[Qemu-devel] [PATCH 0/3] allow online resizing of block devices

[Qemu-devel] [PATCH 0/3] allow online resizing of block devices

[Christoph Hellwig]

This patchset adds support for online resizing of block devices.

The first patch adds a new resize monitor command which call into
the existing image resize code.  This is the meat of the series
and probably needs quite a bit of review and help as I'm not sure
about how to implement the error handling for monitor commands
correctly.  Am I really supposed to add a new QERR_ definition
for each possible error?  And if yes how am I supposed to define
them?  The macros for them aren't exactly self-explaining.

The second patch adds a way to tell drivers about a resize, and the
third one adds a guest notification for config changes to virtio-blk
which allows the guest to pick it up without a rescan.  I've just sent
the corresponding Linux guest driver patch to Rusty.

[Qemu-devel] [PATCH 1/3] block: add resize monitor command

[Christoph Hellwig]

Add a monitor command that allows resizing of block devices while
qemu is running.  It uses the existing bdrv_truncate method already
used by qemu-img to do it's work.  Compared to qemu-img the size
parsing is very simplicistic, but I think having a properly numering
object is more useful for non-humand monitor users than having
the units and relative resize parsing.

For SCSI devices the new size can be updated in Linux guests by
doing the following shell command:

	echo > /sys/class/scsi_device/0:0:0:0/device/rescan

For ATA devices I don't know of a way to update the block device
size in Linux system, and for virtio-blk the next two patches
will provide an automatic update of the size when this command
is issued on the host.

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

Index: qemu/hmp-commands.hx
===================================================================
--- qemu.orig/hmp-commands.hx	2011-01-14 15:11:48.527004132 +0100
+++ qemu/hmp-commands.hx	2011-01-14 15:40:14.407006506 +0100
@@ -53,6 +53,24 @@ Quit the emulator.
 ETEXI
 
     {
+        .name       = "resize",
+        .args_type  = "id:s,size:l",
+        .params     = "device size",
+        .help       = "resize a block image",
+        .user_print = monitor_user_noop,
+        .mhandler.cmd_new = do_resize,
+    },
+
+STEXI
+@item resize
+@findex resize
+Resize a block image while a guest is running.  Usuaully requires guest
+action to see the updated size.  Resize to a lower size is supported,
+but should be used with extreme caution.
+ETEXI
+
+
+    {
         .name       = "eject",
         .args_type  = "force:-f,device:B",
         .params     = "[-f] device",
Index: qemu/blockdev.c
===================================================================
--- qemu.orig/blockdev.c	2011-01-14 15:11:48.539261151 +0100
+++ qemu/blockdev.c	2011-01-14 15:50:35.604293558 +0100
@@ -700,3 +700,41 @@ int do_drive_del(Monitor *mon, const QDi
 
     return 0;
 }
+
+int do_resize(Monitor *mon, const QDict *qdict, QObject **ret_data)
+{
+    const char *id = qdict_get_str(qdict, "id");
+    int64_t size = qdict_get_int(qdict, "size");
+    BlockDriverState *bs;
+
+    bs = bdrv_find(id);
+    if (!bs) {
+        qerror_report(QERR_DEVICE_NOT_FOUND, id);
+        return -1;
+    }
+
+    if (bdrv_get_type_hint(bs) == BDRV_TYPE_CDROM) {
+        error_report("Can not resize CDROM devices\n");
+        return -1;
+    }
+
+    if (size < 0) {
+        error_report("Can only resize to positive sizes");
+        return -1;
+    }
+
+    switch (bdrv_truncate(bs, size)) {
+    case 0:
+        return 0;
+    case -ENOTSUP:
+        error_report("This image format does not support resize");
+        return -1;
+    case -EACCES:
+        error_report("Image is read-only");
+        return -1;
+    default:
+        error_report("Error resizing image");
+        return -1;
+    }
+}
+
Index: qemu/blockdev.h
===================================================================
--- qemu.orig/blockdev.h	2011-01-14 15:11:48.548263456 +0100
+++ qemu/blockdev.h	2011-01-14 15:12:00.965047363 +0100
@@ -53,5 +53,6 @@ int do_change_block(Monitor *mon, const
                     const char *filename, const char *fmt);
 int do_drive_del(Monitor *mon, const QDict *qdict, QObject **ret_data);
 int do_snapshot_blkdev(Monitor *mon, const QDict *qdict, QObject **ret_data);
+int do_resize(Monitor *mon, const QDict *qdict, QObject **ret_data);
 
