[Qemu-devel] [PATCH 00/12] More block-related fixes and cleanups

[Qemu-devel] [PATCH 00/12] More block-related fixes and cleanups

[Markus Armbruster]

I'm working on cleanly separating block device host and guest parts.
I'd like to route all this work through Kevin's block tree.  This is
still just preliminaries.

There will be at least one more round of cleanup & fixes before
blockdev_add proper.  I intend to start with a minimal QMP-only
version, then add features.

Markus Armbruster (12):
  scsi: scsi_bus_legacy_handle_cmdline() can fail, fix callers
  ide: Make it explicit that ide_create_drive() can't fail
  blockdev: Remove drive_get_serial()
  blockdev: New drive_of_blockdev()
  blockdev: Clean up automatic drive deletion
  qdev: Decouple qdev_prop_drive from DriveInfo
  blockdev: drive_get_by_id() is no longer used, remove
  block: Catch attempt to attach multiple devices to a blockdev
  savevm: Survive hot-unplug of snapshot device
  block: Fix virtual media change for if=none
  ide: Make PIIX and ISA IDE init functions return the qdev
  pc: Fix CMOS info for drives defined with -device

 block.c              |   55 +++++++++++++++++++++++++++++
 block.h              |    5 +++
 block_int.h          |    8 ++--
 blockdev.c           |   45 +++++++++++++++---------
 blockdev.h           |    7 +++-
 hw/esp.c             |    3 +-
 hw/fdc.c             |   32 +++++++++-------
 hw/ide.h             |   13 ++++---
 hw/ide/core.c        |   18 +++++----
 hw/ide/internal.h    |    2 +-
 hw/ide/isa.c         |    8 ++--
 hw/ide/piix.c        |    6 ++-
 hw/ide/qdev.c        |   22 ++++++++---
 hw/lsi53c895a.c      |    2 +-
 hw/pc.c              |   94 +++++++++++++++++++++++++++++++------------------
 hw/pc.h              |    3 +-
 hw/pc_piix.c         |   16 ++++++---
 hw/pci-hotplug.c     |   10 ++++-
 hw/qdev-properties.c |   46 +++++++++++++++++++++----
 hw/qdev.h            |    7 ++--
 hw/s390-virtio.c     |    2 +-
 hw/scsi-bus.c        |   19 ++++++----
 hw/scsi-disk.c       |   21 ++++++-----
 hw/scsi-generic.c    |    7 ++--
 hw/scsi.h            |    4 +-
 hw/usb-msd.c         |   29 ++++++++++++---
 hw/virtio-blk.c      |    3 +-
 hw/virtio-pci.c      |    4 +-
 savevm.c             |   31 ++--------------
 29 files changed, 344 insertions(+), 178 deletions(-)

[Qemu-devel] [PATCH 01/12] scsi: scsi_bus_legacy_handle_cmdline() can fail, fix callers

[Markus Armbruster]

None of its callers checks for failure.  scsi_hot_add() can crash
because of that:

(qemu) drive_add 4 if=scsi,format=host_device,file=/dev/sg1
scsi-generic: scsi generic interface too old
Segmentation fault (core dumped)

Fix all callers, not just scsi_hot_add().

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 hw/esp.c         |    3 +--
 hw/lsi53c895a.c  |    2 +-
 hw/pci-hotplug.c |    3 +++
 hw/scsi-bus.c    |   11 +++++++----
 hw/scsi.h        |    2 +-
 hw/usb-msd.c     |    3 +++
 6 files changed, 16 insertions(+), 8 deletions(-)

diff --git a/hw/esp.c b/hw/esp.c
index 7740879..349052a 100644
--- a/hw/esp.c
+++ b/hw/esp.c
@@ -679,8 +679,7 @@ static int esp_init1(SysBusDevice *dev)
     qdev_init_gpio_in(&dev->qdev, parent_esp_reset, 1);
 
     scsi_bus_new(&s->bus, &dev->qdev, 0, ESP_MAX_DEVS, esp_command_complete);
-    scsi_bus_legacy_handle_cmdline(&s->bus);
-    return 0;
+    return scsi_bus_legacy_handle_cmdline(&s->bus);
 }
 
 static SysBusDeviceInfo esp_info = {
diff --git a/hw/lsi53c895a.c b/hw/lsi53c895a.c
index f5a91ba..c2a8010 100644
--- a/hw/lsi53c895a.c
+++ b/hw/lsi53c895a.c
@@ -2176,7 +2176,7 @@ static int lsi_scsi_init(PCIDevice *dev)
 
     scsi_bus_new(&s->bus, &dev->qdev, 1, LSI_MAX_DEVS, lsi_command_complete);
     if (!dev->qdev.hotplugged) {
-        scsi_bus_legacy_handle_cmdline(&s->bus);
+        return scsi_bus_legacy_handle_cmdline(&s->bus);
     }
     return 0;
 }
diff --git a/hw/pci-hotplug.c b/hw/pci-hotplug.c
index c39e640..55c9fe3 100644
--- a/hw/pci-hotplug.c
+++ b/hw/pci-hotplug.c
@@ -90,6 +90,9 @@ static int scsi_hot_add(Monitor *mon, DeviceState *adapter,
      */
     dinfo->unit = qemu_opt_get_number(dinfo->opts, "unit", -1);
     scsidev = scsi_bus_legacy_add_drive(scsibus, dinfo, dinfo->unit);
+    if (!scsidev) {
+        return -1;
+    }
     dinfo->unit = scsidev->id;
 
     if (printinfo)
diff --git a/hw/scsi-bus.c b/hw/scsi-bus.c
index 24bd060..d5b66c1 100644
--- a/hw/scsi-bus.c
+++ b/hw/scsi-bus.c
@@ -83,7 +83,6 @@ void scsi_qdev_register(SCSIDeviceInfo *info)
 }
 
 /* handle legacy '-drive if=scsi,...' cmd line args */
-/* FIXME callers should check for failure, but don't */
 SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, DriveInfo *dinfo, int unit)
 {
     const char *driver;
@@ -98,18 +97,22 @@ SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, DriveInfo *dinfo, int unit)
     return DO_UPCAST(SCSIDevice, qdev, dev);
 }
 
-void scsi_bus_legacy_handle_cmdline(SCSIBus *bus)
+int scsi_bus_legacy_handle_cmdline(SCSIBus *bus)
 {
     DriveInfo *dinfo;
-    int unit;
+    int res = 0, unit;
 
     for (unit = 0; unit < MAX_SCSI_DEVS; unit++) {
         dinfo = drive_get(IF_SCSI, bus->busnr, unit);
         if (dinfo == NULL) {
             continue;
         }
-        scsi_bus_legacy_add_drive(bus, dinfo, unit);
+        if (!scsi_bus_legacy_add_drive(bus, dinfo, unit)) {
+            res = -1;
+            break;
+        }
     }
+    return res;
 }
 
 void scsi_dev_clear_sense(SCSIDevice *dev)
diff --git a/hw/scsi.h b/hw/scsi.h
index b668e27..b1b5f73 100644
--- a/hw/scsi.h
+++ b/hw/scsi.h
@@ -98,7 +98,7 @@ static inline SCSIBus *scsi_bus_from_device(SCSIDevice *d)
 }
 
 SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, DriveInfo *dinfo, int unit);
-void scsi_bus_legacy_handle_cmdline(SCSIBus *bus);
+int scsi_bus_legacy_handle_cmdline(SCSIBus *bus);
 
 void scsi_dev_clear_sense(SCSIDevice *dev);
 void scsi_dev_set_sense(SCSIDevice *dev, uint8_t key);
diff --git a/hw/usb-msd.c b/hw/usb-msd.c
index 003bd8a..8e9718c 100644
--- a/hw/usb-msd.c
+++ b/hw/usb-msd.c
@@ -531,6 +531,9 @@ static int usb_msd_initfn(USBDevice *dev)
     s->dev.speed = USB_SPEED_FULL;
     scsi_bus_new(&s->bus, &s->dev.qdev, 0, 1, usb_msd_command_complete);
     s->scsi_dev = scsi_bus_legacy_add_drive(&s->bus, s->conf.dinfo, 0);
+    if (!s->scsi_dev) {
+        return -1;
+    }
     s->bus.qbus.allow_hotplug = 0;
     usb_msd_handle_reset(dev);
 
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 01/12] scsi: scsi_bus_legacy_handle_cmdline() can fail, fix callers

[Christoph Hellwig]

On Fri, Jun 25, 2010 at 06:53:21PM +0200, Markus Armbruster wrote:
> None of its callers checks for failure.  scsi_hot_add() can crash
> because of that:
> 
> (qemu) drive_add 4 if=scsi,format=host_device,file=/dev/sg1
> scsi-generic: scsi generic interface too old
> Segmentation fault (core dumped)
> 
> Fix all callers, not just scsi_hot_add().

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

[Qemu-devel] [PATCH 02/12] ide: Make it explicit that ide_create_drive() can't fail

[Markus Armbruster]

All callers of ide_create_drive() ignore its value.  Currently
harmless, because it fails only when qdev_init() fails, which fails
only when ide_drive_initfn() fails, which never fails.

Brittle.  Change it to die instead of silently ignoring failure.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 hw/ide/qdev.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/hw/ide/qdev.c b/hw/ide/qdev.c
index 0f9f22e..127478b 100644
--- a/hw/ide/qdev.c
+++ b/hw/ide/qdev.c
@@ -84,8 +84,7 @@ IDEDevice *ide_create_drive(IDEBus *bus, int unit, DriveInfo *drive)
     dev = qdev_create(&bus->qbus, "ide-drive");
     qdev_prop_set_uint32(dev, "unit", unit);
     qdev_prop_set_drive(dev, "drive", drive);
-    if (qdev_init(dev) < 0)
-        return NULL;
+    qdev_init_nofail(dev);
     return DO_UPCAST(IDEDevice, qdev, dev);
 }
 
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 02/12] ide: Make it explicit that ide_create_drive() can't fail

[Christoph Hellwig]

On Fri, Jun 25, 2010 at 06:53:22PM +0200, Markus Armbruster wrote:
> All callers of ide_create_drive() ignore its value.  Currently
> harmless, because it fails only when qdev_init() fails, which fails
> only when ide_drive_initfn() fails, which never fails.
> 
> Brittle.  Change it to die instead of silently ignoring failure.
> 
> Signed-off-by: Markus Armbruster <armbru@redhat.com>

Looks good,


Reviewed-by: Christoph Hellwig <hch@lst.de>

[Qemu-devel] [PATCH 03/12] blockdev: Remove drive_get_serial()

[Markus Armbruster]

Unused since commit 6ced55a5.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 blockdev.c |   12 ------------
 blockdev.h |    1 -
 2 files changed, 0 insertions(+), 13 deletions(-)

diff --git a/blockdev.c b/blockdev.c
index 3b8c606..e0495e5 100644
--- a/blockdev.c
+++ b/blockdev.c
@@ -78,18 +78,6 @@ int drive_get_max_bus(BlockInterfaceType type)
     return max_bus;
 }
 
-const char *drive_get_serial(BlockDriverState *bdrv)
-{
-    DriveInfo *dinfo;
-
-    QTAILQ_FOREACH(dinfo, &drives, next) {
-        if (dinfo->bdrv == bdrv)
-            return dinfo->serial;
-    }
-
-    return "\0";
-}
-
 static void bdrv_format_print(void *opaque, const char *name)
 {
     fprintf(stderr, " %s", name);
diff --git a/blockdev.h b/blockdev.h
index 23ea576..a936785 100644
--- a/blockdev.h
+++ b/blockdev.h
@@ -40,7 +40,6 @@ extern DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit);
 extern DriveInfo *drive_get_by_id(const char *id);
 extern int drive_get_max_bus(BlockInterfaceType type);
 extern void drive_uninit(DriveInfo *dinfo);
-extern const char *drive_get_serial(BlockDriverState *bdrv);
 
 extern QemuOpts *drive_add(const char *file, const char *fmt, ...);
 extern DriveInfo *drive_init(QemuOpts *arg, int default_to_scsi,
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 03/12] blockdev: Remove drive_get_serial()

[Christoph Hellwig]

On Fri, Jun 25, 2010 at 06:53:23PM +0200, Markus Armbruster wrote:
> Unused since commit 6ced55a5.
> 
> Signed-off-by: Markus Armbruster <armbru@redhat.com>

Looks good,


Reviewed-by: Christoph Hellwig <hch@lst.de>

[Qemu-devel] [PATCH 04/12] blockdev: New drive_of_blockdev()

[Markus Armbruster]

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 blockdev.c |   12 ++++++++++++
 blockdev.h |    1 +
 2 files changed, 13 insertions(+), 0 deletions(-)

diff --git a/blockdev.c b/blockdev.c
index e0495e5..8023cfd 100644
--- a/blockdev.c
+++ b/blockdev.c
@@ -78,6 +78,18 @@ int drive_get_max_bus(BlockInterfaceType type)
     return max_bus;
 }
 
+DriveInfo *drive_of_blockdev(BlockDriverState *bs)
+{
+    DriveInfo *dinfo;
+
+    QTAILQ_FOREACH(dinfo, &drives, next) {
+        if (dinfo->bdrv == bs) {
+            return dinfo;
+        }
+    }
+    return NULL;
+}
+
 static void bdrv_format_print(void *opaque, const char *name)
 {
     fprintf(stderr, " %s", name);
diff --git a/blockdev.h b/blockdev.h
index a936785..4bf75b1 100644
--- a/blockdev.h
+++ b/blockdev.h
@@ -40,6 +40,7 @@ extern DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit);
 extern DriveInfo *drive_get_by_id(const char *id);
 extern int drive_get_max_bus(BlockInterfaceType type);
 extern void drive_uninit(DriveInfo *dinfo);
+extern DriveInfo *drive_of_blockdev(BlockDriverState *bs);
 
 extern QemuOpts *drive_add(const char *file, const char *fmt, ...);
 extern DriveInfo *drive_init(QemuOpts *arg, int default_to_scsi,
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 04/12] blockdev: New drive_of_blockdev()

[Christoph Hellwig]

> +DriveInfo *drive_of_blockdev(BlockDriverState *bs)

I'd call this find_drive_by_blockdev.

[Qemu-devel] Re: [PATCH 04/12] blockdev: New drive_of_blockdev()

[Markus Armbruster]

Christoph Hellwig <hch@lst.de> writes:

>> +DriveInfo *drive_of_blockdev(BlockDriverState *bs)
>
> I'd call this find_drive_by_blockdev.

For what it's worth, all externally visible functions dealing with
drives start with "drive_".

[Qemu-devel] Re: [PATCH 04/12] blockdev: New drive_of_blockdev()

[Christoph Hellwig]

On Sat, Jun 26, 2010 at 07:32:09AM +0200, Markus Armbruster wrote:
> Christoph Hellwig <hch@lst.de> writes:
> 
> >> +DriveInfo *drive_of_blockdev(BlockDriverState *bs)
> >
> > I'd call this find_drive_by_blockdev.
> 
> For what it's worth, all externally visible functions dealing with
> drives start with "drive_".

Ok.  How about drive_get_by_blockdev, matching the drive_get_by_id we
have around?

Re: [Qemu-devel] Re: [PATCH 04/12] blockdev: New drive_of_blockdev()

[Markus Armbruster]

Christoph Hellwig <hch@lst.de> writes:

> On Sat, Jun 26, 2010 at 07:32:09AM +0200, Markus Armbruster wrote:
>> Christoph Hellwig <hch@lst.de> writes:
>> 
>> >> +DriveInfo *drive_of_blockdev(BlockDriverState *bs)
>> >
>> > I'd call this find_drive_by_blockdev.
>> 
>> For what it's worth, all externally visible functions dealing with
>> drives start with "drive_".
>
> Ok.  How about drive_get_by_blockdev, matching the drive_get_by_id we
> have around?

drive_get_by_id() gets removed by 07/12.  But there's still drive_get().

drive_get_by_blockdev()?  blockdev_to_drive()?

Re: [Qemu-devel] Re: [PATCH 04/12] blockdev: New drive_of_blockdev()

[Christoph Hellwig]

On Sat, Jun 26, 2010 at 04:46:28PM +0200, Markus Armbruster wrote:
> drive_get_by_id() gets removed by 07/12.  But there's still drive_get().
> 
> drive_get_by_blockdev()?  blockdev_to_drive()?

Oh well, just keep any name - it's not that important aftet all.

[Qemu-devel] Re: [PATCH 04/12] blockdev: New drive_of_blockdev()

[Paolo Bonzini]

On 06/26/2010 04:46 PM, Markus Armbruster wrote:
> drive_get_by_blockdev()?  blockdev_to_drive()?

I like drive_find_by_blockdev or drive_get_by_blockdev, in any case I'll 
make sure my own cpu_get_by_id is named in the same style that you choose.

Paolo

[Qemu-devel] Re: [PATCH 04/12] blockdev: New drive_of_blockdev()

[Markus Armbruster]

Paolo Bonzini <pbonzini@redhat.com> writes:

> On 06/26/2010 04:46 PM, Markus Armbruster wrote:
>> drive_get_by_blockdev()?  blockdev_to_drive()?
>
> I like drive_find_by_blockdev or drive_get_by_blockdev, in any case
> I'll make sure my own cpu_get_by_id is named in the same style that
> you choose.

drive_get_by_blockdev() it is.

As long as this is the only change, I don't think another post to the
list is necessary.  Instead, I'll push to my tree and let Kevin pick it
up from there.

[Qemu-devel] [PATCH 05/12] blockdev: Clean up automatic drive deletion

[Markus Armbruster]

We automatically delete blockdev host parts on unplug of the guest
device.  Too much magic, but we can't change that now.

The delete happens early in the guest device teardown, before the
connection to the host part is severed.  Thus, the guest part's
pointer to the host part dangles for a brief time.  No actual harm
comes from this, but we'll catch such dangling pointers a few commits
down the road.  Clean up the dangling pointers by delaying the
automatic deletion until the guest part's pointer is gone.

