jbd2 batch time tuning behavior question

jbd2 batch time tuning behavior question

[Eric Sandeen]

Question for Ted:  With this commit, you introduced some documentation and some code which are not consistent AFAICT:


commit 30773840c19cea60dcef39545960d541b1ac1cf8
Author: Theodore Ts'o <tytso@mit.edu>
Date:   Sat Jan 3 20:27:38 2009 -0500

    ext4: add fsync batch tuning knobs

<snip>

+                       the transaction.   The commit time is capped by
+                       the max_batch_time, which defaults to 15000us
+                       (15ms).   This optimization can be turned off
+                       entirely by setting max_batch_time to 0.

<snip>

+               case Opt_max_batch_time:
+                       if (match_int(&args[0], &option))
+                               return 0;
+                       if (option < 0)
+                               return 0;
+                       if (option == 0)
+                               option = EXT4_DEF_MAX_BATCH_TIME;
+                       sbi->s_max_batch_time = option;
+                       break;

Docs say "0" turns it off, but option handling turns "0" into 
EXT4_DEF_MAX_BATCH_TIME.  What was the real intent here; should the
docs or the behavior be changed?

Thanks,
-Eric

Re: jbd2 batch time tuning behavior question

[Theodore Ts'o]

On Tue, May 27, 2014 at 11:47:06AM -0500, Eric Sandeen wrote:
> 
> Docs say "0" turns it off, but option handling turns "0" into 
> EXT4_DEF_MAX_BATCH_TIME.  What was the real intent here; should the
> docs or the behavior be changed?

Nice catch!  I would think it's probably better if we make the
behavior change to match the docs.

Want to send a patch, since you found the problem, or shall I create a
patch myself to fix things up?

Thanks,

     	       	      	       	      	  - Ted

Re: jbd2 batch time tuning behavior question

[Eric Sandeen]

On 6/2/14, 9:42 AM, Theodore Ts'o wrote:
> On Tue, May 27, 2014 at 11:47:06AM -0500, Eric Sandeen wrote:
>>
>> Docs say "0" turns it off, but option handling turns "0" into 
>> EXT4_DEF_MAX_BATCH_TIME.  What was the real intent here; should the
>> docs or the behavior be changed?
> 
> Nice catch!  I would think it's probably better if we make the
> behavior change to match the docs.
> 
> Want to send a patch, since you found the problem, or shall I create a
> patch myself to fix things up?
> 
> Thanks,
> 
>      	       	      	       	      	  - Ted
> 

I can send the patch, I just wondered what the intent was.

Thanks,
-Eric