Tux

...making Linux just a little more fun!

BLOCKSIZE unset by default

Mahesh Aravind [ra_mahesh at yahoo.com]


Sat, 21 Jun 2008 08:25:02 -0700 (PDT)

Dear TAG,

I was playing around the command line (learning), and I came across df(1) reporting block size in "1k-blocks". But my dumpe2fs(8) says the block size is 4096 (4K). Shouldn't it be doing thus by default?

It seems this is filed as a Bug under Ubuntu (Bug #180415).

One suggestion is for the install program to calculate the blocksize at install time, and put it somewhere safe (immutable). Like 'export BLOCKSIZE=<whatever>' in /etc/profile.

I also saw that adding ' (apostrophe) before the block size will yield you a digit separator -- cool, eh?

I did:

BLOCKSIZE="'4096" ls -l
           ^ <- see this?

and it gave me size figures separated by commas! :)

My $LANG is en_IN.UTF-8

YMMV

Regards,

Mahesh Aravind


Top    Back


Thomas Adam [thomas.adam22 at gmail.com]


Sat, 21 Jun 2008 16:35:56 +0100

2008/6/21 Mahesh Aravind <ra_mahesh@yahoo.com>:

> I did:
> BLOCKSIZE="'4096" ls -l
> and it gave me size figures separated by commas! :)

Not on my Debian Etch install it didn't. Ah well. Nice thing to know I suppose.

-- Thomas Adam


Top    Back


Ben Okopnik [ben at linuxgazette.net]


Sat, 21 Jun 2008 09:08:49 -0700

On Sat, Jun 21, 2008 at 04:35:56PM +0100, Thomas Adam wrote:

> 2008/6/21 Mahesh Aravind <ra_mahesh@yahoo.com>:
> > I did:
> > BLOCKSIZE="'4096" ls -l
> > and it gave me size figures separated by commas! :)
> 
> Not on my Debian Etch install it didn't.  Ah well.  Nice thing to know
> I suppose.

Worked for me:

ben@Tyr:~$ df 
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/hda1             56214364  39826312  13532500  75% /
varrun                  190748       144    190604   1% /var/run
varlock                 190748         4    190744   1% /var/lock
procbususb              190748       108    190640   1% /proc/bus/usb
udev                    190748       108    190640   1% /dev
devshm                  190748         0    190748   0% /dev/shm
ben@Tyr:~$ BLOCKSIZE="'4096" df
Filesystem           4K-blocks      Used Available Use% Mounted on
/dev/hda1            14,053,591 9,956,578 3,383,125  75% /
varrun                  47,687        36    47,651   1% /var/run
varlock                 47,687         1    47,686   1% /var/lock
procbususb              47,687        27    47,660   1% /proc/bus/usb
udev                    47,687        27    47,660   1% /dev
devshm                  47,687         0    47,687   0% /dev/shm

The magic word is 'Ubuntu', I suppose. Although that seems odd, for something this basic.

-- 
* Ben Okopnik * Editor-in-Chief, Linux Gazette * http://LinuxGazette.NET *


Top    Back


Mahesh Aravind [ra_mahesh at yahoo.com]


Sat, 21 Jun 2008 10:00:02 -0700 (PDT)

Thomas,

> 2008/6/21 Mahesh Aravind <ra_mahesh@yahoo.com>:
> 
> Not on my Debian Etch install it didn't.  Ah well. 
> Nice thing to know
> I suppose.

That would depend very much on your locale...

I think mine's set at LANG=en_IN.UTF-8

It's the locale that determines how the digits are to be separated. If your is "C", then no donuts for you. :D

Try something like.. LANG=en_GB.UTF-8 BLOCKSIZE="'4096" ls -l...

Hey, I also notice that the ".UTF-8" part DOES indeed makes the difference. Any idea why?

LANG=en_GB is NOT the same as LANG=en_GB.UTF-8

Wonderingly,

Mahesh Aravind


Top    Back


Mahesh Aravind [ra_mahesh at yahoo.com]


Sat, 21 Jun 2008 10:17:30 -0700 (PDT)

Ben,

> They magic word is 'Ubuntu', I suppose. Although
> that seems odd, for
> something this basic.
> 

No.. it's coreutils..