Device usb-storage deliberately makes two qdev properties refer to the
same drive, because it automatically creates a second device.  Again,
too much magic we can't change now.  Multiple references worked okay
before, but now free_drive() dies for the second one.  Zap the extra
reference.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 blockdev.c           |   23 +++++++++++++++++++++++
 blockdev.h           |    4 ++++
 hw/qdev-properties.c |   10 ++++++++++
 hw/scsi-disk.c       |    2 +-
 hw/scsi-generic.c    |    2 +-
 hw/usb-msd.c         |   20 ++++++++++++++++----
 hw/virtio-pci.c      |    2 +-
 7 files changed, 56 insertions(+), 7 deletions(-)

diff --git a/blockdev.c b/blockdev.c
index 8023cfd..827ea1c 100644
--- a/blockdev.c
+++ b/blockdev.c
@@ -17,6 +17,29 @@
 
 static QTAILQ_HEAD(drivelist, DriveInfo) drives = QTAILQ_HEAD_INITIALIZER(drives);
 
+/*
+ * We automatically delete the drive when a device using it gets
+ * unplugged.  Questionable feature, but we can't just drop it.
+ * Device models call blockdev_mark_auto_del() to schedule the
+ * automatic deletion, and generic qdev code calls blockdev_auto_del()
+ * when deletion is actually safe.
+ */
+void blockdev_mark_auto_del(BlockDriverState *bs)
+{
+    DriveInfo *dinfo = drive_of_blockdev(bs);
+
+    dinfo->auto_del = 1;
+}
+
+void blockdev_auto_del(BlockDriverState *bs)
+{
+    DriveInfo *dinfo = drive_of_blockdev(bs);
+
+    if (dinfo->auto_del) {
+        drive_uninit(dinfo);
+    }
+}
+
 QemuOpts *drive_add(const char *file, const char *fmt, ...)
 {
     va_list ap;
diff --git a/blockdev.h b/blockdev.h
index 4bf75b1..418ebb6 100644
--- a/blockdev.h
+++ b/blockdev.h
@@ -13,6 +13,9 @@
 #include "block.h"
 #include "qemu-queue.h"
 
+void blockdev_mark_auto_del(BlockDriverState *bs);
+void blockdev_auto_del(BlockDriverState *bs);
+
 typedef enum {
     IF_NONE,
     IF_IDE, IF_SCSI, IF_FLOPPY, IF_PFLASH, IF_MTD, IF_SD, IF_VIRTIO, IF_XEN,
@@ -28,6 +31,7 @@ typedef struct DriveInfo {
     BlockInterfaceType type;
     int bus;
     int unit;
+    int auto_del;               /* see blockdev_mark_auto_del() */
     QemuOpts *opts;
     char serial[BLOCK_SERIAL_STRLEN + 1];
     QTAILQ_ENTRY(DriveInfo) next;
diff --git a/hw/qdev-properties.c b/hw/qdev-properties.c
index 5a8739d..15ca6d3 100644
--- a/hw/qdev-properties.c
+++ b/hw/qdev-properties.c
@@ -293,6 +293,15 @@ static int parse_drive(DeviceState *dev, Property *prop, const char *str)
     return 0;
 }
 
+static void free_drive(DeviceState *dev, Property *prop)
+{
+    DriveInfo **ptr = qdev_get_prop_ptr(dev, prop);
+
+    if (*ptr) {
+        blockdev_auto_del((*ptr)->bdrv);
+    }
+}
+
 static int print_drive(DeviceState *dev, Property *prop, char *dest, size_t len)
 {
     DriveInfo **ptr = qdev_get_prop_ptr(dev, prop);
@@ -305,6 +314,7 @@ PropertyInfo qdev_prop_drive = {
     .size  = sizeof(DriveInfo*),
     .parse = parse_drive,
     .print = print_drive,
+    .free  = free_drive,
 };
 
 /* --- character device --- */
diff --git a/hw/scsi-disk.c b/hw/scsi-disk.c
index 2b38984..d76e640 100644
--- a/hw/scsi-disk.c
+++ b/hw/scsi-disk.c
@@ -1043,7 +1043,7 @@ static void scsi_destroy(SCSIDevice *dev)
     SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, dev);
 
     scsi_disk_purge_requests(s);
-    drive_uninit(s->qdev.conf.dinfo);
+    blockdev_mark_auto_del(s->qdev.conf.dinfo->bdrv);
 }
 
 static int scsi_disk_initfn(SCSIDevice *dev)
diff --git a/hw/scsi-generic.c b/hw/scsi-generic.c
index e31060e..1859c94 100644
--- a/hw/scsi-generic.c
+++ b/hw/scsi-generic.c
@@ -453,7 +453,7 @@ static void scsi_destroy(SCSIDevice *d)
         r = DO_UPCAST(SCSIGenericReq, req, QTAILQ_FIRST(&s->qdev.requests));
         scsi_remove_request(r);
     }
-    drive_uninit(s->qdev.conf.dinfo);
+    blockdev_mark_auto_del(s->qdev.conf.dinfo->bdrv);
 }
 
 static int scsi_generic_initfn(SCSIDevice *dev)
diff --git a/hw/usb-msd.c b/hw/usb-msd.c
index 8e9718c..3dbfcab 100644
--- a/hw/usb-msd.c
+++ b/hw/usb-msd.c
@@ -522,24 +522,36 @@ static void usb_msd_password_cb(void *opaque, int err)
 static int usb_msd_initfn(USBDevice *dev)
 {
     MSDState *s = DO_UPCAST(MSDState, dev, dev);
+    DriveInfo *dinfo = s->conf.dinfo;
 
-    if (!s->conf.dinfo || !s->conf.dinfo->bdrv) {
+    if (!dinfo || !dinfo->bdrv) {
         error_report("usb-msd: drive property not set");
         return -1;
     }
 
+    /*
+     * Hack alert: this pretends to be a block device, but it's really
+     * a SCSI bus that can serve only a single device, which it
+     * creates automatically.  Two drive properties pointing to the
+     * same drive is not good: free_drive() dies for the second one.
+     * Zap the one we're not going to use.
+     *
+     * The hack is probably a bad idea.
+     */
+    s->conf.dinfo = NULL;
+
     s->dev.speed = USB_SPEED_FULL;
     scsi_bus_new(&s->bus, &s->dev.qdev, 0, 1, usb_msd_command_complete);
-    s->scsi_dev = scsi_bus_legacy_add_drive(&s->bus, s->conf.dinfo, 0);
+    s->scsi_dev = scsi_bus_legacy_add_drive(&s->bus, dinfo, 0);
     if (!s->scsi_dev) {
         return -1;
     }
     s->bus.qbus.allow_hotplug = 0;
     usb_msd_handle_reset(dev);
 
-    if (bdrv_key_required(s->conf.dinfo->bdrv)) {
+    if (bdrv_key_required(dinfo->bdrv)) {
         if (cur_mon) {
-            monitor_read_bdrv_key_start(cur_mon, s->conf.dinfo->bdrv,
+            monitor_read_bdrv_key_start(cur_mon, dinfo->bdrv,
                                         usb_msd_password_cb, s);
             s->dev.auto_attach = 0;
         } else {
diff --git a/hw/virtio-pci.c b/hw/virtio-pci.c
index d1303b1..31a68fe 100644
--- a/hw/virtio-pci.c
+++ b/hw/virtio-pci.c
@@ -571,7 +571,7 @@ static int virtio_blk_exit_pci(PCIDevice *pci_dev)
 {
     VirtIOPCIProxy *proxy = DO_UPCAST(VirtIOPCIProxy, pci_dev, pci_dev);
 
-    drive_uninit(proxy->block.dinfo);
+    blockdev_mark_auto_del(proxy->block.dinfo->bdrv);
     return virtio_exit_pci(pci_dev);
 }
 
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 05/12] blockdev: Clean up automatic drive deletion

[Christoph Hellwig]

Looks correct to me,


Reviewed-by: Christoph Hellwig <hch@lst.de>

[Qemu-devel] [PATCH 06/12] qdev: Decouple qdev_prop_drive from DriveInfo

[Markus Armbruster]

Make the property point to BlockDriverState, cutting out the DriveInfo
middleman.  This prepares the ground for block devices that don't have
a DriveInfo.

Currently all user-defined ones have a DriveInfo, because the only way
to define one is -drive & friends (they go through drive_init()).
DriveInfo is closely tied to -drive, and like -drive, it mixes
information about host and guest part of the block device.  I'm
working towards a new way to define block devices, with clean
host/guest separation, and I need to get DriveInfo out of the way for
that.

Fortunately, the device models are perfectly happy with
BlockDriverState, except for two places: ide_drive_initfn() and
scsi_disk_initfn() need to check the DriveInfo for a serial number set
with legacy -drive serial=...  Use drive_of_blockdev() there.

Device model code should now use DriveInfo only when explicitly
dealing with drives defined the old way, i.e. without -device.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 block_int.h          |    6 ++----
 hw/fdc.c             |   22 ++++++++++------------
 hw/ide/core.c        |   17 +++++++++--------
 hw/ide/internal.h    |    2 +-
 hw/ide/qdev.c        |   12 ++++++++----
 hw/pci-hotplug.c     |    4 ++--
 hw/qdev-properties.c |   21 ++++++++++++---------
 hw/qdev.h            |    6 +++---
 hw/s390-virtio.c     |    2 +-
 hw/scsi-bus.c        |    8 ++++----
 hw/scsi-disk.c       |   16 +++++++---------
 hw/scsi-generic.c    |    6 +++---
 hw/scsi.h            |    2 +-
 hw/usb-msd.c         |   15 +++++++--------
 hw/virtio-blk.c      |    2 +-
 hw/virtio-pci.c      |    4 ++--
 16 files changed, 73 insertions(+), 72 deletions(-)

diff --git a/block_int.h b/block_int.h
index b64a009..e60aed4 100644
--- a/block_int.h
+++ b/block_int.h
@@ -210,10 +210,8 @@ void *qemu_blockalign(BlockDriverState *bs, size_t size);
 int is_windows_drive(const char *filename);
 #endif
 
-struct DriveInfo;
-
 typedef struct BlockConf {
-    struct DriveInfo *dinfo;
+    BlockDriverState *bs;
     uint16_t physical_block_size;
     uint16_t logical_block_size;
     uint16_t min_io_size;
@@ -234,7 +232,7 @@ static inline unsigned int get_physical_block_exp(BlockConf *conf)
 }
 
 #define DEFINE_BLOCK_PROPERTIES(_state, _conf)                          \
-    DEFINE_PROP_DRIVE("drive", _state, _conf.dinfo),                    \
+    DEFINE_PROP_DRIVE("drive", _state, _conf.bs),                       \
     DEFINE_PROP_UINT16("logical_block_size", _state,                    \
                        _conf.logical_block_size, 512),                  \
     DEFINE_PROP_UINT16("physical_block_size", _state,                   \
diff --git a/hw/fdc.c b/hw/fdc.c
index 45a876d..08712bc 100644
--- a/hw/fdc.c
+++ b/hw/fdc.c
@@ -80,7 +80,6 @@ typedef enum FDiskFlags {
 } FDiskFlags;
 
 typedef struct FDrive {
-    DriveInfo *dinfo;
     BlockDriverState *bs;
     /* Drive status */
     FDriveType drive;
@@ -100,7 +99,6 @@ typedef struct FDrive {
 static void fd_init(FDrive *drv)
 {
     /* Drive */
-    drv->bs = drv->dinfo ? drv->dinfo->bdrv : NULL;
     drv->drive = FDRIVE_DRV_NONE;
     drv->perpendicular = 0;
     /* Disk */
@@ -1862,10 +1860,10 @@ FDCtrl *fdctrl_init_isa(DriveInfo **fds)
 
     dev = isa_create("isa-fdc");
     if (fds[0]) {
-        qdev_prop_set_drive(&dev->qdev, "driveA", fds[0]);
+        qdev_prop_set_drive(&dev->qdev, "driveA", fds[0]->bdrv);
     }
     if (fds[1]) {
-        qdev_prop_set_drive(&dev->qdev, "driveB", fds[1]);
+        qdev_prop_set_drive(&dev->qdev, "driveB", fds[1]->bdrv);
     }
     if (qdev_init(&dev->qdev) < 0)
         return NULL;
@@ -1884,10 +1882,10 @@ FDCtrl *fdctrl_init_sysbus(qemu_irq irq, int dma_chann,
     fdctrl = &sys->state;
     fdctrl->dma_chann = dma_chann; /* FIXME */
     if (fds[0]) {
-        qdev_prop_set_drive(dev, "driveA", fds[0]);
+        qdev_prop_set_drive(dev, "driveA", fds[0]->bdrv);
     }
     if (fds[1]) {
-        qdev_prop_set_drive(dev, "driveB", fds[1]);
+        qdev_prop_set_drive(dev, "driveB", fds[1]->bdrv);
     }
     qdev_init_nofail(dev);
     sysbus_connect_irq(&sys->busdev, 0, irq);
@@ -1905,7 +1903,7 @@ FDCtrl *sun4m_fdctrl_init(qemu_irq irq, target_phys_addr_t io_base,
 
     dev = qdev_create(NULL, "SUNW,fdtwo");
     if (fds[0]) {
-        qdev_prop_set_drive(dev, "drive", fds[0]);
+        qdev_prop_set_drive(dev, "drive", fds[0]->bdrv);
     }
     qdev_init_nofail(dev);
     sys = DO_UPCAST(FDCtrlSysBus, busdev.qdev, dev);
@@ -2030,8 +2028,8 @@ static ISADeviceInfo isa_fdc_info = {
     .qdev.vmsd  = &vmstate_isa_fdc,
     .qdev.reset = fdctrl_external_reset_isa,
     .qdev.props = (Property[]) {
-        DEFINE_PROP_DRIVE("driveA", FDCtrlISABus, state.drives[0].dinfo),
-        DEFINE_PROP_DRIVE("driveB", FDCtrlISABus, state.drives[1].dinfo),
+        DEFINE_PROP_DRIVE("driveA", FDCtrlISABus, state.drives[0].bs),
+        DEFINE_PROP_DRIVE("driveB", FDCtrlISABus, state.drives[1].bs),
         DEFINE_PROP_END_OF_LIST(),
     },
 };
@@ -2053,8 +2051,8 @@ static SysBusDeviceInfo sysbus_fdc_info = {
     .qdev.vmsd  = &vmstate_sysbus_fdc,
     .qdev.reset = fdctrl_external_reset_sysbus,
     .qdev.props = (Property[]) {
-        DEFINE_PROP_DRIVE("driveA", FDCtrlSysBus, state.drives[0].dinfo),
-        DEFINE_PROP_DRIVE("driveB", FDCtrlSysBus, state.drives[1].dinfo),
+        DEFINE_PROP_DRIVE("driveA", FDCtrlSysBus, state.drives[0].bs),
+        DEFINE_PROP_DRIVE("driveB", FDCtrlSysBus, state.drives[1].bs),
         DEFINE_PROP_END_OF_LIST(),
     },
 };
@@ -2066,7 +2064,7 @@ static SysBusDeviceInfo sun4m_fdc_info = {
     .qdev.vmsd  = &vmstate_sysbus_fdc,
     .qdev.reset = fdctrl_external_reset_sysbus,
     .qdev.props = (Property[]) {
-        DEFINE_PROP_DRIVE("drive", FDCtrlSysBus, state.drives[0].dinfo),
+        DEFINE_PROP_DRIVE("drive", FDCtrlSysBus, state.drives[0].bs),
         DEFINE_PROP_END_OF_LIST(),
     },
 };
diff --git a/hw/ide/core.c b/hw/ide/core.c
index 0b3b7c2..cc4591b 100644
--- a/hw/ide/core.c
+++ b/hw/ide/core.c
@@ -2594,15 +2594,15 @@ void ide_bus_reset(IDEBus *bus)
     ide_clear_hob(bus);
 }
 
-void ide_init_drive(IDEState *s, DriveInfo *dinfo,
+void ide_init_drive(IDEState *s, BlockDriverState *bs,
                     const char *version, const char *serial)
 {
     int cylinders, heads, secs;
     uint64_t nb_sectors;
 
-    s->bs = dinfo->bdrv;
-    bdrv_get_geometry(s->bs, &nb_sectors);
-    bdrv_guess_geometry(s->bs, &cylinders, &heads, &secs);
+    s->bs = bs;
+    bdrv_get_geometry(bs, &nb_sectors);
+    bdrv_guess_geometry(bs, &cylinders, &heads, &secs);
     s->cylinders = cylinders;
     s->heads = heads;
     s->sectors = secs;
@@ -2613,11 +2613,11 @@ void ide_init_drive(IDEState *s, DriveInfo *dinfo,
     s->smart_autosave = 1;
     s->smart_errors = 0;
     s->smart_selftest_count = 0;
-    if (bdrv_get_type_hint(s->bs) == BDRV_TYPE_CDROM) {
+    if (bdrv_get_type_hint(bs) == BDRV_TYPE_CDROM) {
         s->is_cdrom = 1;
-        bdrv_set_change_cb(s->bs, cdrom_change_cb, s);
+        bdrv_set_change_cb(bs, cdrom_change_cb, s);
     }
-    if (serial && *serial) {
+    if (serial) {
         strncpy(s->drive_serial_str, serial, sizeof(s->drive_serial_str));
     } else {
         snprintf(s->drive_serial_str, sizeof(s->drive_serial_str),
@@ -2668,7 +2668,8 @@ void ide_init2_with_non_qdev_drives(IDEBus *bus, DriveInfo *hd0,
         dinfo = i == 0 ? hd0 : hd1;
         ide_init1(bus, i);
         if (dinfo) {
-            ide_init_drive(&bus->ifs[i], dinfo, NULL, dinfo->serial);
+            ide_init_drive(&bus->ifs[i], dinfo->bdrv, NULL,
+                           *dinfo->serial ? dinfo->serial : NULL);
         } else {
             ide_reset(&bus->ifs[i]);
         }
diff --git a/hw/ide/internal.h b/hw/ide/internal.h
index eef1ee1..0125a9f 100644
--- a/hw/ide/internal.h
+++ b/hw/ide/internal.h
@@ -555,7 +555,7 @@ uint32_t ide_data_readw(void *opaque, uint32_t addr);
 void ide_data_writel(void *opaque, uint32_t addr, uint32_t val);
 uint32_t ide_data_readl(void *opaque, uint32_t addr);
 
