[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH 5/5] e1000: using new interface--unmap to unplug
From: |
Stefan Hajnoczi |
Subject: |
Re: [Qemu-devel] [PATCH 5/5] e1000: using new interface--unmap to unplug |
Date: |
Wed, 25 Jul 2012 08:12:13 +0100 |
On Wed, Jul 25, 2012 at 4:31 AM, Liu Ping Fan <address@hidden> wrote:
> From: Liu Ping Fan <address@hidden>
>
> Signed-off-by: Liu Ping Fan <address@hidden>
> ---
> hw/e1000.c | 15 +++++++++++++--
> 1 files changed, 13 insertions(+), 2 deletions(-)
>
> diff --git a/hw/e1000.c b/hw/e1000.c
> index 4573f13..4c1e141 100644
> --- a/hw/e1000.c
> +++ b/hw/e1000.c
> @@ -1192,6 +1192,18 @@ e1000_cleanup(VLANClientState *nc)
> s->nic = NULL;
> }
>
> +static void
> +pci_e1000_unmap(DeviceState *dev)
> +{
> + PCIDevice *p = PCI_DEVICE(dev);
> + E1000State *d = DO_UPCAST(E1000State, dev, p);
> +
> + /* DO NOT FREE anything!until refcnt=0 */
> + /* isolate from memory view */
> + memory_region_destroy(&d->mmio);
> + memory_region_destroy(&d->io);
> +}
It's not obvious to me why a 2-stage cleanup is needed (->unmap(),
->exit()). Explaining things a bit more in the commit description
would help. Here's what I'm thinking:
We want to remove the memory regions at the same time as removing the
device from the tree, but ->exit() is only called when the object is
finalized. Because of the object reference held during dispatch, the
reference might not reach 0 during hotplug and another thread could
still be running this device's code?
This series only applies this change to e1000 and piix pci hotplug.
How/when will all the other devices be converted? Will it be safe to
leave them unconverted once dispatch really happens in parallel?
Stefan
- [Qemu-devel] [PATCH 0/5] prepare unplug out of protection of global lock, Liu Ping Fan, 2012/07/24
- [Qemu-devel] [PATCH 1/5] qom: adopt rwlock to protect accessing dev from removing it, Liu Ping Fan, 2012/07/24
- [Qemu-devel] [PATCH 2/5] exec.c: use refcnt to protect device during dispatching, Liu Ping Fan, 2012/07/24
- [Qemu-devel] [PATCH 3/5] hotplug: introduce qdev_unplug_ack() to remove device from views, Liu Ping Fan, 2012/07/24
- [Qemu-devel] [PATCH 4/5] qom: delay DeviceState's reclaim to main-loop, Liu Ping Fan, 2012/07/24
- [Qemu-devel] [PATCH 5/5] e1000: using new interface--unmap to unplug, Liu Ping Fan, 2012/07/24
- Re: [Qemu-devel] [PATCH 5/5] e1000: using new interface--unmap to unplug,
Stefan Hajnoczi <=