CVS log for src/bin/ps/ps.c

Up to [cvs.NetBSD.org] / src / bin / ps

Request diff between arbitrary revisions

Keyword substitution: kv
Default branch: MAIN

use emalloc and friends, add ktrace flag printing.

sync with head

use parsenum like everywhere else in the code.

Sort options in usage.

PR standards/11223

Add -G to take a single group argument heading towards POSIX.2 compliance.
Patch from jperkin and reviewed by simonb.

Ooops, restore accidently removed files from merge mishap

Sync with HEAD

Mostly merge changes from HEAD upto 20200411

ps(1): Guard freeing the memory of pinfo with __NO_LEAKS.

No more leaks are detected by LSan/NetBSD as of the LLVM snapshot
(clang10svn) from 2019-09-15.

Plug memory leak in ps(1)

pinfo is allocated in setpinfo() with calloc(3).

Free it when no longer used, just before the program termination.

Detected with LSan.

Sync with HEAD, resolve some conflicts

kvm_geterr() already contains errno, use errx.

use EXIT_FAILURE instead of 1

Pull up following revision(s) (requested by kamil in ticket #515):
	bin/ps/ps.c: revision 1.89
Fix an unitialized memory read bug in ps(1)
rawcpu of type int, is declared inside main(){} and it can be passed as
uninitialized to setpinfo().
The setpinfo() function has a switch checking the value of rawcpu:
  if (!rawcpu)
    pi[i].pcpu /= 1.0 - exp(ki[i].p_swtime * log_ccpu);
rawcpu is set to 1 with the command line argument "-C".
   -C           Change the way the CPU percentage is calculated by using a
                "raw" CPU calculation that ignores "resident" time (this
                normally has no effect).
Bug reproducible with an invocation: "ps u". It hides with "ps uC".
Initialize rawcpu by default to 0, before the getopt(3) machinery.
Detected with MSan running on NetBSD/amd64.
Sponsored by <The NetBSD Foundation>

Fix an unitialized memory read bug in ps(1)

rawcpu of type int, is declared inside main(){} and it can be passed as
uninitialized to setpinfo().
The setpinfo() function has a switch checking the value of rawcpu:

  if (!rawcpu)
    pi[i].pcpu /= 1.0 - exp(ki[i].p_swtime * log_ccpu);

rawcpu is set to 1 with the command line argument "-C".

   -C           Change the way the CPU percentage is calculated by using a
                "raw" CPU calculation that ignores "resident" time (this
                normally has no effect).

Bug reproducible with an invocation: "ps u". It hides with "ps uC".

Initialize rawcpu by default to 0, before the getopt(3) machinery.

Detected with MSan running on NetBSD/amd64.

Sponsored by <The NetBSD Foundation>

Sync with HEAD.  (Note that most of these changes are simply $NetBSD$
tag issues.)

simplify logic; there must be no CPU usage when p_swtime is zero

provide a tree like display with -d, from FreeBSD

Calculate CPU usage (pcpu) once per process if it is required. This change
significantly improves performance for slow machines when output is sorted
by pcpu.

ok martin

KNF
ok martin

Drop main() prototype.

remove \n from error messages

Rebase to HEAD as of a few days ago.

Rebase.

Don't print junk errnos. Avoids:
  % ps -tfoo
  ps: /dev/ttyfoo: Undefined error: 0
  Exit 1

sync with head.

for a reference, the tree before this commit was tagged
as yamt-pagecache-tag8.

this commit was splitted into small chunks to avoid
a limitation of cvs.  ("Protocol error: too many arguments")

Don't bother using variables whose value is never changed from the
initialization value.

Remove some pointless inclusions os sys/user.h

Make ps -s use LTIME instead of TIME in the default output format.

Pull up following revision(s) (requested by dholland in ticket #355):
	bin/ps/ps.c: revision 1.78
Push logic to convert a ttyname to a device number into its own
function. Improve dealing with ptyfs by explicitly handling missing
pts/%d entries, if the kernel supports the pts device (PR 40813).

sync with head.

