f.s. corruption ext3, kernel 4.4.8

f.s. corruption ext3, kernel 4.4.8

[Barry Kauler]

I sent this direct to Greg KH yesterday morning. 24 hours later got an
automated reply to send my email to one of the lists.

I'm a user, not a kernel developer, gotta post this somewhere.

I reckon this is urgent.
I have rolled back from 4.4.8 to 4.4.7, have used heavily for 24 hours
since then,
no problem. So, it's the kernel, not my laptop.

Here is my msg to Greg:
-------------------------
Greg,
I thought that I had better alert you to this, just in case.

I posted to my blog, quoting it here:

--------------------
As I reported last night, I compiled the 4.4.8 kernel

This morning, powered-up my laptop, clicked on "sda5", my main working
partition, about 300GB with ext3 filesystem.

I edited 'quicksetup', to fix the firewall checkbox, then went to save
and got an I/O error.

I found that sda5 has been mounted "ro". I unmounted it,
clicked again to mount it, this time it mounted "rw", but looking at
the filesystem with ROX-Filer, I find corruption, some directories have
become I/O errors and replaced by triangle-with-exclamation-mark
icons.

I rebooted, with Quirky 8.0 and 4.4.7 kernel. Mounted sda5 and now the
filesystem looks OK in ROX-Filer. Oh dear.

I unmounted sda5, started "fsck.ext3 -p /dev/sda5", which reported
errors and check is forced. Afterward, filesystem looks ok.

Hard disk or RAM failure, maybe. But it is quite a coincidence that I
compiled 4.4.8 last night. Then, going back to 4.4.7, my filesystem
looks OK again.
-----------------------

Of course, when it mounted "ro" the first time, that should have set
off alarm bells and I should probably have backed up immediately.

It may just be my laptop drive or RAM, but I thought it wise to alert
you, just in case 4.4.8 is the cause. It could be that there is
something with the ext3 driver or elsewhere that is causing the
corruption.

Back on 4.4.7, I seem to be OK. Right now, doing a backup of the partition.

Regards,
Barry Kauler

Re: f.s. corruption ext3, kernel 4.4.8

[Theodore Ts'o]

On Mon, Apr 25, 2016 at 09:46:19AM +0800, Barry Kauler wrote:
> I sent this direct to Greg KH yesterday morning. 24 hours later got an
> automated reply to send my email to one of the lists.
> 
> I'm a user, not a kernel developer, gotta post this somewhere.
> 
> I reckon this is urgent.
> I have rolled back from 4.4.8 to 4.4.7, have used heavily for 24 hours
> since then,
> no problem. So, it's the kernel, not my laptop.

No one else has reported any problems to date, and there are only two
changes in ext4 code between v4.4.7 and v4.4.8, neither of which seem
at all likely to cause file system corruption.

ee8516a ext4: ignore quota mount options if the quota feature is enabled
321299a ext4: add lockdep annotations for i_data_sem

If you can send us the "EXT4-fs error" message that would have been
logged before the file system went read-only, or the output from
e2fsck, that might be helpful in matching up your report from any
others, or to have some insight into what might have happened.

Unfortunately, without any more information there's not much we can do
with your report other than to keep an ear out in case anyone else
complains.

Thanks,

					- Ted

Re: f.s. corruption ext3, kernel 4.4.8

[Barry Kauler]

On 4/25/16, Theodore Ts'o <tytso@mit.edu> wrote:
> On Mon, Apr 25, 2016 at 09:46:19AM +0800, Barry Kauler wrote:
>> I sent this direct to Greg KH yesterday morning. 24 hours later got an
>> automated reply to send my email to one of the lists.
>>
>> I'm a user, not a kernel developer, gotta post this somewhere.
>>
>> I reckon this is urgent.
>> I have rolled back from 4.4.8 to 4.4.7, have used heavily for 24 hours
>> since then,
>> no problem. So, it's the kernel, not my laptop.
>
> No one else has reported any problems to date, and there are only two
> changes in ext4 code between v4.4.7 and v4.4.8, neither of which seem
> at all likely to cause file system corruption.
>
> ee8516a ext4: ignore quota mount options if the quota feature is enabled
> 321299a ext4: add lockdep annotations for i_data_sem
>
> If you can send us the "EXT4-fs error" message that would have been
> logged before the file system went read-only, or the output from
> e2fsck, that might be helpful in matching up your report from any
> others, or to have some insight into what might have happened.
>
> Unfortunately, without any more information there's not much we can do
> with your report other than to keep an ear out in case anyone else
> complains.
>

Thanks for the response.

