Re: Problems with journal size option on mkfs

Re: Problems with journal size option on mkfs

[Steve Pratt]

Chris Mason wrote:
>On Wed, 2002-05-15 at 16:12, Steve Pratt wrote:
>>
>> I am in the process of writing a File System Interface Module (FSIM) for
>> EVMS that will interface with ReiserFS and am having a problem with the
-s
>> (journal size) option on mkfs.reiserfs.  The problem is that if I
specify
>> any journal size the mkfs fails.  By fails, what I mean is that the
device
>> is not mountable, I don't actually get any error message from mkfs.
This
>> happens even when I give it the default size of 8193 blocks.  I didn't
see
>> anything in the faq about this option not working, but I tried it on 2
>> different machines with the same results.  I am running tools level
3.x.1b
>> and stock 2.4.17 kernel + EVMS + JFS + kgdb + vfslock.
>>
>> Here is a sample of the output, the only line that changes is the
reference
>> to non-standard journal

>The kernel side of the dedicated logging device code has not been sent
>for inclusion in 2.4.x yet.  The beta patch below implements it, against
>2.4.19-pre7.

So the -s option only works with offline log devices?  It doesn't say
anything about this in the man pages.  Also, wouldn't the mkfs code just
ignore this option if you did not specify an external log device (like the
example)?

Just looked at man page again, and it indicates that this should work for
inline logs as well.

Steve

EVMS Development - http://www.sf.net/projects/evms
Linux Technology Center - IBM Corporation
(512) 838-9763  EMAIL: SLPratt@US.IBM.COM

Re: Problems with journal size option on mkfs

[Chris Mason]

On Wed, 2002-05-15 at 17:49, Steve Pratt wrote:
> 
> Chris Mason wrote:
> >On Wed, 2002-05-15 at 16:12, Steve Pratt wrote:
> >>
> >> I am in the process of writing a File System Interface Module (FSIM) for
> >> EVMS that will interface with ReiserFS and am having a problem with the
> -s
> >> (journal size) option on mkfs.reiserfs.  The problem is that if I
> specify
> >> any journal size the mkfs fails.  By fails, what I mean is that the
> device
> >> is not mountable, I don't actually get any error message from mkfs.
> This
> >> happens even when I give it the default size of 8193 blocks.  I didn't
> see
> >> anything in the faq about this option not working, but I tried it on 2
> >> different machines with the same results.  I am running tools level
> 3.x.1b
> >> and stock 2.4.17 kernel + EVMS + JFS + kgdb + vfslock.
> >>
> >> Here is a sample of the output, the only line that changes is the
> reference
> >> to non-standard journal
> 
> >The kernel side of the dedicated logging device code has not been sent
> >for inclusion in 2.4.x yet.  The beta patch below implements it, against
> >2.4.19-pre7.
> 
> So the -s option only works with offline log devices?  It doesn't say
> anything about this in the man pages.  Also, wouldn't the mkfs code just
> ignore this option if you did not specify an external log device (like the
> example)?
> 
> Just looked at man page again, and it indicates that this should work for
> inline logs as well.
> 

It works with both, but the kernel doesn't understand things properly
without the patch.

-chris

Re: Problems with journal size option on mkfs

[Steve Pratt]

Oleg Drokin wrote:
>Hello!

>On Wed, May 15, 2002 at 03:12:40PM -0500, Steve Pratt wrote:

>> I am in the process of writing a File System Interface Module (FSIM) for
>> EVMS that will interface with ReiserFS and am having a problem with the
-s
>> (journal size) option on mkfs.reiserfs.  The problem is that if I
specify
>> any journal size the mkfs fails.  By fails, what I mean is that the
device
>> is not mountable, I don't actually get any error message from mkfs.
This
>> happens even when I give it the default size of 8193 blocks.  I didn't
see

>This is because non-standard journal support is only included in 2.5
kernel
>series.
>2.4 support is available as separate patch for now.