 #endif

Re: [Qemu-devel] [PATCH 1/3] block: add resize monitor command

[Stefan Hajnoczi]

On Fri, Jan 14, 2011 at 4:20 PM, Christoph Hellwig <hch@lst.de> wrote:
> +STEXI
> +@item resize
> +@findex resize
> +Resize a block image while a guest is running.  Usuaully requires guest

s/Usuaully/Usually/

> +action to see the updated size.  Resize to a lower size is supported,
> +but should be used with extreme caution.

This resizes the image files.  Resizing an LVM volume is a useful case
too.  One way to integrate that feature is to implement a host_device
.bdrv_truncate() that checks the underlying device size and returns
success if it matches the new value and failure otherwise.  Or perhaps
allow the QEMU resize command without an argument to fetch the size
from the underlying device (aka refresh the size).

> +    if (bdrv_get_type_hint(bs) == BDRV_TYPE_CDROM) {
> +        error_report("Can not resize CDROM devices\n");
> +        return -1;
> +    }

Hrm...BDRV_TYPE_FLOPPY probably too?

Stefan

Re: [Qemu-devel] [PATCH 1/3] block: add resize monitor command

[Christoph Hellwig]

On Mon, Jan 17, 2011 at 11:28:47AM +0000, Stefan Hajnoczi wrote:
> > + ? ?if (bdrv_get_type_hint(bs) == BDRV_TYPE_CDROM) {
> > + ? ? ? ?error_report("Can not resize CDROM devices\n");
> > + ? ? ? ?return -1;
> > + ? ?}
> 
> Hrm...BDRV_TYPE_FLOPPY probably too?

If we want to be consistent, yes.  Or we could remove the CDROM check
and let people shoot themselves in the foot as much as they want.

Re: [Qemu-devel] [PATCH 1/3] block: add resize monitor command

[Kevin Wolf]

Am 14.01.2011 17:20, schrieb Christoph Hellwig:
> Add a monitor command that allows resizing of block devices while
> qemu is running.  It uses the existing bdrv_truncate method already
> used by qemu-img to do it's work.  Compared to qemu-img the size
> parsing is very simplicistic, but I think having a properly numering
> object is more useful for non-humand monitor users than having
> the units and relative resize parsing.
> 
> For SCSI devices the new size can be updated in Linux guests by
> doing the following shell command:
> 
> 	echo > /sys/class/scsi_device/0:0:0:0/device/rescan
> 
> For ATA devices I don't know of a way to update the block device
> size in Linux system, and for virtio-blk the next two patches
> will provide an automatic update of the size when this command
> is issued on the host.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> 
> Index: qemu/hmp-commands.hx
> ===================================================================
> --- qemu.orig/hmp-commands.hx	2011-01-14 15:11:48.527004132 +0100
> +++ qemu/hmp-commands.hx	2011-01-14 15:40:14.407006506 +0100
> @@ -53,6 +53,24 @@ Quit the emulator.
>  ETEXI
>  
>      {
> +        .name       = "resize",
> +        .args_type  = "id:s,size:l",

size should be 'o' instead of 'l'. The latter may be too small on 32 bit
hosts and doesn't support convenient suffixes:

 * 'l'          target long (32 or 64 bit)

 * 'o'          octets (aka bytes)
 *              user mode accepts an optional T, t, G, g, M, m, K, k
 *              suffix, which multiplies the value by 2^40 for
 *              suffixes T and t, 2^30 for suffixes G and g, 2^20 for
 *              M and m, 2^10 for K and k

