Re: [Qemu-devel] [PATCH v3 4/9] ide: Drop redundant IDEState member bs

[Kevin Wolf]

Am 30.01.2014 um 13:16 hat Markus Armbruster geschrieben:
> It's a copy of dev->conf.bs.  The copy was needed for non-qdevified
> controllers, which lacked dev.
> 
> Note how pci_piix3_xen_ide_unplug() cleared the copy.  We'll get back
> to that in the next few commits.
> 
> Signed-off-by: Markus Armbruster <address@hidden>

So this pci_piix3_xen_ide_unplug() is what happens here:

> --- a/hw/ide/piix.c
> +++ b/hw/ide/piix.c
> @@ -169,12 +169,9 @@ static int pci_piix_ide_initfn(PCIDevice *dev)
>  
>  static int pci_piix3_xen_ide_unplug(DeviceState *dev)
>  {
> -    PCIIDEState *pci_ide;
>      DriveInfo *di;
>      int i = 0;
>  
> -    pci_ide = PCI_IDE(dev);
> -
>      for (; i < 3; i++) {
>          di = drive_get_by_index(IF_IDE, i);
>          if (di != NULL && !di->media_cd) {
> @@ -183,7 +180,6 @@ static int pci_piix3_xen_ide_unplug(DeviceState *dev)
>                  bdrv_detach_dev(di->bdrv, ds);
>              }
>              bdrv_close(di->bdrv);
> -            pci_ide->bus[di->bus].ifs[di->unit].bs = NULL;
>              drive_put_ref(di);
>          }
>      }

Probably I'm just missing the obvious, but it seems to me that the
copy was cleared here, while the original was left around. This was
no problem because the original was unused anyway after device
initialisation.

Now that the copy doesn't exist any more, we can't clear it, obviously,
but why don't we have to clear the original? Won't we still run the
"device is attached" code branches even though the device is really
unplugged?

Kevin