headers
Nathan Scott | 1 Oct 2004 04:58
Picon
Favicon

Re: XFS performance issues: O_DIRECT and Linux 2.6.6+

Hi James,

On Tue, Sep 21, 2004 at 02:10:40PM -0500, James Foris wrote:
> 
> Any idea how long befor the patch makes its way into Liunus' tree ?
> 

This will be in 2.6.9.  You can use the -rc releases already
with this fix & I'm updating the cvs tree to 2.6.9-rc3 at the
moment, so that should appear in a few hours too.

cheers.

--

-- 
Nathan
Permalink | Reply | 0 comments |
headers
Nathan Scott | 1 Oct 2004 05:13
Picon

TAKE 904196 - merge direct IO fix

Merge in direct IO fix from mainline 2.6.9-rc3.

Date:  Fri Oct  1 13:13:32 AEST 2004
Workarea:  chook.melbourne.sgi.com:/build/nathans/xfs-linux
Inspected by:  akpm <at> osdl.org

The following file(s) were checked into:
  longdrop.melbourne.sgi.com:/isms/xfs-kern/xfs-linux-melb

Modid:  xfs-linux-melb:xfs-kern:19629a
linux-2.6/xfs_aops.c - 1.81
Permalink | Reply | 0 comments |
headers
Nathan Scott | 1 Oct 2004 05:13
Picon

TAKE 904196 - 2.6.9-rc3 + kdb4.4

Upgrade kernel to 2.6.9-rc3 and kdb to 4.4

Date:  Fri Oct  1 13:02:59 AEST 2004
Workarea:  chook.melbourne.sgi.com:/build/nathans/2.6.x-xfs
Inspected by:  torvalds <at> osdl.org

The following file(s) were checked into:
  longdrop.melbourne.sgi.com:/isms/linux/2.6.x-xfs-melb


Modid:  2.6.x-xfs-melb:linux:19628a
drivers/scsi/megaraid/megaraid_mbox.h - 1.1
split-patches/kdb-i386-4.4-2 - 1.1
split-patches/kdb-common-v4.4-2 - 1.1
sound/usb/usx2y/usx2y.h - 1.1
sound/usb/usx2y/usbusx2yaudio.c - 1.1
sound/usb/usx2y/usbusx2y.h - 1.1
sound/usb/usx2y/usbusx2y.c - 1.1
sound/usb/usx2y/usbus428ctldefs.h - 1.1
sound/usb/usx2y/usX2Yhwdep.h - 1.1
Documentation/RCU/RTFP.txt - 1.1
Documentation/RCU/UP.txt - 1.1
Documentation/RCU/arrayRCU.txt - 1.1
Documentation/RCU/checklist.txt - 1.1
Documentation/RCU/listRCU.txt - 1.1
Documentation/RCU/rcu.txt - 1.1
Documentation/arm/IXP2000 - 1.1
Documentation/arm/Samsung-S3C24XX/EB2410ITX.txt - 1.1
Documentation/arm/Samsung-S3C24XX/Overview.txt - 1.1
sound/usb/usx2y/usX2Yhwdep.c - 1.1
(Continue reading)

Permalink | Reply | 0 comments |
headers
vijaya saradhi uppaluri | 1 Oct 2004 06:17
Picon

Re: xfs setting inode ctime

Dean,

 Thanks for your response. I am currently working on stack of file
systems where I set the attribute ATTR_CTIME and expect the underlying
file-system to set the ctime to the value I pass. It is working on
other file-systems(ext2) on linux.

-Thanks,
 Saradhi.

On Wed, 29 Sep 2004 10:14:03 -0500, Dean Roehrich <roehrich <at> sgi.com> wrote:
> >From:  vijaya saradhi uppaluri <uvsaradhi <at> gmail.com>
> 
> >Is there any logical reason behind this? Why does XFS_AT_CTIME is
> >coupled with ATTR_DMI?
> >
> >Is there any goodway of setting ctime attribute of xfs?
> >
> >I observed that other file-systems sets the ctime accordinlgy if
> >ATTR_CTIME flag is set.
> 
> XFS inherits this restriction from Irix, where userspace is not allowed to set
> ctime.  The Irix utime(2) and utimes(2) system calls allow only atime and
> mtime to be set, and always set ctime to the current time.  I see the same in
> Solaris 9 and RH9 manpages.
> 
> How are you changing ctime?
> 
> Dean
>
(Continue reading)

Permalink | Reply | 1 comment |
headers
Nathan Scott | 1 Oct 2004 07:22
Picon

TAKE 904196 - sync up XFS trees

Remove unused function prototypes.

