Bcache confirmed unsafe with 4.14.0? 4.14.1?

Bcache confirmed unsafe with 4.14.0? 4.14.1?

[Marc MERLIN]

I have bcache | dmcrypt | btrfs, and I was upgrading to 4.14.0 when
someone told me about the corruption problems people noticed in 4.14.0

Is there an understood root cause and something to avoid to not trigger
the bug, or not so much?

I saw a bunch of patches go by and I'm not too clear if 4.14.1 is safe,
or not.

Can someone advise?

Thanks,
Marc
-- 
"A mouse is a device used to point at the xterm you want to type in" - A.S.R.
Microsoft is to operating systems ....
                                      .... what McDonalds is to gourmet cooking
Home page: http://marc.merlins.org/                         | PGP 1024R/763BE901

Re: Bcache confirmed unsafe with 4.14.0? 4.14.1?

[Pavel Goran]

Hello Marc,

Friday, November 24, 2017, 2:41:12 PM, you wrote:

> I have bcache | dmcrypt | btrfs, and I was upgrading to 4.14.0 when
> someone told me about the corruption problems people noticed in 4.14.0

> Is there an understood root cause and something to avoid to not trigger
> the bug, or not so much?

The bug was caused by a change in the block layer code (outside of bcache),
commit 74d46992e0d9dee7f1f376de0d56d31614c8a17a, "block: replace bi_bdev with
a gendisk pointer and partitions index". I don't know if there are any
specific usage patterns that trigger the problem. I strongly suggest not to
run the affected kernel versions on any machine that uses bcache (or, even
better, on any machine at all).

> I saw a bunch of patches go by and I'm not too clear if 4.14.1 is safe,
> or not.

In 4.14.1 (as published here
https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/tag/?h=v4.14.1),
this bug was not fixed yet. However, the very recent 4.14.2 does contain the
fix (commit 62530ed8b1d07a45dec94d46e521c0c6c2d476e6, "bio: ensure
__bio_clone_fast copies bi_partno").

> Can someone advise?

> Thanks,
> Marc

Pavel Goran
  

Re: Bcache confirmed unsafe with 4.14.0? 4.14.1?

[Marc MERLIN]

On Fri, Nov 24, 2017 at 04:22:53PM +0700, Pavel Goran wrote:
> Hello Marc,
> 
> Friday, November 24, 2017, 2:41:12 PM, you wrote:
> 
> > I have bcache | dmcrypt | btrfs, and I was upgrading to 4.14.0 when
> > someone told me about the corruption problems people noticed in 4.14.0
> 
> > Is there an understood root cause and something to avoid to not trigger
> > the bug, or not so much?
> 
> The bug was caused by a change in the block layer code (outside of bcache),
> commit 74d46992e0d9dee7f1f376de0d56d31614c8a17a, "block: replace bi_bdev with
> a gendisk pointer and partitions index". I don't know if there are any
> specific usage patterns that trigger the problem. I strongly suggest not to
> run the affected kernel versions on any machine that uses bcache (or, even
> better, on any machine at all).
> 
> > I saw a bunch of patches go by and I'm not too clear if 4.14.1 is safe,
> > or not.
> 
> In 4.14.1 (as published here
> https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/tag/?h=v4.14.1),
> this bug was not fixed yet. However, the very recent 4.14.2 does contain the
> fix (commit 62530ed8b1d07a45dec94d46e521c0c6c2d476e6, "bio: ensure
> __bio_clone_fast copies bi_partno").

Thanks for clearing that up, I duly appreciate it.

Marc
-- 
"A mouse is a device used to point at the xterm you want to type in" - A.S.R.
Microsoft is to operating systems ....
                                      .... what McDonalds is to gourmet cooking
Home page: http://marc.merlins.org/                         | PGP 1024R/763BE901

Re: Bcache confirmed unsafe with 4.14.0? 4.14.1?

[Justin Cinkelj]

OK. In nekaj besed o teh openmpi/foam testih?
Justin

On 11/24/2017 05:29 PM, Marc MERLIN wrote:
> On Fri, Nov 24, 2017 at 04:22:53PM +0700, Pavel Goran wrote:
>> Hello Marc,
>>
>> Friday, November 24, 2017, 2:41:12 PM, you wrote:
>>
>>> I have bcache | dmcrypt | btrfs, and I was upgrading to 4.14.0 when
>>> someone told me about the corruption problems people noticed in 4.14.0
>>> Is there an understood root cause and something to avoid to not trigger
>>> the bug, or not so much?
>> The bug was caused by a change in the block layer code (outside of bcache),
>> commit 74d46992e0d9dee7f1f376de0d56d31614c8a17a, "block: replace bi_bdev with
>> a gendisk pointer and partitions index". I don't know if there are any
>> specific usage patterns that trigger the problem. I strongly suggest not to
>> run the affected kernel versions on any machine that uses bcache (or, even
>> better, on any machine at all).
>>
>>> I saw a bunch of patches go by and I'm not too clear if 4.14.1 is safe,
>>> or not.
>> In 4.14.1 (as published here
>> https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/tag/?h=v4.14.1),
>> this bug was not fixed yet. However, the very recent 4.14.2 does contain the
>> fix (commit 62530ed8b1d07a45dec94d46e521c0c6c2d476e6, "bio: ensure
>> __bio_clone_fast copies bi_partno").
> Thanks for clearing that up, I duly appreciate it.
>
> Marc