So just to be sure I understand this:the current 2.4 kernel code (2.4.18)
does not include support for ANY non-standard journal options, including
non default size of an inline log??

If this is true, wouldn't it have made sense to put some checking in the
utilities to detect the level of support in the kernel and either put out a
meaningful error message or simply ignore the option?  As it stands now,
from the command line mkfs doesn't even indicate that there was an error.

Even if the ReiserFS utilities don't check, is there some way I can query
the ReiserFS kernel code to detect the presence of these features?

Steve

Re: Problems with journal size option on mkfs

[Oleg Drokin]

Hello!

On Thu, May 16, 2002 at 09:59:46AM -0500, Steve Pratt wrote:
 
> >This is because non-standard journal support is only included in 2.5
> kernel
> >series.
> >2.4 support is available as separate patch for now.
> So just to be sure I understand this:the current 2.4 kernel code (2.4.18)
> does not include support for ANY non-standard journal options, including
> non default size of an inline log??

Yes, this is correct.

> If this is true, wouldn't it have made sense to put some checking in the
> utilities to detect the level of support in the kernel and either put out a
> meaningful error message or simply ignore the option?  As it stands now,
> from the command line mkfs doesn't even indicate that there was an error.

There were no error at mkfs time. May be user is doing mkfs from some bootable
floppy with old kernel and then plans to boot into more modern kernel, for
example.

> Even if the ReiserFS utilities don't check, is there some way I can query
> the ReiserFS kernel code to detect the presence of these features?

Well, the simplest way is to have (or make) small file with reiserfs image
with non-standard journal and try to mount it ;)
Or you can try to mount valid reiserfs fs with jdev= mount option, which
is only present on code with nonstandard journaling support.
In short, there is no easy way. But if you can think of any,
we might implement it if it non-intrusive and easy enough.
Actually this is not very reliable in the sence that you cannot say
which kernel user plans to use later on.

Bye,
    Oleg

Re: Problems with journal size option on mkfs

[Hans Reiser]

Oleg Drokin wrote:

>Hello!
>
>On Thu, May 16, 2002 at 09:59:46AM -0500, Steve Pratt wrote:
> 
>  
>
>>>This is because non-standard journal support is only included in 2.5
>>>      
>>>
>>kernel
>>    
>>
>>>series.
>>>2.4 support is available as separate patch for now.
>>>      
>>>
>>So just to be sure I understand this:the current 2.4 kernel code (2.4.18)
>>does not include support for ANY non-standard journal options, including
>>non default size of an inline log??
>>    
>>
>
>Yes, this is correct.
>
>  
>
>>If this is true, wouldn't it have made sense to put some checking in the
>>utilities to detect the level of support in the kernel and either put out a
>>meaningful error message or simply ignore the option?  As it stands now,
>>from the command line mkfs doesn't even indicate that there was an error.
>>    
>>
>
>There were no error at mkfs time. May be user is doing mkfs from some bootable
>floppy with old kernel and then plans to boot into more modern kernel, for
>example.
>
>  
>
>>Even if the ReiserFS utilities don't check, is there some way I can query
>>the ReiserFS kernel code to detect the presence of these features?
>>    
>>
>
>Well, the simplest way is to have (or make) small file with reiserfs image
>with non-standard journal and try to mount it ;)
>Or you can try to mount valid reiserfs fs with jdev= mount option, which
>is only present on code with nonstandard journaling support.
>In short, there is no easy way. But if you can think of any,
>we might implement it if it non-intrusive and easy enough.
>Actually this is not very reliable in the sence that you cannot say
>which kernel user plans to use later on.
>
>Bye,
>    Oleg
>
>
>  
>
I don't agree with Oleg's answer.  Lack of a meaningful error message is 
a bug.

