linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-1)
@ 2006-09-29 15:18 girish
  2006-09-29 16:31 ` William Pitcock
  0 siblings, 1 reply; 10+ messages in thread
From: girish @ 2006-09-29 15:18 UTC (permalink / raw)
  To: girishvg


Hello.

Could somebody please check if this is acceptable.

Thanks.

Signed-off-by: Girish V. Gulawani <girishvg@gmail.com>


--- linux-vanilla/fs/proc/array.c	2006-09-20 12:42:06.000000000 +0900
+++ linux/fs/proc/array.c	2006-09-30 00:16:59.000000000 +0900
@@ -248,6 +248,8 @@ static inline char * task_sig(struct tas
  	int num_threads = 0;
  	unsigned long qsize = 0;
  	unsigned long qlim = 0;
+	int num_children = 0;
+	struct list_head *_p;

  	sigemptyset(&pending);
  	sigemptyset(&shpending);
@@ -268,9 +270,11 @@ static inline char * task_sig(struct tas
  		qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
  		spin_unlock_irq(&p->sighand->siglock);
  	}
+	list_for_each(_p, &p->children)
+		++num_children;
  	read_unlock(&tasklist_lock);

-	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
+	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads +  
num_children);
  	buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);

  	/* render them all */

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-1)
  2006-09-29 15:18 [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-1) girish
@ 2006-09-29 16:31 ` William Pitcock
  2006-09-29 16:51   ` girish
  0 siblings, 1 reply; 10+ messages in thread
From: William Pitcock @ 2006-09-29 16:31 UTC (permalink / raw)
  To: girish; +Cc: linux-kernel


On Sep 29, 2006, at 10:18 AM, girish wrote:

>
> -	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
> +	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads +  
> num_children);

Personally, I'd prefer the children count to be separate, something  
like:

buffer += sprintf(buffer, "Threads:\t%d (%d children, %d total)",  
num_threads, num_children, num_threads + num_children);

That would be rather nice, indeed.

Also, next time, make sure that linux-kernel is CC'd, not BCC'd.

---
William Pitcock
nenolod@atheme.org
http://people.atheme.org/~nenolod/
http://nenolod.net



^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-1)
  2006-09-29 16:31 ` William Pitcock
@ 2006-09-29 16:51   ` girish
  2006-09-29 17:06     ` Jan Engelhardt
  0 siblings, 1 reply; 10+ messages in thread
From: girish @ 2006-09-29 16:51 UTC (permalink / raw)
  To: linux-kernel; +Cc: William Pitcock, girishvg


On Sep 30, 2006, at 1:31 AM, William Pitcock wrote:

>
> On Sep 29, 2006, at 10:18 AM, girish wrote:
>
>>
>> -	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
>> +	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads +  
>> num_children);
>
> Personally, I'd prefer the children count to be separate, something  
> like:
>
> buffer += sprintf(buffer, "Threads:\t%d (%d children, %d total)",  
> num_threads, num_children, num_threads + num_children);
>
> That would be rather nice, indeed.
>
> Also, next time, make sure that linux-kernel is CC'd, not BCC'd.
>
> ---
> William Pitcock
> nenolod@atheme.org
> http://people.atheme.org/~nenolod/
> http://nenolod.net
>
>

Agree.  It indeed look better. I too had an awk script in mind, to  
parse the line.  I ended up removing such formatting, because not all  
process spawn child thread(s), showing num_children count as zero.  
That looked bit odd. So here it is again - new wine.

Thanks.

Signed-off-by: Girish V. Gulawani <girishvg@gmail.com>
--- linux-vanilla/fs/proc/array.c	2006-09-20 12:42:06.000000000 +0900
+++ linux/fs/proc/array.c	2006-09-30 01:47:25.000000000 +0900
@@ -248,6 +248,8 @@ static inline char * task_sig(struct tas
  	int num_threads = 0;
  	unsigned long qsize = 0;
  	unsigned long qlim = 0;
+	int num_children = 0;
+	struct list_head *_p;

  	sigemptyset(&pending);
  	sigemptyset(&shpending);
@@ -268,9 +270,14 @@ static inline char * task_sig(struct tas
  		qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
  		spin_unlock_irq(&p->sighand->siglock);
  	}
+	list_for_each(_p, &p->children)
+		++num_children;
  	read_unlock(&tasklist_lock);

-	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
+	buffer += sprintf(buffer, "Threads:\t%d", num_threads);
+	if (num_children)
+		buffer += sprintf(buffer, " (%d children, %d total)",  
num_children, num_threads + num_children);
+	buffer += sprintf(buffer, "\n");
  	buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);

  	/* render them all */

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-1)
  2006-09-29 16:51   ` girish
@ 2006-09-29 17:06     ` Jan Engelhardt
  2006-09-29 17:18       ` girish
  0 siblings, 1 reply; 10+ messages in thread
From: Jan Engelhardt @ 2006-09-29 17:06 UTC (permalink / raw)
  To: girish; +Cc: linux-kernel, William Pitcock

>> > 
>> > -	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
>> > +	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads +
>> > num_children);
>> 
>> Personally, I'd prefer the children count to be separate, something like:
>> 
>> buffer += sprintf(buffer, "Threads:\t%d (%d children, %d total)",
>> num_threads, num_children, num_threads + num_children);
>> 
>> That would be rather nice, indeed.

