[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Qemu-devel] [PATCH v2 1/6] block/qdev: Allow node name for drive pr
From: |
Kevin Wolf |
Subject: |
Re: [Qemu-devel] [PATCH v2 1/6] block/qdev: Allow node name for drive properties |
Date: |
Mon, 4 Jul 2016 12:43:54 +0200 |
User-agent: |
Mutt/1.5.21 (2010-09-15) |
Am 02.07.2016 um 17:33 hat Max Reitz geschrieben:
> On 30.06.2016 16:13, Kevin Wolf wrote:
> > If a node name instead of a BlockBackend name is specified as the driver
> > for a guest device, an anonymous BlockBackend is created now.
> >
> > Signed-off-by: Kevin Wolf <address@hidden>
> > ---
> > hw/core/qdev-properties-system.c | 22 +++++++++++++++++++---
> > 1 file changed, 19 insertions(+), 3 deletions(-)
>
> I just noticed this won't work with (at least) USB mass storage devices.
> scsi_bus_legacy_add_drive() creates a new device and uses
> qdev_prop_set_drive() to set its drive, and that function uses
> blk_name() to get the required value for .drive.
>
> That can of course be fixed by using bdrv_get_node_name(blk_bs(value))
> there if blk_name(value) is empty. But in addition we need to make sure
> that the BB is not deleted in usb_msd_realize_storage() (because that
> function has to detach the device from the BB before creating the SCSI
> disk), so we need to wrap the explicitly hacky block there in
> blk_ref(blk) and blk_unref(blk).
I'm not sure if I even want to fix this here... Maybe it's better to
leave node names broken for usb-storage, as a motivation for someone to
try and clean up this mess for good. I mean, it's not the first time
that this hack breaks.
Kevin
pgpJmEREK2ttS.pgp
Description: PGP signature