Push logic to convert a ttyname to a device number into its own
function. Improve dealing with ptyfs by explicitly handling missing
pts/%d entries, if the kernel supports the pts device (PR 40813).

sync with head

Choose better argument name for `-U' (sync with man page).
From Bug Hunting in email.

Use __dead

Fix ps(1) and top(1) to show reasonable CPU numbers i.e. cpu_index() provided
by the kernel, instead of CPU order number, which is generally random.

Sync with HEAD.

Third (and last) commit. See http://mail-index.netbsd.org/source-changes/2009/05/13/msg221222.html

Pull up following revision(s) (requested by mrg in ticket #622):
	bin/csh/csh.1: revision 1.46
	bin/csh/func.c: revision 1.37
	bin/ps/print.c: revision 1.111
	bin/ps/ps.c: revision 1.74
	bin/sh/miscbltin.c: revision 1.38
	bin/sh/sh.1: revision 1.92 via patch
	external/bsd/top/dist/machine/m_netbsd.c: revision 1.7
	lib/libkvm/kvm_proc.c: revision 1.82
	sys/arch/mips/mips/cpu_exec.c: revision 1.55
	sys/compat/darwin/darwin_exec.c: revision 1.57
	sys/compat/ibcs2/ibcs2_exec.c: revision 1.73
	sys/compat/irix/irix_resource.c: revision 1.15
	sys/compat/linux/arch/amd64/linux_exec_machdep.c: revision 1.16
	sys/compat/linux/arch/i386/linux_exec_machdep.c: revision 1.12
	sys/compat/linux/common/linux_limit.h: revision 1.5
	sys/compat/osf1/osf1_resource.c: revision 1.14
	sys/compat/svr4/svr4_resource.c: revision 1.18
	sys/compat/svr4_32/svr4_32_resource.c: revision 1.17
	sys/kern/exec_subr.c: revision 1.62
	sys/kern/init_sysctl.c: revision 1.160
	sys/kern/kern_exec.c: revision 1.288
	sys/kern/kern_resource.c: revision 1.151
	sys/sys/param.h: patch
	sys/sys/resource.h: revision 1.31
	sys/sys/sysctl.h: revision 1.184
	sys/uvm/uvm_extern.h: revision 1.153
	sys/uvm/uvm_glue.c: revision 1.136
	sys/uvm/uvm_mmap.c: revision 1.128
	usr.bin/systat/ps.c: revision 1.32
- - add new RLIMIT_AS (aka RLIMIT_VMEM) resource that limits the total
address space available to processes.  this limit exists in most other
modern unix variants, and like most of them, our defaults are unlimited.
remove the old mmap / rlimit.datasize hack.
- - adds the VMCMD_STACK flag to all the stack-creation vmcmd callers.
it is currently unused, but was added a few years ago.
- - add a pair of new process size values to kinfo_proc2{}. one is the
total size of the process memory map, and the other is the total size
adjusted for unused stack space (since most processes have a lot of
this...)
- - patch sh, and csh to notice RLIMIT_AS.  (in some cases, the alias
RLIMIT_VMEM was already present and used if availble.)
- - patch ps, top and systat to notice the new k_vm_vsize member of
kinfo_proc2{}.
- - update irix, svr4, svr4_32, linux and osf1 emulations to support
this information.  (freebsd could be done, but that it's best left
as part of the full-update of compat/freebsd.)
this addresses PR 7897.  it also gives correct memory usage values,
which have never been entirely correct (since mmap), and have been
very incorrect since jemalloc() was enabled.
tested on i386 and sparc64, build tested on several other platforms.
thanks to many folks for feedback and testing but most espcially
chuq and yamt for critical suggestions that lead to this patch not
having a special ugliness i wasn't happy with anyway :-)

- add new RLIMIT_AS (aka RLIMIT_VMEM) resource that limits the total
address space available to processes.  this limit exists in most other
modern unix variants, and like most of them, our defaults are unlimited.
remove the old mmap / rlimit.datasize hack.

