f4574783dc
KERNEL CHANGES: - The kernel only knows about privileges of kernel tasks and the root system process (now RS). - Kernel tasks and the root system process are the only processes that are made schedulable by the kernel at startup. All the other processes in the boot image don't get their privileges set at startup and are inhibited from running by the RTS_NO_PRIV flag. - Removed the assumption on the ordering of processes in the boot image table. System processes can now appear in any order in the boot image table. - Privilege ids can now be assigned both statically or dynamically. The kernel assigns static privilege ids to kernel tasks and the root system process. Each id is directly derived from the process number. - User processes now all share the static privilege id of the root user process (now INIT). - sys_privctl split: we have more calls now to let RS set privileges for system processes. SYS_PRIV_ALLOW / SYS_PRIV_DISALLOW are only used to flip the RTS_NO_PRIV flag and allow / disallow a process from running. SYS_PRIV_SET_SYS / SYS_PRIV_SET_USER are used to set privileges for a system / user process. - boot image table flags split: PROC_FULLVM is the only flag that has been moved out of the privilege flags and is still maintained in the boot image table. All the other privilege flags are out of the kernel now. RS CHANGES: - RS is the only user-space process who gets to run right after in-kernel startup. - RS uses the boot image table from the kernel and three additional boot image info table (priv table, sys table, dev table) to complete the initialization of the system. - RS checks that the entries in the priv table match the entries in the boot image table to make sure that every process in the boot image gets schedulable. - RS only uses static privilege ids to set privileges for system services in the boot image. - RS includes basic memory management support to allocate the boot image buffer dynamically during initialization. The buffer shall contain the executable image of all the system services we would like to restart after a crash. - First step towards decoupling between resource provisioning and resource requirements in RS: RS must know what resources it needs to restart a process and what resources it has currently available. This is useful to tradeoff reliability and resource consumption. When required resources are missing, the process cannot be restarted. In that case, in the future, a system flag will tell RS what to do. For example, if CORE_PROC is set, RS should trigger a system-wide panic because the system can no longer function correctly without a core system process. PM CHANGES: - The process tree built at initialization time is changed to have INIT as root with pid 0, RS child of INIT and all the system services children of RS. This is required to make RS in control of all the system services. - PM no longer registers labels for system services in the boot image. This is now part of RS's initialization process.
646 lines
19 KiB
C
646 lines
19 KiB
C
/* ps - print status Author: Peter Valkenburg */
|
|
|
|
/* Ps.c, Peter Valkenburg (valke@psy.vu.nl), january 1990.
|
|
*
|
|
* This is a V7 ps(1) look-alike for MINIX >= 1.5.0.
|
|
* It does not support the 'k' option (i.e. cannot read memory from core file).
|
|
* If you want to compile this for non-IBM PC architectures, the header files
|
|
* require that you have your CHIP, MACHINE etc. defined.
|
|
* Full syntax:
|
|
* ps [-][aeflx]
|
|
* Option `a' gives all processes, `l' for detailed info, `x' includes even
|
|
* processes without a terminal.
|
|
* The `f' and `e' options were added by Kees Bot for the convenience of
|
|
* Solaris users accustomed to these options. The `e' option is equivalent to
|
|
* `a' and `f' is equivalent to -l. These do not appear in the usage message.
|
|
*
|
|
* VERY IMPORTANT NOTE:
|
|
* To compile ps, the kernel/, fs/ and pm/ source directories must be in
|
|
* ../../ relative to the directory where ps is compiled (normally the
|
|
* tools source directory).
|
|
*
|
|
* If you want your ps to be useable by anyone, you can arrange the
|
|
* following access permissions (note the protected memory files and set
|
|
* *group* id on ps):
|
|
* -rwxr-sr-x 1 bin kmem 11916 Jul 4 15:31 /bin/ps
|
|
* crw-r----- 1 bin kmem 1, 1 Jan 1 1970 /dev/mem
|
|
* crw-r----- 1 bin kmem 1, 2 Jan 1 1970 /dev/kmem
|
|
*/
|
|
|
|
/* Some technical comments on this implementation:
|
|
*
|
|
* Most fields are similar to V7 ps(1), except for CPU, NICE, PRI which are
|
|
* absent, RECV which replaces WCHAN, and PGRP that is an extra.
