xfs
[Top] [All Lists]

Re: xfs on a sparc64

To: Thomas Graichen <graichen@xxxxxxxxxxxxx>, thomas.graichen@xxxxxxxxxxxxx
Subject: Re: xfs on a sparc64
From: Steve Lord <lord@xxxxxxx>
Date: Fri, 10 Nov 2000 15:23:03 -0600
Cc: linux-xfs@xxxxxxxxxxx
In-reply-to: Message from Thomas Graichen <news-innominate.list.sgi.xfs@innominate.de> of "10 Nov 2000 21:21:11 GMT." <news2mail-8uhos7$fhg$3@mate.bln.innominate.de>
Sender: owner-linux-xfs@xxxxxxxxxxx
> "Martin K. Petersen" <mkp@xxxxxxxxxxxxx> wrote:
> >>>>>> "kris" == kris buggenhout <gast6@xxxxxx> writes:
> 
> > kris> could this be related to 64bit structure ...  page_buf_io.c: In
> > kris> function `pagebuf_file_read': page_buf_io.c:694: `O_DIRECT'
> 
> > Fixed.  SPARC should compile fine now.
> 
> > Please note that you need to create your file system with -bsize=8k
> > because SPARCs use 8KB pages.
> 
> which i assume i have to take care of on the alpha too (which
> also has 8k pages)?
> 
> t
> 


Yes - this is a pagebuf restriction at the moment, the filesystem blocksize is
tied to the page size of the system. This will get fixed eventually, supporting
blocksizes less than the pagesize is one task which is just a matter of
code within XFS and pagebuf. Supporting block sizes larger than a page is
harder as we really need a guaranteed and cheap way of allocating kernel
memory in chunks bigger than a page - or of making multiple pages look
contiguous in the kernel. There is code in there right now to remap the
memory, but it is not deemed an acceptable solution.

Steve



<Prev in Thread] Current Thread [Next in Thread>