Hans

Re: Problems with journal size option on mkfs

[Edward Shushkin]

Hans Reiser wrote:
> 
> Oleg Drokin wrote:
> 
> >Hello!
> >
> >On Thu, May 16, 2002 at 09:59:46AM -0500, Steve Pratt wrote:
> >
> >
> >
> >>>This is because non-standard journal support is only included in 2.5
> >>>
> >>>
> >>kernel
> >>
> >>
> >>>series.
> >>>2.4 support is available as separate patch for now.
> >>>
> >>>
> >>So just to be sure I understand this:the current 2.4 kernel code (2.4.18)
> >>does not include support for ANY non-standard journal options, including
> >>non default size of an inline log??
> >>
> >>
> >
> >Yes, this is correct.
> >
> >
> >
> >>If this is true, wouldn't it have made sense to put some checking in the
> >>utilities to detect the level of support in the kernel and either put out a
> >>meaningful error message or simply ignore the option?  As it stands now,
> >>from the command line mkfs doesn't even indicate that there was an error.
> >>
> >>
> >
> >There were no error at mkfs time. May be user is doing mkfs from some bootable
> >floppy with old kernel and then plans to boot into more modern kernel, for
> >example.
> >
> >
> >
> >>Even if the ReiserFS utilities don't check, is there some way I can query
> >>the ReiserFS kernel code to detect the presence of these features?
> >>
> >>
> >
> >Well, the simplest way is to have (or make) small file with reiserfs image
> >with non-standard journal and try to mount it ;)
> >Or you can try to mount valid reiserfs fs with jdev= mount option, which
> >is only present on code with nonstandard journaling support.
> >In short, there is no easy way. But if you can think of any,
> >we might implement it if it non-intrusive and easy enough.
> >Actually this is not very reliable in the sence that you cannot say
> >which kernel user plans to use later on.
> >
> >Bye,
> >    Oleg
> >
> >
> >
> >
> I don't agree with Oleg's answer.  Lack of a meaningful error message is
> a bug.

mkreiserfs allows to create non-standard journal regardless of 
the current kernel version. If it is not correct, we'll modify it.
Thanks,
Edward.

> 
> Hans

Problems with journal size option on mkfs

[Steve Pratt]

I am in the process of writing a File System Interface Module (FSIM) for
EVMS that will interface with ReiserFS and am having a problem with the -s
(journal size) option on mkfs.reiserfs.  The problem is that if I specify
any journal size the mkfs fails.  By fails, what I mean is that the device
is not mountable, I don't actually get any error message from mkfs.  This
happens even when I give it the default size of 8193 blocks.  I didn't see
anything in the faq about this option not working, but I tried it on 2
different machines with the same results.  I am running tools level 3.x.1b
and stock 2.4.17 kernel + EVMS + JFS + kgdb + vfslock.

Here is a sample of the output, the only line that changes is the reference
to non-standard journal


[root@slpratt root]$ mkfs.reiserfs -f -d -s 8193 /dev/evms/test2
mkfs.reiserfs 3.x.1b (2002)
mkfs.reiserfs: Guessing about desired format..
mkfs.reiserfs: Kernel 2.4.17 is running.
Block 16 (0x3ffe) contains super block. Format 3.6 with non-standard
journal
Count of blocks on the device: 32121
Number of blocks consumed by mkreiserfs formatting process: 8212
Free blocks: 23909
Blocksize: 4096
Hash function used to sort names: "r5"
Number of bitmaps: 1
Root block: 8211
Tree height: 2
Objectid map size 2, max 972
Journal parameters:
      Device [0x0]
      Magic [0x6aaeccf4]
      Size 8193 blocks (including 1 for journal header) (first block 18)
      Max transaction length 1024 blocks
      Max batch size 900 blocks
      Max commit age 30
Filesystem state 0x0
sb_version 2
inode generation number: 0
UUID: a7ee44d1-8af3-43f7-af4b-5a979684eff4
Initializing journal - 0%....20%....40%....60%....80%....100%
Syncing..ok
[root@slpratt root]$
[root@slpratt root]$ mount -t reiserfs /dev/evms/test2 /mnt/tmp
mount: wrong fs type, bad option, bad superblock on /dev/evms/test2,
       or too many mounted file systems
[root@slpratt root]$


and ....


