fs/Kconfig question regarding CONFIG_BLOCK

fs/Kconfig question regarding CONFIG_BLOCK

[Jan Engelhardt]

(this time hopefully with a working To/Cc)
Hi,



fs/Kconfig has:

if BLOCK
menu "DOS/FAT/NT Filesystems"
...
<stuff here>
...
endmenu
endif


Why is it wrapped into BLOCK, or, why are all of the other filesystems 
which require a block device?


	-`J'
-- 

Re: fs/Kconfig question regarding CONFIG_BLOCK

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 599 bytes --]

On Tue, 17 Oct 2006 20:41:41 +0200, Jan Engelhardt said:

> fs/Kconfig has:
> 
> if BLOCK
> menu "DOS/FAT/NT Filesystems"

> Why is it wrapped into BLOCK, or, why are all of the other filesystems 
> which require a block device?

Some filesystems (such as /proc, /sys, and so on - basicaly, the "pseudo" file
systems) are able to stand by themselves.  Filesystems that read actual blocks
of data off actual media will require the services of the block layer to do
that.  So if you've built a tiny embedded kernel that doesn't include the block
layer, you can't read those sorts of filesystems....



[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: fs/Kconfig question regarding CONFIG_BLOCK

[Jan Engelhardt]

>> fs/Kconfig has:
>> 
>> if BLOCK
>> menu "DOS/FAT/NT Filesystems"
>
>> Why is it wrapped into BLOCK, or, why are all of the other filesystems 
>> which require a block device?
>
>Some filesystems (such as /proc, /sys, and so on - basicaly, the "pseudo" file
>systems) are able to stand by themselves.  Filesystems that read actual blocks
>of data off actual media will require the services of the block layer to do
>that.  So if you've built a tiny embedded kernel that doesn't include the block
>layer, you can't read those sorts of filesystems....

Never mind, I see that some filesystems have 'depends on BLOCK' instead 
of being wrapped into if BLOCK. Not really consistent but whatever.


	-`J'
-- 

Re: fs/Kconfig question regarding CONFIG_BLOCK

[Jens Axboe]

On Tue, Oct 17 2006, Jan Engelhardt wrote:
> 
> >> fs/Kconfig has:
> >> 
> >> if BLOCK
> >> menu "DOS/FAT/NT Filesystems"
> >
> >> Why is it wrapped into BLOCK, or, why are all of the other filesystems 
> >> which require a block device?
> >
> >Some filesystems (such as /proc, /sys, and so on - basicaly, the "pseudo" file
> >systems) are able to stand by themselves.  Filesystems that read actual blocks
> >of data off actual media will require the services of the block layer to do
> >that.  So if you've built a tiny embedded kernel that doesn't include the block
> >layer, you can't read those sorts of filesystems....
> 
> Never mind, I see that some filesystems have 'depends on BLOCK' instead 
> of being wrapped into if BLOCK. Not really consistent but whatever.

Feel free to send in patches that make things more consistent.

-- 
Jens Axboe

Re: fs/Kconfig question regarding CONFIG_BLOCK

[Jan Engelhardt]

>> Never mind, I see that some filesystems have 'depends on BLOCK' instead 
>> of being wrapped into if BLOCK. Not really consistent but whatever.
>
>Feel free to send in patches that make things more consistent.

How would you like things? if BLOCK or depends on BLOCK?
Does menuconfig/oldconfig/etc. parse the whole config structure faster 
it it done either way?

	-`J'
-- 

Re: fs/Kconfig question regarding CONFIG_BLOCK

[Jens Axboe]

On Tue, Oct 17 2006, Jan Engelhardt wrote:
> >> Never mind, I see that some filesystems have 'depends on BLOCK' instead 
> >> of being wrapped into if BLOCK. Not really consistent but whatever.
> >
> >Feel free to send in patches that make things more consistent.
> 
> How would you like things? if BLOCK or depends on BLOCK?

Well, if you can hide an entire block with if BLOCK, then that would be
preferred. Otherwise depends on BLOCK.

> Does menuconfig/oldconfig/etc. parse the whole config structure faster 
> it it done either way?

I'd be surprised if if BLOCK wasn't faster over, say, 10 depends on
BLOCK.

