Subject: Re: Importing tmpfs
To: None <tech-kern@netbsd.org>
From: Christos Zoulas <christos@astron.com>
List: tech-kern
Date: 09/09/2005 17:15:19
In article <6b2d1e19050909100779f91b0a@mail.gmail.com>,
Julio M. Merino Vidal <jmmv84@gmail.com> wrote:
>Hi all,
>
>[ I'm CC'ing my SoC project mentor ]
>
>The tmpfs developed during the Summer of Code program is now
>usable, somewhat resource-friendly (improved a lot in the past
>few days) but still has some bugs and misfeatures. The most
>important ones:
>- Does not use anonymous memory for file meta-data.
>- Some random panics caused due to incorrect memory accesses.
> I *suspect* these are related in some way to the usage of wired
> kernel memory and exhausting it, so fixing the previous item properly
> may also fix this one.
>- NFS support is broken.
>
>The thing is that many people (mostly developers) is asking me
>when it will be integrated into NetBSD's main source tree. The
>advantages of importing it now:
>- It is easier for people to test it. Just enable it in the kernel and
> voila!
>- It is easier for other developers to contribute, if they want to.
>- The current sources have a lot of CVS history attached to them,
> all of which will be lost during the import. Importing it sooner
> means we will loose less history. (With "import" I'm referring to
> cvs-adding files.)
>
>The disadvantages... well, as long as it stays disabled by default,
>there are no disadvantages, I think. But if people use it on purpose,
>they can expect problems. (The manual pages warn them about
>experimental code ;)
>
>Note that all the files are licensed under the official TNF license.
>
>So what do you people think? Can/should it be incorporated now
>or should it wait? Until which point?
I am all for importing it, but your sponsor should review it and give
his ok first.
christos