Date:  Fri Oct  1 15:20:25 AEST 2004
Workarea:  bruce.melbourne.sgi.com:/source2/ptools/xfs-linux
Inspected by:  nathans

The following file(s) were checked into:
  longdrop.melbourne.sgi.com:/isms/xfs-kern/xfs-linux-melb

Modid:  xfs-linux-melb:xfs-kern:19636a
xfs_log_priv.h - 1.102
xfs_bmap_btree.h - 1.65
xfs_acl.h - 1.30
xfs_inode.h - 1.194
xfs_attr_leaf.h - 1.35
quota/xfs_qm.h - 1.6
linux-2.6/xfs_fs_subr.h - 1.12
linux-2.4/xfs_fs_subr.h - 1.16

Merge back Al Viros link handlink fixes and put_link inode op.

Date:  Fri Oct  1 15:21:26 AEST 2004
Workarea:  bruce.melbourne.sgi.com:/source2/ptools/xfs-linux
Inspected by:  nathans

The following file(s) were checked into:
  longdrop.melbourne.sgi.com:/isms/xfs-kern/xfs-linux-melb

Modid:  xfs-linux-melb:xfs-kern:19637a
linux-2.6/xfs_iops.c - 1.223
(Continue reading)

Permalink | Reply | 0 comments |
headers
Christoph Hellwig | 1 Oct 2004 09:28
Favicon

Re: xfs setting inode ctime

On Fri, Oct 01, 2004 at 09:47:03AM +0530, vijaya saradhi uppaluri wrote:
> Dean,
> 
>  Thanks for your response. I am currently working on stack of file
> systems where I set the attribute ATTR_CTIME and expect the underlying
> file-system to set the ctime to the value I pass. It is working on
> other file-systems(ext2) on linux.

Dean, I think we should change change XFS to use the value provided
by ->setattr to conform to the inkernel API specified by the Linux kernel.
Permalink | Reply | 0 comments |
headers
Dean Roehrich | 1 Oct 2004 17:59
Picon
Favicon

Re: xfs setting inode ctime

>From:  Christoph Hellwig <hch <at> infradead.org>
>On Fri, Oct 01, 2004 at 09:47:03AM +0530, vijaya saradhi uppaluri wrote:
>> Dean,
>> 
>>  Thanks for your response. I am currently working on stack of file
>> systems where I set the attribute ATTR_CTIME and expect the underlying
>> file-system to set the ctime to the value I pass. It is working on
>> other file-systems(ext2) on linux.
>
>Dean, I think we should change change XFS to use the value provided
>by ->setattr to conform to the inkernel API specified by the Linux kernel.

I agree.

Ug, xfs_setattr is a real beaut, isn't it?

Dean
Permalink | Reply | 3 comments |
headers
bugzilla-daemon | 2 Oct 2004 00:39
Picon
Favicon

[Bug 363] segfault of user-app while triggering dmapi event

http://oss.sgi.com/bugzilla/show_bug.cgi?id=363

------- Additional Comments From mmontour <at> bycast.com  2004-01-10 15:39 PDT -------
I see a similar problem with the current linux-2.4-xfs kernel tree. I found one
issue in fs/xfs/xfs_dmapi.c xfs_dm_send_namesp_event:

	error = dm_send_namesp_event(event, vfsp ? vfsp->vfs_super: NULL,
				    LINVFS_GET_IP(vp1), vp1_right,
				    LINVFS_GET_IP(vp2), vp2_right,
				    name1, name2,
				    mode, retcode, flags);

In a CREATE event, vnode_t *vp2 is NULL. I changed the code to not call
LINVFS_GET_IP on a NULL value (using "vp2 ? LINVFS_GET_IP(vp2) : NULL"), and now
I successfully get CREATE and POSTCREATE events without any kernel oopses or
segfaults. I haven't investigated this further, so I don't know (A) if this is
correct or (B) if this same issue appears elsewhere in the code.

------- You are receiving this mail because: -------
You are the assignee for the bug, or are watching the assignee.
Permalink | Reply | 0 comments |
headers
Andrew Ho | 2 Oct 2004 05:41

2.4.21-20.EL - XFS

I am looking for this kernel for testing.

Thanks,

Andrew
Permalink | Reply | 0 comments |
headers
Nathan Scott | 4 Oct 2004 02:21
Picon

TAKE 907752 - fix userspace dmapi build

Fix DMAPI userspace source to include fewer kernel headers directly,
and include XFS headers via the usual libxfs.h interface.

Date:  Sun Oct  3 17:20:59 PDT 2004
Workarea:  snort.melbourne.sgi.com:/home/nathans/us-xfs-cmds
Inspected by:  nathans

The following file(s) were checked into:
  bonnie.engr.sgi.com:/isms/slinx/xfs-cmds

Modid:  xfs-cmds:slinx:180254a
dmapi/include/dmapi_kern.h - 1.15
dmapi/libdm/dm_handle2path.c - 1.12
dmapi/libdm/dm_handle.c - 1.8
Permalink | Reply | 0 comments |

Gmane