-void ide_init_drive(IDEState *s, DriveInfo *dinfo,
+void ide_init_drive(IDEState *s, BlockDriverState *bs,
                     const char *version, const char *serial);
 void ide_init2(IDEBus *bus, qemu_irq irq);
 void ide_init2_with_non_qdev_drives(IDEBus *bus, DriveInfo *hd0,
diff --git a/hw/ide/qdev.c b/hw/ide/qdev.c
index 127478b..3bb94c6 100644
--- a/hw/ide/qdev.c
+++ b/hw/ide/qdev.c
@@ -39,7 +39,7 @@ static int ide_qdev_init(DeviceState *qdev, DeviceInfo *base)
     IDEDeviceInfo *info = DO_UPCAST(IDEDeviceInfo, qdev, base);
     IDEBus *bus = DO_UPCAST(IDEBus, qbus, qdev->parent_bus);
 
-    if (!dev->conf.dinfo) {
+    if (!dev->conf.bs) {
         fprintf(stderr, "%s: no drive specified\n", qdev->info->name);
         goto err;
     }
@@ -83,7 +83,7 @@ IDEDevice *ide_create_drive(IDEBus *bus, int unit, DriveInfo *drive)
 
     dev = qdev_create(&bus->qbus, "ide-drive");
     qdev_prop_set_uint32(dev, "unit", unit);
-    qdev_prop_set_drive(dev, "drive", drive);
+    qdev_prop_set_drive(dev, "drive", drive->bdrv);
     qdev_init_nofail(dev);
     return DO_UPCAST(IDEDevice, qdev, dev);
 }
@@ -99,14 +99,18 @@ static int ide_drive_initfn(IDEDevice *dev)
     IDEBus *bus = DO_UPCAST(IDEBus, qbus, dev->qdev.parent_bus);
     IDEState *s = bus->ifs + dev->unit;
     const char *serial;
+    DriveInfo *dinfo;
 
     serial = dev->serial;
     if (!serial) {
         /* try to fall back to value set with legacy -drive serial=... */
-        serial = dev->conf.dinfo->serial;
+        dinfo = drive_of_blockdev(dev->conf.bs);
+        if (*dinfo->serial) {
+            serial = dinfo->serial;
+        }
     }
 
-    ide_init_drive(s, dev->conf.dinfo, dev->version, serial);
+    ide_init_drive(s, dev->conf.bs, dev->version, serial);
 
     if (!dev->version) {
         dev->version = qemu_strdup(s->version);
diff --git a/hw/pci-hotplug.c b/hw/pci-hotplug.c
index 55c9fe3..d743192 100644
--- a/hw/pci-hotplug.c
+++ b/hw/pci-hotplug.c
@@ -89,7 +89,7 @@ static int scsi_hot_add(Monitor *mon, DeviceState *adapter,
      * specified).
      */
     dinfo->unit = qemu_opt_get_number(dinfo->opts, "unit", -1);
-    scsidev = scsi_bus_legacy_add_drive(scsibus, dinfo, dinfo->unit);
+    scsidev = scsi_bus_legacy_add_drive(scsibus, dinfo->bdrv, dinfo->unit);
     if (!scsidev) {
         return -1;
     }
@@ -214,7 +214,7 @@ static PCIDevice *qemu_pci_hot_add_storage(Monitor *mon,
             return NULL;
         }
         dev = pci_create(bus, devfn, "virtio-blk-pci");
-        qdev_prop_set_drive(&dev->qdev, "drive", dinfo);
+        qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv);
         if (qdev_init(&dev->qdev) < 0)
             dev = NULL;
         break;
diff --git a/hw/qdev-properties.c b/hw/qdev-properties.c
index 15ca6d3..257233e 100644
--- a/hw/qdev-properties.c
+++ b/hw/qdev-properties.c
@@ -285,33 +285,36 @@ PropertyInfo qdev_prop_string = {
 
 static int parse_drive(DeviceState *dev, Property *prop, const char *str)
 {
-    DriveInfo **ptr = qdev_get_prop_ptr(dev, prop);
+    BlockDriverState **ptr = qdev_get_prop_ptr(dev, prop);
+    BlockDriverState *bs;
 
-    *ptr = drive_get_by_id(str);
-    if (*ptr == NULL)
+    bs = bdrv_find(str);
+    if (bs == NULL)
         return -ENOENT;
+    *ptr = bs;
     return 0;
 }
 
 static void free_drive(DeviceState *dev, Property *prop)
 {
-    DriveInfo **ptr = qdev_get_prop_ptr(dev, prop);
+    BlockDriverState **ptr = qdev_get_prop_ptr(dev, prop);
 
     if (*ptr) {
-        blockdev_auto_del((*ptr)->bdrv);
+        blockdev_auto_del(*ptr);
     }
 }
 
 static int print_drive(DeviceState *dev, Property *prop, char *dest, size_t len)
 {
-    DriveInfo **ptr = qdev_get_prop_ptr(dev, prop);
-    return snprintf(dest, len, "%s", (*ptr) ? (*ptr)->id : "<null>");
+    BlockDriverState **ptr = qdev_get_prop_ptr(dev, prop);
+    return snprintf(dest, len, "%s",
+                    *ptr ? bdrv_get_device_name(*ptr) : "<null>");
 }
 
 PropertyInfo qdev_prop_drive = {
     .name  = "drive",
     .type  = PROP_TYPE_DRIVE,
-    .size  = sizeof(DriveInfo*),
+    .size  = sizeof(BlockDriverState *),
     .parse = parse_drive,
     .print = print_drive,
     .free  = free_drive,
@@ -637,7 +640,7 @@ void qdev_prop_set_string(DeviceState *dev, const char *name, char *value)
     qdev_prop_set(dev, name, &value, PROP_TYPE_STRING);
 }
 
-void qdev_prop_set_drive(DeviceState *dev, const char *name, DriveInfo *value)
+void qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value)
 {
     qdev_prop_set(dev, name, &value, PROP_TYPE_DRIVE);
 }
diff --git a/hw/qdev.h b/hw/qdev.h
index be5ad67..7a01a81 100644
--- a/hw/qdev.h
+++ b/hw/qdev.h
@@ -253,8 +253,8 @@ extern PropertyInfo qdev_prop_pci_devfn;
     DEFINE_PROP(_n, _s, _f, qdev_prop_netdev, VLANClientState*)
 #define DEFINE_PROP_VLAN(_n, _s, _f)             \
     DEFINE_PROP(_n, _s, _f, qdev_prop_vlan, VLANState*)
-#define DEFINE_PROP_DRIVE(_n, _s, _f)             \
-    DEFINE_PROP(_n, _s, _f, qdev_prop_drive, DriveInfo*)
+#define DEFINE_PROP_DRIVE(_n, _s, _f) \
+    DEFINE_PROP(_n, _s, _f, qdev_prop_drive, BlockDriverState *)
 #define DEFINE_PROP_MACADDR(_n, _s, _f)         \
     DEFINE_PROP(_n, _s, _f, qdev_prop_macaddr, MACAddr)
 
@@ -275,7 +275,7 @@ void qdev_prop_set_string(DeviceState *dev, const char *name, char *value);
 void qdev_prop_set_chr(DeviceState *dev, const char *name, CharDriverState *value);
 void qdev_prop_set_netdev(DeviceState *dev, const char *name, VLANClientState *value);
 void qdev_prop_set_vlan(DeviceState *dev, const char *name, VLANState *value);
-void qdev_prop_set_drive(DeviceState *dev, const char *name, DriveInfo *value);
+void qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value);
 void qdev_prop_set_macaddr(DeviceState *dev, const char *name, uint8_t *value);
 /* FIXME: Remove opaque pointer properties.  */
 void qdev_prop_set_ptr(DeviceState *dev, const char *name, void *value);
diff --git a/hw/s390-virtio.c b/hw/s390-virtio.c
index 898f442..c7c3fc9 100644
--- a/hw/s390-virtio.c
+++ b/hw/s390-virtio.c
@@ -262,7 +262,7 @@ static void s390_init(ram_addr_t ram_size,
         }
 
         dev = qdev_create((BusState *)s390_bus, "virtio-blk-s390");
-        qdev_prop_set_drive(dev, "drive", dinfo);
+        qdev_prop_set_drive(dev, "drive", dinfo->bdrv);
         qdev_init_nofail(dev);
     }
 }
diff --git a/hw/scsi-bus.c b/hw/scsi-bus.c
index d5b66c1..2c58aca 100644
--- a/hw/scsi-bus.c
+++ b/hw/scsi-bus.c
@@ -83,15 +83,15 @@ void scsi_qdev_register(SCSIDeviceInfo *info)
 }
 
 /* handle legacy '-drive if=scsi,...' cmd line args */
-SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, DriveInfo *dinfo, int unit)
+SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, BlockDriverState *bdrv, int unit)
 {
     const char *driver;
     DeviceState *dev;
 
-    driver = bdrv_is_sg(dinfo->bdrv) ? "scsi-generic" : "scsi-disk";
+    driver = bdrv_is_sg(bdrv) ? "scsi-generic" : "scsi-disk";
     dev = qdev_create(&bus->qbus, driver);
     qdev_prop_set_uint32(dev, "scsi-id", unit);
-    qdev_prop_set_drive(dev, "drive", dinfo);
+    qdev_prop_set_drive(dev, "drive", bdrv);
     if (qdev_init(dev) < 0)
         return NULL;
     return DO_UPCAST(SCSIDevice, qdev, dev);
@@ -107,7 +107,7 @@ int scsi_bus_legacy_handle_cmdline(SCSIBus *bus)
         if (dinfo == NULL) {
             continue;
         }
-        if (!scsi_bus_legacy_add_drive(bus, dinfo, unit)) {
+        if (!scsi_bus_legacy_add_drive(bus, dinfo->bdrv, unit)) {
             res = -1;
             break;
         }
diff --git a/hw/scsi-disk.c b/hw/scsi-disk.c
index d76e640..9c78979 100644
--- a/hw/scsi-disk.c
+++ b/hw/scsi-disk.c
@@ -1043,26 +1043,24 @@ static void scsi_destroy(SCSIDevice *dev)
     SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, dev);
 
     scsi_disk_purge_requests(s);
-    blockdev_mark_auto_del(s->qdev.conf.dinfo->bdrv);
+    blockdev_mark_auto_del(s->qdev.conf.bs);
 }
 
 static int scsi_disk_initfn(SCSIDevice *dev)
 {
     SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, dev);
+    DriveInfo *dinfo;
 
-    if (!s->qdev.conf.dinfo || !s->qdev.conf.dinfo->bdrv) {
+    if (!s->qdev.conf.bs) {
         error_report("scsi-disk: drive property not set");
         return -1;
     }
-    s->bs = s->qdev.conf.dinfo->bdrv;
+    s->bs = s->qdev.conf.bs;
 
     if (!s->serial) {
-        if (*dev->conf.dinfo->serial) {
-            /* try to fall back to value set with legacy -drive serial=... */
-            s->serial = qemu_strdup(dev->conf.dinfo->serial);
-        } else {
-            s->serial = qemu_strdup("0");
-        }
+        /* try to fall back to value set with legacy -drive serial=... */
+        dinfo = drive_of_blockdev(s->bs);
+        s->serial = qemu_strdup(*dinfo->serial ? dinfo->serial : "0");
     }
 
     if (!s->version) {
diff --git a/hw/scsi-generic.c b/hw/scsi-generic.c
index 1859c94..79347f4 100644
--- a/hw/scsi-generic.c
+++ b/hw/scsi-generic.c
@@ -453,7 +453,7 @@ static void scsi_destroy(SCSIDevice *d)
         r = DO_UPCAST(SCSIGenericReq, req, QTAILQ_FIRST(&s->qdev.requests));
         scsi_remove_request(r);
     }
-    blockdev_mark_auto_del(s->qdev.conf.dinfo->bdrv);
+    blockdev_mark_auto_del(s->qdev.conf.bs);
 }
 
 static int scsi_generic_initfn(SCSIDevice *dev)
@@ -462,11 +462,11 @@ static int scsi_generic_initfn(SCSIDevice *dev)
     int sg_version;
     struct sg_scsi_id scsiid;
 
-    if (!s->qdev.conf.dinfo || !s->qdev.conf.dinfo->bdrv) {
+    if (!s->qdev.conf.bs) {
         error_report("scsi-generic: drive property not set");
         return -1;
     }
-    s->bs = s->qdev.conf.dinfo->bdrv;
+    s->bs = s->qdev.conf.bs;
 
     /* check we are really using a /dev/sg* file */
     if (!bdrv_is_sg(s->bs)) {
diff --git a/hw/scsi.h b/hw/scsi.h
index b1b5f73..4fbf1d5 100644
--- a/hw/scsi.h
+++ b/hw/scsi.h
@@ -97,7 +97,7 @@ static inline SCSIBus *scsi_bus_from_device(SCSIDevice *d)
     return DO_UPCAST(SCSIBus, qbus, d->qdev.parent_bus);
 }
 
-SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, DriveInfo *dinfo, int unit);
+SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, BlockDriverState *bdrv, int unit);
 int scsi_bus_legacy_handle_cmdline(SCSIBus *bus);
 
 void scsi_dev_clear_sense(SCSIDevice *dev);
diff --git a/hw/usb-msd.c b/hw/usb-msd.c
index 3dbfcab..19a14b4 100644
--- a/hw/usb-msd.c
+++ b/hw/usb-msd.c
@@ -522,9 +522,9 @@ static void usb_msd_password_cb(void *opaque, int err)
 static int usb_msd_initfn(USBDevice *dev)
 {
     MSDState *s = DO_UPCAST(MSDState, dev, dev);
-    DriveInfo *dinfo = s->conf.dinfo;
+    BlockDriverState *bs = s->conf.bs;
 
-    if (!dinfo || !dinfo->bdrv) {
+    if (!bs) {
         error_report("usb-msd: drive property not set");
         return -1;
     }
@@ -538,21 +538,20 @@ static int usb_msd_initfn(USBDevice *dev)
      *
      * The hack is probably a bad idea.
      */
-    s->conf.dinfo = NULL;
+    s->conf.bs = NULL;
 
     s->dev.speed = USB_SPEED_FULL;
     scsi_bus_new(&s->bus, &s->dev.qdev, 0, 1, usb_msd_command_complete);
-    s->scsi_dev = scsi_bus_legacy_add_drive(&s->bus, dinfo, 0);
+    s->scsi_dev = scsi_bus_legacy_add_drive(&s->bus, bs, 0);
     if (!s->scsi_dev) {
         return -1;
     }
     s->bus.qbus.allow_hotplug = 0;
     usb_msd_handle_reset(dev);
 
-    if (bdrv_key_required(dinfo->bdrv)) {
+    if (bdrv_key_required(bs)) {
         if (cur_mon) {
-            monitor_read_bdrv_key_start(cur_mon, dinfo->bdrv,
-                                        usb_msd_password_cb, s);
+            monitor_read_bdrv_key_start(cur_mon, bs, usb_msd_password_cb, s);
             s->dev.auto_attach = 0;
         } else {
             autostart = 0;
@@ -610,7 +609,7 @@ static USBDevice *usb_msd_init(const char *filename)
     if (!dev) {
         return NULL;
     }
-    qdev_prop_set_drive(&dev->qdev, "drive", dinfo);
+    qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv);
     if (qdev_init(&dev->qdev) < 0)
         return NULL;
 
diff --git a/hw/virtio-blk.c b/hw/virtio-blk.c
index 0bf929a..71c5d21 100644
--- a/hw/virtio-blk.c
+++ b/hw/virtio-blk.c
@@ -489,7 +489,7 @@ VirtIODevice *virtio_blk_init(DeviceState *dev, BlockConf *conf)
     s->vdev.get_config = virtio_blk_update_config;
     s->vdev.get_features = virtio_blk_get_features;
     s->vdev.reset = virtio_blk_reset;
-    s->bs = conf->dinfo->bdrv;
+    s->bs = conf->bs;
     s->conf = conf;
     s->rq = NULL;
     s->sector_mask = (s->conf->logical_block_size / BDRV_SECTOR_SIZE) - 1;
diff --git a/hw/virtio-pci.c b/hw/virtio-pci.c
index 31a68fe..c6ef825 100644
--- a/hw/virtio-pci.c
+++ b/hw/virtio-pci.c
@@ -547,7 +547,7 @@ static int virtio_blk_init_pci(PCIDevice *pci_dev)
         proxy->class_code != PCI_CLASS_STORAGE_OTHER)
         proxy->class_code = PCI_CLASS_STORAGE_SCSI;
 
-    if (!proxy->block.dinfo) {
+    if (!proxy->block.bs) {
         error_report("virtio-blk-pci: drive property not set");
         return -1;
     }
@@ -571,7 +571,7 @@ static int virtio_blk_exit_pci(PCIDevice *pci_dev)
 {
     VirtIOPCIProxy *proxy = DO_UPCAST(VirtIOPCIProxy, pci_dev, pci_dev);
 
-    blockdev_mark_auto_del(proxy->block.dinfo->bdrv);
+    blockdev_mark_auto_del(proxy->block.bs);
     return virtio_exit_pci(pci_dev);
 }
 
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 06/12] qdev: Decouple qdev_prop_drive from DriveInfo

[Christoph Hellwig]

On Fri, Jun 25, 2010 at 06:53:26PM +0200, Markus Armbruster wrote:
> Make the property point to BlockDriverState, cutting out the DriveInfo
> middleman.  This prepares the ground for block devices that don't have
> a DriveInfo.
> 
> Currently all user-defined ones have a DriveInfo, because the only way
> to define one is -drive & friends (they go through drive_init()).
> DriveInfo is closely tied to -drive, and like -drive, it mixes
> information about host and guest part of the block device.  I'm
> working towards a new way to define block devices, with clean
> host/guest separation, and I need to get DriveInfo out of the way for
> that.
> 
> Fortunately, the device models are perfectly happy with
> BlockDriverState, except for two places: ide_drive_initfn() and
> scsi_disk_initfn() need to check the DriveInfo for a serial number set
> with legacy -drive serial=...  Use drive_of_blockdev() there.
> 
> Device model code should now use DriveInfo only when explicitly
> dealing with drives defined the old way, i.e. without -device.