|
|
* The info is obtained from the following fields of proc, mproc and fproc:
|
|
* F - kernel status field, p_rts_flags
|
|
* S - kernel status field, p_rts_flags; mm status field, mp_flags (R if p_rts_flags
|
|
* is 0; Z if mp_flags == ZOMBIE; T if mp_flags == STOPPED; else W).
|
|
* UID - mm eff uid field, mp_effuid
|
|
* PID - mm pid field, mp_pid
|
|
* PPID - mm parent process index field, mp_parent (used as index in proc).
|
|
* PGRP - mm process group field, mp_procgrp
|
|
* SZ - kernel text size + physical stack address - physical data address
|
|
* + stack size
|
|
* p_memmap[T].mem_len + p_memmap[S].mem_phys - p_memmap[D].mem_phys
|
|
* + p_memmap[S].mem_len
|
|
* RECV - kernel process index field for message receiving, p_getfrom
|
|
* If sleeping, mm's mp_flags, or fs's fp_task are used for more info.
|
|
* TTY - fs controlling tty device field, fp_tty.
|
|
* TIME - kernel user + system times fields, user_time + sys_time
|
|
* CMD - system process index (converted to mnemonic name by using the p_name
|
|
* field), or user process argument list (obtained by reading the stack
|
|
* frame; the resulting address is used to get the argument vector from
|
|
* user space and converted into a concatenated argument list).
|
|
*/
|
|
|
|
#include <minix/config.h>
|
|
#include <minix/com.h>
|
|
#include <minix/sysinfo.h>
|
|
#include <minix/endpoint.h>
|
|
#include <limits.h>
|
|
#include <timers.h>
|
|
#include <sys/types.h>
|
|
|
|
#include <minix/const.h>
|
|
#include <minix/type.h>
|
|
#include <minix/ipc.h>
|
|
#include <string.h>
|
|
#include <stdlib.h>
|
|
#include <unistd.h>
|
|
|
|
#include <minix/com.h>
|
|
#include <fcntl.h>
|
|
#include <a.out.h>
|
|
#include <dirent.h>
|
|
#include <sys/stat.h>
|
|
#include <sys/ioctl.h>
|
|
#include <signal.h>
|
|
#include <stdio.h>
|
|
#include <ttyent.h>
|
|
|
|
#include "../../kernel/arch/i386/include/archtypes.h"
|
|
#include "../../kernel/const.h"
|
|
#include "../../kernel/type.h"
|
|
#include "../../kernel/proc.h"
|
|
|
|
#include "../../servers/pm/mproc.h"
|
|
#include "../../servers/pm/const.h"
|
|
#include "../../servers/vfs/fproc.h"
|
|
#include "../../servers/vfs/const.h"
|
|
#include "../../servers/mfs/const.h"
|
|
|
|
|
|
/*----- ps's local stuff below this line ------*/
|
|
|
|
|
|
#define mindev(dev) (((dev)>>MINOR) & 0377) /* yield minor device */
|
|
#define majdev(dev) (((dev)>>MAJOR) & 0377) /* yield major device */
|
|
|
|
#define TTY_MAJ 4 /* major device of console */
|
|
|
|
/* Structure for tty name info. */
|
|
typedef struct {
|
|
char tty_name[NAME_MAX + 1]; /* file name in /dev */
|
|
dev_t tty_dev; /* major/minor pair */
|
|
} ttyinfo_t;
|
|
|
|
ttyinfo_t *ttyinfo; /* ttyinfo holds actual tty info */
|
|
size_t n_ttyinfo; /* Number of tty info slots */
|
|
|
|
/* Macro to convert memory offsets to rounded kilo-units */
|
|
#define off_to_k(off) ((unsigned) (((off) + 512) / 1024))
|
|
|
|
|
|
/* Number of tasks and processes and addresses of the main process tables. */
|
|
int nr_tasks, nr_procs;
|
|
extern int errno;
|
|
|
|
/* Process tables of the kernel, MM, and FS. */
|
|
struct proc *ps_proc;
|
|
struct mproc *ps_mproc;
|
|
struct fproc *ps_fproc;
|
|
|
|
/* Where is INIT? */
|
|
int init_proc_nr;
|
|
#define low_user init_proc_nr
|
|
|
|
#define KMEM_PATH "/dev/kmem" /* opened for kernel proc table */
|
|
#define MEM_PATH "/dev/mem" /* opened for pm/fs + user processes */
|
|
|
|
int kmemfd, memfd; /* file descriptors of [k]mem */
|
|
|
|
/* Short and long listing formats:
|
|
*
|
|
* PID TTY TIME CMD
|
|
* ppppp tttmmm:ss cccccccccc...
