xfs_read_buf error 5.

xfs_read_buf error 5.

[Nikhil Kulkarni]

Hi,

 

We are running the latest stable kernel 2.6.18 on RedHat Enterprise 4
Update 4.

When I create an xfs partition and assign it a size > 2GB, I get the
following errors during bootup:

 

Sep 21 15:44:06 ssimppi7 kernel:  sda: sda1

Sep 21 15:44:06 ssimppi7 kernel: sda1: rw=0, want=7167993856,
limit=2873026796

Sep 21 15:44:06 ssimppi7 kernel: I/O error in filesystem ("sda1")
meta-data dev sda1 block 0x1ab3ee7ff       ("xfs_read_buf") error 5 buf
count 512

Sep 21 15:43:51 ssimppi7 mount: mount: /dev/sda1: can't read superblock

 

If the partition size is < 2GB then everything works smoothly. The
filesystem mounts successfully.

 

I searched the archives but could not find anything that would help me
in debugging this issue.

 

I was wondering if anyone else has reported the same issue or if I am
missing a step or if this is a defect in the code or

something??

 

Your response is greatly appreciated.

 

Thanks,

Nikhil



[[HTML alternate version deleted]]

Re: xfs_read_buf error 5.

[David Chinner]

On Thu, Sep 21, 2006 at 03:52:18PM -0700, Nikhil Kulkarni wrote:
> Hi,
> 
>  
> 
> We are running the latest stable kernel 2.6.18 on RedHat Enterprise 4
> Update 4.
> 
> When I create an xfs partition and assign it a size > 2GB, I get the
> following errors during bootup:
....
> If the partition size is < 2GB then everything works smoothly. The
> filesystem mounts successfully.

Sounds like a partition problem, not an XFS problem...
What is the size of the partitions in both cases according to
/proc/partitions?

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

RE: xfs_read_buf error 5.

[Nikhil Kulkarni]

Hi David,

Hi David,

Thanks for your prompt response, I really appreciate it!!! I made a
major tying mistake in describing the initial size of the partitions. We
are having issues when we assign a size > 2TB and not 2GB. I'm sorry
about the typo.

Here are the 2 /proc/partition files:

This is the one where the partition size is 3.5T

[nikhil@nkulkarni tmp]$ more xfs-log-3.5T
[root@ssimppi7 log]# cat /proc/partitions
major minor  #blocks  name

   8     0 4093902848 sda
   8     1 1436513398 sda1
   8    16  143257600 sdb
   8    17   10482381 sdb1
   8    18   10482412 sdb2
   8    19    4192965 sdb3
   8    20          1 sdb4
   8    21  118093783 sdb5
   8    32  878837760 sdc
   8    33   12811806 sdc1
   8    34   64010992 sdc2
   8    35       8032 sdc3
   8    36          1 sdc4
   8    37   19197643 sdc5
   8    38  100582933 sdc6
   8    39  682224291 sdc7
 253     0  136445952 dm-0
 253     1  545775616 dm-1

This is the one where the partition size is 2TB:

[nikhil@nkulkarni tmp]$ more xfs-log-2T
root@ssimppi7 log]# cat /proc/partitions
major minor  #blocks  name

   8     0 4093902848 sda
   8     1 2047998298 sda1
   8    16  143257600 sdb
   8    17   10482381 sdb1
   8    18   10482412 sdb2
   8    19    4192965 sdb3
   8    20          1 sdb4
   8    21  118093783 sdb5
   8    32  878837760 sdc
   8    33   12811806 sdc1
   8    34   64010992 sdc2
   8    35       8032 sdc3
   8    36          1 sdc4
   8    37   19197643 sdc5
   8    38  100582933 sdc6
   8    39  682224291 sdc7
 253     0  136445952 dm-0
 253     1  545775616 dm-1
[nikhil@nkulkarni tmp]$


I think you are right. The partitions are not set up correctly.

Do you know on a 2.5 kernel on a 32 bit operating system, which tool can
be used to setup partitions for sizes up to 4TB or 8TB?
fdisk on a 32 bit os does not work correctly. I thought parted worked
but apparently it does not either.

Any suggestions how I can utilize the entire 4TB file system using xfs?

Thanks,
Nikhil

-----Original Message-----
From: David Chinner [mailto:dgc@sgi.com] 
Sent: Thursday, September 21, 2006 5:16 PM
To: Nikhil Kulkarni
Cc: xfs@oss.sgi.com
Subject: Re: xfs_read_buf error 5.

