NetBSD/usr.sbin/pstat/pstat.c

959 lines
23 KiB
C

/* $NetBSD: pstat.c,v 1.100 2007/07/17 21:50:07 christos Exp $ */
/*-
* Copyright (c) 1980, 1991, 1993, 1994
* 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.
*/
#include <sys/cdefs.h>
#ifndef lint
__COPYRIGHT("@(#) Copyright (c) 1980, 1991, 1993, 1994\n\
The Regents of the University of California. All rights reserved.\n");
#endif /* not lint */
#ifndef lint
#if 0
static char sccsid[] = "@(#)pstat.c 8.16 (Berkeley) 5/9/95";
#else
__RCSID("$NetBSD: pstat.c,v 1.100 2007/07/17 21:50:07 christos Exp $");
#endif
#endif /* not lint */
#include <sys/param.h>
#include <sys/time.h>
#include <sys/vnode.h>
#include <sys/ucred.h>
#include <stdbool.h>
#define _KERNEL
#include <sys/file.h>
#include <ufs/ufs/inode.h>
#define NFS
#include <sys/mount.h>
#undef NFS
#include <sys/uio.h>
#include <sys/namei.h>
#include <miscfs/genfs/layer.h>
#include <miscfs/union/union.h>
#undef _KERNEL
#include <sys/stat.h>
#include <nfs/nfsproto.h>
#include <nfs/rpcv2.h>
#include <nfs/nfs.h>
#include <nfs/nfsnode.h>
#include <sys/ioctl.h>
#include <sys/tty.h>
#include <sys/conf.h>
#include <sys/device.h>
#include <sys/sysctl.h>
#include <err.h>
#include <kvm.h>
#include <limits.h>
#include <nlist.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include "swapctl.h"
struct nlist nl[] = {
#define V_MOUNTLIST 0
{ "_mountlist" }, /* address of head of mount list. */
#define V_NUMV 1
{ "_numvnodes" },
#define FNL_NFILE 2
{ "_nfiles" },
#define FNL_MAXFILE 3
{ "_maxfiles" },
#define TTY_NTTY 4
{ "_tty_count" },
#define TTY_TTYLIST 5
{ "_ttylist" },
#define NLMANDATORY TTY_TTYLIST /* names up to here are mandatory */
{ "" }
};
int usenumflag;
int totalflag;
int kflag;
int hflag;
char *nlistf = NULL;
char *memf = NULL;
kvm_t *kd;
static const char * const dtypes[] = { DTYPE_NAMES };
static const struct {
u_int m_flag;
u_int m_visible;
const char *m_name;
} mnt_flags[] = {
__MNT_FLAGS
};
struct flagbit_desc {
u_int fd_flags;
char fd_mark;
};
#define SVAR(var) __STRING(var) /* to force expansion */
#define KGET(idx, var) \
KGET1(idx, &var, sizeof(var), SVAR(var))
#define KGET1(idx, p, s, msg) \
KGET2(nl[idx].n_value, p, s, msg)
#define KGET2(addr, p, s, msg) do { \
if (kvm_read(kd, (u_long)(addr), p, s) != s) \
warnx("cannot read %s: %s", msg, kvm_geterr(kd)); \
} while (/* CONSTCOND */0)
#define KGETRET(addr, p, s, msg) do { \
if (kvm_read(kd, (u_long)(addr), p, s) != s) { \
warnx("cannot read %s: %s", msg, kvm_geterr(kd)); \
return (0); \
} \
} while (/* CONSTCOND */0)
#if 1 /* This is copied from vmstat/vmstat.c */
/*
* Print single word. `ovflow' is number of characters didn't fit
* on the last word. `fmt' is a format string to print this word.
* It must contain asterisk for field width. `width' is a width
* occupied by this word. `fixed' is a number of constant chars in
* `fmt'. `val' is a value to be printed using format string `fmt'.
