Bug 450566

Summary:	FEAT: RHEL5.3 backport fallocate syscall
Product:	Red Hat Enterprise Linux 5	Reporter:	Eric Sandeen <esandeen>
Component:	kernel	Assignee:	Eric Sandeen <esandeen>
Status:	CLOSED ERRATA	QA Contact:	Martin Jenner <mjenner>
Severity:	low	Docs Contact:
Priority:	low
Version:	5.3	CC:	jcm, jlv
Target Milestone:	rc
Target Release:	---
Hardware:	All
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:
Clones:	561396 (view as bug list)		Environment:
Last Closed:	2009-01-20 20:22:28 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:
Bug Blocks:	561396

Description Eric Sandeen 2008-06-09 17:10:46 UTC

The upstream fallocate syscall allows efficient space preallocation for
filesystems which support it (today: ext4, xfs, and ocfs2).

To fully support new ext4 features for the tech preview in 5.3, we should
backport this new syscall.

Relevant upstream commits for the OS infrastructure:

http://git.engineering.redhat.com/?p=linux-2.6.git;a=commitdiff;h=97ac73506c0ba93f30239bb57b4cfc5d73e68a62
http://git.engineering.redhat.com/?p=linux-2.6.git;a=commitdiff;h=0d786d4a2773f06a791e8c3730d049077fb81df6

We should probably also update glibc to wire posix_fallocate to this interface
as well.

Thanks,
-Eric

Comment 1 Eric Sandeen 2008-06-13 18:12:21 UTC

There are 2 tricky bits to this w.r.t. KABI

1. it requires a new inode operation, ->fallocate.  This gets tested:

+       if (inode->i_op && inode->i_op->fallocate)
+               ret = inode->i_op->fallocate(inode, mode, offset, len);

so it is not safe to do for out-of-tree filesystems; ->fallocate will point off
into junk.  One thought is to add a superblock s_flags flag, MS_HAS_FALLOC, and
filesystems which do have ->fallocate can set this.  Then we only test for
fallocate in the presence of that flag.  The only risk is if some other out of
tree filesystem stole that flag for some reason, but I that's an
acceptable/small risk.

2. it requires BH_Unwritten support (unwritten = allocated but uninitialized). 
This is a buffer head flag not present in RHEL5, and we capped useable flags
with BH_PrivateStart for individual filesystem use.  It might be reasonably safe
to put BH_Unwritten up at bit 31, and assume (hope?) that no out of tree
filesystem uses so many flags as to use up that slot... it's far from foolproof
though.

-Eric

Comment 4 Don Zickus 2008-09-15 14:17:45 UTC

in kernel-2.6.18-115.el5
You can download this test kernel from http://people.redhat.com/dzickus/el5

Comment 8 errata-xmlrpc 2009-01-20 20:22:28 UTC

An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on therefore solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.

http://rhn.redhat.com/errata/RHSA-2009-0225.html