And I would suggest three separate lines to keep it parseable!


Jan Engelhardt
-- 

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-1)
  2006-09-29 17:06     ` Jan Engelhardt
@ 2006-09-29 17:18       ` girish
  2006-09-29 18:12         ` Jan Engelhardt
  0 siblings, 1 reply; 10+ messages in thread
From: girish @ 2006-09-29 17:18 UTC (permalink / raw)
  To: Jan Engelhardt; +Cc: linux-kernel, William Pitcock


On Sep 30, 2006, at 2:06 AM, Jan Engelhardt wrote:

>>>>
>>>> -	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
>>>> +	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads +
>>>> num_children);
>>>
>>> Personally, I'd prefer the children count to be separate,  
>>> something like:
>>>
>>> buffer += sprintf(buffer, "Threads:\t%d (%d children, %d total)",
>>> num_threads, num_children, num_threads + num_children);
>>>
>>> That would be rather nice, indeed.
>
> And I would suggest three separate lines to keep it parseable!
>
>
> Jan Engelhardt
> -- 

How about this?

         buffer += sprintf(buffer, "Threads:\t%d", num_threads);
         if (num_children)
                 buffer += sprintf(buffer, " Children: %d Total: %d",  
num_children, num_threads + num_children);
         buffer += sprintf(buffer, "\n");


^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-1)
  2006-09-29 17:18       ` girish
@ 2006-09-29 18:12         ` Jan Engelhardt
  2006-09-29 18:24           ` [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-3) girish
  0 siblings, 1 reply; 10+ messages in thread
From: Jan Engelhardt @ 2006-09-29 18:12 UTC (permalink / raw)
  To: girish; +Cc: linux-kernel, William Pitcock

>
> How about this?
>
> buffer += sprintf(buffer, "Threads:\t%d", num_threads);
> if (num_children)
>                buffer += sprintf(buffer, " Children: %d Total: %d",
> num_children, num_threads + num_children);
> buffer += sprintf(buffer, "\n");
>

No, this:

> if (num_children)                                                             
>                buffer += sprintf(buffer, "\nChildren: %d\nTotal: %d",

the newlines are essential because then you get _one_ field of 
information for _each_ call of fgets().



Jan Engelhardt
-- 

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-3)
  2006-09-29 18:12         ` Jan Engelhardt
@ 2006-09-29 18:24           ` girish
  2006-09-29 22:51             ` Eric W. Biederman
  0 siblings, 1 reply; 10+ messages in thread
From: girish @ 2006-09-29 18:24 UTC (permalink / raw)
  To: Jan Engelhardt; +Cc: linux-kernel, William Pitcock


On Sep 30, 2006, at 3:12 AM, Jan Engelhardt wrote:

>>
>> How about this?
>>
>> buffer += sprintf(buffer, "Threads:\t%d", num_threads);
>> if (num_children)
>>                buffer += sprintf(buffer, " Children: %d Total: %d",
>> num_children, num_threads + num_children);
>> buffer += sprintf(buffer, "\n");
>>
>
> No, this:
>
>> if (num_children)
>>                buffer += sprintf(buffer, "\nChildren: %d\nTotal: %d",
>
> the newlines are essential because then you get _one_ field of
> information for _each_ call of fgets().
>
>
>
> Jan Engelhardt
> --  

This is perfect. Here is the patch.
Thanks.

Signed-off-by: Girish V. Gulawani <girishvg@gmail.com>

--- linux-vanilla/fs/proc/array.c	2006-09-20 12:42:06.000000000 +0900
+++ linux/fs/proc/array.c	2006-09-30 03:18:28.000000000 +0900
@@ -248,6 +248,8 @@ static inline char * task_sig(struct tas
  	int num_threads = 0;
  	unsigned long qsize = 0;
  	unsigned long qlim = 0;
+	int num_children = 0;
+	struct list_head *_p;

  	sigemptyset(&pending);
  	sigemptyset(&shpending);
@@ -268,9 +270,13 @@ static inline char * task_sig(struct tas
  		qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
  		spin_unlock_irq(&p->sighand->siglock);
  	}
+	list_for_each(_p, &p->children)
+		++num_children;
  	read_unlock(&tasklist_lock);

  	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
+	if (num_children)
+		buffer += sprintf(buffer, "Children:\t%d\nTotal:\t%d\n",  
num_children, num_threads + num_children);
  	buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);

  	/* render them all */


^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-3)
  2006-09-29 18:24           ` [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-3) girish
@ 2006-09-29 22:51             ` Eric W. Biederman
       [not found]               ` <E78297DA-5F0F-40FA-A008-89264570B313@gmail.com>
  0 siblings, 1 reply; 10+ messages in thread
From: Eric W. Biederman @ 2006-09-29 22:51 UTC (permalink / raw)
  To: girish; +Cc: Jan Engelhardt, linux-kernel, William Pitcock


This conflicts with the /proc changes in -mm.
Where we have manged to remove the use of the tasklist_lock.