*/
#define PRWORD(ovflw, fmt, width, fixed, val) do { \
(ovflw) += printf((fmt), \
(width) - (fixed) - (ovflw) > 0 ? \
(width) - (fixed) - (ovflw) : 0, \
(val)) - (width); \
if ((ovflw) < 0) \
(ovflw) = 0; \
} while (/* CONSTCOND */0)
#endif
void filemode(void);
int getfiles(char **, int *, char **);
int getflags(const struct flagbit_desc *, char *, u_int);
struct mount *
getmnt(struct mount *);
char * kinfo_vnodes(int *);
void layer_header(void);
int layer_print(struct vnode *, int);
char * loadvnodes(int *);
int main(int, char **);
void mount_print(struct mount *);
void nfs_header(void);
int nfs_print(struct vnode *, int);
void ttymode(void);
void ttyprt(struct tty *);
void ufs_header(void);
int ufs_print(struct vnode *, int);
int ext2fs_print(struct vnode *, int);
void union_header(void);
int union_print(struct vnode *, int);
void usage(void);
void vnode_header(void);
int vnode_print(struct vnode *, struct vnode *);
void vnodemode(void);
int
main(int argc, char *argv[])
{
int ch, i, quit, ret, use_sysctl;
int fileflag, swapflag, ttyflag, vnodeflag;
gid_t egid = getegid();
char buf[_POSIX2_LINE_MAX];
setegid(getgid());
fileflag = swapflag = ttyflag = vnodeflag = 0;
while ((ch = getopt(argc, argv, "TM:N:fghikmnstv")) != -1)
switch (ch) {
case 'f':
fileflag = 1;
break;
case 'M':
memf = optarg;
break;
case 'N':
nlistf = optarg;
break;
case 'n':
usenumflag = 1;
break;
case 's':
swapflag = 1;
break;
case 'T':
totalflag = 1;
break;
case 't':
ttyflag = 1;
break;
case 'k':
kflag = 1;
break;
case 'g':
kflag = 3; /* 1k ^ 3 */
break;
case 'h':
hflag = 1;
break;
case 'm':
kflag = 2; /* 1k ^ 2 */
break;
case 'v':
case 'i': /* Backward compatibility. */
vnodeflag = 1;
break;
default:
usage();
}
argc -= optind;
argv += optind;
/*
* Discard setgid privileges. If not the running kernel, we toss
* them away totally so that bad guys can't print interesting stuff
* from kernel memory, otherwise switch back to kmem for the
* duration of the kvm_openfiles() call.
*/
if (nlistf != NULL || memf != NULL)
(void)setgid(getgid());
else
(void)setegid(egid);
use_sysctl = (nlistf == NULL && memf == NULL);
if ((kd = kvm_openfiles(nlistf, memf, NULL, O_RDONLY, buf)) == 0)
errx(1, "kvm_openfiles: %s", buf);
/* get rid of it now anyway */
if (nlistf == NULL && memf == NULL)
(void)setgid(getgid());
if ((ret = kvm_nlist(kd, nl)) != 0) {
if (ret == -1)
errx(1, "kvm_nlist: %s", kvm_geterr(kd));
for (i = quit = 0; i <= NLMANDATORY; i++)
if (!nl[i].n_value) {
quit = 1;
warnx("undefined symbol: %s", nl[i].n_name);
}
if (quit)
exit(1);
}
if (!(fileflag | vnodeflag | ttyflag | swapflag | totalflag))
usage();
if (fileflag || totalflag)
filemode();
if (vnodeflag || totalflag)
vnodemode();
if (ttyflag)
ttymode();
if (swapflag || totalflag)
if (use_sysctl)
list_swap(0, kflag, 0, totalflag, 1, hflag);
exit(0);
}
#define VPTRSZ sizeof(struct vnode *)
#define VNODESZ sizeof(struct vnode)
#define PTRSTRWIDTH ((int)sizeof(void *) * 2) /* Width of resulting string
when pointer is printed
in hexadecimal. */
void
vnodemode(void)
{
char *e_vnodebase, *endvnode, *evp;
struct vnode *vp;
struct mount *maddr, *mp;
int numvnodes, ovflw;
int (*vnode_fsprint) (struct vnode *, int); /* per-fs data printer */
mp = NULL;
e_vnodebase = loadvnodes(&numvnodes);
if (totalflag) {
(void)printf("%7d vnodes\n", numvnodes);
goto out;
}
endvnode = e_vnodebase + numvnodes * (VPTRSZ + VNODESZ);
(void)printf("%d active vnodes\n", numvnodes);
#define ST mp->mnt_stat
#define FSTYPE_IS(mp, name) \
(strncmp((mp)->mnt_stat.f_fstypename, (name), \
sizeof((mp)->mnt_stat.f_fstypename)) == 0)
