squashfs size in statfs

squashfs size in statfs

[Jan Engelhardt]

Hello list,


# l /mnt
total 36293
drwxr-xr-x   2 root root       20 Jun  5 11:50 .
drwxr-xr-x  31 root root     4096 Jun  5  2006 ..
-rw-r--r--   1 root root 37158912 Jun  5 11:06 mem
# df
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/shm/sc.sqfs         26688     26688         0 100% /mnt
# l sc.sqfs
-rwx------  1 jengelh users 27279360 Jun  5 11:50 sc.sqfs

I think statfs() should show the uncompressed size, no?



Jan Engelhardt
-- 

Re: squashfs size in statfs

[H. Peter Anvin]

Followup to:  <Pine.LNX.4.61.0606051243100.579@yvahk01.tjqt.qr>
By author:    Jan Engelhardt <jengelh@linux01.gwdg.de>
In newsgroup: linux.dev.kernel
>
> Hello list,
> 
> 
> # l /mnt
> total 36293
> drwxr-xr-x   2 root root       20 Jun  5 11:50 .
> drwxr-xr-x  31 root root     4096 Jun  5  2006 ..
> -rw-r--r--   1 root root 37158912 Jun  5 11:06 mem
> # df
> Filesystem           1K-blocks      Used Available Use% Mounted on
> /dev/shm/sc.sqfs         26688     26688         0 100% /mnt
> # l sc.sqfs
> -rwx------  1 jengelh users 27279360 Jun  5 11:50 sc.sqfs
> 
> I think statfs() should show the uncompressed size, no?
> 

No.

	-hpa

Re: squashfs size in statfs

[Jan Engelhardt]

>> Hello list,
>> 
>> 
>> # l /mnt
>> total 36293
>> drwxr-xr-x   2 root root       20 Jun  5 11:50 .
>> drwxr-xr-x  31 root root     4096 Jun  5  2006 ..
>> -rw-r--r--   1 root root 37158912 Jun  5 11:06 mem
>> # df
>> Filesystem           1K-blocks      Used Available Use% Mounted on
>> /dev/shm/sc.sqfs         26688     26688         0 100% /mnt
>> # l sc.sqfs
>> -rwx------  1 jengelh users 27279360 Jun  5 11:50 sc.sqfs
>> 
>> I think statfs() should show the uncompressed size, no?
>
>No.
>
Yes, because CRAM does it that way, and maybe zisofs does it too:

11:50 shanghai:/dev/shm # mount out.cram /mnt -oloop,ro
11:50 shanghai:/dev/shm # df
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/shm/out.cram        14336     14336         0 100% /mnt
11:50 shanghai:/dev/shm # l out.cram
-rw-r--r--  1 root root 110592 Jun 15 11:50 out.cram



Jan Engelhardt
-- 

Re: squashfs size in statfs

[Phillip Lougher]

On 6/15/06, Jan Engelhardt <jengelh@linux01.gwdg.de> wrote:
> >> Hello list,
> >>
> >>
> >> # l /mnt
> >> total 36293
> >> drwxr-xr-x   2 root root       20 Jun  5 11:50 .
> >> drwxr-xr-x  31 root root     4096 Jun  5  2006 ..
> >> -rw-r--r--   1 root root 37158912 Jun  5 11:06 mem
> >> # df
> >> Filesystem           1K-blocks      Used Available Use% Mounted on
> >> /dev/shm/sc.sqfs         26688     26688         0 100% /mnt
> >> # l sc.sqfs
> >> -rwx------  1 jengelh users 27279360 Jun  5 11:50 sc.sqfs
> >>
> >> I think statfs() should show the uncompressed size, no?
> >
> >No.
> >
> Yes, because CRAM does it that way, and maybe zisofs does it too:

Zisofs doesn't (H. Peter Anvin should know as he wrote it :-) ).

root@pierrot:/# ls -la dir.iso
-rw-r--r--  1 root root 366592 2006-06-16 22:41 dir.iso
root@pierrot:/# mount -t iso9660 dir.iso /mnt -o loop
root@pierrot:/# df /mnt
Filesystem           1K-blocks      Used Available Use% Mounted on
/dir.iso                   358       358         0 100% /mnt
root@pierrot:/# ls -la /mnt
total 13
drwxr-xr-x   2 root root     2048 2006-06-16 22:41 .
drwxr-xr-x  32 root root     4096 2006-06-16 22:56 ..
-rw-r--r--   1 root root 51200000 2006-06-16 22:40 zero

Statfs should return the size of the filesystem, not the amount of
data the filesystem  represents.  In this respect the behaviour of
Squashfs and Zisofs is correct.

This is analogous to performing stat on a gzipped file.  The stat
returns the size of the compressed file, not the uncompressed size.

Phillip

Re: squashfs size in statfs

[H. Peter Anvin]

Phillip Lougher wrote:
>> >
>> Yes, because CRAM does it that way, and maybe zisofs does it too:
> 
> Zisofs doesn't (H. Peter Anvin should know as he wrote it :-) ).
> 
> root@pierrot:/# ls -la dir.iso
> -rw-r--r--  1 root root 366592 2006-06-16 22:41 dir.iso
> root@pierrot:/# mount -t iso9660 dir.iso /mnt -o loop
> root@pierrot:/# df /mnt
> Filesystem           1K-blocks      Used Available Use% Mounted on
> /dir.iso                   358       358         0 100% /mnt
> root@pierrot:/# ls -la /mnt
> total 13
> drwxr-xr-x   2 root root     2048 2006-06-16 22:41 .
> drwxr-xr-x  32 root root     4096 2006-06-16 22:56 ..
> -rw-r--r--   1 root root 51200000 2006-06-16 22:40 zero
> 
> Statfs should return the size of the filesystem, not the amount of
> data the filesystem  represents.  In this respect the behaviour of
> Squashfs and Zisofs is correct.
> 
> This is analogous to performing stat on a gzipped file.  The stat
> returns the size of the compressed file, not the uncompressed size.
> 

A better analogy is it is like statting a sparse file on, say, an ext3 filesystem.  stat 
(ls -s) and statfs report the amount of storage consumed.

	-hpa

Re: squashfs size in statfs

[Jan Engelhardt]

>> > Yes, because CRAM does it that way, and maybe zisofs does it too:
>> 
>> Zisofs doesn't (H. Peter Anvin should know as he wrote it :-) ).
>> 
>> Statfs should return the size of the filesystem, not the amount of
>> data the filesystem  represents.  In this respect the behaviour of
>> Squashfs and Zisofs is correct.
>> 
>> This is analogous to performing stat on a gzipped file.  The stat
>> returns the size of the compressed file, not the uncompressed size.
>
> A better analogy is it is like statting a sparse file on, say, an ext3
> filesystem.  stat (ls -s) and statfs report the amount of storage consumed.
>
Too bad. Would have been a good indicator of how much space you need to 
copy an entire disc to disk, rather than running and waiting for du -s to 
complete.

So cramfs would need fixing.


Jan Engelhardt
--