Re: XFS (was: ReiserFS (was: JFS (was: The FreeBSD core team needs your help)))a

2001-07-19 Thread Greg Lehey

On Sunday,  8 July 2001 at 20:47:53 -0500, Dave Uhring wrote:
> On Sunday 08 July 2001 19:32, Matthew Emmerton wrote:
>
>>
>> This "virus-like" aspect of the GPL is also very loudly explained in
>> a 12-page presentation distributed internally to all developers at
>> IBM.  (The virus-like aspect is a big deal.  If someone accidentally
>> *statically* linked a piece of GPL'd object code, such as GNU
>> getopt(), into a major product such as DB2 EEE for Linux, then they
>> would be forced to open the code to DB2.  That would not be a very
>> profitable move for IBM.)
>>
>> The problem for me (as a developer who would love to port XFS to
>> FreeBSD, but by being employed by IBM, cannot), is that IBM would
>> have IP rights over any changes that I would have to submit back to
>> SGI in order to make XFS work on FreeBSD.  For IBM to release that
>> code under the GPL, SGI would have to work with IBM and come to an
>> agreement (which would involve all of the IP laywers from the two
>> firms battling it out).  Since this process would take months and a
>> pile of cash, FreeBSD would never see XFS.
>>
>> Sad, but true.  Maybe I need to switch companies :)
>
> No need to switch companies over something like this.  Fix up JFS for
> FreeBSD :-)  and stay with IBM.

As I've said on the correct group (FreeBSD-fs), if anybody wants to
port JFS to FreeBSD, please contact me.

Greg
--
See complete headers for address and phone numbers

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2001-07-08 Thread Bill Moran

Juha Saarinen wrote:
> 
> On Sun, 8 Jul 2001, Bill Moran wrote:
> 
> > I know my opinion of Wind River has been negatively impacted by
> > the numerous spelling errors I found on their web site the first
> > time I visited.
> 
> That's different though -- one person rubbishes the product because the
> presentation uses a Politically Incorrect typeface, whereas you (quite
> rightly) assume that if they can't be bothered to run a spelling checker
> over their Web page, their quality control is somewhat suspect...

True, as I pointed out later in my previous post:
"On the flip side, the Sistina page displays well even though I
don't have that font, so _I_ wasn't put off by it."

-Bill

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: XFS (was: ReiserFS (was: JFS (was: The FreeBSD core team needsyour help)))a

2001-07-08 Thread Jonathan Belson

Juha Saarinen wrote:
> 
> On Sun, 8 Jul 2001, Matthew Emmerton wrote:
> 
> > So why don't we get a FreeBSD port of XFS done "real soon now", see what
> > kind of kernel mods we have to make, and talk nice to the SGI folks?  If
> > they're running into brick walls at every turn on the Linux path, why don't
> > we make the FreeBSD path look like the yellow brick road?  It may be a great
> > way to get some SGI resources headed the way of FreeBSD.
> 
> I asked about this before... SGI isn't keen on Sun, Microsoft, IBM et al
> "leeching" its intellectual property, so while the developers like the
> idea, the lawyers are dead against it.

Is it possible to have such filesystem compiled as a module to sidestep
licensing issues?


--
C-YA
Jon



To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2001-07-08 Thread Juha Saarinen

On Sun, 8 Jul 2001, Bill Moran wrote:

> I know my opinion of Wind River has been negatively impacted by
> the numerous spelling errors I found on their web site the first
> time I visited.

That's different though -- one person rubbishes the product because the
presentation uses a Politically Incorrect typeface, whereas you (quite
rightly) assume that if they can't be bothered to run a spelling checker
over their Web page, their quality control is somewhat suspect...