maddr = NULL;
vnode_fsprint = NULL;
for (evp = e_vnodebase; evp < endvnode; evp += VPTRSZ + VNODESZ) {
vp = (struct vnode *)(evp + VPTRSZ);
if (vp->v_mount != maddr) {
/*
* New filesystem
*/
if ((mp = getmnt(vp->v_mount)) == NULL)
continue;
maddr = vp->v_mount;
mount_print(mp);
vnode_header();
if (FSTYPE_IS(mp, MOUNT_FFS) ||
FSTYPE_IS(mp, MOUNT_MFS)) {
ufs_header();
vnode_fsprint = ufs_print;
} else if (FSTYPE_IS(mp, MOUNT_NFS)) {
nfs_header();
vnode_fsprint = nfs_print;
} else if (FSTYPE_IS(mp, MOUNT_EXT2FS)) {
ufs_header();
vnode_fsprint = ext2fs_print;
} else if (FSTYPE_IS(mp, MOUNT_NULL) ||
FSTYPE_IS(mp, MOUNT_OVERLAY) ||
FSTYPE_IS(mp, MOUNT_UMAP)) {
layer_header();
vnode_fsprint = layer_print;
} else if (FSTYPE_IS(mp, MOUNT_UNION)) {
union_header();
vnode_fsprint = union_print;
} else
vnode_fsprint = NULL;
(void)printf("\n");
}
ovflw = vnode_print(*(struct vnode **)evp, vp);
if (VTOI(vp) != NULL && vnode_fsprint != NULL)
(*vnode_fsprint)(vp, ovflw);
(void)printf("\n");
}
out:
free(e_vnodebase);
}
int
getflags(const struct flagbit_desc *fd, char *p, u_int flags)
{
char *q = p;
if (flags == 0) {
*p++ = '-';
*p = '\0';
return (0);
}
for (; fd->fd_flags != 0; fd++)
if ((flags & fd->fd_flags) != 0)
*p++ = fd->fd_mark;
*p = '\0';
return (p - q);
}
const struct flagbit_desc vnode_flags[] = {
{ VROOT, 'R' },
{ VTEXT, 'T' },
{ VSYSTEM, 'S' },
{ VISTTY, 'I' },
{ VEXECMAP, 'E' },
{ VXLOCK, 'L' },
{ VXWANT, 'W' },
{ VBWAIT, 'B' },
{ VALIASED, 'A' },
{ VDIROP, 'D' },
{ VLAYER, 'Y' },
{ VONWORKLST, 'O' },
{ 0, '\0' },
};
void
vnode_header(void)
{
(void)printf("%-*s TYP VFLAG USE HOLD TAG NPAGE",
PTRSTRWIDTH, "ADDR");
}
int
vnode_print(struct vnode *avnode, struct vnode *vp)
{
char *type, flags[sizeof(vnode_flags) / sizeof(vnode_flags[0])];
int ovflw;
/*
* set type
*/
switch (vp->v_type) {
case VNON:
type = "non"; break;
case VREG:
type = "reg"; break;
case VDIR:
type = "dir"; break;
case VBLK:
type = "blk"; break;
case VCHR:
type = "chr"; break;
case VLNK:
type = "lnk"; break;
case VSOCK:
type = "soc"; break;
case VFIFO:
type = "fif"; break;
case VBAD:
type = "bad"; break;
default:
type = "unk"; break;
}
/*
* gather flags
*/
(void)getflags(vnode_flags, flags, vp->v_flag);
ovflw = 0;
PRWORD(ovflw, "%*lx", PTRSTRWIDTH, 0, (long)avnode);
PRWORD(ovflw, " %*s", 4, 1, type);
PRWORD(ovflw, " %*s", 6, 1, flags);
PRWORD(ovflw, " %*ld", 5, 1, (long)vp->v_usecount);
PRWORD(ovflw, " %*ld", 5, 1, (long)vp->v_holdcnt);
PRWORD(ovflw, " %*d", 4, 1, vp->v_tag);
PRWORD(ovflw, " %*d", 6, 1, vp->v_uobj.uo_npages);
return (ovflw);
}
const struct flagbit_desc ufs_flags[] = {
{ IN_ACCESS, 'A' },
{ IN_CHANGE, 'C' },
{ IN_UPDATE, 'U' },
{ IN_MODIFIED, 'M' },
{ IN_ACCESSED, 'a' },
{ IN_RENAME, 'R' },
{ IN_SHLOCK, 'S' },
{ IN_EXLOCK, 'E' },
{ IN_CLEANING, 'c' },
{ IN_ADIROP, 'D' },
{ IN_SPACECOUNTED, 's' },
{ 0, '\0' },
};
void
ufs_header(void)
{
(void)printf(" FILEID IFLAG RDEV|SZ");
}
int
ufs_print(struct vnode *vp, int ovflw)
{
struct inode inode, *ip = &inode;
union dinode {
struct ufs1_dinode dp1;
struct ufs2_dinode dp2;
} dip;
char flags[sizeof(ufs_flags) / sizeof(ufs_flags[0])];
char dev[4 + 1 + 7 + 1]; /* 12bit marjor + 20bit minor */
char *name;
mode_t type;
dev_t rdev;
KGETRET(VTOI(vp), &inode, sizeof(struct inode), "vnode's inode");
KGETRET(ip->i_din.ffs1_din, &dip, sizeof (struct ufs1_dinode),
"inode's dinode");
if (ip->i_size == dip.dp1.di_size)
rdev = dip.dp1.di_rdev;
else {
KGETRET(ip->i_din.ffs1_din, &dip, sizeof (struct ufs2_dinode),
"inode's UFS2 dinode");
rdev = dip.dp2.di_rdev;
}
/*
* XXX need to to locking state.