My kernel is still not using EXT4_USE_FOR_EXT23

My config is:

CONFIG_EXT2_FS=y
CONFIG_EXT2_FS_XATTR=y
CONFIG_EXT2_FS_POSIX_ACL=y
# CONFIG_EXT2_FS_SECURITY is not set
CONFIG_EXT3_FS=y
CONFIG_EXT3_FS_POSIX_ACL=y
# CONFIG_EXT3_FS_SECURITY is not set
CONFIG_EXT4_FS=y
CONFIG_EXT4_FS_POSIX_ACL=y
# CONFIG_EXT4_FS_SECURITY is not set
CONFIG_EXT4_ENCRYPTION=y
CONFIG_EXT4_FS_ENCRYPTION=y
# CONFIG_EXT4_DEBUG is not set

...is this deprecated?

Just asking, as if everyone else is using EXT4_USE_FOR_EXT23, then
perhaps they won't hit my problem.

Re: f.s. corruption ext3, kernel 4.4.8

[Eric Sandeen]

On 4/24/16 9:33 PM, Barry Kauler wrote:
> On 4/25/16, Theodore Ts'o <tytso@mit.edu> wrote:
>> On Mon, Apr 25, 2016 at 09:46:19AM +0800, Barry Kauler wrote:
>>> I sent this direct to Greg KH yesterday morning. 24 hours later got an
>>> automated reply to send my email to one of the lists.
>>>
>>> I'm a user, not a kernel developer, gotta post this somewhere.
>>>
>>> I reckon this is urgent.
>>> I have rolled back from 4.4.8 to 4.4.7, have used heavily for 24 hours
>>> since then,
>>> no problem. So, it's the kernel, not my laptop.
>>
>> No one else has reported any problems to date, and there are only two
>> changes in ext4 code between v4.4.7 and v4.4.8, neither of which seem
>> at all likely to cause file system corruption.
>>
>> ee8516a ext4: ignore quota mount options if the quota feature is enabled
>> 321299a ext4: add lockdep annotations for i_data_sem
>>
>> If you can send us the "EXT4-fs error" message that would have been
>> logged before the file system went read-only, or the output from
>> e2fsck, that might be helpful in matching up your report from any
>> others, or to have some insight into what might have happened.
>>
>> Unfortunately, without any more information there's not much we can do
>> with your report other than to keep an ear out in case anyone else
>> complains.
>>
> 
> Thanks for the response.
> 
> My kernel is still not using EXT4_USE_FOR_EXT23
> 
> My config is:
> 
> CONFIG_EXT2_FS=y
> CONFIG_EXT2_FS_XATTR=y
> CONFIG_EXT2_FS_POSIX_ACL=y
> # CONFIG_EXT2_FS_SECURITY is not set
> CONFIG_EXT3_FS=y
> CONFIG_EXT3_FS_POSIX_ACL=y
> # CONFIG_EXT3_FS_SECURITY is not set
> CONFIG_EXT4_FS=y
> CONFIG_EXT4_FS_POSIX_ACL=y
> # CONFIG_EXT4_FS_SECURITY is not set
> CONFIG_EXT4_ENCRYPTION=y
> CONFIG_EXT4_FS_ENCRYPTION=y
> # CONFIG_EXT4_DEBUG is not set
> 
> ...is this deprecated?
> 
> Just asking, as if everyone else is using EXT4_USE_FOR_EXT23, then
> perhaps they won't hit my problem.

Ted asked for the actual error messages; without that, there's really
nothing to go on here.  It's hard to debug from narrative.  ;)

As for ext3 - that was removed in v4.3, so it's unlikely that you are
using the ext3 driver at this point; in 4.4 you can only be using
the ext4 driver for an ext3 filesystem.

Again, actual copies of the error messages might tell us something.  

Thanks,
-Eric