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---
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 2011-01-24 11:55:36.744254374 +0100 +++ qemu/hmp-commands.hx 2011-01-24 11:56:23.619254094 +0100 @@ -53,6 +53,25 @@ Quit the emulator. ETEXI { + .name = "block_resize", + .args_type = "device:B,size:o", + .params = "device size", + .help = "resize a block image", + .user_print = monitor_user_noop, + .mhandler.cmd_new = do_block_resize, + }, + +STEXI +@item block_resize +@findex block_resize +Resize a block image while a guest is running. Usually requires guest +action to see the updated size. Resize to a lower size is supported, +but should be used with extreme caution. Note that this command only +resizes image files, it can not resize block devices like LVM volumes. +ETEXI + + + { .name = "eject", .args_type = "force:-f,device:B", .params = "[-f] device", Index: qemu/blockdev.c =================================================================== --- qemu.orig/blockdev.c 2011-01-24 11:56:20.903004129 +0100 +++ qemu/blockdev.c 2011-01-24 11:56:38.391254165 +0100 @@ -705,3 +705,33 @@ int do_drive_del(Monitor *mon, const QDi return 0; } + +/* + * XXX: replace the QERR_UNDEFINED_ERROR errors with real values once the + * existing QERR_ macro mess is cleaned up. A good example for better + * error reports can be found in the qemu-img resize code. + */ +int do_block_resize(Monitor *mon, const QDict *qdict, QObject **ret_data) +{ + const char *device = qdict_get_str(qdict, "device"); + int64_t size = qdict_get_int(qdict, "size"); + BlockDriverState *bs; + + bs = bdrv_find(device); + if (!bs) { + qerror_report(QERR_DEVICE_NOT_FOUND, device); + return -1; + } + + if (size < 0) { + qerror_report(QERR_UNDEFINED_ERROR); + return -1; + } + + if (bdrv_truncate(bs, size)) { + qerror_report(QERR_UNDEFINED_ERROR); + return -1; + } + + return 0; +} Index: qemu/blockdev.h =================================================================== --- qemu.orig/blockdev.h 2011-01-24 11:55:36.764254165 +0100 +++ qemu/blockdev.h 2011-01-24 11:56:23.627253465 +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_block_resize(Monitor *mon, const QDict *qdict, QObject **ret_data); #endif Index: qemu/qmp-commands.hx =================================================================== --- qemu.orig/qmp-commands.hx 2011-01-24 11:55:36.771253955 +0100 +++ qemu/qmp-commands.hx 2011-01-24 11:56:23.632253884 +0100 @@ -601,6 +601,34 @@ Example: -> { "execute": "netdev_del", "arguments": { "id": "netdev1" } } <- { "return": {} } + +EQMP + + { + .name = "block_resize", + .args_type = "device:B,size:o", + .params = "device size", + .help = "resize a block image", + .user_print = monitor_user_noop, + .mhandler.cmd_new = do_block_resize, + }, + +SQMP +block_resize +------------ + +Resize a block image while a guest is running. + +Arguments: + +- "device": the device's ID, must be unique (json-string) +- "size": new size + +Example: + +-> { "execute": "block_resize", "arguments": { "device": "scratch", "size": 1073741824 } } +<- { "return": {} } + EQMP {
Extend the change_cb callback with a reason argument, and use it to tell drivers about size changes. Signed-off-by: Christoph Hellwig <hch@lst.de> Index: qemu/block.c =================================================================== --- qemu.orig/block.c 2011-01-24 11:56:20.899004129 +0100 +++ qemu/block.c 2011-01-24 11:56:39.796004269 +0100 @@ -645,7 +645,7 @@ int bdrv_open(BlockDriverState *bs, cons /* call the change callback */ bs->media_changed = 1; if (bs->change_cb) - bs->change_cb(bs->change_opaque); + bs->change_cb(bs->change_opaque, CHANGE_MEDIA); } return 0; @@ -684,7 +684,7 @@ void bdrv_close(BlockDriverState *bs) /* call the change callback */ bs->media_changed = 1; if (bs->change_cb) - bs->change_cb(bs->change_opaque); + bs->change_cb(bs->change_opaque, CHANGE_MEDIA); } } @@ -1135,6 +1135,9 @@ 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->change_cb(bs->change_opaque, CHANGE_SIZE); + } } return ret; } @@ -1366,7 +1369,8 @@ int bdrv_enable_write_cache(BlockDriverS /* XXX: no longer used */ void bdrv_set_change_cb(BlockDriverState *bs, - void (*change_cb)(void *opaque), void *opaque) + void (*change_cb)(void *opaque, int reason), + void *opaque) { bs->change_cb = change_cb; bs->change_opaque = opaque; @@ -1411,7 +1415,7 @@ int bdrv_set_key(BlockDriverState *bs, c /* call the change callback now, we skipped it on open */ bs->media_changed = 1; if (bs->change_cb) - bs->change_cb(bs->change_opaque); + bs->change_cb(bs->change_opaque, CHANGE_MEDIA); } return ret; } Index: qemu/block.h =================================================================== --- qemu.orig/block.h 2011-01-24 11:52:25.248004200 +0100 +++ qemu/block.h 2011-01-24 11:56:39.797004269 +0100 @@ -182,7 +182,8 @@ int bdrv_is_locked(BlockDriverState *bs) void bdrv_set_locked(BlockDriverState *bs, int locked); int bdrv_eject(BlockDriverState *bs, int eject_flag); void bdrv_set_change_cb(BlockDriverState *bs, - void (*change_cb)(void *opaque), void *opaque); + void (*change_cb)(void *opaque, int reason), + void *opaque); void bdrv_get_format(BlockDriverState *bs, char *buf, int buf_size); BlockDriverState *bdrv_find(const char *name); BlockDriverState *bdrv_next(BlockDriverState *bs); Index: qemu/block_int.h =================================================================== --- qemu.orig/block_int.h 2011-01-24 11:52:25.255004060 +0100 +++ qemu/block_int.h 2011-01-24 11:56:39.801004059 +0100 @@ -153,7 +153,7 @@ struct BlockDriverState { int valid_key; /* if true, a valid encryption key has been set */ int sg; /* if true, the device is a /dev/sg* */ /* event callback when inserting/removing */ - void (*change_cb)(void *opaque); + void (*change_cb)(void *opaque, int reason); void *change_opaque; BlockDriver *drv; /* NULL means no media */ @@ -203,6 +203,9 @@ struct BlockDriverState { void *private; }; +#define CHANGE_MEDIA 0x01 +#define CHANGE_SIZE 0x02 + struct BlockDriverAIOCB { AIOPool *pool; BlockDriverState *bs; Index: qemu/hw/ide/core.c =================================================================== --- qemu.orig/hw/ide/core.c 2011-01-24 11:56:20.904004339 +0100 +++ qemu/hw/ide/core.c 2011-01-24 11:56:39.807004199 +0100 @@ -1584,11 +1584,15 @@ static void ide_cfata_metadata_write(IDE } /* called when the inserted state of the media has changed */ -static void cdrom_change_cb(void *opaque) +static void cdrom_change_cb(void *opaque, int reason) { IDEState *s = opaque; uint64_t nb_sectors; + if (!(reason & CHANGE_MEDIA)) { + return; + } + bdrv_get_geometry(s->bs, &nb_sectors); s->nb_sectors = nb_sectors; Index: qemu/hw/sd.c =================================================================== --- qemu.orig/hw/sd.c 2011-01-24 11:52:25.281004130 +0100 +++ qemu/hw/sd.c 2011-01-24 11:56:39.813007971 +0100 @@ -422,9 +422,14 @@ static void sd_reset(SDState *sd, BlockD sd->pwd_len = 0; } -static void sd_cardchange(void *opaque) +static void sd_cardchange(void *opaque, int reason) { SDState *sd = opaque; + + if (!(reason & CHANGE_MEDIA)) { + return; + } + qemu_set_irq(sd->inserted_cb, bdrv_is_inserted(sd->bdrv)); if (bdrv_is_inserted(sd->bdrv)) { sd_reset(sd, sd->bdrv);
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");
On Mon, Jan 24, 2011 at 01:32:33PM +0100, Christoph Hellwig wrote:
> 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 2011-01-24 11:55:36.744254374 +0100
> +++ qemu/hmp-commands.hx 2011-01-24 11:56:23.619254094 +0100
> @@ -53,6 +53,25 @@ Quit the emulator.
> ETEXI
>
> {
> + .name = "block_resize",
> + .args_type = "device:B,size:o",
> + .params = "device size",
> + .help = "resize a block image",
> + .user_print = monitor_user_noop,
> + .mhandler.cmd_new = do_block_resize,
> + },
> +
> +STEXI
> +@item block_resize
> +@findex block_resize
> +Resize a block image while a guest is running. Usually requires guest
> +action to see the updated size. Resize to a lower size is supported,
> +but should be used with extreme caution. Note that this command only
> +resizes image files, it can not resize block devices like LVM volumes.
> +ETEXI
> +
> +
> + {
> .name = "eject",
> .args_type = "force:-f,device:B",
> .params = "[-f] device",
> Index: qemu/blockdev.c
> ===================================================================
> --- qemu.orig/blockdev.c 2011-01-24 11:56:20.903004129 +0100
> +++ qemu/blockdev.c 2011-01-24 11:56:38.391254165 +0100
> @@ -705,3 +705,33 @@ int do_drive_del(Monitor *mon, const QDi
>
> return 0;
> }
> +
> +/*
> + * XXX: replace the QERR_UNDEFINED_ERROR errors with real values once the
> + * existing QERR_ macro mess is cleaned up. A good example for better
> + * error reports can be found in the qemu-img resize code.
