sys-utils/readprofile.8

   1 .TH READPROFILE "8" "October 2011" "util-linux" "System Administration"
   2 .SH NAME
   3 readprofile \- read kernel profiling information
   4 .SH SYNOPSIS
   5 .B readprofile
   6 [options]
   7 .SH VERSION
   8 This manpage documents version 2.0 of the program.
   9 .SH DESCRIPTION
  10 .LP
  11 The
  12 .B readprofile
  13 command uses the
  14 .I /proc/profile
  15 information to print ascii data on standard output.  The output is
  16 organized in three columns: the first is the number of clock ticks,
  17 the second is the name of the C function in the kernel where those
  18 many ticks occurred, and the third is the normalized `load' of the
  19 procedure, calculated as a ratio between the number of ticks and the
  20 length of the procedure.  The output is filled with blanks to ease
  21 readability.
  22 .SH OPTIONS
  23 .TP
  24 \fB\-a\fR, \fB\-\-all\fR
  25 Print all symbols in the mapfile.  By default the procedures with
  26 reported ticks are not printed.
  27 .TP
  28 \fB\-b\fR, \fB\-\-histbin\fR
  29 Print individual histogram-bin counts.
  30 .TP
  31 \fB\-i\fR, \fB\-\-info\fR
  32 Info.  This makes
  33 .B readprofile
  34 only print the profiling step used by the kernel.  The profiling step
  35 is the resolution of the profiling buffer, and is chosen during
  36 kernel configuration (through `make config'), or in the kernel's
  37 command line.  If the
  38 .B \-t
  39 (terse) switch is used together with
  40 .B \-i
  41 only the decimal number is printed.
  42 .TP
  43 \fB\-m\fR, \fB\-\-mapfile\fR \fImapfile\fR
  44 Specify a mapfile, which by default is
  45 .IR /usr/src/linux/System.map .
  46 You should specify the map file on cmdline if your current kernel
  47 isn't the last one you compiled, or if you keep System.map elsewhere.
  48 If the name of the map file ends with `.gz' it is decompressed on the
  49 fly.
  50 .TP
  51 \fB\-M\fR, \fB\-\-multiplier\fR \fImultiplier\fR
  52 On some architectures it is possible to alter the frequency at which
  53 the kernel delivers profiling interrupts to each CPU.  This option
  54 allows you to set the frequency, as a multiplier of the system clock
  55 frequency, HZ. Linux 2.6.16 dropped multiplier support for most systems.
  56 This option also resets the profiling buffer, and requires superuser
  57 privileges.
  58 .TP
  59 \fB\-p\fR, \fB\-\-profile\fR \fIpro-file\fR
  60 Specify a different profiling buffer, which by default is
  61 .IR /proc/profile .
  62 Using a different pro-file is useful if you want to `freeze' the
  63 kernel profiling at some time and read it later.  The
  64 .I /proc/profile
  65 file can be copied using `cat' or `cp'.  There is no more support for
  66 compressed profile buffers, like in
  67 .B readprofile-1.1,
  68 because the program needs to know the size of the buffer in advance.
  69 .TP
  70 \fB\-r\fR, \fB\-\-reset\fR
  71 Reset the profiling buffer.  This can only be invoked by root,
  72 because
  73 .I /proc/profile
  74 is readable by everybody but writable only by the superuser.
  75 However, you can make
  76 .B readprofile
  77 set-user-ID 0, in order to reset the buffer without gaining privileges.
  78 .TP
  79 \fB\-s, \fB\-\-counters\fR
  80 Print individual counters within functions.
  81 .TP
  82 \fB\-v\fR, \fB\-\-verbose\fR
  83 Verbose.  The output is organized in four columns and filled with
  84 blanks.  The first column is the RAM address of a kernel function,
  85 the second is the name of the function, the third is the number of
  86 clock ticks and the last is the normalized load.
  87 .TP
  88 \fB\-V\fR, \fB\-\-version\fR
  89 Display version information and exit.
  90 .TP
  91 \fB\-h\fR, \fB\-\-help\fR
  92 Display help text and exit.
  93 .SH FILES
  94 .nf
  95 /proc/profile              A binary snapshot of the profiling buffer.
  96 /usr/src/linux/System.map  The symbol table for the kernel.
  97 /usr/src/linux/*           The program being profiled :-)
  98 .fi
  99 .SH BUGS
 100 .LP
 101 .B readprofile
 102 only works with a 1.3.x or newer kernel, because
 103 .I /proc/profile
 104 changed in the step from 1.2 to 1.3
 105 .LP
 106 This program only works with ELF kernels.  The change for a.out
 107 kernels is trivial, and left as an exercise to the a.out user.
 108 .LP
 109 To enable profiling, the kernel must be rebooted, because no
 110 profiling module is available, and it wouldn't be easy to build.  To
 111 enable profiling, you can specify "profile=2" (or another number) on
 112 the kernel commandline.  The number you specify is the two-exponent
 113 used as profiling step.
 114 .LP
 115 Profiling is disabled when interrupts are inhibited.  This means that
 116 many profiling ticks happen when interrupts are re-enabled.  Watch
 117 out for misleading information.
 118 .SH EXAMPLE
 119 Browse the profiling buffer ordering by clock ticks:
 120 .nf
 121    readprofile | sort \-nr | less
 122
 123 .fi
 124 Print the 20 most loaded procedures:
 125 .nf
 126    readprofile | sort \-nr +2 | head \-20
 127
 128 .fi
 129 Print only filesystem profile:
 130 .nf
 131    readprofile | grep _ext2
 132
 133 .fi
 134 Look at all the kernel information, with ram addresses:
 135 .nf
 136    readprofile \-av | less
 137
 138 .fi
 139 Browse a `frozen' profile buffer for a non current kernel:
 140 .nf
 141    readprofile \-p ~/profile.freeze \-m /zImage.map.gz
 142
 143 .fi
 144 Request profiling at 2kHz per CPU, and reset the profiling buffer:
 145 .nf
 146    sudo readprofile \-M 20
 147 .fi
 148 .SH AVAILABILITY
 149 The readprofile command is part of the util-linux package and is
 150 available from
 151 .UR https://\:www.kernel.org\:/pub\:/linux\:/utils\:/util-linux/
 152 Linux Kernel Archive
 153 .UE .