Subject: Re: keeping the process start time in core at all times...
To: NetBSD Kernel Technical Discussion List <tech-kern@NetBSD.ORG>
From: Greg A. Woods <woods@weird.com>
List: tech-kern
Date: 11/29/2002 22:42:36
[ On Saturday, November 30, 2002 at 13:00:20 (+1100), Simon Burge wrote: ]
> Subject: Re: keeping the process start time in core at all times... 
>
> It went away when ps was changed to use sysctl() calls to fetch all
> the data.  

Ah ha.... yes.... KVM_NO_FILES....  KERN_PROC2, struct kinfo_proc2, etc.

Maybe David Laight's idea of having a way to force a read of the u area
would indeed be the best way to go -- perhaps a third-level name like
KERN_PROC2_FULL and of course yet another command-line flag (-f?) for
'ps' to request it be done.

> That said, I have heard lots of people who mentioned that they miss the
> start time being displayed by ps.  The stats part of the U is either 224
> bytes on a 32bit arch or 448 bytes on a 64bit arch - I think that's too
> big to add to struct proc, but adding just the start time is something
> I'd be for...

Yes, that is a fair bit more crud to force everyone to keep around, at
least for machines with big proc tables (or very little RAM).....

Well, if KERN_PROC2 can't be "fixed" then I'd certainly be happy with
just p_start moved to struct proc....

-- 
								Greg A. Woods

+1 416 218-0098;            <g.a.woods@ieee.org>;           <woods@robohack.ca>
Planix, Inc. <woods@planix.com>; VE3TCP; Secrets of the Weird <woods@weird.com>