*/
(void)getflags(ufs_flags, flags, ip->i_flag);
PRWORD(ovflw, " %*llu", 7, 1, (unsigned long long)ip->i_number);
PRWORD(ovflw, " %*s", 6, 1, flags);
type = ip->i_mode & S_IFMT;
if (S_ISCHR(ip->i_mode) || S_ISBLK(ip->i_mode)) {
if (usenumflag ||
(name = devname(rdev, type)) == NULL) {
snprintf(dev, sizeof(dev), "%d,%d",
major(rdev), minor(rdev));
name = dev;
}
PRWORD(ovflw, " %*s", 8, 1, name);
} else
PRWORD(ovflw, " %*lld", 8, 1, (long long)ip->i_size);
return 0;
}
int
ext2fs_print(struct vnode *vp, int ovflw)
{
struct inode inode, *ip = &inode;
struct ext2fs_dinode dip;
char flags[sizeof(ufs_flags) / sizeof(ufs_flags[0])];
char dev[4 + 1 + 7 + 1]; /* 12bit marjor + 20bit minor */
char *name;
mode_t type;
KGETRET(VTOI(vp), &inode, sizeof(struct inode), "vnode's inode");
KGETRET(ip->i_din.e2fs_din, &dip, sizeof (struct ext2fs_dinode),
"inode's dinode");
/*
* XXX need to to locking state.
*/
(void)getflags(ufs_flags, flags, ip->i_flag);
PRWORD(ovflw, " %*llu", 7, 1, (unsigned long long)ip->i_number);
PRWORD(ovflw, " %*s", 6, 1, flags);
type = dip.e2di_mode & S_IFMT;
if (S_ISCHR(dip.e2di_mode) || S_ISBLK(dip.e2di_mode)) {
if (usenumflag ||
(name = devname(dip.e2di_rdev, type)) == NULL) {
snprintf(dev, sizeof(dev), "%d,%d",
major(dip.e2di_rdev), minor(dip.e2di_rdev));
name = dev;
}
PRWORD(ovflw, " %*s", 8, 1, name);
} else
PRWORD(ovflw, " %*u", 8, 1, (u_int)dip.e2di_size);
return (0);
}
const struct flagbit_desc nfs_flags[] = {
{ NFLUSHWANT, 'W' },
{ NFLUSHINPROG, 'P' },
{ NMODIFIED, 'M' },
{ NWRITEERR, 'E' },
{ NACC, 'A' },
{ NUPD, 'U' },
{ NCHG, 'C' },
{ 0, '\0' },
};
void
nfs_header(void)
{
(void)printf(" FILEID NFLAG RDEV|SZ");
}
int
nfs_print(struct vnode *vp, int ovflw)
{
struct nfsnode nfsnode, *np = &nfsnode;
char flags[sizeof(nfs_flags) / sizeof(nfs_flags[0])];
char dev[4 + 1 + 7 + 1]; /* 12bit marjor + 20bit minor */
struct vattr va;
char *name;
mode_t type;
KGETRET(VTONFS(vp), &nfsnode, sizeof(nfsnode), "vnode's nfsnode");
(void)getflags(nfs_flags, flags, np->n_flag);
KGETRET(np->n_vattr, &va, sizeof(va), "vnode attr");
PRWORD(ovflw, " %*ld", 7, 1, (long)va.va_fileid);
PRWORD(ovflw, " %*s", 6, 1, flags);
switch (va.va_type) {
case VCHR:
type = S_IFCHR;
goto device;
case VBLK:
type = S_IFBLK;
device:
if (usenumflag || (name = devname(va.va_rdev, type)) == NULL) {
(void)snprintf(dev, sizeof(dev), "%d,%d",
major(va.va_rdev), minor(va.va_rdev));
name = dev;
}
PRWORD(ovflw, " %*s", 8, 1, name);
break;
default:
PRWORD(ovflw, " %*lld", 8, 1, (long long)np->n_size);
break;
}
return (0);
}
void
layer_header(void)
{
(void)printf(" %*s", PTRSTRWIDTH, "LOWER");
}
int
layer_print(struct vnode *vp, int ovflw)
