File BlockSize

File BlockSize

[Anish Srivastava]

Hi!!

Is there any way I can have 8K block sizes in ext2, reiserfs or ext3.

I am trying to install Oracle on Linux with 8K DB_Block_size.
But it gives me a Block size mismatch saying that the File BlockSize is only
4K

Maybe, there is a kernel patch available which enables Linux to create 8K
file blocks.
Thanks in anticipation....

Regards,
Anish Srivastava

Re: File BlockSize

[William Lee Irwin III]

On Tue, Feb 12, 2002 at 01:00:07PM +0530, Anish Srivastava wrote:
> Hi!!
> Is there any way I can have 8K block sizes in ext2, reiserfs or ext3.
> I am trying to install Oracle on Linux with 8K DB_Block_size.
> But it gives me a Block size mismatch saying that the File BlockSize is only
> 4K
> Maybe, there is a kernel patch available which enables Linux to create 8K
> file blocks.
> Thanks in anticipation....

Unfortunately filesystem block sizes larger than PAGE_SIZE are unsupported.
I wish they were, though.

Cheers,
Bill

Re: File BlockSize

[William Lee Irwin III]

On Tue, Feb 12, 2002 at 01:00:07PM +0530, Anish Srivastava wrote:
>> Hi!!
>> Is there any way I can have 8K block sizes in ext2, reiserfs or ext3.
>> I am trying to install Oracle on Linux with 8K DB_Block_size.
>> But it gives me a Block size mismatch saying that the File BlockSize is only
>> 4K
>> Maybe, there is a kernel patch available which enables Linux to create 8K
>> file blocks.
>> Thanks in anticipation....

On Mon, Feb 11, 2002 at 11:52:03PM -0800, William Lee Irwin III wrote:
> Unfortunately filesystem block sizes larger than PAGE_SIZE are unsupported.
> I wish they were, though.

Sorry -- I've been corrected elsewhere on this, and it's blocksizes greater
than PAGE_CACHE_SIZE... and there is some plan to eventually allow a
configurable PAGE_CACHE_SIZE.


Cheers,
Bill

Re: File BlockSize

[Hans Reiser]

William Lee Irwin III wrote:

>On Tue, Feb 12, 2002 at 01:00:07PM +0530, Anish Srivastava wrote:
>
>>Hi!!
>>Is there any way I can have 8K block sizes in ext2, reiserfs or ext3.
>>I am trying to install Oracle on Linux with 8K DB_Block_size.
>>But it gives me a Block size mismatch saying that the File BlockSize is only
>>4K
>>Maybe, there is a kernel patch available which enables Linux to create 8K
>>file blocks.
>>Thanks in anticipation....
>>
>
>Unfortunately filesystem block sizes larger than PAGE_SIZE are unsupported.
>I wish they were, though.
>
>Cheers,
>Bill
>-
>To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>the body of a message to majordomo@vger.kernel.org
>More majordomo info at  http://vger.kernel.org/majordomo-info.html
>Please read the FAQ at  http://www.tux.org/lkml/
>
>
I should be more precise, on alpha you can do it with reiserfs.

Hans

Re: File BlockSize

[Mike Fedyk]

On Sat, Feb 16, 2002 at 10:23:39AM +0300, Hans Reiser wrote:
> William Lee Irwin III wrote:
> 
> >On Tue, Feb 12, 2002 at 01:00:07PM +0530, Anish Srivastava wrote:
> >
> >>Hi!!
> >>Is there any way I can have 8K block sizes in ext2, reiserfs or ext3.
> >>I am trying to install Oracle on Linux with 8K DB_Block_size.
> >>But it gives me a Block size mismatch saying that the File BlockSize is 
> >>only
> >>4K
> >>Maybe, there is a kernel patch available which enables Linux to create 8K
> >>file blocks.
> >>Thanks in anticipation....
> >>
> >
> >Unfortunately filesystem block sizes larger than PAGE_SIZE are unsupported.
> >I wish they were, though.

I believe that is _page cache_ size instead of PAGE_SIZE.  Though,
page_cache_size is based on PAGE_SIZE...