-- 
Regards,


Juha

PGP fingerprint:
B7E1 CC52 5FCA 9756 B502  10C8 4CD8 B066 12F3 9544


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: XFS (was: ReiserFS (was: JFS (was: The FreeBSD core team needsyour help)))a

2001-07-08 Thread Juha Saarinen

On Sun, 8 Jul 2001, Matthew Emmerton wrote:

> So why don't we get a FreeBSD port of XFS done "real soon now", see what
> kind of kernel mods we have to make, and talk nice to the SGI folks?  If
> they're running into brick walls at every turn on the Linux path, why don't
> we make the FreeBSD path look like the yellow brick road?  It may be a great
> way to get some SGI resources headed the way of FreeBSD.

I asked about this before... SGI isn't keen on Sun, Microsoft, IBM et al
"leeching" its intellectual property, so while the developers like the
idea, the lawyers are dead against it.


-- 
Regards,


Juha

PGP fingerprint:
B7E1 CC52 5FCA 9756 B502  10C8 4CD8 B066 12F3 9544


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2001-07-08 Thread David Malone

On Sat, Jul 07, 2001 at 08:35:35PM -0500, Dave Uhring wrote:

> You seem to have missed the critical point of that paper.  When the
> system goes completely haywire and either crashes or locks up so hard
> that a manual reset is required, UFS/softupdates requires a substantial
> amount of time to run fsck.

Background fsck is now working in -current, which means that when your
system boots you don't have to fsck the disk immediately. It seems to
work just fine for me so far.

> If you have a very large filesystem, you
> then have to wait until fsck completes.

I believe that giving the right options to newfs can significantly
reduce fsck times too. There's notes on it in the new tuning man
page.

> I use "logging" on Solaris and XFS on Linux and have tried reiserfs on
> Linux.  All are superior to UFS/softupdates when the going gets tough.
> Disk access times may or may not be comparable with UFS/softupdates, but
> the integrity of my filesystems is more important than raw speed.

AFAIK Softupdates shouldn't be any less carefull with your data
than journaling, providing the application calls fsync. One advantage
might be that data is written to the disk twice, which means if
one bit of the disk goes bad you might be able to find it elsewhere.
(Mind you, I guess RAID is the right way to do that sort of thing.)

I dunno which is harder to impliment right - journaling or softupdates.
This may actually be the issue which determines the safety of your
data.

David.

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2001-07-08 Thread Bill Moran

Dag-Erling Smorgrav wrote:
> 
> Dave Uhring <[EMAIL PROTECTED]> writes:
> > I just took a look at www.sistina.com and a web site which has its font
> > set to Arial is suspect, IMHO.  If they have to use Microsoft products
> > to produce a web site..
> 
> Is that the best you could come up with?  I mean, you could criticize
> the ingenuity of their designs, or the quality of their code, or their
> ability to deliver on their promises - or would that require too much
> effort?  It's so much easier to just dismiss them out of hand because
> they use a font you don't like on their web page, isn't it?

Generally, this is what most people do. I'm not defending Dave, or
supporting DES. But many people judge a company by such first
appearances. I know people who will dismiss a company as
amatuer by the quality of their product packaging.
I know my opinion of Wind River has been negatively impacted by
the numerous spelling errors I found on their web site the first
time I visited. Web pages that only display in M$ browsers also
make me feel uncomfortable with a company. and _assuming_ that
everyone will have a font just because it comes with Windows
falls into that category.
On the flip side, the Sistina page displays well even though I
don't have that font, so _I_ wasn't put off by it. Other aspects
of the page design made me decide not to read further into it,
however. (primarily the diffucult-to-read blue on black text on
the nav column)

-Bill

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: XFS (was: ReiserFS (was: JFS (was: The FreeBSD core team needs your help)))a

2001-07-08 Thread Matthew Emmerton

> On Sun, Jul 08, 2001 at 10:02:27AM -0400, Matthew Emmerton wrote:
> >
> > So why don't we get a FreeBSD port of XFS done "real soon now", see what
> > kind of kernel mods we have to make, and talk nice to the SGI folks?  If
> > they're running into brick walls at every turn on the Linux path, why
don't
> > we make the FreeBSD path look like the yellow brick road?  It may be a
great
> > way to get some SGI resources headed the way of FreeBSD.
>
> Great idea!  Why don't you go do it?

