17.2. Kernel Utilities

17.1. Diagnostic Utilities

Mac OS X includes many diagnostic utilities that you can use to monitor your system and investigate problems.

17.1.1. top

The top utility displays memory statistics and a list of running processes. It is divided into two regions: the top region contains memory statistics, and the bottom region contains details on each process.

You can specify the number of processes to show by supplying a numeric argument. By default, top refreshes its display every second and sorts the list of processes by process ID (PID) in descending order. You can set top to sort by CPU utilization with -u, and you can specify the refresh delay with the -s option. Figure 17-1 shows the output of top -u 16 (if you wanted to refresh the output every 3 seconds, you could run top -s3 -u 16).

Table 17-1 describes the values shown in the top region, and Table 17-2 describes the columns in the bottom region (process information).

Figure 17-1. Sample output from top

Table 17-1. Memory information displayed by top

Item number	Item	Description
1	Processes	Number of processes and threads. A running process is currently using CPU time, while a sleeping process is not.
2	Load Avg.	Average system load (the number of jobs vying for the CPU's attention) over the last 1, 5, and 15 minutes.
3	CPU usage	Breakdown of CPU usage, listing time spent in user mode, kernel (sys) mode, and idle time.
4	SharedLibs	Number of shared libraries in use, along with their memory utilization.
5	MemRegions	Number of Mach virtual memory regions in use, along with memory utilization details.
6	PhysMem	Physical memory utilization. Memory that is wired cannot be swapped to disk. active memory is memory that's currently being used, inactive memory is memory that Mac OS X is keeping "on deck" for processes that need it, and free memory is memory that's not being used at all.
7	VM	Virtual memory statistics, including the total amount of virtual memory allocated (the sum of the VSIZE in the process list), as well as paging activity (data paged in and out of physical memory).

Table 17-2. Process information displayed by top

Item number	Item	Description
8	PID	Process ID
9	COMMAND	Program's name
10	%CPU	Percentage of the CPU that the process is using
11	TIME	Total amount of CPU time this process has used
12	#TH	Number of threads in this process
13	#PRTS	Number of Mach ports
14	#MREGS	Number of memory registers
15	RPRVT	Resident private memory
16	RSHRD	Resident shared memory
17	RSIZE	Resident memory
18	VSIZE	Process's total address space, including shared memory

17.1.2. fs_usage

The fs_usage utility shows a continuous display of filesystem-related system calls and page faults. You must run fs_usage as root. By default, it ignores anything originating from fs_usage, Terminal, telnetd, sshd, rlogind, tcsh, csh, or sh.

Figure 17-2 shows the output of fs_usage monitoring the startup of a new bash shell, and displaying the following columns:

• Timestamp

• System call

• Filename

• Elapsed time

• Name of the process

17.1.3. latency

latency measures the number of context switches and interrupts, and reports on the resulting delays, updating the display once per second. This utility must be run as root. Example 17-1 shows a portion of its output.

Figure 17-2. Monitoring filesystem operations with fs_usage

Example 17-1. Partial output from latency

     Sat Apr  2 11:07:35                             0:04:18                          SCHEDULER     INTERRUPTS     ---------------------------------------------     total_samples           209208         336535           delays <  10 usecs      183575         332310     delays <  20 usecs       11999           3197     delays <  30 usecs        2416            836     delays <  40 usecs        1289            158     delays <  50 usecs         624             25     delays <  60 usecs         492              6     delays <  70 usecs         323              2     delays <  80 usecs         329              1     delays <  90 usecs         280              0     delays < 100 usecs         235              0     total  < 100 usecs      201562         336535 

The SCHEDULER column lists the number of context switches and the INTERRUPTS column lists the number of interrupts.

17.1.4. sc_usage

The sc_usage utility samples system calls and page faults, then displays them onscreen. sc_usage must be run by root or by someone with superuser privileges. The display is updated once per second. You must specify a PID, a command name, or a program to execute with the -E switch. For example, to monitor the Finder, use sudo sc_usage Finder. Figure 17-3 shows the output of running sc_usage on the Finder. Table 17-3 explains sc_usage's output.

Figure 17-3. sc_usage monitoring the Finder

Table 17-3. Information displayed by sc_usage

Item number	Row	Description
1	TYPE	System call type
2	NUMBER	System call count
3	CPU_TIME	Processor time used by the system call
4	WAIT_TIME	Absolute time that the process spent waiting
5	CURRENT_TYPE	Current system call type
6	LAST_PATHNAME_WAITED_FOR	Last file or directory that resulted in a blocked I/O operation during a system call
7	CUR_WAIT_TIME	Cumulative time spent blocked
8	THRD#	Thread ID
9	PRI	Scheduling priority

17.1.5. vm_stat

The vm_stat utility displays virtual memory statistics. Unlike implementations of vm_stat in other Unix systems, it does not default to continuous display. Instead, it displays accumulated statistics.

To obtain a continuous display, specify an interval argument (in seconds), as in vm_stat 1. Figure 17-4 shows the output of vm_stat with no arguments.

Figure 17-4. vm_stat displaying accumulated statistics

Figure 17-5 shows the output of vm_stat 1. Table 17-4 describes the information that vm_stat displays (the item numbers correspond to the callouts in both figures).

Figure 17-5. vm_stat's continuous output

Table 17-4. Information displayed by vm_stat

Item number	Accumulated mode	Continuous mode	Description
1	Pages free	free	Total free pages
2	Pages active	active	Total pages in use that can be paged out
3	Pages inactive	inac	Total inactive pages
4	Pages wired down	wire	Total pages wired into memory (cannot be paged out)
5	translation Faults	faults	Number of times vm_fault has been called
6	Pages copy-on-write	copy	Number of faults that resulted in a page being copied
7	Pages zero filled	zerofill	Number of pages that have been zero-filled
8	Pages Reactivated	reactive	Number of pages reclassified from inactive to active
9	Pageins	pagein	Number of pages moved into physical memory
10	Pageouts	pageout	Number of pages moved out of physical memory