This looks good to me.  Assuming the size change of the property can't
cause any problems:


Reviewed-by: Christoph Hellwig <hch@lst.de>

[Qemu-devel] [PATCH 07/12] blockdev: drive_get_by_id() is no longer used, remove

[Markus Armbruster]

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 blockdev.c |   12 ------------
 blockdev.h |    1 -
 2 files changed, 0 insertions(+), 13 deletions(-)

diff --git a/blockdev.c b/blockdev.c
index 827ea1c..3747098 100644
--- a/blockdev.c
+++ b/blockdev.c
@@ -75,18 +75,6 @@ DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit)
     return NULL;
 }
 
-DriveInfo *drive_get_by_id(const char *id)
-{
-    DriveInfo *dinfo;
-
-    QTAILQ_FOREACH(dinfo, &drives, next) {
-        if (strcmp(id, dinfo->id))
-            continue;
-        return dinfo;
-    }
-    return NULL;
-}
-
 int drive_get_max_bus(BlockInterfaceType type)
 {
     int max_bus;
diff --git a/blockdev.h b/blockdev.h
index 418ebb6..a72d335 100644
--- a/blockdev.h
+++ b/blockdev.h
@@ -41,7 +41,6 @@ typedef struct DriveInfo {
 #define MAX_SCSI_DEVS	7
 
 extern DriveInfo *drive_get(BlockInterfaceType type, int bus, int unit);
-extern DriveInfo *drive_get_by_id(const char *id);
 extern int drive_get_max_bus(BlockInterfaceType type);
 extern void drive_uninit(DriveInfo *dinfo);
 extern DriveInfo *drive_of_blockdev(BlockDriverState *bs);
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 07/12] blockdev: drive_get_by_id() is no longer used, remove

[Christoph Hellwig]

Looks good,


Reviewed-by: Christoph Hellwig <hch@lst.de>

[Qemu-devel] [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Markus Armbruster]

For instance, -device scsi-disk,drive=foo -device scsi-disk,drive=foo
happily creates two SCSI disks connected to the same block device.
It's all downhill from there.

Device usb-storage deliberately attaches twice to the same blockdev,
which fails with the fix in place.  Detach before the second attach
there.

Also catch attempt to delete while a guest device model is attached.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 block.c              |   22 ++++++++++++++++++++++
 block.h              |    3 +++
 block_int.h          |    2 ++
 hw/fdc.c             |   10 +++++-----
 hw/ide/qdev.c        |    2 +-
 hw/pci-hotplug.c     |    5 ++++-
 hw/qdev-properties.c |   21 ++++++++++++++++++++-
 hw/qdev.h            |    3 ++-
 hw/s390-virtio.c     |    2 +-
 hw/scsi-bus.c        |    4 +++-
 hw/usb-msd.c         |   11 +++++++----
 11 files changed, 70 insertions(+), 15 deletions(-)

diff --git a/block.c b/block.c
index e71a771..5e0ffa0 100644
--- a/block.c
+++ b/block.c
@@ -659,6 +659,8 @@ void bdrv_close_all(void)
 
 void bdrv_delete(BlockDriverState *bs)
 {
+    assert(!bs->peer);
+
     /* remove from list, if necessary */
     if (bs->device_name[0] != '\0') {
         QTAILQ_REMOVE(&bdrv_states, bs, list);
@@ -672,6 +674,26 @@ void bdrv_delete(BlockDriverState *bs)
     qemu_free(bs);
 }
 
+int bdrv_attach(BlockDriverState *bs, DeviceState *qdev)
+{
+    if (bs->peer) {
+        return -EBUSY;
+    }
+    bs->peer = qdev;
+    return 0;
+}
+
+void bdrv_detach(BlockDriverState *bs, DeviceState *qdev)
+{
+    assert(bs->peer == qdev);
+    bs->peer = NULL;
+}
+
+DeviceState *bdrv_get_attached(BlockDriverState *bs)
+{
+    return bs->peer;
+}
+
 /*
  * Run consistency checks on an image
  *
diff --git a/block.h b/block.h
index 6a157f4..88ac06e 100644
--- a/block.h
+++ b/block.h
@@ -71,6 +71,9 @@ int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags);
 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
               BlockDriver *drv);
 void bdrv_close(BlockDriverState *bs);
+int bdrv_attach(BlockDriverState *bs, DeviceState *qdev);
+void bdrv_detach(BlockDriverState *bs, DeviceState *qdev);
+DeviceState *bdrv_get_attached(BlockDriverState *bs);
 int bdrv_check(BlockDriverState *bs);
 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
               uint8_t *buf, int nb_sectors);
diff --git a/block_int.h b/block_int.h
index e60aed4..a94b801 100644
--- a/block_int.h
+++ b/block_int.h
@@ -148,6 +148,8 @@ struct BlockDriverState {
     BlockDriver *drv; /* NULL means no media */
     void *opaque;
 
+    DeviceState *peer;
+
     char filename[1024];
     char backing_file[1024]; /* if non zero, the image is a diff of
                                 this file image */
diff --git a/hw/fdc.c b/hw/fdc.c
index 08712bc..1496cfa 100644
--- a/hw/fdc.c
+++ b/hw/fdc.c
@@ -1860,10 +1860,10 @@ FDCtrl *fdctrl_init_isa(DriveInfo **fds)
 
     dev = isa_create("isa-fdc");
     if (fds[0]) {
-        qdev_prop_set_drive(&dev->qdev, "driveA", fds[0]->bdrv);
+        qdev_prop_set_drive_nofail(&dev->qdev, "driveA", fds[0]->bdrv);
     }
     if (fds[1]) {
-        qdev_prop_set_drive(&dev->qdev, "driveB", fds[1]->bdrv);
+        qdev_prop_set_drive_nofail(&dev->qdev, "driveB", fds[1]->bdrv);
     }
     if (qdev_init(&dev->qdev) < 0)
         return NULL;
@@ -1882,10 +1882,10 @@ FDCtrl *fdctrl_init_sysbus(qemu_irq irq, int dma_chann,
     fdctrl = &sys->state;
     fdctrl->dma_chann = dma_chann; /* FIXME */
     if (fds[0]) {
-        qdev_prop_set_drive(dev, "driveA", fds[0]->bdrv);
+        qdev_prop_set_drive_nofail(dev, "driveA", fds[0]->bdrv);
     }
     if (fds[1]) {
-        qdev_prop_set_drive(dev, "driveB", fds[1]->bdrv);
+        qdev_prop_set_drive_nofail(dev, "driveB", fds[1]->bdrv);
     }
     qdev_init_nofail(dev);
     sysbus_connect_irq(&sys->busdev, 0, irq);
@@ -1903,7 +1903,7 @@ FDCtrl *sun4m_fdctrl_init(qemu_irq irq, target_phys_addr_t io_base,
 
     dev = qdev_create(NULL, "SUNW,fdtwo");
     if (fds[0]) {
-        qdev_prop_set_drive(dev, "drive", fds[0]->bdrv);
+        qdev_prop_set_drive_nofail(dev, "drive", fds[0]->bdrv);
     }
     qdev_init_nofail(dev);
     sys = DO_UPCAST(FDCtrlSysBus, busdev.qdev, dev);
diff --git a/hw/ide/qdev.c b/hw/ide/qdev.c
index 3bb94c6..b4bc5ac 100644
--- a/hw/ide/qdev.c
+++ b/hw/ide/qdev.c
@@ -83,7 +83,7 @@ IDEDevice *ide_create_drive(IDEBus *bus, int unit, DriveInfo *drive)
 
     dev = qdev_create(&bus->qbus, "ide-drive");
     qdev_prop_set_uint32(dev, "unit", unit);
-    qdev_prop_set_drive(dev, "drive", drive->bdrv);
+    qdev_prop_set_drive_nofail(dev, "drive", drive->bdrv);
     qdev_init_nofail(dev);
     return DO_UPCAST(IDEDevice, qdev, dev);
 }
diff --git a/hw/pci-hotplug.c b/hw/pci-hotplug.c
index d743192..b47e01e 100644
--- a/hw/pci-hotplug.c
+++ b/hw/pci-hotplug.c
@@ -214,7 +214,10 @@ static PCIDevice *qemu_pci_hot_add_storage(Monitor *mon,
             return NULL;
         }
         dev = pci_create(bus, devfn, "virtio-blk-pci");
-        qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv);
+        if (qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv) < 0) {
+            dev = NULL;
+            break;
+        }
         if (qdev_init(&dev->qdev) < 0)
             dev = NULL;
         break;
diff --git a/hw/qdev-properties.c b/hw/qdev-properties.c
index 257233e..caf6798 100644
--- a/hw/qdev-properties.c
+++ b/hw/qdev-properties.c
@@ -291,6 +291,8 @@ static int parse_drive(DeviceState *dev, Property *prop, const char *str)
     bs = bdrv_find(str);
     if (bs == NULL)
         return -ENOENT;
+    if (bdrv_attach(bs, dev) < 0)
+        return -EEXIST;
     *ptr = bs;
     return 0;
 }
@@ -300,6 +302,7 @@ static void free_drive(DeviceState *dev, Property *prop)
     BlockDriverState **ptr = qdev_get_prop_ptr(dev, prop);
 
     if (*ptr) {
+        bdrv_detach(*ptr, dev);
         blockdev_auto_del(*ptr);
     }
 }
@@ -640,11 +643,27 @@ void qdev_prop_set_string(DeviceState *dev, const char *name, char *value)
     qdev_prop_set(dev, name, &value, PROP_TYPE_STRING);
 }
 
-void qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value)
+int qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value)
 {
+    int res;
+
     qdev_prop_set(dev, name, &value, PROP_TYPE_DRIVE);
+    res = bdrv_attach(value, dev);
+    if (res < 0) {
+        error_report("Can't attach drive %s to %s.%s: %s",
+                     bdrv_get_device_name(value),
+                     dev->id ? dev->id : dev->info->name,
+                     name, strerror(-res));
+    }
+    return res;
 }
 
+void qdev_prop_set_drive_nofail(DeviceState *dev, const char *name, BlockDriverState *value)
+{
+    if (qdev_prop_set_drive(dev, name, value) < 0) {
+        exit(1);
+    }
+}
 void qdev_prop_set_chr(DeviceState *dev, const char *name, CharDriverState *value)
 {
     qdev_prop_set(dev, name, &value, PROP_TYPE_CHR);
diff --git a/hw/qdev.h b/hw/qdev.h
index 7a01a81..cbc89f2 100644
--- a/hw/qdev.h
+++ b/hw/qdev.h
@@ -275,7 +275,8 @@ void qdev_prop_set_string(DeviceState *dev, const char *name, char *value);
 void qdev_prop_set_chr(DeviceState *dev, const char *name, CharDriverState *value);
 void qdev_prop_set_netdev(DeviceState *dev, const char *name, VLANClientState *value);
 void qdev_prop_set_vlan(DeviceState *dev, const char *name, VLANState *value);
-void qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value);
+int qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value) QEMU_WARN_UNUSED_RESULT;
+void qdev_prop_set_drive_nofail(DeviceState *dev, const char *name, BlockDriverState *value);
 void qdev_prop_set_macaddr(DeviceState *dev, const char *name, uint8_t *value);
 /* FIXME: Remove opaque pointer properties.  */
 void qdev_prop_set_ptr(DeviceState *dev, const char *name, void *value);
diff --git a/hw/s390-virtio.c b/hw/s390-virtio.c
index c7c3fc9..6af58e2 100644
--- a/hw/s390-virtio.c
+++ b/hw/s390-virtio.c
@@ -262,7 +262,7 @@ static void s390_init(ram_addr_t ram_size,
         }
 
         dev = qdev_create((BusState *)s390_bus, "virtio-blk-s390");
-        qdev_prop_set_drive(dev, "drive", dinfo->bdrv);
+        qdev_prop_set_drive_nofail(dev, "drive", dinfo->bdrv);
         qdev_init_nofail(dev);
     }
 }
diff --git a/hw/scsi-bus.c b/hw/scsi-bus.c
index 2c58aca..9678328 100644
--- a/hw/scsi-bus.c
+++ b/hw/scsi-bus.c
@@ -91,7 +91,9 @@ SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, BlockDriverState *bdrv, int
     driver = bdrv_is_sg(bdrv) ? "scsi-generic" : "scsi-disk";
     dev = qdev_create(&bus->qbus, driver);
     qdev_prop_set_uint32(dev, "scsi-id", unit);
-    qdev_prop_set_drive(dev, "drive", bdrv);
+    if (qdev_prop_set_drive(dev, "drive", bdrv) < 0) {
+        return NULL;
+    }
     if (qdev_init(dev) < 0)
         return NULL;
     return DO_UPCAST(SCSIDevice, qdev, dev);
diff --git a/hw/usb-msd.c b/hw/usb-msd.c
index 19a14b4..008cc0f 100644
--- a/hw/usb-msd.c
+++ b/hw/usb-msd.c
@@ -532,12 +532,13 @@ static int usb_msd_initfn(USBDevice *dev)
     /*
      * Hack alert: this pretends to be a block device, but it's really
      * a SCSI bus that can serve only a single device, which it
-     * creates automatically.  Two drive properties pointing to the
-     * same drive is not good: free_drive() dies for the second one.
-     * Zap the one we're not going to use.
+     * creates automatically.  But first it needs to detach from its
+     * blockdev, or else scsi_bus_legacy_add_drive() dies when it
+     * attaches again.
      *
      * The hack is probably a bad idea.
      */
+    bdrv_detach(bs, &s->dev.qdev);
     s->conf.bs = NULL;
 
     s->dev.speed = USB_SPEED_FULL;
@@ -609,7 +610,9 @@ static USBDevice *usb_msd_init(const char *filename)
     if (!dev) {
         return NULL;
     }
-    qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv);
+    if (qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv) < 0) {
+        return NULL;
+    }
     if (qdev_init(&dev->qdev) < 0)
         return NULL;
 
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Christoph Hellwig]

On Fri, Jun 25, 2010 at 06:53:28PM +0200, Markus Armbruster wrote:
> For instance, -device scsi-disk,drive=foo -device scsi-disk,drive=foo
> happily creates two SCSI disks connected to the same block device.
> It's all downhill from there.

And from some quick testing a while ago the thing seems to actually
work.  Not that I think that it is a good idea, but do we want to change
behaviour in that respect?

> Device usb-storage deliberately attaches twice to the same blockdev,
> which fails with the fix in place.  Detach before the second attach
> there.

Can anyone explain what the hell usb storage is actually trying to do
with the two drives?

Re: [Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Markus Armbruster]

Christoph Hellwig <hch@lst.de> writes:

> On Fri, Jun 25, 2010 at 06:53:28PM +0200, Markus Armbruster wrote:
>> For instance, -device scsi-disk,drive=foo -device scsi-disk,drive=foo
>> happily creates two SCSI disks connected to the same block device.
>> It's all downhill from there.
>
> And from some quick testing a while ago the thing seems to actually
> work.  Not that I think that it is a good idea, but do we want to change
> behaviour in that respect?

Valid question.  I'd answer yes.  It's an easy error to make, and likely
to end in massive file system corruption in the guest.

>> Device usb-storage deliberately attaches twice to the same blockdev,
>> which fails with the fix in place.  Detach before the second attach
>> there.
>
> Can anyone explain what the hell usb storage is actually trying to do
> with the two drives?

It's actually a SCSI controller with a single drive on its single bus.

-device usb-storage,drive=foo creates *two* devices: usb-storage itself,
which serves as SCSI controller, and scsi-disk for the drive.
usb-storage copies its drive property to scsi-disk.

I don't like this.  Each -device should create just one device.

Re: [Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Christoph Hellwig]

On Sat, Jun 26, 2010 at 04:44:11PM +0200, Markus Armbruster wrote:
> Valid question.  I'd answer yes.  It's an easy error to make, and likely
> to end in massive file system corruption in the guest.

I suspect a modern distro in the guest will detect it as a multi-path setup.

> > Can anyone explain what the hell usb storage is actually trying to do
> > with the two drives?
> 
> It's actually a SCSI controller with a single drive on its single bus.
> 
> -device usb-storage,drive=foo creates *two* devices: usb-storage itself,
> which serves as SCSI controller, and scsi-disk for the drive.
> usb-storage copies its drive property to scsi-disk.
> 
> I don't like this.  Each -device should create just one device.

Indeed.  I'd also prefer to get rid of this.  Anthony, how hard are the
rules on backwards compatiblity for things like this?

Note that currently the usb storage emulation is extremly broken anyway,
just writing to it produces I/O errors after a short while.  This means
it can't be used very much at all.

Re: [Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Kevin Wolf]

Am 27.06.2010 11:36, schrieb Christoph Hellwig:
> On Sat, Jun 26, 2010 at 04:44:11PM +0200, Markus Armbruster wrote:
>> Valid question.  I'd answer yes.  It's an easy error to make, and likely
>> to end in massive file system corruption in the guest.
> 
> I suspect a modern distro in the guest will detect it as a multi-path setup.
> 
>>> Can anyone explain what the hell usb storage is actually trying to do
>>> with the two drives?
>>
>> It's actually a SCSI controller with a single drive on its single bus.
>>
>> -device usb-storage,drive=foo creates *two* devices: usb-storage itself,
>> which serves as SCSI controller, and scsi-disk for the drive.
>> usb-storage copies its drive property to scsi-disk.
>>
>> I don't like this.  Each -device should create just one device.
> 
> Indeed.  I'd also prefer to get rid of this.  Anthony, how hard are the
> rules on backwards compatiblity for things like this?

How would breaking compatibility help us? For the user a USB MSD is only
one device, so requiring two -device parameters sounds wrong.

If anything, scsi-disk must change to be able to share code without
requiring a device - and this is a change that would be kept internal
with no change on the user interface.

Or are you just talking about things like migration between versions
which would very likely break? That would probably be okay for USB MSD.

> Note that currently the usb storage emulation is extremly broken anyway,
> just writing to it produces I/O errors after a short while.  This means
> it can't be used very much at all.

When I tried last time, it did produce lots of kernel error messages in
the guest, but in the end the data was written correctly. So it doesn't
seem to be completely unusable.

