usr.bin/limits/limits.1

.\" Copyright (c) 1996 David Nugent <davidn@blaze.net.au>
.\" All rights reserved.
.\"
.\" Redistribution and use in source and binary forms, with or without
.\" modification, is permitted provided that the following conditions
.\" are met:
.\" 1. Redistributions of source code must retain the above copyright
.\"    notice immediately at the beginning of the file, without modification,
.\"    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. This work was done expressly for inclusion into FreeBSD.  Other use
.\"    is permitted provided this notation is included.
.\" 4. Absolutely no warranty of function or purpose is made by the author
.\"    David Nugent.
.\" 5. Modifications may be freely made to this file providing the above
.\"    conditions are met.
.\"
.Dd June 25, 2020
.Dt LIMITS 1
.Os
.Sh NAME
.Nm limits
.Nd set or display process resource limits
.Sh SYNOPSIS
.Nm
.Op Fl C Ar class | Fl P Ar pid | Fl U Ar user
.Op Fl SHB
.Op Fl ea
.Op Fl bcdfklmnopstuvw Op Ar val
.Nm
.Op Fl C Ar class | Fl U Ar user
.Op Fl SHB
.Op Fl bcdfklmnopstuvw Op Ar val
.Op Fl E
.Oo
.Op Ar name Ns = Ns Ar value ...
.Ar command
.Oc
.Sh DESCRIPTION
The
.Nm
utility either prints or sets kernel resource limits, and may optionally set
environment variables like
.Xr env 1
and run a program with the selected resources.
Three uses of the
.Nm
utility are possible:
.Bl -tag -width indent
.It Xo
.Nm
.Op Ar limitflags
.Op Ar name Ns = Ns Ar value ...
.Ar command
.Xc
This usage sets limits according to
.Ar limitflags ,
optionally sets environment variables given as
.Ar name Ns = Ns Ar value
pairs, and then runs the specified
.Ar command .
.It Nm Op Ar limitflags
This usage determines values of resource settings according to
.Ar limitflags ,
does not attempt to set them and outputs these values to
standard output.
By default, this will output the current kernel resource settings
active for the calling process.
Using the
.Fl C Ar class
or
.Fl U Ar user
options, you may also display the current resource settings modified
by the appropriate login class resource limit entries from
the
.Xr login.conf 5
login capabilities database.
.It Nm Fl e Op Ar limitflags
This usage determines values of resource settings according to
.Ar limitflags ,
but does not set them.
Like the previous usage, it outputs these values to standard
output, except that it will emit them in
.Ic eval
format, suitable for the calling shell.
If the shell is known (i.e., it is one of
.Nm sh , csh , bash , tcsh , ksh , pdksh
or
.Nm rc ) ,
.Nm
emits
.Ic limit
or
.Ic ulimit
commands in the format understood by
that shell.
If the name of the shell cannot be determined, then the
.Ic ulimit
format used by
.Xr sh 1
is used.
.Pp
This is very useful for setting limits used by scripts, or prior
launching of daemons and other background tasks with specific
resource limit settings, and provides the benefit of allowing
global configuration of maximum resource usage by maintaining a
central database of settings in the login class database.
.Pp
Within a shell script,
.Nm
will normally be used with eval within backticks as follows:
.Pp
.Dl "eval `limits -e -C daemon`"
.Pp
which causes the output of
.Nm
to be evaluated and set by the current shell.
.El
.Pp
The value of
.Ar limitflags
specified in the above contains one or more of the following options:
.Bl -tag -width ".Fl C Ar class"
.It Fl C Ar class
Use current resource values, modified by the resource entries applicable
for the login class
.Ar class .
.It Fl U Ar user
Use current resource values, modified by the resource entries applicable
to the login class the
.Ar user
belongs to.
If user does not belong to any class, then the resource capabilities
for the
.Dq Li default
class are used, if it exists, or the
.Dq Li root
class if the user is a superuser account.
.It Fl P Ar pid
Select or set limits for the process identified by the
.Ar pid .
.It Fl S
Select display or setting of
.Dq soft
(or current) resource limits.
If specific limits settings follow this switch, only soft limits are
affected unless overridden later with either the
.Fl H
or
.Fl B
options.
.It Fl H
Select display or setting of
.Dq hard
(or maximum) resource limits.
If specific limits settings follow this switch, only hard limits are
affected until overridden later with either the
.Fl S
or
.Fl B
options.
.It Fl B
Select display or setting of both
.Dq soft
(current) or
.Dq hard
(maximum)
resource limits.
If specific limits settings follow this switch, both soft and hard
limits are affected until overridden later with either the
.Fl S
or
.Fl H
options.
.It Fl e
Select
.Dq "eval mode"
formatting for output.
This is valid only in display mode and cannot be used when running a
command.
The exact syntax used for output depends upon the type of shell from
which
.Nm
is invoked.
.It Fl b Op Ar val
Select or set the
.Va sbsize
resource limit.
.It Fl c Op Ar val
Select or set (if
.Ar val
is specified) the
.Va coredumpsize
resource limit.
A value of 0 disables core dumps.
.It Fl d Op Ar val
Select or set (if
.Ar val
is specified) the
.Va datasize
resource limit.
.It Fl f Op Ar val
Select or set the
.Va filesize
resource limit.
.It Fl k Op Ar val
Select or set the
.Va kqueues
resource limit.
.It Fl l Op Ar val
Select or set the
.Va memorylocked
resource limit.
.It Fl m Op Ar val
Select or set the
.Va memoryuse
size limit.
.It Fl n Op Ar val
Select or set the
.Va openfiles
resource limit.