- adds the VMCMD_STACK flag to all the stack-creation vmcmd callers.
it is currently unused, but was added a few years ago.

- add a pair of new process size values to kinfo_proc2{}. one is the
total size of the process memory map, and the other is the total size
adjusted for unused stack space (since most processes have a lot of
this...)

- patch sh, and csh to notice RLIMIT_AS.  (in some cases, the alias
RLIMIT_VMEM was already present and used if availble.)

- patch ps, top and systat to notice the new k_vm_vsize member of
kinfo_proc2{}.

- update irix, svr4, svr4_32, linux and osf1 emulations to support
this information.  (freebsd could be done, but that it's best left
as part of the full-update of compat/freebsd.)


this addresses PR 7897.  it also gives correct memory usage values,
which have never been entirely correct (since mmap), and have been
very incorrect since jemalloc() was enabled.

tested on i386 and sparc64, build tested on several other platforms.

thanks to many folks for feedback and testing but most espcially
chuq and yamt for critical suggestions that lead to this patch not
having a special ugliness i wasn't happy with anyway :-)

fix -Wsign-compare issues

fix dev_t 64 bit issues.

Sync usage with man page.

Add "ps -A" option.  This displays information about all processes,
like BSD "ps -ax" or SYSV "ps -ef".  It's specified in SUSv3.

Also make some minor clarifications to the man page.

Sync with wrstuden-revivesa-base-2.

Remove the \n and tabs from the __COPYRIGHT() strings.

Sync w/ -current. 34 merge conflicts to follow.

sync with head

PR/38755: murray armfield: /bin posix programs missing setprogname and/or
setlocale

sync with head.

Remove clause 3 and 4 from TNF licenses

sync with head.

sync with HEAD

ps -t? needs to imply -x. Otherwise the non-x logic filters all the results
out. Thanks to der Mouse for prompting me to fix it...

Add a CPUID keyword; idea and some of the code from Anon Ymous

sync with HEAD

make -s alter the default format rather than doing parsefmt.
it allows users to use -o to replace it later.

Revert the changes done after newlock2 merge, except for the removal of
the now unused LSDEAD state. The current kernel headers are now
source-compatible with pre-newlock2 kernel headers again.

Approved by Andrew Doran.

Sync with kernel changes introduced by merging the newlock2 branch.

* When all columns are given null customised headers, the blank header
  line is not printed at all.  This is specified in P1003.1-2004
  (SUSv3), and is useful.

* Customised headers may contain embedded space, commas and equals
  signs.  To specify multiple customised headers, use multiple -o or -O
  options.  This is specified (for "-o", not for "-O") in P1003.1-2004
  (SUSv3), and is useful.

* When a column is given a null (blank) customised header, it keeps its
  default minimum width.  This is specified in P1003.1-2004 (SUSv3), and
  seems harmless.

* Fix a bug that made it impossible to print the same keyword multiple
  times, with different customised headers each time.  (Previously, the
  last customised header was used for all instances of the keyword.)

* Make the behaviour of "-O" more useful.  The first -O option adds
  the default keywords only if there have not yet been any formatting
  options, and multiple -O options now insert their keywords in adjacent
  positions.  Now {ps -j -O %cpu} is like {ps -j} with one extra column;
  Previously, it would have had all the columns implied by "-j", plus
  all the default columns, plus the extra column specified by "-O".

* Convert from home-grown linked lists to SIMPLEQ lists.

Discussed in tech-userlevel.

factor out the code that does numeric parsing into a function and check for
errors properly.

sprinkle a little const, and now everything compiles with WARNS=3

appease gcc -Wuninitialized

allow -t ttyp0 -t p0 -t /dev/ttyp0 -t /dev/pts/0

Make ps understand "-tpts/4" by assuming that ptyfs is mounted on
/dev/pts and that the '/' immediately preceding the digits does not
need a 'p' in front of it (ick, kludge_oldps_options()).

Add (unsigned char) cast to ctype function

#define<TAB>

Wrap some long lines.
Consisently use (void) in front of printf(...);
A few more white-space nits.