|
|
*
|
|
* F S UID PID PPID PGRP SZ RECV TTY TIME CMD
|
|
* fff s uuu ppppp ppppp ppppp ssss rrrrrrrrrr tttmmm:ss cccccccc...
|
|
*/
|
|
#define S_HEADER " PID TTY TIME CMD\n"
|
|
#define S_FORMAT "%5s %3s %s %s\n"
|
|
#define L_HEADER " F S UID PID PPID PGRP SZ RECV TTY TIME CMD\n"
|
|
#define L_FORMAT "%3o %c %3d %5s %5d %5d %6d %12s %3s %s %s\n"
|
|
|
|
|
|
struct pstat { /* structure filled by pstat() */
|
|
dev_t ps_dev; /* major/minor of controlling tty */
|
|
uid_t ps_ruid; /* real uid */
|
|
uid_t ps_euid; /* effective uid */
|
|
pid_t ps_pid; /* process id */
|
|
pid_t ps_ppid; /* parent process id */
|
|
int ps_pgrp; /* process group id */
|
|
int ps_flags; /* kernel flags */
|
|
int ps_mflags; /* mm flags */
|
|
int ps_ftask; /* fs suspend task */
|
|
int ps_blocked_on; /* what is the process blocked on */
|
|
char ps_state; /* process state */
|
|
vir_bytes ps_tsize; /* text size (in bytes) */
|
|
vir_bytes ps_dsize; /* data size (in bytes) */
|
|
vir_bytes ps_ssize; /* stack size (in bytes) */
|
|
phys_bytes ps_vtext; /* virtual text offset */
|
|
phys_bytes ps_vdata; /* virtual data offset */
|
|
phys_bytes ps_vstack; /* virtual stack offset */
|
|
phys_bytes ps_text; /* physical text offset */
|
|
phys_bytes ps_data; /* physical data offset */
|
|
phys_bytes ps_stack; /* physical stack offset */
|
|
int ps_recv; /* process number to receive from */
|
|
time_t ps_utime; /* accumulated user time */
|
|
time_t ps_stime; /* accumulated system time */
|
|
char *ps_args; /* concatenated argument string */
|
|
vir_bytes ps_procargs; /* initial stack frame from MM */
|
|
};
|
|
|
|
/* Ps_state field values in pstat struct above */
|
|
#define Z_STATE 'Z' /* Zombie */
|
|
#define W_STATE 'W' /* Waiting */
|
|
#define S_STATE 'S' /* Sleeping */
|
|
#define R_STATE 'R' /* Runnable */
|
|
#define T_STATE 'T' /* stopped (Trace) */
|
|
|
|
_PROTOTYPE(char *tname, (Dev_t dev_nr ));
|
|
_PROTOTYPE(char *taskname, (int p_nr ));
|
|
_PROTOTYPE(char *prrecv, (struct pstat *bufp ));
|
|
_PROTOTYPE(void disaster, (int sig ));
|
|
_PROTOTYPE(int main, (int argc, char *argv []));
|
|
_PROTOTYPE(char *get_args, (struct pstat *bufp ));
|
|
_PROTOTYPE(int pstat, (int p_nr, struct pstat *bufp, int Eflag ));
|
|
_PROTOTYPE(int addrread, (int fd, phys_clicks base, vir_bytes addr,
|
|
char *buf, int nbytes ));
|
|
_PROTOTYPE(void usage, (char *pname ));
|
|
_PROTOTYPE(void err, (char *s ));
|
|
_PROTOTYPE(int gettynames, (void));
|
|
|
|
|
|
/*
|
|
* Tname returns mnemonic string for dev_nr. This is "?" for maj/min pairs that
|
|
* are not found. It uses the ttyinfo array (prepared by gettynames).