{
struct layer_node lnode, *lp = &lnode;
KGETRET(VTOLAYER(vp), &lnode, sizeof(lnode), "layer vnode");
PRWORD(ovflw, " %*lx", PTRSTRWIDTH + 1, 1, (long)lp->layer_lowervp);
return (0);
}
void
union_header(void)
{
(void)printf(" %*s %*s", PTRSTRWIDTH, "UPPER", PTRSTRWIDTH, "LOWER");
}
int
union_print(struct vnode *vp, int ovflw)
{
struct union_node unode, *up = &unode;
KGETRET(VTOUNION(vp), &unode, sizeof(unode), "vnode's unode");
PRWORD(ovflw, " %*lx", PTRSTRWIDTH + 1, 1, (long)up->un_uppervp);
PRWORD(ovflw, " %*lx", PTRSTRWIDTH + 1, 1, (long)up->un_lowervp);
return (0);
}
/*
* Given a pointer to a mount structure in kernel space,
* read it in and return a usable pointer to it.
*/
struct mount *
getmnt(struct mount *maddr)
{
static struct mtab {
struct mtab *next;
struct mount *maddr;
struct mount mount;
} *mhead = NULL;
struct mtab *mt;
struct mount mb;
for (mt = mhead; mt != NULL; mt = mt->next)
if (maddr == mt->maddr)
return (&mt->mount);
KGETRET(maddr, &mb, sizeof(struct mount), "mount table");
if ((mt = malloc(sizeof(struct mtab))) == NULL)
err(1, "malloc");
mt->mount = mb;
mt->maddr = maddr;
mt->next = mhead;
mhead = mt;
return (&mt->mount);
}
void
mount_print(struct mount *mp)
{
int flags;
(void)printf("*** MOUNT %s %s on %s", ST.f_fstypename,
ST.f_mntfromname, ST.f_mntonname);
if ((flags = mp->mnt_flag) != 0) {
int i;
const char *sep = " (";
for (i = 0; i < sizeof mnt_flags / sizeof mnt_flags[0]; i++) {
if (flags & mnt_flags[i].m_flag) {
(void)printf("%s%s", sep, mnt_flags[i].m_name);
flags &= ~mnt_flags[i].m_flag;
sep = ",";
}
}
if (flags)
(void)printf("%sunknown_flags:%x", sep, flags);
(void)printf(")");
}
(void)printf("\n");
}
char *
loadvnodes(int *avnodes)
{
int mib[2];
size_t copysize;
char *vnodebase;
if (memf != NULL) {
/*
* do it by hand
*/
return (kinfo_vnodes(avnodes));
}
mib[0] = CTL_KERN;
mib[1] = KERN_VNODE;
if (sysctl(mib, 2, NULL, &copysize, NULL, 0) == -1)
err(1, "sysctl: KERN_VNODE");
if ((vnodebase = malloc(copysize)) == NULL)
err(1, "malloc");
if (sysctl(mib, 2, vnodebase, &copysize, NULL, 0) == -1)
err(1, "sysctl: KERN_VNODE");
if (copysize % (VPTRSZ + VNODESZ))
errx(1, "vnode size mismatch");
*avnodes = copysize / (VPTRSZ + VNODESZ);
return (vnodebase);
}
/*
* simulate what a running kernel does in in kinfo_vnode
*/
char *
kinfo_vnodes(int *avnodes)
{
struct mntlist mountlist;
struct mount *mp, mount;
struct vnode *vp, vnode;
char *beg, *bp, *ep;
int numvnodes;
KGET(V_NUMV, numvnodes);
if ((bp = malloc((numvnodes + 20) * (VPTRSZ + VNODESZ))) == NULL)
err(1, "malloc");
beg = bp;
ep = bp + (numvnodes + 20) * (VPTRSZ + VNODESZ);
KGET(V_MOUNTLIST, mountlist);