> +        .params     = "device size",
> +        .help       = "resize a block image",
> +        .user_print = monitor_user_noop,
> +        .mhandler.cmd_new = do_resize,
> +    },
> +
> +STEXI
> +@item resize
> +@findex resize
> +Resize a block image while a guest is running.  Usuaully requires guest
> +action to see the updated size.  Resize to a lower size is supported,
> +but should be used with extreme caution.
> +ETEXI
> +
> +
> +    {
>          .name       = "eject",
>          .args_type  = "force:-f,device:B",
>          .params     = "[-f] device",
> Index: qemu/blockdev.c
> ===================================================================
> --- qemu.orig/blockdev.c	2011-01-14 15:11:48.539261151 +0100
> +++ qemu/blockdev.c	2011-01-14 15:50:35.604293558 +0100
> @@ -700,3 +700,41 @@ int do_drive_del(Monitor *mon, const QDi
>  
>      return 0;
>  }
> +
> +int do_resize(Monitor *mon, const QDict *qdict, QObject **ret_data)
> +{
> +    const char *id = qdict_get_str(qdict, "id");
> +    int64_t size = qdict_get_int(qdict, "size");
> +    BlockDriverState *bs;
> +
> +    bs = bdrv_find(id);
> +    if (!bs) {
> +        qerror_report(QERR_DEVICE_NOT_FOUND, id);
> +        return -1;
> +    }
> +
> +    if (bdrv_get_type_hint(bs) == BDRV_TYPE_CDROM) {
> +        error_report("Can not resize CDROM devices\n");
> +        return -1;
> +    }

Hm, is there a real reason except that CD-ROMs are read-only? The code
below seems to take read-only devices into account.

Kevin

Re: [Qemu-devel] [PATCH 1/3] block: add resize monitor command

[Christoph Hellwig]

On Mon, Jan 17, 2011 at 06:36:15PM +0100, Kevin Wolf wrote:
> size should be 'o' instead of 'l'. The latter may be too small on 32 bit
> hosts and doesn't support convenient suffixes:

Fixed.

> Hm, is there a real reason except that CD-ROMs are read-only? The code
> below seems to take read-only devices into account.

I've removed the CDROM check for the next version.

Re: [Qemu-devel] [PATCH 1/3] block: add resize monitor command

[Christoph Hellwig]

On Mon, Jan 17, 2011 at 06:36:15PM +0100, Kevin Wolf wrote:
> > +        .args_type  = "id:s,size:l",
> 
> size should be 'o' instead of 'l'. The latter may be too small on 32 bit
> hosts and doesn't support convenient suffixes:

o actually fails for 2GB or more for me:

(qemu) resize scratch 2047 
resize scratch 2047 
(qemu) 
(qemu) resize scrarch 2048
resize scrarch 2048
invalid size

for l these worked fine.

Re: [Qemu-devel] [PATCH 1/3] block: add resize monitor command

[Kevin Wolf]

Am 19.01.2011 16:35, schrieb Christoph Hellwig:
> On Mon, Jan 17, 2011 at 06:36:15PM +0100, Kevin Wolf wrote:
>>> +        .args_type  = "id:s,size:l",
>>
>> size should be 'o' instead of 'l'. The latter may be too small on 32 bit
>> hosts and doesn't support convenient suffixes:
> 
> o actually fails for 2GB or more for me:
> 
> (qemu) resize scratch 2047 
> resize scratch 2047 
> (qemu) 
> (qemu) resize scrarch 2048
> resize scrarch 2048
> invalid size
> 
> for l these worked fine.

Hm, yeah, 'o' uses ssize_t instead of int64_t, so it's broken on a 32
bit host as well. Though I assume that you use a 64 bit host, and I
can't really see what's the problem there...

Kevin

Re: [Qemu-devel] [PATCH 1/3] block: add resize monitor command

[Christoph Hellwig]

On Wed, Jan 19, 2011 at 04:49:04PM +0100, Kevin Wolf wrote:
> > (qemu) resize scratch 2047 
> > resize scratch 2047 
> > (qemu) 
> > (qemu) resize scrarch 2048
> > resize scrarch 2048
> > invalid size
> > 
> > for l these worked fine.
> 
> Hm, yeah, 'o' uses ssize_t instead of int64_t, so it's broken on a 32
> bit host as well. Though I assume that you use a 64 bit host, and I
> can't really see what's the problem there...

This is a test on a 32-bit host.  The target_long of "l" worked fine
because a kvm enabled qemu always builds for an x86-64 target, even
with a 32-bit host.