|
|
* Tname assumes that the first three letters of the tty's name can be omitted
|
|
* and returns the rest (except for the console, which yields "co").
|
|
*/
|
|
char *tname(dev_nr)
|
|
Dev_t dev_nr;
|
|
{
|
|
int i;
|
|
|
|
if (majdev(dev_nr) == TTY_MAJ && mindev(dev_nr) == 0) return "co";
|
|
|
|
for (i = 0; i < n_ttyinfo && ttyinfo[i].tty_name[0] != '\0'; i++)
|
|
if (ttyinfo[i].tty_dev == dev_nr)
|
|
return ttyinfo[i].tty_name + 3;
|
|
|
|
return "?";
|
|
}
|
|
|
|
/* Return canonical task name of task p_nr; overwritten on each call (yucch) */
|
|
char *taskname(p_nr)
|
|
int p_nr;
|
|
{
|
|
int n;
|
|
n = _ENDPOINT_P(p_nr) + nr_tasks;
|
|
if(n < 0 || n >= nr_tasks + nr_procs) {
|
|
return "OUTOFRANGE";
|
|
}
|
|
return ps_proc[n].p_name;
|
|
}
|
|
|
|
/* Prrecv prints the RECV field for process with pstat buffer pointer bufp.
|
|
* This is either "ANY", "taskname", or "(blockreason) taskname".
|
|
*/
|
|
char *prrecv(bufp)
|
|
struct pstat *bufp;
|
|
{
|
|
char *blkstr, *task; /* reason for blocking and task */
|
|
static char recvstr[20];
|
|
|
|
if (bufp->ps_recv == ANY) return "ANY";
|
|
|
|
task = taskname(bufp->ps_recv);
|
|
if (bufp->ps_state != S_STATE) return task;
|
|
|
|
blkstr = "?";
|
|
if (bufp->ps_recv == PM_PROC_NR) {
|
|
if (bufp->ps_mflags & PAUSED)
|
|
blkstr = "pause";
|
|
else if (bufp->ps_mflags & WAITING)
|
|
blkstr = "wait";
|
|
else if (bufp->ps_mflags & SIGSUSPENDED)
|
|
blkstr = "sigsusp";
|
|
} else if (bufp->ps_recv == FS_PROC_NR) {
|
|
switch(bufp->ps_blocked_on) {
|
|
case FP_BLOCKED_ON_PIPE:
|
|
blkstr = "pipe";
|
|
break;
|
|
case FP_BLOCKED_ON_POPEN:
|
|
blkstr = "popen";
|
|
break;
|
|
case FP_BLOCKED_ON_DOPEN:
|
|
blkstr = "dopen";
|
|
break;
|
|
case FP_BLOCKED_ON_LOCK:
|
|
blkstr = "flock";
|
|
break;
|
|
case FP_BLOCKED_ON_SELECT:
|
|
blkstr = "select";
|
|
break;
|
|
case FP_BLOCKED_ON_OTHER:
|
|
blkstr = taskname(bufp->ps_ftask);
|
|
break;
|
|
case FP_BLOCKED_ON_NONE:
|
|
blkstr = "??";
|
|
break;
|
|
}
|
|
}
|
|
(void) sprintf(recvstr, "(%s) %s", blkstr, task);
|
|
return recvstr;
|
|
}
|
|
|
|
/* If disaster is called some of the system parameters imported into ps are
|
|
* probably wrong. This tends to result in memory faults.
|
|
*/
|
|
void disaster(sig)
|
|
int sig;
|
|
{
|
|
fprintf(stderr, "Ooops, got signal %d\n", sig);
|
|
fprintf(stderr, "Was ps recompiled since the last kernel change?\n");
|
|
exit(3);
|
|
}
|
|
|
|
/* Main interprets arguments, gets system addresses, opens [k]mem, reads in
|
|
* process tables from kernel/pm/fs and calls pstat() for relevant entries.