Kevin

Re: [Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Christoph Hellwig]

On Mon, Jun 28, 2010 at 10:24:49AM +0200, Kevin Wolf wrote:
> How would breaking compatibility help us? For the user a USB MSD is only
> one device, so requiring two -device parameters sounds wrong.

But it is separate devices.  At least the standards compliant usb
storage devices just are a bride of scsi commands over usb and fit into
the SAM device model, which makes a difference between initiator, target
and LUN.  So having a different device for the specific target vs the
initiator port makes a difference. (and yes, we're still totally missing
support for multiple luns, which would require another level of
devices).  Trying to hide this is not all that useful - not anymore
useful than hiding it on a "normal" scsi host controller anyway.

Re: [Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Kevin Wolf]

Am 28.06.2010 12:16, schrieb Christoph Hellwig:
> On Mon, Jun 28, 2010 at 10:24:49AM +0200, Kevin Wolf wrote:
>> How would breaking compatibility help us? For the user a USB MSD is only
>> one device, so requiring two -device parameters sounds wrong.
> 
> But it is separate devices.  At least the standards compliant usb
> storage devices just are a bride of scsi commands over usb and fit into
> the SAM device model, which makes a difference between initiator, target
> and LUN.  So having a different device for the specific target vs the
> initiator port makes a difference. (and yes, we're still totally missing
> support for multiple luns, which would require another level of
> devices).  Trying to hide this is not all that useful - not anymore
> useful than hiding it on a "normal" scsi host controller anyway.

Maybe we need something like composed devices? So when the user asks for
a USB stick, he actually gets all devices that this stick internally
uses? Otherwise it becomes really hard to use -device directly.

I guess the same applies for mainboards, CPUs and probably some more
things, though I don't really know how these are (planned to be) done in
qdev.

Kevin

Re: [Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Markus Armbruster]

Kevin Wolf <kwolf@redhat.com> writes:

> Am 28.06.2010 12:16, schrieb Christoph Hellwig:
>> On Mon, Jun 28, 2010 at 10:24:49AM +0200, Kevin Wolf wrote:
>>
>>> Am 27.06.2010 11:36, schrieb Christoph Hellwig:
>>>> On Sat, Jun 26, 2010 at 04:44:11PM +0200, Markus Armbruster wrote:
[...]
>>>>> -device usb-storage,drive=foo creates *two* devices: usb-storage itself,
>>>>> which serves as SCSI controller, and scsi-disk for the drive.
>>>>> usb-storage copies its drive property to scsi-disk.
>>>>>
>>>>> I don't like this.  Each -device should create just one device.
>>>> 
>>>> Indeed.  I'd also prefer to get rid of this.  Anthony, how hard are the
>>>> rules on backwards compatiblity for things like this?
>>>
>>> How would breaking compatibility help us? For the user a USB MSD is only
>>> one device, so requiring two -device parameters sounds wrong.

-device designed to be simple, stupid and straightforward: you get
exactly what you asked for, no more, no less.  usb-storage breaks this
design maxim.

>> But it is separate devices.  At least the standards compliant usb
>> storage devices just are a bride of scsi commands over usb and fit into
>> the SAM device model, which makes a difference between initiator, target
>> and LUN.  So having a different device for the specific target vs the
>> initiator port makes a difference. (and yes, we're still totally missing
>> support for multiple luns, which would require another level of
>> devices).  Trying to hide this is not all that useful - not anymore
>> useful than hiding it on a "normal" scsi host controller anyway.
>
> Maybe we need something like composed devices? So when the user asks for
> a USB stick, he actually gets all devices that this stick internally
> uses? Otherwise it becomes really hard to use -device directly.

Could be useful.

> I guess the same applies for mainboards, CPUs and probably some more
> things, though I don't really know how these are (planned to be) done in
> qdev.

I'd like to keep -device stupid.  If we need "smarter" controls, let's
layer them on top.

Re: [Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Kevin Wolf]

Am 30.06.2010 13:52, schrieb Markus Armbruster:
> Kevin Wolf <kwolf@redhat.com> writes:
> 
>> Am 28.06.2010 12:16, schrieb Christoph Hellwig:
>>> On Mon, Jun 28, 2010 at 10:24:49AM +0200, Kevin Wolf wrote:
>>>
>>>> Am 27.06.2010 11:36, schrieb Christoph Hellwig:
>>>>> On Sat, Jun 26, 2010 at 04:44:11PM +0200, Markus Armbruster wrote:
> [...]
>>>>>> -device usb-storage,drive=foo creates *two* devices: usb-storage itself,
>>>>>> which serves as SCSI controller, and scsi-disk for the drive.
>>>>>> usb-storage copies its drive property to scsi-disk.
>>>>>>
>>>>>> I don't like this.  Each -device should create just one device.
>>>>>
>>>>> Indeed.  I'd also prefer to get rid of this.  Anthony, how hard are the
>>>>> rules on backwards compatiblity for things like this?
>>>>
>>>> How would breaking compatibility help us? For the user a USB MSD is only
>>>> one device, so requiring two -device parameters sounds wrong.
> 
> -device designed to be simple, stupid and straightforward: you get
> exactly what you asked for, no more, no less.  usb-storage breaks this
> design maxim.

I suppose the real question is: What is a device? qemu's internal view
(dozens of devices that communicate with each other) and the user's view
(it's one USB stick/mainboard/...) may differ.

>> Maybe we need something like composed devices? So when the user asks for
>> a USB stick, he actually gets all devices that this stick internally
>> uses? Otherwise it becomes really hard to use -device directly.
> 
> Could be useful.
> 
>> I guess the same applies for mainboards, CPUs and probably some more
>> things, though I don't really know how these are (planned to be) done in
>> qdev.
> 
> I'd like to keep -device stupid.  If we need "smarter" controls, let's
> layer them on top.

Makes sense. I'm not opposed to having something that deals only with
"atomic devices" or whatever you want to call them.

What users will usually want to have is something that treats an USB
stick as one device, though. So maybe the stupid version should become
-qdev or something and the extended version that is meant for users
should get the easy-to-remember name -device.

Kevin

Re: [Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Gerd Hoffmann]

   Hi,

>> Note that currently the usb storage emulation is extremly broken anyway,
>> just writing to it produces I/O errors after a short while.  This means
>> it can't be used very much at all.
>
> When I tried last time, it did produce lots of kernel error messages in
> the guest, but in the end the data was written correctly. So it doesn't
> seem to be completely unusable.

I think it is actually the linux kernel being at fault here.  I've seen 
I/O errors on real hardware too, on my pretty old laptop which has usb 
1.1 only.  Probably something like timeouts being *way* too low for usb 
1.1 xfer speeds.

cheers,
   Gerd

Re: [Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Markus Armbruster]

Christoph Hellwig <hch@lst.de> writes:

>Markus Armbruster <armbru@redhat.com> writes:
>
>> Christoph Hellwig <hch@lst.de> writes:
>>
>>> On Fri, Jun 25, 2010 at 06:53:28PM +0200, Markus Armbruster wrote:
>>>> For instance, -device scsi-disk,drive=foo -device scsi-disk,drive=foo
>>>> happily creates two SCSI disks connected to the same block device.
>>>> It's all downhill from there.
>>>
>>> And from some quick testing a while ago the thing seems to actually
>>> work.  Not that I think that it is a good idea, but do we want to change
>>> behaviour in that respect?
>>
>> Valid question.  I'd answer yes.  It's an easy error to make, and likely
>> to end in massive file system corruption in the guest.
>
> I suspect a modern distro in the guest will detect it as a multi-path setup.

Really?  The guest sees two disks, different serial numbers, possibly on
different buses (one could be SCSI, the other iDE).

[...]

[Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Kevin Wolf]

Am 25.06.2010 18:53, schrieb Markus Armbruster:
> For instance, -device scsi-disk,drive=foo -device scsi-disk,drive=foo
> happily creates two SCSI disks connected to the same block device.
> It's all downhill from there.
> 
> Device usb-storage deliberately attaches twice to the same blockdev,
> which fails with the fix in place.  Detach before the second attach
> there.
> 
> Also catch attempt to delete while a guest device model is attached.
> 
> Signed-off-by: Markus Armbruster <armbru@redhat.com>
> ---
>  block.c              |   22 ++++++++++++++++++++++
>  block.h              |    3 +++
>  block_int.h          |    2 ++
>  hw/fdc.c             |   10 +++++-----
>  hw/ide/qdev.c        |    2 +-
>  hw/pci-hotplug.c     |    5 ++++-
>  hw/qdev-properties.c |   21 ++++++++++++++++++++-
>  hw/qdev.h            |    3 ++-
>  hw/s390-virtio.c     |    2 +-
>  hw/scsi-bus.c        |    4 +++-
>  hw/usb-msd.c         |   11 +++++++----
>  11 files changed, 70 insertions(+), 15 deletions(-)
> 
> diff --git a/block.c b/block.c
> index e71a771..5e0ffa0 100644
> --- a/block.c
> +++ b/block.c
> @@ -659,6 +659,8 @@ void bdrv_close_all(void)
>  
>  void bdrv_delete(BlockDriverState *bs)
>  {
> +    assert(!bs->peer);
> +
>      /* remove from list, if necessary */
>      if (bs->device_name[0] != '\0') {
>          QTAILQ_REMOVE(&bdrv_states, bs, list);
> @@ -672,6 +674,26 @@ void bdrv_delete(BlockDriverState *bs)
>      qemu_free(bs);
>  }
>  
> +int bdrv_attach(BlockDriverState *bs, DeviceState *qdev)
> +{
> +    if (bs->peer) {
> +        return -EBUSY;
> +    }
> +    bs->peer = qdev;
> +    return 0;
> +}
> +
> +void bdrv_detach(BlockDriverState *bs, DeviceState *qdev)
> +{
> +    assert(bs->peer == qdev);
> +    bs->peer = NULL;
> +}
> +
> +DeviceState *bdrv_get_attached(BlockDriverState *bs)
> +{
> +    return bs->peer;
> +}
> +
>  /*
>   * Run consistency checks on an image
>   *
> diff --git a/block.h b/block.h
> index 6a157f4..88ac06e 100644
> --- a/block.h
> +++ b/block.h
> @@ -71,6 +71,9 @@ int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags);
>  int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
>                BlockDriver *drv);
>  void bdrv_close(BlockDriverState *bs);
> +int bdrv_attach(BlockDriverState *bs, DeviceState *qdev);
> +void bdrv_detach(BlockDriverState *bs, DeviceState *qdev);
> +DeviceState *bdrv_get_attached(BlockDriverState *bs);
>  int bdrv_check(BlockDriverState *bs);
>  int bdrv_read(BlockDriverState *bs, int64_t sector_num,
>                uint8_t *buf, int nb_sectors);
> diff --git a/block_int.h b/block_int.h
> index e60aed4..a94b801 100644
> --- a/block_int.h
> +++ b/block_int.h
> @@ -148,6 +148,8 @@ struct BlockDriverState {
>      BlockDriver *drv; /* NULL means no media */
>      void *opaque;
>  
> +    DeviceState *peer;
> +
>      char filename[1024];
>      char backing_file[1024]; /* if non zero, the image is a diff of
>                                  this file image */
> diff --git a/hw/fdc.c b/hw/fdc.c
> index 08712bc..1496cfa 100644
> --- a/hw/fdc.c
> +++ b/hw/fdc.c
> @@ -1860,10 +1860,10 @@ FDCtrl *fdctrl_init_isa(DriveInfo **fds)
>  
>      dev = isa_create("isa-fdc");
>      if (fds[0]) {
> -        qdev_prop_set_drive(&dev->qdev, "driveA", fds[0]->bdrv);
> +        qdev_prop_set_drive_nofail(&dev->qdev, "driveA", fds[0]->bdrv);
>      }
>      if (fds[1]) {
> -        qdev_prop_set_drive(&dev->qdev, "driveB", fds[1]->bdrv);
> +        qdev_prop_set_drive_nofail(&dev->qdev, "driveB", fds[1]->bdrv);
>      }
>      if (qdev_init(&dev->qdev) < 0)
>          return NULL;
> @@ -1882,10 +1882,10 @@ FDCtrl *fdctrl_init_sysbus(qemu_irq irq, int dma_chann,
>      fdctrl = &sys->state;
>      fdctrl->dma_chann = dma_chann; /* FIXME */
>      if (fds[0]) {
> -        qdev_prop_set_drive(dev, "driveA", fds[0]->bdrv);
> +        qdev_prop_set_drive_nofail(dev, "driveA", fds[0]->bdrv);
>      }
>      if (fds[1]) {
> -        qdev_prop_set_drive(dev, "driveB", fds[1]->bdrv);
> +        qdev_prop_set_drive_nofail(dev, "driveB", fds[1]->bdrv);
>      }
>      qdev_init_nofail(dev);
>      sysbus_connect_irq(&sys->busdev, 0, irq);
> @@ -1903,7 +1903,7 @@ FDCtrl *sun4m_fdctrl_init(qemu_irq irq, target_phys_addr_t io_base,
>  
>      dev = qdev_create(NULL, "SUNW,fdtwo");
>      if (fds[0]) {
> -        qdev_prop_set_drive(dev, "drive", fds[0]->bdrv);
> +        qdev_prop_set_drive_nofail(dev, "drive", fds[0]->bdrv);
>      }
>      qdev_init_nofail(dev);
>      sys = DO_UPCAST(FDCtrlSysBus, busdev.qdev, dev);
> diff --git a/hw/ide/qdev.c b/hw/ide/qdev.c
> index 3bb94c6..b4bc5ac 100644
> --- a/hw/ide/qdev.c
> +++ b/hw/ide/qdev.c
> @@ -83,7 +83,7 @@ IDEDevice *ide_create_drive(IDEBus *bus, int unit, DriveInfo *drive)
>  
>      dev = qdev_create(&bus->qbus, "ide-drive");
>      qdev_prop_set_uint32(dev, "unit", unit);
> -    qdev_prop_set_drive(dev, "drive", drive->bdrv);
> +    qdev_prop_set_drive_nofail(dev, "drive", drive->bdrv);
>      qdev_init_nofail(dev);
>      return DO_UPCAST(IDEDevice, qdev, dev);
>  }
> diff --git a/hw/pci-hotplug.c b/hw/pci-hotplug.c
> index d743192..b47e01e 100644
> --- a/hw/pci-hotplug.c
> +++ b/hw/pci-hotplug.c
> @@ -214,7 +214,10 @@ static PCIDevice *qemu_pci_hot_add_storage(Monitor *mon,
>              return NULL;
>          }
>          dev = pci_create(bus, devfn, "virtio-blk-pci");
> -        qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv);
> +        if (qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv) < 0) {
> +            dev = NULL;
> +            break;
> +        }

I think in error cases we'll leak dev.

>          if (qdev_init(&dev->qdev) < 0)
>              dev = NULL;
>          break;
> diff --git a/hw/qdev-properties.c b/hw/qdev-properties.c
> index 257233e..caf6798 100644
> --- a/hw/qdev-properties.c
> +++ b/hw/qdev-properties.c
> @@ -291,6 +291,8 @@ static int parse_drive(DeviceState *dev, Property *prop, const char *str)
>      bs = bdrv_find(str);
>      if (bs == NULL)
>          return -ENOENT;
> +    if (bdrv_attach(bs, dev) < 0)
> +        return -EEXIST;
>      *ptr = bs;
>      return 0;
>  }
> @@ -300,6 +302,7 @@ static void free_drive(DeviceState *dev, Property *prop)
>      BlockDriverState **ptr = qdev_get_prop_ptr(dev, prop);
>  
>      if (*ptr) {
> +        bdrv_detach(*ptr, dev);
>          blockdev_auto_del(*ptr);
>      }
>  }
> @@ -640,11 +643,27 @@ void qdev_prop_set_string(DeviceState *dev, const char *name, char *value)
>      qdev_prop_set(dev, name, &value, PROP_TYPE_STRING);
>  }
>  
> -void qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value)
> +int qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value)
>  {
> +    int res;
> +
>      qdev_prop_set(dev, name, &value, PROP_TYPE_DRIVE);
> +    res = bdrv_attach(value, dev);
> +    if (res < 0) {
> +        error_report("Can't attach drive %s to %s.%s: %s",
> +                     bdrv_get_device_name(value),
> +                     dev->id ? dev->id : dev->info->name,
> +                     name, strerror(-res));
> +    }
> +    return res;
>  }
>  
> +void qdev_prop_set_drive_nofail(DeviceState *dev, const char *name, BlockDriverState *value)
> +{
> +    if (qdev_prop_set_drive(dev, name, value) < 0) {
> +        exit(1);
> +    }
> +}
>  void qdev_prop_set_chr(DeviceState *dev, const char *name, CharDriverState *value)
>  {
>      qdev_prop_set(dev, name, &value, PROP_TYPE_CHR);
> diff --git a/hw/qdev.h b/hw/qdev.h
> index 7a01a81..cbc89f2 100644
> --- a/hw/qdev.h
> +++ b/hw/qdev.h
> @@ -275,7 +275,8 @@ void qdev_prop_set_string(DeviceState *dev, const char *name, char *value);
>  void qdev_prop_set_chr(DeviceState *dev, const char *name, CharDriverState *value);
>  void qdev_prop_set_netdev(DeviceState *dev, const char *name, VLANClientState *value);
>  void qdev_prop_set_vlan(DeviceState *dev, const char *name, VLANState *value);
> -void qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value);
> +int qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value) QEMU_WARN_UNUSED_RESULT;
> +void qdev_prop_set_drive_nofail(DeviceState *dev, const char *name, BlockDriverState *value);
>  void qdev_prop_set_macaddr(DeviceState *dev, const char *name, uint8_t *value);
>  /* FIXME: Remove opaque pointer properties.  */
>  void qdev_prop_set_ptr(DeviceState *dev, const char *name, void *value);
> diff --git a/hw/s390-virtio.c b/hw/s390-virtio.c
> index c7c3fc9..6af58e2 100644
> --- a/hw/s390-virtio.c
> +++ b/hw/s390-virtio.c
> @@ -262,7 +262,7 @@ static void s390_init(ram_addr_t ram_size,
>          }
>  
>          dev = qdev_create((BusState *)s390_bus, "virtio-blk-s390");
> -        qdev_prop_set_drive(dev, "drive", dinfo->bdrv);
> +        qdev_prop_set_drive_nofail(dev, "drive", dinfo->bdrv);
>          qdev_init_nofail(dev);
>      }
>  }
> diff --git a/hw/scsi-bus.c b/hw/scsi-bus.c
> index 2c58aca..9678328 100644
> --- a/hw/scsi-bus.c
> +++ b/hw/scsi-bus.c
> @@ -91,7 +91,9 @@ SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, BlockDriverState *bdrv, int
>      driver = bdrv_is_sg(bdrv) ? "scsi-generic" : "scsi-disk";
>      dev = qdev_create(&bus->qbus, driver);
>      qdev_prop_set_uint32(dev, "scsi-id", unit);
> -    qdev_prop_set_drive(dev, "drive", bdrv);
> +    if (qdev_prop_set_drive(dev, "drive", bdrv) < 0) {
> +        return NULL;
> +    }

As we do here.

