top - display and update information about the top cpu processes
top [ -CISTabcinqtuv ] [ -dcount ] [ -mmode ] [ -ofield ] [ -stime ] [ -Uusername ] [ number ]
Top displays the top 30 processes on the system and periodically updates this information. Raw cpu percentage is used to rank the pro- cesses. If number is given, then the top number processes will be dis- played instead of the default. Top makes a distinction between terminals that support advanced capa- bilities and those that do not. This distinction affects the choice of defaults for certain options. In the remainder of this document, an "intelligent" terminal is one that supports cursor addressing, clear screen, and clear to end of line. Conversely, a "dumb" terminal is one that does not support such features. If the output of top is redi- rected to a file, it acts as if it were being run on a dumb terminal.
-C, --color Turn off the use of color in the display. -I, --idle-procs Do not display idle processes. By default, top displays both active and idle processes. -S, --system-procs Show system processes in the display. Normally, system pro- cesses such as the pager and the swapper are not shown. This option makes them visible. -T, --tag-names List all available color tags and the current set of tests used for color highlighting, then exit. -a, --all Show all processes for as long as possible. This is shorthand for "-d all all". This option is especially handy in batch mode. -b, -n, --batch Use "batch" mode. In this mode, all input from the terminal is ignored. Interrupt characters (such as ^C and ^\) still have an effect. This is the default on a dumb terminal, or when the output is not a terminal. -c, --full-commands Show the full command line for each process. Default is to show just the command name. This option is not supported on all platforms. -i, --interactive Use "interactive" mode. In this mode, any input is immediately read for processing. See the section on "Interactive Mode" for an explanation of which keys perform what functions. After the command is processed, the screen will immediately be updated, even if the command was not understood. This mode is the default when standard output is an intelligent terminal. -q, --quick Renice top to -20 so that it will run faster. This can be used when the system is being very sluggish to improve the possibil- ity of discovering the problem. This option can only be used by root. -t, --threads Show individual threads on separate lines. By default, on sys- tems which support threading, each process is shown with a count of the number of threads. This option shows each thread on a separate line. This option is not supported on all platforms. -u, --uids Do not take the time to map uid numbers to usernames. Normally, top will read as much of the file "/etc/passwd" as is necessary to map all the user id numbers it encounters into login names. This option disables all that, while possibly decreasing execu- tion time. The uid numbers are displayed instead of the names. -v, --version Write version number information to stderr then exit immedi- ately. No other processing takes place when this option is used. To see current revision information while top is running, use the help command "?". -d count, --displays count Show only count displays, then exit. A display is considered to be one update of the screen. This option allows the user to select the number of displays he wants to see before top auto- matically exits. Any proper prefix of the words "infinity", "maximum", or "all" can be used to indicate an infinite number of displays. The default for intelligent terminals is infinity. The default for dumb terminals is 1. -m mode, --mode=mode Start the display in an alternate mode. Some platforms support multiple process displays to show additional process informa- tion. The value mode is a number indicating which mode to dis- play. The default is 0. On platforms that do not have multiple display modes this option has no effect. -o field, --sort-order=field Sort the process display area on the specified field. The field name is the name of the column as seen in the output, but in lower case. Likely values are "cpu", "size", "res", and "time", but may vary on different operating systems. Note that not all operating systems support this option. -s time, --delay=time Set the delay between screen updates to time seconds. The default delay between updates is 5 seconds. -U username, --user=username Show only those processes owned by username. This option cur- rently only accepts usernames and will not understand uid num- bers. Both count and number fields can be specified as "infinite", indicating that they can stretch as far as possible. This is accomplished by using any proper prefix of the keywords "infinity", "maximum", or "all". The default for count on an intelligent terminal is, in fact, infinity. The environment variable TOP is examined for options before the command line is scanned. This enables a user to set his or her own defaults. The number of processes to display can also be specified in the envi- ronment variable TOP. The options -C, -I, -S, and -u are actually tog- gles. A second specification of any of these options will negate the first. Thus a user who has the environment variable TOP set to "-I" may use the command "top -I" to see idle processes.