> >
> I should be more precise, on alpha you can do it with reiserfs.
> 
> Hans

And ext2 and ext3 and ...

Re: File BlockSize

[Alan Cox]

> Is there any way I can have 8K block sizes in ext2, reiserfs or ext3.

Buy an Alpha 8)

> I am trying to install Oracle on Linux with 8K DB_Block_size.
> But it gives me a Block size mismatch saying that the File BlockSize is only
> 4K
> 
> Maybe, there is a kernel patch available which enables Linux to create 8K
> file blocks.

With current kernels the maximum block size of a file system you can mount
is the page size of the architecture. Generally people limit to 4K to avoid
file systems that only work with some machines.

Going to a block size bigger than page size causes all sorts of fun with 
allocation failures if there are not two pages free adjacent to one another
when allocating, and isn't really worth the cost.

Alan

Re: File BlockSize

[William Lee Irwin III]

On Tue, Feb 12, 2002 at 02:37:43PM +0000, Alan Cox wrote:
> Going to a block size bigger than page size causes all sorts of fun with 
> allocation failures if there are not two pages free adjacent to one another
> when allocating, and isn't really worth the cost.

This sounds like fairly severe memory fragmentation, which seems more
worrisome to me than blocksize constraints. Should I look into that?


Cheers,
Bill

Re: File BlockSize

[Rik van Riel]

On Tue, 12 Feb 2002, William Lee Irwin III wrote:
> On Tue, Feb 12, 2002 at 02:37:43PM +0000, Alan Cox wrote:
> > Going to a block size bigger than page size causes all sorts of fun with
> > allocation failures if there are not two pages free adjacent to one another
> > when allocating, and isn't really worth the cost.
>
> This sounds like fairly severe memory fragmentation, which seems more
> worrisome to me than blocksize constraints. Should I look into that?

Sorry for being dense, but I don't see why an 8 kB block would
need to occupy 2 contiguous pages in ram.

The page cache is indexed in pages, it should be easy enough to
do the disk IO on the 8 kB on-disk block as 2 4 kB IOs, each to
a different page in RAM.

regards,

Rik
-- 
"Linux holds advantages over the single-vendor commercial OS"
    -- Microsoft's "Competing with Linux" document

http://www.surriel.com/		http://distro.conectiva.com/

Re: File BlockSize

[Alan Cox]

> > This sounds like fairly severe memory fragmentation, which seems more
> > worrisome to me than blocksize constraints. Should I look into that?
> 
> Sorry for being dense, but I don't see why an 8 kB block would
> need to occupy 2 contiguous pages in ram.

Because all the kernel code knows that you can add a constant to the
base of a buffer to get anywhere in that block. Also the one buffer
per two page case isnt handled either


Alan

Re: File BlockSize

[Rik van Riel]

On Tue, 12 Feb 2002, Alan Cox wrote:

> > > This sounds like fairly severe memory fragmentation, which seems more
> > > worrisome to me than blocksize constraints. Should I look into that?
> >
> > Sorry for being dense, but I don't see why an 8 kB block would
> > need to occupy 2 contiguous pages in ram.
>
> Because all the kernel code knows that you can add a constant to the
> base of a buffer to get anywhere in that block. Also the one buffer
> per two page case isnt handled either

Is this still the case after blkdev-in-pagecache ?

Rik
-- 
"Linux holds advantages over the single-vendor commercial OS"
    -- Microsoft's "Competing with Linux" document

http://www.surriel.com/		http://distro.conectiva.com/

Re: File BlockSize

[Daniel Phillips]

On February 12, 2002 09:57 pm, William Lee Irwin III wrote:
> On Tue, Feb 12, 2002 at 02:37:43PM +0000, Alan Cox wrote:
> > Going to a block size bigger than page size causes all sorts of fun with 
> > allocation failures if there are not two pages free adjacent to one another
> > when allocating, and isn't really worth the cost.
> 
> This sounds like fairly severe memory fragmentation, which seems more
> worrisome to me than blocksize constraints. Should I look into that?

This is one of the chronic VM problems that rmap is supposed to cure, at least
it will provide a base on which an active physical defragger can be built.

-- 
Daniel