READPROFILE(1)
NAME
readprofile - a tool to read kernel profiling information
SYNOPSIS
readprofile [ options ]
VERSION
This manpage documents version 2.0 of the program.
DESCRIPTION
The readprofile command uses the /proc/profile information
to print ascii data on standard output. The output is
organized in three columns: the first is the number of
clock ticks, the second is the name of the C function in
the kernel where those many ticks occurred, and the third
is the normalized `load' of the procedure, calculated as a
ratio between the number of thicks and the lenght of the
procedure. The output is filled with blanks to ease read-
ability.
Available command line options are the following:
-m mapfile
Specify a mapfile, which by default is
/usr/src/linux/System.map. You should specify the
map file on cmdline if your current kernel isn't
the last one you compiled. If the name of the map
file ends with `.gz' it is decompressed on the fly.
-p pro-file
Specify a different profiling buffer, which by
default is /proc/profile. Using a different pro-
file is useful if you want to `freeze' the kernel
profiling at some time and read it later. The
/proc/profile file can be copied using `cat' or
`cp'. There is no more support for compressed pro-
file buffers, like in readprofile-1.1, because the
program needs to know the size of the buffer in
advance.
-i Info. This makes readprofile only print the profil-
ing step used by the kernel. The profiling step is
the resolution of the profiling buffer, and is cho-
sen during kernel configuration (through `make con-
fig'), or in the kernel's command line. If the -t
(terse) switch is used together with -i only the
decimal number is printed.
-a Print all symbols in the mapfile. By default the
procedures with 0 reported ticks are not printed.
-r Reset the profiling buffer. This can only be
invoked by root, because /proc/profile is readable
by everybody but writable only by the superuser.
However, you can make readprofile setuid 0, in
order to reset the buffer without gaining privi-
leges.
-v Verbose. The output is organized in four columns
and filled with blanks. The first column is the
RAM address of a kernel function, the second is the
name of the function, the third is the number of
clock ticks and the last is the normalized load.
-V Version. This makes readprofile print its version
number and exit.
EXAMPLES
Browse the profiling buffer ordering by clock ticks:
readprofile | sort -nr | less
Print the 20 most loaded procedures:
readprofile | sort -nr +2 | head -20
Print only filesystem profile:
readprofile | grep _ext2
Look at all the kernel information, with ram addresses"
readprofile -av | less
Browse a `freezed' profile buffer for a non current ker-
nel:
readprofile -p ~/profile.freeze -m /zImage.map.gz
BUGS
readprofile only works with an 1.3.x or newer kernel,
because /proc/profile changed in the step from 1.2 to 1.3
This program only works with ELF kernels. The change for
a.out kernels is trivial, and left as an exercise to the
a.out user.
To enable profiling, the kernel must be rebooted, because
no profiling module is available, and it wouldn't be easy
to build. To enable profiling, you can specify "profile=2"
(or another number) on the kernel commandline. The number
you specify is the two-exponent used as profiling step.
Profiling is disabled when interrupts are inhibited. This
means that many profiling ticks happen when interrupts are
re-enabled. Watch out for misleading information.
FILES
/proc/profile A binary snapshot of the profiling buffer.
/usr/src/linux/System.map The symbol table for the kernel.
/usr/src/linux/* The program being profiled :-)