Early 4.1.1 release for image corruption fixes?

Early 4.1.1 release for image corruption fixes?

[Kevin Wolf]

Hi,

the QEMU 4.1.0 release has two problems that can easily cause image
corruption with any qcow2 images (no special configuration needed to
trigger the bugs):

1. A locking bug in the qcow2 code. I just sent a pull request that
   includes the fix for this. The important patch there is:

   'qcow2: Fix corruption bug in qcow2_detect_metadata_preallocation()'

2. A kernel bug in the XFS driver that became visible by new I/O
   patterns the qcow2 implementation started to use in 4.1. Max is
   currently working on a workaround for this.

The Planning/4.1 wiki page tells me that a 4.1.1 release is planned for
end of November, which isn't too far, but I was wondering if want to
have a stable release even earlier, right after the fixes for both
problems are in.

Peter, if we haven't made a decision until then, maybe discussing this
is an agenda item for QEMU Summit, too.

Kevin

Re: Early 4.1.1 release for image corruption fixes?

[Max Reitz]

[-- Attachment #1.1: Type: text/plain, Size: 1379 bytes --]

On 25.10.19 16:29, Kevin Wolf wrote:
> Hi,
> 
> the QEMU 4.1.0 release has two problems that can easily cause image
> corruption with any qcow2 images (no special configuration needed to
> trigger the bugs):

Three, actually.  There’s also b2c6f23f4a9f6d8f1b648705cd46d3713b78d6a2
which fixed 50ba5b2d994853b38fed10e0841b119da0f8b8e5 that was part of
4.1.0, too.

> 1. A locking bug in the qcow2 code. I just sent a pull request that
>    includes the fix for this. The important patch there is:
> 
>    'qcow2: Fix corruption bug in qcow2_detect_metadata_preallocation()'
> 
> 2. A kernel bug in the XFS driver that became visible by new I/O
>    patterns the qcow2 implementation started to use in 4.1. Max is
>    currently working on a workaround for this.

You most likely won’t see the kernel bug on 4.1.0, because 50ba5b2d99 is
basically the same bug, but in qemu...

(I don’t know if it has the same technical reason, but it causes the
same symptoms.)

Max

> The Planning/4.1 wiki page tells me that a 4.1.1 release is planned for
> end of November, which isn't too far, but I was wondering if want to
> have a stable release even earlier, right after the fixes for both
> problems are in.
> 
> Peter, if we haven't made a decision until then, maybe discussing this
> is an agenda item for QEMU Summit, too.
> 
> Kevin
> 



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 488 bytes --]

Re: Early 4.1.1 release for image corruption fixes?

[Daniel P. Berrangé]

On Fri, Oct 25, 2019 at 04:29:15PM +0200, Kevin Wolf wrote:
> Hi,
> 
> the QEMU 4.1.0 release has two problems that can easily cause image
> corruption with any qcow2 images (no special configuration needed to
> trigger the bugs):
> 
> 1. A locking bug in the qcow2 code. I just sent a pull request that
>    includes the fix for this. The important patch there is:
> 
>    'qcow2: Fix corruption bug in qcow2_detect_metadata_preallocation()'
> 
> 2. A kernel bug in the XFS driver that became visible by new I/O
>    patterns the qcow2 implementation started to use in 4.1. Max is
>    currently working on a workaround for this.
> 
> The Planning/4.1 wiki page tells me that a 4.1.1 release is planned for
> end of November, which isn't too far, but I was wondering if want to
> have a stable release even earlier, right after the fixes for both
> problems are in.

IMHO data corruption for our primary image format is an excellent
reason to do a stable release at the soonest viable opportunity.

Regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|