Am 07.05.2020 um 11:33 hat Max Reitz geschrieben:
> On 06.05.20 19:04, Kevin Wolf wrote:
> > Am 18.02.2020 um 13:42 hat Max Reitz geschrieben:
> >> Replace child_file by child_of_bds in all remaining places (excluding
> >> tests).
> >>
> >> Signed-off-by: Max Reitz <mreitz@redhat.com>
> >> Reviewed-by: Eric Blake <eblake@redhat.com>
> > 
> >> diff --git a/block/blkreplay.c b/block/blkreplay.c
> >> index f97493f45a..71628f4d56 100644
> >> --- a/block/blkreplay.c
> >> +++ b/block/blkreplay.c
> >> @@ -27,8 +27,9 @@ static int blkreplay_open(BlockDriverState *bs, QDict *options, int flags,
> >>      int ret;
> >>  
> >>      /* Open the image file */
> >> -    bs->file = bdrv_open_child(NULL, options, "image",
> >> -                               bs, &child_file, 0, false, &local_err);
> >> +    bs->file = bdrv_open_child(NULL, options, "image", bs, &child_of_bds,
> >> +                               BDRV_CHILD_DATA | BDRV_CHILD_PRIMARY,
> >> +                               false, &local_err);
> > 
> > Why isn't blkreplay a filter?
> I don’t know, but that’s pre-existing.  (It doesn’t set .is_filter.)
> 
> I’m afraid I have little insight into what blkreplay actually does.  I
> thought maybe in conjunction with the whole replay stuff it might turn
> out not to be a filter.
> 
> So is it a filter?  If so, no problem.  I suppose we can fix it up in
> the same patch that marks mirror-top and commit-top as filters.

I think it is one. If I understand correctly, it basically just changes
the timing of requests, but nothing about their data.

Adding .is_filter to it in the same patch with mirror-top/commit-top
makes sense to me.

Kevin