|
|
*/
|
|
int main(argc, argv)
|
|
int argc;
|
|
char *argv[];
|
|
{
|
|
int i;
|
|
struct pstat buf;
|
|
int db_fd;
|
|
int uid = getuid(); /* real uid of caller */
|
|
char *opt;
|
|
int opt_all = FALSE; /* -a */
|
|
int opt_long = FALSE; /* -l */
|
|
int opt_notty = FALSE; /* -x */
|
|
int opt_endpoint = FALSE; /* -E */
|
|
char *ke_path; /* paths of kernel, */
|
|
char *mm_path; /* mm, */
|
|
char *fs_path; /* and fs used in ps -U */
|
|
char pid[2 + sizeof(pid_t) * 3];
|
|
unsigned long ustime;
|
|
char cpu[sizeof(clock_t) * 3 + 1 + 2];
|
|
struct kinfo kinfo;
|
|
int s;
|
|
u32_t system_hz;
|
|
|
|
if(getsysinfo_up(PM_PROC_NR, SIU_SYSTEMHZ, sizeof(system_hz), &system_hz) < 0) {
|
|
exit(1);
|
|
}
|
|
|
|
(void) signal(SIGSEGV, disaster); /* catch a common crash */
|
|
|
|
/* Parse arguments; a '-' need not be present (V7/BSD compatability) */
|
|
for (i = 1; i < argc; i++) {
|
|
opt = argv[i];
|
|
if (opt[0] == '-') opt++;
|
|
while (*opt != 0) switch (*opt++) {
|
|
case 'a': opt_all = TRUE; break;
|
|
case 'E': opt_endpoint = TRUE; break;
|
|
case 'e': opt_all = opt_notty = TRUE; break;
|
|
case 'f':
|
|
case 'l': opt_long = TRUE; break;
|
|
case 'x': opt_notty = TRUE; break;
|
|
default: usage(argv[0]);
|
|
}
|
|
}
|
|
|
|
/* Open memory devices and get PS info from the kernel */
|
|
if ((kmemfd = open(KMEM_PATH, O_RDONLY)) == -1) err(KMEM_PATH);
|
|
if ((memfd = open(MEM_PATH, O_RDONLY)) == -1) err(MEM_PATH);
|
|
if (gettynames() == -1) err("Can't get tty names");
|
|
|
|
getsysinfo(PM_PROC_NR, SI_KINFO, &kinfo);
|
|
|
|
nr_tasks = kinfo.nr_tasks;
|
|
nr_procs = kinfo.nr_procs;
|
|
|
|
/* Allocate memory for process tables */
|
|
ps_proc = (struct proc *) malloc((nr_tasks + nr_procs) * sizeof(ps_proc[0]));
|
|
ps_mproc = (struct mproc *) malloc(nr_procs * sizeof(ps_mproc[0]));
|
|
ps_fproc = (struct fproc *) malloc(nr_procs * sizeof(ps_fproc[0]));
|
|
if (ps_proc == NULL || ps_mproc == NULL || ps_fproc == NULL)
|
|
err("Out of memory");
|
|
|
|
if(getsysinfo(PM_PROC_NR, SI_KPROC_TAB, ps_proc) < 0) {
|
|
fprintf(stderr, "getsysinfo() for SI_KPROC_TAB failed.\n");
|
|
exit(1);
|
|
}
|
|
|
|
if(getsysinfo(PM_PROC_NR, SI_PROC_TAB, ps_mproc) < 0) {
|
|
fprintf(stderr, "getsysinfo() for PM SI_PROC_TAB failed.\n");
|
|
exit(1);
|
|
}
|
|
|
|
if(getsysinfo(VFS_PROC_NR, SI_PROC_TAB, ps_fproc) < 0) {
|
|
fprintf(stderr, "getsysinfo() for VFS SI_PROC_TAB failed.\n");
|
|
exit(1);
|
|
}
|
|
|
|
/* We need to know where INIT hangs out. */
|
|
for (i = FS_PROC_NR; i < nr_procs; i++) {
|
|
if (strcmp(ps_proc[nr_tasks + i].p_name, "init") == 0) break;
|
|
}
|
|
init_proc_nr = i;
|
|
|
|
/* Now loop through process table and handle each entry */
|
|
printf("%s", opt_long ? L_HEADER : S_HEADER);
|
|
for (i = -nr_tasks; i < nr_procs; i++) {
|
|
if (pstat(i, &buf, opt_endpoint) != -1 &&
