Diff for /wikisrc/tutorials/how_to_enable_and_run_dtrace.mdwn between versions 1.1 and 1.28

version 1.1, 2011/11/20 20:55:21 version 1.28, 2020/03/30 15:40:27
Line 1 Line 1
 DTrace is a Dynamic Tracing framework developed by Sun and ported to NetBSD. It enables extensive instrumentation of the kernel and user space. See the [DTrace Community Page](http://hub.opensolaris.org/bin/view/Community+Group+dtrace/WebHome) for more information. Also see [DTrace Introduction](http://wikis.sun.com/display/DTrace/Introduction).  DTrace is a Dynamic Tracing framework developed by Sun and ported to NetBSD. It enables extensive instrumentation of the kernel and user space. See the [DTrace Community Page](http://dtrace.org) for more information.
   Also see [DTrace Introduction](http://dtrace.org/guide/preface.html), Brendan Gregg's [DTrace one liners](http://www.brendangregg.com/DTrace/dtrace_oneliners.txt) and his notes for [DTrace on FreeBSD](https://wiki.freebsd.org/DTrace/).
   
 DTrace is a work-in-progress effort and it is for i386 systems only. Two providers are available; the Statically Defined Tracing (SDT) provider and the Function Boundary Tracer (FBT) provider.  # Current status
   
 You can currently run a hello world DScript.  ## Supported platforms
   
 ##  Building DTrace  DTrace is a work-in-progress effort and it is for x86 systems and some arm boards.
   
 You need the following options in your kernel:  * i386 and amd64
      * earm* (evbarm and armv4 based ports)
     options         INSECURE  
     options         KDTRACE_HOOKS   # DTrace support  ## Supported providers
     options         MODULAR  
      * DTrace: What to do when a script BEGINs, ENDs, ERRORs
   * FBT: Function Boundary Tracing
 You also need to build distribution with the options MKMODULAR=yes and MKDTRACE=yes.  * IO: Disk I/O
   * Lockstat: Kernel Lock Statistics
 ##  Running hello world  * Proc: Process and thread related events
   * Profile: Time based interrupt event source for Profiling
 Load the solaris and dtrace modules, and the SDT (Statically Defined Tracing) and FBT (Function Boundary Tracing) modules:  * SDT: Statically Defined Tracing
      * Syscall: System Calls
     modload solaris  * Syscall Linux (32bit & 64 bit): System calls via the Linux binary emulation layer
     modload dtrace  * VFS: Filesystem operations (confined to namecache events at time of writing - 8.99.22)
     modload sdt  
     modload fbt  ## TODO for netbsd-7
      
   * Measure effect of `options KDTRACE_HOOKS` on system performance.
 Make the dtrace device node:  * Determine whether the profile module works and list it here.
      * Integrate [[riz|users/riz]]'s syscall provider patch.
     mkdir /dev/dtrace  
     mknod /dev/dtrace/dtrace c dtrace 0  # How to use
      
   ##  Building DTrace 
 List the dtrace probes  
      You need the following options in your kernel: 
     dtrace -l      
          options         KDTRACE_HOOKS   # kernel DTrace hooks
        ID   PROVIDER            MODULE                          FUNCTION NAME      options         MODULAR
         1     dtrace                                                     BEGIN  
         2     dtrace                                                     END  Optionally:
         3     dtrace                                                     ERROR  
         4        fbt            netbsd             AcpiAcquireGlobalLock entry      options         INSECURE   # permit modules to loaded from user space once system has gone multiuser and securelevel has been raised.
         5        fbt            netbsd             AcpiAcquireGlobalLock return  
         6        fbt            netbsd             AcpiAllocateRootTable entry  A Distribution needs to be built with the options `MKDTRACE=yes` and `MKCTF=yes`, this is taken care of automatically and doesn't need to be specified manually. The list of platforms it is applied to automatically is set in `src/share/mk/bsd.own.mk`
         7        fbt            netbsd                    AcpiAttachData entry  
         .  Set the system to load the solaris and dtrace related modules in `/etc/modules.conf`, for a list of available modules, see `/stand/$MACHINE/$VERSION/modules/`
         .  
     29129        fbt           solaris                   zfs_vop_getattr entry  For example, add the following to `/etc/modules.conf` (the file may not exist already on a system):
     29130        fbt           solaris                   zfs_vop_getattr return      
     29131       proc                                                     create  - `solaris`
     29132       proc                                                     exec  - `dtrace`
         .  - `dtrace_fbt`
         .  - `dtrace_lockstat`
     29140       proc                                                     lwp_start  - `dtrace_profile`
     29141       proc                                                     lwp_exit  - `dtrace_sdt`
      - `dtrace_syscall`
   - `dtrace_syscall_linux`
        
   A `dtrace` device node is created automatically in `/dev/dtrace` when the modules are loaded into place.
       