for (mp = mountlist.cqh_first;;
mp = mount.mnt_list.cqe_next) {
KGET2(mp, &mount, sizeof(mount), "mount entry");
TAILQ_FOREACH(vp, &mount.mnt_vnodelist, v_mntvnodes) {
KGET2(vp, &vnode, sizeof(vnode), "vnode");
if (bp + VPTRSZ + VNODESZ > ep)
/* XXX - should realloc */
errx(1, "no more room for vnodes");
memmove(bp, &vp, VPTRSZ);
bp += VPTRSZ;
memmove(bp, &vnode, VNODESZ);
bp += VNODESZ;
}
if (mp == mountlist.cqh_last)
break;
}
*avnodes = (bp - beg) / (VPTRSZ + VNODESZ);
return (beg);
}
void
ttymode(void)
{
int ntty;
struct ttylist_head tty_head;
struct tty *tp, tty;
KGET(TTY_NTTY, ntty);
(void)printf("%d terminal device%s\n", ntty, ntty == 1 ? "" : "s");
KGET(TTY_TTYLIST, tty_head);
(void)printf(
" LINE RAW CAN OUT HWT LWT COL STATE %-*s PGID DISC\n",
PTRSTRWIDTH, "SESS");
for (tp = tty_head.tqh_first; tp; tp = tty.tty_link.tqe_next) {
KGET2(tp, &tty, sizeof tty, "tty struct");
ttyprt(&tty);
}
}
static const struct flagbit_desc ttystates[] = {
{ TS_ISOPEN, 'O'},
{ TS_DIALOUT, '>'},
{ TS_CARR_ON, 'C'},
{ TS_TIMEOUT, 'T'},
{ TS_FLUSH, 'F'},
{ TS_BUSY, 'B'},
{ TS_ASLEEP, 'A'},
{ TS_XCLUDE, 'X'},
{ TS_TTSTOP, 'S'},
{ TS_TBLOCK, 'K'},
{ TS_ASYNC, 'Y'},
{ TS_BKSL, 'D'},
{ TS_ERASE, 'E'},
{ TS_LNCH, 'L'},
{ TS_TYPEN, 'P'},
{ TS_CNTTB, 'N'},
{ 0, '\0'},
};
void
ttyprt(struct tty *tp)
{
char state[sizeof(ttystates) / sizeof(ttystates[0]) + 1];
char dev[2 + 3 + 1 + 5 + 1]; /* 12bit major + 20bit minor */
struct linesw t_linesw;
const char *name;
char buffer;
pid_t pgid;
int n, ovflw;
if (usenumflag || (name = devname(tp->t_dev, S_IFCHR)) == NULL) {
(void)snprintf(dev, sizeof(dev), "0x%3x:%x",
major(tp->t_dev), minor(tp->t_dev));
name = dev;
}
ovflw = 0;
PRWORD(ovflw, "%-*s", 7, 0, name);
PRWORD(ovflw, " %*d", 3, 1, tp->t_rawq.c_cc);
PRWORD(ovflw, " %*d", 4, 1, tp->t_canq.c_cc);
PRWORD(ovflw, " %*d", 4, 1, tp->t_outq.c_cc);
PRWORD(ovflw, " %*d", 5, 1, tp->t_hiwat);
PRWORD(ovflw, " %*d", 4, 1, tp->t_lowat);
PRWORD(ovflw, " %*d", 8, 1, tp->t_column);
n = getflags(ttystates, state, tp->t_state);
if (tp->t_wopen) {
state[n++] = 'W';
state[n] = '\0';
}
PRWORD(ovflw, " %-*s", 7, 1, state);
PRWORD(ovflw, " %*lX", PTRSTRWIDTH + 1, 1, (u_long)tp->t_session);
pgid = 0;
if (tp->t_pgrp != NULL)
KGET2(&tp->t_pgrp->pg_id, &pgid, sizeof(pid_t), "pgid");
PRWORD(ovflw, " %*d", 6, 1, pgid);
KGET2(tp->t_linesw, &t_linesw, sizeof(t_linesw),
"line discipline switch table");
name = t_linesw.l_name;
(void)putchar(' ');
for (;;) {
KGET2(name, &buffer, sizeof(buffer), "line discipline name");
if (buffer == '\0')
break;
(void)putchar(buffer);
name++;
}
(void)putchar('\n');
}
static const struct flagbit_desc filemode_flags[] = {
{ FREAD, 'R' },