-- 
Jens Axboe

Re: fs/Kconfig question regarding CONFIG_BLOCK

[Randy Dunlap]

On Wed, 18 Oct 2006 09:09:22 +0200 Jens Axboe wrote:

> On Tue, Oct 17 2006, Jan Engelhardt wrote:
> > >> Never mind, I see that some filesystems have 'depends on BLOCK' instead 
> > >> of being wrapped into if BLOCK. Not really consistent but whatever.
> > >
> > >Feel free to send in patches that make things more consistent.
> > 
> > How would you like things? if BLOCK or depends on BLOCK?
> 
> Well, if you can hide an entire block with if BLOCK, then that would be
> preferred. Otherwise depends on BLOCK.
> 
> > Does menuconfig/oldconfig/etc. parse the whole config structure faster 
> > it it done either way?
> 
> I'd be surprised if if BLOCK wasn't faster over, say, 10 depends on
> BLOCK.

Jens,
Has anyone looked at what BLOCK=n does to mm/page-writeback.c ?
It calls blk_congestion_end(), which isn't there.

mm/built-in.o: In function `writeback_congestion_end':
(.text.writeback_congestion_end+0xc): undefined reference to `blk_congestion_end'
make: *** [.tmp_vmlinux1] Error 1
Command exited with non-zero status 2


---
~Randy

Re: fs/Kconfig question regarding CONFIG_BLOCK

[Jens Axboe]

On Wed, Oct 18 2006, Randy Dunlap wrote:
> On Wed, 18 Oct 2006 09:09:22 +0200 Jens Axboe wrote:
> 
> > On Tue, Oct 17 2006, Jan Engelhardt wrote:
> > > >> Never mind, I see that some filesystems have 'depends on BLOCK' instead 
> > > >> of being wrapped into if BLOCK. Not really consistent but whatever.
> > > >
> > > >Feel free to send in patches that make things more consistent.
> > > 
> > > How would you like things? if BLOCK or depends on BLOCK?
> > 
> > Well, if you can hide an entire block with if BLOCK, then that would be
> > preferred. Otherwise depends on BLOCK.
> > 
> > > Does menuconfig/oldconfig/etc. parse the whole config structure faster 
> > > it it done either way?
> > 
> > I'd be surprised if if BLOCK wasn't faster over, say, 10 depends on
> > BLOCK.
> 
> Jens,
> Has anyone looked at what BLOCK=n does to mm/page-writeback.c ?
> It calls blk_congestion_end(), which isn't there.
> 
> mm/built-in.o: In function `writeback_congestion_end':
> (.text.writeback_congestion_end+0xc): undefined reference to `blk_congestion_end'
> make: *** [.tmp_vmlinux1] Error 1
> Command exited with non-zero status 2

Yeah currently known, with Andrew's latest we should be getting closer.

-- 
Jens Axboe

Re: fs/Kconfig question regarding CONFIG_BLOCK

[Horst H. von Brand]

Jens Axboe <jens.axboe@oracle.com> wrote:
> On Tue, Oct 17 2006, Jan Engelhardt wrote:
> > >> Never mind, I see that some filesystems have 'depends on BLOCK' instead 
> > >> of being wrapped into if BLOCK. Not really consistent but whatever.
> > >
> > >Feel free to send in patches that make things more consistent.
> > 
> > How would you like things? if BLOCK or depends on BLOCK?
> 
> Well, if you can hide an entire block with if BLOCK, then that would be
> preferred. Otherwise depends on BLOCK.
> 
> > Does menuconfig/oldconfig/etc. parse the whole config structure faster 
> > it it done either way?
> 
> I'd be surprised if if BLOCK wasn't faster over, say, 10 depends on
> BLOCK.

I'd be /very/ surprised if anybody even noticed...
-- 
Dr. Horst H. von Brand                   User #22616 counter.li.org
Departamento de Informatica                    Fono: +56 32 2654431
Universidad Tecnica Federico Santa Maria             +56 32 2654239
Casilla 110-V, Valparaiso, Chile               Fax:  +56 32 2797513

Re: fs/Kconfig question regarding CONFIG_BLOCK

[Jan Engelhardt]

>> I'd be surprised if if BLOCK wasn't faster over, say, 10 depends on
>> BLOCK.
>
>I'd be /very/ surprised if anybody even noticed...

I am about to do a poorman's benchmark.


	-`J'
-- 

fs/Kconfig question regarding CONFIG_BLOCK

[Jan Engelhardt]

Hi,



fs/Kconfig has:

if BLOCK
menu "DOS/FAT/NT Filesystems"
...
<stuff here>
...
endmenu
endif


Why is it wrapped into BLOCK, or, why are all of the other filesystems 
which require a block device?


	-`J'
--