I would, but since I currently work for IBM, it's not going to happen.
(Under the terms of my current IP agreement with IBM, if I port XFS to
FreeBSD, IBM will "own" the FreeBSD XFS port -- something I doubt SGI would
be happy about.  Believe me, I hate my IP agreement and want to change it,
but I don't have the $$ for lawyers right now.)

--
Matt Emmerton


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2001-07-07 Thread Dave Duchscher

On Sat, Jul 07, 2001 at 08:35:35PM -0500, Dave Uhring wrote:
>
> You seem to have missed the critical point of that paper.  When the
> system goes completely haywire and either crashes or locks up so hard
> that a manual reset is required, UFS/softupdates requires a substantial
> amount of time to run fsck.  If you have a very large filesystem, you
> then have to wait until fsck completes.  And if you are
> lucky, it will not terminate with the suggestion that you run fsck by
> hand.  With a true journalling filesystem this wait is obviated.  The
> last transactions are rerun or truncated and the system boots up.

Just to bring up a point, Softupdates will also avoid the long fsck at
boot.  If I understand the papers I have read and with playing with
Softupdates on current, Softupdates leaves files system in a
consistent state so that the file-system can be mounted after a
crash/lockup/etc immediately and only a background fsck need be run to
free up left over pieces laying around.

You guys also might want to wonder over to Kirk's Softupdates site:

  http://www.McKusick.com/softdep/index.html

DaveD

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



ReiserFS (was: JFS (was: The FreeBSD core team needs your help))

2001-07-07 Thread Greg Lehey

On Saturday,  7 July 2001 at 14:50:21 +0200, A. L. Meyers wrote:
> On Sat, 7 Jul 2001, Greg Lehey wrote:
>
>> On Friday,  6 July 2001 at 11:31:49 +0100, Antony T Curtis wrote:
>>> Greg Lehey wrote:

 On Wednesday,  4 July 2001 at 11:38:08 +0100, Antony T Curtis wrote:
> Greg Lehey wrote:
>>
>> On Tuesday, 12 June 2001 at 19:22:45 +0200, Steve O'Hara-Smith wrote:
>>> On Tue, 12 Jun 2001 12:09:58 +0100
>>> Josef Karthauser <[EMAIL PROTECTED]> wrote:
>>>
 On Fri, Jun 08, 2001 at 08:32:23AM -0700, Eric Parusel wrote:
>>> A journalling FS for those people who just hate waiting for a
> couple
>>> of
>>> TB of slow disks to fsck?
>>
>> Does ReiserFS work with FreeBSD? ? ? ?
>
> (big snip)
>
> Hey, guys and gals, did you forget this part of the original
> post?

No, they didn't notice it, because the Subject: line was pointing
elsewhere.  Piggybacking new questions onto old topics doesn't work
well unless you change the Subject: line.

> Just installed SuSE Linux 7.2 with Reiser FS throughout on an Intel
> SMP box. The FS purrs, even on /, which doesn't mean everything is
> better or worse than FBSD.

I don't know enough about ReiserFS to be able to give a useful
opinion.  The Linux people I know are by no means in agreement about
its merits, but I've heard that it's best as a "special purpose" FS
for small files.  I don't know how valid that statement is.

Greg
--
See complete headers for address and phone numbers

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS (was: The FreeBSD core team needs your help)

2001-07-07 Thread Philip J. Koenig

> Date: Sat, 7 Jul 2001 14:50:21 +0200 (CEST)
> From: "A. L. Meyers" <[EMAIL PROTECTED]>

> As far as I know, ReiserFS is GPL. What would porting it to
> FreeBSD be better or worse than other (newer and less stable)
> alternatives?
> 
> Is this a technical issue or are politics involved?


AFAIK ReiserFS is a work in progress.