On Thu, Sep 21, 2006 at 03:52:18PM -0700, Nikhil Kulkarni wrote:
> Hi,
> 
>  
> 
> We are running the latest stable kernel 2.6.18 on RedHat Enterprise 4
> Update 4.
> 
> When I create an xfs partition and assign it a size > 2GB , I get the
> following errors during bootup:
....
> If the partition size is < 2GB then everything works smoothly. The
> filesystem mounts successfully.

Sounds like a partition problem, not an XFS problem...
What is the size of the partitions in both cases according to
/proc/partitions?

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

Re: xfs_read_buf error 5.

[David Chinner]

On Thu, Sep 21, 2006 at 05:45:32PM -0700, Nikhil Kulkarni wrote:
> Hi David,
> 
> Hi David,
> 
> Thanks for your prompt response, I really appreciate it!!! I made a
> major tying mistake in describing the initial size of the partitions. We
> are having issues when we assign a size > 2TB and not 2GB. I'm sorry
> about the typo.

I was wondering about that ;)

> Here are the 2 /proc/partition files:
> 
> This is the one where the partition size is 3.5T
> 
> [nikhil@nkulkarni tmp]$ more xfs-log-3.5T
> [root@ssimppi7 log]# cat /proc/partitions
> major minor  #blocks  name
> 
>    8     0 4093902848 sda
>    8     1 1436513398 sda1

That says it's only 1.4TB...

> This is the one where the partition size is 2TB:
> 
> [nikhil@nkulkarni tmp]$ more xfs-log-2T
> root@ssimppi7 log]# cat /proc/partitions
> major minor  #blocks  name
> 
>    8     0 4093902848 sda
>    8     1 2047998298 sda1

And that is 2TB.

> I think you are right. The partitions are not set up correctly.
> 
> Do you know on a 2.5 kernel on a 32 bit operating system, which tool can
> be used to setup partitions for sizes up to 4TB or 8TB?
> fdisk on a 32 bit os does not work correctly. I thought parted worked
> but apparently it does not either.

Did you build your kernel with CONFIG_LBD=y? (Block layer option)
This is the option that allows >2TB block devices on 32 bit kernels...

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group

RE: xfs_read_buf error 5.

[Nikhil Kulkarni]

Hi David,

I figured it out. I used the msdos label in parted to while creating
partitions, and that is why they got all screwed up. I then switched to 
"gpt" and now everything works like a charm!!!!

Thanks very much for all your help.

Nikhil

-----Original Message-----
From: David Chinner [mailto:dgc@sgi.com] 
Sent: Thursday, September 21, 2006 7:14 PM
To: Nikhil Kulkarni
Cc: David Chinner; xfs@oss.sgi.com
Subject: Re: xfs_read_buf error 5.

On Thu, Sep 21, 2006 at 05:45:32PM -0700, Nikhil Kulkarni wrote:
> Hi David,
> 
> Hi David,
> 
> Thanks for your prompt response, I really appreciate it!!! I made a
> major tying mistake in describing the initial size of the partitions.
We
> are having issues when we assign a size > 2TB and not 2GB. I'm sorry
> about the typo.

I was wondering about that ;)

> Here are the 2 /proc/partition files:
> 
> This is the one where the partition size is 3.5T
> 
> [nikhil@nkulkarni tmp]$ more xfs-log-3.5T
> [root@ssimppi7 log]# cat /proc/partitions
> major minor  #blocks  name
> 
>    8     0 4093902848 sda
>    8     1 1436513398 sda1

That says it's only 1.4TB...

> This is the one where the partition size is 2TB:
> 
> [nikhil@nkulkarni tmp]$ more xfs-log-2T
> root@ssimppi7 log]# cat /proc/partitions
> major minor  #blocks  name
> 
>    8     0 4093902848 sda
>    8     1 2047998298 sda1

And that is 2TB.

> I think you are right. The partitions are not set up correctly.
> 
> Do you know on a 2.5 kernel on a 32 bit operating system, which tool
can
> be used to setup partitions for sizes up to 4TB or 8TB?
> fdisk on a 32 bit os does not work correctly. I thought parted worked
> but apparently it does not either.

Did you build your kernel with CONFIG_LBD=y? (Block layer option)
This is the option that allows >2TB block devices on 32 bit kernels...

Cheers,

Dave.
-- 
Dave Chinner
Principal Engineer
SGI Australian Software Group