> + */
> +int do_block_resize(Monitor *mon, const QDict *qdict, QObject **ret_data)
> +{
> + const char *device = qdict_get_str(qdict, "device");
> + int64_t size = qdict_get_int(qdict, "size");
> + BlockDriverState *bs;
> +
> + bs = bdrv_find(device);
> + if (!bs) {
> + qerror_report(QERR_DEVICE_NOT_FOUND, device);
> + return -1;
> + }
> +
> + if (size < 0) {
> + qerror_report(QERR_UNDEFINED_ERROR);
> + return -1;
> + }
> +
> + if (bdrv_truncate(bs, size)) {
> + qerror_report(QERR_UNDEFINED_ERROR);
> + return -1;
> + }
Can't resize if block migration is in progress. Don't see a problem
with simply disallowing resize in that case.
Am 24.01.2011 18:44, schrieb Marcelo Tosatti:
> On Mon, Jan 24, 2011 at 01:32:33PM +0100, Christoph Hellwig wrote:
>> 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 2011-01-24 11:55:36.744254374 +0100
>> +++ qemu/hmp-commands.hx 2011-01-24 11:56:23.619254094 +0100
>> @@ -53,6 +53,25 @@ Quit the emulator.
>> ETEXI
>>
>> {
>> + .name = "block_resize",
>> + .args_type = "device:B,size:o",
>> + .params = "device size",
>> + .help = "resize a block image",
>> + .user_print = monitor_user_noop,
>> + .mhandler.cmd_new = do_block_resize,
>> + },
>> +
>> +STEXI
>> +@item block_resize
>> +@findex block_resize
>> +Resize a block image while a guest is running. Usually requires guest
>> +action to see the updated size. Resize to a lower size is supported,
>> +but should be used with extreme caution. Note that this command only
>> +resizes image files, it can not resize block devices like LVM volumes.
>> +ETEXI
>> +
>> +
>> + {
>> .name = "eject",
>> .args_type = "force:-f,device:B",
>> .params = "[-f] device",
>> Index: qemu/blockdev.c
>> ===================================================================
>> --- qemu.orig/blockdev.c 2011-01-24 11:56:20.903004129 +0100
>> +++ qemu/blockdev.c 2011-01-24 11:56:38.391254165 +0100
>> @@ -705,3 +705,33 @@ int do_drive_del(Monitor *mon, const QDi
>>
>> return 0;
>> }
>> +
>> +/*
>> + * XXX: replace the QERR_UNDEFINED_ERROR errors with real values once the
>> + * existing QERR_ macro mess is cleaned up. A good example for better
>> + * error reports can be found in the qemu-img resize code.
>> + */
>> +int do_block_resize(Monitor *mon, const QDict *qdict, QObject **ret_data)
>> +{
>> + const char *device = qdict_get_str(qdict, "device");
>> + int64_t size = qdict_get_int(qdict, "size");
>> + BlockDriverState *bs;
>> +
>> + bs = bdrv_find(device);
>> + if (!bs) {
>> + qerror_report(QERR_DEVICE_NOT_FOUND, device);
>> + return -1;
>> + }
>> +
>> + if (size < 0) {
>> + qerror_report(QERR_UNDEFINED_ERROR);
>> + return -1;
>> + }
>> +
>> + if (bdrv_truncate(bs, size)) {
>> + qerror_report(QERR_UNDEFINED_ERROR);
>> + return -1;
>> + }
>
> Can't resize if block migration is in progress. Don't see a problem
> with simply disallowing resize in that case.
Then we should add a check to bdrv_truncate.
Kevin
On Tue, Jan 25, 2011 at 01:01:04PM +0100, Kevin Wolf wrote:
> > Can't resize if block migration is in progress. Don't see a problem
> > with simply disallowing resize in that case.
>
> Then we should add a check to bdrv_truncate.
What check exactly would that be anyway?
Am 24.01.2011 13:32, schrieb 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---
Thanks, applied all to the block branch.
Kevin
On Mon, Jan 24, 2011 at 12:32 PM, Christoph Hellwig <hch@lst.de> wrote:
> 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.
Did you CC any lists? I didn't see it on
virtualization@lists.linux-foundation.org.
Stefan
On Tue, Jan 25, 2011 at 07:05:37PM +0100, Christoph Hellwig wrote:
> On Tue, Jan 25, 2011 at 01:01:04PM +0100, Kevin Wolf wrote:
> > > Can't resize if block migration is in progress. Don't see a problem
> > > with simply disallowing resize in that case.
> >
> > Then we should add a check to bdrv_truncate.
>
> What check exactly would that be anyway?
Need to add the information to BlockDriver state. drive_del is also
broken, it can be fixed later. Patch looks good to me.