There are some other established journaling FS's out there which 
might be candidates, ie the IBM one and one from SGI.




--
Philip J. Koenig   [EMAIL PROTECTED]
Electric Kahuna Systems -- Computers & Communications for the New Millenium


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2001-07-07 Thread Ted Sikora

Juha Saarinen wrote:
> 
> On Sat, 7 Jul 2001, Dave Uhring wrote:
> 
> > I use "logging" on Solaris and XFS on Linux and have tried reiserfs on
> > Linux.  All are superior to UFS/softupdates when the going gets tough.
> > Disk access times may or may not be comparable with UFS/softupdates, but
> > the integrity of my filesystems is more important than raw speed.
> 
> Hmmm... that's one reason I've not implemented ReiserFS on my Linux box.
> Read too many horror stories about how it eats your file system, and how
> it doesn't work with NFS etc.

Absolutely false. Works perfectly... I have had it on several machines
for about a year with absolutely no problems .. nada. On the other hand 
UFS/softupdates is pretty reliable itself albiet a little pokier. I
haven't seen fsck in action for so long I almost forgot it existed.
Reiserfs would be nice on FreeBSD though. IBM's JFS might be even
better.

--
Ted Sikora
[EMAIL PROTECTED]
http://www.unixos2.org
http://www.powerusersbbs.com

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2001-07-07 Thread Dave Uhring


- Original Message -
From: "Bill Moran" <[EMAIL PROTECTED]>
Cc: <[EMAIL PROTECTED]>
Sent: Saturday, July 07, 2001 4:04 PM
Subject: Re: JFS


> Dave Uhring wrote:
> > You seem to have missed the critical point of that paper.  When the
> > system goes completely haywire and either crashes or locks up so
hard
> > that a manual reset is required, UFS/softupdates requires a
substantial
> > amount of time to run fsck.  If you have a very large filesystem,
you
> > then have to wait until fsck completes.  And if you are
> > lucky, it will not terminate with the suggestion that you run fsck
by
> > hand.  With a true journalling filesystem this wait is obviated.
The
> > last transactions are rerun or truncated and the system boots up.
>
> Actually ... according to the article, the system boots up and _then_
> determines what needs done to repair the filesystem.
>
> Also, the lack of a need for fscking is not the only benefit of
> RieserFS. In fact, it's a _minor_ improvement. If your system is
> going down so often that the speed of a fsck is a major factor in the
> layout of the system, you've got other issues you need to address
> first!
> The other issues that might make Reiserfs a good idea (and a possible
> improvement over UFS) are the various improvements such as small
> file storage and large directory storage. I know that I'm interested
> in seeing performance comparisons with regard to these factors, and
> so far, I've seen none that compare ReiserFS to UFS/softupdates.
>
> My $.02
>
> -Bill
>

As I indicated, my experience with ReiserFS is limited.  I have been
using SGI's XFS for Linux for a short time and am quite pleased with its
performance and filesystem integrity.  If you have a spare system
available, it's easy enough to try it out.

http://oss.sgi.com/projects/xfs/1.0_installer.html

Also requires the 2 install CDs for RedHat-7.1.

SGI's ftp site has a patch available to bring up XFS on Linux-2.4.5
kernel, also.




To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2001-07-07 Thread Juha Saarinen

On Sat, 7 Jul 2001, Dave Uhring wrote:

> I use "logging" on Solaris and XFS on Linux and have tried reiserfs on
> Linux.  All are superior to UFS/softupdates when the going gets tough.
> Disk access times may or may not be comparable with UFS/softupdates, but
> the integrity of my filesystems is more important than raw speed.

Hmmm... that's one reason I've not implemented ReiserFS on my Linux box.
Read too many horror stories about how it eats your file system, and how
it doesn't work with NFS etc.



