Re: uarea swap-out

[Matt Thomas <matt%3am-software.com@localhost>]

On Aug 16, 2009, at 5:07 PM, Mindaugas Rasiukevicius wrote:

> Hello,
>
> The following patch removes the uarea swap-out functionality:
>
> http://www.netbsd.org/~rmind/uarea_swapout_rem.diff
>
> After some thought and discussion with few developers, I believe it is
> worth, because it:
>
> - Addresses the issue described in PR/38828.
> - Gets small, but nice simplification in threading code.
> - Avoids few scans on LWP list and thus potentially long holds of  
> proc_lock.
> - Removes __SWAP_BROKEN cases.
> - Cuts around 1000 lines of code.
>
> Concerns?

If we are going to get rid of swappable uareas, then we should go all  
the way.

finally kill struct user (which just contains the pcb).  Move the pcb  
into the lwp as l_md.

The separate lwp pool should go away and the lwp should just be placed  
at the bottom of the uarea followed by the pcb.  l_addr should be  
renamed l_pcb.

(shouldn't proc0paddr be renamed to lwp0laddr?)

For lwp0, there is no reason why it has to be copied to the lwp0 uarea  
and it can remain
where it is.

As page sizes increase, we also need to deal with subpage uareas.