The way they handle locale, blocksize etc are different...

Trying manipulating the LANG/LC_ALL vars..

-- Mahesh Aravind


Top    Back


Ben Okopnik [ben at linuxgazette.net]


Wed, 25 Jun 2008 11:11:39 -0400

On Sat, Jun 21, 2008 at 10:00:02AM -0700, Mahesh Aravind wrote:

> > 2008/6/21 Thomas wrote:
> > 
> > Not on my Debian Etch install it didn't.  Ah well. 
> > Nice thing to know
> > I suppose.
> 
> That would depend very much on your locale...
> 
> I think mine's set at LANG=en_IN.UTF-8
> 
> It's the locale that determines how the digits are to be separated.
> If your is "C", then no donuts for you. :D
> 
> Try something like.. LANG=en_GB.UTF-8 BLOCKSIZE="'4096" ls -l...
> 
> Hey, I also notice that the ".UTF-8" part DOES indeed makes the
> difference. Any idea why?
> 
> LANG=en_GB is NOT the same as LANG=en_GB.UTF-8
> 
> Wonderingly,
> Mahesh Aravind

It's a different charmap for that locale, thus the different result. Different $LC_COLLATE, etc. One is configured to separate numbers that way, the other one isn't. Frankly, I'm a bit puzzled at your puzzlement. :)

-- 
* Ben Okopnik * Editor-in-Chief, Linux Gazette * http://LinuxGazette.NET *


Top    Back


Mahesh Aravind [ra_mahesh at yahoo.com]


Wed, 25 Jun 2008 10:50:16 -0700 (PDT)

Ben,

--- On Wed, 6/25/08, Ben Okopnik <ben@linuxgazette.net> wrote:

> It's a different charmap for that locale, thus the
> different result.
> Different $LC_COLLATE, etc. One is configured to separate
> numbers that
> way, the other one isn't. Frankly, I'm a bit
> puzzled at your puzzlement.
> :)
> 

Oh.. I didn't know that ".UTF-8" had such a profound change. But then, why isn't Thomas Adam not getting the commas?

But the purpose of this thread was to know WHY is BLOCKSIZE not set by distros. Different partioning software format the filesystem using different block sizes.

df(1), du(1) etc. automatically assumes that the disk is using 1K blocksize. Why isn't somebody doing something about it?

Regards,

Mahesh Aravind


Top    Back


Ben Okopnik [ben at linuxgazette.net]


Wed, 25 Jun 2008 14:05:43 -0400

On Wed, Jun 25, 2008 at 10:50:16AM -0700, Mahesh Aravind wrote:

> Ben,
> 
> --- On Wed, 6/25/08, Ben Okopnik <ben@linuxgazette.net> wrote:
> 
> > It's a different charmap for that locale, thus the
> > different result.
> > Different $LC_COLLATE, etc. One is configured to separate
> > numbers that
> > way, the other one isn't. Frankly, I'm a bit
> > puzzled at your puzzlement.
> > :)
> > 
> 
> Oh.. I didn't know that ".UTF-8" had such a profound change. But then,
> why isn't Thomas Adam not getting the commas?

Perhaps because his locale isn't set to one that uses commas to separate numbers that way?

> But the purpose of this thread was to know WHY is BLOCKSIZE not set by
> distros.  Different partioning software format the filesystem using
> different block sizes.

I'd guess it's because it's an optional feature - not everybody wants to see their data that way, but it's available if you do.

> df(1), du(1) etc. automatically assumes that the disk is using 1K
> blocksize.  Why isn't somebody doing something about it?

Again, why should they? If I'm used to seeing the output of 'df' in 1k blocks, and somebody arbitrarily and suddenly changed it, I'd be pretty annoyed. Doing it the other way - i.e., sticking with what already exists and providing options to do it differently - is a) infinitely expandable and b) not likely to annoy anyone/generate help requests. It makes a lot of sense to me.

-- 
* Ben Okopnik * Editor-in-Chief, Linux Gazette * http://LinuxGazette.NET *


Top    Back