When top is running in "interactive mode", it reads commands from the terminal and acts upon them accordingly. In this mode, the terminal is put in "CBREAK", so that a character will be processed as soon as it is typed. Almost always, a key will be pressed when top is between dis- plays; that is, while it is waiting for time seconds to elapse. If this is the case, the command will be processed and the display will be updated immediately thereafter (reflecting any changes that the command may have specified). This happens even if the command was incorrect. If a key is pressed while top is in the middle of updating the display, it will finish the update and then process the command. Some commands require additional information, and the user will be prompted accord- ingly. While typing this information in, the user's erase and kill keys (as set up by the command stty) are recognized, and a newline ter- minates the input. Note that a control-L (^L) always redraws the cur- rent screen and a space forces an immediate update to the screen using new data. These commands are currently recognized: h or ? Display a summary of the commands (help screen). Version infor- mation is included in this display. C Toggle the use of color in the display. c Display only processes whose commands match the specified string. An empty string will display all processes. This com- mand is not supported on all platforms. d Change the number of displays to show (prompt for new number). Remember that the next display counts as one, so typing d1 will make top show one final display and then immediately exit. f Toggle the display of the full command line. H Toggle the display of threads on separate lines. By default, on systems which support threading, each process is shown with a count of the number of threads. This command shows each thread on a separate line. This command is not supported on all plat- forms. i (or I) Toggle the display of idle processes. k Send a signal ("kill" by default) to a list of processes. This acts similarly to the command kill(1)). M Sort display by memory usage. Shorthand for "o size". m Change to a different process display mode. Some systems pro- vide multiple display modes for the process display which shows different information. This command toggles between the avail- able modes. This command is not supported on all platforms. N Sort by process id. Shorthand for "o pid". n or # Change the number of processes to display (prompt for new num- ber). o Change the order in which the display is sorted. This command is not available on all systems. The sort key names vary fron system to system but usually include: "cpu", "res", "size", "time". The default is cpu. P Sort by CPU usage. Shorthand for "o cpu". q Quit top. r Change the priority (the "nice") of a list of processes. This acts similarly to the command renice(8)). s Change the number of seconds to delay between displays (prompt for new number). T Sort by CPU time. Shorthand for "o time". U Toggle between displaying usernames and uids. u Display only processes owned by a specific username (prompt for username). If the username specified is simply "+", then pro- cesses belonging to all users will be displayed.
The actual display varies depending on the specific variant of Unix that the machine is running. This description may not exactly match what is seen by top running on this particular machine. Differences are listed at the end of this manual entry. The top lines of the display show general information about the state of the system. The first line shows (on some systems) the last process id assigned to a process, the three load averages, the system uptime, and the current time. The second line displays the total number of processes followed by a breakdown of processes per state. Examples of states common to Unix systems are sleeping, running, starting, stopped, and zombie. The next line displays a percentage of time spent in each of the processor states (typically user, nice, system, idle, and iowait). These percentages show the processor activity during the time since the last update. For multi-processor systems, this information is a summation of time across all processors. The next line shows ker- nel-related activity (not available on all systems). The numbers shown on this line are per-second rates sampled since the last update. The exact information displayed varies between systems, but some examples are: context switches, interrupts, traps, forks, and page faults. The last one or two lines show a summary of memory and swap activity. These lines vary between systems. The remainder of the screen displays information about individual pro- cesses. This display is similar in spirit to ps(1) but it is not exactly the same. The columns displayed by top will differ slightly between operating systems. Generally, the following fields are dis- played: PID The process id. USERNAME Username of the process's owner (if -u is specified, a UID col- umn will be substituted for USERNAME). THR The number of threads in the processes (this column may also be labeled NLWP). PRI Current priority of the process. NICE Nice amount in the range -20 to 20, as established by the use of the command nice. SIZE Total size of the process (text, data, and stack) given in kilo- bytes. RES Resident memory: current amount of process memory that resides in physical memory, given in kilobytes. STATE Current state (typically one of "sleep", "run", "idl", "zomb", or "stop"). TIME Number of system and user cpu seconds that the process has used. CPU Percentage of available cpu time used by this process. COMMAND Name of the command that the process is currently running.
Top supports the use of ANSI color in its output. By default, color is available but not used. The environment variable TOPCOLORS specifies colors to use and conditions for which they should be used. At the present time, only numbers in the summay display area can be colored. In a future version it will be possible to highlight numbers in the process display area as well. The environment variable is the only way to specify color: there is no equivalent command line option. Note that the environment variable TOPCOLOURS is also understood. The British spelling takes precedence. The use of color only works on ter- minals that understand and process ANSI color escape sequences. The environment variable is a sequence of color specifications, sepa- rated by colons. Each specification takes the form tag=min,max#code where tag is the name of the value to check, min and max specify a range for the value, and code is an ANSI color code. Multiple color codes can be listed and separated with semi-colons. A missing min implies the lowest possible value (usually 0) and a missing max implies infinity. The comma must always be present. When specifying numbers for load averages, they should be multiplied by 100. For example, the specification 1min=500,1000#31 indicates that a 1 minute load average between 5 and 10 should be displayed in red. Color attributes can be combined. For example, the specification 5min=1000,#37;41 indicates that a 5 minute load average higher than 10 should be displayed with white characters on a red background. A special tag named header is used to control the color of the header for process display. It should be specified with no lower and upper limits, specifically header=,# followed by the ANSI color code. You can see a list of color codes recognized by this installation of top with the -T option. This will also show the current set of tests used for color highligting, as specified in the environment.