Re: [Qemu-devel] [PATCH 1/3] block: add resize monitor command

[Kevin Wolf]

Am 19.01.2011 16:52, schrieb Christoph Hellwig:
> On Wed, Jan 19, 2011 at 04:49:04PM +0100, Kevin Wolf wrote:
>>> (qemu) resize scratch 2047 
>>> resize scratch 2047 
>>> (qemu) 
>>> (qemu) resize scrarch 2048
>>> resize scrarch 2048
>>> invalid size
>>>
>>> for l these worked fine.
>>
>> Hm, yeah, 'o' uses ssize_t instead of int64_t, so it's broken on a 32
>> bit host as well. Though I assume that you use a 64 bit host, and I
>> can't really see what's the problem there...
> 
> This is a test on a 32-bit host.  The target_long of "l" worked fine
> because a kvm enabled qemu always builds for an x86-64 target, even
> with a 32-bit host.

Oh, okay, then it makes perfect sense. We should fix 'o' then to use
int64_t. The patch below should do it.

strtosz (which is used by 'o') has the same problem in git master.
There's a fix by Jes, so be sure to test this on the block branch.

Kevin

diff --git a/monitor.c b/monitor.c
index d291158..0cda3da 100644
--- a/monitor.c
+++ b/monitor.c
@@ -4162,7 +4162,7 @@ static const mon_cmd_t
*monitor_parse_command(Monitor *mon,
             break;
         case 'o':
             {
-                ssize_t val;
+                int64_t val;
                 char *end;

                 while (qemu_isspace(*p)) {

Re: [Qemu-devel] [PATCH 1/3] block: add resize monitor command

[Christoph Hellwig]

On Wed, Jan 19, 2011 at 05:01:19PM +0100, Kevin Wolf wrote:
> Oh, okay, then it makes perfect sense. We should fix 'o' then to use
> int64_t. The patch below should do it.
> 
> strtosz (which is used by 'o') has the same problem in git master.
> There's a fix by Jes, so be sure to test this on the block branch.

The larger sizes work fine on the block branch.

[Qemu-devel] [PATCH 2/3] block: tell drivers about an image resize

[Christoph Hellwig]

Add a new size_changed flag in the BlockDriverState and call the
change_cb callback to notify drivers about a size change.

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

Index: qemu/block.c
===================================================================
--- qemu.orig/block.c	2011-01-14 17:05:49.527003363 +0100
+++ qemu/block.c	2011-01-14 17:07:23.206255143 +0100
@@ -1135,6 +1135,10 @@ int bdrv_truncate(BlockDriverState *bs,
     ret = drv->bdrv_truncate(bs, offset);
     if (ret == 0) {
         ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
+        if (bs->change_cb) {
+            bs->size_changed = 1;
+            bs->change_cb(bs->change_opaque);
+        }
     }
     return ret;
 }
Index: qemu/block_int.h
===================================================================
--- qemu.orig/block_int.h	2011-01-14 17:05:49.537004411 +0100
+++ qemu/block_int.h	2011-01-14 17:06:02.539004271 +0100
@@ -167,6 +167,7 @@ struct BlockDriverState {
     char backing_format[16]; /* if non-zero and backing_file exists */
     int is_temporary;
     int media_changed;
+    int size_changed;
 
     BlockDriverState *backing_hd;
     BlockDriverState *file;

Re: [Qemu-devel] [PATCH 2/3] block: tell drivers about an image resize

[Ryan Harper]

* Christoph Hellwig <hch@lst.de> [2011-01-14 10:27]:
> Add a new size_changed flag in the BlockDriverState and call the
> change_cb callback to notify drivers about a size change.
> 
> Signed-off-by: Christoph Hellwig <hch@lst.de>
> 
> Index: qemu/block.c
> ===================================================================
> --- qemu.orig/block.c	2011-01-14 17:05:49.527003363 +0100
> +++ qemu/block.c	2011-01-14 17:07:23.206255143 +0100
> @@ -1135,6 +1135,10 @@ int bdrv_truncate(BlockDriverState *bs,
>      ret = drv->bdrv_truncate(bs, offset);
>      if (ret == 0) {
>          ret = refresh_total_sectors(bs, offset >> BDRV_SECTOR_BITS);
> +        if (bs->change_cb) {
> +            bs->size_changed = 1;
> +            bs->change_cb(bs->change_opaque);
> +        }

Do we want to check to ensure the size_changed flag isn't set before
doing a second resize event?  I'm wondering if the truncate takes
$longtime and user gets impatient and issues a second resize command.
How should we respond?  Ignore it?  queue it up? 

