[9fans] undocumented fs config options?

[9fans] undocumented fs config options?

[anothy]

in reading through /sys/src/fs/worms and the file
server source, there seem to be two undocumented
device types: r and x.

x is Devswab - i believe this bit-swaps whatever's
on the physical disk, presumably for moving a file
system between different-endian servers. correct?

i'm not clear on r; it seems related to be the same
as w, but for optical disks (described in
port/portdat.h as "scsi video drive"), or l with no
label (l is described as "scsi video drive
(labeled)"). in the examples, r, like l, is only
used within a j device to describe the unmountable
platters. it seems either r or l is required in the
device3 section of a j definition; fsconfig(8)
makes no mention of that.

can l or r be used on regular disks? what's the
difference between r and w devices?
-α.

Re: [9fans] undocumented fs config options?

[Boyd Roberts]

From: <nigel@9fs.org>
> They might, but I'm confused. Why would I want to sync
> to the German signal when I can use the UK signal?

you probably wouldn't.  i was thinking more genrally.
IIRC that xmitter has a 1600km range from frankfurt.

Re: [9fans] undocumented fs config options?

[nigel]

[-- Attachment #1: Type: text/plain, Size: 110 bytes --]

They might, but I'm confused. Why would I want to sync
to the German signal when I can use the UK signal?


[-- Attachment #2: Type: message/rfc822, Size: 2041 bytes --]

From: "Boyd Roberts" <boyd@planete.net>
To: <9fans@cse.psu.edu>
Subject: Re: [9fans] undocumented fs config options?
Date: Thu, 10 May 2001 07:31:40 +0200
Message-ID: <00c101c0d912$7dc36510$e8b7c6d4@SOMA>

From: "Nigel Roles" <nigel@9fs.org>
> 
> The configuration I use is an ARC Galleon Rugby clock receiver
> connected to the cpu server, a small daemon to read the clock
> and set the rtc, and sntpd to serve time to whomever requires it.

do they sell the oregon scientific clocks in the uk?  they're the
one's which are synched with an xmitter in germany.

they cost nothing and should be easily converted to do the job.

then again, there's the TAC.

Re: [9fans] undocumented fs config options?

[Boyd Roberts]

From: "Nigel Roles" <nigel@9fs.org>
> 
> The configuration I use is an ARC Galleon Rugby clock receiver
> connected to the cpu server, a small daemon to read the clock
> and set the rtc, and sntpd to serve time to whomever requires it.

do they sell the oregon scientific clocks in the uk?  they're the
one's which are synched with an xmitter in germany.

they cost nothing and should be easily converted to do the job.

then again, there's the TAC.

Re: [9fans] undocumented fs config options?

[Nigel Roles]

I believe the sntp option is also undocumented....

If you enter the config

ipsntp <address>

the fileserver will periodically set the clock from an sntp server
running on the machine with the given address. This allows
a variety of clock sources to be used, rather than relying on
something attached to the fileserver.

The configuration I use is an ARC Galleon Rugby clock receiver
connected to the cpu server, a small daemon to read the clock
and set the rtc, and sntpd to serve time to whomever requires it.

> On Tue May  8 11:20:27 EDT 2001, anothy@cosym.net wrote:
> > in reading through /sys/src/fs/worms and the file
> > server source, there seem to be two undocumented
> > device types: r and x.
> >
> > x is Devswab - i believe this bit-swaps whatever's
> > on the physical disk, presumably for moving a file
> > system between different-endian servers. correct?
>
> yes, 'x' does the non-trivial swap between big and little
> endian filesystem formats. this allowed us to move our old
> sony jukeboxes from a big-endian mips magnum to a little-endian
> pc. since then we've copied the data off the sonys and
> turned them off.
>
> >
> > i'm not clear on r; it seems related to be the same
> > as w, but for optical disks (described in
> > port/portdat.h as "scsi video drive"), or l with no
> > label (l is described as "scsi video drive
> > (labeled)"). in the examples, r, like l, is only
> > used within a j device to describe the unmountable
> > platters. it seems either r or l is required in the
> > device3 section of a j definition; fsconfig(8)
> > makes no mention of that.
>
> my understanding is that 'r' and 'l' only differ in that
> with 'l' one block is reserved on the platter for a label.
>
> >
> > can l or r be used on regular disks? what's the
> > difference between r and w devices?
> > -α.
>
> i believe 'r' and 'l' devices can only be used as the third
> part of a 'j' device. there is no individual 'init' routine for
> 'r' and 'l' devices, it's part of the 'j' initialisation.
>
> that's not to say it might be possible to concoct a filesystem
> specification with 'r' or 'l' devices not associated with a
> jukebox that would pass the configuration parsing, but it wouldn't
> work.

Re: [9fans] undocumented fs config options?

[jmk]

On Tue May  8 11:20:27 EDT 2001, anothy@cosym.net wrote:
> in reading through /sys/src/fs/worms and the file
> server source, there seem to be two undocumented
> device types: r and x.
> 
> x is Devswab - i believe this bit-swaps whatever's
> on the physical disk, presumably for moving a file
> system between different-endian servers. correct?

yes, 'x' does the non-trivial swap between big and little
endian filesystem formats. this allowed us to move our old
sony jukeboxes from a big-endian mips magnum to a little-endian
pc. since then we've copied the data off the sonys and
turned them off.

> 
> i'm not clear on r; it seems related to be the same
> as w, but for optical disks (described in
> port/portdat.h as "scsi video drive"), or l with no
> label (l is described as "scsi video drive
> (labeled)"). in the examples, r, like l, is only
> used within a j device to describe the unmountable
> platters. it seems either r or l is required in the
> device3 section of a j definition; fsconfig(8)
> makes no mention of that.

my understanding is that 'r' and 'l' only differ in that
with 'l' one block is reserved on the platter for a label.

> 
> can l or r be used on regular disks? what's the
> difference between r and w devices?
> -α.

i believe 'r' and 'l' devices can only be used as the third
part of a 'j' device. there is no individual 'init' routine for
'r' and 'l' devices, it's part of the 'j' initialisation.

that's not to say it might be possible to concoct a filesystem
specification with 'r' or 'l' devices not associated with a
jukebox that would pass the configuration parsing, but it wouldn't
work.