TOP user-configurable defaults for options. TOPCOLORS color specification
As with ps(1), things can change while top is collecting information for an update. The picture it gives is only a close approximation to reality.
kill(1), ps(1), stty(1), mem(4), renice(8) @MAN_SUPPLEMENT@
DEC OSF/1 NOTES
Original author was Anthony Baxter, <firstname.lastname@example.org>. Derived originally from m_ultrix, by David S. Comay <email@example.com>, although by now there is hardly any of the code from m_ultrix left. Helped a lot by having the source for syd(1), by Claus Kalle, and from several people at DEC who helped with providing information on some of the less-documented bits of the kernel interface. Patches from Rainer Orth <ro@TechFak.Uni-Bielefeld.DE> Theory of operation: Use Mach calls to build up a structure that con- tains all the sorts of stuff normally found in a struct proc in a BSD system. Then everything else uses this structure. This has major per- formance wins, and also should work for future versions of the O/S.
Priorities are shown the same as they exist in process data structures, ranging from 0 to 255. Note that this is not the same as the ps(1) "pri" column, which subtracts 84 from each number before displaying it. Priority numbers fall in to priority classes as follows: 0 - 63 Interrupt threads 64 - 127 Top half kernel threads 128 - 159 Realtime user threads 160 - 223 Time sharing user threads 224 - 255 Idle user threads
Starting with FreeBSD 8.0 the display of individual threads can be tog- gled with the synonymous commands t and H. Information about state, flags, CPU time and percent cpu are shown for each individual thread. Other information is identical for all threads in the same process.
FreeBSD ALTERNATE DISPLAY
FreeBSD supports an alternate process display which shows i/o informa- tion. Since this information is tracked per process and not per thread, the per-thread display is not supported in this mode. All fields calculate the number of operations observed since the last update and are displayed as a per-second rate. The fields in this dis- play are as follows: VCSW Voluntary context switches IVCSW Involuntary context switches READ Number of blocks read WRITE Number of blocks written FAULT Number of page faults TOTAL Total number of i/o operations PERCENT Percentage of total i/o attributed to this process. If no i/o occured then this field is 0 for all processes.
FreeBSD KERNEL SUMMARY
All rates are shown per-second. Ctx Number of context switches. Trap Number of kernel traps. Intr Number of device interrupts. Soft Number of software interrupts. Fork Number of forks, vforks, and rforks. Flt Total number of page faults. Pgin Number of pages paged or swapped in to physical memory. Pgout Number of pages paged or swapped out from physical memory. Fr Total number of pages freed.
FreeBSD MEMORY SUMMARY
Memory: 10M Act 1208K Inact 3220K Wired 132K Free 25% Swap, 2924Kin 2604Kout K: Kilobyte M: Megabyte G: Gigabyte %: 1/100 Act: number of pages active Inact: number of pages inactive Wired: number of pages wired down Free: number of pages free Swap: swap usage Kin: kilobytes swap pager pages paged in (last interval) Kout: kilobytes swap pager pages paged out (last interval) See /usr/include/sys/vmmeter.h and /sys/vm/vm_meter.c. Contributors: Christos Zoulas, Steven Wallace, Wolfram Schneider, Monte Mitzelfelt. This module was retrofitted from FreeBSD 4.6.2 sources.
HPUX 9 INFORMATION
Under HP/UX 9, the kernel symbol _ccpu was eliminated. The author believe that _cexp is a suitable substitute, but cannot be positive. This seems to be confirmed by the fact that information produced using this assumption correlates well with that produced by HP's version of top. This port was adapted from the port for HP/UX version 8 (written by Christos Zoulas). The adaptation was performed by Kevin Schmidt <firstname.lastname@example.org>.
HPUX 10 INFORMATION
The process information layout has changed slightly since previous ver- sions. The CPU percentage column reports weighted cpu as calculated directly by the kernel. The WCPU column is no longer present in the output and a TTY column has been added to indicate the name of the process's controlling terminal. The definition of an idle process has been relaxed to include those processes that have only just gone to sleep. This version of top does not display a per-cpu breakdown of processor state. Perhaps a later version will add this sophistication across all platforms. The HP/UX 10 port has greatly benefitted from the diligent efforts of the following individuals: John Haxby <email@example.com>, Rich Hol- land <firstname.lastname@example.org>, and <email@example.com>.
