Question about JBD2 transactions

Question about JBD2 transactions

[Abraham Choi]

Hi all,

Frist of all, thank you for Ext4's developer.

I have encounterd a crash problem of ext4 file system on Kernel 3.18.20 
with mounting 
rw,seclabel,nosuid,nodev,relatime,discard,noauto_da_alloc,data=ordered 0 0.
[   22.361511] EXT4-fs error (device sda8): ext4_mb_generate_buddy:757: 
group 0, block bitmap and bg descriptor inconsistent: 21208 vs 21219 free 
clusters
I think, the reason is that recovered transactions has not recovered over 
39809 during booting.
Booting N-2
	[   16.416979] (jbd2_journal_recover, 279): JBD2: recovery, exit 
status 0, recovered transactions 39665 to 39728
	Last commit_sequence=39796
Booting N-1
	[   16.626582] (jbd2_journal_recover, 279): JBD2: recovery, exit 
status 0, recovered transactions 39730 to 39797
	Last commit_sequence=39866
Booting N
	[   14.118542] (jbd2_journal_recover, 279): JBD2: recovery, exit 
status 0, recovered transactions 39799 to 39809
	
The root casue will be explained the more detail in the follwing.
Jounral Blocks
... <Descriptor block #39808> <Data blocks> <Commit block #39808> 
<All zero block> <Descriptor block #39809> <Data blocks> <Commit block 
#39809> ... 
I'm wondering that journal has all zero data, only 1 block.
How it can be possible?
Any tips or help would be appreciated.

Regards,
Abraham Choi

Re: Question about JBD2 transactions

[Theodore Ts'o]

On Mon, Mar 21, 2016 at 02:23:55AM +0000, Abraham Choi wrote:
> Hi all,
> 
> Frist of all, thank you for Ext4's developer.
> 
> I have encounterd a crash problem of ext4 file system on Kernel 3.18.20 
> with mounting 
> rw,seclabel,nosuid,nodev,relatime,discard,noauto_da_alloc,data=ordered 0 0.
> [   22.361511] EXT4-fs error (device sda8): ext4_mb_generate_buddy:757: 
> group 0, block bitmap and bg descriptor inconsistent: 21208 vs 21219 free 
> clusters
> I think, the reason is that recovered transactions has not recovered over 
> 39809 during booting.
> Booting N-2
> 	[   16.416979] (jbd2_journal_recover, 279): JBD2: recovery, exit 
> status 0, recovered transactions 39665 to 39728
> 	Last commit_sequence=39796
> Booting N-1
> 	[   16.626582] (jbd2_journal_recover, 279): JBD2: recovery, exit 
> status 0, recovered transactions 39730 to 39797
> 	Last commit_sequence=39866
> Booting N
> 	[   14.118542] (jbd2_journal_recover, 279): JBD2: recovery, exit 
> status 0, recovered transactions 39799 to 39809
> 	
> The root casue will be explained the more detail in the follwing.
> Jounral Blocks
> ... <Descriptor block #39808> <Data blocks> <Commit block #39808> 
> <All zero block> <Descriptor block #39809> <Data blocks> <Commit block 
> #39809> ... 
> I'm wondering that journal has all zero data, only 1 block.
> How it can be possible?
> Any tips or help would be appreciated.

Can you reproduce this?  And how did you dump out the journal blocks?
There shouldn't be an all zero block after the commit block, unless
that really was the last transaction.

				- Ted