-- 
Regards,


Juha

PGP fingerprint:
B7E1 CC52 5FCA 9756 B502  10C8 4CD8 B066 12F3 9544


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2001-07-07 Thread Bill Moran

Dave Uhring wrote:
> You seem to have missed the critical point of that paper.  When the
> system goes completely haywire and either crashes or locks up so hard
> that a manual reset is required, UFS/softupdates requires a substantial
> amount of time to run fsck.  If you have a very large filesystem, you
> then have to wait until fsck completes.  And if you are
> lucky, it will not terminate with the suggestion that you run fsck by
> hand.  With a true journalling filesystem this wait is obviated.  The
> last transactions are rerun or truncated and the system boots up.

Actually ... according to the article, the system boots up and _then_
determines what needs done to repair the filesystem.

Also, the lack of a need for fscking is not the only benefit of
RieserFS. In fact, it's a _minor_ improvement. If your system is
going down so often that the speed of a fsck is a major factor in the
layout of the system, you've got other issues you need to address
first!
The other issues that might make Reiserfs a good idea (and a possible
improvement over UFS) are the various improvements such as small
file storage and large directory storage. I know that I'm interested
in seeing performance comparisons with regard to these factors, and
so far, I've seen none that compare ReiserFS to UFS/softupdates.

My $.02

-Bill

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



RE: JFS

2001-07-07 Thread A. L. Meyers

Hi again!

I'd like to see continuing cooperation amoung the various more or
less open source Unixes and their clones. If anyone is capable
and competent to port ReiserFS to FreeBSD, it would be Hans
Reiser himself and his friends.

Which is not at all stating or implying that UFS is not a very
good FS. But what Robert says is simply a statement of fact.
Experiencing one-on how quickly a JFS (e. g. Reiser) recovers
is really impressive.

Good night.

Lucien


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS (was: The FreeBSD core team needs your help)

2001-07-05 Thread Greg Lehey

On Wednesday,  4 July 2001 at 11:38:08 +0100, Antony T Curtis wrote:
> Greg Lehey wrote:
>>
>> On Tuesday, 12 June 2001 at 19:22:45 +0200, Steve O'Hara-Smith wrote:
>>> On Tue, 12 Jun 2001 12:09:58 +0100
>>> Josef Karthauser <[EMAIL PROTECTED]> wrote:
>>>
>>>> On Fri, Jun 08, 2001 at 08:32:23AM -0700, Eric Parusel wrote:
>>>>>>> A journalling FS for those people who just hate waiting for a
>>>>> couple
>>>>>>> of
>>>>>>> TB of slow disks to fsck?
>>>>>>
>>>>>> Does ReiserFS work with FreeBSD?
>>>>>
>>>>>> From what I've read, XFS is quite good as well  (Whether or not it
>>>>> could ever work with *BSD, I don't know)
>>>>
>>>> Apparently XFS would run better on FreeBSD than on Linux, from what
>>>
>>>   Whatever happened to the open source release of JFS, or is JFS really
>>> bad ?
>>
>> The open source version of JFS was based on OS/2, not AIX.  It's not
>> an overly good fit to UNIX.
>
> That was only because AIX's JFS implementation was so closely bound into
> their kernel that there was no easy way to "port" it out of it. Also,
> AFAIK, it was written in a mesh of different languages too, including
> POWER architecture assembly.
>
> The OS/2 version was the first clean implementation to plug into OS/2's
> IFS driver model - and being written in C, it is much more 'portable'.
> (AFAIK, it was supposed to be able to be recompiled for OS/2 for CHRP
> PowerPC)

Since writing that (quite some time ago, IIRC) I have joined IBM and
am now working with the people who did the JFS port.  They
substantially confirm your viewpoint, with the added information that
the "old" JFS, now called JFS 1, is being phased out under AIX, and
the "new" AIX JFS, JFS 2, is based on the same code base as the OS/2
port.  With that background, IBM's approach makes a lot more sense.
It's a pity that this issue wasn't clarified earlier.