[root@slpratt root]$ fsck.reiserfs /dev/evms/test2
reiserfsck 3.x.1b (2002)
Will read-only check consistency of the filesystem on /dev/evms/test2
Will put log info to 'stdout'

Do you want to run this program?[N/Yes] (note need to type Yes):Yes
reiserfsck: either make journal device available or use
--no-journal-available
            if you have the standard journal or if your device is available
            and you specified it correctly, you have to run rebuild-sb
[root@slpratt root]$


Let me know if you need more information.

Steve

EVMS Development - http://www.sf.net/projects/evms
Linux Technology Center - IBM Corporation
(512) 838-9763  EMAIL: SLPratt@US.IBM.COM

Re: Problems with journal size option on mkfs

[Oleg Drokin]

Hello!

On Wed, May 15, 2002 at 03:12:40PM -0500, Steve Pratt wrote:

> I am in the process of writing a File System Interface Module (FSIM) for
> EVMS that will interface with ReiserFS and am having a problem with the -s
> (journal size) option on mkfs.reiserfs.  The problem is that if I specify
> any journal size the mkfs fails.  By fails, what I mean is that the device
> is not mountable, I don't actually get any error message from mkfs.  This
> happens even when I give it the default size of 8193 blocks.  I didn't see

This is because non-standard journal support is only included in 2.5 kernel
series.
2.4 support is available as separate patch for now.

Bye,
    Oleg

Re: Problems with journal size option on mkfs

[Hans Reiser]

Oleg Drokin wrote:

>Hello!
>
>On Wed, May 15, 2002 at 03:12:40PM -0500, Steve Pratt wrote:
>
>  
>
>>I am in the process of writing a File System Interface Module (FSIM) for
>>EVMS that will interface with ReiserFS and am having a problem with the -s
>>(journal size) option on mkfs.reiserfs.  The problem is that if I specify
>>any journal size the mkfs fails.  By fails, what I mean is that the device
>>is not mountable, I don't actually get any error message from mkfs.  This
>>happens even when I give it the default size of 8193 blocks.  I didn't see
>>    
>>
>
>This is because non-standard journal support is only included in 2.5 kernel
>series.
>2.4 support is available as separate patch for now.
>
>Bye,
>    Oleg
>
>
>  
>
journal resizing and relocating will go into the next pre1.

Hans

Re: Problems with journal size option on mkfs

[Yury Umanets]

Hans Reiser wrote:

> Oleg Drokin wrote:
>
>> Hello!
>>
>> On Wed, May 15, 2002 at 03:12:40PM -0500, Steve Pratt wrote:
>>
>>  
>>
>>> I am in the process of writing a File System Interface Module (FSIM) 
>>> for
>>> EVMS that will interface with ReiserFS and am having a problem with 
>>> the -s
>>> (journal size) option on mkfs.reiserfs.  The problem is that if I 
>>> specify
>>> any journal size the mkfs fails.  By fails, what I mean is that the 
>>> device
>>> is not mountable, I don't actually get any error message from mkfs.  
>>> This
>>> happens even when I give it the default size of 8193 blocks.  I 
>>> didn't see
>>>   
>>
>>
>> This is because non-standard journal support is only included in 2.5 
>> kernel
>> series.
>> 2.4 support is available as separate patch for now.
>>
>> Bye,
>>    Oleg
>
> journal resizing and relocating will go into the next pre1.
>
progsreiserfs (libreiserfs for reiser3) could relocate and resize 
journal three versions ago.