>      if (qdev_init(dev) < 0)
>          return NULL;
>      return DO_UPCAST(SCSIDevice, qdev, dev);
> diff --git a/hw/usb-msd.c b/hw/usb-msd.c
> index 19a14b4..008cc0f 100644
> --- a/hw/usb-msd.c
> +++ b/hw/usb-msd.c
> @@ -532,12 +532,13 @@ static int usb_msd_initfn(USBDevice *dev)
>      /*
>       * Hack alert: this pretends to be a block device, but it's really
>       * a SCSI bus that can serve only a single device, which it
> -     * creates automatically.  Two drive properties pointing to the
> -     * same drive is not good: free_drive() dies for the second one.
> -     * Zap the one we're not going to use.
> +     * creates automatically.  But first it needs to detach from its
> +     * blockdev, or else scsi_bus_legacy_add_drive() dies when it
> +     * attaches again.
>       *
>       * The hack is probably a bad idea.
>       */
> +    bdrv_detach(bs, &s->dev.qdev);
>      s->conf.bs = NULL;
>  
>      s->dev.speed = USB_SPEED_FULL;
> @@ -609,7 +610,9 @@ static USBDevice *usb_msd_init(const char *filename)
>      if (!dev) {
>          return NULL;
>      }
> -    qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv);
> +    if (qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv) < 0) {
> +        return NULL;
> +    }
>      if (qdev_init(&dev->qdev) < 0)
>          return NULL;

And here.

Kevin

[Qemu-devel] Re: [PATCH 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Markus Armbruster]

Kevin Wolf <kwolf@redhat.com> writes:

> Am 25.06.2010 18:53, schrieb Markus Armbruster:
>> For instance, -device scsi-disk,drive=foo -device scsi-disk,drive=foo
>> happily creates two SCSI disks connected to the same block device.
>> It's all downhill from there.
>> 
>> Device usb-storage deliberately attaches twice to the same blockdev,
>> which fails with the fix in place.  Detach before the second attach
>> there.
>> 
>> Also catch attempt to delete while a guest device model is attached.
[...]
>> diff --git a/hw/pci-hotplug.c b/hw/pci-hotplug.c
>> index d743192..b47e01e 100644
>> --- a/hw/pci-hotplug.c
>> +++ b/hw/pci-hotplug.c
>> @@ -214,7 +214,10 @@ static PCIDevice *qemu_pci_hot_add_storage(Monitor *mon,
>>              return NULL;
>>          }
>>          dev = pci_create(bus, devfn, "virtio-blk-pci");
>> -        qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv);
>> +        if (qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv) < 0) {
>> +            dev = NULL;
>> +            break;
>> +        }
>
> I think in error cases we'll leak dev.

Correct.

>>          if (qdev_init(&dev->qdev) < 0)
>>              dev = NULL;
>>          break;
[...]
>> diff --git a/hw/scsi-bus.c b/hw/scsi-bus.c
>> index 2c58aca..9678328 100644
>> --- a/hw/scsi-bus.c
>> +++ b/hw/scsi-bus.c
>> @@ -91,7 +91,9 @@ SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, BlockDriverState *bdrv, int
>>      driver = bdrv_is_sg(bdrv) ? "scsi-generic" : "scsi-disk";
>>      dev = qdev_create(&bus->qbus, driver);
>>      qdev_prop_set_uint32(dev, "scsi-id", unit);
>> -    qdev_prop_set_drive(dev, "drive", bdrv);
>> +    if (qdev_prop_set_drive(dev, "drive", bdrv) < 0) {
>> +        return NULL;
>> +    }
>
> As we do here.

Yes.

>>      if (qdev_init(dev) < 0)
>>          return NULL;
>>      return DO_UPCAST(SCSIDevice, qdev, dev);
>> diff --git a/hw/usb-msd.c b/hw/usb-msd.c
>> index 19a14b4..008cc0f 100644
>> --- a/hw/usb-msd.c
>> +++ b/hw/usb-msd.c
>> @@ -532,12 +532,13 @@ static int usb_msd_initfn(USBDevice *dev)
>>      /*
>>       * Hack alert: this pretends to be a block device, but it's really
>>       * a SCSI bus that can serve only a single device, which it
>> -     * creates automatically.  Two drive properties pointing to the
>> -     * same drive is not good: free_drive() dies for the second one.
>> -     * Zap the one we're not going to use.
>> +     * creates automatically.  But first it needs to detach from its
>> +     * blockdev, or else scsi_bus_legacy_add_drive() dies when it
>> +     * attaches again.
>>       *
>>       * The hack is probably a bad idea.
>>       */
>> +    bdrv_detach(bs, &s->dev.qdev);
>>      s->conf.bs = NULL;
>>  
>>      s->dev.speed = USB_SPEED_FULL;
>> @@ -609,7 +610,9 @@ static USBDevice *usb_msd_init(const char *filename)
>>      if (!dev) {
>>          return NULL;
>>      }
>> -    qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv);
>> +    if (qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv) < 0) {
>> +        return NULL;
>> +    }
>>      if (qdev_init(&dev->qdev) < 0)
>>          return NULL;
>
> And here.

Yes.

I double-checked the entire patch series, and could not find further
leaks.

Thanks a lot!

[Qemu-devel] [PATCH v2 08/12] block: Catch attempt to attach multiple devices to a blockdev

[Markus Armbruster]

For instance, -device scsi-disk,drive=foo -device scsi-disk,drive=foo
happily creates two SCSI disks connected to the same block device.
It's all downhill from there.

Device usb-storage deliberately attaches twice to the same blockdev,
which fails with the fix in place.  Detach before the second attach
there.

Also catch attempt to delete while a guest device model is attached.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
v2: plug leaks pointed out by Kevin Wolf
    fix qdev_prop_set_drive() to set the property only when attach
    succeeded (free_drive() will die without this)

Also available at git://repo.or.cz/qemu/armbru.git
Tag block-fixes-v2

 block.c              |   22 ++++++++++++++++++++++
 block.h              |    3 +++
 block_int.h          |    2 ++
 hw/fdc.c             |   10 +++++-----
 hw/ide/qdev.c        |    2 +-
 hw/pci-hotplug.c     |    6 +++++-
 hw/qdev-properties.c |   22 +++++++++++++++++++++-
 hw/qdev.h            |    3 ++-
 hw/s390-virtio.c     |    2 +-
 hw/scsi-bus.c        |    5 ++++-
 hw/usb-msd.c         |   12 ++++++++----
 11 files changed, 74 insertions(+), 15 deletions(-)

diff --git a/block.c b/block.c
index e71a771..5e0ffa0 100644
--- a/block.c
+++ b/block.c
@@ -659,6 +659,8 @@ void bdrv_close_all(void)
 
 void bdrv_delete(BlockDriverState *bs)
 {
+    assert(!bs->peer);
+
     /* remove from list, if necessary */
     if (bs->device_name[0] != '\0') {
         QTAILQ_REMOVE(&bdrv_states, bs, list);
@@ -672,6 +674,26 @@ void bdrv_delete(BlockDriverState *bs)
     qemu_free(bs);
 }
 
+int bdrv_attach(BlockDriverState *bs, DeviceState *qdev)
+{
+    if (bs->peer) {
+        return -EBUSY;
+    }
+    bs->peer = qdev;
+    return 0;
+}
+
+void bdrv_detach(BlockDriverState *bs, DeviceState *qdev)
+{
+    assert(bs->peer == qdev);
+    bs->peer = NULL;
+}
+
+DeviceState *bdrv_get_attached(BlockDriverState *bs)
+{
+    return bs->peer;
+}
+
 /*
  * Run consistency checks on an image
  *
diff --git a/block.h b/block.h
index 6a157f4..88ac06e 100644
--- a/block.h
+++ b/block.h
@@ -71,6 +71,9 @@ int bdrv_file_open(BlockDriverState **pbs, const char *filename, int flags);
 int bdrv_open(BlockDriverState *bs, const char *filename, int flags,
               BlockDriver *drv);
 void bdrv_close(BlockDriverState *bs);
+int bdrv_attach(BlockDriverState *bs, DeviceState *qdev);
+void bdrv_detach(BlockDriverState *bs, DeviceState *qdev);
+DeviceState *bdrv_get_attached(BlockDriverState *bs);
 int bdrv_check(BlockDriverState *bs);
 int bdrv_read(BlockDriverState *bs, int64_t sector_num,
               uint8_t *buf, int nb_sectors);
diff --git a/block_int.h b/block_int.h
index e60aed4..a94b801 100644
--- a/block_int.h
+++ b/block_int.h
@@ -148,6 +148,8 @@ struct BlockDriverState {
     BlockDriver *drv; /* NULL means no media */
     void *opaque;
 
+    DeviceState *peer;
+
     char filename[1024];
     char backing_file[1024]; /* if non zero, the image is a diff of
                                 this file image */
diff --git a/hw/fdc.c b/hw/fdc.c
index 08712bc..1496cfa 100644
--- a/hw/fdc.c
+++ b/hw/fdc.c
@@ -1860,10 +1860,10 @@ FDCtrl *fdctrl_init_isa(DriveInfo **fds)
 
     dev = isa_create("isa-fdc");
     if (fds[0]) {
-        qdev_prop_set_drive(&dev->qdev, "driveA", fds[0]->bdrv);
+        qdev_prop_set_drive_nofail(&dev->qdev, "driveA", fds[0]->bdrv);
     }
     if (fds[1]) {
-        qdev_prop_set_drive(&dev->qdev, "driveB", fds[1]->bdrv);
+        qdev_prop_set_drive_nofail(&dev->qdev, "driveB", fds[1]->bdrv);
     }
     if (qdev_init(&dev->qdev) < 0)
         return NULL;
@@ -1882,10 +1882,10 @@ FDCtrl *fdctrl_init_sysbus(qemu_irq irq, int dma_chann,
     fdctrl = &sys->state;
     fdctrl->dma_chann = dma_chann; /* FIXME */
     if (fds[0]) {
-        qdev_prop_set_drive(dev, "driveA", fds[0]->bdrv);
+        qdev_prop_set_drive_nofail(dev, "driveA", fds[0]->bdrv);
     }
     if (fds[1]) {
-        qdev_prop_set_drive(dev, "driveB", fds[1]->bdrv);
+        qdev_prop_set_drive_nofail(dev, "driveB", fds[1]->bdrv);
     }
     qdev_init_nofail(dev);
     sysbus_connect_irq(&sys->busdev, 0, irq);
@@ -1903,7 +1903,7 @@ FDCtrl *sun4m_fdctrl_init(qemu_irq irq, target_phys_addr_t io_base,
 
     dev = qdev_create(NULL, "SUNW,fdtwo");
     if (fds[0]) {
-        qdev_prop_set_drive(dev, "drive", fds[0]->bdrv);
+        qdev_prop_set_drive_nofail(dev, "drive", fds[0]->bdrv);
     }
     qdev_init_nofail(dev);
     sys = DO_UPCAST(FDCtrlSysBus, busdev.qdev, dev);
diff --git a/hw/ide/qdev.c b/hw/ide/qdev.c
index a5fdab0..b34c473 100644
--- a/hw/ide/qdev.c
+++ b/hw/ide/qdev.c
@@ -83,7 +83,7 @@ IDEDevice *ide_create_drive(IDEBus *bus, int unit, DriveInfo *drive)
 
     dev = qdev_create(&bus->qbus, "ide-drive");
     qdev_prop_set_uint32(dev, "unit", unit);
-    qdev_prop_set_drive(dev, "drive", drive->bdrv);
+    qdev_prop_set_drive_nofail(dev, "drive", drive->bdrv);
     qdev_init_nofail(dev);
     return DO_UPCAST(IDEDevice, qdev, dev);
 }
diff --git a/hw/pci-hotplug.c b/hw/pci-hotplug.c
index d743192..fe468d6 100644
--- a/hw/pci-hotplug.c
+++ b/hw/pci-hotplug.c
@@ -214,7 +214,11 @@ static PCIDevice *qemu_pci_hot_add_storage(Monitor *mon,
             return NULL;
         }
         dev = pci_create(bus, devfn, "virtio-blk-pci");
-        qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv);
+        if (qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv) < 0) {
+            qdev_free(&dev->qdev);
+            dev = NULL;
+            break;
+        }
         if (qdev_init(&dev->qdev) < 0)
             dev = NULL;
         break;
diff --git a/hw/qdev-properties.c b/hw/qdev-properties.c
index 257233e..ce6e002 100644
--- a/hw/qdev-properties.c
+++ b/hw/qdev-properties.c
@@ -291,6 +291,8 @@ static int parse_drive(DeviceState *dev, Property *prop, const char *str)
     bs = bdrv_find(str);
     if (bs == NULL)
         return -ENOENT;
+    if (bdrv_attach(bs, dev) < 0)
+        return -EEXIST;
     *ptr = bs;
     return 0;
 }
@@ -300,6 +302,7 @@ static void free_drive(DeviceState *dev, Property *prop)
     BlockDriverState **ptr = qdev_get_prop_ptr(dev, prop);
 
     if (*ptr) {
+        bdrv_detach(*ptr, dev);
         blockdev_auto_del(*ptr);
     }
 }
@@ -640,11 +643,28 @@ void qdev_prop_set_string(DeviceState *dev, const char *name, char *value)
     qdev_prop_set(dev, name, &value, PROP_TYPE_STRING);
 }
 
-void qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value)
+int qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value)
 {
+    int res;
+
+    res = bdrv_attach(value, dev);
+    if (res < 0) {
+        error_report("Can't attach drive %s to %s.%s: %s",
+                     bdrv_get_device_name(value),
+                     dev->id ? dev->id : dev->info->name,
+                     name, strerror(-res));
+        return -1;
+    }
     qdev_prop_set(dev, name, &value, PROP_TYPE_DRIVE);
+    return 0;
 }
 
+void qdev_prop_set_drive_nofail(DeviceState *dev, const char *name, BlockDriverState *value)
+{
+    if (qdev_prop_set_drive(dev, name, value) < 0) {
+        exit(1);
+    }
+}
 void qdev_prop_set_chr(DeviceState *dev, const char *name, CharDriverState *value)
 {
     qdev_prop_set(dev, name, &value, PROP_TYPE_CHR);
diff --git a/hw/qdev.h b/hw/qdev.h
index 7a01a81..cbc89f2 100644
--- a/hw/qdev.h
+++ b/hw/qdev.h
@@ -275,7 +275,8 @@ void qdev_prop_set_string(DeviceState *dev, const char *name, char *value);
 void qdev_prop_set_chr(DeviceState *dev, const char *name, CharDriverState *value);
 void qdev_prop_set_netdev(DeviceState *dev, const char *name, VLANClientState *value);
 void qdev_prop_set_vlan(DeviceState *dev, const char *name, VLANState *value);
-void qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value);
+int qdev_prop_set_drive(DeviceState *dev, const char *name, BlockDriverState *value) QEMU_WARN_UNUSED_RESULT;
+void qdev_prop_set_drive_nofail(DeviceState *dev, const char *name, BlockDriverState *value);
 void qdev_prop_set_macaddr(DeviceState *dev, const char *name, uint8_t *value);
 /* FIXME: Remove opaque pointer properties.  */
 void qdev_prop_set_ptr(DeviceState *dev, const char *name, void *value);
diff --git a/hw/s390-virtio.c b/hw/s390-virtio.c
index c7c3fc9..6af58e2 100644
--- a/hw/s390-virtio.c
+++ b/hw/s390-virtio.c
@@ -262,7 +262,7 @@ static void s390_init(ram_addr_t ram_size,
         }
 
         dev = qdev_create((BusState *)s390_bus, "virtio-blk-s390");
-        qdev_prop_set_drive(dev, "drive", dinfo->bdrv);
+        qdev_prop_set_drive_nofail(dev, "drive", dinfo->bdrv);
         qdev_init_nofail(dev);
     }
 }
diff --git a/hw/scsi-bus.c b/hw/scsi-bus.c
index 2c58aca..b84b9b9 100644
--- a/hw/scsi-bus.c
+++ b/hw/scsi-bus.c
@@ -91,7 +91,10 @@ SCSIDevice *scsi_bus_legacy_add_drive(SCSIBus *bus, BlockDriverState *bdrv, int
     driver = bdrv_is_sg(bdrv) ? "scsi-generic" : "scsi-disk";
     dev = qdev_create(&bus->qbus, driver);
     qdev_prop_set_uint32(dev, "scsi-id", unit);
-    qdev_prop_set_drive(dev, "drive", bdrv);
+    if (qdev_prop_set_drive(dev, "drive", bdrv) < 0) {
+        qdev_free(dev);
+        return NULL;
+    }
     if (qdev_init(dev) < 0)
         return NULL;
     return DO_UPCAST(SCSIDevice, qdev, dev);
diff --git a/hw/usb-msd.c b/hw/usb-msd.c
index 19a14b4..65e9624 100644
--- a/hw/usb-msd.c
+++ b/hw/usb-msd.c
@@ -532,12 +532,13 @@ static int usb_msd_initfn(USBDevice *dev)
     /*
      * Hack alert: this pretends to be a block device, but it's really
      * a SCSI bus that can serve only a single device, which it
-     * creates automatically.  Two drive properties pointing to the
-     * same drive is not good: free_drive() dies for the second one.
-     * Zap the one we're not going to use.
+     * creates automatically.  But first it needs to detach from its
+     * blockdev, or else scsi_bus_legacy_add_drive() dies when it
+     * attaches again.
      *
      * The hack is probably a bad idea.
      */
+    bdrv_detach(bs, &s->dev.qdev);
     s->conf.bs = NULL;
 
     s->dev.speed = USB_SPEED_FULL;
@@ -609,7 +610,10 @@ static USBDevice *usb_msd_init(const char *filename)
     if (!dev) {
         return NULL;
     }
-    qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv);
+    if (qdev_prop_set_drive(&dev->qdev, "drive", dinfo->bdrv) < 0) {
+        qdev_free(&dev->qdev);
+        return NULL;
+    }
     if (qdev_init(&dev->qdev) < 0)
         return NULL;
 
