[OASIS Issue Tracker] (VIRTIO-144) VIRTIO_BLK_F_CONFIG_WCE removed from block devices spec

[OASIS Issues Tracker <workgroup_mailer@lists.oasis-open.org>]

    [ https://issues.oasis-open.org/browse/VIRTIO-144?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=60497#comment-60497 ] 

Paolo Bonzini commented on VIRTIO-144:
--------------------------------------

There are still two points being debated:

* whether the case of transient device backend storage should be mentioned in the spec. For example, "If the device backend storage is permanent, the device MUST ensure that stable writes are committed to it before reporting completion of the write (cases~\ref{item:flush1} and~\ref{item:flush2}) or the flush (case~\ref{item:flush3})."  I'm leaning towards this formulation.

* whether a device that does not offer the VIRTIO_BLK_F_FLUSH feature "MAY" (Paolo) or "SHOULD" (mst) care about data safety. Regarding this, my opinion is that the device already SHOULD have offered VIRTIO_BLK_F_FLUSH, but it didn't. So there's no reason to believe that the device will follow this other recommendation, and MAY is good enough.

> VIRTIO_BLK_F_CONFIG_WCE removed from block devices spec
> -------------------------------------------------------
>
>                 Key: VIRTIO-144
>                 URL: https://issues.oasis-open.org/browse/VIRTIO-144
>             Project: OASIS Virtual I/O Device (VIRTIO) TC
>          Issue Type: Bug
>    Affects Versions: virtio 1.0 cs02
>            Reporter: Paolo Bonzini
>            Priority: Critical
>              Labels: AdoptionBlocker
>
> VIRTIO_BLK_F_CONFIG_WCE is important in order to achieve good performance (up to 2x, though more realistically +30-40%) in latency-bound workloads. However, it was removed by mistake together with VIRTIO_BLK_F_FLUSH.
> The lack of this feature means that virtio 1.0 cs02 does not achieve parity with 0.9.



--
This message was sent by Atlassian JIRA
(v6.2.2#6258)