Re: Problems with journal size option on mkfs

[Hans Reiser]

Yury Umanets wrote:

> Hans Reiser wrote:
>
>> Oleg Drokin wrote:
>>
>>> Hello!
>>>
>>> On Wed, May 15, 2002 at 03:12:40PM -0500, Steve Pratt wrote:
>>>
>>>  
>>>
>>>> I am in the process of writing a File System Interface Module 
>>>> (FSIM) for
>>>> EVMS that will interface with ReiserFS and am having a problem with 
>>>> the -s
>>>> (journal size) option on mkfs.reiserfs.  The problem is that if I 
>>>> specify
>>>> any journal size the mkfs fails.  By fails, what I mean is that the 
>>>> device
>>>> is not mountable, I don't actually get any error message from 
>>>> mkfs.  This
>>>> happens even when I give it the default size of 8193 blocks.  I 
>>>> didn't see
>>>>   
>>>
>>>
>>>
>>> This is because non-standard journal support is only included in 2.5 
>>> kernel
>>> series.
>>> 2.4 support is available as separate patch for now.
>>>
>>> Bye,
>>>    Oleg
>>
>>
>> journal resizing and relocating will go into the next pre1.
>>
> progsreiserfs (libreiserfs for reiser3) could relocate and resize 
> journal three versions ago.
>
>
>
>
>
>
>
and the kernel code....

Re: Problems with journal size option on mkfs

[Chris Mason]

On Thu, 2002-05-16 at 06:47, Hans Reiser wrote:

> journal resizing and relocating will go into the next pre1.
> 

That makes for a huge set of changes we have planned for pre1.  We need
to limit ourselves to one significant change per prerelease, otherwise
debugging is very difficult.

The journal relocation stuff needs to be tossed onto an FTP site and
widely tested before we submit for a stable kernel series.  It clearly
has not been widely tested 2.5.x ;-)

-chris

Re: Problems with journal size option on mkfs

[Hans Reiser]

Chris Mason wrote:

>On Thu, 2002-05-16 at 06:47, Hans Reiser wrote:
>
>  
>
>>journal resizing and relocating will go into the next pre1.
>>
>>    
>>
>
>That makes for a huge set of changes we have planned for pre1.  We need
>to limit ourselves to one significant change per prerelease, otherwise
>debugging is very difficult.
>
>The journal relocation stuff needs to be tossed onto an FTP site and
>widely tested before we submit for a stable kernel series.  It clearly
>has not been widely tested 2.5.x ;-)
>
>-chris
>
>
>
>
>  
>
Yes, I admit I was thinking as I pressed send about just how much stuff 
would be going into pre1 and wondering if it should be broken into pre1, 
pre2, and pre3.

Probably the journal relocation stuff, being the most tested in 2.5 (do 
any reiserfs users use it in 2.5 though?), should go into pre3.

The bitmap code and the file write code should go into pre1, along with 
the cleanups.

Feel free to suggest different orderings, though maybe we should only 
think about this when it is time for it....

Hans

Re: Problems with journal size option on mkfs

[Chris Mason]

On Thu, 2002-05-16 at 08:00, Hans Reiser wrote:
>
> Yes, I admit I was thinking as I pressed send about just how much stuff 
> would be going into pre1 and wondering if it should be broken into pre1, 
> pre2, and pre3.
> 
> Probably the journal relocation stuff, being the most tested in 2.5 (do 
> any reiserfs users use it in 2.5 though?), should go into pre3.
> 
> The bitmap code and the file write code should go into pre1, along with 
> the cleanups.
> 
> Feel free to suggest different orderings, though maybe we should only 
> think about this when it is time for it....

I think we should start with the speedups only.  Lets get the bitmap
code in 2.5.x (ASAP, I'm willing to merge if Oleg is busy) and the file
write changes seem to have a little ways to go.

I believe the relocation stuff should be tested in andrea's and/or
Alan's kernels before we submit for 2.4.21preX.  We shouldn't mix
must-have speedups with nice to have features (as large as the
relocation anyway) in the same 2.4.x series.

-chris