Ext3 users - Apr 2004 - Strange Fedora Booting problem: can not mount "LABEL=*"partitions

Hi, Theodore,

 

 Thanks for your tests. But I got the problem and got it FIXED by
flushing out the contents of /etc/blkid.tab before cloned clients
reboot.

 

 I'm using the e2fsprogs-1.34-1 comes with Fedora Core 1.

hello06.alexa.com root 135% rpm -qf /sbin/fsck

e2fsprogs-1.34-1

hello06.alexa.com root 136%

 

 and the original contents in /etc/blkid.tab are attaching below, while
my cloned machine is in fact, a PATA IDE machine. All the /dev/sd*
should be /dev/hd* instead. Thanks.

 

hello06.alexa.com root 168% cat blkid.tab

<device DEVNO="0x0806" TIME="1079722753"
TYPE="swap">/dev/sda6</device>

<device DEVNO="0x0807" TIME="1079722753"
TYPE="swap">/dev/sda7</device>

<device DEVNO="0x0808" TIME="1079722753"
TYPE="swap">/dev/sda8</device>

<device DEVNO="0x080a" TIME="1079722753"
LABEL="/var"
UUID="a4e5efd0-a648-472d-a70d-737461f2acf6" SEC_TYPE="ext3"
TYPE="ext2">/dev/sda10</device>

<device DEVNO="0x0811" TIME="1079722753"
LABEL="/1"
UUID="9b90d679-275e-4656-916f-21ce963677e7" SEC_TYPE="ext3"
TYPE="ext2">/dev/sdb1</device>

<device DEVNO="0x2201" TIME="1079658729"
LABEL="/1"
UUID="9b90d679-275e-4656-916f-21ce963677e7" SEC_TYPE="ext3"
TYPE="ext2">/dev/hdg1</device>

<device DEVNO="0x2106" TIME="1079658729"
TYPE="swap">/dev/hde6</device>

<device DEVNO="0x2107" TIME="1079658729"
TYPE="swap">/dev/hde7</device>

<device DEVNO="0x2108" TIME="1079658729"
TYPE="swap">/dev/hde8</device>

<device DEVNO="0x210a" TIME="1079658729"
LABEL="/var"
UUID="a4e5efd0-a648-472d-a70d-737461f2acf6" SEC_TYPE="ext3"
TYPE="ext2">/dev/hde10</device>

<device DEVNO="0x0801" TIME="1079722753"
LABEL="/"
UUID="a3866268-8cf6-4b88-b3d9-e9623a1763a4" SEC_TYPE="ext3"
TYPE="ext2">/dev/sda1</device>

<device DEVNO="0x0805" TIME="1079722753"
LABEL="/usr"
UUID="6ae04a60-5e15-44d7-a009-dca14f2ad01b" SEC_TYPE="ext3"
TYPE="ext2">/dev/sda5</device>

<device DEVNO="0x0809" TIME="1079722753"
LABEL="/alexa"
UUID="06009234-95d5-4088-87c0-5bde8470a8ba" SEC_TYPE="ext3"
TYPE="ext2">/dev/sda9</device>

<device DEVNO="0x080b" TIME="1079722753"
LABEL="/0"
UUID="655e3381-7157-463e-aaed-fe45a9617e79" SEC_TYPE="ext3"
TYPE="ext2">/dev/sda11</device>

<device DEVNO="0x0821" TIME="1079722753"
LABEL="/2"
UUID="b0971594-1529-406a-b969-278d4f770cb3" SEC_TYPE="ext3"
TYPE="ext2">/dev/sdc1</device>

<device DEVNO="0x0831" TIME="1079722753"
LABEL="/3"
UUID="46eaafeb-0fb7-4647-91e1-c9fe9a1374a7" SEC_TYPE="ext3"
TYPE="ext2">/dev/sdd1</device>

hello06.alexa.com root 169%

 

 

Thanks.

--Guolin Cheng

 

 

 

-----Original Message-----
From: Theodore Ts'o [mailto:tytso at mit.edu] 
Sent: Monday, April 19, 2004 3:28 PM
To: Guolin Cheng
Cc: Stephen C. Tweedie; ops; Fedora (E-mail); Redhat Ext3 (E-mail); Jeff
Garzik
Subject: Re: Strange Fedora Booting problem: can not mount
"LABEL=*"partitions

 

On Thu, Apr 15, 2004 at 10:50:03AM -0700, Guolin Cheng wrote:
> 
>  Thanks. But the problem got debugged&fixed, the answer was post on
> fedora-list about 2 weeks ago. 
> 
> The problem is: the /etc/blkid.tab file works as an old unappropriate
> disk partitions cache for fsck|blkid commands when stystem image is
> installed to a different arch (scsi->ide) machine, the old cache will
> mislead fsck|blkid at the first run and only the first run, since the
> first run will update /etc/blkid.tab file. 
 

What version of e2fsprogs were you testing with?  I've just tested

using the latest version of e2fsprogs, and it works just fine.  In the

test below, I corrupt /etc/blkid.tab by swapping /dev/hda and

/dev/hdc.  This might correspond with might happen after disks get

switched around.  I run fsck in debugging mode to make sure it gets

the correct devices, despite the corrupted /etc/blkid.tab file.  As

you can see, it works fine:

 

# sed -e 's/hda/hdz/g' -e 's/hdc/hda/g' -e 's/hdz/hdc/'
-e
's/0x03/0xf3/' -e s'/0x16/0x03/' -e 's/0xf3/0x16/'
/etc/blkid.tab  >
/tmp/blkid.tab.broken

 

# cp /tmp/blkid.tab.broken /etc/blkid.tab

 

# fsck -AVN -a

fsck 1.35 (28-Feb-2004)

Checking all file systems.

[/sbin/fsck.ext3 (1) -- /] fsck.ext3 -a /dev/hda1

[/sbin/fsck.ext3 (1) -- /usr] fsck.ext3 -a /dev/hda3

[/sbin/fsck.ext3 (1) -- /debian] fsck.ext3 -a /dev/hda5

 

# e2label /dev/hda1

root

 

# e2label /dev/hda3

usr

 

# e2label /dev/hda5

Debian

 

The only thing I can think of is that you might have been using an

older version of e2fsprogs that was buggy....

 

                                    - Ted

-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://listman.redhat.com/archives/ext3-users/attachments/20040419/bd703222/attachment.htm>

On Mon, Apr 19, 2004 at 05:03:25PM -0700, Guolin Cheng
wrote:
> Hi, Theodore,
> 
>  Thanks for your tests. But I got the problem and got it FIXED by
> flushing out the contents of /etc/blkid.tab before cloned clients
> reboot.
> 
>  I'm using the e2fsprogs-1.34-1 comes with Fedora Core 1.
Can you do me a favor and please try replicating my test using the
e2fsprogs-1.34-1 with Fedora Core 1, instead of just asserting that
the problem exists?  What you are describing should not be occurring
at all; you should not need to do the workaround.  

In fact, if you need to do it, it's a bad, Bad, BAD problem in the
blkid library.  This is why I'm so interested in tracking this down.

I have not been able to replicate the problem using the most recent
version of e2fsprogs, and I would like to know the problem is still
there.  If you could try replicating the test which I did in my last
e-mail message, and the problem exists in Fedora Core 1, but does not
exist in recent versions of e2fsprogs, then I can be confident in
knowing that the problem has been fixed.  (Or was introduced in the
Fedora Core 1 changes, in which case it's not my problem.  :-)

						- Ted