[Devel] Re: [PATCH 6/6] Explain reason for task being uncheckpointable

[Serge E. Hallyn]

Quoting Sukadev Bhattiprolu (sukadev at linux.vnet.ibm.com):
> 
> From: Sukadev Bhattiprolu <sukadev at linux.vnet.ibm.com>
> Date: Sat, 14 Mar 2009 10:21:07 -0700
> Subject: [PATCH 6/6] Explain reason for task being uncheckpointable
> 
> Try to give an useful message on why a task is uncheckpointable.

As I said somewhere else, I think this is what makes it worth
having a procfile (or debugfs file) over just doing a trial
sys_checkpoint() with an invalid fd.

However,

> Signed-off-by: Sukadev Bhattiprolu <sukadev at linux.vnet.ibm.com>
> ---
>  checkpoint/checkpoint.c    |   13 +++++++++----
>  fs/proc/base.c             |   19 ++++++++++++++++++-
>  include/linux/checkpoint.h |    7 +++++++
>  3 files changed, 34 insertions(+), 5 deletions(-)
> 
> diff --git a/checkpoint/checkpoint.c b/checkpoint/checkpoint.c
> index ad0956c..c8c377d 100644
> --- a/checkpoint/checkpoint.c
> +++ b/checkpoint/checkpoint.c
> @@ -264,18 +264,23 @@ static int cr_write_all_tasks(struct cr_ctx *ctx)
>  	return ret;
>  }
> 
> -int task_checkpointable(struct task_struct *t)
> +int __task_checkpointable(struct task_struct *t)
>  {
>  	if (t->state == TASK_DEAD) {
>  		pr_warning("c/r: task %d is TASK_DEAD\n", task_pid_vnr(t));
> -		return 0;
> +		return CR_DEAD;
>  	}
> 
>  	/* Verify that task is frozen, unless it is self-checkpoint */
>  	if (t != current && !frozen(t))
> -		return 0;
> +		return CR_NOT_FROZEN;
> +
> +	return CR_CHECKPOINTABLE;
> +}
> 
> -	return 1;
> +int task_checkpointable(struct task_struct *t)
> +{
> +	return __task_checkpointable(t) == CR_CHECKPOINTABLE;
>  }
> 
>  static int cr_may_checkpoint_task(struct task_struct *t, struct cr_ctx *ctx)
> diff --git a/fs/proc/base.c b/fs/proc/base.c
> index 989ca93..6e4c07c 100644
> --- a/fs/proc/base.c
> +++ b/fs/proc/base.c
> @@ -2502,9 +2502,26 @@ static int proc_pid_personality(struct seq_file *m, struct pid_namespace *ns,
>  }
> 
>  #ifdef CONFIG_CHECKPOINT_RESTART
> +static char *explain_checkpointable(int reason)
> +{
> +	char *message;
> +
> +	switch(reason) {
> +	case CR_CHECKPOINTABLE: message = "1"; break;
> +	case CR_DEAD: 		message = "0 (dead)"; break;
> +	case CR_NOT_FROZEN: 	message = "0 (not frozen)"; break;
> +	default:		message = "0 (unknown)"; break;

I don't think splitting it up this way is going to work long-term.
It means you need one code for each type of message, and can't give
any more detail beyond that (i.e. the value of
current->nsproxy->uts_ns.refcount, or whatever).

I think it would be better to pass a buffer into __task_checkpointable,
and let task_checkpointable() pass in a NULL buffer meaning "I
(sys_checkpoint) don't care about the explanation".

> +	}
> +	return message;
> +}
> +
>  static int proc_pid_checkpointable(struct task_struct *task, char *buffer)
>  {
> -	return sprintf(buffer, "%d\n", task_checkpointable(task));
> +	int rc;
> +
> +	rc = __task_checkpointable(task);
> +
> +	return scnprintf(buffer, PAGE_SIZE, "%s\n", explain_checkpointable(rc));
>  }
>  #endif
> 
> diff --git a/include/linux/checkpoint.h b/include/linux/checkpoint.h
> index d1f53a6..a1eba73 100644
> --- a/include/linux/checkpoint.h
> +++ b/include/linux/checkpoint.h
> @@ -142,6 +142,13 @@ static inline int cr_enabled(void)
>  	return 1;
>  }
> 
> +enum cr_uncheckpointable_reason {
> +	CR_CHECKPOINTABLE = 0,
> +	CR_DEAD,
> +	CR_NOT_FROZEN
> +};
> +
> +int __task_checkpointable(struct task_struct *t);
>  int task_checkpointable(struct task_struct *t);
> 
>  #else /* !CONFIG_CHECKPOINT_RESTART */
> -- 
> 1.5.2.5
> 
> _______________________________________________
> Containers mailing list
> Containers at lists.linux-foundation.org
> https://lists.linux-foundation.org/mailman/listinfo/containers
_______________________________________________
Containers mailing list
Containers at lists.linux-foundation.org
https://lists.linux-foundation.org/mailman/listinfo/containers