 Put the following into the file hello.d  List the dtrace probes 
          
     BEGIN      dtrace -l
     {      
         trace("Hello world");         ID   PROVIDER            MODULE                          FUNCTION NAME
         exit(0);          1     dtrace                                                     BEGIN
     }          2     dtrace                                                     END
              3     dtrace                                                     ERROR
           4        fbt            netbsd             AcpiAcquireGlobalLock entry
 Run the hello world script:          5        fbt            netbsd             AcpiAcquireGlobalLock return
              6        fbt            netbsd             AcpiAllocateRootTable entry
     dtrace -s hello.d          7        fbt            netbsd                    AcpiAttachData entry
              . 
     dtrace: script './hello.d' matched 1 probe          .
     CPU     ID                    FUNCTION:NAME      29129        fbt           solaris                   zfs_vop_getattr entry 
       0      1                           :BEGIN   Hello world      29130        fbt           solaris                   zfs_vop_getattr return
          29131       proc                                                     create
       29132       proc                                                     exec 
 A more complex example that traces the execution of a sleep operation in the kernel:          .
              .
     #pragma D option flowindent      29140       proc                                                     lwp_start
          29141       proc                                                     lwp_exit
     fbt::syscall:entry  
     /execname == "sleep" && guard++ == 0/  
     {  ## Running hello world 
             self->traceme = 1;  
             printf("fd: %d", arg0);  Put the following into the file hello.d:
     }      
          BEGIN
     fbt:::      {
     /self->traceme/          trace("Hello world");
     {}          exit(0);
          }
     fbt::syscall:return      
     /self->traceme/  
     {  Run the hello world script: 
             self->traceme = 0;      
             exit(0);      dtrace -s hello.d
     }      
          dtrace: script './hello.d' matched 1 probe
       CPU     ID                    FUNCTION:NAME
 Start the script running (dtrace -s <scriptname.d>) and then execute a sleep 2 in another shell.        0      1                           :BEGIN   Hello world
   
   The same script could be executed as a one liner on the shell, using
   
       dtrace -n 'BEGIN { trace("Hello world"); exit(0); }'
   
   ## A more complex example
   
   The following script traces the execution of a sleep operation
   in the kernel. Put it in sleep.d:
       
       #pragma D option flowindent
   
       syscall::nanosleep:entry
       /execname == "sleep" && guard++ == 0/
       {
               self->traceme = 1;
       }
   
       fbt:::
       /self->traceme/
       {}
   
       syscall::nanosleep:return
       /self->traceme/
       {
               self->traceme = 0;
               exit(0);
       }
   
   Start the script running:
   
       dtrace -s sleep.d
   
   This will take a while as the script instruments every function in the
   kernel. When it's ready, it will print a message like "dtrace: script
   'sleep.d' matched 59268 probes".  Then execute a "sleep 2" in another
   shell.
   
   ## Tools included in base
   
   Starting with NetBSD-8, on builds where `MKDTRACE=yes` is set, scripts from
   [Brendan Gregg's DTrace toolkit](https://github.com/opendtrace/toolkit/) are installed in base as standard.
   
   At present, the following scripts are installed in `/usr/sbin`: 
   
   - `dtruss` - An implementation of the truss utility in DTrace which traces the system calls
   made by a process
   - `execsnoop` - snoop on execution of processes as they occur
   - `opensnoop` - snoop on openning of files as they occur
   - `procsystime` -  print process system call time details.
   
   ## Troubleshooting
   
   The Compact C Type Format (CTF) has a 2^15 limit on types which can overflow, this prevents DTrace from
   working correctly.
   
   Check the number of types using `ctfdump` e.g
       ctfdump -S /netbsd
   
   Note the line which states `total number of types`, the value should by less than 32768.
   
   If overflow is not an issue, `libdtrace(3)` can provide some insight into what is going on via an
   environment variable. Define `DTRACE_DEBUG` before tracing.
   
        env DTRACE_DEBUG= execsnoop
   

Removed from v.1.1  
changed lines
  Added in v.1.28


CVSweb for NetBSD wikisrc <wikimaster@NetBSD.org> software: FreeBSD-CVSweb