maximilian attems | 3 Sep 09:52 2003
Picon

[Kernel-janitors] [PATCH] list_for_each fs/

again pure cosmetical change :)
- for (list = ymf_devs.next; list != &ymf_devs; list = list->next) {
+ list_for_each(list, &ymf_devs) {

tested, please apply

a++ maks

 dquot.c       |    2 +-
 jffs/intrep.c |    4 ++--
 namespace.c   |    2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)

--- linux-2.6.0-test4/fs/jffs/intrep.c	Sat Aug 23 01:57:01 2003
+++ linux-2.6.0-maks4/fs/jffs/intrep.c	Wed Sep  3 09:27:52 2003
 <at>  <at>  -1624,7 +1624,7  <at>  <at> 

 	D3(printk("jffs_find_file(): ino: %u\n", ino));

-	for (tmp = c->hash[i].next; tmp != &c->hash[i]; tmp = tmp->next) {
+	list_for_each(tmp, &c->hash[i]) {
 		f = list_entry(tmp, struct jffs_file, hash);
 		if (ino != f->ino)
 			continue;
 <at>  <at>  -2022,7 +2022,7  <at>  <at> 

 	for (pos = 0; pos < c->hash_len; pos++) {
 		struct list_head *p, *next;
-		for (p = c->hash[pos].next; p != &c->hash[pos]; p = next) {
+		list_for_each(p, &c->hash[pos]) {
(Continue reading)

Szonyi Calin | 4 Sep 11:15 2003
Picon

Bug in jfs.fsck

Hello

I had a kernel panic and forced to press the reset button.
After that -- surprise --
fsck says that filesystem is clean but mount says:
mount: / not mounted already, or bad option 
when trying to remount the root filesystem with the command:
/sbin/mount -w -v -n -o remount /
Strace results of the command:
/sbin/mount -n -o rw,remount / 
are attached in the file strace-mount

If I try to do a 
/sbin/fsck -f  / 
the results are in the file fsck-f (attached)
If I try 
/sbin/fsck -v -f /
the results are in the file fsck-f-v (attached)

The _only_ way to recover my partition (root) is to boot into another 
linux (in this case on /dev/hda5) and run fsck from there.
The results of this fsck are attached in the file mdk-fsck-f-v

It seems that  fsck is trying to open the partition exclusively 
and because it can't it doesn't want to operate.
The root filesystem cannot be opened exclusively because other 
programs are using it. See atached file lsof-hda6 for details.

It should have an option: "Are you sure you want to repair the filesystem ? 
type yes  if you really know what you are doing"
(Continue reading)

sundeep saini | 4 Sep 16:20 2003

Re: Re: jfs directory structure

 
 
-------Original Message-------
 
Respected Sir, 
            I again stuck in problem. I am sending my query as a attachment
as .rtf file.
Please guide me.

With regrds
Sundeep saini

 
 
-------Original Message-------
 
Respected Sir,
            I again stuck in problem. I am sending my query as a attachment as .rtf file.
Please guide me.
 
With regrds
Sundeep saini
(Continue reading)

Dave Kleikamp | 4 Sep 17:19 2003
Picon

Re: Bug in jfs.fsck

On Thu, 2003-09-04 at 04:15, Szonyi Calin wrote:
> It seems that  fsck is trying to open the partition exclusively 
> and because it can't it doesn't want to operate.
> The root filesystem cannot be opened exclusively because other 
> programs are using it. See atached file lsof-hda6 for details.

Okay.  It looks like a recent patch to fs/block_dev.c now disallows
O_EXCL opens when a file system is mounted.  I'll have to make a change
to the utilities, since the kernel change seems reasonable.  A quick fix
would be to remove the O_EXCL flag from the open call in
jfsutils/fsck/fsckpfs.c.  I will make a proper fix to attempt the O_EXCL
call when appropriate and open without it when necessary.

> It should have an option: "Are you sure you want to repair the filesystem ? 
> type yes  if you really know what you are doing"
> for such cases 

When the file system is mounted read-only, I don't think a confirmation
is necessary.  It just used to work with the O_EXCL flag, and now it
doesn't.

Thanks for the very descriptive report,
Shaggy
--

-- 
David Kleikamp
IBM Linux Technology Center
Szonyi Calin | 4 Sep 17:32 2003
Picon

RE: Bug in jfs.fsck


> 
> A quick fix
> would be to remove the O_EXCL flag from the open call in
> jfsutils/fsck/fsckpfs.c.  I will make a proper fix to attempt 
> the O_EXCL
> call when appropriate and open without it when necessary.
> 

I'll do that.

There is still a problem with the fact that the filesystem was marked clean 
when actually it should be marked dirty. 
Fsck says that "Filesystem is dirty but is marked clean". 
See atached file (fsck-f) for details.

> 
> Thanks for the very descriptive report,
Thanks for your answer 

> Shaggy
> -- 
> David Kleikamp
> IBM Linux Technology Center
> 

Calin

Attachment (fsck-f): application/octet-stream, 1040 bytes
(Continue reading)

Dave Kleikamp | 4 Sep 19:47 2003
Picon

RE: Bug in jfs.fsck

On Thu, 2003-09-04 at 10:32, Szonyi Calin wrote:

> There is still a problem with the fact that the filesystem was marked clean 
> when actually it should be marked dirty. 
> Fsck says that "Filesystem is dirty but is marked clean". 
> See atached file (fsck-f) for details.

Hmm.  "but is marked clean" is misleading.  It's not clean...the journal
needs replaying.  I need to see about fixing that message.  Once the
journal was replayed, the rest of the fsck was clean.  The only other
thing that bothers me is the warning when replaying the journal,
"Invalid segment offset".  I'll have to look into that too.