-- 
1.6.6.1

[Qemu-devel] [PATCH 09/12] savevm: Survive hot-unplug of snapshot device

[Markus Armbruster]

savevm.c keeps a pointer to the snapshot block device.  If you manage
to get that device deleted, the pointer dangles, and the next snapshot
operation will crash & burn.  Unplugging a guest device that uses it
does the trick:

    $ MALLOC_PERTURB_=234 qemu-system-x86_64 [...]
    QEMU 0.12.50 monitor - type 'help' for more information
    (qemu) info snapshots
    No available block device supports snapshots
    (qemu) drive_add auto if=none,file=tmp.qcow2
    OK
    (qemu) device_add usb-storage,id=foo,drive=none1
    (qemu) info snapshots
    Snapshot devices: none1
    Snapshot list (from none1):
    ID        TAG                 VM SIZE                DATE       VM CLOCK
    (qemu) device_del foo
    (qemu) info snapshots
    Snapshot devices:
    Segmentation fault (core dumped)

Move management of that pointer to block.c, and zap it when the device
it points to goes away.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 block.c  |   25 +++++++++++++++++++++++++
 block.h  |    1 +
 savevm.c |   31 ++++---------------------------
 3 files changed, 30 insertions(+), 27 deletions(-)

diff --git a/block.c b/block.c
index 5e0ffa0..34055e0 100644
--- a/block.c
+++ b/block.c
@@ -63,6 +63,9 @@ static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
 static QLIST_HEAD(, BlockDriver) bdrv_drivers =
     QLIST_HEAD_INITIALIZER(bdrv_drivers);
 
+/* The device to use for VM snapshots */
+static BlockDriverState *bs_snapshots;
+
 /* If non-zero, use only whitelisted block drivers */
 static int use_bdrv_whitelist;
 
@@ -660,6 +663,9 @@ void bdrv_close_all(void)
 void bdrv_delete(BlockDriverState *bs)
 {
     assert(!bs->peer);
+    if (bs == bs_snapshots) {
+        bs_snapshots = NULL;
+    }
 
     /* remove from list, if necessary */
     if (bs->device_name[0] != '\0') {
@@ -1772,6 +1778,25 @@ int bdrv_can_snapshot(BlockDriverState *bs)
     return 1;
 }
 
+BlockDriverState *bdrv_snapshots(void)
+{
+    BlockDriverState *bs;
+
+    if (bs_snapshots)
+        return bs_snapshots;
+
+    bs = NULL;
+    while ((bs = bdrv_next(bs))) {
+        if (bdrv_can_snapshot(bs)) {
+            goto ok;
+        }
+    }
+    return NULL;
+ ok:
+    bs_snapshots = bs;
+    return bs;
+}
+
 int bdrv_snapshot_create(BlockDriverState *bs,
                          QEMUSnapshotInfo *sn_info)
 {
diff --git a/block.h b/block.h
index 88ac06e..012c2a1 100644
--- a/block.h
+++ b/block.h
@@ -193,6 +193,7 @@ const char *bdrv_get_encrypted_filename(BlockDriverState *bs);
 void bdrv_get_backing_filename(BlockDriverState *bs,
                                char *filename, int filename_size);
 int bdrv_can_snapshot(BlockDriverState *bs);
+BlockDriverState *bdrv_snapshots(void);
 int bdrv_snapshot_create(BlockDriverState *bs,
                          QEMUSnapshotInfo *sn_info);
 int bdrv_snapshot_goto(BlockDriverState *bs,
diff --git a/savevm.c b/savevm.c
index 20354a8..f1f450e 100644
--- a/savevm.c
+++ b/savevm.c
@@ -83,9 +83,6 @@
 #include "qemu_socket.h"
 #include "qemu-queue.h"
 
-/* point to the block driver where the snapshots are managed */
-static BlockDriverState *bs_snapshots;
-
 #define SELF_ANNOUNCE_ROUNDS 5
 
 #ifndef ETH_P_RARP
@@ -1575,26 +1572,6 @@ out:
     return ret;
 }
 
-static BlockDriverState *get_bs_snapshots(void)
-{
-    BlockDriverState *bs;
-
-    if (bs_snapshots)
-        return bs_snapshots;
-    /* FIXME what if bs_snapshots gets hot-unplugged? */
-
-    bs = NULL;
-    while ((bs = bdrv_next(bs))) {
-        if (bdrv_can_snapshot(bs)) {
-            goto ok;
-        }
-    }
-    return NULL;
- ok:
-    bs_snapshots = bs;
-    return bs;
-}
-
 static int bdrv_snapshot_find(BlockDriverState *bs, QEMUSnapshotInfo *sn_info,
                               const char *name)
 {
@@ -1674,7 +1651,7 @@ void do_savevm(Monitor *mon, const QDict *qdict)
         }
     }
 
-    bs = get_bs_snapshots();
+    bs = bdrv_snapshots();
     if (!bs) {
         monitor_printf(mon, "No block device can accept snapshots\n");
         return;
@@ -1769,7 +1746,7 @@ int load_vmstate(const char *name)
         }
     }
 
-    bs = get_bs_snapshots();
+    bs = bdrv_snapshots();
     if (!bs) {
         error_report("No block device supports snapshots");
         return -EINVAL;
@@ -1833,7 +1810,7 @@ void do_delvm(Monitor *mon, const QDict *qdict)
     int ret;
     const char *name = qdict_get_str(qdict, "name");
 
-    bs = get_bs_snapshots();
+    bs = bdrv_snapshots();
     if (!bs) {
         monitor_printf(mon, "No block device supports snapshots\n");
         return;
@@ -1863,7 +1840,7 @@ void do_info_snapshots(Monitor *mon)
     int nb_sns, i;
     char buf[256];
 
-    bs = get_bs_snapshots();
+    bs = bdrv_snapshots();
     if (!bs) {
         monitor_printf(mon, "No available block device supports snapshots\n");
         return;
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 09/12] savevm: Survive hot-unplug of snapshot device

[Christoph Hellwig]

Looks good,

Reviewed-by: Christoph Hellwig <hch@lst.de>

Of course specifying an explicit medium for snapshot, be that the
snapshot section of a qcow2 image or just a separate flat file and
managing that one explicitly would be even better.

[Qemu-devel] Re: [PATCH 09/12] savevm: Survive hot-unplug of snapshot device

[Markus Armbruster]

Christoph Hellwig <hch@lst.de> writes:

> Looks good,
>
> Reviewed-by: Christoph Hellwig <hch@lst.de>
>
> Of course specifying an explicit medium for snapshot, be that the
> snapshot section of a qcow2 image or just a separate flat file and
> managing that one explicitly would be even better.

Indeed.

[Qemu-devel] Re: [PATCH 09/12] savevm: Survive hot-unplug of snapshot device

[Kevin Wolf]

Am 25.06.2010 18:53, schrieb Markus Armbruster:
> savevm.c keeps a pointer to the snapshot block device.  If you manage
> to get that device deleted, the pointer dangles, and the next snapshot
> operation will crash & burn.  Unplugging a guest device that uses it
> does the trick:
> 
>     $ MALLOC_PERTURB_=234 qemu-system-x86_64 [...]
>     QEMU 0.12.50 monitor - type 'help' for more information
>     (qemu) info snapshots
>     No available block device supports snapshots
>     (qemu) drive_add auto if=none,file=tmp.qcow2
>     OK
>     (qemu) device_add usb-storage,id=foo,drive=none1
>     (qemu) info snapshots
>     Snapshot devices: none1
>     Snapshot list (from none1):
>     ID        TAG                 VM SIZE                DATE       VM CLOCK
>     (qemu) device_del foo
>     (qemu) info snapshots
>     Snapshot devices:
>     Segmentation fault (core dumped)
> 
> Move management of that pointer to block.c, and zap it when the device
> it points to goes away.
> 
> Signed-off-by: Markus Armbruster <armbru@redhat.com>
> ---
>  block.c  |   25 +++++++++++++++++++++++++
>  block.h  |    1 +
>  savevm.c |   31 ++++---------------------------
>  3 files changed, 30 insertions(+), 27 deletions(-)
> 
> diff --git a/block.c b/block.c
> index 5e0ffa0..34055e0 100644
> --- a/block.c
> +++ b/block.c
> @@ -63,6 +63,9 @@ static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
>  static QLIST_HEAD(, BlockDriver) bdrv_drivers =
>      QLIST_HEAD_INITIALIZER(bdrv_drivers);
>  
> +/* The device to use for VM snapshots */
> +static BlockDriverState *bs_snapshots;
> +
>  /* If non-zero, use only whitelisted block drivers */
>  static int use_bdrv_whitelist;
>  
> @@ -660,6 +663,9 @@ void bdrv_close_all(void)
>  void bdrv_delete(BlockDriverState *bs)
>  {
>      assert(!bs->peer);
> +    if (bs == bs_snapshots) {
> +        bs_snapshots = NULL;
> +    }

This should probably be in bdrv_close() instead. A BlockDriverState can
be closed, but not deleted yet; it can't handle snapshots in this state,
though.

>      /* remove from list, if necessary */
>      if (bs->device_name[0] != '\0') {
> @@ -1772,6 +1778,25 @@ int bdrv_can_snapshot(BlockDriverState *bs)
>      return 1;
>  }
>  
> +BlockDriverState *bdrv_snapshots(void)
> +{
> +    BlockDriverState *bs;
> +
> +    if (bs_snapshots)
> +        return bs_snapshots;

I know that this function is just moved with no changes, but while we're
at it and you need to respin anyway, can we add braces here?

> +
> +    bs = NULL;
> +    while ((bs = bdrv_next(bs))) {
> +        if (bdrv_can_snapshot(bs)) {
> +            goto ok;
> +        }
> +    }
> +    return NULL;
> + ok:
> +    bs_snapshots = bs;
> +    return bs;
> +}

And instead of a goto we could do the right thing directly in the if block.

Kevin

[Qemu-devel] Re: [PATCH 09/12] savevm: Survive hot-unplug of snapshot device

[Markus Armbruster]

Kevin Wolf <kwolf@redhat.com> writes:

> Am 25.06.2010 18:53, schrieb Markus Armbruster:
>> savevm.c keeps a pointer to the snapshot block device.  If you manage
>> to get that device deleted, the pointer dangles, and the next snapshot
>> operation will crash & burn.  Unplugging a guest device that uses it
>> does the trick:
>> 
>>     $ MALLOC_PERTURB_=234 qemu-system-x86_64 [...]
>>     QEMU 0.12.50 monitor - type 'help' for more information
>>     (qemu) info snapshots
>>     No available block device supports snapshots
>>     (qemu) drive_add auto if=none,file=tmp.qcow2
>>     OK
>>     (qemu) device_add usb-storage,id=foo,drive=none1
>>     (qemu) info snapshots
>>     Snapshot devices: none1
>>     Snapshot list (from none1):
>>     ID        TAG                 VM SIZE                DATE       VM CLOCK
>>     (qemu) device_del foo
>>     (qemu) info snapshots
>>     Snapshot devices:
>>     Segmentation fault (core dumped)
>> 
>> Move management of that pointer to block.c, and zap it when the device
>> it points to goes away.
>> 
>> Signed-off-by: Markus Armbruster <armbru@redhat.com>
>> ---
>>  block.c  |   25 +++++++++++++++++++++++++
>>  block.h  |    1 +
>>  savevm.c |   31 ++++---------------------------
>>  3 files changed, 30 insertions(+), 27 deletions(-)
>> 
>> diff --git a/block.c b/block.c
>> index 5e0ffa0..34055e0 100644
>> --- a/block.c
>> +++ b/block.c
>> @@ -63,6 +63,9 @@ static QTAILQ_HEAD(, BlockDriverState) bdrv_states =
>>  static QLIST_HEAD(, BlockDriver) bdrv_drivers =
>>      QLIST_HEAD_INITIALIZER(bdrv_drivers);
>>  
>> +/* The device to use for VM snapshots */
>> +static BlockDriverState *bs_snapshots;
>> +
>>  /* If non-zero, use only whitelisted block drivers */
>>  static int use_bdrv_whitelist;
>>  
>> @@ -660,6 +663,9 @@ void bdrv_close_all(void)
>>  void bdrv_delete(BlockDriverState *bs)
>>  {
>>      assert(!bs->peer);
>> +    if (bs == bs_snapshots) {
>> +        bs_snapshots = NULL;
>> +    }
>
> This should probably be in bdrv_close() instead. A BlockDriverState can
> be closed, but not deleted yet; it can't handle snapshots in this state,
> though.

Right.  I was thinking about the dangling pointer only.

My patch works as advertized: it fixes the crash.  But zapping
bs_snapshots in bdrv_close() is even better.  I'll respin.


>>      /* remove from list, if necessary */
>>      if (bs->device_name[0] != '\0') {
>> @@ -1772,6 +1778,25 @@ int bdrv_can_snapshot(BlockDriverState *bs)
>>      return 1;
>>  }
>>  
>> +BlockDriverState *bdrv_snapshots(void)
>> +{
>> +    BlockDriverState *bs;
>> +
>> +    if (bs_snapshots)
>> +        return bs_snapshots;
>
> I know that this function is just moved with no changes, but while we're
> at it and you need to respin anyway, can we add braces here?
>
>> +
>> +    bs = NULL;
>> +    while ((bs = bdrv_next(bs))) {
>> +        if (bdrv_can_snapshot(bs)) {
>> +            goto ok;
>> +        }
>> +    }
>> +    return NULL;
>> + ok:
>> +    bs_snapshots = bs;
>> +    return bs;
>> +}
>
> And instead of a goto we could do the right thing directly in the if block.

Separate patch.  I hate it when people squash code motion and change
together.

[Qemu-devel] [PATCH 10/12] block: Fix virtual media change for if=none

[Markus Armbruster]

BlockDriverState member removable controls whether virtual media
change (monitor commands change, eject) is allowed.  It is set when
the "type hint" is BDRV_TYPE_CDROM or BDRV_TYPE_FLOPPY.

The type hint is only set by drive_init().  It sets BDRV_TYPE_FLOPPY
for if=floppy.  It sets BDRV_TYPE_CDROM for media=cdrom and if=ide,
scsi, xen, or none.

if=ide and if=scsi work, because the type hint makes it a CD-ROM.
if=xen likewise, I think.

For the same reason, if=none works when it's used by ide-drive or
scsi-disk.  For other guest devices, there are problems:

* fdc: you can't change virtual media

    $ qemu [...] -drive if=none,id=foo,... -global isa-fdc.driveA=foo
    QEMU 0.12.50 monitor - type 'help' for more information
    (qemu) eject foo
    Device 'foo' is not removable

  unless you add media=cdrom, but that makes it readonly.

* virtio: if you add media=cdrom, you can change virtual media.  If
  you eject, the guest gets I/O errors.  If you change, the guest sees
  the drive's contents suddenly change.

