Re: [PATCH 1/2] tracing: Add a trace for task_exit

[Eric W. Biederman]

Peter Enderborg <peter.enderborg@xxxxxxxx> writes:

> This is the peer functions to task_rename and task_newtask.
> With this we get hole "life-cycle" of task and can easily
> see short livied task and their exit status.

This patch is incorrect.  The location you are dealing with is not part
of task exit.  The location you have instrumented is part of reaping a
task which can come arbitrarily long after the task exits.

There are some special rules associated with task_comm so I don't know
if your change to __string from a fixed size character array is safe.

Certainly something like that needs an explanation of why such a type
change is safe.

Eric


> Format might look like:
>             bash-1144    [006] ....  1306.601707: task_newtask: pid=1181 comm=bash clone_flags=1200000 oom_score_adj=0
>            <...>-1181    [007] ....  1306.602080: task_rename: pid=1181 oldcomm=bash newcomm=ls oom_score_adj=0
>             bash-1144    [006] d...  1306.785960: task_exit: pid=1181 oom_score_adj=0 exit_signal=17 exit_code=0 exit_state=0x10 comm=ls
>
> For a sequence when a bash shell runs the ls command.
>
> Signed-off-by: Peter Enderborg <peter.enderborg@xxxxxxxx>
> ---
>  include/trace/events/task.h | 32 ++++++++++++++++++++++++++++++++
>  kernel/exit.c               |  3 +++
>  2 files changed, 35 insertions(+)
>
> diff --git a/include/trace/events/task.h b/include/trace/events/task.h
> index 64d160930b0d..2e977d2935e1 100644
> --- a/include/trace/events/task.h
> +++ b/include/trace/events/task.h
> @@ -56,6 +56,38 @@ TRACE_EVENT(task_rename,
>  		__entry->newcomm, __entry->oom_score_adj)
>  );
>  
> +TRACE_EVENT(task_exit,
> +
> +	TP_PROTO(struct task_struct *task),
> +
> +	TP_ARGS(task),
> +
> +	TP_STRUCT__entry(
> +		__field(pid_t,	pid)
> +		__field(short,	oom_score_adj)
> +		__field(int,	exit_signal)
> +		__field(int,	exit_code)
> +		__field(int,	exit_state)
> +		__string(comm, task->comm)
> +
> +	),
> +
> +	TP_fast_assign(
> +		__entry->pid = task->pid;
> +		__entry->oom_score_adj = task->signal->oom_score_adj;
> +		__entry->exit_signal = task->exit_signal;
> +		__entry->exit_code = task->exit_code;
> +		__entry->exit_state = task->exit_state;
> +		__assign_str(comm, task->comm);
> +	),
> +
> +	TP_printk("pid=%d oom_score_adj=%hd exit_signal=%d exit_code=%d exit_state=0x%x comm=%s",
> +		  __entry->pid,
> +		  __entry->oom_score_adj, __entry->exit_signal,
> +		  __entry->exit_code, __entry->exit_state,
> +		  __get_str(comm))
> +);
> +
>  #endif
>  
>  /* This part must be outside protection */
> diff --git a/kernel/exit.c b/kernel/exit.c
> index 04029e35e69a..3ab0944e5dfc 100644
> --- a/kernel/exit.c
> +++ b/kernel/exit.c
> @@ -68,6 +68,7 @@
>  #include <linux/uaccess.h>
>  #include <asm/unistd.h>
>  #include <asm/mmu_context.h>
> +#include <trace/events/task.h>
>  
>  static void __unhash_process(struct task_struct *p, bool group_dead)
>  {
> @@ -107,6 +108,8 @@ static void __exit_signal(struct task_struct *tsk)
>  		posix_cpu_timers_exit_group(tsk);
>  #endif
>  
> +	trace_task_exit(tsk);
> +
>  	if (group_dead) {
>  		tty = sig->tty;
>  		sig->tty = NULL;