Issue #2449 has been updated by Venkatesh Srinivas.

On Wed, Nov 07, 2012 at 09:21:41PM -0800, Adam Sakareassen via Redmine wrote:
>
>Issue #2449 has been updated by Adam Sakareassen.
>
>File 0001-Fix-INVARIANT-build-issue.patch added
>
>> * I think the "++spin->countb;" can be moved up into the INVARIANTS too.
>
>I agree.  A new patch is attached with such changes. (Disregard the previous 
>patch file)

In the version of the spinlock just before the s/x mode went in, the
++spin->countb was actually important for performance on a 48-core
(4 socket x 12-core) K10 machine I believe; there should be logs of
discussion on IRC from October or November 2011 discussing it.

>countb is not referred to anywhere in the kernel code outside of Invariants 
>code.  I have left it's definition in the header files at this stage.  It 
>looks like some of the comments in those header files could use with an 
>update.  It appears the code increments or decrements counta.  Rather than use 
>a chase counter as the comments suggest.
>

The chased counter comment was a relic from a past version of the spinlock code.

-- vs;
----------------------------------------
Submit #2449: INVARIANTS option fix.
http://bugs.dragonflybsd.org/issues/2449

Author: Adam Sakareassen
Status: Closed
Priority: Normal
Assignee: 
Category: 
Target version: 


With the INVARIANTS option defined the kernel does not build.  
This small patch fixes the issue.


-- 
You have received this notification because you have either subscribed to it, 
or are involved in it.
To change your notification preferences, please click here: 
http://bugs.dragonflybsd.org/my/account

Reply via email to