[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH 0/2] Pointer properties and device_add
From: |
Marcel Apfelbaum |
Subject: |
Re: [Qemu-devel] [PATCH 0/2] Pointer properties and device_add |
Date: |
Mon, 02 Dec 2013 10:52:56 +0200 |
On Sun, 2013-12-01 at 16:14 +0100, Andreas Färber wrote:
> Am 01.12.2013 14:13, schrieb Marcel Apfelbaum:
> > On Fri, 2013-11-29 at 10:43 +0100, address@hidden wrote:
> >> From: Markus Armbruster <address@hidden>
> >>
> >> Pointer properties can be set only by code, not by device_add. A
> >> device with a pointer property can't work with device_add only unless
> >> the property may remain null. cannot_instantiate_with_device_add_yet
> >> needs to be set then. PATCH 1/2 sets it when needed and else
> >> documents why not. PATCH 2/2 documents this for future users of
> >> pointer properties.
> >>
> >> This applies on top of my "[PATCH v4 00/10] Clean up and fix no_user"
> >> series.
> >
> > Even that I am not familiar with this code, I've checked all the changes
> > and I agree with them.
> >
> > Reviewed-by: Marcel Apfelbaum <address@hidden>
> >
> > Anyway, I do have a question:
> > Why not asserting on qdev_device_add if we have a pointer property?
>
> When we do device_add / device-add, the guest is usually running and we
> shouldn't kill a running guest just because the user is trying something
> stupid that we can easily prevent. ;)
>
> The alternative BTW is dropping all those pointer properties and
> replacing them with link<> properties. Paolo tried that for the OMAP
> timers once but I fear that series was never picked up...?
I heard about these link<> properties, can someone point me to their
implementation?
>
> > Instead of checking only cannot_instantiate_with_device_add_yet,
> > we can go over properties and if we have a pointer property, assert or
> > return...
>
> Raising an error for certain property types may be an option. Although
> theoretically the existence of an incompatible property would not
> necessarily indicate incompatibility to instantiate the device, in
> practice I believe we don't have such excess properties.
If a pointer property does not necessary mean: "Can't be used with device_add",
I am not so sure anymore...
Thanks Andreas!
Marcel
>
> Regards,
> Andreas
>
> >
> > Thanks,
> > Marcel
> >
> >>
> >> Markus Armbruster (2):
> >> hw: cannot_instantiate_with_device_add_yet due to pointer props
> >> qdev: Document that pointer properties kill device_add
> >>
> >> hw/audio/marvell_88w8618.c | 2 ++
> >> hw/dma/sparc32_dma.c | 2 ++
> >> hw/gpio/omap_gpio.c | 4 ++++
> >> hw/i2c/omap_i2c.c | 2 ++
> >> hw/i2c/smbus_eeprom.c | 2 ++
> >> hw/intc/etraxfs_pic.c | 4 ++++
> >> hw/intc/grlib_irqmp.c | 2 ++
> >> hw/intc/omap_intc.c | 4 ++++
> >> hw/net/etraxfs_eth.c | 2 ++
> >> hw/net/lance.c | 2 ++
> >> include/hw/qdev-properties.h | 17 +++++++++++++++++
> >> 11 files changed, 43 insertions(+)
> >>
> >
> >
> >
>
>