thread_cow_update(struct thread *td);
int thread_create(struct thread *td, struct rtprio *rtp,
int (*initialize_thread)(struct thread *, void *), void *thunk);
void thread_exit(void) __dead2;
void thread_free(struct thread *td);
void thread_link(struct thread *td, struct proc *p);
void thread_reap_barrier(void);
int thread_single(struct proc *p, int how);
void thread_single_end(struct proc *p, int how);
void thread_stash(struct thread *td);
void thread_stopped(struct proc *p);
void childproc_stopped(struct proc *child, int reason);
void childproc_continued(struct proc *child);
void childproc_exited(struct proc *child);
void thread_run_flash(struct thread *td);
int thread_suspend_check(int how);
bool thread_suspend_check_needed(void);
void thread_suspend_switch(struct thread *, struct proc *p);
void thread_suspend_one(struct thread *td);
void thread_unlink(struct thread *td);
void thread_unsuspend(struct proc *p);
void thread_wait(struct proc *p);
void stop_all_proc(void);
void resume_all_proc(void);
static __inline int
curthread_pflags_set(int flags)
{
struct thread *td;
int save;
td = curthread;
save = ~flags | (td->td_pflags & flags);
td->td_pflags |= flags;
return (save);
}
static __inline void
curthread_pflags_restore(int save)
{
curthread->td_pflags &= save;
}
static __inline int
curthread_pflags2_set(int flags)
{
struct thread *td;
int save;
td = curthread;
save = ~flags | (td->td_pflags2 & flags);
td->td_pflags2 |= flags;
return (save);
}
static __inline void
curthread_pflags2_restore(int save)
{
curthread->td_pflags2 &= save;
}
static __inline bool
kstack_contains(struct thread *td, vm_offset_t va, size_t len)
{
return (va >= td->td_kstack && va + len >= va &&
va + len <= td->td_kstack + td->td_kstack_pages * PAGE_SIZE);
}
static __inline __pure2 struct td_sched *
td_get_sched(struct thread *td)
{
return ((struct td_sched *)&td[1]);
}
extern void (*softdep_ast_cleanup)(struct thread *);
static __inline void
td_softdep_cleanup(struct thread *td)
{
if (td->td_su != NULL && softdep_ast_cleanup != NULL)
softdep_ast_cleanup(td);
}
#define PROC_ID_PID 0
#define PROC_ID_GROUP 1
#define PROC_ID_SESSION 2
#define PROC_ID_REAP 3
void proc_id_set(int type, pid_t id);
void proc_id_set_cond(int type, pid_t id);
void proc_id_clear(int type, pid_t id);
EVENTHANDLER_LIST_DECLARE(process_ctor);
EVENTHANDLER_LIST_DECLARE(process_dtor);
EVENTHANDLER_LIST_DECLARE(process_init);
EVENTHANDLER_LIST_DECLARE(process_fini);
EVENTHANDLER_LIST_DECLARE(process_exit);
EVENTHANDLER_LIST_DECLARE(process_fork);
EVENTHANDLER_LIST_DECLARE(process_exec);
EVENTHANDLER_LIST_DECLARE(thread_ctor);
EVENTHANDLER_LIST_DECLARE(thread_dtor);
EVENTHANDLER_LIST_DECLARE(thread_init);
#endif /* _KERNEL */
#endif /* !_SYS_PROC_H_ */