Is there any other path that invokes bdrv_truncate?  If so, do we want
invoke the call back in those scenarios or only if the truncate is
initiated from the monitor invocation?

>      }
>      return ret;
>  }
> Index: qemu/block_int.h
> ===================================================================
> --- qemu.orig/block_int.h	2011-01-14 17:05:49.537004411 +0100
> +++ qemu/block_int.h	2011-01-14 17:06:02.539004271 +0100
> @@ -167,6 +167,7 @@ struct BlockDriverState {
>      char backing_format[16]; /* if non-zero and backing_file exists */
>      int is_temporary;
>      int media_changed;
> +    int size_changed;
> 
>      BlockDriverState *backing_hd;
>      BlockDriverState *file;

-- 
Ryan Harper
Software Engineer; Linux Technology Center
IBM Corp., Austin, Tx
ryanh@us.ibm.com

Re: [Qemu-devel] [PATCH 2/3] block: tell drivers about an image resize

[Christoph Hellwig]

> Do we want to check to ensure the size_changed flag isn't set before
> doing a second resize event?  I'm wondering if the truncate takes
> $longtime and user gets impatient and issues a second resize command.
> How should we respond?  Ignore it?  queue it up? 

The truncate is synchronous and the monitor command won't return before
it's done.  The interfact might be a bit cleaner if we'd just pass the
reason for the change callback as another argument.  I'll do that for
the next respin.

> Is there any other path that invokes bdrv_truncate?  If so, do we want
> invoke the call back in those scenarios or only if the truncate is
> initiated from the monitor invocation?

Currently only the qemu-img and qemu-io tools invooke bdrv_truncate.
But if we grow new users we should always tell the driver/guest about it
so that it knows about the size change.

[Qemu-devel] [PATCH 3/3] virtio-blk: tell the guest about size changes

[Christoph Hellwig]

Raise a config change interrupt when the size changed.  This allows
virtio-blk guest drivers to read-read the information from the
config space once it got the config chaged interrupt.

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

Index: qemu/hw/virtio-blk.c
===================================================================
--- qemu.orig/hw/virtio-blk.c	2011-01-14 17:00:07.468262896 +0100
+++ qemu/hw/virtio-blk.c	2011-01-14 17:07:06.897257239 +0100
@@ -504,6 +504,16 @@ static int virtio_blk_load(QEMUFile *f,
     return 0;
 }
 
+static void virtio_blk_change_cb(void *opaque)
+{
+    VirtIOBlock *s = opaque;
+
+    if (s->bs->size_changed) {
+        virtio_notify_config(&s->vdev);
+        s->bs->size_changed = 0;
+    }
+}
+
 VirtIODevice *virtio_blk_init(DeviceState *dev, BlockConf *conf)
 {
     VirtIOBlock *s;
@@ -546,6 +556,7 @@ VirtIODevice *virtio_blk_init(DeviceStat
     register_savevm(dev, "virtio-blk", virtio_blk_id++, 2,
                     virtio_blk_save, virtio_blk_load, s);
     bdrv_set_removable(s->bs, 0);
+    bdrv_set_change_cb(s->bs, virtio_blk_change_cb, s);
     s->bs->buffer_alignment = conf->logical_block_size;
 
     add_boot_device_path(conf->bootindex, dev, "/disk@0,0");

Re: [Qemu-devel] [PATCH 0/3] allow online resizing of block devices

[Luiz Capitulino]

On Fri, 14 Jan 2011 17:20:44 +0100
Christoph Hellwig <hch@lst.de> wrote:

> This patchset adds support for online resizing of block devices.
> 
> The first patch adds a new resize monitor command which call into
> the existing image resize code.  This is the meat of the series
> and probably needs quite a bit of review and help as I'm not sure
> about how to implement the error handling for monitor commands
> correctly.  Am I really supposed to add a new QERR_ definition
> for each possible error?  And if yes how am I supposed to define
> them?  The macros for them aren't exactly self-explaining.

