[9fans] Ken fs updated

[9fans] Ken fs updated

[geoff]

As you've probably noticed if you've pulled today, the file server
kernel, manual pages and paper have been updated to match my 64-bit
file server kernel.  I've removed /n/sources/contrib/geoff/fs64.tgz.
/sys/src/fs/doc has some examples, advice and a list of changes.

The emelie kernel is now called 9emeliefs.  Examining port/portdat.h,
you'll see that defining OLD will compile a 32-bit kernel compatible
on disk with existing file systems.  There's a little more noise at
boot time (but the file server has always been chatty), including an
announcement of the largest file size.

Diffs against the previous kernel are voluminous, but mostly due to
two things: type parameterisation and goto elimination.

I'm running the 64-bit kernel and ran the 32-bit kernel from this tree
during development.  9netics has one each (a 32-bit and a 64-bit
kernel) and Kenji Okamoto is running the 64-bit kernel with filename
components cranked up to 256 bytes maximum.

Re: [9fans] Ken fs updated

[Chris Silva]

Is there a install .iso? (last time I installed it was changed to fossil only)

On 2/18/06, geoff@collyer.net <geoff@collyer.net> wrote:
> As you've probably noticed if you've pulled today, the file server
> kernel, manual pages and paper have been updated to match my 64-bit
> file server kernel.  I've removed /n/sources/contrib/geoff/fs64.tgz.
> /sys/src/fs/doc has some examples, advice and a list of changes.
>
> The emelie kernel is now called 9emeliefs.  Examining port/portdat.h,
> you'll see that defining OLD will compile a 32-bit kernel compatible
> on disk with existing file systems.  There's a little more noise at
> boot time (but the file server has always been chatty), including an
> announcement of the largest file size.
>
> Diffs against the previous kernel are voluminous, but mostly due to
> two things: type parameterisation and goto elimination.
>
> I'm running the 64-bit kernel and ran the 32-bit kernel from this tree
> during development.  9netics has one each (a 32-bit and a 64-bit
> kernel) and Kenji Okamoto is running the 64-bit kernel with filename
> components cranked up to 256 bytes maximum.
>

Re: [9fans] Ken fs updated

[geoff]

As far as I know, there's no ISO image to install Ken's fs.  Note that
this is the standalone file server found in /sys/src/fs, not kfs.
It's officially deprecated, but some of us old-timers still run it.

Re: [9fans] Ken fs updated

[Russ Cox]

> Is there a install .iso? (last time I installed it was changed to fossil only)

The /sys/src/fs file server is not intended for use in new systems.
We don't even use it much inside Bell Labs anymore.  It's included
in the distribution mainly for those who must maintain old servers.

It's true that fossil had stability problems early on (though thanks
to venti I've never lost any data), but it is pretty solid now.
I fixed some performance bugs related to sync and remove 
recently, but otherwise we haven't been significant changes.

The only outstanding problem I know about in fossil is that 
it doesn't handle filling up gracefully.  But neither does the
old file server.

Russ

Re: [9fans] Ken fs updated

[Russ Cox]

> Please address this. When ir screws wirth me it's real.

I have, off and on, been thinking about how to do this.
I do plan to put in a fix eventually, but I haven't decided
what a good fix is.

Russ

Re: [9fans] Ken fs updated

[Brantley Coile]

> The /sys/src/fs file server is not intended for use in new systems.

I still perfer it.

	bc
	

Re: [9fans] Ken fs updated

[Bruce Ellis]

Please address this. When ir screws wirth me it's real.

brucee

On 2/18/06, Russ Cox <rsc@swtch.com> wrote:
> > Is there a install .iso? (last time I installed it was changed to fossil only)
>
> The /sys/src/fs file server is not intended for use in new systems.
> We don't even use it much inside Bell Labs anymore.  It's included
> in the distribution mainly for those who must maintain old servers.
>
> It's true that fossil had stability problems early on (though thanks
> to venti I've never lost any data), but it is pretty solid now.
> I fixed some performance bugs related to sync and remove
> recently, but otherwise we haven't been significant changes.
>
> The only outstanding problem I know about in fossil is that
> it doesn't handle filling up gracefully.  But neither does the
> old file server.
>
> Russ
>
>