|
|
(opt_all || buf.ps_euid == uid || buf.ps_ruid == uid) &&
|
|
(opt_notty || majdev(buf.ps_dev) == TTY_MAJ)) {
|
|
if (buf.ps_pid == 0 && i != PM_PROC_NR) {
|
|
sprintf(pid, "(%d)", i);
|
|
} else {
|
|
sprintf(pid, "%d", buf.ps_pid);
|
|
}
|
|
|
|
ustime = (buf.ps_utime + buf.ps_stime) / system_hz;
|
|
if (ustime < 60 * 60) {
|
|
sprintf(cpu, "%2lu:%02lu", ustime / 60, ustime % 60);
|
|
} else
|
|
if (ustime < 100L * 60 * 60) {
|
|
ustime /= 60;
|
|
sprintf(cpu, "%2luh%02lu", ustime / 60, ustime % 60);
|
|
} else {
|
|
sprintf(cpu, "%4luh", ustime / 3600);
|
|
}
|
|
|
|
if (opt_long) printf(L_FORMAT,
|
|
buf.ps_flags, buf.ps_state,
|
|
buf.ps_euid, pid, buf.ps_ppid,
|
|
buf.ps_pgrp,
|
|
#if 0
|
|
off_to_k((buf.ps_tsize
|
|
+ buf.ps_stack - buf.ps_data
|
|
+ buf.ps_ssize)),
|
|
#else
|
|
0,
|
|
#endif
|
|
(buf.ps_flags & RTS_RECEIVING ?
|
|
prrecv(&buf) :
|
|
""),
|
|
tname((Dev_t) buf.ps_dev),
|
|
cpu,
|
|
i <= init_proc_nr || buf.ps_args == NULL
|
|
? taskname(i) : buf.ps_args);
|
|
else
|
|
printf(S_FORMAT,
|
|
pid, tname((Dev_t) buf.ps_dev),
|
|
cpu,
|
|
i <= init_proc_nr || buf.ps_args == NULL
|
|
? taskname(i) : buf.ps_args);
|
|
}
|
|
}
|
|
return(0);
|
|
}
|
|
|
|
char *get_args(bufp)
|
|
struct pstat *bufp;
|
|
{
|
|
int nargv;
|
|
int cnt; /* # of bytes read from stack frame */
|
|
int neos; /* # of '\0's seen in argv string space */
|
|
phys_bytes iframe;
|
|
long l;
|
|
char *cp, *args;
|
|
static union stack {
|
|
vir_bytes stk_i;
|
|
char *stk_cp;
|
|
char stk_c;
|
|
} stk[ARG_MAX / sizeof(char *)];
|
|
union stack *sp;
|
|
|
|
/* Phys address of the original stack frame. */
|
|
iframe = bufp->ps_procargs - bufp->ps_vstack + bufp->ps_stack;
|
|
|
|
/* Calculate the number of bytes to read from user stack */
|
|
l = (phys_bytes) bufp->ps_ssize - (iframe - bufp->ps_stack);
|
|
if (l > ARG_MAX) l = ARG_MAX;
|
|
cnt = l;
|
|
|
|
/* Get cnt bytes from user initial stack to local stack buffer */
|
|
if (lseek(memfd, (off_t) iframe, 0) < 0)
|
|
return NULL;
|
|
|
|
if ( read(memfd, (char *)stk, cnt) != cnt )
|
|
return NULL;
|
|
|
|
sp = stk;
|
|
nargv = (int) sp[0].stk_i; /* number of argv arguments */
|
|
|
|
/* See if argv[0] is with the bytes we read in */
|
|
l = (long) sp[1].stk_cp - (long) bufp->ps_procargs;
|
|
|
|
if ( ( l < 0 ) || ( l > cnt ) )
|
|
return NULL;
|
|
|
|
/* l is the offset of the argv[0] argument */
|
|
/* change for concatenation the '\0' to space, for nargv elements */
|
|
|
|
args = &((char *) stk)[(int)l];
|
|
neos = 0;
|
|
for (cp = args; cp < &((char *) stk)[cnt]; cp++)
|
|
if (*cp == '\0')
|
|
if (++neos >= nargv)
|
|
break;
|
|
else
|
|
*cp = ' ';
|
|
if (cp == args) return NULL;
|
|
*cp = '\0';
|
|
|
|
return args;
|
|
|
|
}
|
|
|
|
/* Pstat collects info on process number p_nr and returns it in buf.