KNF, ANSFify, de-__P, unC++ify, white space nits.

g/c SDEAD

Move UCB-licensed code from 4-clause to 3-clause licence.

Patches provided by Joel Baker in PR 22249, verified by myself.

- add keywords for utime, stime and ctime (output sss.ssssss if small)
- fiddle with keyword table - mainly to keep width under 80 cols)
- add -k keylist so sort output
- don't use logs to work out field width, get -MAXINT right
- user defined headers can be 1 char (actually they can be zero length)
  so ?? in tty column isn't guaranteed to be wider than header
- use PRId64 to remove some long long casts
- use kvm_openfiles(NULL, NULL, NULL, KVN_NO_FILES, ..) unless core file
  specified
(approved by christos)

Merge the nathanw_sa branch.

Catch up to -current.

Small typo fix.

Treat SSTOP and SACTIVE the same way in pick_representative_lwp(), and
include LSSTOP in the state hierarchy.

Consider LWPs in LSSUSPENDED as a worst-case for a representative LWP.

g/c procfs fallback code, it's redundant now that we use sysctl to get
process info

Use the real process state when picking a representative LWP, not the
old-style state that was faked up for us by sysctl.

Don't puke on zero-LWP processes (zombies).

Note the presense of multiple LWPs or scheduler activations under "state".
Add a "lstate" that prints LWP-specific state and display it instead
of "state" under ps -s.

Add -s to synopsis.

Don't need to free the kinfo_lwp pointer; libkvm does that.

Don't shadow global symbol "sleep".

Add support for handling multi-LWP processes by selecting one LWP as
"representative" of the process.
Also add a -s option that prints a line for each LWP.
While I'm here, remove the /proc fallback code.

file ps.c was added on branch nathanw_sa on 2002-04-23 20:41:14 +0000

Sync getopt() / man page with actual getopt options.

fix minor WARNS=2

Remove duplicate `if'. Noted by Geoff C. Wing in bin/13551.

Add a small optimizetion in savevars().  break out of the loop once needcomm
is set since setting it multiple times only wastes cpu time.

If there were no processes to list, it meant that our criteria failed
and didnt match anything.  Dont make our caller think we succeeded.

Sync w/ netbsd-1-5-base.

Don't check namelist when choosing to use the sysctl interface.

Fix handling of "ps U<user>" where <user> ended it 't' - now if the
first argument doesn't start with a '-' and ends in 't', the 't' only
gets converted to a 'T' (for listing processes on the current tty)
if it doesn't contain any other command line option characters that
take an argument.  Problem noticed by ITOH Yasufumi on tech-userlevel.

If the 'start' format was the last column, the width wouldn't be set
and started() output a zero length string instead of a string the right
length - fix header width calculations so the the last column is handled
the same all others.  This only showed up for 'start' because started()
knew that the column header is the same width and the field itself...

Update copyrights.

Calculate field widths on the fly so that all columns line up nicely
and make more effective use of screen real estate when some columns
(eg USER and VSZ) didn't need the full default width.

Use new sysctl/kvm interfaces.  This will stop the "proc size mismatch"
errors when internal kernel structures change size.  Also remove
the sgid bit - all live kernel data is accessed through the sysctl
interface.

Pull up revision 1.37 (requested by abs):
  Only check the terminal size via STDOUT, not STDERR or STDIN.

If ps cannot get the terminal size from STDOUT, it tried STDERR and STDIN.
What this does to the POLS when a script that is parsing ps output just happens
to run in a narrow terminal cannot be described in polite company.
If STDOUT is not a tty, leave width at 80.

Use tty "?" to select processes with no controlling tty and tty "-" to
select processes with a revoked controlling tty.

Idea from SunOS/Solaris, suggested by Alexis Rosen.

make -M imply -K.  the current system is not likely to be
a reasonable approximation of a crash dump.

Pull up to last week's -current.

Document -U switch (displays processes belonging to given user/uid).
Reported in PR 8943 by WL <wliao@dura.spc.uchicago.edu>

"ps -e" only shows the environment for the processes owned by the
current user id or all process if run by root.