> All said, I would be interested in a JFS port for FreeBSD 

I'm going to be doing a lot of work on JFS in the next few months.  I
don't think I'll port it to FreeBSD, but I'll be available for
questions, and I'll have a better understanding.


>> unix soit qui mal y pense

You're aware that the original word of this phrase, "hon(n)i", means
"ashamed"?

Greg
--
See complete headers for address and phone numbers

To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



RE: More than just logging, Re: JFS

2000-01-31 Thread Steve O'Hara-Smith

> Why logging filesystems don't work:
> 
> You generally (with the hardware available in PCs now) 
> can't tell the difference between:
> 
> 1) loss of power (ok!)
> 2) crash where the filesystem datastructures weren't corrupted (ok!)
> 3) crash where the filesystem datastructures were corrupted (ouch)
> 4) crash where the disk/bus got scrambled (ouch)

Nice analysis. I have a vague memory (I can't seem to find the message
in the archives) of somebody on -current talking about a proposal for a log
device that emitted events from the filesystem. Something like that and a WORM
might make a reliable logging filesystem.



To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: More than just logging, Re: JFS

2000-01-31 Thread Daniel C. Sobral

Alfred Perlstein wrote:
> 
> Bad news: people running depending on _only_ logging are kidding themselves.

Yeah. Funny how they have been doing so without getting bitten, eh?

--
Daniel C. Sobral(8-DCS)
[EMAIL PROTECTED]
[EMAIL PROTECTED]

"If you consider our help impolite, you should see the manager."


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2000-01-31 Thread Daniel C. Sobral

Cy Schubert wrote:
> 
> > > Due to the lack of interest, FreeBSD's LFS has fallen into disrepair
> > > over the years.  With the implementation of softupdates in FreeBSD I
> > > don't think there is any need for LFS any more.
> >
> > Repeat that over and over the next time you wait fsck finish a 40 Gb
> > filesystem checkup, and see if you manage to convince yourself of that.
> 
> Read the CVS logs and tell me how you could interpret the comments any
> differently:
> 
> 1.21 Fri Jan 30 11:33:40 1998 UTC by phk
> CVS Tags: HEAD
> Diffs to 1.20
> FILE REMOVED
> 
> Retire LFS.
> 
> If you want to play with it, you can find the final version of the
> code in the repository the tag LFS_RETIREMENT.
> 
> If somebody makes LFS work again, adding it back is certainly
> desireable, but as it is now nobody seems to care much about it,
> and it has suffered considerable bitrot since its somewhat haphazard
> integration.
> 
> R.I.P
> 
> Obviously you don't know what you're talking about either.  Do you?

I do know. The main reason why LFS was never updated isn't that it was
made obsolete by softupdates, as claimed above, but that it was made
obsolete by JFS. Why work on LFS if it is not up to a JFS?
Unfortunately, the people who have to suffer enourmous waits after
crashes usually have way more to do, even if they have the skills to fix
LFS.

With the disks getting bigger and bigger, this is due to change.

BTW, NetBSD is happy with _their_ _functional_ LFS.

--
Daniel C. Sobral(8-DCS)
[EMAIL PROTECTED]
[EMAIL PROTECTED]

"If you consider our help impolite, you should see the manager."


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



Re: JFS

2000-01-31 Thread Cy Schubert

In message <[EMAIL PROTECTED]>, "Daniel C. Sobral" writes:
> Cy Schubert - ITSD Open Systems Group wrote:
> > 
> > 4.4BSD has something like JFS, LFS (Log Structured Filesystem).  LFS
> > developed from a paper by John Ousterhout, the same fellow who
> > developed Tcl & Tk.  All other log structured filesystems, JFS, AdvFS,
> > Veritas Filesystem, are based on Ousterhout's work.
> 
> A journalled structured filesystem is NOT a log structured filesystem.

I stand corrected on this point.