The system-wide limit on the maximum number of
open files per process can be viewed by examining the
.Va kern.maxfilesperproc
.Xr sysctl 8
variable.
The total number of simultaneously open files in the entire
system is limited to the value displayed by the
.Va kern.maxfiles
.Xr sysctl 8
variable.
.It Fl o Op Ar val
Select or set the
.Va umtxp
resource limit.
The limit determines the maximal number of the process-shared locks
which may be simultaneously created by the processes owned by the
user, see
.Xr pthread 3 .
.It Fl p Op Ar val
Select or set the
.Va pseudoterminals
resource limit.
.It Fl s Op Ar val
Select or set the
.Va stacksize
resource limit.
.It Fl t Op Ar val
Select or set the
.Va cputime
resource limit.
.It Fl u Op Ar val
Select or set the
.Va maxproc
resource limit.
The system-wide limit on the maximum number of processes
allowed per UID can be viewed by examining the
.Va kern.maxprocperuid
.Xr sysctl 8
variable.
The maximum number of processes that can be running simultaneously
in the entire system is limited to the value of the
.Va kern.maxproc
.Xr sysctl 8
variable.
.It Fl v Op Ar val
Select or set the
.Va virtualmem
resource limit.
This limit encompasses the entire VM space for the user process
and is inclusive of text, data, bss, stack,
.Xr brk 2 ,
.Xr sbrk 2
and
.Xr mmap 2 Ns 'd
space.
.It Fl w Op Ar val
Select or set the
.Va swapuse
resource limit.
.El
.Pp
Valid values for
.Ar val
in the above set of options consist of either the
string
.Dq Li infinity ,
.Dq Li inf ,
.Dq Li unlimited
or
.Dq Li unlimit
for an infinite (or kernel-defined maximum)
limit, or a numeric value optionally followed by a suffix.
Values which relate to size default to a value in bytes, or one of the
following suffixes may be used as a multiplier:
.Pp
.Bl -tag -offset indent -width 4n -compact
.It Li b
512 byte blocks.
.It Li k
kilobytes (1024 bytes).
.It Li m
megabytes (1024*1024 bytes).
.It Li g
gigabytes.
.It Li t
terabytes.
.El
.Pp
The
.Va cputime
resource defaults to a number of seconds, but a multiplier may be
used, and as with size values, multiple values separated by a valid
suffix are added together:
.Pp
.Bl -tag -offset indent -width 4n -compact
.It Li s
seconds.
.It Li m
minutes.
.It Li h
hours.
.It Li d
days.
.It Li w
weeks.
.It Li y
365 day years.
.El
.Bl -tag -width ".Fl C Ar class"
.It Fl E
Cause
.Nm
to completely ignore the environment it inherits.
.It Fl a
Force all resource settings to be displayed even if
other specific resource settings have been specified.
For example, if you wish to disable core dumps when starting up
the Usenet News system, but wish to set all other resource settings
as well that apply to the
.Dq Li news
account, you might use:
.Pp
.Dl "eval `limits -U news -aBec 0`"
.Pp
As with the
.Xr setrlimit 2
call, only the superuser may raise process
.Dq hard
resource limits.
Non-root users may, however, lower them or change
.Dq soft
resource limits
within to any value below the hard limit.
When invoked to execute a program, the failure of
.Nm
to raise a hard limit is considered a fatal error.
.El
.Sh EXIT STATUS
The
.Nm
utility
exits with
.Dv EXIT_FAILURE
if usage is incorrect in any way; i.e., an invalid
option, or set/display options are selected in the same invocation,
.Fl e
is used when running a program, etc.
When run in display or eval mode,
.Nm
exits with a status of
.Dv EXIT_SUCCESS .
When run in command mode and execution of the command succeeds, the exit status
will be whatever the executed program returns.
.Sh EXAMPLES
Show current stack size limit:
.Bd -literal -offset indent
$ limits -s
Resource limits (current):
	  stacksize              524288 kB
.Ed
.Pp
Try to run
.Xr ls 1
with 1 byte of
.Va datasize
limit:
.Bd -literal -offset indent
$ limits -d 1b ls
Data segment size exceeds process limit
Abort trap
.Ed
.Pp
Produce
.Ql eval mode
output to limit
.Va sbsize
to 1 byte.
Output obtained when command is run from
.Xr sh 1 :
.Bd -literal -offset indent
$ limits -e -b 1b
ulimit -b 512;
.Ed
.Pp
Same as above from
.Xr csh 1
.Bd -literal -offset indent
% limits -e -b 1b
limit -h sbsize 512;
limit sbsize 512;
.Ed
.Sh SEE ALSO
.Xr csh 1 ,
.Xr env 1 ,
.Xr limit 1 ,
.Xr sh 1 ,
.Xr getrlimit 2 ,
.Xr setrlimit 2 ,
.Xr login_cap 3 ,
.Xr login.conf 5 ,
.Xr rctl 8 ,
.Xr sysctl 8
.Sh HISTORY
The
.Nm
utility first appeared in
.Fx 2.1.7 .
.Sh AUTHORS
The
.Nm
utility was written by
.An David Nugent Aq Mt davidn@FreeBSD.org .
.Sh BUGS
The
.Nm
utility does not handle commands with equal
.Pq Ql =
signs in their
names, for obvious reasons.
.Pp
The
.Nm
utility makes no effort to ensure that resource settings emitted or displayed
are valid and settable by the current user.
Only a superuser account may raise hard limits, and when doing so
the
.Fx
kernel will silently lower limits to values less than
specified if the values given are too high.