[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH v2] qom: Reject attempts to add a property that
From: |
Anthony Liguori |
Subject: |
Re: [Qemu-devel] [PATCH v2] qom: Reject attempts to add a property that already exists |
Date: |
Mon, 08 Oct 2012 12:06:16 -0500 |
User-agent: |
Notmuch/0.13.2+93~ged93d79 (http://notmuchmail.org) Emacs/23.3.1 (x86_64-pc-linux-gnu) |
Peter Maydell <address@hidden> writes:
> On 8 October 2012 14:29, Anthony Liguori <address@hidden> wrote:
>> This is wrong.
>>
>> Container properties are added by the user. You will turn a gracefully
>> failure (during hotplug) into an abort().
>
> No, it's turning a bug into an abort -- we don't handle trying to
> create two identically named properties correctly today.
Killing a guest because of something a user mistypes is not very friendly.
>
>> Please limit this to static properties as they are not added by a user.
>
> Adding two dynamic properties of the same name is also not
> going to work and we need to do something with it...
Raise an error.
> What is the code path for properties being added by a user?
qdev_device_add().
> If it's qdev_device_add() then that code presumably doesn't
> care about graceful failures because it passes NULL as an
> error pointer.
Then we should handle the error there gracefully.
> container_get() seems to assume that adding the
> child property will always succeed and will not do the right
> thing if there already exists a child property of the relevant
> name but wrong type.
>
> Basically it seems to me that any code which might actually
> be hit by this assert() rather needs examination and rewriting
> to handle the error case anyway...
There are only two cases that actually matter today:
1) static properties
2) qdev_device_add().
Yes, (2) is not doign error checking today. It should. I would be very
happy with an abort() in (1) since that's clearly a programming bug.
Regards,
Anthony Liguori
>
> -- PMM