{ FWRITE, 'W' },
{ FAPPEND, 'A' },
#ifdef FSHLOCK /* currently gone */
{ FSHLOCK, 'S' },
{ FEXLOCK, 'X' },
#endif
{ FASYNC, 'I' },
{ 0, '\0' },
};
void
filemode(void)
{
struct file *fp;
struct file *addr;
char flags[sizeof(filemode_flags) / sizeof(filemode_flags[0])];
char *buf, *offset;
int len, maxfile, nfile, ovflw;
KGET(FNL_MAXFILE, maxfile);
if (totalflag) {
KGET(FNL_NFILE, nfile);
(void)printf("%3d/%3d files\n", nfile, maxfile);
return;
}
if (getfiles(&buf, &len, &offset) == -1)
return;
/*
* Getfiles returns in malloc'd memory a pointer to the first file
* structure, and then an array of file structs (whose addresses are
* derivable from the previous entry).
*/
addr = ((struct filelist *)offset)->lh_first;
fp = (struct file *)(offset + sizeof(struct filelist));
nfile = (len - sizeof(struct filelist)) / sizeof(struct file);
(void)printf("%d/%d open files\n", nfile, maxfile);
(void)printf("%*s%s%*s TYPE FLG CNT MSG %*s%s%*s USE IFLG OFFSET\n",
(PTRSTRWIDTH - 4) / 2, "", " LOC", (PTRSTRWIDTH - 4) / 2, "",
(PTRSTRWIDTH - 4) / 2, "", "DATA", (PTRSTRWIDTH - 4) / 2, "");
for (; (char *)fp < offset + len; addr = fp->f_list.le_next, fp++) {
if ((unsigned)fp->f_type >= sizeof(dtypes) / sizeof(dtypes[0]))
continue;
ovflw = 0;
(void)getflags(filemode_flags, flags, fp->f_flag);
PRWORD(ovflw, "%*lx", PTRSTRWIDTH, 0, (long)addr);
PRWORD(ovflw, " %-*s", 9, 1, dtypes[fp->f_type]);
PRWORD(ovflw, " %*s", 6, 1, flags);
PRWORD(ovflw, " %*d", 5, 1, fp->f_count);
PRWORD(ovflw, " %*d", 5, 1, fp->f_msgcount);
PRWORD(ovflw, " %*lx", PTRSTRWIDTH + 1, 2, (long)fp->f_data);
PRWORD(ovflw, " %*d", 5, 1, fp->f_usecount);
PRWORD(ovflw, " %*x", 5, 1, fp->f_iflags);
if (fp->f_offset < 0)
PRWORD(ovflw, " %-*lld\n", PTRSTRWIDTH + 1, 2,
(long long)fp->f_offset);
else
PRWORD(ovflw, " %-*lld\n", PTRSTRWIDTH + 1, 2,
(long long)fp->f_offset);
}
free(buf);
}
int
getfiles(char **abuf, int *alen, char **aoffset)
{
size_t len;
int mib[2];
char *buf;
size_t offset;
/*
* XXX
* Add emulation of KINFO_FILE here.
*/
if (memf != NULL)
errx(1, "files on dead kernel, not implemented");
mib[0] = CTL_KERN;
mib[1] = KERN_FILE;
if (sysctl(mib, 2, NULL, &len, NULL, 0) == -1) {
warn("sysctl: KERN_FILE");
return (-1);
}
/* We need to align (struct file *) in the buffer. */
offset = len % sizeof(off_t);
if ((buf = malloc(len + offset)) == NULL)
err(1, "malloc");
if (sysctl(mib, 2, buf + offset, &len, NULL, 0) == -1) {
warn("sysctl: KERN_FILE");
return (-1);
}
*abuf = buf;
*alen = len;
*aoffset = (buf + offset);
return (0);
}
void
usage(void)
{
(void)fprintf(stderr,
"usage: %s [-T|-f|-s|-t|-v] [-ghkmn] [-M core] [-N system]\n",
getprogname());
exit(1);
}