Fixes PR security/5967 from Todd Vierling.

Since our gcc doesn't warn about NULL format strings anymore, we can
fix the incorrect err(1, "%s", "") et al.
Closes PR bin/7592 by cgd.

pull up revs 1.29-1.31 from trunk (requested by jdolecek):
  When using the procfs for extracting process information, extract process
  start time, arguments, and session leadership status.  Also, fall back to
  procfs when kvm_openfiles() completely fails (e.g. when /dev/mem is not
  readable).  Fixes PR#7772.

command(): don't start using procfs_getargv() instead of kvm_getargv()
just because the latter returned a NULL - that may normally happen,
when the process is, for example, a kernel thread

procfs_getargv(): if the name is brackened in '(' and ')', just
	return NULL immediately - the code in command() will DTRT
	and the output will be same for procfs- and kvm- based lookup
	in all cases

getkinfo_kvm(): fix bug in previous - handled size_t i as if it would be signed
saveuser(): don't check kd - if we get here, kd is never NULL

When using the procfs for extracting process information, extract
also process start time, process arguments and session leadership status.

The procfs fallback is also used when kvm_openfiles() completely fails
(e.g. when /dev/mem is not readable).

Solves PR 7772, though the final implementation is different.

Minor fixes:
1.  Use statfs() to verify that /proc is a procfs.
2.  Add -K option to disable /proc-based method.
3.  Make warnings less verbose, but still accurate.

Added experimental fallback /proc-based lookup.  Warnings are printed when the
fallback method is used, as the results could be untrustworthy if an
intruder is present.  It is highly likely that NetBSD-1.5 will have
an improved kvm interface for reading process information, at which
point this code can be garbage-collected.  Also added a word to the
man page -x option description while I was here.

Fixed segfaulting for "ps -T".  There is no optarg with -T, so instructions
under label "tty:" need to use ttname instead.

Add a -U option to select by user/uid.

Be more retentive about use of NOTREACHED and noreturn.

Delint.

__AUDIT__ cleanup.

- change setgid kmem programs (that lend themselves to this) so setegid(getgid())
  and the top, and then set the effective gid back to kmem around the call to
  kvm_openfiles().  this reduces the time group kmem is available.
- for those above that also allow this, setgid(getgid()) after the call to
  kvm_openfiles() to fully revoke priviledges.
- some KNF
- use err(3) over fprintf(3) in some places

getopt returns -1 not EOF
cleanup manpage

Fix compiler warnings
Add WARNS=1

note that strcpy is safe.

Message buffer passed to kvm_openfiles() should _POSIX2_LINE_MAX long; PR#3266

Fix SEGV when kludging old-style options.
From Greg Stark <gsstark@mit.edu>, PR #2550.

Create and strvis(3) the argument and environment lists only as needed.
Do not keep extra copies around.

Gather the argument and environment lists when we actually want to print them.

convert to new RCS id conventions.

Integrate `-c' option, from Arne Juul.

clean up import

new ps, to match libkvm

update for changed flags

fix from Thomas Eberhardt <thomas@mathematik.uni-Bremen.de>:
don't try to display the resident text size; we can't do that anymore

Add RCS identifiers.

from Brian Somers <brian@awfulhak.demon.co.uk>:
The kvm entry isn't updated for command unless 'time' is given,
therefore
	ps -o command
outputs blank lines [or, in NetBSD, seg faults]
	ps -o time,command
works.

file ps.c was added on branch netbsd-0-9 on 1993-07-19 11:02:11 +0000

from Brian Somers <brian@awfulhak.demon.co.uk>:
The kvm entry isn't updated for command unless 'time' is given,
therefore
	ps -o command
outputs blank lines [or, in NetBSD, seg faults]
	ps -o time,command
works.

fix alien 'w' handling.  you really don't want to know...

add support and documentation for the 'e' option, to print out
the process environment.

changed "Id" to "Header" for rcsids

added rcs ids to all files

after 0.2.2 "stable" patches applied

initial import of 386bsd-0.1 sources

Initial revision