Hi Thiemo, Thank you for comments. You probably right about the "child" -> "is_parent" renaming, But the "parent_open" variable (I know, the name is not the best. you can rename it) is not a duplicate of the "is_parent".
The only one reason for existence of "parent_open" it's set/reset the
"is_parent".
I think we can't use the "is_parent" instead, because we need to know at every
moment if it's a "parent" or "leaf" image.
I mean, if we'll use "is_parent" instead of "parent_open":
+ s->is_parent = 1;
+ if (bdrv_open(s->hd->backing_hd, parent_img_name, BDRV_O_RDONLY) < 0)
goto failure;
+ s->is_parent = 0;
We'll reset it after the file opening. It's mean, later during the
"get_cluster_offset()" we'll not be able to know if it's a "parent" or
"leaf" image.
So, I attach a new patch with you comments.
Regards,
Igor Lvovsky
-----Original Message-----
From: Thiemo Seufer [mailto:[EMAIL PROTECTED]
Sent: Sunday, May 20, 2007 12:42 AM
To: Igor Lvovsky
Cc: qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH] Fix a race condition and non-leaf images
growing in VMDK chains.
Thiemo Seufer wrote:
> Igor Lvovsky wrote:
> >
> > Hi,
> > The bug was in my last patch.
> > This is a new one.
> > I'll very appreciate any comments.
>
> Appended is a variant which gets rid of the confusing "child" variable
> which means "is_parent", and the duplicate of it (parent_open).
>
> There was also an uninitialized variable "tmp" which was stored as
> offset in m_data.
>
> I quite possibly missed some bits which broke the patch, please test if
> it still works for you.
Next try, this time with the latest version of the patch.
Thiemo
block-vmdk-patch.diff
Description: block-vmdk-patch.diff