Well, what happens is this: we screwed up with that interface and we
should replace it soon.

I see you're not adding the new command in QMP (only in the human monitor),
is that intentional? (qmp commands are added to the qmp-commands.hx file).

If it's intentional, then using only error_report() should be ok. If you
plan to have a qmp version, then we'll have to choose between reporting
a generic error version to the client, which is what's going to happen if
you use error_report(), or add the QERR_ macros.

Markus, what do you think? I feel it's pretty urgent for us to replace
that interface.

Re: [Qemu-devel] [PATCH 0/3] allow online resizing of block devices

[Christoph Hellwig]

On Tue, Jan 18, 2011 at 10:35:39AM -0200, Luiz Capitulino wrote:
> Well, what happens is this: we screwed up with that interface and we
> should replace it soon.
> 
> I see you're not adding the new command in QMP (only in the human monitor),
> is that intentional? (qmp commands are added to the qmp-commands.hx file).
> 
> If it's intentional, then using only error_report() should be ok. If you
> plan to have a qmp version, then we'll have to choose between reporting
> a generic error version to the client, which is what's going to happen if
> you use error_report(), or add the QERR_ macros.

No, I hoped this would also add the QMP interface.  Why do we need to
declare the commands twice?  Especially as hmp-commands.hx is full of
magic that makes little sense only for human readable commands.

Any help on how to define the QERR_ macros?

Re: [Qemu-devel] [PATCH 0/3] allow online resizing of block devices

[Luiz Capitulino]

On Tue, 18 Jan 2011 13:48:06 +0100
Christoph Hellwig <hch@lst.de> wrote:

> On Tue, Jan 18, 2011 at 10:35:39AM -0200, Luiz Capitulino wrote:
> > Well, what happens is this: we screwed up with that interface and we
> > should replace it soon.
> > 
> > I see you're not adding the new command in QMP (only in the human monitor),
> > is that intentional? (qmp commands are added to the qmp-commands.hx file).
> > 
> > If it's intentional, then using only error_report() should be ok. If you
> > plan to have a qmp version, then we'll have to choose between reporting
> > a generic error version to the client, which is what's going to happen if
> > you use error_report(), or add the QERR_ macros.
> 
> No, I hoped this would also add the QMP interface.  Why do we need to
> declare the commands twice?

Because they're different interfaces and should not be tied to each other,
they have different name spaces for example and commands in QMP might have
no relation to commands in HMP and vice-versa.

> Especially as hmp-commands.hx is full of
> magic that makes little sense only for human readable commands.

Yes, the reason for that is that we're in the middle of a refactoring,
but I have around 100 patches that makes a clean separation between the
two and creates a cleaner interface (specially for HMP).

> Any help on how to define the QERR_ macros?

I think we should emit a generic error in QMP for now, I don't think it's
a good idea to expend the bad interface we have today.

[Qemu-devel] [PATCH 0/3 v2] allow online resizing of block devices

[Christoph Hellwig]

This patchset adds support for online resizing of block devices.

The first patch adds a new resize monitor command which call into
the existing image resize code.  This is the meat of the series
and probably needs quite a bit of review and help as I'm not sure
about how to implement the error handling for monitor commands
correctly.  Am I really supposed to add a new QERR_ definition
for each possible error?  And if yes how am I supposed to define
them?  The macros for them aren't exactly self-explaining.

The second patch adds a way to tell drivers about a resize, and the
third one adds a guest notification for config changes to virtio-blk
which allows the guest to pick it up without a rescan.  I've just sent
the corresponding Linux guest driver patch to Rusty.

Changes from version 1 to version 2:
 - also add a QMP command (block_resize)
 - use the o format for the size in the monitor command
 - fix typos
 - use QERR_UNDEFINED_ERROR for errors instead of unstructured strings
 - remove the CDROM hint check
 - add a reason argument to the change callback

[Qemu-devel] [PATCH 3/3] virtio-blk: tell the guest about size changes

[Christoph Hellwig]

Raise a config change interrupt when the size changed.  This allows
virtio-blk guest drivers to read-read the information from the
config space once it got the config chaged interrupt.

