Re[2]: Corrupted ASFS patch.

Re[2]: Corrupted ASFS patch.

[Pavel Fedin]

Hello Phillip,

Thursday, February 12, 2009, 7:38:04 PM, you wrote:

> Yes, replace all uXX with __beXX in on-disk structures.

 Done. Take it.

-- 
Best regards,
 Pavel                            mailto:sonic.amiga@gmail.com

Corrupted ASFS patch.

[Pavel Fedin]

 Ouch, really sorry. Unfortunately i have Internet access only at work
and they block all ports except HTTP, FTP and (luckily) POP3. So i can
send mail only via web interface.
 Documentation says that i can also place the file somewhere on the
Web and provide a link. Here is it:
http://rapidshare.com/files/195946803/asfs-1.0b12_patch_2.6.24.diff.html
 As to endianess signing, sorry, i can't help here because actually
i'm not the filesystem's author. The author is Marek Szyprowski
<march@staszic.waw.pl>, he also tried to post this patch some time
ago, but gave up because someone said that SFS patch won't be accepted
because "Linux doesn't need one more exotic FS".
 I just adapted the driver to 2.6.24 kernel (this patch works on
2.6.28 without changes).
 You can contact him about the details.

Re: Corrupted ASFS patch.

[maximilian attems]

On Mon, Feb 09, 2009 at 04:18:39PM +0300, Pavel Fedin wrote:
>  Ouch, really sorry. Unfortunately i have Internet access only at work
> and they block all ports except HTTP, FTP and (luckily) POP3. So i can
> send mail only via web interface.
>  Documentation says that i can also place the file somewhere on the
> Web and provide a link. Here is it:
> http://rapidshare.com/files/195946803/asfs-1.0b12_patch_2.6.24.diff.html
>  As to endianess signing, sorry, i can't help here because actually
> i'm not the filesystem's author. The author is Marek Szyprowski
> <march@staszic.waw.pl>, he also tried to post this patch some time
> ago, but gave up because someone said that SFS patch won't be accepted
> because "Linux doesn't need one more exotic FS".
>  I just adapted the driver to 2.6.24 kernel (this patch works on
> 2.6.28 without changes).
>  You can contact him about the details.

a MIA fs dev is not a bonus point in order to get something merged.

as debian linux-2.6 maintainer we used to have this fs, but dropped as
the author showed no will or sign to go upstream.
the small but important critic does need to get worked on.
so if you wana help to get this somewhere get the fs fixed and
properly send in patches.

kind regards

-- 
maks

Re[2]: Corrupted ASFS patch.

[Pavel Fedin]

Hello maximilian,

Monday, February 9, 2009, 4:31:20 PM, you wrote:

> as debian linux-2.6 maintainer we used to have this fs, but dropped as
> the author showed no will or sign to go upstream.
> the small but important critic does need to get worked on.

 Hm, he told me the opposite thing...
 Well, at least I am interested in merging it into the kernel. Well,
i'll try to get things fixed (however i don't have much time to do
it).

> so if you wana help to get this somewhere get the fs fixed and
> properly send in patches.

 First, i've worked around my company's firewall and now i'm at last able to
use e-mail client for sending messages again. :-)
 Second, sorry for re-posting the patch (it was still laying in The
Bat's outbox and when i pressed 'Send mail' it was succesfully sent.
Well, at least this would let you to review the patch better...
 Third, this is the question fot Matthew, about annotating endianess
in on-disk structures. What if the filesystem is bi-endian? Original
SFS (http://sourceforge.net/projects/smartfilesystem) is bi-endian. I
still don't know how much of support for little-endian version is implemented
in Linux version.

-- 
Best regards,
 Pavel                            mailto:sonic.amiga@gmail.com

Re: Corrupted ASFS patch.

[Phillip Lougher]

Pavel Fedin wrote:

>  Third, this is the question fot Matthew, about annotating endianess
> in on-disk structures. What if the filesystem is bi-endian? Original
> SFS (http://sourceforge.net/projects/smartfilesystem) is bi-endian. I
> still don't know how much of support for little-endian version is implemented
> in Linux version.
> 

This issue has come up before (over Squashfs which used to be
'bi-endian').  The general consensus is that all Linux filesystems
should be one endianness only, as it potentially simplifies code and
makes it more efficient (i.e. the swap code can be conditionally
compiled).

As SFS is not a native Linux filesystem you could argue that the
bi-endian layout is a legacy feature over which you have no control.
However, unless there are a lot of little-endian SFS filesystems
out there, supporting them is likely to give you more pain than
it's worth, and it will present an additional barrier to mainlining.

Phillip

Re[2]: Corrupted ASFS patch.

[Pavel Fedin]

Hello Phillip,

Thursday, February 12, 2009, 12:42:42 PM, you wrote:

> As SFS is not a native Linux filesystem you could argue that the
> bi-endian layout is a legacy feature over which you have no control.
> However, unless there are a lot of little-endian SFS filesystems
> out there, supporting them is likely to give you more pain than
> it's worth, and it will present an additional barrier to mainlining.

 I've asked Michal Schulz at AROS IRC channel, he told me that
little-endian version was introduced as experimental and it did not
gave any significant speedup (just 1%). This ended its life - support
for little-endian SFS exists only in theory, it is not adviced for
production use, not supported by formatting and installation tools,
etc.
 So far the issue is closed. SFS is bigendian. End of story.
 Do i understand right that i just need to replace all 'uXX' with
'beXX' in on-disk structures definition ?


-- 
Best regards,
 Pavel                            mailto:sonic.amiga@gmail.com