Thanks again,
Shaggy
--

-- 
David Kleikamp
IBM Linux Technology Center
Dave Kleikamp | 5 Sep 20:48 2003
Picon

[ANNOUNCE] JFS 1.1.3

Release 1.1.3 of JFS was made available today.

It is important that you update the JFS utilities if you are using the
2.6.0 kernel.  A recent change in the kernel changed the behavior of
the O_EXCL flag in an open call to a block device.  Older versions of
fsck.jfs may fail to replay the journal when a file system is mounted
read-only.  (This will be the case if the root file system is JFS.)

Drop 66 on September 5, 2003 includes fixes to the file system and
utilities.

Utilities changes

- jfs_fsck can now find external journal on evms release 2 volumes
- code cleanup: fsck messaging is much cleaner
- Don't right-justify when printing volume label
- jfs_fsck should return zero when replaying the journal is successful
- jfs_fsck should not require that the device is opened with O_EXCL

File System changes

- Fix compat.h so that compile does not break on unlikely and dump_stack
- jfs_lookup should check for bad inode returned by iget
- Prevent rare deadlock
- Fix compile problem with recent Redhat kernels
- Code cleanup suggested by static analysis tool
- Peformance improvement
- Use block device inode/mapping instead of direct_inode/direct_mapping
- Support for kernels older than 2.4.18 has been removed
- Fix resize errors
(Continue reading)

Dave Kleikamp | 5 Sep 21:10 2003
Picon

Re: [ANNOUNCE] JFS 1.1.3

Oops.  The subject should read "1.1.3"
Erwin Burgstaller | 6 Sep 13:57 2003
X-Face

Space from deleted files not freed


Just making my first steps with JFS on Linux, I've noticed following:

root <at> lbox:~# uname -a
Linux lbox 2.4.21-lvm #1 Sat Sep 6 12:06:56 CEST 2003 i686 GNU/Linux
root <at> lbox:~# mount /dev/hdb8 /mnt/hdb8
root <at> lbox:~# df /mnt/hdb8
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/hdb8              1943408       372   1943036   1% /mnt/hdb8
root <at> lbox:~# cd /mnt/hdb8
root <at> lbox:/mnt/hdb8# cp /home/cdimage/knoppix-v3.2 .
root <at> lbox:/mnt/hdb8# df .
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/hdb8              1943408    716960   1226448  37% /mnt/hdb8
root <at> lbox:/mnt/hdb8# ls -l
total 716588
-rw-r--r--    1 root     root     733786112 Sep  6 13:35 knoppix-v3.2
# rm *
root <at> lbox:/mnt/hdb8#
# ls -l
total 0
root <at> lbox:/mnt/hdb8#
# df .
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/hdb8              1943408    716960   1226448  37% /mnt/hdb8
root <at> lbox:/mnt# umount hdb8
root <at> lbox:/mnt# df /dev/hdb8
Filesystem           1K-blocks      Used Available Use% Mounted on
/dev/hda4             17803524  10985800   5913268  66% /
root <at> lbox:/mnt# mount /dev/hdb8 /mnt/hdb8
(Continue reading)

szonyi calin | 8 Sep 11:30 2003
Picon

Re: Re: [ANNOUNCE] JFS 1.1.3

 --- Dave Kleikamp <shaggy <at> austin.ibm.com> a écrit : > Oops. 
The subject should read "1.1.3"

it doesn't compile (jfs-utils) with gcc-2.95.3.
Only with gcc-3.x.x

Bye
Calin

=====
--
A mouse is a device used to point at 
the xterm you want to type in.
Kim Alm on a.s.r.

___________________________________________________________
Do You Yahoo!? -- Une adresse  <at> yahoo.fr gratuite et en français !
Yahoo! Mail : http://fr.mail.yahoo.com
Mon Tue Wed Thu Fri Sat Sun
1234567
891011121314
15161718192021
22232425262728
2930     
Language
Options
Current view: Threads only / Showing only 20 lines / Not hiding cited text.
Change to All messages, whole messages, or hide cited text.

Post a message
NNTP Newsgroup
Classic Gmane web interface
XML RSS Feed
List Information

About Gmane
Recent entries
Archives
 
10
 
13
 
9
 
7
 
97
 
19
 
10
 
5
 
23
 
5
 
18
 
23
 
2
 
2
 
10
 
2
 
7
 
16
 
17
 
7
 
36
 
17
 
10
 
7
 
1
 
23
 
14
 
3
 
3
 
5
 
5
 
13
 
25
 
21
 
9
 
4
 
7
 
20
 
13
 
6
 
2
 
12
 
17
 
6
 
10
 
8
 
22
 
29
 
42
 
10
 
5
 
18
 
6
 
43
 
23
 
10
 
18
 
10
 
1
 
1
 
2
 
36
 
25
 
31
 
26
 
35
 
39
 
36
 
20
 
25
 
20
 
35
 
58
 
31
 
48
 
27
 
9
 
11
 
2
 
29
 
9
 
32
 
24
 
16
 
16
 
14
 
17
 
20
 
21
 
8
 
21
 
8
 
38
 
22
 
11
 
25
 
49
 
50
 
36
 
51
 
50
 
15
 
36
 
54
 
25
 
62
 
53
 
25
 
14
 
52
 
5
 
63
 
30
 
33
 
43
 
75
 
9
 
31
 
15
 
7
 
14
 
64
 
8
 
10
 
19
 
23
 
14
 
17
 
20
 
41
 
6
 
30
 
57
 
40
 
29
 
49
 
6
Design
Your Own Design
Paste the URL of your CSS below.
Download CSS template