|
|
* It is assumed that tasks do not have entries in fproc/mproc.
|
|
*/
|
|
int pstat(p_nr, bufp, endpoints)
|
|
int p_nr;
|
|
struct pstat *bufp;
|
|
int endpoints;
|
|
{
|
|
int p_ki = p_nr + nr_tasks; /* kernel proc index */
|
|
|
|
if (p_nr < -nr_tasks || p_nr >= nr_procs) {
|
|
fprintf(stderr, "pstat: %d out of range\n", p_nr);
|
|
return -1;
|
|
}
|
|
|
|
if (isemptyp(&ps_proc[p_ki])
|
|
&& !(ps_mproc[p_nr].mp_flags & IN_USE)) {
|
|
return -1;
|
|
}
|
|
|
|
bufp->ps_flags = ps_proc[p_ki].p_rts_flags;
|
|
|
|
if (p_nr >= low_user) {
|
|
bufp->ps_dev = ps_fproc[p_nr].fp_tty;
|
|
bufp->ps_ftask = ps_fproc[p_nr].fp_task;
|
|
bufp->ps_blocked_on = ps_fproc[p_nr].fp_blocked_on;
|
|
} else {
|
|
bufp->ps_dev = 0;
|
|
bufp->ps_ftask = 0;
|
|
bufp->ps_blocked_on = FP_BLOCKED_ON_NONE;
|
|
}
|
|
|
|
if (p_nr >= 0) {
|
|
bufp->ps_ruid = ps_mproc[p_nr].mp_realuid;
|
|
bufp->ps_euid = ps_mproc[p_nr].mp_effuid;
|
|
if(endpoints) bufp->ps_pid = ps_proc[p_ki].p_endpoint;
|
|
else bufp->ps_pid = ps_mproc[p_nr].mp_pid;
|
|
bufp->ps_ppid = ps_mproc[ps_mproc[p_nr].mp_parent].mp_pid;
|
|
/* Assume no parent when the parent and the child share the same pid.
|
|
* This is what PM currently assumes.
|
|
*/
|
|
if(bufp->ps_ppid == bufp->ps_pid) {
|
|
bufp->ps_ppid = NO_PID;
|
|
}
|
|
bufp->ps_pgrp = ps_mproc[p_nr].mp_procgrp;
|
|
bufp->ps_mflags = ps_mproc[p_nr].mp_flags;
|
|
} else {
|
|
if(endpoints) bufp->ps_pid = ps_proc[p_ki].p_endpoint;
|
|
else bufp->ps_pid = NO_PID;
|
|
bufp->ps_ppid = NO_PID;
|
|
bufp->ps_ruid = bufp->ps_euid = 0;
|
|
bufp->ps_pgrp = 0;
|
|
bufp->ps_mflags = 0;
|
|
}
|
|
|
|
/* State is interpretation of combined kernel/mm flags for non-tasks */
|
|
if (p_nr >= low_user) { /* non-tasks */
|
|
if (ps_mproc[p_nr].mp_flags & ZOMBIE)
|
|
bufp->ps_state = Z_STATE; /* zombie */
|
|
else if (ps_mproc[p_nr].mp_flags & STOPPED)
|
|
bufp->ps_state = T_STATE; /* stopped (traced) */
|
|
else if (ps_proc[p_ki].p_rts_flags == 0)
|
|
bufp->ps_state = R_STATE; /* in run-queue */
|
|
else if (ps_mproc[p_nr].mp_flags & (WAITING | PAUSED | SIGSUSPENDED) ||
|
|
fp_is_blocked(&ps_fproc[p_nr]))
|
|
bufp->ps_state = S_STATE; /* sleeping */
|
|
else
|
|
bufp->ps_state = W_STATE; /* a short wait */
|
|
} else { /* tasks are simple */
|
|
if (ps_proc[p_ki].p_rts_flags == 0)
|
|
bufp->ps_state = R_STATE; /* in run-queue */
|
|
else
|
|
bufp->ps_state = W_STATE; /* other i.e. waiting */
|
|
}
|
|
|
|
bufp->ps_tsize = (size_t) ps_proc[p_ki].p_memmap[T].mem_len << CLICK_SHIFT;
|
|
