headers
Nathan Scott | 1 Jul 2002 03:10
Picon

TAKE - kdev_val

Date:  Sun Jun 30 18:09:33 PDT 2002
Workarea:  snort.melbourne.sgi.com:/home/nathans/2.4.x-xfs

The following file(s) were checked into:
  bonnie.engr.sgi.com:/isms/slinx/2.4.x-xfs

Modid:  2.4.x-xfs:slinx:122586a
linux/include/linux/kdev_t.h - 1.7
	- add missing brace to kdev_val() - assume this will get picked up
	  by Marcelo before 2.4.19-final...

linux/fs/xfs/linux/xfs_linux.h - 1.74
linux/fs/xfs/pagebuf/page_buf_internal.h - 1.9
	- make our kdev_val() 2.5 compat macro dependent on kernel version as
	  2.4.19 looks like it will have a kdev_val now.  this fixes up some
	  warnings introduced in the 2.4.19-rc1 merge.
Permalink | Reply | 0 comments |
headers
Keith Owens | 1 Jul 2002 03:38
Picon
Favicon

Re: patch file corrupt?

On Sun, 30 Jun 2002 16:55:43 +1000, 
Ian Cumming <ian <at> ids.org.au> wrote:
>I just tried to apply this patch to my 2.4.18 tree:
>ftp://oss.sgi.com/projects/xfs/download/patches/2.4.18/xfs-2.4.18-all-i386.bz2
>
>After extracting the patch, it seems that the diff had "line wrapping"
>applied to it. It won't patch, because lines longer than 72 characters
>have had carriage returns inserted into it, and patch breaks as a
>result.

I just checked that patch, it is fine and it patches cleanly against a
pristine 2.4.18 tree.  md5sums

0411a3921a553c8566f8dc1770f5bf34  xfs-2.4.18-all-i386.bz2
f154a5ad3376cf54095cb158e5f54338  xfs-2.4.18-all-i386
Permalink | Reply | 0 comments |
headers
Simon Matter | 1 Jul 2002 09:28

Crash and some more problems

Hi

One of my servers running RedHat 7.2-XFS with kernel 2.4.9-34SGI_XFS_1.1
has crashed yesterday. With crash I mean really dead, nothing in the
logs, no keyboard leds, no output on the console. I don't know what
happened, my only guess is that temperature was getting too high for
this quite old box.

Unfortunately there were two more things:

1) When I restarted the box (after cooling it down), it started all the
soft raid devices, mounted root and died again when 'mounting local
filesystems'. HD leds kept flashing which means the raid was synced but
everything else was dead. I finally booted with 'lilo init=/bin/sh',
waited for all raid devices to finish sync. I mounted and unmounted all
filesystems to replay logs, ran xfs_repair on them without showing up
any problem. I have then rebooted and it came up as expected.

2) I have checked the filesystems and logs. Everything seems fine except
that I found some lines of /var/log/messages in the file
/var/log/squid/store.log from squid. Not really nice from a security
point of view.

Any comments?

Simon

file /etc/fstab:

LABEL=/boot             /boot                   xfs     defaults
(Continue reading)

Permalink | Reply | 0 comments |
headers
mdew | 1 Jul 2002 13:54
Picon
Favicon

Error: uffix or operands invalid for `bsf'

While making the latest linux-2.4-xfs tree...I came across this error.

make[4]: Entering directory
`/usr/src/linux-2.4-xfs/linux/fs/xfs/pagebuf'
gcc -D__KERNEL__ -I/usr/src/linux-2.4-xfs/linux/include  -Wall
-Wstrict-prototypes -Wno-trigraphs -O3 -fno-strict-aliasing -fno-common
-fomit-frame-pointer -pipe -mpreferred-stack-boundary=2 -march=i686 
-I.. -nostdinc -I /usr/lib/gcc-lib/i386-linux/3.1.1/include
-DKBUILD_BASENAME=page_buf_locking  -c -o page_buf_locking.o
page_buf_locking.c
{standard input}: Assembler messages:
{standard input}:416: Error: suffix or operands invalid for `bsf'
make[4]: *** [page_buf_locking.o] Error 1
make[4]: Leaving directory `/usr/src/linux-2.4-xfs/linux/fs/xfs/pagebuf'
make[3]: *** [first_rule] Error 2
make[3]: Leaving directory `/usr/src/linux-2.4-xfs/linux/fs/xfs/pagebuf'
make[2]: *** [_subdir_pagebuf] Error 2
make[2]: Leaving directory `/usr/src/linux-2.4-xfs/linux/fs/xfs'
make[1]: *** [_subdir_xfs] Error 2
make[1]: Leaving directory `/usr/src/linux-2.4-xfs/linux/fs'
make: *** [_dir_fs] Error 2

Debian Sid/GCC 3.1.1pre2/Binutils 2.12.90.0.9-1

--

-- 
ph33r!
Linux mdew 2.4.19-rc1-xfs #5 Tue Jun 25 17:35:28 NZST 2002 i686 unknown
GPG Key: http://mdew.orcon.net.nz/gpg
Permalink | Reply | 0 comments |
headers
Nathan Scott | 2 Jul 2002 01:18
Picon
Favicon

Re: XFS NFS and umask

On Mon, Jun 24, 2002 at 09:09:28PM -0700, John Stevens wrote:
> I am running kernel 2.4.18 XFS, a clean install using the SGI installer.  I
> have searched the archive for a solution to the umask not being set properly
> on NFS mounted volumes.   I am mounting the Linux box on multiple Onyx and
> Onyx 2, is there a solution to this problem??

We've recently discussed a solution to this which the ext2/ext3
patch maintainer is using - we will do a similar thing to that.
This change should go into CVS later today.

cheers.