> 
> > Due to the lack of interest, FreeBSD's LFS has fallen into disrepair
> > over the years.  With the implementation of softupdates in FreeBSD I
> > don't think there is any need for LFS any more.
> 
> Repeat that over and over the next time you wait fsck finish a 40 Gb
> filesystem checkup, and see if you manage to convince yourself of that.

Read the CVS logs and tell me how you could interpret the comments any 
differently:

1.21 Fri Jan 30 11:33:40 1998 UTC by phk 
CVS Tags: HEAD
Diffs to 1.20 
FILE REMOVED 

Retire LFS.

If you want to play with it, you can find the final version of the
code in the repository the tag LFS_RETIREMENT.

If somebody makes LFS work again, adding it back is certainly
desireable, but as it is now nobody seems to care much about it,
and it has suffered considerable bitrot since its somewhat haphazard
integration.

R.I.P


Obviously you don't know what you're talking about either.  Do you?


Regards,   Phone:  (250)387-8437
Cy Schubert  Fax:  (250)387-5766
Sun/DEC Team, UNIX GroupInternet:  [EMAIL PROTECTED]
ITSD
Province of BC
"COBOL IS A WASTE OF CARDS."




To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message



More than just logging, Re: JFS

2000-01-31 Thread Alfred Perlstein

* Gary Palmer <[EMAIL PROTECTED]> [000130 22:55] wrote:
> Tom wrote in message ID
> <[EMAIL PROTECTED]>:
> > On Sun, 30 Jan 2000, Daniel C. Sobral wrote:
> > 
> > > > Due to the lack of interest, FreeBSD's LFS has fallen into disrepair
> > > > over the years.  With the implementation of softupdates in FreeBSD I
> > > > don't think there is any need for LFS any more.
> > > 
> > > Repeat that over and over the next time you wait fsck finish a 40 Gb
> > > filesystem checkup, and see if you manage to convince yourself of that.
> > 
> >   Actually, one of the goals of the softupdates development is a fsck'less
> > filesystem.  I'm not sure how this is to be achieved.  Probably a metadata
> > journal, though that is just speculation.  All the work on metadata update
> > ordering in softupdates would probably apply very nicely to a journal.
> 
> The way I understand it is that SoftUpdates is meant to leave the
> metadata consistant enough that the filesystem can be mounted
> read/write immediately at boot, and then have a background fsck go
> through and remove blocks which are allocated in the bitmaps, but
> aren't really used.  The only thing you lose by not running the
> background daemon is space.  I don't think anyone's running like this
> today, but that is Kirks plan.

(This is me (hopefully accurately) repeating something that Terry
Lambert brought to my attention, credit where credit is due.  The
safe-shutdown is my rambling on about a solution.)

Why logging filesystems don't work:

You generally (with the hardware available in PCs now) 
can't tell the difference between:

1) loss of power (ok!)
2) crash where the filesystem datastructures weren't corrupted (ok!)
3) crash where the filesystem datastructures were corrupted (ouch)
4) crash where the disk/bus got scrambled (ouch)

the problems with 3 and 4 really make a logging filesystem a
"shot in the dark" because you never know if "safe" areas on the
disk really are safe because there was a chance for corruption.

  How do you know that a bug in some other code didn't trounce on
  the filesystem's data and write to a spot that's supposedly
  "committed"?

Since you generally can't tell between 1,2,3 and 4 you really ought
to fsck in the background anyway.

Good news: afaik Kirk is working on this capability!
Bad news: people running depending on _only_ logging are kidding themselves.

Mini-fix for this problem:
Providing safe shutdown points for the filesystem, if something
"weird" is found, disable further access to that file/directory or
possibly shutdown the entire filesystem.

Basically instead of panic'ing when invalid structures are read from
disk, just disallow further access to the objects.

patches anyone? :)

-- 
-Alfred Perlstein - [[EMAIL PROTECTED]|[EMAIL PROTECTED]]


To Unsubscribe: send mail to [EMAIL PROTECTED]
with "unsubscribe freebsd-stable" in the body of the message