* scsi-generic: if you add media=cdrom, you can change virtual media.
  I didn't test what that does to the guest or the physical device,
  but it can't be pretty.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 block.c           |    8 ++++++++
 block.h           |    1 +
 hw/fdc.c          |   10 ++++++++--
 hw/ide/core.c     |    1 +
 hw/scsi-disk.c    |    5 ++++-
 hw/scsi-generic.c |    1 +
 hw/virtio-blk.c   |    1 +
 7 files changed, 24 insertions(+), 3 deletions(-)

diff --git a/block.c b/block.c
index 34055e0..2ae4275 100644
--- a/block.c
+++ b/block.c
@@ -1292,6 +1292,14 @@ BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read)
     return is_read ? bs->on_read_error : bs->on_write_error;
 }
 
+void bdrv_set_removable(BlockDriverState *bs, int removable)
+{
+    bs->removable = removable;
+    if (removable && bs == bs_snapshots) {
+        bs_snapshots = NULL;
+    }
+}
+
 int bdrv_is_removable(BlockDriverState *bs)
 {
     return bs->removable;
diff --git a/block.h b/block.h
index 012c2a1..3d03b3e 100644
--- a/block.h
+++ b/block.h
@@ -162,6 +162,7 @@ int bdrv_get_translation_hint(BlockDriverState *bs);
 void bdrv_set_on_error(BlockDriverState *bs, BlockErrorAction on_read_error,
                        BlockErrorAction on_write_error);
 BlockErrorAction bdrv_get_on_error(BlockDriverState *bs, int is_read);
+void bdrv_set_removable(BlockDriverState *bs, int removable);
 int bdrv_is_removable(BlockDriverState *bs);
 int bdrv_is_read_only(BlockDriverState *bs);
 int bdrv_is_sg(BlockDriverState *bs);
diff --git a/hw/fdc.c b/hw/fdc.c
index 1496cfa..6c74878 100644
--- a/hw/fdc.c
+++ b/hw/fdc.c
@@ -1847,10 +1847,16 @@ static void fdctrl_result_timer(void *opaque)
 static void fdctrl_connect_drives(FDCtrl *fdctrl)
 {
     unsigned int i;
+    FDrive *drive;
 
     for (i = 0; i < MAX_FD; i++) {
-        fd_init(&fdctrl->drives[i]);
-        fd_revalidate(&fdctrl->drives[i]);
+        drive = &fdctrl->drives[i];
+
+        fd_init(drive);
+        fd_revalidate(drive);
+        if (drive->bs) {
+            bdrv_set_removable(drive->bs, 1);
+        }
     }
 }
 
diff --git a/hw/ide/core.c b/hw/ide/core.c
index cc4591b..ebdceb5 100644
--- a/hw/ide/core.c
+++ b/hw/ide/core.c
@@ -2629,6 +2629,7 @@ void ide_init_drive(IDEState *s, BlockDriverState *bs,
         pstrcpy(s->version, sizeof(s->version), QEMU_VERSION);
     }
     ide_reset(s);
+    bdrv_set_removable(bs, s->is_cdrom);
 }
 
 static void ide_init1(IDEBus *bus, int unit)
diff --git a/hw/scsi-disk.c b/hw/scsi-disk.c
index 9c78979..2211245 100644
--- a/hw/scsi-disk.c
+++ b/hw/scsi-disk.c
@@ -1049,6 +1049,7 @@ static void scsi_destroy(SCSIDevice *dev)
 static int scsi_disk_initfn(SCSIDevice *dev)
 {
     SCSIDiskState *s = DO_UPCAST(SCSIDiskState, qdev, dev);
+    int is_cd;
     DriveInfo *dinfo;
 
     if (!s->qdev.conf.bs) {
@@ -1056,6 +1057,7 @@ static int scsi_disk_initfn(SCSIDevice *dev)
         return -1;
     }
     s->bs = s->qdev.conf.bs;
+    is_cd = bdrv_get_type_hint(s->bs) == BDRV_TYPE_CDROM;
 
     if (!s->serial) {
         /* try to fall back to value set with legacy -drive serial=... */
@@ -1072,7 +1074,7 @@ static int scsi_disk_initfn(SCSIDevice *dev)
         return -1;
     }
 
-    if (bdrv_get_type_hint(s->bs) == BDRV_TYPE_CDROM) {
+    if (is_cd) {
         s->qdev.blocksize = 2048;
     } else {
         s->qdev.blocksize = s->qdev.conf.logical_block_size;
@@ -1081,6 +1083,7 @@ static int scsi_disk_initfn(SCSIDevice *dev)
 
     s->qdev.type = TYPE_DISK;
     qemu_add_vm_change_state_handler(scsi_dma_restart_cb, s);
+    bdrv_set_removable(s->bs, is_cd);
     return 0;
 }
 
diff --git a/hw/scsi-generic.c b/hw/scsi-generic.c
index 79347f4..3915e78 100644
--- a/hw/scsi-generic.c
+++ b/hw/scsi-generic.c
@@ -509,6 +509,7 @@ static int scsi_generic_initfn(SCSIDevice *dev)
     DPRINTF("block size %d\n", s->qdev.blocksize);
     s->driver_status = 0;
     memset(s->sensebuf, 0, sizeof(s->sensebuf));
+    bdrv_set_removable(s->bs, 0);
     return 0;
 }
 
diff --git a/hw/virtio-blk.c b/hw/virtio-blk.c
index 71c5d21..f0b3ba5 100644
--- a/hw/virtio-blk.c
+++ b/hw/virtio-blk.c
@@ -500,6 +500,7 @@ VirtIODevice *virtio_blk_init(DeviceState *dev, BlockConf *conf)
     qemu_add_vm_change_state_handler(virtio_blk_dma_restart_cb, s);
     register_savevm("virtio-blk", virtio_blk_id++, 2,
                     virtio_blk_save, virtio_blk_load, s);
+    bdrv_set_removable(s->bs, 0);
 
     return &s->vdev;
 }
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 10/12] block: Fix virtual media change for if=none

[Christoph Hellwig]

Looks good,


Reviewed-by: Christoph Hellwig <hch@lst.de>

[Qemu-devel] [PATCH 11/12] ide: Make PIIX and ISA IDE init functions return the qdev

[Markus Armbruster]

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 hw/ide.h      |   11 ++++++-----
 hw/ide/isa.c  |    8 ++++----
 hw/ide/piix.c |    6 ++++--
 3 files changed, 14 insertions(+), 11 deletions(-)

diff --git a/hw/ide.h b/hw/ide.h
index 0e7d540..f0cb320 100644
--- a/hw/ide.h
+++ b/hw/ide.h
@@ -1,17 +1,18 @@
 #ifndef HW_IDE_H
 #define HW_IDE_H
 
-#include "qdev.h"
+#include "isa.h"
+#include "pci.h"
 
 /* ide-isa.c */
-int isa_ide_init(int iobase, int iobase2, int isairq,
-                 DriveInfo *hd0, DriveInfo *hd1);
+ISADevice *isa_ide_init(int iobase, int iobase2, int isairq,
+                        DriveInfo *hd0, DriveInfo *hd1);
 
 /* ide-pci.c */
 void pci_cmd646_ide_init(PCIBus *bus, DriveInfo **hd_table,
                          int secondary_ide_enabled);
-void pci_piix3_ide_init(PCIBus *bus, DriveInfo **hd_table, int devfn);
-void pci_piix4_ide_init(PCIBus *bus, DriveInfo **hd_table, int devfn);
+PCIDevice *pci_piix3_ide_init(PCIBus *bus, DriveInfo **hd_table, int devfn);
+PCIDevice *pci_piix4_ide_init(PCIBus *bus, DriveInfo **hd_table, int devfn);
 
 /* ide-macio.c */
 int pmac_ide_init (DriveInfo **hd_table, qemu_irq irq,
diff --git a/hw/ide/isa.c b/hw/ide/isa.c
index b6c6347..10777ca 100644
--- a/hw/ide/isa.c
+++ b/hw/ide/isa.c
@@ -75,8 +75,8 @@ static int isa_ide_initfn(ISADevice *dev)
     return 0;
 };
 
-int isa_ide_init(int iobase, int iobase2, int isairq,
-                 DriveInfo *hd0, DriveInfo *hd1)
+ISADevice *isa_ide_init(int iobase, int iobase2, int isairq,
+                        DriveInfo *hd0, DriveInfo *hd1)
 {
     ISADevice *dev;
     ISAIDEState *s;
@@ -86,14 +86,14 @@ int isa_ide_init(int iobase, int iobase2, int isairq,
     qdev_prop_set_uint32(&dev->qdev, "iobase2", iobase2);
     qdev_prop_set_uint32(&dev->qdev, "irq",     isairq);
     if (qdev_init(&dev->qdev) < 0)
-        return -1;
+        return NULL;
 
     s = DO_UPCAST(ISAIDEState, dev, dev);
     if (hd0)
         ide_create_drive(&s->bus, 0, hd0);
     if (hd1)
         ide_create_drive(&s->bus, 1, hd1);
-    return 0;
+    return dev;
 }
 
 static ISADeviceInfo isa_ide_info = {
diff --git a/hw/ide/piix.c b/hw/ide/piix.c
index dad6e86..fa22226 100644
--- a/hw/ide/piix.c
+++ b/hw/ide/piix.c
@@ -160,22 +160,24 @@ static int pci_piix4_ide_initfn(PCIDevice *dev)
 
 /* hd_table must contain 4 block drivers */
 /* NOTE: for the PIIX3, the IRQs and IOports are hardcoded */
-void pci_piix3_ide_init(PCIBus *bus, DriveInfo **hd_table, int devfn)
+PCIDevice *pci_piix3_ide_init(PCIBus *bus, DriveInfo **hd_table, int devfn)
 {
     PCIDevice *dev;
 
     dev = pci_create_simple(bus, devfn, "piix3-ide");
     pci_ide_create_devs(dev, hd_table);
+    return dev;
 }
 
 /* hd_table must contain 4 block drivers */
 /* NOTE: for the PIIX4, the IRQs and IOports are hardcoded */
-void pci_piix4_ide_init(PCIBus *bus, DriveInfo **hd_table, int devfn)
+PCIDevice *pci_piix4_ide_init(PCIBus *bus, DriveInfo **hd_table, int devfn)
 {
     PCIDevice *dev;
 
     dev = pci_create_simple(bus, devfn, "piix4-ide");
     pci_ide_create_devs(dev, hd_table);
+    return dev;
 }
 
 static PCIDeviceInfo piix_ide_info[] = {
-- 
1.6.6.1

[Qemu-devel] Re: [PATCH 11/12] ide: Make PIIX and ISA IDE init functions return the qdev

[Christoph Hellwig]

Looks good,


Reviewed-by: Christoph Hellwig <hch@lst.de>

[Qemu-devel] [PATCH 12/12] pc: Fix CMOS info for drives defined with -device

[Markus Armbruster]

Drives defined with -drive if=ide get get created along with the IDE
controller, inside machine->init().  That's before cmos_init().
Drives defined with -device get created during generic device init.
That's after cmos_init().  Because of that, CMOS has no information on
them (type, geometry, translation).  Older versions of Windows such as
XP reportedly choke on that.

Split off the part of CMOS initialization that needs to know about
-device devices, and turn it into a reset handler, so it runs after
device creation.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 hw/ide.h      |    2 +
 hw/ide/qdev.c |    7 ++++
 hw/pc.c       |   94 +++++++++++++++++++++++++++++++++++---------------------
 hw/pc.h       |    3 +-
 hw/pc_piix.c  |   16 +++++++---
 5 files changed, 81 insertions(+), 41 deletions(-)

diff --git a/hw/ide.h b/hw/ide.h
index f0cb320..4ccb580 100644
--- a/hw/ide.h
+++ b/hw/ide.h
@@ -23,4 +23,6 @@ void mmio_ide_init (target_phys_addr_t membase, target_phys_addr_t membase2,
                     qemu_irq irq, int shift,
                     DriveInfo *hd0, DriveInfo *hd1);
 
+void ide_get_bs(BlockDriverState *bs[], BusState *qbus);
+
 #endif /* HW_IDE_H */
diff --git a/hw/ide/qdev.c b/hw/ide/qdev.c
index b4bc5ac..2d9acbb 100644
--- a/hw/ide/qdev.c
+++ b/hw/ide/qdev.c
@@ -88,6 +88,13 @@ IDEDevice *ide_create_drive(IDEBus *bus, int unit, DriveInfo *drive)
     return DO_UPCAST(IDEDevice, qdev, dev);
 }
 
+void ide_get_bs(BlockDriverState *bs[], BusState *qbus)
+{
+    IDEBus *bus = DO_UPCAST(IDEBus, qbus, qbus);
+    bs[0] = bus->master ? bus->master->conf.bs : NULL;
+    bs[1] = bus->slave  ? bus->slave->conf.bs  : NULL;
+}
+
 /* --------------------------------- */
 
 typedef struct IDEDrive {
diff --git a/hw/pc.c b/hw/pc.c
index 1848151..0cea196 100644
--- a/hw/pc.c
+++ b/hw/pc.c
@@ -25,6 +25,7 @@
 #include "pc.h"
 #include "apic.h"
 #include "fdc.h"
+#include "ide.h"
 #include "pci.h"
 #include "vmware_vga.h"
 #include "monitor.h"
@@ -275,14 +276,65 @@ static int pc_boot_set(void *opaque, const char *boot_device)
     return set_boot_dev(opaque, boot_device, 0);
 }
 
-/* hd_table must contain 4 block drivers */
+typedef struct pc_cmos_init_late_arg {
+    ISADevice *rtc_state;
+    BusState *idebus0, *idebus1;
+} pc_cmos_init_late_arg;
+
+static void pc_cmos_init_late(void *opaque)
+{
+    pc_cmos_init_late_arg *arg = opaque;
+    ISADevice *s = arg->rtc_state;
+    int val;
+    BlockDriverState *hd_table[4];
+    int i;
+
+    ide_get_bs(hd_table, arg->idebus0);
+    ide_get_bs(hd_table + 2, arg->idebus1);
+
+    rtc_set_memory(s, 0x12, (hd_table[0] ? 0xf0 : 0) | (hd_table[1] ? 0x0f : 0));
+    if (hd_table[0])
+        cmos_init_hd(0x19, 0x1b, hd_table[0], s);
+    if (hd_table[1])
+        cmos_init_hd(0x1a, 0x24, hd_table[1], s);
+
+    val = 0;
+    for (i = 0; i < 4; i++) {
+        if (hd_table[i]) {
+            int cylinders, heads, sectors, translation;
+            /* NOTE: bdrv_get_geometry_hint() returns the physical
+                geometry.  It is always such that: 1 <= sects <= 63, 1
+                <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
+                geometry can be different if a translation is done. */
+            translation = bdrv_get_translation_hint(hd_table[i]);
+            if (translation == BIOS_ATA_TRANSLATION_AUTO) {
+                bdrv_get_geometry_hint(hd_table[i], &cylinders, &heads, &sectors);
+                if (cylinders <= 1024 && heads <= 16 && sectors <= 63) {
+                    /* No translation. */
+                    translation = 0;
+                } else {
+                    /* LBA translation. */
+                    translation = 1;
+                }
+            } else {
+                translation--;
+            }
+            val |= translation << (i * 2);
+        }
+    }
+    rtc_set_memory(s, 0x39, val);
+
+    qemu_unregister_reset(pc_cmos_init_late, opaque);
+}
+
 void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
-                  const char *boot_device, DriveInfo **hd_table,
+                  const char *boot_device,
+                  BusState *idebus0, BusState *idebus1, 
                   FDCtrl *floppy_controller, ISADevice *s)
 {
     int val;
     int fd0, fd1, nb;
-    int i;
+    static pc_cmos_init_late_arg arg;
 
     /* various important CMOS locations needed by PC/Bochs bios */
 
@@ -351,38 +403,10 @@ void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
     rtc_set_memory(s, REG_EQUIPMENT_BYTE, val);
 
     /* hard drives */
-
-    rtc_set_memory(s, 0x12, (hd_table[0] ? 0xf0 : 0) | (hd_table[1] ? 0x0f : 0));
-    if (hd_table[0])
-        cmos_init_hd(0x19, 0x1b, hd_table[0]->bdrv, s);
-    if (hd_table[1])
-        cmos_init_hd(0x1a, 0x24, hd_table[1]->bdrv, s);
-
-    val = 0;
-    for (i = 0; i < 4; i++) {
-        if (hd_table[i]) {
-            int cylinders, heads, sectors, translation;
-            /* NOTE: bdrv_get_geometry_hint() returns the physical
-                geometry.  It is always such that: 1 <= sects <= 63, 1
-                <= heads <= 16, 1 <= cylinders <= 16383. The BIOS
-                geometry can be different if a translation is done. */
-            translation = bdrv_get_translation_hint(hd_table[i]->bdrv);
-            if (translation == BIOS_ATA_TRANSLATION_AUTO) {
-                bdrv_get_geometry_hint(hd_table[i]->bdrv, &cylinders, &heads, &sectors);
-                if (cylinders <= 1024 && heads <= 16 && sectors <= 63) {
-                    /* No translation. */
-                    translation = 0;
-                } else {
-                    /* LBA translation. */
-                    translation = 1;
-                }
-            } else {
-                translation--;
-            }
-            val |= translation << (i * 2);
-        }
-    }
-    rtc_set_memory(s, 0x39, val);
+    arg.rtc_state = s;
+    arg.idebus0 = idebus0;
+    arg.idebus1 = idebus1;
+    qemu_register_reset(pc_cmos_init_late, &arg);
 }
 
 static void handle_a20_line_change(void *opaque, int irq, int level)
diff --git a/hw/pc.h b/hw/pc.h
index ccfd7ad..63b0249 100644
--- a/hw/pc.h
+++ b/hw/pc.h
@@ -104,7 +104,8 @@ void pc_init_ne2k_isa(NICInfo *nd);
 void pc_audio_init (PCIBus *pci_bus, qemu_irq *pic);
 #endif
 void pc_cmos_init(ram_addr_t ram_size, ram_addr_t above_4g_mem_size,
-                  const char *boot_device, DriveInfo **hd_table,
+                  const char *boot_device,
+                  BusState *ide0, BusState *ide1,
                   FDCtrl *floppy_controller, ISADevice *s);
 void pc_pci_device_init(PCIBus *pci_bus);
 
diff --git a/hw/pc_piix.c b/hw/pc_piix.c
index 68040b7..7f07e47 100644
--- a/hw/pc_piix.c
+++ b/hw/pc_piix.c
@@ -79,6 +79,7 @@ static void pc_init1(ram_addr_t ram_size,
     IsaIrqState *isa_irq_state;
     DriveInfo *hd[MAX_IDE_BUS * MAX_IDE_DEVS];
     FDCtrl *floppy_controller;
+    BusState *idebus[MAX_IDE_BUS];
     ISADevice *rtc_state;
 
     pc_cpus_init(cpu_model);
@@ -132,11 +133,16 @@ static void pc_init1(ram_addr_t ram_size,
     }
 
     if (pci_enabled) {
-        pci_piix3_ide_init(pci_bus, hd, piix3_devfn + 1);
+        PCIDevice *dev;
+        dev = pci_piix3_ide_init(pci_bus, hd, piix3_devfn + 1);
+        idebus[0] = qdev_get_child_bus(&dev->qdev, "ide.0");
+        idebus[1] = qdev_get_child_bus(&dev->qdev, "ide.1");
     } else {
         for(i = 0; i < MAX_IDE_BUS; i++) {
-            isa_ide_init(ide_iobase[i], ide_iobase2[i], ide_irq[i],
-	                 hd[MAX_IDE_DEVS * i], hd[MAX_IDE_DEVS * i + 1]);
+            ISADevice *dev;
+            dev = isa_ide_init(ide_iobase[i], ide_iobase2[i], ide_irq[i],
+                               hd[MAX_IDE_DEVS * i], hd[MAX_IDE_DEVS * i + 1]);
+            idebus[i] = qdev_get_child_bus(&dev->qdev, "ide.0");
         }
     }
 
@@ -144,8 +150,8 @@ static void pc_init1(ram_addr_t ram_size,
     pc_audio_init(pci_enabled ? pci_bus : NULL, isa_irq);
 #endif
 
-    pc_cmos_init(below_4g_mem_size, above_4g_mem_size, boot_device, hd,
-                 floppy_controller, rtc_state);
+    pc_cmos_init(below_4g_mem_size, above_4g_mem_size, boot_device,
+                 idebus[0], idebus[1], floppy_controller, rtc_state);
 
     if (pci_enabled && usb_enabled) {
         usb_uhci_piix3_init(pci_bus, piix3_devfn + 2);
-- 
1.6.6.1