bufp->ps_dsize = (size_t) ps_proc[p_ki].p_memmap[D].mem_len << CLICK_SHIFT;
|
|
bufp->ps_ssize = (size_t) ps_proc[p_ki].p_memmap[S].mem_len << CLICK_SHIFT;
|
|
bufp->ps_vtext = (off_t) ps_proc[p_ki].p_memmap[T].mem_vir << CLICK_SHIFT;
|
|
bufp->ps_vdata = (off_t) ps_proc[p_ki].p_memmap[D].mem_vir << CLICK_SHIFT;
|
|
bufp->ps_vstack = (off_t) ps_proc[p_ki].p_memmap[S].mem_vir << CLICK_SHIFT;
|
|
bufp->ps_text = (off_t) ps_proc[p_ki].p_memmap[T].mem_phys << CLICK_SHIFT;
|
|
bufp->ps_data = (off_t) ps_proc[p_ki].p_memmap[D].mem_phys << CLICK_SHIFT;
|
|
bufp->ps_stack = (off_t) ps_proc[p_ki].p_memmap[S].mem_phys << CLICK_SHIFT;
|
|
|
|
bufp->ps_recv = _ENDPOINT_P(ps_proc[p_ki].p_getfrom_e);
|
|
|
|
bufp->ps_utime = ps_proc[p_ki].p_user_time;
|
|
bufp->ps_stime = ps_proc[p_ki].p_sys_time;
|
|
|
|
bufp->ps_procargs = ps_mproc[p_nr].mp_procargs;
|
|
|
|
if (bufp->ps_state == Z_STATE)
|
|
bufp->ps_args = "<defunct>";
|
|
else if (p_nr > init_proc_nr)
|
|
bufp->ps_args = get_args(bufp);
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* Addrread reads nbytes from offset addr to click base of fd into buf. */
|
|
int addrread(fd, base, addr, buf, nbytes)
|
|
int fd;
|
|
phys_clicks base;
|
|
vir_bytes addr;
|
|
char *buf;
|
|
int nbytes;
|
|
{
|
|
if (lseek(fd, ((off_t) base << CLICK_SHIFT) + addr, 0) < 0)
|
|
return -1;
|
|
|
|
return read(fd, buf, nbytes);
|
|
}
|
|
|
|
void usage(pname)
|
|
char *pname;
|
|
{
|
|
fprintf(stderr, "Usage: %s [-][aeflx]\n", pname);
|
|
exit(1);
|
|
}
|
|
|
|
void err(s)
|
|
char *s;
|
|
{
|
|
extern int errno;
|
|
|
|
if (errno == 0)
|
|
fprintf(stderr, "ps: %s\n", s);
|
|
else
|
|
fprintf(stderr, "ps: %s: %s\n", s, strerror(errno));
|
|
|
|
exit(2);
|
|
}
|
|
|
|
/* Fill ttyinfo by fstatting character specials in /dev. */
|
|
int gettynames()
|
|
{
|
|
static char dev_path[] = "/dev/";
|
|
struct stat statbuf;
|
|
static char path[sizeof(dev_path) + NAME_MAX];
|
|
int index;
|
|
struct ttyent *ttyp;
|
|
|
|
index = 0;
|
|
while ((ttyp = getttyent()) != NULL) {
|
|
strcpy(path, dev_path);
|
|
strcat(path, ttyp->ty_name);
|
|
if (stat(path, &statbuf) == -1 || !S_ISCHR(statbuf.st_mode))
|
|
continue;
|
|
if (index >= n_ttyinfo) {
|
|
n_ttyinfo= (index+16) * 2;
|
|
ttyinfo = realloc(ttyinfo, n_ttyinfo * sizeof(ttyinfo[0]));
|
|
if (ttyinfo == NULL) err("Out of memory");
|
|
}
|
|
ttyinfo[index].tty_dev = statbuf.st_rdev;
|
|
strcpy(ttyinfo[index].tty_name, ttyp->ty_name);
|
|
index++;
|
|
}
|
|
endttyent();
|
|
while (index < n_ttyinfo) ttyinfo[index++].tty_dev= 0;
|
|
|
|
return 0;
|
|
}
|