The Linux port was written by Richard Henderson <firstname.lastname@example.org>. The CPU% calculation was brazenly stolen from the Solaris 2 port and should be attributed to one of the many names listed in its man page. The order support was stolen from SUNOS 5 port by Alexey Klimkin <email@example.com> Made to work under 2.4 by William LeFebvre.
This module has been tested on NetBSD 1.6, NetBSD 2.0 and NetBSD 3.0. It should also work on NetBSD 1.5, and probably any newer releases of NetBSD with little or no changes.
SUNOS 4 DIFFERENCES
On multiprocessor machines, the amount of time the processors spend in a spin lock is displayed along with the other processor state percent- ages. The percentages shown for processor states are averages across all processors. A process in run state also has its current processor displayed in the STATE column, for example "run/2" indicates running on processor 2. There is an extra column in the process display indicat- ing which processor each running process is assigned to. Information about physical memory is displayed on the memory status line, but information about virtual memory is not available. Due to incompatabilities in kernel data structures, a top executable compiled on a Sun 4 multiprocessor architecture machine (sun4m) will not run correctly on a uniprocessor architecture machine (sun4), and vice versa. You will have to compile and maintain separate executables for these architectures. Yeah, I don't like it either. Some processes may show up with a resident set size (RES column) larger than total virtual memory size (SIZE column). This seems odd at first, but is a consequence of shared libraries: shared memory is counted as resident but is not counted in total size. The SunOS 4 port was written by William LeFebvre.
SUNOS 5 NOTES
CPU percentage is calculated as a fraction of total available computing resources. Hence on a multiprocessor machine a single threaded process can never consume cpu time in excess of 1 divided by the number of pro- cessors. For example, on a 4 processor machine, a single threaded process will never show a cpu percentage higher than 25%. The CPU per- centage column will always total approximately 100, regardless of the number of processors. The kernel summary line shows the following information, all displayed as a per-second rate: ctxsw Context switches. trap Number of traps. intr Number of interrupts. syscall Number of system calls. fork Number of forks and vforks. flt Number of page faults. pgin Number of kilobytes paged in to physical memory. pgout Number of kilobytes paged out from physical memory. The memory summary line displays the following: phys mem Total amount of physical memory that can be allocated for use by processes (it does not include memory reserved for the kernel's use). free mem The amount of unallocated physical memory. total swap The total amount of swap area allocated on disk. free swap The amount of swap area on disk that is still available. Unlike previous versions of top, the swap figures will differ from the summary output of swap(1M) since the latter includes physical memory as well. The column NLWP indicates the number of lightweight processes in a process. This usually corresponds to the number of threads in that process. The display of individual threads can be toggled with the synonymous commands t and H. Information about state, priority, CPU time and per- cent CPU are shown for each individual thread. Other information is identical for all threads in the same process. In this display the column LWP replaces NLWP and shows the lightweight process id. The column names LWP and NLWP are consistent with ps(1). In BSD Unix, process priority was represented internally as a signed offset from a zero value with an unsigned value. The "zero" value was usually something like 20, allowing for a range of priorities from -20 to 20. As implemented on SunOS 5, older versions of top continued to interpret process priority in this manner, even though it was no longer correct. Starting with top version 3.5, this was changed to agree with the rest of the system. Long options are not currently available in Solaris. The SunOS 5 (Solaris 2) port was originally written by Torsten Kasch, <firstname.lastname@example.org>. Many contributions have been pro- vided by Casper Dik <Casper.Dik@sun.com>. Support for multi-cpu, cal- culation of CPU% and memory stats provided by Robert Boucher <email@example.com>, Marc Cohen <firstname.lastname@example.org>, Charles Hedrick <email@example.com>, and William L. Jones <jones@chpc>.
The SVR4 port was initially written by Andrew Herbert. He was guided by a SVR4 port of top version 2.1 which was done by Andy Crump (firstname.lastname@example.org). Robert Boucher (email@example.com) adapted it to top version 3.1. Ported to System 3000 Release 2.03 by Jeff Janvrin (jeff.janvrinColumbiaSC.NCR.COM)
The SVR5 port was generated by Mike Hopkirk from the SVR42 port by David Cutter with lots of help from Kurt Gollhardt and Doug Souders
Copyright (C) 1984-2007 William LeFebvre. For additional licensing information, see http://www.unixtop.org/license/ 4th Berkeley Distribution Local TOP(1)