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

Index: qemu/hw/virtio-blk.c
===================================================================
--- qemu.orig/hw/virtio-blk.c	2011-01-19 17:47:10.335004828 +0100
+++ qemu/hw/virtio-blk.c	2011-01-19 17:50:10.748272881 +0100
@@ -504,6 +504,15 @@ static int virtio_blk_load(QEMUFile *f,
     return 0;
 }
 
+static void virtio_blk_change_cb(void *opaque, int reason)
+{
+    VirtIOBlock *s = opaque;
+
+    if (reason & CHANGE_SIZE) {
+        virtio_notify_config(&s->vdev);
+    }
+}
+
 VirtIODevice *virtio_blk_init(DeviceState *dev, BlockConf *conf)
 {
     VirtIOBlock *s;
@@ -546,6 +555,7 @@ VirtIODevice *virtio_blk_init(DeviceStat
     register_savevm(dev, "virtio-blk", virtio_blk_id++, 2,
                     virtio_blk_save, virtio_blk_load, s);
     bdrv_set_removable(s->bs, 0);
+    bdrv_set_change_cb(s->bs, virtio_blk_change_cb, s);
     s->bs->buffer_alignment = conf->logical_block_size;
 
     add_boot_device_path(conf->bootindex, dev, "/disk@0,0");

[Qemu-devel] [PATCH 0/3 v3] allow online resizing of block devices

[Christoph Hellwig]

This patchset adds support for online resizing of block devices.

The first patch adds a new resize monitor command which call into
the existing image resize code.  This is the meat of the series
and probably needs quite a bit of review and help as I'm not sure
about how to implement the error handling for monitor commands
correctly.  Am I really supposed to add a new QERR_ definition
for each possible error?  And if yes how am I supposed to define
them?  The macros for them aren't exactly self-explaining.

The second patch adds a way to tell drivers about a resize, and the
third one adds a guest notification for config changes to virtio-blk
which allows the guest to pick it up without a rescan.  I've just sent
the corresponding Linux guest driver patch to Rusty.

Changes from version 2 to version 3:
 - add missing braces
 - use device as the argument name for the device string
 - also rename the HMP version of the command to block_resize

Changes from version 1 to version 2:
 - also add a QMP command (block_resize)
 - use the o format for the size in the monitor command
 - fix typos
 - use QERR_UNDEFINED_ERROR for errors instead of unstructured strings
 - remove the CDROM hint check
 - add a reason argument to the change callback
---end quoted text---

[Qemu-devel] [PATCH 3/3] virtio-blk: tell the guest about size changes

[Christoph Hellwig]

Raise a config change interrupt when the size changed.  This allows
virtio-blk guest drivers to read-read the information from the
config space once it got the config chaged interrupt.

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

Index: qemu/hw/virtio-blk.c
===================================================================
--- qemu.orig/hw/virtio-blk.c	2011-01-20 10:32:50.455032625 +0100
+++ qemu/hw/virtio-blk.c	2011-01-24 11:56:42.833004269 +0100
@@ -504,6 +504,15 @@ static int virtio_blk_load(QEMUFile *f,
     return 0;
 }
 
+static void virtio_blk_change_cb(void *opaque, int reason)
+{
+    VirtIOBlock *s = opaque;
+
+    if (reason & CHANGE_SIZE) {
+        virtio_notify_config(&s->vdev);
+    }
+}
+
 VirtIODevice *virtio_blk_init(DeviceState *dev, BlockConf *conf)
 {
     VirtIOBlock *s;
@@ -546,6 +555,7 @@ VirtIODevice *virtio_blk_init(DeviceStat
     register_savevm(dev, "virtio-blk", virtio_blk_id++, 2,
                     virtio_blk_save, virtio_blk_load, s);
     bdrv_set_removable(s->bs, 0);
+    bdrv_set_change_cb(s->bs, virtio_blk_change_cb, s);
     s->bs->buffer_alignment = conf->logical_block_size;
 
     add_boot_device_path(conf->bootindex, dev, "/disk@0,0");