--

-- 
Nathan
Permalink | Reply | 0 comments |
headers
Nathan Scott | 2 Jul 2002 02:36
Picon

TAKE - ACLs

Several changes to reduce the overlap we have with the rest of the
kernel due to ACLs - eg. we no longer add a new field to the struct
super_block.  Also use Andreas' idea to work around the problems in
our handling of the umask/mode for inodes coming to us from NFS.

cheers.

Date:  Mon Jul  1 17:12:19 PDT 2002
Workarea:  snort.melbourne.sgi.com:/home/nathans/linux-xfs

The following file(s) were checked into:
  bonnie.engr.sgi.com:/isms/slinx/2.4.x-xfs

Modid:  2.4.x-xfs:slinx:122601a
linux/include/linux/posix_acl_xattr.h - 1.1
	- the minimal types and macros we need in order to support the
	  posix_acl extended attributes.

linux/include/linux/fs.h - 1.151
	- remove use of additional s_xattr_flags, use the existing inode
	  flags field instead if ACLs are enabled.

linux/fs/namei.c - 1.44
	- use a naming scheme consistent with the other inode flags.

linux/fs/xfs/linux/xfs_linux.h - 1.75
	- rework acl headers so that only one, simpler acl header is needed.

linux/fs/xfs/linux/xfs_vnode.c - 1.84
linux/fs/xfs/linux/xfs_super.c - 1.185
(Continue reading)

Permalink | Reply | 0 comments |
headers
Nathan Scott | 2 Jul 2002 03:51
Picon

TAKE - experimental capabilities code

Early days for capability support, but this will be helpful to
people investigating this stuff.

cheers.

Date:  Sun Jun 30 18:18:23 PDT 2002
Workarea:  snort.melbourne.sgi.com:/home/nathans/2.4.x-xfs

The following file(s) were checked into:
  bonnie.engr.sgi.com:/isms/slinx/2.4.x-xfs

Modid:  xfs-cmds:slinx:122587a
cmd/xfsprogs/doc/CHANGES - 1.71
cmd/xfsprogs/mkfs/proto.c - 1.8
	- Fix typo in mkfs realtime summary inode alloc failure message.

Date:  Mon Jul  1 18:36:32 PDT 2002
Workarea:  snort.melbourne.sgi.com:/home/nathans/2.4.x-xfs

The following file(s) were checked into:
  bonnie.engr.sgi.com:/isms/slinx/2.4.x-xfs

Modid:  2.4.x-xfs:slinx:122603a
linux/include/linux/posix_cap_xattr.h - 1.1
	- initial, experimental structure definitions for capability extended
	  attribute.  almost guaranteed that this will change...

linux/fs/xfs/xfs_cap.c - 1.1
	- initial version of the backend filesystem code which converts ondisk
	  extended attribute format to the kernels native capability structure.
(Continue reading)

Permalink | Reply | 0 comments |
headers
Hiroyuki Nakano | 2 Jul 2002 08:49
Picon
Picon

Re: XFS force shutdown : EFSCORRUPTED

Hi Steve,

I wanted to confirm XFS commands are executed to normaly
on mounted device. So I tried to access to mounted device.
Now I see commands are no problem to executed on mounted
device.

Thanks a lot.

-nak

> 
> I happened to be sitting next to another filesystem developer
> when I read this, and he remembered an old ext2 race condition.
> I think I understand what is happening - and how to fix it, the
> only problem is that fixing it will potentially be a major
> performance issue - so the question is, why do you need this to
> work? Would a solution which disallowed block access to the
> device when it was mounted be sufficient, or do you really 
> want to look at metadata live via the block device. The
> xfsdump program does not need this, the only thing which does
> is running xfs_db on a live filesystem, but that is not really
> of use to anyone but developers.
> 
> Steve
>
Permalink | Reply | 1 comment |
headers
Andrew Morton | 2 Jul 2002 23:33
Picon
Picon

Re: [PATCH] fix qsort breakage

Christoph Hellwig wrote:
> 
> Qsort in the XFS tree has two problems:
> 
>  o allocates memory using GFP_KERNEL although called from under i_sem
>    (possible deadlock)

The page allocator will never take an i_sem.  Note how generic_file_write()
performs tons of GFP_HIGHUSER allocations inside i_sem.  No deadlocks
there.

qsort should take gfp_flags as an argument. Or not perform any allocation,
as you've described.

But the code seems OK as-is.  Suggest you stick with the more
robust GFP_KERNEL.

(It'd be nice to stick that qsort in lib/qsort.c, rather than making
it XFS-private, btw).

-
Permalink | Reply | 2 comments |
headers
Tatsu Yasugahira | 3 Jul 2002 07:01
Picon
Picon

mount as read-only and XFS commands

Hi,

I'd like to confirm the behavior of XFS.

I found that the XFS command (xfs_growfs) changes the filesystem
which is mounted by read-only.
This phenomenon occurs on the read-only filesystem and 
on the filesystem that goes to read-only by the remount operation.

I think it should be the error with EROFS or EBUSY.

How do you think?

--------

I did the following (tp_readonly.sh) :

	MNT=/mnt/xfs
	DEV=/dev/hda8
	mkfs -t xfs -f $DEV -d size=12000b
	mount -t xfs -o ro $DEV $MNT
	xfs_growfs -d $MNT
	xfs_growfs -n $MNT
	umount $MNT
	mount -t xfs $DEV $MNT

This results in the following messages:

	+ MNT=/mnt/xfs
	+ DEV=/dev/hda8
(Continue reading)

Permalink | Reply | 1 comment |

Gmane