Potential ext4 corruption in 6.1.64-cip10

Potential ext4 corruption in 6.1.64-cip10

[Jan Kiszka]

Hi all,

seems also our 6.1 CIP kernel is affected by [1]. Please do not use in
production if you have an ext4 filesystem!

Jan

[1] https://lore.kernel.org/stable/20231205122122.dfhhoaswsfscuhc3@quack3/

-- 
Siemens AG, Technology
Linux Expert Center

Re: Potential ext4 corruption in 6.1.64-cip10

[Pavel Machek]

[-- Attachment #1: Type: text/plain, Size: 543 bytes --]

Hi!

> seems also our 6.1 CIP kernel is affected by [1]. Please do not use in
> production if you have an ext4 filesystem!

This seems to affect O_SYNC / O_DIRECT users, so I'm not sure how
common problem that is.

6.1.66 is out, and there's matching 6.1.66-rt. I can release one or
both, but I don't see any confirmation that it is fixed in .66 or how.

Best regards,
								Pavel
-- 
DENX Software Engineering GmbH,        Managing Director: Erika Unter
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

Re: Potential ext4 corruption in 6.1.64-cip10

[Jan Kiszka]

On 11.12.23 10:36, Pavel Machek wrote:
> Hi!
> 
>> seems also our 6.1 CIP kernel is affected by [1]. Please do not use in
>> production if you have an ext4 filesystem!
> 
> This seems to affect O_SYNC / O_DIRECT users, so I'm not sure how
> common problem that is.

Well, Debian considered it grave:
https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1057843

> 
> 6.1.66 is out, and there's matching 6.1.66-rt. I can release one or
> both, but I don't see any confirmation that it is fixed in .66 or how.
> 

I would expect some reference to this commit in 6.1.y as well, but there
is nothing yet, no commit after dc4542861ec8dde92c3c8a5139bc412860aebe60
touching that file.

Jan

-- 
Siemens AG, Technology
Linux Expert Center

Re: Potential ext4 corruption in 6.1.64-cip10

[Pavel Machek]

[-- Attachment #1: Type: text/plain, Size: 1574 bytes --]

Hi!

> >> seems also our 6.1 CIP kernel is affected by [1]. Please do not use in
> >> production if you have an ext4 filesystem!
> > 
> > This seems to affect O_SYNC / O_DIRECT users, so I'm not sure how
> > common problem that is.
> 
> Well, Debian considered it grave:
> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1057843

Aha, thanks for a hint, there are good pointers in that.

> > 6.1.66 is out, and there's matching 6.1.66-rt. I can release one or
> > both, but I don't see any confirmation that it is fixed in .66 or how.
> > 
> 
> I would expect some reference to this commit in 6.1.y as well, but there
> is nothing yet, no commit after dc4542861ec8dde92c3c8a5139bc412860aebe60
> touching that file.

So this should be fixed in

commit ea574927fc0bc343016ea3337fcfc0b3fb26fe08
Author: Christoph Hellwig <hch@lst.de>
Date:   Thu Jun 1 16:58:54 2023 +0200

    iomap: update ki_pos a little later in iomap_dio_complete
    
    commit 936e114a245b6e38e0dbf706a67e7611fc993da1 upstream.
    
    Move the ki_pos update down a bit to prepare for a better common helper
    that invalidates pages based of an iocb.
    
    Link: https://lkml.kernel.org/r/20230601145904.1385409-3-hch@lst.de

...and that's indeed in 6.1.66, and looks like it fixes relevant area.

I'll just do 6.1.66 based -cip and -cip-rt releases, unless someone
stops me.

Best regards,
								Pavel
-- 
DENX Software Engineering GmbH,        Managing Director: Erika Unter
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

Re: Potential ext4 corruption in 6.1.64-cip10

[Jan Kiszka]

On 11.12.23 10:53, Pavel Machek wrote:
> Hi!
> 
>>>> seems also our 6.1 CIP kernel is affected by [1]. Please do not use in
>>>> production if you have an ext4 filesystem!
>>>
>>> This seems to affect O_SYNC / O_DIRECT users, so I'm not sure how
>>> common problem that is.
>>
>> Well, Debian considered it grave:
>> https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1057843
> 
> Aha, thanks for a hint, there are good pointers in that.
> 
>>> 6.1.66 is out, and there's matching 6.1.66-rt. I can release one or
>>> both, but I don't see any confirmation that it is fixed in .66 or how.
>>>
>>
>> I would expect some reference to this commit in 6.1.y as well, but there
>> is nothing yet, no commit after dc4542861ec8dde92c3c8a5139bc412860aebe60
>> touching that file.
> 
> So this should be fixed in
> 
> commit ea574927fc0bc343016ea3337fcfc0b3fb26fe08
> Author: Christoph Hellwig <hch@lst.de>
> Date:   Thu Jun 1 16:58:54 2023 +0200
> 
>     iomap: update ki_pos a little later in iomap_dio_complete
>     
>     commit 936e114a245b6e38e0dbf706a67e7611fc993da1 upstream.
>     
>     Move the ki_pos update down a bit to prepare for a better common helper
>     that invalidates pages based of an iocb.
>     
>     Link: https://lkml.kernel.org/r/20230601145904.1385409-3-hch@lst.de
> 
> ...and that's indeed in 6.1.66, and looks like it fixes relevant area.
> 

Ah, that missing dependency of the patch - subtle as it was written.

> I'll just do 6.1.66 based -cip and -cip-rt releases, unless someone
> stops me.

That sounds good.

Jan

-- 
Siemens AG, Technology
Linux Expert Center

WIFI in 6.1.X was Re: Potential ext4 corruption in 6.1.64-cip10

[Pavel Machek]

[-- Attachment #1: Type: text/plain, Size: 674 bytes --]

Hi!

> Ah, that missing dependency of the patch - subtle as it was written.
> 
> > I'll just do 6.1.66 based -cip and -cip-rt releases, unless someone
> > stops me.

Ok, this is not good week for 6.1.X.

https://www.phoronix.com/news/Linux-6.6.6-Released

There's wifi regression in 6.1.66. I guess that's better than ext4
regression, and 6.1.67 is not nice round number, so I guess we do not
need to do anything just now... but if someone is affected by the
problem let me know.

Best regards,
								Pavel
-- 
DENX Software Engineering GmbH,        Managing Director: Erika Unter
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 195 bytes --]

Re: WIFI in 6.1.X was Re: Potential ext4 corruption in 6.1.64-cip10

[Jan Kiszka]

On 11.12.23 21:27, Pavel Machek wrote:
> Hi!
> 
>> Ah, that missing dependency of the patch - subtle as it was written.
>>
>>> I'll just do 6.1.66 based -cip and -cip-rt releases, unless someone
>>> stops me.
> 
> Ok, this is not good week for 6.1.X.
> 
> https://www.phoronix.com/news/Linux-6.6.6-Released
> 
> There's wifi regression in 6.1.66. I guess that's better than ext4
> regression, and 6.1.67 is not nice round number, so I guess we do not
> need to do anything just now... but if someone is affected by the
> problem let me know.

Yeah, time to generally reflect on testing and other workflows, it looks
like.

Jan

-- 
Siemens AG, Technology
Linux Expert Center