Bug 105805 - Add functional XFS filesystem to kernel (or at least hooks)
Summary: Add functional XFS filesystem to kernel (or at least hooks)
Keywords:
Status: CLOSED NEXTRELEASE
Alias: None
Product: Fedora
Classification: Fedora
Component: kernel
Version: rawhide
Hardware: All
OS: Linux
medium
medium
Target Milestone: ---
Assignee: Arjan van de Ven
QA Contact: Brian Brock
URL:
Whiteboard:
Depends On: 66513
Blocks:
TreeView+ depends on / blocked
 
Reported: 2003-09-27 20:45 UTC by Eric Sandeen
Modified: 2007-11-30 22:10 UTC (History)
1 user (show)

Fixed In Version:
Doc Type: Enhancement
Doc Text:
Clone Of:
Environment:
Last Closed: 2004-02-19 21:40:57 UTC
Type: ---
Embargoed:


Attachments (Terms of Use)
3 new exports for xfs I/O (2.12 KB, patch)
2003-11-07 00:57 UTC, Eric Sandeen
no flags Details | Diff
XFS doesn't need union at bottom of inode, save the memory (885 bytes, patch)
2003-11-07 00:58 UTC, Eric Sandeen
no flags Details | Diff

Description Eric Sandeen 2003-09-27 20:45:46 UTC
From Bugzilla Helper:
User-Agent: Mozilla/5.0 Galeon/1.2.11 (X11; Linux i686; U;) Gecko/20030417

Description of problem:
The current Fedora Core test2 kernel is not very far away from
compiling & running the current XFS filesystem code.

In terms of compilation, we only need 2 more exported symbols, and a
slight reorganization of init_inode_once(), splitting the function
into two parts so that XFS does not need to allocate memory for
the unused union at the bottom of the inode.  (the init_inode_once
change is not strictly necessary, we could work around this).

Steve Lord is looking at fixing up the O_DIRECT path with sct's new
O_DIRECT code, and this may take a few more exported symbols, but
so far it does not look like any core kernel code is affected.

So, with some very minor changesto core kernel code, we can make current XFS
code compile against this kernel - and if you wish to stop there, we could 
simply provide xfs-module RPMS that would work wit the released kernel.

However, if we've gone that far, we may as well proceed to updating
the snapshot xfs code that is already in your tree, I think.

not attaching any patches yet, just checking on the possibility of
this at this stage in the release.  If you're interested, I'll attach
some real code, even specfile patches.  :)

Also, as you proceed to a 2.6 kernel in your next release, I can help
look at anaconda to make the installer xfs-capable, if desired (most of the
code is already there)

Thanks,

-Eric

Version-Release number of selected component (if applicable):


How reproducible:
Always

Steps to Reproduce:
1. try to mount an xfs filesystem, or build with CONFIG_XFS_FS enabled
2. fail :)
    

Additional info:

Comment 1 Eric Sandeen 2003-11-07 00:57:20 UTC
Created attachment 95781 [details]
3 new exports for xfs I/O

Comment 2 Eric Sandeen 2003-11-07 00:58:28 UTC
Created attachment 95782 [details]
XFS doesn't need union at bottom of inode, save the memory

Comment 3 Eric Sandeen 2003-11-07 01:04:03 UTC
Kernel from FC1 + XFS at ftp://oss.sgi.com/projects/xfs/testing/FC1

Only 1 patch, exporting 3 symbols, would be needed to allow sgi to
ship our own xfs-module rpms.  Another trivial patch lets us
save quite a lot of memory in the inodes.  Both are attached, work
with the released FC1 kernel.

Similar things would work for RHEL, too.

I understand that perhaps you can't ship xfs in your kernels at 
this point.  But these 3 tweaks would make it easy for us to provide
your customers who want to use XFS with a simple means of doing so.

Thanks,

-Eric

Comment 4 Anthony 2003-12-30 16:18:32 UTC
I too would love to see future versions of Fedora Core (test2 and FC2)
have built-in XFS support, especially if Anaconda has support as well.
     Could this bug report be updated when new information as to the
status of this request in any future versions of Fedora.

Comment 5 Dave Jones 2004-02-19 21:40:57 UTC
FC2 already has it enabled.



Note You need to log in before you can comment on or make changes to this bug.