Bug #19

usched questions and a patch

Added by belczyk over 8 years ago. Updated over 5 years ago.

Status:ClosedStart date:
Priority:LowDue date:
Assignee:-% Done:

0%

Category:-
Target version:-

Description

Hello hackers!

I've got a question about the TAILQ_ENTRY(lwp) lwp_procq in struct lwp - why
isn't it a part of lwp_usdata union? Isn't it a usched-dependent data? What
if a different usched wants to hold its lwps in e.g. red-black tree or a
heap?

Another thing I noticed is that usched_set() syscall does not zero the
content of lwp_usdata which makes impossible for the newly selected usched
to distinguish between an lwp it already had on its queue/tree/whatever and
had already put some data into its lwp_usdata, and a 'new' lwp it will get in
the following call to acquire_curproc() from userexit() IIRC.

What do You think about the attached patch?

And one more thing - what do You think about adding another usched API call,
e.g. exiting_curproc() called from exit1() instead of release_curproc() so
that a usched could (1) actually trace lwps during their lifetime and
(2) differentiate between lwps 'only' going into kernel and the ones that do
exit1()?

usched-misc.diff Magnifier (3.52 KB) belczyk, 12/21/2005 06:41 PM

History

#1 Updated by alexh over 5 years ago

Is the attached patch of any interest whatsoever? I'm not familiar enough with the
scheduler.

#2 Updated by belczyk over 5 years ago

On Sun, May 10, 2009 at 10:17:25AM +0000, Alex Hornung (via DragonFly issue tracker) wrote:
>
> Alex Hornung <> added the comment:
>
> Is the attached patch of any interest whatsoever? I'm not familiar enough with the
> scheduler.

It is so old... please close this issue.

Also available in: Atom PDF