The information in Children: is racy as it may change immediately
after you drop the lock.

Why is it interesting to report this information?
A process that cares can keep track of this in user space?

Eric

> Signed-off-by: Girish V. Gulawani <girishvg@gmail.com>
>
> --- linux-vanilla/fs/proc/array.c	2006-09-20 12:42:06.000000000 +0900
> +++ linux/fs/proc/array.c	2006-09-30 03:18:28.000000000 +0900
> @@ -248,6 +248,8 @@ static inline char * task_sig(struct tas
>  	int num_threads = 0;
>  	unsigned long qsize = 0;
>  	unsigned long qlim = 0;
> +	int num_children = 0;
> +	struct list_head *_p;
>
>  	sigemptyset(&pending);
>  	sigemptyset(&shpending);
> @@ -268,9 +270,13 @@ static inline char * task_sig(struct tas
>  		qlim = p->signal->rlim[RLIMIT_SIGPENDING].rlim_cur;
>  		spin_unlock_irq(&p->sighand->siglock);
>  	}
> +	list_for_each(_p, &p->children)
> +		++num_children;
>  	read_unlock(&tasklist_lock);
>
>  	buffer += sprintf(buffer, "Threads:\t%d\n", num_threads);
> +	if (num_children)
> +		buffer += sprintf(buffer, "Children:\t%d\nTotal:\t%d\n",
> num_children, num_threads + num_children);
>  	buffer += sprintf(buffer, "SigQ:\t%lu/%lu\n", qsize, qlim);
>
>  	/* render them all */
>
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-3)
       [not found]               ` <E78297DA-5F0F-40FA-A008-89264570B313@gmail.com>
@ 2006-09-30 17:11                 ` Jan Engelhardt
  2006-10-01 11:34                   ` girish
  0 siblings, 1 reply; 10+ messages in thread
From: Jan Engelhardt @ 2006-09-30 17:11 UTC (permalink / raw)
  To: girish; +Cc: Eric W. Biederman, linux-kernel, William Pitcock

>
>  PID COMMAND      %CPU   TIME   #TH #PRTS #MREGS RPRVT  RSHRD  RSIZE  VSIZE
> 22429 top         16.6%  0:21.12   1    18    20  1.35M   684K  1.77M  26.9M
> ------------------------------------------------------------------------
> --------------------------------------------------------
>
> Comments/opinions?

I fail to see which column you mean.

#TH perhaps? I think, that can be calculated under linux by

 (a) counting the directories in /proc/22429/task using readdir
or
 (b) get the nlink of /proc/22429/task and subtract 2, which should give 
     the same as (a), and, better than (a), should also be atomic


Jan Engelhardt
-- 

^ permalink raw reply	[flat|nested] 10+ messages in thread

* Re: [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-3)
  2006-09-30 17:11                 ` Jan Engelhardt
@ 2006-10-01 11:34                   ` girish
  0 siblings, 0 replies; 10+ messages in thread
From: girish @ 2006-10-01 11:34 UTC (permalink / raw)
  To: Jan Engelhardt; +Cc: Eric W. Biederman, linux-kernel, William Pitcock

>>  PID COMMAND      %CPU   TIME   #TH #PRTS #MREGS RPRVT  RSHRD   
>> RSIZE  VSIZE
>> 22429 top         16.6%  0:21.12   1    18    20  1.35M   684K   
>> 1.77M  26.9M
>> --------------------------------------------------------------------- 
>> ---
>> --------------------------------------------------------
>>
>> Comments/opinions?
>
> I fail to see which column you mean.
>
> #TH perhaps? I think, that can be calculated under linux by
>
>  (a) counting the directories in /proc/22429/task using readdir
> or

I have implemented the child_count script, in this way. I was  
wondering what is more convenient.


>  (b) get the nlink of /proc/22429/task and subtract 2, which should  
> give
>      the same as (a), and, better than (a), should also be atomic

This one, is good. I did not think about this approach. 

^ permalink raw reply	[flat|nested] 10+ messages in thread

end of thread, other threads:[~2006-10-01 11:34 UTC | newest]

Thread overview: 10+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2006-09-29 15:18 [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-1) girish
2006-09-29 16:31 ` William Pitcock
2006-09-29 16:51   ` girish
2006-09-29 17:06     ` Jan Engelhardt
2006-09-29 17:18       ` girish
2006-09-29 18:12         ` Jan Engelhardt
2006-09-29 18:24           ` [PATCH] include children count, in Threads: field present in /proc/<pid>/status (take-3) girish
2006-09-29 22:51             ` Eric W. Biederman
     [not found]               ` <E78297DA-5F0F-40FA-A008-89264570B313@gmail.com>
2006-09-30 17:11                 ` Jan Engelhardt
2006-10-01 11:34                   ` girish

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).